Studies in Inorganic Chemistry 21

Luminescence and
the Solid State
2nd Edition
Studies in Inorganic Chemistry 21
Other titles in this series
1. Phosphine, Arsine and Stibine Complexes of the Transition Elements
by C.A. McAuliffe and W. Levason

2. Phosphorus: An Outline of its Chemistry, Biochemistry and Technology (Second Edition)

by D.E.C. Corbridge

3. Solid State Chemistry 1982

edited by R. Metselaar, H.J.M. Heijligers and J. Schoonman

4. Gas Hydrates
by E. Berecz and M. Balla-Achs

5. Sulfur: lts Significance for Chemistry, for the Geo-, Bio-, and Cosmosphere and Technology
edited by A. Müller and B. Krebs

6. Phosphorus: An Outline of its Chemistry, Biochemistry and Technology (Third Edition)

by D.E.C. Corbridge

7. Inorganic High Pressure Chemistry: Kinetics and Mechanisms

edited by R. van Eldik

8. Graphite Fluorides
by N. Watanabe, T. Nakajima and H. Touhara

9. Selected Topics in High Temperature Chemistry: Defect Chemistry of Solids

edited by Ø. Johannesen and A.G. Andersen

10. Phosphorus: An Outline of its Chemistry, Biochemistry and Technology (Fourth Edition)
by D.E.C. Corbridge

11. Chemistry of the Platinum Group Metals

edited by F.R. Hartley

12. Luminescence and the Solid State

by R.C. Ropp

13. Transition Metal Nuclear Magnetic Resonance

edited by P.S. Pregosin

14. Chemistry of Inorganic Ring Systems

edited by R. Steudel

15. Inorganic Polymeric Glasses

by R.C. Ropp

16. Electron Paramagnetic Resonance of d Transition Metal Compounds

by F.E. Mabbs and D. Collison

17. The Chemistry of Artificial Lighting Devices. Lamps, Phosphors and Cathode Ray Tubes
by R.C. Ropp

18. Structure and Chemistry of the Apatites and Other Calcium Orthophosphates
by J.C. Elliott

19. Molybdenum: An Outline of its Chemistry and Uses

edited by E.R. Braithwaite and J. Haber

20. Phosphorus 2000: Chemistry, Biochemistry & Technology

by D.E.C. Corbridge
Studies in Inorganic Chemistry 21

Luminescence and the

Solid State
2nd Edition

R. C. Ropp
138 Mountain Avenue,
Warren, NJ 07059, U.S.A.

2004

ELSEVIER
Amsterdam - Boston - Heidelberg - London - New York - Oxford
Paris - San Diego - San Francisco - Singapore - Sydney - Tokyo
ELSEVIER B.V. ELSEVIER Inc. ELSEVIER Ltd ELSEVIER Ltd
Sara Burgerhartstraat 25 525 B Street, Suite 1900 The Boulevard, Langford Lane 84 Theobalds Road
P.O. Box 211,1000 AE San Diego, CA 92101-4495 Kidlington, Oxford OX5 1GB London WC1X 8RR
Amsterdam, USA UK UK
The Netherlands

© 2004 Elsevier B.V. All rights reserved.

This work is protected under copyright by Elsevier B.V., and the following terms and conditions apply to its use:

Photocopying
Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the
Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for
advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational
institutions that wish to make photocopies for non-profit educational classroom use.

Permissions may be sought directly from Elsevier's Rights Department in Oxford, UK: phone (+44) 1865 843830, fax (+44)
1865 853333, e-mail: [email protected]. Requests may also be completed on-line via the Elsevier homepage (http://
www.elsevier.com/locate/permissions).

In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood
Drive, Danvers, MA 01923, USA; phone: (+ 1) (978) 7508400, fax: (+ 1) (978) 7504744, and in the UK through the Copyright
Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London W1P 0LP, UK; phone: (+44) 20
7631 5555; fax: (+44) 20 7631 5500. Other countries may have a local reprographic rights agency for payments.

Derivative Works
Tables of contents may be reproduced for internal circulation, but permission of the Publisher is required for external resale or
distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and
translations.

Electronic Storage or Usage

Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter
or part of a chapter.

Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by
any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher.
Address permissions requests to: Elsevier's Rights Department, at the fax and e-mail addresses noted above.

Notice
No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products
liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the
material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug
dosages should be made.

Second edition 2004

First edition 1991, ISBN 0-444-88940-X

Library of Congress Cataloging in Publication Data

A catalog record is available from the Library of Congress.

British Library Cataloguing in Publication Data

A catalogue record is available from the British Library.

ISBN: 0-444-51661-1

® The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper).
Printed in The Netherlands.
 PREFACE

Since the first date of publication of this book in 1991, the subject of
phosphors and luminescence has assumed even more importance in the
overall scheme of technological development. Many new types of displays
have appeared which depend upon phosphors in their operation. Some of
these were pure conjecture in 1991 but are a reality in 2003. The
computer has continued to mature and manipulation of bit-rates by several
billion per second are common in desktop applications (compared to
kilobits per second in 1991). Many, if not all, displays now use computers
(or programmable controllers) to operate properly. I have included
descriptions of the newer (as well as the older) types of displays in this
edition along with an annotated portrait of the phosphors used in each
category. Many of these new light sources promise to displace and make
obsolete our current light sources, i.e.- incandescent lamps, fluorescent
lamps and the ubiquitous color Cathode Ray Tube now used in TV and
computer monitors.

The original volume of this book was written to introduce the reader to
the science and art of preparing inorganic luminescent materials, namely
"phosphors". In order to understand how and why luminescent materials
exhibit such specific intrinsic properties, one needs to be thoroughly
versed in the science of the solid state. Since that time, I have published a
separate volume entitled:

"The Chemistry of Artificial Lighting Devices" - Lamps, Phosphors

and Cathode Ray Tubes" - published by Elsevier in 1993

In this volume, I presented the exact formulas and conditions required to

make all of the phosphors known at that time. Each formula was the
result of many hours of experimentation to optimize the final phoshor
composition. You may wish to consult this volume from time to time in
order to get an idea of how a given phosphor was made.

Because I believe that one needs to know what the history of any given
subject entails, I have retained much of the material presented in the

V
vi Preface

initial edition, while adding information to update the current state of

affairs concerning phosphor usage. I have tried to be as comprehensive as
possible.

I trust that you will be able to completely understand the complexities of

phosphors once you have studied this current volume. It is clear to me
that the use of phosphors is changing and will continue to change as new
types of electronic devices and displays continue to mature. If you are
working in the field, I hope that this presentation is of value in your
project.

I have enjoyed preparing this manuscript and hope that you find reading it
both profitable and enlightening.

R. C. Ropp,
October, 2003
 ACKNOWLEDGEMENT

This book is dedicated to my wife, Francisca Margarita, who has staunchly

supported me over the number of years that it has taken to compose the
material in this book.

vii
This page is intentionally left blank
 TABLE OF CONTENTS
PAGE
PREFACE v
ACKNOWLEDGEMENT vii
INTRODUCTION xvii
CHAPTER 1. - INTRODUCTION TO THE SOLID STATE 1
1.1. Changes of State 3
1.2. Energetics of Changes of State 5
1.3. Propagation Models and the Close-Packed Solid 8
1.4. The Structure of Solids 12
1.5. Determination of Structure of Compounds 25
1.6. The Defect Solid 30
CHAPTER 2. - THE POINT DEFECT 39
2.1. Types of Point Defects 39
I. The Point Defect in Homogeneous Solids 41
II. The Point Defect in Heterogeneous Solids 45
2.2. The Plane Net 49
2.3. Defect Equation Symbolism 57
2.4. Some Applications for Defect Chemistry 58
I. Phosphors 59
II. Lithium Niobate 60
III. Bubble Memories 62
IV. Calcium Sulfide Phosphor 64
2.5. Thermodynamics of the Point Defect 65
I. Statistical Mechanics Approach 65
II. Schottky and Frenkel Defects 66
III. Defect Thermodynamics 68
IV. Combined Approach to Defect Formation 70
a. Interstitial Atoms 72
b. Frenkel Pairs 72
c. Schottky Defects 73
V. Defect Equilibria 73
2.6. Defect Equilibria in Various Types of Compounds 75
I. Stoichiometric Binary Compounds of MXs 76
II. Defect Concentrations in MXs Compounds 78
III. Non-Stoichiometric Binary Compounds 80
IV. Defect Concentrations in MXs± 6 82
V. Ionization of Defects 85
2.7. Brouwer's Approximation Method 88
2.8. Analyses of Real Crystals using Brouwer's Method 95
- Comparison to The Thermodynamic Method
I. The AgBr Crystal with a Divalent Impurity, Cd2+ 95

IX
x Table of Contents

II. Defect disorder in AgBr 100

- A Thermodynamic Approach
2.9. Summary and Conclusions 105
2.10. The Effects of Purity (And Impurities) 107
2.11. Nanotechnology and The Solid State 109

CHAPTER 3 - THE SOLID STATE- MECHANISMS OF NUCLEATION.

SOLID STATE DIFFUSION. GROWTH OF PARTICLES
AND MEASUREMENT OF SOLID STATE REACTIONS 113

3.1. Solid State Reactions and Nucleation Mechanisms 114

I. Types of Solid State Reactions 115
II. Solid State Nucleation Processes and Models 122
III. Diffusion Controlled Solid State Reactions 133
a. The Tarnishing Reaction 133
b. Fick's Laws of Diffusion 136
IV. Role of Defects in Solid State Diffusion Mechanisms 139
V. Analysis of Diffusion Reactions in the Solid State 143
a. Analysis of Diffusion Reactions in Spinel 144
b. Phosphors Based on Spinels 147
c. Diffusion in Silicates and Silicate Phosphors 152
VI. Diffusion Mechanisms Where the Cation Changes Its 161
Valence State
3.2. Homogeneous Nucleation Processes - Particle Growth 164
3.3. Nucleation in Precipitation Reactions 168
3.4. Sequences in Particle Growth 169
I. Ostwald Ripening of a Calcium Phosphate Product 171
II. Sintering and Sintering Processes 172
III. Grain Growth 174
3.5.Methods of Measurement of Solid State Reactions 183
I. Differential Thermal Analysis 184
II. Differential Scanning Calorimetry 198
III. Utilization of DTA and DSC 200
a. Applications of DTA 200
b. Uses of DSC 203
IV. Thermogravimetry 205
V. Determination of Rate Processes in Solid State 211
Reactions
a. Types of Solid State Reactions 212
b. The Freeman-Carroll Method Applied to DTA 215
c. The Freeman-Carroll Method Applied to TGA 216

CHAPTER 4 - MEASURING PARTICLE SIZE AND

GROWING SINGLE CRYSTALS 219
4.1. Measurement of Particle Sizes and Shapes 219
I. Particle Shapes 223
II. Measurement of Particle Shape 228
 Table of Contents xi

III. Measurement of Particle Size 232

a. The Binomial Theorem-Particle Distributions 233
b. Measuring Particle Distributions 236
IV. Analysis of Particle Size Distribution Parameters 243
a. The Histogram 243
b. Frequency Plots 244
c. Cumulative Frequency 245
d. Log-Normal Probability Method 246
e. Types of Log Normal Particle Distributions 248
1. Unlimited Particle Distributions 248
2. Limited Particle Distributions 249
3. Log Normal Distributions- Discontinuous 250
Limits
4. Multiple Particle Distributions 251
f. Using Particle Distribution Parameters in
Industry 252
1. Fluorescent Lamp Phosphor Particles 252
2. Tungsten Metal Powder 253
V. A Typical PSD Calculation 254
VI. Methods of Measuring Particle Distributions 257
a. Sedimentation Methods 258
b. Electrical Resistivity- The Coulter Counter 261
c. Laser Diffractometry 264
d. Other Methods of Measuring Particle Size 266
4.2.- Growth of Single Crystals 268
I. Heating Elements and Crucibles 269
II. Growth of Single Crystals from the Melt 273
a. Necking in a Seed 274
b. Melt Growth Using the Czochralski Method 276
1. Growing a Single Crystal 278
2. Effects of Rotation in the Czrochralski 281
System
3. Design of a Crystal Pulling Apparatus 283
4. Heat Flows in the Czrochralski System 285
5. An Automated Czrochralski System 289
6. Growing Laser Crystals 291
c. Melt Growth Using the Bridgeman-Stockbarger 298
Method
d. Melt Growth Using the Kyropoulos Method 300
e. Crystal Growth Using a Variation of the 301
Kyropolous Method
f. Edge Defined Crystal Growth from the Melt 305
g. Zone Melting and Refining 307
1. Crystal Purification Using Zone Refining 308
2. The Impurity Leveling Factor 311
h. Crystal Growth from the Melt- 315
The Verneuil Method
xii Table of Contents

i. Melting and Stoichiometry of the Crystal 319

1. Dystectic Crystals 321
2. Defect-Free Single Crystal Silicon 322
j . Actual Imperfections of Crystals Grown 324
from a Melt
k. Crystal Growth from a Liquid Phase 327
1. Growth of Single Crystals From 327
a Molten Flux
2. Hydrothermal Growth of Crystals 330
1. Vapor Methods Used for Single Crystal Growth 340
CHAPTER 5 - OPTICAL AND ELECTRONIC PROPERTIES OF SOLIDS 345
5.1. The Nature of Light 345
5.2. Absorbance, Reflectivity and Transmittance 351
5.3. Electronic Aspects of Crystals 360
I. Band Models and the Reciprocal Lattice 360
II. Calculation of Energy Bands in Crystals 370
III. Point Defects and the Energy Band Model 382
a. Crystal Structures 382
b. Brillouin Zones 382
c. Energy Bands in Solids 383
5.4.- Phonons as Quantized Lattice Vibrations 387
I. Phonon dispersion Equations 389
II. The Case of the Impurity Activator Center 395
in a Phosphor
5.5.- Electronic Aspects of Phosphors 396
I. Energy Processes in a Phosphor 398
II. Properties Associated with Phosphors 406
a. Notation 407
b. Quantum efficiency 408
c. Decay times 409
d. Band shapes 411
e. Spectroscopic notation 413
III. Organic Phosphors and their Luminescent 416
Mechanisms
5.6.- Factors Associated with Excitation Energy Conversion 418
I. Prediction of Electronic Transition Intensities 420
a. Multipole Interactions 420
b. Einstein Absorption and Emission 421
Coefficients
c. Electronic Transition Moments 423
1. Concise Hamiltonian Calculation 424
of Transition Moments
2. Detailed Calculation of Electronic 425
Transition Moments
d. Dipole and Multipole Oscillator Strengths 428
 Table of Contents xiii

5.7.- Mechanisms of Energy Transfer Mechanisms in Solids 431

I. Radiative Transfer (Radiation Trapping) 432
II. Energy Transfer by Resonance Exchange 433
III. Energy Transfer by a Spatial Process 436
IV. Energy Exchange by Spin Coupling 439
V. Energy Transfer by Non-Resonant Processes 442
5.8.- Summary of Phonon Processes as Related to Phosphors 444
I. Excitation of the Activator Center 444
a. Phonon process 444
b. Energy flow 444
II. Excited State Relaxation 444
a. Phonon process 444
b. Energy flow 444
III. Center Ready to Emit 444
a. Direct relaxation 444
b. Emission of a photon 444
c. Resonant and non-resonant processes 445
d. Actual phonon process 445

CHAPTER 6 - DESIGN OF PHOSPHORS 447

6.1.- The Luminescent Center in Inorganic Phosphors 447

6.2.- The Ground State Perturbation Factor 452
6.3.- Design of a Phosphor 456
I. Choice of the Host Components 456
II. Choice of the Activator(s) 462
III. Quenchers or "Killers" of Luminescence 468
6.4.- Factors Affecting Phosphor Efficiencies (Brightness) 471
6.5.- Preparation of Phosphors 480
I. Phosphor Parameters 480
a. Independent Variables 480
b. Dependent Variables 486
c. The Firing Cycle as a Dependent Variable 488
d. Size and Mass Fired 492
e. Firing Atmosphere 493
f. Effect of Host Structure 493
g. Effect of Preparation Method 499
h. Ratios of Components and Use of a Flux 500
6.6.- Commercial Phosphors 501
I. Cathode-ray Phosphors 501
a. How Cathode-Ray Phosphors Are Used 502
b. JED EC Phosphors 506
II. Fluorescent Lamp Phosphors 511
6.7.- Measurement of Optical Properties of Phosphors 516
I. Measurement of Phosphor Brightness 517
II. Measurement of Spectral Energy Distributions 518
III. Measurement of Quantum Efficiencies 520
IV. Specification and Measurement of Color 520
a. The Human Eye 521
xiv Table of Contents

b. The Nature of Chroma 523

c. The Standard Observer 526
d. Color Measurement 532
e. Color Matching 541
f. Color Spaces 543
g. The Munsell Color Tree 544
h. Color Matching and MacAdam Space 548
6.8.- Rare Earths and Their Spectral Properties 552
I. History of the Lanthanides 552
II. Chemistry of the Lanthanides 554
III. Rare Earth Energy Levels and Electronic States 559
a. Calculating Rare Earth Energy Levels 565
b. The Concept of Fractional Parentage 568
c. Energy Transitions and Mixed States 573
d. Spectroscopic Rules for the Rare Earths 575
e. Experimental Stark States 577
f. The Free-Ion Approach to Rare Earth 582
Energy Levels
g. Charge Transfer States and 5d Multiplets 585
h. Phonon Assisted Relaxation in Phosphors 587
g. Anti-Stokes Phosphors 594
h. The Solid State Laser 603
CHAPTER 7 - CURRENT PHOSPHOR DEVICE TECHNOLOGY 615
7.1.- Current Progress in CRT Display and Discharge Lighting 615
I. Cathode-Ray Tubes (CRT) 615
II. Fluorescent Lamps 622
7.2.- Current Display Devices Based Upon Electron Generation 628
I. CRT's for Display, Measurement and Specialty Types 630
a. Oscilloscopes & Storage Tubes-Measurement 630
b. Flying Spot Scanner 632
c. Radar Displays 633
d. Image Intensifier Screens 635
II. Thin Film Technology 637
III. Devices Using Internal Electron Beam Generation
a. Vacuum Fluoresecent Displays 645
b. Field Emission Displays (FED) 648
c. Large Outdoor Displays 651
d. Light Emitting Diodes (LED) 652
e. Light Emitting Diode Lasers 659
f. Organic Light Emitting Diode Displays (OLED) 667
g. Electroluminescent Displays (EL) 673
7.3. Current Display Devices Based Upon Photon Generation 676
I. Fluoresecent Lamps for Special Purposes 677
a. Liquid Crystal Back-Light Displays 677
b. Lamps for Copying Machines 683
c. Blacklight Lamps for Water Treatment 683
d. Medical Uses for Blacklight Lamps 684
 Table of Contents xv

II. Display Devices Based Upon Vacuum UV Photons 684

a. Plasma Display Panels-Vacuum UV 685
b. Neon Signs 690
III. Devices Utilizing High Energy Photons 691
a. X-Ray Intensifying Screens 691
b. Scintillation Phosphors for Imaging 696
c. Long Decay Phosphors 701
SUBJECT INDEX 703
This page is intentionally left blank
 INTRODUCTION

As many of you may not have had the opportunity to investigate the body of
knowledge concerning phosphors, I have continued the survey and
summarization of the important parts of solid state science in the
introductory chapters of this 2nd edition. Many of these chapters have
been completely rewritten or modified from the original descriptions.
Each chapter has a special contribution to make in your overall
understanding of the solid state science of phosphors and luminescence.

For example, Chapter 1 begins with a survey of factors needed to define

the solid state. It also contains a survey of x-ray techniques which one
needs to know concerning solid state lattice structures of phosphors.
Chapter 2 deals with the point-defect aspects of solids in general, which
are shown to be thermal in origin. Examples taken from phosphors are
presented. The thermodynamics and equilibria of point defects is
presented, as well as practical ways to evaluate the effects of point defects
in solids. Chapter 3 is new in that mechanisms of nucleation, solid state
diffusion and sequences in particle growth are described together.
Chapter 4 presents methods of measuring particle shapes and sizes,
particle size distributions and growing single crystals. Chapter 5 is very
important since the factors associated with energy conversion, electronic
transition intensities, energy transfer mechanisms and phonon processes
(vibronic coupling) concerning phosphors are presented. Chapter 6
presents the factors identified with design of phosphors and those
required to obtain efficient luminescence. The factors affecting phosphor
efficiencies are delineated as well. Chapter 7, the final one, describes first
the recent advances made in both CRT's and fluorescent lamps, then
vacuum fluorescent displays, field emission displays, light emitting diodes
and lasers, x-ray intensifying screens and plasma display tubes. All of
these utilize phosphors in their construction in some way. I have divided
these new displays and light sources into 2 categories: those that use
electrons to excite the phosphors and those that depend upon generated
high energy photons to excite the phosphors to operate as a device. I have
enjoyed the preparation of this manuscript and hope that you find reading
it both profitable and enlightening.

xvii
This page is intentionally left blank
 Introduction to The Solid State

This treatise is written to elucidate and explain the characteristics of

phosphors and the phenomenon of luminescence. That is, the nature of
materials which absorb energy and refimit it as light. We encounter
phosphors every day when we watch television. The television tube (or
plasma panel as it may be) has dots (or lines) composed of red, green and
blue-emitting phosphors. These produce the active pictures that we view
from Cable or TV-Network broadcasts such as ABC, CBS and NBC. We also
use phosphors in fluorescent lamps as inside (and outside) lighting in our
homes and offices. However, in order to understand how phosphors
perform, we need to know something about the nature of solids in order
to comprehend the luminescent state. This chapter will summarize the
nature of solids as related to phosphors in general. Of necessity, we
cannot be completely comprehensive. Many scientific texts have been
written on each of these individual subjects and you can refer to them to
get more specific answers to certain questions that cannot be covered
here. This chapter will cover the basics of:

1. A comparison of the three states of matter

2. How one determines the structure of solids
3. An introduction to the defect solid state

Because of the importance of the defect solid (all phosphors fall into this
category), we will devote a chapter to the defect-state. Since we need to
know how phosphors are made, a chapter covering solid state reactions is
also mandatory. Finally, we will cover both single crystal formation and
particle size determinations. Some phosphors are single crystal. A good
example is that of the laser. The first laser invented by Maiman used a
single crystal of ruby, i.e.- aluminum oxide or sapphire (corundum)
activated by trivalent chromium (Cr*). A great variety of laser crystals are
now in use including YAGiNd3^, i.e.- Y3Al5On: Nd3+. Since most phosphors
are used as powders, we will also include a chapter showing how one
determines particle size and particle size distributions. We will also

1
2 Introduction to the solid state

include a description of the nature of light and photons in general since

phosphors emit photons once they are excited by absorption of energy.
Color is also an important component of photons as we perceive them.

Because I believe that one needs to be aware of the history of science in

order to thoroughly understand any part of it, I will try to include as much
background and results of investigations as they pertain to phosphors. For
example, the methods of determining structures of solids did not occur
overnight but were developed slowly over a period of time once Rontgen
discovered x-rays. Nowadays, if you are in this field, you probably use a
computer-controlled program to collect diffraction data automatically and
to do the calculations for you. Yet, you must to know how the calculations
need to be accomplished in order to ascertain whether you have obtained
an accurate depiction of the structure of any material, or not.

When a detailed study of solids is undertaken, one quickly determines

that solids have properties which differ profoundly from those of the
other states of matter. We will be concerned mainly with inorganic
solids, although the same principles apply to organic solids as well. It is
well to note here that OLEDs, or "Organic Light Emitting Diodes" are the
promise of the future to replace inorganic phosphors as light-emitting
displays in certain applications. Inorganic solids are the primary materials
of construction for use in electronics, lighting, communications,
information storage and display, superconductors, and others at this time,
but have been supplanted by organic based materials in some applications.

In comparing the three states of matter, our approach will be as follows.

First, we will compare the three (3) basic states of matter, namely gases,
liquids and solids. We will then contrast these states energetically and
atomistically. Next, we will discuss structures of solids and the factors
involved in determination of crystal structure. Finally, we will introduce
the concept of the defect solid and how such defects affect the
macroscopic properties of the solid state.

We can summarize differences between the three states of matter as

follows:
 1.1 Changes of state 3

a. The solid is the most condensed phase of the three normally

possible, and contains the least energy.
b. The solid has but three (3) degrees of freedom in contrast to
those usually present in the gaseous state. (The number actually
present in a gas depends upon the number of atoms in the
molecule).
c. All solids contain defects. There is no such thing as a perfect
solid.

1.1.- CHANGES OF STATE

Of the three states of matter, gaseous, liquid and solid, the major
difference involves their atomistic freedom of movement. For example, a
gas molecule can move in three (3) dimensions without restriction and
has full vibrational and rotational modes within the molecule. We know
that it is possible to condense gases to liquids to solids by removal of
energy. Each state represents a succeedingly lower degree of
condensation, and has a characteristic temperature range at which it
exists, the gas being the "hottest", and solids the "coldest". Thus, solids
contain the least energy. The actual temperature range for the gaseous,
liquid and solid state of a given material depends, of course, upon the
nature of the atoms involved. Even elements differ quite widely. Gallium,
for example, is a metal that is liquid slightly above room temperature, but
does not vaporize until above 1100 °C. Tungsten, also a metal, does not
melt until the temperature is above 3850 °C.

An elementary way to understand how the solid differs from the gaseous
or liquid states involves the following example. Take the MX2 molecule. In
the gaseous state, there will be a maximum of 3 x 3, or nine degrees of
freedom, for the molecule composed of these three atoms (there are
three dimensions, x, y, & z, in which to operate). These can be divided
into 3-translational, 3-vibrational and 3-rotational degrees of freedom. But
as we change the state of matter, we know that the translational degrees
of freedom present in gases disappear in solids (and are restricted in
liquids). For gaseous molecule, the 3 vibrational degrees of freedom will
have (2J+1 = 7) rotational states superimposed upon them (J here is the
4 Introduction to the solid state

number of atoms in the molecule, assuming quantized vibrational states).

If we measure the absorption spectra of our molecule, MX2 , in the infra-
red region of the spectrum, we obtain results similar to those shown as
follows:

1.1.1.-

The gaseous phase shows seven well separated rotational and vibrational
states, the liquid phase broadened vibrational plus rotational states (they
are no longer distinct), while the solid exhibits only one broad featureless
band in this region of the spectrum. We conclude that we are limited to
three degrees of freedom in the solid state. The 3-translational and 3-
rotational degrees of freedom have disappeared because of the imposition
of long range ordering of MX2 molecules in the solid. This point cannot
be over-emphasized since long range ordering is the major difference
between solids and the other states of matter.
 1.2 Energetics of changes of state 5

1.2.- ENERGETICS OF CHANGES OF STATE

When heat energy is added to a solid, one of two changes will occur.
Either the internal heat content (temperature) of the material will change
or a change of state will occur. On an atomistic level, this involves an
increase in vibrational energy which manifests itself either as an increase
in internal temperature, or the breaking of bonds in the solid to form
either a liquid or gas. The most common example used to illustrate
energy involved in changes of state is that of water. We know that water,
a liquid, will change to a solid (ice) if its internal temperature falls below
a certain temperature. Likewise, if its temperature rises above a certain
point, water changes to a gas (steam). Because water is so plentiful on the
Earth, it was used in the past to define Changes of State and even to
define Temperature Scales.

The invention of the thermometer is generally credited to Galileo. In his

instrument, built about 1592, the changing temperature of an inverted
glass vessel produced the expansion or contraction of the air within it,
which in turn changed the level of the liquid with which the vessel's long,
open-mouthed neck was partially filled. This general principle was
perfected in succeeding years by experimenting with liquids such as
alcohol or mercury and by providing a scale to measure the expansion and
contraction brought about in such liquids by rising and falling
temperatures.

By the early 18th century as many as 35 different temperature scales had

been devised. The German physicist, Daniel Gabriel Fahrenheit, in the
period of 1700-30, produced accurate mercury thermometers calibrated
to a standard scale that ranged from 32, the melting point of ice, to 96
for body temperature. The unit of temperature (degree) on the
Fahrenheit scale is 1/180 of the difference between the boiling (212) and
freezing points of water. The first centigrade scale (made up of 100
degrees) is attributed to the Swedish astronomer Anders Celsius, who
developed it in 1742. Celsius used 0 for the boiling point of water and
100 for the melting point of snow. This was later inverted to put 0 on the
cold end and 100 on the hot end, and in that form it gained widespread
6 Introduction to the solid state

use. It was known simply as the centigrade scale until in 1948 the name
was changed to honor Celsius. In 1848 the British physicist William
Thompson (later Lord Kelvin) proposed a system that used the degrees
that Celsius used, but was keyed to absolute zero (-273.15 °C). The unit of
this scale is now known as the Kelvin, i.e.- °K. The Rankine scale employs
the Fahrenheit degree keyed to absolute zero (-459.67 °F) , i.e.- °R.
These are the four temperature scales that we employ today.

The factors involved in energy change, i.e.- temperature change, include:

"heat capacity", i.e.- Cp or Cv, and "heat of transformation", H. The former
is connected with internal temperature change whereas the latter is
involved in changes of state. The actual names we use to describe H
depend upon the direction in which the temperature change occurs, vis:

1.2.1.- CHANGE OF STATE H. HEAT OF: TEMPERATURE

fC, !F
ice to water Fusion 0 32
water to ice Solidification 0 32
water to steam Vaporization 100 212
steam to water Condensation 100 212

Note that when a change of state, i.e.- water to ice or ice to water, occurs,
there is no change in temperature while this is occurring (this is a
concept that is sometimes difficult for beginning scholars to grasp). Heat
capacity was originally defined in terms of water. That is, heat capacity
was defined as the amount of heat required to raise the temperature of
one cubic centimeter of water (whose density is defined as 1.0000 @ 4
°C.) by one(l) degree. "Heat" itself is defined as the amount of energy
required to raise 1 cc. of water by one degree, i.e.- one (1.0) calorie. The
calorie in turn was originally defined as the amount of heat required to
raise the temperature of one gram of water from 14.5 °C to 15.5 °C at a
constant pressure of one (1) atmosphere. Heat capacity was also known as
thermal capacity. We thus label heat capacity as Cp, meaning the thermal
capacity at constant pressure. Originally, Cv, the thermal capacity at
constant volume was also used, but its use is rare nowadays. This is due to
the fact that some materials do not have a linear temperature expansion.
 1.2 Energetics of changes of state 7

These are two types of "heat" involved in the thermal changes of any given
material. We specify the "heat" of a material, i.e.- its internal energy, by
Hs,L,G- where S,L, or G refer to solid, liquid or gas. Thus, the relation
between Hs,L,G and Cp i S:

1.2.2.- {H == C p • T}s,L,G, or AHs,L,G = Cp(S,L,G) ATs,L,G

where the actual state of matter is either S, L, or G.

Note that: The internal temperature of a material does not change

as the material undergoes a change of state. All of the energy goes
into forming a new state of matter.

For a change of state between solid and liquid, we would have:

1.2.3- {Hs - HL} = Cp (T(S) - T(L)) or AH = Cp AT

AH here is a heat of transformation involved in a change of state whereas

AHs,L,G (sometimes written AEs,L,G) refers to change in internal heat
for a given state of matter.

We can now calculate the amount of energy required to raise one gram of
ice at - 10 °C. to form one gram of steam at 110 °C:

1.2.4.- CALORIES REQUIRED TO CHANGE 1 Gm OF ICE TO STEAM

FORM AT c E> C n AT AHx TOTAL

Ice -10 to 0 0. 5 5.0 cal. ... 5.0 cal.
(Ice 0 ... ... 80 cal. 80.0 cal.
to water) (fusion)
Water 100 1. 0 100 cal. ... 100.0 cal.
(Water 0 ... ... 540 cal. 540.0 cal.
To steam) (vaporization)
Steam 10 0. 5 5.0 cal. 5.0 cal.

730.0 cal.
8 Introduction to the solid state

Note that in this case AB^ is specified in terms of the type of change of
state occurring, while Cp AT (= AH) is the change in internal heat which
occurs as the temperature rises. At a given change of state, all of the
energy goes to the change of state and the temperature does not change
until the transformation is complete.

1.3. - PROPAGATION MODELS AND THE CLOSE-PACKED SOLID

We have already said that the solid differs from the other states of matter
in that long range ordering of atoms or molecules has appeared. To
achieve long range order in any solid, one must stack atoms in a
symmetrical way to completely fill space. If the atoms are all of one kind,
i.e.- one of the elements, the problem is straight forward. Sets of eight
atoms, each set arranged as a cube, will generate a cubic structure. Two
sets of three atoms, each set of three arranged in a triangle, will
propagate a hexagonal pattern with three dimensional symmetry. The
following diagram illustrates this point:
 1.3 Propagation models and the close-packed solid 9

This diagram shows a perspective of cubic and hexagonal close-packing

propagation units in their space-filling aspects. What we mean by-
propagation units are solid state building blocks that we can stack in a
symmetrical form to infinity. We do not consider 1-, 2- or 3-atom units
since they are trivial. That is, they are only 1- or 2-dimensional at best.
But, 4-atoms will form a 3-dimensional tetrahedron (half a cube is only 2-
dimensional) which is a valid propagation unit. This means is that we can
take tetrahedrons and fit them together 3-dimensionally to form a
symmetrical structure which extends to infinity. However, the same is
not true for 5-atoms, which forms a four-sided pyramid. This shape
cannot be completely fitted together in a symmetrical and space-filling
manner. Even if we stack these pyramids, we find that their translational
properties preclude formation of a symmetrical structure (There is too
much lost space!) However, if one more atom is added to the pyramid, we
then have an octahedron which is space-filling with translational
properties.

Going further, combinations of seven atoms are asymmetrical, but eight

atoms form a cube which can be propagated to infinity. Note that by
turning the top layer of four atoms (see 1.3.1.) by 45° , we have a
hexagonal unit which is related to the hexagonal unit composed of 6
atoms, two triangles atop of each other. By taking Ping-Pong balls and
gluing them together to form the propagation units as shown in 1.3.1.,
one can get a better picture of these units.

Although we can continue with more atoms per propagation unit, it is

easy to show that all of those are related to the four basic propagation
units found in the solid state (and depicted in 1.3.1.), to wit:

1.3.2.- Tetrahedron (4) Octahedron (6)

Hexagon (6 or 8) Cube (8)

We thus conclude that structures of solids are based, in general, upon four
basic propagation units, which are stacked in a symmetrical and space-
filling form to near infinity. Variation of structure in solids depends upon
whether the other atoms forming the structure are larger or smaller than
10 Introduction to the solid state

the basic propagation units composing the structure. In many cases they
are smaller and will fit into the interstice of the propagation unit. A good
example is the phosphate tetrahedron, PO4. The P4+ atom is small enough
to fit into the center (interstice) of the tetrahedron formed by the four
oxygen atoms. If we combine it with Er^+ (which is slightly smaller than
PO4), we obtain a tetragonal structure, a sort of elongated cube of high
symmetry. But if we combine it with La^+ , which is larger than PO43- , a
monoclinic structure with low symmetry results.

Now, you may think that this discussion is mainly conjecture since the
case in nature may differ considerably. Actually, it is easy to ascertain that
most of the elements (which are homogeneous, i.e.- having the all of the
same type of atoms) form structures that are either cubic or hexagonal.
This is shown in the following diagram:

1.3.3.-

It should be clear that the metal elements have either a face-centered,

 1.3 Propagation models and the close-packed solid 11

body-centered or hexagonal close-packed structure. These three

structural units are shown in the following diagram:

A better perspective is given in the 1.4.14 diagram, given below. Therein,

we show the seven types of close-packed structures, i.e.- Bravais lattices,
that appear in the solid state. Of those elements whose structures are not
indicated in the above diagram, the following table shows their structure:

Table 1-1
Crystal Structures of Some Elements
Gases at Room Temp Liquids at Room Temp. Other Solid Metals
H, He = hexagonal
Ne, Ar, Kr, Xe, Rn Ga = orthorhombic Rhombohedral
(Solid = fee) Hg = rhombohedral B, As, Sb, Bi
F = cubic Orthorhombic
Cl, Br = orthorhombic I, S, Np
N = hexagonal P, Po - Monoclinic
O = cubic In, Sn- Tetragonal
fcc= face-centered Co, Ni, Cu, etc.
cubic

You will note that even though some elements are gases or liquids at room
temperature, the structure that they form in the solid has been
determined. Some of the temperatures required to condense them into
the solid phase are close to absolute zero.

But, if the atoms are not all of the same kind, i.e.- heterogeneous, we have
12 Introduction to the solid state

a different problem. Part of the problem lies in the fact that 6.023 x
10^3 molecules comprise one mole and we must stack these in a
symmetrical manner to form a close-packed solid. For CSCO3, this
represents 100. 1 grams of powder. We find in general that solid
structures are based on the largest atom present, as well as how it stacks
together (its valence) in space filling-form. For most inorganic
compounds, this is the oxygen atom, e.g.- oxides, silicates, phosphates,
sulfates, borates, tungstates, vanadates, etc. The few exceptions involve
chalcogenides, halides, hydrides, etc., but even in those compounds, the
structure is based upon aggregation of the largest atom, i.e.- the sulfur
atom in ZnS. Zinc sulfide exhibits two structures, sphalerite- a cubic
arrangement of the sulfide atoms, and wurtzite- a hexagonal arrangement
of the sulfide atoms. Divalent zinc atoms have essentially the same
coordination in both structures.

This brings us to a discussion of the structure of solids and how structure

is determined.

1.4.- THE STRUCTURE OF SOLIDS

We have indicated that solids can have several symmetries. Let us now
examine the structure of solids in more detail. In 1895, ROntgen
experimentally discovered "x-rays" and produced the first picture of the
bones of the human hand. This was followed by work by von Laue in 1912
who showed that solid crystals could act as diffraction gratings to form
symmetrical patterns of "dots" on a photographic film whose arrangement
depended upon how the atoms were arranged in the solid. It was soon
realized that the atoms formed "planes" within the solid. In 1913, Sir
William Henry Bragg and his son, William Lawrence Bragg, analyzed the
manner in which such x-rays were reflected by planes of atoms in the
solid. They showed that these reflections would be most intense at
certain angles, and that the values would depend upon the distance
between the planes of atoms in the crystal and upon the wavelength of the
x-ray. This resulted in the Bragg equation:

1.4.1.- nX = 2d sine
 1.4 The structure of solids 13

where d, the distance, is in angstroms (A = 1O~8 cm) between planes and

6 is the angle in degrees of the reflection. In Bragg's x-ray diffraction
equation, i.e.- the angle , 8 , is actually the angle between a given plane of
atoms in the structure and the path of the x-ray beam. The unit, "d", is
defined as the distance between planes of the lattice and X is the
wavelength of the radiation. It was Georges Friedel who in 1913
determined that the intensities of reflections of x-rays from these planes
in the solid could be used to determine the symmetry of the solid. Thus,
by convention, we usually define planes, not points, in the lattice. We can
define the structure of any given solid in terms of its lattice points. What
this means is that if we substitute a point for each atom (ion) composing
the structure, we find that these points constitute a lattice, i.e.- an
internal configuration, having certain symmetries.

A lattice is not a structure per se. A lattice is defined as a set of three-

dimensional points. These points may, or may not, be totally occupied by
the atoms composing the structure. Consider a cubic structure such as
that shown in the following diagram:

1.4.2.-
14 Introduction to the solid state

Here we have a set of atoms (ions) arranged in a simple three-

dimensional cubic pattern. The lattice directions are defined, by
convention, as x , y & z. Again, by convention, "x" is defined as the right
hand direction from the origin, "y" is in the vertical direction and "z" is at
an angle to the plane of x & y (In our cube, the angle is 90°) . Note that
there are eight (8) cubes in our example. The unit-cell is the smallest
cube. The unit-cell directions are defined as the "lattice-unit-vectors".
That is, the x, y, & z directions of the unit cell are vectors having
directions corresponding t o x @ a ; y @ T b ; z @ c with lengths of
each unit- vector being equal to 1.0. (Our notation for a vector henceforth
is a letter which is "outlined", i.e.- the a unit-cell translation vector).
T, the translation vector , is then:

1.4.3.- T = iij a + n2 to + n 3 c

where ni , n2 , and n 3 are intercepts of the unit-vectors, a , to , c , on

the x , y , z - directions in the lattice, respectively. The unit cell volume is
then:

1.4.4.- ¥ = {a • b x e } (This is a "dot - cross" vector product).

Thus, it is easy to see that in Bragg's x-ray diffraction equation, the angle,
9 , is actually the angle between a given plane of atoms in the structure
and the path of the x-ray beam. The unit, "d", is defined as the distance
between planes of the lattice and X is the wavelength of the radiation.

As we said, we usually define planes, not points, in the lattice. The reason
that we do this is that the waves of electromagnetic radiation are
constructively diffracted by planes of atoms in the solid rather than points
in the lattice. One system that has come into general use is that of
"MILLER INDICES" which is represented by: { h , k , 1}.

In this system, we use a , b, & c (not unit-cell vectors) to represent the

lengths of intercepts which define the planes within the unit cell. Miller
Indices are the reciprocals of the intercepts, a , b & c, of the chosen
plane on the x , y , z - directions in the lattice.
 1.4 The structure of solids 15

To illustrate this concept, examine the symmetry elements of our cube,

shown as follows:
1.4.5.-
16 Introduction to the solid state

Note that in all cases, we have the intercepts of the a, b & c specified as
Miller Indices. This simply means that given these indices, we can
determine where the atoms lie within the unit cell of the lattice. This is
indicated in the following:

1.4.6.- a. b . c MILLER INDICES

(1/2, 0 ,. 0 ) (200)
(0 , 1/2 , 0 ) (020)
( 0, 0, 1 12) (002)

Note also the (100), (110) and (111) planes are illustrated. Planes are
important in solids because, as we will see, they are used to locate atom
positions within the lattice structure. The TETRAD, TRIAD , AND DIAD
AXES ARE ALSO SHOWN IN Part D. These are rotational symmetry axes.
That is, the triad axis must be rotated 3-times in order to bring a given
corner back to its original position.

The final factor to consider is that of the angle between the x , y , and z
directions in the lattice. In our examples so far, angles were 90° in all
directions. If the angles are not 90°, then we have additional lattices to
define. For a given unit-cell defined by the axes a, b and c, the
corresponding angles are defined as: a , p , y , where a is the angle in
the x-direction, etc.

In 1921, Ewald developed a method of calculating the sums of diffraction

intensities from different planes in the lattice by considering what is
called the "Reciprocal Lattice". The reciprocal lattice is obtained by
drawing perpendiculars to each plane in the lattice, so that the axes of
the reciprocal lattice are perpendicular to those of the crystal lattice.
This has the result that the planes of the reciprocal lattice are at right
angles (90°) to the real planes in the unit-cell. Ewald used a sphere to
represent how the x-rays interact with any given lattice plane in three
dimensional space. He employed what is now called the Ewald Sphere to
show how reciprocal space could be utilized to represent diffraction of x-
rays by lattice planes. Ewald originally rewrote the Bragg equation as:
 1.4 The structure of solids 17

1.4.7.- sinq = n 1/ = l/d{hkl}

2(1 {hkl} 2/1

Using this equation, Ewald applied it to the case of the diffraction sphere
which we show in the following diagram:

1.4.8.-

In this case, the x-ray beam enters the sphere enters from the left and
encounters a lattice plane, L. It is then diffracted by the angle 26 to the
point on the sphere, P, where it is registered as a diffraction point on the
reciprocal lattice. The distance between planes in the reciprocal lattice is
given as 1/dhkl which is readily obtained from the diagram. It is for these
reasons, we can use the Miller Indices to indicate planes in the real
lattice, based upon the reciprocal lattice.

The reciprocal lattice is useful in defining some of the electronic

properties of solids. That is, when we have a semi-conductor (or even a
conductor like a metal), we find that the electrons are confined in a band,
defined by the reciprocal lattice. This has important effects upon the
conductivity of any solid and is known as the "band theory" of solids. It
turns out that the reciprocal lattice is also the site of the Brillouin zones,
i.e.- the "allowed" electron energy bands in the solid. How this originates
is explained as follows.
18 Introduction to the solid state

The free electron resides in a quantized energy well, defined by k (in

wave-numbers). This result can be derived from the SchrSdinger wave-
equation. However, in the presence of a periodic array of electromagnetic
potentials arising from the atoms confined in a crystalline lattice, the
energies of the electrons from all of the atoms are severely limited in
orbit and are restricted to specific allowed energy bands. This potential
originates from attraction and repulsion of the electron clouds from the
periodic array of atoms in the structure. Solutions to this problem were
made by Bloch in 1930 who showed they had the form (for a one-
dimensional lattice):

1.4.9.- *P = e^ujx) - one dimensional

*Pk (a) = e^Uklx) - three dimensional

where k is the wave number of the allowed band as modified by the

lattice, a may be x, y or z, and u^x) is a periodical function with the same
periodicity as the potential. One representation is shown in the following
diagram, given as 1.4.10. on the next page.

We have shown the least complicated one which turns out to be the
simple cubic lattice. Such bands are called "Brilluoin" zones and , as we
have said, are the allowed energy bands of electrons in any given
crystalline lattice. A number of metals and simple compounds have been
studied and their Brilluoin structures determined.

However, when one gives a representation of the energy bands in a solid,

a "band-model" is usually presented. The following diagram, presented as
1.4.11. on a succeeding page, shows three band models as used to depict
energy states of any given solid. They include insulators, semi-conductors
and metals (conductive). This is. In the solid, electrons reside in the
valence band, (as defined by the Brilloin zones in the reciprocal lattice)
but can be excited into the conduction band by absorption of energy. The
energy gap of various solids depends upon the nature of the atoms
comprising the solid. Semi-conductors have a rather narrow energy gap
(forbidden zone) whereas that of insulators is wide (metals have little or
no gap). That is, the allowed bands in metals overlap with the valence
 1.4 The structure of solids 19

1.4.10.- The Allowed Energy Bands (Brillouin Zones) in a Crystal

band. Note that energy levels of the atoms "A" in 1.4.11. are shown in the
valence band. These will vary depending upon the nature atoms present.
Electrons can be excited into these low-lying states, depending upon the
temperature of the material.
Thus, we find that a total of three (3) factors are needed to define a given
lattice and its structure. This is shown as follows:

1.4.12.- I - unit-cell axes , intercepts and angles

II - rotational symmetry
III - localized space group symmetry
20 Introduction to the solid state

1.4.11.- Energy Band Models Used to Depict Energy States of Solids

It is these three factors (see 1.4.12.) which give rise to the different
symmetries of solids:

Factor I gives rise to the 14 Bravais lattices

Factor II generates the 32 point- groups
Factor III creates the 232 space- groups

In these three factors, each contributes to the total number of

symmetrical lattices that can appear in the solid state. This will become
evident in the following discussion: In Factor I (which we have already
considered), if a * b * c and a * p * y, then we have a different lattice
than that defined by a = b * c ; a = p * Y The number of combinations
that we can make from these 3-lengths and 3-angles is seven (7) and
these define the 7 unique lattice structures, called BRAVAIS LATTICES.
These have been given names, as shown in 1.4.13. on the next page.

Each of these seven lattices may have sublattices, the total being 14. If we
arrange the crystal systems in terms of symmetry, the cube has the
 1.4 The structure of solids 21

1.4.13.- CUBIC TETRAGONAL HEXAGONAL

ORTHORHOMBIC TRIGONAL MONOCLINIC
TRICLINIC

highest symmetry and the triclinic lattice, the lowest symmetry. We can
therefore arrange the seven (7) systems into a hierarchy as shown in the
following diagram:

The highest symmetry lattice is at the top, while the lowest is at the
bottom. These 14 Bravais lattices are shown in 1.4.15. given on the next
page.

If we now apply rotational symmetry (Factor II) to the 14 Bravais lattices,

we obtain the 32 Point-Groups with the factor of symmetry imposed upon
them. The symmetry elements that have been used are shown as follows:

1.4.16.- Rotation axes

Plane symmetry < horizontal
J J
vertical
Inversion symmetry (mirror)

Table 1-2, lists these on the following page with corresponding symbols,
and the relation between axes and angles associated with each structure.
These are the 14 Bravais Lattices which are unique in themselves.
22 Introduction to the solid state

Note that only cubic, tetragonal and orthorhombic lattices have 90° angles
in all lattices directions. Monoclinic and hexagonal lattices have two 90°
directions while trigonal and triclinic lattices have none. Table 1.2. lists
the angles of all of the 14 Bravais lattices, with lattice types, i.e.- body-
centered, etc., symbols and angles within the lattices defined. Note that
in trigonal- and hexagonal- P symmetries, y = 120 °.
 1.4 The structure of solids 23

TABLE 1-2
The Fourteen f 14) Bravais Lattices in Three Dimensions
RESTRICTIONS
LATTICES ON UNIT CELL
SYSTEM IN SYSTEM LATTICE SYMBOLS AXES OR ANGLES

CUBIC 3 P (primitive) a = b = c
I (body - centered) a = p = y = 90°
F(face - centered)

HEXAGONAL 1 P a=b *c

(a = |3 = 90 ° ; Y = 120 °)

TETRAGONAL 2 P, I a=b * c
<x= p = Y = 9 0 °)

TRIGONAL 1 R a=b=c
(Rhombohedral) a = p= y < 1 2 0 °
90 ° > y < 120 °

ORTHORHOMBIC 4 P,C a *b #c
I ,F a = p = Y = 90°

MONOCLINIC 2 P, C a*b*c
a = y= 90°* p
TRICLINIC 1 P a*b *c
a *p* Y

In 1965, Weinreich developed a set of figures which illustrate the 32

point-group symmetries. These are shown in 1.4.16., given on the
following page, along with appropriate Schoenflies symbols.

These are a type of short-hand used to indicate the number and type of
symmetry elements present in a given point-group. In this nomenclature,
the elements are defined as shown in 1.4.17. on afollowing page.
24 Introduction to the solid state

1.4.16.

There is one other factor contributing to the overall symmetries of the

lattice structure. This factor is that of the local symmetry of the atomic
 1.5 Determination of structure of compounds 25

1.4.17.- C = rotation axis only

D = dihedral (rotation plus dihedral rotation axes)
I = inversion symmetry
T = tetrahedral symmetry
O = octahedral symmetry

groups which actually form the structure. Examples are the "solid-state
building blocks" given above, i.e.- the tetrahedron- like group, PO4.3- ,
and the octahedron- like NbO6". It is easy to see that if a structure is
composed of such building blocks, they will impose a local structural
symmetry on the lattice, in addition to the other symmetries already
present. The result is that Factor III of 1.4.11. imposes further symmetry
restrictions on the 32 point groups and we obtain a total of 231 space
groups. We do not intend to delve further into this aspect of lattice
contributions to crystal structure. It is sufficient to know that they exist.
Having now covered the essential parts of lattice structure, we will
elucidate how actual structure is determined for a given solid.

1.5. - DETERMINATION OF STRUCTURE OF COMPOUNDS

We want to review how one goes about actually determining the structure
of a given solid. There are two factors we need to consider: 1) what steps
are required in actually determining a structure?; 2) what kind of
information do we obtain?

A crystal is a periodic array of atoms in which the interatomic distances

and interplanar spacing are of the same order of magnitude as the
wavelengths of the readily available x-rays, e.g.- the K a radiations : Mo =
0.711 A ; Cu= 1.5418 A ; and Cr = 2.291 A. A crystal therefore acts as a
three-dimensional diffraction grating for x-rays, and three equations (the
Laue equations) must be satisfied if there is to be constructive
interference of monochromatic x-rays. The Laue equations are:

1.5.1.- a(a-ao)=hX; b (p-p 0 ) = kX ; c (y - yo ) = !*•

26 Introduction to the solid state

where a, b and c are the repeat distances of the lattice, a and ao (etc.),
are the direction cosines for the diffracted and incident beams
respectively, and h, k, and 1 are integers defining the order of the
particular diffracted beam. W. L. Bragg showed that the above equations
are equivalent to the condition for reflection of the x-rays by the plane
with indices hkl, namely:

1.5.2.- nX = 2 d sine

8 is the angle between the incident (or reflected) beam and the plane,
which we define using hkl values. Here, d is the perpendicular distance
between successive planes. In structure and formulation determinations
of a given compound, we proceed as follows. First of all, using x-ray
diffraction equipment available, one obtains a series of diffraction lines.
One then calculates the value of d, the distance between adjacent planes
in the crystal lattice by using the Bragg Equation. The intensities are read
from the diffraction chart, scaled to the most intense line. If we do not
know what the nature of the material is, then we use the obtained pattern
to determine composition.

Using the set of diffraction lines, we pick the three most intense lines in
the pattern. One example is shown in 1.5.3. for LaA^O^. We wish to

identify these lines from the pattern and then to identify the compound.

1.5.3.- Diffraction Pattern of a Lanthanum Aluminate

By referring to the "POWDER DIFFRACTION FILE", published by the

 1.5 Determination of structure of compounds 27

AMERICAN SOCIETY FOR TESTING AND MATERIALS 1916 Race St.,

Philadelphia, Penna. 19103, we can look up the most probable
composition. This turns out to be LaAlnO18. Usually, we will know how
the material was made and the components used to make it. If not, we
can analyze for constituents. In this case, we would find La and Al, and
would surmise that we have an oxidic compound. We find that there are
two compounds possible, viz.- LaAlO3 and LaAli 1O18 • However, the 3
most intense lines of LaAlO3 are:

1.5.4.- d = 2.66 A(100), 3.80 A(80) and 2.19 A(80).

This does not fit our pattern. We do find that the pattern for LaAli 1O18
is identical to the x-ray pattern we obtained.

Using 1.5.3., d-values can then be calculated from the 29 values using the
Bragg equation, for the hexagonal composition, LaAlnO18 . The {h,k,l}
values can be calculated from special formulas developed for this purpose.
These are given in Table 1-3, presented on the next page. These values
are obtained by trying certain values in the hexagonal formula, and seeing
if the results conform. What this means is that we substitute Miller
Indices into the formula and see if the calculated value matches 1/d2.
Once this is done, we have characterized our material. Because of the
physical geometry of the x-ray diffraction goniometer (angle-measuring
device), one obtains values of 26 directly.

Note that the equations in Table 1-3 for the high symmetry lattices are
rather simple while those for low symmetry lattices are complicated.
Since we know that the compound, LaAlnO18 is hexagonal, we use the
equation:

1.5.5.- l/d2 = 4/
3 [(h 2 + k 2 + l 2 )/ a 2 ]

If we do this, we obtain the data in Table 1-4 also shown on the next
page. This allows us to determine the unit cell lengths for our compound
as:
Hexagonal: s^ = 5.56. A ; b 0 = 22.04 A
28 Introduction to the solid state

Table 1- 3
Plane Spacings for Various Lattice Geometries

CUBIC TETRAGONAL
1 / d 2 = h 2 + k2 + 12/ a 2 1/ d2 = h2 + k 2 / a 2 + l2 / c
2

HEXAGONAL ORTHORHOMBIC
1/d 2 = 4 / 3 [(h2 + k2 + 12)/ a 2] 1/d2=(h2/a2)+(k2/D2)+(i2/c2)+ i2/c2

RHOMBOHEDRAL
1/d 2 = f h 2 + k 2 + j2) S j n 2 a + 2fhk + kl + hlUcos 2 a - cos a 1
a 2 (1 - 3 cos 2 a + 2 cos^ a )

MONOCLINIC
1/d 2 = { 1/sin 2 b}{ h 2 /a 2 + (k 2 sin 2 b) / b 2 + I 2 / c 2 - 2hl cos b / a c }

TRICLINIC
1/d 2 =1/V2{ S n h 2 + S22 k 2 + S33 I 2 + 2 S12 hk + 2 S23 kl + 2 S13 hi

where: V = volume of unit cell ; S n = b 2 c 2 sin 2 a ; S22 = a 2 c 2

sin 2 b ; S33 = a 2 b 2 sin 2 g ; S12 = abc 2 (cos a cos b - cos g ) ;
S23 = a 2 be (cos b cos g - cos a) ; S13 = ab 2 c ( cos g cos a - cos b )

TABLE 1 -4
D
Coiwersion o f 26 Values of the Diffraction ] attern to {hkl} Valiies
26 I/Io d {hkl} 2q I/Io d {hkl}
8.02 16 11.02 002 36.18 74 2.48 114
16.1 6 4.81 004 39.39 27 2.29 023
18.86 32 4.71 001 40.94 29 2.20 0010
20.12 28 4.41 012 42.79 66 2.11 025
22.07 10 4.03 013 45.01 46 2.01 026
24.23 21 3.67 001 53.36 15 1.72 029
24.55 11 3.63 014 58.57 36 1.58 127
32.19 44 2.78 110 60.07 60 1.54 0211
32.50 15 2.76 008 67.35 48 1.39 220
33.23 15 2.70 112 71.61 18 1.32 0214
34.02 100 2.64 017 95.42 14 1.04 2214
 1.5 Determination of structure of compounds 29

Additionally, we can list the x-ray parameters and convert them to

structural factors as shown. These values are averaged over all of the
reflections used for calculation. Note that this pattern has several planes
where the Hd" value is more than ten.

Let us consider one other example. Suppose we obtained the following set
of 28 values and intensities for a compound:

TABLE 1-5
DIFFRACTION LINES AND INTENSITIES OBTAINED

Intensity 2 6 in degrees
96 29.09
100 33.70
56 48.40
50 57.48
14 60.28
18 78.38
10 90.55
14 80.78
17 97.80
19 118.26
11 121.00
7 144.34
9 148.49

The first three values are the strongest diffraction lines. After calculating
"d" values and looking up the set of strong lines which correspond to our
set, we find that the probable compound is CdO2 , or cadmium peroxide.
This compound turns out to be cubic in structure, with ao = 5.313 A .
When we calculate the {h,k,l} values of the diffracting planes, the
strongest line is found to be {200}. We can then make the determination
that since Cd 2+ is a strongly diffracting atom (it has high atomic weight,
which is one way of stating that it has many electron shells, i.e.- I s 2 2s 2
2p 6 3s 2 3p 6 3d!0 4s 2 4p 6 4d 1 0 , the structure is probably face-centered
30 Introduction to the solid state

cubic. Indeed, this turns out to be the case. In the unit cell, Cd atoms are
in the special positions of : {0,0,0}, {1/2,1/2,1/2}; 0,1/2.1/2};
{1/2,1/2,0}. There are four molecules per unit cell. We could continue
further and calculate intensities, I c , using atomic scattering factors
already present in prior literature. We would scale calculated intensities
to our observed intensities by S I c = S I o . We then calculate a reliability
factor, called R, from R = S (Io - Ic ) /S Io. Alow value indicates that our
selection of lattice parameter was correct. If not, we choose a slightly
different value and apply it. The details of the procedure for determining
exact structure and atomic positions in the lattice are well known, but are
beyond the scope of this Chapter. However, for your own edification, I
request that you index the lines (calculate the {hkl} values) given above,
similar to that presented in Table 1-4.

Summarizing to this point, we have shown that only certain propagation

units can be stacked to infinity to form close-packed solids. We have also
shown how the units fit together to form specific solids with specific
symmetries. The nature of the structure of solids has also been reviewed
in some detail. Now let us look at the solid from the standpoint of
stacking and stacking defects.

1.6. -THE DEFECT SOLID

We have shown that by stacking atoms or propagation units together, a

solid with certain symmetry aspects results. If we have done this
properly, a perfect solid should have resulted with no defects in it. Yet,
defects are related to the entropy of the solid, and a perfect solid would
violate the second law of thermodynamics. This law states that zero
entropy is only possible at absolute zero temperature. Thus, most of the
solids that we encounter are defect-solids. It is natural to ask concerning
the nature of these defects. It should be obvious that most of them will be
stacking defects or faults. Nature (and indeed we in our propagation unit
example as well) finds it impossible to stack atoms (molecules) in perfect
order to infinity. Moreover, even if we could obtain a perfect solid, it
would likely be non-reactive and would be singularly stable. The reason
 1.6 The defect solid 31

for this is that it is the defects in solids which give them special
properties.

Consider the surface of a solid. In the interior, we see a certain symmetry

which depends upon the structure of the solid. As we approach the
surface from the interior, the symmetry begins to change. At the very
surface, the surface atoms see only half the symmetry that the interior
atoms do (and half of the bonding as well). Reactions between solids take
place at the surface. If there were some way to complete the symmetry of
the surface atoms, then they too would likely be nearly non-reactive.

In a three-dimensional solid, we can conceive of three major types of

defects, one-, two- and three- dimensional in nature. These are called:
point, line (edge) and volume (plane), respectively. Point defects are
changes at atomistic levels, while line and volume defects are changes in
stacking of groups of atoms (molecules). An easy way to visualize point
defects is shown in the following:
1.6.1.-

A bubble raft is made by creating bubbles in a soap solution which float to

the surface of the liquid (in this case, water) to create a raft. The trick is
32 Introduction to the solid state

to get the size of the bubbles all the same. If not, one does not get a way
to compare the effects of close packing using bubbles to simulate atom
positions in a lattice structure. This problem was worked on until the
investigator was successful.

The first thing one notices is that packing defects have a significant effect
on the close packing of a bubble raft. Two types of defects can be seen.
One is a "vacancy", that is, the bubble supposed to be there, is missing.
The other is an "impurity", here as a larger bubble. Note its effect on the
degree of ordering. Again, lattice compensation is the norm and the close
packing is compensated by an adjustment in the lattice. Note that most of
the bubbles are all the same and are hexagonally-close packed. You can
imagine the effect on stacking in the lattice of a very impure solid where
the compound is only 95% pure.

In the following is another view of defects:

1.6.2.-

In this diagram, we see three types of point defects. In addition to the

vacancy, we also see two types of substitutional defects. Both are direct
substitutions in the "lattice", or arrangement of the atoms. One is a
smaller atom, while the other is larger than the atoms comprising the
lattice. Note the difference, due to size of the impurity, upon the ordering
 1.6 The defect solid 33

in the hexagonally close-packed lattice. In addition, we have also included

the "interstitial" atom, that is, one that is able to insinuate itself into the
interstices of the lattice. This completes the list of possible point defects
in a mono-atomic lattice. In the next Chapter, we will show other point
defects possible when both cation and anion atom-species are present in
the lattice. In the following diagram, we see the effect of an edge
dislocation upon hexagonal close packing:

1.6.3.-

At "1" in 1.6.3., the hexagonal packing has changed to that of cubic close-
packing, causing a line defect. At "2", the line defect is caused by lack of
ordering along the plane. This may also be regarded as a series of point
defects affecting ordering. What has happened is that many defect-
vacancies have congregated and have caused a boundary between the
"grains". Within the grain-boundary, the ordering is uniform but differs
from the next grain. It is actually a line imperfection (2-dimensional), or
an edge imperfection.

A better perspective of an edge dislocation is shown in 1.6.4., presented

on the next page.
34 Introduction to the solid state

1.6.4 -

Note that in its simplest form, an edge dislocation is an omission of a line

of atoms composing the lattice. Another perspective is presented in the
following, shown as 1.6.5. on the next page.

In this diagram, the area within the lattice around the line defect is under
both compression and tension due to the difference in atom-density as
one passes through it in a direction perpendicular to the line defect.
Thus, there is excess energy in the lattice due to the compressive-tensile
forces present. This has the effect of causing the edge dislocations to
propagate through the solid by a hopping motion until they reach the
surface of the solid. There have been several cases where it has been
possible to directly observe line imperfections by suitable preparation and
microscopic examination of the surface in reflected light. One example is
the MgO crystal. MgO is a cubic crystal and it is possible to etch it along
the {100} direction (this is the direction along the x-axis). What is
observed is a series of surface lines of specific length. The reason that
 1.6 The defect solid 35

these defect lines show up is that they are more easily etched by acid
along the direction of the edge dislocation.

The volume defect is somewhat more difficult to visualize. In the

following figure, given as 1.6.6. on the next page, a "screw dislocation" is
shown, so-called because of its topography.

This is an actual representation of a paraffin single crystal. The top-view

shows the edges of the diamond-shaped crystal which are shaded for
better clarity. Aside-view is also shown. At the bottom of 1.6.6. is shown
how the dislocation line (line defect) becomes a screw-dislocation. If the
growth pattern is spiral, there is an offset "b". The length of the spiral is,
of course related to "r", the radius of growth, and to dr, the change in
offset radius of growth. Referring back to 1.6.4., one sees how the line
defect is related to the spiral growth pattern, since the line defect
propagates to the surface where the actual crystal growth takes place.
Thus, the volume defect is one which is generated from the interior to
the surface of the crystal.
36 Introduction to the solid state

It is the intrinsic defects that have the most interest for us, since they
affect the chemical properties of the solid while extrinsic defects have
little effect. Extrinsic defects are the proper study for those interested in
the mechanics of solids, particularly metals. A broad variety of
 1.6 The defect solid 37

commercial products are based upon controlled-point-defects. These

include transistors, integrated circuits, photosensors, color-television,
fluorescent lamps, just to name a few. None of these would be possible
without point defects. It is for this reason that we devote the next
chapter to the point defect.

SUGGESTED READING

1. G. Weinreich - "Solids-Elementary Theory for Advanced Students"- , J.

Wiley & Sons, Inc., New York (1965).

2. A.J. Dekker - "Solid State Physics" - , Prentice-Hall, Inc., Englewood

Cliffs, NJ (1958). (See Chaps. 1,2 &3 in particular).

3. J.M. Honig, "Imperfections in Crystal", J. Chem. Ed., 34, 224 (1957).

4. J.A. McMillan, "Stereographic Projections of the Colored

Crystallographic Point Groups", Am. J. Pfiys., 35, 1049 (1967).
This page is intentionally left blank
 The Point Defect

There are two types of defects associated with phosphors. One involves
controlled point defects in which a foreign activator cation is
incorporated in the solid in defined amounts. The other involves line and
point defects inadvertently formed in the solid structure because of
impurity and entropy effects. This chapter will define and characterize
the nature of all of these point defects in the solid, their thermodynamics
and equilibria. It will become apparent that the type of defect present will
depend upon the nature of the solid in which they are incorporated. That
is, the characteristics of the point defects in a given phosphor will
depend upon its chemical composition. Of necessity, this chapter is not
intended to be exhaustive, and the reader is referred to the many
treatises concerned with the point defect.

2.1. -TYPES OF POINT DEFECTS

Let us now consider the defect solid from a general perspective. Consider
the case of semi-conductors, where most of the atoms are the same, but
the total of the charges is not zero. In that case, the excess charge (n- or
p- type) is spread over the whole lattice so that no single atom, or group
of atoms, has a charge different from its neighbors. However, most
inorganic solids are composed of charged moieties, half of which are
positive (cations) and half negative (anions). The total charge of the
cations equals, in general, that of the anions. If an atom is missing, the
lattice readjusts to compensate for this loss of charge. If there is an extra
atom present, the charge-compensation mechanism again manifests itself.
Another possibility is the presence of an atom with a charge larger or
smaller than that of its neighbors. In a given structure, cations are usually
surrounded by anions, and vice-versa (Remember what we said in Chapter
1 wherein it was stated that most structures are oxygen-dominated).
Thus, a cation with an extra charge needs to be compensated by a like
anion, or by a nearest neighbor cation with a lesser charge. An example of

39
40 The point defect

charge-compensation for a divalent cation sub-lattice would be the

following defect equation: :

2.1.1.- 2M2+ H M+ + M3+

where the M 3+ and M+ are situated on nearest neighbor cation sites,

which were originally divalent.

Thus, the charge compensation mechanism represents the single

most important mechanism which operates within the defect solid.

Because of this, the number and types of defects, which can appear in the
solid, are limited. This restricts the number of defect types we need to
consider, in both elemental (all the same kind of atom) and ionic lattices
(having both cations and anions present). We have shown that by stacking
atoms or propagation units together, a solid with specific symmetry
results. If we have done this properly, a perfect solid should result with
no holes or defects in it. Yet, the 2nd law of thermodynamics demands
that a certain number of point defects (vacancies) appear in the lattice. It
is impossible to obtain a solid without some sort of defects. A perfect solid
would violate this law. The 2nd law states that zero entropy is only
possible at absolute zero temperature. Since most solids exist at
temperatures far from absolute zero, those that we encounter are defect-
solids. It is natural to ask what the nature of these defects might be,
particularly when we add a foreign cation (activator) to a solid to form a
phosphor.

Consider the surface of a solid. In the interior, we see a certain symmetry

which depends upon the structure of the solid. As we approach the
surface from the interior, the symmetry begins to change. At the very
surface, the surface atoms see only half the symmetry that the interior
atoms do. Reactions between solids take place at the surface. Thus, the
surface of a solid represents a defect in itself since it is not like the
interior of the solid.

In a three-dimensional solid, we can postulate that there ought to be

 2.1 Types of point defects 41

three major types of defects, having either one-, two- or three-

dimensions. Indeed, this is exactly the case found for defects in solids, as
we briefly described in the preceding chapter. We have already given
names to each of these three types of defects. Thus a one-dimensional
defect of the lattice is called a "point" defect, a two-dimensional defect a
"line" or "edge" defect and a three-dimensional defect is called a "plane"
or "volume" defect. We have already described, in an elementary way, line
and volume defects and will not address them further except to point out
how they may arise when certain point defects are present. It is sufficient
to realize that they exist and are important for anyone who studies
homogeneous materials such as metals.

Point defects are changes at atomistic levels, while line and volume
defects are changes in stacking of planes or groups of atoms (molecules)
in the structure. The former affect the chemical properties of the solid
whereas the latter affect the physical properties of the solid. Note that the
arrangement (structure) of the individual atoms (ions) are not affected,
only the method in which the structure units are assembled. That is, the
structure of the solid remains intact in spite of the presence of defects.
Let us now examine each of these defects in more detail, starting with the
one-dimensional lattice defect and then with the multi-dimensional
defects. We will find that specific types have been found to be associated
with each type of dimensional defect which have specific effects upon the
stability of the solid structure. It should be clear that the type of point
defect prevalent in any given solid will depend upon whether it is
homogenous (same atoms) or heterogeneous (composed of differing
atoms).

I. The Point Defect in Homogeneous Solids

We begin by identifying the various defects which can arise in solids and
later will show how they can be manipulated to obtain desirable properties
not found in naturally formed solids. Let us look first at the homogeneous
type of solid. We will first restrict our discussion to solids which are
stoichiometric, and later will examine solids which can be classified as
"non-stoichiometric", or having an excess of one or another of one of the
42 The point defect

building blocks of the solid. These occur in semi-conductors as well as

other types of electronically or optically active solids.

Suppose you were given the problem of identifying defects in a

homogeneous solid. Since all of the atoms in this type of solid are the
same, the problem is somewhat simplified over that of the heterogeneous
solid (that is- a solid containing more than one type of atom or ion). After
some introspection, you could speculate that the homogeneous solid could
have the following types of point defects:

2.1.2.- Types of Point Defects Expected in a Homogeneous Solid

0 Vacancies * Substitutional Impurities

* Self-interstitial % Interstitial Impurities

On the left are the two types of point defects which involve the lattice
itself, while the others involve impurity atoms (Note that interstitial atoms
can involve either an impurity atom or the same atom that makes up the
lattice structure itself). Indeed, there do not seem to anymore than these
four, and indubitably, no others have been observed. Note that we are
limiting our defect family to point defects in the lattice and are ignoring
line and volume defects of the lattice. These four point defects, given
above, are illustrated in the following diagram, given as 2.1.3. on the next
page.

Note that what we mean by an "interstitial" is an atom that can fit into the
spaces between the main atoms in the crystalline array. In this case, we
have shown a hexagonal lattice and have labeled each type of point defect.
Observe that we have shown a vacancy in our hexagonal lattice, as well as a
foreign interstitial atom which is small enough to fit into the interstice
between the atoms of the structure. Also shown are two types of
substitutional atoms, one larger and the other smaller than the atoms
composing the principal hexagonal lattice. In both cases, the hexagonal
packing is disrupted due to a "non-fit" of these atoms in the structure.
Additionally, we have illustrated another type of defect that can arise
 2.1 Types of point defects 43

within the homogeneous lattice (in addition to the vacancy and

substitutional impurities that are bound to arise). This is called the "self-
interstitial". Note that it has a decisive effect on the structure at the
defect. Since the atoms are all the same size, the self-interstitial
introduces a line-defect in the overall structure. It should be evident that
the line-defect introduces a difference in packing order since the close
packing at the arrows has changed to cubic and then reverts to hexagonal
in both lower and upper rows of atoms. It may be that this type of defect
is a major cause of the line or edge type of defects that appear in most
homogeneous solids. In contrast, the other defects produce only a
disruption in the localized packing order of the hexagonal lattice, i.e.- the
defect does not extend throughout the lattice, but only close to the
44 The point defect

specific defect. It should be evident that metals or solid solutions of

metals (alloys) show such behavior in contrast to heterogeneous lattices
which involve compounds such as ZnS. This accounts for the tremendous
discrepancy between theoretical and actual strength of certain alloys in
practical applications due to "fatigue" failure when the object is being
used.

Now, suppose that we have a solid solution of two (2) elemental solids.
Would the point defects be the same, or not? An easy way to visualize such
point defects is shown in the following diagram:

2.1.4.- Defects in the Homogeneous Solid Containing 2 Solids in Solution

Here, we use a hexagonally-packed representation of atoms to depict the

close-packed solid. Both types of homogeneous solids are shown, where
one solid is composed of the same sized atoms while the other is
composed of two different sized atoms. On the right are the types of point
 2.1 Types of point defects 45

defects that could occur for the same sized atoms in the lattice. That is,
given an array of atoms in a three dimensional lattice, only these two types
of lattice point defects could occur where the size of the atoms are the
same. The term "vacancy" is self-explanatory but "self-interstitial" means
that one atom has slipped into a space between the rows of atoms. In a
lattice where the atoms are all of the same size, such behavior is
energetically very difficult unless a severe disruption of the lattice occurs
(usually a "line-defect" results). This behavior is quite common in certain
types of homogeneous solids. In a like manner, if the metal-atom were to
have become misplaced in the lattice and were to have occupied one of
the interstitial positions, as shown in the different sized atom solid (see
2.1.4.) then the lattice is disrupted by its presence at the interstitial
position. This type of defect has also been observed. Note that the atoms
are usually not charged in the homogeneous lattice.

Summarizing, three types of point defects are evident in a homogeneous

lattice. In addition to the Vacancy, two types of substitutional defects can
also be delineated. Both are direct substitutions in the "lattice", or
arrangement of the atoms. One is a smaller atom, while the other is larger
than the atoms comprising the lattice. In both cases, the lattice
arrangement affects the hexagonal ordering of the lattice atoms around it.
The lattice packing is seen to be affected for many lattice distances.

It is for this reason that compounds containing impurities sometimes have

quite different chemical reactivities than the purest ones. However, the
interstitial impurity does not affect the lattice ordering at all. Now, let us
look at the heterogeneous lattice

II. The Point Defect in Heterogeneous Solids

The situation concerning defects in heterogeneous inorganic solids is

similar to that given above, except for one very important factor, that of
charge on the atoms. Covalent inorganic solids are a rarity while ionicity
or partial ionicity seems to be the norm. Thus, heterogeneous solids are
usually composed of charged moieties, half of which are positive (cations)
and half negative (anions). In general, the total charge of the cations will
46 The point defect

equal that of the anions (Even in the case of semi-conductors, where the
total of the charges is not zero, the excess charge (n- or p- type) is spread
over the whole lattice so that no single atom, or group of atoms, ever has a
charge different from its neighbors. Note also that most of the semi-
conductors that we use are homogeneous in nature, modified by
homogeneous additions to form p- or n-type electrically charged areas). In
a given structure, cations are usually surrounded by anions, and vice-versa.
Because of this, we can regard the lattice as being composed of a cation
sub-lattice and an anion sub-lattice. (Remember what was stated in
Chapter 1 concerning the fact that most structures are oxygen-
dominated). What we mean by a "sub-lattice" is illustrated in the following
diagram:

2.1.5.-

In this case, we have shown both the cubic cation and anion "sub-lattices
separately, and then the combination. It should be clear that all positive
charges in the cation sub-lattice will be balanced by a like number of
negative charges in the anion sub-lattice, even if excess charge exists in
one or the other of the sub-lattices. If an atom is missing, then the overall
lattice readjusts to compensate for this loss of charge. If there is a
different atom present, having a differing charge, the charge-
compensation mechanism again manifests itself. Thus, a cation with an
extra charge needs to be compensated by a like anion, or by a nearest
neighbor cation with a lesser charge. An example of this type of charge-
 2.1 Types of point defects 47

compensation mechanism for a divalent cation sub-lattice would be the

following defect equation:

2.1.6.- 2Ca 2 + % Li+ + Sb 3+

where the Sb 3+ and Li+ are situated on nearest neighbor cation sub-lattice
sites, in the divalent Ca2+ sub-lattice. Note that a total charge of 4+ exists
on both sides of the above equation.

Thus, the charge compensation mechanism represents the single most

important mechanism which operates within the defect ionic solid.

We find that the number and types of defects, which can appear in the
heterogeneous solid, are limited because of two factors:

1) The charge-compensation factor

2) The presence of two sub-lattices in the ionic solid.

These factors restrict the number of point defect types we need to

consider in ionic heterogeneous lattices (having both cations and anions
present). For ionic solids, the following types of defects have been found
to exist:

* Schottky defects (absence of both cation and anion)

* Cation vacancies
* Anion vacancies
* Frenkel defects (Cation vacancy plus same cation as
interstitial)
* Interstitial impurity atoms (both cation and anion)
* Substitutional impurity atoms(both cation and anion)

These defects are illustrated in the following diagram, given as 2.1.7. on

the next page.

Note that, in general, anions are larger in size than cations due to the
extra electrons present in the former. A hexagonal lattice is shown in
48 The point defect

2.1.7. with both Frenkel and Schottky defects, as well as substitutional

defects. Thus, if a cation is missing (cation vacancy) in the cation sub-
lattice, alike anion will be missing in the anion sub-lattice. This is known
as a Schottky defect (after the first investigator (1935) to note its
existence).

In the case of the Frenkel defect, the "square" represents where the
cation was supposed to reside in the lattice before it moved to its
interstitial position in the cation sub-lattice. Additionally, "Anti-Frenkel"
defects can exist in the anion sub-lattice. The substitutional defects are
shown as the same size as the cation or anion it displaced. Note that if
 2.2 The plane net 49

they were not, the lattice structure would be disrupted from regularity at
the points of insertion of the foreign ion.

To summarize, the categories of point defects possible for these two types
of lattices are illustrated are:

1). In an elemental solid, we may have:

^ Vacancies
* Self-interstitial
^ Interstitial Impurities
# Substitutional Impurities

2). In the ionic solid, i.e.- heterogeneous solid, we may have:

$fc Schottky defects (absence of both cation and anion)

3fc Cation or anion vacancies
0 Frenkel defects (Cation vacancy plus the same cation as
interstitial)
Sfc Interstitial impurity atoms (both cation and anion)
0 Substitutional impurity atoms(both cation and anion)

All of these point defects are intrinsic to the solid. The factors
responsible for their formation are entropy effects (point defect faults)
and impurity effects. At the present time, the highest-purity materials
available still contain about 1.0 part per billion of various impurities, yet
are 99.9999999 % pure. Such a solid will contain about 10 1 4 impurity
atoms per mole. So it is safe to say that all solids contain impurity atoms,
and that it is unlikely that we shall ever be able to obtain a solid which is
completely pure and does not contain defects.

2.2. - THE PLANE NET

Now, let us consider how such defects arise in any given solid. The easiest
way to visualize how intrinsic defects occur in the solid is to study the
PLANE NET. Imagine that we have M as cations and X as anions (we shall
50 The point defect

ignore formal charge for the moment). This is shown in the following
diagram:

It is easy to see that we can stack a series of these "NETS" to form a

three-dimensional solid. Note that we have used the labeling: V = vacancy;
i = interstitial ; m = cation site; x= anion site and s = surface site. One
might think that perhaps V"m and V+x ought to be included in our list of
vacancies. However, a negatively-charged cation vacancy alone, particularly
when it is surrounded by negative anions, would not be very stable.
Neither should a positively-charged anion vacancy be any more stable.
Either arrangement would require high energy stabilization to exist.
Therefore, we do not include them in our listing.

However, a V m could capture a positive charge to become V+m and

likewise for the Vx which then becomes a V' x Both of these are stable
when surrounded by oppositely charged sites. We have already stated that
surface sites are special. Hence, they are included in our listing of
intrinsic defects. The same criteria apply concerning charge on the
defect.

For the plane net, we can expect the following types of intrinsic defects:
 2.2 The plane net 51

2.2.2.- Vacancies Charged Particles

V M ,V X> V + M , V X e- , ;r+

Interstitials Surface Sites

Ms , Xs , M+, , X s

The positive hole requires more explanation. 7r+ is the electronic

equivalent of the electron, e- , in solids and they annihilate each other
upon reaction:

2.2.3.- e" + 7T+ = energy.

The following diagram shows how the positive hole can exist in the solid:

2.2.4-

Here, we represent Ge which has 32 electrons. In the solid, it forms

4(sp)3 hybrid bonds as the semi-conductor. Ga has only 31 electrons and
can form but three bonds. If it is incorporated into the structure, the
52 The point defect

lattice responds by forming a positive electronic "hole", i.e.- 7r+.

Still another type of electronic defect is the color-center, shown in the

following diagram:
2.2.5.-

In order to compensate for the loss of negative charge, the lattice has
captured an electron and charge-compensation has resulted at the anion
vacancy. This combination is called an "F-center". This is a special case
where an electron is localized at the vacancy and is optically active. That
 2.2 The plane net 53

is, it absorbs light within a well-defined band and is called a color-center

since it imparts a specific color to the crystal. Note that the peak of the
band changes as the size of the cation in the alkali halides increases.
There appears to be an inverse relation between the size of the cation
(actually, the polarizability of the cation) and the peak energy of the
absorption band of these F-centers.

We can now proceed to write a series of defect reactions for our plane net
in terms of specific defects, including that of the F-Center, where we use
6 as a small fraction:

2.2.6.- SCHOTTKY : MX = M i . 5 X i . j + 6 V m + 6 V x
FRENKEL : MX = M i . 6 X + 6 Mi + SVm
ANTI-FRENKEL: MX = MXl-8 + 8 V x + SX,
F - CENTER: MX = MX1-+ {Vx/e"} + 6 / 2 X2 *

These equations are valid for the defects in a heterogeneous lattice where,
for the F-Center, the brackets enclose the complex consisting of an
electron captured at an anion vacancy, i.e.- {Vx/e~}. Actually, the formation
of an F-center is more complicated than this. A more complete
explanation is given as follows.

It is well known that F-centers can be formed by exposing the NaCl crystal
to sodium metal vapor. The following defect reactions take place:

2.2.7.- Na° = Na+ + e-

Na+ + { Na+, Cl-} = 2 {Na+ [ Cl- , V c l ]}

V
C1 + e- = [Vci/e-I

These equations illustrate how the crystal responds to the presence of

sodium vapor, i.e.- excess Na+ , by forming anion vacancies, to form the F-
center. However, if one anion vacancy can be formed which captures an
electron which is optically active, then you might think that more than
54 The point defect

one anion-vacancy complex might be possible. Indeed, this is the case.

This is illustrated in the following diagram:

2.2.8.-

A better representation of the M-center is shown in the following

diagram:

2.2.9.-

Under certain precise experimental conditions, two anion sites can be

conjoined (in the salt-structure, along the {1,1,0} plane) to form what is
 2.2 The plane net 55

termed the "M-Center" as shown above in the two diagrams, 2.2.8. and
2.2.9.

The optical properties of these two types of centers are given in the
following diagram:

2.2.10.-

The defect equation for formation of the M-center is also given as follows:

2.2.11.- M-center: KC1 = KCli-6 + 5/2 CI2 • + 6/4 [ V'ci I V'Ci ]

Note that each vacancy in 2.2.9. has captured an electron, in response to

56 The point defect

the charge-compensation mechanism which is operative for the defect

reactions. These associated, negatively-charged, vacancies have quite
different absorption properties than that of the F-center.

There are other impurity systems to which this notation can be applied.
For the case of an AgCl crystal containing the Cd2+ cation as an impurity,
we have:

2.2.12.- 2Ag+ % [Cd 2 +,VAg]

This is an example of a heterotype system. Another such system is CdCl2

containing Sb 3+ .

Here, we can write at least three different equations involving defect

equilibria:

2.2.13.- 2Cd 2 + % Sb3+ + p+ + Vcd

2Cd 2 + % Sb3+ + V + C d
2 Cd 2+ % Sb3+ + Li+

In the last equation, charge-compensation has occurred due to inclusion

of a monovalent cation. All of these equations are cases of impurity
substitutions.

Another type is the so-called homotype impurity system. The substance,

nickelous oxide, is a pale-green insulator, when prepared in an inert
atmosphere. If it is reheated in air, or if a mixture of NiO and Li+ is
reheated in an inert atmosphere, the NiO becomes a black semi-
conductor. This is a classical example of the effect of defect reactions
upon the intrinsic properties of a solid:

2.2.14.- 2Ni 2 + % [Ni 3 +/VNi] + p+

2 Ni 2+ % Ni3+ + Li+

This behavior is typical for transition metals which easily undergo changes
in valence in the solid state.
 2.3 Defect equation symbolism 57

Up to this point, we have only investigated stoichiometric lattices. Let us

now examine non-stoichiometric lattices in light of our symbolism.
Consider the semi-conductor, Ge. The defect reactions associated with
the formation of p-type and n-type lattices are:

2.2.15.- n-type: Ge + 5As = [Ge/Ass] + 6 e-

p-type: Ge + 8 Ga = [Ge/Gas ] + 5 p+

The excess charges shown are spread over the entire lattice, as stated
before.

2.3. - DEFECT EQUATION SYMBOLISM

Whether you realize it or not, we have already developed our own

symbolism for defects and defect reactions based on the Plane Net. It
might be well to compare our system to those of other authors, who have
also considered the same problem in the past. It was probably Rees
(1930) who wrote the first monograph on defects in solids. Rees used DM
to represent the cation vacancy, as did Libowitz (1974). This has certain
advantages since we can write the first equation in 2.2.12. as:

+ + P+
2.3.1- 2Cd2+ % {^Cd ^ C d }

Likewise, the other equations become:

2.3.2.- 2Cd 2 + % lSfeJ

Cd +
^ C d

and:
+ + P +
2.3.3, 2Cd2 + % ®cd ^ C d

Although the results are equal as far as utility is concerned, we shall

continue to use our symbolism, for reasons which will become clear later.
58 The point defect

The following compares defect symbolism, as used by prior Authors. Note

that our symbolism most resembles that of Kroeger, but not in all aspects.

2.3.4- Reesf 19301 Krreger (19541 Libowitz f 19741

Cation Site Vacancy: Dq^ VM Qto

Anion Site Vacancy DlJc Vx d x

Cation Interstitial AM Mi , M+j Mj

Anion Interstitial Ax Xi , Xi Xj

Negative Free Charge e e- e"

Positive Free Charge p h+ h+

Interstices --- a V. a V.
Unoccupied Interstitial --- i A

Anti-structure Occupation --- --- MM,XX, MX.XM

These prior authors have considered some intrinsic defects that we have
not touched, namely interstices and the so-called "anti-structure"
occupation. The latter deals with an impurity anion on a cation site
coupled with an impurity cation on an anion site, both with the proper
charge.

We have mentioned interstices but not in detail. They appear as a function

of structure. There is one site in a tetrahedron, four in a body-centered
cube, and six in a simple cube. Thus, a in {a Vi} is 1, 4 or 6, respectively.
We shall need this symbol later, as well as Vj , the unoccupied interstitial.

2.4. - SOME APPLICATIONS FOR DEFECT CHEMISTRY

Before we proceed to analyze defect reactions by a mathematical

approach, let us consider two applications of solid state chemistry. We
begin with a description of some phosphor defect chemistry.
 2.4 Some applications for defect chemistry 59

I. - Phosphors

In the prior literature, it was found (Kinney- 1955) that Ca2 P2 O7 could
be activated by Sb 3+ to form the phosphor: Ca2P2O7: Sb.02 (this formalism
actually means a solid-solution of two pyrophosphate compounds, i.e.-
[(Ca.99,Sb.oi)2P207l). The brightness response of this phosphor was
moderate when excited by ultraviolet radiation but was improved four
times by the addition of Li+ . The optimum amount proved to be that
exactly equal to the amount of Sb 3 + present in the phosphor. The defect
reactions occurring were:

2.4.1.- Defect Reactions Occurring in Calcium Pyrophosphate Phosphor

PHOSPHOR BRIGHTNESS

2Ca 2 + %Sb 3 + C a + V+Ca 25%

or (2 Ca2+ % Sb + Ca3
+ v C a + P+)

2 Ca2+ % Sb 3+ ca +
Li+Ca 100

It is well known that phosphor brightness in a phosphor is proportional to

the numbers of activator ions, i.e.- Sb 3+ ions, actually incorporated into
the pyrophosphate structure. Phosphors are prepared by heating the
ingredients at high temperature (> 1000° C.) to obtain a compound having
high crystallinity. The sintering process decreases entropy and is
counterproductive to the formation of vacancies in the pyrophosphate
lattice. In the absence of Li+ , lack of vacancy-formation actually decreases
the amount of Sb 3+ incorporated into activator sites. Apparently, four
times as many activator ions were incorporated into the lattice when the
charge-compensating Li+ ions were present on nearest neighbor sites.

Note that we have written two defect reactions for the case of vacancy
formation in 2.4.1. Pyrophosphate is an insulator and the formation of a
positively-charged vacancy is much more likely than the vacancy plus a
free positive charge.
60 The point defect

THUS, ALTHOUGH MORE THAN ONE DEFECT REACTION MAY BE

APPLICABLE TO A GIVEN SITUATION, ONLY ONE IS USUALLY FAVORED
BY THE PREVAILING THERMODYNAMIC AND ELECTRICAL
CONDITIONS.

II.- Lithium Niobate

Lithium niobate, LiNbO3, is a photorefractive material, discovered in 1966

(Ashkin et al). That is, it is an electro6ptic material in which the indices
of refraction can be changed by an applied electric field. It is used in
optical devices which employ its nonlinear optical and electrooptic
properties. As a single crystal, LiNbO3 has high transparency to
electromagnetic radiation. If a laser beam is directed down the length of
such a crystal, its frequency is doubled. YAG:Nd3+, i.e.- Y3Al5O12:Nd3+, is a
commonly used laser crystal, whose emission lies at 10,600 A. (near-infra-
red radiation). The combination of a LiNbO3 crystal and YAG:Nd3+,
produces a laser beam at 5,300 A (green light). If a second crystal is also
incorporated in the optical setup, radiation at 2,650 A (ultraviolet
radiation) is obtained. Although the radiation frequency changes, so does
the intensity of light produced. Losses of 100 times or more are common.
However, this detrimental factor is overcome by increasing the power of
the laser beam. Fortunately, LiNbO3 has a high resistance to damage by a
laser beam, unlike many other similar crystals. It gains its unique
characteristics because the crystal structure has "built-in" defects.

LiNbO3 is a ferroelectric structure related to that of the cubic perovskite

(CaTiO3) structure. That is, electric polarization of the lattice electrons
occurs upon application of an electric field. Although a voltage will induce
such polarization, so will the electric vectors of a beam of light,
particularly that of a laser beam. However, LiNbO3 consists of distorted
oxygen-octahedra sharing faces so that a planar hexagonal arrangement
results. The pile-up of the octahedra along the perpendicular direction, c-
axis, follows the cation sequence "Li, Nb and vacant site". The point
symmetry group is C3 (see 1.4.14. and 1.4.16 of the first chapter) with
the trigonal axis along the cation rows. It is very close to a C3V (3m)
configuration.
 2.4 Some applications for defect chemistry 61

The following diagram shows the arrangement of the three ions in the
LiNbO3 structure.

2.4.2.-

Note that we have a structure with a "built-in" crystal defect, a vacancy.

Both the lithium and niobium cations are in an octahedral coordination. In
fact, the two ions, Li+ and Nb5+, have nearly the same radius and occupy
octahedral sites with the same C3V symmetry. The lithium deficiency in
congruent crystals is accommodated by means of Nb u anti-sites and Nb5*
vacancies in a relative concentration that guaranties overall electrical
neutrality. Note that many physical properties depend upon
stoichiometry, e.g.- Curie temperature, absorption spectra, lattice
parameters and photorefractive yield.

However, it has been found that impurities play a major role in the
operation of a frequency-doubling crystal like LiNbO3. Many impurities
have ionic radii similar to that of lithium. They substitute at Li+ sites
rather than Nb5+ sites (possibly because the NbO4 coordination is stronger
at the niobium site). Among these are: Mn*, Fe34" and Ni*. As we have
already seen, substitution of such multivalent cations on a monovalent site
results in lattice compensation such as oxygen vacancies and the
formation of color-centers at the oxygen vacancies. These interfere with
the photorefractive properties of such defect crystals since the ease of
electric polarization of the lattice is impaired during use. Although , as we
62 The point defect

will see, the crystal-growing process is also a purification process, it is not

able to exclude all of the impurities as the crystal grows.

III. Bubble Memories

"Bubble memory" is the term applied to the device which uses a "soft"
magnetic material to carry information. If a ferromagnetic film such as
europium gallium garnet is grown epitaxially upon a suitable substrate
such as gadolinium gallium garnet, i.e.- Gd3 Gas On (= GGG), it forms
magnetic domains in which the electron spins of the cations are aligned
in the same direction in the same domain. This is shown in the following:

The following diagram, given as 2.4.4. on the next page, illustrates how
these would look under polarized light (the Faraday effect) using crossed
Nicol polarizers. (The black and white parts are domains of opposite
polarity).

When a magnetic field is applied, with the field vector horizontal to the
film, the domains collapse to form separated cylinders within the film, as
shown. These appear to be "bubbles" when viewed from the top, hence
the name. The bubbles then become mobile under the influence of a
separate electric field and will move. Actually, the electric field causes the
domain-wall to collapse by a spin-flip mechanism, while the cylinder
volume is maintained by the magnetic field.
 2.4 Some applications for defect chemistry 63

This causes the apparent movement of the wall. Bubbles densities as high
as 1O7 per cm 2 , and bubble velocities of up to 10 3 -10 4 cm./sec. have been
reported. Obviously, bubble velocity depends upon E, the electric field
strength, as well as the composition of the epitaxially grown film. The use
of a vapor-phase-deposited metal grid upon the surface of the film serves
to switch bubbles from site to site. The presence (1) or absence (0) of a
bubble is detected by polarized light beams. Thus, information can be
stored, and retrieved, on a "chip" in binary language.

However, the epitaxial film must be defect-free. GGGis difficult to grow as

a defect-free single crystal. Frenkel defects appear, and give rise to line
dislocations. When the crystal is cut into wafers, the line dislocations
remain and can be revealed by chemical polishing in hot phosphoric acid.
There is a strong chance that these dislocations will be propagated into
the epitaxial film when it is deposited. The resulting film-defects then
"pin" one or more bubbles to one location on the film, making it (them)
64 The point defect

unusable. In addition, a lattice mismatch tends to accentuate the defects

of the GGG substrate onto the film grown on it. The aj, lattice parameter
for GGG is 12.53 A (this is the length of the side of the cube enclosing the
19 atoms) while that of EU3 Gas ° l l i s 12.48 A. This mismatch of lattice
parameter is about 50x greater than that desired so that a composition
such as Y2.45 EU5.5 Fe3.s Gai.2 O12 is generally more suitable to correctly
match the lattice parameter of GGG. In this way, defects present in the
substrate can be avoided in the epitaxially grown film, if proper growth
conditions are maintained. Nevertheless, although the "bubble-memory"
approach to higher density memories in computers was thought to hold
much promise, the problems associated with intrinsic defects in the GGG
film and those involving the substrate proved to be too daunting and today
the bubble-memory approach has been abandoned in favor of more
advantageous methods.

IV. Calcium Sulfide Phosphor

Lenard in 1928 reported on a calcium sulfide phosphor, i.e.- CaS:Bi3+,

having a blue emission. Actually, he probably prepared: CaS:Bi3+:V"
initially. As you can see, the divalent cation site would have to be occupied
by a negative vacancy since charge compensation dictates that the vacancy
will appear. Lenard found that the emission intensity of CaS:Bi3+:V~ was
very low. It is likely that the V" site is a color center which dissipated
most of the excitation energy. He found that the use of chloride fluxes
greatly improved the emission intensity and that the use of KCl produced
the brightest phosphor. He concluded that K+Ca was a coactivator, i.e.- as
CaS:Bi3+:K+ , since he used first used NaCl as a flux during its preparation
and then KCl. Even today, the phosphor is referred to in that manner
although it is obvious that the charge compensation mechanism is most
likely the correct mechanism in the formation of the phosphor.

This concludes our consideration of defect applications. Let us now

consider a more mathematical approach to the description of point
defects.
It has been said: "If you cannot calculate the properties of any given
theory, you really do not understand it".
 2.5 Thermodynamics of the point defect 65

2.5. -THERMODYNAMICS OF THE POINT DEFECT

We shall use two approaches to derive some working values for the point
defect in solids, namely that of Statistical Mechanics and that of the
Thermodynamics of Defects. There are those who have some familiarity
with statistical mechanics. For others, some explanation is due. Statistical
Mechanics as a discipline was originally derived in the early 1920's when
it was realized that one had to deal with large populations of atoms or
molecules in various energy states (particularly gases at that time). These
states arise because each molecule, for example, is vibrating in a manner
slightly different than its neighbor.

What we describe as the energy state for a given set of conditions is

actually the average of that of a Boltzmann population. The discipline best
suited for handling such a system is statistics, hence the name. The
approach used for manipulating molecular populations in Statistical
Mechanics is quite involved, and we shall touch very briefly on the
mathematics involved. We will first describe each of these approaches
separately and then a combined version. Hopefully, this will aid in your
understanding of the two methods of determining the effect of the point
defect upon the properties of the solid.

I. Statistical Mechanics Approach

The language of Statistical Mechanics evolved over a considerable period

of time. For example, the term "ensemble" is used to denote a statistical
population of molecules; "partition function" is the integral, over phase-
space of a system, of the exponential of {-E/kT} [where E is the energy of
the system, k is Boltzmann's constant, and T is the temperature in °K].
From this "function", all of the thermodynamic functions can be derived.
The definitions that we shall need are given as follows in 2.5.1. on the
next page.

Here, Nj is the sum of the individual atoms times an entropy factor and N
is the sum of all of the N J 's.
66 The point defect

2.5.1.- Statistical Mechanics Definitions Needed

W S thermodynamic probability
G = ensemble of related atoms(molecules)
Nj = 2 m
N = 2NJ
Q = Partition Function

Using these, we can derive the following equation for the total energy of
any given system:

2.5.2.- E T = total energy of system = N E(ave.) = 2 Nj Ej

where Nj is the total number of atoms (molecules) involved. Using the

methods of Statistical Mechanics, we can derive by probability relations:

2.5.3.- W = A l 7 h 3 Nj !
Q = 2 exp- Ej / k T (for states)
Q = a. 2 e x p - E , / k T (for levels)

In these equations, r is a so-called "partition coefficient", co is a

degeneracy, k is Boltzsmann's constant, and T is in degrees Kelvin.
The first equation in 2.5.3. is a statistical mechanical definition of work,
whereas the last two describe total energy states. Having these definitions
and equations allows us to define point defects from a Statistical
Mechanical viewpoint.

II. Schottkv and Frenkel Defects

Consider a plane net having N sites, of which NL are lattice sites, N v are
vacancies, and Ni are Interstitials. (Note that we do not consider charge
at the sites for the moment). Using these, we have the following two
equations:
 2.5 Thermodynamics of the point defect 67

2.5.4.- NL = N + N v - Ni

2.5.5.- 2 Ni = «NL

Equation 2.5.4. holds if some fraction of Ni is associated with NL , the

number of lattice sites, i.e.- Frenkel defects. According to the Binomial
Theorem, we can combine pairs of these sites as:

2.5.6.- W(Nv,Ni) = Ni!/N V !(N L -Nv) !

2.5.7.- Wi(Ni,aNl) = aN L ! / (aNL-Ni) !

Note that these equations are simply the Combinatorial Equation as

applied to these two sets of defects. We can set up a partition function(see
the above definition), using equation 2.5.3. We apply this to the Schottky
and Frenkel defects as examples:

2.5.8.- SCHOTTKY: Qv (T) = Wi(Nv , NL) exp -Nv Ev / kT

2.5.9.- FRENKEL: Qi (T) = Wi(Ni , aNL) exp -Ni E4 / kT

We now solve these equations using certain approximations employed in

Statistical Mechanics, including Stirling's Approximation for large
numbers. If the energy required to form vacancies and Interstitials is
greater than kT, and if the defects are only a fraction of the total number
of lattice sites, then we get:

2.5.10.- SCHOTTKY- Nv = N exp -Ev / kT

2.5.11.- FRENKEL- Ni = a N exp - Et / kT

Note that we now have Boltzmann distribution equations for each type of
defect., and that the energy, E, to form the defect is like an activation
energy. The fraction of defects present , either N v / N or Ni / N , is an
exponential function of this activation energy.
68 The point defect

This leads to the following conclusion:

It is clear without a doubt that both Schottky and Frenkel defects

are thermal in origin.

Although you may have wondered why we considered Statistical

Mechanics at all in relation to the point defect, the above observation is
critical to our understanding of the formation of point defects in solids.

It is a fact that in many cases the activation energy required to form

vacancies and/or interstials is approximately that of room temperature:

2.5.12.- Ev.Ei * kT

It is this mechanism which produces defects in the lattice. Temperatures

slightly above room temperature will produce defects in the solid, even
though the rate of production may be extremely slow below about 500 ° K.

The conclusion that we reach is that defect formation is favored in

the solid. It is much more difficult to obtain a "perfect" solid, so that
the defect-solid results.

We can summarize this statistical mechanical approach to the solid state

in that it gives us a method to evaluate the fraction of intrinsic defects
present under specified conditions, and also gives us a measure of the
deviation from absolute stoichiometry. If we can evaluate the energy of
formation of vacancies and interstitials at a given temperature for a given
crystal lattice, then we can calculate the numbers of defects formed at any
other temperature.

Ill - Defect Thermodynamics

The reason why we first investigated the Statistical Mechanics approach

to defect formation is that it gives us a good basis for understanding the
application of chemical thermodynamics to the defect solid state.
 2.5 Thermodynamics of the point defect 69

We begin this approach by defining the total numbers of defects present

as Nd. In addition, we need the following definitions:

2.5.13.- Definitions Needed to Define the Thermodynamic Approach

ENTROPY OF DEFECTS = Sd
ENTHALPY OF DEFECTS = Hd
FREE ENERGY OF DEFECTS = Gd

According to the laws of thermodynamics, we can write:

2.5.14.- AG = AH - ATS

Note that we are using the Gibbs free energy rather than the Helmholtz
free energy at this point. For the defect solid, we must define free-energy
in terms of the free-energy of the perfect solid, Go , as related to the

free-energy of the defect solid, vis:

2.5.15.- AG = Gd - Go
Note that due to increased entropy, the energy of the defect state is
higher than that of the perfect solid. For this reason, we must distinguish
between the sources which contribute to the total entropy, and multiply
by the number of defects present. The applicable equation is:

2.5.16.- Gd = Go + N d H d - TN d A Svib. - T A SCOnflg.

where ASvib. is the entropy change caused by changes in vibrational

frequency within the solid by the presence of defects and A SCOnfig. is the
change of entropy caused by configurational changes in the vicinity of
defects.

This aspect of defect solid state chemistry has been thoroughly studied
and the change in the various quantities of 2.5.16., as a function of
numbers of defects, is shown in the following diagram:
70 The point defect

2.5.17.-
Numbers of Defects as a Function of Gibbs Free Energy

You will note that when it is possible to plot Gd as a function of G, the

Gibbs free energy, a minimum occurs which is the number of defects at
the given temperature for the given solid. Note also that the enthalpy of
defects increases continuously as a function of Nd while both of the
entropy-contributions decrease. The net effect is a minimum value for Gd ,
at which point we can get the number of intrinsic defects present. Thus,
we have shown again that there is a specific number of intrinsic defects
present in the solid at any given temperature, this time by
Thermodynamics.

Let us now consider a combined (Statistical Mechanics and

Thermodynamic) approach to the problem of calculating defect numbers.

IV.- Combined Approach to Defect Formation

In this method, we will delineate the exact procedure used to derive

 2.5 Thermodynamics of the point defect 71

equations for specific defects in the solid. To show how this is

accomplished, we let N L be the total number of lattice sites and N v t h e
number of vacancies. The way that N v vacancies can be arranged upon NL
lattice sites is given by combinatorial statistics as:

2.5.18.- W = NL !
N V !(N V -N V ) !

The configurational or mixing entropy will be defined by:

2.5.19.- ASM = k l n W

Using Stirling's approximation for large numbers (this method is used

extensively in Statistical Mechanics), we can get:

2.5.20.- ASM = k [N L lnN L - N v In N v - (N L -N V ) In (N L -N V ) ]

You will note that we have the change in entropy as a function of t h e

differences between normal lattice sites and vacancies. Since we know
that NL » N v , we can write for the entropy of mixing:

2.5.21.- ASM <= [N L lnN L - N v In N v ]

and ASM = - k [Nv In N v / NL ] = - R ^ In X,

where Xy is the fraction of defects actually present. As expected, for non-

interacting defects, the entropy of mixing is ideal so that the free energy
of the system can be written as:

2.5.22.- AFV = N v (AEV - T ASVV) + RT [NL- N v / NL (In N L - N v / N L )

+ Nv/NLln-Nv/NJ]

where ASVV is the change in vibrational energy of the lattice arising from
the change in vibrational frequency around the vacant lattice site. If we
minimize the free energy, i.e.-

2.5.23.- 5Fv/aNv = 0 = AEV - T ASVV) + RT In (Nv / NL- Nv)

and: In (Nv / NL- Nv) = In X^ = AEV - T ASVV) / RT
72 The point defect

Thus, we can write for the atomic fraction of vacancies present:

2.5.24.- X, = exp - AF°V /RT =e"AVRT

You may wonder why we have examined this method for specifying
vacancies in general terms. The reason is that we can apply the same
method to the problem of point-defect pairs. This is intended to help you
understand how these various types of defects in the solid arise.

a. Interstitial Atoms

The process of creating an interstitial is just the opposite of creating a

vacancy. If there are N, possible interstitial positions in the lattice, and if
AFj is the free energy needed to move the atom into its interstitial
position, then the interstitial concentration at equilibrium will be:

2.5.25.- N, / N, = X, = exp (- AF, / kT)

The same approach applies to Frenkel pairs.

b. Frenkel Pairs

We have, in this case, both a vacancy associated with an interstitial atom.

Using the approach shown in 2.5.18., we have:

2.5.26.- W = NLJ x N,J

N V !(N L -N V )! NjKN.-NJ!

Now, we have two combinatorial functions, one for the vacancies and one
for the interstitials. Since the numbers of vacancies and interstials are
equal, i.e.- NF = NL = N,, we can use the same steps given in 2.5.19., ,20,
21., and 2.5.22. to get:

2.5.27.- AS M =RT [(N L lnNJ - (2NFlnNp) + (N,lnN,)

-{(N^-NpJlntN.-N,) ln(NL-NF)}]
 2.5 Thermodynamics of the point defect 73

Minimizing the free energy as before, we can write for the Frenkel pairs:

2.5.28.- N F = (N L N,) 1/2 exp (-AF iV / 2RT)

Note that AF applies to both interstitial and vacancy sites. Using 2.5.23.
and 2.5.24., we can get the fraction of Frenkel defects as:

2.5.29.- Xp = exp (AFiV /2RT

since the free energies are approximately the same, i.e.- AFiV s AF, + AFV

c. Schottkv Defects

The calculation of Schottky defects follows the same method already

given. We use cv for cation-vacancy and av for the associated anion
vacancy. The free energy of this defect is then:

2.5.30.- AFsh = N cv AECT + NAV AEAV - T (N cv ASVCV + NAV ASVAV)

- RT In N ! x N !
N ^ ! (N'-N cv ) ! N AV !(N'-N AV )!

Again, we have two combinatorial factors, one each for the associated
cation vacancies. Here, the total number of lattices sites, N » N a = N cv =
NAV . Minimizing the free energy gives:

2.5.31.- N sh / (N - NSh ) = exp (-AF^ / 2RT)

and: Xsh = exp (-AFsh / 2RT)

You will note that this approach uses Statistical Mechanics to approximate
the thermodynamic constants for the number of defects present.

V. Defect Equilibria

Just as chemical reactions can be described and calculated in terms of

74 The point defect

thermodynamic constants and chemical equilibria, so can we also

describe defect formation in terms of equilibria. This is given as follows:

2.5.32- Law of Mass Action:

bB + cC % dD + eE ; K = adD a e E / &bB &CC

Using this equation, we can calculate the numbers of defects for various
defects in the MX crystal as:

2.5.33.- Frenkel Defects (for the MX crystal)

Mx % Mi + VM ; KF = &ui &VM / &M

2.5.34.- Schottkv Defects ffor the MX crystal)

MX % VM + V X : Ksh = &VM &VM / &MX

Note that we have specified the equilibrium constants in terms of the

activity, a, of the associated defects. We can also write thermodynamic
equations for these defects:

2.5.36.- Chemical Thermodynamics

AG = AH - T AS = - RT In K ; K = exp AS/R • exp - AH/RT

2.5.37.- Defect Thermodynamics

Kd = exp ASd /R • exp - AHd /RT

where d refers to the specific defect.

We may summarize the knowledge we have already developed for the

Schottky and Frenkel defects:

1. We have shown by Statistical Mechanics that we can calculate

numbers of defects present at a given temperature.
 2.6 Defect equilibria in various types of compounds 75

2. There is an Activation Energy for defect formation. In many-

cases, this energy is low enough that defect formation occurs at, or
slightly above, room temperature.
3. Defects may be described in terms of thermodynamic constants
and equilibria. The presence of defects changes both the local
vibrational frequencies in the vicinity of the defect and the local
lattice configuration around the defect.

One question we may logically ask is how are we to know what types of
defects will appear in a given solid? The answer to this question is given
as follows:

IT HAS BEEN FOUND:

"There are two associated effects on a given solid which have opposite
effects on stoichiometry. Usually, one involves the cation site and the
other the anion site. Because of the differences in defect-formation-
energies, the concentration of other defects is usually negligible".

Thus, if Frenkel Defects predominate in a given solid, other defects are

usually not present. Likewise, for the Schottky Defect. Note that this
applies for associated defects. If these are not present, there will still be 2
types of defects present, each having an opposite effect upon
stoichiometry. Thus, we conclude that intrinsic defects usually occur in
pairs. This conclusion cannot be overemphasized. The following
discussion shows how this occurs in the real world of defects in solids.

2.6.- DEFECT EQUILIBRIA IN VARIOUS TYPES OF COMPOUNDS

Up to now, we have been concerned with the MX compound as a

hypothetical example of the solid state. We will now undertake more
concrete examples as found in the real world, using the concepts
developed for the simple MX compound. For the sake of simplicity, we
restrict ourselves to binary compounds, that is- one cation and one anion.
An example of a ternary compound is ABXs , where A and B are different
cations, and S is a small whole number.
76 The point defect

Our example of a binary compound will be:

MXs

We will distinguish between four states for this hypothetical compound, to

wit:
stoichiometric vs: non-stoichiometric
non-ionized vs: ionized

I. Stoichiometric Binary Compounds of MXs

In the real world of defect chemistry, we find that in addition to the

simple defects, other types of defects appear, depending upon the type of
crystal we are dealing with. These may be summarized as shown in the
following. According to our nomenclature, VM is a vacancy at an M cation
site, etc. The first five pairs of defects given above have been observed
experimentally in solids, whereas the last four have not.

2.6.1- Defects in the MXs Compound

PAIRS OF DEFECTS
Schottky VM + V X
Frenkel VM 4• M i
Anti-Frenkel Xi + vx
Anti-Structure XM + MX
Vacancy-Structure v M + MX
Structure-Vacancy VX + XM
Interstitial Mj + XI
Interstitial-Structure M X + Xi
Structure-Interstitial Mi + XM

This answers the hypothesis posed above, namely that defects in solids
occur in pairs. Study these defect-pairs carefully so that you become
familiar with them. They represent the type of structure defects found in
most solids. We have now introduced into our nomenclature a distinction
between structure and anti-structure defects. What this means is that
stacking faults can sometimes result in XM and Mx defects, which are
 2.6 Defect equilibria in various types of compounds 11

high-energy defects since they exist as cations (anions) surrounded by a

complete positive (negative) charge. In fact, this would appear to be
contrary to established modes of charge compensation in a solid.
However, it has been noted that these defects generally exist as a vacancy
plus the anti-structure defect, not as two anti-structure defects together.

Table 2-1 summarizes the various pairs of defects possible for binary
compounds.
TABLE 2-1
POSSIBLE DEFECTS IN THE MXg_COMPOUND
1. Constant Number of Sites
Defect Type Defect Exam Defect Equation Equilib. Constant
Pair pie
Schottky VM + VX TiO 0%V]y i + S V X + a V i Ksh = V M Vx Vi a
a
Frenkel VM + Mi ZnO MM +aV i % V M + M j K F = V M Mi/ M M Vj
Anti- Xj + V X LaH 2 x x +p Vi % X, + v x K A F = X.Vx/XxViP
Frenkel
Anti- XM+MX AuZn M M + X X % M X + XM K AS = *I X X M / M M X X
Struct.
2. Excess Number of Sites
Vac.-Struc V M + M X NiAl 6M M %5M X +(1-6)V M Kys= M»YVM Via
[ M M 8 + a] Vi
Structure- v x +x M 6XX % Ksv=XMVx1+6Via/Xx8
Vacancy XM(l+6)V X + aVi
Interstitial Mi + Xi 6(M M +SXx)% Ki = MiXi S Vi 1 + 6 + a /
Mi+SXi- (l+6+a)V, MM6 XXS6
Interst.- M x +Xi [M M +(1+6)X X % M X
Struct. +(1+6) X, -(l+6+a)V, (l+6)Xi- (l+5+a)Vi]
Struct- MJ + XM — [(1+6) M M + S X X % Ksi = Mi Xmd /
Interst. (1 +5) Mi + 6 XM] MM1+6XM8

Equilibria are given along with the appropriate equilibrium constant. You
will observe that these pairs of defects span the possible combinations of
defects, taken two at a time. The first four are based upon a constant
78 The point defect

number of sites, whereas the last five are based upon an excess in the
number of sites available. This excess we call "6". Note that we are not
speaking of the ratio of cations to anions, i.e.- stoichiometry, but of an
excess of cations or anions to the normal concentration of cations or
anions.

II. Defect Concentrations in MXg Compounds

It is of interest to be able to determine the number of intrinsic defects in

a given solid. As we have shown, pairs of defects predominate in any given
solid. Thus, the number of each type of intrinsic defects, Nj (M) or N4 (X),
will equal each other, For Schottky defects in the MXs crystal, we have:

2.6.2.- Ni(V M ) = Ni ( S V X )

This makes our mathematics simpler since we can rewrite the Schottky
equation of Table 2-1 as:

2.6.3.- 0 H Ni(vM) +SN 1 ( v M ) +aVj

Here, we have expressed the concentration as the ratio of defects to the

number of M- atom sites (this has certain advantages, as we will see). We
can than rewrite the defect equilibria equations of Table 2-1 in terms of
numbers of intrinsic defect concentrations. These are given as follows:

2.6.4.- Equilibrium Constants = Function of Numbers of Intrinsic Defects

SCHOTTKY: Kgh = Ns (S Ni ) 2 = S
FRENKEL: KF = N ^ / (1- N}) (a- Ni)
ANTI-FRENKEL: KAF = N { 2 / (S- Nj) (a- Ni)
ANTI-STRUCT: KAS = Ni 2 / (1-Ni )(S-Ni )S
VAC.-STRUCT: KVs = S s (S + 1 ) s + 1 • Ni 2 S + 1 / (S-Ni -S Nt ) s
STRUCT-VAC: Ksv = ( S H - I ) ^ 1 NiS+2 / (S - Ni - S Ni)
INTERSTITIAL: Ki = S s Ni S + 1 / (a - Ni - SNi) a + s + 1

Some of these equations are complicated and we need to examine them in

 2.6 Defect equilibria in various types of compounds 79

more detail so as to determine how they are to be used. Equation 2.5.16.

given above shows that intrinsic defect concentrations will increase with
increasing temperature and that they will be low for high enthalpies of
defect formation. This arises because the entropy effect is a positive
exponential while the enthalpy effect is a negative exponential.

Consider the following practical example:

TiO is cubic with the NaCl structure. A sample was annealed at 1300 °C.
Density and X-ray measurements revealed that the intrinsic defects were
Schottky in nature (V-n + Vo) and that their concentration was 0.140. In
this case, S = 1 so that:

2.6.5.- Ksh =0.0196=2xl0-2

This crystal is quite defective since 1 out of 7 Ti-atom-sites (0.14 1 ) is a

vacancy, and likewise for the oxygen-atom-sites.

Another example is:

CeHg . From thermodynamic measurements, it was found that the

intrinsic defects were Anti-Frenkel in nature, i.e.- {Hi + VH )• An
equilibrium constant was calculated as:

2.6.6.- KAF = 3.0 x 10 -4

at a temperature of 600 °C. This compound has the cubic fluorite

structure with one octahedral interstice per Ce atom. Therefore, a = 1,
and S = 2 for CeH2. We can therefore write:

2.6.7.- kAF = Ni 2 / (2-Ni )(1-Nj) = 3.0 x 10 "4

or Ni = 2.4 x 10 2 (600 °C.)

This means that 1 out of 42 hydride atoms is interstitial, and 1 out of 84

hydride-atom-sites is vacant.
80 The point defect

Let us review what we have covered concerning stoichiometric binary

compounds:

1. We have shown that defects occur in pairs. The reason for this lies in
the charge-compensation principle which occurs in all solids.
2. Of the nine defect-pairs possible, only 5 have actually been
experimentally observed in solids. These are: Schottky, Frenkel,
Anti- Frenkel, Anti-Structure, Vacancy-Structure.
3. We have given defect-equations for all nine types of defects, and the
Equilibrium Constant (EC) thereby associated. However, calculation of
these equilibria would require values in terms of energy at each site,
values which are sometimes difficult to determine.

A better method is to convert these EC equations to those involving

numbers of each type of intrinsic defect, as a ratio to an intrinsic cation or
anion. This allows us to calculate the actual number of intrinsic defects
present in the crystal, at a specified temperature.

III. Non-Stoichiometric Binary Compounds

We will now extend our treatment of intrinsic defects to the non-

stoichiometric non-ionized compounds, as represented by:

2.6.9.- MXs± 6

where 6 is a small increment. The question is: "How do we obtain non-

stoichiometry in the solid?". Consider a compound governed by either or
both the following equilibria:

2.6.10.- Xx # 1 / 2 X2(gas)ft +VX

2.6.11.- M(external phase) % MM + S V X + VMI

One example might be a halide crystal which has become non-

stoichiometric due to its being heated to a temperature sufficient to cause
a small amount of the halide to become volatile.
 2.6 Defect equilibria in various types of compounds 81

Another case might be an oxide, heated in the presence of excess metal,

e.g.- ZnO + Zn.

For a non-stoichiometric crystal, the concentration of each point defect,

in each conjugate pair, is no longer equal. If there is an excess of VM , Xj ,
or Xx , then the compound will have a surplus of X (or deficiency of M,
which is the same thing) over the ideal stoichiometric composition. This
is called a positive deviation from stoichiometry. Conversely, for a negative
deviation, there will be an excess of Vx, M, , or MM • This explains the
plus and minus in equation 2.6.9. In terms of the above given defects, 6
may be expressed as shown in the following Table:
TABLE 2-2.
Non-Stoichiometry, 6 , as a Function of Specific Types of Defects
in MXs + 5 Binary Compounds
Vacancv Formation Defect Eauation Eauilibrium Constan
Vx from external M M H MM + S[Vx] + aVi K Vx = M M [Vx] s Vi«/M
VM from gaseous X l/2X 2 %SX x + [VM]+aVi KvM=Xxs[VM] Vi« / p ^ l /2S
{if Xx •• 1 and Vi =a, then 6 = S[VM]/ 1-VM}
X-Interstitials l/2X 2 +aVi %[Xi] Kxi= PQ] / ( a - X O p j ^ 1 ' 2
or: {px 2 1 / 2 =l/K X i- 6 /a-5}
M-Interstitials MM + SXx+( 1+a)Vi % KMt= [M{ ] p x 2 S / 2 / M M X X 8
[Mi] + S/2X 2 [a -MJ (1+a)

or: P^® = ^Mi («(S +6) +

6)1+6 /(-8(S+6)«)1/S
(from gaseous X2)
X-Substitutionals l/2(S+l)X 2 %SX X +aVi K^HXMlaa/p^1'2'
(from gaseous X2 on • Vid+a) {or: px 2 ( S + 1 / 2 ) = 1/ K X M •
an M-site) 8/(l+S+5)l/S+l}
f(S+1)/2)
M-Substitutionals (S+l)X x + M M +aVi% KM X = px 2 [Mx]/
(gaseous X2 formed) ((S+1J/2) X 2 + Mx (l-Mx) a S+1 a

fr\f • TTv l(S+1) / 2) —

(KMV,1/S+1 (S+l)(S+d)/
82 The point defect

Note the various mechanisms which give rise to the specific combinations
of defects. These mechanisms have been thoroughly studied as a function
of specific compounds. It is sufficient for us, at this point, to observe
which defect equations govern both the equilibria and the non-
stoichiometry of the general compound, MXs ± 6 •

The equation used to calculate the non-stoichiometry factor, 6 , in the

general case is:

2.6.12.- 6 = (Xi-Vx)+ S(VM-Mi) + (S+1)(XM -M x )/1 + Mj + M x - VM - XM

We can express relationships between defect formation, the influence of

various external factors, and the equilibrium constant thereby related. We
do this in terms of 8, the degree of non-stoichiometry, as given in Table 2-
2. Even though these equations are rather formidable-looking, we shall be
able to use them to good advantage.

Note that each case corresponds to the influence of a reacting external

factor on a stoichiometric solid, which contains intrinsic defects. These
factors produce additional defects because of nonstoichiometry and
charge-compensation. The defect produced is enclosed in brackets in
Table 2-2.

Consider this factor carefully by again examining Table 2-2. Also given is
the reaction producing the defect, with its corresponding equilibrium
constant. In most cases, the deviation, 8, is presented in terms of the
equilibrium constant and the partial pressure of the external gaseous
reactant.

Thus, if the number of defects produced can be measured and an

equilibrium constant calculated, then 8 can be determined both as a
function of partial pressure, px 2 . and temperature (see 2.5.22.).

IV. Defect Concentrations in MXs+s

We now proceed as we did for the stoichiometric-case, namely to develop

 2.6 Defect equilibria in various types of compounds 83

defect-concentration equations for the non-stoichiometric case, i.e.-

MXs±6. Consider the effect of Anti-Frenkel defect production. From Table
2-1, we get KAF , with its associated equation, ICAF • In Table 2-2, we use
Kxi for X-interstitials. Combining these, we get:

2.6.13.- KAF = Ky x • Kxi = Ni2 / (S-Ni)(a-Ni)

When both Vx and Xj coexist in the lattice, the deviation from

stoichiometry (from 2.6.12.) becomes:

2.6.14.- 6 = [Xi] - [VX]

Using the equilibrium constant of 2.6.13., i.e.-

2.6.15.- Kvx= PX2 1/2 [VX] fXx = PX21/2 [Vx] / S - [ V X ]

and the appropriate one from Table 2-2 (i.e.- Kxi ), we get (assume for
simplicity that S = a = 1) :

2.6.16.- 8 = a px2 1 / 2 KX1/PX2KXI+ 1 - S Kv x / K v x + p x 2 1 / 2

We can rearrange terms in 2.6.16. to obtain: Kxi (1- 5)px2 - 8( Ky x Kxi + 1)

p x 2 1 / 2 and if : K y x 2 (1- 8 ) = 0, we can, by using 2.6.13. and Nj «1, obtain:

2.6.17.- Nj2 (1- 6 ) p X 2 - 8 K V x P X 2 1 / 2 - K v x 2 ( l - 8 ) = 0 .

Solving for px2 yields :

2.6.18.- pX2 = K y x 2 ( 8 2 + 2 Ni (1- 82) ± 8 [82+4 N4 (1-8 2 )]*/ 2 / 2 N i 4 ( l - 8 ) 2

Since at stoichiometric composition, 8 must equal zero, this rather

formidable equation reduces to:

2.6.19.- p X2 ° = K v x 2 / N i 2
84 The point defect

where p X2 ° is the pressure of X2 gas in equilibrium with the MXs

crystal at the stoichiometric composition. This gives us the opportunity to
divide 2.6.18. by 2.6.19. to obtain:

2.6.20.- PX2/PX20 = S2 + 2 N
> (1 " 6 2 ) ± 8 62
t +4Ni(l-52)/2Ni2(l-5)2

an( N
We can therefore calculate 6 in terms of the ratio of p X 2 to Px 2 ° i i•
shown as follows:

2.6.21.-
Effect of External Pressure of X 2 Gas on Non-Stoiehiometry
of t h e Hypothetical Compound, MXg+5

Although we will not treat the other types of pairs of defects, it is well to
note that similar equations can also be derived for the other intrinsic
defects. What we have shown is that external reactants can cause further
changes in the non-stoichiometry of the solid.
 2.6 Defect equilibria in various types of compounds 85

V. Ionization of Defects

We have already covered, albeit briefly, non-ionized stoichiometric and

non-ionized non-stoichiometric intrinsic-defect compounds. Let us now
consider the ionization of defects in these compounds. In the MXs
compound, if we remove some of the X-atoms to form Vx, the electrons
from the removed X-atom (or from the bond holding the X-atom in the
crystal) are left behind for charge compensation reasons. At low
temperatures, these electrons are localized near the vacancy but become
dissociated from the point defect at higher temperatures. They become
free to move through the crystal, and we say that the intrinsic defect has
become ionized. We can write the following equations for this mechanism:

2.6.22.- VACANCIES

VX % VX+ + e- KVx* = IVX |+ I e- I / VX

VM%VM"+p+ Kv M * = |VM-|IP+|/VM

In the equilibrium constant equations, each symbol is actually a

concentration, i.e.- numbers of specified defects and electrons, etc.

In alike manner, we write for interstitials and anti-structure defects:

2.6.23.- INTERSTITIALS

Mi % M i + + e - KMi* = |Mi+|| e-| /Mi

Xi % Xt- +p+ Kxi* = |Xi | |p+ |/Xi

2.6.24.- ANTI -STRUCTURE

XM % XM + + e- KXM = |XM+ I I e- | /XM

MX % MX" +P+ KMx = |MX- I | p + | / MX

A useful example for understanding the above equations, and the

ionization of defects, is that of cobaltous oxide, CoO. External oxygen
86 The point defect

pressure will affect stoiehiometry and produce cobalt vacancies, but the
vacancies are ionized at room temperature:

2.6.25.- 1/2 O2 % Oo +V Co " +JT+ + a V s

The n+ is free to migrate throughout the lattice. However, at pressures

below 1 0 6 (obtained by application of a vacuum), the Co vacancies become
doubly-ionized:

2.6.26.- V C o " % Vco2" + it+

This illustrates how the ionized defect arises.

We have already illustrated how vacancies arise through the use of a

PLANE NET. ( SEE 2.2.1.). Therein, we used MX as the molecule to build
the NET. Let us now return, using the ions, M 2+ and X2- as the ionic
forms with which to build the NET. This is shown as follows:
2.6.27.-
|A Plane Net for the Ionized MX Compound |
n 2+x2-M2+>;2- M 2+X2-M2-X2-

X -M X M X M2+X2-M2+
2 3+ a 2+ 2

M 2 - X 2 - M 2 i X 2 M 2 + ) 0 : M 2 + X2- . V a c a n c u ,

Hole > 1 ^ 2 - X 2 - n 2 + K ^ > ^ 2 - M 2 + X 2 ' Complex

x 2 M 2 + X 2 < H 5 > K 2 " M 2 * X 2 " M2+
L M 2 ; X 2 T 1 2 * X ^ - M 2 + X 2 " M 2+ X 2 ~
X 2 l [M 5 '|X 2 M 2 + X 2 ' h 2 t X 2 ' M Z t
M2tX2- M2tX2'M2ti<2'|FP1X2- IVacancy-

I lonizeTT M 2 + X 2 - M 2 + X 2 - M 2 + X 2 - [ r T J X 2 - Comp1ex
Electron! X 2 " M 2 + X=" M 2 + X 2 " M 2 ^ X2" M 2 +
^~_ MVX2~ M 2 "X 2 " M 2+ X 2 - M 2 + X 2 -
" X ^ [ H 3 X 2 ~ M 2 + X 2 " M 2 * X 2 ~M 2+
M 2 + X 2 " M 2+ X 2 " M 2 + X 2 " M 2 + X 2 -
X 2 - M 2 *X 2 - M 2+ K 2 - M 2 + X 2 - M 2 +
 2.6 Defect equilibria in various types of compounds 87

A vacancy-hole complex is shown, as well as a vacancy-trapped electron

complex. In addition, an example of an ionized hole and an ionized
electron-site is also given. We must therefore add to our list of ionization
equations the following:

2.6.28.- M M +Z % MM+2'1 + P +

M M +Z ^ M M +Z+1 + e-

Because of the way that we have defined our MXs crystal, the analogous
ionization of the anion, X, does not occur.

Let us now summarize what we have covered:

1. We have shown that defect equations and equilibria can be

written for the MXs compound, both for the stoichiometric and
non-stoichiometric cases.

2. In addition, we have shown that further defect formation can be

induced by external reacting species, and that these act to form
specific types of defects, depending upon the chemical nature of the
crystal lattice.

3. We have also shown that the intrinsic defects can become

ionized.

Because most of what we have covered to this point has assumed lack of
ionization, we need to proceed further so as to develop equations more
suitable for the real world.

Most of the cases we encounter in the real world involve ionic lattices
containing charged cations and charged anions. Up to now, we have
considered simple hypothetical compounds such as MX and delineated
various kinds of non-ionized intrinsic defects that would be present.
Section 2-6 was concerned with various types of defects intrinsic to the
stoichiometric MXs compound and the non-stoichiometric compound,
88 The point defect

MXs+6- Additionally, we examined the ionized MXs compound. The non-

ionized types are the easiest to illustrate, including the mathematics
involved. Nevertheless, even those equations (see Tables 2-1 and 2-2)
became somewhat cumbersome at times. Such equations are applicable to
metallic, and possible some covalent, crystals. Equations including
ionization and electric charge are needed for ionic crystals and semi-
conductors. But, the mathematics involved starts to become unsolvable.
The main problem is the need to have a complete description, utilizing
equations describing all possible processes, including:

Intrinsic defect formation

Effect of extrinsic factors

Ionization of intrinsic defects

Effects of charge-compensation.

By the time we are finished, we find that we have derived a set of multi-
variant equations in several unknowns. Brouwer (1954) considered this
case and was able to formulate a method of solution. The following is a
summary of that method.

2.7. - BROUWER'S APPROXIMATION METHOD

Let us consider an MXs crystal with Schottky defects (this is one of the
more easily defined types, mathematically). We define S=l. The types of
intrinsic defects was given in 2.6.1., the equilibrium constants of intrinsic
defects was given in Table 2-1, numbers of intrinsic defects was given in
2.6.4., ionization of vacancies in 2.6.22, 2.6.23 & 2.6.24, and 2.6.21. gave
the effect of an external factor (X2 gas) on the production of non-
stoichiometry. We can rewrite these equilibrium constant equations for
the MX compound in the form given in 2.7.1. on the next page.

This set of equations completely defines the defect concentrations in an

MX crystal containing Schottky defects.
 2.7 Brouwer's approximation method 89

2.7.1- a In Ks = In V M + In V x
b. In KV * = In e + In Vx* - In V x
A.
c. In KyM* = In p + In VM" - In VM
d. lnK V v = l / 2 1 n p v + In V x
e. lnKy., = lnVM - 1 / 2 In p,,
The intrinsic ionization constant is:

f. In Kion = In e + In p (where KiOn is the ionization

equilibrium constant)
and the electroneutrality condition is :

g. e + VM" = p + VX+

It is important to realize that e is the sum of the electrons, not a single

negative charge, i.e.-

2.7.2- e = 2 e- and p = 2 p+

This set of equations completely defines the defect concentrations in an

MX crystal containing Schottky defects.

There are seven equations in 2.7.1. and if the equilibrium constants are
known, there remain seven unknowns in these simultaneous equations,
i.e.- These involve: VM, VX , VM" , Vx+ , e , p , and p . The first six
equations are linear relations between logarithms of concentrations, and
logarithms of equilibrium constants, but the last equation is not. The
solution to this set of equations can be accomplished, but the methods are
complex. However, if there is further ionization of defects, or the case
where more than one pair of defects results, the situation becomes
hopeless. The equations can be written, but the set of equations cannot be
easily solved.

In 1954, Brouwer proposed a graphical method for solving these

equations. The method has been adopted because of further development
by Kroeger and Vink (1956). The method entails dividing the range of
90 The point defect

defect concentrations into regions, such that charge-compensation

involves only two defects, so that the electro-neutrality equation (2.7.1-g.)
is simplified. Brouwer's Method has been applied to the case of silver
chloride where impurities like cadmium chloride may be present. The
photographic properties can be enhanced or deteriorated, depending
upon the state of the impurity added, inadvertently or not.

For a large negative stoichiometry deviation, VM' and p+ become

negligible compared to Vx+and e- Then, Equation 2.7.1.- g. becomes:
In e = In Vx+ . By using this relation, all defect concentrations can be
plotted in terms of X2 pressure and the equilibrium constants. This is
shown as aBrouwer plot, given as 2.7.3. on the next page.

It is more convenient to use Nd , the numbers of a given defect, and to

define a quantity, R, as:

2.7.4.- R = KyM p X2 l/2

for the MX compound containing Schottky defects. We can then express

Nd in powers of R, so that:

2.7.5.- lnNd = 6 1nR + C

We find that 8, the slope, is either ± 1/2 or ± 1. This simplifies matters

greatly, as can be seen in 2.7.3.

Here, the defect equations for our MX compound (containing Schottky

defects) are plotted for the case: KiOn > Ksh . In obtaining this plot, we
have derived the following equations from equations 2.7.1. and 2.7.2.:

2.7.6.- VM = R
V x = Ks / R
e = Vx+ = (Kvx* V X ) 1 / 2 = ( Kvx- Ks / R ) 1/2
p =Ki ( R / K y x * - Ks) 1 / 2
VM" = Kvx* • VM / p = (R Kyx* Ks) KvM* ) / Kion
 2.7 Brouwer's approximation method 91

We can readily see that concentrations are now expressed in powers of R,

and that by taking logarithms , we get slopes of ± 1/2 or ± 1. As shown in
2.7.3., we can now differentiate between 3 regions of defect
concentrations, namely:

REGION I - VX+ > VM

REGION II - Vx+ = VM~
REGION III - VM" > V x +
92 The point defect

At large positive deviations from the stoichiometric composition (REGION

III), VM" » e and p » V x + . In this REGION , 2.7.1.-g becomes:

2.7.7. - hi VM" = In p

In the vicinity of the stoichiometric composition (REGION II), the

concentration of defects depends upon whether Ks > KiOn , or vice-
versa. The former usually holds for large band-gap ionic compounds.
Then, 2.7.1.- g becomes:

2.7.8.- lnV M " = lnVx+

We can also show that the change from near stoichiometry to a negative
deviation of stoichiometry (REGION I) occurs when:

2.7.9.- (RKsKv x * / K l o n ) 1 / 2 = (Ks K lon / Kvx* R) 1 / 2

This gives us:

2.7.10.- Rn~ i = K ion / Kyx*

and in a like manner:

2.7.11.- Rn~ m = Ks /Kv x *

Because the stoichiometry is rigidly defined in REGION II by the

condition:

2.7.12.- VM + V M - = VX + Vx+

changes in R or p X 2 do not greatly affect deviation from stoichiometry.

But, there are large changes in both 2 e~ and 2 p+ in this region. Thus,
the stoichiometric composition is probably best defined when e = p .

This occurs, as shown in 2.7.3., at the value:

 2.7 Brouwer's approximation method 93

2.7.13.- Ro = ( K s K i o n ) l / 2 /Ky x *

An example of a compound where Ks » KiOn is KBr. Kroeger (1964)

obtained values (at 600 °C.) of:

KS = 8 x 1 0 1 4

K i o n = 3 x 10-35

If we subtract 2.7.9. from 2.7.10. , we find that the Br2 gas - pressure
changes by 10 4 3 (since R « PBro 1/2 ) o v e r Region II. Moreover, the
deviation from stoichiometry, 8 , changes only by 1 0 3 , or remains
essentially constant. However, 2 c (or 2 p+ ) changes by a factor of
approximately 10 2 1 over Region II. In electronic semi-conductors, the
condition: Ki 0n » Ks , usually prevails. We usually get a Brouwer analysis
like that of 2.7.3.

Now, consider the case where ionization is the norm. This case, shown in
2.7.14. (next page), is for the hypothetical semi-conductor alloy MX,
where M acts as a cation , and X acts more like an anion. That is, when we
get ionization of defects, M loses an electron and X is positively ionized.
An example could be GaAs. In this case, electronic charge is relatively
insensitive (Region II) to composition, whereas the deviation from
stoichiometry, d , varies considerably. In the positive-deviation direction
(Region III), the major defects are VM" , p , and VM .

This gives the relation:

2.7.15.- VM % VM" + P+

In the negative-deviation direction (Region I), e and VM" predominate.

For the major defects of this system, we have the conditions shown in
2.7.16. on the next page.
94 The point defect

2.7.14.-

Region I - e> p
Region II - e » p
Region III - p > e

Note that we have rather well defined the defects present for this type of
semi-conductor, using relationships defined by Brouwer's method. It
 2.8 Analyses of real crystals using Brouwer's method 95

should be apparent that this method adds considerable power to our

ability to analyze intrinsic defects where ionization is the norm.

2.8.- ANALYSES OF REAL CRYSTALS USING BROUWER'S METHOD-

COMPARISON TO THE THERMODYNAMIC METHOD

The silver halide series of compounds have been extensively studied

because of their usage in photographic film. In particular, it is known that
if silver bromide is incorporated into a photographic emulsion, any
incident photon will create a Frenkel defect. When the film is developed,
the Agi+ is reduced to Ag metal. These localized atoms act as nuclei to
cause metal crystal formation at the points "sensitized" by the photon
action. (Note that this description is an oversimplification of the actual
mechanism. Nevertheless, it should be apparent that a knowledge of
defect chemistry of the compound, AgBr, should prove to be very
important in understanding the chemistry of photographic films).

I. The AgBr Crystal with a Divalent Impurity. Cd 2+

Consider the crystal, AgBr. Both cation and anion are monovalent, i.e.- Ag+
and Br- . The addition of a divalent cation such as Cd 2+ should introduce
vacancies, VAg , into the crystal, because of the charge-compensation
mechanism. To maintain electro-neutrality, we prefer to define the
system as:

2.8.1.- (1-6) Ag+ Br- : 6 Cd 2+ S=

Fortunately, AgBr is easy to grow as a single crystal, using Stockbarger

Techniques. Possession and measurement of a single crystal greatly
facilitates our measurement of defects. The imperfections we expect to
find are:

2.8.2.- VAg , Agi , e- , p+ , CdAg , and [CdAg , VAg ]

The last defect is one involving two nearest neighbor cation sites in the
lattice.
96 The point defect

The following table gives the defect reactions governing this case:

TABLE 2 = 3
DEFECT REACTIONS IN THE AgBr CRYSTAL CONTAINING Cd 2+

a. 0 % e- + p+
*>. AgAg + « Vi H VAg + Agi
C AgAg + a V j % VAg" + Agi+
d. Agj *=t Agi+ + e-
e. 1/2 Br 2 % Br B r + V B r
f. CdAg % Cd A g + + e-
g. CdAg + V A g % [Cd Ag , VAg]
h. CdAg+ + V A g " % [CdAg, VAg]

i. For Constant Cd
Cd T = CdAg + Cd A g + + [Cd Ag . V Ag ] = K
). For Electro-neutrality
e- +V A g" = p+ +Agi+ + Cd Ag +

We would normally plot In Nd vs: In KAgi + In KyAg • However, we find it

more convenient to plot In Nd vs: 1/T . The reason for this is as follows.
Experimentally, we find that if we fix the Cd2+ content at some convenient
level, it is necessary to anneal the AgBr crystals at a fixed temperature for
times long enough to achieve complete equilibrium. If the temperature is
changed, then both type and relative numbers of defects may also change.

Thus, we plot In Nd vs: 1/T, as in the Brouwer diagram of 2.8.3., given on

the next page. At low temperatures (Region III), singly-charged defects
predominate, i.e.- 2 VAg = Agi+ + CdAg+. At the junction of III -
II, the charged moieties begin to cluster to form CdAg+ and Vag~ . These
in turn may form the complex:

2.8.4.- CdAg+ + V Ag % [CdAg , VAg]

At the same time, the concentration of Agi+ drops dramatically.

 2.8 Analyses of real crystals using Brouwer's method 97

At the juncture of II - I in the Brouwer diagram, the complex clearly

dominates.
98 The point defect

While the above results show how intrinsic defects are affected by
temperature, we still do not know how the electrons and holes vary as a
function of temperature. (Note that temperature, as specified, is a
preparation temperature, not measurement temperature. Measurement of
intrinsic conductivity, a , is shown at the bottom of 2.8.3. At low
temperatures, conductivity due to vacancies, ay , appears to be the major
contributor:

2.8.5.- VAg % VAg" + p+

Conductivity decreases at higher temperatures because Agi+

concentrations decrease. Finally, at the very high temperature region
(Region I), conductivity is relatively low, and approaches zero, because the
complex [CdAg , VAg] predominates as the number of charged moieties,
(CdAg+ + VAg') decreases.

Varying the Cd 2+ content in the AgBr crystal affects the relative defect
ratios, as shown in the following diagram, shown as 2.8.6. on the next
page. Again, we can identify 3 Regions as a function of Cd- concentration:

> v
2.8.7.- REGION I - Agi+ A g"

REGION II - Agi+ @ VAg"

>
REGION III - VAg" Agi+

Region I of 2.8.7. corresponds closely to that of Region III of 2.8.5.

Teltow (1949) also studied this crystal. He measured conductivity of AgBr
crystal containing various amounts of Cd 2+ , as a function of measurement
temperature. Up to 175 °C, he obtained conductivity curves similar to
those of the middle of 2.8.6. But, as the measurement temperature
increased, the pronounced dip seen in 2.8.6. tended to flatten out. At the
highest measurement temperature of 410 °C, the conductivity was flat.
He concluded that elevated temperatures preclude the formation of
clusters and/or complexes, so that conductivity due to Agj+ remains the
2.8 Analyses of real crystals using Brouwer's method 99
100 The point defect

major contributor to the conductivity, as according to the defect reaction

"6" of Table 2-3.

LET US NOW SUMMARIZE WHAT WE HAVE COVERED TO DATE:

1. By rewriting the equilibrium constants of Table 2-1 and 2.7.20.

(ionization of vacancies) as logarithms, we obtained linear relations
among the set of defect equations.

2. By defining sets of defects as a ratio, R, we can then plot the

ratios so as to show how the relative numbers vary as a function of
the type of defect present in the chosen crystal lattice. This is the
Brouwer Method.

3. We also illustrated the method for a AgBr crystal containing Cd2+ .

The set of defect reactions were given, so as to illustrate the
possible defects present. Then, a Brouwer diagram illustrated the
numbers and types of defects actually present as a function of Cd2+
content in the crystal.

II. Defect Disorder in AgBr- A Thermodvnamic Approach

To illustrate yet another approach to analysis of defect formation, consider

the influence of Br2 - gas upon defect formation in AgBr. The free energy
of formation, AG, is related to the reaction:

2.8.8.- Ag° +l/2Br 2 (g) % AgBr(g) {AGAgBr}

This can be rewritten as:

2.8.9.- aAg P 1 / 2 B r = exp AGAgBr/RT

It makes no difference as to which of the activities we use. If we now fix

pBr2 a t some low value, we find that the possible defects in our AgBr
crystal, as influenced by the external factor, pBr2- "ill be:
 2.8 Analyses of real crystals using Brouwer's method 101

2.8.10.- Agj+ , VAg" , Brr , VBr+ , e- and re+

where we use n for the positive charge to differentiate between pressure,

p, of the external gas. Because of the high electrostatic energy required to
maintain them in an ionic crystal such as AgBr, we can safely ignore the
following possible defects:

2.8.11.- AgBr+ , AgBr++ , BrAg" , BrAg= .

If we have thermal disorder at room temperature (I do not know of any

crystal for which this is not the case), then we can expect the following
defect reaction relations:

2.8.12.- Agi+ = VAg-

Brr = Ag}+

VAg" = V Br +

Brr = V Br +

At equilibrium, the following equations arise:

2.8.13.- A g ^ + l / 2 B r 2 ( g ) % AgBr(s) + V Ag ' + jr+

Kd = VAg-'^/p1'28^

This gives us a total of eight (8) concentrations to calculate. They involve

the following crystal defects:

2.8.14.- Ag^, Br Br , Agi , Bri , VAg- , VBr+ , e" , rc+

Our procedure is to set up a site balance in terms of lattice molecules,

i.e.- for AgBr:
102 The point defect

2.8.15.- AgAg , + VAg- + e - =1 (e- S Ag"Ag)

+
Br B r + VBr + Jt+ = 1 (jt+ = Br+Br)

Since Br2 (gas) is the driving force for defect formation, we need also to
consider deviation from stoichiometry, 6 . Thus, we also set a Agi-s Br
balance:

2.8.16.- AgAg+Agi+ + e- = 1+6

Brer + Brr + jt+ = 1

To maintain electroneutrality:

2.8.17.- Agi+ + V B r + +Jt+ = Bri" + VAg" + e"

We also set up the following equations:

2.8.18.- AgAg + «Vj % Agi+ + VAg- Ke = Agj+ VAg" / Vi«

2.8.19.- Br Br + «Vj % Br r + VBr+ Kg = Brr V B r +/ Vi«

and
2.8.20.- e- +p+ % 0 Kb = (e')(jt + )

Note that we have distinguished between three (3) situations, to wit:

a. Electroneutrality

b. Thermal Disorder

c. Non-stoichiometry (excess cation)

These are the eight equations (2.8.12. to 2.8.20.) required to calculate

the defect concentrations arising from the effects of the external factor,
PBr2 • From measurements of conductivities, transfer numbers
(electromigration of charged species), lattice constants and experimental
 2.8 Analyses of real crystals using Brouwer's method 103

densities, it has been shown that Frenkel defects predominate (Lidiard -

1957). This means that:

2.8.21.- Agi+,V A g - » Brr , V Br +

Furthermore, VAg" = Agi+ so that in terms of our equilibrium constants

we get:

2.8.22.- FOR FRENKEL DEFECTS: Ke » Kg and Ke » Kd

Thus, we need only to consider the above two (2) defects, namely - VAg'
and Agj+ , since they are the major contributors to non-stoichiometry. By
calculating p° as before (when 5 = 0 , see 2.7.22 & 2.7.23.), we can
2
express our overall defect equation as:

2.8.23.- P 1 / 2 B r /(P° Br ) 1/2 ={5/2e+[(l+5/2e)2]l/2}{6/2p+[l+(8/2p)2]l/2}

2 2

Because of the conditions given in 2.8.16., the first half of the equation
can be set equal to one. Note that we are using e , p , and y as t h e
equilibrium constants . i.e. -

2.8.24.- e = Kei/2 p = K p i/2 Y =K Y 1 / 2

In the remaining part of the equation, 8 » p . By taking logarithms, we

can then obtain:

2.8.25.- 1/2 In p / p° = In 5 - In p
2 2

This result then leads us to a plot of the effect of partial pressure of Br2
on the deviation from stoichiometry , d , for the AgBr crystal, as shown in
2.8.26. on the next page (this work is due to Greenwood - 1968).

For p = 0 , there is a point of inflection where the slope of the line is

defined by the equilibrium constant, i.e.- p = K p 1 / 2 .
104 The point defect

The larger this value, the flatter is the curve. All relations regarding the
defects can now be derived.

The major defects turn out to be:

2.8.27.- VAg- % A&+

*+ % K 6 (l/Ke)l/2 pl/2^
e- % (Ke)l/2 (1/Ke) Kb{ 1/p )l/2
This shows that both P+ and e- are minority defects dependent on p .

The following gives the standard enthalpies and entropies of these defect
reactions, according to Kroeger (1965):
 2.9 Summary and conclusions 105

2.8.28.- DEFECT REACTION AS AH

Cal./mol/°K Kcal./mol.

AgAg % Agi+ + VAg" (Frenkel) 25.6 29.3

AgAg+Br B r^ VAg" + VBr+ -13.3 36

0% e- + p+ 25 78

1/2 Br2 + AgAg % AgBr + VAg' + p+ 4.9 25.4

It is apparent that the Frenkel process coupled with the electronic

process are the predominating mechanisms in forming defects in AgBr
through the agency of external reaction with Br2 gas.

A final comment: we can use these thermodynamic values to calculate the

equilibrium constants according to:

2.8.29.- Ki = exp - AGi° / RT

and can also obtain the activity of the silver atom in AgBr from 2.8.29. By
using equation 2.8.9., we can show :

For ciAg = 1 , @ T = 277 °C. ; 5 = + 10 12

For p B r = 1 atm. @ 277 °C. ; 8 = - 10' 7

where the plus or minus indicate an excess or deficit of the silver atom in
AgBr. This result is due to Wagner (1959).

2.9.- SUMMARY AND CONCLUSIONS

Let us now summarize the major conclusions reached regarding the

defect solid. You will note that we have investigated the following
hypothetical compounds: MX, MXs and MXs+ 6- But, when we investigated
crystals in the real world, we found that actual defects in such solids did
106 The point defect

not conform entirely to those of our hypothetical compounds.

Nonetheless, in order to comprehend and form a foundation to

understand how defects affect the properties of actual solids, it was
necessary to study those hypothetical compounds. The following is a
summary of the conclusions we reached regarding the defect solid state:

1. The charge compensation mechanism represents the single most

important mechanism which operates within the defect solid.
2. We have shown that defect equations and equilibria can be
written for the MXs compound, both for the stoichiometric and
non-stoichiometric cases.
3. The conclusion that we reach is that defect formation is favored
in the solid because of the entropy factor. It is much more
difficult to obtain a "perfect" solid, so that the defect-solid
results. We have also shown that the intrinsic defects can become
ionized.
4. Although more than one defect reaction may be applicable to a
given situation, only one is usually favored by the prevailing
thermodynamic and electrical conditions. Thus, we conclude
that intrinsic defects usually occur in pairs. This conclusion
cannot be overemphasized.
5. We have shown that defects occur in pairs. The reason for this
lies in the charge-compensation principle which occurs in all
solids.
6. Of the nine defect-pairs possible, only 5 have actually been
experimentally observed in solids. These are: Schottky, Frenkel,
Anti- Frenkel, Anti-Structure, Vacancy-Structure.
7. There can be no doubt that both Schottky and Frenkel defects
are thermal in origin.
8. We have shown by Statistical Mechanics that we can calculate
numbers of defects present at a given temperature.
9. There is an Activation Energy for defect formation. In many
cases, this energy is low enough that defect formation occurs at,
or slightly above, room temperature.
 2.10 The effects of purity 107

lO.Defects maybe described in terms of thermodynamic constants

and equilibria. The presence of defects changes both the local
vibrational frequencies in the vicinity of the defect and the local
lattice configuration around the defect.
11 .There are two associated effects on a given solid which have
opposite effects on stoichiometry. Usually, one involves the
cation site and the other the anion site. Because of the
differences in defect-formation-energies, the concentration of
other defects is usually negligible.
12. In examining the defect state of real crystals such as AgBr, we
find that we can write, using equilibrium constants and defect
thermodynamics derived from Statistical Mechanics and classical
Thermodynamics, valid equations for the numbers and types of
various associated defects present. However, we also find that we
cannot solve for the value of the unknown quantities in a set of
simultaneous equations since the equations are not linearly
solvable. The equations can be written, but the set of equations
cannot be easily solved. It is for this reason that we have resorted
to graphical method like that of Brouwer, even though it is not
entirely satisfactory in its solutions to the numbers and types of
defects present in real crystals.

Thus, it should be clear that lattice defects in the solid state is the normal
state of affairs and that it is the defects which affect the physical and
chemical properties of the solid.

2.10. - THE EFFECTS OF PURITY (AND IMPURITIES)

Our study has led us to the point where we can realize that the primary
effect of impurities in a solid is the formation of defects, particularly the
Frenkel and Schottky types of associated defects. Thus, the primary effect
obtained in purifying a solid is the minimization of defects. Impurities,
particularly those of differing valences than those of the lattice, cause
charged vacancies and/or interstitials. We can also increase the reactivity
of a solid to a certain extent by making it more of a defect crystal by the
addition of selected impurities.
108 The point defect

It is not so apparent as to what happens to a solid as we continue to purify

it. To understand this, we need to examine the various grades of purity as
we normally encounter them. Although we have emphasized inorganic
compounds thus far (and will continue to do so), the same principles
apply to organic crystals as well. COMMERCIAL GRADE is usually about
95% purity (to orient ourselves, what we mean is that 95% of the material
is that specified, with 5% being different (unwanted-?) material.
Laboratory or "ACS-REAGENT GRADE" averages about 99.8% in purity.

2.11.1.- GRADES OF PURITY FOR COMMON CHEMICALS

GRADE % ppm IMPURITIES IMPURITY ATOMS

PER MOLE OF
COMPOUND.
Commercial 95 50,000 3.0 x 10 2 2
Laboratory 99.8 2000 1.2 x 10 2 1
Luminescent 99.99 100 6 x 10 1 9
Semi-conductor 99.999 10 6 x 10 1 8
Crystal Growth 99.9999 1 6 x 10 17
Fiber-Optics 99.999999 0.01 6 x 10 1 5

The GRADES listed above are named for the usage to which they are
intended, and are usually minimum purities required for the particular
application. Fiber-optic materials are currently prepared by chemical
vapor deposition techniques because any handling of materials introduces
impurities. Furthermore, this is the only way found to date to prepare the
required materials at this level of purity.

The frontiers of purity achievement of solids presently lie at the fraction

of parts per billion level. However, because of Environmental Demands,
analytical methodology presently available far exceeds this. We can now
analyze metals and anions at the femto level (parts per quadrillion= 10" 15 )
if we wish to do so.

Nevertheless, it is becoming apparent that as high purily inorganic solids

are being obtained, we observe that their physical properties may be
 2.11 Nanotechnology and the solid state 109

different than those usually accepted for the same compound of lower
purity .

The higher-purity compound may undergo solid state reactions somewhat

differently than those considered "normal" for the compound. If we
reflect but a moment, we realize that this is what we might expect to
occur as we obtain compounds (crystals) containing far fewer intrinsic
defects.

It is undoubtedly true that many of the descriptions of physical and solid

state reaction mechanisms now existing in the literature are only partially
correct. It seems that part of the frontier of knowledge for Chemistry of
The Solid State lies in measurement of physical and chemical properties
of inorganic compounds as a function of purity. A case in point is that of
the so-called "Nano-Technology", the vanguard of research into chemical
and physical properties of materials in the research community today.

2.11.- Nanotechnology and The Solid State

In the next chapter, we shall examine the methods of characterizing

solids including: the properties of individual particles (including single
crystals); the solid state reactions that are used to form various solids; and
methods used to describe an assembly of particles (particle size). We will
find that most solid materials are composed of particles in the 1- 300 \im.
range. This is 1-300 x 10"6 meters. Most inorganic materials are produced
having particles in this size range. These are the familiar powders such as
coal dust, inorganic chemicals, silt and fine sand, and even bacteria.

Current research defines nano-technology as the use of materials and

systems whose structures and components exhibit novel and significantly
changed properties when control is achieved at the atomic and/or
molecular level. What this means is that when a given material is
produced having particle sizes at fractions of a pm (micron), it displays
novel properties not found in the same material whose particles are larger
than 1.0 pm (micron). Nanotechnology involves dimensions where atoms
and molecules, and interactions between them, influence their chemical
110 The point defect

and physical behavior. Authentic nano-particles are so small that there are
many more atoms on the surface of each particle than the normal particle
of 1.0 jjm. Particles of 1.0 pm, i.e.- 1000 nm or 1000 x 10"9 m, may seem
small but those atoms on the surface of each particle are only about
0.0015% or 15 in a million of the atoms composing the lattice. A nano-
particle with dimensions of 10 nm. brings the surface atoms to about 15%
of the total atoms composing the particle. At this size range, quantum
physics and quantum effects determines the primary behavior of such
particles.

Consider that atoms have a size range of about 1-2 A. Most inorganic
solids, with the exception of halides, sulfides (and other pnictides), are
based upon the oxygen atom, i.e.- oxide = O~~ , whose atomic radius does
not change even when sulfates, phosphates and silicates are formed.
Oxide has an atomic diameter of 1.5 A or 0.15 nm. = 0.00015 nm. Nano-
particles are clumps of 1000 to 10,000 atoms. The latter would be a
particle of 0.15 (am. in diameter. They can be metal oxides, semi-
conductors, or metals with novel properties useful for electronic, optical,
magnetic and/or catalytic uses.

When light meets particles this small, it behaves differently. One example
is TiO2 (titanium dioxide), which has been used as an ultra-violet absorber
for sun-screen products. The usual product is applied to the skin as a
white-reflecting cream. The process for making titanium dioxide varies
but usually employs TiCl4 and its hydrolysis under controlled conditions.
When particles of 50 nm. are formed, the sun-screen cream now is
transparent since the particles absorb and scatter visible light much less
than the larger particles previously used. However, the ultraviolet light
absorption is not changed, only the reflection of white light.

As we shall see in the next chapter, particles are formed first as

"embryos" which are minute particles of the nano-particle class. These
then grow into "nuclei" which then grow into particles. The science of
particle growth has been a major source of our understanding of particles.
As we have already shown, lattice defects, due to thermal effects, are the
norm when a crystal grows to sizable proportions. However, when nano-
 2.1 1 Nanotechnology and the solid state 111

crystals are formed, the numbers of embryos allowed to form, with

corresponding nuclei, are controlled. The nuclei growth is then confined
to atomic dimensions. Much of this growth forms by "Spontaneous
Assembly". That is, when nano-particles are formed, atoms are added one
at a time to form the embryo and then the nucleus. It is the size of the
nucleus that is restricted.

I submit that the prediction given in the previous section, i.e.- see p. 107,
has already been realized. That is, nanoparticles form by self-assembly of
atoms (ions) into defect-free crystals. It is this lack of intrinsic defects
that give such particles their unique chemical and physical properties.
Note that if normal growth were allowed to proceed further, then we
would have the normal defect-crystal.

Suggested Reading

1. A.C. Damask and G.J. Dienes, Point Defects in Metals, Gordon &
Breach, New York (1972).

2. G.G. Libowitz, "Defect Equilibria in Solids", Treatice on Solid State

Chem.- (N.B. Hannay-Ed.), 1,335-385, (1973).

3. F.A. Krceger, The Chemistry of Imperfect Crystals, North-Holland,

Amsterdam (1964).

4. F.A. Kroeger & H.J. Vink in Solid State Physics, Advances in Research
and Applications (F. Seitz & D. Turnbull-Eds.), pp. 307-435 (1956).

5. J.S. Anderson in Problems of Non-Stoichiometry (A. Rabenau-Ed.),

pp. 1-76, N. Holland, Amsterdam (1970).

6. W. Van Gool, Principles of Defect chemistry of Crystalline Solids,

Academic Press, New York (1964).
112 The point defect

7. G. Brouwer, "A General Asymmetric Solution of Reaction Equations

Common in Solid State Chemistry", Philips Res. Rept., 9 , 366-376
(1954)

8. A. B. Lidiard, "Vacancy Pairs in Ionic Crystals", Phys. Rev., 112, 54-55

(1958).

9. J.S. Anderson, "The Conditions of Equilibrium of Nonstoichiometric

Chemical Compounds, Proc. Roy. Soc. (London) , A185, 69-89 (1946).

10. N.N. Greenwood, Ionic Crystals, Lattice Defects & Non-Stoichiometry,

Butterworths, London (1968).

11. Hayes and Stoneham, "Defects and Defect Processes in Non-Metallic

Solids"- J. Wiley & Sons, New York (1985).
The Solid State: Mechanisms of Nucleation, Solid State Diffusion, Growth
of Particles and Measurement of Solid State Reactions

In the previous Chapter, we examined intrinsic defects in the solid. In

this chapter, we will examine solid state reactions as related to nucleation
and diffusion mechanisms. We will also investigate how particles grow
during reactions to produce solids. Solids are generally created as
crystallites (particles) unless special methods to produce large single
crystals are employed. When we have particles, they will undergo solid
state reactions and change form, habit and structure if heated. The solid
state reaction occurs at the boundary of direct contact between adjacent
crystallites.

We will also examine methods of characterizing such solid state reactions,

including DTA (Differential Thermal Analysis) and Thermogravimetric
Analysis, i.e.-TGA). We will reserve the discussion of how size of particles
and a particle size distribution is determined until the next chapter.

If we wish to create a phosphor, there are two general methods to do so.

One is by precipitation from an aqueous (rarely non-aqueous) solution. The
other is by solid state reaction. Sometimes, both methods are employed.
The phenomena concerning particle growth in solution differ quite widely
from those governing growth of particles caused by solid state reactions.
Yet they retain many similarities. The formation and growth of nuclei
during particle formation and growth in either precipitation or solid state
reactions will be of major concern in this chapter.

Both single crystal and crystallite formation employ very similar methods
but differ in technique. They include formation from:
Single Crystal Crystallites (Powders!
Liquid solvent Vapor
Vapor Melt
Melt Flux
Molten salt Precipitation from solution

113
114 Nucleation, diffusion, growth of particles and measurement of solid state

Obviously, the major difference in the single-crystal and polycrystalline

(crystallite) state is a matter of size. For the single-crystal, the size is large
(> 10 cm), whereas in the polycrystalline state, the size of the crystals is
small (10 urn = 0.001 cm.) The methods for obtaining one type of crystal
or the other differ considerably.

Most solids that we encounter in the real world are in the form of
powders. That is, they are in the form of discrete small particles of
varying size, i.e.- crystallites. Each particle has its own unique diameter
and size. Additionally, their physical proportions can vary in shape from
spheres to needles. For a given powder, all particles will be the same
shape. However, their sizes can be altered by the method used to create
them in the first place. Methods of particle formation include:

Solid State Reaction

Precipitation
Condensation or Evaporation

These methods involve the formation of nuclei and the rate of their
growth. We will examine such mechanisms as they relate to both solid
state reactions and precipitation as methods for obtaining discrete
particles (crystallites). Still other processes exist after particles have
formed, including sequences in particle growth like sintering.

Nucleation can be homogeneous (no outside influences) or heterogeneous

(by specific outside coercion). In precipitation processes to form a
distribution of particle sizes, it is probably a combination of both
mechanisms. Since nucleation differs in either precipitation or solid state
reactions, we will address each separately. But, before we can investigate
nucleation in solid state reactions and compare it to nucleation during
precipitation processes to form a solid, we need to determine how solid
state reactions occur.

3.1.- SOLID STATE REACTIONS AND NUCLEATION MECHANISMS

Solid state reaction mechanisms include phase changes, formation of

 3.1 Solid state reactions and nucleation mechanisms 115

phase boundaries, nucleation, nuclei formation rate processes, diffusion

processes, and finally diffusion-controlled solid state reactions. First, we
examine heterogeneous (two different phases) and then homogeneous
(single phase) solid state reactions. The reason that we do this is that the
former is easier to understand and that the latter applies more to
precipitation processes.

I. - Types of Solid State Reactions

We can distinguish seven (7) types of phase transformations in any given

solid. They include the following:

3.1.1.- PHASE TRANSFORMATIONS (SINGLE PHASE)

CHANGE DESIGNATION
Gas to Liquid Condensation G=>L
Gas to Solid Condensation G=>S
Liquid to Gas Evaporation L=s>G
Liquid to Solid Solidification L=*S
Solid to Gas Sublimation S=*G
Solid to Liquid Melting S=>L
Solid to Solid Polymorphic Transformation S=>S

These seven transformations are familiar to most. We can classify most

solid state reactions in terms of the CHANGES given in 3.1.1. This is
shown in the following table, given on the next page.

Here, S , L & G refer to solid, liquid and gas, respectively. Note that just
one composition can experience any or all of these phase changes, i.e.- a
homogeneous solid state reaction. You will note that only decomposition
reactions can be classified as homogeneous. All of the others are
heterogeneous in nature.

Now consider phase transformations between two solids, i.e.- a

heterogeneous reaction, and the types of reaction mechanisms which may
116 Nucleation, diffusion, growth of particles and measurement of solid state

TABLE 3-1
TYPES OF SOLID STATE REACTIONS
Reaction Final Product Example
Decomposition
I—"

S +S 3 AuCl => AuCl3 + 2 Au

A=>B+ C S +G CaCO3 => CaO + CO2 t
G+G NH4C1 => NH3 + HC1 ft
2. Synthesis S 2 Agl + Hgl2 =* Ag2 Hgl4
A+ B=> C MgO + A12O3 =* MgAl2 O4
MgO + SiO2 =* MgSiO3
3 . Substitutional S +S Cu + AgCl => CuCl + Ag
A+ B=>C+ D S +G BaCO + TiO => BaTiO + CO ft
4. Consecutive S i) La2O3 + 11 A12O3 =* 2 LaAlO3 + 10Al2O3
ii) 2 LaAlO3+ 10A1O => LaAl O|ft

occur. If the reaction occurs between particles, it can only progress if the
particles are in direct contact. This phenomenon is illustrated in the
following diagram:

3.1.2.-

In the case where two particles are involved, those that are not in close
proximity will not react. But, those that adjoin each other will undergo
solid state reaction with ease. This is due to the fact that cations and/or
anions from one structure must be transported, or interchange by some
mechanism, to the other structure in order to form a completely new
compound. Thus, the degree of dispersion and mixing of one reacting
solid with another is important to the overall mechanism of solid state
reaction. This cannot be overemphasized since lack of dispersion, i.e.-
incomplete mixing of two solids, is the primary cause for production of
 3.1 Solid state reactions and nucleation mechanisms 117

unwanted phases. That is, the solid state reaction may not proceed
according to expectation if the two phases to be reacted are not mixed
thoroughly.

Suppose we have a crucible half-filled with a powder. We now fill the

crucible with another powder of different composition and then heat the
filled crucible. Any solid state reaction which may occur can only do so at
the boundary of the two layers of powders. If the reaction is: A + B => AB,
then we find that the reaction product, which is also a solid, forms a
phase boundary between the two layers. This is illustrated in the following
diagram:

It is this same condition that exists in a solid state reaction between two
crystallites of differing composition. When A reacts with B, and vice versa,
a phase boundary is formed at each interface, i.e.- between A and AB, and
between B and AB. This grows outward as shown above. Once the phase
boundary is established, then each reacting specie must diffuse through
the phase {AB} to reach its opposite phase boundary in order to react.
Note that in our example, A diffuses through AB to the right while B
diffuses to the left. Thus, the overall solid state reaction is dependent
upon the rate of diffusion of the two (2) species. These two rates may, or
may not, be the same. The reason that A and/or B do not react in the
middle, i.e.- the phase {AB}, is that AB has a certain ordered structure.
Thus, in our model, diffusion by A through AB occurs until A encounters B
118 Nucleation, diffusion, growth of particles and measurement of solid state

on the other side. Diffusion by B occurs in the same manner but in the
opposite direction.

Although the above diagram is a simple model, this is not how it occurs in
Nature. We have presented the above concept because it is easier to
understand than the actual conditions which occur. It should be clear that
the overall solid state reaction is dependent upon the rate of diffusion of
the two (2) species, and the two rates may, or may not, be the same. The
reason that A and/or B do not react in the middle, i.e.- the phase {AB}, is
that AB has a certain ordered structure which probably differs from either
A or B. But there is a more important reason which is not easily illustrated
in any two-dimensional diagram, particularly since each compound has a
three-dimensional lattice structure.

The actuality is that any given atom of A does not move through AB to the
phase boundary of A and AB. What really happens is that any given atom
displaces one of the A atoms in {AB}. The displaced A-atom then causes
another displacement by a "hopping" motion. The stated displacement
then travels to the right until the other edge of the phase boundary is
encountered. There, reaction of the final displaced A-atom occurs with a
B-atom to form AB. Simultaneously, displaced B-atoms are diffusing to the
left by the same "hopping" motion. Thus, the walls of {AB} move in a
three-dimensional direction at the expense of both of the atomic volumes
of A and B. A depiction of this mechanism is given in the following
diagram, given as 3.1.4. on the next page.

In this diagram, we have shown an "A" ion (shown as a black sphere)

which approaches the surface of "AB" and displaces an "A" ion in this solid
phase. A series of "hops", i.e.- from "1" to "7", then occurs in the AB-phase
with the final "A" ion ending within the B-phase where a displacement in
the normally cubic "B" phase occurs. At the same time, the displaced "B"
ion is diffusing in the opposite direction by a series of "hops", i.e.- from
"a" to "e" to the interface of "AB" with "A". Note that the "A" phase is
shown as a hexagonal phase while "B" is a cubic phase (as is the "AB"
phase).
 3.1 Solid state reactions and nucleation mechanisms 119

It should be clear that the rate of diffusion of "A" will differ from that of
"B". Note also that a change in structure has occurred as the AB
arrangement grows.

Let us take another example, using the same type of model, in which A is
a gas and B is a solid. A diagram illustrating this mechanism is given as
follows:

Here, only one interface forms as the reaction takes place initially at the
surface of "B". Once the AB-phase has formed, gaseous A atoms must then
diffuse through to the phase boundary in order for the reaction to occur.
The wall of {AB} thus moves in two (2) dimensions instead of three, in
contrast to the case of 3.1.3.
Consider another example involving diffusion and reactivity, such as the
reaction between particles of:
120 Nucleation, diffusion, growth of particles and measurement of solid state

3.1.6.- BaO + SiO2 =*• BaSiO3

In this case, the structure of BaO consists of discrete atoms of Ba2+ and
O=, arranged in a cubic structure. SiO2 , on the other hand, consists of
SiO4 tetrahedra bound at the apices through mutual oxygen atoms, and
arranged in a three-dimensional network. This is shown in the following
diagram:

3.1.7.-

We find that this solid state reaction is very slow, even at 800 °C. , and
occurs at the interface of the two types of particles. The reaction is slow
because it is diffusion-limited. Since the silica- network is three-
dimensionally bound, the only reaction that occurs is caused by the
diffusion of Ba2+ atoms within the network, as shown in the following:

Note that diffusion occurs only in one direction because the silica
 3.1 Solid state reactions and nucleation mechanisms 121

tetrahedra are not free to move. What is actually happening is that the
three-dimensional network of tetrahedra is being rearranged to form
another structure. This denotes the fact that the actual structure and
composition of the two reacting species are the major factor in
determining the nature and speed of the solid state reaction.

On the other hand, if we use BaCC>3 as a reactant, namely:

3.1.9.- BaCO3 +SiO 2 => BaSiO3 + C O 2 t

we find that the reaction is very fast. This has nothing to do with the gas,
GO2 , produced, but with the form of the BaO particles. In this case, the
initial reaction is:

3.1.10.- BaCO3 => BaO + CO2 ft

The BaO is produced in the form of very small particles of nearly atomic
proportions (nano-particles) which react immediately to form the silicate.
Actually, the rate of reaction is proportional to the number of nuclei
produced per unit volume. A nucleus is a point where atoms or ions have
reacted and begun the formation of the product structure. In the case of
the BaO + SiO2 reaction, the number of nuclei formed per unit of time is
small and formation of the structure is diffusion limited. In the case of
Ba(X>3 decomposition, the atomic-proportioned BaO reacts with SiO2
nearly as fast as it is formed so that the number of nuclei per unit volume
is enormously increased. It is thus apparent that if we wish to increase
solid state reaction rates, one way to do so is to use a decomposition
reaction to supply the reacting species. We will further address this type
of reaction later on in our discussion.

Note also that we have just introduced the concepts of nuclei and
nucleation in our study of solid state reaction processes. Our next step
will be to examine some of the mathematics used to define nucleation. We
will find many areas of correspondence in nucleation rate processes
between solid state reactions and precipitation processes.
122 Nucleation, diffusion, growth of particles and measurement of solid state

II. - Solid State Nucleation Processes and Models

Let us now examine how the nucleation of particles occurs during the
initial stages of a solid state reaction. In general, the rate of solid state
nucleation is similar to the classical rate of chemical reaction, and is given
by.

3.1.11.- Rates of Nucleation

Homogeneous: r = l/V(dni/dt)
Heterogeneous: r = l/V«,(dVt/dt)

where ni is initial mols, Vt is volume at a given time, t, and Voo is the final
product. Why we should use volume as a limiting factor in reactions
between solids should be clear in light of the examples given above.

We will examine heterogeneous nucleation first and homogeneous

nucleation (no outside influences) later. We do this because the former is
easier to understand and forms a basis for understanding the latter.

If we are interested in the solid state nucleation process to form a particle

prior to completing the reaction, we can distinguish between surface and
volume nucleation of the particle. Such cases are shown in 3.1.12. on the
next page as follows.

Starting with a particle, we can transform it either to a gas, a liquid, or a

solid. Model A is determined by a phase transformation wherein the initial
solid is transformed to a fluid (melting) or a gas (sublimation), i.e.-

3.1.13.- As + B L => (AB)L

or:
As +BQ => (AB)G

An example of this reaction is A1(NC>3) 3 • 9 H2O which dissolves in its own

water of hydration, i.e.- BL. When heated, the salt loses water at slightly
less than 100 °C. The water does not evaporate but begins to dissolve the
 3.1 Solid state reactions and nucleation mechanisms 123

salt, so that at 200 °C, one has a fluid which is stable up to about 400 °C.
Yet, if one starts with the anhydrous salt, A1(NC>3) 3 , one obtains AI2O3 at
400 °C. instead.

Model B involves surface nucleation wherein the surface becomes

124 Nucleation, diffusion, growth of particles and measurement of solid state

nucleated in one spot, i.e.- two reacting particles in close proximity,

shown as follows:

3.1.14.-

Two stages usually occur, nucleation and then nuclei growth. If surface
nucleation is fast (Model C) it is likely due to reaction of a gas with the
solid particle. The reaction of a liquid is the other possibility:

3.1.15.- Ag + Bg => (AB)S

As + BL => (AB)S

Model D is an example of the decomposition of a solid whereas Model E is

that involving gas or liquid nucleation of the solid. Note that if nucleation
does not occur (this mechanism is not usually the norm), the solid reacts
uniformly throughout its whole volume (Model F).

In order to show how the theory for the number of nuclei formed per unit
time was developed, we consider the following. First, we assume that a
nucleus will be spherical (to minimize surface energy). The volume will
be:

3.1.16.- V = 4/3jrr3

We use xdt as a radius to define a growth rate and define N as the number
of nuclei formed in t = y. Then the volume, V, formed as a function of time
will be:

3.1.17.- V(t) = J { 0 [ J x d t ] 2 (dN/dt) t = y } dy

This is the general form in case we do not have spherical nuclei. If we

do:
 3.1 Solid state reactions and nucleation mechanisms 125

3.1.18.- V(t) = J {6 J { 4^/3 [ /xdt ]3 (dN/dt)t = y } dy

It is easily seen that e is a geometrical or shape factor and we define X as

an exponent for the radius of the nuclei, xdt (it is 2 in 3.1.17. and 3 in
3.1.18.). In this case, we have both anuclei volume and anumber factor.

Now, consider the transformation to form a nucleus in terms of the rate

instead of a volume. We define:

3.1.19.- Rate of nucleation(I) = # of nuclei formed/unit time/unit volume

If the nucleation is random, we will have the homogeneous case. If it is

specific (with foreign or phase boundary walls), we have the
heterogeneous case, as shown in the following:

3.1.20.-

This drawing illustrates the formation of an embryo B within the

crystallite, A. (an embryo is an incipient nucleus - not drawn to scale).
The change in free energy in going from A to B will be a function of both
volume and surface area (shape). We must discriminate between the
interfacial surface energy between A and B, and the internal energy of B.

Thus, the change in free energy will be:

126 Nucleation, diffusion, growth of particles and measurement of solid state

3.1.21.- A G = N (GV + o) X + p N2/3 Y

where Gy is the difference of free energy of B within A; N is the number

of nuclei; X is the nuclei volume; p is a shape factor; a is the interfacial
tension or surface energy; and Y is the interface exchange energy. The
first part of 3.1.21 is the volume factor while the last part is the surface
factor. But this is the general case, i.e.- for 3.1.17. For a spherical
nucleus, i.e.- 3.1.18., we get:

3.1.22.- A G = Nj 4 / 3 7T r3 (G v + a) + 47rr 2 N2/3y

We now differentiate so as to obtain zero surface energy:

3.1.23.- d(A G)/dr = 4n N4 r 2 (G v + a) + 8 n N 2 /3 y

If we have zero surface energy , then d(AG)/dr = 0 and a= 0. Thus we

have:

3.1.24.- rCrit = - 2 II AGV

which is the Gibbs-Thompson equation defining the critical radius

required for an embryo to form. Note that we have assumed that N 2 / 3 /N a
1 in the calculation. The internal energy (A Gmax = W) is found in the
same way:

3.1.25.- W = 1/3 (Ait r c r i t Y) = 16/3 { JT Y3 / AG V }

Since from thermodynamics:

3.1.26.- AGV = RT In P/P equ il • 1/Vcrit ,

then:
AGV = HV { (T - T e q u i i / Tequil }

where equil. refers to the equilibrium state, V cr it is the critical volume,

 3.1 Solid state reactions and nucleation mechanisms 127

and both Qy and Hy still refer to the difference between the states of A
and B. Using our equations already developed, it is easy to arrive at:

3.1.27.- I = Z exp (- E o / RT) exp (- W/RT)

where E o is the embryo motion energy. If N o is the number of original

embryo nuclei present, then:

3.1.28.- I = dN/dt = k i ( N 0 -N) 6

where ki is a rate constant and 5 is equal to the number of steps involved

in the interface change of structure. Thus:

3.1.29.- N =N O [ 1-exp (-kit)] or: I = k i N 0 exp (-kit)

We note that the growth of nuclei is associated with the motion of a phase
boundary.

We can distinguish between the interface structure of the embryo, B, and

the particle, A, as shown in 3.1.30., given on the next page. In this case,
we have classified the nucleation processes in terms how well the lattices
match each other. That is, how well the embryo lattice matches that of
the nucleus and the ultimate particle. You will note that the embryo has to
form first before it is transformed into the larger nucleus, which in turn is
the basis for growth of the particle.

In the first case, the lattices match fairly closely and are coherent.

In the second case, there is some correspondence between the lattices.

But the incoherent case shows little matching of the two lattices.

It is therefore not surprising that the interface structure has a large effect
not only on the interfacial energy, but also upon the thermodynamics of
change from A to B, and the rate of nuclei formed.
128 Nucleation, diffusion, growth of particles and measurement of solid state

3.1.30.-

We can now correlate 3.1.17. and 6 of 3.1.28. with the cases given in
3.1.30. This is given as 3.1.31. as follows on the next page.

Note the increase in complexity of embryo formation, the interface

structure and the number of steps, 6, needed to do so. In the simple case,
B grows at the expense of A. However, in the second case involving
parallel reactions, two embryos form from A. In the third case, at least
two steps are involved in the two consecutive reactions where "A"
transforms to "B" , which transforms to "C".
 3.1 Solid state reactions and nucleation mechanisms 129

3.1.31.- Embrvo Growth Rate Equations for a Simple Nth Order

Let us now reconsider our nucleation models of 3.1.12., specifically

Models B, D and E. These are examples of phase-boundary controlled
growth involving random nucleation. Assume an exponential embryo
130 Nucleation, diffusion, growth of particles and measurement of solid state

By suitable manipulation, we can get (by Taylor Series expansion):

3.1.32.- RANDOM NUCLEATION

P = Vt/Voo = 8JT N o k 2 3 /Vo ki3 exp [(-kt) - (-kt) 2 /2! + (-kt)3/3! ]

where we have only given the first three terms of the series. If N o is large,
then ki will be small:

3.1.33.- p = {TT N o k 2 3 k i / 3 Vo} t 4

where p is the fraction reacted. If ki is large, then the first term

predominates, and we get:

3.1.34.- p= { 4 j t N o k 2 3 k i / 3 Vo} t 3 ,
and,
dp /dt = {4JT N o k 2 3 k i / 3 Vo} t 2

The effective interfacial area for our nuclei would be (1- a), so that:

3.1.35.- p = l - e x p { 4 j t N o k 2 3 k i / 3 Vo} t 3

But since:

3.1.36.- dp =(1-P) dPex

(where dp e x is defined as the rate of extended nuclei formation), we can

rearrange 3.1.34. and integrate to obtain:

3.1.37.- J dpex = J d p / ( l - p ) = - l n ( l - p )

which is one of the forms we assumed to be correct when we started this

discussion.

We note that the above equations enable us to calculate p, the fraction of

nuclei formed, in terms of a rate constant and the time, t. Note again that
 3.1 Solid state reactions and nucleation mechanisms 131

ki is the rate constant for nuclei formation whereas k 2 is the rate

constant for nuclei growth.

There is another way to approach the same problem. We can define a

volume of domains (nuclei) ready to grow at time, t. This means that ki is

large. Following the same method as given above, we arrive at:

3.1.38.- Pex = 4/3 TT k 2 3 J (t - y)3 I dy

where I = rate of nudeation for a unit volume so that Idy is set equal to
the total number of nuclei present. All other terms have been previously
defined. Thus, if ki is truly a constant, then:

3.1.39.- (1-P) =TT /3 k 2 3 k i t 4

But if nuclei are already present, then:

3.1.40.- - l n ( l - p ) = 4 / 3 T T N 0 k 2 3 t 3 = k 2 1 t r = Z exp (-E/RT) t r

where r is an exponent which depends upon the shape factor, 6 , of the

reaction. We can show how various mechanisms and shape factors affect
the time factor and energies involved in nuclei formation and growth.

All of these equations allow us to calculate numbers and rate of creation of

nuclei in mechanisms of crystal growth. We can also show the differences
in both phase-boundary controlled and diffusion-controlled nudeation
reactions as a function of both constant and zero rate of nudeation. This is
given in Table 3-2, given on the next page.

Referring to 3.1.17. , we see that the shape factor, 0, has a major effect
upon r and the exponent of time, t, as well as on the activation energies.
Ei in the table is the activation energy for nuclei formation, whereas E 2 is
that for nuclei growth. ED is the activation energy for diffusion growth.
This is the first time we have mentioned diffusion as a mechanism in
nuclei formation.
132 Nucleation, diffusion, growth of particles and measurement of solid state

TABLE 3-2
NUCLEATION AS A FUNCTION OF THE SHAPE FACTOR, 6

PHASE BOUNDARY CONTROLLED DIFFUSION CONTROLLED

[p(t) == k 2 V ] IP(t) = D/t ]
e I E_ r E

spheres 4 3 E2+Ei 2.5 3/2 ED + Ei

Constant
Rate of plates 3 2 E2 + Ei 2.0 E 2 + ED
Nucleation
needles 2 E 2 + Ei 1.5 1.5 E D + E I

spheres 4 3 E2 1.5 1.5 ED

Zero Rate
of plates 3 2 E2 1.0 ED
Nucleation
needles 2 E2 0.5 0.5 ED

Actually, there are three (3) major classifications for nuclei growth in
heterogeneous solid state reactions, namely:

3.1.41.- Nuclei Growth in Solid State Reactions

Phase-boundary Controlled
Unlimited Nuclei Growth
Diffusion Controlled

These mechanisms should now be self-evident. We have now described

the formation of a phase boundary and how it grows (Section I.) and have
just completed Section II. concerning nuclei growth. Now, we will
consider diffusion mechanisms and diffusion-controlled solid state
reactions without concern about how nuclei form or grow in the following
section.
 3.1 Solid state reactions and nucleation mechanisms 133

HI. - Diffusion Controlled Solid State Reactions

Let us now examine some solid state reactions which are regulated by
diffusion mechanisms. One such is the "rusting" of iron exposed to the
weather. Over a period of time, we have observed that iron-based parts of
machinery become coated with a layer of iron oxide.

a. The Tarnishing Reaction

Wagner (1902) published an analysis of the oxidation of the surface of a

metal, based on diffusion reactions, which has remained a classic. This
mechanism is called by various names including "Tarnishing". The surface
of a metal consists of metal atoms bound to the inner structure by a series
of hybrid-bonds. If oxygen gas is present (air), the metal, particularly iron,
will form an oxide coating, vis:

3.1.42. - M + l / 2 O 2 => MO

In several known cases, the oxide-coating protects the surface from

further oxide buildup. Yet, there are many cases where the oxide layer
continues to grow until the metal is totally consumed. How is this
possible?

Wagner hypothesized that both metal and oxide ions diffused through the
metal oxide layer so as to build up the layer thickness from both sides, as
follows:
3.1.43.-
134 Nucleation, diffusion, growth of particles and measurement of solid state

In this diagram, oxidation, i.e.- loss of electrons, occurs at the metal

surface while reduction, i.e.- acquisition of electrons, occurs at the metal
oxide-oxygen gas interface. That is, M, in the metallic state, is "oxidized",
i.e.- forms an ion, plus requisite electrons, e~ . Both diffuse through the
metal oxide layer to the surface where oxygen gas, O2, is "reduced" to
oxide, O" , by the electrons which accompany the metal ion because of
charge-compensation requirements in the diffusion mechanism.
Generally, the electrons have a much higher mobility than the metal ions
formed so that the outer surface of the film acquires a negative charge
while the inner surface becomes positive in surface charge. These effects
are the primary cause of charged-ion diffusion. In most cases, the metal
ions diffuse more rapidly than the oxide ion and fix the rate of growth of
the oxide layer. Charge compensation requires that no net flow of current
occurs but the electric field gradient does affects the mobility of the two
diffusing ions, M2+ and Qr . Note that we have a classical "oxidation-
reduction" reaction, but in the solid state.

It should be apparent, from the above diagram, that the thicker x

becomes, the slower is the change in the x-dimension, i.e.-

3.1.44.- dx/dt « l / x or: d x / d t = K - 1/x

The integrated form is:

3.1.45.- x 2 = K t + C, or: x = D(t) 1 ' 2

where D is the Diffusion Coefficient. Equation 3.1.45. is called the

Parabolic Law of Diffusion. If the growth of a phase can be fitted to this
equation, then it is likely that the primary reaction mechanism involves
simple diffusion of ions. However, other mechanisms of tarnishing other
than simple diffusion have also been observed. Obviously, this depends
upon the nature of the metal involved.

The mechanisms appropriate for the growth of thin films less than
approximately 10 ~4 cm. do not apply to the growth rate of thicker oxide
films and layers. This is shown in the following:
 3.1 Solid state reactions and nucleation mechanisms 135

3.1.46- Growth Rate Equations for Tarnishing Reactions

Linear = Ax = kL t
Parabolic = Ax = kp t1/2
Cubic = Ax = k c t1/a
Logarithmic = Ax = k e hi (kt + t j
Inverse Logarithmic = Ax = [ke hi (kt + t j ] ' 1

In the above equations, x is the thickness, t is the time and k is a rate

constant. All of these rate laws have been determined experimentally.
Sometimes, combinations of two or more may be involved. Nevertheless,
it is often nearly impossible to equivocally decide which equation applies
to a given case.

If the molar volume of the newly formed oxide coating is less than that of
the source metal, the film tends to be incoherent and non-protective.
This behavior provides no barrier between the fresh metal surface and the
oxygen gas. The corrosion proceeds as if the oxide layer or scale was not
present. The rate of reaction is therefore constant and is described as:

3.1.47.- dx/dt =k L

This integrates to give the linear growth rate law. Such behavior explains
how the tarnishing mechanism can completely destroy a metal if left
unchecked. It should be clear that what seems to be a simple case of
diffusion may be in actuality very complex.

Diffusion mechanisms also involve the following defect reactions:

3.1.48. - Interchange of Atoms

Vacancy Hopping
Interstitial Movement
Dissociative Exchange

In the first mechanism, Mx and XM interchange may be involved, whereas

the last involves a mechanism similar to that given in 3.1.43.
136 Nucleation, diffusion, growth of particles and measurement of solid state

b. Fick's Laws of Diffusion

Following the success of the Tarnishing Law formulated in 1903. Fick, in

1932, formulated laws which described diffusion processes in solids.
Consider a lattice such as that shown as follows:

With this three-dimensional lattice, it is easy to see that there are six (6)
ways for an atom (ion) in the x-plane to move to the y-plane. Using this
concept allows us to write:

3.1.50.- J x => y = 1/6 • 1/ T nAR • a

Thus, the concentration of A-atoms in the y-plane will be:

3.1.51.- nA(y) = nA+achiA/dx

Combining the two equations gives us:

3.1.52.- J x => y = (nA+ adnA/dx) (aR/6T)

 3.1 Solid state reactions and nucleation mechanisms 137

If we assume that J y => x is also valid, then we have:

3.1.53.- J = J x =>y - J y =s> x

where J is the total number of jumps that occur. Thus:

3.1.54.- J = aR/6T -(nA - n A '+ a dn A /dx- aR/6T = - a 2 R/ 6x • dn A /dx

We now define a diffusion coefficient as:

3.1.55.- D= a2/6X o r D = a a2 / T (interstitial)

where a is the number of interstices in the lattice, for the interstitial case
(see 3.1.48.). This gives us:

3.1.56.- J = - D R d n A / dx

If we now define concentration in terms of n A , i.e.-

3.1.57.- c E Rn A , so that: dc/dx = R dn A /dx

then we get PICK'S First Law:

3.1.58.- J = - Ddc/dx

in which we have assumed that discrete jumps take place, and there are
no concentration gradients between planes. If such exists, then dc/dx *
constant, and we must then subtract two different jump rates:

3.1.59.- J1-J2 = - AxdJ/dx= - Ax (dc/dt)

If the last term is taken only in respect to the x dimension, then:

3.1.60.- [dc/dtJx = e p a c / a x } or (dc/dt)x = D (32 c / dx2)

138 Nucleation, diffusion, growth of particles and measurement of solid state

which is Fick's Second Law.

We can also show that the diffusion coefficient, D, is temperature

dependent, and can fit it to an Arrhenius equation:

3.1.61.- D=Doexp-E/kT

Sometimes, a more useful term is:

3.1.62.- ED = -kid(lnD)/d(l/T)

Let us now reexamine 1IX , the jump frequency, so as to relate diffusion

processes to lattice vibration processes. The reciprocal of the time of stay
is the jump frequency and is related to the diffusion coefficient by:

3.1.63.- D = a2 /6T = a a2/ X aaroa2

where w = 1IX , and is the jump frequency, w can be related to the free
energy of the lattice via a Boltzmann distribution, i.e.-

3.1.64.- tn - wo aexp (-AG0 /rt) = w o aexp (-AS/R - AH/RT)

by a series of mathematical manipulations. It turns out that TOO is the

atomic vibrational frequency or the zero phonon mode of the lattice, and:

3.1.65.- tno = 1/2JT /(K/m)

where K is the elastic constant of the lattice and m is the reduced mass of
the atoms composing the lattice. Note that we have just related the
diffusion coefficient, D, to the zero-phonon mode of the lattice. This is
reasonable since the jump frequency ought to be a function of the phonon
vibrational constant as well.

We will now examine the role of defects in solid state diffusion reactions
to form nuclei and eventually particles.
 3.1 Solid state reactions and nucleation mechanisms 139

IV«- Role of Defects in Solid State Diffusion Mechanisms

To this point, we have examined diffusion growth in terms of nucleation

and embryo formation. Let us now scrutinize the actual species which
diffuse.

WE WILL FIND THAT ALL DIFFUSION MOTION OCCURS BY DEFECT

MOVEMENT IN THE LATTICE .

Consider the following diagram:

Self diffusion can occur by at least two mechanisms, vacancy and

interstitial. Both are "hopping" motions, as described above. Actually, we
must account for all types of defects, including charged species (see
2.2.1.). To do this, let us reconsider the tarnishing reaction (Section D-I),
with the general reaction:

3.1.68.- M+ 1/2X2 =>MX

Such a reaction would occur if we exposed a metal surface to either

oxygen or chlorine. A MX film would build up on the metal surface and
growth would occur by diffusion. In the initial description, we ignored
vacancy and interstitial diffusion and presented only the charged
particles, M 2+ andO= as the diffusing species.
140 Nucleation, diffusion, growth of particles and measurement of solid state

In actuality, the metal diffuses as Mi 2+ and the anion as O r (Note that

interstitials are indicated).

One would expect that equal amounts of each specie would diffuse in
opposite directions, thus preserving electroneutrality. While this is true
for the reaction in 3.1.68., i.e. - M 2+ => : <^= X r , what of the case for
BaSiO3 where diffusion is limited to one direction? It is not reasonable
that the solid should build up a charge as the M2+ ions are diffusing and we
must search for compensating species elsewhere. It turns out that charge
compensation occurs by diffusion of charged vacancies.

To illustrate this, let us use the reaction in 3.1.68. and the following
diagram:

3.1.69.-

We must have electroneutrality in diffusion, i.e.- pairs of defects , and in

the above example, we can write nine equations, of which the following is
just one case:

3.1.70.- Mi+ =>= -s=Xr

We can also write equations involving charge only, or vacancy plus charged
species. However, we have not considered rates of diffusion in our model.
Because of the vast differences in the types of diffusing species, there is
no reason to expect all of them to diffuse at the same rate, particularly
when we compare electrons and vacancies. The former will be very fast
 3.1 Solid state reactions and nucleation mechanisms 141

while the latter will be very slow. The Kirchendall Effect deals with this
manifestation.

Suppose we have a situation where A reacts with B to form a solid

solution, AB. Let us further suppose that the diffusion reaction is:

3.1.71.- Ai +V A = B; + VB

In our case, we will ignore charged species. Let us also suppose that:

3.1.72.- Dy(A) » Dv(B) so that: dcy(A)/dt » dcy(B) /dt

Given these restraints, we obtain a rather queer result. The mechanism

gives rise to the creation of new sites across the diffusion zone and
actually causes a deformation in the solid because the VA defects pile up
and finally become annihilated because of clustering. Actual holes appear
in the solid due to vacancy diffusion, vis-

Note that in this case, we are not forming anew compound through solid
state reaction, but are forming a solid solution of A and B. The
Kirchendall Effect has been observed many times, but occurs most often
in reactions between metals to form alloys.

There are three (3) types of diffusion-controlled reactions possible for

heterogeneous solid state reactions. They are given in 3.1.74. on the next
page:
142 Nucleation, diffusion, growth of particles and measurement of solid state

3.1.74.- TYPES OF DIFFUSION REACTIONS

Simple Diffusion: P = ki (t) 1 / 2

Phase-Boundary Controlled p = k>2 t
Material Transport

Note that these equations relating diffusion reactions resemble the

mechanisms already given for nuclei growth in 3.1.41.

For simple diffusion-controlled reactions, we can show the following

holds:

3.1.75.- x 2 = 2 D c o t V M + (a) 1 / 2

where c o = concentration of constituents at interface; V = volume of

product AB per mol of reactants; a = surface layer thickness at interface
when t = o. It is also well to note that the final volume of the product, AB,
may not be the same as that of the reactants, vis.-

3.1.76.-

For phase-boundary controlled reactions, the situation is somewhat

different. Diffusion of species is fast but the reaction is slow so that the
diffusing species pile up. That is, the reaction to rearrange the structure
is slow in relation to the arrival of the diffusing ions or atoms. Thus, a
phase-boundary (difference in structure) focus exists which controls the
overall rate of solid state reaction. This rate may be described by:

3.1.77.- dx/dt = k i ( s t / V o ) - 1- (l-x) n = k i t / r 0

where St = instantaneous surface area; Vo = original volume of particles

reacting; r 0 = original radius of particles; and n = 1, 1/2 , 1/3 for a l - , 2-,
 3.1 Solid state reactions and nucleation mechanisms 143

or 3- dimensional reaction. Material transport frequently involves an

external gaseous phase, but a general formula has not evolved.

The above equations summarize the three (3) diffusion mechanisms, one
of which usually predominates in any given case (see 3.1.74.). But these
equations contain quantities that are hard to measure, or even estimate. A
much better way is to follow the method of Hancock and Sharp ( ~ 1938).
If one can measure x, the amount of reactant formed in time, t, then the
following equation applies:

3.1.78.- -log find-x)) = m l n T + In k

range of m
simple diffusion 0.57 - 0.62
nuclei growth 1.00-1.15
phase boundary 1.25-3.00

If [- log (ln(l-x)] is plotted against In t, one obtains a value for the slope,
m, of the line which allows classification of the most likely diffusion
process. Of course, one must be sure that the solid state reaction is
primarily diffusion-limited. Otherwise, the analysis does not hold.

V.- Analysis of Diffusion Reactions in the Solid State

Let us now turn to diffusion in the general case, without worrying about
the exact mechanism or the rates of diffusion of the various species. As an
example to illustrate how we would analyze a diffusion-limited solid state
reaction, we use the general equation describing formation of a compound
with spinel (cubic) structure and stoichiometry (Spinel is the name
applied to both one particular compound, i.e.- MgAl2O4 , and a group of
compounds all having the same cubic structure and composition).

3.1.79.- AO + B2O3 => AB2O4

Here, A and B are two different metallic ions and O is the oxygen atom (as
the oxide). Spinel is a cubic mineral composed of magnesium aluminum
oxide or any member of a group of rock-forming minerals. All of these are
144 Nucleation, diffusion, growth of particles and measurement of solid state

metal oxides with the general composition AB204 , in which A may be

magnesium, calcium, iron, zinc, manganese, or nickel. B may be
aluminum, tin, chromium, or iron, and O is oxygen (Note that A is divalent
and B is trivalent and sometimes quadrivalent).

Blasse (1964) listed close to 200 spinels having either a "normal" or

"inverted' spinel structure. What this means is that the cations normally
occupying the "A" site would occupy the "B" site would be exchanged,
depending upon the ionic radius of the two cations. Thus, if we could
make Mg8* smaller in radius, and A P were made larger, we would have:
Al2MgO4 as an inverted spinel. In normal spinels, the divalent cations
occupy tetrahedral sites while the trivalent cations occupy the octahedral
sites. The inverted state depends upon which cations are involved and
their relative ionic size. Thus, we have two cation "sub-lattices" in the
spinel lattice, the tetrahedral or A-sublattice and the octahedral B-
sublattice.

a. ANALYSIS OF DIFFUSION REACTIONS IN SPINEL

There are two (2) different cases that we can distinguish, both of which
represent possible diffusion mechanisms in spinel. These are: Ion
diffusion or Gaseous. Let us consider the ion-diffusion mechanism first.
This is shown in the following diagram.

Here, we have shown A2+ and B 3+ diffusing through AB2O4. The actual
reactions involved are:
 3.1 Solid state reactions and nucleation mechanisms 145

3.1.81.- 3A 2 + +4B2O3 => 3AB 2 O 4 + 2B3+

2 B3+ + 4 AO =* AB2O4 + 3 A2+

4 AO + 4 B2O3 => 4 AB2O4

Notice that the partial reactions in 3.1.81. are balanced both as to

material and charge. There are at least two other possible mechanisms, as
shown in the following diagram:

In one case, the diffusion of A2+ is much faster than B3+ and in the other
the opposite is true. Note that charge-compensation of migrating species
is maintained in all cases by the diffusion of oxide species.

We can also illustrate gaseous transport reaction in the same manner, as

shown in 3.1.83., given on the next page. The partial reactions are:

3.1.84.- O= =*> 1/2 O2 + 2 e-

A2+ + 2 e- + 1/2 O2 + B2O3 =*- AB2O4

In this mechanism where DA2+ » DB3+, transport of external oxygen gas

is involved in the solid state, accompanied by electronic charge diffusion.

There is still one more mechanism to consider. That is the one where the
diffusion rate of B3+ is much greater than that of A, i.e.- DB3+ » DA2+ .
146 Nucleation, diffusion, growth of particles and measurement of solid state

3.1.83.-

Then transport of O2 is just opposite that shown in 3.1.84., and the

reactions are:

3.1.85.- 6O= =>3/2O2+6e-

AO + 3/2 O2 + 6 e- + 2 B3+ => AB2O4

We have now presented all of the possible diffusion reactions in spinel

synthesis. The next step will be to determine which of them is the valid
one.

To start, we measure formation rate in air, i.e.- in pure oxygen gas, and
then in an inert gas. If the rates do not differ significantly, then we can
rule out gaseous transport mechanisms. There are other tests we can
apply, including electrical conductivity, transference numbers and
thermal expansion. Although these methods have been investigated in
detail, we shall not present them here. It suffices to say that:

"For the most part, in oxygen-dominated hosts, diffusion

by cations (small) prevails".

Thus, we write the solid state diffusion reaction for spinel as follows in
3.1.86. on the next page. In this case, Frenkel pair diffusion
predominates and is fester than any other possible mechanism in the
spinel lattice.
 3.1 Solid state reactions and nucleation mechanisms 147

3.1.86.- 3 Mg2+ + 4 A12O3 => 3 MgAl2O4 + 2 A13+

2 Al 3+ + 4 MgO => MffAfoOA + 3 Mg 2+
4 MgO + 4 AI2O3 => 4 MgAl2O4

Frenkel Pain AIAI + °°Vi % VAI + Alj 3+

b. PHOSPHORS BASED ON SPINELS

The spinel group is divided into three immiscible series, the spinel in
which Bis aluminum (aluminum-spinel), the chromite (chromium-spinel)
series in which B is chromium; and the magnetite (iron-spinel) series, in
which B is iron.

For reasons which will become clear later, we cannot use chromium-
spinels or iron-spinels as a phosphor base. Thus, we are limited to
aluminum, tin and the like. We will limit our discussion to tin and gallium-
based compounds, i.e.- Mg2 SnO4 and Mg2GaO4.

Both of these materials can be "activated" by Mn4* which also forms a

spinel, i.e.- Mn2SnO4. The latter asaphosphor, i.e.- Mg2GaO4:Mn2+, is used
as the source of green light in most copying machines today. It is
interesting to note that the cubic spinel structure is formed by a majority
of dibasic cations like Mg2* and that many trivalent or quadrivalent cations
can form oxide-based anions like stannate, aluminate, titanates, vanadates
and the like.

To prepare a phosphor, we need to follow a certain series of steps. We

will go into more detail in another chapter, but here we only present a
bare outline of the preparation of the phosphor- Mg 2 Sn0 4 : Mn2+ , where
the actual formula is a solid solution of: Mg2SnO4 • Mn 2 SnO 4

As shown in the last chapter, "luminescent-grade" materials must have no

more than a total of 100 ppm of impurities. The reaction that we will use
to prepare our phosphor is:

3.1.87.- 2 (MgO + MnO) + SnO2 => (Mg,Mn)2SnO4

148 Nucleation, diffusion, growth of particles and measurement of solid state

We will also use a magnesium compound that will decompose upon

heating to form the oxide, thus ensuring that the solid state reaction will
proceed to 100% completion, as we have already illustrated. However,
magnesium does not form a carbonate directly so we have to use a
hydroxy-carbonate, i.e.- 3MgCO3 • Mg(OH)2 • 3 H2O. A suitable phosphor
formulation is given in the following:

3.1.88.- 2.500 mol of MgO (as 3MgCO3 Mg(OH)2 3 H2O)

1.000 mol SnO2
0.025 mol MnCO3

Since we intend to use this phosphor in a fluorescent lamp, we need to

control its ultimate ultraviolet absorption properties. It is axiomatic that
lamp phosphors must not contain an excess of a UV-absorbing
constituent. In our case, this is SnO2. We thus add an excess of MgO.
However, the magnesium compound only produces about 45 gm of MgO
per 100 gm. of the hydroxy-carbonate. That is, it has an "assay" of 44.6%.
The above formulation thus needs to be adjusted to compensate for this
factor. Additionally, we would normally assay all of the other components
as well. This is done by firing them to obtain the requisite oxides which
are actually the reacting components in the above reaction of 1.3.86. Note
that in some cases, the assay reflects only the amount of adsorbed water,
i.e.- SnO2 would have an assay of about 99.4%.

The procedure is as follows:

1. Weigh out 100.8 gm of MgO (corrected, this is 226.0 gm of

hydroxy-carbonate), 134.7 gm of SnO2 and 2.125 gm of MnCO3.
2. Mix these powders thoroughly by use of a hammermill and/or
other solid state blender.
3. Prefire the mix in a covered silica crucible at 2000 °F (1,093 °C)
in air for 2.0 hours.
4. Allow to cool in air
5. Crush the "cake", blend and remix it.
6. Retire in a reducing atmosphere of 80% nitrogen and 20%
hydrogen for 1.5 hours.
 3.1 Solid state reactions and nucleation mechanisms 149

7. Allow to cool in a nitrogen atmosphere (Note that a "tube" furnace

is normally used to fire phosphors in controlled atmospheres).
8. Remove the phosphor cake, crush it and blend thoroughly.
9. Since we have an excess of MgO in the fired product, we need to
remove it. A 3 liter solution of 1.0 molar acetic acid is prepared.
10. Heat the acetic acid solution to 80 °C and add the phosphor
powder slowly with stirring.
11. Stir for about 10 minutes and then allow the particles to settle.
12. Pour off the supematent liquid and resuspend the powder in
deionized water.
13. Allow to settle and resuspend 2 more times.
14. Filter the particles in a Buchner funnel and dry the wet cake
overnight at 105°C.
15. Sieve the dried phosphor cake through a 325 mesh screen.

The resulting phosphor has a green emission band, as shown in the

following diagram:

In this case, we see the emission band which peaks at 5000 A and two
excitation bands at 2330 A and 2680 A.
150 Nucleation, diffusion, growth of particles and measurement of solid state

What this means is that these UV-radiation wavelengths will excite the
phosphor which then emits visible green light. When the phosphor is
prefired in air, the resulting phosphor does not respond to UV excitation.
However, cathode-ray excitation (an electron-beam like a television tube)
produces the same green emission. Emission occurs from Mn2+ centers in
the spinel structure. These centers are not intrinsic defects as such since
the divalent manganese is able to substitute directly at the Mg2* sites in
the spinel structure. The only difference is the radius of the two cations at
the tetrahedral site. It is because of this difference that increasing the
M n a concentration leads to less efficient, i.e.- "duller", phosphors.

That is, there is an "optimum" activator concentration (as is true for all
phosphors). However, UV excitation is not obtained until the prefired
phosphor is subjected to a reducing atmosphere. It has been determined
that if the reacting oxides are prefired directly in a reducing atmosphere,
a black mass is obtained containing tin in the metallic state.

The two UV excitation peaks are due to Sn2* centers in the phosphor. If
one looks at the excitation peaks of the SrjP.p^Sn2* phosphor, the peaks
are identical. No other cations having optical absorption in the UV are
present other than Sn2+ (Mn3* does not). Thus, the UV absorption results
in excited Sn2* centers which transfer this excitation energy to Mn2*
centers which emit the characteristic green emission.

It is clear that some of the Sn4+ cations in the B-site SnO4 groups are
reduced but no one has determined the extent of such reduction. It is
apparent that only a small fraction of such groups are affected. Otherwise
the lattice structure would not be maintained. Note that a significant
difference in ionic radius occurs when stannic tin is reduced to stannous
tin. X-ray analysis shows no difference between the air-fired and the
reduced phosphors. K. Th. Wilke studied this phenomenon in 1957 and
proposed the following mechanism, given as 3.1.90. on the next page.

The fact that oxygen vacancies are formed is an important factor in our
analysis of the defect chemistry of this phosphor.
 3.1 Solid state reactions and nucleation mechanisms 151

3.1.90.-
SnO2 + H2 => SnO + H2O
SnO + 2 MgO => Mg2SnO3
Mg2SnO3 + H2O => Mg2SnO4 + H2
and:
Mg2SnO3 + O2 =*> Mg2SnO4

According to our analysis given above, the intrinsic defect chemistry of

this spinel, fired in air, would likely involve:

3.1.91.- Mg24-, Mn*, Sn4* , Or , VSn , Sn,4*

The prefired phosphor would then have the following composition:

3.1.92.- (Mg^ , M nMg ) 2 [Sn Sn , (VSn , SnJ] O4

i.e.- the Frenkel defect, where Mg and Mn are divalent and Sn is

quadrivalent. It is likely that the embryo formation involves both MgO and
MnO since the carbonates decompose to form nano-sized oxides. Nuclei
then grow to form the ultimate particles. Note that the reactions:

3.1.93.-
2 MgO + SnO2 => Mg2 SnO4

2 MnO + SnO2 =*> Mn2 SnO4

occur but that MgO is 100 times more prevalent than MnO. Nevertheless,
it is remarkable that the Mn2* cation becomes uniformly distributed within
the tetrahedral sites in the lattice (Or does it?).

However, the re-fired phosphor presents another problem. When the

reduction step takes place, we have the formation of the divalent cation,
Sn2* , which prefers a tetrahedral environment in the lattice. But, the Sn4*
cation resides on an octahedral site. Furthermore, it is likely that oxygen
vacancies are created via:
152 Nucleation, diffusion, growth of particles and measurement of solid state

3.1.94- H 2 + {(Sn Sn , ( V ^ , SnJ O4 )} => {(Sn Sn , (V^.V^.Sn.) O4 )} + H2O

Thus, it is the interstitial Sn,4+ atoms that are reduced to Sn,2+. These
cations then are free to move to tetrahedral sites in the final product
phosphor where there is more room. As stated above, when the phosphor
is excited by ultraviolet light, it is the tin cations which become excited
but do not emit. They transfer the excitation energy to the Mn21 sites
which then emit green light. How do we know this? If we compare other
phosphors which are activated by Mn" alone and those activated by Sn2*
alone, we find that the relatively narrow green emission of (Mg,Mn)2SnO4
is typical of the d 5 electronic configuration of Mn^ whereas the s ^ 6
configuration of Sn** produces a much broader emission band, typically
blue or blue-green. We will describe electronic transition moments and
spectroscopic states later on in Chapter 5.

C. - DIFFUSION IN SILICATES AND SILICATE PHOSPHORS

We have presented spinel synthesis because these systems were studied

first and can be understood in a simple manner. Various silicate systems
have also been studied and it has been determined that they are more
representative of the general case involving solid state synthesis reactions
than spinel. Let us examine the following simple silicate reaction:

3.1.95.- 2 CaO + SiO2 => Ca2SiO4

Calcium oxide, in proper proportion, reacts with silica to form calcium

orthosilicate. As in terms of the spinel case, we would expect to see the
following diffusing species:

3.1.96.- Ca2+ , Si 4+ , O= , e" , p+

where e~ = 2 electrons, etc. Using our model, we would illustrate the

diffusion reactions as shown in 3.1.97., shown on the next page. The
diffusion conditions are given in 3.1.98, also shown on the next page.

The three mechanisms shown in 3.1.98. seem to be valid but further

 3.1 Solid state reactions and nucleation mechanisms 153

investigation reveals that actual diffusion is controlled by the nature of the

lattice structure. Thus, the rate of diffusion has been found to be:

3.1.99.- RATES OF DIFFUSION FOR SILICATE-BASED COMPOUNDS

Rate of species diffusion: Ca2+ a O= » Si 4+

This is caused by the fact that the Si4+ is tied up in the form of SKD4
tetrahedra, as shown in 3.1.100., given on the next page.

Here, we show two silicate tetrahedra tied together to form a disilicate.

On the left is a "stick and ball" diagram while that on the right shows the
oxide atoms in their positions in the tetrahedra. One apex oxide has been
made transparent to show the position of the Sr*+ in the center. These
tetrahedra are tied together at the corners so that a silicate "backbone"
forms the structure. In a compound like Ca3Si2O7 (a disilicate or
pyrosilicate), the metal cations form "bridges" between backbone-layer
154 Nucleation, diffusion, growth of particles and measurement of solid state

silicates and are much more free to move. However, it is well to note that
a small amount of silicate does move, but the exact nature of the diffusing
specie is not certain (It may depend upon the nature of the compounds
being formed. Most probably, it is actually SiOn but the charge of the
actual specie may vary). We will continue to use Si 4+ to indicate silicate
diffusion.

Thus the predominating diffusion reaction for Ca2SiC>4 synthesis is:

3.1.101.-
2Ca 2 +
* 2 Ca 2 + + 2 0=+ SiO, -* Ca 2 Si0 4
2 0= *

We now come to the most important point of this Section. Up to now, we

have assumed that the phase within the phase boundary, as given in the
diagrams, is inviolable and not subject to change. That is, once the
Ca2SiO4 has formed, no further reaction can occur. While this may be true
for some solids, it certainly is not true for Ca2SiC>4, and indeed most other
systems. The way we determine if further reaction is possible is to consult
the phase diagram of the system. If a compound is stable, then it has a
finite probability of forming during the solid state reaction. Let us now
suppose that further reaction does take place at the phase boundary of
CaO and Ca2SiO4 t and also at Ca2SiC>4 and SiC>2. We will call these new
phases "X" and "Y".

This gives us the situation shown in the following diagram:

 3.1 Solid state reactions and nucleation mechanisms 155

It would appear that "X" should be Ca2SiOs and "Y" should appear as
Ca3Si2O7. A further look at the phase diagram shows that the former
stoichiometry does not exist, but the latter does. Because the diffusing
species, Ca2+ and O= , cannot react according the above-given reaction,
they continue to diffuse until they reach the vicinity of the pyrosilicate,
i.e.- Ca3Si2O7. There, the reaction at this phase-boundary is:

3.1.103.- Ca2+ + O= + Ca3Si2O7 => 2 Ca2SiO4

But there is nothing to stop further diffusion of these species and they
continue to diffuse to the vicinity of the SiO2 phase boundary, where the
reaction is:

3.1.104.- Ca2+ + o= + SiO2 =* CaSiO3

Thus, the final possible reaction is that which forms the metasilicate
stoichiometry. In the opposite direction, we also have:
156 Nucleation, diffusion, growth of particles and measurement of solid state

Note that we have two diffusion reactions which form the metasilicate.
This has an amazing effect on the overall reaction as shown in the
following diagram:

3.1.106.-

Note the sequence of intermediate reactions that have occurred. Let us

now recall the prevailing diffusion conditions, as given in 3.1.98., i.e.-

3.1.107.- Rate of Diffusion: Ca2+ a O= » Si 4+

Because of this diffusion mechanism, we observe the above sequence of

diffusion reactions which occur with time, as given above. It is actually
possible to observe these reactions by the use of a polarizing microscope
hot stage wherein the reactions are caused to occur by heating while
observing the final structures formed via the x-ray diffraction patterns.

The above diagram is somewhat cumbersome and it is much easier to

 3.1 Solid state reactions and nucleation mechanisms 157

illustrate these reactions by a graph. This is shown in the following

diagram:

In the sequence of diffusion reactions, we note that Ca2SiO4 is formed

immediately, followed by Ca3Si2O7 . Both begin to disappear when CaSiC>3
begins to form. Near the end, CaSiO3 becomes the major phase present,
but we never reach the point where just ONE COMPOUND remains. We
always obtain a mixture! This point cannot be overstressed.

The overall reaction mechanisms are diffusion-controlled, and the total

reaction is given as follows in 3.1.109. Although these partial solid state
reactions are written to show the formation of the metasilicate, we already
know that none of these reactions come to completion. There are
competing side-reactions.

3.1.109.- 8 CaO + 4 SiO 2 => 4 Ca 2 SiO4

Ca 2 SiO 4 + SiO 2 => 2 CaSiO 3
3 Ca 2 SiO 4 + SiO 2 => 2 Ca 3 Si 2 O 7
2 Ca 3 Si 2 O 7 + 2 SiO 2 => 6 CaSiO 3

8 CaO + 8 SiO 2 = 8 CaSiO3

158 Nucleation, diffusion, growth of particles and measurement of solid state

Thus, we start with a 2:1 stoichiometry of CaO and SiO2 , but end up with
a stoichiometry which is mostly 1:1, that of the metasilicate. Another way
to look at this phenomenon is that the diffusion conditions favor the
formation of the metasilicate, and that the "excess Ca" is taken up in the
formation of compounds having a "calcium-rich" stoichiometry (in
comparison to that of the metasilicate).

However, if we start with a stoichiometry of 1:1 CaO to SiO2, we discover

that the same compounds form as before, and a mixture of compounds is
still obtained. The only difference is that there are smaller amounts of
Ca2SiO4 and Ca3 Si2C>7 present! The only conclusion that we can draw is
that diffusion-controlled solid state reactions tend to produce mixtures of
compounds, the relative ratio of which is related to their thermodynamic
stability at the reaction temperature. Obviously then, if we change the
temperature of reaction, we would expect to see somewhat different
mixtures of compounds produced.

Let us now look briefly at another similar system where we will start with
a stoichiometry of: 1.00 BaO to 1.00 SKD2 (in mols). The mixture would
be expected to react as:

3.1.110.- BaO+SiO 2 => BaSiO3

with the diffusion condition: Ba2+ , O= » Si 4+ . We show the solid state

reaction behavior again by a chart, given as follows in 3.1.111. on the next
page. One might think that since Ba2+ is a much larger ion than Ca2+ , it
would diffuse slower. Such is not the case as can be seen by comparing
the x- axis of 3.1.108. with that of 3.1.111. in terms of reaction time. It is
remarkable that in this case , we started with a 1:1 stoichiometry and
ended up with a compound that has a 2:1 stoichiometry!

The series of diffusion-controlled reactions are, for the case of:

1.0 BaO+ 1.0 SiO2=> :

3.1.112.- (a) 2 Ba2+ + 2 O= + SiO2 => Ba2SiO4

(b) Ba2SiO4 + Si 4+ +2O= => BaSiO3
 3.1 Solid state reactions and nucleation mechanisms 159

3.1.111.-

Because the rate of Reaction a) is so much faster than that of (b), we end
up with Ba2SiO4 as the major product. But, we always end up with a
mixture of stoichiometries. THIS BRINGS US TO A MAJOR AXIOM FOR
THIS CHAPTER REGARDING SOLID STATE REACTIONS, VIS-

"REACTIONS BETWEEN REFRACTORY OXIDES WHICH ARE

DIFFUSION-CONTROLLED ALWAYS RESULTS IN MIXTURES OF
COMPOUNDS "

WHAT THIS MEANS IS THAT WE CANNOT PREPARE PURE

COMPOUNDS BY REACTING OXIDES TOGETHER!

There ought to be a way to evade this situation, and indeed there is. We
can find at least four (4) methods to do so. They include:

3.1.113.- 1. Using a gaseous reactant

2. Using a flux
3. Promoting a super-reactive component
4. Using a precipitated product to act as the reaction base
160 Nucleation, diffusion, growth of particles and measurement of solid state

Note that all of these methods attempt to bypass the dependence of the
solid state reaction upon diffusion of reacting species. But, using a gaseous
reactant may not be practical in all cases. And, sometimes it is hard to
find a flux which does not interfere with the reaction. A flux is defined as:

"A material whose melting point is lower than the solid state
reaction temperature, dissolves one or more of the components and
allows material transport to the reaction zone, without entering into
the solid state reaction. Preferably, the end-product should be
insoluble in the flux".

We can also employ a super-reactive product so as to obtain 100% of the

desired product. Examples are:

3.1.114.- 2BaCO 3 + SiO2 => Ba2SiO4 (100% yield)

BaCO3 + SiO2 => BaSiO3 (100% yield)
3 BaCO3 + SiO2 =*• Ba3SiO5 (100% yield)

The mechanism involved is as we have already described in 3.1.6., i.e.-

The very fine particles of nearly atomic proportions react almost

immediately. Because the product, BaO, has an extremely large surface
area, the solid state reaction is not diffusion limited.

In the last method (see 3.1.113.), we might use a precipitated product to

use as the basis to form the desired compound. We might precipitate:

3.1.116.- Ba2+ + HPO4= => BaHPO4 tt

which is then reacted to form the orthophosphate as the desired product:

 3.1 Solid state reactions and nucleation mechanisms 161

3.1.117.- 2 BaHPO4 + BaCO3 => Ba3(PO4)2 + H2O ff + CO2 ff

If barium orthophosphate is formed in this manner, we can guarantee

that the final solid state reaction product will consist of 100 % of the
desired product.

Before we leave this discussion, we should mention several of the silicate-

based phosphors which are in current use today. These include:

3.1.118.- Silicate Phosphors and Their Uses

Phosphor Name Formula Emission Commercial Uses

Color
Zinc Silicate ZnSiO4 :Mn& Green Lamps & CRT
Barium Silicate BaSL,Os :Pb UV Blacklight lamps
Calcium Silicate CaSiOa:Pb UV Erythemal lamps
Calcium Silicate CaSiOa:Pb:Mn Orange Lamps & CRT
Yttrium Silicate Y2SiO5:Ce:Eu2+ Blue-green Lamps for high color
rendition

These are but a few of the known and different silicate phosphors
currently used commercially today. They all are prepared by reacting SiO2
with the requisite oxides to form the final "composition". We will not try
to analyze the defect nature of any of these phosphors since each
manufacturer has his own method of preparation and any description
would fit that phosphor only. It should be clear that the above given
formulas represent the major compound in the formula but that each is
composed of more than one composition, albeit in small amounts.

VI. - Diffusion Mechanisms Where the Cation Changes its Valence State

Before we leave the subject of heterogeneous solid state diffusion and

nucleation during a solid state reaction, let us examine the case where a
change of valence state of one of the reactants occurs. In formal studies of
162 Nucleation, diffusion, growth of particles and measurement of solid state

such reactions, this situation continues to be the subject of much

investigation because of the unusual situations which are encountered.
Consider the reaction:

3.1.119.- MO + A12O3 => NiAl2O4

This would give us the following diffusion diagram:

But, suppose some of the Ni 2+ oxidizes to Ni 3+ , vis-

3.1.121.- Ni2+ = Ni3+ + e -

This would give us the case where a trivalent cation is present in a

divalent lattice, (see Chapter 2). The operative defect reaction would be:

3.1.122.- 2Ni 2 + N 1 % Ni3+Ni + VNi + p+

YNI + e- = VNI"
2
2 Ni +Ni + e- % Ni3+N1 + VNi" + p+
 3.1 Solid state reactions and nucleation mechanisms 163

This gives us the following diffusion diagram:

This situation gives rise to a complicated set of diffusion conditions:

a. The charged vacancy is one of the migrating species.

b. The charged vacancy combines with Ni 3 + to form a defect
compound.
c. Since DNI 3 + » Dv(Ni) . Ni 3+ diffuses in an opposite direction to
A13+ , as shown in the following set of actual diffusion reactions,
shown as 3.1.124. on the next page.

Note that the charged vacancy diffuses as one of the reacting species to
form the defect compound. This situation is quite common in the solid
state chemistry of compounds containing multivalent cations. The
trivalent Ni 3+ also gives rise to a new compound, MAIO3 . Yet the same
reactions given in 3.1.120.. are also OPERATIVE.

Thus, the overall reaction actually taking place is a combination of

3.1.120. and the following:

3.1.124.-

3+
Ni
•
0= N i 3+ + 0 = +A1
2 °3+ V
iii ""* N i V Ni A1
2°4

Ni 3 +
* Ni 3 + +A1_O T -+ NiA10, + A l 3 +
3
A, 3+ 2 3
3+ 2+
« Al + 4 N i 0 - > NiAl2O4 + 3 Ni
164 Nucleation, diffusion, growth of particles and measurement of solid state

It is well to note that these types of defect solid state reactions are
prevalent in many solid state reactions.

3.2- HOMOGENEOUS NUCLEATION PROCESSES - PARTICLE GROWTH

We will now examine homogeneous nucleation that may occur during a

solid state reaction. One example is a decomposition reaction where a
compound is heated to decompose it to form another material. In one
example, a carbonate decomposes to form an oxide: BaCO3 => BaO. Nuclei
formation and growth is homogeneous since only heat is involved in the
decomposition reaction, not another compound or external reactant. We
have already stated that homogeneous nucleation can be contrasted to
heterogeneous nucleation in that the former is random within a single
compound while the latter involves more than one phase or compound.

Homogeneous nucleation also applies to precipitation processes where

homogeneous nucleation must occur spontaneously before precipitation
can occur.

To begin, we know that a change in free energy must occur as a change in

phase (nucleation) occurs. This will be related to the total volume, V, and
the total surface area, 2, vis-

3.2.1- AG= VAg + So

where AG is the free energy per unit volume and a is the surface energy of
the individual nuclei. If we have spherical nuclei, then:

3.2.2- AG = 4 / 3 ; r r 3 A g + 47t r 2 o

If the total free energy, AG, does not change with nuclei radius, i.e.-
d(AG)/dr is defined as zero, then:

3.2.3.- r*Crit = - 2 a / A g

where r *crit is the critical radius for nucleus formation.

 3.2 Homogeneous nucleation processes - particle growth 165

We can combine equations to obtain:

3.2.4.- AG* = 16TT o3 / 3(Ag)2

where AG is the change in free energy at the critical radius for nucleus
formation. If the following holds:

3.2.5.- Ag = Ah(T-T0) / T o = Ah (AT/ T o )

/where Ag and Ah refer to individual free energy and heat of

transformation of nuclei, then:

3.2.6.- AG* = 16 it a 3 T o 2 / 3 (Ah)2 AT2

Since we know that the number of nuclei can be described by a Boltzmann

distribution, then:

3.2.7.- N* = N exp AG* / kT

We define dN/dt as a frequency (t is time):

3.2.8.- dN/dt = S* fe exp AGD / kT

where GD is the free energy of the defect (nucleus), S is the critical

surface interface, and fe is the interfacial energy. Combining the two
above equations gives us:

3.2.9.- I = dN*/dt=NS* f e exp (-AG* / kT + AGD / kT)

The quantity, NS* f e , is approximately equal to 10 3 6 per cc. per second.

We can solve this graphically, as shown in the following, given as 3.2.10.
on the next page.

It can be seen in A that the value of r* is remarkably constant over a

wide range of temperatures, but that it starts to approach infinity at some
166 Nucleation, diffusion, growth of particles and measurement of solid state

3.2.10.-

critical temperature, T . In contrast, the production of nuclei in B is

maximum at some particular temperature.

Let us now examine some growth kinetics for the general case, that is, the
growth of nuclei after they have formed. The general equation for growth
after formation is given as:

3.2.11.- dx/dt = ki(l-x)n

and:
x = 1- exp (- ki t) 1

where all terms in 3.1.11. are the same as already given. We solve these
equations by aTaylor series, i.e.-

3.2.12.- -ln(l-x) = l + kt/2! + k t 2 / 3 ! + kt 3 / 4!

If we now define X as an instantaneous moment of time, i.e.-

3.2.13.- (dx/dt)t=T =Kne-ki'

This can be arranged to:

3.2.14.- ( dx/dt) t = x = nkit 11 " 1 (1-x)

 3.2 Homogeneous nucleation processes - particle growth 167

If x is small, then:

3.2.15.- dx/dt= nkit 11 " 1

This can be integrated to an equation similar to that of 3.1.11.

Let us now examine transformation kinetics for the homogeneous case of

nucleation and growth.

We assume:

a. The nucleation is random.

b. The nucleation and growth rate are independent of each other.
c. The spheres grow until impingement.

Then, if spheres are nucleated at time, T , the time of growth becomes a

function of (t -t).

If we now define Xext as the extrinsic amount per nucleus before growth
(note that it is not a radius; in feet, it is the same x given in 3.2.11.), then
we have:

3.2.16.- Xext = J* 4/3 7TU3 (t -x)Idx

where U = volume of the nuclei. Note the similarity of this equation to

3.1.18. If we then further define Xf as the final amount of volume per
sphere at impingement where growth stops, i.e.- xf = 1 - Xext ,then we
get the Johnson-Mehl equation for slow nucleation:

3.2.17.- xf = 1- exp {( - 1 / 3 T T I U 3 ) t 4 }

If the nucleation is fast, the Avrami equation is operative:

3.2.18.- Xf = 1- exp {( - l / 3 j r N v U 3 ) t3}

168 Nucleation, diffusion, growth of particles and measurement of solid state

where Nv is defined as the number of nuclei per unit volume.

These two equations have been used to describe homogeneous nucleation,

particularly in precipitation processes.

3.3.- NUCLEATION IN PRECIPITATION REACTIONS

Before we begin a discussion of nucleation during a precipitation, we need

to explain some of the intricacies of precipitation. We first start with two
solutions, one containing the cation of interest and the other the anion.
Both involve soluble substances like nitrates or chlorides for the cations
and acids or sodium compounds (or the like) for the anions. When the two
solutions are combined, a precipitate will form if its aqueous solubility, as
defined by its solubility product, is less than either of the cation and /or
anion solubilities. What this means is that if we use two soluble solutions
and combine them, a precipitate will form if it is insoluble in the aqueous
solution. There are many examples of precipitate formation. We need to
define the effect of exactly how the precipitation should be carried out in
terms of phosphor preparation.

Consider two solutions. One contains, for example, CaCl2 and the other
(NHJ2HPO4 . if we combine these, we will obtain a precipitate of CaHPO4.
The solubility product of CaHPO4 is Ksp = 1.2 x 10 7 mol/L. However, the
exact nature of this precipitate depends upon how the process is
performed. The following diagram, given as 3.3.1. on the next page,
illustrates this factor. You will note that we have shown two processes,
both using a tank containing either CaCL, or NH4HPO4. If we follow the
method shown on the left, the first Ca^ cations added will see a large
excess of HPO42 anions. The precipitate will thus contain a very slight
excess of phosphate, adsorbed on the surface of the so-formed particles.
As the ratio of Ca^/HPO/' in the tank approaches 1.00, the precipitate
composition then approaches a 1.000:1.000 ratio in the solid. The same
mechanism applies to the case shown on the right except that a slight
excess of Ca^ is obtained in the final precipitate.

This effect has a decisive effect upon the precipitate obtained.

 3.4 Sequences in particle growth 169

3.3.1.-

Even though the amount of excess Ca^ is so small that it is difficult to

measure, the phosphor prepared from such a precipitate is inferior in its
properties to that prepared from the precipitate containing a small excess
of phosphate. As we have indicated in the previous chapter, nucleation
processes can be homogeneous (no outside influences) or heterogeneous
(by specific outside constraints). In precipitation processes to form a
particle size distribution, it is probably a combination of both mechanisms.
It should be clear that as Ca^ is added to HPO42, the first to form is an
embryo of CaHPO4 . This rapidly changes into a nucleus which then grows
until all of the nutrient (in this case, addition of calcium ion) is used up to
form the precipitate. Finally, the particles are so large that they impinge
upon each other. We will discuss this in more detail in the following
section where particle growth can be related to both precipitation and
solid state reaction.

3.4. - SEQUENCES IN PARTICLE GROWTH

We have already shown that embryo formation leads to nucleation, and

that this nucleation precedes any solid state reaction or change of state.
170 Nucleation, diffusion, growth of particles and measurement of solid state

In a like manner, nuclei must form in order for any precipitation process
to proceed. Once formed, these nuclei then grow until impingement of
the growing particles occurs. Impingement implies that all of the nutrient
supplying the particle growth has been used up. This mechanism applies
to both solid state reaction and precipitation processes to form product
particles. Then a process known as Ostwald Ripening takes over. This
process occurs even at room temperature.

The sequences in particle growth are:

3.4.1.- Embryo formation

Nucleation
Nuclei Growth
Impingement
Ostwald Ripening (coarsening)
Sintering
Formation of Grain Boundaries

We have already covered the first three in some detail. Impingement

involves the point where the growing particles actually touch each other
and have used up all of the nutrient which had originally caused them to
start growing in the first place. This mechanism occurs in both
precipitation and solid state reaction mechanisms. It is for this reason
that we did not discuss precipitation until we examined solid state
reaction mechanisms.

Ostwald ripening usually occurs between particles, following

impingement, wherein larger particles grow at the expense of the smaller
ones. One example of this would be if one had a precipitate already
formed in solution. In many cases, the smaller particles redissolve and
reprecipitate on the larger ones, causing them to grow larger. Interfacial
tension between the particles is the driving force, and it is the surface
area that becomes minimized. Thus, larger particles having lower surface
area increase at the expense of numerous smaller particles which have a
relatively high surface area.
 3.4 Sequences in particle growth 171

Ostwald ripening differs from nuclei growth in that the relative size and
numbers of particles change, whereas in nuclei growth, the numbers of
particles growing from nuclei do not change. Sintering, on the other
hand, is an entirely different process, and usually occurs when external
heat is applied to the particles.

I. - Ostwald Ripening of a Calcium Phosphate Product

It should be clear that the precipitation process needs to be controlled

carefully in order to produce a material composed of particles of a desired
size and configuration. Both Ostwald ripening and sintering can be utilized
to obtain a particle of desired size, dimensions and particle habit.
Industrial technologists have taken advantage of these particle forming
and altering mechanisms. One example of this type of particle growth is
described as follows.

In fluorescent lamps, a layer of phosphors is applied to the inside of a

glass tube by means of a suspension of particles, i.e.- the halophosphate
phosphor having a composition of: Ca5F,Cl(PO4)3: Sb3+ : Mn2+ (plus minor
amounts of other phosphors to achieve certain lamp "colors"). It has been
determined that the lamp brightness and duration of light output
(maintenance) is highly dependent upon how well the internal surface of
the glass is covered by the particles. By maintaining precipitation
conditions so that small thin squares of CaHPO4 result from the
precipitating solution, a maximum coverage of the glass is achieved.
Precipitation occurs by adding a solution of (NH4)2HPO4 to a solution of
CaCl2. The resulting precipitate at 20° C. is CaHPO4 2H2O which consists
of very fine particles and is not very crystalline. As a matter of fact, the
particles were usually ill-defined and bordered upon amorphous.

The solution temperature is then raised so that Ostwald ripening can

occur. During this process, CaHPO4 becomes the stable form at the
solution temperature of 80° C. As the larger crystallites begin to grow, the
smaller ill-defined crystallites, having a much larger surface area, dissolve
and reprecipitate upon the larger ones. The resulting single-crystal
squares, i.e.- • , have an average size of about 25pm. It has been
172 Nucleation, diffusion, growth of particles and measurement of solid state

determined that diamond-shaped crystals can also be obtained in more

concentrated solutions. What this means is that Ostwald ripening causes
the crystals to grow in the x and y- direction, but that a higher
concentration in the "mother-liquor" containing the precipitate, CaHPO4
2H2O, also grows in the z-direction of the lattice comprising the crystals.
The solid state reaction to form the halophosphate phosphor is:

3.4.2.- 6 CaHPO4 + 3 CaCO3 + CaF2 =* 2 Ca5 F (PCy3

where we have not shown the Sb2O3 and MnCO3 added as "activators". The
halophosphate thus produced follows the crystal habit of the major
ingredient, in this case that of CaHPO4 habit. When applied, the thin
squares lie flat and overlap on the glass surface. During sintering at about
1200 °C, the CaHPO4 does not disintegrate but undergoes an internal
rearrangement to form Ca^Oy while maintaining the same crystal habit.
The CaCO3 disintegrates into small particles (like BaCO3) while CaF2
exhibits a sublimation pressure at the firing temperature. The resulting
solid state reaction to form the halophosphate product thus depends upon
the crystal habit of the major ingredient, CaHPO4.

n - Sintering and Sintering Processes

Sintering of particles occurs when one heats a system of particles to an

elevated temperature. It is caused by an interaction of particle surfaces
whereby the surfaces fuse together and form a solid mass. It is related to a
solid state reaction in that sintering is governed by diffusion processes,
but no solid state reaction, or change of composition or state, takes place.

The best way to illustrate this is to use pore growth as an example. An

example is shown in 3.4.3. as follows on the next page. When a system of
particles, i.e.- a powder, is heated to high temperature, the particles do
not undergo solid state reaction, unless there is more than one
composition present. Instead, the particles that are touching each other
will fuse together and form one larger particle, as shown above. As can be
seen, voids arise when the particles fuse together. The void space will
depend upon the original shape of the particles.
 3.4 Sequences in particle growth 173

3.4.3.-

Here, we have shown spherical particles which produce only a few voids.
Additionally, an overall change in the total volume of the particles, and
that of the fusion product, occurs. Mostly, the change is negative, but in a
few cases, it is positive. Experimentally, this change is a very difficult
problem to measure. What has been done is to form a long thin bar or rod
by putting the powder into a long thin mold and pressing it in a hydraulic
press at many tons per square inch. One can then measure the change of
volume induced by sintering, by measuring a change in length as related
to the overall length of the rod. It has been found that the sintering of
many materials can be related to a power law, and that the shrinkage, AL,
can be described by:

3.4.4.- AL / LQ = k tm

where k is a constant, LQ is the original length, t is the time of sintering,

and the exponent, m , is dependent upon the material being investigated,
it has also been further demonstrated that a considerable difference exists
between the sintering of fine and coarse particles, shown on the next
page as 3.4.5.
174 Nucleation, diffusion, growth of particles and measurement of solid state

3.4.5.-

In this case, the "fine" particles and the "coarse" particles were separated
so that the difference in size between individual particles was minimized.
That is, most of the individual particles in each fraction were almost the
same size. Both the fine and coarse particles have a sintering slope of 1 / 2
but it is the coarse particles which sinter to form a solid having a density
closest to theoretical density. This is an excellent example of the effect of
pore volume, or void formation, and its effect upon the final density of a
solid formed by powder compaction and sintering techniques. Quite
obviously, the fine particles give rise to many more voids than the coarser
particles so that the attained density of the final sintered solid is much
less than for the solid prepared using coarser particles. It is also clear
that if one wishes to obtain a sintered product with a density close to the
theoretical density, one needs to start with a particle size distribution
having particles of varied diameters so that void volume is minimized.

III.- Grain Growth

The next subject we will discuss is that of grain growth. The simplest way
to illustrate this factor is through the sintering behavior of aggregates, as
shown in the following diagram:
 3.4 Sequences in particle growth 175

In this diagram, two steps are implicit. An aggregate is defined a large

particle, composed of many small particles, It is the aggregates, made up
of very fine specks, which sinter to form larger grains (particles). But,
since many particles are growing simultaneously, growth occurs until
impingement, with the formation of boundaries between the particles
(grains). It is these grains which form the final sintered whole. Note that
the crystallographic orientation of each grain differs from that of its
neighbors.

When a system of very fine particles is formed, the interfacial tension is

high due to the very high surface area present. Agglomerates (weak
surface energy interchange) will form, or aggregates (strong surface
energy interchange) can result, especially if the Ostwald ripening
mechanism is slow, or is inhibited. Immediate removal of a precipitate
from its "mother liquor" is one example where the likelihood of aggregate
formation is enhanced. Sintering then produces both pore growth and
grain growth. This mechanism also applies to powder compaction
processes where aggregates may be present in the powder. Sintering then
leads to grain growth as well.

Even where a metal is melted and then cast, nucleation leads to formation
of many fine particles in the sub-solidus (partially solidified) state. This
leads to grain growth in the solid metal, thereby lowering its strength.
Sometimes, special additives are added to the melt to slow nucleation
176 Nucleation, diffusion, growth of particles and measurement of solid state

during cooling, thereby increasing the strength of the metal product. As

we stated, once a particle forms, it may undergo certain reactions
involving the solution from which it was formed. This includes formation
of "embryos", "nuclei" in solution and growth of crystals from a
precipitated nucleus. Precipitated particles usually grow from a nucleus to
which ions are added in a regular manner to form a three-dimensional
structure. Such crystals cease growing when the "nutrient" (the material
which serves to form the particle) becomes depleted in the solution. If a
multitude of nuclei are formed initially, then their overall size of growth
will be limited. Such particles are known as "crystallites". Each will
consist of several grains, having a differing orientation of the crystal
lattice, within each individual particle, as shown in the following diagram:

Note that this involves a precipitation mechanism where aggregates form.

Grain boundaries form junctions between grains within the particle, due
to vacancy and line-defect formation. This situation arises because of the
2nd Law of Thermodynamics (Entropy). Thus, if crystallites are formed by
precipitation from solution, the product will be a powder consisting of
many small particles. Their actual size will depend upon the methods
used to form them. Note that each crystallite can be a single-crystal but, of
necessity, will be limited in size.

Another example is our old friend, BaCO3. If we fire this solid compound
in air at a very high temperature for a long time, we get several changes.
First, it decomposes to very fine particles of BaO. These fine particles
 3.4 Sequences in particle growth 177

have a large surface area, and with continued firing sinters to form larger
particles. Eventually, the particles get big enough, and the porosity
decreases to the point where grain boundaries begin to form between
particles. The grains sinter together to form a large particle with many
grain boundaries. It should be again be emphasized that each grain in the
large particle is essentially a small single crystal with its lattice oriented
in a slightly different direction from that of it neighbors.

Let us now consider the thermodynamics of sintering. There are two

types of sintering which are distinguished by the change in volume which
occurs. These are:

3.4.8.- NO SHRINKAGE : dV/dt = 0

WITH SHRINKAGE : dV/dt = f(V)

As we have already said, the change in volume, from initial state to final
state, can be positive or negative, but is usually negative. The driving force
is a decrease in Gibbs free energy, AG. It is related to both the interfacial
tension (surface energy), g , and the surface energy of the particles (which
is related to their size), vis-

3.4.9.- AG= g dA

Consider a more familiar example, that of a droplet sitting upon the

surface of a liquid. The droplet has a radius, r, and there are n-moles of
liquid within it with a molal volume, V. To form the droplet requires an
amount, nV, of the liquid, where V is the fractional molar volume of the
droplet.

This gives us:

3.4.10.- n V = 4 / 3 7rr3

and the change in free energy to form the droplet is:

3.4.11.- dG =AGdn = g dA
178 Nucleation, diffusion, growth of particles and measurement of solid state

If we now differentiate these equations, we obtain the equations given in

the following:

3.4.12.- Volume of Droplet: n = 4 i r 3 /3 V so that: dn = 4TT /V r 2 dr

and:
Surface Area of Droplet: A = 4JT r 2 so that: dA = 8n rdr

We can put all of the equations together so as to yield the Kelvin equation
for change in free energy as a function of the radius of the spherical
particle, vis-

3.4.13.- AG = 2 g V / r

One can do the same for a cubic particle, in fact for any shape factor.
Since the chemical potential, p , is related to AG, we use the following:

3.4.14.- p - po = AG = RT In p / p o
2 g V / r = RT In p / p 0
or: P = Po exp (2 g V / r RT)

If we now define Ap = p - p o , then:

3.4.15.- p/po = 1+Ap/po

Mathematically, In (1 + Ap/p o ) « Ap/p o (within 5% if Ap/p o >0.1).

Thus we get the approximate equation:

3.4.16.- Ap/po = 2g V / r • 1/RT

It is this equation which has been used more than any other to evaluate
sintering. To evaluate it use, consider the following example:

Alumina: AI2O3 is a very refractory compound. It melts above 1950 °C.

and is not very reactive when heated. If we attempt to sinter it at 1730°C,
we find the following values shown in the following to apply:
 3.4 Sequences in particle growth 179

3.4.17.- Ap/po @ 0.1

g s 2000 dyne/cm.
V = M / d = 25.4 cc. / m o l
R = 8.3 x 10" 7 erg/mol

r = 6 x 1 0 ' 6 cm = 0.06 micron

What this means is that if the alumina particles are smaller than about
0.06 micron, they will not sinter. Even though the above is an
approximation, the specific case for alumina has been confirmed
experimentally.

THUS, WE HAVE SHOWN THAT IF THE PARTICLES ARE TOO

SMALL, THEY WILL NOT SINTER. NORMAL NUCLEI GROWTH
THROUGH DIFFUSION PROCESSES REMAINS THE NORM UNTIL
THE PARTICLES GET LARGE ENOUGH TO SINTER.

Let us now examine why this mechanism might be true. In the case of
sintering of spheres, we can define two cases as before, that of no
shrinkage and that of shrinkage, both as a function of volume. If we have
two spheres in direct contact, we can define certain parameters, as
shown in the following diagram, given as 3.4.18. on the next page.

We start with spheres of radius, r, in direct contact. The two cases shown
are:
1) no shrinkage and 2) with shrinkage.

Actual sintering occurs by flow of mass from each sphere to the mutual
point of contact, which gradually thickens. We can estimate the volume of
mass, V, at the contact area, A, in terms of the following parameters: r,
the radius of the spheres; r , the thickness of the layer buildup; and x, the
radius of contact of the built-up layer. If we have shrinkage, then we must
also evaluate h, the amount of shrinkage, shown above as the height of
interlinking layer. This model brings us to an important point, given as
follows:
180 Nucleation, diffusion, growth of particles and measurement of solid state

"If the spheres are too small, there can be little mass flow to the
area of joining (sintering) of the spheres".

Actually, it is the rate of flow of mass to the joining area that is important
and the area of touching of the spheres will determine this.

It is therefore logical that a size limitation should apply to the case

of sintering and its mechanisms.

If the spheres are too small, then there is not enough touching area and
volume for the sintering mechanism to occur.

The actual values of the sintering parameters have been found to be those
given in the following:

4.3.19.-
VALUES OF SINTERING PARAMETERS
V h A
No Shrinkage 2
7TX / 2r 0 2 2
3T X /r x 2 /2r

With Shrinkage ;rx2 /2r x 2 /2r JT2X3 /2r x 2 /4r

 3.4 Sequences in particle growth 181

Most systems exhibit shrinkage in sintering and it has been found that the
following equation applies:

4.3.20.- AL/Lo=h/r andx=ctm

where t is the time of sintering, h is a characteristic sintering constant

for shrinkage, and m is an exponent dependent upon the mechanism of
sintering.

We have indicated that flow of mass is important in sintering. There are

several operative mechanisms which are determined by the type of
material involved. Actually, the study of sintering deserves a separate
CHAPTER, but we will only summarize the major mechanisms that have
been observed for the constants of 4.3.19. This is shown as follows:

4.3.21.- SINTERING MECHAN][SMS

Mechanism m h Material
Viscous flow 1/2 1 Glass
Evaporation-Condensation 1/3 AL/Lo= 0 NaCl
Grain boundary diffusion by- 1/4 1/2
Vacancy Formation
Volume diffusion 1/5 2/5 Cu,Ag
Grain boundary diffusion by 1/6 1/3
Interstitial Formation
Surface diffusion 1/7 AL/Lo=0 Ice

Note that the values of h depends upon the material and the mechanism
of sintering. Also, these values depend upon r and x, the radii of the
particles and the radius of joining.

Obviously, these equations apply only to spheroids and a shape factor must
be considered as well. The derivation becomes much more difficult and is
beyond the scope of this chapter. The reader is referred to specific
volumes on sintering and sintering mechanisms.
182 Nucleation, diffusion, growth of particles and measurement of solid state

LET US NOW SUMMARIZE THE SUBJECTS WE HAVE COVERED SO FAR

IN THIS CHAPTER.

a. Types of Solid State Reactions

b. Formation of Phase Boundaries between Reacting Solids
c. Diffusion Processes in Solids
d. Nucleation Rate Processes
e. Role of Defects in Solid State Reactions
f. The Critical Radius and Transformation from Embryo to Nucleus,
e. Diffusion Laws and Mechanisms
g. Analysis of Diffusion-Limited Solid State Reactions
h. Precipitation Reactions
i. Sequences in Particle Growth
j . Ostwald Ripening Mechanisms
k. Sintering Processes
1. Grain Growth Mechanisms

Let us summarize what we have determined :

1) We can define the exact nature of a solid state reaction.

2) Before any solid state process can occur, nuclei must form so as
to allow the reaction to go to completion.
3) In some cases, an embryo must form before the nucleus
formation is complete.
4) Nucleation can occur within a given solid (homogeneous
nucleation) or be induced by any outside factor (heterogeneous),
including the presence of a foreign solid.
5) In a diffusion-limited reaction, movement of species occurs by a
"hopping" motion.
6) If a solid state reaction is diffusion-limited, if is unlikely that we
can obtain 100% of any product, and will always obtain a mixture of
compounds whose relative ratio will depend upon their
thermodynamic stability at the firing temperature.
7) Several different types of species, including various solid state
defects, diffuse and form a phase boundary of reaction, which may
further react to form specific compositions.
 3.5 Methods of measurement of solid state reactions 183

8) Precipitation is usually homogeneous. Formation of embryos

which grow into nuclei which grow into particles suspended in the
mother-liquor will vary according to the precipitation: temperature,
concentration of reacting components and method of addition.
9. Particles can grow into larger particles either due to Ostwald
Ripening (in solution) or by sintering (solid state).
10. Particles that are too small cannot sinter to grow into larger
particles.
11. Particles that are formed by either precipitation or sintering
form grain boundaries within each individual crystallite. This is due
to consolidation of the very small crystallites, having a large surface
area, into larger crystallites, having a decreased surface area.

We have now examined the factors involved in solid state reactions and
particle growth. Nucleation, particle growth and diffusion mechanisms
were the main theme of our discussion of solid state reactions, but we
have not addressed how we can obtain a measure of solid state reactions
as they occur. That is, how do we measure how the reaction occurs? The
next section shows how DTA and TGA are applied to the decomposition
and reaction of solids.

3.5. - METHODS OF MEASUREMENT OF SOLID STATE REACTIONS

If we wish to follow a solid state reaction from initial compound(s) to final

product(s), there are only a few methods we can use. Of primary
importance is x-ray identification since we must know what we started
with, and what we end up with. We have already discussed the x-ray
method in some detail and how one goes about using the method.

If a solid is stable at room temperature, it will remain in that state until it

is heated. We find that two effects occur simultaneously, a thermal change
and a weight change. As an example, consider CaCO3. When it is heated to

about 800 °C, it forms calcium oxide, CaO, by solid state reaction, vis:

3.5.1.- CaCO3 -i-heat = CaO + CO2 ft

184 Nucleation, diffusion, growth of particles and measurement of solid state

The arrow indicates that the gas formed, (X>2, is volatile and that a weight
loss occurs. The orthorhombic structure of CaCO3 changes to the cubic
form of CaO. Thermal energy is required to rearrange the atoms. What has
actually happened is that we have exceeded the bonding energy of one
compound (CaCO3) by increasing the vibrational energy of the atoms to the
point where chemical bonds are broken. This occurs at about 840 °C. and
OO2 gas is formed which is stable (and volatile). When we cool the
product, we find that we have CaO. Because this change requires heat
which is absorbed, the overall process is endothermic. If it had released
heat during the change, it would be called exothermic.

Measurement of the weight change is called thermogravimetric analysis

(TGA) whereas measurement of the thermal change accompanying the
structural metamorphosis is called thermal analysis (TA). ALL CHANGES
IN PHASE involve a release or absorption of calories. One reason for this is
that each solid has its own heat capacity. That is, there is a characteristic
heat content for each material which depends upon the atoms composing
the solid, the nature of the vibrations within it, and its structure. The total
heat content, or enthalpy, of each solid is defined by:

3.5.2.- AH = J Cp dT , where : Cp = (aq/arr) p

Thus, as we go from one solid to another, we see a change in caloric

content.

I. Differential Thermal Analysis

In 1821, Seebeck discovered that by joining two wires of different

chemical composition together to form a loop (two junctions), a direct
current (DC) would flow in the circuit, namely-

3.5.3.-
 3.5 Methods of measurement of solid state reactions 185

Seebeck used antimony and copper wires and found the current to be
affected by the measuring instrument (ammeter). But, he also found that
the voltage (EMF) was directly proportional to the difference in
temperature of the two junctions. Peltier, in 1834, then demonstrated
that if a current was induced in the circuit of 3.53., it generated heat at
the junctions. In other words, the SEEBECK EFFECT was found to be
reversible. Further work led to the development of the thermocouple,
which today remains the primary method for measurement of
temperature. Nowadays, we know that the SEEBECK EFFECT arises
because of a difference in the electronic band structure of the two metals
at the junction. This is illustrated as follows:

3.5.4.- The Fermi Level at a Junction of Two Dissimilar Metals

In this diagram, we show the band model structure at the juncture of two
metals, each of which has its own Fermi Level. Flow of electrons is
indicated by the arrow. Since the height of the Fermi Level is proportional
to temperature, then the EMF generated is a function of temperature also.
It is thus apparent that a thermocouple (TC) will consist of a negative and
a positive "leg".

The common thermocouples in use today are listed in Table 3-3 along
with the temperature range over which they are useful. Also listed is the
approximate EMF generated over this range, as well as the nature of each
"leg", i.e.- positive or negative. The compositions of the alloys used to
make the thermocouples listed in Table 3-3 are given on the next page as
3.5.5. on the next page. In differential thermal analysis, i.e.- DTA, we use
one thermocouple "bucked" against another of the same composition to
186 Nucleation, diffusion, growth of particles and measurement of solid state

TABLE 3-3
USEFUL TEMPERATURE RANGES FOR COMMON THERMOCOUPLES

Output Range Useful Temp. Useful

Composition Code (millivolts) Range, °F. Atm.*
(+)Copper-Constantan(-) T -5.28 to 20.81 - 300 to 750 A, N, R
(+)Iron-Constantan(-) J -7.52 to 50.05 -300 to 1600 R
(+)Chromel-Alumel(-) K -5.51 to 56.05 -300 to 2300 A, N
(+)Chromel-Constantan(-) E Oto 75.12 32 to 1800 A, N, R
(-)Platinum-Pt(10%Rh)(+) S 0 to 15.979 32 to 2900 A N
(-) Pt- Pt (13% Rh)(+) R 0 to 18.636 32 to 3100 A N
(-}Tungsten(5%Re) - C 0 to 38.45 32 to 5000 N, R
Tungsten(26%Re) ( + )
• A = air or oxidizing; N = neutral; R = reducing

3.5.5.- Compositions Used to Make Thermocouples

CHROMEL: 90% Ni - 10% Cr

ALUMEL: 95% Ni - 5% Al , Si , Mn
CONSTANTAN: 57% Cu - 43% Ni

produce a "net" EMF. What this means is that either the positive (or
negative) legs of both thermocouples are electrically connected so that
the net EMF at any given temperature is zero. Only if one thermocouple
temperature differs from that of the other does one obtain an EMF
response.

3.5.6.- A Thermocouple Used for Differential Thermal Analysis

 3.5 Methods of measurement of solid state reactions 187

If we put a sample next to one thermocouple and a "standard" next to the

other, we can follow any thermal changes that may take place as the
sample is heated since each TC generates its own EMF as the temperature
changes.

Thus, we put a reference material, R, directly in contact with the "(1)"

thermocouple junction (hereinafter, we will refer to this thermocouple
junction as "TC-1") and a sample, S, at the other, i.e.- TC-2, and can
detect thermal changes occurring in the sample as compared to the
reference. Note that if both TC-1 and TC-2 of 3.5.6. are at the same
temperature, no EMF is generated. Actually what we are measuring are
changes in heat flow as related to Cp (see 3.5.2.). For inorganic materials,
the best reference material to use is a- AI2O3 . Its heat capacity remains
constant even up to its melting point (1930 °C). In DTA, we want to
measure ACP , but find that this is actually:

3.5.7.- [Cp(Sf) - Cp (Si )ldTs + [Cp]dT R

where Si is the initial state and Sf is final state for a given solid state
reaction of the sample, S (No reaction occurs for R).

It should be apparent that we must maintain an equal heat flow into both
R & S simultaneously, at a uniform rate. If we raised the temperature by
steps, we would find that the actual heat flow in both R & S lags behind
the furnace temperature considerably, as shown in the following diagram,
shown as 3.5.8. on the next page.

But if we program the temperature in a linear manner, the sample

temperature also increases linearly. The rate of heating generally used for
most inorganic materials ranges between about 2° to 20°/min. while that
for organic compounds lies between about 15° to 100°/min.

Thus, any signals from the DTA thermocouple will mirror differences in
the heat capacity between the sample and the standard material. Since we
use a-Al2O3 as the reference material (which has a constant heat capacity
from room temperature to its melting point of 1830 °C), the DTA
188 Nucleation, diffusion, growth of particles and measurement of solid state

3.5.8.-

thermocouple output will reflect changes occurring in the sample itself

Exothermic and/or endothermic peaks result due to changes in heat
capacity as the sample reacts upon heating.

Thus, CaCO3 reacts to form CaO and an endothermic peak can be

measured (since heat is required to cause CaCO3 to decompose). The
parameters associated with the DTA method include:

3.5.9.- PARAMETERS ASSOCIATED WITH THE DTA METHOD

1. dT/dt = k
2. Sample Size
3. Rate of Heating
4. Degree of Crystallinity of Sample
5. Effects of External Atmosphere

In the following diagram, shown as 3.5.10. on the next page, the

arrangement of the sample, S, and the reference, R, across the
differential TC is shown, and a typical DTA analysis is also given. Note that
at low temperatures, the DTA peaks are endothermic, that is, heat is
absorbed. Such peaks are similar to those obtained when water-of-
hydration is lost, or when the solid state reaction undergoes a loss of
water such as the following reaction, shown as 3.5.11. on the next page:
 3.5 Methods of measurement of solid state reactions 189

3.5.10.-

3.5.11.- CaHPO4 • 2 H2O => CaHPO4 + 2 H2O ft

The broad endothermic peak shown in the above diagram is similar to

that we might see for a change in composition such as:

3.5.12.- 2CaHPO 4 => Ca2 P2 O7 + H2O ft

The exothermic peak may be a change in structure such as:

3.5.13.- p-Ca2P2O 7 => a - C a 2 P 2 O 7

In general, solid state decomposition reactions occur as endothermic

peaks (AH is negative and heat is absorbed) while phase changes, i.e.-
changes in structure, occur as exothermic peaks (AH is positive and heat
is evolved). Note that temperature change is programmed to be linear
190 Nucleation, diffusion, growth of particles and measurement of solid state

with time, in the above diagram.

The components of a simple DTA Apparatus are shown in the following

diagram:

3.5.15.-

Note that you could build your own DTA apparatus if you so wished. The
procedure to obtain a DTA run is given as 3.5.16. on the next page. The
sample size (a function of its crystallinity) is generally determined by trial
and error. About 500 milligrams is usually sufficient. The reason for this is
practical. One must maintain a constant heat flow across the DTA head.
 3.5 Methods of measurement of solid state reactions 191

3.5.16.- GENERAL PROCEDURE FOR OPERATING A DTA APPARATUS

a. Load both sample and standard material into the DTA holder.
b. Put the DTA holder in place in the furnace.
c. Set rate for temperature programming.
d. Set DC Amplifier gain.
e. Set recorder gain and time drive
f. Begin temperature program.
g. Record Temperature changes

If the sample size is too large, the heat cannot diffuse evenly across the
entire sample. The effect of heating rate is shown in the following:

3.5.16.-

In general, the inorganic system cannot adjust fast enough to maintain a

constant heat flow. This is, of course, a matter of system-design. However,
for organic compounds, a rate of 100 °C./min may be required. For
inorganics, a rate of 2 °C. spreads out the peak whereas a rate of 20 ° C.
per minute seems to be about correct for most systems.
192 Nucleation, diffusion, growth of particles and measurement of solid state

One can program at higher rates, even up to 100 ° C.I min.

It is possible to analyze the heat flow involved within the system in terms
of the DTA peak produced. Consider the following diagram, given as
follows:

3.5.17.-

The sample, S, is within the furnace which is at a temperature, T. The

heat flow is dQ/dt and r is the thermal resistance . The sample undergoes
an enthalpy change , AH, at its solid state reaction point. Thus, the heat
flow is a function of the differences in temperature of the sample and that
of the furnace, i.e.- T > T s . Then:

3.5.18.- dQ/dt = T - T s / r

where t is the time. And the enthalpy change is the difference between
that of the sample and the heat flow:

3.5.19..- dH/dt = Cp(S) dT /dt - dQ/dt = Cp(S) dT/dt - [T- T s / r]

If we set up the same equations for the reference material, combine these
equations and rearrange, we obtain:

3.5.20.-

r[dH/dt] = [ T S - T R ] +[r(Cp(s) -Cp(R))] dT R /dt+ [rCp(S){d(Ts - TR)/dt}]

 3.5 Methods of measurement of solid state reactions 193

Note that we have divided the equation into three (3) parts, each
surrounded by a bracket, i.e. [ ]. This allows us to interpret aDTA peak, as
shown in the following:

3.5.21.-

The deviation from the base line (at 0) is a function of both I and II, that
is- the difference between sample and reference temperatures (I) and
differences in heat capacities of sample and reference. Thus, if the
apparatus is properly designed, I is not a problem, but n cannot be
controlled. It is II that causes the deviation from linearity which results in
a peak. The slopes are a function of the heat capacity differences between
sample plus reference and product plus reference. The slopes (III)
obtained are a function of differences between Ts and T R . Note that at
the top of the peak, we still have approximately 1/2 sample (as reactant)
and 1/2 product.

Referring to 3.5.20. and 3.5.21., it would seem that the area of the DTA
peak should be proportional to AH. This is indeed the case and a
comparison of peak areas does yield an experimental value for AHjsj, vis:

3.5.22.- AH(s) = AHstd. • A(S) / A<std)

This equation assumes a relation between A and AH. Actually, this is not
too hard to prove, as we can show in the following. Consider a heat sink as
a block containing both S and R (sample and reference). This is illustrated
in the following diagram, presented as 3.5.23. on the next page.
194 Nucleation, diffusion, growth of particles and measurement of solid state

3.5.23.-

In this diagram, we define three temperatures, T B , Ts & T R , where each

subscript refers to block, sample and reference, respectively. We also have
two heat capacities, Cp(S) and Cp(R). What we wish to prove is:

3.5.24.- AH = AREA/a

where a is the thermal diffiisivity. For convenience, we define the total

heat flow in terms of heat flow between S & B, R & B, and R & S.

We also use a STANDARD SAMPLE and run it against the reference, so

that we can determine how many calories per gram are required for a
given transition so as to calibrate the system.

Table 3-4, given on the next page, shows some materials suitable for such
calibration, along with the temperature at which the solid state change
occurs. Note that the reference material, a-Al2C>3, is thermally inert
whereas the standard reference materials are not. Appropriate values of
AH are available for the standard reference materials given in Table 3-4.
Returning to our description of an analysis of the heat flows present in
DTA, this gives us the following equations , namely:

3.5.25.-
dT/dt = K
KS = (dQ/dt)SB
KR = (dQ/dt)RB
k = (dQ/dt) RS
 3.5 Methods of measurement of solid state reactions 195

TABLE 3-4
THERMOMETRIC FIXED POINTS
FIXED POINTS* TEMPERATURE
°C. ° F.
B. P. Of O 2 - 183.0 - 297.3
Sublimation Point of OO2 - 87.4 - 109.2
F.P. - Hg -38.9 -38.0
Triple Point of Water 0.01 32.0
Ice Point 0.00 32.0
B.P. - Water 100.0 212.0
Triple Point of Benzene 122.4 252.4
B. P. of Naphthalene 218 424.3
F.P. of Sn 231.9 449.4
B.P. of Benzophenone 305.9 582.6
F.P. of Cd 321.1 610
F.P. of Pb 327.5 621.5
F.P. of Zn 419.6 787.2
B.P. of S 444.7 832.4
F.P. of Sb 630.7 1167.3
F.P. of Al 660.4 1220.7
F.P. of Ag 961.9 1763.5
P.P. of Au 1064.4 1948
F.P. of Cu 1084.5 1984.1
F.P. of Pd 1554 2829
F.P. of Pt 1772 3222
M.P. ofIr 2493 4520
M.P. of Ta 2988 5410
M.P. of W 3438 6220
| • F.P. = Freezing Point :
[ M.P. = Melting Point
! B.P. = Boiling Point IL.. . . . . . . . . . .
:

We can immediately write ( see 3.5.2., 3.5.18. & 3.5.19.) :

196 Nucleation, diffusion, growth of particles and measurement of solid state

3.5.26.- Cp(S) d T s / d t = K s ( T B - T s ) +k((T R -T S ) + d(AH)/dt

and:
CP(R) dT R /dt = KR (TSB- TR) + k (T S - TR)

The equation for R is simplified because R is thermally inert and there is

no change in enthalpy involved. To simplify matters further, we define:

3.5.27.- Os = Cp( S )/K s

OR = Cp(R) / KR
HS = k / KS
HR = k / KR

Making these substitutions, we get:

3.5.28. - O s dTs / d t + (1+ H S )T S - H S T R = T B + (1 / Ks) d(AH)/dt

and:
OR dT R /dt + (1 +HR) T R - H R T S = T B

Now if k = 0, i.e.- there is no heat exchange between R & S (as in a

properly designed apparatus), and if:

3.5.29.- TS =T0 +t(dT/dt)

as it will be if we are programming the temperature. Then we can define

T o as being equal to zero, so as to obtain the following equations:

3.5.30.- <3>S d T s / d t + T s = t (dT/dt) +(1 /Ks)d(AH)/dt

and:
OR dT R /dt + T R = t (dT/dt)

If we are not in a region where a solid state reaction is taking place, then
d(AH)/dt = 0.

The change in baseline temperature (see above) is now:

 3.5 Methods of measurement of solid state reactions 197

3.5.31.- ATB = dT/dt ( O s O R ) = TS - TR

In other words, the change in baseline temperature is caused by a

difference in the relative temperatures of sample and reference, which is
related to their relative heat capacities. Thus, one needs to choose the
reference material very carefully. If we subtract the equations in 3.5.30.,
we can get:

3.5.32.- O s d(T s - T R ) / d t + ( T S - T R ) - (dT R /dt) = (1 / Ks) d(AH)/dt

This can be rearranged to:

3.5.33.- O s d ATS /dt + AT - (dT/dt) = (1 / Ks) d(AH)/dt

AT is the difference in temperature between the reference and the

sample. In other words, it is AT that creates the DTA peak. Since we are
not measuring absolute values of the temperatures, we can define a
relative temperature:

3.5.34.- ATRei = AT - ATB - AT - (T s - T R)

Using this relation, we get:

3.5.35.- <J>sdATRei/dt+AT R +(dT/dt-dT s /dt)(Os-OR) = (1/Ks) d(AH)/dt

If we choose a suitable reference material (such as<x-Al2O3for inorganics),

then dTR /dt will be equal to dT/dt. Our equation is thus simplified to:

3.5.36.- O s J (d ATRei /dt) + ATRei = (1/ Ks) J d AH /dt

However, the first term is equal to zero, so:

3.5.37.- J d AH = Ks J ATRei dt
198 Nucleation, diffusion, growth of particles and measurement of solid state

This is what we started out to prove, i.e.- AH equals the area of the peak
times the total heat flow to the sample.

II. Differential Scanning Calorimetry

Since AH is proportional to the area of the DTA peak, one ought to be able
to measure heats of reaction directly, using the equation: 3.5.22. Indeed
we can and such is the basis of a related method called Differential
Scanning Calorimetry (DSC), but only if the apparatus is modified suitably.
We find that it is difficult to measure the area of the peak obtained by DTA
accurately. Although one could use an integrating recorder to convert the
peak to an electrical signal, there is no way to use this signal in a control-
loop feed-back to produce the desired result.

A more practical way to do this is to control the rate of heating, i.e.-

dT/dt, and provide a separate signal to obtain a heating differential. One
such way that became the basis of DSC is shown in the following diagram:

3.5.38.-

The apparatus consists of a DSC-head within a furnace, like the DTA

apparatus. However, there is also a silver block which encloses the DSC
head as well. This ensures complete and even heat dispersion. There are
 3.5 Methods of measurement of solid state reactions 199

individual heaters for both the reference (R) and sample (S) pan-holders.
What is measured is the current required to keep the differential
thermocouple balanced, i.e. AT = 0. This signal can be amplified and
recorded.

We use the same approach for DSC as we did for DTA. We start with the
thermal heat flow equation which is similar to Ohm's Law, vis-

3.5.39.- dQ/dt = T B - T s /r

We can define the heat change involved with the sample as dh/dt so as to
get the equation:

3.5.40.- dh/dt = C p (s) [dT s /dt] - dQ/dt = Cp (S) [dT s /dt] - [T s - TB] / r

We are using the same terminology for DSC as we did for DTA. Following
the methods given above, we arrive at:

3.5.41.- dq/dt = {Cp(s) - Cp(R)) dT B /dt + 1 /r{dT B /dt • t}

We can thus "interpret" a DSC peak in terms of this equation, as we did

for the DTA peak, as shown in the following diagram:

3.5.42.-
200 Nucleation, diffusion, growth of particles and measurement of solid state

As can be seen, the initial part of the reaction up to (dq/dt) m a x is linear

whereas the curve becomes exponential past the peak. This is due to the
difference in Cp between the reactant and product. Thus, Ai * A2.

Actually, the reaction peak in 3.5.42. is an idealized curve since the

baseline is a function of the difference in heat capacities between
reference to sample and reference and product. Usually, we have a
different baseline, vis:

3.5.43.-

This presents a problem since it is difficult to estimate the area of the

peak. One cannot simply extend the baseline as in Case I. A much better
solution is that shown in Case II, wherein the two very asymmetrical
peaks, i.e.- the initial and final parts of the overall thermal reaction taking
place, are delineated.

III.- Utilization of DTA and DSC

We will now compare these two methods of thermal analysis of solids.

a. Applications of DTA

One of the major uses of DTA has been to follow solid-state reactions as
they occur. All decomposition reactions (loss of hydrates, water of
constitution, decomposition of inorganic anions, e.g.- carbonate to carbon
dioxide gas, etc.) are endothermic and irreversible. Likewise are the
synthesis reactions such as CaO reacting with AI2O3 to form calcium
 3.5 Methods of measurement of solid state reactions 201

aluminate, CaAl2O4. Phase changes, on the other hand, are reversible, but
maybe endothermic or exothermic.

Thus, if we follow a solid state reaction by DTA and obtain a series of

reaction peaks, it is easy to determine which are phase changes by
recording the peaks obtained during the cooling cycle. Whereas DTA data
are qualitative, those from DSC are quantitative and give information
concerning the heat change (change in enthalpy) accompanying the
exothermic or endothermic reaction. For example, one can obtain a value
for melting of a solid state reaction product in terms of calories/gram or
Kcal./mole.

DTA is especially suited in the construction of unknown phase diagrams of

binary compounds. A hypothetical phase diagram and the DTA curves
which would be used to construct it are shown in the following diagram,
shown as 3.5.44. on the next page. In this diagram, the endothermic
(Endo) peaks point to the right while the exothermic (Exo) peaks point to
the left.

Consider a system with two components, A and B (see 3.5.44.). They form
an incongruently melting compound, AB. The compound, AB, forms only a
limited solid solution with A. Most of the composition range is a two-
phase region, with a eutectic. The DTA runs are superimposed on the
specific composition points of the diagram. Thus at (1) on the diagram
(about 10% A and 90% AB), we see one Exo and two Endo peaks. The Exo
peak is the point where the two-phase mixture, A + AB, changes to a
single phase, i.e. - a solid state solution of A in AB. Further on, an Endo
peak indicates the melting point of AB, and finally that of A. At (2), we see
only the two melting points, first that of AB and then A. But at (3), only
the melting point of the eutectic is seen, that is, both A and AB melt at
the same temperature. In our Phase Diagram, the compound, AB, melts
incongruently, that is - it decomposes at its melting point. Therefore, at
(6), a double peak is seen representing the decomposition of AB and the
melting of A. However, B melts at a later time. Note that one can pinpoint
changes in the phase diagram quite accurately by running a DTA
thermogram at specific composition points.
202 Nucleation, diffusion, growth of particles and measurement of solid state

3.5.44.-

For the most part, the thermal changes observed are specific, but it is
wise to cool reversibly, while observing the DTA peaks in cooling so as to
be sure exactly what the original peak represents.

Still another use to which DTA has been employed is the characterization
of amorphous materials. The following, given as 3.5.45. on the next page,
shows a typical DTA thermogram obtained when a powdered sample of
glass is run.
 3.5 Methods of measurement of solid state reactions 203

3.5.45.-

Note that nearly all of the characteristic "glass points" can be determined.
These are:
Tg = Glass transition temperature
T S P = Glass softening point temperature
T D = Glass devitrification temperature
TM = Melting temperature of crystallized product

The only one that is not readily accessible by DTA is the expansion
coefficient. It is determined by use of a thermal expansion apparatus, i.e.-
a dilatometer.

b. Uses of DSC

The greatest use for DSC has turned out to be for characterization of
organic polymers. It has been found that most polymers are amorphous
and have a characteristic T g , i.e.- a "glass" transition temperature which
leads to a "crystalline" phase. Other common uses include determination
of melting points, boiling points, Tg , % crystallinity and oxidative
204 Nucleation, diffusion, growth of particles and measurement of solid state

stability. In obtaining boiling points by DSC, it is necessary to use a closed

pan having an extremely small hole to allow the vapor to escape. The hole
is made by use of a laser and should not be more than 50-80 n in
diameter. When a semi-volatile material is heated, an equilibrium will be
established between material in the gas phase and in the condensed
phase. As the material is being heated, the pressure exerted by the
volatile phase, i.e.- the vapor pressure, increases. The rate of heating is
important and should be kept between about 6-10 °C. per minute. It is
important to have the two phases in equilibrium as the sample increases
in temperature, hence the use of an escarpment in the sample pan to
retard escape of the volatile material. This arrangement allows the vapor
produced to sweep out the air and replace it. At the temperature where
the vapor pressure of the sample exceeds the total pressure of its
surroundings, the material boils. If the outside pressure is kept constant,
there will be an endothermic heat flow associated with condensed phase
material entering the vapor phase. As the temperature increases, the rate
of boiling also increases. When all of the material is in the vapor stage, it
remains in that state until all of the material has boiled off. If the hole is
not small enough, then all of the material will be evaporate and be lost
before the equilibrium condition is attained. If an equilibrium between
vapor and material is not achieved, then the boiling point measured will
not be the true boiling point.

The pan has a small hole on the top to limit the amount of water escaping
at the boiling point. Keep in mind that the heat flow (which is related to
the degree of vapor change achieved) is low in the beginning, but rises
rather fast as the boiling point is reached. The flat leading edge of the
endotherm represents the point where the sample temperature is
constant at the boiling point.
If one measures the boiling points at several pressures, including that of
atmospheric pressure, one can then extrapolate to obtain the vapor
pressure of a material at ambient temperature. This is done using the
Clausius-Clapeyron equation:

3.5.46.- E o = RT 2 ( d l n k i /dt)
and: E o = AH* + RT
 3.5 Methods of measurement of solid state reactions 205

As an example of how the data are obtained, the following diagram shows
the behavior of water at 1.0 atmosphere when it is subjected to the above
conditions:
3.5.47.-

The flat leading edge of the endotherm represents the point where the
sample temperature is constant at the boiling point. If one measures the
boiling points at several pressures, including that of atmospheric
pressure, one can then extrapolate to obtain the vapor pressure of a
material at ambient temperature.

IV. Thermogravimetry

Thermogravimetric analysis (TGA) measures changes in weight of a

sample being heated. A typical apparatus Thermogravimetric analysis
(TGA) is shown on the following page as 3.5.48. The weight is monitored
in real time and changes, either gains or losses, are evident immediately.
The apparatus itself consists of the sample situated within a crucible,
which is enclosed within a temperature-controlled furnace. The sample,
plus crucible, is counterbalanced on a sensitive ANALYTICAL BALANCE.
Weight changes are directly plotted on a two-pen recorder. Weight
readout is usually accomplished by one of two methods, a linear
transducer or a capacitance change between two flat plates, one of which
is free to move with the balance swing. A resistance-capacitance tank
206 Nucleation, diffusion, growth of particles and measurement of solid state

3.5.48.-

circuit completes the electronics, producing a readable voltage. We usually

employ a crucible to hold the powder sample, although flat pans are also
suitable. Furnace temperature is controlled in a linear manner and
recorded. In some cases, the TC is mounted directly at the sample
position. It is important that sample and furnace temperatures be nearly
equal, so as to record accurate weight losses and gains. Operational
parameters for TGA are:

3.5.49.- 1. Sample Size (buoyancy)

2. Sample Closure
3. Heating Rate
4. Heating Mode
5. External Atmosphere
 3.5 Methods of measurement of solid state reactions 207

The SAMPLE SIZE is important because most balances have a limited

range of weighing, as well as a limited sensitivity, i.e.- milligrams per
gram of weight detectable. Many balances feature automatic counter-
weight loading. If a sample is fluffy and a large crucible is used, then the
buoyancy factor must be accounted for. At high temperatures, i.e.- > 800
° C the density of air within the furnace is sufficiently lower than that of
the outside, so that the apparent weight of the sample plus crucible
appears lower than it actually is. And, the larger the crucible, the more air
is displaced within the furnace. For 10.000 grams of total weight, the
buoyancy factor will be about 0.002, enough that a correction needs to be
made for precision work. Sample closure is important since it affects the
rate of solid state reaction. Consider the following solid state reaction:

3.5.50.- CaCO3 % CaO + CO2 ft

If the gas is restricted from escaping, then the equilibrium is shifted to

the left and the CaCO3 does not decompose at its usual temperature. A
higher temperature is required to effect decomposition.

Likewise, an external atmosphere, such as CO2 in the above reaction,

restricts the apparent temperature of decomposition, the rate of reaction,
and sometimes, the mode of decomposition (depending upon the nature
of the compound under investigation). CaCO3 normally decomposes at 860
°C. In a 1.0 atmosphere pressure of CO2, the decomposition temperature
is raised to about 1060 °C.

The HEATING RATE is important from a practical aspect. Usually, we are

measuring furnace temperature and cannot program the temperature too
fast, for fear that the sample temperature will lag the furnace temperature
by too great a degree. It is also for this reason that we use as small a
sample as is practical to obtain a weight loss or gain which the balance can
discriminate. This again depends upon the sensitivity of the balance and
the nature of the reaction we are examining. In general, weight gains (due
to oxidation) require more sensitivity and larger sample sizes than those
of decomposition. Usually, we restrict heating rates to > 15 °C and use a
heating rate of about 6-8 °C/ min. at most. The HEATING MODE to be
208 Nucleation, diffusion, growth of particles and measurement of solid state

used depends upon the results we wish to achieve. A typical TGA run is
given in the following diagram:

3.5.51.-

In the solid state reaction depicted, A begins to decompose to B at T i and

the reaction temperature for decomposition is T2 , with a weight loss of
Wi. Likewise, the reaction of B to form C begins at T3 and the reaction
temperature (where the rate of reaction is maximum) is T4 . Note that the
weight loss becomes constant as each reaction product is formed and the
individual reactions are completed.

If we program the temperature at 6 °C/min., we would obtain the results

in 3.5.50. This is called dynamic thermogravimetry. However, if we set
the furnace temperature just slightly greater than T2 , we would obtain a
reaction limited to that of A decomposing to B, and thus could identify
the intermediate reaction product, B. This technique is called isothermal
thermogravimetry.

Thus, we can follow a solid state reaction by first surveying via dynamic
TGA. If there are any intermediate products, we can isolate each in turn,
 3.5 Methods of measurement of solid state reactions 209

and after cooling (assuming each is stable at room temperature) can

identify it by x-ray analysis. Note that we can obtain an assay easily:

3.5.51.- Assay • final weight / original weight

If there is a gaseous product, we can also identify it by converting weight

loss to mols/mol of original reactant. In 3.5.50. above, 1.00 mol of OC>2 is
expected to be lost per mol of reacting CaC»3 . The actual number of mols
lost depends upon the original sample size.

The most recent TGA apparatus includes what is called EGA, i.e.- effluent
gas analysis. Most often, this consists of a small mass spectrograph
capable of identifying the various gasses most often encountered in TGA.
Gaseous weight losses can be classified according to the nature of the
effluent gases detected. These include the following:

3.5.52.- Gaseous Products = CO2 , N 2 O 4 , H2O , SO2 , SO3 , CO.

Water - of hydration
- of constitution
- adsorbed

Regardless of how well a sample has been dried, it will always have an
adsorbed monolayer of water on the surface of the particles. If we run the
DTA carefully, we will see a small endothermic peak around 100 °C.
Additionally, if we run the TGA properly, we will see a small loss plateau
before the major losses begin. As a matter of fact, if we do not see the loss
of adsorbed water, either the apparatus is not operating properly, or we
do not have sufficient sensitivity to observe the reactions taking place.

The different types of water which can be present during any inorganic
solid state reaction is easily illustrated by the following example:

Dibasic calcium orthophosphate can be formed as a dihydrate or as an

anhydrous product, depending upon the conditions used in its
precipitation:
210 Nucleation, diffusion, growth of particles and measurement of solid state

3.5.53.- CaHPO4 • 2 H2O - brushite

CaHPO4 - monetite
Brushite reacts to form monetite which then reacts to form
pyrophosphate:

3.5.54.- 2 CaHPO4 • 2 H2O = 2 CaHPO4 + 2 H2O ft

2CaHPO 4 = G12P2O7 +H 2 O1t

We can illustrate the type of calculations needed in order to determine

the parameters involved in TGA runs. In the following Table, presented on
the next page, atypical problem that one encounters in TGA is shown and
the calculations needed to produce the desired results.

We start with the reactions given in 3.5.54. for the reactions of calcium
phosphate, since this also illustrates how assays are calculated. The steps
include:
1. Determine assay
2. By subtracting actual assay from theoretical assay,
obtain amount of water actually adsorbed on particle
surfaces
3. Determine losses incurred by stages, if one wishes to
determine the actual reactions occurring during the
solid state reaction

In the following, we present a typical problem one encounters in TGA,

and the calculations needed to produce the desired results. In this case,
we start with a known material for which we have already determined the
fired product by x-ray analysis.

Table 3-5
A TYPICAL PROBLEM IN TGA
METHOD ORIGINAL FIRED PRODUCT
X-ray Analysis: brushite Ca2P2O 7
Weight: 17.311 gram 12.705 gram
Molecular Weight: 172.09 254.11
 3.5 Methods of measurement of solid state reactions 211

Our first step is to analyze the solid state reaction by means of the values
determined in the TGA analysis run. The reaction products are given in
the following, along with the requisite molecular weights:

OVERALL SOLID STATE REACTION:

2 CaHPO4 • 2 H2O = Ca2 P2 O7 + 3 H2O ft
Molecular Weight: 2(136.059) + 2(18.015) 254.11 + 54.045
WEIGHT: 17.311 gram 12.705 gram
(Assay = 73.39%;
Theor. = 73.83%)
ANALYSIS: 12.705 gram Ca2P2O7 « 17.208 gram of
2 CaHPO4 • 2 H2O

By calculation, we find that we have 99.41% of 2 CaHPO4 • 2 H 2 O and

0.59% adsorbed water (by subtraction). We can also determine from the
TGA run the amount of water loss by stages in the overall reaction:

Losses by Stages: 1st loss = adsorbed H2O = 0.102 gram

2nd loss = = 3.532 gram H2O
brushite to monetite = 3.92 mols
3rd loss = monetite to = 0.972 gram H2O
pyrophosphate = 1.08 mols
Total wt. of water lost = 4.606 gram

Note that the second loss corresponds to 3.92 mol of water per mol of
reactant, whereas the 3rd loss is 1.08 mol. This illustrates a serious
problem that can be encountered in dynamic TGA, If the rate of heating is
too fast and not enough time occurs during programming to achieve true
equilibrium between successive solid state reactions, then the loss of
water from one reaction carries over into the next succeeding reaction.

V. Determination of Rate Processes in Solid State Reactions

We have presented two methods useful in following solid state reactions.

In order to completely classify a reaction, we need to obtain an estimate of
212 Nucleation, diffusion, growth of particles and measurement of solid state

the reaction kinetics and order of the solid state reaction. Both DTA and
TGA have been used to obtain reaction rate kinetics. But first, we must
reexamine kinetic theory in light of solid state reactions, which differ
from those involving gaseous components usually quoted in books dealing
with reaction rate kinetics.

a. Types of Solid State Reactions

In general, we can classify solid state reactions as being either

homogeneous or heterogeneous. The former involves reactions by a single
compound whereas the latter involves reactions between two different
compounds. There are at least four (4) types of solid state reactions (as we
have already presented previously):

3.5.55.- Types of Solid State Reactions

1. decomposition: A =>B+C
2. synthesis : A + B => C
3. substitutional: A + B => C + D
4. consecutive: A => B => C

All four may be heterogeneous, but only #1 (and sometimes #4) will be
homogeneous. Rate processes are defined in terms of a rate, r , and a
volume, V, usually a molar volume. Thus, we have:

3.5.56.- Homogeneous: r = 1/V"t • dni/dt

Heterogeneous: r = 1/Vf • dVt/dt

where ni is initial mols, Vt is volume at time, t, and Vf is final volume.

The fraction decomposed at any time is:

3.5.57.- x = Vt / Vf

and this gives us:

3.5.58.- dx/dt = ki (T) • f(x)

 3.5 Methods of measurement of solid state reactions 213

where ki is the rate constant for the reaction described in 3.5.58. This
equation is the general form for kinetic equations.

From the Kinetic Theory, when a gaseous system goes from an initial state
to a final state, the reacting species must come close enough to react. At
the moment of "joining", we have the "activated complex". By using this
concept, we can obtain some general equations useful to us. The concept
of the "activated complex" is illustrated in the following:

3.5.59.-

The free energy of the activated complex, G*. is higher than that of the
initial state. K* is the equilibrium constant of the activated complex, and
k is the Boltzmann constant, then, we can write:

3.5.60.- AG = RT In K*

ki = kT/h exp (-AG»/RT) = kT/h exp (ASVR) exp (-AHVRT)

where the star (*) refers to the activated species. These equations ought
to be familiar to Physical Chemistry students. From the Clausius-Clapyeron
equation, we have:

3.5.61.- E o = RT 2 d l n k i / d t

E o = AH* +RT
214 Nucleation, diffusion, growth of particles and measurement of solid state

where Eo is an internal energy. By defining a frequency factor as:

3.5.61.- Z = kT/h exp (-E0*/RT)

we can simply by means of 3.5.60. to:

3.5.62.- ki = Zexp (-E0*/RT)

which is the ARRHENIUS EQUATION for the rate constant, ki. We can
transform the above equations into general equations as well, and use
them to define the various types of reaction given above. This is given in
the following:

3.5.63.- RATE EQUATIONS FOR SOLID STATE REACTIONS

1. Simple nth Order Rate Equation

ki
nA =* B+ C -dxA /dt = ki x n A

2. Parallel Reactions Rate Equation

ki
A => B
-dxA/dt = ki XA + k2 XA
k2
{A => C}

3. Consecutive
ki k2
A => B => C -dxA/dt = ki XA - k2XB

These rate equations can be used for quite complicated reactions, but a
specific method or approach is needed.

Many authors have tried to devise methods for obtaining rate constants
and orders of reaction for given solid state reactions. None have been
wholly successful, except for Freeman and Carroll (1948). The Freeman
 3.5 Methods of measurement of solid state reactions 215

and Carroll method has been shown by Fong and Chen to be the only one
which gives satisfactory answers to known reactions, whether zero order,
1st order, 2nd order, or even higher. Even fractional orders of reaction
may be determined. This method can be used with either DTA or TGA
data. The following is a description of the Freeman-Carroll method
applied to DTA data:

b. The Freeman-Carroll Method Applied to DTA

Consider the following DTA peak in which AT is plotted vs: time, t, :

3.5.64.-

At any time, a is the fraction decomposed while a * is the fraction which

has yet to react. We can set up equations as before:

3.5.65.- da/dt = ki(l-a)n = Z/ $ exp -E/RT (1- a ) n

The first part of the equation is the general kinetic equation, from which
it is easy to obtain the last part (as we have shown). If we perform the
mathematical operations of : 1) take the naperian log; 2) differentiation,
and then 3) integration, we obtain the following equation:

3.5.66.- A In d a / d t = n (A In (1- a ) - E / R A ( l / T )

If we now set: AT/A = d a / d t and a * - A - a (where A is the area under

216 Nucleation, diffusion, growth of particles and measurement of solid state

the peak) , we can then obtain:

3.5.67.- A (In AT) = n (Aln a *) - E/R A (1 IT)

This allows us to plot: A (In AT) / A (In a *) vs: A (1 IT In a *) so as to

obtain a straight line. The slope is E/R and the intercept is n , the order
of reaction. Having thus obtained the activation energy and order directly,
we can then calculate the reaction rate.

c. The Freeman-Carroll Method Applied to TGA

In this case, we define our equation in terms of weight, w :

3.5.68.- -dw/dt = ki f(w)

As in the DTA method, we define weight reacted in time, t, as wt and final

weight, wf. This allows us to define weight to be reacted as: w = wt - Wf .
Our weight function is then to be defined:

3.5.69.- f(w) = w n

where n is a simple integer, representing the order of reaction. Thus, we

obtain in the same manner as for the DTA data:

3.5.70.- - dw/dt = Z exp (-E/RT) wn

We now do the mathematical manipulations in the order given above to

obtain:

3.5.71.- A (In (-dw/dt)) - E/R (A (1/T)) +nAlnw

This allows us to plot, as before:

A (In - dw/dt) / A In w vs: A(lIT) / A In w .

 3.5 Methods of measurement of solid state reactions 217

However, in many cases, it is easier to use a , the original weight and

calculate: (a - x) and d x / d t , where x is the fraction decomposed at time,
t. Then we plot:

3.5.72.- A (In dx/A) / A (In (a-x)) vs: A (1/T) / A In (a-x)

Thus if we take a well defined curve from either DTA or TGA and find
points on the curve, we can calculate all of the kinetic parameters. Note
that the main concept is to use the amount left to react at any given
instant. In DTA, this was a * while in TGA, it was (a- x). Both thermal
methods give equally satisfactory results.

Recommended Reading

1. W.W. Wendlandt, Thermal Methods of Analysis, Interscience-Wiley,

New York (1964).

2. P.D. Garn, Thermoanalytical Methods of Investigation, Academic Press,

New York (1965).

3. W.J. Smothers and Y Chiang, Handbook of Differential Thermal

Analysis, Chemical Publishing Co., New York (1966).

4. E.M. Barral and J.F. Johnson, in Techniques and Methods of Polymer

Evaluation, P. Slade & L. Jenkins- Ed., Dekker, New York (1966).

5. W.W. Wendlandt, Thermochim. Acta 1 , (1970)

6. D.T.Y. Chem, J. Thermal Anal, 6 109 (1974) - Part I

7. Chen & Fong, loc. cit. 7 295 (1975) - Part II

8. Fong & Chen, loc. cit., 8 305 (1975) - Part III

9. C. Duval, Inorganic Thermogravimetric Analysis, 2nd Ed., Elsevier,

Amsterdam (1963).
218 Nucleation, diffusion, growth of particles and measurement of solid state

10. C. Keattch, An Introduction to Thermogravimetry, Heydon, London

(1969).

11. E. P. Manche & B. Carroll "Thermal Methods", Chapter 4 in Analytical

Methods, , pp. 239-344, Marcel Dekker , New York (1972)

12. Robert E. Reed-Hill, "Physical Metallurgy Principles"- Van Nostrand,

Princeton, New Jersey (1964).

13. A.J. Dekker, "Solid State Physics" - Prentice-Hall, Englewood Cliffs,

New Jersey (1958).
I CHAPTER 4 1
Measuring Particle Size and Growing Single Crystals

In the last chapter, we investigated solid state reactions relating to

formation and growth of crystallites (particles). We examined nucleation
mechanisms during reaction between solids (heterogeneous) and
nucleation mechanisms during precipitation (homogeneous). We will now
determine how the size of such crystallites (as powders) are measured.
We will also investigate how single crystals are grown from a melt and
how stacking faults can arise as a result. Stacking faults are the result of
line defects which occur during particle growth, whether it is from a
melt, by solid state reaction or during precipitation. Of necessity, we will
not be comprehensive in our examination of these topics but will include
enough material to familiarize those who do not have the background
concerning particle distributions and how single crystals are grown.

However, we reserve description of the formation of solids as thin films

to a following chapter. Since the original publication of this volume in
1991, use of thin film technology has grown to be a major part of solid
state chemistry. LEDs (light emitting diodes) and diode lasers have
become an integral part of our lives as witness the traffic lights now in
use. Prior to the development of high-brightness LED lamps, the use of
incandescent lamps combined with an appropriate filter in traffic lights
was widespread. The LED is in the process, or has, replaced such traffic
lamps. The main advantage of the LED traffic light is that it lasts over
25,000 hours before it needs to be replaced. Contrast this to the average
replacement time of incandescent lamps of less than 2000 hours.
Furthermore, the LED uses milliwatts of electrical power while the
incandescent lamp uses watts of power.

4.1.- MEASUREMENT OF PARTICLE SIZES AND SHAPES

Most solids that we actually deal with are powders. The powder is
composed of discrete particles, and each particle may be single crystal or
contain grain boundaries. The size of particles is of interest to us because

219
220 Measuring particle size and growing single crystals

many of their physical (and to a certain extent- chemical) properties are

dependent upon particle size. For example, the "hiding power" of
pigments is dependent upon the size of the pigment particles, and in
particular the distributions of sizes of the particles. There is an optimum
particle size distribution to obtain maximum hiding power, depending
upon the specific application. Pharmaceuticals are another product where
particle size and particle size distribution (PSD) are important. The
effectiveness of dosage and rate of ingestion by the human body depends
upon the proper PSD. Particles are usually defined by their diameter since
many are spheroid. We use the micron as abase, i.e.-

4.1.1.- 1 p (micron) = 10~6 meter = 1 0 ' 4 centimeter = 10 mm.

The following defines particle ranges that we usually encounter in solid

state chemistry-.

4.1.2.- PARTICLE RANGES

Range Centimeters Microns, p Description
Macro 1.0 - 0.05 10 4 - 500 Gravel
Micro 0.01- 0.0001 100 - 1.0 "Normal"
Sub-Micro 0.0001- 10" 7 1.0 -0.001 Colloidal

It is easier to describe particle size in terms of microns rather than

inches, meters or even millimeters. The following diagram, given as
3.1.3. on the next page, summarizes the type of particles that we
encounter in the real world. At the top the particle diameters in microns
are given. Immediately below are standard screen sizes, including both
U.S. and Tyler standard mesh. Screens are made by taking a metal wire of
specific diameter and cross-weaving it to form a screen with specific hole
sizes in it. Thus, a 400 mesh screen will pass 37p particles or smaller, but
hold up all those which are larger. A 60 mesh screen will pass up to 250p
particles, etc. (U. S. Screen Mesh). This diagram also places and defines
the size of most of the particles that we are likely to encounter in the real
world. On the left is a comparison of angstroms (A) and microns up to 1 p
 4.1 Measurement of particle sizes and shapes 221

41.3.-
[Size Range of Typical Particles and Dispersoids[

MICRONS
O.O1 0.1 1.0 10.0 100 1,000 10,000
I B 3+5E1J3] B 3M i. I T H J B 3^5iTHl| 8 i^SETII 8 ] ^ 5 BT i a | S 3 T 5 ITB3J
1
US. STANDARD
^ BoNi sy s i Mini
&E [jro| |m] ra Irt]
SCREEN SEES ^ £&] |25 P | g|]

r™i |i»i dp] i d

IOO ioao ID^OO iao,ooo |p S M^ E
SOLIDS h FUME 4> DU3T
DISPERSOIDS
LIQUIDS I MIST # I SPRAY 4

SOIL I CLAY ++ SILT " • • " f r t SAPC—••"SAND » • GRAVEL

PARTICLES | | I I
ATMOSPHERIC SMOG • ••"CLOUDS & FOC ••J-DRI2ZLE—••"RAIN

¥ ROSIN SMOKE # FERTILIZER GROUND LIMESTONE

# - 0 L SMOKES 44 FLY A3H #

4-TOB hCCO SMOKE 4 4 "COAL DUET •

•"METALLURGICALDUSTS MID FUMES 1 4

• AMMONIUM CHLORIDE FUME< 4 CEHENT D U S T - 4

• ««B^ •-B™ ^4 ¥ ifJS^Lo, „

•-CARBDN BLfcCK— 4 4 _ P a V E R ISED CO A L 4 4-BEACH SAPO—4
I I
4 — P h N T PIGMENTS 4 4 — FLOTATION O R E S - 4
I
COMBUSTION LUHGMMAGIHG
HUCLEI
r 1 I • IDUST • I I I
MICRONS 0.01 0.1 1.0 10 100 1000 10,000

which is = 10, 000 A . In the middle of the section, "Equivalent Sizes" , is

a comparison of sizes and "theoretical mesh". Why the latter term is
222 Measuring particle size and growing single crystals

sometimes used will be discussed later. In "Technical Definitions", size

ranges are given for various types of solids in gases, liquids in gases, and
those for soil. In addition, we have a separate classification for water in air
(fog). Finally, there is a classification for various commercial products and
by-products we might encounter, including viruses and bacteria. Note the
varied sizes of particles that we normally encounter. This listing is not
meant to be all- inclusive but outlines many areas of technical interest
where the sizes of particles are important. Now let us begin to more
clearly define the properties of particles.

Let us take a 1.00 cm. cube and cut it into fourths, as shown:

4.1.4.-

NUMBERS = 4 3 = 6 4

SURFACE = (64)(6)(1/4] 2 = 0.0096 sq. m.

Now we take the same cube and divide it into l . O p cubes, with the
following result:

4.1.5.- l|i cubes = 10 4 cuts in 3-dimensions

Diameter = 1.0 x 10~6 cubic-meters for each particle
Number =(104)3 = 1 . 0 x l 0 l 2 particles
Surface = (6) (1.0 x 10 1 2 ) (1.0 x 10'6)2 = 6.0 sq. Meters

We have discovered two facts about particles:

a) there are large numbers present in a relatively small amount of

material (we started with a 1-cm. Cube).

b) the total surface area can be quite large and depends upon the
size of the particles.
 4.1 Measurement of particle sizes and shapes 223

Since the rate of solid state reaction depends upon surface area, we can
see that very small particles ought to react much faster than large
particles. Another observation is that solid state reactions involving very
fine particles ought to be very fast in the beginning, but then will slow
down as the product particles become larger. There are other properties
of particles which we can think of, as follows:

4.1.6.- PROPERTIES OF PARTICLES

Numbers Shape Size

Aggregate Porosity (pore size) Size distribution
Agglomerate Density Surface area

At this point, we are most interested in the size of particles and how the
other factors relate to the problem of size. However, particle shape will
determine how we define size. Most of the particles that we will
encounter are spheroidal in shape. But, if we discover that we have
needle-like (acicular) particles, how do we define their average diameter?
Is it an average of the sum of length plus cross-section, or what?

I. Particle Shapes

When we determine the size of a particle, we do so based upon its shape.

It is therefore logical that the three-dimensional properties of a particle
directly relate to how the distribution of an ensemble of particles is
measured. To illustrate exactly what we mean by particle shape, consider
the following diagram, given on the next page as 4.1.7.

In this diagram, USP refers to "United States Pharmacopoeia", Vol. USP

24, published in 2000. Note that these shapes are variations in cubic
growth in one or more of the x, y, or z directions of the cubic lattice.
Although we have shown these shapes as regular in the three
crystallographic directions, usually they are not. For example, if one looks
at a "flake" under the microscope, the edges are usually not straight but
are "ragged". There are many particles which grow according to these
examples in which rate of growth is about equal in all crystallographic
224 Measuring particle size and growing single crystals

4.1.7.-
Shapes of Particles According to USP Definitions

directions. However, these are probably the "exception to the rule". When
we have other crystallographic lattices such as tetragonal, hexagonal,
monoclinic or triclinic, the rate of growth along the lattice planes
becomes complex indeed. The USP organization defines these shapes in
terms of length, width and thickness, as follows:

4.1.8.- Definitions of Shape Parameters

• CUBE: particle of similar length, width and thickness in all

directions
• FLAKE: thin, flat particle of similar length and width
• ACICULAR: narrow, needle-like particle of similar width and
thickness
• COLUMNAR: long thin particles similar to acicular except that
the width and thickness are greater
• TABLET: flat particles of similar length and width but thicker
than flakes
• PLATE: flat particles similar to tablets but of differing length to
width
• BLADE: similar to acicular particles, except wider and longer
 4.1 Measurement of particle sizes and shapes 225

In the following diagram, we show how these basic shapes are modified in
growth as particles (crystals) grow along certain preferred planes of the
crystallographic lattice, i.e.- non-cubic:

4.1.9.-

Although these shapes represent regular growth as defined by the

crystallographic lattice, it is rare to find such well-defined crystallites. In
general, crystal growth will follow the direction of the crystal planes of
the lattice and so define the crystal shape. However, rate of crystal growth
may well depend upon the {h,k,l} planes involved. For example, cubic
YAG, i.e.- Y3A15OU , grows very slowly from a melt, especially from a
{1,0,0} plane. If growth is induced from a {1,1,1} plane, the growth of the
single crystal is accelerated by more than ten times. The reason is that
the density of cations along the {1,1,1) plane is much greater than the
226 Measuring particle size and growing single crystals

{1,0,0} plane. Thus, YAG grows at the rate of about 1.0 cm per hour if
growth along the {1,1,1} direction is initiated. However, if neodymium,
i.e.- Nd*1", is added to the melt, the rate drops to 0.10 mm per hour.
YAG:Nd is the solid state laser workhorse used in industry, comparable to
the OO2 (carbon dioxide) laser used in heavy industry.

Concerning particle shapes, one usually does not observe solitary particles
alone but particles that have assembled into more complex arrangements.
The USP volume also provides definitions of such particle mixtures, as
shown in the following:

4.1.10.- Particle Assemblages and Associations

• Aggregate- mass of adhered particles

• Agglomerate- fused or cemented mass of particles
• Spherulite- radial cluster of particles
• Conglomerate- mixture of two or more types of particles
• Drusy- particle covered with tiny particles
• Lamellar Particles- stacked plates
• Particle Conditions including:
1. Edges: angular, rounded, smooth, sharp, fractured &
irregular
2. Optical: color, transparency, translucency & opaqueness
3. Defects: occlusions & inclusions
• Surface Conditions including:
4. Cracked: partial split, break or fissure
5. Smooth: free of irregularities, roughness or projections
6. Rough: bumpy, uneven, not smooth, projections
7. Porous: having surface openings or passageways
8. Pitted: having small dimples or depressions on the surface.

To classify particle shapes in terms of their shape, surface condition and

condition after they are formed by either solid state reaction or by
precipitation is daunting indeed. This is especially true if they are to be
used in medicine where the individual particle shape and condition has a
major effect upon the efficacy of the medicine. The shape and physical
 4.1 Measurement of particle sizes and shapes 227

condition of the particles which are compacted to form a "pill" control

how well the medical molecules are absorbed by the human body. This
fact is well understood by researchers in the biochemical field and is of
significant concern during development of methods of manufacture of
medicinal drugs. Compaction controls how well the medicine is absorbed
in the human body.

In general, the shape and size of the particles determine how well they
can be compacted (under pressure) to form various configurations.
Examples are pharmaceutical pills, pressed forms to be used in plasma-
deposition to construct thin film designs like a LED or a solar-panel, and
metal forms used manufacturing wires like copper and tungsten. In the
latter case, a metal powder is compacted and then sintered to form a
solid rod for manufacture of electrical products. How the powder
particles pack together is of major concern for many technologists. To
define how particles pack together, we need the following definitions:

4.1.11.- Definitions Used in Defining Packing Densities

• Density: d = m/v = mass per unit volume = gm/ml

• Bulk Density: weight of specific volume = gm/in 3
• Relative Density: bulk density/ true density (d)
• Specific Volume: v / m = 1/d = ml/gm
• Absolute Density: true density = m/molecular volume

The first two items are important in inorganic systems while the latter
two are important in organic pharmaceutical assemblages of particles.
The following table presents a comparison of a few commercial materials:

4.1.12.- Bulk Densities and True Densities of Some Common Materials

Material True Density Bulk Density Relative Density Porosity
CaHPO4 2.80 gm/ml 1.43 gm/ml 0.51 49%
Steric Acid 0.99 0.59 0.60 40%
Lactose 1.55 0.67 0.43 67%
Cellulose 1.57 0.46 0.29 71%
228 Measuring particle size and growing single crystals

Note that we have only presented a few examples. Bulk density is a

measure of how well particles pack together in a defined volume. It
depends on the shape of the particles comprising the powder. Most of
the organic compounds (Pharmaceuticals) will exhibit elongated shaped
particles. Note that at least 50% of the volumes presented in 4.1.12 are,
in most cases, unoccupied space, i.e.-porosity. We conclude that if
someone measures the particle size and the size distribution, we need to
know exactly what was measured, often in terms of the particle shape.
Particle size is customarily measured in terms of two-dimensional
(physical aspect) or three-dimensional (volume) values. However, the
particle size is often described in terms of spheroidal diameters even
though the particles may be acicular in nature.

II. Measurement of Particle Size

Fortunately, there are methods which we can use to describe how particle
shape and particle size are measured. An obvious method is to use the
microscope, as shown in the following:

4.1.13.-

filCROSCOPIC COUNTING - PARTICLE DIAMETER EYEPIECE

 4.1 Measurement of particle sizes and shapes 229

At the proper magnification and with a special eye piece, one can directly
measure particle diameters. The eye-piece must have internally-marked
concentric circles so that a given particle will fall within one or more of
the circles. The diameter then can then be read and/or estimated
directly. In this case, three particles are shown, a 40 \x, 20 \i and a 10 p
particle. The most important step is sample preparation on the
microscope slide. Since only a pinch of material is used, one must be sure
that the sample is uniform and representative of the material. Also, since
most materials tend to agglomerate due to accumulated surface charge in
a dry state, one adds a few drops of alcohol and works it with a spatula,
spreading it out into a thin layer which dries. Too much working breaks
down the original particles, particularly the larger ones, and too little
working leaves agglomerates. With a little practice, the preparation of a
proper slide of particles becomes easy.

One then proceeds to count particles. It has been found that if one wishes
to determine an accurate value of the mean diameter, one must count at
least 450 particles. As the number of counts increases, the mean comes
closer and closer to the true value, as shown in the following:

4.1.14.-

If we wish to obtain an average of the diameters, we sum the number of

230 Measuring particle size and growing single crystals

particles between a certain range, i.e.- from 4.0p to 6.Op. The number of
particles at di, then those at d2, etc., are then summed and then are
averaged:

4.1.15.- d"= 2 ni di / 2 ni

where d is the average diameter of particles in that range, n, is the

number of individual particle ranges and d, the diameters of those
particles. Another method is to count particles using an automatic
particle counting instrument which sells for about $ 50-60,000. The
instrument consists of a microscope, a scanning device (usually a flying-
spot scanner), a television display and a programmed microprocessor. All
of the particles within a given frame can be counted and grouped
automatically. Thus, several frames per minute can be counted by an
operator, or the instrument can be set to operate automatically according
to a preset program. Since the instrument scans images of the particles, a
problem of orientation arises. This is shown in the following diagram:

4.1.16.-
PARTICLE ORIENTATION

The microprocessor is programmed to recognize length and width, but

the question of the correct diameter of irregular particles has remained
controversial. For example, Martin's diameter is defined as the shortest
line which divides the area of the image in half. Other definitions have
been used as well. A depiction of this is shown as follows in 4.1.17., given
on the next page. Yet, it is obvious that this is not the correct diameter
for the these particles. Another school of thought would use the average of
the length and width, but for the above particles, the actual value of the
length that ought to be used remains questionable.
 4.1 Measurement of particle sizes and shapes 231

4.1.17.-

This problem has remained the most serious barrier to obtaining PSD by
optical microscopy until recently.

With the advent of computers which can process data at the rate of 700
billion bits per second, this problem has been solved satisfactorily. This
instrument (Malvern Instruments, 10 Southville Road, Southborough, Ma.
01772 - Sysmex FPIA-2100) is a fully automated particle size and shape
analyzer. It uses CCD, i.e.- "charge-coupled device", technology (the
optical basis of a digital video camera) to capture a series of images of
particles suspended in a liquid medium. Particles are sampled from a
dilute suspension which is forced through a "sheath-flow" cell. This
insures that the largest area of the particle is oriented toward the video
camera and that all of the particles are in focus. The cell is illuminated via
a stroboscopic light source and images are captured at 30 Hz. per second.
A computer program then processes the images in real time by the
following steps: digitization, edge highlighting, binarization, edge
extraction, edge tracing and finally storage. The following diagram, given
on the next page as 4.1.18., shows how this is accomplished.

All this requires a high speed computer with large memory storage
capacity and RAM (something not possible until recently). Image analysis
software then calculates the area and perimeter of each of the captured
images and then calculates the particle diameter and circularity.
232 Measuring particle size and growing single crystals

4.1.18.-

The particle shape is easily recognized in this diagram. The area of the
circle is equal to that of the digitized particle so that the particle
circularity (shape) can be classified.

Circularity, C, is determined by:

4.1.19.- C = perimeter of circle /perimeter of particle

Once the measurement is complete, the particle size and circularity can
be displayed in both graphical and table form. A typical report includes:
particle size distribution, circularity distribution and a scattergram of
particle size vs: circularity. All these can be displayed and printed in hard
copy. Individual particle images can be displayed and then classified into
categories including uni-particle, bi-particle, and agglomerate. This
rather new instrument has solved the problem of particle shape and
corresponding size through optical digitization methods.

III. Measurement of Particle Size

Fundamental to particles is the size distribution. By this, we mean the

number of each size in a given collection of particles. From a close
examination of 4.1.5., it is evident that we must deal with large numbers
of particles, even when we have a small starting sample. This problem
 4.1 Measurement ofparticle sizes and shapes 233

becomes clearer if we correlate numbers of separated particles with size,

starting with a cube, one (1) cm in diameter, as in 4.1.4. , and make the
cuts indicated as follows:

4.1.20.- Size Number/ cm 3

1000 fj l.OxlO3
500 ji 8.3 x l O 3
250 ji 6.4 x 10 4
100 n 1.0 x l O 6
50 p 8.3 x 10 6
10 p 1.0 x 10 9
lp 1.0 x l O 1 2

It should be clear that we need to deal with substantial numbers of

discrete particles when we have particles of 100n or smaller. Note that
lOp particles derived from a one centimeter volume number a billion and
even those of lOOp equal a million in number.

Note that this distribution consists of specific (discrete) sizes of particles.

In NATURE, we always have a continuous distribution particles. This
means that we have all sizes, even those of fractional parentage, i.e.-
18.56|i , 18.57n , 18.58 n , etc. (supposing that we can measure 0.01 (i
differences). The reason for this is that the mechanisms for particle
formation, i.e.- precipitation, nucleation and growth, Ostwald ripening,
sintering, are random processes. Thus, while we may speak of the
"statistical variation of diameters" and we use whole numbers for the
diameters, the actuality is that the real diameters are fractional in nature.

Since the processes are random in nature, we find that the use of
statistics to describe the properties of a population of particles, and the
particle size distribution is well suited for this purpose since it was
originally designed to handle large numbers in a population.

A. The Binomial Theorem-Particle Distributions

To describe particle distributions, we will use nomenclature relating to

234 Measuring particle size and growing single crystals

the science of statistical probability. If we are given n-things where we

choose x variables, taken r at a time, the individual probability of
choosing (1 + x) will be:

4.1.21.- Pi = (1 + x ) n

This is the BINOMIAL THEOREM. By Taylor Expansion, we can find the

total probability, P, for items, n, taken r at a time as:

4.1.22.- P{r) = U U . P r (1 - P ) n " r

{r}

where p r is the individual probability for a particle taken r at a time and

P(r) is the total probability. We define n / r a s the combination of particles
within a given size range and use the combinatorial theory which applies
to statistical mechanics as well to an ensemble of particles:

4.1.22.- {jnj = nCr = n!/r!(n-r)!

{r}

If we now let n approach infinity, we get:

4.1.23.- P(r) = 2 pi = 1 / [ 2 : r n p ( l - p ) ] 1 / 2 • exp (- x 2 / 2np (1-p)

One will immediately recognize this as a form of the BOLTZMANN

equation, or the GAUSSIAN LAW. We can modify this equation and put it
into a form more suitable for our use by making the following definitions.

First, we define a mean (average) size of particles in the distribution as d,

given as:

4.1.24.- d" = 2 n d , /Z n

and then define what we call a "standard deviation" as a, for the

distribution of particles. From statistics, we know that this means that
68% of the particles are being counted (34% on either side of the mean,
 4.1 Measurement ofparticle sizes and shapes 235

which we denote by d. This is shown in the following:

4.1.25.- 2(B"+d a ) =0.68

where da is the diameter measured to give 34% of the total number of

particles. In alike manner:

4.1.26.- 2a = 0.954 = 2 (d" ± d2 o )

3 a = 0.999 = 2 (T ± d 3o )

But note that these standard deviations only apply of we have a Gaussian
distribution. We use these formulas to specify what fraction we have of the
total distribution, or to locate points in the distribution.

By further defining:

4.1.27. B~ = 2(npi/n)

o = ( np (1-p ) l / 2 = std. dev.

r Enp + x = d + x

we obtain (d - d) since x is a deviation from the mean. This then brings us

to the expression for the GAUSSIAN PARTICLE SIZE DISTRIBUTION:

4.1.28.- Pr = 1 exp f- f d - B" ]2 i

(2JT)1/2 a 2a2

if: { - °° < d < + °°}

Let us now examine a Gaussian distribution, as shown in 4.1.29. on the

next page. What we have is the familiar "Bell-Shaped" curve. This
distribution has been variously called:

4.1.30.- Gaussian Log Normal Boltzmann Maxwell - Boltzmann

236 Measuring particle size and growing single crystals

4.1.29.- Gaussian Distribution

We shall use the term "Log Normal" for reasons which will clear later. It
should be now apparent that we use terms borrowed from statistics and a
statistical approach to describe a distribution of particles. The two
disciplines are well suited to each other since statistics is easily capable of
handling large assemblages, and the solid state processes with which we
deal are random growth processes which produce large numbers of
particles.

Let us now give some examples of the log-normal distribution as shown in

the diagram given as 4.1.31. on the next page. In this case, we can
compare the parameters given and realize that log-normal distributions
can look quite similar or quite different, depending upon the values of the
parameters involved.

In Case A, the average particle sizes are all the same. The statistical
spread of particle sizes, i.e.- a, are all equal but the total number of
particles differ in the three cases given. In Case B, the average particle
size is the same as is a and the numbers of particle measured.

b. Measuring Particle Distributions

Our next task is to determine how to measure a PSD. Particle sizes are
generally measured by the parameters given in 4.1.32. (see next page):
 4.1 Measurement of particle sizes and shapes 237

4.1.31.-

4.1.32.- Particle Size Parameters

1. Physical Dimensions
2. Weight
3. Volume

We have already discussed measurement of particle size by the use of

particle dimensions in Section II. where particle shape as relating to
particle size was a major concern. Note that even though the shape of the
particle may be quite elongated, the stated size will be predicated upon a
spheroidal shape. The stated (measured) size of a PSD based upon weight
will depend upon how well the particles were separated into fractions.
For example, if we use a set of sieves to separate particle fractions, the
fractions of particles measured may, or may not, be a function of shape.
Acicular particles will not pass through a given screen the same way as
238 Measuring particle size and growing single crystals

particles with regular features even though the average size of spherical
particles was measured to be equivalent.

For the most part, the measured size of sieved (screened) particles will be
based upon a spheroidal shape since this is easiest to calculate, i.e.- V =
4 / 3 7r r 3 . In the case of measurement by volume, spherical shapes are the
only way to specify the apparent size of the particle. Let us now examine
how we can obtain a PSD using screens to separate particles into
fractions.

Suppose we obtain a sample of beach sand. From 4.1.3., we can see that
the particles are liable to range from about 8p to 2400 \i. In order to
generate a PSD, we must separate the particles. But we find that sieves
are only available in certain sizes, as also shown in 4.1.3. These are:

4.1.33.-COMPLETE LISTING OF SIEVE SIZES AVAILABLE-(U.S. STD.)

Sieve No. Microns Sieve No. Microns

3.5 5660 ± 3 % 35 500 + 5%

4 4760+ 3 40 420 ± 5
5 4000 ± 3 45 350 + 5
6 3360 + 3 50 297 + 5
7 2830 ± 3 60 250 ± 5
8 2380 ± 3 70 210 ± 5
10 2000 ± 3 80 177 ± 6
12 1680+ 3 100 149 ± 6
14 1410+ 3 170 88 ± 6
16 1190± 3 200 74 + 7
18 1000 + 5 230 62 ± 7
20 840 ± 5 270 53 ± 7
25 710 + 5 325 44 + 7
30 590 ± 5 400 37 + 7

There are two types of standard screens, the U.S. and the Tyler standard
screens. We have given the U.S. screen values. Those of the Tyler are very
similar, e.g. - #5 Tyler = 3962p , # 20 = 833u, and #400 = 38p. To screen
 4.1 Measurement of particle sizes and shapes 239

our sand sample, we choose to use the following US sieves: #10, # 18,
#20, #30, #40, #50, #80, #100, #170, #200, #270, #325 & #400. Now if
we screen the sand with the #10 screen, all particles larger than 2000 ±
60 p will be retained upon the screen. Now we resieve the part that
passed through the #10 screen, using the #18 screen, to obtain a fraction
which is > 1000 p ± 50p. We then repeat this procedure to obtain a series
of fractions. This is shown as follows:

4.1.34.- Fractions Obtained Mean diameter d

f0 > 2000 p ?
lOOOp < fi < 2000 p 1500 p
840p < f2 < 1000 p 920
590p < f3 < 840 p 715
420p < f4 < 590 n 505
297p < f5 < 420 n 358
177p < f6 < 297 p 237
149p < f7 < 177 p 163
88 p < f8 < 149 p 118
77 p < fg < 88 p 83
53 p < fiQ < 77 p 68
44 p < fi i < 53 p 49
37 p < f 12 < 44 p 41
fl3 < 3 7 p ?

We now have 12 fractions that we can use. The mean diameter is apparent
from the screens used to separate the fractions. In one case, d 12 = 41 ±
2.8 p . We can now weigh each fraction and calculate the % weight in
each fraction, providing we use the total weight of all 14 fractions.
Knowing the density of the sand, and assuming the particles to be
spheres, we can then obtain the number of particles in each fraction.
(Note that we have assumed that all particles small enough to pass
through a given screen has done so. In many cases, this is not true, and
we have to be cognizant of this error - this factor is the main source of
error in the SIEVE METHOD of particle size analysis). There are two
general methods for data-reporting, namely:
240 Measuring particle size and growing single crystals

4.1.35.- METHOD I = % of total weight

METHOD II = Cumulative weight-%

In Method I, we can use the average particle size between the extremes of
the sieve-sizes or we can plot a bar-graph showing the spread of the
individual fractions. In Method II, we add each fraction to the next and
then "normalize" the total to 100%. Each method has its advantages.

In Method I, we calculate the total weight and assign each fraction a %-

value. Mean diameters , d, are easy to obtain in the first method because
we know the screen diameters used to separate the fractions. In the
second, we add f2 to f i , then f3 to f2 + f 1 . then f4 to f3 + f2 + f 1,

etc. To obtain d", we take the average of the added fractions, as follows:

4.1.36.- d"
fl 1500 n
f2 + fl 1210
f3 +f2 + fl 963
U + f3 + f2 + fl 734

We continue with all the fractions we have. We can now plot the data as:

4.1.37.-
 4.1 Measurement of particle sizes and shapes 241

We can see the spread of particles in Method I, but to determine the

mean, d", is a guess. Method II allows us to obtain the mean from the 50%
point, that is, the point where 50% of the particles are smaller than, and
50% are larger than. Actually, the data of Method I can also be plotted as
shown in the following:

4.1.38.-

Methods I , II and III are called "Frequency Plot", "Cumulative Frequency

Plot" , and "Histogram" respectively.

What this means is that the particle data in Method I is plotted to show
the percentage of particles in each particle range. By connecting the dots
in the curve, we get an estimate of those particles which we did not
directly measure. But, getting the average particle size is difficult.

Method II is more direct since we get d and a directly. The data in

Method III, called a "Histogram Plot", have been plotted using a particle
spread that is equal from fraction to fraction. If we use the actual spread
of particles in each range generated by the use of selected sieves, we get
the following diagram shown in 4.1.39. on the next page.

Note that our sand is composed of rather large particles but, more
importantly, when we use screens, the particle spread is very large at
large particle sizes and very narrow at the smallest size. This illustrates
one major factor that we encounter when trying to depict a true particle
242 Measuring particle size and growing single crystals

4.1.39.-

size distribution. Thus, the use of screens for particle size determinations
has seen limited usage.

There is one important point which needs to be emphasized, that is:

"ALL METHODS FOR PRESENTING DATA FROM THE MEASUREMENT

OF PARTICLE SIZE DISTRIBUTIONS, WHETHER INSTRUMENTAL,
SIEVING, SEDIMENTATION, OR PHOTOMETRIC METHODS. MEASURE
 4.1 Measurement of particle sizes and shapes 243

FRACTIONS OF THE TOTAL PARTICLE DISTRIBUTION. IF THE METHOD

IS SENSITIVE, THE FRACTION-SEGMENTS CAN BE SMALL, AND THE
MEASURED PARTICLE DISTRIBUTION WILL BE CLOSE TO THE ACTUAL
ONE. IF THE MEASUREMENT IS LESS SENSITIVE, THERE MAY BE
SIGNIFICANT DEVIATIONS FROM THE CORRECT PSD".

The data for the fraction-steps can be in terms of numbers of particles,

weight, %-weight or even a packed volume. It should be noted here that
the most frequently used method for reporting particle size distributions
is Method II.

IV. Analysis of Particle Size Distribution Parameters

Each of the methods of data presentation has its own special problems in
terms of the amount of extractable information one can obtain. It is
mandatory that we use a method of measurement in which the particle
size ranges are uniform. Otherwise, we obtain the results shown in
4.1.39. Fortunately, most of the PSD instrumentation in use today employ
computer-generated data to display the particle size. Furthermore, the
particle-range spread be measured can be adjusted to suit the material at
hand. Thus, we can use the histogram if we so desire, especially if it has
certain advantages over the other methods of displaying PSD.

a. The Histogram

The following diagram, given as 4.1.40. on the next page, shows atypical
histogram, along with an analysis of the parameters which can be
calculated. As shown in this diagram, the first step is to calculate the
individual areas of the steps, Ai. The area is obtained by multiplying the
difference of measured diameters, d2 and dj by the %. These are then
summed and then divided by the number of individual areas, and then
multiplied by the sum of (d2 - dx) / ni to find the mean, cf. Admittedly,
the method is cumbersome, but sometimes this is all the data we have.
244 Measuring particle size and growing single crystals

4.1.40.-

B. Frequency Plots

Many particle-measuring methods use Stake's Law to determine particle

distributions. By suitable manipulation (see below), we can obtain an
equation relating the Stokes diameter, M, with the particle density, rx ,
and the liquid density, r 2 . Stokes Law is given as follows:

4.1.41.- M = 18 rjh / a t (5 1 - r 2 ) g

where r\ is the liquid viscosity, h the distance the particles settle, a is a

shape factor (a = 1 for a spherical particle), and g is the gravitational
constant. There are at least two methods of using Stokes Law to obtain
data concerning particles sizes. One is to continuously measure the
weight of the particles as they settle through the liquid medium. One
method used in the past employed an analytical balance in which one pan
was set in the bottom of the container and the cumulative weight was read
at intervals corresponding to selected particle sizes (the times of which
were calculated from Stokes Law). Although this method was popular 50
years ago, it is rarely used today.

The photometric method gives % at a specific particle diameter and

hence can be directly plotted as frequency, as presented in 4.1.42. on the
next page. As shown, d o can be calculated from the mean diameter if the
distribution is symmetrical, but they rarely are.
 4.1 Measurement of particle sizes and shapes 245

4.1.42.-

Note also that a, which is defined as the diameter limits equal to 68% of
the total particle population, cannot be obtained, only d o .

c. Cumulative Frequency

The most popular method of data presentation is that of cumulative

frequency, as follows:

4.1.43.-

Note that both d and s are easily obtained. But no other information can
be derived. What we need is a more versatile method. Fortunately, this is
available but the method requires that the particle data be fixed so that
246 Measuring particle size and growing single crystals

the particle-ranges measured are uniform. This method plots uniform

particle ranges as a function of their conformance to a log probability or
Gaussian distribution curve.

d. Log Normal Probability Method

The most versatile method for displaying and analyzing particle size
distributions is by plotting them as a log normal distribution. What this
means is that we use log normal probability paper to plot the data. It is
this method that we shall investigate in detail since it produces
information about the distribution not obtainable by any other method. In
fact, one can distinguish between natural and artificial PSD's and make
inferences concerning the origin of the PSD. Also, we can easily
distinguish the bimodal case , i.e.- two PSD's present at the same time.
This is impossible with other methods of data presentation.

Consider the log normal (Gaussian) distribution:

4.1.44.-

Here, we have shown the spread of particles in such a distribution. Our

approach is to assume that all PSD's have an origin in growth that tends to
produce a log normal population. This is reasonable since nearly all, if not
all, particle growth mechanisms are random in nature. Hence, we expect
to see a log normal PSD as the usual case. What we then do is to look for
 4.1 Measurement of particle sizes and shapes 247

deviations from log normality. It is these which supply additional

information about the PSD. To plot such a distribution, we use log
probability paper. An sample is given in the following diagram:

4.1.45.

The diameter of the particles is located on the y-axis and is plotted at the
proper point on the x-axis of the chart as "% less than" (see Method II of
4.1.16. and 4.1.21, given above).

A truly log normal distribution will give a straight line when plotted on
this type of paper. This means that the PSD is not limited, i.e.- all sizes of
particles are present from - °° to +°° .

However, if the PSD is growth-limited, it will readily apparent from the

graph. Ostwald ripening, WHERE LARGE PARTICLES GROW AT THE
248 Measuring particle size and growing single crystals

EXPENSE OF SMALL ONES, is one mechanism that gives rise to growth

limits. Growth limits are defined in the following:

4.1.47.-
^ - Upper Growth Limit
U
• ^ - Lower Growth Limit
Acx, - Upper Discontinuous Limit
3£0 - Lower Discontinuous Limit

The discontinuous limit is that in which all of the particles beyond a

specific size have been removed or do not exist.

e. Types of Log Normal Particle Distributions

We shall present some examples of particle distributions and how to

interpret them.
1. UNLIMITED PARTICLE DISTRIBUTIONS

The following diagram shows a typical PSD where the distribution does
not have limits:

4.1.48.-
 4.1 Measurement of particle sizes and shapes 249

This is the type of distribution usually found as a result of most

precipitation processes. Following that are distributions that do have
limits to the "normal" distribution. What this means is that the
distributions do conform to the limits defined in 4.1.27.

2. LIMITED PARTICLE DISTRIBUTIONS

If a precipitate is allowed to undergo Ostwald ripening, or is sintered, or

is caused to enter into a solid state reaction of some kind, it will often
develop into a distribution which has a size limit to its growth. That is,
there is a maximum, or minimum limit (and sometimes both) which the
particle distribution approaches. The distribution remains continuous as it
approaches that limit. The log-probability plot then has the form shown as
follows:

4.1.49.-

That is, there is a maximum, or minimum limit (and sometimes both)

which the particle distribution approaches. The distribution remains
continuous as it approaches that limit. In this case, both upper and lower
limits are shown. However, one may have one or the other, or both.
Ostwald ripening tends to use up all of the small particles, without
limiting the upper size. Then we would have the lower limit but not the
upper.
250 Measuring particle size and growing single crystals

However, there are cases where discontinuous limits apply. This is shown
in the following as "Discontinuous Limits".

3. Log Normal Distributions- Discontinuous Limits

Suppose we use a 200-mesh screen to remove particles larger than 74n

and a 400-mesh screen to remove those smaller than 37\\. The PSD
would now look as follows:

4.1.50.-

Note that the distribution is linear between 37 and 74 n but that it

abruptly shifts to ± °° at these points. This is a particle distribution for
which it is impossible to obtain an accurate picture by any other means.
A frequency plot of the same data is shown in the following diagram,
given as 4.1.51. on the next page.

If we were using this method to present data and got this curve, we would
think that the curve was just asymmetrical and that the distribution was
not log-normal. Yet, it is obvious from 4.1.50. that it is log-normal.
 4.1 Measurement of particle sizes and shapes 251

4.1.51.-

4. Multiple Particle Distributions

Log probability plots are particularly useful when the distribution is

bimodal, that is, when two separate distributions are present. Suppose we
have a distribution of very small particles, say in suspension in its mother
liquor. By an Ostwald ripening mechanism, the small particles redissolve
and reprecipitate to form a distribution of larger particles. This would
give us the case shown as follows:

4,1.52.-
252 Measuring particle size and growing single crystals

In this illustration, the size distribution parameters of both distributions

are readily apparent. A similar determination is almost impossible with
other types of data presentation methods. Although a large discontinuous
gap between the two distributions is shown in 4.1.52., it is rarely the
case.
f. Using Particle Distribution Parameters in Industry

As examples of the usefulness of particle distributions, consider the

following. Two actual cases encountered in Industry are described in the
following examples:

1. FLUORESCENT LAMP PHOSPHOR PARTICLES

Fluorescent lamps are manufactured by squirting a suspension of

phosphor particles in an ethyl cellulose lacquer upon the inner surface of
a vertical glass tube. Once the lacquer drains off, a film of particles is
formed. The lacquer is then burned off, leaving a layer of phosphor
particles. Electrodes are sealed on; the tube is evacuated; Hg and inert gas
is added; and the lamp ends are added to finish the lamp. Lamp
brightness and lifetime are dependent upon the particle size distribution
of the phosphor particles. The number of small particles is critical since
they are low in brightness output but high in light scattering and
absorption. A lamp brightness improvement of 10% is easily achieved by
removal of these particles. In 4.1.53. (given on the next page), we show
the PSD that resulted when a certain mechanical method was used to
remove the "fines".

In this application showing how the log-normal plot can be utilized, we

can readily see that the "mechanical" separation of "fines" has created a
new particle distribution with d 2 = 2 n. Even the value of o2 differs from
that of the major particle distribution. In the "fines" fraction, it appears
that the largest particle does not exceed about 5 p. Needless to say, lamps
prepared from this phosphor were inferior in brightness. Armed with this
information, one could then recommend that the method of "fines"
removal be changed.
 4.1 Measurement of particle sizes and shapes 253

4.1.53.-

2. TUNGSTEN METAL POWDER

The lifetime of a tungsten filament in an incandescent light bulb depends

a great deal on the grain size within the wire used to make the filament.
W-metal powder is pressed into a bar as close as possible to theoretical
density. The bar is sintered in an inert atmosphere and then "swaged"
into a long rod. This rod is then drawn into a fine wire (~ 2- 5 mil =
0.002-0.005") which is wound into the filament form. Metal powder
particle size has a major effect upon wire quality since fine particles form
small grains in the wire while large particles will form large grains.

When the filament is operated at 3250 °K to produce light, the wire is hot
enough that vacancy migration will occur in the wire. Simultaneously,
grain growth and an increase in grain size also occurs. The grain
boundaries between large grains will "pin" vacancies, but grain boundaries
between small grains do not appear to do so. If they do, the effect is an
order of magnitude smaller. As the vacancy concentration builds up at the
large grain boundaries (see the Kirchendall effect given in Chapter 3),
local electrical resistance increases. Eventually, a "hot spot" develops and
the tungsten wire melts at that point, with failure of the filament and the
lamp. If only small metal particles are used to make the wire, then the
small grains produced must grow into large grains before "hot spots"
254 Measuring particle size and growing single crystals

cause its eventual failure . Therefore, incandescent lamp manufacturers

exercise very close control of the tungsten-metal powder PSD and the
sintering processes (pore elimination and actual density) used to produce
the wire.

A typical PSD for a tungsten metal powder where the large particles have
been removed is shown as follows. In this case, the PSD is suitable and
particles larger than about 5 n appear to have successfully removed.

4.1.54.-

Note that we can easily determine the spread of the particles as well as
the average size. It is these two parameters which are most important in
the selection (and modification) of tungsten metal powders to be used to
manufacture filaments used in all electronic equipment. Let us now
examine a method of calculating a PSD.

V. A Typical PSD Calculation

Let us suppose that we have a particle counting instrument which sorts

and counts the number of particles at a given particle size. There are
several commercial instruments available including the "Coulter Counter"
and those based on "Laser-Refractometry". We will use one of these to
obtain the following particle data, given as follows:
 4.1 Measurement of particle sizes and shapes 255

4.1.55. - EXPERIMENTAL COUNTS AT PARTICLE DIAMETER

d_ n Counts d n Counts
2.2 n 2 10 ix 902
3.0 4 15 1402
5.0 28 25 1902
7.0 232 30 1982
9.0 602 40 1998
50 2002

Having listed these data, our next step is to calculate the average size in
each particle interval.
TABLE 4 - 1
ATYPICAL PARTICLE SIZE DISTRIBUTION CALCULATION
Diameter n-Counts An Ave. d %. 2%
2.2 \i 0
2 2.6 fi 0.1 0.1
4.0 2
24 4.5 1.2 1.3
5.0 26
206 6.0 10.2 11.5
7.0 232
370 8.0 18.5 30.0
9.0 602
300 9.5 15.0 45.0
10 902
500 12.5 25.0 70.0
15 1402
500 20.0 25.0 95.0
25 1902
80 27.5 4.0 99.0
30 1982
18 35.0 0.9 99.9
40 2000
256 Measuring particle size and growing single crystals

For particles less than 2.2 p, we have 2 particles. At 3.0 p, we have 4

particles. Therefore, in the range 2.2 - 3.0 p, we have two (2) particles
whose average size is 2.6 p. In the range 3.0 - 5.0 p, we have four (4)
particles whose average size is 4.0 p. We have a total of 2000 particles that
we have measured and we can calculate the percent (%) of total counts for
each size range.

We therefore continue to calculate An as shown in the above table. Note

that the same methods of arranging size data are used regardless of
whether we use screens to separate the particles into size fractions or
whether we use PSD instrumentation to do so. This gives the following
log-normal PSD plot, as shown in the following:

4.1.56.-

Here, the parameters of the distribution are given. This material appears
to have been precipitated, or it may have been obtained by a grinding
process, since the parameters indicate a Gaussian unlimited distribution.
Particles grown by solid state reaction, including Ostwald ripening,
generally have PSD characteristics of the original PSD from which they
were formed. That is, there will be limits according to the original
distribution from which it has arisen.

If the original PSD had limits, then the progeny PSD will also have such
 4.1 Measurement of particle sizes and shapes 257

limits. The basis for this behavior lies in the fact that if one starts with a
certain size range of particles as the basis of particle reaction and growth,
one will end with the same size range of particles in the PSD of the
particles produced by the solid state reaction. Such a case is shown in the
following:

4.1.57.-

In this case, the oxalate was prepared by addition of oxalic acid to a

Gd(NO3) solution to form a precipitate. There is an upper limit of about
23 \i in size. This may be due to the fact that the precipitation was
accomplished at 90 °C, or from the fact that rare earth oxalates tend to
form very small particles during precipitation which then grow via
Ostwald ripening and agglomeration to form larger ones. Nevertheless, it
is clearly evident that when the oxalate is heated at elevated temperature
(~ 900 °C), the oxide produced retains the PSD characteristics of the
original precipitate.

VI. Methods of Measuring Particle Distributions

There are four (4) primary methods used to obtain data concerning
particle size distributions. These are shown on the next page:
258 Measuring particle size and growing single crystals

SEDIMENTATION
ELECTRICAL
OPTICAL
ABSORPTION AND PERMEATION

We have already covered the microscopic method of determining particle

size. We will now examine other methods in more detail, including the
sedimentation method, in order to show that some procedures allow one
to obtain information about particles not easily obtained by any other
method. For example, the BET method gives an average particle size but,
more importantly, allows one to determine the porosity of the particles as
well.
A. Sedimentation Methods

There are several particle sizing methods, all based upon sedimentation
and Stokes Law. If a particle is suspended in a fluid (which may be gas, or
any liquid), the force of resistance to movement by the particle will be
proportional to the particle's velocity, v, and its radius, r, namely -

4.1.58.- f = 67rr TI v

where TI is the viscosity of the fluid, providing the particle is spherical. If

the particle settles under the influence of gravity, then we can write:

4.1.59.- f = 4/3/7rr3 (p s - pi ) g = 6JT r r\ v

where p s and Pi are the densities of the solid particle and the liquid,
respectively. Since the distance settled, D, is: h = v t, then:
1/2
4.1.60.- D= { 18 h / a t (r s - r i ) g}

This is Stokes Law for sedimentation where we have added a , a shape

factor, just in case we do not have spherical particles. You will note that
we have already given a formula for Stokes Law in 4.1.42., given above.
This equation is more suited for determination of particle size using a
sedimentation method.
 4.1 Measurement of particle sizes and shapes 259

One way to measure particles is to weigh them as they settle out from
suspension. Such an apparatus is called a "sedimentation balance" and is
designed as shown in the following:

4.1.61.-

The weight of the particles builds up with time and is proportional to 1 / d.

If we assume spherical particles, then we can convert the above curve to
particle diameter from Stokes Law. Although we have added the particle
suspension to a "water cushion" as shown above, it might not seem that
the settling of the particles would strictly adhere to Stokes Law, which
assumes the terminal velocity to be constant.

But, as shown in Table 4-1, given on the next page, the approximate
distance a particle needs to travel to reach terminal velocity in a given
liquid is very short.

These data show that the particles separate very fast due primarily to
their differing terminal velocities. Thus, in the graph shown as 4.1.61.,
the initial weight is due to the large particles, while the upper part of the
curve is due to the smaller sizes. For particles less than about 2.0
microns, settling times under gravity becomes extremely long and this
size remains a lower limit for sedimentation methods of determining
260 Measuring particle size and growing single crystals

Table 4 - 1
Rate of Fall in Water at 25 °C. for Particles Having a Specific Gravity of 2.0
Size (microns) Approximate Distance in cm. Approximate
Traveled prior to Terminal
Reaching Velocity
Terminal Velocity (cm/sec)
2
2000 1.2 x 10 240
200 1.2xlO- 2 2.4
6
20 1.2 x l O " 2.4xlO-2
10
2 1.2X10' 2.4xlO-4
0.2 1.2 x l O " 1 4 2.4xlO-6
16
0.02 1.2 x l O " 2.4xlO-8

particle size. 2000 p is probably the upper limit since one cannot adjust
the balance to operate quickly enough before these large particles are lost.
This discussion has been presented so that the factors involved in
separation of particles by sedimentation can be comprehended and
perhaps used in other contexts.

Another sedimentation method used is the so-called MSA-analyzer. If the

value of "g" in 4.1.40. is increased (such as the use of a centrifuge) one
can analyze the very small particles in any given distribution in a short
time. The problem of course lies in accurate determination of the weight
accumulated at a given time under a specific centripetal force.
Nevertheless, this method was used for a limited time as an improvement
to the sedimentation balance for determining the diameters of small
particles.

Sedimentation methods for determining particle size distributions are no

longer used since the advent of the computer and the laser. It is much
easier to obtain a PSD using these instruments since the machine does all
of the fundamental work.

Another sedimentation method involves the ANDREASEN PIPETTE. This

glass apparatus is inexpensive and consists of a bottle having an internal
sampling tube and calibrated sampling volume (5 ml). One draws a sample
 4.1 Measurement of particle sizes and shapes 261

and then expels it into an external 5 ml beaker. One then dries each
sample and weighs it. This is shown as follows:
4.1.62.-

We will not dwell further on this method since electrical conductivity,

i.e.- Coulter Counter and optical methods, i.e.- Laser-Diffraction have
totally supplanted all of the sedimentation methods for determining
particle sizes.

b. Electrical Resistivity- The Coulter Counter

For quite a while, the most useful method for determining particle
distributions was that of electrical conductivity. The most widely used
instrument is the Coulter Counter (named after the Inventors), although
there are now other similar instruments on the market. It has largely
been supplanted by the Laser-Diffractometer. Originally, the Coulter-
Counter was designed to measure blood corpuscles which are 2-8 p in
size. It has proven to be very suited to measure micron and submicron
particles easily, accurately, and reproducibly. Consider a conductive
solution consisting of water with a soluble salt, i.e.- 1% NaCl, and a
262 Measuring particle size and growing single crystals

dispersing agent used to prevent agglomeration of particles in suspension.

Two electrodes are placed in solution with a non-conducting orifice
between them. If the solution, containing particles in suspension, is
pumped through the orifice, the change in volume, i.e.- electrical
conductivity, produces an electrical pulse as each particle passes through.
The basic design of the Coulter-Counter is shown in the following:

4.1.63.-

A ceramic block such as alumina is generally used because of its chemical

inertness and the orifice is bored or drilled to precise dimensions. The
block is placed so that two (2) chambers result. Then if a DC voltage is
applied across the electrodes, a current will flow only through the orifice,
and an effective resistance arises which depends upon the voltage applied
and the size (volume of the conducting solution) of the orifice. Particles
are added to one side of the two chambers created by the ceramic block,
and the suspension is pumped through to the other side.

There will be an electrical pulse as each particle passes through the

orifice. Each pulse will be proportional to the size of the particle because
the particle volume displaces part of the conducting solution during its
passage through the orifice. In we have properly established steady state
conditions, we will obtain a drop in current as each particle passes
through Electronically, this can be converted to a positive pulse whose
intensity is proportional to each particle volume. We assume sphericity of
the particles. If V is constant, then:

4.1.64.- I = V / A R and: AR = r 0 V p / A2 • 1/{1/1- (r o /r e )} -a/A

 4.1 Measurement of particle sizes and shapes 263

where r o is the actual particle radius, r e is the effective radius (in case
the particle is not spheroid), Vp is the volume of the particle, A is the
cross-sectional area of the orifice, and a is a constant which depends upon
the solvent and solute used to make the conducting solution.

The size of the orifice must be fitted to the range of particles present. If
the particles are too large, they will not pass through the orifice. Too
small and the pulse is not easily detected, Since large numbers of
particles can be present, the powder content must be controlled. Too
high a particle concentration-density will result in "coincidence
counting". That is, two small particles can be counted as one larger one.
This phenomena has been addressed and tables are provided to correct
for incidence counting. The electronic part of the Coulter Counter has
several hundred "channels" to accept pulses. These are then displayed via
an oscilloscope a s :

4.1.65.-

The "threshold" is electrical, and movable, so that one counts the large
particles first (large pulses) and then moves the threshold lower to count
the smaller particles. The threshold is actually a pulse height analyzer
which converts heights to numbers. The data obtained are easily
converted to a frequency plot or a cumulative plot. Since AR response is
264 Measuring particle size and growing single crystals

linear, one can readily calibrate the instrument at one point, using
monosized particles such as pollen or specially prepared plastic particles.
The only requirement is that the calibrating particles ought to be in the
range of the particles to be measured. The Coulter Counter is particularly
useful and unique in that steps as small as 1.0 micron may be used if so
desired.

The latest Coulter Counter instruments incorporate a microprocessor and

the operator can specify the form of the output data to be obtained. The
instrument converts count-data to practically any format desired, with the
exception of log-probability plots.

c - Laser Diffractometrv

Particle Size by Laser Refractometry is based upon Mie scattering of

particles in a liquid medium. There are two primary methods, earlier
ones based on sedimentation and later ones upon scattering of light
(refractometry). Both employ Mie scattering. The earlier instruments
were dependent upon settling time of the particles in the liquid medium.
The later instruments utilize an instantaneous detection of refracted light,
determined by the refractive index of the particle (and its size) and the
liquid medium employed. Up until about 1985, the power of computers
supplied with laser diffraction instruments was not sufficient to utilize the
rigorous solution for homogeneous spherical particles formulated by
Gustave Mie in 1908. Laser particle instrument manufacturers therefore
used approximations conceived by Fraunhofer.

The hypotheses made in Mie Theory include:

1. The particle is an optically homogeneous smooth sphere whose

real and imaginary refractive indices are both known.
2. The spherical particle is illuminated by a plane wave of infinite
extent and of defined wavelength.
3. The real and imaginary refractive indices of the medium
surrounding the particle are both known.
 4.1 Measurement of particle sizes and shapes 265

The Fraunhofer approximation includes:

1. The incident light-wave is plane, of infinite extent and of known

wavelength.
2. The scattering is from a circular aperture in a thin opaque
screen.
3. The extinction coefficient for all sizes is 2.0

Fraunhofer rules do not include the influence of refraction, reflection,

polarization and other optical effects. Early laser particle analyzers used
Fraunhofer approximations because the computers of that time could not
handle the storage and memory requirements of the Mie method.

For example, it has been found that the Fraunhofer-based instrumentation

cannot be used to measure the particle size of a suspension of lactose (R.I.
= 1.533) in iso-octane (R.I. = 1.391) because the relative refractive index
is 1.10. i.e.- 1.533/1.391. This is due to the fact that diffraction of light
passing through the particles is nearly the same as that passing around
the particles, creating a combined interference pattern which is not
indicative of the true scattering in the far field where the detector is
located. The Mie solution anticipates this.

A laser is required to provide a single wavelength so that photons passing

through or around any given particle is diffracted and scattered, free from
any other optical interference. The Mie theory is rigorous and is used to
predict the scattering from particles whose size range from Angstroms to
centimeters.

Mie originally devised his theory to better define the properties of "fog",
i.e.- a dispersion of very small circular drops of water in the atmosphere.
It has now been adapted to particle counting and particle distributions.
Mie theory requires that the real and imaginary refractive be known or be
measured. This may require additional work initially. Experimentation
may be required to choose the proper liquid medium and/or dispersing
agents initially. This may be especially true if one is trying to determine
the PSD of pharmaceutical agents such as new drugs. Laser diffraction
266 Measuring particle size and growing single crystals

particle size instruments have now become one of the major procedures
for particle size determination, particularly for the Pharmaceutical
Industry, i.e.- organic based particles.

d. Other Methods of Measuring Particle Size

Permeability is one other method for obtaining information about particle

diameters. If one packs a tube with a weight of powder exactly equal to its
density, and applies a calibrated gas pressure through the tube, the
pressure-drop can be equated to an average particle size. The instrument
based on this principle is called the "Fisher Sub-Sieve Sizer™". Only one
value can be obtained but the method is fast and reproducible. The
instrument itself is not expensive and the method can be applied to
quality control problems of powders. Permeametry is useful in the particle
range of 0.5 to 50 p.

Gas adsorption is one other method sometimes used for determining

average particle size. In this case, one is usually interested in the surface
area of a powder and calculates the average size of the particles
secondarily. The method is called the BET-method after its developers,
Brunauer, Emmett and Teller. The procedure is time-consuming and is
accomplished as follows. A gas analysis train is used in which gas volumes
can be recycled and measured very accurately. A weighed sample is placed
in the sample tube and allowed to come to equilibrium. Since all solid
materials have a monolayer of water on the surface of the particles (we live
in a wet world), the sample is heated to 300 °C. to expel the water, in a
nitrogen gas flow. Next, it is cooled to liquid nitrogen temperature. At
this point, the sample will adsorb a monolayer of N2-gas molecules on the
surface of each particle. By allowing the temperature of the sample to
return to room temperature, while measuring the volume of nitrogen gas
released, one obtains a value which can be converted to a surface area of
the particles being measured. In practice, one recycles the system several
times, measuring adsorption and desorption volumes successively so as to
obtain an average value. It is found that the unoccupied fraction of the
surface (1- y) approaches a constant value where it is assumed that y = 0.
Nitrogen gas is essential since it is the only gas which forms monolayers
 4.1 Measurement of particle sizes and shapes 267

easily without the tendency to form more than one layer of gas molecules
on the surface. If the gas pressure is kept between strict limits, the BET
equation is:

4.1.66.- pg / (Vg [ 1- pg ] = 1/ (V m o n o • C) + (C - 1) pg / (V monO • C )

where: C = exp { (FAds. - FCond.) / RT} , and pg is the gas pressure. Vg

is the gas volume measured; V m o n o is the gas volume of the monolayer
calculated from the known dimensions of the nitrogen-gas molecules;
FAds. is the Gibbs free energy of adsorption, and Fcond. is the free energy
of condensation, pg is measured as the ratio of the equilibrium pressure
at specific temperatures, i.e.- room and liquid nitrogen temperatures.

It has been found that a plot of pg / {Vg [1-pg] vs: pg is linear for the
pressure range of 0.05 to 0.4, with aslope of (C - 1) / (V m o n o • C ) and
intercept of 1/ (V m o n o • C ). Let us now do a simple calculation using
BET data obtained.

Suppose we have a 20.00 gm. sample having a density of 2.0. We measure

the surface area as 6.0 m^ . From the area of a sphere, A = JT D^ , and the
volume of a sphere, V = 4 / 3 n D^ >, we find the total volume of n spheres
to be 10 cc, i.e.- n{4/3 it D^} = 10. The surface area of n{7iD2} spheres is
6 m3. And n, the total number of spheres present is the same in both
formulas. Therefore, by substitution, we find D= 10 p.

If we obtain a particle diameter by some other method and find that it is

much smaller than that of the BET method, we infer that the particles are
porous. We thus speak of the porosity and need to correct for the pore
surface area if we are to make a reasonable estimate of the true diameter
by the BET method. Note that the BET method is the only method
whereby one can obtain an estimate of the porosity of any given set of
particles.

We are now ready to undertake the art and science of growing a single
crystal. A great many methods are available and the choice one makes will
268 Measuring particle size and growing single crystals

depend upon the physical properties of the material which we use to grow
a single crystal.

4.2.- GROWTH OF SINGLE CRYSTALS

We will now describe the growth of single crystals. That is, how single
crystals are obtained. It will become apparent that most single crystals are
obtained from a melt of the material desired. Although many melt-growth
methods were proposed and tried in the past, only a few have survived to
the present and are in use.

Many of the devices that we use on a daily basis employ a single crystal
element as the heart of the mechanism. A good example is the "quartz-
crystal" watch that you wear on your wrist. Here, a quartz crystal is made
to vibrate under an applied voltage. Its vibrations are coupled to a sensing
circuit and translated into seconds, minutes and hours by using and
counting the known resonant frequency of vibration of the crystal. The
resonant vibrational frequency is determined by the angle of "cut" (or
angle in relation to the crystallographic axes) and the size of the crystal.
There are many other sensing devices based upon a crystal component
such as a device which controls heating, translates force into electrical
voltage, modulates light, or regulates motion. We use them as electrical
heaters, strain gauges, laser controllers or piezoelectric devices.

Obviously, the major difference in the single-crystal and polycrystalline

(crystallite) state is a matter of size. For the single-crystal, the size is large
(> 10 cm), whereas in the polycrystalline state, the size of the crystals is
small (10 nm = 0.001 cm.) The methods for obtaining one or the other
differ considerably. They include formation and growth from:

4.2.1- Single Crystal Crystallites fPowders)

Liquid solvent Vapor

Vapor Melt
Melt Flux
Molten salt Precipitation from solution
 4.2 Growth of single crystals 269

If we are to grow a single crystal, we will need some sort of furnace,

which is simply a closed space, heated by electrical elements wherein the
internal temperature is controlled.

I. Heating Elements and Crucibles

Before we consider crystal growth in detail, let us examine the hardware

needed to obtain a melt. Many of the furnace components available have
temperature limitations, and many of the crystals we might wish to grow
have high melting points, i.e.- >1600 °C. Thus, we may have to choose
certain combinations in order to get a furnace which will adequate for the
task at hand.

There are many ways to build a furnace. Basically, a furnace consists of a

few essential parts, each of which can be varied according to the final
operating requirements needed for the furnace. The following shows the
essentials required for the proper design of a furnace, as shown in 4.2.2.
presented on the next page:

The power source, relay, set-point controller and sensor need not be
discussed here. However, the insulation, heating elements, and crucibles
involve materials, and need to be examined in more detail.

4.2.2.- Elements of Furnace Design

a. A power source,
b. Heating elements
c. Thermal insulation,
d. A crucible to hold the melt
e. A sensor for temperature-feedback
f. A temperature set-point controller
g. A relay or other power-control device for the power source.

Heating elements can be metal wires or ceramic rods which become

incandescent when an electrical current flows through them. Their
compositions are critical and have been developed over many years to
optimize their performance as heaters. Insulation is used to retain the
270 Measuring particle size and growing single crystals

heat generated and to disperse the heat uniformly throughout the heated
space, namely the internal cavity of the furnace. Table 4-2 summarizes the
thermal properties of Insulation and Heating Elements for furnace-
construction.
TABLE 4-2
INSULATION HEATING ELEMENTS
Max. Temp. In Air Max. Power Needed
Usable Temo. Voltage Current
Glass Wool 600 °C. Nichrome Wire 900 °C. med. med.
Fiberfrax 1350 "Globars"™(SiC) 1475 low high
Quartz wool 1100 Kanthal Wire 1300 med. med.
Fire Brick 1100 to 1650°C MoSi2 1700 low high
Alumina(foam& beads) 1850°C Pt (40% Rh) 1800 low high
Zirconia (ZrO2) 2400 ZrO2 :Y 1900 low high
Magnesia 2800 LaCrO3 1900 low high

In furnace insulation, Fiberfrax™ (fibers of aluminum silicate) and "fire-

brick" (bricks made from insulating silicate compounds) are the two most
commonly used materials. For very high temperature work, Zirconia ( =
ZrO2) in the form of beads, "wool", and boards pressed from fibers are
used for temperatures above 1600 °C. Alumina, i.e.- AI2O3 , is much
cheaper, but does not have the thermal shock, or very high temperature
capability (i.e.- > 2000 °C.) of Zirconia. Those heating elements given in
Table 4-2 are generally used in air atmosphere, up to about 1800 °C. If
one needs to produce a melt above 1800 °C, it is necessary to use
refractory metals which must be used in an inert atmosphere. The
following lists some of these heating elements:
Table 4-3
Furnace Elements For NON- ATR Usage Only
Voltage Current
Mo or W wire 2400 to 2800°C low high
Iridium wire 2400 low high
Graphite 3400 low high
R.F. Current 2800 low high
Oxy-hydrogen flame 4000 NA NA
 4.2 Growth of single crystals 271

The temperatures given here are approximate, and are presented solely
for comparison purposes.

Wire-wound heating elements are used most frequently. They are usable
in air up to about 1200 °C. and consist of a heating wire or coil, wound
upon an insulating- preform. They are cheap and will last for considerable
lengths of time, especially if they are run at < 1200 °C. (i.e.-Kanthal™
wire). On the other hand, wire made from precious metal, i.e.- 60% Pt -
40% Rh = {Pt(Rh)}, can be operated up to 1800 °C. in air, and will operate
continuously at 1700 °C. for long periods. However, it is expensive and
special care is required when wrapping a furnace core to form the furnace
element. Another idiosyncrasy of this type of heating element is that the
power-source needs to be especially designed. The Pt(Rh) wire has a very
low resistance at room temperature. As it warms to a few hundred
degrees, its resistivity changes considerably. Thus, a current-limiter is
needed in the power control circuit during start-up, or
the {Pt(Rh)} wire will melt. "Globars"™ (silicon carbide rods) are the
next most frequently used furnace heating elements. They will operate
continuously at 1450 °C. and intermittently at 1500 °C. A newer type of
element. Mo wire coated with silicide, i.e.- MoSi2 (to protect the Mo wire
against oxidation), has become common. Such heating elements have
appeared as "hairpins" and will operate at 1750 °C. in air on a continuous
basis. Another heating element coming into use is the defect conductor,
ZrO2 , doped with Y2O3 to make it conducting. However, this type of
heating element has special operating characteristics which hinder its
wide-spread usage. It does not become conductive until 600 °C. is
reached. But, it can be operated continuously at 1800 ° C. in air for long
periods of time, with a maximum of 1900 °C. It does require very high
currents and low voltage to operate satisfactorily. Still another type of
heating element is that of lanthanum chromite, LaCrO3 . These are
available in the form of rods and will operate at 1800 °C. in air for long
periods.

Most heating elements fail due to the development of internal flaws in

their structure. For example, a "globar" is formed by compressing SiC
particles to form a rod, and then sintering it. During operation (especially
272 Measuring particle size and growing single crystals

if it is operated at the upper end of its temperature-operating range), it

develops "hot-spots". These are due to oxidation and formation of
localized resistive areas within the rod (due to diffusion and collection of
vacancies at grain boundaries). These areas dissipate power locally, so that
the rod eventually fails, i.e.- melts locally at the hot-spot and becomes
non-conductive. Wire-wound elements fail in a similar manner, except
that formation of vacancies within the metallic structure is the most
prevalent mechanism which causes "hot-spots" in the operational heating
coil. These vacancies also migrate to grain boundaries. The grain-
boundary-junction decreases in conductivity, due to vacancy defect
formation. Thereupon, hot-spots form within the wire during operation,
causing ultimate failure of the heating element.

Among those heating elements which require the use of neutral, reducing
atmospheres, or vacuum. Mo- wire or W-wire in the form of heating coils,
graphite in the form of rods or semi-cylinders, are most often used.
Iridium wire is also used but it is very expensive. Both Mo and W wire are
usable up to 2800 °C while Ir can be used only to 2400 °C. Graphite
heating elements can be used above 3000 °C.

We have also included R.F. (radio-frequency) current as a heating

element, although it is only a heating method when employed with a
suitable succeptor. Finally, one other method is listed for the sake of
completeness, that of the oxy-hydrogen flame. It generates combustion
products (H2O) but the RF- method can be used in any atmosphere
including vacuum.

The most critical element in melt-growth of single crystals is the

container, or crucible. The first requirement for selection of a suitable
crucible is that the crucible does not react with the melt. The second is
that it be thermally shock- resistant. A third is that of operational-
temperature capability. A fourth is that it be stable in the chosen
atmosphere. These requirements eliminate many potential crucible
materials for a given application. For use in air, the silica crucible (SiO2)
has no peer when used with oxide-based materials up to 1200 °C. It is
almost non-reactive, thermally shock-resistant, and inexpensive. Mullite,
 4.2 Growth of single crystals 273

i.e.- aluminum silicate, is more reactive, less shock-resistant but cheaper

than silica. It is used in making glass-melts for the most part. Table 4-6,
given on the next page, lists the crucible materials most often used and
their temperature capabilities.

Most metals can be used as crucible materials, but only the precious
metals seem to possess the non-reactivity required for melts at the higher
temperatures. Alumina is less shock- resistant than silica but can be used
at higher temperatures. Shock-resistance relates to how fast one can heat
the crucible and its contents up to the melting point of the material
without cracking the crucible. ZrC>2 and BN are most useful for metal
melts. Ft is used for melting in air and remains chemically inert to most
melts.
TABLE 4- 4
COMPOSITIONS SUITABLE FOR CRUCIBLES

FOR USE IN AIR FOR USE IN NON-OXIDIZING ATMOSPHERE

MATERIAL MAX. TEMP. MATERIAL MAX. TEMP.

Silica 1200 °C. Iridium 2400 °C.
Alumina 1700 Zirconia* 2400
Platinum 1750 Magnesia* 2800
Mullite 1400 Carbon* 2800
Boron 1400 Platinum* 1700
Nitride * Not in hydrogen

For preparation of melts in inert atmosphere, Ir stands alone. It is non-

reactive, has a very high temperature limit, and is not subject to thermal
shock or stress. Pt can also be used but at lower temperatures. Ir is about
4 times as expensive as Pt. A good rule of thumb is to use a metal crucible
for oxide melts and an oxide crucible for metal melts, whenever possible.

II. Growth of Single Crystals From the Melt

For the most part, single crystals are grown from a melt of the compound,
provided that melting the compound does not cause it to decompose. The
274 Measuring particle size and growing single crystals

following parameters are very important when considering growth of a

single crystal from a melt:

4.2.3.- Melting point

Partial vapor pressure of melt
Thermal stability of solid phase
Reactivity of liquid or gaseous phase
Thermal conductivity of both liquid and solid phases.

Materials tend to grow polycrystalline. This behavior is related to the 2nd

Law of Thermodynamics and the Entropy of the system. What happens is
that a large number of nuclei begin to appear as the melt temperature
approaches the freezing point (but before it freezes). All of these nuclei
grow at about the same rate, and at the freezing point produce a large
number of small crystals. If we wish to restrict the growth to just one
crystal, we would like a single nucleus to grow preferentially at the
expense of the others. But usually, it does not. However, if we use a "seed"
crystal and set up conditions so that it will grow, then we can obtain our
desired single crystal. A number of techniques have been developed
including the Kyropoulous, the Czochralski Method and the Bridgeman
(Stockbarger) methods. However, they are dependent upon obtaining a
stable "seed" upon which to start and grow the single crystal.

a. Necking In a Seed

The problems of obtaining a seed-crystal are not simple. We can freeze

the melt to a polycrystalline state. When cool, we examine the boule (after
first removing the crucible) to try to find a single crystal large enough for
a seed (« 3 - 6 mm.). We could also cast the melt into a mold and then
look for seeds. We could also freeze a polycrystalline rod by pulling it
vertically from the melt. We would use a small loop of Pt wire to catch
part of the melt by surface tension. By rotating the wire loop while pulling
vertically, we find that a polycrystalline rod of small dimension builds up.
Once we have the polycrystalline rod, we can reheat it next to the melt
surface so that it remelts and "necks-in" as shown in the following
diagram:
 4.2 Growth of single crystals 275

4.2.4.-

What happens is that the crystallites melt and fuse into a small tip. If we
do this carefully, we will have our "seed". The tip's small size limits
regrowth of the remelted part to that of a single crystal. Then, returning
to the melt, we can initiate the growth of a single crystal, provided that
growth-conditions are suitable.

Another way to obtain a seed is to dip a capillary tube into the melt.
Surface tension causes the tube to fill and when it freezes inside, a small
seed results. The difficulty with this method is that it is difficult to obtain
a tube of proper diameter, made of the proper composition. Glass softens
at too low a temperature and quartz melts around 1400-1500 °C. Usually,
we are restricted to metals and even then, we must be able to cut the tube
to obtain the seed, since it is confined within the tube. Once in a while,
we can use the tube directly and obtain growth directly upon the seed,
even though it has remained within the tube.

There is one other method that can be employed to obtain a seed. We use
a metal crucible having a small tip and cause a melt to form as shown in
4.2.5., presented on the next page.

The tip acts in the same way to form the seed. Nevertheless, we have the
same problem as when we use the metal capillary. We need to obtain the
seed free from its holder. Thus, the crucible must be sacrificed, or else
the whole mass must be extracted from the crucible with the seed in the
tip intact.
276 Measuring particle size and growing single crystals

4.2.5.- Crucible for Use in Obtaining A Seed from a Melt

If there is a sufficient difference in contraction between the mass and the

crucible, perhaps then we can obtain the whole mass intact, including our
seed. But this is rarely the case.

It should be pointed out that once we have grown at least one single
crystal of a given composition, we then can save a small piece to use as a
seed. It is only when we try to grow a single crystal from a material which
has never been attempted and accomplished before that we incur the
problems which we have outlined above for obtaining a seed. Nonetheless,
once a seed is available, then we have the problem of getting the crystal to
grow in a proper manner.

b. Melt Growth Using The Czochralski Method

The most common method for growing single crystals, whether they be a
metal or a complex mixture of oxides, is to pull a single crystal from a
melt. This is the so-called Czochralski Method, using the technique
invented in 1918 by the Polish scientist Jan Czochralski (which he called
crystal pulling). Large crystals can be grown from the liquid formed by
melting any given material (providing that the material does not
decompose upon melting). One attaches a seed crystal to the bottom of a
vertical arm such that the seed is barely in contact with the material at
the surface of the melt and allows the crystal to slowly form as the arm is
lifted from the melt. Let us examine this method in more detail to
illustrate its versatility in crystal growth:
 4.2 Growth of single crystals 277

4.2.6.-
278 Measuring particle size and growing single crystals

As an example, we will grow a crystal from a melt of oxides whose melting

points exceed 1800 C. Automatically, we are limited to use of an Ir
crucible and we will use an R.F.- generator for the power source.

In 4.2.6. (see above), we illustrate the typical setup for the crucible and
melt in the Czochralski apparatus. We place the Ir crucible containing the
mixture of oxides on a circular ZrO2 platform. This acts as a thermal
barrier for the bottom of the apparatus. We may then place a larger ZrO2
cylinder around the crucible for further insulation (Note that the choice of
a thermal insulator is not critical except that it must be able to withstand
the anticipated temperatures to be used). An R.F. coil is placed around
the outside of the insulation in a position where it can electrically couple
with the metal crucible. Finally, an outside wall of insulation is put into
place and atop cover plate is put into position. At 2000 °C, the outer wall
thickness of ZrO2 needs to be at least 2.5-3.0 cm. The whole is then
covered by a bell-jar having a hole at the top. The dome and its contents
are flushed for several minutes with an inert gas before the R.F.-
generator is turned on. Gas flow is maintained throughout the entire
operation because we are using an iridium crucible. As the crucible heats
ups, it's entire contents degas. As the materials heat up and melt, they
contract so that it may be necessary to add more material to the melt
until the crucible is full before actual crystal growth is attempted. This
usually takes 3-4 crucible volumes of material. Thus, the top-hole needs to
be large enough to accommodate this step.

1. Growing a Single Crystal

Pulling a good single crystal is not easy. There are many factors involved,
each of which imposes restraints upon the others. Some depend upon the
crystal pulling system design, while others depend upon the nature of the
material being grown. These factors, for a given system, become
parameters. That is, they are inter-related.

The parameters which control CZOCHRALSKI GROWTH are listed in the

order of their importance in 4.2.7., given on the next page. All of these
 4.2 Growth of single crystals 279

4.2.7.- CZOCHRALSKI CRYSTAL GROWTH PARAMETERS

1. Melt temperature and its thermal conductivity

2. Chemical stability of the melt
3. Degree that melt will "super-cool"
4. Temperature control possible and temperature gradient
achievable
5. Interface angle which forms between crystal and melt
6. Degree of diameter control possible
7. Rotation rate used
8. Rate of pulling
9. Melt level maintained

factors are dependent upon the nature of the crystal composition being
grown, including its thermal, physical and chemical properties.

Because of the importance of R.F. generators to crystal growth, let us now

examine them in more detail. A radio-frequency generator is essentially a
gigantic R.C.-tank circuit (R.C. = resistance- capacitance) which uses
mercury-pool diodes as oscillators. The frequency is fixed and the
generator is self-regulating in power-output. As a result, change of power-
output, which affects temperature regulation, is slow. The actual output of
power occurs through the R.F. coil. The coil itself is a carefully-wound
copper tube helix which has cooling water running through it, as shown in
the following diagram:

4.2.8.-

If a susceptor such as a metal crucible is placed within the coil, The R.F.
power induces "eddy-currents" in the crucible, causing it to heat up.
Eddy-currents are circular electrical currents induced within the metal by
280 Measuring particle size and growing single crystals

the R.F. field of the coil. It is essentially a "skin" effect, and the depth of
penetration, i.e.- depth of eddy-current generation within the crucible is
defined by:

4.2.9.- depth = 1/2JC [ p / \d x 1O9 ] 1 / 2

where \i is a permeability , p is a resistivity of the susceptor (crucible) and

f is the frequency of the R. F. field. R. F. generators come in various sizes,
to wit:

4.2.10.- 10 kc - 20 KW
50 kc - 25 KW
100 kc - 50 KW
2 meg. - 75 KW

The size to use depends upon the susceptor being used, the temperature
of operation desired and the heat losses within the system. For an Ir
crucible, the 10 kc-20KW generator works better than the 2 meg.- 75KW
generator. The spacing of the coils is extremely critical. One usually
speaks of the "coupling ratio" for a given crucible- R.F. coil combination.
That is the ratio of power transmitted to the crucible divided by the
power delivered to the coil.

Returning to Czochralski crystal growth, of the 9 parameters given in

4.2.7. above, the first 3 are functions of the material whose single crystal
we are trying to grow. #4 relates both to the material and the physical
design of the melt- furnace plus the electrical characteristics of the coil.
Parameters #5 through #9 relate almost entirely to the operation of the
system. These can be adjusted to accommodate the type of crystal being
grown. There are some crystals which will grow at a rate of several
centimeters per hour while others grow at a small fraction of that rate. As
we have indicated, there are certain steps required to operate the
Czochralski apparatus.

These are given in the following:

 4.2 Growth of single crystals 281

4.2.11.- OPERATION OF THE CZOCHRALSKI APPARATUS

1. Load crucible and start melt. Maintain inert atmosphere as

required for stability of the melt.
2. Add material to melt until crucible is full.
3. When the crucible is full, gradually lower the melt temperature to
achieve supercooling and incipient nucleation.
4. Allow system to equilibrate and dip seed into melt. Rotate seed at
predetermined speed.
5. Allow seed to grow outward to desired diameter.
6. Begin pulling incipient crystal out of melt at rate which maintains
crystal size and growth.

2. Effects of Rotation in the Czrochralski System

We find that the rotation rate affects both the "interface angle" and control
of crystal diameter, as shown in the following diagram:

4.2.12.-

Note that we have defined "interface angle" as the angle between the
growing crystal and the residual melt. Rate of pulling also affects the
282 Measuring particle size and growing single crystals

quality of the crystal as well as the actual number of intrinsic defects

which may appear in the final crystal. In the upper left of 4.2.12., a flat-
interface between the crystal and melt surface has been maintained by
regulating the rate of rotation. If it is not maintained correctly, the edges
the crystal may grow too fast (resulting in a hollow crystal- upper right) or
the edges may round (lower left) or facet (lower right).

If faceting is allowed to continue, we might obtain a result as shown in the

following diagram, where the facet has continued to the point where a
spiral crystal can be seen, as shown in the following diagram:

4.2.13.-

Diameter control, Parameter #6, is important to the final quality of the

obtained crystal. The sides of the crystal need to be straight because they
reflect the regularity of the lattice planes within the crystal. Effects of
deviation from "correct" growing conditions on the quality of the crystal
so-produced, i.e.- its outward appearance are shown as follows in 4.2.14.,
shown on the next page. It should be clear that proper growth of the
crystal has a definitive effect upon the lattice of the single crystal as it
grows.
 4.2 Growth of single crystals 283

4.2.14.-Defects Produced in a Crystal as a Function of Growth Conditions

If the sides of the crystal are kept straight during growth, the lattice
planes remain intact. However, if the crystal is "necked-in" or "pulled-in"
during growth, line-defects appear in the lattice structure. If this is a
laser-crystal that we are growing, the line defects have a localized effect
upon the refractive index of the crystal, making it useless as a laser
crystal.
3. Design of a Crystal Pulling Apparatus

To grow a single crystal, a proper design consists of a heavy base to

minimize effects of external vibrations on the surface of the melt, a
precision screw driven by a controlled reversing-motor (so as to control
rate of pulling precisely), and a precision motor controlling rate of
rotation. Such a design is shown as 4.2.14., presented on the next page.

Note that we have not shown the R.F. generator. However, we have shown
one design of the pulling apparatus (there are many). Here, we have
shown the mechanical parts of the apparatus in addition to the
arrangement of the crucible and melt already shown in 4.2.6. A reversing
motor is used to turn a precision screw to move the pull rod away from
the melt in a precise manner. The function of the "pull-motor" is to turn
the crystal so that it grows in a regular manner. You will note that we did
not show a temperature sensor at the bottom of the crucible. This control
is optional and can be used to monitor the bottom temperature of the
284 Measuring particle size and growing single crystals

4.2.14- One Design of a Czochralski Crystal Pulling Apparatus

Top Plate

Zirconia Insulation

melt. Not shown is a fixed temperature sensor which monitors the

temperature of the top of the melt. This sensor must be used since
temperature adjustments are based upon its readings. There are other
refinements to this apparatus that can be made. We have only shown the
basic components needed to grow a single crystal.

The next most important parameters in Czochralski growth of crystals

are: the heat flow and heat losses in the system. Actually, all of the
parameters (with the possible exception of #2 and #9) are strongly
affected by the heat flow within the crystal-pulling system. The next
section addresses this factor.
 4.2 Growth of single crystals 285

4. Heat Flows in the Czrochralski System

A typical heat-flow pattern in a Czochralski system involves both the

crucible and the melt, as shown in the following diagram:

4.2.15.-

In this case, we see that heat radiates from both the sides and the top of
the melt even though the crucible is being heated by the eddy-currents
from the RF-generator coupling. Note also that the crystal being pulled
also radiates heat. However, this is for a static crucible. The circular
currents set up in the melt are the results of the crystal being turned as it
grows. The circular heat flow pattern causes the surface to radiate heat.

The crystal also absorbs heat and reradiates it further up on the stem. If
the crystal is kept stationary, we see that the heat flow pattern is uneven.
Thus, it is mandatory that the crystal be rotated so as to control crystal
diameter, so as to obtain a defect-free crystal. The above description
applies to the system where only the growing crystal is rotated.

There is at least one other way to "stir" the melt so as to control heat flow.
This is illustrated as follows:
286 Measuring particle size and growing single crystals

6.4.11.- ± crystal rotation = ± crucible rotation

± crystal rotation * ± crucible rotation

We can rotate the crucible by itself, or in conjunction with the crystal. But
another complexity arises, namely what direction of rotation and what
relative speed of rotation should we use for both, or either?

If we rotate the crystal clockwise and the crucible counter-clockwise,

then the heat flow patterns become complex indeed. These complexities
have been studied in detail. A simplified version of the effects of rotation
upon stirring patterns in the melt are given in the following diagram:

4.2.16.- Stirring Patterns Observed for a Static and Rotating Crucible

Here, we show some of the changes that occur in heat flow patterns as the
crystal grows. These patterns are indicative of the heat-flow patterns and
 4.2 Growth of single crystals 287

thermal gradients that exist in the melt. However, if we turn both the
crucible and the crystal, we have at least 8 combinations from which we
can make a choice. The effect on stirring patterns is shown as 4.2.17. on
the following page.

In this diagram,, we have shown clockwise rotation as E , while counter-

clockwise rotation is given as B . the speed of rotation has not been
defined except as "slow" and "fast". If we rotate in the same direction, but
at a slow rate for the crystal and turn the crucible much faster, we get the
stirring pattern shown in # 1.

Note that a completely different stirring pattern occurs if crystal and

crucible are turned in opposite directions (compare #1 to #6). We have
indicated the cases where the crucible rotation is dominant and those
where the crystal rotation dominates.

The best compromise seems to be fast- rotation for the crystal and slow or
no rotation for the crucible. Of all the possible methods of stirring the
melt, the static-crucible method seems to be the best, and this is the
method used by most crystal-growers.

The next best method seems to be rotating the crucible at a slow rate,
counter to the direction of the crystal rotation. It is clear that crystal-
rotation needs to dominate the stirring pattern so that mixing of the melt
continues while the crystal is growing.

However, we have not shown all of the complexities that can occur while
the crystal is growing. To do that would require more space than we have
at the moment. You will note, however, that the problem of building a
Czrochralski crystal pulling apparatus is not easy and requires a number of
factors and controls if the apparatus is to be capable of growing several
different kinds of single crystals, i.e.- from low temperature melting
compounds to high melting compounds. It is for this reason that several
commercial varieties of Czrochralski crystal pulling machines are now
available. These machines incorporate many of the improvements that we
have been discussing.
288 Measuring particle size and growing single crystals
 4.2 Growth of single crystals 289

5. An Automated Czrochralski System

The Czochralski Method has remained the most frequently employed

method for obtaining single crystals. The seemingly formidable
complexities involved have been conquered by the use of computer-
control of the crystal-growing Parameters. One such system, available
commercially, is shown in 4.2.18., presented on the next page.

Here, we show only a bare outline of the individual components in the

overall system. This SYSTEM is capable of operation in inert atmosphere
or vacuum. A melt /seed contact monitor is provided as well as a CCTV
camera for observing and controlling the crystal diameter as it grows.
Note that both the crucible and crystal rotation can be controlled. In
order to control the heat-convection patterns which normally appear in
the melt, an external cryomagnet is supplied. Its magnetic field controls
heat losses, plus it maintains a better control of the crystal growth. A slave
micro-processor controls both crystal diameter and meniscus-contact of
the growing crystal.

As we have stated, this is most important if we wish to obtain a crystal

essentially free from ingrown defects, i.e.- line defects. This takes the
form of a melt/seed contact monitor and a separate monitor is provided
for the operator as a CCTV camera for observing the crystal diameter as it
grows. There is also a crystal annealing furnace to remove any crystal
strain that may have been induced by the crystal-growing conditions. A
base heater helps to maintain a uniform temperature gradient in the melt-
crucible during crystal growth.

Note that both the crucible and crystal rotation can be controlled. In
order to control the heat-convection patterns which normally appear in.
For the melt, an external cryomagnet can be supplied. Its magnetic field
helps to control heat loss to maintain a better control of the crystal
growth. This has not been shown but includes a source of liquid nitrogen
for the cryomagnet as well as a controller to maintain the correct
magnetic field needed for defect free growth of the crystal.
290 Measuring particle size and growing single crystals

4.2.18.-
 4.2 Growth of single crystals 291

6. Growing Laser Crystals

It is the Czrochralski Method that has been primarily utilized for growing
crystals for use as lasers. The first systems were designed by the crystal-
grower who utilized his knowledge of equipment to build the desired
assembly. However, as we have shown above, the Czrochralski Method
involves many complex parameters, all of which interact during the
pulling stage. Although the early apparatus was adequate for study of single
crystals, it was not satisfactory when an optically transparent,
homogeneous and defect-free crystal was mandatory. Defects such as:
chemical (impurities), physical (growth induced lattice defects) and
optical (refractive index inhomogeneities) limited the power that the
laser crystal could generate. "Lensing", i.e.- bending of generated light,
occurred at high power levels which resulted in destruction of the crystal
by overheating.

As we have shown, crystal growth from a liquid melt is accomplished by

establishing an equilibrium between the solid crystal and the liquid melt.
By displacing the equilibrium in slight favor of the solid, the crystal will
grow. The crystal growth rate is determined by two factors:

1. Heat flow through the crystal

2. Mass flow within the melt.

We have already shown how stirring of the melt affects heat flow. The rate
of crystallization, i.e.- the growth rate, is the major crucial factor in the
growth of laser crystals. Although most crystals can be "pulled" at a rate of
several millimeters per hour, the laser crystal can only be pulled at a rate
of 0.4 mm./hr. or less. What this means is that if we wish to have a laser
crystal of about 4.0 centimeters, it would take about 100 hours to grow it.
The costs of growing even larger crystals increases exponentially.

The crystal most studied has been the YAG, i.e.- Y3AlgOn , activated by
Nd3+. As we said, cubic YAG, grows very slowly from a melt, especially
from a {1,0,0} plane. If growth is induced from a {1,1,1} plane, the growth
of the single crystal is accelerated by more than ten times. The reason is
292 Measuring particle size and growing single crystals

that the density of cations along the {1,1,1) plane is much greater than
the {1,0,0} plane. Thus, YAG grows at the rate of about 1.0 mm per hour if
growth along the {1,1,1} direction is initiated. However, if neodymium,
i.e.- Nd*\ is added to the melt, the rate drops to 0.10 mm per hour.
YAG:Nd is the solid state laser workhorse used in industry, comparable to
the CO2 (carbon dioxide) laser used in heavy industry.

For undoped crystals, the limiting factor is heat flow through the crystal.
But, for "doped" crystals, i.e.- activated with Nd3+, the limiting factor is
the distribution of N d3+ ions between the liquid and the solid crystal. This
addition to the melt, which must segregate at the growth interface,
creates the need for mass flow in the melt and lowers the growth rate. To
understand this, we must look at the phase diagram of Y3Al5On and
Nd3AlgOu. This is shown in the following diagram:

On the right, we have shown a "stylized" version of the phase diagram

between the two aluminates. There is limited solid state solubility at high
 4.2 Growth of single crystals 293

yttrium contents as well as high neodymium contents. The exact limit of

each is not exactly known (at least to me). Nor is the position of the
eutectic temperature exactly known. However, the diagram on the left
shows the change in Nd* content that must occur due to liquid freezing
and its effect on the limited phase solubility of the two materials. The
concentration of Nd^ in the liquid is Q whereas that in the solid phase is
Cg. There is a shift in Nd3* content as the liquid freezes and the solid
single crystal is formed. It is this factor which controls the growth rate of
this laser crystal and, indeed, most of the "doped" laser crystals now being
grown.

The addition of Nd** requires a lower growth for the following reasons:

1. The rules of phase equilibrium require that segregation must

occur between liquid and solid.
2. Due to the segregation effect, excess neodymium is rejected into
the liquid and forms a boundary layer at the growth interface.
3. The diffusion of Nd3*, as an activator of the YAG:Nd laser, through
this boundary layer is the mass transport step that limits the growth
rate of the YAG crystal.
4. If the rejected Nd3* exceeds a certain critical level, constitutional
supercooling occurs.
5. Constitutional supercooling takes place when the rejected Nd31"
builds up to the point where liquid ahead of the growth interface, at
the boundary of liquid and solid, becomes supercooled due to its
composition, i.e.- the freezing point of the liquid is changed because
the ratio of Nd3* to YAG has changed. This is easily seen by
reexamining the left side of 4.2.19. Note that the freezing point of
the liquid is a function of its constitution, i.e.- it drops as the Nd3*
concentration builds up during crystal growth. Note also that if the
crystal being grown were: NdaAlgO^Y^, i.e.- on the other side of the
phase diagram, the freezing point would increase as the Y3*"
concentration would build up in the boundary layer.
6. If supercooling occurs due to change in effective composition due
to growth of the crystal, the rate of freezing at the interface
increases dramatically so that the solid-liquid interface breaks
294 Measuring particle size and growing single crystals

down. When this occurs, defects enter the crystal.

7. The critical growth rate is given by:

R c = GD / m C 0 ( l - k )

where G is the temperature gradient at the interface, D is the

diffusion coefficient of N d^ across the boundary layer, m is the slope
of the liquidus line in the left hand side of 4.2.19., Co is the
concentration of Nd3*" in the liquid and k is the segregation
coefficient.

This analysis suggests that the slow growth rates used to grow YAGiNd*1"
are unavoidable since the factors that control growth rate are beyond
regulation by the operator. That is, as long as the growth rate is kept
below Rc, and/or the gradient is kept high, supercooling due to
composition at the interface can be avoided.

Periodic fluctuations at the melt-crystal interface arise from a number of

factors (You will note that we have already discussed some of these,
somewhat perfunctorily, as given above):

1. Crystal rotation in a thermally asymmetric field. That is, if the

crystal is not rotated in a thermally balanced field while it is being
pulled, any point on the solid-liquidus interface can experience a
sinusoidal fluctuations in growth rate. These fluctuations cause
growth striations in the crystal and are the source of "lensing"
mentioned above. If the fluctuations exceed R c , more severe defects
occur in the crystal, as we have already shown.
2. Temperature fluctuations arising from melt turbulence. Note that
we have already shown the melt stirring patterns that occur as a
result of direction and degree of rotation of both the solid crystal
and crucible holding the melt (see 4.2.17.). It is these factors that
greatly influence temperature fluctuations during the pulling step.
3. Daily variations in power available in the RF coil. That is, if the
electrical power varies, deviations in crystal growth will show up in
the growing crystal.
 4.2 Growth of single crystals 295

4. The induction heating coil is a helix so that the RF field is not

symmetrical. Thus, one side of the iridium crucible will be hotter
than the other side.
5. Another source of variation is the temperature of the cooling
water. The power delivered to the Ir crucible is: W= I2R, where r is
made up of two components, i.e.- RD + R,,, where the former is the
resistance of the crucible (and its contents) and the latter is the
resistance of the coil itself. It has been found that variation in
temperature of the cooling water itself has a definitive effect of the
power output of the RF heating source.

At this point you might think that the problems associated with growing
laser crystals are nearly impossible to overcome. However, this was the
point of the diagram given in 4.2.18. By automating the crystal-growing
system, we find that thermal variations at the liquidus interface can be
diminished and nearly abated. Now, we can show the parts of the overall
crystal-pulling system of 4.2.18., not shown but mentioned above. This is
shown in 4.2.20., given on the next page.

Note that the crystal is being weighed in real time so that a crystal
diameter can be calculated. This assumes that the crystal remains round
and no voids occur during growth. By keeping track of the diameter as the
crystal grows, avoidance of ingrown defects can be avoided as well as
being able to remain as close as possible to the critical pulling rate, Rc.
The power to the RF coil is thereby controlled precisely by a separate
feed-back loop.

Not shown is a further improvement consisting of a water-temperature

control for the cooling water in the RF coils. It has been found that even
changes of a few degrees (° C) is mirrored in changes of crystal diameter
in the growing crystal. It is these changes that have made possible an
increase in pulling rate of most laser crystals, including that of YAGrNd3*,
without exceeding the critical growth rate mandated by the individual
phase diagrams of the solids. In this way, critical defects in the crystals
have also been avoided.
296 Measuring particle size and growing single crystals

4.2.20.- Control Loops Used to Control Crystal Diameter

In order to show you what is meant here by critical defects, consider the
following diagram, which shows growth defects in a crystal before
processing into laser rods.

4.2.21.- Growth Defects in a Crystal Before Processing into Laser Rods.

 4.2 Growth of single crystals 297

In many cases the constitutional defects (which are only seen under
polarized light) are nearly abated and in some cases completely
eliminated. In the cases shown above, four rods of selected diameter can
be cored from the pulled crystal rod on the left. The rod on the right
allows much larger diameter rods to be cored, or if desired, several more
rods of smaller diameter to be obtained.

A final comment on the Czochralski Method: GaAs has become important

in construction of integrated circuits for computers because of the
promise of speed of response and density of components, compared to
silicon wafers. Both As and Ga tend to oxidize in air as they approach the
melt stage and AS2O3 sublimes. Both elements are toxic to man. In the
melt stage, they have high partial pressures as well, so that the use of an
inert atmosphere is not sufficient to allow growth of a single crystal.

One method that has been used for this case has employed a "liquid
encapsulating agent". As shown in the following diagram, this consists of
a lower melting agent used to form a liquid barrier, floating on the surface
of the melt which serves as the source for the crystal (in this case, Ga-As).

4.2.22.-

The arrow indicates the liquid barrier layer. This use of a barrier melt
illustrates that there are several ways to grow crystals which would be
difficult to obtain under "ordinary" means of crystal growth, i.e.-
prevention of oxidation and evaporation of GaAs during crystal growth.
298 Measuring particle size and growing single crystals

c. Melt Growth Using the Bridgeman-Stockbarger Method

Whereas the Czochralski method requires rather elaborate equipment to

obtain single crystals of good quality, the Bridgeman (Stockbarger)
method uses a fairly simple apparatus. This is shown in the following
diagram:
4.2.23.-

Note that we are using a shaped crucible in order to induce a seed to

make the crystal grow. Under optimum conditions, one can grow quite
good crystals by this method. However, one must experiment before the
final conditions become apparent for the particular crystal being grown.
 4.2 Growth of single crystals 299

The major components needed to grow single crystals by this method are:

* A two (2) zone furnace

* Two (2) set point temperature- controllers
* A crucible with "seed" tip
s§§ A constant-speed elevating and lowering device for
the crucible.

The procedure involves obtaining a full crucible of melt whose

composition mirrors that of the crystal we wish to grow. We then lower
the crucible containing the melt through a baffled zone within the furnace
at a slow rate. The baffles within the furnace produces a uniform
temperature decrease between the two zones, resulting in a temperature
profile like that given on the right in the Figure. The temperature set-
point of the upper zone is just above the melting point of the material to
be grown, and the lower zone is set just below the melting point of the
material. This results in a smooth temperature profile, above and below
that of the crystal fusion point. Thus, if we can induce a seed to form in
the tip of the crucible, then, as the crucible is lowered through the
freezing part of the temperature zone, the rest of the crucible will freeze
as a single crystal. The parameters for the Bridgeman Method are:

4.2.24.- PARAMETERS INVOLVED IN THE BRIDGEMAN METHOD

1. Temperature Profile within Furnace

2. Rate of Crucible Raising or Lowering
3. Melt Temperature
4. Supercooling of Melt
5. Annealing Temperature
6. Thermal Conductivity of Melt and Crystal
7. Thermal Expansion of Crystal
8. Crucible Material Used
9. Chemical Stability of Material Being Grown

Parameter #1 is a matter of furnace design. There should be provision for

300 Measuring particle size and growing single crystals

several baffles so as to adjust the temperature- gradient distance. Then, it

is a matter of adjusting the two set-point temperatures to achieve the
proper gradient. For some crystals, the gradient needs to be sharp,
whereas for others it can be more gradual. The two set-point parameter
depends primarily upon the degree of supercooling experienced in the
melt, prior to nucleation. This in turn may be affected by purity of the
constituents. The raising and lowering mechanism can be simple, that of a
gear-driven motor with a counterweight for the crucible-melt mass. It is
convenient to use the lower set- point as the annealing temperature for
the crystal. But, for very high melting points, this may not be possible.
Then, we need to add a lower annealing furnace. As the crystal freezes, it
may not do so uniformly. If so, internal strain results (a polariscope will
reveal this). Furthermore, if the expansion coefficients of the crystal and
crucible are too disparate, then external strain on the crystal will generate
internal strain. The crystal will than have to be annealed to relieve this
strain. The Bridgeman Method is useful for many types of crystals. The
key to getting this method to work is to induce a seed to form in the
"seed-tip". If the crucible "seed- tip" is not properly designed, the method
will not work.

d. Melt Growth Using the Kvropoulos Method

A comparable way to grow crystals is the Kyropoulos Method, which uses

a type of apparatus similar to the Bridgeman-Stockbarger method. In the
Kyropoulos method, we first form a melt and then introduce a seed. By
raising the melt through the temperature gradient, a single crystal will
grow from the point where the seed engages the melt. The apparatus is
shown on the next page as 4.2.25.

Note that we have reversed the temperature gradient within the furnace
in 4.2.22. and that the top is cooler then the bottom, where the melt is
first formed. In other variations of the Kyropoulos method, we raise the
melt through the freezing point of the melt, by raising it past the baffles of
the furnace. Another possible method is one where we use a single- zone
furnace, stabilize the melt, lower the furnace temperature to incipient
 4.2 Growth of single crystals 301

4.2.25.-

nucleation and then cool to form the crystal. However, this works only for
a few systems. The reason is probably supercooling. There are only a few
melts where a large degree of supercooling occurs. Addition of a seed will
then cause very rapid growth of the single crystal. But, it is usually
strained and must be carefully annealed to obtain the mass intact. If the
apparatus shown in 4.2.25. is used and the crucible is exactly positioned
so that a vertical temperature gradient is maintained exactly along the
crucible height, the melt, when seeded, will grow into a single crystal
more slowly. Then, the degree of supercooling is not as important. As a
matter of fact, melts with little or no degree of supercooling can be
induced to form single crystals by following this procedure.

e. Crystal Growth Using a Variation of the Kvopolous Method

The problems in getting a "seed" to form in the Bridgeman crucible was,

perhaps, the basis for a new and different type of crystal growth based
upon a variation of the Kyropolous method. It was noted that if one could
get the melt to "super-cool" (that is- not crystallize when the melt
temperature was lowered just slightly below the crystallization
temperature), one could then induce rapid crystallization by "touching"
the surface of the super-cooled melt with a seed of the same material. In
302 Measuring particle size and growing single crystals

this method, the seed is introduced after the melt temperature has been
stabilized and then brought to incipient nucleation. This is illustrated in:

4.2.26.

It is this method which is currently being used to manufacture sapphire

boules as large as 18 inches in diameter. The boule is then cut into slabs
which are polished and used for UV transmitting windows, substrates for
various electronic devices and even non-scratching faces on your watch.
Note that sapphire is about as hard as diamond and has a heat-
transmitting capability almost equal to many metals, while remaining
essentially chemically inert and electrically non-conductive.

Sapphire has become the material of choice for many applications,

including windows for watches and other devices where other materials
like glass would be easily scratched. In addition to hardness, sapphire has
high mechanical strength for high pressure and high shock loading. It is
also transparent from 1420 A to 50,000 A (far infra-red). This makes it a
good candidate for optical components in the integrated circuit design
field. However, sapphire has a rhombohedral lattice structure which
results in anisotropic properties. That is, it refracts light at different
angles according to the wavelength of the light. Such behavior is called
birefringence. Sapphire can be used if a {0,0,0,1} oriented crystal is used.
However, getting the crystal boule to freeze along the (0001) planes is
very difficult. Usually, the preferred growth orientation is on the (1120)
or (1010) axis. The growth of a large boule such as shown in 4.2.26.
allows one to orient the crystal and cut optical parts along the (0001) axis
 4.2 Growth of single crystals 303

to obtain high quality sapphire optical elements such as lenses, prisms

and shaped optics having little or no birefringence.

Although we have illustrated how the crystal is frozen in place within its
melt container, the actual situation is not so simple. It has been only
within the past few years that the technique for obtaining a melt at the
desired supercooling stage has been realized. The method is called the
"heat-exchanger method" (HEM). The following diagram only shows a part
of the total design of the HEM apparatus:

4.2.27.-

In this case, we have not shown all of the parts of the apparatus, just the
essential ones. You will note that the heating elements create the melt by
304 Measuring particle size and growing single crystals

radiative heating in a vacuum. Once the melt is obtained, the vacuum is

turned off and helium gas is flowed through the heat exchanger to
progressively cool and create a supercooled melt. In this case, there is a
temperature gradient formed in the melt from top to bottom. By carefully
measuring this gradient (the instruments are not shown), one can achieve
a supercooled melt which will freeze upon command into an oriented
single crystal. In some cases, a single crystal seed is placed on the bottom
of the crucible before the melt is to be formed but care has to be made not
to completely melt the seed crystal before the freezing step is initiated.
Sapphire single-crystal boules of 33 cm in diameter have been routinely
fabricated which weigh up to 65 kg.

The other crystal being grown in this way is calcium fluoride, CaF2. Such
crystals are as large as 30 cm in diameter and 10 cm thick. Because of its
low index of refraction and low dispersion, clear, colorless fluorite of
optical quality is used for apochromatic lenses. The crystal is cut into
pieces and fashioned into optical components. CaF2 is transparent from
0.1pm (1000 A) to 7.0 nM (70,000 A - mid-infra-red). Lithography is the
technology used. This is an optical technique which uses a photo-mask to
isolate certain areas of a photo-resist from radiation which polymerizes
any part of the photoresist exposed. The un-polymerized resist is then
washed off, leaving the desired circuit consisting of patterned lines. This
is repeated many times. Integrated circuit manufacturers form line
circuits on a single crystal disc of silicon. They have progressed from
relatively thick lines to 0.18 jam to 0.13 jim lines.

Line-widths in the integrated circuit alternate with "trenches" so that the

metal "lines" connect individual transistors in the circuit, be it a CMOS or
MOSFET type. The alternation looks like this: n i l i i n n n n n 11111111
The density of lines for a 0.13 ym (130 nm) spacing is thus about 38,500
lines per centimeter (1/2 of the total space, assuming a linear cross-
section). If a tighter spacing is sought, the light source is restricted since
its diffraction limit is close to the Franhofer limit and must be changed.
Originally, ordinary light sources, i.e.- incandescent or discharge lamps
were used but soon met their limitations. Even monochromatic line-
 4.2 Growth of single crystals 305

sources, i.e.- lasers, have approached their diffraction limits using quartz
optics. Although Excimer, i.e.- rare gas, lasers are available, quartz absorbs
too much of the optical power, i.e.- does not transmit the ultraviolet light,
to be of any use. Currently, excimer lasers operating at 193 nm. are being
used with CaF2 optics. This allows lines to be formed as small as 100 nm.
When the 157 nm laser becomes the workhorse of the microlithography
process to form IC lines, it is expected that lines as small as 50-70 nm
will become possible.

f. Edge Defined Crystal Growth From the Melt

Many times, the form of the crystal obtained is not suited for the end use
contemplated. For example, we have already shown how a flat plate of a -
AI2O3 (sapphire) is manufactured, using the HEM method. This material
is often used as a base for integrated circuits (IC's) because of its high
thermal conductivity and its low electrical conductivity. Si is vapor
deposited on the surface of the alumina plate, and through various
photographic techniques and selective etching (with suitable additives
diffused into the silicon layer), an integrated circuit is built up. The IC
relies upon the high thermal conductivity of the alumina base for its long
life since "hot-spots" can quickly destroy the IC. In the past, the
Czrochralski method was used to grow an a - AI2O3 crystal. It was then
sliced into wafers, polished on both sides and then into smaller flat plates
of the required size. However, C1-AI2O3 (sapphire or corundum) is
extremely hard (9.5 on the Mohs scale) and is difficult to work with.

One method that was used was to grow a - AI2O3 directly as a flat ribbon.
By using edge defined growth, one can do this. The apparatus consists of a
normal Czochralski melt with an anvil at the surface of the melt, as shown
in the following diagram, given as 4.2.28. on the next page.

Once the crystal has started to grow, we pull it through the anvil, thus
defining its size. Once it is in the form of a strip, as shown below, it can
be drawn and wound over a large wheel. The strip is later cut directly into
plates, with no polishing required. Because of the high melting point of
a-Al2O3 (M.P. = 1920 °C), an iridium crucible and anvil are needed.
306 Measuring particle size and growing single crystals

4.2.28.-

But, the method is more versatile than we might suppose. We can grow
crystals in the form of tubing, rods or strip, as shown by the anvils at the
right. In fact, we can grow in nearly any configuration we might wish.

Tubing in the form of a - AI2O3 is used for the construction of the familiar
high- pressure sodium- vapor lamps used for street- lighting. End- caps of
niobium metal are sealed on the tubing; the capped tubing is evacuated,
sodium metal is added and the whole sealed off and mounted. Operation
of the lamp occurs at ~ 800 °C and about 15 atmospheres internal
pressure of sodium vapor. These operating conditions mandate the use of
an optically transparent, chemically- and thermally- stable tubing such as
a - ALjOg . In fact, no other material is known that will successfully
withstand these operating conditions.

Edge-defined growth of a single-crystal tubing of <X-A12O3 for use in high-

pressure sodium- vapor lamps was initially tried and found to be
successful. However, this method was deemed too expensive since
polycrystalline tubing formed by high-pressure sintering at 1700 °C and
100 atmospheres pressure performs as well and at much lower cost per
tubing. Additionally, the HEM method has largely supplanted edge-
defined growth of a-Al2O3 strips or wafers for use in integrated circuit
design.
 4.2 Growth of single crystals 307

g. Zone Melting and Refining

There is nothing to say that we must melt all of the material at one time.
If we have along rod of polycrystalline material, and melt apart of it, i.e.-
a "zone", we can sweep this zone down the length and ultimately melt and
freeze all of the rod. The major problem is, again, inducing a "seed" of
single- crystal material to form (and to continue). In the forward direction
of travel, the zone is melting whereas at the back of the zone, freezing
occurs. If we can induce the back part to freeze as single- crystal, then we
can transform the whole rod. This means that we can start with
polycrystalline material and end up with a single- crystal, namely:

4.2.29.- Powder => Polycrystalline =*• Single Crystal

For the case of Zone-Melting to form a single- crystal, we can distinguish

between two separate cases, namely- horizontal and vertical zone
melting. This is shown as follows in the following:
4.2.30.-

In the horizontal case, we fill a boat with powder whereas the vertical
method requires a pressed rod, made from powder. The heating- source
illustrated is a R.F. - coil. This is the best heat- source, providing that the
308 Measuring particle size and growing single crystals

material is conducting and can be heated by induced eddy-currents.

Silicon is the material for which this method has been most employed.
Note that our choice among heating elements is limited to the R.F.- coil, a
hot-wire coil or perhaps the MoSi2 "hairpin" element. The reason is that
we must restrict the melt zone as much as possible, while moving the
zone through the material. Heating by radiation does not perform very
well since it occurs i n a 4 n direction.

One possible solution is to use a heat reflector around the wire heating
element to limit the heating length. Another limitation of the method is
that the power losses in the system are large because we must use a
protective tube or other insulation around the material so as to be able to
move the heat source. Unless we are willing to squander power input, we
cannot reach the very high temperatures of melting required for some
materials. Therefore, the zone-melting method of single crystal growth
has usually been confined to materials of moderate to low melting
temperatures such as silicon.

The vertical method shown above has sometimes been called the "floating
zone" method. The thickness of polycrystalline rod determines the
maximum length of melt zone that can be used successfully. This relation
can be calculated from:

4.2.31.- THIN RODS THICK RODS

ZMax = 2 (3r)l/2 ZMax = 2.8 [ y I pg ]

where r is the radius in cm., y is the surface tension of the melt, p is its
density , g is the gravitational constant, and ZMax is the maximum length
of the floating zone.

1. Crystal Purification Using Zone Refining

When a crystal freezes from the melt, it tends to reject impurities. If

Zone- Refining is applied to such a crystal, the resulting single crystal can
be purer than the original. If we make more than one pass, we can
 4.2 Growth of single crystals 309

approach a high degree of purity. Such a procedure is called "zone-

refining".

It has been determined that there is a distribution coefficient for the

impurities between crystal and melt which favors the melt. We can see
how this arises when we reflect that impurities tend to cause formation of
intrinsic defects within the crystal and lattice strain as a result of their
presence. In the melt, no such restriction applies. Actually, each impurity
has its own distribution coefficient. However, one can apply an average
value to better approximate the behavior of the majority of impurities.

One calculates the concentration of impurities left in the single crystal, as

a function of a single pass, from:

4.2.32.- ci = c 0 [1 - d-ko) exp {- k o L I Z}\

where c 0 is the original concentration of impurities, L is the length the

zone travels (in cm.), and Z is the melt-zone length. Note that we are
taking a ratio of L to Z. We need to minimize Z in order to make the
process efficient in segregation of impurities. k 0 is the distribution
coefficient, defined by:

4.2.33.- k o = Ci (solid) / a (melt)

We plot Cj / c o vs: L / Z, the number of zone-lengths solidified. This is

illustrated in the following diagram, given as 4.2.34. on the next page. In
this case, we have shown how the impurities change as a function of the
distribution coefficient for one pass. Note that at values of k o > 1, we do
not obtain any significant improvement in impurity content. For example,
we start at k = 5.0, but end up at 1.0. Even at ko = 0.5, the further we
move the molten zone, the less we purify the material. This has been
called the impurity-leveling effect. At the very low values of ko, one still
cannot improve the impurity level to any significant degree, using just one
pass.
310 Measuring particle size and growing single crystals

We find that multiple passes are required to obtain purification. This

effect is shown in the following:
 4.2 Growth of single crystals 311

Here, we show how the impurities change as a function of the number of

passes, given a specific distribution coefficient for the impurities in the
melt relative to those in the growing crystal. Note that as the number of
passes reaches 10, we obtain a change of some 90 times in the impurity
content at the front of the rod. But, at the back, i.e.- L/Z = 8, we approach
an ultimate distribution with 20 passes (the dotted line). Here, we have
used:

4.2.36.- k0 =0.5andL/Z=8

If we could see the internal distribution of impurities after the rod had
been zone-refined by 20 passes, we would see the distribution of
impurities as shown in the following:

4.2.37.-

In this diagram, we show the actual distribution of impurities in the

crystal after it has zone refined by 20 passes. The density of dots
represents the concentration of impurities. What has happened is that we
have moved many of the impurities from the front to the back, but not all.
and we finally end up with an impurity distribution we cannot change. To
understand this, we must examine the impurity-leveling factor in more
detail.
2. The Impurity Leveling Factor

When a melt-zone is moved through a long crystal, impurity

concentrations build up in the melt zone due to rejection by the growing
crystal. We can also say that the distribution coefficient favors a
purification process, i.e.- k « 1. Another reason (at least where metals
are concerned) is that a solid-solution between impurity and host ions
exists. It has been observed that the following situation occurs:
312 Measuring particle size and growing single crystals

4.2.38.-

The impurity, x , builds up at the solid- liquid interface as the zone moves.
We can write for the distribution coefficient:

4.2.39.- k x = CxL / c x S

where xLis the impurity- concentration in the liquid, etc.,

and we estimate k from:

4.2.40.- kNi / N M X exp - ( E M - E X ) / k T

where I is the impurity, MX is the compound under consideration, and

EM , Ex are the activation energies for formation of solid solution of MX •
IX . Note that we have assumed that I is a cation impurity in the crystal.
We can differentiate between 3 separate cases, as shown in 4.2.41.,
presented on the next page.

It should be clear that if the impurity freezes into the solid from the melt,
there must be a certain amount of solid solution formation, even though
the impurity content be less than 0.01%. Case I & II show simple solid
solution behavior for k < 1. For c ^ > CxS we will get an improvement in
purity in the crystal. But for C^L < c x s the opposite effect is seen. Thus,
two different impurities could manifest opposite behaviors, depending
upon what host they were in.

Nevertheless, simple solid-solution for impurity systems is rarely the

norm. The most prevalent case is that of Case III of 4.2.42. Limited solid
solution occurs, and we get a two-phase system, as shown in that diagram,
also given on the next page.
 4.2 Growth of single crystals 313

4.2.41.- Behavior of Impurities as a Function of Type of Solid Solution

Formed
314 Measuring particle size and growing single crystals

In 4.2.42., we show the phase diagram of the crystal composition and two
cases for impurity segregation between melt and crystal as it grows in
time (given on the right of the diagram). Note that an initial purification
occurs in both cases but the distribution coefficient for the upper case is
such that the amount of impurity actually incorporated into the crystal, kj-
c o , rapidly approaches the original impurity concentration, c 0 . At ti, it
actually increases over the original concentration at some point within the
solid crystal.

The same thing occurs for the case at the bottom of 4.2.42. except that
the distribution coefficient is such that the impurity buildup is slower.
This illustrates the fact that impurity segregation and purification
processes are dependent upon the type of impurity involved and its
individual segregation coefficient. As we illustrated above in 4.2.34., the
problem is that the impurity is initially rejected from the solid, but its
concentration builds up in front of the growing crystal. The segregation
coefficient, ki , then operates on that increased concentration and the
product, ki c o , increases.

If we use individual zone lengths, as shown in 4.2.34., we would have:

4.2.43.- ZONE IMPURITY CONCENTRATION

1. (c o - k c 0 )
2. (c o - k c o ) + k(c o - k c 0 )
3. (c o - k c o ) + k(c o - k c o ) + k 2 (c o - k c o ), etc.

This is the reason for the behavior shown in 4.2.38. The impurity front
builds up until its concentration surpasses the original concentration
(sometimes by manyfold). Thereupon, the impurity concentration levels
out (but not necessarily to the same c 0 we have used for illustration in
4.2.38.). It is for these reasons that zone-refining has found limited
application, since actual purification by this method does not produce a
crystal with complete uniform impurity distribution.

The conclusion we must draw is that we should purify our raw materials
 4.2 Growth of single crystals 315

before we attempt to grow our single crystal, not after. Zone refining has
been found to be effective only in a few limited cases. It is not applicable
to the general case or to high melting crystals. But if we wish to do zone-
leveling of an added impurity, then we have a very useful technique. A
good example would be to add a small amount of Ga to Ge . We can add a
small volume of Ga to the front zone and will find that it has become
evenly incorporated into the Ge crystal after several passes, using zone-
refining.

While the above description might seem quite complex, one needs to
know what is happening as one attempts to grow a single crystal, even if
the description is somewhat limited in scope, as is the above narrative. It
should be clear that the distribution of impurities between the melt and
solidifying solid is a function of the phase diagram applicable, i.e.- the
phases present, and the nature of the material, both chemically and
physically, being used. If you need to know more concerning impurity
distribution in crystals, a number of publications dealing solely with this
subject can be found in the library.

h. Crystal Growth from the Melt- The Verneuil Method

In our discussion to this point, we have formed a melt and then processed
it to grow a single crystal. In the Verneuil method, only a small amount of
material is actually melted at one instant and a single crystal is gradually
constructed. This method has also been called the "flame-fusion" method
of crystal growth. If a powder is blown through a flame, it will melt if the
flame is hot enough. The only flame hot enough to do this is the oxy-
hydrogen flame. However, a specially designed burner is required so that
the crystallites will melt during the short time that they pass through the
flame-front. The original work was accomplished by Verneuil (1931) and
the apparatus is named after this investigator.

The tip of the torch is important. One design used for such a torch is
shown in 4.2.44, given on the next page. The center tube is used for
oxygen gas which transports the powder. The burner is designed so that
the outer tubes contain only hydrogen gas, with the interstices between
316 Measuring particle size and growing single crystals

4.2.44.-

the H2 - tubes transporting additional oxygen gas. This design will melt
Y2O3 powder which has a melting point of 2380 °C.

As shown in the following diagram, given as 4.2.45 on the next page, the
overall Verneuil apparatus consists of a sealed hopper to contain the
powder, the TORCH itself, a refractory pedestal to hold the growing
crystal, and an after-furnace to anneal the crystal. Thermal stability is
important in this method because the high temperatures reached may be
sufficient to cause decomposition of the material.

The Verneuil Method of crystal growth is not generally applicable to all

types of crystals. There are serious deficiencies in the method. For
example, there is a large temperature drop of hundreds of degrees over a
few millimeters within the crystal. This causes a large difference in
thermal expansion within a limited space, and consequent strain. Many
crystals are not refractory enough to withstand the stress buildup and so
crack into many smaller parts. This makes it very difficult to obtain a
single crystal of any size. Thus, if a crystal is grown by this technique, it
must be annealed carefully in order to obtain it intact.

Generally, the method is restricted to crystals like AI2O3 whose thermal

conductivity is high and whose refractive nature makes it possible to
obtain a crystal. For the most part, one is restricted to growing simple
oxides by this methods. Complex oxides such as Ca2SiO4 or CCI2P2O7
generally cannot be grown easily and the Czoehralski Method becomes the
method of choice.
 4.2 Growth of single crystals 317

4.2.45.-
318 Measuring particle size and growing single crystals

The procedure for using the apparatus shown in 4.2.45. is given as

follows:

4.2.46.- Operating Instructions for Verneuil Apparatus

1. Fill powder hopper, close and attach to burner.

2. Light burner and adjust flame to oxidizing.
3. Adjust pedestal to proper height and begin rotation.
4. Slide after-furnace into position.
5. Turn on and adjust oxygen flow in powder hopper.
6. Begin tapping.
7. As melt begins to build up, adjust both height and speed of
rotation so that crystal begins to form in a regular manner.
8. When the proper crystal size has been reached, stop
powder flow by shutting off the oxygen flow to the hopper.
Continue flame operation and pedestal rotation.
9. Lower the crystal into the after- furnace, and anneal crystal,
Gradually lower furnace temperature and let cool to room
temperature.

The parameters involved in the Verneuil method are:

1. Melt Temperature
2. Thermal Stability at Melt Temperature
3. Temperature Gradient present in Furnace and after-
furnace.
4. Sintering Volume Losses
5. Thermal Conductivity of Crystal
6. Annealing Temperature Required
7. Effects of Reducing Flame on Material
8. Chemical Stability of Material being Used

Although this method was used initially to grow very refractive crystals of
high melting point (because Czrochralski equipment was not adequate to
attain the high temperatures needed), recent improvements in
 4.2 Growth of single crystals 319

Czrochralski growth equipment have made the Verneuil method nearly

obsolete.
i. Melting and Stoichiometrv of the Crystal

Although we have described the use of melts to grow single crystals, we

have said little concerning how such melts are achieved and the
stoichiometry obtained. In general, one adds the correct ratio of oxides
and melts them to obtain a melt of the compound desired. We may also
prereact the oxides to form the compound and then melt it. When a
crystal is grown from either melt, little difference can be noted in
individual crystal properties. However, this applies only to congruently
melting compounds. What this means is that the melt retains the same
chemical stoichiometry as the materials used to make the melt. As a
matter of fact, it is possible to use an excess of cationic- oxide, or anionic-
oxide, in the melt and still obtain a stoichiometric crystal.

There are some cases where a non- stoichiometric composition will be

obtained even though the compound is congruently- melting. An example
of such a phase - system is shown in 4.2.47., given on the next page.

Even though this phase diagram is complicated, it is easy to understand.

We start with compounds "A" and "B", one on each end of the phase
diagram, i.e.- 100% A or 100% B. There are two regions in between these
limits where solid solutions of A & B occur. More importantly, a region
exists where the composition freezes and melts congruently (that is- the
melt does not decompose). When it melts incongruently, however, the
melt then decomposes into two separate phases. On either side of the
congruently- melting composition is a phase region of partial- solid-
solution. In those regions, one solid freezes before the other, so that it is
impossible to grow a crystal from a melt of these mixed oxides. Note that
even at the eutectic composition (there are two of them), one still obtains
a mixtures of oxides.

However, there is a congruently- melting composition and a range of

compositions associated with it. The low temperature stoichiometric
composition melts incongruently but it is not easily achieved from the
320 Measuring particle size and growing single crystals

melt. It shifts composition upon melting or freezing! One example, not

generally known, is that of the YVO4 crystal.
4.2.47-

If we make a powder by reacting yttrium oxide with vanadium pentoxide,

i.e.-

4.2.48- Y2O3 + V2O5 =* 2 YVO4

we find that YVO4 decomposes upon melting to YVO3 plus oxygen. YVO4 is
 4.2 Growth of single crystals 321

used as a laser crystal because of its unique optical properties. The

vanadate group is optically active and has an optical absorption coefficient
of several thousand because the transition has a charge-transfer character.
In general, laser crystal "activators" like neodymium, Nd3+ , have
absorption coefficients of less than 100. Thus, the absorption band of
YVO4 is very wide which makes it easier to couple to many types of
excitation light sources. However, the melting decomposition of YVO4 has
been solved by incorporating excess vanadium pentoxide in the melt.

1. Dvstectic Crystals

It turns out that this behavior is more prevalent than one might realize. It
is, in fact, typical for many systems and some of these are listed in Table
4-3. The maximum composition- existence range is given along with their
dystectic composition, that is, the most usual composition.

TABLE 4-3
Nonstoichiometric Nature of Some Binary
Congruently Melting Compounds

COMPOUND MAX. EXISTENCE RANGE DYSTECTIC COMPOSITION

Gao.5+xPo.5-x + 0.00025 to - 0.00005 + 0.00006

Gao.5+x Aso.5-x + 0.00023 to - 0.00018 + 0.00004
Pt>o.5+x So.5-x +0.0005 to - 0.00005 +0.0005
Pt>O.5+xScO.5-x NOT KNOWN -0.00005
Pbo.5+xTeo.5-x + 0.0005 to - 0.00013 - 0.00012
Sno.5+xSo.5-x +0 to - 0.000008 - 0.000005
Cdo.5+xTeo.5-x + 0.000002 to - 0.000008 - 0.000005
Mgi+xAl2-2xO4-2x + 0.47 to - 0.42 NOT KNOWN
Gd3+xGa5-x O12 + 0.30 to - 0.002 + 0.05
Lii-5xNbi+xO3 +0.031 toO- +0.0092
Lii-5xTai+xO3 + 0.025 to O" + 0.0066

A closer examination of this table reveals a rather startling conclusion. It

can be seen that most of these compounds are subject to structure-
322 Measuring particle size and growing single crystals

vacancy defects. Certainly all of the pnictides fall into this class. We are
even given the defect concentration. Thus in GaP, we have:

4.2.49.- 6 pp H p G a + (1+8) V p + aVj

where 8 is usually 0.00006. We can conclude that the melting and freezing
behavior of a compound has a significant effect of its stoichiometry.
However, it appears to manifest itself mainly as point defects in the
crystal produced. The point is that although we melt the components,
whether they be oxides or otherwise, to form a desired compound or
composition, there are many cases where we do not obtain the
stoichiometry we expect to get.

This brings us to an important part of our discussion of stoichiometry of a

crystal formed from a melt. As we have shown above, we can sometimes
obtain a dystectic composition but most likely we will obtain a crystal
containing "ordinary" types of defects.

2. Defect-Free Single Crystal Silicon

A particularly important crystal is that of single crystal silicon. Pure Si is

used in the form of flat, polished plates no more than 0.060 inches thick,
i.e.- 1.5 millimeters. Growth of silicon crystals begins with the
preparation of extremely pure polycrystalline silicon having fewer than 1
dopant atom per 10 billion silicon atoms. This silicon is melted in a
quartz-lined furnace. The temperature of the molten silicon is reduced to
just above the melting point, and a small bar (the seed) of silicon in
single-crystal form is introduced into the surface of the melt. The seed is
slowly withdrawn, while rotating under carefully controlled conditions, to
form a single-crystal cylindrical ingot of silicon. This ingot is greater than
200 millimeters in diameter and weighs up to 100 kilograms (220
pounds). After growth, the silicon crystal is ground to a smooth cylindrical
shape and sliced into thin wafers, using diamond tools. These are usually
about 0.6 millimeter thick. The surfaces of the wafers are polished flat by
a series of successively finer abrasives until one side has a perfect mirror
finish. For the manufacture of integrated circuits, the methods used to
 4.2 Growth of single crystals 323

achieve planarity and smoothness of the individual discs are extremely

important. It has been found that any crystal defects which appear in the
surface of the silicon disc are mirrored in the integrated circuit being
formed on the surface of the disc. Thus, many "dice", cut from the disc to
make operating transistors do not function properly and must be
discarded.

Under conventional crystal growth conditions of silicon boules, vacancies

agglomerate to form small, low-density octahedral voids, commonly called
"D-defects". Interstitials can form distributed dislocation clusters. Both
have serious effects upon the process of forming integrated circuits. The
void defects have been clearly associated with dielectric breakdown
failures in integrated circuits, while the dislocation defects have been
associated with controlling the rate of the electrical defect producing
reactions due to integrating the management of defects throughout the
entire growth process. In other words, the dislocation defects tend to
cause the formation of defects in the integrated circuit at the point of the
dislocation defect. Additionally, such a defect may not appear until the IC
is being operated. Such defects are known to be related to certain classes
of leakage current failures. The void defects have recently become a par-
ticular concern throughout the industry in spite of their extremely low
density.

In the growth of the single-crystal silicon, the crystal boule is pulled from
the melt and is allowed to cool as it is being pulled. By annealing the
upper part of the boule and finally subjecting it to a controlled cooling
rate, this "zone engineering" ensures that the reactions that produce
either the void or dislocation cluster defect are completely suppressed.
The result is crystals that are completely free of both void and dislocation
cluster defects. Eliminating one or the other can be readily achieved.
Crystals have been made that are completely free of both classes of defects
and have earned this material the name "perfect silicon,".

At present, defect-free silicon crystals have been achieved at only at

diameters of 200 mm. Comparisons of crystal quality were made among
three techniques: a typical conventional Czrochralski crystal growth
324 Measuring particle size and growing single crystals

technique, a slow-cooled controlled reaction and the "perfect silicon"

process. The quality levels achieved in D-defect levels of the material is
mirrored in the gate oxide integrity of these materials.

j . Actual Imperfections of Crystals Grown from a Melt

Now we can summarize all of the imperfections likely to appear in crystals

grown from a melt. Some of these, particularly stacking- faults, were
discovered only when single crystals were grown large enough so that the
deviation from long range order became apparent, as we discussed for the
case of silicon.

The intrinsic defects found in crystals include vacancies, interstitials,

impurities and impurity compensations, reverse order, and combinations
such as V- S and I- S, etc. Their numbers are well described by:

4.2.50.- Nt = N 0 exp - A E i / k T

where i refers to the intrinsic defect. In addition, we have dislocations,

both edge (line) and spiral (screw), the latter being three- dimensional.
Their numbers are well described by:

4.2.51.- AGi = nAHi - T (ASconfig + n ASyib)

This equation arises because both of these extrinsic defects affect the
energy of the crystal. We can also have grain boundaries which may be:
clustering of line defects or mosaic blocks. The latter may be regarded as
very large grains in a crystallite.

Another imperfection in crystals is called "twinning". This usually happens

when enantiomorphs are present, or possible. A good example is quartz,
i.e.-

4.2.52.- a-quartz % p- quartz

573 °C
 4.2 Growth of single crystals 325

In this case, two crystals grow and are joined at a given plane, each being
a mirror image of the other.
The other crystal imperfection we have not covered is "stacking faults". A
good example is SiC. Here, we have two sublattices, one based on Si and
the other on C, each of which is hexagonal. In stacking alternate layers of
identical atoms, we first stack Si and then C. If we refer to "A" as the 1st
layer, "B" as the 2nd layer, and "C" as the 3rd layer, then the normal
stacking sequence for the hexagonal lattice is:

4.2.53.- ABC- ABC- ABC

where we have shown three sequences. For SiC, "A" in the first sequence
is Si, "B" is C, and "C" is Si. We still have the same packing since only "A"
atoms are over "A" atoms, as shown in the following diagram:

4.2.54.-
ISTACKING SEQUENCES IN THE HEXAGONAL LATTICEI

LAYERS

A A A A A A A
c c c c c c
B B B B B B
A A A A A A A
C C C C C C
B B B B B B
A cjA^ A A A A A
a
o

Here, we have arranged the layers on a two-dimensional structure, even

though the layers are arranged in three dimensional order. Note that only
two crystallographic axes are indicated, We call this the natural stacking
sequence because of the nature of the hexagonal close- packed lattice.

This behavior is more common than one might think, particularly those
having the hexagonal crystal structure. Thus, in addition to point defect
deficiencies in the lattice, we also may have stacking defects, i.e.- line
326 Measuring particle size and growing single crystals

defects and spiral defects, i.e.- growth around a line defect. However, SiC
also exhibits other stacking sequences, as shown in the following:

4.2.55. p - SiC s ABC ABC

SiC - 4H = ABCA ABCA
SIC - 6H = ABCACB ABCACB
SiC - 15R = ABCBACABACBCACB ABCBACABACBCACB

These arranged layers are called "polytypes" and are prevalent where
simple compounds such as SiC and SiN are involved. In many cases, the
properties of such compounds depend, to a large extent, upon t h e
specific stacked layers obtained during formation. For SiC, we can also
have "polytypes" where two stacking sequences like 4H - 6H can combine
to form a unit. Another "polymorph" is 4H -15R. This phenomenon has
been thoroughly studied and polymorphs of 87R and 270R have been
reported.

Another type of stacking fault is called "polystructure". A good example is

ZnS, which is dimorphic (has two forms). The cubic form of ZnS is called
sphalerite, whereas the hexagonal form is called wurtzite {These are
mineral names, after the first geologist who discovered them). The
stacking sequence for sphalerite is: AB or ABBA; that for wurtzite is: ABC.
A polystructure sometimes results when sphalerite is converted to
wurtzite:
~ 1200 °c.
4.2.55.- ZnS cubic = ZnS hexagonal

The polystructure sequence is: ABC AB •ABC--AB ABC. This may

be regarded as cubic- close- packing within a hexagonal- close- packed
structure. We may also note that ZnS is subject to the same "polytype"
stacking faults as those given for SiC above.

Thus, the stacking fault patterns noted are a function of the type of lattice
involved, not on the chemical composition of the material. For the most
part, these stacking faults are found only in the high symmetry lattices
such as hexagonal close-packed and cubic close-packed structures.
 4.2 Growth of single crystals 327

k. Crystal Growth from a Liquid Phase

You will note that we have formed a liquid phase for growing a single
crystal by melting the component(s). There are other ways of growing
single crystals which use a liquid phase which does not have the same
composition as that of the crystal we wish to obtain.

1. Growth of Single Crystals From a Molten Flux

This method is related to single crystal growth from a melt in that it

employs a molten flux which dissolves the material and redeposits it upon
a selected substrate. That is, the molten flux acts as a transport medium.
The temperature of the flux can be varied to suit the material and to
promote high solubility of the solute material in the molten solvent.
However, the method is limited in that one is restricted by the
temperature at which one can use the molten flux. One example is "YIG",
yttrium iron garnet, i.e.- Y3FesOi2. This material is used in the
Electronics Industry as single crystals for microwave generating devices.
It can be grown via the molten flux method. A typical molten-flux
apparatus is shown in 4.2.56., as follows on the next page.

The apparatus is simple and consists of a furnace whose temperature can

be accurately controlled, a crucible to hold the molten flux, and a method
of rotating the crucible to stir the flux containing the dissolved material.
Single crystals can be induced to grow along the sides of the crucible by
controlling the temperature and supercooling of the molten flux.

The steps involved in using this method are as follows:

1. A flux such as lead borate is melted. PbB2O4 is useful in this
method because it will undergo supercooling rather easily.

2. The material which is to form single crystals is dissolved in the

molten flux to near saturation (Note that this requires prior
knowledge regarding solubility of compound in molten flux).
328 Measuring particle size and growing single crystals

3. The crucible, which is usually platinum, is rotated to obtain a

uniform temperature distribution within the melt- compound
solution.
4.2.56.- A Molten-Flux Apparatus

4. The solution temperature is gradually lowered to incipient

nucleation. At this point, because of the physical arrangement of the
heating elements, a temperature gradient will exist along the length
of the crucible, from top to bottom.

5. Single crystals will begin growing along the bottom of the

crucible, and sometime later along the edges.

6. The crucible is kept rotating to maintain a uniform mixing and

heat flow while the crystals are growing.
 4.2 Growth of single crystals 329

The crystal-growing parameters for this method are given as follows:

4.2.57.- Molten Flux Growth Parameters

1. Flux melt temperature

2. Solubility of MX in BX2
3. Degree of supercooling
4. Temperature gradient achieved
5. Rate of rotation used

This method has serious deficiencies for use as a general method. We find
that if we dissolve MX in a BX2 flux and grow an MX crystal, it is likely to
be contaminated with B, or even EK2.

Naturally, the flux to be used depends upon the crystal to be grown. Many
times, it is a matter of trial and error to establish the proper flux needed.
The two most important parameters are: the degree of supercooling and
the temperature gradient achieved. If these are not within the correct
range, one does not obtain single crystal growth. This method has been
used in the past only because of its relative simplicity of apparatus and
materials. However, the quality of crystals so-produced has been rather
poor. Crystals produced by this method are suitable for structure
determinations, but are poor in optical quality and are not suited for
electronic applications.

Nonetheless, there is one area where molten flux growth has been very
successful. That is epitaxial growth of a single crystal film on a substrate,
such as that used for production of bubble memories. This is shown in
the following, given as 4.2.58. on the next page.

In this case, we use molten lead tetraborate (PbB4O7), which melts at

about 960 °C. The melt is raised to 980 °C. and the requisite oxides are
dissolved therein The temperature is lowered to near 960 °C. and
supercooling begins. It should be obvious that the degree of supercooling
that the melt will undergo before the dissolved oxides precipitate out
330 Measuring particle size and growing single crystals

depends upon the amount of oxides dissolved. This is generally

determined fay trial and error.

4.2.58.- Epitaxial Growth on a GGG Crystal Wafer

We lower the temperature to just above that point. A substrate slice of

single- crystal GGG (gadolinium gallium garnet = Gd3Ga5 O12) is then
submerged in the molten flux, while being rotated.

Since the substrate is cooler, a garnet film will grow on its surface under
these conditions. The surface of the GGG slice will have been carefully
polished to minimize surface defects. It takes about 6 minutes to grow a
single-crystal film about 50 - 100 [x thick. The film grows in an epitaxial
manner, that is- it builds up on the crystallographic planes of the
substrate itself.

In this case, purity is another parameter for epitaxial growth in addition

to those given above. The purity of the film grown, and its intrinsic
defects, affects the operation of the bubble memory film. Obviously,
impurities cannot be tolerated.

2. Hvdrothermal Growth of Crystals

Single crystal growth by hydrothermal methods utilizes water as the

material-transport medium. The method is most often used for growing
single crystal quartz. Quartz (SiO2) is not very soluble in water, but its
 4.2 Growth of single crystals 331

solubility increases considerably at higher temperatures. Thus, growth is

accomplished at high pressures in a sealed autoclave. These quartz
crystals are used as resonant frequency "tuning forks" for timing
applications in wrist watches. As shown in the following, seed crystals are
hung within an autoclave on a revolving hanger.

Nutrient (high purity sand or natural crystal quartz) is contained at the

bottom and dissolves as the autoclave is heated. A temperature gradient is
most often used so that the nutrient dissolves and is transported to the
cooler area where it is redeposited as single crystal material. Note the
temperature gradient at the baffle.

The solubility of the solute (in this case, quartz) is a function of both
pressure and temperature. Pressure could be in theory be used as the
controlling parameter rather than temperature. However, it is difficult to
332 Measuring particle size and growing single crystals

design an apparatus with a pressure gradient, whereas obtaining a

temperature gradient is fairly easy.

Actually, what is controlled is the critical volume from the equilibrium:

4.2.60.- H2O (g) «> H2O(i)

where 1 and g refer to the liquid and gaseous state, respectively.

The phase diagram for water is shown at the lower left of the following
diagram:

4.2.61.-

Above the critical temperature, water exists as a gas at all pressures. But,
if we fix the volume of both gas and liquid (as in an autoclave), then no
liquid is possible below the critical volume (~ 30 % by volume). By filling
the available volume greater than 30% with water, we can go to very high
pressures (i.e.- > 3000 atmospheres) and still maintain a liquid volume.

In hydrothermal growth, the materials usually grown as single crystals are

those classified as insoluble at standard temperature and pressure (STP).
For quartz, growth is usually accomplished at 85% fill and 2000
atmospheres at 350 °C. The parameters for hydrothermal growth of single
 4.2 Growth of single crystals 333

crystals are given on as follows:

4.2.62.- Parameters Controlling Hydrothermal Growth of Single Crystals.

1. Operating Pressure
2. Operating Temperature
3. Solubility in super-critical water
4. Degree of supersaturation
5. Degree of supercooling
6. Purity of Nutrient

The last three factors require some further consideration. We will discuss
these in light of the growth of quartz, since it is this crystal for which the
most experience has been gained. The rate of crystal growth is a function
of seed-crystal orientation. The rate of growth, rhkl . may vary several
orders of magnitude, depending upon the {hkl} plane orientation.
However, a certain degree of supersaturation is mandatory as the nutrient
dissolves, passes through the baffle, and moves to the seed area.
Otherwise, the nutrient would precipitate before reaching the seed
crystals. It has been determined that the rate of crystal growth is a direct
function of the degree of supersaturation, A Sat., as follows:

4.2.63.- rhkl = a • khkl • A Sat

where khkl is the seed orientation factor and a is a constant dependent

upon the crystal system. However, it has been found that the degree-of-
supercooling factor is contra-indicative to the growth rate. To understand
this, imagine the face of a growing crystal. Rejection of impurities at the
growing interface causes a localized increase of impurity concentration.

This is shown in the following diagram, given as 4.2.64. on the next page.
If the superimposed temperature gradient is T a , i.e.- dT/dl is a constant,
and T e is the equilibrium curve, i.e.- dC/dT e , then supercooling will
occur. The rate of deposition thereby changes. If it becomes slower,
crystal growth slows whereas if it speeds up, then dendritic growth or
334 Measuring particle size and growing single crystals

faceting occurs. Note that the impurity- rejection mechanism, observed

for the case of Molten Flux growth, also occurs here.

4.2.64.-

For all these reasons, purity of materials is one of, if not the, most
important of the parameters controlling hydrothermal growth.

While we have discussed how to produce quartz crystals by hydrothermal

growth, we have not addressed the applications in which they are used.
Two very important technical fields use quartz crystals. These are: as
frequency oscillators and as thin-film monitors. Both of these technologies
take advantage of the fact that quartz can be act as a crystal oscillator by
applying a controlled voltage.

Piezoelectricity was discovered in 1880 by Pierre and Paul-Jacques Curie,

who found that when they compressed certain types of crystals including
quartz, tourmaline, and Rochelle salt, along certain axes, a voltage was
produced on the surface of the crystal. The next year, they observed the
converse effect, the elongation of such crystals upon the application of an
electric current.

Some solids, notably certain crystals, have permanent electric

polarization. Other crystals become electrically polarized when subjected
 4.2 Growth of single crystals 335

to stress. In electric polarization, the center of positive charge within an

atom, molecule, or crystal lattice element is separated slightly from the
center of negative charge. Piezoelectricity (literally "pressure electricity")
is observed if a stress is applied to a solid, for example, by bending,
twisting, or squeezing it. If a thin slice of quartz is compressed between
two electrodes, a potential difference occurs. Conversely, if the quartz
crystal is inserted into an electric field, the resulting stress changes its
dimensions. This was the basis for the discovery that quartz resonators
could control the frequency of an electrical oscillating circuit.

The operating frequency of a particular resonator, usually called a

"thickness-shear" resonator, is primarily determined by the thickness of
the crystal plate, i.e.- the thinner the plate, the higher is the frequency of
operation. But, a thin rectangular plate with large lateral dimensions is
subject to breakage. This led to larger and thicker flat plates as large as
1.0 inch on a side. Obviously, the dimension of the original quartz crystal
presented a problem, particularly if the crystal was limited in size. The
early market was for commercial radio stations and ham operators in the
1920's and 1930's. Crystal control was used by the Military in the 1940's
for controlling Radar. As the demand for higher frequencies arose, it was
found that round plates could withstand the stresses of high-frequency
flexion much better, and these became the standard at that time. With the
advent of larger hydrothermally grown quartz crystals with improved
lattice properties, i.e.- lack of growth-induced defects, this problem was
solved. Nowadays, the quartz resonators, driven by improvements in the
ever-expanding market for smaller-sized packages in surface-mount
technology for integrated circuits, are cut in highly defined modes from
the crystal. The terms "strip resonator", "AT strip resonator" and "BT
strip resonator" are commonly used to describe the low profile, surface
mount quartz crystals used in oscillator circuits.

To understand this terminology, consider the diagram given on the next

page as 4.2.65. Quartz exists in two forms: a-, or low, quartz, which is
stable up to 573 °C (1,063 °F), and p-, or high, quartz, stable above 573 °C.
The two are closely related with only small movements of their
constituent atoms during the alpha-beta transition.
336 Measuring particle size and growing single crystals

The structure of a-quartz is trigonal, with either a right- or left-handed

symmetry group.

The structure of beta-quartz is hexagonal with again either a left- or right-

handed symmetry group, equally populated in crystals. At the transition
temperature, the localized tetrahedra, i.e.- SiO4 , of high-quartz twist,
resulting in the lower symmetry of low-quartz. That is, atoms move from
special space group positions to more general positions.

What is shown in the above diagram of 4.2.65. is an alpha-quartz crystal,

oriented along the y-axis. The z-axis is at 90° while the x-axis is
perpendicular to the field of view. Y-cut plates are cut from the crystal in
away so that the thickness dimension is parallel to the y-axis. In this way,
the resonant frequency of a Y-cut plate is primarily determined by its
thickness. However, Y-cut plates exhibit some undesirable traits, such as
temperature instability and fragility, that make them unsuitable for
commercial use. So most plates today are cut at specific angles to the z-
axis (the y-axis is the major mechanical direction in the trigonal lattice) as
shown for the AT and BT cuts. This results in plates having more
desirable traits of frequency stability versus temperature. The difference
in AT and BT cuts are nearly opposite in crystal lattice orientation and
results in substantial differences in their operating characteristics. The
term "resonator" simply means a crystal plate sandwiched between two
electrodes. Most of today's demand for surface-mount resonators is for
these AT and BT crystals because they can resonate at very high
frequencies.
 4.2 Growth of single crystals 337

The following diagram shows the main types of vibrational modes that can
be induced in a crystal resonator plate:

AT and BT plates are made to vibrate primarily in the thickness shear

mode. However, it is important to realize that quartz crystals can be made
to vibrate in any one or combination of these modes. The vibrational
modes can be induced electrically, acoustically, thermally or by some
combination of all of these factors. The thickness shear frequency
response of an AT plate can be described in terms of the fundamental
frequency. For a rectangular AT cut plate, the equation for calculating the
approximate frequency of vibration is:

4.2.67.- fm,n,p = 0.5 (l/p) 1 / 2 {Cm nrVT2 + C u m2/L2 + G^ m2AV2 } 1/2

where m, n and p are integers, p is the density of quartz (2.65 gm/cc), T

is the thickness in mm, W is the width in mm, and Cm , C n and C^ are the
elastic constants of quartz in the x, y and z directions. Ifm = n = p = l , we
have the fundamental frequency. However, it is easy to obtain unwanted
"overtones" such as third or fifth harmonics, i.e.- (3,1,1) or (5,1,1)
overtones. Overtones are suppressed by careful sizing and cuts of the
crystal plate.

The operation of a crystal quartz resonator can be explained as follows:

338 Measuring particle size and growing single crystals

4.2.68.- Equivalent Circuit for a Resonating AT-cut Quartz Plate

The actual resonator is made by attaching electrodes to opposite sides of

the quartz plate. The metal can be evaporated upon the surface or a silver
paste attached. Application of a voltage and current causes the crystal to
vibrate. The fundamental frequency can be estimated from the induction,
capacitance and resistance of the crystal plate in an operating mode, i.e.
e
^CRYSTAL > ^CRYSTAL AND RCRYCTAL • Th crystal is assumed to be vibrating at
a specific frequency and order of overtone. C CIRCUrr is the capacitance of
the circuit to which the resonator is electrically connected. Then:

4.2.69.- f = 1/2* (LCRYSrAL . CcRYCTAL ) 1/2

where L and C are the motional inductance and capacitance during

vibration as a resonator. L is in millihenries, C is in picofarads and f is in
mega-hertz.

The other application of quartz resonators is the quartz-microbalance.

These devices monitor layers of films deposited by changing frequency as
the layer adds thickness to the quartz plate. Since the resonant frequency
entirely depends upon the thickness, i.e.- mass, of the plate, the quartz
resonator is uniquely suited for this application. Although an anti-
reflection coating of MgF2 was sufficient 20 years ago, nowadays current
designs call may call for a 24-layer stack of films with alternating
refractive indices. With high speed optical communication systems, the
required stack may consist of up to 256 layers, all of which must be
monitored as to thicknesses deposited sequentially.

Quartz resonant sensors measure film thickness by monitoring change in

fundamental frequency of the quartz microbalance. The sensitivity is
 4.2 Growth of single crystals 339

remarkable. A uniform coating of as little as 10 A of aluminum will cause a

change of 20 mega-hertz in frequency, easily measured by today's
electronics. The useful life of this quartz sensor depends upon the
thickness and type of coating monitored. If a low stress metal such as Al is
deposited, layers as thick as 1,000,000 A have been measured. At the
other extreme, high stress dielectric films such as MgF2 will reach less
than 2000 A before the crystal malfunctions. What this means is that the
quartz microbalance sensor is incorporated within the thin-film coating
chamber, as shown in the following diagram, and is subjected to repeated
coatings until the weight builds up to the point where it cannot no longer
vibrate at its fundamental or overtone frequencies:

4.2.70.-

You will note that we have presented only a bare outline of the
components in this diagram. We will address these techniques in more
detail in a later chapter on thin film technology and its ramifications to
the advancement of our "Information Age".

The quartz microbalance has some short comings in its use as a thin-film
monitor. Frequency shifts can be positive or negative, depending upon the
film being deposited. When coatings are added to the crystal surface, the
resonant frequency decreases linearly. If the coating is removed, the
340 Measuring particle size and growing single crystals

resonant frequency increases. The following are a summary of the factors

which affect the resonant frequency of the microbalance when used as a
thin-film monitor:

4.2.71.- Causes of Resonant Frequency Changes in a Quartz Sensor

1. Real time additions of a thin film from the coating source

2. Stresses induced from deformation by deposited high
stress films like MgF2 and SiO2.
3. Vibrations induced through the mounting hardware.
4. Variations in the voltage used to oscillate the crystal.
5. Chemical changes in the film being monitored as it is being
deposited
6. Adhesion failure of the monitored coating on the quartz
electrodes.
7. Radiofrequency interference in the monitoring circuit.
8. Temperature changes in the crystal when high temperature
coatings are being deposited.
9. Splatters of material from a faulty coating source as it is
being ablated by the various means used to do so.
10. Ablation of the crystal surface by high-energy plasmas used
to clean the substrate before film deposition.

All of these contribute to early crystal failure. In the case of thin-film

stacks of 100 or more, the coating chamber cannot be opened to the air
before the coating is completed. If the crystal fails, then the process is
ruined and will have to be started again with a new crystal sensor. The
latest efforts to prevent early crystal failure are beginning to bear fruit, but
we will not address these here at this point

1. Vapor Methods Used for Single Crystal Growth

If the material whose single crystal we want is volatile or sublimable, then

we may choose a vapor-method of crystal growth. These methods have
been used for a variety of crystals including ZnS and CdS. In this method,
a carrier- gas is most often used for material transport and for the
 4.2 Growth of single crystals 341

sulfides, H2S is the gas of choice. The following diagram, given as 4.2.72.
on the next page, shows a simple apparatus to grow crystals by the vapor
transport method:

4.2.72.- A Furnace for Growing Crystals from the Vapor Phase

In this method, we start the gas flow and then begin heating. It is
important to heat the material to just below the point of sublimation
(volatilization) and let the system come to equilibrium. Gas flow need not
be rapid but needs to be sufficient to carry the volatiles to the cooler part
of the furnace. The tube used in the furnace is most often a silica-tube,
although metal tubes have sometimes been used. The choice depends
upon the nature of the material being sublimed and crystallized.

For ZnS and CdS, it is important to exclude all traces of oxygen since
these materials are easily oxidized, namely-

4.2.73.- 2 ZnS + 3 O2 = 2 ZnO + SO2 t

The operating parameters for the vapor phase method of growing crystals
are shown in 4.2.74., presented on the next page.

Obviously, whether the material has a low (high vapor pressure) or a high
temperature of sublimation is important because of furnace- construction
material considerations.
342 Measuring particle size and growing single crystals

4.2.74. - Operating Parameters for Vapor Phase Growth of Crystals

1. Temperature of sublimation (volatilization) of material

2. Amount of gas flow used
3. Degree of furnace- temperature set- point above material
sublimation temperature.
4. Temperature gradient at crystal-growing junction of tube.

In general, we cannot use this method for materials which are volatile
above about 1200 °C. because most materials of construction cannot
withstand the corrosive nature of the vapors at this high temperature. If
we set the temperature of the furnace above the sublimation point of the
material, it will volatize all at once. This would necessitate setting the gas-
flow such that all of the material is transported by the gas. It would be
better to set the temperature just below the vaporization (sublimation)
point. There will be enough material transported to begin growth of single
crystals. In fact, it has been determined that the slower the growth, the
better are the crystals obtained.

There is another method that has been sometimes employed in the vapor
phase growth of crystals. This method uses an evacuated capsule as
shown:

4.2.63- Another Method Sometimes Used for Vapor Phase Growth

The capsule is generally made from quartz, although platinum is

 4.2 Growth of single crystals 343

sometimes used. The capsule needs to be evacuated to remove any

residual gas before heating is started. Otherwise, the internal pressure
would build until the capsule would explode. Even though it is evacuated,
capsules have been known to explode because the quartz (metal) walls
could not contain the internal vapor pressure of the material being grown
as single crystal. Care must be exercised not to handle the capsule hot
before and after crystal growth.

In addition to sublimation or vaporization, we can also use chemical

transport as a method of single crystal growth. For example, we could use
the apparatus of 4.2.60. to grow crystal of ZnCl2 by the following
reactions:

4.2.64.- ZnO + 3 HI = Znl 2 ft + H2O

Zn + I2 = Znl2 ft

Metal carbonyls are also convenient for growth of certain METAL single
crystals, as shown in the following:

4.2.65.- M(S) + x CO(g) = M(CO(g))x

M(CO(g))x = M(S) +xCO(g)

In these cases, one is limited to the growth of single crystals of the

transition metals and those metals which can form volatile carbonyl
compounds. The alkali metals, alkaline earth metals and certain of those
which are allotropic in nature are not at all suited for this type of crystal
growth.

Suggested Reading

1. R.R. Irani and Clayton F. Callis, "Particle Size- Measurement,

Interpretation and Application" - J. Wiley & sons, New York (1963).
2. John Wulff et al, "The Structure and Properties of Materials - Volumes
I, II, III, & IV" - J. Wiley & Sons, New York (1964).
3. Polakowski and Ripling, "Strength and Structure of Engineering
Materials" - Prentice-Hall, Englewood Cliffs, NJ (1966).
344 Measuring particle size and growing single crystals

4. R.E. Newnham, "Structure-Property Relations" - Springer-Verlag, New

York (1975).
5. "Optical Microscopy"- General Test (776), Vol. USP 24, pp. 1965-67,
Publ. By The United States Pharmacopoeia Society, Rockville, Md. (2000)
6. Robert E. Reed-Hill, "Physical Metallurgy Principles"- Van Nostrand,
Princeton, New Jersey (1964).
7. A.J. Dekker, "Solid State Physics" - Prentice-Hall, Englewood Cliffs,
New Jersey (1958).
8. Perelomova & Tagieva, "Problems in Crystal Physics"- MIR Publ.,
Moscow, Eng. Transl. - (1983).
9. W.W. Wendlandt, Thermal Methods of Analysis, Interscience-Wiley,
New York (1964).
10. P.D. Garn, Thermoanalytical Methods of Investigation, Academic
Press, New York (1965).
11. W.J. Smothers and Y Chiang, Handbook of Differential Thermal
Analysis, Chemical Publishing Co., New York (1966).
12. E.M. Barral and J.F. Johnson, in Techniques and Methods of Polymer
Evaluation, P. Slade & L. Jenkins- Ed., Dekker, New York (1966).
13. W.W. Wendlandt, Thermochim. Acta 1 , (1970)
14. D.T.Y. Chen, J. Thermal Anal, 6 109 (1974) - Part I
15. Chen & Fong, loc. cit. 7 295 (1975) - Part II.
16. Fong & Chen, loc. cit., 8 305 (1975) - Part III
17. C. Duval, Inorganic Thermogravimetric Analysis, 2nd Ed., Elseveir,
Amsterdam (1963).
18. C. Keattch, An Introduction to Thermogravimetry, Heydon, London
(1969).
19. "Vapor Pressure Determination using DSC", A. Brazens, R.B. Cassel,
C.W. Schaumann & R. Seyler, Proceeding. N. Am. Therm. Analysis Soc.
22nd Conf. Denver, Colo. (1993).
 Optical and Electronic Properties of Solids

In order to fully understand the nature of phosphors, it is necessary to

consider light and the optical behavior of solids, i.e.- the interaction of
solids with electromagnetic radiation, including the resonant transfer of
absorbed (excitation) energy from one site to another in a solid. We will
emphasize phosphors and discuss theories of electronic and vibrational
properties of solids as well as solid state luminescence mechanisms in
detail. This will establish the groundwork for a specific discussion
concerning the design of phosphors and how phosphors are prepared and
used in the next chapter. We will then enumerate the many phosphors
that have been developed for myriad purposes in the following chapter,
particularly those in use since 1991, the date of publication of the first
printing of this treatise.

5-1 THE NATURE OF LIGHT

In the early 17th century, the nature of light was not well understood.
Nowadays, we know that light is comprised of "photons", which are
quantized waves having some of the properties of particles. However, this
concept was not made clear until the advent of quantum electrodynamics
in the 1930's, following Einstein's concept of relativity in the 1920's.

Such was not the case originally. Beginning with Kepler's Paralipomena in
1604, the study of optics had been a central activity of the scientific
revolution in Europe. Descartes' statement, in 1634, of the sine law of
refraction, relating the angles of incidence and emergence at the
interfaces of the material through which light passes, had added a new
mathematical understanding to the science of light. Descartes had also
made light central to a description of the mechanized theory of nature.

The concept of photons with wave properties has its roots in the study of
optics and optical phenomena. Until the middle of the 17th century, light
was generally thought to consist of a stream of some sort of particles or

345
346 Optical and electronic properties of solids

corpuscles emanating from light sources. Newton and many other

scientists of his day supported the idea of the corpuscular theory of light.
It was Newton in 1664 demonstrated that sunlight could be separated
into colors. He also showed in 1703 that "ordinary" light, such as that
from aflame, could be dispersed into its constituent colors by a prism, but
the phenomenon was not clearly grasped at that time.

About the same time, the idea that light might be a wave phenomenon
was proposed by Huygens and others. Huygens demonstrated that every
point on a wavefront may be regarded as a source of spherical wavelets,
the envelope of which accounts for the position of the wavefront at a later
time. Huygens was thus able to explain rectilinear propagation and those
involving the laws of reflection and refraction. Fresnel added to this
hypothesis by showing that the wavelets can interfere, and this led to a
theory of diffraction. However, Newton still regarded light as a stream of
particles which were either refracted or reflected.

It remained for Lord Rayleigh in 1887 to explain the wave-nature of light

by considering such subjects as electromagnetism, color, acoustics, and
diffraction gratings. His most significant early work was his theory in
1871 explaining the blue color of the sky as the result of scattering of
sunlight by small particles in the atmosphere. The Rayleigh scattering
law, which evolved from this theory, has since become classic in the study
of all kinds of wave propagation. Indeed, diffraction effects that are now
known to be associated with the wave nature of light were observed by
Grimaldi as early as 1665, but the significance of Grimaldi's observations
were not understood at the time.

It was thus that many scientists described light as a beam of particles

whereas others regarded it as a series of waves. This disagreement
remained until the 19th century because various early investigators who
had experimented with light had shown that light could be reflected or
diffracted by optical means. Some interpreted their observed behavior as
that expected from a beam of particles while others felt that their data
could only be explained by assuming wavelike behavior for a beam of light.
Those who made the most significant contributions were Newton,
 5.1 The nature of light 347

Descartes, Huygens and Rayleigh. Although other scientists made

additions to the total body of knowledge of optics as well, these four
established the major principles regarding propagation and interaction of
electromagnetic radiation with matter. Later contributors to our overall
understanding of the nature of photons included:

1) Fresnel and Thomas Young (1815) on interference and diffraction

respectively
2) Maxwell in 1873 who postulated that an oscillating electrical
circuit should radiate electromagnetic waves
3) Heinrich Hertz in 1887 who used an oscillating circuit of small
dimensions to produce electromagnetic waves which had all of the
properties of light waves
4) Einstein in 1905 who explained the photoelectric effect (He did
so by extending an idea proposed by Planck five years earlier to
postulate that the energy in a light beam was concentrated in
"packets" or photons.. The wave picture was retained in that a
photon was considered to have a frequency and that the energy of a
photon was proportional to its frequency).

Experiments by Milliken in 1908 soon confirmed Einstein's predictions.

In 1921, A.H. Compton succeeded in determining the motion of a photon
and an electron both before and after a collision between them. He found
that both behaved like material bodies in that both kinetic energy and
momentum were conserved in the collision. The photoelectric effect and
the Compton effect, then, seemed to demand a return to the corpuscular
theory of light. The reconciliation of these apparently contradictory
experiments has been accomplished only since about 1930 with the
development of quantum electrodynamics, a comprehensive theory that
includes both wave and particle properties of photons. Thus, the theory of
light propagation is best described by an electromagnetic wave theory
while the interaction of a photon with matter is better described as a
corpuscular phenomenon.

All bodies emit some form of electromagnetic radiation, as a result of the

thermal motion of their molecules. This radiation, called thermal
348 Optical and electronic properties of solids

radiation, is a mixture of wavelengths. At a temperature of 300 °C, the

most intense of these waves have a wavelength of 50,000 A which is in the
far infrared region of the electromagnetic spectrum. When the
temperature is raised to about 800 °C, such a body emits enough visible
radiation to be self-luminous and appears "red-hot". However, by far the
most energy is still carried by photons having wavelengths in the infra-red
region of the spectrum. But at 3150 °C, which is the temperature of a
tungsten filament in an incandescent light bulb, the body then appears
"white-hot" and a major part of the energy is in the visible region of the
spectrum. The following diagram, given as 5.1.1. on the next page, shows
the emission of radiation as a function of "black-body" temperatures. A
"black-body" can be simulated by a sphere internally coated with a "black"
surface and having a small hole for light to escape. One then observes the
wavelengths that escape as a function of the internal temperature of the
sphere. For the highest temperatures, the curves are derived from a fit of
the equations which correspond to lower temperatures.

Note that even bodies at liquid-air temperatures emit photons between 10

and 100 microns in wavelength, i.e.- 100,000 and 10 6 A in wavelength.
The earth itself at a temperature of 300 °K. has an emission between
about 20,000 and 300,000 A in wavelength, i.e.- 2000 nm. and 30,000
nm. What this means is that even you emit infra-red wavelengths as a
consequence of being warm. Note that this is the basis of "night-vision"
headgear which can detect human bodies even when it is so dark that
other objects are invisible to the naked eye.

In vacuum, all electromagnetic radiation travels at the speed of light. This

is given by:

5.1.2.- c = {l/eO(xo}1/2

where e o is the permittivity of free space, and \IQ is the permeability of

free space. The former comes from Gauss's Law and the latter from
Faraday's Law. However, the speed of light in media other than vacuum is
always slower than in space. This is believed to be due to resonance
interactions between the electromagnetic fields of the electrons
 5.1 The nature of light 349

associated with the transparent media and that of the traveling photon.
When electrons or other charged particles are accelerated to relativistic
speeds, i.e.- a fraction of the speed of light, they emit photons as they
travel in a transparent medium, i.e.- Cerenkov effect. Thus, it should be
clear that both photons and particles of mass are inexorably interwoven by
the matrix of space-time. What this means is that when particles having a
given "rest" mass are caused to accelerate near to a limiting speed, i.e.-
the speed of light, they are prone to release that excess energy gained
350 Optical and electronic properties of solids

through the emission of photons. Einstein was the first to realize this
phenomenon, which has since been proven many times over.

Because particles, i.e.- electrons, have wave properties similar to those of

photons, we need to differentiate between them. It was de Broglie in
1906 who first postulated the wave nature of particles. In 1927, Davisson
and Germer first showed that electrons are reflected from the surface of a
solid in the same way that x-rays are reflected. The wave hypothesis
clearly required sweeping revisions of our fundamental concepts
regarding the nature of matter. The best explanation to date seems to be
that a particle must be regarded as an entity not entirely localized in
space whereas the photon is a point-source, i.e.- is localized in space.
What this means is that an electron is strongly attracted or repelled by
electromagnetic fields as it moves through space whereas a photon, being
a localized point source, is only weakly affected.

Thus, a photon moves at a constant speed through space while an electron

does not. Yet, both have electromagnetic fields thereby associated with
each, which are subject to reflection, diffraction etc. The photon-packet
thus interacts with the space-time continuum as it moves through space
at a constant speed. In contrast, the particle interacts with both the time-
space continuum and the electromagnetic fields thereby associated with
mass, and its speed is not constant, but subject to mass-mass (gravity)
interactions as well. The wavelength of the photon (which has no mass) is
a function of its internal energy as it moves through space. The particle
has a mass which is determined by how much it is spread out in the
space-time continuum. Consequently, it has properties we normally
associate with "mass".

From the above, it should be clear that a photon is a "force-carrier" while

an electron is a "matter-constituent". The notion of particles as mediators
of force in nature has provided a framework for testing and developing the
Standard Model and the associated "Big-Bang" theory of the formation of
the Universe. It has also been important for the exploration in depth of
many other important questions about the physical world. The most
familiar interaction is Electromagnetism. Electromagnetic radiation in its
 5.2 Absorbance, reflectivity and transmittance 351

various forms (including radiowaves, microwaves, infra-red-light, visible-

light, ultraviolet-light, and x-rays) can be thought of as the exchange of
massless photons between electrically charged particles, either quarks or
leptons. We can summarize all of the above in that a photon is a quantum
of radiation, whereas an electron is composed of matter-quanta.

5.2.- ABSORBANCE, REFLECTIVITY AND TRANSMITTANCE.

When a beam of photons strikes a solid, specific interactions take place

which are related to the Quantum Theory. These interactions have been
measured in the past and certain formulas have been found to apply.
According to HUYGHEN'S principle of electromagnetic radiation
scattering, when photons come into close contact with a solid, the
electric and magnetic field vectors of the incident photons couple with
those of the electrons associated with the atoms comprising the solid.
This , of course, depends upon the nature of the solid. The radiation can
be reflected, transmitted or absorbed fully. In the case of absorption, the
energy of the photon changes the energy of the atom or molecule in the
solid, resulting in heating at the site of absorption. In transmission of the
photon through the solid (assuming it is transparent to the photonic
wavelength), no interaction occurs. In reflection (scattering), the photon
can undergo either an elastic or inelastic collision with the atoms of a
solid. In the former mechanism, the wavelength is unchanged, whereas
the inelastic collision changes the wavelength of the photons. That is, part
of the energy of the photon is absorbed at the site of atom or molecule of
the solid, resulting in an "excited" state in which an electron is promoted
into an upper energy level. If the wavelength of the emitted photon is not
changed, then the photon is "scattered" and the reflection is an elastic
collision.

This is the basis for the so-called Rayleigh scattering in which blue-light is
scattered while all of the other colors of visible light result in inelastic
collisions in which the wavelength is changed. Rayleigh showed that the
intensity of the scattered light was proportional to the wavelength. We
now know that this is due to the Raman scattering effect. That is, the
atom or molecule absorbs part of the energy via vibrational (phonon)
352 Optical and electronic properties of solids

interaction and refimits the photon with an altered wavelength. These

interactions have been measured in the past and certain equations have
been found to apply. The following gives some of these formulas applicable
to concepts given later in this Chapter.

5.2.1.- Formulas Applicable to Optical Properties of Solids

a. Absorbance : A = log 1 /T = log Io / 1

b. Transmittance: T = I / I o (I = measured intensity; I o is original)
c. Absorptivity: A = A / be (A is measured absorption; b is the
optical path-length; c is the molar concentration)
d.Reflection: Reflectivity = specular reflection, i.e.- at specific
angles
Reflectance = diffuse reflection, i.e.-scattered radiation
e. Intensity: I is defined as the energy / unit area of abeam of
photons, i.e.-electromagnetic radiation.

According to Huygen's principle of electromagnetic radiation scattering,

when photons come into close contact with a solid, the electric and
magnetic field vectors of the incident photons couple with those of the
electrons associated with the atoms comprising the solid. This interaction
leads to at least four (4) components, namely: R- the radiation reflected,
A- the radiation absorbed, T- the radiation transmitted, and S - the
radiation scattered. This mechanism is illustrated in 5.2.2., given on the
next page, as follows.

The original intensity of the radiation is defined as Io- A part of the

intensity is absorbed, another part is transmitted, still another part is
scattered, and a part of the total intensity is reflected. The components, S
and T, are processes which are independent of the wavelength
(frequency) of the incident photons, whereas R and A are mostly
wavelength dependent.

The exact amount of energy extracted from Io by each process is a

complex set of variables depending upon the type and arrangement of
 5.2 Absorbance, reflectivity and transmittance 353

atoms composing the solid. The processes, A & S, are related in that both
involve absorption of energy, but S is a resonance process.

5.2.2.- Photon Interactions within a Solid

That is, the same amount of energy absorbed in the scattering process is
regmitted by the atom(s). Contrast this to absorption, where apart of the
energy is changed into vibrational energy within the solid. We can
summarize these properties as:

5.2.3.- Interaction Parameters in Terms of Radiation Properties

S = f(L-S) -f(I)
A =f (X) • f (L-S) -f(I)
R =f (X) • f(0 ,S) -f (I)

where X is wavelength, (L • S) relates to electric field vectors of the

electrons within the atoms comprising the solid (i.e.- angular momentum
and spin vectors), I is the intensity (number of photons impinging per
second), 8 is the angle between the beam and the atomic lattice, and 6 is
the atom density. Once you have studied these equations, it becomes
obvious that in order to predict and calculate the optical properties of a
354 Optical and electronic properties of solids

solid, we need to know more about the discrete factors which control the
optical properties of that solid.

Consider an optically homogeneous thin film. By optically homogeneous,

we mean one that is thin enough so that no scattering can occur. If abeam
of photons is incident to the surface at a given angle (but less than that
where all of the beam is transmitted- the so-called Brewster angle), part
of the beam will be reflected and part will be absorbed. This is shown as
follows:

5.2 A.- Single Reflection in a Thin Film

The reflectance, R, is a consequence of the difference in refractive

indices of the two media, (1) - air, and (2)- the semi- transparent thin
film. The amount of absorption is a function of the nature of the solid. In
this case, A, the absorbance, is defined as:

5.2.5.- A = (1- R) - T

It is easy to see how this equation is derived from the diagram since T,
the transmittance is equal to (1- R) - A. Then, the original beam intensity,
Io = 1.00, is found to be diminished according to the Beer-Lambert law:

5.2.6.- A = lnlo/l = Eel

 5.2 Absorbance, reflectivity and transmittance 355

where 1 is the pathlength (width of film traversed), 8 is a molar extinction

coefficient of the absorbing species, and c is the concentration of the
absorbing species. Although we have shown but one reflection at X,
another is equally likely to occur at Y as well. This problem of multiple
reflections was worked out by Bode (1954) and we will use it to prove the
validity of the equation of 5.1.4. for the case of the homogeneous thin
film. First, we define total values in terms of intrinsic, or individual,
values for a homogeneous medium, as shown in the following:

5.2.7.- Intrinsic Values of R , A & T

R = Er T = 2 t = In I o / I A =2a

For the case of multiple reflections, Fresnel's equation may be used:

5.2.8.- r =

where r| is the refractive index and r is the intrinsic reflection at a single

plane. For multiple reflections, we have the following situation:

5.2.9.- Multiple Reflections in aThin Film

356 Optical and electronic properties of solids

We can show that r is reflected at each interface, and that only (1- r)
enters the medium, i.e.- the thin film, at each reflection. The total
transmission is a series function:

5.2.10.- 2 t = (1- r)2t + r2(l- r ) 2 t 3 + r 4 ( l - r ) 2 t 5 + r6 (l - r ) 2 t 7 + ..

and: 2 r = r + r(l - r ) 2 t 2 + r3 (l - r ) 2 t 4 + r5 (l - r ) 2 t 6 +

We can rearrange these equations to obtain:

5.2.11.- T = (1 - r ) 2 / ( I - r 2 t 2 )
R = r + {r (1 - r ) 2 t 2 } /{I - r 2 t2}
A = {(1 - r)(l - t)}/ (1- rt) = (1 - R) - T

which is the result we want. This exercise also shows the relation
between A, R, & T in greater detail. However, this derivation only applies
to an optically homogeneous thin film. The derivation of scattering
coefficients to obtain absorbance from reflectance values of a powder was
accomplished by Kubela and Monk in 1930. Several other approaches to
the particle scattering problem have also been taken, but we will not
address any of these at the moment.

Consider the energy interactions which occur when a photon strikes a

solid such as a phosphor. The timeframe of interaction is about 1 0 1 8
seconds, as is easily seen:

5.2.12.- Speed of photon = v= 10 1 0 cm./sec.

Distance between lattice planes = d = 10~8 cm. = ~ 1 A
Time for photon to traverse lattice = d/v= ~ 10 - 1 8 seconds

The fact that the photon does traverse the lattice planes does not mean
that the photon will be absorbed or even scattered by the solid. The
reflectance of the photon is a function of the nature of the compositional
surface, whereas absorption depends upon the interior composition of the
solid. A "resonance" condition must exist before the photon can transfer
energy to the solid (absorption of the photon). In the following, we show
 5.2 Absorbance, reflectivity and transmittance 357

in general terms the resonance condition of both R & A, and how this
relates to the perception of light as we see it.

5.2.13.- Energy Transfer to a Solid By a Photon

R A ENERGY TRANSFER EXAMPLES

High High Moderate "Colored" solid

Low High Very High "Black" solid

High Low Nil "White" solid

Low Low Nil "Transparent"

Note the difference between a "colored" solid and a "black" solid. In the
first case, only parts of the visible spectrum are absorbed or reflected. In
the latter case, all parts of the visible spectrum are absorbed and little or
none are reflected, hi the case of the "white" solid, all visible wavelengths
are reflected.

As an example of controlled absorption, consider the case of a pigment. It

is quite common to add controlled amounts of a transition metal to a
transparent solid to form an inorganic pigment. In one such case, we add
~ 1% of chromium oxide to aluminum oxide to obtain a pink solid, i.e.-
"ruby". The Cr 3+ ion in the AI2O3 lattice absorbs blue and green light and
reflects mostly the red wavelengths. Thus, both processes, R & A, are
wavelength dependent.

Since we have discussed colors, we need to further define the photons

that embody them in terms of wavelength, "k, and other symbols
commonly used in optics and spectroscopy. This is shown as follows in
5.2.14. given on the next page. Note that the visible region of the
spectrum lies between about 1.75 and 3.00 ev. Thus, in a "colored" solid
where absorption occurs from a ground state to an upper absorbing state,
358 Optical and electronic properties of solids

5.2.14.- Part of the Electromagnetic Spectrum

the separation of the upper energy state from the ground state cannot be
more than about 3.00 ev. Furthermore, in such a solid, resonant
absorption of the photon does not occur unless that photon has a
minimum energy between 1.75 and 3.00 ev. Moreover, if the solid
reflects photons having a wide range of energies, we say that we have a
"white" solid. We can, as a general rule, divide inorganic solids into two
(2) classes: those which absorb light (pigmented), and those which emit
light (phosphors and solid state lasers). That is, we can have pigments
(for paints) and phosphors (for TV or lamps). In both cases, we can add
controlled amounts of a transition metal ion to the solid in order to
control the photon absorption and/or photon emission properties of that
solid.

We have now presented a general picture of how photons were discovered

and identified as "force" carriers between electrons and the like, i.e.-
"leptons". We have also shown, in general terms, how solids interact with
 5.2 Absorbance, reflectivity and transmittance 359

photons, whether by transmission, reflection or absorption. The structure

of solids was covered in Chapter 1 and the growth of crystals in the last
chapter. We have also referred to phonons (which are quantized vibrations
within the crystalline lattice) but we have not explained why lattice
vibrations are also quantized. What we find is that the lattice, being finite
in terms of repeating unit cells, allows only certain modes of quantized
vibrations to occur. That is, only certain resonant vibrations are allowed
within the unit cell of the crystalline lattice.

Note that each atom in the solid state has its own inner electron shells,
unique to its own atomic number, plus those that it contributes to the
overall structure through bonding by ionic or covalent means (or some
combination or both). In addition, when two compounds react in the solid
state to form a new compound, the bonding electrons of each compound
become redistributed into well-defined energy bands of the new
compound (as defined by the crystal lattice structure). Such bands are
called Brillouin zones after a French investigator, Leon Brillouin (1931)
who first worked on this problem.

The Hartree method, i.e.- the Hartree-F6k method, of using the

SchrOdinger equation(1915), for calculating the electronic structure of a
given atom or group of atoms, provided a theory for a common framework
which described the varying properties of different kinds of solids. In this
method of calculation, the electron orbitals are allowed to spread out over
the entire solid. Effectively, one deals with an infinite number of electrons
and orbitals, which can be handled with mathematical techniques
developed by the Swiss-born U.S. physicist Felix Bloch and others (1928).
The orbitals display shell-like features even in a very large system. There
are groups of orbitals with energies close together. Known as bands, they
are separated by gaps in energy.

The characteristics of any such solid were shown by Brillouin to occupy

only certain energy zones, well defined by the reciprocal lattice. Such
zones depend upon the lattice structure and the type of electron bonding
present in the solid. Some solids are dielectric, some are semi-
conducting and others, notably metals, are conductors.
360 Optical and electronic properties of solids

These differences are reflected in the type of Brillouin zones present and
their occupation by the available electrons.

5.3.- ELECTRONIC PROPERTIES OF CRYSTALS

It would behoove us to consider the factors that cause change in

electronic properties of solids as a function of structure and bonding. It is
evident that this aspect is very important, since the electrical nature of
solids varies according to both composition and structure. Some materials
are semi-conducting and others are dielectric. Note that "electrical"
refers to the conduction properties of the solid whereas "electronic"
refers to the interaction of photons, electrons and the like, i.e.- excitation
within the lattice. Since we intend to focus on the optical properties of
solids, notably phosphors in later chapters, and since the optical
properties of solids depend, in large part, on the electronic properties of
the solid, we need to know more on this particular aspect of solids. A
good example of this is the crystal, p- BaB2O4- This crystal belongs to the
trigonal system with point symmetry of C3V. It is a negative uniaxial
crystal, used to double the frequency of a laser beam. There is a high
temperature form, notably a-BaB2O4 (transition temperature: p t o a = 925
°C), but this form is not optically active. It is natural to ask why this
should be so. The answer lies in what happens to the electrons associated
with each atom as that atom enters into the lattice to form the solid. We
know that some of the electrons form bonds with the anion, either as
covalent or ionic bonds (or somewhere in between). What happens to the
rest? Do they remain around the parent atom? Why are some solids
conductive while others are not? The following discussion addresses
these questions.

I. Band Models and The Reciprocal Lattice

When electrons enter a solid lattice, they occupy certain energy levels
arranged in well-defined zones. What this means is that there is an energy
zone within the crystal where the electron wave density is high, and a
resonance condition has arisen. The electrons are free to move within
this zone and electroneutrality is preserved because the electron wave
 5.3 Electronic properties of crystals 361

density is spread throughout the crystal. These allowed zones are called
Brillouin Zones., as we have already indicated. Their geometry is
determined by the lattice structure.

To understand this, consider a free electron with energy, e , and wave

vector, k . As shown on the next page in 5.3.1., e varies with + k. This
curve is a potential energy diagram of the free electron. Note that we have
converted sin 8 (the linear sinusoidal function) into n (the radial function).
If we apply a mono-atomic lattice, arranged in a linear manner, and having
a lattice constant of a , to our electron, we can show that Bragg reflection
occurs at:

5.3.2.- k = ±n/a

Observe also that there are three allowed zones in 5.3.1., labeled as shown
in the following:

5.3.3.- # 1st allowed band

* forbidden band
# 2nd allowed band

It is thus evident that even in a simple linear structure (i.e.- a row of M-

atoms), the electrons associated with the structure are restricted to an
allowed band.

Joining the atoms into a structure causes a perturbation of the electron

wave-functions which results in restricted energy zones for the electrons
associated with the structure. The general case is shown in Part la & b of
5.3.1. The lattice constant is "d" (from Bragg's Equation). Note that we
have used a 1-dimensional structure so as to show its perturbation effect
upon a single electron. In Part 2a & 2b, we have the same free electron
function and linear lattice as given in Part la & b above, except that we
have arranged the lattice in a 3-dimensional pattern. In Part 3a & b , we
show the 3-dimensional structure effects on the allowed zones when
occupied by several electrons in the crystal. Note that an allowed zone has
resulted. Thus, we conclude that the effect of structure on free electrons
362 Optical and electronic properties of solids
 5.3 Electronic properties of crystals 363

in the solid, results in an allowed band structure (or Brillouin Zone) to

which the electrons are confined in the solid. This work was first
accomplished by Felix Bloch. His equations described the wave-functions
of the electrons, as a function of the diffraction conditions imposed by the
periodic lattice. The result is called a BLOCH function:

5.3.4.- Wk = U k (r) exp [j h • r ]

where h is a general lattice vector and Uk (r) is a periodic function. The

function "r" is a general direction in the lattice, and we require three for
the solid. In order to mathematically describe electron energy bands in
solids, we find that we must examine the reciprocal lattice. The best way
to understand this concept is to examine actual structures and the
Brillouin Zones associated with them, as shown in 5.3.5., given on the
next page.

The reciprocal lattice is just what the name says it is. If we have unit cell
vectors in the real lattice, the imaginary reciprocal lattice vectors are 90°
to the real ones. These imaginary vectors define the Brillouin Zone for
that structure.

In 5.3.5., we show the allowed electron energy zones in two different

cubic crystals. At the top of this figure, we have a face-centered cubic
lattice (fee) with the 1st Brillouin Zone (cell) drawn in. Note that the cells
are in reciprocal space and the reciprocal lattice is body-centered. The
cell-faces of the Brillouin Zone are at right angles to the points of the real
lattice. It is a truncated dodecahedron and is derived by cutting off the
corners of the cube on the {111} plane to derive the Zone faces at 90 ° to
the points of the cubic lattice. At the bottom of the Figure is the Brillouin
Zone for the body-centered cubic lattice (bcc). Note the difference
between the two reciprocal lattices. In the bcc reciprocal lattice, the
Brillouin Zone impinges on the faces of the unit-cell cube, whereas in the
fee reciprocal lattice, it is flat at these points. The faces of the zone
represent energy gaps in velocity space, and the surface is sometimes
called "k" space. Although Brillouin Zones define the extent of k-space,
the maximum energy of the electrons (as a function of temperature)
364 Optical and electronic properties of solids

defines the Fermi energy. Thus, the Fermi Surface of the "sea of
electrons" in a solid is the upper energy of these electrons, confined to
reciprocal space by the Brillouin Zone. To better understand how this
 5.3 Electronic properties of crystals 365

works, consider a band model for metallic copper, as shown in the

following diagram:

5.3.6.-

For Cu° , the 4 s 1 electron-subshell is half-filled. This is the top of the

filled band, or Fermi energy of the sea-of-electrons in the solid. Copper
has a fee lattice structure. This diagram illustrates the same band
366 Optical and electronic properties of solids

structure of copper, but from a different aspect, i.e.- that of energy. Thus,
the Brillouin band model shows the placement of the allowed energy
zones in the crystal whereas the energy band model of 5.3.6. shows the
relative energy of the electron bands, including those of the bonding
electrons and "free" electrons in the metal solid.

We are usually more interested in the energy levels of the solid and tend
to use the band model when dealing with phosphors rather than the
Brillouin band model. To illustrate what we are discussing, consider the
model given as 5.3.7. as follows on the next page. Compare it to the band
model given above as 5.3.6.

In this diagram, we see the experimentally-determined Fermi surface of

the copper electrons within the Brillouin Zone. The "necks" actually
define the positions of individual atoms in the real fee lattice. Compare
the actual Fermi Surface to the Brillouin Zone geometry. Also shown is
the Fermi Surface of the element, germanium- Ge4* , heavily doped with
arsenic, As5+ , to make it a n-type semiconductor. It has the same
Brillouin Zone as copper but far fewer electrons are available for
conduction. This results in "wells", or energy depressions, at the atomic
positions, in contrast to the "necks" of copper which extend out to the
surface of the real lattice. Thus, the actual occupation of the Brillouin
Zone by electrons depends not only upon geometry, but also upon the
numbers and energies of the available electrons. However, we conclude
that the energy band model is more adaptable to our needs in that it tells
us the relative energy of the individual electrons in the solid.

One other example is the compound, TiO. Calculated energy bands are
given in the following, presented as 5.3.8 on the following page. TiO has
the cubic NaCl structure but is not a dielectric. It is nearly metallic in
electronic character. This diagram presents the results of experimental
measurements, and the electron bands calculated from that data,
presented in terms of energy and direction around the cubic lattice. In
this case, lattice intercepts are used rather than Miller Indices. The
individual energies, as a function of the electron-band, are labeled at each
intercept. It is easy to see why TiO conducts like a metal, since the 3d
 5.3 Electronic properties of crystals 367

5.3.7.- FERMI SURFACES AND BRILLOUIN ZONES IN COPPER

and 4s electron bands overlap extensively in the solid. The conduction

band is the 2nd Brillouin Zone in the solid (see 5.3.8. on the next page).
The Fermi Energy rises and falls with temperature and so may be level
with the valence band, or above or below it.
368 Optical and electronic properties of solids

5.3.8.-

In a semi-conductor with proper doping, thermal energy is sufficient to

bridge the forbidden gap so that conduction results. It is easier to use a
two dimensional band model to represent Fermi Levels and Brillouin
Zones in the crystal lattice, just as long as we realize that the actual case is
three-dimensional in nature.

The ENERGY BAND MODEL has arisen to serve as an easy method of

representing electron states in the solid. The Fermi level usually defines
the top of the valence band, and is related to the 1st Brillouin Zone in the
solid. We generally draw a band model as in 5.3.9.{next page) which
 5.3 Electronic properties of crystals 369

shows a comparison of band models for insulators, semi-conductors and

conductors.

5.3.9.- BAND MODELS FOR INSULATORS AND CONDUCTORS

A final comment regarding Brillouin Zones: we generate the faces of these

zones at 90 ° to the faces of the real lattice. To do this for the fee lattice,
we cut off the corners of the cube at right angles to the 4-triad axes (see
5.3.5.). If we cut the cube at 90 ° to the 6-diad axes, we generate a
rhombic dodecahedron which is the Brillouin Zone for the bec lattice. Our
direction of cutting is 90 ° to the real lattice vectors, but parallel to the
reciprocal lattice vectors.

We ought to discuss the mathematics connected with construction of the

reciprocal lattice and those of the Brillouin Zone. We will do so in the
following Section.

We will start by examining the methods of calculation of energy bands in

solids.
370 Optical and electronic properties of solids

II. Calculation of Enerffv Bands in Crystals

As stated above, the basic equations used to describe electron energy

bands in solids are called the BLOCH FUNCTIONS . These are a part of
quantum mechanics, so we begin by borrowing one of its more famous
concepts, that of "The Particle in a Box".

Consider a free electron. It has both particle and wave properties. The
general SCHROEDINGER equation describing its wave properties is:

5.3.10.- HWe = EWe

where H is the hamiltonian operator which "operates" upon the wave-

function of the electron, *Pe . and E is an energy function. There may be
more than one solution to the E- function which fits the equation. Indeed,
a number of "quantized" solutions to equation 5.3.10. have been shown to
exist. But, if we confine the electron to a box having a length, 1 , we find
that the number of solutions diminishes considerably. What happens is
that a resonance condition arises for the electron when it is confined to
the box. Standing waves result whose individual wavelengths are a
function of "1". The wavevector is an energy solution for the resonance
conditions. Let us define:
5.3.11.- k = any wavevector of the electron

We = wavefunction of that wavevector

If we set up a Schroedinger equation for the particle in a box, we get:

5.3.12.- [-h2/2m V2] W k (r) = E W k (r)

where r is any direction in the crystal, h = h/2?r , m is the mass of the
electron and V 2 is of course the second derivative of the partial equation
relating the three directions, x, y, & z in the lattice:

5.3.13.- v2 = (a2/ax2 + a2/ay2 + a2/ay2

 5.3 Electronic properties of crystals 371

For the particle in a box, the only allowed solutions for k are:

5.3.14.- kx = 0, 2 K /I , 4?i / I , etc.

Similar functions exist for k y and k z . The general solution for the energy,
E k , is:

5.3.15.- Ek = h 2 / 2 m [ k x 2 + ky2 + k a 2 ]

where k x , etc. is a vector and the momentum, p = h k x> y i8 is

intrinsically included. Solutions for the wavefunction, Wk , are:

5.3.16.- W k (r) = (1 / V)112 exp [-i k- r ]

where the boldface indicates vector functions. Note that we have

expressed the wavefunction in one direction only, and in terms of the
potential energy, V.

BLOCH FUNCTIONS, U, are wavefunctions similar to those given above,

except that a periodic lattice is the perturbing medium, instead of the
box. We use U as our potential energy, in order to avoid confusion. For a
periodic lattice, the potential energy is defined one-dimensionally as:

5.3.17.- U (x) = U{x + a)

where x is a lattice direction and a is a unit cell dimension. One solution

is:

5.3.18.- U (x) = 2 U ( G j e x p i @ - x

where © is the reciprocal lattice vector. From 5.3.10., we obtain by

combination: Kinetic energy + potential energy = total energy

5.3.19.- l / 2 m ip2 + u (x) + [l/2m ip)2 + 2 U(G) exp i © • x] <P(x) = E «P(x)

372 Optical and electronic properties of solids

This is a one-electron approximation. By solving for the momentum

operator, i.e.-

5.3.20.- ip =-ihd["F(x)]/dx

and using appropriate boundary conditions, we get a Fourier Series:

5.3.21.- W k (r) = Zk C (Ik) exp [-i Ik-x ]

where C is the Fourier coefficients for which we need a value. It turns out
that the only allowed values of *Pk are:

5.3.22.- Wfc (x) * C(Ik - ®)

This brings us to:

5.3.23.- W& (x) = 2 G C(!k - ®) exp [i (k- @) • x]

We separate this into parts:

5.3.24.- W& (x) = {2 C(Ik - ®) exp [i(k- ®) • x]} exp {i k • x}

where @ is the reciprocal lattice vector. The first term was defined as the
potential energy. If we now define "r" as the general lattice direction, in
contrast to the specific direction, Y . We then have the BLOCH
THEOREM for electron waves in aperiodic lattice, i.e.-

5.3.25.- W& (x) = exp i k • x Uk (r)

The wave equation for our periodic lattice is now:

5.3.26.- h 2 / 2m(k2 - e ) Cfc (k) + 2 UG Ck ( Us - ®) = 0

We can now reexamine 5.3.21. in terms of our derived results.

 5.3 Electronic properties of crystals 373

The diffraction points are in terms of 2n/a or some integral fraction

thereof. They correspond to the diffraction condition:

5.3.27.- Ak = G

which in "k"-space is the Brillouin Zone. One might logically ask of what
use are these equations and what have we really demonstrated. To answer
this, let us review what we have studied in this chapter.

In order to make the above description more relevant, consider the

following description of how one goes about actually calculating the
dynamics of the real lattice. Note that we used Bloch functions and
Hamiltonians to describe how one begins the calculations which define
the reciprocal lattice, but this does not tell us exactly how one should
approach the problem. One actually uses the lattice parameters of the real
lattice in order to obtain a meaningful result.

Consider the real lattice. Let us redefine the following:

5.3.28.- Lattice Directions : x, y, z .

Unit cell vectors : SE , j , %,

Translation vectors: a , h , c

We can define any lattice as:

5.3.29.- a = axX + a y j + az^

to = b x x + b y ] f + bzg

C = CX I + C y J + C z l

Using these equations, we can determine the equation of a plane as

follows.
Consider the following two-dimensional lattice:
374 Optical and electronic properties of solids

5.3.30.-

The x and y directions are defined as given above. The unit cell also has
the same definition. We define the set of planes having the Miller indices
{310} as:

5.3.31.- 3 x +y = a

We have also defined "d", the separation between the planes, and "r", the
vector to the nearest neighbor. We find that the equation of a plane is:

5.3.32.- dl = n • jr

A closer view of these two vectors relating d, from the origin, "o", to a
point "p", is provided by the following diagram:

5.3.33.-

Using the terminology given above:

5.3.34.- a = h a + kb + 1c

where {h, k, 1} are Miller indices or lattice intercepts.

 5.3 Electronic properties of crystals 375

If the lattice is cubic, then we have:

5.3.35.- d= 1 /(h2+k2+12)1/2 • (h a + k b + 1 c)

which is the equation relating lattice planes (Miller indices) to "d", the
distance between selected planes.

In the early history of x-ray diffraction, Laue (1905) discovered that by

placing a crystal oriented in the proper manner in the path of a "white" x-
ray beam, he could obtain a series of spots , arranged as concentric
circles, on a photographic film. His geometrical arrangement was:

5.3.36.-

The Laue equations which characterize the x-ray diffraction in this case
are:

5.3.37.- T = ni a + n2 b + n3 c
Ak= kscatt ' Jko

where kscatt is the scattering factor and k 0 is the original x-ray beam
vector. This leads to:

5.3.38.- ©• Ak = 2 ji n i ; 1b • A k = 2 j t n 2 ; © • A k = 2 Jt n 3

We can solve for A k as follows:

376 Optical and electronic properties of solids

5.3.39.- Ak = 2 ; i ( n i / A a + 112/B b +113C c)

where A , B, and C are reciprocal lattice intercepts. Laue found that by

using the reciprocal lattice concept, he could easily label the x-ray points
on the film in terms of {h,k,l}. Furthermore, all {O,k,l} , {h,O,l} , or
{h,k,O} points lay on separate concentric circles, thus simplifying
identification. It remained for Ewald (1908) to explain this phenomenon.
Ewald rearranged the Bragg equation to:

5.3.40.- sine = n/d / 1/X

This equation defines so-called "Ewald Space". Radiation of appropriate

wavelength, impinging upon ordered rows of atoms, produces cones of
scattered radiation, as follows:

These cones intersect a sphere of radius, 1/K, and define planes which are
separated by 1/d. 8 is the angle between the cones. The distance between
 5.3 Electronic properties of crystals 377

the center of a cone and its edge is the reciprocal lattice vector, ko and
kscatt are as defined above. We can set up a scattering equation in terms of
the reciprocal lattice vectors, as follows:

5.3.42.- Ak = q A + r B +sC

where A , B, and C are the reciprocal lattice vectors. Note the similarity
to 5.3.36. and 5.3.38.

It is easy to show , using the Ewald construction, that the following

combination of vectors, are true:

5.3.43.- A - a = 2 it B-a=O C- a = 0
A • b =0 B - b = 2it C•b = 0
A-c=O B•e = 0 C - C = 2 J I

We can now equate the reciprocal lattice vectors to those of the real
lattice vectors:

5.3.44.- A= 2it ( b x c /a-bxc)

B= 23t(SLxc / a - b x c )

C= 2 i (axb /a-bxc)

where A , etc., is in c m 1 and a is in cm. Thus, for the real lattice, the
translation vector is :

5.3.45.- T = ni a + n2b +113C

and that for the reciprocal lattice is:

5.3.46.- G = h A + k B +1G

where n 1 . n 2 . n 3 , and h, k, and 1 are intercepts in the respective lattices.

Note that {hkl} are the same Miller Indices referred to in Chapter 1.
378 Optical and electronic properties of solids

It should be recognized that if we multiply G • T, we get a scalar product,

since we are multiplying reciprocals.

The wave equation for Laue diffraction is:

5.3.47.- Ak = G

It is also the same for wave propagation, namely:

5.3.48.- l o + G = kscatt

If we square both sides of this equation, we get:

5.3.49.- (k0 + G)2 = (k s c a tt) 2

If we have elastic scattering (as we do in x-ray scattering), then k 0 =

fcscatt • Thus, we obtain the WAVE EQUATION for the reciprocal lattice:

5.3.50.- 2 k 0 • G + G2 =0

In this case, G connects points in the reciprocal lattice. We can now

proceed to calculate reciprocal lattice vectors for the several figures
illustrated above.

BODY-CENTERED CUBIC frefer to 5.3.5.1

Primitive lattice vectors are:

a = a/2 (s + j - %,)
b = a/2 (- ss + j + %)
c = a/2 (x - j + z)

where a is the unit cell dimension. The unit cell volume is therefore:

5.3.51.- ¥ = a • b x c = 1/2 a3
 5.3 Electronic properties of crystals 379

We set up the reciprocal lattice vectors as:

5.3.52.- A = 2 it /a(x + y )

B = 2 jt/a( y + z)

C = 2 at /a(x + z)

From the translation vector of 5.3.45., we can get:

5.3.53.- G = 2 it /a [(h + 1) x + (h + k) y + {k + 1) % ]

Using units numbers for intercepts, i.e.- the first Brillouin Zone, and
ignoring h, k, or 1 equal to zero, we finally get 12 vectors which define the
Zone in the bcc lattice. These are:

5.3.54.- 2 it /a( ± x ± j )
2 r e / a ( ± y ± z)
2 3i /a ( ± x ± z)

Using all combinations of the values allows us to construct the 1st

Brillouin Zone of the bcc lattice.

FACE-CENTERED CUBIC LATTICE (Refer to 5.3.5.-1

First, we define the real lattice as:

5.3.55.- si= a/2 [x + j )

to = a/2 (y + z)
o = a/2 (x + z)

For the reciprocal lattice, we get:

5.3.56.- A = 2 n / a ( x + J- z)
B = 2 3 i / a ( - x + y + z)
C = 2 it /a(x - y + z)
380 Optical and electronic properties of solids

We then set up the reciprocal lattice equation as before:

5.3.57.- G = 2 rc /a [(h - k + 1) x + (h + k -1) y + ( - h + k + 1 ) ^

]
In this case, we have more vector values than we can use. Therefore, we
take 6 of the shortest-non-zero vectors to plot the Figure. These are:

5.3.58.- 2n/a(ii )
2n/a(± y)
2n/a(i x )

Plotting these gives the results shown in 5.3.5. It is well to note that the
same treatment can be given to other lattices. We chose the cubic lattices
because they have high symmetry which simplifies the calculation needed
to derive the reciprocal lattice and the associated Brillouin Zones.

An alternative way to obtain BLOCH functions is illustrated in the

following. This method is essentially the same as that given above but uses
a slightly different approach to the problem. The Schrodinger equation
can be written in terms of the total energy of the system, in contrast to
that of "The Particle in a Box". If we do so, we get.

5.3.59.- - h 2 / 2m V 2 ] W k (r) + V(r) = E W k (r)

where V(r) is an effective potential applied to each electron in the lattice

and r is any direction in the lattice, i.e.- either x, y or z. V(r) has the
property of:

5.3.60.- V(r) = V(r + ISn)

where 3Sn is a lattice vector indicating the nth position of atoms in the
translational symmetry of the lattice.

Note that this approach relates to the actual circumstances that we find in
crystalline solids, in contrast to that already presented above. We
 5.3 Electronic properties of crystals 381

presented the data in this manner so that you could understand exactly
what was meant by the Reciprocal lattice and BLOCH functions.

To obtain a BLOCH function, we can write 5.3.60. in the form of a Fourier

function, as follows:

5.3.61.- Vr= 2>n e i<Vr

where © is the reciprocal lattice vector and SP is the same vector defined
above for the lattice. Although 5.3.59. cannot be solved explicitly for the
lattice values (because of the infinite number of possibilities of a finite
lattice, an indirect solution was found by Bloch as:

5.3.62.- W( -s) = e ' h Uk (if)

where Ik is the wave vector and Uk (s*) is the periodic function of the
lattice translation. This can be expanded by a Fourier series as:

5.3.63.- Uk (if) = 2 C n & e i Cn ' s

the wave vectors, Ik , are well-defined quantum numbers of the electronic

states of a given crystal (remember that Ik is in terms of Jf which defines
all three lattice directions). Using the above equations, we can rewrite
5.3.59. in the form:

5.3.64.- {- h2 / 2m (& + © e) 2 - E} Q. 2 (Cn V 1-n) = 0

This allows us to get energy values, i.e.- Eigen (E) values by neglecting the
Fourier components, Q, and (Cn V i n ) , but using the Fourier component,
Cn. One can start with any order approximation of 5.3.60. such as: CQ = 1,
CQ = 0 (n * 0) so as to get:

5.3.65.- E = h 2 / 2m & 2 = E o (Ik)

This is the free electron model of 5.3.1. Higher orders are obtained in the
382 Optical and electronic properties of solids

same manner. For n = 0,1, i.e.- two electrons, the equations, 5.3.61. and
5.3.62., become:

5.3.66.-

E - 1/2 {E (&) + E 0fe * fifo}*/{[E Ifo1* - E fflg A m\} + V(@)>

Thus, the energy gap at the wave vectors satisfies:

5.3.67.= Ik 2 = (k + © a) 2

This is the so-called Bragg condition.

III. Point Defects and the Energy Band Model

Even though we have made considerable effort to explore the reciprocal

lattice because it allows us to determine the actual location of the bonding
electrons in any given three-dimensional lattice, it should be clear that
the electron-band model is easier to understand and easier to use than
the Reciprocal lattice band model. To date in this discussion, we have
covered:
a. Crystal Structures

Using propagation models (see Chapter 1), it is easy to see that only cubic
and hexagonal close-packed lattices exist if the angles are 90 ° and unit
cells lengths are equal. Tetragonal and orthorhombic cells arise if the
unit-cells lengths are not equal. When the angles of the unit cell are
allowed to vary, then we generate monoclinic, triclinic and trigonal
lattices. There are a total of 7 primitive lattices. When we impose unit-cell
symmetry upon these, we get a total of 14 Bravais or Space Lattices.
If we apply rotational symmetry factors to these, we then obtain the 32
Point Groups. Local atom-symmetry then generates the total of 231 Space
Groups.
b. BriUniiin Zones

We were able to show that the perturbation of an electron wave-function

 5.3 Electronic properties of crystals 383

by a periodic lattice resulted in certain allowed three-dimensional zones

within the crystal. These are intrinsic to the structure. We also showed
that the electron occupation of the Brillouin Zone may not be complete,
and will depend upon the nature and numbers of electrons involved in the
structural lattice.

c. Energy Bands in Solids

Finally, we derived the BLOCH FUNCTION to show that these energy

bands, in reciprocal space, do have some validity in quantum mechanics.
It also gives insight as to the nature of the Fermi level. We also illustrated
band models in 5.3.9. What is important to realize that the valence band
there is drawn in two dimensions. Actually, it follows the Brillouin Zone or
k-space of the crystal lattice in three-dimensions. The Fermi level surface
is also affected by both k-space and temperature. It is constrained by
reciprocal space, just as the Brillouin Zone is. We use the band model to
illustrate certain aspects of each unique crystal. Otherwise, the required
model would be quite complex, particularly those crystals with low
symmetry. We usually illustrate some specific defect and the band model
immediately adjoining it. For a phosphor, this would be the activator
(impurity) center. Since we have already (Chapter 2) examined various
point defects, let us now illustrate them within the periodic lattice as a
function of the energy bands and the Band Model.

The effect of a periodic lattice upon the wavefunctions of an electron is

illustrated in the following diagram, given as 5.3.68. on the next page.
Note that the positions of the individual atoms within the lattice are
shown, i.e.- "lattice sites".

At the left side, we have shown the energy levels of a single electron as
the energy, E , increases. When lattice sites are imposed, we get a series
of bands {shown horizontally) as E increases. Note also the unoccupied
zones, indicated by the hatched area. Occupation of the allowed bands is a
function of the types and numbers of electrons present. At the bottom is
the wavefunction associated, as a function of atom-position in the lattice.
384 Optical and electronic properties of solids

5.3.68.-

Let us now construct a band model for our semi-conductor, Ge. We

illustrate both p-type and n-type defects, as follows:

5.3.69.-

Ga is an acceptor while As is the donor in these diagrams. The individual

energy levels of the defects are also shown. Note that they are not the
same as those of the lattice atoms themselves. For Ga3+, the positive "hole"
 5.3 Electronic properties of crystals 385

is superimposed upon the energy "hump" caused by the defect, VM. In the
case of As5+ , the electron is attached to the conduction band and is
available for conduction. The energy required to promote this electron
into the conduction band is of the order, < kT>, or room temperature.
There is an energy well immediately below the electron.

The case for the color center is somewhat different. Here, an electron has
been captured by the vacancy, but has its own energy levels, as follows:

5.3.70.-

The reason for this difference is that the host crystal is a dielectric or
insulator (NaCl), while Ge is a semi-conductor. The vacancy, Vx , is thus
seen as a donor, while VM is an acceptor. Thus, in 5.3.70., we have the
example of the defect pair:

5.3.71.- [ VX , e- ]

When an interstitial is formed, it becomes a donor upon ionization, as

shown in 5.3.72, given on the next page.

We have shown the case for the defect pair: [Mi+, VM" ]. Firstly, the VM" is
seen as a acceptor , while Mj+ is a donor (But this occurs after ionization
of the intrinsic defects). They are most likely to be nearest neighbors, so
that they tend to cancel each other, as we suspect they ought to do.
386 Optical and electronic properties of solids

5.3.72.-

Finally, we show the effect on the energy band of the solid when several
defects cluster, as shown in the following:

5.3.73.-

Here, we have used the defect pair:

5.3.74. [Vx+ , e-]

which is an anion vacancy which has ionized. Note how this differs from
that of 5.3.70. which is the band model for the color-center. Whereas, in
the color-center, the vacancy accepts an electron from an external source,
 5.4 Phonons as quantized lattice vibrations 387

intrinsic defects such as the vacancy in 5.3.70. ionize to form their own
associated electrons. Color centers cannot, and do not, form associated
clusters, but anion vacancies can do so. When this happens, a new set of
bands are localized, whose energy levels are characteristic of the defects,
as shown in 5.3.70.

We can thus make an important deduction from our studies, namely that
associated defects have definite effects upon the energy band structure of
solids. It is not surprising, then, that reactivity of solids depends
considerably upon the degree of purity of the solid itself.

5.4.- PHONONS AS QUANTIZED LATTICE VIBRATIONS

We have referred to phonons as quantized lattice vibrations without really

showing why such vibrations arise. We have also shown how a photon
interacts with a solid, whether by transmission, reflection or absorption.
Although it has been described as a particle with wave properties, it is
evident to the erudite that the true nature of the photon is neither, to the
exclusion of the other. It may be in fact a manifestation of more
fundamental natural quantities or combinations of sub-particles such as
quarks or combinations of such. Nevertheless, all matter having a
temperature above that of absolute zero continues to emit a flux of photons
while it exists. This may be ascertained by considering the fact that
"black-body" radiation emanates from all solids having a temperature above
absolute zero, , i.e.: 0 °K. Most of the time, the radiation is in the far-infra
red region, but photons are emitted on a continuous basis. And, when the
temperature of that matter is raised, the energy of the black-body
radiation rises as well. We define the radiation in terms of comparable
temperature. The radiation of the Sun is about 10,000 °K., that is, the
spectrum of the radiation is the same as that of a hot body of matter
having that same temperature.

There are certain equations which can be used to describe the observed
properties of the fundamental unit of quantized energy, the photon. It
can be shown that the vectors, F • M , of a photon (these vectors are
388 Optical and electronic properties of solids

electric and magnetic field vectors, respectively) interact with the

electrons of a solid and lead to a dispersion equation:

5.4.1.- G s F +k'

where G is the reciprocal lattice vector, F is a lattice-wave vector, and k'

is the scattered photon. Since F , the lattice-wave vector, is involved in
photon interactions, it is natural to ask "What does this mean?". The
elastic wave vector promotes lattice vibrations of defined energy when a
photon interacts with the atoms (ions) of a solid. We have evidence that F,
the elastic wave in the lattice, is quantized. First, as both the following
quantities approach zero, i.e.- the internal heat, as defined by Q, and the
absolute temperature in °K, :

5.4.2.- {Cp (S) =* 0} ; {T (°K) =* 0}

the value of F approaches a unitary value.

Secondly, if one uses monochromatic radiation, i.e.- a laser, one can

obtain evidence of Stokes' and anti-Stokes1 scattering of phonons, i.e.-
lattice vibrations, as shown in the following:

5.4.3.-

You will note that the frequency shift is quantized.

 5.4 Phonons as quantized lattice vibrations 389

That is, the interaction of a photon with lattice vibrations (phonons) is a

quantum unit which is either positive or negative in regard to frequency.

A general dispersion equation would be:

5.4.4.- k*' + Ik = t +®

where It' and It are the interaction vectors of the photon and electrons of
the lattice. Ik is the real lattice vector and @ is the reciprocal lattice
vector (These terms have already been discussed above). We can then
write:

5.4.5.- k* - k*1 = ©± Ik

Thus, the interaction vector of the photon subtracts from the lattice wave
vector of the electron in terms of the real lattice vector, adding and
subtracting from the reciprocal lattice vector. What this means is that any
photon-phonon interaction can add or subtract to the total energy of
either the photon or phonon, or both. (See 5.4.3. given above). We can
then write:

5.4.6.- h2 k*2 / 2 m = h2 (T*)2 / 2 m ± hcok

where m is the reduced mass of the atom(s) in the lattice and oo = 2rc v
(h = h/27t). Using these equations, our next step is to develop phonon
dispersion equations for the monatomic lattice and then for a diatomic
lattice, so as to illustrate the properties of phonons and phonon
dispersion interactions for the case of a phosphor. Finally, we will show
how absorbed energy in a phosphor is transferred by resonance processes
via phonon dispersion interactions.

I. Phonon Dispersion Equations

Consider a monatomic lattice. This maybe a metal or other element. Our

approach is to apply Hooke's Law for an elastic vibrator, in which the
390 Optical and electronic properties of solids

return force is equal to that of the displacement force so that no energy is

lost in the vibrational process. We define the following:

5.4.7.- HOOKE'S1 LAW FOR AN ELASTIC VIBRATION

F S force of displacement u = a coordinate in the lattice

t E time x S lattice direction

s = "spring" constant p = amass density

HOOKES LAWS

F = s (du/ax) ; s (a 2 u /a2x) = p (02u / a 2 t)

We can rewrite these equations in the form of a traveling wave, i.e.-

5.4.8.- u = uo exp (- i [CD t + k* x]}

where: k* = co/v s a n d v s = ( s / p ) 1 / 2 .

Now consider a cubic monatomic lattice. We can use a plane net to show
the position of each atom , and the displacement can be shown as crosses,
so as to show how each atom is displaced from a given equilibrium point.
This is illustrated in 5.4.9. on the next page.

We have illustrated two types of vibration, one in which the atoms vibrate
together in an equal distance (shown as a longitudinal vibration) and the
other where the atoms vibrate an unequal distance, relative to one another
(here shown as a transverse vibration). Note however, that the atoms
vibrate in phase, regardless of the direction and the distance. We can
write equations using these terms by considering u as a coordinate and p
as a displacement, in terms of the total force required, Ft , and the
individual forces on each atom, Fi t as a form of Hookes Law:

5.4.10.- Ft = 2 Fi (up+x - ux ) where x i s a lattice parameter.

 5.4 Phonons as quantized lattice vibrations 391

5.4.9.- VIBRATIONS IN A CUBIC MONATOMIC LATTICE

From 5.4.7., we have:

5.4.11.- M d u x / d t = 2 ^ 1 (up+x - Ux )

where M is amass. For a traveling wave, we get:

5.4.12.- up+x = uo exp [i(p + x) al? - ico t]

and:
- CD2 M UQ exp [i x "k* a] = 2 Fi (exp [ia> t])

where a is the unit lattice vector. What we have done is to combine 5.4.8.
and 5.4.11. By suitable manipulation, we can get the equations shown as
follows:

5.4.13.- co2 M = - 2 Ft {(exp [i p a k* ]) - exp (i p a k* ) - 2} or:

co2 = 2/M 2 Fi (1 - cos p ait*)

What this means is that a standing wave will occur for nearest neighbor
planes of the reciprocal lattice. Since all atoms in this example are
equivalent, we can replace 2 Fi by a constant, C, as follows::

5.4.14.- to2 = 2 C / M ( 1 -cospak*)

392 Optical and electronic properties of solids

Note that we are still using I*' and k* as the interaction vectors of the
photon and electrons of the lattice. A plot of the 5.4.14. equation in terms
of a traveling wave in a lattice gives:

5.4.15.- A Standing Wave in the Real Lattice

We can also plot this wave in terms of the critical points of the reciprocal
lattice, as shown in the following:

5.4.16.- A Standing Wave in the Reciprocal Lattice

It is obvious that a standing wave exists between - rc/a < k < jr/a.
Furthermore, these waves are quantized in energy, as we have said.
 5.4 Phonons as quantized lattice vibrations 393

Now consider a diatomic lattice. The situation is quite similar and we can
use the same approach. In this case, two different atoms are present in
the unit cell. An example would be NaCl. If we proceed through the
mathematical analysis, using m for one mass and M as the other, we arrive
at the result shown in the following:

5.4.17.-

Now we have two (2) phonon dispersion curves, a so-called optical branch
and a lower energy acoustical branch. The standing waves are better
understood in terms of the actual displacement the atoms undergo:

5.4.18.-

It is easily seen that, for the acoustical modes of lattice phonon vibration,
both types of atoms vibrate together (the displacement is in the same
direction - which may account for the lower energy required). In contrast,
in the higher energy optical modes, each type of atom vibrates together,
but in opposite direction to the other, i.e.- the displacement is opposite
394 Optical and electronic properties of solids

for each type of atom, the two types being denoted by black or white
balls). Thus, we can conclude that in lattices having more than one type of
atom, there are both optical and acoustical modes of both longitudinal
and transverse vibration. We label these phonon vibrational modes in a
given lattice as:

5.4.19.- QUANTIZED PHONON VIBRATIONAL MODES IN A LATTICE

Longitudinal Transverse

LO TO - O = Optical

LA TA - A = Acoustical

This brings us to the general case of phonon modes in a lattice. If there

are "y" atoms per primitive cell, i.e.- y different kinds of vibrating groups,
then there are:

5.4.20.- NUMBER OF BRANCHES OF PHONON DISPERSION

For y - atoms: Acoustical = 3

Optical =3y-3

Thus, for the case of three atoms present in the lattice, we would get the
following number of phonon branches, as shown in 5.4.21. on the
following page.

It becomes apparent that we can predict, and calculate the expected

number of phonon modes present in a given lattice, by knowing the types
and numbers of atoms present. We have now shown how phonons have
been defined as quantized lattice vibrations. It should be clear that
phonon energy is completely defined by Quantum Mechanics and that if
we wish to add or subtract energy during a photon interaction (such as an
excitation process in a phosphor), the amount that can be added is a
direct function of the type of atoms (ions) comprising the solid and its
structure.
 5.4 Phonons as quantized lattice vibrations 395

5.4.21.-

II. The Case of the Impurity Activator Center in a Phosphor

We can now show, in view of our discussion of phonon modes given above,
how the phonon frequency around the activator center differs from that of
the general phonon modes of the lattice. We already know that the
activator center in an inorganic phosphor is a substitutional defect,
located at a cation site in the lattice. In this case, we will have localized
phonon modes because of the induced perturbation by the presence of the
defect on the host lattice. What is not so apparent is that the local phonon
modes around the activator site differ from those of the lattice phonon
waves. This arises because the mass of the activator cation differs from
those of the cations of the lattice. We use the same mathematical
approach as given above, as follows:

5.4.22.- LOCAL PHONON MODES AROUND THE ACTIVATOR CENTER

Coordinates
lattice cation : M ui , U2
activator : Mo uo

Equations : M o d 2 uo / dt 2 = c [Uj + u-i - 2 uo]

M d 2 ui / d t 2 = c [u2 + uo -2 u-i]
396 Optical and electronic properties of solids

Combining these, we get:

5.4.23.- (o2M(o j = co^ax (M* / 2MMQ - Mo2)

where a>max is a cutoff frequency. In other words, there is a set of local

phonon modes surrounding the impurity activator site which has an upper
cutoff frequency.

5.5.- ELECTRONIC ASPECTS OF PHOSPHORS

We can, as a general rule, divide inorganic solids into 2 classes: those that
absorb light (pigmented) and those that emit light (phosphors and solid
state lasers). In both cases, we add controlled amounts of a transition
metal ion to control the absorption and/or emission properties of that
solid. At this point, it should be apparent that emission refers to:

1. Absorption of energy (whether by absorption of a photon or other

means such as energy from an electron)
2. Reemission of a photon (usually of lower energy, particularly if the
original source of energy was a photon).

Thus, one view of a phosphor is that it is a "photon-converter". An

inorganic phosphor consists of two parts:

(1) The HOST or inorganic compound

(2) The ACTIVATOR or the added transition metal cation

In general, the host needs to be transparent, or non- absorbing, to the

radiation source used for the "excitation" process. The activator does just
what the name implies, i.e.- it "activates" the host. This combination has
several advantages, the most important being that both the type and the
amount of activator can be precisely controlled. The details concerning
those elements that can be used to form the host, and those which can be
used as activators will be delayed until we reach the next chapter.

There are three terms we need to become familiar with, as follows:

 5.5 Electronic aspects of phosphors 397

5.5.1.- Terms Relating to Phosphors and Phosphor Emission

LUMINESCENCE - Absorption of energy with subsequent emission of light

FLUORESCENCE- Like luminescence, but visible emission

PHOSPHORESCENCE- Luminescence or fluorescence, but with delayed

emission of light

Luminescence is the general case in which a higher energy photon is

absorbed and a lower energy photon is emitted (such a process is called a
Stokes process). In this case, the excess energy is absorbed by the solid
and appears as lattice vibrational (heat) energy. We shall use fluorescence
and luminescence interchangeably, even though, strictly speaking, they
are not equivalent. Phosphorescence involves a process where photon
absorption occurs but the reemission process is delayed. This delay may
be a function of the type of transition metal employed, or a function of the
action of solid state defects, including vacancies and the like which "trap"
the energy for a time. This occurs most often in the sulfide phosphors.

Thus, we need to define a characteristic time of decay associated with

luminescence and phosphorescence. In general, we find that the rate of
decay is an exponential process. The types of decay that we find in
phosphors includes the following:

5.5.2.- Decay Times Associated with Luminescent Processes

RANGE OF DECAY TIMES (t = 1/e)

Fluorescence 10' 9 sec. to 10~ 3 sec.

Phosphorescence 10*3 sec. to 100.0 sec.

The decay time, t, is defined by convention as the time for steady state
fluorescent intensity to decay to 0.3679 (1/e) of its original value.
398 Optical and electronic properties of solids

Note that there are some materials which glow for hours after being
excited. This is not due to phosphorescence but to the presence of traps,
i.e.- lattice defects, put there deliberately during their manufacture. Such
phosphors, notably CaS activated by various cations including Bi3+, are
sometimes called "daylight" phosphors since they can be excited by visible
light. Once activated, they will glow for hours.

We can now define a phosphor as a solid state material which emits in the
visible part of the electromagnetic spectrum (this is not strictly true since
some phosphors emit ultraviolet light and some emit infra-red light). Let
us now examine the configurational aspects of phosphors and then the
energy transfer mechanisms which occur once the phosphor is excited.
Later in the following chapter, we will examine the solid state chemistry
of phosphors.

I. Energy Processes in a Phosphor

Phosphors may be classified as organic (fluorescent organic dyes, as used

in dye-lasers) and inorganic (as used in television or fluorescent lamps).
We restrict ourselves to the latter for a very practical reason. Although
organic fluorescent dyes have found wide spread usage, it remains
indisputable that the specifics of the "activator" center have not yet been
described, except in general terms of electronic structure of the organic
compound. It will become obvious, in light of our later discussions, that
the actual activator center in organic dyes must, of necessity, involve an
electron or electrons, isolated from the vibrational- electronic (vibronic)
structure of the organic molecule. Yet, the exact electronic structure of
the center in this class of luminescent materials has not yet been
established, even at this late date. We will address the approach used to
describe organic phosphor excitation-emission mechanisms later in this
discussion.

A phosphor is an energy converter. If a photon of 4.95 ev. (2500 A or

40,000 cm- 1 = ultraviolet light) is absorbed by a phosphor, then the
phosphor might emit a photon of say 2.48 ev. (5000 A = 20,000 cm" 1 =
green light). This is a so-called Stokes process where less energy is
 5.5 Electronic aspects ofphosphors 399

emitted than is absorbed. Note that there are also anti-Stokes processes
which will be discussed in a following chapter. (The "extra" energy comes
from lattice vibrations). The specific energy processes occurring in a
phosphor will, of course, depend upon the exact composition of the
phosphor.

The sequence of energy processes that occur within a phosphor include:

5.5.3.- Energy Dissipation in a Phosphor

a. Absorption of energy (from a variety of sources)

b. Excitation within the activator center to form an excited state
c. Relaxation of the excited state (it is here that energy is lost to the
vibrational states of the lattice)
d. Emission of a lower energy photon from the excited state, and
relaxation to the original (ground) state

In a phosphor, absorption of energy may occur in the host, or directly in

the activator center (the added transition metal ion). But it is the activator
center (or site) which becomes excited, whether by transfer of energy
from the host, or not. That is, the activator center absorbs the energy and
changes its electronic energy state from a ground state to an excited
state.

There may be several excited states possible, but only the lowest excited
state is involved in photon emission. If more than one excited state does
occur, each excited state relaxes to the next lower energy excited state
until the final excited state is reached, whereupon photon emission
occurs in due time. Whether upper excited states are involved depends
upon the original energy of the exciting photon. This energy may exceed
the minimum energy needed for excitation and the excess energy causes
the upper excitation states to become populated or even ionized. It is
common to represent the energy states of a phosphor in terms of a
potential energy diagram containing only the ground state and the lowest
energy excited state, as follows:
400 Optical and electronic properties of solids

5.5.4.-

This is a representation for a single activator center. Herein, we use

spectroscopic notation for the ground and lowest energy excited states
(We will explain this later). Thus, the electronic energy (involving a one
electron transition) is:

iSo => !Pi

Let us now interpret the above diagram so that we can understand the
electronic processes which occur in a phosphor.

According to the Franck-Condon principle, a change in electronic state

takes place in a time nearly instantaneous, as compared to other
processes within the activator center, i.e.- 10' 18 sec. (see 5.2.12. given
above). In our diagram of 5.5.4., we have shown an isolated activator
center within a given host lattice. Take note that there are several ground
state energies, of sharply defined radius values (the horizontal lines, i.e.-
<r>). The lowest ground state energy is the lowest energy occupied state.
If we heat up the lattice, i.e.- increase its vibrational energy, then the
other ground state levels will become occupied. Note that their limits of
 5.5 Electronic aspects ofphosphors 401

<r> are wider than the lowest energy state. Each of these horizontal lines
denotes a separate, so-called, Stark state. When sufficient energy is
absorbed, then the center will become excited. Again, the values of <r>
are sharply defined in the excited state, but are different than those of the
ground state. This effect of vibrational modes of the host lattice upon the
electronic states of the activator center is called vibronic coupling, and is
caused by phonon wave propagation throughout the host lattice, and
interaction at the activator center. (Phonon waves are quantized lattice
vibrational waves present in all solids). Remember that unless the host
lattice is at 0 °K, then vibrational waves are always present. Once the
activator center has become excited, then it relaxes through the set of
Stark states until it reaches the lowest energy excited-state from which
emission occurs. Note that there is a range of excited states and that the
final excited state can also relax, under certain conditions, directly to the
ground state without emission of a photon. Then, all of the excitation
energy is dissipated by vibrational processes throughout the lattice.

Probably the most important point to realize is that this diagram is a static
one, whereas the reality is a dynamic one in which there is a large
plurality of centers in various energy states which vibrate according to the
effective values of <r> for that particular energy state each center finds
itself in. This gives rise to a density of states whose occupation centers
around the original excitation energy transition, as shown above, but
whose occupation density is affected, in the excited states, by phonon
(vibrational) waves of the host lattice. This variation in <r> is a
consequence of host lattice vibrations which affect the relative positioning
of the activator center. It is essential to realize that the variation in <r>
arises from a random interaction of phonons with the excited center.
That is, the density of states is Gaussian because the interaction
perturbation (vibrational + electronic = vibronic coupling) is a random
process.

Since the activator center is usually a substitutional defect in the lattice,

its vibrational modes will differ markedly from those of the hostfas shown
below). If both host and activator center were cooled to 0 °K, then the
electronic transition would be between a single Stark state of the ground
402 Optical and electronic properties of solids

state to a single Stark state of the excited state. This would result in
essentially a single energy line (and is called the "zero-phonon" mode).
However, vibronic coupling broadens the energy states both in the
transition from the ground(absorption) and in the excited state (as shown
by the variations which can occur in <r> at each Stark state in the
diagram). This action results in a broad band both for excitation and for
emission. We emphasize that the 4f emitting states of the rare earths are
shielded from the perturbation effects of the lattice by the outer valence
electron shells, so that line emission occurs rather than broad bands.

Note that in a broad band, each photon absorbed, or emitted, has an

incremental difference of energy from the other photons, depending upon
the vibrations, or phonon modes present during the instantaneous
moment of electronic change. It is this instantaneous change, as
secondarily affected by the slower phonon modes, which cause broad
bands of absorption (and/or excitation) and those of emission. It is
because phonon energies are quantized that they can add and subtract
from the electronic energy process.

From the above description, it should be evident that the electronic

excitation-emission transition is a dynamic process which is perturbed by
vibronic coupling of the phonon spectrum present in the host lattice.
Thus, the host is just as important as the activator center. Another way to
describe the overall process is to state that the electronic transition in
the activator center involves the zero-phonon line, broadened above
absolute zero temperatures by quantized phonon interactions to form a
band of permissible excitation and emission energies.

Now, what we have not said is that to have an efficient phosphor, we must
have a luminescence center that is isolated from vibronic coupling from
the host lattice. We will address this factor in some detail in the next
chapter.

We can represent the energy transition in the activator center of 5.5.4. as:

5.5.5.- AEexcitation = { I xPi I => I ^ o 1} ± n h a^

 5.5 Electronic aspects of phosphors 403

where h = h/2jr (h is Planck's constant) , ^ P i l represents the virtual

energy of the excited electronic state in the crystal lattice and hcoj is the
energy of a single phonon state (remember, there is a plurality of sharply
defined phonon energies present, i.e.- the phonon spectrum of the
lattice). Note that a relaxation occurs in the excited state. This results in a
change of <r> (actually how far the excited center can change in effective
radius <r>), as compared to that of the ground state). This shift is a
virtual one and is called the "Stokes Shift". Thus the effective position of
the excited state has changed relative in energy to that of the ground state
and results in a broad band for absorption (excitation in the case of
phosphors) and a broad band for emission, as shown in the following
diagram:

5.5.6.- Energy Bands Associated with Phosphors

In this diagram, we have shown the absorption band and the emission
band as being completely separated from each other for clarity. Usually,
they are not. The zero phonon line is the same as the electronic
transitions shown in 5.5.4., i.e.- the vertical arrows. Thus, the Stokes
shift arises because of a change in <r>, as we have said, with the added
hcoj phonon energy perturbations (vibronic coupling) which cause a broad
band of absorption (excitation) and emission. Note that the rare-earth
activators do not exhibit this behavior because the 4f° emitting energy
404 Optical and electronic properties of solids

levels are completely shielded by the outer 5s2 , 5p6 electron shells. Thus,
they emit sharp line spectra, in contrast to the "ordinary" activators which
are coupled to the phonon modes of the host lattice.

The time-frame in which these energy processes take place are:

5.5.7. Times Associated with Energy Processes in a Phosphor

Excitation = 10' 18 sec. (Franck-Condon principle)

Relaxation = 1 0 8 sec. (vibrational)
Emission = 10' 6 to 10"3sec. (electronic transition)

In view of what we have already stated, it is now easy to see that for
emission to occur, vibronic coupling between the excited activator center
and the lattice must be minimized, since at least one hundred or more
lattice vibrations will occur before the photon is emitted. Otherwise, a
radiationless relaxation will occur. It is this fact that mandates careful
choice of the host used as a basis for preparing the phosphor.

We find that only certain cations of predetermined valence state can

function as activators if we wish to obtain high efficiency in the phosphor.
What this means is that certain phosphors may emit photons, once
excited, but a sizable fraction of the excitation energy goes to excited
phonon states, i.e.- heat in the lattice. Thus, to obtain a phosphor of high
brightness, we must choose an activator in which the combined spin and
angular momentae are zero, i.e.- AJ = 0. This will be explained in more
detail in the next chapter. The energy sources useful in excitation of
phosphors in general are shown in 5.5.8., given on the next page.

There are several things we need to explain. In the case of particles,

kinetic energy is lost by the particle and is transferred to the host lattice
via excited phonon states, i.e.- lattice states. This energy is then
transferred to the activator center and subsequently appears as excitation
and then emission energy. This phenomenon has been extensively studied
because of its importance to the TV industry.
 5.5 Electronic aspects of phosphors 405

5.5.8.- Sources Useful in Exciting Phosphors

PARTICLE ABSORPTION PROCESS

High speed electrons Cathodoluminescence, i.e. -TV
p - rays Cathodoluminescence
a - rays Radiation detection
PHOTON ABSORPTION
Ultraviolet and visible Photoluminescence
X - ray
Y - rays
Infra-red " " (Anti-Stokes1)
PHONON PROCESSES
Heat caused by conduction = Thermoluminescence
Heat caused by friction = Triboluminescence
Heat transferred by flame = Candoluminescence

It has been found that when an energetic electron strikes a solid particle,
it loses its energy in a number of ways. The following diagram shows some
of these mechanisms:

5.5.9.- How an Electron Loses Its Energy in a Solid

406 Optical and electronic properties of solids

Depending upon the energy of the electron, the penetration depth

depends somewhat upon the constituent atoms of the host. Each electron
recoils from collisions with atoms (ions) in the lattice, slowly losing
energy with each ricochet until it can no longer lose further energy. Heavy
atoms are ionized easier than the lighter elements. At the left, "A" shows
how the electron depth is distributed when the energy of the electron is
low. As the electron energy increases, the depth also increases until
finally at greater depths, ionization of the constituent atoms begins to take
place. This mechanism is shown on the right hand side of the diagram.
The first excitations of the lattice are caused by the creation of optical
phonons. These phonons then travel to an activator site and transfer their
energy to the emitting center. At higher electron energies, plasmons (a
collective excitation of valence electrons of the host lattice) and /or
excitons (which are electron-hole pairs) are created.

Note that these mechanisms apply mostly to semi-conducting phosphors

like ZnS:Ag and the like. The plasmon soon decays (~ 10 15 sec.) to an
excited state electron in the conduction band. If ionization occurs, large
numbers of secondary electrons appear in the conduction band. This
accounts for the high cathode-ray efficiency of the sulfides in general.
Most lamp phosphors are insulators and do not respond well to electron
excitation.

When lamp phosphors are excited by ultraviolet photon energy, the

photon energy may be dissipated by the lattice or directly in the center.
Additionally, phonon energy may be the sole source of excitation in some
phosphors. Actually, some phosphors are capable of energy storage, as by
absorption of visible light photons, which is subsequently released by
phonon energy. However, this description is only a general depiction of
energy dissipation in phosphors and we need to lay a firmer basis of
associative properties before we can proceed further.

II. Properties Associated with Phosphors

The following are specific aspects of phosphors we need to discuss before

we can delve further into energy dissipation processes.
 5.5 Electronic aspects of phosphors 407

a. Notation

A language has grown up around the science of phosphors. Most

phosphors consist of a host composition plus the activator, added in
carefully controlled quantities. The activator itself is a substitutional
DEFECT and is subject to lattice phonon perturbations. Therefore, it is
essential that the charge on the substitutional cation is equal to that of the
host lattice cations. Otherwise, an efficient phosphor does not result.

We denote a phosphor as: MaYOb : N x , where M is the cation, YOb is

the anion (assuming we have an oxygen-dominated phosphor and not a
fluoride or sulfide), and N is the activator. It is understood that the lo-
cation is in solid solution in the host-matrix, and the above formula is
actually: [(l-x)Ma)Y0b)*xNYOb]- Thus for a tin-activated strontium
pyrophosphate phosphor, we would write:

5.5.10.- Sr2P2O7 : Sno.o2 = 0.99 Sr 2 P2O 7 * 0.01 S n ^ O y

In this case, the host cation is Sr 2+ and the activator cation is Sn 2+ . Thus,
the first part is a short-hand notation for the latter.

However, suppose we have the phosphor: Sr2P2O7: Sb3+o.O2 • In this case,

the actual formula is considerably different than the short-hand formula.
From our study of defect reactions (Chapter 2), we know that if a trivalent
ion substitutes on a divalent ion, it must do so in conjunction with a
substitutional defect, usually a vacancy. Thus the actual formula is:

5.5.11.- 0.99 Sr2P2O7 • 0.01 {(Sb3+, V+ ) 2 P2O 7 .

where both Sb3+ and V+ are on Sr 2+ lattice sites. If we added Na+ for
charge compensation, we would have:

5.5.12.- 0.99 Sr2P2O7 • 0.01 {(Sb3+, Na+ ) 2 P2O7

or Sr2P2O7 : Sb 3 * : Na+ .
408 Optical and electronic properties of solids

However, it is well to note that Na+ is not an activator in the strictest

sense but is a member of the lattice structure, placed there to avoid
formation of a lattice vacancy. If a vacancy is formed, it is likely that the
resulting "phosphor" may be non-luminescent or will exhibit a low degree
of luminescent efficiency. Although you will find that many phosphors like
ZnS:Ag", AT* : K+ are listed as requiring the K+ activator, in reality this
cation is present for charge compensation in the lattice. This fact has not
been recognized by many prior workers in the field. Furthermore, some
zinc (cadmium) sulfides have several listed "activators", part of which are
undoubtedly there for charge compensation. Examples include:

ZnS: Ag'iAl3*
ZnS:Zn° IAI9" :C1".

b. Quantum Efficiency

We have already stated that proper choice of host and activator is essential
to obtain an efficient phosphor. Now consider the case where one-
hundred (100) photons are incident upon the phosphor. From our
discussion given above, we know that a few are reflected, some are
transmitted, and, if the phosphor is an efficient combination of host and
activator, most of the quanta are absorbed. But not all the absorbed quanta
will result in an activated center, and, once these centers become
activated, not all will emit a subsequent photon. Some become deactivated
via relaxation processes. To determine just how efficient a phosphor may
be, we measure what is termed the "quantum efficiency", i.e.- QE. The
quantum efficiency is defined as:

5.5.13.- QE 5 photons emitted/ photons absorbed

To obtain specific values, we measure total energy of emission and the

total energy absorbed. It is easier to measure intensity of photons emitted
as a function of wavelength. This gives us:

5-5.14.- QE = {(I dMemission / fIdX) absorption^ 1 ' R^absorbed

(1 - R Reflected}
 5.5 Electronic aspects of phosphors 409

where I dX. is the integrated energy of the emission band or that of the
absorption band and (1-R) is the integrated energy of the absorption or
reflectance band. In this equation, we correct for reflection as well.
Generally, phosphors which have QE's of 80%, or greater , are considered
to be efficient phosphors. This subject will be discussed at greater length
in the next chapter.

c. Decay Times

We have already briefly discussed luminescence decay times. Reiterating,

the decay time of a phosphor has been defined as the time for the steady
state luminescence intensity to decay to 1/e, or 0.368, of its original
intensity. It has been found that the intensity of photon emission builds up
in the order of microseconds, i.e.- ~ 10' 6 sec. to a specific value, i.e.- the
excitation process takes only a few microseconds. Since the intensity also
decays in microseconds (if the excitation source is removed), there is an
equilibrium value attained in the presence of the excitation source, which
is a combination of both excitation time and decay time. This so-called
steady state is called I o , and is promulgated by the population of emitting
sites in the phosphor which are simultaneously being excited and
decaying back to the ground state (with emission occurring during the
process).

Two types of decay are normally encountered. We have already mentioned

exponential decay. The other is a logarithmic decay. It is essential to
realize that both types of decay arise from a Gaussian array of emitters.
That is, the energy states of the excited centers and the energy spectrum
of the emission band are the consequence of a random process of phonon
perturbations of the excited state of the activator center. Hence a random
distribution of excited states results. The distribution is Gaussian and the
decay times of this Gaussian array of potential emitters can be described
via the binomial derivation, as shown in 5.5.15. on the next page:

This equation presents the change in numbers of excited states with time
in relation to the decay time, x , and is a Gaussian expression for a
410 Optical and electronic properties of solids

5.5.15.- EXPONENTIAL DECAY

I(t) = - dn/dt = n / T and n = n 0 exp (- t/ x)

population of emitters where n is the number of emitters, x is the decay

time and t is a specific time. I(t) is the intensity of photons emitted per
second at that specific time. We can also make the substitution:

5.5.16.- I (t) = [no / T ] exp (-t 1%)

which may be more useful in some cases.

The other type of decay encountered is a logarithmic decay:

5.5.17.- I(t) = - d n / d t = a n 2 and I(t) « 1 / 1 2

The coefficient, a , has been determined to be a fractional constant. By

combining as above, we get:

5.5.18.- n(t) = n o / (n o a t + 1) and I(t) = n 2 / (n a t + 1 )

Fortunately, we do not encounter logarithmic decays very often and

sometimes it is difficult to differentiate between the two types. The only
way we can really differentiate between the two is to obtain the decay
curve and to subject it to mathematical analysis.

However, there is a very useful reason for determining the type of

luminescence decay curve present. Confirming the presence of an
exponential decay curve means that only one type of emitter is present. If
a logarithmic decay process is found, it usually means that more than one
type of emitting center is present, or that two or more decay processes
are operative. While this does not occur very often, it is useful to know if
such is present. This phenomenon occurs more in cathode-ray phosphors
than in lamp phosphors, i.e.- sulfides vs: oxide- hosts.
 5.5 Electronic aspects of phosphors 411

d. Band Shapes

As we have said, the excitation process results in a density of states

arising from a random process of phonon perturbation of the excited
state, both before it relaxes and afterwards as well. It is this random
formation of Gaussian energy states that give rise to a broad band in
excitation, and to a broad band in emission. The zero-phonon line arises
from the nature of the electronic transition taking place which is
broadened by the vibronic coupling process. As the temperature rises,
increased phonon branching results and the emission band is even further
broadened, as follows:

5.5.19.-

In the above diagram, the emission intensities of the three bands remain
constant. It is the peak intensity which changes as the band broadens. To
this point, we have accepted the fact that vibronic coupling leads to broad
band excitation and emission in a phosphor. Take note that the above
diagram is the result of experimental measurements which prove that as
the temperature increases, the phonon spectrum becomes broadened,
thereby leading to broadening of the bands. Thus, at - 200 °C, the number
of phonon vibrations is restricted and a rather sharp emission band is
seen. As the temperature rises, the number of separate phonon branches
increases (the empty phonon levels become occupied) and the emission
(excitation) band is further broadened. Note that at some temperature
above 300 °C. (in this example), the phonon vibrations increase to the
412 Optical and electronic properties of solids

point where the emission intensity decreases to zero. In this case, what
has happened is that relaxation processes through vibronic coupling have
dominated the overall electronic excitation and emission processes and
the phosphor loses brightness, i.e.- a loss in the number of emitted
photons actually emitted.

It is important to realize that, in our examples, we have dealt with a single

activator site. The actuality is that there is a plurality of activator sites
present in the host, each of which, at any given instant, has a slightly
different energy state from its neighbor, because the randomness of the
vibronic coupling process. Thus, each activator site emits a photon of
slightly different energy (wavelength) from its relaxed excited state,
leading to a broad emission band. Thus, there is a Gaussian array
(Statistical Ensemble) of absorbers and emitters. The energy differences
among them comes from:

1) Vibronic coupling of the electronic excited state(s) of the

activator center

2) Phonon addition and subtraction to cause occupation of adjacent

Stark states, particularly in the excited state

3) Relaxation processes.

On an energy basis, we end up with an array of centers in very slightly

different energy states, which are further perturbed by the vibronic
coupling present at the site.

At this point you may have wondered why we have always stated that it is
the excited state which is subject to vibronic coupling, and not the ground
state as well. As will become apparent in the next chapter, only certain
types of activators can be used to prepare a phosphor. These involve
cations which have a ground state, iSo . In this type of electron structure,
vibronic coupling is minimized or is zero because of the closed electron
shells present (the electrons spins and angular momentae are all coupled)
which prevents phonon perturbation from occurring.
 5.5 Electronic aspects ofphosphors 413

e. Spectroscopic Notation

Optically active cations generally encompass the transition metal series of

the periodic table. Atoms or ions with closed electron shells are not
optically active, that is- they do not interact with photons having energies
lower than about 5.00 ev. Examples of closed shell ions are:

5.5.20- Na+ = Is 2 2s 2 2p6

Ca2+ = Is 2 2s 2 2p63s 2 3p 6 (Ar Core)
Zn 2 + = (Ar Core) 3d 10
Sn 4 + = (Kr Core) 4d 10

On the other hand, optically active ions have electrons external to the
closed shells which can be excited by a photon of proper energy.
Examples are:

5.5.21.- V3+ = (Ar Core) 3d 1

Mn2+ = (Ar Core) 3d 5
Sn2+ = (Kr Core) 4d 1 0 5s 2

Note that Sn 2+ is optically active whereas Sn 4+ is not. In order to

describe the energy states of an optically active cation, we must turn to
the field of spectroscopy. The most often used method of notation is that
of Russell-Saunders terminology.

It is well known that in a multi-electron atom, each electron present is

characterized by its own set of unique quantum numbers. These numbers
involve: 1 - the angular momentum of the electron; s - the spin quantum
number; and m - the magnetic quantum number. The total momentum of
any electron is well characterized by: j = 1 + s for the single electron.
(The magnetic quantum number is ignored here because it only appears in
the presence of a magnetic field). When there are more than one electron
present in the atom, we must combine the angular momentae and the
spins of the electrons to give a single value (which hopefully corresponds
to that observed for the atom containing the number of electrons we are
trying to couple). In Russell-Saunders coupling, the spin-orbit coupling is
414 Optical and electronic properties of solids

done as follows. The individual angular momentae are vectoriaily summed

together, as are the individual spin momentae. Then, the sums are
combined to give an overall value, namely:

5.5.22.- L=2l ; S = 2 s ; J = L ± S

Russell-Saunders coupling is used because it agrees with experimental

values most of the time. However, there are cases where it has been found
better to sum the individual "j" values and then sum these to give the final
value.

In Russell-Saunders coupling, sometimes called L-S coupling, the

individual the spectroscopic terms are based on the L - values obtained
(which are small whole numbers). The spin term is a resultant of 2S + 1.
The central term is thus defined like this:

2S+lLj

L- values and tenns are given in the following:

5.5.23.- RUSSELL-SAUNDERS TERMS

L - Values Term

0 S
1 P
2 D
3 F
4 G
etc.

Actually, these terms originated with Balmer (1890) who first organized
the spectra of the ionized hydrogen atom. He named each succeeding
energy series as:

S = singlet; P = principle ; D = diffuse and F = fine.

 5.5 Electronic aspects of phosphors 415

Nowadays, we use the same letters but S is a singlet state, P is a doublet

state and D is a triplet state, etc. Thus in the doublet state, 3 Pi , S = 1 , L
= 1, and J = 1. It has been found that L-S coupling adequately describes
the optically active energy states for nearly all atoms and ions, with the
exception of the rare earths.

It is usual to write a "Hamiltonian" so as to calculate the energy states of

the atom. The method most often used is the so-called "central-field"
approximation. Therein, it is assumed that the Hamiltonian for the
optically active electrons is composed of several terms: the central field
Hamiltonian- Ho , the coulomb field-He , the spin-orbit field-Hso . and
the crystal field-Hcf. For L-S coupling, crystal-field energies are greater
than spin-orbit energies. The following diagram gives the relation
between the various Hamiltonians:

5.5.24.- HAMILTONIANS FOR A TWO-ELECTRON ATOM- Is 2 2s 2 2p 2

The first thing to note is that only 2 p-electrons are involved. The others
are closed-shells. For two p- electrons, 3 terms arise in L-S coupling, 6
levels and 21 Stark states. It is easy to see that quite complicated spectra
could arise from a cation incorporated into a lattice under the influence of
416 Optical and electronic properties of solids

its crystal field, especially if all transitions between all levels were
possible. Fortunately, there are "selection rules" for electronic transitions.
These selection rules govern observed intensities and the ease by which
these transitions take place are determined by limitations placed upon
changes in quantum numbers. The allowed transitions are:

5.5.25.- SELECTION RULES FOR ALLOWED TRANSITIONS

AL=±1 AS = 0 AJ=±l,0

but not J=0 to J=0. Actually these selection rules apply only to DEPOLE
TRANSITIONS, i.e.- transitions between electric dipole states. There are
other states as well.

The total number of types of "multipole" states that can exist are:

5.5.26.- MULTIPOLAR RADIATION TRANSITIONS POSSIBLE

Electric dipoles Magnetic dipoles

Electric quadrupoles Magnetic quadrupoles
Higher multipole orders

There are also various combinations of the above, and as the atomic weight
of the atom increases, the above selection rules do not hold rigorously,
because the radiation states become "mixed", i.e.- they are not pure "spin-
orbit" coupling states any more. That is, L-S coupling does not hold. For
this reason, prediction of spectral intensities by selection rules is not
exact for heavy atoms, as we shall see for rare earth (lanthanide) cations.
Note that the above description applies only to inorganic cations, where
the electronic transitions (excitation) occur within the outer electron
shells. Phosphors based on organic compounds show quite different
mechanisms.

III. Organic Phosphors and Their Luminescence Mechanisms

The luminescent center in organic phosphors has been described in

 5.5 Electronic aspects of phosphors All

terms of molecular orbitals of the compound, rather than by spectroscopic

notation of an isolated center. The energy levels are usually designated as
"HOMO andLUMO", i.e.- "highest occupied molecular orbital" and "lowest
unoccupied molecular orbital", respectively. However, as we will show,
this explanation is not entirely satisfactory from an energetic standpoint.
To show what we are addressing, consider the following diagram:

5.5.27.-

In this diagram, we show the energy levels related to organic

luminescence mechanisms as explained by molecular orbitals. Conjugated
molecules like Anthracene, Pyrene and Perlylene show high quantum
efficiencies. That is, their response to ultraviolet light rivals that of the
inorganic phosphors. There are other configurations of luminescent
molecules as well. Yet the exact configuration of the luminescent center
has remained obscure.

At the left, both the HOMO and LUMO levels are shown in relation to the
paired bonding orbitals of the molecule. At the right, the HOMO and
LUMO energy levels contain So , the ground state, and two "singlet" states,
the excited states. All of these states are perturbed by structure vibrations
418 Optical and electronic properties of solids

and so have vibrational levels superimposed upon both the ground state
and the excited state(s). Note that excitation occurs in the molecule from
the ground state to the singlet states since the transition to the triplet
state is strictly forbidden by spectroscopic rules. Once energy has been
absorbed and the transition to the excited state(s) has occurred, a
relaxation among the vibronic levels occurs, as shown by the wavy lines in
the diagram. Relaxation can occur from S 2 to Sj or even to Tj. Emission
then occurs from either S t or T : .

However, this tells us little about the nature of the activator center. It is
obvious that the organic phosphor center must be isolated from the
vibrations within the molecule. Otherwise, any excitation energy would be
dissipated by vibronic coupling to the structure as heat. Thus, the exact
localized structure of the organic luminescent center remains to be
elucidated. Let us now examine the factors involved with transfer of
excitation energy from a lattice point in inorganic phosphors and its
conversion to emission. We will find that this mechanism is quite involved
in that quantum mechanical resonance and energy exchange is one of the
main methods by which absorbed energy is transferred from an excited
point in the lattice to an emitting center.

5.6.- FACTORS ASSOCIATED WITH EXCITATION ENERGY CONVERSION

There are two types of energy transfer from a point in a lattice to a

luminescent center. For semi-conducting phosphors like ZnS, migration
of electrons in the conduction band, or holes in the valence band, conveys
the excitation energy to localized luminescent centers. Excitons (bound
electron-hole pairs) is another mechanism. For insulators (which are
generally oxygen dominated compositions), the excitation energy can be
transferred from an excited point in the lattice, or from an excited
luminescent center to other unexcited centers in the lattice by Quantum
Mechanical resonance.

Resonant energy transfer mechanisms include those given in 5.6.1.,

presented on the next page.
 5.6 Factors associated with excitation energy conversion 419

5.6.1.- Multipolar interaction

Exchange interactions
Phonon assisted energy transfer
Excitation diffusion
Sensitization of luminescence

We will address all of these in the following discussion.

We can summarize the material in Section 5.5. as follows. Essentially, we

have treated a phosphor as a one-electron process. The sequences for
changes in electronic states are:

5.6.2.- ELECTRONIC CHANGES

ABSORPTION EXCITATION EMISSION

ForSn2+: hvi lS 0 => J Pi !

Pi => ^ o + hv 2

The energy dissipation changes associated with these electronic changes

in the Sn 2+ center are shown in the following:

5.6.3.- ENERGY DISSIPATION PROCESSES

=
^absorbed E o r igi n a i - Ereflected

Eexcitation = Eabsorbed - Eiattice absorption

Eiattice absorption = Eexcitation relaxation + Eemission relaxation

+ Eexcited-center quenching relaxation

And the total energy involved is the sum total of these:

5.6.4.- E or iginal = Eemission + EeXcitation + Eiattice absorption + E re fi ec t e( j

where lattice absorption is all the relaxation processes which result in

loss of energy. However, this equation does not include vibronic coupling
where energy flow from the lattice to the center occurs. Now, we can
420 Optical and electronic properties of solids

finally present all of the time processes as they relate to phosphors as

follows. This is shown in the following:

5.6.5.-Time Scale for Energy Dissipation Processes in Phosphors

Eabsorption = 10" 18 sec.

Eexcitation = 10" 11 sec. (electronic transition)
Elattice absorption = 1O' 8 sec (vibronic coupling)
Eemission = ICC9 to 10'2 sec. (electronic transition)

We have repeated this data for emphasis. Note that the time required for
absorption and then excitation to take place is about one thousandth of
the time required for one lattice vibration.

Or, in other words, the activator center is not perturbed until it absorbs
energy and transforms to an excited state. Once it has done so, then it can
relax to a equilibrium excited state from which emission can occur almost
instantaneously, within 100 vibrations or even longer, i.e.- as much as
10,000 lattice vibrations.

I. Prediction of Electronic Transition Intensities

We now embark upon one of the more difficult of the concepts to be

presented. We have already presented the types of "poles" possible
without defining their nature. There are a variety of electronic transitions
based upon the spacial configuration of the electrons involved. These
interpolar interactions are responsible for transfer of electronic energy
from site to site in the lattice. As shown below, there are "dipole-dipole"
interactions, i.e.- "dd" as well as "dipole-quadrupole" interactions, i.e.-
"dq". There are even "quadrupole-quadrupole" transitions. I.e. "qq".

a. Multipole Interactions

Perhaps the easiest way to visualize this is to imagine a "p-electron" and

its electronic density lobes in space, as shown in the following:
 5.6 Factors associated with excitation energy conversion 421

5.6.6.- ELECTROPOLE RADIATORS

Note that the electric field vector, F , points in the direction of greatest
electron field density. In contrast, the magnetic field vector, M , will be at
right angles to that density. This affects the degree of coupling that can
take place so that transition intensities between electric and magnetic
"poles" are orders of magnitude lower than those involving like kinds of
poles, i.e.- electric dipole to electric dipole vs: electric dipole to electric
quadrupole. One dipole can affect another at fairly great distances in the
lattice, i.e.- F • F , and this is called an electric dipole (d,d) interaction.
Obviously, we can have transitions between any two types of poles given in
5.6.5. but the transition intensities will vary according to the type of pole
interaction involved. Einstein (1903) was interested in this problem and
defined many of the terms we use today.

II. Einstein Absorption and Emission Coefficients

Consider an atom capable of absorption and emission. Einstein defined

the following:

5.6.7.- DEFINITIONS APPLICABLE TO TRANSITION INTENSITIES

Ground State Excited State
States: n m
Numbers of Atoms Nn Nm
Energies of Atoms En Em

You will note that he specified both excited and ground states in terms of
both numbers and energies of the atoms involved.
422 Optical and electronic properties of solids

Einstein then proceeded to define an intensity of absorption as:

5.6.8.- I absorption = N n Bnm h m Al p(m) dco

where: E = hn ; Al = pathlength ; o = frequency of radiation;

p (oo) = frequency density, and:

Bnm = Einstein coefficient of Absorption

Bjnn = Einstein coefficient of Induced Emission
Ajnn = Einstein coefficient of Spontaneous Emission
(from one pole to the other in the order given)

Thus, n is the ground state here while m is the upper energy state.
Einstein noted two types of emission and had to define two (2) types of
emission coefficients. These were delineated as two types of emission
processes:
1) spontaneous
2) induced.

The former occurs in phosphors while the latter is the basis for laser
action. From Beer's Law, we get:

5.6.9.- I = I o exp (- £ c 1 )

and: N n = N o exp (- E n /kT) ;

N m = N o exp ( - E m /kT).

That is, the numbers of atoms in both the ground state and excited state
is an exponential function of the energy of either. Combining these
equations gives us:

5.6.10.- N n / N m = exp [ - { E n - E m }/ kT]

As in 5.6.8., Einstein's definition of emission intensity for spontaneous

emission is:
 5.6 Factors associated with excitation energy conversion 423

5.6.11.- Iemission= N n Amnht»C

where c is a constant, i.e.- the velocity of light and h is h/2n. Using this,
we can show that:

5.6.12.- N n B n m p M =N m B m n p(co) - NnAmn

at a steady state condition of equilibrium. Note that these Einstein

equations are in terms of a frequency density, p(m). We now redefine the
equations in terms of p(v) which is a radiation density. This gives us:

5.6.13.- p(v) = (N n Am n )/[B nm (N m - N n )]

= Amn/Bnm {{exp[(- Em + En / kT) -1] A}

Note that these two equations are equal to p(v). They state that the
radiation density (N n A mn ) divided by the difference in numbers of excited
and ground states is equal to the ratio of the emission coefficient divided
by the absorption coefficient times the exponential value of the reciprocal
of the differences of energy of the emission and excitation. In comparison,
one form of Planck's Law is:

5.6.14.- p(v) = (8rc h v 3 ) / c 3 • l/[exp[(hv/kT) -1]

Making this substitution gives us:

5.6.15.- Amn = (87T h v 3 ) / C 3 ) Bnm = 8n hco2 Bn m

Einstein's equations give us a way of obtaining values for the two types of
coefficients which we can use to further develop other concepts regarding
energy transfer in solids, particularly those processes associated with
phosphors.
c. Electronic Transition Moments

Electronic transition moments are defined as the probability for a given

excitation energy transition to take place. It should be evident that the
transition moment depends upon the spin-orbit coupling of the electrons
424 Optical and electronic properties of solids

in both the ground and excited states. This mandates the use of wave-
equation methods and quantum mechanics. At first, we shall restrict our
discussion, while attempting to be lucid and concise. We will then present
the details of the required calculations to be complete. You may wish to
skip the latter part now but will have it readily available for future use.

1. Concise Hamiltonian Calculation of Transition Moments

By use of the Schroedinger equation, i.e.

5.6.16.- HW=E¥

where H is the so-called Hamiltonian operator and E is the eigen (energy)

states, we can solve for the wavefunctions, W, in terms of specific time-
dependent coefficients, ak (t), as:

5.6.17.- W= 2ak(t) ^ k

This allows us to define a probability of emission in terms of these

coefficients as: am* • am , where am* is a conjugate (see the details
given in the next section). By suitable manipulation, where we solve for
the energy of the eigenstates, £ :

5.6.18.- £ = J(W\ lEkl * F k d T ) / J {Wk * Wk dT ) =

(a*k ak Ek)/ I (a*k ak)

This is the so-called Hartree-F6k method used in quantum mechanics,

where T is a characteristic time as differentiated from t, an undefined
time. By defining a dipole moment as Hi • 2 ei xi , where xi is a unit
vector in the x-direction, and using the electric and magnetic vectors, F
and M, already defined, we can get an equation relating Bmn and m, as
follows :.-

5.6.19.- 4TT2 / h.2c (m x J V . (PX(<») ) 2 t = Bmn [6/4TT (PX (CD) ) 2 t

and: Bmn = 87r2/3h2 c (m x )2mn
 5.6 Factors associated with excitation energy conversion 425

This gives us:

5.6.20- A™ = Sn hco)2 B m = 647T4 w3 /3h (m x ) 2 mn

Note that we now have both Einstein coefficients in terms of the dipole
moment. If we can obtain a value for either, we can then calculate the
dipole moment, or vice-versa. However, the most important part is the
dipole transition moment which can also be written as an integral:

5.6.21.- (m x ) m n = < W* m | Z e i xi I ^ n d T >

Sometimes, this form is more useful.

2. Detailed Calculation of Electronic Transition Moments

We start with the one-dimensional, time-dependent, SchrOedinger

equation:

5.6.22.- H ? = - (h/2:r i) d»P/ dt

A "ket" (half of a total integral) for this Hamiltonian is | L S J J' z > . It is

well known that the Hamiltonian actually consists of two parts, the
unperturbed part and the time-perturbation part, i.e. -

5.6.23.- HOW + H'(t) W = - (h/27r i) &VI dt

To solve for the wavefunction, W, we obtain values in terms of specific

coefficients, ak , i.e.-

5.6.24.- W= 2 ak (t) WOk

This involves the so-called Hartree-F6k, or Slater determinant, methods,

named after the originators of the calculations. (Actually, Hartree-FOk
methods use wavefunction integrals whereas Slater shortened the method
of approximation by use of determinants). For the time-perturbed
Hamiltonian, we get:
426 Optical and electronic properties of solids

5.6.25.- Zak(t) Wo^ H'(t) Wok =-h/2jri-2(d[a k (t)]/dt) WOk

One of the rules In quantum mechanical calculations is that we must

always multiply by the conjugate in order to maintain symmetry. In this
case, the conjugate is: Y°k*. We also define the probability of emission as:
am- aM* (where the conjugate is also used). By combining and separating
the equations, we get:

5.6.26.- - d[ak (t)]/dt = -2TT i/h exp 2JT i (Em - E n ) (t/h) J**P*m H ' m ^ n dT

where we have introduced the subscripts, m & n, as defined above, along

with T , a characteristic time. As t approaches T, we get the following
eigenstate solutions:

5.6.27.- £ = IT V k | E | Wk dT ]/ [ W m Wn dT ] = £a* k s* E k / I a*k %

We now define a dipole moment as:

5.6.28.- m = I ei Xi

for a uni-dimensional operator, with xi as a unit vector in the x- direction.

Then:

5.6.29.- ~t = electric field vector = £ei

M = magnetic field vector

such that the Hamiltonian:

5.6.30.- H'(t) ="?(w, t) m x or: (IllxJid, = / < W k | m x | W k dT>

In this case, m x is ahermetian operator. We thus get:

5.6.31.- p(ca) doo = 1/87C [ (t{w, t)} 2 +{ M (co, t)2} ] = { t (ro, t ) 2 } /4TT

where F is defined as:

 5.6 Factors associated with excitation energy conversion 427

5.6.32.- "? (ro.t) = 2 "F°x cos(2n c cot) = ? ° x [exp {lit ceo t) + exp(-2jr cm t)]

and where all quantities have been set including c = speed of light. We can
now redefine our general equation into terms of the dipole moment and
the electric field vector:

5.6.33.- < W m [ H'(t) | Wm dT > = % (co, t) • t% (m, n)

Upon integrating:

5.6.34.- am(t) = Mx (m, n) ~?x («) { 1- exp f27r i fEm/h - En/hl + c oo t}

(Em - E n ) + he to
+ 1 + exp f27r i [Ern/h - En /hi + c co t\
(Em - E n ) + he co

If we now square the above equation, we get:

5.6.35.-

am(t) a' m (t) = 4[Mx (m, n)]2 J | ? x (co) |2 { sin^ fwt fEm/h - En/hl - c m )
(E m - E n ) - he co

Note that we used the conjugate and that F x (co) is now an operator in the
integral. If F x (co) is allowed to become a constant, we get:

5.6.36.- am(t) a*m(t) = 47r2/h2c [ Mx (m. n)]2 (fx (co)]2 t = Bmn p'(co) co t
and: Bran = 8jr2/3h2c (m x ) 2 m n

We then arrive at the following:

5.6.37.- Aatn = 8JT heco2 Bmn = 6 4 ^ w 3 / 3 h (II1X ) 2 m n

Now we have both Einstein coefficients in terms of the transition

moment, m. If we can obtain values of either, then we can calculate the
transition moment, or vice-versa.
428 Optical and electronic properties of solids

However, the most important part is that the transition moment can also
be written as an integral:

5.6.38.- ( m x ) m n = < ^ * m I 2 e i xi | W n d T >

Note that this equation is identical to 5.6.19., already derived above.

Now, we will consider oscillator strengths in light of how they contribute

to energy exchange by various mechanisms between excited and
unexcited sites in the lattice of a phosphor.

d. Dipole and Multipole Oscillator Strengths

If there is an exchange of energy between two atoms, and/or two energy

states, the electric field interaction can be characterized by both electric
pole and oscillator strengths. Both of these relate to the transition
moment, which in turn relates to the selection rules governing the
electronic transition. If (Illx)mn is large, then Dmn • the dipole strength,
will be large. Additionally, the oscillator strength, will also be large.
Usually, the electric field strength of the individual electrons of a given
atom, ei, will be equal. This allows us to rewrite 5.6.38. as:

5.6.39.- (life )mn = ei < W*m | Exi | Wn dT > = e (Q mn )

where Qmn is now defined as an oscillator strength. In 5.5.24., we

showed the effect of increasing electric field strength upon the
spectroscopic states which appear. However, not all states may appear.

For example, the 3P2 state has five Stark states which can appear, but if
only two states appeared (due to a diminished electric field), one state
would be doubly degenerate and the other triply degenerate. For this
reason, we need to rewrite 5.6.19. to include G, the degeneracy (which is
a small whole number), as follows:-

5.6.40.- Amn = 647T4 /3h G(m) 2 m n

Bmn =8JT /2h2iG(m)2 m n
 5.6 Factors associated with excitation energy conversion 429

Having this result allows us to define the dipole strength as:

5.6.41.- Darn = G (Q m n) 2 (in units of cm 2 )

Note that at this point, we have defined the dipole strength in terms of
the oscillator strength. We do this because the oscillator strength, while a
valid quantity, is much harder to quantify than the dipole strength. The
quantities are quite different. Dipole strengths refer to electric field
strengths between radiating dipoles, i.e.- different atoms which may be
separated by several lattice sites, whereas oscillator strengths refer to
interactions between states, usually of the same atom.

Thus, Dam is actually an electric field density in terms of the transition

moment, m m n . We can thus derive the equations relating dipole strengths
with the Einstein coefficients as:

5.6.42.- ^ n n = (64n 4 w 3 e 2 /3h) Dn,n = 7.24 x 10 1 0 oo2 Dmn

Bmn = (8*3 e 2 /3h 2 c) E W = 14.50 x lO 2 ^ E W

We have now connected dipole strengths with all of the above

relationships. In studies of phosphors, one finds these relationships to be
very useful, since they give an insight into the energy changes taking place
within the phosphor, particularly those that affect quantum efficiency and
brightness.

To relate oscillator strengths to the above derived equations, we now

return to our definition of absorption intensity, as follows:

5.6.43.- Iabsorption = N n Bnm h e r o p(co)

By suitable manipulation, we get:

5.6.44.- Iabsorption = N n (8:r 2 /3h 2 c) G(m x ) 2 m n h c t o p M

= N n 7r/(e 2 /me c) G (m x ) 2 mn
430 Optical and electronic properties of solids

where the quantity,/, has been derived, like Dmn. an<^ shown to be:

5.6.45.- / = (8TT2 <ame c/3hc 2 ) G (m x ) 2 mn

Here, e is the electron charge, / i s the oscillator strength, and m e is the

mass of the electron. It has been found that a better definition of oscillator
strength can be given in terms of the electric field which exists between
states, as follows:

5.6.46.- / = mc 2 /jre 2 J Ema dy

where E m n is the electric field strength between the two states. This
allows us to transpose values o f / i n t o terms of the dipole strength and
Einstein coefficients:

5.6.47.- / = (Siikne /3h) Dmn = 1.085 x 10 1 1 Dmn

/ = (he 2 m e / 7t e2) co D m n = 7.483 x I O 1 5 co Dmn

At this point, you may be somewhat overwhelmed with all of these

formulas. Yet, their derivations are necessary because they form a basis of
discussion for following subjects whose understanding would be
impossible without reference to the foregoing equations.

From our definition of intensity as a function of time, as follows: -

5.6.48.- I(t) = I o exp (-t / T)

When t = % then we will have:

5.6.49.- I(t) = I o /e = x = 1.5/fv 2 = 1 / Amn

In our case, T turns out to be the decay time for the spontaneous emission
and is a very accessible and useful quantity. By measuring T, we can get all
of the above-derived quantities, including Io. What we have really
accomplished is to show that rather complicated concepts relating to
 5.7 Mechanisms of energy transfer in solids 431

energy absorption and energy transfer in a solid can be translated to fairly

simple terms which we can realize and grasp.

We find that the intensity of emission is a function of the oscillator

strength (for spontaneous emission). This leads us to multipolar selection
rules, which are a useful way to classify electronic transitions and
interactions. We can classify such transitions as: dipole - dipole (dd) and
dipole-quadrupole (dq) interactions (see 5.6.6. given above).

The multipolar selection rules are shown in the following:

5.6.50.- MULTIPOLAR SELECTION RULES

Electric Rules Intensity Magnetic Rules Intensity

dd AJ = ± 1 1.0 dd AJ=±l,0 0.1

(g -* u) [(u -* u) or (g -* g)j
4
dq 10- dq 10" 7
qq 10" 7

Note that we have normalized dd- transitions to 1.0, so as to show the

relative intensities of other types of transitions, "g" refers to the German
"gerade" = even, while "u" is ungerade = uneven states. Thus, the
transition ^ o => 3 Pi is aAJ transition of 0 -> 1 , or (g -» u), which makes
it an electric dipole transition. In general, most of the transitions that we
encounter are dd or dq in nature. Once in a while, we encounter a
magnetic dd transition, particularly in the case of certain rare earth
transitions.

5.7. - MECHANISMS OF ENERGY TRANSFER IN SOLIDS

Considerable effort has been made in the past to clarify and quantify how
energy is transferred between lattice sites within a solid. The literature is
filled with specific examples in the area of non-radiative energy transfer.
Sometimes, combinations of two, or more, separate mechanisms operate
simultaneously. We are going to classify the various mechanisms by types,
432 Optical and electronic properties of solids

according to a general mechanism. You will note that we have already

presented a listing of resonant energy transfer mechanisms among
excited centers or states in 5.6.1. which include the general types of
interactions and exchanges that occur in phosphors. We now will describe
these in more detail as follows.

I. Radiative Transfer fRadiation Trapping)

Consider two (2) atoms in a solid, separated by a few unit cell distances
and capable of emitting a photon. Let us suppose that one atom, Ni,
absorbs enough energy to become excited. It will then emit a photon. This
photon may then be captured by the other atom, N2 , which then becomes
excited itself. The following diagram shows two possible cases of radiation
transfer by trapping.

R 7 1 .

Note that we have used the term "Virtual". Virtual is used here in the
context that the photon is ephemeral and only exists momentarily as the
exchange is made. On the left side of this diagram, the energy levels are
equal while those excited states on the right are not. The excess energy
in this exchange is supplied by phonon absorption from the lattice. This
process, radiation trapping or transfer, is common in a luminescent solid.
It might be thought, therefore, that the quantum efficiency in phosphors
could exceed 100%. However, it rarely, if ever, does. That is, if one has an
 5.7 Mechanisms of energy transfer in solids 433

assembly of radiators, the excitation energy from a single emitted photon

may be passed from site to site until it reaches a non-radiative energy sink
where the energy is dissipated as phonon energy to the lattice.

It is for this reason that quantum efficiency does not usually exceed 100%.
There are, however, some phosphors which absorb a high energy photon
and then emits two photons. However, the efficiency does not exceed
100%. Only 1 or 2 phosphors have been said to have an efficiency of
100% but in view of the above discussion, there is some doubt as to the
accuracy of the measurements.

II. Energy Transfer by Resonance Exchange

By resonance transfer, we mean two radiators which are radiating at the

same frequency. This is quite different from the example given above for
radiative exchange. Note that we are not speaking of lattice vibrational
frequencies but of radiation frequencies. We thus speak of two "coupled"
oscillators. To further illustrate this concept, suppose we take two
dipoles. there will be an electric moment, £(8) , which is a vector
product whose strength is a function of the angle, 6 , between the vector
moments of the dipoles:

5.7.2.- £(8) = (w 4 r 2 / 4JT Ro c 2 ) sin 2 8

where Ro is a critical distance in the lattice. We can show this function

to be proportional to the degree of overlap between energy states of the
two different dipoles. To show how this arises, we define the following as
given in 5.7.3. on the next page. Then, from equations developed before:

5.7.4.- Emission = h c3 / M a>3 and: U = m x 2 / Ro 3

where Emission is the characteristic lifetime of emission. If Site 1.

becomes excited, and if its Emission is sufficiently long, there can be a
transfer of excitation energy to Site 2., with a characteristic transfer
lifetime, transfer-
434 Optical and electronic properties of solids

5.7.3.- DEFINITIONS FOR THE CASE OF RESONANCE EXCHANGE

Mx = dipole moment
Ro = critical distance
U S interaction energy
E S total energy
Qy = overlap energy
P = probability of energy transfer
Uw = specific energy transferred
X = 27T c /to (co is a frequency)

This lifetime is related to the other factors as:

5.7.5.- transfer - E/ U w - h /U - h R o 3 / m x 2

This gives us: P = U/ h Q • [Ql/ &2 ]

We can immediately see that the probability of energy transfer is a direct

function of the overlap of the energy states between the two sites
involved. To further illustrate this phenomenon, consider two optically
active sites in a lattice: [ 1* > and [ 2 >, where the star (*) indicates the
excited site. We have already shown that the optical line (zero phonon
transition between spectroscopic ground and excited states) will
homogeneously broaden via vibronic coupling to produce an assemblage of
excited oscillators within a specific band of energy as defined by:

5.7.6.- Eexcitation = ± h 2 n coi

where 2 n coj is a summation of all phonon modes of the lattice. In our

case, we have the situation given in 5.7.7. presented on the next page.

This process is a resonance process between matched oscillators and is

non-radiative in nature. One important consequence is sometimes called
"sensitized luminescence". Note that 11> has a higher energy than | 2>
and that there is an overlap area of energies caused by vibronic coupling
broadening of the zero-phonon transition.
 5.7 Mechanisms of energy transfer in solids 435

5.7.7.-

It is this overlap of energies which allows transfer of energy from one

oscillator to another, even though they are separated by several lattice
plane sites. The overall process may be described by:

5.7.8.- TtransferMQi.2) 2 h2
' U2 &1 - (Ql,2) 2 h 2 Ro6 / ®1 m 4

where Qi, 2 is the overlap between Ensemble 1 states, i.e.- 11* > , ready
to emit, and Ensemble 2 states { 12 >} ready to absorb. There is a critical
radius between any excited state site in 11* >, and the receiving site in
I 2 > which is described as:

5.7.9.- Ro - c/co (Qi/(^ m i SS i O n Q i , 2 ) 1 / 2 ~ l/2ir (Qi /(^mission [Gl.2] 2 ) 1 / 2

Actually, 5.7.7. only applies to dd-coupling processes which are resonant.

We must also classify resonant energy transfer processes as defined in
5.7.10., presented on the next page.

We have already illustrated Direct Exchange Coupling. The dd-dq

exchange is very similar, except that the intensity is lower because of the
oscillator strengths. Virtual phonon exchange involves phonon wave
coupling at the two sites where a part of the energy is transmitted
436 Optical and electronic properties of solids

5.7.10.- Coupling Processes in Phosphors

Electric Multipole Coupling, including dd-dq, i.e.- Sb 3+ : Mn 2+

Magnetic Dipole Coupling

Direct Exchange Coupling (between like Poles)

Virtual Phonon Exchange Coupling

through lattice phonon waves. We must further sub-divide each type given
in 5.7.10. into resonant (matched) and non-resonant (unmatched) energy
coupling. The latter sub-division requires additional energy which is
supplied by the localized phonon modes present at the receiving site. The
overall process is therefore a direct (resonant) overlap of multipole fields
or an indirect process where part of the excitation energy is supplied by
phonon modes at the receiving site.

III. Energy Transfer bv a Spatial Process

In this process, the energy exchange occurs over a rather long distance in
the lattice, and is executed as an excited energy state carried by a phonon
wave to the receiving site. This energy mechanism was first developed by
Dexter (1962). If we define 'W as a rate of energy transfer, it can be
shown that:

5.7.11.- <W= 2TT/ h 2 {< 1,2* | H Q | | l ' , 2 > } 2

where H Q is a Hamiltonian operator describing the overlap. Note that we

did not go into detail concerning how the formula was derived. Dexter
defined the excited ion (which we have called 11*> ) as a sensitizer (S)
and the absorber (A) ion (herein called 12 > ) as an activator. Thus, the
probability of energy transfer would be, in this terminology, PSA • Using
appropriate values substituted into 5.7.11., we can obtain equations of
probability of energy transfer, in terms of the type of coupling involved.
This is shown in the following:
 5.7 Mechanisms of energy transfer in solids 437

5.7.12.-

PSA (dd)=[9T,4c4QA8/1287rh T s TA R6SA] [(e/[(k)l/2 e v ] 4 |J(/s (EJdE/E*)]

PSA(dq)=[1357tr19e8/4h6<&<rAR8SA] [gA*/gA] [(e/[(k)l/2 e]4 j

[ ( J ( / S (E) FA (E) dE/E4)/ E8]

These equations look formidable but can be solved rather easily since the
variables in the two equations can be obtained experimentally. They are
defined as follows:

TI = refractive index of the host

k = dielectric constant of the host
e/ey = electric field strength (host compared to vacuum)
9 A = JOA (E) dE (OA is the optical cross-section of the activator)
FA (E) = a shape factor for the activator band,
where A is: A (E) = QA FA (E) and: J FA(E) dE = 1.0.
q = orientation between poles
E = total energy
c = speed of light
g = magnetic coupling vector of ground state

g* = magnetic coupling vector for excited state

By using appropriate values, we can reduce the equations in 5.7.12. to:

5.7.13.- P S A (dd) = ( 2 7 / R S A ) 6 • 1/ T s

PSA(dq) = (14/R A )8 . i/Ts

The important part for us is the fact that we can deduce the critical radius
beyond which energy transfer does not occur. In a normal lattice, this
distance is approximately:

5.7.14.- R S A (dd) = 25.0 A and: R S A (dq) = 13.4 A

438 Optical and electronic properties of solids

As an example, we can use a simple cubic lattice such as NaCl, where ao =

2.814 A. By defining a reduced concentration in terms of the total sites
occupied, compared to total sites available, i.e.-

5.7.15.- Xs =NS/NT and: XA = N A / N T

we can obtain an estimate of the energy transfer efficiency of the two

processes, but only if the average transfer lifetimes are approximately
equal, i.e.-

5.7.16.- TsA(dd) »TSA(dq)

Using this assumption gives us:

5.7.17.- Effective Receptor Sites for Energy Transfer from an Excited

Sensitizer Site

•TsAfdd) =0.5 Xs = 1 X A = 2900 SITES

T S A (dq) = 0 . 5 Xs = 1 XA = 100 SITES

There are 4 sites per unit cell in NaCl. For a given sensitizer site, there
are about 24 nearest neighbor sites. Thus, a sensitizer can affect an
activator site as far away as about 120 lattice planes. This is indeed along
distance for resonant energy transfer to occur. It is also evident that the
number of activator sites to which energy transfer may take place from a
single sensitizer site will vary according to the host, i.e.- the phonon
spectrum of the host, the lifetime of the excited states of S and A, and the
lifetime of the energy transfer process.

We can also express the same energy transfer mechanisms in terms of

spherical harmonics and the oscillator strengths. The actual formulas and
derivation have been given at the end of Section 5.6. Using spherical
harmonics, we obtain quantities like those of 5.7.12. and 5.7.13.:
 5.7 Mechanisms of energy transfer in solids 439

5.7.17.- CRITICAL RADIUS FOR ENERGY TRANSFER FROM SPHERICAL

HARMONIC CALCULATIONS
Process Scritical Relative Intensity
{electric or dd (1 / R S A ) 6 1-00
magnetic)
electric dq (1 / R S A ) 8 10" 4

electric qq (1 / R S A ) 1 0 10" 7

These results are the same as we derived before. Thus, we have covered
all of the multipole resonant energy processes possible.

IV. Energy Exchange bv Spin Coupling

There is another process which is important in some cases. If two ions

are on adjacent sites (no further than about 2-3 lattice sites apart), a
direct energy exchange can occur through the coupling of electron spins.
Thus, we have an exchange Hamiltonian between electron spins on two
different sires, as follows :-

5.7.19.- Hexchange = Jeff. Si • §2

where Jeff is an efficiency of coupling and Si are the individual spin

vectors. At 90° , J = 1.00 (bilinear). In a bilinear situation such as dd or
dq coupling, L-S coupling prevails., as follows:-

5.7.20.- J(LS) = Jo exp ( - 2d / In Si,2 )

where: d = distance between sites, <Si,2 is the overlap of the spin

integral, and J s 1 / d 2 . But for a qq-process, we have:

5.7.21.- Hspin =*Hso =*• final state

Here, the spin exchange goes from pure spin to a spin-orbit process to a
final state. We can further illustrate this type of spin exchange process by
using a band model.
440 Optical and electronic properties of solids

Consider two ions, both initially in an excited state, as shown in the

following:

5722-

In this case one ion at Site-1 spin-couples to another ion at Site-2. In the
above diagram, i refers to the initial state of each and f is the final state.
Thus, the "2-ion" goes from the "B" level to the "A" level, while the "1-ion"
simultaneously transforms from the "B" level to the "C" level . Thus, from
two excited ions, we end up with one in the ground state and the other in
an energy state double that of the initial state. This process has been
observed in rare earth activated phosphors which can absorb infrared
radiation and convert it to visible light (an Anti-Stokes process).

It occurs only between nearest neighbors and is not a spatial process. In

contrast, multi-polar resonance processes do involve transfer over
relatively large distances through the lattice. We can compare qq-coupling
with a spin-exchange process by using ruby, i.e.- AI2O3 : Cr 3+ , as an
example. This is illustrated in 5.7.23., given on the next page. For ruby.
5.7.19. applies, and for S = 2.0, Huang (1954) has shown the following to
apply:

5.7.24.- < 2 Eg ,4 A2 | Hexchange. I 4 A 2 , 2 Eg > = 3/2 Jeff.

 5.7 Mechanisms of energy transfer in solids 441

5.7.23.-

For a (qq) transition process, we use the method and formulas of

spherical harmonics equations given above. For quadrupolar coupling,
spin-orbit coupling is necessary to break the spin selection rule (see
5.7.21. above). Thus, the process goes by:

2
5.7.25.- E g ( H q q ) =*> 2x 2 (H S d=> 4 A2

In this case, Vj becomes equivalent to H Q , the spatial overlap operator.

The transfer matrix element is reduced and becomes:

5.7.26.- | i , 2 / E(2T2) - E(4A2)

where E(2Tz) and E(4A2) are the energies of the states and §1,2 is a
coupling coefficient. The total coupling element then becomes:

5.7.27.- [< 2 Eg s 4 A 2 |Hqq|4A2 , 2 Eg >]2 qq • £ x> 2 / E(2T2) - E(4A2)]2

The upper limit for H q q for a 13 A separation of Cr 3+ in AI2O3 is about 8

xlCT3 cm"1- hi ruby, a good estimate of l i , 2 is about 170 cm- 1 while
442 Optical and electronic properties of solids

E(2Ta) - E(4A2) = 21,000 c m 1 . Hence, the (qq) coupling matrix is no

larger than about: 6 x 1 0 7 cm-1- For spin- exchange, the energy transfer
will take place for Cr3+(i) to Cr3+(2) through one oxygen link.

Let us now compare the spin exchange process with that of the
quadrupolar coupling process. Although exchange will occur through one
oxygen link, the Cr 3+ cations are actually four nearest neighbors apart, so
that Jeff. = 10 c m 1 . At 13 A, there are three oxygen links, so that
Hexchange. (This comes from 5.7.19. and 5.7.24.) can be estimated as: Jeff.
=10 c m 1 (Si) 2 • (S2)2) = 10-3 c m 1 . This is three orders of magnitude
larger than the qq-process. We can therefore conclude that the spin-
exchange process is strong in many lattices and may even be equivalent to
dipole-quadrupole processes found in many phosphors. Nevertheless, it Is
a short-range process and occurs between nearest neighbors only. For
very dilute solid solutions of Cr 3+ in AI2O3 it has been determined that the
spin-exchange process does not occur at all.

V. Energy Transfer bv Non-Resonant Processes

In order to be complete, we shall now describe non-resonant processes

where there is a large mismatch in energy of the oscillators. The first
analogue for energy transfer was formulated by Orbach (1967). This
corresponds to the situation shown as follows:
5.7.28.-
 5.7 Mechanisms of energy transfer in solids 443

Even though there is no overlap of vibronically-coupled energies, energy

transfer can occur as (dd), (dq) or Hexchange. through coupling of each
state to the lattice viaphonon exchange. This involves emission of a virtual
phonon at Site 1 (which is in an excited state) to the lattice which is
immediately absorbed at Site 2 , with simultaneous spatial transfer of
resonant energy by an allowed multipole process.

In other words, two mismatched multipoles can readjust their energy

states (within certain limits as determined by h 2 n <a\ the energy of the
phonon branches of the lattice) so that a resonant energy transfer process
can occur. The mathematics are complicated so that we will describe the
process simply as a modification of 5.7.11., wherein h I n q been
included:

5.7.30- W = 2TT/ h2 {< n (It*. 3*) + (A , 1,2* )| Heff. | n (t, 3*)(1*,2) > }2

x { fi (X) f2 (X) dX/ gi (on ) g 2 [on - Aco {t)] d aq

where A = ± 1 , corresponding to the creation and/or destruction of a

phonon and the other quantities have already been defined.

In addition, n (which is a function of allowed directions in the lattice),

and I* and 3 (which are the phonon coupling vectors) describe the
number of phonons involved in the process. Note that what we have done
is to include phonon processes as part of the energy transfer process
integral. The actual evaluation of Heff. is very complicated and will not be
further elucidated.

Suffice it to say that both Heff. (actually Hdd , Hdq as the case maybe) and
Heiec- vibr. (a Hamiltonian describing vibronic coupling) must be included.
Nevertheless, this type of energy transfer involving phonon creation and
destruction at two sites where the oscillators are not originally coupled
has been observed experimentally and occurs in many cases where energy
transfer between excited and unexcited centers occurs.

We can now summarize these processes as they relate to phosphors.

444 Optical and electronic properties of solids

5.8.- SUMMARY OF PHONON PROCESSES AS RELATED TO PHOSPHORS

As a summary to this Chapter, let us summarize the phonon coupling

processes which can occur during luminescence :

I. Excitation of the Activator Center

Absorption of a photon causes activator center to become excited, i.e.- the

ground state is converted to the excited state by absorption of energy.

a. Phonon process: Homogeneous broadening of zero-phonon line

(a transition between Stark States)

b. Energy flow: Lattice to center, with phonon energy causing

secondary Stark states to become excited so
that a population of energy states exists, each
varying by the amount of energy present in
the phonon that was absorbed.

II. Excited State Relaxation

The excited center relaxes to an equilibrium excited energy state:
a. Phonon process: Description of energy results in a change
of effective radius of the excited center.
b. Energy flow: Center to lattice via phonon energy
exchange

III. Center Ready to Emit: Excitation energy can be dissipated in

one of three (3) ways.
a. Direct relaxation
i. Phonon process: Direct conversion of excitation
energy to phonon energy
ii. Energy flow: Excited center to lattice
b. Emission of a photon
i. Phonon process: Relaxation of individual Stark ground
state to equilibrium ground state,
ii. Energy flow: Center to lattice
 5.8 Summary of phonon processes as related to phosphors 445

c. Resonant and Non-Resonant Processes : This includes :

i. Virtual photon exchange
ii. Resonant energy exchange by overlap exchange
between two coupled oscillators,
iii. Electric multipole coupling,
iv. Direct exchange coupling,
v. Coupled multipole radiators,
vl. Spin coupling,
vii. Coupled multipole oscillators involving creation and
destruction of phonons at the coupled sites, allowing the
transition to take place.

d. Actual Phonon Process: all of the above cases involve phonon

interactions at both sites, including creation and destruction of
phonons (the non- resonant case)
i. Energy Flow: #1 - none;
#11 - lattice to centers;
#111 - lattice to centers where energy
is carried by phonon waves;
#IV - lattice to centers;
#V- lattice to centers;
#VI - none;
#VII- self evident.

What we can conclude is that the luminescence process involving a

phosphor is indeed complicated and is a dynamic process in which the
excited center undergoes many perturbations by the lattice during its
excited lifetime.
446 Optical and electronic properties of solids

Suggested Reading

1. F. Williams, Ed.- "Proc. International Conf. on Luminescence", N.

Holland Publ., Amsterdam (1970).

2. F.A. Kroger, "Some Aspects of the Luminescence of Solids"- Elsevier

Publ., Amsterdam, The Netherlands (1948).

3. "Phosphor Handbook", Edited by: S. Shionoya and W.M. Yen, pp. 921,
CRC Press, Boca Raton, Fla. (1999).
|1 CHAPTER 6 ||
Design of Phosphors

In the last chapter, we explored the interaction of an electron or a photon

with the lattice of a solid like a phosphor. We concluded that a photon is a
"force-carrier" while an electron is a "matter-constituent". We also
investigated how excitation energy from these two "forces" was absorbed,
i.e.- transmitted to the lattice, and then dissipated. This mechanism
involved either reemission of a photon of lower energy or by phonon
modes as heat. The modes by which energy was absorbed and then
transmitted within the lattice of the solid included resonant energy
transfer between various types of "centers" including "activators" and
"energy traps" (the former emitting photons and the latter "heat" as low
energy phonons). In this Chapter, we shall investigate the design of
phosphors, methods of preparation of phosphors, the electronic devices
in which they are used, a list of cathode-ray phosphors and a list of those
used in lamps. We shall also explore ancillary areas such as color spaces
and the specification of color. In order to elucidate the design of
phosphors, particularly new phosphor compositions, we must return to
our examination of the luminescent center and the energy processes
associated with it.

6.1- THE LUMINESCENT CENTER IN INORGANIC PHOSPHORS

We have already discussed the energy processes which occur at the

activator site during absorption, excitation, and emission transformations.
Both interaction with a particle (electron) and a photon was explored.
Briefly, they include phonon processes with vibronic coupling between
electronic states of the activator and the phonon energies of the crystal
host lattice. Following absorption and excitation at the activator site,
phonon creation and destruction at the site populates adjacent Stark
levels in the excited center which are further broadened by vibronic
coupling. The isolated activator center can be thought of as interacting
with its crystalline environment via the crystal field, which is time-varying
because the lattice is vibrating. The effective radius of the excited state

447
448 Design of phosphors

wavefunctions also changes as vibronic perturbation by the lattice

continues. These two mechanisms account for the homogeneous
broadening of the zero-phonon line, as discussed previously. Real time
phonon energy flow processes continue to affect the activator center, both
prior to, and following the emission of a photon. Since each photon
emitted has a very slightly different energy, we end up with a band of
emission energies. The emitted-photon energy differences are so small
that no known instrumentation has sufficient resolution to differentiate
between them.

Although we have previously shown in Chapter 5 that the excitation

transition for Sn 2+ is: iSo => iPi, the actual mechanism is, in reality,
more complicated than originally presented. The precise process has
been found to be:

J
6.1.1.- hvi + iSo => Pi-(nan ) =>3Pl ^ l s0 +hV2

where h mi is a relaxation involving phonon energy. The ground state for

Sn 2 + is d 1 0 s 2 (iSo ). When this activator cation absorbs photon energy, the
d 10 sp excited state is split by the crystal field into a singlet state, xPi , and
a triplet state, 3 P i , of lower energy. Each excited state has 2 J + 1 Stark
states, for a total of seven. The potential energy diagram of the excitation-
emission mechanism is shown in 6.1.2., given on the next page.

The excitation transition is shown as: iSo => 1 Pi. It is this state which
relaxes to the 3 Pi state from which emission occurs. However, if we
measure the excitation band for several different Sn 2+ activated
phosphors, we usually will obtain two closely connected bands. There are
two possibilities. One is that the excitation transition also involves: iSo =>
3
Pi , a transition which is "not allowed" since AS = 2. The other is that
they are crystal field states.

If we compare the excitation bands of various Sn 2+ - activated phosphors,

i.e. - those having various host crystals, we find that activated phosphors,
the excitation bands vary significantly, but the high energy band, i.e.- the
1
S 0 => *Pi transition, seems to predominate.
 6.1 The luminescent center in inorganic phosphors 449

6.1.2.-

Because the two states are so close in peak energy, one finds it difficult to
delineate the actual excitation transition.

Now consider a different type of activator cation such as Cr3+. Suppose we

use AI2O3: Cr 3+ as the example. We do this to contrast:

1). atomic energy level schemes with crystal field splitting

2). potential energy diagrams

with the intent of comparing the relative merits of these methods of

displaying the energy levels involved in absorption, excitation and
emission of phosphors.

The energy diagram for Cr 3+ in Al2O3is given as 6.1.3. on the next page.
450 Design of phosphors

6.1.3.-

Whereas in Chapter 5, we presented only the energy levels and states of

Cr 3+ in the AI2O3 crystal, now we illustrate all of the factors involved,
namely the effect of the crystalline field upon the actual states which
appear. The electrons of the Cr3+ ion are 3d 3 • The free ion states are:
4
F , 4 p , 2Gand2H.

Under the influence of the crystal field, these states split, as shown for a
trigonal field, into crystal field components. The crystal field strength of
AI2O3 is 20,000 c m 1 as shown, and this accounts for the exact energies
which appear for the excitation transitions observed. We have already
discussed these in light of spin-coupling and trigonal field transitions.
Cr 3+ does exhibit an emission transition which is: 2 E g => 4A2g. However,
the QE efficiency is < 5%.

The major point of this discussion is that it is necessary to specify the

 6.1 The luminescent center in inorganic phosphors 451

influence of the crystal field upon the free ion states. Thus for Sn 2+ , the
two excitation bands, discussed above, may also be attributed to crystal
field components, depending upon the nature of the host in which the
Sn 2 +finds itself.

Since the QE of ruby is so low, let us use a potential energy diagram to

determine how this might be so, as shown in the following:

6.1.4.-

In this diagram, all three (3) excitation transitions are shown as well as
that of the emission transition. What is different in 6.1.4., compared to
that of 6.1.3., is that the relaxation processes are also apparent. Thus, it is
easily seen that the excitation transition, 4 A2g => 2 Ti will result in
emission, whereas the other transitions, i.e.- 4A2g => *T2 , 2T2 & ^ i
will mostly result in relaxation processes, without emission. Note also that
the *T2 state appears to be the major path for relaxation back to the
ground state. It should be clear, then, why the QE of ruby is rather low.
452 Design of phosphors

We have illustrated two methods of displaying energy levels of optically

active ions. Both have their shortcomings and advantages. In addition, we
have examined two different types of ions that are optically active. Both
possess energy gaps between ground and excited states far enough apart
so that emission of a photon in energy is possible. However, as we will
discover later, those ions which have a plurality of energy levels available
in a crystal field, or those having numerous spin states such as Cr3+, do
not exhibit high quantum efficiencies. The reason for this, we will find, is
that phonon coupling of the excited state to the lattice phonon modes is
the preferred pathway of energy dissipation. We will also find that the
phonon-creation mode of energy dissipation is far easier that the emission
of a larger packet of energy such as a photon. Furthermore, "impurity"
cations may be present in the host lattice, and form an "energy sink"
where the excitation energy is dissipated without emission of a photon.
Such ions are called "killers" of luminescence.

6.2 - THE GROUND STATE PERTURBATION FACTOR

Our next step is define the conditions in which we would expect to obtain
a phosphor having a luminescent process of high quantum efficiency. We
then will be able to discuss how to design a phosphor, the experimental
parameters involved, and the solid state chemistry of these defect-
controlled materials.

We have already stated that both a suitable host and an activator cation are
needed to form a phosphor. The activator needs to be optically active and
capable of absorbing energy, without dissipating it before the emission
process can occur. The host must be carefully matched to the chosen
activator because of the effect of the lattice phonon-energy branches on
the localized modes of the activator. We have also shown that for cations
having unpaired spins where spin-coupling mechanisms predominate, the
QE cannot be, and is not, high. This eliminates a large number of potential
transition metal cations.

What we have not emphasized heretofore is that only certain valence

states of specific cations will be of use to us. Consider the following.
 6.2 The ground state perturbation factor 453

Our first rules for the design of high efficiency phosphors are:

CHOICE OF ACTIVATOR CATION

a. Must be optically active

b. Possess a large energy gap between emitting and
ground state, in its crystalline environment.
c. Have the proper valence state which is easily
established in its crystalline environment.
d. Stabilize that valence state within the crystalline host.

We find that we can divide the Periodic Table into two classes of cations,
optically active and optically inactive. The latter possess the closed-shell
electron configuration of the rare gases, i.e.- d 10 , as represented by the
monovalent state of the alkali metals. Moreover, all of the transition
metals have d-electrons present. Most of these are spin-coupled, have
closely spaced energy levels, and thus act as "killers" for luminescence.
However, it is possible to find valence states where the luminescence
process proceeds with high efficiency.

A good example is Mn 2+ (3d 5 = 6 S5/2 ). This state produces strong

luminescence in many hosts while Mn 7+ , Mn 6 + , Mn4* (a special case),
and Mn 3+ do not. The electron configurations for these ions are: d° , d 1 ,
d 3 , and d 4 , respectively. All are spin coupled to the phonon modes of the
host lattice. Note that the Mn 2+ ion involves the half-filled electron shell
which has a singlet state, 6 S5/2 .

We also need to stabilize the chosen activator state in its crystalline host.
It will do no good to introduce an activator into the host, if subsequent
events cause a change in its valence state, thereby rendering it incapable
of efficient luminescence. One way to do this is to introduce the wrong
firing atmosphere during the firing process to crystallize the phosphor.
The other factor involves the valence state of the host as compared to that
of the activator. If the two do not match, then the activator cation will be
incorporated into the host along with a vacancy which may act as an
"energy sink" to dissipate the excitation energy as phonons. We have
454 Design of phosphors

already given the example of CaS:Bi3+:K+ where it was shown that K+ was
required for efficient luminescence. Many workers have not realized the
fact that K+ is optically inert (K+ has the argon closed shell electronic
configuration) and is present only for charge compensation in this
phosphor.

The last factor to be considered is the so-called "ground state

perturbation". We have already shown that the activator site is under the
influence of vibronic coupling involving localized phonon modes. This
includes both phonon creation and destruction mechanisms and the time-
varying electric field of the lattice, which produces vibrational
perturbation at the activator site. Because of these influences, we must
minimize the vibronic coupling effects at the site by proper choice of the
activator. Otherwise, we will not obtain an efficient luminescence process.
To do this, we must choose a cation whose electronic ground state has no
uncoupled spins, no angular momentae, and in which the sum of all
electron moments, J, is zero. What this means is that the ground state
must be a spherical set of eigenstate energies which are symmetrical to
the crystal field, namely-

6.2.1.- Required Electronic Parameters for Minimizing the Ground State

Perturbation Factor

S = 0}
L =0} = the ! So state
J = 0}

All cations having this electronic ground state configuration will function
as efficient activators, whereas the others will not. Let us examine why
this is so.

Basically, an ion with a 1 So state has 2J + 1 Stark states or a single Stark

state in the ground state configuration. Therefore, the phonon
perturbation operator cannot populate adjacent levels via vibronic
coupling of the localized vibrational modes at the activator site, because
none exist. Additionally, with the absence of a total momentum operator
 6.2 The ground state perturbation factor 455

(J = 0), the ground state electron wavefunctions are spherically

symmetric, making vibronic coupling practically non-existent at the site.
In this way, we can obtain an activator center, substituted within a host
crystal, which is not coupled to the phonon modes of the lattice in its
ground state. However, once it becomes excited, it is subject to the full
perturbation of vibronic coupling. Therefore, we have achieved our goal of
eliminating the ground state perturbation factor. All other cations have
electronic states with either spin coupling factors or angular moments
which can couple via vibronic processes to the phonon modes of the host
lattice. This results in excessive loss of excitation energy before the
emission process can take place.

Let us compare the cation, Mn 2+ , to a !So cation. The ground state, which
is 6 S5/2 , is a singlet with six (6) Stark states. The lowest energy
excitation state is 4 G5/2 so that the emission transition is dipole allowed.
However, the plurality of levels in the ground state does not allow this
cation to function effectively as an activator. The absorbance of Mn 2+ is
very low, and if one exposes it to photons of the proper energy so as to
cause excitation, one finds that the emission intensity is very low because
of the lack of efficient excitation processes. (There are in fact five (5)
weak absorption lines which appear in the Mn 2+ absorption spectra). Yet,
if excitation energy is transferred to Mn 2+ sites via some energy exchange
process, they exhibit very high quantum efficiencies in emission because
the transition is dipole allowed.

It is well to note that some of the best phosphors (which have QE's
exceeding 85%) are based on combinations of two activators, e.g.- Sn 2+ -
Mn 2+ or Sb 3+ - Mn 2+ . In this case, either Sn 2+ or Sb 3+ acts as a sensitizer
and Mn 2+ is the activator, i.e.- the SA pairs mentioned in Chapter 5.
Strong absorption and excitation occurs at the Sn 2+ or Sb 3+ site while,
with the proper composition, the emission can occur wholly from the
Mn 2+ site. The Mn 2+ activator does not show a zero phonon line which
can be perturbed in the excited state by vibronic coupling processes
because it is excited by the sensitizer, not through the lattice. Thus, all
those cations with a half-filled electron shell, including Mn 2+ , are able to
456 Design of phosphors

function efficiently as activators (but only in the presence of a sensitizer

cation).

At this point, we repeat the fact that the most important factor in the
design of a phosphor is the choice of an activator cation with an electronic
ground state wherein the perturbation by localized phonon modes of the
host lattice is non- existent. This mandates the choice of activators with
the iSo ground state. There are no other choices. The other factor is
matching the valence state of the activator to that of the host to prevent
lattice defects such as vacancies from forming.

6 - 3 : DESIGN OF A PHOSPHOR

Our next subject involves the factors required to design a phosphor. We

must consider: 1) choice of the host components, 2) choice of the
activator , and 3) "killers" of luminescence.

I. Choice of the Host Components

In designing a phosphor, we find that the inorganic host consists of one of

several different cations combined with one of several different anions.
The anion may be simple or it may consist of a group of atoms, e.g.-
borates, silicates or phosphates. It should be apparent that both the cation
and anion selection must be optically transparent, since we want the
absorption- excitation process to take place at the activator site. The
exception is those phosphors in which the absorption of energy, and/or
photons, occurs in the anion- group. However, these phosphors, called
"self-activated", cannot function if an activator cation is also present.
Because of the above restrictions, we find that activators can be divided
into two (2) classes, those involving a single optically active cation and
those involving an anionic group which functions as an activator.

The host cation is therefore limited to those ions which have the rare gas
electron configuration, or closed electron shells, and therefore are
optically inactive. In the following diagram, we show those cations which
can function to form a phosphor having the desired quantum efficiency:
 6.3 Design of a phosphor 457

6.3.1-
1The Periodic Fable as Related to IPhosphor Composition
• •

H Cations That Can Be C3+) C4+) He

(2+)
Li +
Mg
Used to Form He
Phosphors
Na + Mg (3+) (4+) (2+) Ar

K + Ca Sc Ti Zn Ga Ge Kr

Rb + Sr V Zr Cd In Sn Xe

Cs+ Ba La Hf Hg Tl Pb Rn
+
Fr Ra Ac 104

La 3 ¥
Gd3] Lu 3 ]
I
Ac 3 Cm3t Lw 3 f
i

In this diagram, the required valence state of the cations are shown in the
parentheses at the top. The suitable cations are those in the following
subgroups of the periodic table:

1A, IIA, IIIB, IVB IIB, IIIA, IVA

including the lanthanides of IIIB (which are a special case). Thus, we can
choose any of these cations, but they must be stabilized in the host crystal
in the required valence. For example, we can choose Zr 4 * to form a
crystal. We will need an anion having a valence state of (4-) . Suitable host
compositions would be: ZrSiO4 , ZrGeC>4, etc. All of the above cations have
the closed shell configuration, s 2 p 6 , of the rare gases. Since the alkali
metal ions, Na+ etc., are mostly water soluble, we do not find them very
useful to form phosphors such as Na4SiO4 or Na4GeO4. However, if we
should wish to use such a phosphor in a water-free environment, then
phosphors based on the alkali metal cations would exhibit the required
high energy efficiencies of luminescence. Note also that certain of the
lanthanides can be used to form host crystals, e.g.- Lu3+ = 4f14 5s 2 5p 6
because of the outer closed electron shells. Although certain cations such
458 Design of phosphors

as V5+, Nb 5+ , Cr6*, or Mo 6+ possess the required electronic configuration,

they are not included in the diagram because these cations do not exist as
separate entities, but are found in certain anions such as NbO43-, etc. It is
somewhat capricious that a phosphor could be formed as:

Sn 4+ SiO4 : Sn 2+ = (1 -x) SnSiO4 • x Sn2 SiC>4

b u t t h e actual preparation of this material would b e extremely difficult. (It

might b e accomplished by initially preparing t h e crystals separately a n d
t h e n combining t h e m in a n inert a t m o s p h e r e ) .

In t h e choice of anions for design of phosphors, we have already

established that anions c a n b e divided into two classes, those t h a t a r e
optically inert a n d those t h a t are optically active. T h e optically i n e r t
anions useful in designing phosphors are shown in t h e following diagram:

6.3.2.-
|The Periodic Table as Related to Phosphor Composition

H Anions that Can <3 - ) C4-) (3-) C2-) C1-) He

be Used to Form
Phosphors B03 F Ne
AIO3 SiO< P 0 4 SO 4 Cl Ar
3aO3 GeO. AsO- SeO^ Br Kr
lnO 3 SnO4SbO- T.0, I Xe
La =b0jBi04 Rn

Ac 104

La
Ac

Each represents a class of anions. That is, the orthosilicate is: SiO44".
But this anion is one of many silicate anions which include: Si2O76~, SiO3",
 6.3 Design of a phosphor 459

SLiOn5, etc. You will note that we have specified oxygen-dominated

anions, involving the subgroups:

IIIA, IVA, VA, VIA & VIIA

It is the valence state of the central ion which is important. The actual
formula for these anions, that is - the ratio of oxygen atoms to the central
ion, will encompass many anionic forms and structures of the final
crystalline host. Nevertheless, the valence state of the central ion remains
constant.

There is one other class of anions which are important. These are the so-
called "self-activated" anions. In this case, if we combine certain of the
optically-inert cations of 6.3.1. with those of the following diagram, we
obtain a phosphor which does not require an activator cation:

6.3.3.-
|~The Periodic Table as Related to Phosphor Composition!

Anions That Are Optically

H He
Active - "Self-Activation"
Ne

(4-) C3-) C2-) (1-) Ar

Ti04'v04|cr04|Mn0i Kr

NbO4MoO. Xe

La HfoJ Ta0 W0
4 4 ReO4 Rn

JAc 104

In the above diagram, the required valence of the anions is shown at the
top of each column. Note that these are all oxygen-dominated anions. The
subgroups involved are:
IVB, VB & VIB

In 6.3.3., each of these optically active anions can be combined with the
460 Design of phosphors

optically-inert cations of 6.3.1. to form the host crystal. The anions of

6.3.3. are optically active because of an allowed transition within the
group orbitals, called a "charge-transfer" transition. It is akin to molecular
orbitals of organic phosphors where the excited state involves a shift in
electron-density toward the central cation. Thus, CaWO4 is a phosphor as
is YVO4. We can clarify this point by examining the molecular
energy levels of the VO43- group as shown in the following diagram:

6.3.4.-

The VO43" anion group is tetrahedral and the V 5+ cation is surrounded by

four (4) oxygen atoms which are bound by covalent bonds to the central
metal ion. The electronic configuration of the vanadate group is made up
of metal orbitals, shown at the left, and ligand orbitals of the oxygens,
shown at the right. The excitation transition of the complex grouping
occurs from 3 t2 , an orbital derived primarily from p- electron orbitals of
oxygen to 3 aif an orbital derived mostly from 3d electron orbitals of the
metal, V 5+ . Thus, the excited state consists of an increased electron
 6.3 Design of a phosphor 461

density in the vicinity of the metal ion, along the tetrahedral bonds. For
this reason, the mechanism is called a "charge-transfer" transition. This is
an allowed transition and the oscillator strength is extremely high, i.e.- in
the high thousands, as contrasted to the low hundreds of "ordinary"
activators.

Following relaxation (the Stoke's shift), emission occurs, from the

transition: H2 =* 3 t2 , a dipole-allowed transition. Because these
transitions involve molecular orbitals and because vibronic coupling is also
present, broad bands of high efficiency are observed both in absorption
and emission. This completes the list of possible cations and anions which
can be chosen to form a phosphor having high quantum efficiency when
combined with a suitable activator.

However, there is also one other class of host which we have not yet
addressed. This class comprises those anions which are not oxygen-
dominated. Thus, we find that phosphors can be formed from:

I-VII compounds, II-VI compounds, & III-IV compounds

The following table lists some of these types of compounds, a phosphor

example and physical properties.

Table 6-1
Type Example & Emission Physical Properties
+
I A-VII A KCl: Tl = near UV Insulator
2+
II A- VI A CaS: Mn == yellow Insulator
2+
II B - VI A ZnS: Mn = orange Semi-conductor
III A- V A GaP, GaAs, InP Semi-conductor

Note that when we combine subgroup "A" cations & anions, we obtain a
compound which is an insulator (as illustrated by KCl and CaS). But if we
change the cation to one from the IIB group, we get a semi-conductor,
ZnS. This illustrates the requirement of carefully choosing the proper
cation for our phosphor, depending upon the application for which we
462 Design of phosphors

envisage its usage. It is well to note that the IIIA-VA class of phosphors
are not used as crystallite powders like those based on ZnS and the like.
They are utilized for many optoelectronic devices such as light-emitting
diodes, semi-conducting lasers and photo-diodes in the form of crystalline
thin-films. We will address such devices based on these compositions,
i.e.- GaP, in more detail in the next chapter.

II. Choice of the Activatorfs)

Now we come, in our design of phosphors, to the choice of suitable

activators. Because of the limitations imposed by the ground state
perturbation factor, we are limited to certain cations in specific valence
states. As we have stated before, this limits us to the electron
configuration: d 1 0 s 2 t 1 So). The following diagram shows these choices:

6.3.5^
[The Periodic Table as Related to Phosphor Composition]
Cations That Can Be
Used as Activator
H Centers (i+J (2+)(3+) He

Me
f Half-filled Shells
(1+X2+) (3+) C1-) (0) Ar
Cr lin Fe Cu Zn Ga Ge As Kr

Tc Ag Cd In Sn Sb Xe

La Re Au Hg Tl Pb Bi Rn

Ac 104

Ce Pr Nd Sm Eu Gd Tb oy Ho Er Tm Vb
(3+)
Th U

Also shown are the cations having a half-filled shell, and their valence
state. We have already pointed out that this type of activator does not show
strong absorption, but does show strong and efficient emission once it
 6.3 Design of a phosphor 463

becomes excited. It is for this reason that some authors have called them
"activators" and the others shown at the right-hand side of 6.3.5.
"sensitizers". In any case, these "half-filled shell" activators cannot be used
singly, but must always be employed in combination with one of the
sensitizer cations. Even though we have listed them, there is considerable
doubt that the ions, Tc 2 + and Re 2+ can actually be stabilized in a selected
host lattice. The same may be said for Cr+. In spite of this, if any of these
cations could be stabilized in a host lattice in the valence state shown
above, they would function as efficient activators. There are other ions
having this same electronic configuration, but have not been presented
because of the perceived difficulty of stabilization of the required valence
state, regardless of the host crystal. The cations, Mn 2+ and Fe 3+ , are well
known states, produce efficient luminescence, and have high QE's when
incorporated into the proper phosphor host. Note that Fe^ is not used
generally as an activator because it is easily reduced to the ferrous state,
which is an efficient "killer" of luminescence.

One important point needs to be emphasized. All of the indicated valence

states in 6.3.5. are stable states of that element (except as noted). One
could probably use Se4"1" as an activator, if one could obtain that state in a
crystalline environment. Unfortunately, one cannot. The selenates, as
anions, are the stable forms of the selenium ion, and the uncombined
form has yet to be observed in a crystal. This brings us to a critical
proposition, namely:

IN CHOOSING AN ACTIVATOR, ONE MUST BE ABLE TO FORM AND

STABILIZE THE ACTIVATOR IN THE PROPER VALENCE STATE WITHIN
THE HOST CRYSTAL IN THE REQUIRED ELECTRONIC CONFIGURATION
IF EFFICIENT LUMINESCENCE IS TO RESULT.

Let us further illustrate rules for designing phosphors with yet another
example. Suppose we choose Mg 2+ as the cation, SiO44' as the anion and
Sn 2+ as the activator. The crystal, Mg2 SiO4 is well known and could serve
as the host composition. In addition, Sn2SiO4 is also a known form. Thus,
if we attempt to combine them, either in the form of oxides, e.g.- MgO +
SnO + SiO2 , or directly as the compounds themselves, a question remains
464 Design of phosphors

as to the success of the combination. We have already said that in order

for such combinations of compounds to occur, a solid solution must result.

According to Vegard's Law, in order to form a solid solution, the cations

and/or anions of two compounds must have ionic radii within ± 15% of
each other. The rule is not so rigid if the solid solution content is kept
within a limit of no more than 5% of one in the other. That is: In the
composition, (1- x) MSiO4 • x NSiC>4 , if x is kept below 5% , then a solid
state solution will form. Since most activator concentrations are usually no
more than 5%, solid solutions can be formed for many of the possible
combinations already shown in 6.3.1., 6.3.2. & 6.3.5.

It is for this reason that the phosphor, Mg2SiO4: Sno.02 can be obtained,
even though the ionic radius of Mg 2+ = 0.65 A and that of Sn 2+ = 1.12 A .
However, we find that the efficiency of this phosphor is rather low and
the QE is low as well. The rationale for this is that a cationic radius
mismatch has resulted in lattice strain at the activator site and increased
loss of excitation energy due to increased vibronic coupling at the site.
Conversely, if we choose Sr 2+ whose cationic radius is 1.13 A, the so-
produced phosphor, Sr2SiO4: Sno.02 , exhibits high energy efficiency and
has a high QE as well.

Therefore, it is apparent that there are other rules for phosphor design
that need to be elucidated:

a. The activator cation and the host cation need to be matched

in size so as to obtain maximal efficiency in the produced
phosphor. Mismatch creates strain in the lattice, and limits
the actual solubility of the activator in the host lattice.

b. The activator cation needs to be of the same valence state as

that of the host cation. Otherwise, the activator will substitute
into the lattice with the formation of cation vacancies (for
charge-compensation). An alternative is the introduction of a
compensating cation for the same purpose.
 6.3 Design of a phosphor 465

One other point involves the defect solid state reaction behavior of certain
selected hosts. One might wonder how one would stabilize an activator
cation in the metallic state, e.g.- Zn° . If one calcines, i.e.- "fires" at an
elevated temperature, the compound, ZnO, in a reducing atmosphere, the
following defect reactions take place:

6.3.6.- DEFECT REACTIONS OF ZnO IN A REDUCING ATMOSPHERE

1. ZnO + x H 2 = ZnfOx-x.xVo) + x H2O + (2 e ) x + x Zn2+j

2. x Zn2+i + (2 e-)x = x Zn°
3. x Zn° + x Vo= = x Zn°o=
4. ZnOi. x + x Zn°o= = ZnO: x Zn°o=

Note that the defect formation behavior is one where the hydrogen gas
reduces the host lattice and forms an oxygen vacancy plus two free
electrons. These electrons then reduce the interstitial Zn 2+ cation (which
became an interstitial cation in order to satisfy the charge compensation
mechanism when the oxygen anion was removed by the reducing
atmosphere to form water) to the metal valence state. Zn° • This atom is
too large to remain as an interstitial, and therefore occupies the oxygen
vacancy already created. Note that the radius of Zn° is 1.48 A while that of
the oxygen vacancy is 1.50 A. This is one of the few cases where the
cation in its metallic form functions as an activator. The semi-conducting
efficiency of this phosphor is high, as is its QE.

The hosts that we have discussed to date have been ionic in nature, and
dielectric as well (they are non-conductive). There is also another class of
phosphor hosts which are covalent and semi-conductive in nature, namely
the zinc and cadmium sulfides and/or selenides. The criterion for
selection of a semi-conducting host for use as a phosphor includes choice
of a composition with an energy band gap of at least 3.00 ev. This
mandates the use of an optically inactive cation, combined with sulfide,
selenide and possibly telluride. The oxygen-dominated groupings such as
phosphate, or silicate or arsenate, etc. are not semi-conductive in nature.
And, none of the other transition metal sulfides have band gaps sufficiently
466 Design of phosphors

large to be used as phosphors. We have already shown energy band

diagrams in Chapter 5.

We find that the ZnS and ZnSe host structures are excellent semi-
conductors and exhibit very high efficiencies of photoluminescence and
cathodoluminescence, especially when activated by copper, silver and
gold. There is a large body of prior literature wherein the valence state of
these activators is considered to be: Cu+ , Ag+ and Au+. These are known
stable states of these elements. The electron configuration of the
elemental state is: d ^ s 1 . Note that the d 1 0 electron configuration would
be optically inert. Therefore, it is likely that the monovalent ion has the
ground state of:

6.3.7.- Ag+ = d 9 s 1 = 2
Si/2

However, in our scheme of activator ground states, this electron

configuration would be subject to substantial ground state perturbation,
particularly when used in dielectric host crystals. Indeed, they do not
function very well as activators when incorporated into such hosts.
However, they do function well in the semi-conducting hosts, ZnS (CdS)
and ZnSe. The lowest energy excited state for this electron configuration
is: 4 Di/2 but the transition is not dipole allowed.

A much more plausible explanation considers the silver activator to have

the Ag- electronic configuration. The ground state would then be:

6.3.8.- Ag- = d 1 0 s 2 = 1 So

But the Ag- center would have to substitute on the sulfide site, namely:

6.3.8.- ZnS: Ag s =

where the same mechanism as already given for the formation of the ZnO:
Zn° phosphor applies. But, this does not seem likely due to the mis-match
in ionic sizes. Indeed, the ZnS phosphors are "self-activated", and exhibit
very high luminescence efficiency under both photoexcitation and
 6.3 Design of a phosphor 467

cathode-ray excitation. The band gap of ZnS (or fundamental absorption

edge as it is sometimes called) is about 3.71 ev (or 3350 A). Silver
introduces an energy level at about 0.31 ev in the band gap (3.40 ev
below the conduction level). The "self-activated" phosphor likely has the
formula:

6.3.10.- ZnS : x Zn° s =

In both of the above cases, the ionic radii of the activator states and the
sulfide site are:

6.3.11- S= = 1.84 A
Zn° = 1.48 A
Ag- = 1.62 A (estimated)

Thus, the states for these activators as given in 6.3.5. and discussed
herein would conform to our rules for functioning efficient activators.
However, it is well to note that the mechanism given in 6.3.8. has a
negative monovalent ion substituting at a negative divalent site. This
would imply that an additional negative charge is required for charge
compensation in the crystal. The emission band of silver occurs at 4500 A
(2.76 ev). The "self-activated" ZnS phosphor emits at about 4450 A.
Therefore, delineation between these two types of sites is nearly
impossible. Furthermore, the probability is high that the "self-activated"
center forms first in the crystal and that the Ag- center forms after that.
The best explanation for the formation of silver centers seems to be that
when a sulfide atom is lost, as in 6.3.6., a corresponding zinc atom
becomes interstitial. This would give rise to the following defect
reactions, namely-

6.3.12.- Defect Reactions for Formation of the ZnS:Ag Phosphor

1. ZnS + xH 2 = Zn(Si- x , x V s ) + x H 2 S + (2e" ) x + Zn2+i

2. Ag+ + 2e- = Ag-
3. Zn 2+ i + e- lattice = Zn+i
4. Zn(Si_ x , x V s ) = (Zn,Zn+i) Si- X : Ag"S=
468 Design of phosphors

However, the above explanation applies to phosphors used in the past in

cathode-ray tubes.

The current phosphors being used in color-television tubes are:

Blue = ZnS:Ag :A13+

Green = ZnS:Cu":Al3+
Red = Y2O2S:Eu3+

Note that Al**" is listed as an "activator". It is not. It has been found that if
Al3+ is incorporated within the ZnS lattice, the resulting phosphor retains
its original brightness over longer periods of operation of the television
tube and is less subject to degradation. That is, it does not "discolor" over
the lifetime of the operation of the display. It should now be obvious that
the combination of Al 3 ^ 2 * and Ag"s= represents a pair of charge-
compensated sites which conform to our rules for defect formation in a
divalent lattice. It is clear that, after many years of experimentation, the
best and most stable ZnS phosphor has been shown to be one where the
lattice is free from inherent defects. This was accomplished by
substitution of a trivalent cation along with a negative luminescent cation
(Al3+Zn2+ and Ag"s=) in neighboring divalent sites.

HI. Quenchers, or "Killers" of Luminescence

Cations, having unpaired electron spins, function as quenchers of

luminescence. Some of these are shown in the following diagram, given as
6.3.13. on the next page.

While this may seem incongruous to some (because we have already cited
sub-group IB as an activator-group), it is the actual valence state that is
important. Usually, their electronic configuration involves d- electrons as
well. Note that certain of these cations are optically active in the proper
valence state. In the wrong valence state, however, they function as
quenchers.
 6.3 Design of a phosphor 469

6.3.13.-

[The Periodic Table as Related to Phosphor Composition

Cations with Unpaired Spins

H Which Function as Quenchers He
of Luminescence Ne

(3+M3+) (3+) (2+) (2+) Ar

Ti V Cr Fe2* Co2* Ni Cu Kr

Zr Nb Mo Ru3* Rh Pd Xe

La Hf Ta W Re4* Os4* Ir** Pt Rn

Ac 104

These encompass the periodic table sub-groups:

IIIB, IVB, VB, VIB, VIIIB & IB

For example, in ZnS:Cu, if the Cu-cation is not stabilized in the Cu" state, it
acts as a killer of luminescence. This is particularly true in oxygen-
dominated phosphor compositions. This insight has important
consequences since the method of preparation thus becomes important.
If these cations become stabilized in the wrong valence state, they become
energy traps and dissipaters of excitation energy.

These cations are essentially acceptors, even for resonant energy (dq)
transfer. Since they cannot undergo an excitation transition because of
ground state coupling to the local phonon modes, they function conversely
to the luminescence process. Many of them exhibit strong absorption
bands at the frequencies of light commonly used for excitation of
phosphors. Therefore, excitation energy is dissipated to the lattice by
phonon processes, once it has been captured by this type of site.

An important consequence to this observation is given in 6.3.14.,

presented on the next page.
470 Design of phosphors

6.3.14.- TO OBTAIN A PHOSPHOR HAVING A HIGH DEGREE OF

LUMINESCENCE EFFICIENCY, IT IS IMPORTANT TO EXCLUDE ALL
TRANSITION METALS FROM THE HOST CRYSTAL. THAT IS, A HIGH
DEGREE OF PURITY IS MANDATORY IF ONE IS TO OBTAIN A
PHOSPHOR WITH A HIGH ENERGY EFFICIENCY AND A HIGH QUANTUM
EFFICIENCY.

As a matter of practical consequence, we find that concentrations of

impurities much larger than about 1 ppm. have serious, deleterious effects
upon any given phosphor composition. We can summarize our rules for
Design of phosphors as:

6.3.15.- Cation Anion Activator Option

OPTICALLY: Inert Inert Active (1)

Inert Active Anion (2)

We can choose one or the other option, but not both at the same time. As
an example, we can choose Y2WO6 as a self-activated phosphor
composition. It shows emission from the tungstate group as a broad band
in the blue region of the visible spectrum. Again, the emission is strong
because the excitation-emission transitions involve charge-transfer
mechanisms, which exhibit extremely high oscillator strengths
(extinction coefficients), as for the case of orthovanadate, given above. Yet,
the addition of any of the activators given in 6.3.5. quenches the emission
intensity of the tungstate center. In fact, the loss in intensity is directly
proportional to the concentration of the activator added.

The only activators which can be used with self-activated phosphors are
the rare earths. The reason for this is that the rare earths have the
general electronic configuration: 4fn 5s 2 5p 6 . These outer shielding
electron shells isolate the Ln 3+ center from the influence of the tungstate
center, and vice-versa. We will discuss this type of phosphor in more
detail later in this chapter.
 6.4 Factors affecting phosphor efficiencies All

6-4: FACTORS AFFECTING PHOSPHOR EFFICIENCIES (BRIGHTNESS)

Let us now examine the factors which affect phosphor efficiencies. As we

have said, the presence of an emission band implies an assembly of
emitters (or excited states) having various energy states just prior to
emission, as illustrated in the following:

6.4.1.-

The intensity, I, at any single wavelength is proportional to the number of

centers having that particular energy. Because we have a Gaussian array of
energy emitters, we have an emission band exhibiting a Gaussian
distribution of emission energies. This arises because the phonon energy
flow, although quantized, is random. For this reason, the energy, E(A.) is
proportional to the number of centers, N, having that particular energy at
the time of emission. The intensity of I at any particular time, t, then is:

6.4.2.- I(t) = N E(X).

We can conclude that the probability of emission, at time t, will be related

to the overall half-life, TExcitation » of the excited state, namely-

6.4.3.- P E (t) = f(l/TExcitation)

According to the Binomial theorem (considering that we have either

472 Design of phosphors

ground or excited states), we will have a Gaussian array of emitters

described by:

6.4.4.- N(t) = N o exp - t/^Excitation

where N o is the equilibrium number of excited states present. This

implies that there will be non-radiative transitions present as well, since
ground states are also involved.

We can illustrate this using a potential diagram similar to

that given before, namely:

6.4.5.-

EQ is an activation energy for cross-over (relaxation and dissipation of

energy to the lattice) from excited to ground state. It is easy to show that
energy transferred from the excited state to the lattice has a certain
probability. Pi, to occur, namely-

6.4.6.- Pi = S exp [- EQ / kT]

 6.4 Factors affecting phosphor efficiencies All

where S is an overlap function between the two states. Thus, the

efficiency of emission, i.e.- the number of emitters (excited states) which
actually emit as compared to the original number, N o , is:

6.4.7.- Tieff = N/N o = PEmission/(PEmission + Pi ) = 1/ ^Excited

and: Tteff = { (1 + S /Pi) exp - E Q / kT}" 1

It was SchOn (1931) who first demonstrated that the low temperature
broad band spectra are associated with zero-point vibrational energy
effects (the zero-phonon line transition we have already discussed).
Experimentally, one observes changes in the widths of a given emission
band as a function of temperature:

6.4.8.-

We have already shown these curves in Chapter 5 and repeat them here
for emphasis. On the right is shown the so-called "temperature-
dependence of luminescent intensity" while the left side show the
frequency-broadening of the emission bands due to increased
temperature, i.e.- the temperature-dependent phonon-coupling factor.
The emission band is quite sharp at low temperatures and broadens as the
temperature rises. This is a general property of all phosphors.

However, there is another factor not elucidated before as shown in the

474 Design of phosphors

following description. If one prepares both CaWO4 : W , and BaWO4 : W

(where the W implies self-activation, or emission from the tungstate
group), we find that the former composition is a very efficient ("bright")
phosphor where the latter is essentially inert (at room temperature).

The reason for this is shown in the following diagram:

6.4.8.-

It is easily seen that while the CaWO4 :W phosphor has its maximum
efficiency of emission close to room temperature, the BaWO4:W is most
efficient close to liquid nitrogen temperature. Therefore, the former is
nearly inert at the temperature of maximum emission of the latter and
vice-versa. This phenomenon is quite common for many phosphor
systems, particularly the self-activated ones.

By measuring the temperature dependence of emission, one can estimate

Eg for "temperature quenching". That is, one can measure the
temperature at which emission ceases, or is lost entirely, due to vibronic
coupling to give an estimate of EQ , the activation energy for cross-over
 6.4 Factors affecting phosphor efficiencies 475

(relaxation and dissipation of energy to the lattice) from excited to ground

state.. One of the first cases where this theory was applied by Johnson
and Williams (1952) to KC1:TI, as shown in the following diagram:

6..4.10.-

Here, AE is our Eg of 6.4.5. and is obtained from the slope of the line.
The so-called "quenching temperature" is obtained from the intersection
of the two lines, as shown. It should be clear, then, that temperature
quenching of emission will vary upon the nature of the phosphor, the
activator and the host lattice. Those phosphors involving "optically-active"
cations will be expected to possess lower quenching temperatures
whereas those involving "shielded" energy levels like the rare earths will
have very high quenching temperatures.

For example, Y2O3 : Euo.06 has a maximum brightness output at about 750
- 780 °C (1053 °K) and the emission is clearly visible above the dull-red
heat of the furnace. Yet, the phosphor still glows brightly at 1100 °C,
even though its emission is effectively masked by the full radiation of the
furnace.
476 Design of phosphors

The phenomenon of "temperature-broadening" was also first solved by

Williams and Hebb (1951). By using a Gaussian harmonic oscillator, these
authors solved the problem of obtaining the shape of absorption
(excitation) and emission bands for the quantum mechanical case, namely -

6.4.11.- PE = (K/2^kT)!/ 2 exp { (-KX 2 /2kT)(dX/dE)

where Xwas defined as the change in <r> , and 2X = width of the band =
&. Also, dE is the change in overall energy, and K is a force-constant. (One
can also use temperature dependence effects on bandwidth to get dE).
Williams and Hebb showed that 6.4.11. is satisfactory, provided one
defines an "effective" temperature, namely-

6.4.12.- Teff = ecoth8/T ( 6 = hy / 2k)

This gives us a band width as:

6.4.13.- &(ev) = A(coth [hy /2k]) 1 / 2

where A is a constant characteristic of the material. This result was

confirmed by Russell andKlick (1956) for the case of the F-center in KC1,
as shown in 6.4.14., presented on the next page.

In this diagram, AE is the half-width or I)/2 , as defined above.

Temperature is plotted as T 1 / 2 to show high temperature effects.
Maximum energy of absorption occurs at the intersection of the extension
of the slopes of both sides of the curve.

A slightly different approach uses the maximum energy of the band as a

starting point. This is shown in 6.4.15., also presented on the next page.
Then, the energy of any given point on the curve will be:

6.4.16.- E = E m a x exp -(Am/ c|>)2

where the same assumptions used before are valid, namely that we have a
Gaussian array of absorbers and/or emitters.
 6.4 Factors affecting phosphor efficiencies All

6.4.14.-

6.4.15.-

If we wish to quantify the absorption, then we use:

6.4.17.- A= J"E(co)dco = E m a x Jexp -(Aco/ H) 2 dco

478 Design of phosphors

By obtaining H and (j> experimentally, we can then calculate what the

absorption intensity (oscillator strength) ought to be.

There are two other factors which affect phosphor efficiencies. One we
have already discussed, that of high purity. Reiterating, we must use high
purity materials to prepare phosphors because the presence of very small
amounts of impurities are killers of luminescent efficiency (energy
dissipation traps). The other factor involves "optimum" activator
concentrations actually present in the host lattice.

We have already stated that a general formula for a phosphor is:

6.4.18.- Mi. x XO 4 :A x

where M is the cation, XO4 is the anion and A is the activator cation. It
has been determined that the "optimum" activator concentration should
lie between about:

5 x 1 0 4 < x < 0.20 mols per mol of phosphor.

There is always an "optimum" activator concentration, but the exact

number of sites will vary from host to host. That is, the optimum in one
host, being a function of the plurality of localized phonon modes present
in the lattice, will not be the same in a different host lattice.

The reason for this is that the emission intensity, Emission, is directly
proportional to the number of activator centers present, NA, and is
inversely proportional to (an exponential function) the number of
quenching centers, NQ.These quenching centers maybe nearest neighbor
activator sites which dissipate energy by direct exchange, or, they may be
distant sites accessible by resonant exchange mechanisms. What we find
is that if the activator concentration becomes too high, a loss in energy
efficiency (QE as well) is experienced.

We can illustrate this effect as follows in 6.4.18., given on the next page.
 6.4 Factors affecting phosphor efficiencies 479

6.4.18.-

Initially, the emission intensity is linearly proportional to the number of

emitting activator sites, NA , present, and dl/dx is a constant value. At
some point, the density of these sites becomes too large and the overall
emission intensity falls, according to an exponential rate, as concentration
increases. The reason for this is that the original nearest neighbors
become quenching sites because of direct exchange coupling, or as given
in 5.5.1. of Chapter 5 for the virtual photon exchange mechanism.

Thus, if the number of activator sites increase beyond a specific, but not
easily defined, density, then the energy dissipation mechanisms become
dominant.

There is a further complication, not yet clarified, which involves the

actual "optimum" activator concentration as a function of the host
employed. This is illustrated in the following diagram, given on the next
page as 6.4.20. The activator concentration peak can be quite narrow and
as little as ± 5 x l O 3 mols per mol of host can change the emission
intensity by ± 50% of the peak value in some extreme cases. The
"optimum" activator concentration, being a function of the nature of the
480 Design of phosphors

activator chosen and the nature of the host is not easily predicted or
calculated. It is usually experimentally determined.

6.4.20.-

9-5 - PREPARATION OF PHOSPHORS

In this section, we will first describe the preparation of phosphors in

terms of the parameters involved, followed by measurements usually
required to ascertain the energy efficiency of the phosphors so- produced,
and finally how to make such measurements.

I. Phosphor Parameters

Phosphor preparation parameters can be divided into two (2) classes,

independent and dependent variables.

a.- Independent Variables

The following lists the independent variables of phosphor preparation:

 6.5 Preparation of phosphors 481

6.5.1.- Independent Variables

1. Purity of materials 4. Matching of ionic size of

activator and host cation
2. Choice of host lattice 5. Charge compensation
methods in host
3. Choice of activator 6. Ratio of reacting components

We have already emphasized the importance of purity within the finished

phosphor. REITERATING, we must exercise all caution to choose pure
raw materials to form the phosphor and prevent further contamination
from occurring during processing to form the phosphor.

Let us choose a phosphor composition. For the cation, we choose Ba2+ , a

silicate as the anion, and an activator such as Pb2+-. This could give us a
formula such as: (Bai- x Pb x ) SiO4 . We find that Vegard's Law holds for
these two cations and they will form a solid solution. There is no need for
charge-compensation since both cations have the same valence state.
However, by reference to the phase-diagram for BaO and SiC-2, we find
that the formation of several compounds is feasible. In this diagram, given
as 6.5.2. on the next page, the existence of four(4) compounds is
indicated, each of which could serve as our host. It is evident that by
changing the ratio of BaO to SiO2 , one can obtain the various
compositions. We could prepare the mesosilicate compound by solid state
reaction of:

6.5.3.- BaCC-3 + 2 SiO2 = BaSi2C>5 + CO2

The different compounds which result are:

6.5.4.- MPLS PRODUCT

BaO S1Q2
2 1 Ba2SiO4
1 1 BaSiO3
2 3 Ba2Si3O8
1 2 BaSi2O5
482 Design of phosphors

6.5.2.-

The major question we need to ask is which of these compositions will

produce the most efficient phosphor. This is a difficult question to answer
if we do not already know the answer! It turns out that the BaSi2Os
composition is the one which works best with the Pb 2+ activator.

In contrast, Ba2SiC>4 works better with Sn 2+ . The exact reason for this is
not known, but it is suspected that the phonon spectrum of the host
and/or the crystal structure is the answer. This brings us to another
major rule in phosphor preparation, namely:

6.5.5.- WHEN CHOOSING A SPECIFIC HOST FOR A SELECTED

ACTIVATOR, ONE MUST DO SO BY EXPERIMENTAL MEANS, SINCE THE
METHOD OF PROPER COMBINATION OF HOST WITH A GIVEN
ACTIVATOR IS NOT COMPLETELY UNDERSTOOD, THAT IS - WHY A
PARTICULAR HOST WORKS WELL WITH A PARTICULAR ACTIVATOR IS
NOT FULLY UNDERSTOOD AT THIS TIME.

Therefore, in our model phosphor, we would ordinarily prepare all four

 6.5 Preparation of phosphors 483

compositions, and finalize the composition according to our experimental

results.

Another independent variable we wish to consider is that of the ratio of

reacting components. As we have already shown in 6.5.2., a 1:2 ratio of
reacting components produces the BaSi2Os composition. However, as we
have also shown in Chapter 3, the extent of the solid state reaction
between two components (i.e.- the "completeness" of the reaction)
depends upon the relative diffusion rate of the reacting species. In our
case, this is the Ba 2+ cation. Thus, if we use a 1.00 : 2.00 ration, we would
find that the reaction would proceed to about 98-99% and then nearly
stop. The reasons for this have already been discussed in Chapter 3, but
can be summarized as follows. The rate of reaction between two solids is
an exponential one, rapid in the beginning, but slowing as the
components are used up. An asymptote is gradually approached, but the
reaction never becomes 100% complete. Therefore, in a solid state
reaction to form a given compound, we always end up with a small amount
of unreacted components . Usually, this amount is in the 100 -1000 ppm.
range.

Nevertheless, any such "impurity" severely affects the efficiency of the so-
produced phosphor. We find that, for the present case of BaSi2O5 , the
presence of even very small amounts of BaO has a very strong absorption
band in the ultraviolet which detracts form the overall performance of the
phosphor when used in a fluorescent lamp. In contrast, an excess SiO2
is completely transparent in the UV. Even if the phosphor is to be used in
a cathode-ray tube, the presence of excess cation is deleterious to its
successful operation as an efficient phosphor. Our solution is to use an
excess of the non-optically active anion to promote the completeness of
the reaction. In general, the excess required to do so is not more than a
few mol%. An example of this is given in 6.5.6., presented on the next
page.

If we performed both of these reactions and incorporated Pb 2+ as an

activator, we would find that the phosphor produced by Reaction (1) was
much superior in performance to that produced by Reaction (2).
484 Design of phosphors

6.5.6.- Reaction (1):

1.00 BaCO3 + 2.025 SiO2 = 1.00 BaSi2C>5 + 0.025 SiO2 + 1.00 OO2

Reaction f2)

1.025 BaCO3 + 2.00 SiO2 = 1.00 BaSi2O5 + 0.025 BaO + 1.025 CO2

Yet any analysis we might perform to discern the difference between the
two phosphors would yield no answer because the two phosphor
compositions are the same, with the exception of the small amount of
excess reactant in each case. Therefore, another rule for preparing
phosphors is:

6.5.7.- IN FORMULATING A PHOSPHOR COMPOSITION, ONE ALWAYS

EMPLOYS A SMALL EXCESS OF THE ANIONIC REACTANT SO AS TO
AVOID THE PRESENCE OF STRONGLY ABSORBING CATIONIC SPECIES
IN THE END-PRODUCT.

This rule is a general one since all of the anions used (as given in 6.3.2.)
are optically transparent. The only case that one will encounter is that
where an optically active anion is chosen. Even in that case, one
formulates the composition so that only a very slight excess (about 0.005 -
0.01 mol%) of the anion is present in the reacting species to form the
final composition.

After we have followed all of these rules, our phosphor formulation turns
out to be:

6.5.8.- 0.99 BaCO3

0.01 PbCO3
2.02 SiO2

The last factor to consider is that of charge compensation. Suppose we

substitute Sb 3+ for Pb 2+ . This would require a formulation such as that
given in 6.5.9. on the next page.
 6.5 Preparation of phosphors 485

6.5.9.- Formulation for BaSi2O?:Sb3+:K+

0.99 BaCO3
0.005 Sb 2 O 3
0.005 K2CO3
2.04 SiO2

The slight increase in silica content is required to maintain a 0.02 mol

excess, since the (Sb,K) cations use up 0.02 mol SiO2 more than the
reaction of 6.5.8.

Nevertheless, there are many cases where charge-compensation cannot

be effected, in spite of anything we might try to do. A case in point is
CaF2= Dy3+ • We might try to use Na+ as a charge-compensating ion, but we
will find that it is not effective. Defect reactions of Dy3+ with this lattice
are:

6.5.10.- Dy3+ => Dy3+ca + V >

Dy3+ + 0= => Dy3+Ca + O >

Even if all oxygen atoms are excluded during the reaction to form the
crystal (including that of water), lattice defects will be formed. In fact, the
presence of charge-compensating ions do not affect these defect
reactions. The phosphor compositions we obtain because of the
predominance of the defect reactions are:

6.5.11.- {(1 -x) Ca,x[Dy3+ C a + V > ]} F 2

{(1 -x) Ca, x [ Dy3+Ca + o > ]} F 2

The phosphor represented in the first formula is a good

thermoluminescent (TL) material for measuring exposure to radiation
dosage. The vacancies function as traps for energy, and that energy is
released by heating the phosphor and measuring the amount of energy
trapped (obviously, one has to calibrate the dosimeter before using it
experimentally). But, the second is not a good TL phosphor. Thus, we
must be very careful to completely exclude the presence of oxygen from
486 Design of phosphors

any source , if we are to obtain a good thermoluminescent dosimeter

material.

Finally, we should reemphasize the effect of impurities on luminescent

efficiencies. As we have already said, even parts per million of excess
transition metal cations cause unwanted absorption and seriously affect
phosphor energy efficiencies. These cations are already present as
Impurities. However, the inclusion of excess anion in the phosphor
composition precludes this adverse effect.

b. Dependent Variables

What we mean by dependent variables are those parameters which are

interdependent, i.e.- as one is changed, the others are affected as well.
They include:

6.5.12.- Dependent Variables

1. Firing temperature 6. Firing time

2. Size of mass being fired 7. Firing atmosphere

3. Activator concentration 8. Effect of host structure

4. Effect of method of preparation 9. Ratios of reacting components

5. Use of a flux

It may seem somewhat surprising to see the same parameters listed

herein when we have already discussed them as independent variables.
However, as we shall see, they are not completely independent
parameters and so must be included as dependent parameters as well.

Before we can proceed to an in-depth examination of the dependent

variables, we need to examine the steps required to prepare a phosphor.
These are shown in 6.5.13., given on the next page.
 6.5 Preparation of phosphors 487

6.5.13.- STEPS IN PREPARING A PHOSPHOR

1. Selection of materials
2. Assay
3. Weighing
4. Blending
5. Firing cycle
6. Dispersion
7. Evaluation

SELECTION OF MATERIALS - We require raw materials having an

impurity level less than 10 ppm. of total impurity cations. In particular,
we need to avoid the "killer" cations given in 6.3.13. It is advantageous to
choose a compound which decomposes to form an oxide. A case in point
is our chosen phosphor composition where we will choose BaCO3 as a
reacting component rather than BaO, because of the reasons stated in
Chapter 3 regarding the nature of diffusion reactions and their effect upon
the final product. If we chose Ba(NO3)2 , we find that the acidic N2O4
produced during decomposition of the nitrate has a deleterious effect on
the produced phosphor. This does not occur with the carbonate. In a like
manner, if we were to choose Na2SiO3 as a reactant, we could do so.
However, we would have to use BaCl2 in conjunction with this compound
so as to produce NaCl as a final reactant. Then we would remove the NaCl
so-produced by washing the product in water. Nevertheless, we would
find that very small amount of Na+ cations had been incorporated into the
host lattice, BaSi2Os , in spite of any precautions we might have taken. For
these reasons, we choose certain special compounds, so chosen as to
produce innocuous by-products such as gases.

ASSAY: To maintain strict stoichiometry in the end-product, we

determine the amount of adsorbed water present on each of the reacting
components. We then recalculate the exact weight needed to compensate
for the amount of water present.

WEIGHING: The exact amounts of components are weighed and blended

together.
488 Design of phosphors

BLENDING: - This is one of the most important steps in the process.

Usually, we prefer to use a hammermill and partially grind the materials
together to form a homogeneous mixture, on a particle scale. A mortar
and pestle can serve for very small amounts of materials. In the final mix,
the individual particles need to be tangent to one another so that the solid
state reaction can proceed.

FIRING CYCLE - This will be discussed in detail below.

DISPERSION - We need a fine powder in most cases, and after firing, we

break up the "cake", inspecting it under ultraviolet light to discard any
inert parts. We then crush the "cake", and sift it to obtain a fine power. If
a flux is used, we then wash the powder to remove any soluble flux
present, and then dry the powder.

EVALUATION - There are a series of measurements which we will make

to determine the energy efficiency and QE of our final product.

c. The Firing Cycle as a Dependent Variable

Each phosphor composition has its own unique, but required, firing cycle.
Note that we consolidate firing time and firing temperature together,
because they are inter-dependent. First of all, we choose a temperature
that is above the minimum temperature required for the solid state
reaction. In many cases, this is a matter of experience, but this minimum
is easily obtained from DTA and TGA data. A temperature at least 200 °C.
above this minimum is generally required. What we find is that there are
certain combinations of firing time and temperature which produce the
best results in terms of quality of phosphor so-produced.

An example for the phosphor, YNbO4 :Eu is shown in 6.5.14., given on

the next page. This phosphor was produced by the solid state reaction:

6.5.15.- Y2 (C2O4)3 • 4 H2O + Nb 2 O 5 = 2 YNbO4 + 4 H2O + 3 CO + 3 CO2

where the Eu 3+ is not shown for simplistic reasons.

 6.5 Preparation of phosphors 489

6.5.14.-

The measured responses (a measure of relative energy efficiency) are

shown in the diagram as slopes, where the topology varies in the "z"
direction. The "x" direction is the time variable and the "y" direction is
the temperature variable. What we have is a "ridge" (denoted by the
straight line in the middle) where the firing cycle effects, i.e.- the
combination of T and t, are linear.

The optimized firing times are:

For the YNbC>4 :Eu phosphor: T(°C) = - 8.33 t(hrs) + 1225

For the LaNbO4:Eu phosphor: T(°C) = - 12.5 t(hrs) + 1305

From this diagram, we can choose a proper firing cycle for either of these
phosphors, providing that we use the oxalates of the rare earths as raw
materials.

As an example of how the solid state reaction proceeds to form the

490 Design of phosphors

luminescent state, the following diagram is presented:

6.5.16.-

The solid state reactions which occur and the stages of development
which occur during phosphor formation are remarkably alike from system
to system. These stages are shown for both Srs F(PO4)3:Sb and
Sr5Cl(PO4)3:Sb. The emission colors for these phosphors is blue-green.

The stages for the phosphor, SrsF(PO4) 3:Sb are as follows:

 6.5 Preparation of phosphors 491

(1) At 1/2 hour firing time ® 1100 °C., the mass is mostly inert with
a few luminescent "spots"
(2) At the 1.0 hour firing time, a blue-emitting phase is obtained,
with only a small amount of inert material present
(3) By 2 hours firing time, the blue-emitting phase has shrunk and a
green-blue emitting phase has proceeded to grow from the outside;
At 3 hours, the transformation is complete to the green-blue
emitting phase. When the firing temperature is 1200 °C, the same
mechanisms are observed, except that they occur faster.

The stages for the Sr5Cl(PC>4)3:Sb phosphor are more revealing:

(1) At the 1100 °C. firing temperature, the growth of the blue-green
emitting phase is slow and becomes complete at about 4 hours of
firing time, while the inert phase diminishes as the luminescent
phase grows.
(2) At the 1200 ° C. firing temperature, the blue-green emitting
phase has already passed its peak growth and has begun to change
into a different inert phase. This inert phase increases with firing
time. Thus, we have discovered that it is possible to "overfire" a
phosphor.

What we have learned is that there are at least four(4) stages which occur
in the formation and development of the luminescent phase, to wit:

1. Solid state reaction to form the host structure - inert

2. Sintering to form the crystalline lattice - inert
3. Formation of the luminescent phase, including
rearrangement of host crystal components.
4. Continued reaction to form a different luminescent phase,
including that of a separate post- luminescent inert phase.

In the case of our strontium fluoroapatite phosphors, we can explain the

various changes rather easily. Initially, the host lattice forms and grows
more crystalline. However, the host-activator combination does not
become luminescent until the activator cation is incorporated at the
492 Design of phosphors

correct lattice site to form the luminescent center. Originally, there is a

fluorine-rich structure, but as firing time continues, oxygen diffuses into
the structure, thereby changing the emission color. This may be due to a
change in the site-structure of the activator, Sb 3+ , or the formation of
oxygen- vacancy sites, as in the example of CaF2:Dy> given above. The
chloroapatite phosphors are even less stable, and are subject to loss of
chloride atoms forming the host lattice, thereby resulting in an inert
material. The presence of oxygen in the lattice may also be a major
contributor to obtaining the "over-fired" inert phase.

WHAT THIS ALL MEANS IS THAT THERE IS AN "OPTIMUM" FIRING

CYCLE REQUIRED SO AS TO OBTAIN MAXIMUM ENERGY EFFICIENCY
IN THE SO-PRODUCED PHOSPHOR.

d. Size and Mass Fired

Many cases will be encountered during the preparation of phosphors

where the activator forms a volatile compound. If we useAs 3 + , Sb 3+ , Bi 3+ ,
Pb 2 + , T1+ and Zn° or Cd° as activators, then the probability is high that
volatile compound formation will ensue. This will lead to loss of activator,
thereby causing a lower concentration than intended. However, there is
also a beneficial side-effect in that the activator becomes better dispersed
throughout the host structure when a volatile compound is formed.

We can control this loss by control of the mass being fired. Many times,
this critical parameter leads to improved energy efficiency of the
phosphor thereby produced. The significant parameter is surface area per
unit mass. Consider the following:

6.5.17.- Surface Area to Volume Ratios fS/V)

Diameter Cubes Spheres

1.0 cm 6 cm 2 /cc. 6 cm 2 /cc.

10 cm 0.6 " 0.6 "
100 cm 0.06 " 0.06 "
 6.5 Preparation of phosphors 493

Note that shape is not a factor. Therefore, the lower the value of S/V, the
lower will be the loss of any volatile component during firing.
Consequently, we may deliberately choose a chemical form of the activator
which is volatile so as to take advantage of the dispersion mechanism.
Then we must adjust the mass parameter and sometimes the firing cycle
as well.

e. Firing Atmosphere

If we examine the oxidation states of the activators of 6.3.5., it is apparent

that most require one of their lower valence states to function as an
activator. Therefore, one needs to provide a neutral, or slightly reducing,
atmosphere during firing, since the normal air atmosphere would
promote the higher valence states. Suitable neutral atmospheres include
N2 , A , Ne, etc. A reducing atmosphere would usually consist of 95-98%
N 2 and 2-5% H 2 .

However, there are cases where it is necessary to fire in air. A case in

point is the tin-activated phosphors, involving the Sn 2+ activator. These
are fired in air first, and result in an optically inert material. They are
then refired in a reducing atmosphere to develop the luminescence of the
Sn 2+ activator. The reason for this is that most tin compounds are easily
reduced to the metallic form. By first firing in air, the tin is incorporated
within the host lattice in an optically inert form. Refiring in a reducing
atmosphere then converts the tin to the required lower valence state, and
phosphors having high energy efficiencies and brightness result. We may
conclude that the chosen firing procedure will depend upon the chemical
properties of the reactants which are to form the phosphor composition.

f. Effect of Host Structure

Little has been said concerning this parameter because of its complexity.
The subject is so complicated that a separate treatise could be written on
this subject alone. Essentially, it involves two factors: 1) the site symmetry
at the cation site, and 2) the strength of the crystal field at that site.
Consider the activator, Sn 2+ . As we change the host, the emission colors
494 Design of phosphors

change from the ultraviolet to the deep red. That is, we obtain a series of
phosphors having various emission colors, even though the activator
remains the same. This is easily explained by considering the following
diagram:

6.5.18.-

The emitting state for Sn 2+ is the 3 P i state. The free-ion energy level lies
above 40,000 c m 1 . But in a crystal field, this level splits into three field
components. For this reason, we can vary the emission band of Sn 2+ by
varying the nature of the host employed.

The alkaline earth pyrophosphates are polymorphic in nature. The low

temperature form is called J5- M2P2O7 while the high temperature form is
a- M2P2O7 for M= Ca and Sr. For Ba, a- is the low temperature form,
while 6-Ba2P2C>7 is the high temperature form. The firing temperature
required to change one form to the other (with differing crystal structure)
varies from 1150 °C.(Ca) to 750 °C. (Ba). The requisite crystal structure
factors for these crystal forms are shown in 6.5.18., given on the next
page. Because of the similarity of these compounds, we can formulate
mixtures of the cations to form a solid solution, and thereby "fine-tune"
the strength of the crystal field at the activator site.
 6.5 Preparation of phosphors 495

6.5.18.- Alkaline Earth Pvrophosphates

LOW TEMPERATURE FORMS

Crystal Structure Lattice symmetry Cation Site Cation

Symmetry Coordination
fi- Ca2P2C>7 Tetragonal (^4 (P4i) Ci {7 oxygens or
- 8 oxygens}
J3- Sr2P2O7 •• " " {9 oxygens

a- Ba2P2O7 Orthorhombic C92v (Pna2i) Ci 6 oxygens

HIGH TEMPERATURE FORMS

a- Ca2P2C-7 Orthorhombic C92v (Pna2i) Ci 6 oxygens

a-Sr2P2O7

6- Ba2P2O7 NOT KNOWN

Note that we have not included the magnesium pvrophosphates. The high
temperature form is a-Mg2P2O7 which has the thortvieite structure
(monoclinic with space group: C32h (C2/m) and cation site symmetry =
Ci). It has the unusual feature that it transforms to the low temperature
form at 68 °C. The low temperature form is P-Mg2P2C-7 (monoclinic with
space group: C52h (P2i/c) with cation symmetry = Ci). Neither form
produces a tin-activated phosphor as shown byRopp (1960).

This brings up a significant point, namely that when we have a crystal host
with a low transition temperature, we will not obtain a luminescent
material. The reason for this is that once we have formed a given crystal
structure during the firing cycle, the activator is situated in a cation site
having a specific symmetry. We normally "quench" that phase by removing
the fired mass from the furnace. The rate of cooling is generally fast
enough to ensure that the desired crystal structure and site symmetry
results. However, if a very low transition temperature exists, the mass
496 Design of phosphors

cannot be cooled fast enough to prevent the crystal structure

transformation from occurring. This results in a host lattice having
considerable strain and distortion at the cation site. In our case, the
problem is further compounded by the disparity in cationic size between
Mg 2+ and Sn 2+ . Thus, another rule concerning phosphor preparation is:

6.5.19- IN THE SELECTION OF HOST COMPOUNDS TO PREPARE A

PHOSPHOR, WE AVOID THOSE WHICH POSSESS A LOW TEMPERATURE
FORM, ESPECIALLY IF THEY HAVE A LOW TRANSITION TEMPERATURE.

This explains the non-luminescent behavior of the Mg2P2O7 crystal

forms.

Note that as we change crystal structures in the other alkaline earth

pyrophosphates, changes in lattice symmetry, and cation coordination
occur, even though the local cation symmetry remains the same, i.e.- Ci =
no symmetry. Nevertheless, there are major effects of these seemingly
similar crystal structure parameters upon the phosphor so-produced, as
shown in 6.5.20., given on the next page.

These data illustrate the several effects that the crystal lattice parameters
have upon both the intensity (energy efficiency) and energy content
(wavelength) of the emission bands. Thus, for the low temperature forms,
P-M2P2O7, both Ca and Sr have the same lattice symmetry and site
symmetry but differ slightly in cation coordination. This results in
essentially the same emission band for both.

The "cation-fit" parameter, however, plays a major role in determining

relative emission intensity, i.e.- energy efficiency, in that the Sr-
compound is by far the "brightest" phosphor.

For the a-Ba2P2O7 composition, the crystal structure and cation

coordination differs from the others. This, plus the larger size of the Ba2+
cation, causes a high crystal field intensity to appear at the activator site,
and results in a major shift in emission wavelengths.
 6.5 Preparation of phosphors 497

6.5.20.- EMISSION BANDS OF Sn2+ IN ALKALINE EARTH

PYROPHOSPHATES

Emission Cation Relative Intensity

Radius
2
Color Peak (Sn + = 1.12 A)

«- M2P2O7 - High Temperature Forms

Ca U.V. 3620 A 0.99 A 17

Blue 4290 15

Sr blue 4630 1.13 100

S-Ba deep-red 6760 1.35 40

tMsE2Q7_- Low Temperature Forms

Ca U.V. 3690 A 0.99 A 14

Sr U.V. 3640 1.13 80

a-Ba green 5050 1.35 65

For the high temperature forms, a - M2P2O7 , two emission bands appear
in the Ca compound denoting the presence of two cationic sites. The
reason for this is not known.

We cannot compare the effect of lattice symmetry or cation coordination

on the emission bands, but can observe the effect of the cationic size on
the energy position of the emission band (the strength of the crystal field
at the activator site) as well as the "cation- fit" parameter on the emission
intensity. A major shift in energy position of the bands is evident. Ba2+ has
more electrons than Ca2+ so that the crystal field (10 dQ) at the activator
site is stronger and the emission band is shifted to lower energies.
However, part of the effect may also be attributable to a change in the
498 Design of phosphors

crystal structure. We cannot evaluate this effect because the structure of

6- E&2P2O7 is not known.

For the general phosphor composition, (Sr x Cay Baz )2p2C>7 :Sn, where x *
y * z, there is a linear relation between peak emission energy (the zero-
phonon line) and the cationic radius, namely-

6.5.21.- E (cm 1 ) = - 9610 R(A) + 32,696

What this means is that we ought to be able to "fine-tune" the emission

band between about 4300 A and 5000 A. by forming the requisite solid
state solution containing the various sized cations. In practice, this
approach has limited utility because as one shifts the ratio of : (Sr + Ca +
Ba) = 2.00, either the a - or J3- M2P2O7 form becomes the dominant one.
However, if x » y or z, that is- if the a- Sr2P2O7 structure is maintained,
then one can shift the emission band to some degree.

The above example of the effect of the host crystal is generally applicable
to all cases we might encounter. Even our illustrative phosphor,
BaSi2O5:Pb, follows these rules. The structure is orthorhombic
(sanbornite) with a lattice symmetry of: D72h (Pmna) and a site symmetry
of C2h- Site coordination is six (oxygens). The cationic radii are: Ba2+=
1.35 A and Pb 2+ = 1.20 A. The emission band occurs at 3660 A, with a QE
of > 75% (2537 A excitation).

Our last example of the effect of host lattice on emission properties of

phosphors is that of Ce 3+ This activator has a ground state of 2 F5/2 arising
from an electron configuration of 4f 1 5s 2 5p 6 . The lowest excited state is
2
D3/2, arising from the 4f°5s 2 5p 6 5d x configuration. This is an allowed
dipole transition with the excited-"d"-state lying outside of the shielding
electron shells. Thus, it is subject to the influence of the crystal field of
the host lattice. In YPO4, the emission band is narrow and peaks at 3260 A
(Ultraviolet). In Y2SiOs, the emission broadens somewhat and peaks at
4200 A (Deep-blue). But in Y2AI5O21, the emission peak is very broad and
peaks close to 5460 A (Yellow). All of these phosphors have comparable
 6.5 Preparation of phosphors 499

QE's close to 80%. Quite obviously, the strength of the crystal field at the
activator site has a major effect on the position of the emitted band.

g. Effect of Preparation Method Employed

We now return to our phosphor example, BaSi2Os :Pb. There are at least
three (3) methods we can employ in the general preparation of
phosphors, these are:

6.5.22.- POSSIBLE METHODS OF PREPARING PHOSPHORS

1. Reaction between Mixed Oxides - BaO+ SK>2

2. Precipitation of a Raw Material Containing the
Activator
3. Direct Precipitation of the Phosphor

We have already discussed # 1. To coprecipitate a raw material, we would

use Na2SiO3 as a solution in water along with the proper combination of
BaCl2 and PbCb in solution. We then precipitate at 100 °C. so as to obtain
a crystalline product, adding the silicate solution to the cation solution.
The precipitate is filtered, and the NaCl removed by washing. We then
must add 1.00 mol of SK>2 to the raw material and then fire the mixture.
The solid state reaction is:

6.5.23.- (Ba,Pb)SiO3 + SiO2 = (Ba,Pb)Si2O5

Note that we are not addressing the cases where we use raw materials to
prepare the phosphor, but of coprecipitated material containing both the
cation and the activator. An example of a raw material used to prepare a
phosphor would be SrHPO4 which reacts:

6.4.24.- 2 SrHPO4 = Sr2P2O7 + H2O

But here we would have to add SnO to the SrHPO4, before firing, to obtain
the phosphor, Sr2P2O?:Sn. In this case, we find that it is very difficult to
500 Design of phosphors

attempt to coprecipitate the phosphates from solution because of the

disparity in their chemical solution behavior.

There are a few phosphors which can be directly formed by precipitation

from solution. One such is YVO4: Eu. The chemistry is complex, since one
must start with the ions, VO2+ , Y3* , and Eu 3+ , and convert these to the
final product. The precipitate itself has a brightness of about 40% of the
final fired phosphor.

This brings us to another rule for phosphor preparation, namely -

6.5.25.- FOR THE MOST PART, WE USE AN INTERMEDIATE

PRECIPITATED RAW MATERIAL WHICH SERVES AS A BASE IN THE
FORMULATION. IN THIS WAY, WE AVOID THE DIFFUSION-LIMITED
SOLID STATE REACTIONS THAT MIGHT PRODUCE BY-PRODUCTS
AND/OR LATTICE DEFECTS IN THE FINAL FIRED PHOSPHOR PRODUCT.

h. Ratios of components and Use of a Flux

We have already discussed the effects of varying the ratios of reacting

components, both in this chapter and in Chapter 3. Reiterating,
sometimes the solid state reaction does not proceed as predicted,
particularly if we use mixtures of oxides, or if we use an exactly
stoichiometric mixture of reactants, regardless whether the mixture
includes a precipitated raw material which is to serve as a base for the
overall solid state reaction. It is for this reason that we always include a
small excess of the anion so as to drive the reaction to completion.

A flux is defined as an additive which does not affect the reactants during
the firing cycle, but which promotes the crystal growth of the final
product. It does so by providing a liquid phase for transport of material,
including ions, so as to increase the crystallinity and size of the particles
so-produced. The final product should also be non-reactive with the flux
at the firing temperature. Because the flux must be removed after the
reaction is complete, it also needs to be water-soluble. An example of the
action of a flux is given as 6.5.26. on the next page.
 6.6 Commercial phosphors 501

6.5.26.-

Particle (3) is the recrystallized product produced by liquid transport

through the flux, from material transported from particles (1) and (2).

6-6 : COMMERCIAL PHOSPHORS

There are two classes of phosphors that we will discuss in the chapter.
One is designed for photoluminescence and is used in fluorescent and
high-pressure mercury lamps, while the other is designed for
cathodoluminescence and is used in cathode-ray tubes, including Color
Television. The former must have a high energy efficiency for 2537 A
photon excitation, the resonance radiation of a mercury-vapor discharge,
while the latter must exhibit high energy efficiency when bombarded with
a high-voltage electron beam. The design requirements for these two
classes of phosphors are quite different, and we will discuss each in turn,
including the special technology that has grown around each.

I. Cathode Ray Phosphors

The cathode-ray tube is now over 100 years old. In 1880, William Crookes
had demonstrated that cathode-rays, i.e.- an electron beam, would cast
sharp shadows on a fluorescent screen if a mask was present within the
tube. It remained for Karl Braun to invent the device that changed the
world, a working cathode-ray tube (CRT) having internal electrodes which
502 Design of phosphors

could deflect the electron beam when a voltage was applied. To reach this
phase of CRT technology, phosphors, vacuum technology and discovery of
the electron and its properties had to precede the CRT. It was 1603
when Vincenzio Cascariolo described the first phosphor. Cascariolo's
phosphor evidently was a barium sulfide, made by firing a barium sulfate
("Bologna Stone") with graphite. The first commercially available
phosphor (1870) was "Balmain's paint," a calcium sulfide preparation. In
1866, the first stable zinc sulfide phosphor was described. Braun's CRT
contained all of the components of the modern CRT including: an
electron source, electron beam focusing and deflection, acceleration of
the electron beam and a phosphor screen. These are condensed into two
main components of the CRT, namely the electron gun and the phosphor
screen assemblies. These two have received the most effort as the CRT
has been improved over the years. We will address the CRT phosphors
used up to 1990 in this chapter and the latest advancements in display
technology in the next chapter.

a. How Cathode-Rav Phosphors Are Used

In a cathode-ray tube, an electron-beam is generated within an "electron-

gun", and caused to sweep across a faceplate containing a phosphor
"screen". This screen is actually a uniform layer of phosphor particles
deposited by special techniques. The beam "paints" a series of lines on the
phosphor-screen to form a rectangle called a "raster" which emits visible
light, as shown in 6.6.1. on the next page.

The electron gun has a cathode which emits electrons. The electrons are
controlled and bunched into a beam as it exits the electron gun. The
deflection coil varies the motion of the electron beam, generated within
the electron gun, to form the raster. The number of separate lines within
the raster is about 525, but the separate rasters are interwoven so they
overlap to form the complete picture.

The raster on the phosphor screen is excited in real time so that we see
moving pictures. The human eye has a movement perception of about
1 / 20 of a second so that the raster, reformed about every 1 / 30 of a second
 6.6 Commercial phosphors 503

6.6.1.-

gives the appearance of continuous motion. A close-up of the typical

electron-gun is shown in the following:

6.6.2.-

The acceleration electrode is connected to the high voltage anode (~3O

kV) while the other anode are connected to the electron gun voltage
sources. The color television tube operates on the same principle except
that three colors of phosphors are used, as shown in 6.6.3., presented on
the next page. This mandates the need for three separate electron guns
or means of exciting each color separately from the other colors
504 Design of phosphors

6.6.3.-

(if one gun is used). It has proven easier to use three guns rather than
one, since the registration problems of each color screen, excited
separately by the one gun, was found to be very difficult. The electron
beam from the single gun had to be directed by a signal from a fast-decay
phosphor in order to follow the TV registration signal. Thus, the three-
gun system has dominated to this day.

The two systems shown above are in present usage. One is based on a
phosphor dot pattern, arranged so that a metallic screen, containing
holes in it, is placed between the electron-gun assembly and the phosphor
dots (which are 0.017 inch in diameter, on the average). This screen is
called a "shadow-mask" and contains exactly 1 / 3 of the total number of
dots actually present on the phosphor screen. Each shadow-mask hole is
positioned exactly at the interstice of each three-dot triad. Since there
are three electron-guns, each gun excites just one color of phosphor.
Thus, we have three color-rasters which are superimposed on each other,
each raster being generated by its own electron-gun controlled by the
color signal being transmitted to the color T.V. set. The line-raster
system is similar to the dot-raster system except that vertical lines are
etched in the shadow-mask and a single electron-gun can be used. The
line pattern exposure is generally more open than that of the hole pattern
exposure so that more electrons get through the shadow mask, resulting
in a brighter picture at the faceplate of the television tube. What we are
discussing is shown more clearly in the following diagram:
 6.6 Commercial phosphors 505

6.6.4.-

Here, we show the three electron guns and their arrangement in relation
to the three phosphor stripes. The Trinitron™ system has come into
general use because the slits in the shadow mask are more open, resulting
in a brighter picture on the screen with much better color separation and
clarity, i.e.- contrast in the picture.

The most recent T.V. tubes have eliminated the shadow mask and use a
single gun. The position of the electron beam, i.e.- the instantaneously
color being generated, is controlled by a secondary phosphor layer, on top
of the emitting phosphor layer, which consists of a U.V. phosphor with a
microsecond decay. Since the color phosphors decay in milliseconds,
there is time for the feed-back signal to be detected and used to control
the color generated in real time, as each raster is being generated.

From this discussion, it should be obvious that the two most important
properties of cathode-ray phosphors are the response to electron-beam
excitation (brightness) and the decay time. We require a long-decay
phosphor for radar applications and a short-decay phosphor for television
usage. Nearly all the cathode-ray phosphors are based on the zinc and
cadmium sulfides because they exhibit the highest efficiency to cathode-
ray excitation. ZnS forms a series of solid solutions with CdS whose
emission band can be shifted from the blue (ZnS:Ag) to the red phosphor,
506 Design of phosphors

i.e.- [(Zno.20. Cdo.8o)S:Ag]. This is shown in the following diagram:

6.6.5.-

When a 20 KV electron strikes the ZnS phosphor, its energy is absorbed

by the lattice, resulting in band-gap excitation. This mechanism is
sometimes described as an "exciton" wave. The lattice energy then travels
until it reaches an energy sink, usually the activator center. The overall
energy efficiency for ZnS:Ag is about 20-22%. The next best Cathode-Ray
phosphor is about 6-8%, except for some of the newer rare earth
phosphors (see below). The continuous shift in emission with Cd/Zn ratio
has led to the suggestion that the luminescent centers in this system are
associated with the sulfide, rather than the specific activators employed
(Leverenz-1950). As we have already discussed, because of the defect
reactions which certainly proceed during the preparation process (due to
the presence of oxygen and water), the formation of the "self-activated"
phosphor (i.e.- ZnS: Zn°s= : Emission peak = 4350 A) is nearly impossible

to avoid, even when other activators are present.

b. JEDEC Phosphors

The various cathode-ray phosphors have been classified over the years
 6.6 Commercial phosphors 507

according to a JEDEC number (Joint Electron Device Engineering

Council). This is the designation most often referred for any cathode-ray
tube application. Table 6-2 gives a complete listing as follows:

TABLE 6-2
CHARACTERISTICS OF JEDEC PHOSPHORS

COMPOSITION COLOR PERSISTENCE APPLICATION

p-1 Zn 2 SiO 4 :Mn green medium-20 ms oscilloscope
P-2 ZnS:Cu blue- medium(70 ms) oscilloscope
green
P-3 Zn2BeSiO4:Mn green- medium(20ms) oscilloscope
yellow
P-4 ZnS:Ag white medium short B&W television
+ZnCdS2:Ag (100 us)
P-5 CaWO4:W blue short (25|is) oscilloscope
P-7 ZnS:Ag + blue medium(57psec) radar
ZnCdS2:Cu + yellow long (400 ms)
P-10 KChVci dark very long electrochrome
trace
P-ll ZnS:Ag blue med short oscilloscope
(34ps)
P-12 (Zn,Mg)F2:Mn orange long (210 msec) radar
P-13 MgSiO3:Mn red- med long military radar
orange (52ms)
P-14 ZnS:Ag + blue short(27(isec) + radar
ZnCdS 2 :Cu + orange very long
(500 ms)
P-15 ZnO:Zn° blue- very short (2.8 flying spot scan
green ps)
P-16 (Ca,Mg)SiO3:Ce purple very short(0.12p flying spot scan
s)
P-17 ZnS:Ag + blue- short (5.2nsec) radar
ZnCdS2^u white + + very long
yellow (429 ms)
508 Design of phosphors

TABLE 6-2fContinuedl
CHARACTERISTICS OF JEDEC PHOSPHORS

COMPOSITION COLOR PERSISTENCE APPLICATION

P-18 (Ca,Mg)SiO3:Ti white medium monochrome
+ CaSiO3:Pb long(13msee) television
P-19 KMgF3:Mn orange long(220 msec) radar
P-20 ZnCdS2:Ag yellow med short(0.3 ms) storage tubes
P-21 MgF2:Mn orange long (220 msec) radar
P-22 ZnS:Ag + blue + green m(200 usec) color T.V.
ZnCdS2:Ag + + red medium short
Y2O3:Eu
P-23 ZnS:Ag + blue+ yellow medium short sepiatone TV
ZnCdS:Ag (200 usec)
P-24 CaS:Ce green short (1.5 us) flying spot
P-25 CaSiO3:Pb:Mn orange medium (60 psec) radar
P-27 Zn 3 (PO 4 )2:Mn red medium (27 ms) color TV
P-28 (Zn,Cd)S:Ag:Cu green very long (500 ms) radar
P-31 ZnCdS2:Cu:Ni green short (40 usec) high brightness
monitor
P-32 CaMgSiO3:Ti + purple-blue long (800 msec) radar
(Zn,Cd)S:Cu
P-33 MgKF2:Mn orange very long (3.0 sec) radar
P-34 ZnS:Pb:Cu blue-green very long(50 sec) oscilloscope
P-35 Zn(S,Se):Ag blue-white short (0.85 usec) photography
P-36 (Zn,Cd)S:Ag:Ni yellow-green very short(10 us) flying spot
P-37 ZnS:Ag:Ni green-blue very short flying spot
P-38 MgZnF2:Mn orange long ( 1 - 4 sec) radar
P-39 Zn2SiO4:Mn:As green med. long(400ms) radar
P-41 (Zn,Mg)F2:Mn+ orange- yellow long (100 msec) light pens, etc
Ca2MgSi2O7:Ce
P-42 ZnCdS:Ag + blue + green short(27nsec) + penetration
ZnCdS2:Cu very long (500 ms) screen
P-43 Gd2O2S:Tb green medium(20 msec) displays
 6.6 Commercial phosphors 509

TABLE 6-2fContinued)
CHARACTERISTICS OF JEDEC PHOSPHORS
COMPOSITION COLOR PERSISTENCE APPLICATION
P-44 La2O2S:Tb green medium displays
(20 msec)
P-45 Y2O2S:Tb green short (1.8 msec) displays
P-46 Y3Al5Oli:Ce green white very short( 120n) flying spot
P-47 Y2SiO5:Ce purple blue very short flying spot
(82 nsec)
P-48 YgAlgOj^Ce + yellowish very short flying spot
Y 2 SiO 5 :Ce green (120 nsec) scanners
P-49 YVO4:Eu red medium graphic
display
P-50 Y2O3:Eu red medium graphic
display
P-51 Y 2 O 3 :Eu + reddish medium Multicolor
(Zn,Cd)S:Ag:Ni orange displays
P-52 Zn,SiO 4 :Ti purplish short (28 msec) military
P-53 Y 3 Al 5 O u :Tb green medium Heads-up
display
P-54 Y2O,:Eu red medium displays
P-55 ZnS:Ag + blue + green medium projection
Y 2 O 2 S:Tb + + red television
(Zn,Cd)S:Cu:Al
P-56 Y2O3=Eu red medium projection
tube
P-57 Zn 2 SiO 4 :Mn + yellowish medium radar
MgF,,:Mn green

It should be noted that these JEDEC registrations, begun in 1945 by EIA

in the U.S., i.e.- Electronic Industries Association, have been largely
supplanted by WTDS designations. The various "P" registrations can be
grouped as shown in 6.6.6., given on the next page.
Note that most of these phosphors are based on the zinc and cadmium
sulfides. But there are some specialty phosphors based on silicates and
510 Design of phosphors

6.6.6.- Classification of JEDEC Phosphor Applications

Electrochrome display = 1 Oscilloscope display = 6

Radar display = 1 6 Television display= 6
Flying spot scanner = 8 Display monitor = 7
Storage monitor = 1 Photography display = 1
Light pen on display = 1 "Penetration" display = 1

P-6, 8, 9, 26, 29, 30 & P-40 are either obsolete or were withdrawn.

fluorides. The P-22 phosphors are those presently used for color
television. Many of the phosphor screens used as display monitors for
computers are the P-4 variety (white), or various silver-activated sulfides
containing Zn/Cd ratios to produce green or yellow emission. Many of the
JEDEC phosphors are not being used at this time because the newer rare
earth activated phosphors have been proven superior to those based on
zinc and cadmium sulfides. Table 6-3 shows a comparison of sulfide and
rare earth activated phosphors as to their efficiency and performance in
cathode-ray tubes.
Table 6-3
Comparison of Cathode-Ray & Rare Earth Phosphors in CRT's
Composition Energy Peak Color Brightness Relative
Eff. Wavelength Loss Strength
Zn2SiO4:Mn 8% 525 nm green ~ 35 % W
CaWO4:Pb 4 425 blue - 30 % VW
Y2O3:Eu 9% 611 nm red 3% S
Y2O2S:Eu 12 626 red 12% S
Gd2O2S:Tb 15 545 Y-green 5% S
ZnS:Cu-.Al 23 530 green 37 % VW
ZnS:Ag:Cl 21 450 blue 31% VW
Y2SiO5:Tb 1.5 545 Green 1.5 % VW
YgALA^Tb 2 545 green 2.1 % VS
Y3Al3Ga2O1j:Tb 1.5 545 green 1.7% VS
CaS:Ce 22 520 green 30% VW
LaOBr:Tb 20 544 nm green
 6.6 Commercial phosphors 511

Many of these measurements were made in conjunction with use in

projection CRT systems. The projection television system uses three
separate cathode-ray tubes, each containing a red or green or blue
phosphor screen. Each CRT must be operated at its maximum output over
a long period of time since the brightness output of the three are
combined and projected upon a much larger viewing screen. Because the
loss in brightness from the Schmidt-lens to the viewing screen is several
orders of magnitude, any loss in screen brightness in the projection tube
during operation is magnified in the viewing screen. The brightness loss
given in Table 6-3 was accomplished using a standard electron beam
current and voltage. Since many of these phosphors discolor, that is- they
develop a dark cast or coating on the surface of the phosphor particles,
this factor must be considered. It was arbitrarily defined to give a
subjective "relative strength", applicable to each phosphor. As we can see,
this factor shows how each phosphor screen maintains its relative
brightness during operation of a CRT.

II. Fluorescent Lamp Phosphors

The fluorescent lamp is basically a low pressure mercury discharge lamp

with a layer of phosphor particles on the inside surface of the glass tube,
as shown in 6.6.6., presented on the next page.

The lamp has an internal gas-fill pressure of about 1 -2 mm pressure and

contains a mixture of Ne, A, and Kr gases. The internal radiation
generated is mostly (85%) 2537 A , which is the resonance wavelength of
the mercury vapor discharge. The other 15% is distributed between 1850
A, 3150 A, 3650 A, 4300 A, 5460 A, and 5785 A. The general phosphor
system used has the apatite structure with the general composition:

Cas F,C1(PO4)3 :Sb:Mn

where Sb 3+ is the sensitizer and Mn 2+ is the activator. This system is

used because the emission colors can be changed from bluish-white to
reddish-white by variation in the Sb/Mn ratio and content. The phosphor
coating is always a blend of phosphors with the major part composed of
512 Design of phosphors

6.6.6.-

the apatite. In the operation of the lamp, the ballast controls the amount
of operating current (it is essentially a "choke-coil") while the rapid-start
switch starts the filaments. When the filaments glow so as to ionize the
mercury at the electrodes, the voltage is switched, thereby initiating the
discharge. Fluorescent lamps normally were manufactured in two sizes, a
48 inch lamp (40 watts) and a 96 inch lamp (80 watts). We will describe,
in the next chapter, the latest changes taking place as a result of lamp and
phosphor improvement.

Another type of lamp is the so-called high pressure mercury vapor

(HPMV) lamp. If one increases the internal pressure of the lamp from a
few mm. to several atmospheres, i.e.- » 760 mm. pressure, the number
of collisions per second between mercury atoms increases enormously in
the discharge. This has the effect of shifting the resonance radiation to
longer wavelengths, including the visible wavelengths: 4300 A, 5460 A,
and 5785 A., resulting in a greenish-white emitting lamp. There is also a
 6.6 Commercial phosphors 513

considerable amount of 3650 A (ultraviolet) radiation present. In order to

achieve the required internal pressure needed for operation of this lamp,
a quartz envelope is employed, as shown in the following:
6.6.5.-

The HPMV lamp has been used extensively in street-lighting applications.

Electrodes are sealed to each end of the quartz tubing to form a special
pressed seal, thereby forming the lamp. Mercury plus a rare gas mixture
is introduced into the evacuated tube at the "seal-off point at a
predetermined pressure. A phosphor coating is then applied to the inner
side of the glass envelope, the quartz lamp is then mounted within and
the whole is sealed with end-caps. This type of lamp runs very hot,
typically between 700 °C. and 800 °C. at the outer surface of the quartz
lamp. The reason for this is that a high temperature is required within
the quartz tube to get the internal pressure up to the necessary operating
conditions of several atmospheres of internal pressure. Therefore, this
type of lamp dissipates about 400 watts of electrical power. However, the
temperature at the surface of the glass envelope is never more than about
250-300 °C. Because the discharge emission is a greenish-white, a red
phosphor is needed to correct for the lack of red emission in the
discharge radiation. In addition to being red, the phosphor must also have
good temperature dependence of emission and respond to 3650 A.
514 Design of phosphors

Typical lamp phosphors are shown in the following table:

TABLE 6-4
COMMERCIAL LAMP PHOSPHORS
Composition Emission Excitation Q.E.
Color Peak Half-width Peak (2537 A)
Ba2P2O7:Ti blue-green 4940 A 1700 A 2590 A 85%
Sr 2 P2O 7 :Sn blue 4600 1050 2580 86
(Ca,Zn)3(PO4)2:Sn pink 6100 1360 2580 82
(Sr,Mg) 3 (PO 4 )2:Snpink 6280 1250 2590 84
MgGa2O4:Mn green 5040 3 0 0 2700 78
(Sr,Mg)2P2O7:Eu U.V. 3930 245 3250 ...
Sr 3 (PO 4 )2: Eu deep blue 4060 340 3210 ...
Ca5F(PO4)3:Sb blue 4760 1360 2550 71
Sr5F(PO4)3:Sb blue-green 5090 1530 2560 76
CaSiO3:Pb:Mn red 6150 940 2600 85
BaSi2O5:Pb U.V. 3510 410 2600 75
SrFBO2:Eu U.V. 3710 190 3380 ...
Zn2SiO4:Mn green 5280 410 2580 70
MgWO4:W blue 4730 1360 2880 83
Cd(BO2)2:Mn red 6180 750 2740 78
YVO4:Eu red 6200 40 3290 89
YVO4:Dy yellow 5750 60 3290 90
CaWO4:W blue 4330 1140 2540 75
Y2O3:Eu red 6195 30 2600 92
Cas(F,Cl)(PO4), :Sb:Mn various varies 2600 85
CaSiO3:Pb UV 3300 560 2500 75
CaWO4:Pb deep blue 4330 1140 2600 76
Mg4FGeO6:Mn deep red (5 lines) 6575 163 4200-3000-2200 82
Mg5AS2On:Mn deep red (5 lines) 6400 163 4400-3300-2300 80
Sr 5 Cl(PO 4 ) 3 : Eu blue 4450 375 3000-2580-2300 91
BaMg2Ali6O27: Eu blue 4540 600 4 peaks-2600 8 8
SrMg2Ali8O39: Eu blue 4560 640 4 peaks-2600 100?
Mg3AlnOig: Ce:Tb green 5520 120 2800 92
BaMg2Ali6O27: Eu:Mn blue-green 5150-4250 1300 4 peaks-2600 100?
 6.6 Commercial phosphors 515

Excitation bands are broad so that most of the incident 2537 A photons
generated within the lamp are absorbed. The electrical efficiency of the
LPMV lamp for generating 2537 A photons is only about 32%. Thus with a
phosphor having a QE of 80%, a LPMV lamp would have a 22% energy
conversion efficiency for generation of visible light. However, this is
considerably better than its competitor, the incandescent lamp, whose
energy efficiency is slightly less than 8%.

There are a number of specialty phosphors that have been used to make
LPMV lamps for special purposes. These are listed in Table 6-5:

TABLE 6-5
COMMERCIAL SPECIALTY LAMP PHOSPHORS
Composition Emission Excitation Q.E.
Color Peak Half-width Peak (2537 A)
BaSi2O5:Pb UV 3510 A 410 A 2540 A 76%
(Ba,Sr)2(Mg,Zn)Si2O7:Pb UV 3560 1000 2537 78
Y3Al5On:Ce "white" 5000 very broad 3500-2600 82
Sr2P2O7:Eu deep blue 4200 290 2600 88
YMg2AlUO19: CeUV 3400 465 2600 70
MgGa2O4: Mn green 5040 120 2760 82

These phosphors are used primarily in "Black-Light" lamps and photocopy

machines such as those marketed by Xerox Corp.

HPMV lamp phosphors include YVO4:Eu , YVO^Dy and (Sr,Mg)3(PO4)2:Sn.

A better HPMV phosphor is the Y(Po.2oVo.8O)C>4:Eu composition which has
a superior temperature dependence of emission. Although Y2O3:Eu also
has a good temperature dependence of emission, it does not respond to
3650 A radiation. Before we leave this section, we should point out that
new improvements have been made in both HPMV and LPMV lamps. For
example, in HPMV lamps, the mercury vapor has been augmented with
other metal vapors such as Tl, In and others. Advantage is taken of the
fact that the iodides of these metals are volatile so that the metal-vapor
emission spectra, containing a relatively high content of red radiation,
516 Design of phosphors

results. This obviates the use of phosphors for most cases. The lamps are
called metal-vapor lamps, even though they still rely on the mercury vapor
emission for the major part of the visible light emitted.

For LPMV lamps, the newest lamps contain three (3) essentially "line-
emitters, using Eu 2+ for blue emission, Tb 3 + for green emission, and Eu 3+
for red emission. This is the so-called "high output lamp" which has line
emission in contrast to the broad band emission of the phosphors given
in Table 6-4. That is, the light output has large gaps in the wavelengths
emitted. However, the human eye integrates the wavelengths so that the
emission is perceived as "white" light. Note that the last four phosphors
presented in Table 6-4 are based upon the magnetoplumbite structure,
i.e.- aluminate compounds, which are related to the spinels. It is these
phosphors which have caused a revolution in the manufacture of LPMV
lamps. Such phosphors can be used at much higher current densities
within the fluorescent lamp and are more efficient than the older lamp
phosphors listed above. The current type of lamp is a 1.0" lamp, i.e.-
33T5, where the first number is the watts of energy dissipation for a four
foot lamp. The old lamp, which has nearly been replaced, is a 40T12
lamp. We will address the changes which have taken place in current
lamp manufacture since 1991 (the date of the prior edition of this
manuscript) in the next chapter.

It should be noted that all of the above phosphors were described in a

companion book, "The Chemistry of Artificial Lighting Devices- Lamps,
Phosphors and Cathode-Ray Tubes", published by Elsevier in 1993 (ISBN
0-444-81709-3). In this volume, the exact methods for manufacture of
these phosphors were presented along with the exact formulas and
materials required to do so.

6-7-MEASUREMENT OF OPTICAL PROPERTIES OF PHOSPHORS

Although methods of preparing several phosphor compositions have been

explored, we have not yet examined methods of measuring phosphor
properties. The need to measure the "brightness" or light output under a
controlled excitation source should be apparent if one is to optimize any
 6.7 Measurement of optical properties of phosphors 517

given phosphor composition. Other measurements needed are spectral

energy distributions, color measurements and particle size. All of these
are necessary before a lamp test is made to determine the stability of a
test phosphor in an actual LPMV or HPMV lamp.

I. Measurement of Phosphor Brightness

Let us suppose that we have prepared a series of phosphor compositions

wherein the activator concentration was the variable. We would need to
measure these samples to determine which concentration was near to the
optimum. Let us further suppose that we wish to use our phosphor in a
LPMV lamp application. Although we could manufacture a test-lamp,
usually this is not very feasible. What we do is to prepare a thick layer on a
holder and compare it to a "standard" phosphor having similar emission
characteristics. The layer is called a "plaque" and the apparatus is called a
"plaque-tester", as shown in the following:

6.7.1.-

Essentially it is a light-tight box equipped with ultraviolet lamps, having a

suitable wavelength for excitation of the phosphors being tested. For the
most part, we use lamps with 2537A output, although others can be used,
as required. The plaque-tester also has a suitable detector to measure the
intensity of the light emitted by the phosphor, a collimator to eliminate
any "stray" radiation, and has a U.V. filter to prevent the excitation
518 Design of phosphors

wavelengths from reaching the detector. Generally, an ordinary window

pane will serve.

The detector can be one of several, including a photomultiplier tube

(which is that illustrated above). The photomultiplier tube has the
disadvantage of requiring a high voltage supply to be operative. The main
requirement for the detector is that it must have the proper response to
the wavelengths being measured. Other usable detectors include a GaAsAl
diode or a barrier-layer cell. In addition to the apparatus shown above, we
also need a regulated voltage supply so as to regulate the intensity of the
2537 A illumination of the individual plaques. Note that this is just one of
many possible configurations. However, it does show the required
components needed to measure relative brightness of a series of
phosphors. The next parameter we need to establish is the emission
spectrum of any given phosphor.

II. Measurement of Spectral Energy Distributions

One of the more critical measurements required for phosphors is

determining the spectral distribution of the emission and/or that of
excitation, that is- the wavelengths which best excite the phosphor. Such
an instrument is called a Spectrofluorimeter and has the general optical
design, as shown in 6.7.2., presented on the next page.

Two (2) monochromators are required, an excitation monochromator and

an emission monochromator. The source of excitation is generally a
xenon lamp since it provides useful radiation from about 1750 A to beyond
10,000 A at intensities not obtainable from other types of sources. To
obtain an emission spectrum, we set the excitation monochromator at a
wavelength which provides good response from the phosphor sample.
This can be done visually. We then operate the emission monochromator
to obtain the spectrum of the emitted light, taking care that the switch is
set for the correct side of the input from the operating monochromator.
Having done this, we next set the emission monochromator at the
wavelength of maximum emission of the phosphor, and operate the
 6.7 Measurement of optical properties of phosphors 519

6.7.2.-

excitation monochromator to obtain the excitation spectrum of the

phosphor.

In this optical design, it is difficult to obtain an absorption spectrum, the

best that we can do is to operate both monoehromators simultaneously so
as to obtain a reflectance spectrum. Even then, we need to compare the
reflectance signal of the phosphor to that of optical-grade BaSO4. Barium
sulfate has a reflectance of 98.99% throughout the visible and ultraviolet
portions of the spectrum and has the added advantage that it does not
degrade under U.V. radiation. In our case, we must first obtain the
reflectance of BaSO4 , and without changing instrumental setting, measure
our phosphor. We will find that the instrument does have energy losses
within the monoehromators and detector (they are not linear in energy
throughput). It is these losses, detected by BaSO4 , that we must apply to
our phosphor reflectance spectrum to obtain the true spectrum.

There are commercial instruments which measure excitation and

emission spectra directly in pwatts/A bandwidth, thereby obviating the
necessity of measuring absorption spectra. The quantity we wish to obtain
is the total energy emitted (so as to be able to compare separate
phosphors), namely:
520 Design of phosphors

6.7.3.- E = hv = he/ X = k I I dX

With the commercial instruments, we can measure this value directly, if

we have an energy integrator built into the instrument. There are other
instruments which automatically correct for instrumental energy losses.

III. Measurement of Quantum Efficiencies

Once we have obtained the measured values discussed above for a given
phosphor, it is relatively easy to calculate the QE of our phosphor. Here
we use S = standard and U = unknown, namely-

6.7.4.- (QE)u = fQEtef fY.E dk^jtaHnnls M YE dX^iec^lTT Iffl - Rial

{ (IE dXexCitation)u }{ I E demission) S) }{U - R)u}

Note that we have assumed that EdXexCitation is equal to absorption. This

is usually true, but not always. Sometimes, absorption does not cause
excitation. For this reason, we measure the reflectance spectrum in the
region of the excitation band for comparison and possible correction of
QE values of the phosphors. Do not confuse the (1 - R) values. What we
really need is the integrated values of reflectance over the entire range of
the absorption band. But this is extremely difficult to measure. Therefore,
we use (1- R) values. One other comment is needed. What we have
specified in 6.7.4. is a single wavelength in R and the peak wavelength of
excitation and emission, namely:

6-7.5.- (YE dXexcitationiu Hfl - R)n_h (emission) = set value

I E dX e m j S S i o n )jj }x (excitation) = set value

What we do not know, from the spectral data that we have, is how X
affects the QE as a function of R, the reflectance.

IV. Specification and Measurement of Color

An important measurement for specifying the properties of phosphors is

 6.7 Measurement of optical properties of phosphors 521

that of color. We find that even though our vision is sufficient to

categorize color as a gross feature, we still need to be able to measure
small differences of color. One method, used in the past, was to assemble
a series of samples having small but observable differences in color, and
then compare an unknown to these. Because the human eye is a superb
color instrument, this approach was perfectly feasible.

I. The Human Eve

Let us examine the human eye as a prelude to setting up a color-

specification method. The optical parts of the human eye are:

6.7.6.-

The cornea is the outer entrance to the eye. The iris is controlled by
muscles so as to adjust the pupil of the eye according to ambient light
intensity. The lens focuses light upon the retina and its shape can be
somewhat altered by musculature control. The retina consists basically of
a mosaic of light-sensitive cells, each operating on a photo-chemical
principle. There are two (2) types present: the "rods" and the "cones".
The basic difference between the two is that cones are sensitive to color
whereas rods have little color response and are operative primarily when
522 Design of phosphors

a very low light intensity is present, i.e.- night-vision. When the

luminance of the visual field is high, vision is said to "photopic" and the
eye is "light-adapted". A low luminance condition results in "scotopic"
vision, the eye being "dark-adapted". The range of luminance discernible
by the human eye is about 50 :1. There are many other factors which
contribute to human sight, such as binocular vision, but we are only
interested in color perception at the moment.

The mechanism of color vision is fairly well understood. Basically, it

involves three (3) types of chromophores (each having a unique molecular
structure) which react with a photon to produce an electrical signal,
detectable by the optic nerve. We will not delve into this chemistry. It has
been demonstrated that three separate types of cones exist, giving three
sets of color preceptors. Each set of preceptors is active over a different
range of wavelengths, these ranges being centered in the blue, green and
red regions of the visible spectrum. Thus, we speak of having three
primary colors, red, green, and blue.

But the retina does not respond equally to all wavelengths. For equal
energies, a yellow-green light produces a much stronger response in the
human eye than a red or blue light. Thus, we say that the yellow-green
light is "brighter" than the red or blue lights. This is called the luminosity
response of the eye. By measuring a number of individual observers, we
can obtain what we call a "Standard Luminosity Curve". Apparently,
photopic vision relates to "sunlight", to which the human had adapted
through evolution, while scotopic vision related to "moonlight", that is,
sunlight modified by reflection from the Moon's surface.

Photopic vision peaks at 5500 A whereas scotopic vision peaks at 5200 A.

The following, given as 6.7.7. on the next page, shows both the photopic
and scotopic response curves for the human eye, as determined from a
number of observers. In this case, the relative response of the observers
are summed into a response called "THE STANDARD OBSERVER" and is
normalized for easier usage. You will note that these eye-response curves
are the result of an average of many human eye response curves.
 6.7 Measurement of optical properties of phosphors 523

6.7.7.-

Now, let us examine the effects of colors as perceived by the human eye.
b. The Nature of Chroma

The visible spectrum extends from about 4000 A to 7000 A. We find that
the eye acts as an integrating instrument. Thus, two colors may appear
equal to the eye even though one is monochromatic light and the other
has aband of wavelengths. This is shown in 6.7.8., given on the next page.
In this case, we may see the same color, but the photon energies are
much different.

It was Newton, using a glass prism plus slits, who first demonstrated that
sunlight consisted of colors or chroma. However, the idea of
monochromatic light compared to a band of color, i.e.- an assembly of
lines, completely escaped the notice of investigators until the 20th
century. Subsequent work then showed that colors could be duplicated by
mixing the three primaries, red, green and blue to obtain the various
chroma, including shades of "white". Actually, these shades involved the
luminous intensity, that is- the amount of light falling upon a surface.
524 Design of phosphors

6.7.8.-

Let us first define some of the terms of measurement of luminosity and

then proceed to determine how to measure the chroma, or chromaticity.

Consider a thin flat plate which has no absorption, only reflectance and
transmittance. We will find that light falling upon its surface with a certain
intensity has a luminance, L, while the light transmitted is defined as H,
the exittance, or emittance. H will equal L if no absorption takes place or
if there is no scattering at the surface of the thin plate. However, even at
atomic distances, a certain amount of scattering does take place. We find
that two types of scattering are possible. If the surface is perfectly smooth
on an atomic level, then a light wave would be back- scattered along the
same exact path, and we would have a perfect diffuser, which we call S o .
However, there is always an angle associated with the scattering, which
we call S a (where a is defined as the scattering angle), and we have an
imperfect diffuser, as shown in 6.7.9., presented on the next page.

If we view the thin plate from the left where it is illuminated with
intensity, L, what we see is the scattered light, or light diffusion from the
surface. If the plate is a perfect diffuser, then we will see the exact
amount of L scattered back along the same plane as a diffuse component.
Note that we are not speaking of reflection (which is an entirely different
mechanism where the wavelength of the light is affected) but of scattering
(where the light is absorbed, then reemitted at the same wavelength). For
scattering by a perfect diffuser, LQ = I o /S o . However, this is never the
case.
 6.7 Measurement of optical properties of phosphors 525

6.7.9.-

What we find is that there is an angle dependence of scattering, and that:

6.7.10.- 1= I a /S a =Ia/(Socosa)

where a is the angle of scattering. Therefore, H does not equal L. If we

define $ as the flux of light, i.e.- the number of photons incident per
second, we find:

6.7.11.- H = <>j / S a = <>j / S o cos a where: 4> = 47t I (point source)

<>
( = 7TI (flat surface)

The intensity units for <>( are related to a primary radiation source, the
candela, Cd. The definition of a candela is:

6.7.12.- CANDELA: a unit of luminous intensity, defined as 1/60 of the

luminous intensity per square centimeter of a black-body radiator
operating at the temperature of freezing platinum (1772 °C), Formerly
known as a candle.

This gives us the following intensity units, as shown in 6.7.13., given on

the next page. Thus, we have two units of measurement of intensity. One
is related to scattering from a surface, L, i.e.- in foot-lamberts and the
other is related to emittance, H, i.e.- in lumens per square foot. It is well
to note the differences between these two units. Many times, they are
confused with one another and not used correctly.
526 Design of phosphors

6.7.13.- INTENSITY UNITS

1.0 Cd (at a one-foot distance) = L In (foot-lamberts)

= H/jr (lumens)

Although we have assumed "white" light up to now, either of these two can
be wavelength dependent. If either is wavelength dependent, then we
have a pigment (reflective- but more properly scattering) with intensity in
foot-lamberts, or a phosphor (emittance) with intensity in lumens.

In terms of color, we have additive processes (emittance) and subtractive

processes (reflectance). If we wish to match colors, the primary colors
are quite different, namely -

6.7.14.- ADDITIVE PRIMARIES SUBTRACTIVE PRIMARIES

red magenta
green yellow

blue cyan

Let us now consider how to set up a color matching system,

c. The Standard Observer

Since color matching is meant for humans, it is natural to define color in
terms of an average, or "Standard Observer". Our first step is to build an
instrument which contains three colored lamp sources, a place for the
observer, intensity detectors, and a monochromator. One design is shown
in 6.7.15., given on the next page.

There are two (2) sources of light to be compared. One is from a set of
three lamps whose emission is modified by means of suitable filters to give
a red beam , a green beam and a blue beam. These are mixed at the
screen to form a single spot (although we have not illustrated it in that
way, so as to be more discernible). The other source comes from a
monochromator so as to obtain a monochromatic beam of light.
 6.7 Measurement of optical properties of phosphors 527

6.7.15.-

There are controls to adjust the individual beams of red, green and blue
light, as well as that of the monochromatic beam. In this way, the mixed
beams of light can be directly compared to the monochromatic spot. In
the back of the screen are detectors to measure the energy intensity of
the beams of light being compared.

We need about 5000 observers to obtain a satisfactory average, both for the
dark-adapted and the light-adapted human eye. Note that we can compare
any color in terms of red + green + blue to a monochromatic color.
There are three (3) things that we need to accomplish:
528 Design of phosphors

6.7.16.- 1. Define eye response in terms of color at equal energy

2. Define shades of "white" in terms of % red, % green and %
blue, at equal energies of those "whites".
3. Define "color" in terms of % red, % green and % blue, as
compared to monochromatic radiation.

The difficulty in setting up the initial system for color comparisons cannot
be underestimated. The problem was enormous. Questions as to the
suitability of various lamp sources, the nature of the filters to be used, and
the exact nature of the primary colors to be defined occupied many years
before the first attempts to specify color in terms of the Standard
Observer were started. But before we further delve into these problems,
let us first establish a better background into the factors controlling this
investigation. Some of this work we have already discussed. It is
presented again to create an integrated approach to the problem of color
specification.

It had long been known that an incandescent solid emits electromagnetic

radiation. Work by Planck (1889) on the concept of a "black-body" led to
Planck's Law: E = h y , which was the beginning of the quantum theory.
The "black-body" is a theoretical concept of a radiating body which
absorbs all radiation incident upon it, but reemits that radiation according
to certain laws, including its absolute temperature. Experimentally, we
approximate the black-body by a blackened sphere with a small hole in it
for the internal radiation to emerge. When the sphere is heated, the
energy emitted can be calculated from:

6.7.17.- I (X) = 3.703 x 10"5 X dX I [exp {(1.432 / XT) -1}]

which reduces to : Total radiation = 5.73 x 1 0 5 T 4 ergs/sec/cm 2 .

where T is in °K, and X is the spectrum of the black-body emission. We

therefore speak of the "color-temperature" of an emitting source
(including lamps and phosphors) as related to a black-body at that same
temperature. The wavelength distribution of a black-body at several
temperatures is shown in 6.7.. 18., given on the next page.
 6.7 Measurement of optical properties of phosphors 529

6.7.18.-

It is easily seen that temperatures around 6500 °K. are necessary to bring
the peak maximum into the middle of the visible spectrum. Although the
Sun is a black-body radiator of about 10,000 °K (as viewed directly from
space), scattering and reflection within the Earth's atmosphere is
sufficient to lower the effective black-body radiation perceived to 6500 °K.
Thus, the Sun is a 6500 °K. source which we call "daylight". The direct
viewed brightness of the Sun at the Earth's surface is about 165,000
candela/cm 2 , that of the Moon - 0.25 candela/cm 2 , and a clear sky is
about 0.8 candela/cm 2 . For these reasons, DAYLIGHT has been defined as:
"The northern skylight at 11:30 am. at Greenwich, England on October
31, 1931". This is also the definition of ILLUMINANT- C. The other
standard illuminant that we use is ILLUMINANT- A, which is the radiation
emitted from an incandescent tungsten filament operating at 3250 °K.
The spectra of these sources are shown in the following diagram, given as
6.7.19. on the next page.

Referring back to our Color Comparator of 6.7.15., we use these concepts

to calibrate our lamps in terms of spectra and relative energy in terms of
these standard sources. ILLUMINANT - B , by the way, was originally
defined as "average sunlight" but it was soon determined that "average" is
not the same at all parts of the Earth's globe.
530 Design of phosphors

6.7.19.-

Our next step in using the Color Comparator is to set up proper filters so
as to obtain and use "primary" color lamp sources. We find that by using
the monochromator of the Color Comparator, we can approximate the
wavelength response of our so-called "Standard Observer", in the red
region, the green region and the blue region of the visible spectrum.

But we also find that we need a band of wavelengths for each color, since
a monochromatic beam is not at all suitable. This is where the choice
becomes subjective, since we are relying upon the spoken response of
individuals. We find that by choosing a blue filter peaking at 4400 A, a
green filter peaking at 5200 A, and a red filter peaking at 6200 A (but
having a lesser peak in the blue), we have a "Blue"-blue, a "Green"-green
and a "Red"-red which will satisfy most observers. Note that the original
single color was chosen by approximating the human eye response in the
three-color regions, using monochromatic light to obtain a brightness
response, i.e.- 6.7.7., and then adjusting the broad-band transmission
properties of the three filters of the lamps to obtain the proper colors. A
final check of these lamp filters would be to mix all three colors
additively, and then to evaluate the "white" thereby produced. We can
then substitute a Standard Lamp for the monochromator and see if we can
 6.7 Measurement of optical properties of phosphors 531

reproduce its exact color temperature. If not, then we need to modify the
transmission characteristics of our filters used on the source lamps.

Once we have done this, we now have our three primary colors in the
form of standard lamps, and can proceed to determine Items 1,2 & 3,
given in 6.7.16. To do this, we vary the wavelength of the monochromatic
light, and determine relative amounts of red, green and blue light
required to match the monochromatic color. This is done, as stated
before, for about 5000 observers. The result is finalized response curves
for the Standard Observer, also called "Tristimulus Response curves".

6.7.20.-

We finally arrive at the result we want, since we can now set up

"Tristimulus Filters" to use in defining colors. We can now define ¥ as our
standard luminosity curve for the human eye (photopic vision). Note that
~x , the red tristimulus value, has a certain amount of blue in it in order to
duplicate the response of the red preceptor in the retina. Note that these
colors are the result of the "Standard Observer" measurements that we
started in the first place.
532 Design of phosphors

d.- Color Measurement

Our next step is to define colors in terms of tristimulus responses. We

know that we can define the energy of any spectral curve as a summation
of intensities times wavelengths, i.e.-

6.7.21.- ER = I (IdX)R EG = I (IdX)G EB = I (IdX)B

Therefore if we take the spectral curve, and multiply it by the overlap of

each tristimulus response curve, we get TRISTIMULUS VALUES, i.e.-

6.7.22.- Emittance Reflective (scattering)

(additive) (subtractive)

X = x I IR dX X =x £ IR RR dk

Y = ? i i G dx Y = "y i i G RG dx

Z = z IIBdX Z =z" £ I B R B dX

However, we find that these values are difficult to use since each color
gives a set of tristimulus values, but each set does not have a specific
relation to any other. The reason for this is that the intensity of IR * IG *
IB . Therefore, we define a set of chromaticity coordinates where:

6.7.23.- CHROMATICITY COORDINATES:

x + y + z = 1.00

x = X/X+Y+Z y=Y/X+Y+Z z = Z/X+Y+Z

Note that we now have three (3) different sets of values related to color
specification, and three different types of symbols.

In order not to get confused, we reiterate them again. They are shown in
6.7.24, presented on the next page.
 6.7 Measurement of optical properties of phosphors 533

6.7.24.- VALUES RELATED TO COLOR SPECIFICATION

Tristimulus
Color Response Values Chromaticitv Coordinates
red: x X x
green: U Y y
blue: z" Z z

The advantage of chromaticity coordinates is that we now have a set of

normalized values which we can use to compare colors having different
intensity values (and thus different energy values as well). Furthermore,
we need only specify x and y since x + y + z = 1.00. This allows us to
specify monochromatic radiation in terms of our chromaticity
coordinates. Since monochromatic radiation is a boundary of color-
mixing, then we can construct a CHROMATICITY COORDINATE DIAGRAM
in terms of x and y , namely:
6.7.24.-
534 Design of phosphors

The above is known as the C.I.E. chromaticity diagram (Commission

Internationale de l'Eclairage). First we note that the boundary of this x
and y diagram is bounded, as we have already stated, by the values of
monochromatic light. Thus, we can find any color, be it monochromatic
or not, in terms of its x and y coordinates. It would be more dramatic to
print the various hue areas in color but it is difficult to accurately print
reflectance hues. It is better to name the colors directly. This was done by
Kelly (1940). Note also that we do not use the term "color" anymore but
use the term "hue".

Any hue can be specified by x and y. For example, we can specify the locus
of black-body hues and even Illuminants A, B & C, namely-

6.7.25.- x and v Chromaticitv Coordinates

6000 A 0.640 0.372

5200 0.080 0.850
4800 0.140 0.150

ILLUMINANT A (3250 A) 0.420 0.395

ILLUMINANT B (4500 A) 0.360 0.360
ILLUMINANT C (6500 A) 0.315 0.320

These are shown in the following diagram, given as 6.7.26. on the next
page. Note that all of these are emitters, similar to that of a phosphor.

Let us now summarize the results we have achieved. We have measured

the luminosity response of the human eye, in terms of photopic and
scotopic behavior. We also defined a "black-body" and its wavelength
emission, stipulating its absolute temperature. We then defined Standard
Sources. We next designed a Color Comparator and then determined the
transmission characteristics of three (3) filters required to duplicate the
response of the three color preceptors of the human eye. These we called
the tristimulus response of the Standard Observer.
 6.7 Measurement of optical properties of phosphors 535

6.7.26.-

Finally, we obtained Chromaticity Coordinates which we could use to plot

various hue values. If we have a Color Comparator, then we can measure
any hue and compare it to any other. But, if we do not, we either build or
buy one, or resort to an instrumental method. The Color Comparator uses
a human observer to specify color. If we wish to use an instrumental
method, then we must correct each component of the instrument to the
chroma response of the Standard Observer. While the Color Comparator
given above in 6.7.15. was perfectly satisfactory for setting up a system of
chromaticity coordinates, it was difficult and awkward to use. What was
really required was an instrumental method of color measurement.

The requirements for an instrumental method of specifying color include

a light source, the colored object and a detector, arranged as follows in
6.7.27., shown on the next page.
536 Design of phosphors

6.7.27.- Components Required for an Instrument to Measure Color

Note that we must be able to define the output of the source in terms of
its wavelength. This includes any source, whether it is an incandescent
lamp or a discharge lamp. The output must be stable. We also must define
the reflectance of the object whose color we are trying to measure. This
mandates a detector whose response to wavelength has been
standardized. These are parameters of the instrument which are not
always recognized until one begins to develop an instrument to measure
color, particularly that of reflectance. While this measurement might
seem simple enough to perform, it is not.

Since the response characteristics of these optical components are not

linear, nor flat, we need an analogue system in order to be able to measure
color. The analogue system simply corrects for the non-linearity of the
source and detector, as shown in the following for emittance. This is
shown in 6.7.28., on the next page.

This diagram illustrates the optical analogue process in which the

properties of an instrument light source, a properly selected filter, and a
photo-detector are combined to provide an optical analogue of the similar
properties of a CIE standard illuminant (here ILLUMINANT C) and a CIE
standard observer response function (here the U response).
 6.7 Measurement of optical properties of phosphors 537

6.7.28.-

You might question why we need a light source to measure an emitter.

The light source acts as a standard for comparison, using the reflectance
from a standardized reflectance material so as to be able to use the
STANDARD OBSERVER response we have already developed for color
matching.

The corresponding optical analogue for reflectance is:

6.7.28.-

This diagram shows schematically how the spectral power distribution of

a CIE source, the spectral reflectance, R, of an object, and the spectral
color-matching functions, x, y and z, combine by multiplication (each
wavelength by each wavelength), followed by summation across the
spectrum, to give the CIE tristimulus values. However, these analogues are
538 Design of phosphors

actually hypothetical. The actual values we need to measure reflectance

are:
6.7.30.-

When we introduce the color response characteristics of the source,

combined with the detector of our instrument, we find that we must
drastically modify the transmission characteristics of our filters in order
to duplicate the CIE color matching functions for the equal-energy
spectrum. However, this is not an impossible task and we find that an
excellent match can be obtained to the transmission functions of 6.7.20.
This is typical for commercially available instruments. Now, we have an
instrument, called a Colorimeter, capable of measuring reflective color.

The process for measuring emissive color is somewhat different. First, we

obtain an emission spectrum by means of a spectrofluorimeter. We can
now integrate I dX to obtain the energy and then specify this in terms of x
and y (see 6.7.20. and 6.7.22.). However, it is nearly impossible to use
each and every wavelength present. Referring to 6.7.20., we can see that
if we draw vertical lines on each color function, the density of which are
spaced according to the peak height of each function, we would obtain
what we call "weighted functions". We would then have a set of lines
(wavelengths) useful for calculating chromaticity coordinates. That is, We
would have a set of lines whose spacing was inversely proportional to peak
height (the higher the intensity, the closer would be the spacing), where
the line spacing would have to be close between the wavelengths:

6.7.31.- x = 5500 - 6500 A y = 5100 - 6000 A

z = 4200 - 4800 A
 6.7 Measurement of optical properties of phosphors 539

A practical method for doing this is to weight the line heights by using
tristimulus values times the height of our experimental curve, as
determined by measurement at the specified wavelength, times the
energy distribution of a Standard Source such as ILLUMINANT C.

In the following, we show sixteen (16) weights, spaced 200 A apart, for all
three (3) tristimulus values, X, Y, Z. By multiplying the line heights of a
spectrum by these values and then summing them, we can obtain values of
X, Y& Z.

6.7.32.- STANDARD WEIGHTS FOR ILLUMINANT C

Wavelength X-Weight Y-Weight Z-Weight

4000 A 0.044 0.001 0.187

4200 2.926 0.085 14.064
4400 7.680 0.513 38.643
4600 6.633 1.383 38.087
4800 2,345 3.210 18.464
5000 0.069 6.884 5.725
5200 1.193 12.882 1.450
5400 5.588 18.268 0.365
5600 11.751 18.606 0.074
5800 16.801 15.989 0.026
6000 17.896 10.684 0.012
6200 14.031 6.264 0.003
6400 7.437 2.897 0.000
6600 2.728 1.003 0.000
7000 0.175 0.063 0.000
Sum 98.046 100.001 118.100

x= 0.3101 y= 0.3163

For example, the following, given as 6.7.33. on the next page, is a

spectrum of a Cool-White fluorescent lamp, with the required wavelengths
already marked off.
540 Design of phosphors

6.7.33.-

The following gives the standard weights to be applied for this spectrum:

6.7.34.- STANDARD WEIGHT CALCULATIONS FOR A COOL WHITE LAMP

Wavelength X-Weight Y-Weiffht Z-Weight
4000 A -0.042 -0.005 -0.233
4200 1.849 0.073 8.997
4400 6.032 0.361 30.107
4600 2.719 0.602 15.796
4800 1.107 1.485 8.162
5000 0.023 3.390 2.834
5200 0.495 6.855 0.813
5400 5.937 18.304 0.331
5600 14.018 23.752 0.111
5800 24.837 23.454 0.038
6000 23.501 14.167 0.017
6200 12.849 5.717 0.002
6400 3.974 1.527 0.000
6600 0.773 0.280 0.000
6800 0.081 0.029 0.000
7000 -0.002 -0.001 0.000
Sum 98.151 100.00 67.975
 6.7 Measurement of optical properties of phosphors 541

This gives the final color coordinates of x = 0.3688 and y = 0.3757 for
this lamp.

Note that we can now measure and specify color for the emissive mode, as
well as the reflective mode. If we measure a reflectance spectrum of an
object using a spectrophotometer, then we can use the Weighted Line
Method as well as a Colorimeter (which has special filters to compensate
for the preceptor response of the human eye). Either of these two
methods produce comparable results.

Note that these methods apply only to wavelengths in the visible portion
of the spectrum. If we wish to measure a lamp with an ultraviolet output, a
completely different approach is needed since the human eye cannot
detect this radiation.

e. Color Matching

As a final topic in this Section on the MEASUREMENT OF OPTICAL

PROPERTIES OF PHOSPHORS, let us consider the following diagram,
given as 6.7.35. on the next page.

In this diagram are shown the chromaticity coordinates for the three
phosphors used in most color-television tubes. What we wish to do is to
obtain a "black and white" picture having a 6500 °K. hue, like "Daylight".
Assuming all three color-phosphors have the same emission intensity
(which they do not), we adjust the emission intensity (by changing the
electron beam intensity within the tube) so that a mix of about 60% blue
and 40% red results. That is, the color coordinates of the blue to red ratio
lie at: x = 0.359 and y = 0.200. When we mix this with about 66% of the
green, we get a 6500°K. white emission. The actual % ratio of the
individual colors is: 0.666 green, 0.199 blue and 0.134 red. In addition,
we can reproduce any color within the triangle bounded by these three
hues. But, there are many hues that cannot be reproduced. Note that if we
had a phosphor with chromaticity coordinates of x = 0.100 and y = 0.800,
it would reproduce hues over a much wider range of colors.
542 Design of phosphors

6.7.35.-

But such a phosphor is not currently available. It has been estimated that
only about 85% of the hues found in Nature can be reproduced by a color
picture tube.

We can also calculate and mix colors predictably and accurately. We use
what has been called the "lever principle". We can mix phosphors or solid
dyes or any other colored material, provided that we know, or can
measure, the x and y values. Let us take the blue sulfide: x = 0.150 and y
= 0.100, and mix it with a yellow emitting phosphor having x = 0.450 and
y = 0.500. Note that we can reproduce any hue from yellow to blue,
including bluish-white and yellowish-white. We draw a straight line
between the two points and then measure the distance between the two
points. If we start with 100% yellow, adding blue shifts the hue toward
 6.7 Measurement of optical properties of phosphors 543

the blue. At 50%-50% blue to yellow, we are on the bluish white side. The
exact figures required to reproduce the 6500 °K white are 45.7% blue
and 54.3% yellow, providing the efficiencies of the phosphors are about
equal. If we were using the three hues given in 6.7.35., we first use the
baseline between the red and blue phosphor and draw a second line that
intersects within the 6500°K. ellipse of ILLUMINANT - C.

f.- Color Spaces

If we have a certain color, a change in intensity has a major effect on what

we see (in both reflectance and emittance). For example, if we have a
blue, at low intensity we see a bluish-black, while at high intensity we see
a bluish-white. Yet, the hue has not changed, only the intensity. This
effect is particularly significant in reflectance since we can have a "light-
blue" and a "dark-blue", without a change in chromaticity coordinates.

The concept of "lightness" involves the reflective power of materials. If

the reflectance approaches 100%, we say the material is "white", whereas
complete absorption (0.00% reflectance) produces a "black" material. Let
us examine exactly what is meant by these terms, particularly those of
subtractive color mixing. When colors are prepared by mixing dyes or
pigments, the resultant reflective hue is controlled by a subtractive
process of the three (3) primaries, whose reflectance spectrum is given
above. When these are mixed, the resulting hue is that where the curves
overlap, as is easily seen in the following diagram, given on the next page
as 6.7.38.

When we mix these primary colors, their reflectances remove more of the
incident light, and we see the part where the reflectances are reinforced.
Thus, we can get red, green and blue, but they are not primary colors in
the subtractive system. Intermediate hues can be obtained also in this
process when the subtractive primaries are used in less than full
concentration. That is, they are "lightened". Although we can explain this
effect on a spectrophotometric basis, we do not have a way of specifying
hues in terms of saturation, using the CIE system. This is one of the
failings of this system
544 Design of phosphors

6.7.36.-

Nevertheless, we do not have a better system for specifying color at this

time. There are other methods, which we will now discuss, in which this
deficiency was addressed. However, as you will see, none have been
entirely successful. It is the CIE system that is used mostly for color
specification, particularly for emissive materials like phosphors.

g. The Munsell Color Tree

One of the first attempts to specify reflective colors, color mixing and
saturation was accomplished by Munsell (1903). He devised a color
system based on factors he called hue, chroma and value. Munsell set up a
three-dimensional arrangement based upon minimum perceptual color
 6.7 Measurement of optical properties of phosphors 545

difference steps. He based these upon direct observation because he did

not have the instrumental means to do so. Therefore, his results are not
the same as those we are now using for the CIE system.

Munsell set up a cylinder whose vertical axis began with "black" and
ended with "white". This is the factor "Value" in Munsell notation, each
layer of which is given a specific number. He also defined five (5) primary
hues. These are: purple, red, yellow, green and blue. They are arranged
around a circle as shown in the following, and a series of circles are
"stacked" to form a "tree", namely:

6.7.37.-

In Munsell's system, hues are specified in a circular fashion with the same
set of hues on the same level. As the value, i.e.- "lightness", changes, one
jumps to the next highest circle. The amount of color "deepness" is
specified as chroma and becomes less as one approaches the edge of each
color circle.

The advantage of this SYSTEM is that in addition to being able to specify

hue, we can also specify Chroma, which is the degree of saturation of a
specific hue, as we move from the center outwards. Value is the relative
546 Design of phosphors

amount of "blackness" or "whiteness" (equal parts of black and white

produce gray) in the particular reflective color being specified. It is this
system that gives us access to the degree of "grayness" and/or saturation
of any hue.

On a practical basis, if we wish to set up this system, we would assemble a

set of "color-chips" at each point on the Color-Tree. Each color-chip
would be specified by two factors, H = hue, and V/C, which is value
(grayness) modified by chroma (saturation). The actual number of layers in
the Munsell Color Tree was determined by "minimum perceptual
difference". That is, the minimum change in color that produces a
perceptible difference. This arrangement specifies all light colors as well
as the dark ones. To use such a system, one would choose the color-chip
closest to the hue and saturation of the test color and thus obtain values
for H and V/C.

However, it was soon discovered that the system was not perfect. Reasons
for this include the facts that the hues defined by Munsell are not those of
the primaries of the human preceptor. Furthermore, Munsell was
somewhat subjective in his definitions of hues. And, the system did not
take into account the luminosity response of the human eye in regard to
color. If one does so, then the Munsell System becomes "bulged" in the
direction of the yellow-green regions and shrunken in the blue and red
regions. We can illustrate this by taking the Munsell colors for Value-5 and
Chroma-8 and plotting them on aCIE diagram, as shown in 6.7.38., given
on the next page.

What we should get is a circular spacing of the ten Munsell colors, if the
Munsell system is truly accurate in regard to the luminosity factor.
However, they are not. On the CIE diagram, the spacing is considerably
distorted. Since we already know the CIE method to be corrected for the
luminosity factor, the conclusion is obvious. In 1920, Priest showed that if
the Munsell-Chips were viewed on a white-background, the "brightness",
i.e.- lightness as viewed by the human eye, could be related to the Munsell
system by:
6.7.39.- V = 10 Yl/2
 6.7 Measurement of optical properties of phosphors 547

6.7.38.-

where Y in 6.5.39. was defined as "brightness" or eye response. Over the

years, the Munsell System has been modified in an attempt to agree with
actual human perception, using a "middle-gray" background". Y was
defined as the Tristimulus Value of the CIE color system and used in the
following formulae, given as follows:

6.7.40.- Munsell- Godlove (1927): V = [1.47 Y - 0.474 Y2 ] l^

N.B.S.- Judd (1943) : Y/YMgo = 1.2219 V - 0.2311 V2 - 0.2395

V3 -0.0201 V4 + 0.008404 V$

Glasser (1958): V = 25.29 Y 1 / 3 - 18.38

The Judd formula given here is too difficult to use and it is the Glasser
formula which came into general use to modify the Munsell Color System
back towards the cylindrical form. Nevertheless, it is well to note that it
was the subjective observation of the lack of correction for luminosity in
the Munsell System that gave impetus to the original development of the
CIE Color System. The major problem with the Munsell system was that
548 Design of phosphors

each person attempting to match colors could not produce the exact
same result. So color matching became dependent upon the person.
Consequently, the concept of the Standard Observer was born.

h. Color Matching and MacAdam Space

When Munsell devised his color space, he did so on the basis of minimum
observable color perception steps. But the problem with the Munsell
System was one of reproducibility, which the CIE Standard Observer
cured. In formulating a color match, one wants to be able to predict the
correct concentration of colorants required, whose scattering and
absorption properties are known, i.e.- the lightness, so as to match the
sample submitted, starting with their spectrophotometric curves. In
practice, this is not so simple, since two colors must have identical
spectrophotometric curves to be exactly equal.

It turns out that the human eye will identify the two colors to be equal if
their spectrophotometric reflectances are reasonably close. Two colors
may appear to be equal under Daylight illumination, but quite different
under incandescent lamp illumination (see 6.7.18.). These colors are
known as "metamers" and the phenomenon "metamerism". It was soon
determined that the 1931 CIE chromaticity diagram, and luminance
function, Y, are not representative of equal visual spacing. That is, equal
changes in Y do not represent equal changes in visual perception for all
values of Y. Nor do equal increments of x and y represent the same visual
effect for all locations on the chromaticity diagram. In other words, there
is a minimum perceptual difference on both x and y (i.e.- Ax and Ay).
But, the size of Ax and Ay is not the same at all parts of the chromaticity
diagram. This is the same problem that Munsell encountered and is due
to the fact that the human eye is an integrating instrument, not a
dispersive one. MacAdam (1942) studied this problem and established
limits of Minimum Perceptual Difference steps (MPD) in the form of
ellipses on the CIE diagram, given on the next page as 6.7.41.

The distance from the center to a point on the circumference of these

ellipses represents the same visually perceived color difference
 6.7 Measurement of optical properties of phosphors 549

6.7.41.-

(here they have been made ten (10) times too large so as to be readily
visible on the diagram). Note that very large changes in Ax and Ay occur
in the green region as compared to that of the other regions. This is, of
course, not tolerable and represents a serious deficiency in the 1931 CIE
system. MacAdam's work led to a new system for color specification and a
decided improvement in color matching techniques.

It turns out that a linear transformation of x and y coordinates to a new

set of coordinates, u and v, is sufficient to do the job. MacAdam's
transformation equations were:
550 Design of phosphors

6.7.42.- u = 4x/(-2x+ 12y+3)

v = 6y / (-2x+ 12 y+ 3)

Plotting these values gave the following chromaticity diagram:

6.7.43.-

The true test is how well the Munsell hues plot out on the CIE diagram.
As can be seen on the right hand side of the diagram, the Value 5 -
Chroma 8 hues do construct a nearly perfect circle. Thus, the MacAdam
transformation is a definite improvement over the 1931 CIE system. In
1960, the CIE adopted the MacAdam System, having defined the
equations (with MacAdam's help):

6.7.44.- u= 4 X / (X+ 15Y+ 3Z)

v= 6 Y/ (X+ 15 Y+ 3Z)

The following diagram shows MacAdam's Uniform Chromaticity Scale, and

has Minimum Perceptible Color Difference steps plotted as ellipses:
 6.7 Measurement of optical properties of phosphors 551

6.7.45.-

Note that the sizes of the ellipses are about equal in all parts of the
diagram and in most regions of the color spectrum. This system was fully
adopted by the CIE in 1964 and remains the recommended system for
color matching.

As a final step, let us now return to phosphors and show the various types
of fluorescent lamps manufactured in terms of emittance hues and
tolerances. In the following diagram, given as 6.7.48 on the next page, we
have plotted the Standard Specifications for fluorescent lamps, in terms
of the chromaticity as defined by MacAdam ellipses.

It is these which are in use in the Industry for quality control of

phosphors and fluorescent lamps. The ellipses show the color tolerance
for emittance of each of the lamp hues. Thus, they are actually the
Minimum Perceptible Color Difference, for each of these lamp colors.

Because the lanthanides ("rare earths") have become prominent in the

manufacture of fluorescent lamps, we will address this technology next to
show how they differ from the 'SQ activators given above. We will find that
these cations can be used in situations where no other activators can be
applied.
552 Design of phosphors

6.7.46.

6.8- RARE EARTHS AND THEIR SPECTRAL PROPERTIES

The lanthanides (rare earths) have found use in regard to phosphors, solid
state lasers (which are specialized phosphors) and Anti-Stokes phosphors.
Indeed, if rare earths were not available, we would not have useful lasers,
i.e.- YAG:Nd3+ or infra-red converting (Anti-Stokes) phosphors. We begin
with a short history of the lanthanides and their electronic properties.

I. History of the Lanthanides

Lanthanides have historically been called "rare earths" because it was

originally believed that these elements were sporadically distributed in
nature. Actually, they are relatively abundant in the Earth's crust, even
 6.8 Rare earths and their spectral properties 553

though only a few ore bodies exist where they constitute the major
metallic content. The isolated oxidic product has the feel of Fuller's
Earth, hence the use of the term "earth". Their chemistry is unique and
major difficulties in separating them have been encountered. This
behavior relates to their electronic configuration:

6.8.1. |Xe core| 6s 2 = | 4d lo 4f n 5s 2 5p 6 | 6s 2

The optically active electrons involve the 4f shell, buried deep within the
Xe-core of electrons. Because of the shielding effect of the outer closed
shells (5s 2 5p 6 ), the chemistry of each lanthanide is quite similar to the
others. Thus, the separation of each rare earth, free from the others, has
proven to be very difficult. The most stable valence state is Ln 3+ , where
three (3) electrons (4f* and 6s2) have been removed. Most separations
were accomplished in the 1800's by repeated fractional crystallization.
Sometimes, literally thousands of recrystallization steps were required
before a reasonably pure product could be isolated and demonstrated to be
pure. A number of false starts have been recorded where the investigator
thought he had a pure product, only to be shown that he did not. The first
rare earth to be documented, gadolinium - Gd, was actually named around
1805. But, the product was later shown to be a mixture of rare earths. Gd
was finally isolated in a pure form some 75 years later. It was named after
the Swedish investigator, J. Gadolin (1794) who characterized the first
rare earth ore.

The following table, presented on the next page, lists the lanthanides and
the dates of their first isolation and characterization. Promethium is
radioactive, with a very short half-life, and does not occur in Nature at all.
It was first synthesized in an atomic reactor and is now used in human
circulatory analysis via "CAT" scans when it is injected into the body.

Y & Sc are not lanthanides, but their chemical properties are so nearly
identical that they occur in the same ore bodies and have the same
separation difficulties as those of the lanthanides. Hence, they are
included in the rare earth category.
554 Design of phosphors

Table 6-6
Name Trivalent State Investigator

Lanthanum La= 4f° Mossander - 1839

Cerium Ce = 4f! Berzelius - 1814
Praseodymium Pr = 4fl Welsbach - 1885
Neodymium Nd = 4f3 Welsbach - 1885
(Promethium) (Pm) = 4f* Fermi Labs - 1969
Samarium Sm = 4f5 de Boisbandran- 1879
Europium Eu = 41 6 Demarcy - 1869
Gadolinium Gd = 4f de Marignac - 1881
Terbium Tb = 4f8 Delafontaine - 1878
Dysprosium Dy= 4f9 de Boisbandran - 1886
Holmium Ho = 4f*0 Cleve - 1879
Erbium Er = 4f n Cleve - 1879
Thulium Tm = 4f12 Cleve - 1879
Ytterbium Yb = 4f*3 de Marignac - 1878
Lutecium Lu = 4f14 Urbain, Welsbach, James - 1907
Yttrium Y=4f° Mosander - 1843
Scandium Sc = 4f° Nilsson - 1879

II. Chemistry of the Lanthanides

Because each rare earth has essentially the same electronic configuration,
it is very difficult to separate them from each other. Indeed, even a given
rare earth said to be 99.99% pure will have several hundred parts per
million of the others which comprise the ore from which they were
refined.

The rare earths can be divided into two classes, the "lights" and "heavies",
i.e.- < 4I7 and > 4 ^ , respectively. There are slight differences in chemical
reactivity which follow this rule as well. Lanthanides form cations in
solution and are not amphoteric. That is, they do not form anions. The
oxides are the usual form, and most have the formula, Ln2O3 . There are
some mixed valence states which result in formulas like Pr6 On (which
contains Pr3+ and Pr 4 ^ , and Tb4O7 (Tb3+ and Tb4+, i.e. - Tb2O3 + 2TbO2).
 6.8 Rare earths and their spectral properties 555

The rare earths follow the same pattern of solubility encountered in most
other "non-rare earth" compounds, i.e.-

6.8.2.- SOLUBILITY OF THE RARE EARTH SALTS

SOLUBLE INSOLUBLE
Ln(NO 3 ) 3 Ln(OH)3
LnCl3 LnF3
LnBr3 LnBO3 • 2 H2O
Lnl 3 L11PO4 • n H2O
Ln 2 (SO 4 ) 3 LnVO4
Ln 2 (C 2 O 4 ) 3 n H 2 O

This list is not all-inclusive but is illustrative. The oxalate is the most
insoluble salt and is most often used for recovery of rare earths from
solution. The commercial ores mined for extraction of rare earths
include:

6.8.3.- LANTHANIDE ORES

NAME GENERAL FORMULA

Monazite (Ce,La,Th,Ln)PO4
Bastnaesite (Ba,Sr,La,Ce,Ln) F n OD3
Xenotime (Y,Gd,Ln)PO4
Euxenite (Ln,Y)NbTaTiO6

Only the major cations are shown in the above formulae, even though each
ore contains all of the rare earths in various proportions. One commercial
chemical process used for initial separation of the rare earths follows this
sequence of steps, as shown in 6.8.4., presented on the next page.

Note that this involves ordinary laboratory procedures. The separation of

the individual rare earths in a pure form is another matter. This problem
was not solved until the late 1930's and early 1940 1 S, when Spedding and
co-workers (University of Iowa-Ames Laboratories) applied ion- exchange
chromatography to the problem.
556 Design of phosphors

6.8.4. - CHEMICAL SEPARATION OF THE RARE EARTHS

1. ORE: Leach crushed rock with 96% H2SO4 = 98% dissolution.

2. Dissolve into water = dissolved rare earth cations.
3. Precipitate with oxalic acid = 99 % + Recovery
4. Fire oxalates to oxides (> 800 °C).
5. Product = Mixed rare earth oxides.

To do so, one uses a hydrogen-ion exchange resin which is loaded into a

long glass column. For commercial separation, this consists of several 12"
by 25 foot long columns in tandem, and the process takes several weeks
(typically 4 to 6) to complete the process. After considerable work with
several types of resins, Spedding was able to establish the following
retention series on an ion-exchange column:

6.8.5.- Th*+ > La3+ > |Ce 3+ |> |Y3+|> Lu3+> Ba2+>Sr2+>Ca2+>K+>NH4+> H+

where |Ce 3 + | and|Y 3+ | represent the light and heavy fractions,

respectively. The process includes loading a column with a lanthanide
solution, and then eluting with a 5% ammonium citrate solution at a pH of
2.5 to 3.2 . Under these conditions, one gets a separation of each rare
earth, which moves through the columns as overlapping bands, one
following the other. We can characterize such a separation as:

6.8.6- Ce (lights) Fraction

Y (heavies) Fraction

The Ce-fraction is the familiar off-color beige powder, long used as a glass-
polishing agent. Rare earth metals can be prepared by fused-salt
electrolysis, starting with the fluorides. The product is a "mixed-metal" ,
or "mischmetal", used to produce cigarette lighter-flints.

However, this method was much too slow and costly. In the 1950's, a new
method was developed by the Bureau of Mines in Colorado. This method
involved solvent extraction with an aqueous nitrate solution of lanthanide
cations, shaken together with an immiscible organic phase such as CCI4
 6.8 Rare earths and their spectral properties 557

containing tri(n-butyl)-phosphate. Its extraction coefficient was a = 1.50.

The most useful, and expensive, rare earth is Eu3+ and it was soon
determined that DEHPA extracted more europium from the ore. The
latest process uses DEHPA, a di(2-ethyl hexyl) phosphoric acid, as a
complexing and extraction agent. The extraction coefficient for DEHPA is
a =2.50. By pH control, one can select and extract a single lanthanide as
desired. This process has allowed the production of products containing
99.99% of a single lanthanide and even 99.999% purity (these levels of
purity are in terms of the dominant rare earth present). All of the
lanthanides are presently available in a purity up to, and including,
99.9999%. The commercial process used by Molycorp is shown in the
following diagram given as 6.8.7. on the next page. The extraction
process involves an organic insoluble phase in contact with an acidic
aqueous phase of controlled pH.

Note that aqueous flotation (with appropriate detergent compounds) is

used to separate the rare-earth-rich particles from the siliceous materials.
After drying, the ore is calcinated (fired in air to form the oxides) and
then treated with concentrated HC1. A cerium-rich concentrate is left
after the other rare earths have dissolved. We have not shown how the
cerium concentrate is processed since the main object is to obtain the
more expensive oxides as a final product. The rare earth chloride solution
is then adjusted to pH = 1.0, purified with activated carbon to remove
some of the heavy metals and then extracted with DEHPA. The first
extraction separates the "heavy" and "light" fractions of the rare earths.
The resulting extracted solutions are then subjected to further extraction
processes to produce the individual products, including La, Pr and Nd.

The organic solution contains the La-concentrate which is purified,

extracted further and then precipitated as single hydroxides. The other
fraction is readjusted to pH = 3.5 and further extracted. After the first 2
extractions (which are returned to the pH = 1 . 0 step), concentrated HC1
is added and the last 2 extractions are used to separate the Eu, Sm and
Gd fractions.

Any heavy metals left over are precipitated and separated by means of H 2 S.
558 Design of phosphors

6.8.7.-

The details of this extraction process is shown in the following, given as

6.8.8. on the next page. A reduction step serves to form the Eu^ state so
that it can be further purified from the Sm and Gd remains that may
contaminate it.

Precipitation as EuSO4 is the final step of separation. Calcination forms the

oxide, the usual commercial form of europium, i.e.- Eu2O3.
 6.8 Rare earths and their spectral properties 559

6.8.8.-

III. Rare Earth Energy Levels and Electronic States

Although the lanthanides can possess up to 14 identical f-electrons, each

rare earth differs from the others in its optical and magnetic properties.
For the general case, in terms of effective radius, <r>, we have the
following relation of electronic energy levels, shown as follows:

6.8.9.-
560 Design of phosphors

We can thus write the electronic configuration in terms of ascending

energy as: Xe 4P1 (5s25p6) 6s 2 5d, where the 5d orbital is ionized first,
then the 6s 2 orbitals. Electronic transitions (ground to excited state) take
place among the 4P1 orbitals. Because of the shielding nature of the
5s 2 5p 6 shells, perturbation by surrounding neighbors does not take place
and sharp line spectra result.

But, since 1, the azimuthal quantum number, can take on values from 0 to
1,2,3,4, etc., with s = 3 1/2 (i.e.- 7 x 1/2), the observed spectra can
become quite complex. The following table shows neutral atom electron
configurations and observed valences:

TABLE 6- 7
Electron Configurations of the Rare Earth Elements
Number Symbol Electronic Configuration Observed
Neutral Atom Trivalent Atom Valence

57 La 4fO6s 2 5d 4fO 3
58 Ce 4f! 6s 2 5d 4f* 3,4
59 Pr 4f3 6s 2 4f2 3
60 Nd 4f4 6s 2 4f3 3
62 Sm 4f6 6s 2 4f4 2,3
63 Eu 4f7 6s 2 4f6 2,3
64 Gd 4f8 6s 2 5d 4f? 3
65 Tb 4f3 6s 2 4f8 3,4
66 Dy 4f!06s 2 4f9 3,4
67 Ho 4fH6s 2 4fW 3
68 Er 4fl 2 6s 2 4fH 3
69 Tm 4fl36s 2 4f*2 3
70 Yb 4f146s2 4f!3 2,3
71 Lu 4f146s25d 4f*4 3

For any given atom, the number of possible states can be calculated from
combinatorial theory, i.e.-

6.8.10.- C k n = C k i4 = 14 ! / (f ! [14-fj !)
 6.8 Rare earths and their spectral properties 561

where f is the number of electrons present in the particular electron state

of the specific rare earth. This gives:

6.8.11.- NUMBER OF ELECTRONIC ENERGY LEVELS POSSIBLE

Ln3+
f= _1 3 7
Spin and Orbit: 14 364 3432
J-Levels: 2 41 327

It turns out that the spin and orbit numbers are actually the degeneracy of
the energy levels in question and that they represent the possible Stark
State splitting pointed out in a previous chapter. The full set of values are
shown in the following Table, namely-

TABLE 6-8

NUMBER OF STATES WHICH ARISE FROM THE 4f CONFIGURATION

f= 1 2 . 3 4 5 6 7

Total Number of J-Levels: 2 13 41 107 198 295 327

Multiplets: 1 7 17 47 73 119 119

Degeneracy: 14 91 364 10012002 3003 3432

(spin and orbital)
GROUND STATE: 2F 3H 4j 5j 6H 7F 8S
MULTIPLET

Obviously, for f = 8, 9, the total number of J-levels will equal 295, 198,
with the same degree of degeneracy, etc.

As we showed in Chapter 5, the Hamiltonian for the 4f - field will likewise

consist of several parts:
562 Design of phosphors

6.8.12.- HAMILTONIAN FOR 4f ELECTRONS

Central Field Coulomb Spin-Orbit Crystal field

H4f = Ho + Hc + Hso + Hcf

electrons terms J-levels Stark components

(multiplets) (~ 2000 c m 1 ) (~ 200 cm' 1 )

The rare earths are unique in that their individual Stark States are
separated by no more than 200 cnr 1 - The following shows part of the
diagram for the 4f6 electron configuration:

6.8.13.-

A Table is presented on the next page which shows how these states arise
in L-S coupling, in terms of the number of levels and the number of
multiplets. As we shall see, they are not valid because the only good
quantum number for the lanthanides is J, the azimuthal quantum number.

The data in this Table lists the number of multiplets (Terms) and Levels
(Spin-orbit: J = L ± S) for both odd and even electron configurations.
 6.8 Rare earths and their spectral properties 563

TABLE 6-11
Number of J Levels for Configurations Consisting of Equivalent 4f- electrons
ODD MULTIPLETS
Number of NUMBER OF J LEVELS
Configuration Multiplicity Multisets 1/2 .3/2 5/2 7/2 9/2 1/2 13/2 1 5/2 17/2 19/2 21/2 23/2 25/2 Total
f1andf13 doublet 1 - 1 1 - - - - - - - 2

fSandfH doublet 12 1 3 4 4 4 2 2 1 - - - - 24
3

quartet 5 1 3 3 3 3 2 1 1 - - - - - 17

TOTAL 17 2 6 7 7 7 2 3 3 1 - - - - 41

f5 and f9 doublet 46 4 9 12 13 14 14 12 9 7 4 2 1 91

quartet 24 5 10 13 14 14 12 9 7 4 2 1 - 92

sextet 3 1 2 3 3 2 2 1 1 - - . 15

TOTAL 73 10 21 28 30 29 26 20 16 9 5 3 1 198

f doublet 72 7 12 17 20 19 18 16 12 6 3 2 1 - 142

quartet 40 8 15 20 23 22 20 16 12 8 5 2 1 152

sextet 6 2 4 5 6 5 4 3 2 1 - - - - 32

octet 1 - - 1 - . - - - - - - - 1

TOTAL 119 17 31 42 50 46 35 26 18 11 5 3 1 - 327

EVEN MULTIPLETS
Number of NUMBER OF J LEVELS
Configuration Multiplicity Multiplets 0 1 2 3 4 5 6 7 8 9 10 11 12 Total
f2andf12 singlet 4 1 _ 1 1 _ 1 _ _ _ 4
triplet 3 1 1 2 1 2 1 1 9
Total 7 2 1 3 1 3 1 2 _ _ _ _ _ _ 13
HandflO singlet 20 2 4 1 4 2 3 1 2 _ 1 _ _ 20
triplet 22 3 5 9 9 11 9 8 5 4 2 1 _ _ 66
quintet 5 1 2 4 3 4 3 2 1 1 _ _ 21
Total 47 6 7 17 13 19 14 13 7 7 2 1 _ 91
f6andf8 singlet 46 4 1 6 4 8 4 7 3 4 2 2 _ 1 46
triplet 56 6 11 20 21 25 22 21 15 12 7 5 2 1 168
quintet 16 3 6 10 11 12 10 9 6 4 2 1 _ 74
sextet 1 1 1 1 1 1 1 1 _ . _ _ _ 7
Total 119 14 19 37 37 46 37 38 24 20 11 8 2 2 295
564 Design of phosphors

We have already listed the total number of Stark States possible in Table
6-8 as Degeneracy (spin and orbital).

It is easily seen, in view of the above, that the number of energy levels
which can appear in rare earth spectra are quite complex and numerous.
In 1963, Dieke measured the experimental energy levels of Ln 3+ in LaF3
single crystals. We will not present this data here except to note that
many levels appear in the experimental spectra, even though the Stark
States of each individual level are not shown in Dieke's diagram.

Following the appearance of Dieke's data, and that of others, attempts

were made to calculate the theoretical energy levels of the trivalent rare
earths. The first attempts used the Hamiltonians given in 10.3.4.
However, the results were not even close to the experimental levels
measured, despite the various crystal field parameters applied.
Considerable effect was expended in applying those methods so as to
transmute the calculated levels to resemble those experimentally
determined, but with little success.

It turns out that the electrons buried below the 4f energy levels do exert a
definitive effect upon the allowed 4f energy states. Russell-Saunders
coupling is no longer dominant, and L * £ 1 , and S * I s for the electron
shells present. In the 4f shell, the effect of the spin-orbit coupling factor,
KM . is greater than that of the crystal field splitting factor. This has the
effect of mixing the character of the levels which appear, so that the
excited states possess part of the ground state character, and vice-versa
(this will be explained below). A surprising result is that the intensities of
rare earth spectra are much stronger than those that would be predicted
by Hund's Rules for quadrupole (qq) transitions. Most are in fact classified
as dipole transitions, both electric and magnetic. Thus, L-S terminology,
i.e.- Russell-Saunders, is no longer valid in describing the spectroscopic
energy states of the trivalent rare earths.

It turns out that what is required to make the calculated energy levels
conform to the experimental levels is to use some sort of "intermediate"
coupling scheme, that is- a method of coupling spin and orbit states that
 6.8 Rare earths and their spectral properties 565

is "intermediate" to the extremes of the coupling methods, namely L - S

and jj - coupling. To see how this arises, we must delve into methods of
calculating such energy levels.

a. Calculating Lanthanide Energy Levels

We will present an abbreviated synopsis of how one proceeds to calculate

the energy levels of the lanthanides. This problem remained extremely
difficult, and in some cases unsolvable, until the advent of computers. A
complete description was given in the first edition of this volume.

First, let us set up a J-matrix to calculate the 4f energy levels in L-S

coupling, using the Hamiltonians of 6.8.10. Although we have already said
that this method does not work, we still need to examine the general
method of calculation. WE FIRST SET UP A SERIES OF MATRICES, using
the Hartree-F6k methods already developed for Quantum Mechanics. The
next step is to then reduce them in terms of Slater determinants, namely-

6.8.14.-

=
J / M i~J§_ f L 2 1 - S * L 3 1 - S . l _ 4 1 - S • • • • • •
L12 S . l ^ S ,
j-i3-|.L 2 3-S-.t 3 3-.S.
L,4 S . . . . t^jS •••
::
V :::::::::- '"J
What we are doing is solving for one energy term in the above L - S
matrix. Remember that the simple case, p 2 , has at least 100 terms, and
that 4f14 may have thousands of terms per multiplet. We diagonalize the
matrix as shown and solve for the energy term. We repeat this for all of
the terms. Once we have accomplished this formidable task, we find that
our energy values for the rare earths are off as much as 50% from the
experimental values. Note that we have used a vector model for the 4f
566 Design of phosphors

shell where L and S commute with H4f , i.e.- the Hamiltonian in

Schroedinger's equation of 6.8.12.:

6.8.15.- H (L-S) W4f = E *F4f

Slater determinants for two equivalent f-electrons are written in terms of

aROOTHAN operator, f^f, where:

6.8.16.- fW Ui = £i Ui

where Ui is a spin-orbital function relating to the Coulomb and Exchange

operators, J i and Ki , respectively. 8j is the energy of the state under
calculation. Interaction energies for pairs of identical 4f-electrons are:

6.8.17.- E o = F o - 4/195 F 2 - 2/143 F 4 - 1000/5577 F 6

The F- values are Slater radial integrals. Using these is easier than it
appears, since Slater has tabulated all possible values for use in matrix
diagonalization.

In our matrix of 6.8.14., we set all non-diagonal terms to zero and used
only the diagonal terms to obtain J z . Since this did not work, let us now
use the off-diagonal elements to re-calculate adjusted- J z terms and
energies.

This is equivalent to stating that we no longer have L - S coupling.

For a given 4P1 configuration, we can get a number of J- values, many with
the same labeling, but with different energies (remember what we said
about intermediate coupling).

Our next step is to arrange all J's of the same term into a J-matrix, so as
to calculate the energies. For 4f2 (Pr3+) or 4f12 (Tm 3+ ), we are looking for
13 solutions of the j-matrix, as shown in 6.8.18, given on the next page.
 6.8 Rare earths and their spectral properties 567

6.8.18.-
pfhe J z Matrix

J
^ 21^2^23 • " •
2 J J
J z = "fi 31 32^3^***
J J J
41 42 43 ^*K
• • •
• • •
\ : : : /

At this point, the Slater parameters can be described as:

6.8.19.- E = F o - 2/45 F 2 - 1/33 F 4 - 50/1287 F 6

It is convenient to use operators developed by Racah (1950) which

considerably simplify the work. For 4f electrons, we have:

6.8.20.- ENERGY PARAMETERS FOR CALCULATION

Eo = Fo - 2/45 F 2 - 1/33 F 4 - 50/ 1287 F 6

Ei = F o - 14/405 F 2 + 7/297 F 4 + 350/11,583 F 6
E2 = F o - 1/2025 F 2 + 1/3267 F 4 + 175/1656369 F6
E3 = F o - 1/135 F 2 + 2/1089 F 4 - 175/42471 F 6
etc.
For 4f2 , we can have up to 91 energy states, which illustrates the
complexity to which we have subjected ourselves. The Racah parameters
are given in the following:

Ak = F o - 49 F 4 {qAk} = crystal field parameter.

Bk = F 2 - 5 F 4 {qBk} = crystal field parameter = qAk <r k >
Ck= 35 F 4 {qCk} = spherical harmonic function = (2k+l/4jr) 1 / 2 qYk

where we have all the requisite values of k available for calculating these
parameters. Nielson and Koster (1963) gave eigenfunctions and multiplet
568 Design of phosphors

energies for all of the 4P1 configurations in terms of the parameters: qBk
and q Ck.
b. The Concept of Fractional Parentage

While we are carrying out these calculations, we find that we have

introduced, through necessity, two new concepts, that of "intermediate
coupling" and that of "fractional parentage". To explain what we are
discussing, consider the following. While the original work on energy
(spectroscopic) states of an atom was being formulated, it was realized
that a free-atom's electronic states could be adequately described by
either of these two states which differed by the method of coupling its
azimuthal and spin states:

6.8.21.- Russell-Saunders Coupling j - j Coupling

L = £1 S = Is j = 1± S
J = L + S = (II ± Is) U = j + j . . . . = I (1 + s)

where 1 and s are individual quantum states of any given electron in the
atom. In the first case, the individual electron vectors are first summed,
and then coupled together to give the final J. In the second case (which
has been found to apply to heavy atoms), the individual electron vectors
are first coupled, and then summed to give j . Each j is then summed with
the other j's of the separate electrons to give the final spectroscopic value.
As we said, both of these methods represent the extremes in
spectroscopic coupling notation. While Russell-Saunders coupling terms
have been well documented, no such terms have been formulated for j-j
coupling. Fractional parentage refers to the fact that the original system of
specifying spectroscopic terms, i.e.- 2G, 4H, etc., was modified and
adjusted using L - S coupling rules. In intermediate coupling, L and S are
no longer good quantum numbers, and only J is valid (because of our
method of calculation). However, even though we developed a new
method of calculating energies, we did not, and have not, developed anew
method of specifying terms in intermediate coupling. Therefore, we refer
to the fractional parentage of L - S terms, namely:
 6.8 Rare earths and their spectral properties 569

6.8.22.- % SLJ (J = 15/2) = 93% 6 H, 7% 4 I

which happens to be the ground state of Dy3+. The same fractional

parentage occurs for the upper energy excited states as well. It is for this
reason that the energy transitions in the lanthanides do not follow Hund's
Rules for L-S coupling, and that we observe dd and dq transitions for rare
earth transitions (with the correspondingly strong oscillator strengths)
rather than the weak qq transitions expected. That is, because the
excited states contain a fractional part of the ground state term, and vice-
versa, the transitions actually involve transitions between coupled ground
and excited states.

The complete intermediate coupling energy diagram for 4f2 electrons is

presented in the following diagram, given as 6.8.23. on the next page.

Note that the energies are in terms of Slater parameters and the spin-
orbit coupling parameter, ^.f and are plotted against the crystal-field
parameter, also in terms of §4f. This parameter has been defined in terms
of a set of constant values of the Racah parameters, as shown. On the right
of the diagram (|4f = 1), are shown the hypothetical levels of the infinitely
strong cubic field, which are the j - j values. It has been shown that £4f
varies in a linear manner for 4P1 electrons. Using these data permits us to
predict and use the proper values for establishing Intermediate Coupling
diagrams for all of the rare earths, without having to begin the calculations
from first principles.

We can then consolidate the effects of both the spin-orbit and crystal field
parameters into a single diagram, as shown in 6.8.24., presented on a
following page. In this diagram, we show the L - S states at the left hand
side. The actual experimental energy values determined are plotted as
points on the individual curves for the several Ln 3+ ions. While the
agreement is good, it is not perfect.

In comparing these values to those of 6.8.23., we see that overlapping

states in 6.8.23. have been simplified to a single state in 6.8.24., thus
illustrating the concept of fractional parentage.
570 Design of phosphors

6.8.23.-

The most complicated spectra will be that of 4I 7 , which is the electron

configuration for Eu 2+ , Gd 3+ a n d T b 4 + . By reference to 6.8.23. and the
left hand part of 6.8.24., it is easy to see that if F2 and Sj4f are adjusted,
one can get the energy levels for several ions. If we do this, we get the
values shown in 6.8.25., presented on the next page. The formulas used
to do the adjustment are:

6.8.26.- T] =(E -F 2 /0.06)/[(F 2 /0.06) 2 + £ 2 ] 1 / 2

and: § = X / 1+X where X = 0.06 £/ F 2
 6.8 Rare earths and their spectral properties 571

6.8.24.-

6.8.25.- Spectroscopic Parameters for Several Related Ions

F-^incm-1 f=4f in cm' 1
Eu 2+ 345 1220
Gd3+ 405 1581
Tb 4+ 435 2080

These are the same as used in the above diagrams. We find that several
overlapping states do exist so that we have difficulty in justifying any
labeling. What has been done is to specify each of the energy levels in
terms of their composite states. In the following Table, we list an
abbreviated list of energy states for Dy3+ in terms of the fractional
parentage of each state. J varies from J = 1/2 to J = 17/2 and we have
only shown 3 of the 12 J-terms and 22 of the 56 energy levels.
572 Design of phosphors

TABLE 6-12
FRACTIONAL PARENTAGE OF Dy3+ ENERGY LEVELS BELOW 4 0 , 0 0 0 c m 1
After Wyboume (1963)
J Calculated Observed % SLJ
Energy Energy
1/2 13,723 916F,94D
31,991 54 4 D, 28 4 P, 112P, 76F
37,764 70 4 P, 284D, 26F

3/2 13,175 13,148 91 6 F, 7 4 D, 14F

27,715 30 4 D , 29 4 P, 22 2P,9 6 P, 5 6F,4 2 D,1 4 F
32,129 55 6P,144 D, 102D,104P,54F,32P,2 ^F
33,617 79 4 F, 9 2 D, 7 6 P, 4 4 D
37,696 84 4F, 8 2 D, 3 6 P, 3 4 P, 2 2 P
39,604 35 4 P, 22 6p, 19 4D,11 2 D,6 4 F, 5 2 P

5/2 10,155 10,155 92 6 H, 7 4 G

12,389 93 6F, 4 4 D, 3 4 F
28,756 27 6 P, 26 4 P, 20 4 F,10 4 D,10 4 G,4 2 D,l6H
30,013 38 6 P, 28 4 F, 16 4 D,10 4 G, 7 4 P,1 2D
31,874 38 4 D, 33 4 G,11 2 F,9 2 D,4 4 P,3 6 H,3 6 F
34,231 48 4 F, 21 4 G,14 4 D,8 2 F, 4 6F.2 2 D,2 4 P
36,878 46 4G, 144F, 13 2 F,10 6p, 8 2 D,7 4 P,1
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

15/2 0 0 93 6H,6 4 I
22,087 22,086 54 4 I, 162K, 14 4K, 56H, 52L,4 4L,2 4M
28,742 47 4M, 204I, 15 4L, 152L, 34K
31,402 564K, 15 4M, 114L, 72K, 6 4 I, 5 2L
36,794 564L, 214K, 12 4M, 92K, 12L
38,361 292L, 292K, 17 4M, 114I, 104L, 2 4K

We have already indicated the ground state and excited states to be

coupled, i.e.- the spin-orbital states are mixed. The ground state level
consists of: J = 15/2 (93 6H,6 4I). What we do is to label the energy states
in terms of the spectroscopic label which makes the major contribution to
its intermediate coupling makeup. Thus, for the J= 5/2 levels, we would
 6.8 Rare earths and their spectral properties 573

call some of these 4 D and 4 G, even though the labels only contribute about
half of their total makeup.

The data given in Table 6-12 for Dy3+ have been replicated for the other
rare earths as well. Having this data allows us to form what is called a
Groatrian Diagram of the indicated energy levels, according to the J-
values (which are the only valid quantum numbers in intermediate
coupling). Let us show this for Dy3+ so that we can examine the energy
processes which take place for this trivalent rare earth ion.

In the following Groatrian Diagram, given as 6.8.27. on the next page, the
energy levels have been plotted according to each J-value. The strongest
emission for Dy3+ in all hosts consists of a set of lines around 5720 A , a
yellow emission. The energy of the emission transitions is centered at
about 17,500 cm- 1 and phosphors have been produced with quantum
efficiencies in excess of 85%. The emission transition is always:

6.8.28.- 4Fg/2 ^ 6Hl3/2

Crystal field splitting accounts for the set of emission lines observed. In
this diagram, all of the energy levels have been labeled with the term
which contributes the greatest degree to that level.

C. Energy Transitions and Mixed States

Our next step is to further examine selection rules for "forced electric
dipole" transitions. We need to account for the actual energy observed, in
terms of the two terminal states possible, i.e.- yellow = [17560 cm' 1 :
6
Hi3/2] vs: [green = 21,020 c n r 1 : 6 Hi5/2]. We can see that if an excitation
energy of 40,000 cm' 1 were to be used, the 2 Hn/2 level would be
excited. This excited state would then relax, through phonon processes,
to various states until the emitting 4Fg/2 state is reached. Then photon
emission must result if the excited state is to return to its ground state.

You might ask why this state is the emitting state. Remember that a
photon is a significant amount of energy and Nature would prefer to emit
574 Design of phosphors

6.8.27.-

phonons if possible. But the energy gap is too large for this. It turns out
that photons are only emitted from large energy gaps. Note that other
rare earth ions such as Nd3* , which emit in the infra-red, are used in a
variety of lasers. Their ground states are generally at 0 energy. But, we
 6.8 Rare earths and their spectral properties 575

still do not know why the 6 H13/2 state of Dy3+ is the terminal state, rather
than 6 Hi5/2-

The composition of the three states, i.e.- emitting and terminal states, are
shown in the following:

6.8.29.- MIXED STATES FOR Dy3+ IN INTERMEDIATE COUPLING

4
EMITTING STATE F 9 / 2 = 64 4 F, 19 4 G,9 6 F, 4 2 G,3 6 H

6
GROUND STATES Hi3/2 : 96 6H, 3 4 I
6
H15/2 : 93 6H, 6 4 I

Note that the excited state has a 6 H component. It turns out that strong
transition intensities in the lanthanides occur due to a transition termed:
"forced-electric-dipole" transitions, where AJ = 2 . This rule holds in all
cases where strong intensities have been observed. Thus, for our case, the
transition involves the 6 H13/2 state, because of the AJ = 2 restriction.

We have already shown in Chapter 5 that spectral intensities, I, are

proportional to oscillator strength, f, which is inversely proportional to
half-life, TH , namely -

6.8.30.- I » f a 1/ T 4 f

Transition between 4f =s> 4f energy levels are strictly forbidden, according

to Hund's Rules, and are classified as quadrupole (qq) transitions. Their
intensities ought to be: I » 10~7, but experimentally, I « 1.0, i.e.-
intensities rivaling dd- transitions.

D. Spectroscopic Rules for the Rare Earths

It was Van Vleck (1958) who showed that mixed states, i.e.- intermediate
coupling, lead to "enforced dipole" transitions because of the above
described coupling of states. He described this phenomenon as arising
from lattice perturbation of the states wherein the odd (u) crystal field
576 Design of phosphors

terms become mixed in, thus making the transition "allowed". The
Enforced Dipole Selection Rules are then somewhat similar to L - S
coupling dipole transition rules. These rules are shown as follows.

6.8.31.- FORCED DIPOLE SELECTION RULES

Electric: AJ = ± 2 (rarely ± 1)
(0*0) (u%g)
Magnetic: AJ = +1 (rarely ± 2)
(0 * 0) (u # u)
Limits: AJ < 6

EXPERIMENTAL INTENSITIES OBSERVED

Electric: I « 1.0
Magnetic: I » 0.1 to 1.0

Using these rules, we can now reexamine the emission transition of Dy3+.
For example, a transition from the 4 F9/2 emitting state to the 6 F states is
strictly forbidden (0 * 0). The other possibility includes the 6 H states,
i.e.- 6 H13/2 , 6 H11/2 , 6 H9/2 , 6 H7/2 , and 6H5/2 . However, the 6 H7/2
and 6 H5/2 states have no 4 I composition. This leaves us with the 6 H13/2
and 6 H11/2 states. While it is true that in absorption (excitation) AJ can
equal up to six, the change in J for emission is generally limited to AJ = 2,
with lesser intensities for AJ =1. Because transition to the 6 Hi5/2 state
involves AJ = 3, it is not possible. This accounts for the observed
luminescent transition for Dy3+.

The determining factors for intense rare earths spectra (dipole

transitions) are:

6.8.32.- DETERMINING FACTORS FOR RARE EARTH SPECTRA

1. Mixed states of like parentage

2. Forced dipole selection rules
3. Symmetry at the lattice site
 6.8 Rare earths and their spectral properties 577

The last factor is rather easily explained. Suppose we put a Ln 3+ ion into a
site having inversion symmetry, i.e.- there are two or more mirror planes
present at the site. Lattice perturbation is then uniform at the site, and no
"odd" terms "become mixed in". Thus, the transition reverts to a (qq)
transition, with resulting low transition intensity. Many examples exist in
the literature that show that the emission intensity of Ln 3+ ions are
sensitive to the nature of the site symmetry of the host.

In general, low symmetry cation sites produce the most intensely

emitting lanthanide phosphors.

While we have examined calculated levels, as compared to experimental

ones, we have not determined exactly how the Stark States arise. Let us
now examine this phenomenon.

e. Experimental Stark States

Up to now, we have adjusted the spin-orbit coupling factor, t^{, to fit the
experimental energy levels. One problem with this approach is that while
most of the levels seem to fit, there are one or two which do not. If we
examine 6.8.24. in terms of the "dots" representing a fit of experimental
data, we see that correspondence is fairly good. However, the problem is
more demanding than this. Let us further examine the fit of these energy
levels, using Er 3 + because its energy levels are far enough apart so that no
discrepancy, or mistake, in identification of the individual levels can be
made. To do so, we set up a spectrophotometer so that we can measure
the reflectance of a powder containing the ion. We use the diffuse-
reflectance mode so as to avoid spurious spectral effects. With a specially
designed instrument, we find that we can obtain a resolution of about ± 2
A of the spectral diffuse reflectance lines (Ropp -1970). This is sufficient
to resolve the existing Stark States.

In order to minimize any contribution to the spectra from the host, we

use a phosphate compound, ErPO4 , because the PO4 group is transparent
through out the visible and ultraviolet spectrum, up to about 1750 A. The
578 Design of phosphors

following diagram, shows a high-resolution powder reflectance spectrum

of ErPO4, reduced from a 30-foot chart:

6.8.33.-

There are a number of bands which appear in absorption. Only a few of the
transitions are identified. However, it is clear that a number of individual
lines (Stark States) can be identified.

In this case, we have shown the calculated levels of Er3+ , according to

their J-value. Note that these levels are far apart, in terms of energy. This
corresponds to our experimental criterion given above.

The overall spectral transitions and their identification are shown in the
following Groatrian Diagram of Er3+, given as 6.8.34. on the next page.
 6.8 Rare earths and their spectral properties 579

6.8.34.-

At the right is shown the experimented halfwidths of the bands shown in

the main part of the spectra. Note that the correspondence between
580 Design of phosphors

experimental bands and calculated levels, in terms of the indicated

transitions, is only fair. That is, the location of the bands do not agree
entirely with the location of the calculated levels, nor do the half-widths
of the bands correspond with what we might think they ought to be. It is
not even possible to assign transitions to bands which are relatively wide
to two or more simultaneous transitions.

Let us now examine one of the bands at a much higher degree of clarity, in
a spectral region not hindered by overlapping bands, using a higher
degree of resolution in the instrument. The following diagram is that of
the 4 S3/2 band wherein the individual Stark States can be easily seen:

6.8.35.-

There are six (6) Stark States evident here, each having a half-width of
about 7 to 9 A. These have arisen because of the crystal field splitting, but
more importantly represent transitioiis between Stark states of the 4115/2
and 4S3/2 multiplets. This ion belongs to the so-called "Kramers Ions"
which consist of atoms having an "odd", i.e.- uneven ("u") number of
electrons.
 6.8 Rare earths and their spectral properties 581

The rules for crystal field splitting are:

6.8.36.- # OF STARK STATES POSSIBLE FOR ODD AND EVEN ATOMS

Even: Stark States = 2J + 1

Odd: Stark States = J + 1 /2

In our case, there are eight (8) Stark states for 4 Iis/2 and two (2) for
4
S3/2 . Thus, we should observe sixteen (16) transitions between these
states. But, we see only six (6), due to crystal field restrictions upon
coupling between ground and excited Stark States. The same restrictions
apply for the other transitions as shown in the following:

6.8.37.- Transition Stark States

Possible Actual
4
Il5/2=>4F7/2 32 10

4
Il5/2^2G9/2 40 7

It is therefore evident that only a part of the Stark Lines appear in the
spectrum of a given Ln3+ ion , incorporated in a given crystal. The above
lines are those of Er 3 + in ErPO4. If we change the host, so as to change
the crystal site symmetry, we find that the number, and position of the
Stark Lines change, but we never obtain all of the possible Stark States.
In fact, if we sum all of the different Stark Lines present in a number of
crystals, we can find about 24 of the possible 32 in one case, and 32 of the
40 in another case, but never in a single spectrum. If we wish to identify
the energy of the 4Ii5/2=> 4S3/2 transition, we find that what we must do
is to identify the middle point, or baricenter, by weighting the positions
and strengths of the energy absorption lines, thereby identifying the
"middle" of the band.

The following, given as 6.8.38. on the next page, shows the baricenters of
two such bands wherein the baricenters are directly compared to the
calculated energy levels. Here, AE = Eexperimentai - E ca i cu i ate d- What we
have established is that the exact fit of calculated to experimental energy
582 Design of phosphors

6.8.38.- COMPARISON OF EXPERIMENTAL BARICENTERS TO

CALCULATED ENERGY LEVELS AS A FUNCTION OF SEVERAL
HOSTS AND SITE SYMMETRIES

Compound Site Baricenter

2
Symmetry IE7/2 AE G9/ 2 AE
ErBO 3 DM 4854 A 27 A 4070 53 A
Er 2 Ti 2 O 7 Oh 4892 65 4077 60
ErPO 4 D2d 4910 83 4083 66
Er 2 O 3 C2v 4920 93 4101 84

levels is nearly impossible to accomplish. Although the fit is close, it still

is not exact. This presents us with a dilemma since we want to be able to
identify and specify energy transitions in both absorption (excitation) and
emission. What we have done so far is to adjust the spin-orbit coupling
factor, XAi. to fit the experimental energy levels so as to obtain a set of
energy levels.

Nonetheless, this approach has not been successful because of the factors
that we have presented above. It has proven impossible to exactly match
the energy levels of a given trivalent lanthanide in any given crystal host
because of the effect of the crystal field on the number of stark states that
can appear. Fortunately, another approach has arisen.

f. The "Free-Ion" Approach to Rare Earth Energy Levels

It turns out that another method exists where the approach has been to
calculate so-called "free-ion" levels, so that the spin-orbit coupling factor,
XAi and the crystal field parameter, |4f, are not adjusted to give a
"satisfactory" fit to the experimental data. Wybourne (1963) and others
have accomplished this task by using a Hartree-Fock approach (and a high
speed computer) to calculate the energy levels of all of the trivalent rare
earth ions.

These levels are shown in the following diagram, given on the next page
as 6.8.39. In general, these levels show many similarities and only a few
 6.8 Rare earths and their spectral properties 583

6.8.39.- Calculated Free-Ion Energy Levels of the Trivalent Rare Earths

584 Design of phosphors

discrepancies to most experimental spectra. Still, they represent a better

method of identifying energy transitions because they are self-consistent,
and not dependent upon the crystal in which the rare earth finds itself.

Before we leave this subject, we should point out that the Ln 3+ ions which
produce the brightest phosphors are those which have the largest energy
gap between excited and ground states. These are in general: Eu3+ , Gd3+
and Tb 3+ , i.e.- 4^, 4f7 and 4I 8 . These trivalent states also have the most
complicated energy levels. This was first demonstrated by Thomaschek
(1933), as shown in the following diagram:

6.8.40.-
Relative Emission Intensities of the Trivalent Rare Earths

Excitation of these ions using 40,000 c m 1 energy results in various

excited states for each ion. Each state relaxes by phonon processes until
the terminal excited state is reached, from which emission of a visible
photon results. But we need to remember that any time we put a trivalent
rare earth ion into a host crystal to form a phosphor, we will modify its
free-ion energy levels, depending upon the site symmetry and crystal field
strength.
 6.8 Rare earths and their spectral properties 585

g. Charge Transfer States and 5d Multiplets

Although we have investigated the 4f electron energy states, we chose to

ignore the possibility that other excited states might exist. It was Dieke
(1964) who first pointed out that three (3) overlapping spectroscopic
manifolds existed for the lanthanides. These are: the 4f -manifold, the 5d
manifold and the charge-transfer manifold. All three exist within the same
energy-space and it was Ropp and Carroll (1965) who first showed that
these could be measured and sorted on the basis of the half-width of the
measured band. But, it was Jorgenson (1962) who first showed that a
specific trend exists for the 5d states which is exactly opposite to the
trend shown for charge-transfer states in the trivalent lanthanides. The
equations he derived were:

6.8.41.- CHARGE - TRANSFER STATES

W - q(E - A) + Ki D + K2 E + K3 C4f

5d EXCITATION STATES

W2 + (q - 1)(E - A 2 ) + K4E 3 + K6 &f

where W and W2 are reference standards for comparison of energies, q is

the number of 4f electrons, (E - A) is the difference between core
attraction and interelectronic repulsion, D gives the change of q with
spin-pairing energy, E 3 , and the various K's are adjustment parameters.

When the proper values are plotted vs the number of 4f electrons, the
following curves result, as shown in 6.8.42., given on the next page.

It is easy to see that the lowest energy 5d-excited states are exhibited by
Ce 3+ and Tb 3 + . The lowest energy charge-transfer excited states (which
are also a function of the host crystal) are Eu 3+ and Yb3+. All the others are
intermediate in energy to those extremes of the two curves. In addition,
we find that charge- transfer states occur for those trivalent ions which
586 Design of phosphors

6.8.42.-

also exhibit a divalent state while 5d states are prevalent for those ions
which also have a tetravalent state.

In 1967, Ropp and Carroll demonstrated that the trivalent rare earths
followed the trends given in 6.8.42. for both the 5d states and the charge-
transfer bands. It is well to note that the following trivalent ions exhibit
specific excitation bands in their spectra, and that these bands will be the
dominant feature in the spectra being measured. That is, within the
spectral range of 1800 A to 4000 A, , which is the usual range measured
for excitation spectra of phosphors being developed for lamp purposes.

This is summarized as follows:

 6.8 Rare earths and their spectral properties 587

6.8.43.- 5d Excitation States Charge Transfer Excitation states

Ce4+ , Tb 3+ Sm3+ , Eu 3+ , Yb3+

As an example of a charge transfer band, the following shows the

spectrum of the phosphor, Y2O3 : Eu, given as follows:

6.8.44.-

Since we cited Anti-Stokes phosphors as one area dominated by rare

earths, we need to address this technological field. We will find that these
phosphors are unique and only made possible by the unusual energy levels
encountered in these 14 components of the Periodic Table.

h. Phonon Assisted Relaxation in Phosphors

Because the energy gaps for many of the trivalent rare earths are often no
more than 2000 c m 1 and many of the Stark States are as close as 200
588 Design of phosphors

c m 1 , phonon-assisted relaxation (emission to, and absorption from, the

host lattice) may be expected to predominate in certain cases. While this
is certainly true in all phosphors, the close-spaced energy levels of the
rare earths allow us to experimentally examine such electronic processes.

Examine the diagram, given as 6.8.40 above, showing calculated free-ion

energy levels again. For the most part, the energy levels are rather closely
spaced. When an upper energy level of a rare earth ion in a crystal is
excited, it may decay to a lower state by:

1) Emission of a photon
2) Energy transfer to another site by a multipole process
3) Emission of several phonons to the host lattice.

When we say phonon-emission, we are actually stating that infra-red

photons matching the phonon spectrum of the lattice are emitted and
that lattice absorption then occurs- this is equivalent to virtual photon
emission since the emitted photons never appear outside of the lattice).
The last process has been studied extensively and we shall summarize the
results herein. The rare earths are unique in that, for energy gaps of
several thousands of wavenumbers, several simultaneous phonon emission
processes may be involved.

Multiphonon relaxation processes are usually studied by determining both

the transition rates for the process, and the phonon spectrum of the host
crystal. Total transition rates (reciprocals of fluorescent lifetimes) are
measured and multiphonon (MP) decay rates are extracted from these. In
practice, one does not observe the MP transition rate between two energy
levels, but that between two J- manifolds, i.e.- the sets of Stark States The
decay rate, W, obeys the relation:

6.8.45.- W= Bexp(aAE) = Bexp(a/t<On)

where k = h/2n and n is the number of phonons of a given frequency,

while con , B and a are constants characteristic of the host lattice. The
temperature dependence of the transition probability (decay rate) for
 6.8 Rare earths and their spectral properties 589

processes involving the emission of n-phonons of a single frequency, en, is

given by:

6.8.46.- Wn = Wo[[l- exp (-/too/kT)]-n

Note that this equation is like a dispersion equation (which we have

already discussed). Phonon relaxation rates follow these equations only at
low temperatures, i.e.- 4,2 °K.

The phonon spectrum determined for YVO4:Eu for the absorption

transition, 7 F 0 => 5 D i , is shown in the following:

6.8.47.-
590 Design of phosphors

The zero-phonon line is also indicated, along with its relative intensity of
4340. This spectrum was obtained at 77 °K, by measuring the absorption
of the 7 F 0 => 5 Di transition, while monitoring the 5 D 0 => 7F2 emission
transition. The phonon lines shown are essentially vibronic side-bands. If
we now assume that the rate of adjustment among the individual Stark
States of a J-multiplet is much faster than the relaxation of the multiplet
itself to the next lowest energy multiplet, i.e.- the lifetime of the
emission, we can determine the actual number of phonons involved in the
process {refer to 5.5.7. in Chapter 5). What we do is to arrange 6.8.46.
into a Bose-Einstein occupation probability equation, namely:

6.8.48.- n = [(exp (-/tto/kT) - I]" 1

Combining this with 6.8.46. gives the form most useful for determining
the multiphonon (MP) rate of the J-multiplet:

6.8.49.- WMp = I Wo g (exp [-dm/ kT)/ £ g (exp (-600/ kT)

We determine the experimental rate of phonon decay "W-fop , for several

levels of various ions (activators) in YVO4 and then plot them versus the
energy gap. For the selected ions, we have the transitions:

6.8.50.- Eu 3 + : 5
Di => 5 D 0 + n /too =>7F2 + hv
5
D3 => 5DO + n h co =>7F2 + hv

Er 3+ : 4
S3/2 => 4Fg/2 + n fim =>4Ii5/2 + hv
4
Ill/2 => 4 Il3/2 + n /tco =*>4Ii5/2 + hv

Ho 3+ : 5
S2 => 5F5 + n h co => 5 Is + hv
5p 3 ^ 5g 2 + n ft co => 5 l8 + hv

The phonon emission rates are seen to fit a straight line as shown on the
following diagram, given on the next page as 6.8.51.
It is easily seen that the smaller the energy gap between levels, the larger
is the phonon rate.
 6.8 Rare earths and their spectral properties 591

6.8.51.-

This is as we might expect since it should be easier for the phonon energy
to span smaller energy gaps.

To complete the picture, we now need an estimate of the temperature

dependence of the rate of phonon emission in YVO4. The following
measurements show experimental data obtained by Reed et al (1971) for
the temperature dependence of the transition: 4S3/2 => 4Fg/2 of Er 3+ in
YVO4 . This is shown in 6.8.52. on the following page.

Note that when one increases the temperature in a lattice, the intensity of
vibrations, i.e. density of phonons, increases correspondingly.
592 Design of phosphors

6.8.52.-

One can fit this curve above 50 °K by a function like 6.8.49. by using the
data given in 6.8.50. We thus obtain:

6.8.53.- Transition: 4Sct/9^>4Fq/9

•WMP = 5.00 x 10 4 [n(270) + l] 2 [n (880) + 1]3/ 1+3 exp (-7OO/kT)

where n(270) in 6.8.53. means the phonon-mode occupation probability.

This result shows that two 270 c m 1 phonons and three 880 c m 1
phonons are involved in the relaxation process.

We can now summarize the numbers of phonons involved in the relaxation

processes for all of the lanthanides given above in 6.8.50., as shown in the
following, given as 6.8.54. on the next page.

It should be clear that the sum of the energies involved the controlling
factor here. Note that a flve-phonon process is indicated in two of the
 6.8 Rare earths and their spectral properties 593

6.8.54.- PHONON EMISSION TO THE LATTICE BY RELAXATION

PROCESSES FOR VARIOUS ACTIVATORS IN YVO4

Ion Phonon Emitting n co_ Actual Energy Gap in

State Crystal

Eu3+ 5Dl 2 880 cm-l 1758 c m ' 1

5D3 3 880 2860
1 271

Er3+ 4S3/2 2 271 3091

3 880
4
Ill/2 4 880 3464

Ho 3 + 5
S2 1 880 2914
4 456
5F5 2 880 2108
1 456

transitions shown. These are phonons emitted during relaxation of an

upper excited state to the photon-emitting state. It has been determined
that a six (6)- phonon process represents the upper limit for any given
relaxation process. This probably arises because of physical limitations.
That is, the lattice can only absorb only so many phonons during the time
that the relaxation is taking place.

It should be clear that we have now confirmed that phonon absorption and
emission from the activator site actually occurs. We have alluded to this
fact in prior chapters but no specific data had been presented which
actually confirmed this mechanism. The above discussion has proven this.

Now that we have examined the case where the excited center emits
phonons to the lattice, let us now study the opposite case, where phonon
absorption from the lattice occurs, i.e.- a phonon assisted electronic
transition. In this case, phonon absorption occurs so that the mismatched
energy transition can take place.
594 Design of phosphors

As should be apparent by now, the rare earths are one of the few cases
where such Stokes processes can be studied at room temperature. If we
were to study phonon processes using other activators, we would find that
we must study them at 4.2 °K. We would also find that the phonon spectra
were rather complex and difficult to interpret, instead of being simple.
Now we will address the so-called "Anti-Stokes" phosphors where the
dominant mechanism is absorption of several phonons to produce a single
photon of higher energy. We will find that only certain trivalent
lanthanides are suitable when combined with a selected host lattice if we
wish to have high efficiency, i.e.- "brightness".

g. Anti-Stokes Phosphors

The unique energy levels of the rare earths have been utilized to produce
Anti-Stokes phosphors. These are essentially "infra-red to visible" light
converters. Two, or more, infra-red photons are absorbed to produce one
visible photon. Reexamine the energy levels of Pr 3 + , Nd 3 + , Ho 3+ and Er 3+
again. All of these ions absorb in the infra-red region of the spectrum. If
we started with two ions, say Er 3+ , each of which absorbed an infra-red
photon of 6600 c m 1 , we would end up with two ions in the 4 Ii3/2 excited
state. By an exchange process, one ion could then donate its energy to the
other to produce one ion in the 4Ig/2 state, while the other reverted back
to the ground state. This exchange process would look like the following,
as shown in the diagram given as 6.8.55. on the next page.

The exchange process may be direct, or may occur by reabsorption of an

emitted photon. A three-body process for Er 3 + would bring us up to the
4
S3/2 state, from which we could expect a visible photon to be emitted.
But the probability of photon emission from the intermediate states, 4113/2
or 4Ig/2 , is just as high as that for the exchange process. Therefore, it is
not surprising that we see little infra-red to visible light conversion in
such a system. That is, photon emission in the infra-red is equally likely to
that of the exchange excitation process. Remember, the exchange
Hamiltonian depends upon the coupling coefficient, Jeff , times the
overlap operators, Sy. However, Jeff is only effective for nearest
neighbors.
 6.8 Rare earths and their spectral properties 595

6.8.55.-

It drops off in relation to 1/r3 , the radius of the ion in the crystal. A
three-body exchange process has too low a probability to be of practical
value in an Anti-Stokes phosphor. If we reexamine the calculated free-ion
energy level diagram, we notice that Yb3+ has a single energy level with no
adjacent levels to which relaxation could occur. Therefore, if we add this
ion to our material, we should be able to obtain a phosphor which
functions as an Anti-Stokes material. The energy level of Yb3+ lies at
10,300 c m 1 (9710 A). It just happens that the output of a GaAsrSi laser
diode occurs at 9500 ± 500 A.

The "up-conversion" (Anti-Stokes) phosphors have been used in solid

state indicator lamps incorporated within various electronic equipment
for several years before being supplanted by the visible-emitting GaAlAs:Si
solid state diode.

A large number of hosts were studied in which Yb3+ and Ln 3+ were

combined. Obviously, the phonon spectrum will be critical to the overall
operation of up-conversion phosphors. Some of the hosts studied are
given in the diagram shown on the next page as follows:
596 Design of phosphors

6.8.56.- CaWO 4 BaF2 Y2O2 S LaF 3

NaY(WO4)2 SrF2 YF 3
YVO4 CdF 2 BaYF5

It t u r n s out t h a t E r 3 + performs best with Yb 3 + in any given host. A sample

composition would b e : (Y0.79 Ybo.20 Ero.01) F3 . Note t h a t t h e ratio of Yb
to Er is 20:1 so a s to maximize t h e energy exchange process between t h e
two activators.

After considerable work, it w a s determined that t h e hosts which w o r k

best in this application are those in which vibronic coupling is minimized.
This w a s established by determining a n d comparing t h e p h o n o n
dispersion b r a n c h e s of t h e various c o m p o u n d s . It w a s found t h a t t h o s e
hosts which have low energy optical branches in t h e p h o n o n s p e c t r u m
function best for up-conversion phosphor applications. Note t h a t this is
akin to minimizing ground state perturbation at t h e activator site by
proper choice of host. The best hosts were found to b e :

6.8.57.- Compositions Used for Infra-red Converting P h o s p h o r s

Hosts Activators
Oxyhalides: YOF Er3+ - green
YOC1 - red
Y3OCI7

Halides: L11F3 Tm3+ - blue

BaLnFs - infra-red
LiLnF4
NaYF4

where Ln = Y, Gd, La and Lu. Note that the host-ions are "closed-shell"
cations (Gd3+ = 4I7) and conform to our rules for choice of host. Four
hosts proved to be most useful:

6.8.58.- YOC1 YF3 Y3OCI7 NaYF4

 6.8 Rare earths and their spectral properties 597

The following gives a comparison of measured properties, based upon the

hosts:

6.8.59.- MEASURED PROPERTIES OF ANTI-STOKES PHOSPHORS

Color Phosphor Emission Output Efficiency

(watts/watt input)
Red Yo.74Ybo.25Ero.10Oa 6600 A 1.0%

Green Yo.84Ybo.15Ero.o1F3 5500 0.11%

Blue Yo.65Ybo.35Tmo.ooiF3 4700 0.011%

Infra-red Yo.65Ybo.35Tmo.ooiF3 8100 0.031%

Note that the green phosphor output is lower by a factor of ten, compared
to the red phosphor. But, its actual brightness to the eye is still 24 times
that of the red phosphor. An interesting property of the
Y2,74Ybo,25Ero,QiOCl7 phosphor is that its emission is "tunable". That is, it
can be varied from green to red by varying the excitation intensity.

At low power levels of the Y2t74Ybo,25Ero,oiOCl7 phosphor, the emission is

predominantly green, 4S3/2 => 4Ii5/2. At higher power levels, the red
emission, 4Fg/2 =* 4115/2 predominates.

One can also change the composition of Y3OCI7 so as to cause one

emission color to predominate:

6.8.60.- Red Green to Red Green

Y=0.59 Y=0.59 Y=0.91

Yb=0.40 Yb = 0.20 Yb = 0.08
Er = 0.01 Er = 0.01 Er = 0.01

The reason that these hosts produce good infra-red converting phosphors
is that their phonon-cutoff energy (the maximum energy of the optical
598 Design of phosphors

branches of the phonon spectrum) is very low. This arises because the
atomic masses of the hosts are relatively low.

If we measure the phonon spectrum as deduced from vibronic coupling,

we get the following:

6.8.61.- Host Color Phonon Cutoff Energy

(for Er 3+ ) Calculated Measured

4
Transition = S 3 / 2 =>4115/2

NaY(WC-4)2 green 360 c m 1 350 c m 1

LaF3 green 360 350

YF3 green 360 375

NaYF4 green 360 405

4
Transition = Fg/2 => 4 Ii5/2

Y3OCI7 red 580 600

Y2O3 red 520 550

YOC1 red 570 620

With these phonon energies, i.e.- less than 400 c m 1 for green emission
and less than 600 cm-1 for red emission, the multiphonon relaxation
process is nearly impossible because of the energy difference between
multiplets would require more than a sextet of simultaneous phonon
emissions. Therefore, these hosts exhibit a high efficiency for infra-red to
visible light conversion.

The following diagram, given as 6.8.62. on the next page, shows the lower
part of the Er 3+ and Yb3+ energy level diagrams.
 6.8 Rare earths and their spectral properties 599

6.8.62.-

You will note that the two Er 3 + levels, 4 I n / 2 and 2Fs/2 , are very closely
matched in energy. Irradiation at 10,300 c m 1 will excite either of these
centers. Deexcitation by phonon emission to the lattice would take, for
these hosts, a 7 to 9 phonon process for Er 3 + ( 4 Fn/2 - 4 Ii5/2 ). and a 16
to 25 phonon process for Yb 3+ ( 2 Fs/2- 2F7/2). Both are obviously physically
impossible as a process.

Keep in mind that if both Er 3 + and Yb3+ become excited and if Yb3+
transfers its energy , the final state of the Er 3 + ion will be 4F7/2. This state
then relaxes to the emitting state, 4S3/2 , from which comes a green
photon (about 18,400 c n r 1 in energy).

Alternately, two Yb3+ ions can sequentially transfer energy to the Er 3 + ion
which then follows the same energy path.
600 Design of phosphors

The following diagram shows the excitation routes that have been cited
for both green emission from Er3+ and for blue emission from Tm 3+ in
Anti-Stokes phosphors:

6.8.63.- Transitions Associated with Green and Blue Emitting Phosphors

Note that the case for emission of a blue photon by Tm 3+ is slightly

different than that of Er3+ . A three-step process is indicated. This may
consist of excitation of the Tm 3+ ion and energy transfer from two excited
Yb3+ ions, or excitation of the Tm 3+ ion by exchange (sequential
excitation) with three excited Yb3+ sites. Note that two relaxation steps
(phonon emission to the host lattice) would be involved.

The excitation routes for red emission (energy = 15,200 cm 1 ) are not as
clear. Four (4) different paths have been postulated for excitation from
Yb3+ to the 4Fg/2 level of Er3+. None of these have been proven to
dominate. The whole process for these 4 postulated mechanisms for red
emission can be summarized as shown in 6.8.64., as follows:
 6.8 Rare earths and their spectral properties 601

6.8.64.- The Four Postulated Mechanisms for Red Emission in YOC13

It shows the details of the four paths stated that have been postulated for
excitation from Yb3+ to the 4 F 9 / 2 emitting level of Er3+.. Note that these
mechanisms appear quite complex but actually involve only a few phonons.
The exact mechanism remains today not fully defined.

Two relaxation steps are shown in (a), and one in (b). In (c), a three -
photon (i.e.- infra-red photons) process is shown, with corresponding
relaxation steps. Once the Er3+ center has reached the 2G7/2 state, an
exchange process (shown by the serrated line)with an unexcited Yb3+ ion
may take place, leaving the Er3+ ion in the red-emitting state, 4F9/2 . In
(d), still another possible Er3+ excited-state-exchange process is shown,
wherein the terminal excited state is 4Ii3/2 . which then goes to the red-
emitting state, 4Fg/2 , by a final exchange process. Thus, (d) is a variation
of (c). The whole process for (c) and (d) can be summarized as shown in
6.8.65., given on the next page, where the "stars" indicate the number of
individual excitation-energy steps involved in the exchange process.
602 Design of phosphors

6.8.65.-
Er + 2 Yb + 3 h c vi ^ 2 Yb* + Er*
Yb* + Er* => Yb + Er* *
Yb* + Er* * ^ Yb* + Er* * *
Yb + Er* * * => Yb* + Er* *

Er* * => Er + h c V2
{h c vi = 10,300 c m - l ; h c v 2 = 15,200 c m 1 }

We can now compare the relative efficiencies of Er 3 + for green emission,

by up-conversion of infra-red radiation from a GaAs:Si laser diode, as a
function of the host employed. This is shown in the following:

6.8.66.- Relative Efficiencies of Anti-Stokes Phosphors Using E r *

Host Lattice Relative Green Emission Intensity

a- NaYF4 100%
(3- NaYF4 40
Y3OCI7 105
YF3 60
BaYF5 50
LaNaF4 40
LaF3 30
La2Mo06 15
LaNbO4 10
NaGdO2 5
La 2 O 3 5
NaYW2O8 5

Note that those hosts which have anions with low atomic numbers (and
weights) perform best for Anti-Stokes conversion, i.e.- they have the
lowest phonon-cutoff frequencies. There is a structure effect which may
be ascertained by a comparison of a- NaYF4 and p- NaYF4 Actually, in p-
NaYF4 , the red emission is 100% in intensity, not the green. One would
 6.8 Rare earths and their spectral properties 603

suspect that differences already given for Y3OCI7 are due to symmetry
changes at the E r ^ site, as affected by the surrounding Yb3+ neighbors.

One should also note that these phosphors were used for a time to
enhance the output of LED light sources. With the advent of vastly
improved LED components, they have been superseded and are not used
anymore. We will address the LED's currently used and the applications to
which they have found employment. The improvement has been such that
LED's are likely to replace both incandescent lamps and fluorescent lamps
in the near future, provided that a suitable phosphor can be found to use
with them.

The final subject we will explore is that concerning solid state lasers, the
spectral basis for establishing a solid state laser, and their properties.

h. The Solid State Laser

The solid state laser is actually a specialized phosphor. It consists of a

host and an activator, but the host must be single crystal. What we have to
say about the solid state laser applies equally to all lasers, be they gaseous
or solid state.

Consider the properties of a photon in more detail. We know that it has

both particle and wave-properties. We have determined that if the wave
vibrates in one plane, i.e.- the direction of propagation, we have plane-
polarized light. It may also vibrate in a circular motion, transverse to its
direction of propagation, and then we have circularly polarized light. We
also know that that if these photons are reflected from a surface, the
polarization is destroyed, probably because each photon ends up on a
different plane of vibration of the lattice. This effect arises because the
atoms of the solid are vibrating, and the reflected waves will all have a
slightly different orientation from each other, due to the reflection-
interaction, i.e.- the Helmholtz scattering mentioned in Chapter 5.
Furthermore, if they are not all of the same energy, we know that some
photons will be scattered, and some absorbed and/or reflected (a non-
resonant process).
604 Design of phosphors

If an assemblage of photons have identical energies, and if they all vibrate

in the same plane, we say that we have "coherent" light. But, if they have a
wide disparity of energies and vibrate at all possible orientations of the
waves, we say that we have "incoherent" light. Most light sources produce
incoherent light. It is the laser which produces coherent light.

You may now wonder what is meant by coherent light. Consider the
"Particle in a Box" in Chapter 5, which we used as an introduction to the
calculations of Quantum Mechanics. The only energy states which appear
for the electron are those for which "standing waves" appear as a function
of "1", the length of the "box". To generate coherent light, we set up
exactly the same conditions in the form of a resonant cavity. All
commercial lasers are based upon this principle, namely that the heart of
the laser is this cavity.

Return now to the generalized phosphor: MXO4: A. Let us further suppose

that the activator has a transition of : ^ o => J Pi for excitation. We now
know that the ground state electronic wavefunctions are symmetrical,
whereas the lP\ state will have a vectored orientation due to the nature

of the p-electron orbital, as shown in the following diagram:

6.8.67.-

To this point, we have been concerned with phosphors consisting of many

small particles, each of which lies at a different orientation to all of the
others. This results in incoherent emission. But, if we now grow a single
crystal containing the activator, we would find that, upon excitation of the
crystal phosphor, the emitted light is now polarized. The plane of
polarization has a specific relation to the crystallographic axes of the
 6.8 Rare earths and their spectral properties 605

crystal. By changing the host crystal, we further conclude that the excited
states at all activator sites of the same symmetry relax and maintain a
specific orientation relative to the lattice, just prior to emitting a photon.

For the *Pi state, the polarization of the emitted photon is normal to the
orientation of the lobes of the excited p-electron state. For a uni-axial
crystal, the photons emerge along a lattice direction consistent with the
direction of minimum electric field of the atoms of the structure. This
tells us that the *Pi state has relaxed to an orientation at the cation site
consistent with the electric field orientation of the crystal field gradient
present at that site in the crystal. The photon is emitted normal to that
orientation, resulting in plane-polarized light. But, because the crystal is
vibrating, perturbation of the excited state still results in a band of
emission energies, as determined by the phonon spectrum of the crystal.
Thus, we can conclude that having a single crystal will result in plane-
polarized emission. There is still another requirement to achieve a solid
state laser. We have already examined the wave properties of an electron
in "The Particle in a Box" in Chapter 5. Therein, we showed that only
certain energy states were possible, because resonance effects. Because of
wave-vector addition and subtraction effects, only certain energy wave
states exist. What we find is that if we place our single crystal in such a
"box", we can set up a similar resonance state, so as to produce light
waves with restricted frequencies. Such a box is called a "resonant cavity".

Obviously, the resonant-cavity must conform to a frequency mode of

emission of our crystal. To build such a cavity, we find that we must
carefully control its physical dimensions so that a specific wavelength of
emission will build in intensity and control the wavelength of coherent
emission. To do this, we insert mirrors (first surface) at both ends of the
crystal so that many reflections of a photon can occur through the crystal,
as illustrated in the following diagram, given as 5.8.68. on the next page.

This requires that we polish the ends of the crystal, and that the mirrors
are of the highest optical quality that we can obtain. One mirror must
reflect all of the light back through the crystal, while the other reflects
all but 0.2 % of the light back through the crystal.
606 Design of phosphors

6.8.68.-

Under such conditions, any single photon emitted must have been
reflected through the crystal many times before it finally escaped. We
might now expect to obtain coherent emission, but we do not. This
situation arises because we have an activator which emits a broad band of
energies, and although we can restrict the emission so that coherent
photons result, the intensity (and efficiency) is so low as to make the
whole device almost useless. Actually, we get only about 1 coherent
photon out of about 10,000 generated. The rest of the energy ends up as
heat. Although we could cool our single crystal phosphor to minimize
vibronic coupling, the extremely low quantum efficiency of the device
precludes any practical usage.

This leaves us with two other possible choices. One, we can change to a
rare earth where the energy transitions involve line emission, or two, we
can choose an activator having a relatively long half-life in its excited state,
so as to be able to build up a sizable population of activators in their
excited state, before they are able to relax and emit a photon. If such a
single-crystal phosphor is then placed in the resonant cavity, the
resonance condition then precludes the emission of any photons other
than those that conform to the resonance condition of the cavity. That is,
the resonance wave builds in intensity and "stimulates" each excited
activator site to revert to a specific vibronic state of coupling, dictated by
the cavity, from which the stimulated emission emerges It is this
 6.8 Rare earths and their spectral properties 607

mechanism, which is called "stimulated emission", that is the basis of the

solid state laser. Our conditions for a solid state laser include:

6.8.69.- Conditions Requisite for Construction of a Solid State Laser

1. Construct a resonant cavity to match the chosen emission

2. Achieve a minimum spread of emission energies
3. Long half-life of the excited state
4. Achieve an inverted population, i.e.- of excited states over
ground states, so as to obtain stimulated emission.

Of these conditions, the most important is #4 in which the activators

centers must become excited and remain in the excited state until
"stimulated" to emit under the resonant conditions to emit stimulated
emission.

Rare earths as activators fit the first three criteria very nicely so as to
produce condition (4). Remember that we are considering the solid state
where the activator has essentially one degree of freedom, i.e.- vibration.
Thus, if the activator has a relatively long half-life, has a narrow band of
emission energies, and is incorporated within a host where the excited
state electron wave states are all in the same site symmetry, then we will
be able to achieve stimulated emission.

Gas lasers, on the other hand, have three (3) degrees of freedom so that
each excited gas molecule can instantaneously orient itself to the
resonance condition and thereby achieve population inversion, and
coherent light emission. The last condition we find essential if we are to
obtain any power in the beam.

It turns out that the central rare earths, which exhibit strong enforced-
dipole emission, do not perform well as laser activators, where a long half-
life is essential to obtain a population inversion. Population inversion
requires an excitation source strong enough so that the number of
exciting photons absorbed are sufficient to bring the population to the
point where there are more activator centers excited than in the ground
608 Design of phosphors

state. This means that there is a threshold of input energy, above which
we obtain laser action, i.e.- stimulation of coherent light. Below this point,
we obtain polarized emission, but not stimulated coherent emission. The
best rare earth for laser application has been found to be Nd 3+ which
emits at 1.086 fi (10,860 A). Other rare earths which have found some
usage are: Er 3 + and Tm 3+ . These ions function best in fluoride-based
materials. The hosts which function best as laser crystals are few indeed.

The criteria required are shown as follows. Of these four, the most
important ones (for high powered lasers) is #1 and #4.

6.8.70.- REQUIRED CRYSTAL PROPERTIES FOR LASER CRYSTALS

1. High thermal conductivity

2. High site symmetry at the activator site, i.e.- a cubic crystal
3. An activator site having inversion symmetry so as to
preclude enforced dipole transitions.
4. Be thermally stable under high power levels of excitation
and emission.

The following table, presented on the next page, lists the single crystal
hosts now being used for solid state lasers.

Table 6-13
Solid State Hosts Used to Construct Lasers

Y3AI5O11 - Yttrium Aluminum Garnet (YAG)

Yttrium Lithium Fluoride - YIiF4
La2BeO4 - Lanthanum Berylate (Alexandrite)
Yttrium Vanadate - YVO4
A12O3 :Ti3+
TiO2

with Nd 3+ as the activator for YAG, YVO4 and YLiF4 , and Cr3+ as the
activator for La2BeO4. A YAG crystal is tough, has a high thermal
conductivity and can withstand very high power levels. Although La2BeO4
 6.8 Rare earths and their spectral properties 609

is not so thermally stable, it has the advantage that the Cr 3+ activator can
be tuned among several wavelengths. YVO4 shows the highest efficiency as
a host. Also used was: (Y, ScteGasOn , but it does not have the thermal
stability of YAG.

The performance of a solid state laser is measured as "slope efficiency".

Above the threshold, an increase in excitation energy density produces a
like increase in laser power output. While phosphors may approach 90%
in quantum efficiency, lasers are no more than 10%. The rest of the
energy input is dissipated as heat. This means that we must cool the laser
cavity. Slope efficiency is measured as an energy conversion figure. A good
solid state laser will have aslope efficiency between about 0.5 to 5.0 %.
This means that if we use an excitation source of 100 watts in energy, the
best that we can expect is about 5.0 watts output The familiar gas laser
based on neon (Ne) has a typical output of 0.5 watts (500 milliwatts) for a
100 watt input. A major limitation of the solid state laser is the power of
the excitation source. Originally, Nd 3+ was used because it is well-excited
by the radiation emitted by the incandescent lamp. But, the inefficiency of
the tungsten filament for conversion of energy to light is well known, and
it is necessary to cool the laser cavity because the major portion of energy
conversion occurs as heat.

The overall energy losses in a laser cavity are given s follows:

6.8.71.- Energy Conversion Efficiency

Light source 10%

Crystal absorption 20%
Optical conversion losses 50%
Quantum efficiency 10%

By adding these numbers together, we get, for 100 watts input, about 100
milliwatts of output energy, or about 0.1% efficiency. However, the energy
is in the form of coherent light.
The most recent "advances" in solid state laser technology has been the
use of solid state laser diodes to "pump" the 4 Gn/2 and 2 Gn/2 states of
610 Design of phosphors

neodymium at about 5860 A. It is these states which exhibit the strongest

absorption bands in the overall Nd 3+ absorption spectrum. They are
directly coupled to the emitting state, 4F3/2, and the laser transition is
4
F3/2 => 4 Ii3/2 . having an energy of 10,580 A . In a side-pumped design
that uses a fiber-lens to couple light from diode-laser bars into a block of
YAG:Nd (size - 5.9 x 0.8 x 0.3 cm), a power input of 11.3 watts to the
block produced 3.5 watts of coherent emission, i.e.- 31% efficiency.

Let us now examine some of the mechanisms for achieving laser action in
a crystal. Under conditions of thermal equilibrium, the number of atoms
in any electronic state is proportional to a Boltzmann factor, so that for a
total of N atoms, a particular state, N{ , would be defined by:

6.8.72. - Ni/N = exp (- E4 / kT) / £ exp (- E4 / kT)

where Ei is the energy of that state. We can now define:

6.8.73.- coy = spontaneous transition rate

Wy = induced transition rate

where i is the initial state and j is the terminal state. Actually, we ought to
multiply these rates by the appropriate Einstein coefficients, Ay , the
transition probability of spontaneous emission, and By , the transition
probability of induced absorption. The total number of states is:

6.8.74.- N T = N i + N 2 + N 3 + N 4 + N{ =1.0

If we have true equilibrium, then:

6.8.75.- Ni/Nj = exp (- hv/kT) and Ni coy = Nj CDJI

where coy / COJI = exp ( - hv/ kT)

Let us now define the energy levels of a laser. From the ground state,
energy absorbed transforms the activator to an excited state. In a laser,
the excited state can relax before or after reaching the emitting state.
 6.8 Rare earths and their spectral properties 611

The relaxation step is needed so that our emitting state will have an
appropriate lifetime before stimulated emission takes place. In other
words, we cannot use a two-level laser, i.e.- excited state and ground
state, because the excited state will not possess a lifetime consistent with
that required to achieve an inverted population. Thus, a two-level system
is fine for spontaneous emission (fluorescence) but not for a laser.

What we use is either a three-level or four-level energy system to obtain

stimulated emission, as shown in the following:

6.8.76.-

Note that the levels are numbered. 1 =*• 3 is the excitation transition while
3 => 2 and 4 =*> 1 are relaxation transitions. Let us take the simpler case of
the three level laser. Let N i be the number of activator sites in the ground
state, N3 in the excited state, etc. We can then set up a series of
differential equations involving coy and the number of sites changing per
unit of time. This is shown as follows in 6.8.77., presented on the next
page.

In this case, relaxation takes place from a level close to the ground state.
And, we have specified rates in terms of Einstein probability coefficients.
This allows us to determine under what conditions the threshold for
stimulated emission can be reached. Note that in our three level diagram
relaxation from Level 2. involves a phonon-assisted transition or a phonon
emission to the lattice.
612 Design of phosphors

6.8.77.-

At a steady state of equilibrium:

6.8.78.- dNi/dt=dN2/dt=dN3/dt = 0

But during the initial excitation stage, if we increase the excitation

intensity, hvi3 , i.e.- the density of photons, pvi 3 , then the number of
excited activator centers will be a function of:

6.8.79.- dN 3 /dt= N 3 [B13 pvi 3 ] - N 3 A 3 i - N 3 A 3 2 + N 3 [ B13 pv 23 ]

Obviously, we want Bi 3 » A31 and A32 » B23 , so as to build up a

population at State 3. This will occur if we pick our activator with energy
levels far enough apart and with a sufficient half-life of State 3. to prohibit
the transitions: 3 => 1 and 2 => 3. Likewise, for the other states, we can
write:

6.8.80.- dN 2 /dt= N 3 A32 - N 2 [B23 pv 23 ] - N2A21 + N 3 i Bi 2

dNi/dt = N 3 A31 + N2A21 - Ni[ B13 pvi 3 ]

In the last equation, we have included an exchange process, Pi 3 pvi 3 .

This tells us that we must control the number of activator centers per
cubic centimeter of crystal host or we will get a 2 =*• 3 transition by
exchange of resonance energy or by direct absorption of radiated energy
 6.8 Rare earths and their spectral properties 613

(and a lower efficiency laser crystal). Under steady state conditions,

wherein only spontaneous emission is involved, we can solve for the ratios
of the steady state populations for fluorescence:

6.8.81.- N3/N1 = B13 P vi3 /A31 + B12 N3/N2 = Ai 2 /A 32

Now, for a population inversion to occur, the following conditions must

hold:

6.8.82.- B13 » A32 - strong absorption, long half-life of excited state

A21 » A32 - rapid relaxation of terminal laser State 2.

This means that we shall have to choose our host-activator combination

carefully. It is for these reasons that only a few combinations of host-
activator such as YAG:Nd have been found to be satisfactory. To get an
efficient laser system, an inverted excited state population must exist in
order to get stimulated emission. If an oscillation is to build up, then:

6.8.83.- N 3 - N 2 / Ni = [B31 pv3i ] / [A13 + B12] {1 - (A32/A21)}

For this fraction to be high, the conditions of 6.8.78 must hold. This
means that we need a strong and continuous excitation source to obtain a
lasing action in our crystal.

Solid state lasers have become ubiquitous in our society. Pulsed YAG:Nd
lasers are used for welding of metals and plastics. Among the applications
are: medical apparatus catheters where the weld size must be very small;
fiber optics communication systems; and hermetic sealing of integrated
circuit packages.

Disc lasers based on YAG:Yb are used for welding and higher cutting
speeds than other lasers. It uses an optical fiber to deliver the cutting
power with improved beam quality and total power achievable. Fiber
lasers, which are glass fibers doped with Yb3+, are also being used in the
614 Design of phosphors

same applications. Their advantage is that the fiber is flexible and can be
placed in regions impossible to reach by other lasers.

In the next chapter, we will show how fluorescent lamp technology has
matured and the area of studies that are being pursued in an effort to
further improve or replace fluorescent lamps. We will also examine
display technology and the many new types of displays now available that
did not exist when the first edition of this volume was published in 1991.

SUGGESTED READING

lJ_"Some Aspects of the Luminescence of Solids"- F.A. KrOger, Elsevier

Publ., Amsterdam, The Netherlands (1948).
 =
II
I' . . i I,
CHAPTER 7
i I, 11. i I ,
I
M

Current Phosphor Device Technology

In the last chapter, we emphasized how phosphors are designed, the

parameters involved, and how phosphors can be measured. In this
chapter, we will present the latest devices (some are 10 years old or
more) and the phosphors currently employed in them. Included will be
the many phosphors which have been developed for particular display
purposes. We will also show how the light sources we now depend upon,
i.e.- incandescent and fluorescent lamps, are going to become obsolete as
the quality of light-emitting diodes (LEDs) improves. Nonetheless, it is
clear that phosphors will continue to be used to make "white" emitting
LED light sources. The main advantage of the LED lamp is the fact that it
uses milliwatts of power, compared to watts of power for the
incandescent and fluorescent lamp.

7.1.- CURRENT PROGRESS IN CRT DISPLAY AND DISCHARGE LIGHTING

We will begin our survey of present-day devices utilizing phosphors by

first addressing recent improvements made in cathode-ray tubes followed
by fluorescent lamps and then other lighting devices. We will then
address the devices that depend upon thin film deposition for
manufacture of the appliance. All of these utilize phosphors in some way
for light output.

I. Cathode Ray Tubes fCRT)

In the last chapter, we presented a synopsis of the technological advances

made in CRT's, particularly those for the color TV tube up to about the
mid-1960's. Since that time, a number of innovations have been
accomplished. This change has been in response to market demand for a
less bulky display device as well as a larger display size, i.e.- the "flat"
television tube. Major improvements have been created in the electron
gun, deflection yoke and the glass faceplate. What we are addressing is the
changes in type of display which has resulted. In the ordinary CRT of the

615
616 Current phosphor device technology

past, the picture size was limited to about 27" diagonal because of the
deflection limitation of the electron beam. The first color-TV tubes used a
curved faceplate with around tube construction of no more than 18". This
was mandated by the shadow-mask which had to be carefully positioned so
that the three color phosphors could be properly excited by the electron
beam, i.e.- "color registration". The following diagram shows what we are
discussing:

7.1.1.-

Color registration became exacerbated as the shape of the faceplate

changed from round to rectangular. This remained a significant problem.
A rectangular tube was mandated by the semantics of the picture being
viewed. That is, people were accustomed to a rectangular picture like a
photograph and the round tube soon became obsolete. The innovation by
Sony who introduced vertical stripes on the shadow mask, with
accompanying phosphor lines, alleviated the problem somewhat. The
Trinitron™ method of vertical phosphor stripes has become the norm
today, along with the specialized electron gun needed to produce the TV
picture (see 6.6.4. of the last chapter). Improvements such as a "black-
surround" of the phosphor dots or stripes and darkening of the glass
faceplate were made to improve "picture contrast" as well. Nonetheless,
demand for larger TV displays continued.

In 6.6.1., we showed the typical CRT construction. What we did not state
was that the maximum angle of deflection of the electron beam was about
 7.1 Current progress in CRT display and discharge lighting 617

30° vertically. This limited the size of the image that could be generated.
The market demanded larger pictures. Work on the electron gun and the
deflection yoke resulted in much larger faceplates and TV pictures. This
is shown in the following diagram:

7.1.2.-

In this diagram, we have somewhat over-emphasized the curved glass

faceplate on the left. You will note that the edges of the TV picture would
be slightly distorted because of the curvature of the glass. This problem
persisted until the advent of the flat glass faceplate of the CRT shown on
the right of the diagram. This type of CRT was mandated for several
reasons, one of which was color registration and minimization of
achromatic aberration due to differences in refractive indices of air and
glass. The difficulties in obtaining such a color CRT cannot be
overemphasized.

Excerpted from United States Patents 5,536,995 (issued July 1996) and
5,964,364 (issued Oct. 2000- reissue 36,838- Aug. 2000), "Glass Panel for
a Cathode Ray Bulb & Glass Bulb for a Cathode Ray and a Method of
618 Current phosphor device technology

Producing the Same", the following is a description of the inventions

leading to this CRT innovation:

"In the production of a flat face-plate panel, a differential stress

(strain) pattern is imposed upon a glass panel with a glass-temper
factor less in the skirt than in the face, both in the longitudinal
mode and in cross-section (thickness) of the glass panel faceplate,
by controlled cooling, immediately after press-shaping a mass of hot
glass to a form, thereby obviating an annealing step, (which is
usually) followed by a tempering step. Compressive stress layers are
formed in the inner and outer surfaces of the skirt portion wherein
the compressive stress value of the face portion is larger than the
compressive stress value of the skirt portion and the compressive
stress value of the outer surface of the face portion is larger than the
inner surface. The so-formed glass faceplate is superior in impact
strength as well as having a reduced thickness of glass at the center
of the panel".

What this means is that the flat faceplate is a tempered-glass panel with
reduced thickness at the center and increased tempering at the face of
the tube. This configuration also reduced achromatic aberration at the
edges of the glass faceplate. It was determined that this configuration was
superior to the prior ones where the glass thickness was constant, as
shown in the following diagram:
7.2.3.-

The flat faceplate panel on the left, with a constant thickness, was found
to be subject to easy breakage during final fabrication of the color TV CRT
 7.1 Current progress in CRT display and discharge lighting 619

whereas the new configuration was much stronger. It also allowed the
construction of a color TV having superior color registration and less
color aberration when viewed under normal light conditions. As shown in
7.2.2., the maximum angle of the electron beam of the tube having a flat-
glass faceplate is near to 60°. Note that we are describing the vertical
angle, not the horizontal beam angle which is much greater. It should be
clear that the flat-faceplate CRT posed much greater problems in
achieving color registration because the "holes" in the shadow mask could
no longer be of uniform size. This mandated a major change in the
manufacture of shadow masks, which depended upon both size and shape
of the CRT viewing screen. In order to achieve such changes in electron
beam generation within the CRT, a number of improvements were made
in the electron gun, as well:
Table 7-1
Milestones in CRT Electron Gun Design
1. Change from "delta-shaped", i.e.- 0 0 to "in-line" configuration,
i.e.- 0 0 0 substantially improved the relative alignment of the
three individual guns, red, green and blue.
2. Realization that beam angle had significant effects upon spherical
aberration of color registration, magnification of beam diameter and
space charge increase at large angles, led to improvement in
electron-grids within each gun.
3. In-line gun design with a common lens (electronic) critically
reduced color aberration of the main lens and enabled most CRT's
to achieve the limits of system performance in which the color
registration remained even when picture contrast and brightness
controls were set at their maximum.
4. A design of a shunt/enhanced module in the guns enabled the in-
line color CRT to have perfect color convergence even at the
corners and edges of the screen.
5. Dispenser cathodes were designed to have a current-loading
density many times that of the conventional triple-oxide cathodes.
This enabled the CRT to run at much higher beam currents with
brightness and resolution much better than prior CRT's because the
electron beam could be focused to a smaller spot.
620 Current phosphor device technology

The current "high-definition" color-TV's incorporate all of these

improvements. However, major changes were made in how such
improved CRT's were used. Most of these CRT's were manufactured in
21", 27", 31" and 36" sizes for use in homes.

For larger sizes, the CRTs were "stacked" so that each CRT showed only a
part of the total picture. This is shown as follows:

7.1.4.-

Here, 16 CRT's are stacked to produce a large display picture. Many TV

stations use such displays and they are used in outdoor displays as well.
The flat-faceplates fit together seamlessly and no internal edges within
the picture can be seen by the viewer.

With the advent of rare earth phosphors, an entirely different approach to

a large screen TV became possible. Instead of generating all three color
screens in the same CRT, three separate tubes were used and their
images were combined on a projection screen so that a much larger
picture could be viewed. This is shown as 7.1.5., presented as follows:
 7.1 Current progress in CRT display and discharge lighting 621

7.1.5-

This made possible a viewing screen as large as 60 inches diagonally. Note

that an internal reflecting mirror is used to enhance the emitted light
coming from the CRT. Each color CRT is then combined with the other
two to produce the picture (which must be aligned for proper color
registration). The tube is operated under extremely high current density
and, in most cases, the faceplate needs to be cooled by flowing air. The
phosphors currently being used in cathode-ray tubes are:

7.1.6.-
Projection Tubes Color Television
Y2O3:Eu - red Y2O2S ; Eu - red
ZnS:Ag,Cl - blue ZnS:Ag, Cl - blue

The green phosphors include Gd2O2S:Tb and Y2SiO5:Tb. Because stability,

brightness output with current density and body color changes are a major
problem in projection CRT's, phosphors for this application have been
evaluated in terms of:

7.1.7.- Relative brightness (Br) = Ic v

where I c is the cathode current density and v is a figure of merit.

622 Current phosphor device technology

The following compares several phosphors currently used for projection

tubes, where the figure of merit has already been defined.

Table 7-2
Phosphor Color Fig. of Merit Br. Loss Rel. Stability
Y2O3:Eu3+ red 0.92 3% excellent
Y 2 O 2 S:Eu 3+ red 0.85 12% excellent
3+
Gd 2 O 2 S:Tb green 0.84 5% excellent
Y 2 SiO 5 :Tb s+ green 0.91 1.5% poor
Y 3 Al 5 O li: Tb 3+ green 0.91 2.0% very strong
Zn 2 SiO 4 :Mn 2+ green 0.85 34% very poor
CaS:Ce 3+ green 0.82 30% very poor
ZnS:Cu:Al green 0.67 37% inferior
ZnS:Ag,Cl blue 0.75 31% inferior

Brightness loss is that incurred over a period of operating time of the CRT
while Stability is a subjective measure based primarily upon brightness
loss and development of "body color", i.e.- a color developed by
decomposition of the surface layers of the phosphor particles (in many
cases, the cation of the lattice is reduced to a dark metallic state). You will
note that the best red phosphor is Y2O3:Eu3+ while the best green
phosphor is either Gd2O2S:Tb3+ or YgAlgO^Tb3*. Only one blue-emitting
phosphor has been found to be suitable for these CRT's. The phosphors
currently being used in Color Television tubes are:

ZnS:Ag,Cl - blue
ZnS:Cu,Al - green
Y2O2S ; Eu - red

II. Fluorescent Lamps

There are two areas where discharge lamps are currently used in
improved lighting applications: 1) Home lighting, and 2) Commercial
lighting (as in factories offices and medical facilities). The former uses
fluorescent lamps while the latter uses both fluorescent lamps and "high-
 7.1 Current progress in CRT display and discharge lighting 623

energy discharge" (HID) lamps. The latter are primarily high-pressure

mercury vapor (HPMV) discharge lamps. One design was shown in 6.6.5.
of the last chapter. We did not mention "metal halide" lamps which are
HPMV lamps with selected metals added to the discharge. One such lamp
was the Metalarc™ lamp which added: Indium, Thallium, Sodium and
Scandium lines to the 3650 A, 4060 A, 4380 A, 5470 A, 5770 A and 5790
A lines of the HPMV discharge. The resulting spectrum is shown on the
follows:

7.1.8.-

Note that these metals added the needed "redness" so that metal-halide
lamps now produce a "white" radiation in contrast to the "greenish white"
of the old HPMV lamp (you will note that, in the past, certain phosphors
were coated on the inside of the glass envelope to improve the "redness",
but this addition was never entirely satisfactory. The construction of the
HID lamp is shown in the following diagram, given as 7.1.9. on the next
page. Shown is the construction of a HPMV-HID lamp using a quartz arc-
tube. A significant advance in HID lamps has been the replacement of the
quartz arc-tube by a ceramic tube. Since the lamp runs hot, i.e.- 600 to
800 °C, quartz was never entirely satisfactory as a lamp envelope. The
ceramic construction is the same except that end-caps are sealed upon
the ceramic tubing whose composition is usually A12O3. Zirconia is also an
alternative.
624 Current phosphor device technology

7.1.9.-

This type of lamp was used extensively for lighting in warehouses, offices
and medical facilities until the advent of improved fluorescent lamps
which used much less power.

Major advances in fluorescent lamps resulted with the advent of rare earth
activated phosphors. In 6.6.6. of the last chapter, we showed the general
construction and operation of a fluorescent lamp. The primary fluorescent
lamp manufactured in the 1960's through the 1980's was the 40T12 lamp
of various colors (see 6.7.46. of the last chapter). Cool White was the
largest type of color purchased by consumers. 40 is the lamp wattage and
"12" is the diameter in 1/8" inches, i.e.- 1.5 inches in diameter.

The following table, shown on the next page, gives a comparison of the
various sizes of fluorescent lamps currently available.
 7.1 Current progress in CRT display and discharge lighting 625

Table 7-3

Lamp Length/ Est. Ambient Rated Lumens

Type Lamp Watts Diameter Life (hr) Op. Temp Lumens per Watt
T-12 F-40T-12 40 48"/1.5 20,000 77 °F 3200 80
T-8 F-32T8 32 48/1.0" 20,000 77 °F 2850 89
T-8 F-32T8 R 32 48/1.0" 24,000 77 °F 3100 97
T-5 F-39T5 28 46/.63" 20,000 90 °F 2900 103
T12HO 48T12HO 60 48/1.5" 12,000 77 °F 4250 71
T8HO 48T8HO 44 48/1.0" 18,000 77 °F 4000 91
T5HO 54T5HO 54 46/.63" 20,000 90 °F 5000 93

Whereas the first two lamps shown in the table used conventional
phosphors, i.e.- halophosphate mixtures, the rest of these lamps utilized
the tri-phosphor rare earth blends of three colors. These rare earth
blends provide 17 to 28% improved brightness over the traditional
phosphors, with high color rendition of 92% in most cases. The old
"high-output" = HO lamps were much lower in lumens per watt light
output and had much greater loss in brightness as a function of lamp life.
The small diameter lamps have led to lower wattage at about the same
brightness and lamp life. The tri-phosphors currently in use include:

7.1.10.- Tri-band Phosphors Used in Fluorescent Lamps

BaMgAl10O17: Eu2+ - BLUE

MgAlnOjg: Ce3+ :Tb3+ - GREEN
Y2O3: Eu3+ - RED

What has not been addressed is the operating characteristics of the two
types of lamps, i.e.- conventional halophosphate lamps compared to tri-
band rare earth lamps. Conventional lamps in 1950 lost up to 50% of their
initial brightness at 4000 hours of operation and lasted no more than
8,000-10,000 hours. 50% of these 40T12 lamps failed by 8000 hours due
to electrode problems. By 1990, these figures were improved to 75% at
626 Current phosphor device technology

4000 hours with about the same failure rate. In contrast, the F32T8 lamp,
with improved ballasts, maintained its lumen output at 98% at 4000 hours
operation and 91% at 30,000 hours. Lamp mortality was 1% at 4000
hours, 7% at 12,000 hours and 50% at 30,000 hours of operation. This, of
course, depended upon the conditions of operation where the lamp was
not turned off and on too many times. A 12 hour burn cycle per day was
used to evaluate these lamps.

Commercial lighting has traditionally used so-called "high-intensity-

discharge" (HID) lamps, i.e.- high-pressure mercury vapor discharge
lamps. As we stated in the last chapter, such lamps were improved by the
addition of selected metals to the discharge, resulting in "metal-halide"
HID lamps. But even these 400 watt HID lamps are being replaced by rare
earth fluorescent lamps. The following table gives a comparison of the two
types of lamps.

Table 7-4 - Loss in Brightness of Metal HID and Rare Earth Fluorescents
Hours of Operation
Lamp Watts 0 4000 12,000 20,000 30,000
Metal-halide 400 28,100 22,460 15,400 8,140
% Br. Loss 0 85% 72 % 68%
Mortality 3% 17% 50%
6- F32T8 RE 192 20,580 19450 17,970 16,570 8,370
% Br. Loss 0 98% 95% 93% 91%
Mortality 1% 3 % 7 % 50%

You will note that the table compares six 32T8 lamps to one HID metal
halide lamp. But, the wattage is about half for the 32T8 lamps. The total
output as a function of time is a function of the loss in lumens with time
and the remaining lamps. Thus, the 32T8 lamps lasted until over 20,000
hours when one of them failed. By 30,000 hours of operation (2500 days
@ 12 hours/day = 6.8 years), three of these lamps had failed.

It should be clear that the 32T8 lamps with rare earth phosphors perform
much better than the prior fluorescent lamps used for commercial
 7.1 Current progress in CRT display and discharge lighting 627

lighting. Part of this performance is due to improved ballasts.

Four types of ballasts have been, or are being, used: electromagnetic,

"rapid start" "instant start", and "electronic". Electromagnetic ballasts use
a "choke-coil" to suppress current while the filament- electrodes are
being heated to start the lamp. Once the electrodes are hot, the voltage is
switched across the lamp to start the discharge. This method is relatively
inefficient and was a major loss in fluorescent lamp operation. Rapid-start
ballasts still preheat the lamp electrodes at a lower voltage but are very
similar in control to electromagnetic ballasts. Instant start ballasts apply a
voltage high enough to directly start the mercury arc. Many fluorescent
lamp installations still use the "instant start" method of lamp control. The
electronic ballast is essentially micro-processor controlled and is about
30% more efficient than any of the other types. The program allows time
for filament preheat and then switches to a high frequency waveform of
25,000 to 40,000 cycles per second. Electronic ballasts can operate the
32T5 HO lamps more efficiently than any other ballast. The high
frequency eliminates lamp flicker and extends lamp life. Internal losses
are only 1-2 watts compared to 10-12 watts for the old electromagnetic
ballasts. The newest fluorescent lamp ballast uses micro-wave energy to
excite the lamp. Here, the mercury arc is maintained by electromagnetic
radiation having a frequency within the range of 1 gigahertz to 1 terahertz
(10 9 to 10 12 cycles per second) with a wavelength between 1 mm. and 1
meter. The lamp has an increased lifetime of up to 100,000 hours.

The newest lamp technology, the T5 fluorescent lamp, is having a major

impact on the lighting industry. The F32T5 lamp is brighter and more
visible (color rendering index = 92 +) and produces 30% more light than
the F32T8 lamp. They have become a viable alternative to HID lamps and
particularly the HID metal halide (HID-MH) lamps in commercial
installations since the cost of operating T5 lamps is about 50% that of the
HID lamp annually. Whereas the HID-MH lamps lose up to 45% of their
initial brightness over time, the T-5 lamps lose only 5%. The new
fluorescent technology is now preferred for lighting warehouses and
adjoining buildings, offices of all kinds and medical facilities. However,
HID and HID-MH lamps are still preferred for outdoor lighting because of
628 Current phosphor device technology

their ability to operate in and withstand a wide range of temperatures.

7.2.- CURRENT DISPLAY DEVICES BASED UPON ELECTRON CREATION

The next subject we will undertake will include those devices which
depend upon thin film technology for their manufacture. Since the
publication of this work in 1991, the use of thin film technology has
burgeoned. In order to appreciate exactly what has happened since that
time, we will address the many uses which have evolved for thin film
devices in relation to phosphor usage. Thin films are formed by many
methods and are used to make many devices that we now use in every day
life. Many of these devices were not available in 1991 or were too
expensive for many consumers. One good example is the portable cell
phone which uses integrated circuitry made by thin film techniques.
Another is the advent of flat display systems such as plasma displays. This
has made possible the flat television tube which can be hung on a wall
(although most people who have such a device usually set it up like an
"ordinary" television display). We will emphasize display technology since
this volume is meant to thoroughly describe the unique properties of
phosphors as "energy converters" for use in lighting and display devices.

In the past 12 years, a number of illuminating and display devices have

appeared in the marketplace. As we have already shown, only 2 types of
excitation can serve to excite solid state phosphors. These are the
electron and the photon. We have already presented the essential
differences between these two types of energy sources in the last chapter.
The following chart shows the current types of devices which depend
upon electron energy for phosphor excitation:

Note that we have shown that phosphor excitation can occur via one of
two paths, that of electron bombardment (cathode-rays) or by electric
field generation of electrons. The former has been further divided into
external (electron gun) and internal (electric field) generation of an
electron beam. Since we have already described color TV CRT's, we will
first address the other types of CRTs first and then internal electron
generation in selected devices such as the vacuum fluorescent display.
 7.2 Current display devices based upon electron creation 629

7.2.1.-

Finally, we will describe the properties of phosphors needed for devices

in which the electron excitation is generated by electric fields. This
includes LED's, Laser LEDs and electroluminescence (EL). We will first
address the CRT's used for specialty applications.
630 Current phosphor device technology

I. CRT's for Display. Measurement and Specialty Types

We have already addressed CRT's for television, including color, black &
white and color projection systems. Color display CRTs are used primarily
as computer monitors. Here, the same rare earth phosphors are used as
for color TV. Monochrome displays refers to CRT monitors where only
one color screen is used for special purposes, as described in the last
chapter. The following is a repeat of 6.6.6. given in the last chapter:

7.2.2.- Classification of JEDEC Phosphor Applications

Electrochrome display = 1 Oscilloscope display = 6

Radar display = 1 6 Television display = 6
Flying spot scanner = 8 Display monitor = 7
Storage monitor = 1 Photography display = 1
Light pen on display = 1 "Penetration" display = 1

Note that all of the classification of CRT applications of 7.2.1. are included
here.
a Oscilloscopes and Storage Tubes- Measurement

These CRT's are used to detect and analyze electrical signals and
waveforms. The same basic structure is used similar to CRTs employed as
color TV's except that the electron gun is highly sophisticated with
special electron optics. The oscilloscope CRT is required to have superior
beam deflection and sensitivity, linearity of response to voltage and
frequency changes and linear response to very high frequencies. The
latest oscilloscopes use a micro-channel plate (which is an electron
multiplier- that is, the signal from just one electron is multiplied many
times by secondary electron emission) to further improve deflection
sensitivity. The microchannel plate is placed just between the electron
beam and the phosphor screen to increase the beam current so that
frequencies above several hundred megahertz can be detected easily. Most
oscilloscopes are useful up to about one giga-hertz (10 9 cycles per sec.)
frequency.
 7.2 Current display devices based upon electron creation 631

The storage tube is similar to the oscilloscope CRT except that it has a
memory function fabricated within the CRT. There are two electron gun
components, a writing gun and a reading electron gun. The writing gun
uniformly scans the dielectric target which is placed between the gun and
the phosphor screen. The target consists of a gold-plated mesh on which
a thin layer (1-5 JJ) of MgF2 is deposited. A picture image of the signal is
drawn on the target by the writing beam and stored as a pattern of
electrical charges. That is, the target becomes charged only where the
writing beam strikes the matrix, which retains the image of the signal
being analyzed. Thus, quite fleeting signals can be detected. The
electrostatic charge patterns remain on the target and are reinforced by
the non-focused electron beam of the reading gun so that those areas
remain charged. The image is transferred onto the phosphor screen as
the reading gun maintains the charge. The excess is transferred to the
phosphor screen. Since the electrostatic image is not erased but
reinforced by the reading gun during operation, the stored image can be
maintained for extended periods of time until it is intentionally erased.
The image is updated every 1/30 of a second. Since the reading beam is
supplied on a continuous basis, the screen brightness can be maintained
at a high level. Thus, a storage tube has been used in aircraft cockpits for
so-called "daylight display".

Phosphors used for these applications are shown as follows and include
the following phosphors:

7.2.3.- Phosphors for Oscilloscopes and Storage Tubes

Oscilloscopes Storage Tubes

PI- ZnSiO4: Mn P20- ZnCdS:Ag, Cl
P2- ZnS:Ag, Cl
P31- ZnS:Cu, Cl

Since high resolution is required for these screens, they must be

extremely uniform in phosphor particle size. That is, the PSD must be
narrow in size and no more than 6-8 p in average size. You will note that
these devices still employ the old JEDEC phosphors. More recently, the
632 Current phosphor device technology

rare earth green-emitting phosphors, Gd2O2S:Tb, YgAlgOnrTb and

Y2SiO5:Tb have also been used. The major criterion has been the decay
time of the chosen phosphor since a time less than milliseconds is
required in the oscilloscope application.

b. Flying Spot Scanner

From the beginnings of commercial television, flying spot scanner CRT's

have been used commercially to broadcast still or moving images to
television sets in homes via an electrical signal. This form of television
pickup device, used to transmit motion picture images from film
transparencies, is called the flying spot scanner. The following diagram
illustrates the components required in the apparatus:

7.2.4.- Flying Spot Scanner System

Here, we show just the basics of the system. The light source is a cathode-
ray tube in which a beam of electrons, deflected in a standard scanning
pattern, impinges on a fluorescent phosphor surface. The beam produces
thereby a focused spot of light that moves in the scanning pattern on the
face of the tube. The light from this spot is focused optically on the
surface of the photographic film transparency to be televised. As the
image of the spot moves, it traces out a scanning line across the film, and
the amount of light emerging from the other side of the film at each point
 7.2 Current display devices based upon electron creation 633

is determined by the degree of transparency of the film at that point. The

emerging light is collected and caused to enter a photoelectric cell,
which produces a current proportional to the light entering it. For color
films, three separate color CRTs are needed, each with its own color
filter. The response of each color, red, green and blue is combined
electronically to form the color signal transmission.

Obviously, a fast-decay phosphor is mandatory for this application. Decay

times of nano-seconds, i.e.- 10" 9 second, are required for this scanning
mode to achieve sufficient detail since the film is moving at the rate of 30
frames per second. The phosphors used for flying spot scanners include:

7.2.5.- Phosphors for Flying Spot CRT Scanners

P24 -ZnO:Zn° = 2.8 {is. P16 - C^lVIgS^O^Ce = 120 ns.

P46 - Y3Al5Oli:Ce = 150 ns. P47 - Y2SiO5:Ce = 80 ns.
YgtAl.GakOnrCe =120 ns. YPO4:Ce = 45 ns.

One of these phosphors, YPO4:Ce, was developed by the author. Whether it

was ever used in the flying spot application remains unknown. Perhaps,
the fastest decay phosphor known to date is: Lu2SiO5:Ce. whose decay to
10% is 1.5 ns.

c. Radar Displays

Phosphors used in radar displays require the opposite of flying spot

scanner CRT's, i.e.- a long decay. Essentially, a beam of electromagnetic
radiation (whose frequency lies between 3 mega-hertz and 300 giga-hertz)
is swept in a 360 ° circle from the transmitting antenna. The reflected
radar signal is displayed upon a CRT screen. Since time is required to
rotate the antenna, a long decay phosphor is required. Radar systems may
be categorized according to the function they perform. These include:
aircraft traffic control, aircraft navigation, ship safety and navigation on
the high seas, surface (ground or sea) surveillance, tracking of moving
ground targets for law enforcement, weapon control, missile guidance,
instrumentation, remote sensing of the environment, intruder detection,
634 Current phosphor device technology

or underground probing. The most common type of radar signal consists

of a repetitive train of short-duration pulses. A ground-based radar system
with a range of about 50 to 60 nautical miles (or 90 to 110 kilometers),
such as the kind used for air traffic control at airports, would have a pulse
width of 1.0 psec. However, there would be 1,000 cycles within the pulse
which corresponds to a pulse repetition frequency of 1,000 hertz. In this
example, the average power is 1,000 watts (1 kilowatt). The average
power, rather than the peak power, is the measure of the capability of a
radar system. Radar systems have average powers from a few milliwatts to
as much as one or more megawatts, depending on the application. A weak
echo signal from a target might be as low as one trillionth of a watt (10
watt). In short, the power levels in a radar system can be very large (at the
transmitter) and very small (at the receiver).

Timing of the extremes encountered in a radar system is critical. An air-

surveillance radar (one that is used to search for aircraft) might scan its
antenna 360 degrees in azimuth in a few seconds, but the pulse width
might be about one microsecond in duration. (Some radar pulse widths
are 1,000 times smaller-i.e., of nanosecond duration.) The range to a
target is determined by measuring the time that a radar signal takes to
travel out to the target and back. Radar waves travel at the same speed as
light--roughly 186,000 miles per second. The range to the target is equal
to c times T/2, where c = velocity of propagation of radar energy, and T =
round-trip time as measured by the radar. From this expression, the
round-trip travel of the radar signal is at a rate of 150 meters per
microsecond. For example, if the time that it takes the signal to travel out
to the target and back were measured by the radar to be 600
microseconds (0.0006 second), then the range of the target would be 90
kilometers or 56 miles.

To preserve the reflection signal, a long decay phosphor is required. The

phosphors used in the past were generally composed of two phosphors,
one with a long decay. Thus, the screen exhibits two colors, that of the
target and the other forming the background which is refreshed in every
360 ° rotation of the radar antenna. Radar phosphors include:
 7.2 Current display devices based upon electron creation 635

7.2.6.- Phosphors for Radar CRT's

Phosphors Colors Decay Times

P-7 ZnS:Ag + ZnCdS2:Cu blue + yellow med.(57(is)+ long(400ms)
P-12 (Zn,Mg)F2:Mn orange long (210 msec)
P-14 ZnS:Ag + ZnCdS2:Cu blue + orange short(27ps)+v.long(500ms)
P-17 ZnS:Ag + ZnCdS2:Cu blue + yellow short(5.2|is)+v.long(429ms
P-19 KMgF3:Mn orange long(220 msec)
P-28 (Zn,Cd)S:Ag:Cu green very long (500 ms)
P-39 Zn2SiO4:Mn.As green med. long(400ms)
P-57 Zn,SiO4: Mn+MgF9: Mn yellow- green medium

All of these were given in Table 6-2 of the last chapter. More recently,
digital technology has replaced this type of CRT using long decay
phosphors. A computer-controlled display utilizing a full color CRT has
been developed for aircraft traffic control. Therein, the direction, size of
target and altitude of aircraft are directly displayed upon the phosphor
screen.

IV. Image Intensifier Screens

The image intensifier tube is one type of CRT but it does not utilize a
scanning electron beam. Image intensifiers can be separated into two
classes: 1) electronic tubes that incorporate an apparatus designed to
operate in very low light levels, and 2) phosphor screens intended to
detect and amplify radiation like x-rays. The former is used primarily as
"night-vision" goggles and relies upon electron emission from a
photocathode to amplify the available light as a picture. The latter is used
for x-ray tomography in medical establishments wherein the phosphor
screen is passive but responds to x-ray excitation.

The image intensifier tube, shown on the next page as 7.2.7., generally
has the construction shown. All internal spaces in the image intensifier
are evacuated to a high vacuum of at least 10"8 mm. to ensure that the
operation of the device is not degraded by residual gasses.
636 Current phosphor device technology

7.2.7.-

Note that an external lens focuses light upon a fiber optic having a
photocathode on the internal side of the fiber optic. A 300 VDC potential
channels emitted electrons directly onto the microchannel plate. What we
have not shown is the transparent electrodes between the photocathode
and microchannel plate and that on each side of the plate. This electrode
is usually In2O3-SnO2 and is formed by a sputtering process. This is the
first time we have encountered a thin-film in a device. The next devices
that we will describe utilize thin film processes for their manufacture and
we will address this method in more detail below.

As we have said, the microchannel plate multiplies each photo-electron by

causing secondary electron emission, thereby multiplying the cathode
current. Once the photo-electrons reach the phosphor screen via the
1000 VDC drop, the resulting emission passes through the viewing fiber
optic to form a picture. Since the incoming photons are kinetic, the
resulting scene is viewed in real time. Note that fiber optics are used both
in the front and rear of the tube. This ensures that the very low level of
photon excitation on the photocathode which forms the image remains
coherent so that the resolution of the picture viewed remains clear and
not blurred. This type of passive cathode ray tube can be reasonably large
or can be miniature as in night-vision goggles used world-wide.
 7.2 Current display devices based upon electron creation 637

The phosphors used for this type of CRT include:

7.2.8.- Phosphors for Image Intensifier Tubes

P-31 ZnCdS2:Cu:Ni green short (40 usec)

P-36 (Zn,Cd)S:Ag:Ni yellow-green very short(10 (is)

Because the image light intensity is often very low, these tubes use
phosphors which have emission close to the peak of the eye sensitivity
curve, i.e.- green or yellow-green. This maximizes the response of the
device for its intended application, namely night-vision under lack of
visible light. Note that some night-vision goggles also detect near infra-
red radiation as well. This is a function of the composition used for the
photocathode.

We will now summarize the techniques required to form thin films.

II. Thin Film Technology

The devices utilizing internal electron beam generation require the

formation of thin films in most cases. Although we have described the
growth of crystals in some detail in a previous chapter, we have not
considered the fact that we can grow single or polycrystalline crystals in
the form of a thin film. Such films can exhibit much different optical and
electrical properties from the same bulk crystal. This is particularly true
since we can form a series of stacks of thin films of varying composition
whose overall behavior will be a combination of those films comprising the
stack. It is the electronic properties of such materials that are useful in
industrial applications. Therefore, it would behoove us to consider the
factors that cause change in electronic properties of solids as a function of
structure and bonding. It is evident that this aspect is very important,
since the electrical nature of solids varies according to both composition
and arrangement of atoms in the structure. Some materials are semi-
conducting and others are dielectric. Still others are good conductors of
charge (metals). Still others have unique optical properties, due to
arrangement of electrons and atoms within the solid state. In addition,
638 Current phosphor device technology

the size of crystallites, i.e.- as "nano-particles" and their form, i.e.- as

"thin films" determine many of their optical and electronic properties.
There are many types of electronic devices whose properties are
dependent upon being able to employ thin film technology. These are
shown as follows:

7.2.11.- Electronic Devices Employing Thin Film Technology

1. Integrated Circuits
2. Tantalum Capacitors Used in Integrated Circuits
3. Transparent Conductors Used in a Variety of Devices
4. Light Emitting Diodes
5. Light Emitting Diode Lasers
6. Displays Including"
a. Field Effect Displays (FED)
b. Active Matrix Liquid Crystal Displays (AMLCD)
c. Electroluminescent Displays (EL)
d. Light Emitting Diode Displays (LED)
e. Cathode Ray Tube Displays (CR)
f. Organic Light Emitting Diode Displays (OLED)
g. Plasma Display Panels (PDP)
h. Liquid Crystal Displays (LCD)
7. Solar windows
8. Thin film Transistors
9. Anti-Reflection Coatings

Items 1, 2, 3, 7, 8 & 9 do not employ phosphors in their operation even

though they depend upon thin film methods for their formation. To
choose the proper thin-film coating process, one much consider the
following factors such as: material efficiency, edge effects, substrate
suitability and effectiveness, homogeneity, uniformity, particle defects,
and product robustness. All of these factors influence the choice of the
method to be used in forming such thin films. In general, the finished
thin film should have: well defined thickness, no voids, no pinholes, no
particle contamination or defects and a detailed homogeneity for the
particular application.
 7.2 Current display devices based upon electron creation 639

Thin films are made by a variety of techniques including:

1. Evaporation
2. Chemical-Vapor Deposition
3. Mechanical Film Deposition
4. Sputtering
5. Ion Beam Deposition
6. Electron Beam Deposition
7. Spin-Coating

We will not discuss the last method since it is merely a mechanical

procedure which uses centripetal force on a spinning disc to disperse the
slurry material. All of the other techniques require a substantial vacuum in
order to form the thin film and to prevent oxidation and contamination of
the film during its formation on a selected substrate. The history of
vacuum pumps covers many years. Both incandescent and gas-discharge
lamps would not have been developed and made a part of our life if the
vacuum pump had not been invented and made practical. Nor would we
have the color television if vacuum pumps had not been invented. All
three of these devices use thin films in one form or another. The past 12
years have accelerated the use of thin films for AMLCD's, OLED's, PDP's,
LED's and LCD's, to the degree that the vacuum pump is indispensable.

Since the formation of most thin films is dependent upon being able to
form a vacuum, we will first examine the use of vacuum pumps and how
they are used to form thin films. The vacuum pump has a long history. In
1855, H. Geissler invented the mercury piston pump which was widely
used. In 1865, Sprengel invented the mercury siphon pump. Using a
combination of these two pumps, a "good" vacuum of about 10' 6 torr could
be achieved (a millionth of an atmosphere- nowadays, we achieve 10"9 torr
routinely). Edison used these techniques in 1879 during his experiments
with carbon filament lamps. McLeod invented his "McLeod Vacuum
Gauge" in 1874 which was capable of exact measurements of internal
pressure in glass bulbs. The mercury diffusion pump was invented by
Sprengel in 1865 and made practical by Irving Langmuir of GE in 1913.
In 1926, C.R. Burch replaced the mercury with a low vapor pressure oil
640 Current phosphor device technology

but it was not until 1937 that an all metal oil-diffusion pump came into
practical use in industry. All diffusion pumps up to that time were made
from glass. One of the early major problems with oil diffusion pumps was
that the oil was seriously carbonized and degraded during operation of the
pump. W. Gaede invented the modern oil-sealed rotary-vane mechanical
vacuum pump in 1904. By 1910, electric motor-driven rotary-vane pumps
were in common use. Before that time, various types of pumps, including
the mercury diffusion pump, were very inefficient in terms of "pumping
power". The mechanical rotary-vane pump, operating in an oil-bath where
the spinning vanes were sealed, increased the pumping power by many
hundreds of times over the old vacuum-pumping systems. Rotary-vane
pumps became the "fore-pump" in 1916 as used by Langmuir and
Dushman of GE in their experiments on incandescent tungsten filaments.
The forepump lowered the internal pressure to about 10"3 torr at which
point the diffusion pump takes over to lower the internal pressure to at
least 10" 6 torr.

Mercury diffusion pumps were used widely from about 1920 to 1946
when they began to be replaced by oil diffusion pumps. Today, oil-diffusion
pumps use special oils in which the vapor pressure at room temperature
is 10"10 torr (polyphenol ether) in contrast to earlier oils whose vapor
pressure about equaled that of mercury at 10' 3 torr. Other types of vacuum
pumps in use today include: "Turbopumps" (which utilize a high speed
turbine for molecular drag from the pumping volume); "Cryopumps"
(which use low temperatures generated by liquid helium (- 268 °C) to
remove gases from the pumping volume) and "Getter" or "Ion" pumps
(which trap gas molecules by molecular bonding). Extreme vacuums have
been achieved up to 10"15 torr by using combinations of these pumps. At
this point, diffusion of helium gas through glass from the outside
atmosphere becomes the major deterrent to achieving lower pressures.
The gas pressure regions where these pumps are used are shown in
7.2.12., given on the next page.

Note that this list is not exhaustive. In all cases, a forepump (oil-sealed
rotary vane) is needed in addition to the specific high-vacuum pump.
 7.2 Current display devices based upon electron creation 641

7.2.12.- Types of Vacuum Pumps Needed to Produce a Specific Vacuum

Type of Rough= Med= High Vacuum = Ultrahigh=

Vacuum 1 to 1000 1 to 10"3 10" 3 to 10"7 10" 7 to 10 10
Diaphragm Vane rotary Turbomolecular Turbomolecular
Vane rotary Getter Vane rotary Sputter ion
Piston Oil diffusion Oil diffusion Cryopump
Diffusion Adsorption Adsorption Sublimation

We will now address the specific categories of methods used to produce

thin films. As we said, thin films are formed by:

7.2.13.- Methods Used to Form Thin Films

Evaporation - This is a technique where a material is heated in an

electrode to its boiling point in a vacuum. It works best with metals
and uses a tungsten coil as the electrode. A few low-melting
compounds can be used to form a thin film but decomposition
during melting has remained a serious problem. During evaporation,
all internal surfaces of the vacuum chamber are coated. Color TV's
and the like use a thin film of Al as a charge- dissipation mechanism
to prevent charge accumulation in the phosphor screen or the
shadow mask within the CRT.
Chemical-Vapor Deposition - This technique involves the use of
metal-organic compounds and the like which can be decomposed
upon a hot substrate. They are generally delivered to the hot surface
via a flowing inert gas. Molecular-beam epitaxy, commonly known as
MBE, is a form of vapor growth. The field began when the American
scientist John Read Arthur reported in 1968 that gallium arsenide
could be grown by sending a beam of gallium atoms and arsenic
molecules toward the flat surface of a crystal of the compound or a
similar material. The amount of gas molecules could be controlled to
grow just one layer, or just two, or any desired amount. This method
is slow, since the gas stream has a low density of atoms. Chemical
vapor deposition (CVD) was developed to replace MBE and is
642 Current phosphor device technology

another form of growth that makes use of the vapor technique. Also
known as vapor-phase epitaxy (VPE), it is much faster than MBE
since the atoms are delivered in a flowing gas rather than in a
molecular beam. Synthetic diamonds are grown by CVD. Rapid
growth occurs when methane (CH4) is mixed with atomic hydrogen
gas, which serves as a catalyst. Methane dissociates on a heated
surface of diamond. The carbon remains on the surface, and the
hydrogen leaves as a molecule. Liquid-phase epitaxy (LPE) is similar
and uses the solution method to grow a thin film on a substrate. The
substrate is placed in a solution with a saturated concentration of
solute. This technique is used to grow many materials employed in
modern electronics and optoelectronic devices, such as gallium
arsenide, gallium aluminum arsenide, and gallium phosphide.
Precursors like tri-ethyl gallium, arsenic hydride and phosphine are
generally used in the gaseous phase. The composition is controlled
by the rate of gas flow to the hot target substrate. CVD processes can
take place over a wide range of temperatures and pressures, from
600 °C to 1100 °C and from 760 mm to a few milliTorr. It has found
wide use because of its versatility.
3. Mechanical Film Deposition - In this technique, a doctor blade is
used to mechanically deposit a thin film (usually particles in a liquid
matrix) upon a surface as shown in the following diagram:

This diagram is a representation of the actual process. The actual

machine is usually much more complex. The particle slurry (like an
ink) flows upon the substrate but its depth is controlled by the
doctor blade. The thin film thus formed is then dried and then fired
to form a composite film. Obviously, any thin film thus formed must
 7.2 Current display devices based upon electron creation 643

be stable to heating in air or atmosphere. Reducing atmospheres

during firing are sometimes used.
4. Sputtering - This is a process in which atoms, ions, and
molecular species in the surface of a target material are ejected
under the action of an ion-beam irradiation. Energies typical of ion
implantation are employed and, while any ion type may be used, rare
gases such as argon and neon are most common because they avoid
unwanted chemical interactions between the ions of the beam and
the substrate. A simple apparatus is shown in the following:

In this case, the argon atoms are ionized and attracted to the
cathode having the material to be sputtered. Layers of material are
stripped off the surface of the target by the action of the kinetic
energy of the gas molecules. The released atoms and/or molecules
are attracted to the substrate to form a thin adherent layer. Note
that the material is not vaporized or ionized but is transferred to the
substrate with the same composition of the original material.
Sputtering results from several interaction mechanisms.
Conceptually, the simplest is rebound sputtering, in which an
incident ion strikes an atom on the surface of the target, causing it
to recoil. The recoiling atom promptly collides with a neighboring
atom in the target, rebounds elastically, and is ejected from the
surface. By means of any of these various mechanisms, several atoms
may be sputtered for each ion incident on the target. The number of
atoms sputtered per incident ion is called the sputtering yield. Note
644 Current phosphor device technology

that sputtering does not require a high temperature like evaporation

techniques to form a thin film. Materials like alumina (with a
melting point of 1840 °C) are easily sputtered. Even tungsten (MP >
3560 °C) is amenable to this method of forming a thin film.

The latest technique for sputtering is the use of a "magnetron"

which is a high frequency AC device operating at 10 to 100 mega-
hertz. The induced gaseous ions are controlled by a magnetic field
instead of an electric field (shown in the above diagram). Large
"Climate-Control" windows are coated by this technique. Sputtering
to form thin films has become one of the most common processes
used to do so.
5. Ion Beam Deposition Methods - This method involves formation
of an ion beam of high energy. Compared with thermal energy
processes, a 100 ev ion has a temperature equivalent to 1,000,000
°K. Ion beam systems have been used for: cleaning substrates, in-situ
cleaning, etching of low volatility materials, reactive etching, sputter
deposition, reactive sputter deposition, ion beam assisted coating of
evaporated films, CVD coatings and modification of surface
properties of materials before they are coated. All of these involve
thin film formation. The following lists some of the areas where ion-
beam thin-film technology has been employed:

Magnetron Sputtering Systems

Electron-Beam Ion-Beam Pre-clean Systems
Pulsed Laser Deposition System
Inert Ion Beam Etching
Chemically Assisted Ion Beam Etching
Reactive Ion Beam Etching
Dual Ion Beam System Configuration
Ion Beam Sputter Deposition
Ion Beam Coating System

These ion-beam systems have been used to manufacture:

superconducting films, thin film heads for electronic equipment,
optical thin films, ring-laser gyro-mirrors, dielectric films of all
 7.2 Current display devices based upon electron creation 645

sorts, encapsulation films of all sorts, optical anti-reflection coatings

of all sorts, interconnect films on integrated circuits, sensors of all
kinds, advanced magnetic heads for DVD players, x-ray optic
devices, laser facets, high reflectance mirrors and field-divisional
wave multiplexers.
6. Electron Beam Deposition- this technique uses an electron beam
instead of an ion beam. In this respect, it is not as versatile as the
ion-beam method or other methods used to manufacture thin films.

This completes our survey of the methods of achieving a vacuum, the

types of vacuum pumps available and the techniques used to form thin
films. We are now ready to continue our description of Devices which
utilize phosphors in their operation. This was shown in 7.2.1. given above
and includes displays that are not very prevalent at the present.

III. Devices Using Internal Electron Beam Generation

Whereas the devices described above used an external source or

mechanism to generate electrons for phosphor excitation, the devices
described in this section use an internal electric field to do so.

a. Vacuum Fluorescent Displays

The vacuum fluorescent display is a type of CRT except that the electron
beam is diffuse instead of being focused. Its construction consists of: 1) an
anode substrate, 2) an electrode assembly and 3) a front glass plate. One
construction is shown in the following diagram, given as 7.2.14. on the
next page.

In this diagram, we have shown one segment of the display which can
indicate only one number. The anode segment is made by evaporating
metal strips on the glass substrate (including the electrical lead-ins) and
then coating them with a suitable phosphor. Conductive leads are
attached, sometimes during the formation of the metal strips. The
cathode electrode is formed from a tungsten wire coated with tri-
carbonates. The whole is sealed together to form the display with the grid
646 Current phosphor device technology

7.2.14.- Construction of a Flat Vacuum Fluorescent Display

positioned between the two electrodes. The grid is present as an

electrode for controlling and accelerating the diffuse electron beam. We
have not shown the exhaust tube used to form the vacuum since several of
these segments would be joined together to form the actual display used
to indicate a set of numbers. Once the vacuum is achieved, the triple
carbonate is then heated to form the triple oxide coating which is the
active layer emitting electrons. When a voltage is applied between the
anode and cathode assemblies, emitted electrons are accelerated to the
anode segment connected at that moment, exciting the phosphor coating.
The complete hermetically sealed assembly is essentially a triode vacuum
tube with multiple anodes and directly heated cathodes. The usual
number of segments is no more than six. However, since alphabetic
characters can also be displayed, the total number of segments can be as
high as 10-12. Vacuum fluorescent display panels have been used since
 7.2 Current display devices based upon electron creation 647

1972 in cash registers, "point of sale" displays and other similar

applications. In operation, the filamentary cathode is heated to about 600
°C thereby releasing thermally-excited electrons which are directed
toward the number segment to be displayed. Since low voltages are
involved, i.e.- 2 to 50 VDC, the best phosphor for this tube is conductive
phosphor, ZnO:Zn°, which has a blue-green emission. Demand for other
colors resulted in modification of other CRT phosphors. The phosphor
characteristics needed for this device include:

1. Conductivity to prevent surface charging during operation

(this prevents electron excitation of the phosphor by negating
low voltage electron penetration of the phosphor surface and
generation of secondary electrons).
2. Low threshold voltage for emission.
3. Efficient luminescence by low voltage electrons without
emission saturation during operation.
4. Long lifetime stability without degradation during operation
from harmful materials from the hot cathode.

The first requirement resulted in mixtures of a conducting medium as

InSnO2. The mixture proved more efficacious than either of the
components. Phosphor mixtures such as ZnS:Au,Al + InSnO2 proved to be
much better than Y^SiEu^ + InSnO2. The latter mixture degraded to
almost zero after a few hundred hours of operation. The following table
shows a comparison of some of the phosphors used:

7.2.15.- Phosphors Used in Vacuum Fluorescent Displays

Composition Color Effic- L/watt Peak.nm Stability
ZnO:Zn° green 10.1 505 excellent
ZnS:Zn°, Cl + In9O3 blue 1.2 465 fair
ZnS:Cu,Al + In9O3 yellow-green 2.8 548 good
ZnS:Au,Al + In,O3 green-yellow 3.2 554 very good
ZnCdS:Au,Al + In,O3 yell.-orange 2.8 592 very good
ZnCdS:Ag,Cl + In,A orange 2.2 605 very good
ZnCdS:Ag,Al + In9O3 red-orange 1.8 662 fair
648 Current phosphor device technology

Note the rare earth based compositions are not represented. None of
these phosphors, except the ZnO:Zn° composition have been entirely-
adequate. It is, perhaps, for this reason that this display has become
nearly obsolete. With the advent of flat panel displays such as the
computer controlled plasma display or the back-lighted liquid-crystal
displays (which we will describe later), the vacuum fluorescent display has
been largely superseded. Such computer controlled displays can show the
item purchased, the amount of the sale and keep a running total for the
buyer.
b. Field Emission Displays fFEDi

The field emission display is a relatively new type of display device. It

relies upon "cold-cathode" emission from special electrode "tips" and is
aimed at the color TV market. As such, it has not yet found complete
acceptance and is still in the developmental stage. The following diagram
shows the general structure that has evolved:

7.2.16.-
Construction of a Field Emitting Display (FED)- Cross Section

Note that this device depends upon cold-cathode emission of electrons to

excite the phosphor dots. The field-emitting tips have been constructed
 7.2 Current display devices based upon electron creation 649

from a variety of materials but diamond (tetrahedral carbon) seems to be

the best material. It is hard, resistant to field effects and is conductive.
The methods used to form diamond tips have been many. There are four
types of carbon structures. These are based upon the bonding between s-
and p- orbitals of the carbon atom: diamond (sp 3 - tetrahedral), graphite
(sp 2 - two dimensional layers), linear (sp 1 - linear and polymeric carbons)
and amorphous (where a mixture of bonding types exist). The sp 3 bonding
type is typical not only for diamond but for aromatic carbon compounds in
general. The linear sp 1 type is typical for aliphatic and polymeric carbon
compounds. When pure carbon is used to form FED tips, it has been
found that the method used controls the type of field-emitting tip that is
produced:

7.2.17.- Field Emitting Tips and Type of Tip Produced

The tetrahedral diamond structure is preferable for several reasons.

However, the actual carbon tip structure is always a mixture of the three
s-p hybrid electron bond types. The closer that one can get to the pure
sp 3 tetrahedral diamond structure, the better performing are the tips
produced. This has remained a significant problem in the manufacture of
FED displays. The C-H in this diagram indicates that methane was used as
the precursor which was decomposed to a carbon layer, but that a certain
650 Current phosphor device technology

amount of hydrogen remained in the formed composition. This does not

affect the conductivity of the tips but the product is not as hard or
conductive as is the diamond-like composition. Nevertheless, the pure
diamond structural composition has not been achieved when thin-film
techniques are used to form the tip structure.

Note that an array of several thousand tips is used to excite each individual
phosphor dot or "pixel". Note that each tip in 7.2.16. is insulated from
the voltage control electrodes to prevent "cross-arcing". The extraction-
grid is voltage-biased to control the direction of electron acceleration.
The advantages of the FED are that it can be made in large display sizes,
even over 72 inches across, no scanning is involved, and that it is thin.
The disadvantages are the complexity of manufacture of the tips and the
extraction grid, as well as an upper limit to the accelerating voltage (< 6
KV DC) that can be applied between the two electrodes. At very high
accelerating voltages, internal breakdown (discharges and shorting
between components) can occur. Therefore, the FED is limited to fairly
low voltages such as a few thousand DC volts at most.

At such voltages, the phosphors that can be used are limited and
conductive phosphors seem to perform best. Although the type of
phosphor mixtures given in 7.2.15. for vacuum fluorescent displays have
been tried, none seem to be entirely satisfactory. The phosphors that have
been tried or are being used in the FED display are:
7.2.18.-
ZnO:Zn° - blue-green = 505 nm.
(Zn,Mg)O:Zn° - blue = 476 nm.
ZnGa2O4:Mn2t - green = 510 nm.
CaTiO3:Pr* - red = 610 nm.
Y2NbO4:Bi^ - blue
SrGa2S4:Eu2f - green
BaZnS^Mn^ - red

Of these, the brightest and most efficient phosphor is ZnO:Zn°, but not
even this phosphor is completely satisfactory since it saturates at 200 VDC
 7.2 Current display devices based upon electron creation 651

and beyond. The area where the FED has found to be advantageous is the
small display and the very large display. For military purposes, the "heads-
up" display, where the display is worn by the user, has been satisfactory
because the screen brightness does not have to rival that of the CRT TV
screen.
c. Large Outdoor Displays

Large outdoors displays such as a "billboard" constitute the last category

using internal electron beam generation as a method of phosphor
excitation. Actually, the type of display used for this purpose is one where
small individual cathode-ray tube's are employed. Each small CRT employs
a "Flood-Gun" approach to excite the phosphor screen. A standard screen
is about 35 x 24 feet in size. Since the individual tubes are about 1.125
inches in diameter, 38,400 tubes are required to assemble the display.
Note that these tubes are of a simple design and do not require to be
turned off or on continuously. Nor do they require a scanning raster of the
phosphor screen. Once the billboard picture is established by lighting
selected tubes, the picture remains until deliberately changed. The
advantage to this type of outdoor display is that it is easily viewed under
full sunlight as well as night-time conditions. Two types of Outdoor
Displays have been made available, one a mono-color and the other a
tricolor display.

The other competitor for billboard size displays is that of a modified FED
where the display utilizes millions of "tips" to form a pixel-display having
sufficient brightness for daylight viewing. This display operates
continuously. No scanning electronics or electrical connections are
required in contrast to that which is intended for color-TV viewing. Thus,
the large outdoor FED display has high brightness since the picture
remains constant and the emission tips operate on a constant emission
basis. Nevertheless, the CR tube outdoor display has seen more use
because of monetary reasons. However, the lifetime of the FED outdoor
display does not depend upon the tungsten cathode like the CR tube, but
upon the degradation of the emission tips (which is minimal during
operation unless they become "poisoned"). However, the ultimate usage of
652 Current phosphor device technology

these two types of outdoor displays will depend the relative "lifetime" of
each, a factor still has to be determined.

d. Light Emitting Diodes fLEDl

LED's became commercially available in the late 1960's with the advent of
the green-emitting GaP:N and the red-emitting GaP:Zn,0 diodes. GaP is
an indirect band-gap semi-conductor. That is, there is a "trap" that exists
near the conduction band that acts as a recombination center for light
emission. This mechanism is shown in the following diagram:

7.2.19.-

When a voltage is applied, electrons are forced through the p-n junction
where they are available for recombination with the holes. The N and Zn,O
centers in GaP are iso-electronic traps that provide very efficient
recombination centers for electron-hole pairs and subsequent light
emission. Nonetheless, the actual brightness for these LED's left
something to be desired in relation to other light sources. The only use
found for these devices was in signal lights on instruments or on kitchen
appliances in homes.

More recently, the advent of improved LED devices, emitting red, orange,
green, yellow and blue emission have become available. The progress
made is shown in the following diagram, shown on the next page:
 7.2 Current display devices based upon electron creation 653

7.1.20.-

All of these LED's are based on direct- band gap materials. Thus, LED's
covering the entire visible spectral range are now available. The AlGalnP
double-heterostructure LED's have been fabricated using a technique
where the GaAs absorbing substrate was removed and replaced by a n-type
GaP substrate which was non-absorbing. This made possible luminous
efficiencies that exceed unfiltered 60W tungsten bulb light in the yellow
green to red regions (> 15 lumens/watt). This advance opened the way for
use in traffic signal lights where the 60 watt bulb with yellow or red filter
was replaced by a 20 watt LED bulb. In 1996, an actual installation showed
a 38% energy savings and a 1.7 million watt reduction in electrical usage.
Lifetimes of such LED's exceed 50,000 hours while incandescent lamps
rarely exceed 1200 hours.

The LED's are grown by metal vapor phase epitaxy (MOCVD) on GaAs
substrates for matched lattice growth. A p-type GaP window layer serves
to form the semi-conductor. What we are addressing is shown in the
following diagram, given on the next page as 7.2.21. Typical compounds
used for MOCVD epitaxial growth include: tetraethyl silicate, trimethyl
654 Current phosphor device technology

metal chlorides, hydrides like arsine, diborane or phosphine, and tributyl-

gallium or triethyl-aluminum.

7.2.21.-

Originally, the LED shown above was grown by thin film techniques upon
a sapphire substrate. Nowadays, it is usual to use a neutral substrate, i.e.-
neither n- or p- doped, composed of the active lasing composition. We
have already addressed the MOCVD method of forming thin films. The
wavelength that any given LED will emit (in the case of the direct band-
gap materials) is predicated upon the band-gap, that is, the energy gap
between the valance band and the conduction band of the semi-conductor.
Table 7-1, given on the next page shows these electronic properties.

As can be seen, a wide variety of energy gaps exists for these semi-
conductors. However, it is the electron-mobility which is the deciding
factor in the selection of materials for LED's. That and their relationship
to the energy required for visible emission. These include: ZnSe, ATP,
GaP, AlAs, AlSb, CdTe, GaAs, and InP. However, AsAs, GaP and AlSb are
indirect band-gap semi-conductors and have not been found useful for
high brightness LED's.

The family of III-V and II-VI semi-conductors has a wide-range of lattice

constants and corresponding band-gaps. Those with smaller lattice
 7.2 Current display devices based upon electron creation 655

Table 7-1
Intrinsic Properties of HI-V and II-VI Semi-Conductors
Semi- Lattice Const Energy Gap - Electron Hole
conductor A electron volt Mobility * Mobility*
Si 5.4310 1.11 1,400 470
Ge 5.6461 0.67 3,900 1,900
GaP 5.4506 2.26 110 75
A1P 5.4625 2.45
GaAs 5.6535 1.42 8,500 400
AlAs 5.6605 2.17 280 ~ 20
InP 5.8688 1.35 5,000 150
InAs 6.0584 0.36 33,000 460
GaSb 6.0954 0.72 5,000 850
AlSb 6.1355 1.58 900 450
InSb 6.4788 0.17 80,000 1,250
ZnSe 5.6676 2.80 530 ~ 70
ZnTe 6.0880 2.20 530 130
CdTe 6.4816 1.49 700 60
2
* Mobility @ 300 °K in cm /Vsec.

constants tend to produce visible emission. Nonetheless, these "single"

compositions have not proven to be satisfactory in other respects such as
stability and change in refractive index with temperature.

It is for these reasons that the compositions producing the brightest

LED's are always a mixed composite. The following diagram, given on the
next page as 7.2.22., shows the wavelength range available as a tuned
energy for several composite LED compositions. We have also indicated
the visible range of radiation.

LED's are constructed from a p-type of semi-conductor and a layer of n-

type semiconductor, deposited upon a substrate of a similar semi-
conductor to form a p/n junction, as shown above. When voltage is applied
in a forward bias (positive to the p-side and negative to the n-side) which
656 Current phosphor device technology

7.2.22.-

is slightly above the voltage corresponding roughly to the bandgap energy,

current begins to flow and light is emitted. This recombination creates
photons with energies corresponding to the composite semi-conductor's
bandgap (Eg). This gives rise to the following relation:

7.2.23.- k = 1240 / Eg

Devices comprise layers of specific composition like AlGaAs, an alloy of

GaAs and AlAs in which the electronic bandgap and refractive index can
be varied through changing the alloy composition. GaAs has an Eg of 1.4
ev. so it produces wavelengths of about 8500 A. Changing the semi-
conductor layers to alloys of AlGaAs or InGaP, which have higher
bandgaps, created photons from the near-infrared (8500 A) to the red
(6300 A). Changing the semi-conductor layers to a lower bandgap material
creates 8500 to 16,000 A radiation. Only "direct-bandgap" materials
create radiation efficiently. The colors that can be produced by LEDs are
shown in the chromaticity diagram given on the next page as 7.2.24. Note
that the range of colors is larger than those covered by display
(Television) phosphors. What is not stated is that the brightness leaves
something to be desired. The current goal for LED's is to achieve an
 7.2 Current display devices based upon electron creation 657

7.1.24.-

brightness equivalent to fluorescent lamps whose output is close to 80-90

lumens per Watt. The major impetus has been the use of a PHOSPHOR in
conjunction with a blue LED which produces a "white" light. The
phosphor currently being used is YAG:Ce3+ whose excitation band overlaps
the output of the blue LED. Improvements in semi-conductor composition
and packaging have produced a chip of 1.0 by 1.0 mm with a "white"
brightness of 25 lumens using 0.350 amps, at 3.0 volts DC. A 5.0 watt
lamp produces 120 lumens or 20 L/W. The LED of just a few years ago
produced 1.5 L/W. The color- rendering index is just 70 but is expected
to be improved with the use of improved phosphors. White LEDs are
currently 15-20% more efficient in brightness than halogen or
incandescent lamps. The main advantages of LEDs, in addition to their
low power consumption compared to other lamps, is that they only lose
30% of their initial brightness after operating over 50,000 hours. Note
658 Current phosphor device technology

that we have already said that fluorescent lamps last only 4,000 hours at
most and that incandescent lamps last no more than 1200 hours. A major
innovation occurred in 1996 when researchers at Hewlett-Packard were
able to form a LED whose output outshone the light sources now used
commercially. They had grown a red-emitting LED using a GaAs substrate
with the appropriate buffer layers similar to that shown in the following
diagram:

7.2.26.-

The same composition shown in 7.2.22. was used except that the GaAs
buffer layer was etched, removed and replaced with a layer of GaP which
was transparent to the emitted red radiation. A change in brightness and
efficiency of 2.5 times resulted because the emitting layer was exposed to
a greater degree. This decreased the number of internal reflections that
occurred within the thin layer structure and allowed more light to escape.
Note that the yellow LED is now 3 times more efficient than the yellow-
filtered incandescent lamp and the red LED is 9 times more efficient than
its incandescent counterpart. High brightness LED's are currently being
used (or planned) in the following commercial applications, shown as
7.2.26. on the next page.
 7.2 Current display devices based upon electron creation 659

7.2.26.- Commercial Applications for High Brightness LEDs

- Aircraft lighting - Automotive taillights

- Display backlights - Traffic lights (colors)
- Indoor accent lighting and display
- Medical devices - Automotive headlights
- Lighting "Tiles" for indoor lighting

The current materials of choice are:

- AlInGaP : red, yellow or green

-GaN - green or blue

White LEDs are typically coated with the YAG:Ce phosphor that causes
them to emit a whitish light. The Ce3* activator in YAG has an overlapping
excitation-emission band, making it possible to use a blue LED to excite
this phosphor. Indeed, the YAGiCe3* phosphor is "daylight" excited. That
is, it luminesces in sunlight due to the blue radiation present.

Current phosphor research is aimed at obtaining an improved material for

use in producing a white LED. This will make possible LED tiles which
can be used in home lighting.

d. Light Emitting Diode Lasers

Although LEDs and laser diodes appear to be similar, in fact- they are
made from the same semi-conductor materials, they are very different in
that they are designed to behave in much different ways. The light from
diode lasers is coherent in contrast to the incoherent light emitted from
LEDs. Emission from LEDs result in a An pattern whereas the diode laser
emits a very narrow beam. As a result, diode lasers can be used to read the
pits on a compact disc or for scanning a barcode in a checkout line, jobs
that LEDs cannot perform.

The following diagram, given as 7.2.27. on the next page, shows a typical
construction of a laser diode:
660 Current phosphor device technology

7.2.27.-

Note that waveguides and mirrors are provided to form a resonant

chamber to generate coherent emission. When voltage is applied in a
forward bias (positive to the p-side and negative to the n-side) above a
voltage corresponding roughly to the active semi-conductor's bandgap,
current begins to flow. Holes injected into the n-side layer recombine
with electrons and electrons injected into the p-side recombine with the
holes (see 7.2.1. given above). Resonant photons are thus emitted if the
proper chamber is present.

First developed in 1962, laser diodes have achieved powers that allow
them to be used in applications not thought possible a few years ago. The
following table, presented on the next page, shows some of the laser
compositions, wavelengths and applications now used for these versatile
devices. Note that the bandgap of these materials could be calculated from
the equation:
X = 1240 / Eg

where "k is in nano-meters. This is the same equation already given in

7.2.21. You will note that diode lasers cover a wide range of coherent
emission wavelengths, from the visible to the far infrared. Also, the same
materials used as for LED's are employed here except that a resonant laser
chamber equipped with totally reflecting mirror and an output mirror of
 7.2 Current display devices based upon electron creation 661

99.8% reflectivity is provided to achieve the resonant condition required

for coherent emission of light.

Table 7-5
Semiconductor Laser Materials
Compound Wavelength in nm. Applications
GaN/AlGaN 448-490 Disc data storage
Zn(S,Se) 447-480 Disc data storage
ZnCdSe 490- 525 Disc data storage
AlGalnP/GaAs 620-680 Bar code scanners
InGaAsP/InP 1100-1650 Chemical sensors
In0.73^a0.27-^S0.58Po.52 1330 Optical communication
* I1 0.73t*O.27" S 0.9(r 0.10 1550 Metrology
AlGaAsSb /InGaAsSb 2000-3000 Chemical sensors
GaAs/AlGaAs 6000-12,000 Quantum cascade laser
PbEuSeTe 330-5800 High res. spectroscopy
PbSnTe 6300-29,000 Cryogenic spectroscopy

Over recent years, diode lasers have achieved higher output powers,
smaller sizes, higher efficiencies and increased reliability. They have
made possible abroad range of electronic devices not possible in 1991.

These include: "compact discs" (CD) for read and write devices for
computers, and playing music; laser printers; CVD (compact video discs)
for use with TV sets to play movies and the like; hand held compact
computers; fiber-optic switching chambers; and medical diagnostic
devices.

One main advantage of diode lasers is their small size. They can be used
singly or fabricated into a pattern for use as a display. One usage has been
a circular arrangement for use as traffic lights. Another has been use as a
single diode as a signal light. A typical construction is shown in the
following diagram, given as 7.2.28. on the next page. Here, we show a
"can" in which a single diode is installed. These can be quite small and are
replacing the neon-discharge lamps of the past. Diode lifetimes are in
662 Current phosphor device technology

excess of 20,000 hours, in contrast to incandescent lamps at 2,000 to

5,000 hours.
7 9 9.8 -

The can size can be quite small and no more than 1/8 inches across. This
allows use in device applications not possible before.

Diode lasers also appear in other configurations. One area is their use in
pumping solid state crystal lasers. The strongest absorption line of Nd3+
lies at 831 nm. That is, the YAG:Nd laser can be best excited at this
wavelength. It is relatively easy to configure a diode having this output
which would requires a bandgap of 1.5 ev. (See Table 7-1 and 7.2.20.).
The construction used is shown in the following diagram:
7.2.29.-
 7.2 Current display devices based upon electron creation 663

Here, we show how an array of diode lasers are arranged in a stack so that
they can be used to generate very high optical power. What we have not
shown is the details of the active layer comprising the individual diode
lasers. This is shown in the following diagram:

7.2.30.-

As you can see, a number of individual steps in the thin film process is
required to form the individual diode lasers in the array. Such processes
are used also to form traffic lights and the like. Note that a mask is
needed for each individual step.

Diode lasers have virtually replaced flashlamps and the like as "laser
pumps" for high power applications of solid state lasers including all solid
state lasers based upon single crystals. What we have shown are "edge-
emitting" diode lasers. There are also other configurations including
"surface emitting" diodes such as would fit in the "Can" shown in 7.2.28.
One such construction is shown in the following diagram, given on the
next page as 7.2.31.

Note that the laser emission occurs from the front surface, in contrast to
the edge-emission we have already illustrated. It is such configurations
664 Current phosphor device technology

7.2.31.-

that will make possible the use of a blue diode laser, combined with a
proper phosphor, to replace fluorescent lamps in home lighting. That,
and the phenomenon known as the "quantum-well", discovered around
1988 in relation to singular diode lasers. The front surface diode laser
also makes possible high brightness arrays in almost any configuration,
including around array suitable for traffic lights.

It was the "quantum well" structure that revolutionized the power output
of the diode lasers. This is shown in the following diagram, given as
7.2.32. on the next page.

You will note that we have only shown the so-called "quantum wells" of the
laser diode. These correspond to the "active layer" of 7.2.27. and 7.2.29.
That is, it was found that the "active layer" of the laser could be made
much thinner and could be stacked together in the same space as prior
diode lasers. It has been determined that, in very thin layers of semi-
conductors, i.e.- < 20 nm., quantum effects dominate as the electrons
become confined.
 7.2 Current display devices based upon electron creation 665

7.2.32.- Semi-Conductor Laser Diode Progress

This results in the conduction and valence band becoming quantized into
"sub-bands" or quantum wells. These strained layers result in a vastly
improved electrical efficiency as the bias current is raised in the diode
laser. Even a coherent single array has been improved markedly. The
coherent array of 1988 had been improved by 100 times in 2003 while
the semi- and incoherent arrays have been improved even more, up to 10 5
times. What this means is that instead of lining up the arrays as shown in
7.2.29., the arrays were arranged in a non-synchronous but cooperative
fashion (it does not mean that the diode emission was non-coherent).
This resulted in tremendous gains in laser power outputs.

However, a major part of this power improvement was also due to massive
gains in diode laser efficiency. Diverse commercial lasers, based upon
single crystals, are typically no more than 10% to 20% in efficiency, but
usually much less. It is typical for a crystal laser to have a "slope" efficiency
of 2-5% with outputs less than 10 milliwatts. InGaAsP diode lasers are
efficient by these standards with efficiencies greater than 20%. Strained
layer quantum-well structures have been measured with efficiencies in
excess of 85% at 1500 nm. wavelength. Typical threshold currents for
666 Current phosphor device technology

such diode lasers (the current needed to start the lasing action) is only
20-30 milliamps. Because of the thinness of the active lasing composition,
it was found that the electrons and radiation emitted "tunneled" between
the layers, thus multiplying the total output from a given current level
operation. The optical output was multiplied by several orders of
magnitude.

In the following table, we list the level of operating power that the several
applications require:
Table 7-6
Wavelength Application Approx. Power (Watts)
600 -1000 nm Illumination 0.04 to 0.20
Commercial Processing 0.25 to 0.95
Medical Devices 3.0 to 25.0
Aviation Uses 15 - 100
1000 - 2000 nm. Gas Sensing 0.35 - 2.5
Eye Safe Illumination 0.90 - 7.5
Lidar 15 - 9 0
2000 - 5000 nm. Gas Sensing 0.30 - 2.5
Eye Safe Illumination 0.90 - 7.5
Missile Jamming 5.0 - 30.0
Lidar 40.0 - 200

You will note that we have not listed all of the possible applications for
high output diode lasers. Most of the applications listed as Illumination in
the 600 to 1000 nm range are those already dominated by solid state or
gaseous lasers like the familiar He-Ne red laser. Nevertheless, diode
lasers are rapidly replacing those lasers in certain applications.

Diode lasers are revolutionizing medical diagnostics. For example, they

have been used in brain-scanning where two wavelengths, 780 and 830
nm., are directed towards a patient's head. The light passes through the
scalp and is reflected from the cerebellum. Differences in blood
hemoglobin can be made which show the blood flow in tissues. Devices
are being used in doctor's offices and hospitals to measure the amount of
 7.2 Current display devices based upon electron creation 667

blood-borne oxygen, as a measure of the capability of a patient's

respiratory system.

Two methods now under development include: 1) measurement of blood

flow during epileptic seizures in the brain and 2) non-invasive
measurement of blood glucose levels using three different wavelengths:
834, 1304 and 1554 nm. by diabetic patients. Compared to conventional
methods of medical imaging, NIR (near infra-red) optical tomography is
relatively safe (since it has no ionizing radiation) and cheap since it does
not require large magnets or radiation shields.

f. Organic Light Emitting Diode Displays (OLED)

The OLED display is based upon certain organic compounds which emit
photons in a manner already described for LED's. When a forward-biased
current is applied, light is emitted according to the energy gap between
the LUMO (lowest unoccupied molecular orbital) and the ground state,
LOMO (lowest occupied molecular orbital). We have already discussed this
mechanism in Chapter 5 (see 5.5.27.). As shown in the following, only
certain types of organic compounds are useful in this application:

7.2.33.- Organic Electroluminescent Compounds

These compounds are used to manufacture OLED displays that are used
668 Current phosphor device technology

in numerous devices including cell phone displays. Note that all of these
compounds are organic semi-conductors. The OLED display is formed by
thin film techniques similar to those for LED's and Diode Lasers:

7.2.34.-

As you can see, the fabrication is rather simple. Yet, the light output rivals
many of the LED's and is certainly more than the output of phosphor
screens in certain devices. A conductive layer is required for electron
injection into the electron transport layer (ETL). The metal electrode
(ME) is composed of Al, Ag or Ag-Mg, and is usually light reflective. The
hole transport layer (HTL) injects holes from the transparent electrode
(TE) into the HTL. The point where the excitons, i.e.- pairs of holes &
electrons, recombine is the light emitting (LE) region. This can vary
according to the type of compounds used to form the two organic layers.
These devices can be classified as:

TE: HTL(LE): ETL: ME

TE: HTL: ETL(LE): ME

When bipolar materials are available, i.e.- have the ability to carry both
holes and electrons, the device structure can have a three layer
configuration wherein the LE layer is sandwiched between the HTL and
ETL, i.e.-
TE: HTL: LE: ETL: ME.
 7.2 Current display devices based upon electron creation 669

An aromatic diamine: N,N'-diphenyl-N,N'-bis(3-methylphenyl), l,l'-4,4'-

diamine has been used as a HTL. Also used has been the compound shown
to the right in 7.2.33. The organic compound, Aluminum tri-hydroxy-
quinolate (also given in 7.2.33.) is probably the best ETL material. Laser
dyes such as: 4-dicyanomethylene,2-methyl,6,9-dimethylstyrl,4H pyran,
eCeob
have been used as LE layers. Quinadone, i.e.- NH O also is a good LE
compound used as the emitting layer in OLED's.
The size of the display is limited only by the size of the thin-film apparatus
used to make the film. A large number of organic materials have been
examined for this application. Most of the successful ones are those based
upon the modification of the basic backbone of those compounds given in
7.2.33. The requirements for a successful OLED are:

1. The compound must produce a stable and uniform thin film

under vacuum sublimation.
2. The compound must have a large affinity for electrons, i.e.-
be a "hole-conductor"- a p-semiconductor.
3. The compound must have a high electron mobility for
transport of electrons to the emitting center.
4. The ionization potential of the compound must be high to
prevent loss of hole-energy at the cathode before recombining
with electrons, i.e.- the excitons formed during electron
energy transport must remain stable until the energy is
transmitted to the emitting center.
5. The compound should have large exciton energies to
forestall energy transfer from the EM layer to the ETL, i.e.-
the exciton energy should be large enough to prevent exciton
recombination within the ETL and loss of excitation energy at
the emitting centers.

A recent OLED design involved a three-color emission to produce a white-

light emitting OLED, as shown in 7.2.35., given on the next page. Note
that the construction is similar to that we have already presented. The
only difference is that three different colored emitting strata are involved:
670 Current phosphor device technology

7.2.35.- Configuration of a White-Emitting OLED

This 3- layer device achieved a luminance of 4,200 candela/ square meter

at a relatively low voltage of 16 VDC. In contrast, the surface brightness of
a cathode-ray tube is only 100 candela/square meter, while a fluorescent
lamp has a value of 8,000 candela/square meter. Whether such a device
can reach a surface brightness of 10,000 candela/ square meter or not has
not yet been determined.

An active research area is currently the use of conducting polymers as

OLED devices. These polymeric-organic electronic materials are varied in
compositions. In 1957, the first intrinsic electrically conduction polymer,
 7.2 Current display devices based upon electron creation 671

polyacetylene was reported by Chiang et al. Since that time a number of

such compounds have reported, as shown in the following:

7.2.36.- Electrically Conducting Polymers

These are just a few of the many compounds that have been developed.
However, they are not electroluminescent (EL) and cannot be used
directly for OLED devices. What has been done is to modify the organic
structure to produce light-emitting polymers. Polythiophene does not
emit whereas derivatives exhibit strong EL emission. There is another
advantage as shown in the following, given as 7.2.34. on the next page.

Here, we have shown four derivatives of polythiophene and their spectral

properties. Note that the emission band can be shifted across the entire
range visible to the human eye. The peak of the band can be shifted to
optimize the light output in any desired wavelength, something that the
OLED compounds described above cannot do.

The semi-conducting properties of conjugated polymers originate from

the delocalized n orbitals formed in these carbon-containing compounds
such as poly(phenylene vinylene), polythiophene and poly(paraphenylene).
Like the OLED's described above, the polymer-OLED consists of a
luminescent film, sandwiched between an anode and a cathode.
672 Current phosphor device technology

7.2.37.- Emission of Some Polythiophene Derivatives

Glass coated with a thin layer of transparent metal and a transparent

coating of indium-tin-oxide are most frequently used. The polymer can be
applied by a number of techniques including use of a inkjet printer or
sublimation methods as given above.. The films are stable and have high
quantum efficiency. At this point in time, it is not clear which organic
conducting films will be used. Lifetime studies, coupled with relative light
output, will probably determine the final decisions made

OLED displays are replacing liquid-crystal displays in many areas. Among

these are mobile phones and "personal assistant" displays. Conventional
LCD's suffer from limited viewing angle clarity and are difficult to see in
ambient light (unless they are "back-lit"). The possibility of using OLED's
for full size TV's and computer displays has not yet become realized.
However, there is a good likelihood that this display area will be realized.
Since the OLED screen can also be printed by inkjet technology, the main
problem of forming lines or dots on a TV screen has been circumvented.
Because the OLED display is easier to manufacture than either the backlit
 7.2 Current display devices based upon electron creation 673

color-LCD or the LED (mono-chrome or color) display, the impetus to

replace these applications is strong. OLED displays also require much less
power than either of these technologies. The major problem will be the
range of colors that can be displayed and the color-reproduction possible
in comparison to the already well-developed phosphor display now used
in color-TV, and the filters used in the color-LCD displays. Note that all of
these OLED compounds are organic in nature and are fabricated using
organic reactions well known since the early 1800"s. Thus, the largest
chemical companies like DOW, BASF, Eastman Kodak and DuPont are
actively working on OLED displays.

It is well to point out that the use of phosphors has been circumvented in
the OLED devices. It may be that, in future years, phosphors will become
obsolete as the use of conductive organic compounds and polymers
mature.

g. Electroluminescent Displays fEL)

The electroluminescence phenomenon was originally discovered in 1936

by Destriau. Light emission was observed from a ZnS phosphor powder.
The device consisted of the powder layer dispersed in castor oil, with the
layer between two high voltage electrodes. When the transparent
conductive films, SnO2.x or InSnO2.x became available, Sylvania Electric
Corp. manufactured an AC device called "PanelLight" that was used as a
night light. We have already described the technology of OLED materials
(which are actually electroluminescent (EL) substances).

The first EL devices were based upon powdered semi-conductors like

ZnS :Mn2+which emits green light. Next came LED materials as described
above and finally, OLED EL-devices spanned the gap between the two
types of materials. ZnS:Mn2+ based EL devices require a rather high
voltage to perform, in contrast to that needed for LED's and OLED's. The
EL-device consists of an EL-phosphor, sandwiched between two insulating
layers. When a DC-voltage is applied, light is emitted. This construction is
very similar to 7.2.25. or 7.2.27., as shown in the following diagram,
given as 7.2.38. on the next page:
674 Current phosphor device technology

7.2.38.-

You will note that the layers are quite thin, an advantage of EL displays
over some of its competitors. A considerable amount of work was aimed at
producing thin layers of the zinc sulfide phosphors, using the techniques
listed in 7.2.13. The result has been that thin layer EL displays produce a
brighter screen than those based upon a powder layer. While EL displays
have replaced liquid-crystal displays in some areas, they have been
surpassed in performance by LED's and particularly OLED's. EL displays
have been used as backlighting in wristwatches and key-pads. In spite of
intense research work on EL screens for color television, the results have
not been entirely satisfactory. For "Heads-Up" displays, i.e.- military or
aviation displays, where the screen is directly in front of the eyes of the
viewer, the EL thin-layer (transparent) screen is entirely satisfactory.
However, for home television or computer terminal displays, the EL
display has been supplanted by either the backlit AMLCD (Active matrix
liquid crystal) or the plasma discharge displays. OLED displays are rapidly
overtaking the EL display as well. Although the original EL displays were
direct-current driven, the latest type is the ACTFEL (alternating current
thin film EL) display. Its main advantage is the very high contrast, i.e.- low
scattering, resulting in a low reflectance ratio, of the image produced.
Phosphors that have been used in ACTFEL displays include those given in
7.2.39., listed on the next page.
 7.2 Current display devices based upon electron creation 675

7.2.39.- Phosphors Used in ACTFEL Devices.

Phosphor Color Lum.@ 60 Hz Eff.-(Lum/W) Peak (nm)

21
ZnSiMn - yellow 300 cd./m2 5.0 585
ZnSiSnr*. Cl red 12 0.08 610
Ca&Eu3* red 12 0.05 650
CaSSe: Eu3* red 25 0.25 620
ZnSiTb^ OF green 100 1.30 545
Zn^iO^Mn^ green 10 0.20 510
Zn,GaA:Mir* green 10 0.20 501
ZnSiCe^.Mn^ bl.-green 105 1.63 506
C^Ga^Ce 3 * blue 10 0.04 460
ZnS;Tm3t,F blue 0.2 <0.01 450
SrSiCe^.Eu3* white 36 0.20 490
ZnS:Mn,SrS:Ce white 450 1.6 580

As you can see, the brightness of these display screens is rather low in
comparison to the most advanced screens like the OLED screen. As we
said, the EL display has seen limited use in recent years in favor of the
more efficient light producing screens like the LED and OLED types. The
major advantage of ACTFEL screens continues to be the very high contrast
present in the image displayed. For this reason, this type of display will
continue to be used when a sharp image is required, particularly in small
displays like the "Heads-Up" display or individual, head-worn, monitors
where brightness is not the key factor in usage.

This completes our survey of devices operating by electron generation in

an electric field (as shown in 7.2.1.). you will note that we included
"Organic-Thin Film & Powder" in conjunction with the discussion of LED's
and Diode lasers. This was done because of the very close similarity of
these displays to the OLED EL displays.

The next survey of Phosphor Devices for Electronic Display will cover
those in which generated photons are directly involved in phosphor
excitation.
676 Current phosphor device technology

7.3.- CURRENT DISPLAY DEVICES BASED UPON PHOTON CREATION

The types of illuminating and display devices that we will now address are
shown in the following:

7.3.1.

In this case, phosphors as screens or thin-films are excited by photons

having various energies, from near UV to x-rays. We will address all of
these devices in connection with the phosphors employed in their
manufacture and the thin-film structure used to form the screen. We have
 7.3 Current display devices based upon photon creation 677

already addressed "Fluorescent Lamps for Illumination" earlier. This

included substitution of 3-narrow-band phosphors for conventional broad-
band phosphors which resulted in smaller, but brighter, fluorescent
lamps. Such lamps use much less power than their predecessors, as
compared to the 40T12 lamp previously used for general illumination. In
addition, the latest fluorescent lamps show much better maintenance of
output over their entire life. Improved ballasts have also contributed to
this remarkable improvement. The F32T5 lamp is now replacing the HID
metal halide (HID-MH) lamps in commercial installations since the cost of
operating T5 lamps is about 50% that of the HID lamp annually. Whereas
the HID-MH lamps lose up to 45% of their initial brightness over time,
the T-5 lamps lose only 5%. The new fluorescent technology is now
preferred for lighting warehouses and adjoining buildings, offices of all
kinds and medical facilities.

I. Fluorescent Lamps for Special Purposes

These lamps include those used for "back-lights" in computer terminals

and those used in copying machines.

a. Liquid Crystal Back-Light Displays

In order to show how these lamps are used, it is necessary to describe

liquid crystal displays (LCD). We will not describe this technology
thoroughly since it does not directly relate to the use of phosphors in the
display, except in the fluorescent lamps used for "backlighting" the LCD. A
"liquid crystal"- (LC) is an organic compound whose molecules align
themselves with an applied electric field. An LC substance flows as a liquid
but maintains some of the ordered structure characteristic of a crystal.

Because the molecular forces producing liquid crystalline states are very
weak, the structures are easily affected by changes in mechanical stress,
electromagnetic fields, temperature, and chemical environment. Three
main categories have been recognized: smectic, nematic, and cholesteric:

Smectic liquid crystals consist of flat layers of cigar-shaped

678 Current phosphor device technology

molecules with their long axes oriented perpendicularly to the

plane of the layer.
Nematic liquid crystals are also oriented with their long axes
parallel, but they are not separated into layers, and they behave like
toothpicks in a box, maintaining their orientation but free to move
in any direction. Nematic substances can be aligned by electric and
magnetic fields, resulting in a number of characteristics such as the
ability in some cases to be electrically switched from clear to
opaque. This property gives rise to many technical applications such
as in display systems.
Cholesteric liquid crystals form in thin layers, each one molecule
thick; and within each layer the molecules are arranged with their
long axes in the plane of the layer and parallel to each other, as a
two-dimensional nematic structure.

LCD displays are formed by thin-film techniques in which the liquid

crystal layer is surrounded by electrodes which control the alignment of
the LC's. If the LCD is a color display, color-filters are used to form a red,
green or blue pixel, with a thin-film-transistor at each pixel to control the
transmission of light at that point. This construction is shown in the
following:

7.3.2.-

In this case, a "backlight" is used for the light source. A "twisted nematic"
liquid crystal is used as a shutter for light transmittal through the glass
and the Si TFT transistor controls the shutter. If colored filters are used,
 7.3 Current display devices based upon photon creation 679

then we have a color television or computer monitor display. Note that we

have presented a simple structure in order to promote understanding of
how such configurations are formed by thin film technology. The actual
structure is usually much more intricate. The TFT's control the electrodes
which align the liquid crystals. We have not shown the polarizers which
are sometimes used in conjunction with the color filters. These are
placed just before the backlight and after the filters.

Let us now examine the fabrication of a color TFT-LCD display. A color

television uses a cathode-ray tube (CRT) as a display as does a computer
monitor. However, the cathode-ray tube is being slowly superseded by
liquid-crystal displays (LCD) which are considerably thinner than the glass
CRT. LCD devices can be transmissive or reflective. They are made with a
demanding set of processes similar to those used for making IC's. But,
while IC makers save money by making their chips smaller, display users
insist on having their displays larger. This leads to a struggle between cost
and utility for display makers. Most color displays rely upon a thin-film
transistor (TFT) to control a liquid-crystal (LCD) bit or pixel. The cathode
ray television tube uses a pattern of phosphor dots (pixels) with selective
excitation to form an image. The TFT-LCD forms an image by controlling
the transmission of an external light source by means of the TFT. For
producing a red image, all of the red-filter TFT pixels are turned "on" so
that the image seen is red. Obviously, color reproduction is dependent
upon the quality of the color-filter. Nonetheless, the steps to achieve the
final TFT array are many and include steps similar to IC fabrication except
that glass is used as the substrate instead of silicon. The following
diagram, shown as 7.3.3. on the next page, presents the diagram of a
single TFT-pixel on the overall glass substrate. Obviously, the actual
number of TFT pixels needed for any given display amounts to many
thousands. The overall process involved the use of many "masks" similar
to the many steps involved in Integrated Circuit manufacture.

Storage condensers are needed to operate the pixel (active-matrix). Some

storage capacitors can be constructed by using part of the gate electrode
as a storage capacitor electrode (which is called aCs-ongate). Note that
680 Current phosphor device technology

7.3.3.-

the ITO areas on the left are transparent and are the source of the
transmitted light, i.e.- pixels. The three layers are "etched-backed" to
define the a- Si islands which form the TFT part of the structure. This
process thus defines the island, or "active matrix", which controls each
pixel in the display. The transparent electrodes are also not shown in this
diagram. Note that this process uses photolithography and plasma dry
etching to define the islands by use of masks, the latter called an etch-
protect mask. We have not described the steps used to produce a mask. A
better view of a single pixel construction is shown in the following:
7.3.4.-
 7.3 Current display devices based upon photon creation 681

Note that we have now shown all of the details of a single pixel associated
with the TFT. The overall construction was given in 7.3.3. and the single
transistor and pixel in 7..3.4. Note that the entire space surrounding the
TFT is filled with liquid crystal. The same TFT compositions are shown in
both diagrams.

We have not yet described the Backlight. This is simply a serpentine

fluorescent lamp made in a flat configuration:

This lamp is usually formed from a lead-glass which can be easily shaped
at relatively low softening temperatures. The lead-glass tube is coated
with a phosphor mixture, then evacuated and sealed. It is then bent into
shape. The phosphors most used are the 3-line emitting rare earth
compositions given above for the 32T5 lamps, i.e.-

BaMgAl10O17: Eu2+ - BLUE

MgAluO19: Ce3+ :Tb3+ - GREEN
Y2O3: Eu3+ - RED

The electrodes are mostly of the "cold-cathode" type. That is, they consist
of unheated metal electrodes, which emit electrons under a high voltage.
More recently, electronic ballasts, described above, have been used as well
as micro-wave high-frequency ballasts. The problems of getting these
lamps to operate properly upon a DC voltage (using a battery) have been
enormous.

Because of the complexity of manufacturing AMLCD displays, it is unlikely

682 Current phosphor device technology

that they will ever be used for applications other than computer monitors
or "point-of-sale" displays. Because of such complexity, these displays are
limited in how big they can be made. This factor involves one of
equipment size required more than the actual procedures needed to
complete the display. One estimate of the ultimate size that any of the
feasible display functions is shown in the following:

7.3.5.- Estimate of Ultimate Market Penetration of Several Display Types

Note that the CRT covers a sizable portion of the current, and estimated
future market for displays. This ranges from the very small (one inch or
less) to the very large, up to 48 inches across (I have seen such a CRT
tube (48") and it is enormous. I estimate that the vertical deflection is
close to 80 °- see above discussion on CRT's). For computer monitor
displays, the LCD display has taken over a major share of the market while
the very small displays are a combination of these two, including EL
displays. The larger "Heads-Up" displays are dominated by the EL type of
display. Nevertheless, the PDP ("Plasma Display Panel) has become the
major type for larger displays and large color-TV's. The color is superior
and high resolution displays are easier to manufacture. For the very large
displays, the projection TV seems to dominate at this moment. Whether
this dominance remains will be determined by how much the PDP and/or
 7.3 Current display devices based upon photon creation 683

projection-TV can be improved. Note that we have not yet discussed the
PDP. This will be done below. It should be clear that each type of display
is more suitable for certain applications than any of the others.

b. Lamps for Copying Machines

The xerographic process utilizes a toning drum coated with a

photosensitive layer. This layer is first charged, usually positive, before
scanning begins. Green radiation from a fluorescent lamp is most often
used. The lamp is mechanically scanned over the object to be reproduced
(most often a sheet of paper containing alphabetical characters and/or
images) while the drum is simultaneously turned. When the light from the
lamp is reflected from the sheet of paper being scanned, certain areas on
the layer on the drum becomes discharged except at the points or lines
where the black ink does not reflect the light. Toner particles, being
negative, are attracted to these still charged areas and are then
transferred to a blank sheet of paper to make the copy. Because the lamp
is used in a scanning mode, it is usual to produce an aperture lamp in
which a "slot" is induced by w i p i n g a n a r r o w l d n e a c r o s s t h e p h o s p h o r
coating. The lamp looks like this; ^mm^^m^^^^mmmmmrMmm^^

Radiation from an aperture lamp is emitted in tight "fan" configuration.

For scanning a sheet of printed paper, it is ideal since the emission is
concentrated in a relatively narrow band. The phosphor most used for this
application is: Mg2Ga2O4:Mn2+. More recently, some of the terbium
activated phosphors like: (Mg,Ce)AlnO19:Tb3+, LaPO4:Ce3+:Tb3+, and
Y2SiO5:Ce3+:Tb3+ have been employed because the major part of their
emission is concentrated in the 543 nm line.

c. Blacklight Lamps for Water Treatment and Viewers

The "Blacklight" lamp is essentially a fluorescent lamp that emits in the

near Ultraviolet. The BaSLjO^Pb^ phosphor has no peer in this
application. This UV-emitting lamp (emission peak at 355 nm.) is used at
sewage treatments plants to destroy bacteria and fungi spores in the
incoming sewage stream. A combination of lamps surrounds the stream
684 Current phosphor device technology

and the ultraviolet emission penetrates the aqueous suspension of sludge

to destroy all pathogens present. Although other lamp emissions have
been tried, even the quartz low pressure emission at 254 nm., the rate of
success has not equaled that of the 355 nm lamp. The phosphor emitting
at 370 nm., (Ba, Sr, Mg)2Si2O7:Pb^, is almost as good but the phosphor,
(Ba, Sr, Zn^Sip^Pb 2 *, emitting at 295 nm. is much inferior.

d. Medical Uses for Blacklight Lamps

"Blacklight" is also used to view luminescent paints and dyes. There are
numerous medical uses where "fluorescent tags" consisting of selected
organic complexes of Eu3* and Tb3* are used to identify certain compounds
of medical interest, including tagged DNA fragments.

Note that we have been discussing lamps utilizing 254 nm. photon as an
excitation energy. Since we have already discussed HPMV lamps which
utilize 254-413 nm radiation to excite phosphors (see above), we are now
ready to explore the vacuum-UV region which uses photons generated at
147 nm to 173 nm wavelengths to excite phosphor display devices. These
include 2 types, plasma display panels (PDP) and Neon Signs.

II. Display Devices Based on Vacuum UV Photon Generation

These devices are based upon a so-called "plasma-discharge" which is a

misnomer for the same type of discharge used to create 254 nm radiation.
A stable plasma, in physics, consists of an electrically conducting medium
in which there are roughly equal numbers of positively and negatively
charged particles, produced when the atoms in a gas become ionized.
Thus, the low pressure mercury discharge used in fluorescent lamps is a
plasma. The modern concept of the plasma state is of recent origin,
dating back only to the early 1950s. Its history is interwoven with many
disciplines. Three basic fields of study made unique early contributions to
the development of plasma physics as a discipline: electric discharges,
magnetohydrodynamics (in which a conducting fluid such as mercury is
studied), and kinetic theory. Of the two types of light generating devices
 7.3 Current display devices based upon photon creation 685

utilizing plasma-generated vacuum-UV radiation in use today, the Neon

Sign is the oldest.

Since both operate on the same principles, we will address the Plasma
Display Panel (PDP) first.

a. Plasma Display Panels- Vacuum UV Photon Generation

Whereas most of the display types that we have described are based upon
either electron- generation by electric fields of some sort or by mercury-
discharge (254 run photon generation), the newest display is based upon
generation of vacuum ultraviolet photons to excite the phosphors. The
gaseous discharge uses one or more of the noble gases, i.e.- Ne, Ar, Kr,
and/or Xe. Of these, Xe (xenon) seems to be the best, but its strongest
emission line lies at 147 run., in the so-called "vacuum UV" region. Most
materials strongly absorb this radiation but some phosphors are not
excited at all by this radiation line. The resonant line of atomic Xe is 147
nm but the resonant line of Xe-Xe molecules lies at 173 nm.

The design of a PDP display is complex since the plasma itself is hard to
start. Once it is started, it is necessary to maintain it by a voltage less than
that needed for the full discharge but enough to keep some gaseous atoms
ionized, i.e.- Xe => e" + Xe+ . The following diagram, given on the next
page as 7.3.6., shows one construction that is being used.

Note that this display is essentially two plates of glass in which separators
are used to form a series of channels along the width and breadth of the
device. Within the channels, a series of conductive metal "bus" bars are
located. It is these that are covered with sequential layers of phosphors.
Note that the entire inside surface is covered. The top glass faceplate
consists of transparent bus-electrodes, along with "sustain" electrodes, at
right angles to the phosphor-covered bus-bars. The dielectric layer is
essentially a layer of low-softening glass particles. The final step to form
the PDP is to sinter the dielectric layer to form the integral PDP product.

Note that the plasma is only generated at the point where the top and
686 Current phosphor device technology

7.3.6.- Construction of a PDP Display

bottom bus-bar electrodes cross each other. Actually, this is a very

simplified diagram, since scan electrodes, priming slits, and other
features are required for proper operation as a TV monitor and/or display.
However, we have not shown these in the interest of simplicity.

The following, given as 7.3.7. on the next page, shows a simple version of
the construction of one pixel in the PDP display.
 7.3 Current display devices based upon photon creation 687

7.3.6.-

Here, we have only shown the basic parts of the pixel. Note that, in a 48"
PDP display, there will be up to several million of these pixels present.

Originally, a PDP based upon the "eximer" discharge was proposed. This
involves "Excimer" lasers which had been developed based upon the
molecules: ArF, KrF, XeF, and XeCl, which exhibit laser action at: 193
run., 248, 308 and 350 nm., respectively. The lasing action was obtained
by passing an electrical discharge through a suitable mixture of a noble gas
and fluorine gas. However, it was soon determined that silicate glass could
not be used to form a PDP because of the well known chemical action of F2
on silica. Although alumina could have been used as the containment for
the plasma discharge, it proved to be too expensive, especially in the
larger sizes. Thus, Xe gas remains the major one used, with small amounts
of the other noble gases sometimes added. Although some of the original
PDP displays were based upon a DC voltage to excite each pixel of the
display, the most recent ones use AC voltage and current to originate and
maintain the plasma at the pixels being excited. You will note that we have
not attempted to describe the electrical characteristics and the electronic
parts of the PDP design. For example, a 48 " display requires at least a
800 line-scan to address the display at > 200 volts, with access times of
3-4 usec. That is, the repetition rate must occur in that time to sustain a
688 Current phosphor device technology

moving color-picture without apparent blurring to the human eye. Large

area high-resolution PDP panels are driven at about 1.0 psec, and at a
repetition rate of 3 usec.

Although a great body of work had been directed towards the excitation
characteristics of lamp and cathode-ray phosphors, little had been
published concerning the vacuum-UV (VUV) excitation attributes,
particularly at 147 and 173 ran. The major reason for this is that both
nitrogen and particularly oxygen strongly absorb at those wavelengths.
The obvious solution to this dilemma is to construct a VUV "plaque-tester"
(see 6.7.1. of the last chapter. However, the design will be simpler since
the plasma discharge is generated internally. One design is shown in the
following:

7.3.8.- Design of a VUV Plaque Tester

We have not shown the auxiliary equipment needed, just the essentials.
One of the most important features is maintaining a vacuum-tight seal.
The other is being able to evacuate to > 10 8 torr while eliminating all
traces of water vapor. This means operating in a controlled atmosphere
room where the humidity is kept low (or in a "glove-box" where the
humidity is very low). The results of such measurements started the
development of a spectroradiometer where the excitation-emission
monochromators were evacuated as well. The final results are shown in
the following table, given on the next page, compares several phosphors
under VUV excitation.
 7.3 Current display devices based upon photon creation 689

This is not a comprehensive listing of all of the phosphors investigated.

The most important part for a color-television display is the b r i g h t n e s s
produced at the "Daylight" color, i.e.- the Hluminant-C color

Table 7-7
Phosphor Composition Color Rel. Lum. Output
Zn^SiO^Mn^ green 1.00
3+
Y2O3:Eu RED 0.71
YBO3:Eu3+ 1.00
GdBO 3 :Eu 3+ 0.92
YGdBO3:Eu3+ 1.20
Zn 3 (PO 4 ) 2 :Mn 2+ 0.34
Zn^Ga/VMn2* GREEN 0.23
2+
BaMgAl 1 4 O 2 3 : M n 0.92
BaMgAl 1 2 O 1 9 : M n 2 + 1.15
YBO 3 :Tb 3 + 1.12
CaWO 4 BLUE 0.76
2+
BaMgAl 1 0 O 1 7 : E u 1.65
Y 2 SiO 5 :Ce 3 + 1.14

.Some of the combinations used are shown in the following:

7.3.9.- Tri-Color Phosphors Used in PDP Display

Phosphors Rel. Brightness
Blue Green Red
CaWO4 Zn 2 SiO 4 :Mn a+ Y2O3:Eu3+ 65
3+
Y2SiO5:Ce Zn 2 Si0 4 :Mn^ Y 2 0 3 :Eu 3+ 105
BaMgAl10O17: Eu 2+ Zn 2 SiO 4 :Mn^ YBO3:Eu3+ 150
2+
BaMgAl14O23: Eu Zn^iO^Mn* YGdBO3:Eu3+ 182

As you can see, the luminance of the PDP display is predicated upon the
output of the phosphors when excited by the plasma discharge. What we
have not stated is that some phosphors are not as stable under the
690 Current phosphor device technology

processing conditions used to make the PDP. At the present time, the
phosphors being used for the PDP display are:

BaMgAl10O17: Eu2+ - blue

Zn2SiO4:Mn2f - green
YGdBO3:Eu3+ - red

Of the displays that can be used for color-television, the PDP panel
outperforms all of its competitors. The screen is brighter, resulting in
superior contrast for high definition display.

b. Neon Signs

Signs are ubiquitous in our society. The word "sign" originally meant a
device used for advertising purposes. The ancient Egyptians, Greeks and
Romans created signboards by whitewashing convenient sections of walls
for suitable inscriptions. Early shop signs were developed when
tradesmen, dealing with a largely illiterate public, devised certain easily
recognizable emblems to represent their trades. Some signs like the
three golden balls of the pawnbroker, and the red and white stripes of the
barber early became identified with particular trades.

When practical electric generators were invented in the late 19th

century, illumination became possible for shop signs and billboards, and
by 1910 the French scientist Georges Claude was experimenting with the
neon tube and other gas-filled illuminating devices. In less than a decade,
signs were being fashioned of glass tubes bent to form words and designs
that glowed red when the gases inside them were subjected to an electric
current. A typical neon sign tubing consists of a 10-15 mm (0.59") glass
tubing at least 3 meters (118.5") long. The tubing has end-caps attached
and then is evacuated. The glass is usually a lead-glass with a low
temperature softening point. The evacuated tubing is heated to remove all
traces of water and air and then is "backfilled" with 0.8-2.3 torr of neon
gas. The tubing was then bent into a desired shape and operated to
produce a reddish orange plasma discharge. The plasma discharge
originally used a cold cathode operating mode at high AC voltage (but low
 7.3 Current display devices based upon photon creation 691

AC current). Nowadays, the high frequency ballasts are used, similar to

those used for fluorescent lamps. Because colors other than the red
plasma discharge of neon were needed to make the sign more effective
for advertising, the tubing was internally coated with phosphors. After
bakeout, mercury and rare gases was added and the tubing was then
sealed with endcaps similar to a miniature fluorescent lamp. The only
difference was in the glass used since the soda lime glass used in
fluorescent lamps could not be easily bent into shapes. As a matter of fact,
the flat "backlights" described above were derived from neon-sign
technology. The electrodes are mostly of the "cold-cathode" type. That is,
they consist of unheated metal electrodes, which emit electrons under a
high voltage. More recently, electronic ballasts, described above, have
been used as well as micro-wave high-frequency ballasts. A variety of
phosphors are used similar to those cited above for the PDP display.

The nighttime skyline of every city in the world has changed as electric
neon signs came to dominate the main commercial streets. The city with
the most "neon-lights" is, of course, Las Vegas, Nevada.

III. Devices Utilizing High Energy Photons

We will first describe devices which use phosphors which respond to high
energy photons like x-rays and gamma rays (high energy x-rays).

a. X-Ray Intensifying Screens

There are two types of x-ray intensifying screens. One is passive and the
other active. The passive type of X-ray phosphor intensifying screen is
used to increase the speed of radiographic films using light emission from
the phosphor. The phosphor must have high response to x-ray excitation
since the x-ray absorption of the radiographic film is very low, at best 1 %.
Its construction is shown in the following diagram, given as 7.3.10 on the
next page.

The construction of the cassette uses two reflective light layers, one a
highly reflecting layer for the phosphor emission and the other highly
692 Current phosphor device technology

absorbing. The film is sandwiched between these optical layers. This

arrangement reduces blurring of the image due to light scattering and/or
movement by the patient. It also makes it possible to use a small focus x-
ray tube. Note that the film has two protective layers coated with silver
halide emulsion layers on both sides. The image quality of x-ray
intensifying screens is appraised by sharpness of the image, graininess of
the image and contrast obtained. Contrast can be altered by changing the
7.3.10.-

x-ray absorption attributes of the phosphor. The characteristics of

phosphors required for this application are:

1. High emission efficiency

2. Strong x-ray absorption
3. Emission spectrum matched to the spectral sensitivity of
the x-ray film
4. Short emission decay time
 7.3 Current display devices based upon photon creation 693

5. Stability to x-ray stimulation- no degradation over time

6. Good physical properties including narrow PSD and low
particle size, and good coating dispersion to form the screen.

The following lists some of the phosphors used for this application.

7.2.11.- Phosphors used in X-Ray Image Intensifier Applications

Composition Color Efficiency - Peak.nm X-ray

X-ray Excit. Response
BaFCl:Eu2+ violet 13% 380 49.3
BaSO4: Eu 2+ violet 6 390 45.5
CaWO4 blue 5 420 61.8
Gd2O2S:Tb3+ green 13 545 59.5
LaOBr:Tm3+ blue 14 360,460 49.3
La2O2S:Tb3+ green 12.5 545 52.6
3+
YTaO4:Nb blue 11 410 67.4
ZnS:Ag°, Cl blue 17 450 9.7
(Zn,Cd)S:Ag°,Cl green 19 530 38.4

The best phosphors for this application have been found to be:
LaOBr:Tm3+, Gd2O2S:Tb3+ and YTaO4:Nb3+. The active type of X-Ray
intensifier screen, utilizing phosphors, uses a combination of excitation
sources, x-rays and electrons. It is used to detect x-rays under a number
of conditions such as direct medical viewing at low x-ray intensity and in
airports to view baggage contents. The active x-ray device consists of a
quasi-CRT, except that it is constructed so that electrons are emitted
somewhat like the image intensifier of 7.2.7. However, instead of
detecting photons, like the night-vision goggles, it detects x-rays in a
coherent pattern to form a dynamic picture. Its construction is shown on
the next page as follows.

The qualifications for this type of display include:

1. High image brightness;
2. Sufficient image resolution.
694 Current phosphor device technology

7.3.12.-

The phosphors used for the input screen must have:

1. A large x-ray absorption coefficient;

2. High luminescent efficiency when exposed to x-rays;
3. An emission spectrum matched to that of the photocathode.

Since the focused beam current in this type of modified CRT display is
much lower than in the ordinary CRT, the output phosphor must have:

1. High luminous efficiency at low current inputs;

2. High resolution within the image, especially if this type of display
is used for medical imaging (Note that the primary usage is
predicated upon maintaining a low x-ray dosage to the patient). In
general the output screen is prepared by thin film methods in order
to achieve the high resolution required.

The following, given as 7.3.13. on the next page, lists some of the
phosphors used for this application. You will note that some of the
phosphors used for the image intensifier tubes of 7.2.7. like P-31 and P-
36 have been used for this application.
 7.3 Current display devices based upon photon creation 695

7.3.13.- Phosphors Used for Image X-Ray Intensification Screens

Input Phosphors Formula Emission color

(Zn,Cd)S:Ag green
ZnCdS2:Cu:Ni green
Gd2O2S:Tb3+ green
Output Phosphors ZnS:Cu green
ZnCdS2:Cu green-yellow
YTaCVNb blue

It is well to note here that the differences between x-ray:

Fluoroscopic screens
Intensifying screens
Image intensifier screens
Radiographic image screens

is not great. All of these utilize phosphors that respond to x-ray excitation.
The following lists some of the phosphors that have been developed for
detecting x-rays in various applications:

7.3.14.- Phosphors for X-Ray Detection

Formula Emission color-peak X-RAY Application
(Zn,Cd)S:Ag green Fluoroscopic screens
ZnCdS2:Cu:Ni green Fluoroscopic screens
Gd2O2S:Tb3+ green-545 nm Intensifying screen
BaFCL:Eu2+ purple-380 nm Intensifying screen
LaOBr:Tm3+ blue- Fluoroscopic screens
YTaO4:Nb blue-410 nm Radiographic image
(Gd,Y)2O3:Eu3+ red-610 nm Fluoroscopic for CT

X-ray intensifying screens are used in radiological diagnosis and in

industrial non-destructive testing. X-ray fluoroscopic screens are used
mainly in health examinations and in radiological examination of luggage
696 Current phosphor device technology

at airports. X-ray image intensifying screens are used mainly for medical
objectives.

We have not presented all of the phosphors developed for such purposes
but only those that are currently most important for the applications given
above. You will note that in most cases, the decay time of these phosphors
is not important since the images are mostly static.

In the case of dynamic images, short decay cadmium sulfide-based

phosphors are used.

b. Scintillation Phosphors for Imaging

The next subject we will address is that of scintillators. These are

phosphors used to detect a, |3, and y rays from incident sources. This
application has become very important with the advent of CT-scanners
and the PET-scanner, i.e.- CT = "computerized tomagraphy" and PET =
"positron emission tomography". In all cases, radioactive isotopes are used
as the source of these "rays", including positrons, i.e.- positive electrons,
p + . Although scintillation, i.e.- detection of high energy radioactive-decay
particles, may occur for nearly all phosphors, efficient scintillators must
satisfy the following requirements for practical use.

Most inorganic scintillators consist of transparent single crystals, whose

dimensions range from a few millimeters to many centimeters. This
requirement arises because the crystals are situated next to the detector
window. Some inorganics, such as ZnS:Ag°, are good scintillators but
cannot be grown in the form of optical-quality large crystals. The
inorganic materials that produce the highest light output unfortunately
have relatively long decay times. The most common inorganic scintillator
is NaI:Tl+ which has an unusualty large light yield corresponding to a
scintillation efficiency of about 13 percent. Its decay time is 230 nanosec.
This is acceptable for many applications but too long when extremely high
counting rates or fast timing measurements are involved. Its emission
matches the peak response of photomultiplier tubes. This phosphor
probably attains its high quantum yield from matched ionic radii (Na+ =
 7.3 Current display devices based upon photon creation 697

2.23 A , Tl+ = 2.08 A), a simple cubic structure and the high absorption
coefficients of both I" and Tl+ for high energy particles. It also is
transparent to its own emission. CsI:Tl+ also has a good quantum efficiency
for detection of a,|3, and y rays but has two relatively long decay
components with a decay time of 680 nano-seconds. Its emission
spectrum is shifted toward the longer-wavelength end of the visible
spectrum and is a better match to the spectral response of photodiodes.
Both have iodine, with an atomic number of 53, as a major constituent.
Therefore the photoelectric cross section in these materials is large
enough to make them attractive in gamma-ray spectroscopy. You will note
that we have mentioned several requirements for a good scintillation
detector. These include:

High quantum efficiency for a,|3, p + and y rays

Short decay times, i.e.- < 1 usec.
Emission matched to photodetectors
High cross-section for radiation absorption

The mechanism for excitation-emission can be described as follows. In

certain types of transparent materials, the energy deposited by a highly
energetic particle creates excited ionic states that quickly decay through
the emission of visible or ultraviolet light. The amount of light generated
from a single charged particle of a few MeV kinetic energy is very weak
and cannot be seen with the unaided eye. There are four distinct steps
involved in the production of a pulse of light due to a single energetic
charged particle:

1. The particle slows down (and may stop in the scintillator),

leaving a trail of excited atomic or molecular species along its track.
The particle may be incident on the scintillator from an external
source, or it may be generated internally by the interaction of
uncharged quanta such as gamma rays or neutrons. Typical excited
states require only a few electron volts for their excitation. Thus
many thousands are created along a typical charged particle track.
2. These excited activator sites return to their ground state,
emitting energy in the form of a photon of visible or ultraviolet light.
698 Current phosphor device technology

They are emitted in all directions. The total energy represented by

this light (given as the number of photons multiplied by the average
photon energy) is a small fraction of the original particle energy
deposited in the scintillator. This fraction, called scintillation
efficiency ranges from about 3 to 15 percent for common
scintillation materials. The photon energy (or the wavelength of the
light) is distributed over an emission spectrum that is characteristic
of the particular scintillation material, i.e.- modified by vibronic
coupling. The excited species have a characteristic mean lifetime,
and their population decays exponentially. The decay time
determines the rate at which the light is emitted following the
excitation and is also characteristic of the particular scintillation
material. Decay times range from less than one nanosecond to
several microseconds and generally represent the slowest process
in the several steps involved in generating a pulse from the
detector. There is often a preference for collecting the light quickly
to form a fast-rising output signal pulse, and short decay times are
therefore highly desirable in some applications.
3. Some fraction of the light leaves the scintillator through an exit
window provided on one of its surfaces. The remaining surfaces of
the scintillator are provided with an optically reflecting coating so
that the light that is originally directed away from the exit window
has a high probability of being reflected from the surfaces and
collected. As much as 90 percent of the light can be collected under
favorable conditions.
4. Only a fraction of the emerging light photons are converted to
charge in a light sensor normally mounted in optical contact with
the exit window. This fraction is known as the quantum efficiency of
the light sensor. In a silicon photodiode, as many as 80 to 90
percent of the light photons are converted to electron-hole pairs,
but in a photomultiplier tube, only about 25 percent of the photons
are converted to photoelectrons at the wavelength of maximum
response of its photocathode.

The net result of this sequence of steps, each with its own inefficiency, is
the creation of a relatively limited number of charge carriers in the light
 7.3 Current display devices based upon photon creation 699

sensor. A typical pulse will correspond to at most a few thousand charge

carriers, i.e.- electrons. This figure is a small fraction of the number of
electron-hole pairs that would be produced directly in a semiconductor
detector by the same energy deposition. One consequence is that the
energy resolution of scintillators is rather poor owing to the statistical
fluctuations in the number of electrons actually obtained. For example, the
best energy resolution from a scintillator for 0.662 MeV gamma rays (a
common standard) is about 5 to 6 percent. You will also note that the
major action of a highly energetic particle within a symmetric crystal
structure is the creation of Frenkel pairs along the charged particle track.
NaI:Tl+ is unique in that it "self-repairs" itself at room temperature over
time. Most other scintillators like ZnS:Ag° form irreversible vacancies and
Zn+ interstitials which lower its efficiency over the time it is being used.

There are many characteristics that are desirable in a scintillator,

including high scintillation efficiency, short decay time, linear
dependence of the amount of light generated from deposited energy, good
optical quality, and availability in large sizes at modest cost. No known
material meets all these criteria, and therefore many different materials
are in common use, each with attributes that are best suited for certain
applications.

In Table 7-10, we show the properties of some scintillators in use today.

Positron emission tomography (PET) is a highly sensitive technique for
diagnosing stroke and other neurological diseases such as multiple
sclerosis and epilepsy. Positron-emitting radionuclides with short half-
lives (Notably 97 Tc 43 with a half-life of 6 minutes) are used to detect
cerebral blood flow, oxygen utilization, and glucose metabolism, providing
both qualitative and quantitative information regarding metabolism and
blood flow, such as in the heart.

The introduction of computed tomography (CT scan) in 1972 was a major

advance in visualizing almost all parts of the body. Particularly useful in
diagnosing tumors and other space-occupying lesions, it uses a tiny X-ray
beam that traverses the body in an axial plane. Detectors record the
strength of the exiting X rays.
700 Current phosphor device technology

Table 7 - 1 0
Phosphor Peak, n m . Decay, nsec Yield * Absorp" Use
NaI:Tl+ 415 230 38,000 2.22 XCT,PET,Nucl.
CsI:Tl+ 560 680 32,000 3.24 XCT,
LilrEu2* 470 1,400 11,000 2.71 neutrons
ZnS:Ag° 450 200 49,000 0.84 a-ray
CaF^.Eu2*- 435 900 24,000 0.47 p-ray
YAlO^Ce3" 370 28 15,000 1.37 XCT
CuWO4 470 5,000 14,000 7.68 XCT
PbWO4 480 300 1,500 12.9 Nucl
Bi4GeaO,, 4 8 0 3 0 0 3,800 10.8 XCT.PET.Nucl
*- Number of photons/ MeV of particle energy; **- Absorption coefficient
at 150 kev particle energy (cm1); Use: XCT = x-ray computer tomography,
PET = Positron emission tomography, Nucl = nuclear emission detection

This information is then processed by a computer and a cross-sectional

image of the body produced. CT is the preferred examination for
evaluating stroke, particularly subarachnoid hemorrhage, as well as
abdominal tumors and abscesses. CT scans provide two-dimensional views
of cross sections of the body, and these images must be viewed in
sequence by the radiologist. Computer technology now makes it possible
to construct holograms that provide three-dimensional images from
digital data obtained by conventional CT scanners. These holograms can
be useful in locating lesions more precisely and in mapping the exact
location of coronary arteries when planning bypass surgery or angioplasty.

Scintillators used for y-ray detection from: 97 Tc 43 (141 kev), ^ I ^ (159

kev) and 210Tl81 (75 kev), include NaI:Tl and Bi4Ge3O12. The former is
available in 80 cm single-crystal plates, a form not easily achieved by other
scintillator phosphor compositions.

Many new scintillators like: I^SiOg^e*", L^SiO^Ce3*, and YAlC^Ce^ are

being investigated because of the known very low decay times, as short as
1 nano-second. There use would allow much shorter exposure times for
patients to x-rays. The key is the ability to grow good optical grade
 7.3 Current display devices based upon photon creation 701

crystals. Although a bed of phosphor particles has been tried, the packing
density achieved has always been too low and the pulse output never
approaches that of the single crystal NaI:Tl phosphor.

c. Long Decav Phosphors

This is the last subject we will examine. Long decay phosphors are used in
phosphorescent paints as distinguished from radioluminescent paints. In
the latter, radium ( 226Ra28) is added to a suspension of a phosphor like
ZnS:Cu or ZnS:Ag. The emitted alpha particles excite the phosphor and
the paint glows on a continuous basis since the half-life of Ra is 1,600
years. In contrast, phosphorescent paint glows because of the phosphors
used as a pigment in its manufacture. These phosphors are prepared so
that deep-traps appear in its Brilloiun bands of the solid. As a result, when
they are excited by visible (mostly blue) light or near UV, they glow for
hours because the deep traps are emptied very slowly at room
temperature. We will not attempt to describe the exact mechanism of how
these traps are depopulated. Briefly, it involves "hole-traps" near the
valance band. The phosphors now in used for making such paints is shown
as follows:
Table 7-8
Phosphor Color Peak-nm After-Glow* Persistance*
(10 min) (min)
3+
CaSrS:Bi blue 450 5 90
3+ 3+
CaAl2O4:Eu :Nd blue 440 35 1000
3+
SrAl2O4:Eu green 520 30 2000
3+ 3+
SrAl2O4:Eu :Dy green 520 ~ 100 > 2000
ZnS:Cu,Cl y-green 530 45 200
ZnS:Cu:Co y-green 530 40 > 500
3
CaSrStEu^Tm * red 650 12 45
* Initial After-Glow = 100 when exposed to 100 lux of 300-400 nm light
# Persistence is defined as time to decay to 0.3 mcd./m2, the limit of light
perception by the human eye.

When these phosphors are mixed with a varnish resin to form a paint, the
702 Current phosphor device technology

result is used for marking stairwells and the like where a sudden loss of
light is compensated by the phosphorescent paint. The output is sufficient
to allow easy egress by any stranded person from the darkened vantage
point.
 Subject index 703

(h,k,l) planes 15 Bridgeman parameters 299

ACTFEL displays 674 Brillouin energy zones 359
ACTFEL phosphors 675 Brillouin Zones 364
Activator choices 453 Brillouin Zones in copper 367
Activator choices 462 Brilluoin zones 18
Activator concentrations 480 Brouwer's method 88
Activators as impurities 395 Bubble memories 62
Adsorbed water 209 Bulk density 227
AgBr ex. -Brouwer's method 92 Calcium phosphate ripening 171
Allowed band structure-solids 363 Calcium sulfide phosphor 64
Analogue reflectance- Optical 538 Calculation of energy bands 370
Analogue system-Emittance 537 Calculation of Frenkel pairs 72
Analogue system-Emittance 539 Calculation of Schottky pairs 72
Analogue system-Reflectance 537 Candela- definition 525
Anion choices-phosphors 458 Cathode ray tubes 615
Anion sub-lattice 46 Cation choices- phosphors 457
Anion vacancies 48 Cation sub-lattice 46
Anti-Stokes phosphors 594 Cation vacancies 48
Arrhenius equation 214 Ce and Y fractions 556
Associated defects 75 CeH2 example 79
Backlight displays 677 Celsius 5
Band model for copper 365 Changes of state 1
Bandgaps in diodes 652 Charge compensation 40
Beer-lambert law 354 Charge compensation 47
BET equation 267 Charge compensation 106
Binomial theorem 224 Charge transfer orbitals 460
Black body radiation 528 Charge transfer states 585
Black-body radiation 349 Chemical vapor deposition 641
Blacklight lamp phosphors 684 Chroma 523
Bloch function 363 Chromaticity coordinates,x,y,z 532
Bloch functions 371 CIE chromaticity diagram 533
Boltzmann equation 234 CIE MacAdam system-1964 551
Bragg equation 12 Clausius-Clapeyron equation 204
Bravais lattices-14 22 Color AM-LCD phosphors 681
Bridgeman components 299 Color AM-LCD structure 680
Bridgeman method- crystals 298 Color center 52
704 Subject index

Color center band model 384 Czochralski rotation 281

Color Comparator 527 Daylight defined 529
Color instrument 536 Daylight phosphors 398
Color matching 542 Decay times 397
Color measurement 520 Decay times- phosphors 409
Color mixing-sub tractive 544 Defect disorder- AgBr 100
Color phosphor triads 504 Defect equation 40
Color registration 616 Defect equations 53
Color spaces 543 Defect equilibria 73
Color TV- current phosphors 622 Defect free energy, Gd 70
Color- lever principle 542 Defect pairs 80
Color-emissive & reflective 541 Defect pairs 106
Commercial CRT phosphors 501 Defect solid 30
Commercial lamp phosphors 514 Defect thermodynamics 68
Computer monitors 630 Defect-free single crystal Si 322
Copy machine phosphors 683 Degrees of freedom 1
Copying machines 683 Density of states 401
Cr3+ in ruby 450 Differential Thermal Analysis 184
CRT flat faceplate 618 Diffusion by small cations 146
Crucible compositions 273 Diffusion controlled mixtures 159
Cryopumps 640 Diode displays- organic 667
Crystal field splitting rules 581 Diode laser arrays 662
Crystal freezing- incipient 302 Diode laser construction 660
Crystal growth and shape 225 Diode laser power output 666
Crystal stoichiometry 319 Diode laser progress 665
Cube 9 Diode lasers 659
Cumulative Frequency plot 244 Diode lasers-medical uses 667
Czochralski automatic system 289 Diode lasers-quantum well 664
Czochralski crystal defects 283 Dipole moments 425
Czochralski crystal rotation 286 DSC and boiling points 204
Czochralski heat flows 285 DSC block 198
Czochralski interface angle 281 DSC uses 203
Czochralski method 276 DTA applications 200
Czochralski operation 281 DTA heating rates 191
Czochralski parameters 279 DTA of glass points 203
Czochralski pulling apparatus 284 DTA parameters 188
 Subject index 705

DTA thermocouple 187 F-center: NaCl 53

3+
Dy energy levels-% SLJ 572 Fahrenheit 5
Dynamic TGA 208 FED outdoor display 651
Dystectic crystal composition 320 FED phosphors 650
Edge defined crystal growth 306 Ferefs diameter 231
Edge dislocation 33 Fermi surface in solids 364
Effluent gas analysis 209 Fick's first law 137
Einstein coefficients 422 Field emission display (FED) 648
EL displays 674 Firing atmosphere 493
Electric moment 433 Fluorescent lamp sizes-2003 625
Electromagnetic spectrum 358 Flying spot scanner 632
Electron as particle of matter 350 Forced dipole selection rules 576
Electron beam deposition 645 Forced-electric-dipole 575
Electron losses in a solid 405 Fractional parentage 568
Electron restricted energy 361 Franck-Condon principle 400
Electron-gun components 503 Free electron energy 361
Electronic devices- thin film 638 Freeman & Carroll method 215
Embryo 125 Frenkel defects 48
Embryo motion energy 127 Frenkel- Thermal origin 68
Emittance & reflectance 526 Frequency Plot 244
Endothermic-exothermic 184 Fresnel's equation 356
Energy band models 20 Furnace design 269
Energy band models 369 Furnace elements- non air 271
Energy transfer overlap 433 Furnace insulation 270
Energy transfer- spacial 436 d 0 reciprocal lattice vector 372
Energy transfer-critical radii 439 Miller indices 374
Energy transfer-spin coupling440 Gaussian PSD 235
Energy transition moments 424 Ge:As energy band models 384
Entropy change- ASvib. 69 Getter pumps 640
ErPO4 reflectance spectrum 578 Gibbs-Thompson equation 126
Evaporation 641 Grain boundaries 176
Ewald and Ewald Sphere 17 Grain growth 174
Ewald construction 376 Ground state energy 401
Excess cations in phosphors 483 Ground state perturbation 454
Exchange process 595 Hamiltonian operator 370
Excitation sources-phosphors 405 Hancock & Sharp method 143
706 Subject index

Hartree-F6k method 359 Intrinsic defect band models 386

Heat capacity 6 Intrinsic defects 75
Heat exchanger method 303 Ion beam deposition 644
Heat- definition 6 Ionization of defects 85
Heating elements 269 Isothermal TGA 208
HEM apparatus 303 JEDEC & RE CRT phosphors 510
HEM apparatus for CaF2 304 JEDEC phosphor types 510
Heterogeneous solids 45 JEDEC phosphors 507
Heterotype system 56 Jump frequency 138
HID & Triband Lamps 626 Kelvin 6
HID lamps 623 Kirchendall effect 141
Hierarchy- crystal systems 21 Kubela and Monk equation 356
Histogram 244 Kyropoulos method 300
Homogeneous nucleation 167 L-S coupling 414
Homotype impurity system 56 Lamp phosphors-specialty 515
Hooke's law 389 Lamps- 32T6 performance 626
Hopping mechanism 119 Lamps- F32T5 performance 627
Hopping motions 139 Lamps- micro-wave energy 627
Host choices 456 Lanthanide configurations 560
Host structure 493 Lanthanide chemistry 554
HPMV lamp 513 Lanthanide coupling scheme 564
HPMV lamp phosphors 515 Lanthanide electron shells 559
Human eye 521 Lanthanide energy states 561
HUYGHEN'S principle 351 Lanthanide free-ion approach 582
Hydrothermal growth-quartz 330 Lanthanide history 552
Hydrothermal parameters 333 Lanthanide J-levels 563
Illuminant C 529 Lanthanide levels possible 561
Image intensifier tube 636 Lanthanide ores 555
Impurity effects 107 Lanthanide salt solubility 555
Impurity leveling effect 309 Lanthanide spectra- factors 576
Impurity leveling factor 312 Lanthanides- 5d states 585
Infra-red converting 596 Lanthanides- process 557
Interfacial energy 165 Lanthanides-relative intensity 584
Intermed. coupling diagram 570 Laser crystals 290
Internal heat 8 Laser energy losses 609
Interstitial atoms 72 Laser slope efficiency 609
 Subject index 707

Laser- 3 level 612 Milestones in CRT guns 619

Lattice points 13 Miller indices 14
Lattice-unit-vectors 14 Min. Perceptual Differences 549
Laue equations 26 Mismatched energy exchange 443
Laue equations 375 Mixed states 573
LED construction 656 Mixed states of Dy3+ 575

LED formula 656 Molar extinction coefficient 355

LED progress 653 Molecular orbitals 417
LED- phosphors in use 659 Molten flux method 328
LED-commercial applications 659 Molten flux parameters 329
LED-inverted pyramid 658 Multiphonon rate equations 590
Light intensity 526 Multiphonon relaxation 588
Light interaction in solids 353 MULTIPOLAR selection rules 431
Light- particle nature 347 Multipole interactions 421
Light- wave nature 346 Multipole types 416
Line defect 34 Munsell color tree 544
Liquid crystal types 677 Munsell- hue.chroma & value 544
Lithium niobate 60 MXg defect equilibria 78
Log normal probability plot 246 MXg defects 76
Log normal PSD types 248 NaI:Tl+ scintillator 696
Long decay phosphors 701 Nanotechnology & purity 109
LPMV line emitters 516 Nature of light 345
Luminescence killers 469 Neon signs 690
Luminescent center 402 Non-stoichiometry factor, 6 82
Luminosity 524 Non-stoichiometric MXs±s 88
M-center 54 Nucleation of particles 122
MacAdam- MPD steps 548 Nucleation rates 122
Market penetration- displays 682 Nucleation-shape factor 132
Martin's diameter 231 Nucleation-random 125
McLeod gauge 639 Nuclei & phase boundary 127
Mechanical film deposition 642 Nuclei and sintering 179
Melt growth parameters 274 Nuclei formation and growth 131
Mercury diffusion pumps 639 Octahedron 9
Metal element structure 10 Oil diffusion pumps 640
Metal halide lamp 624 OLED construction 670
Migrating charged vacancy 163 OLED devices 668
708 Subject index

OLED devices and phosphors 673 Phosphor coupling processes 436

OLED displays 672 Phosphor dependent variable 486
OLED displays-design 668 Phosphor energy bands 403
OLED prerequisites 669 Phosphor energy dissipation 399
Optical formulas-solids 352 Phosphor energy processes 398
Organic phosphors 417 Phosphor firing cycle 488
Oscillator strengths 428 Phosphor luminescent center 447
Oscilloscopes 630 Phosphor methods 499
Ostwald ripening 170 Phosphor notation 407
Outdoor displays- CRT 651 Phosphor parameters 480
Parabolic law of diffusion 134 Phosphor preparation steps 486
Particle Assemblages 226 Phosphor size and mass 492
Particle counting instrument 232 Phosphor terms 397
Particle growth limits 248 Phosphors for projection TV 622
Particle in a box 370 Phosphors for Storage tubes 631
Particle mean size 234 Phosphors- flying spot 633
Particle orientation 230 Phosphors-image intensifiers 637
Particle packing densities 227 Phosphor energy dissipation 419
Particle properties 223 Photon as force-carrier 350
Particle ranges 219 Photon emission from solids 387
Particle shapes 224 Photon energy transfer 357
Particle size parameters 237 Photon interaction in solids 356
Particle size ranges 221 Photon interactions in solids 353
Particle sizes in Nature 233 Photopic & Scotopic vision 523
Particles by microscope 228 Piezoelectricity 334
PDP panel 686 Plane spacing 28
Phase boundary 117 Plaque tester 517
Phase transformations 115 Plasma discharge 685
Phonon acoustical and optical 394 Plasma display construction 686
Phonon branches in solids 394 Point defects 31
Phonon elastic waves 388 Point defects- homogenous 42
Phonon emission in YVO4 593 Point group symmetries 24
Phonon processes-phosphors 444 Polystructures 326
Phonon resonant vibrations 359 Polythiophene emission 672
Phonon vibrational modes 394 Polytypes 326
Phosphor band shapes 411 Pore growth 172
 Subject index 709

Potential energy diagram 400 Resonance ex.- definitions 433

Powder Diffraction File 26 Resonance exchange 433
Precipitation mechanisms 169 Resonant energy transfer 419
Probability of emission 426 Roentgen 12
Projection TV 621 Rotational symmetry 21
Propagation models 8 Rotational symmetry axes 15
PSD by Andreasen pipette 260 Rules for phosphor design 464
PSD by BET method 266 Russell-Saunders terms 414
PSD by Coulter Counter 261 Scanning calorimetry 198
PSD by gas adsorption 266 Schottky - Thermal origin 68
PSD by laser diffractometry 264 Schottky defects 48
PSD by permeability 266 Schroedinger equation 370
PSD by sedimentation 258 Scintillator mechanisms 697
PSD calculations 254 Scintillator phosphors 700
PSD in Industry 252 Scintillator uses 699
PSD-bimodal types 251 Scintillators 697
PSD- growth limited 247 Screen sizes- standard 220
Pyrophosphate phosphors 495 Screw dislocation 35
Quantum efficiency 408 Seebeck effect 185
Quantum efficiency 520 Seed crystal 275
Quartz AT and BT cuts 336 Seed crystal by crucible 276
Quartz resonators- watches 335 Seive method - PSD 239
Quartz twinning 324 Selection rules 416
Quartz vibrational modes 337 Self interstitial 40
Quartz-microbalance 338 Self-activated phosphors 459
Racah operators 567 Semiconducting Lasers 661
Radar phosphors 635 Sensitized luminescence 435
Radiation scattering 352 Sensitizers-activators 463
Radiative transfer 432 Shape parameters 224
Radiofrequency generators 279 Silicate diffusion 153
Random nucleation 130 Silicate phosphors 161
Rankine 6 Silicon boule defects 323
Rayleigh scattering in air 351 Single crystal growth rate 291
Reciprocal lattice 16 Single crystal vapor methods 340
Reciprocal lattice 360 Sintering equation 178
Reciprocal lattice wave Eq. 378 Sintering mechanisms 182
710 Subject index

Sintering of aggregates 174 Tetrahedron 9

Sintering of particles 172 TFT-LCD construction 678
Slater determinants 566 TFT-LCD pixels 680
Slater integrals-F values 566 TGA analysis 208
Solid state lasers 607 TGA parameters 206
Solid state rate equations 214 TGA water types 209
Solid state rate processes 211 TGA-solid state reactions 210
Solid state reactions 116 The plane net 49
Space groups-231 20 The Standard Observer 522
Spectrofluorimeter 519 Thermocouples- common 186
Spectroscopic notation 413 Thermogravimetric analysis 205
Speed of light 348 Thermolum. phosphor 485
Spin exchange 440 Thermometer 5
Spin-orbit & crystal field 571 Thermometric fixed points 195
Spinel 145 Thin film techniques 639
Spinel phosphors 147 Thin film technology 637
Sputtering 643 Thomaschek 584
Stacked CRTs 620 Time frame in phosphors 404
Stacking faults in crystals 324 TiO example 79
Stacking sequences-crystals 325 Translation vector 14
Standard deviations, a 235 Tri-color PDP phosphors 690
Standard Observer 526 Triband phosphors - 2003 626
Stark states- experimental 577 Tristimulous response, ~x,~Q,~z 532
Statistical Mechanics 65 Tristimulous values, X,Y,Z 532
Std. Illuminating sources 530 Types of defects 41
Stimulated emission 607 Unit cell 14
Stoichiometric MXs 87 Unit cell volume 14
Stokes law 244 Up-conversion phosphors 595
Stokes process 398 US standard seive sizes 238
Stokes shift 403 Use of a flux 500
Super-cooled melt 301 Vacancy 32
Surface and volume nucleation 123 Vacuum fluorescent display 645
Symmetry elements of a cube 15 Vacuum forepump 640
Tarnishing 133 Vacuum pumps 639
Temperature broadening 478 Vacuum UV generation 685
Temperature quenching 474 Vapor growth parameters 341
 Subject index 711

Vegard's law 464 X-ray intensifying phosphors 693

Verneuil method 315 X-ray intensifying screens 691
Verneuil parameters 318 X-ray phosphor attributes 692
VFD phosphors 647 X-ray summary 382
Vibronic coupling 401 Xerographic process 683
Vibronic spectrum- YVO4:Eu 589 YAG:Nd crystal growth 292
Volume defect 35 Zero phonon mode 138
Volume of domains 131 Zero phonon mode 402
Volume sintering 177 Zero vibronic coupling 412
Von Laue 12 ZnCdS2:Ag phosphors 506
VUV phosphors 689 ZnO- defect reactions 465
VUV Plaque Tester 688 ZnS bandgap 467
William Henry Bragg 12 ZnS:Ag charge-compensation 467
X-Ray Detection phosphors 695 ZnS:Ag defect reactions 467
X-ray image tube 694 Zone melting 307
X-ray image tube phosphors 695 Zone refining of crystals 308
This page is intentionally left blank