Electromigration 1: Tim Turner Keithley Instruments

Electromigration 1
Tim Turner
Keithley Instruments
[email protected]
Electromigration
Al
Self Diffusion
Vacancy
Black’s Equation
Lifetime = A (1/J)n e-ε /KT a
Where: J = Current Density

n = Current Density Exponent
εa = Thermal Activation Energy
K = Boltzmann’s Constant
T = Absoulte Temperature
Grain Boundary Diffusion
Grain
Grain
Boundary
Electromigration - Effect of Temp
Extracting fundamental physical parameters for use in reliability & TCAD

models from wafer level results – e.g., bulk and grain diffusion constants.
Bamboo Effect
N 99%
o
r
m 90% Wide Line
a
l
68%
P
r 50%
o Narrow Line
b
a 32%
b
i
l 10%
i
t
y 1%
Log Time
Phase Diagrams of Binary Alloy Systems
AI-Cu Phase Diagram
G. Borelius et. al. Handlinger NR 169, 1943
Al/Si Phase Diagram
Al
450 C
Al + Si
100% 0.5%
Al Si
Interface Diffusion
Ti/TiN
INTERMETALLIC
Al/Cu(0.5%)
INTERMETALLIC
Ti/TiN/Ti
INTERMETALLICS
Ti/TiN
INTERMETALLIC
Al/Cu(0.5%)
INTERMETALLIC
Ti/TiN/Ti
Ti/TiN
INTERMETALLIC
Al
INTERMETALLIC
Ti/TiN/Ti
Electromigration Test Structure 1
Electromigration Test Structure 2
Lower Average Temperature

Steep Temperature Gradient
Electromigration Lifetime
Distribution vs. Log Time
N 99%
o
r
m 90% Structure 2
a
l
68%
P
r 50%
o Structure 1
b
a 32%
b
i
l 10%
i
t
y 1%
Log Time
Log-Normal Distribution
• Lifetimes of a Line Follow a Gausian Distribution
• Distributions are a function of Line Length
Accelerated Electromigration
• SWEAT Test - Drives Constant Acceleration
Factor - Recommended for Processes with Poor
Line Width Control (Wet Etched Metal)
• Isothermal Test - Forces Constant “Temperature”

- Recommended for Dry Etched Metal Processes
• Constant Current Test - Most Easily Extrapolated

Test Results - Not Recommended for Highly
Accelerated Tests Which Have high Joule
Heating
Highly Accelerated Tests
• Joule Heating Produces Temperature Gradients
• Temperature Gradients Produce Flux
Divergancies
• Flux Divergancies Reduce Test Lifetime

• Black’s Equation Assumes Consistent
Temperatures and Current Densities - Can not be
applied to Joule Heated Tests
Cu Electromigration
• Higher Recrystalization Temperature
• Intermetallics on Three Sides
• Adhesion to Top Silicon Nitride Layer Not Good
(surface migration)
• Line Thickness Effect

• Sidewall Short Issue
Electromigration Sidewall Fracture
Expansion of Wide Cu Line Causes Fracture of

Sidewall Oxide Allowing Shorts
Expansion Test for Sidewall Strength
• Test Structure: Wide Metal Line Between 2 other
Wide Metal Lines
• Ramp up Current Through Line to Heat the Line

• Measure Temperature by the Change in
Resistance and the TCR
• Calculate Stress based on Thermal Expansion

and Measured Metal Space
Electromigration Interactions
• Line Width and Thickness Impacts Current
Density
- Photolithography Effects
-Etch Effects
-CMP Effects
METAL LINE WIDTH
MEASUREMENT
Wide Measure Low

Wide Measure High
Wide Line
Wide Force High
Force Low
Narrow Force High
Narrow Line
Narrow Measure Low Narrow Measure High
WIDE AND NARROW LINES ARE EACH AT

LEAST 800u LONG
2 Resister Line Width
Cross Bridge Line Width
Joule Heating Effects
• Joule Heating Effects are Generally Seen at
About 0.022 degrees C/watt/sq. u of area/u of
Oxide Thickness Under the Conductor
• All Conductors Have a TCR (Temperature

Coefficient of Resistance)
• Heating Causes Resistance Increase Which

Causes Higher Power Dissipation (Constant
Current) and Thus More Heating
• Heating Causes Inaccurate and Unstable

Resistance Measurements Leading to Inaccurate
and Unstable Line Width, Contact Resistance...
Joule Heating Effects
Temperature Van der Pauw Cross for
@10uA = 25 C Sheet Resistivity Measurement
@1mA = 43.1 C
R @ 25 C = 10 ohms
Voltage Drop
@10uA = 100uV
R @ 43 C = 10.82 ohms
@1mA = 10mV
(assume 0.5%/degree C)
8.2% error
Required Voltage Resolution
Voltage Resolution Required to Measure a Metal Line

Width with 0.5% Instrument Resolution
V o ltage Re s o lution (uV )
10
0.1
0.1 0.12 0.14 0.16 0.18 0.2
Line Width (u)
Line Width and Resistivity
Measurements on Al Lines
SMU 2 SMU 3
Measure V1 Measure V2
SMU 1
Force
I
Length of Test Line Drawn so that V1 - V2 = 100mV
Accuracy of Voltage Measure = 0.04% range +
0.04% Reading + (Rmat X Io) = +/- 870uV
Box Cross Resistivity Requires a
Differential Voltage Measurement
Conductor
V1 V1 - V2 = 7µV
870 µV Inst. Accuracy
is Now Very Significant
Box
I in Cross
V2
Current Limited
to prevent Joule
R = 4.53 (V/I)
Heating of
Connecting Lines
I out
PHOTOLITHOGRAPHY ISSUES
PHOTORESIST EROSION
PHOTORESIST
METAL 2
METAL 1
POLY
FIELD OXIDE
PHOTORESIST
METAL 2
METAL 1
POLY
FIELD OXIDE
Reflective Notching
REFLECTIVE NOTCHING
LIGHT
MASK
PHOTORESIST
METAL
POLY POLY
CMP Dishing
Wide Lines are Worst Case

Cu Line Width and Resistivity
Measurements
Cu Cu
Barrier Metal Barrier Metal
Sidewall Barrier Layers Complicate

Cu
Resistivity and Line Width Measurements
Grain Boundary Barrier Metal
Cu Cu Sidewall Barrier
is thinned for
Narrow Lines
Cu Grain Size is a Function of Line Width,
this impacts Resistivity
Cu Cu
CMP Dishing Can Increase Resistivity for

Wide Lines
Cu Resistivity Changes with Line Width
Al
Sheet
Resistivity
Cu
A B C D
0.2 2.0 5.0

Metal Line Width (µ)
Section A: Barrier Sidewall Thinning
Section B: Cu Grain Size Reduction
Section C: Increasing Effect of Sidewall
Section D: CMP Dishing
Cu Sheet Resistivity Measurement
V
Cu Sheet Resistivity = 4.53 V/I
Measures Cu/Barrier Sheet Resistivity. I I
Sidewall Barrier Has Negligible Effect.

Dishing Effect Detected with 2 Different Sizes
Measurement Noise Levels Required to Measure Sheet Resistivity
of Cu Metal Lines (Nominal 0.02 ohms/square)
Connecting Box Side Max. I for Max.

Line Width Length Measurement Noise
Min. Box Width (100X Barrier 1 5 4.77mA 1.1V
Layer Thickness of 0.05)
Typical Structure 3 15 14.3mA 3.2V
Max. Box Width (Limited by 7.2 36 34mA 7.5V

80 Scribe Lane)
Cu Line Width Measurement by
Capacitance
Metal 2 Metal 2
A A
Poly Poly
Scribe Lane Fingered Capacitors of Two Sizes

Used to Measure Line Width
Capacitance Noise and Resolution < 2.7fF
Required to make this measurement on a typical
Scribe Lane Capacitor
Electromigration Process Control
• Highly Accelerated Test to Measure Material
Effects (Isothermal with Wide/Narrow Structure)
- Grain Size and Texture (Wide Lines)
- Intermetallic Growth (Narrow Lines)
• Cu Lines Need Sidewall Strength Test

• Multiple Metal Line and Thickness Measurement
Structures to Look for Process Interactions that
will Impact Current Density

Electromigration 1: Tim Turner Keithley Instruments

Uploaded by

Electromigration 1: Tim Turner Keithley Instruments

Uploaded by

Electromigration 1

Lifetime = A (1/J)n e-ε /KT a

Where: J = Current Density

εa = Thermal Activation Energy

Extracting fundamental physical parameters for use in reliability & TCAD

Lower Average Temperature

• Isothermal Test - Forces Constant “Temperature”

• Constant Current Test - Most Easily Extrapolated

• Flux Divergancies Reduce Test Lifetime

• Line Thickness Effect

Expansion of Wide Cu Line Causes Fracture of

• Ramp up Current Through Line to Heat the Line

• Calculate Stress based on Thermal Expansion

Wide Measure Low

WIDE AND NARROW LINES ARE EACH AT

• All Conductors Have a TCR (Temperature

• Heating Causes Resistance Increase Which

• Heating Causes Inaccurate and Unstable

Voltage Resolution Required to Measure a Metal Line

Wide Lines are Worst Case

Sidewall Barrier Layers Complicate

CMP Dishing Can Increase Resistivity for

0.2 2.0 5.0

Measures Cu/Barrier Sheet Resistivity. I I

Sidewall Barrier Has Negligible Effect.

Connecting Box Side Max. I for Max.

Max. Box Width (Limited by 7.2 36 34mA 7.5V

Scribe Lane Fingered Capacitors of Two Sizes

• Cu Lines Need Sidewall Strength Test

You might also like