The Rietveld Method

Juan Rodriguez-Carvajal
Institut Laue-Langevin
Diffraction Group

What is a Rietveld refinement?

Structural and profile parameters,
standard deviations and R-factors
 The calculated profile of
powder diffraction patterns
yci   I h (Ti  Th )  bi
{h}

Ih  Ih   I  Contains structural information:

atom positions, magnetic moments, etc
Contains micro-structural information:
  ( xhi ,  P ) instr. resolution, defects, crystallite size…

bi  bi   B  Background: noise, incoherent scattering

diffuse scattering, ...
 The calculated profile of
powder diffraction patterns
yci   I h (Ti  Th )  bi
{h}

The symbol {h} means that the sum is extended only to those
reflections contributing to the channel “i” .

This should be taken into account (resolution function of the

diffractometer and sample broadening) before doing the
actual calculation of the profile intensity.

This is the reason why some Rietveld programs are run in

two steps
 Several phases ( = 1,n) contributing
to the diffraction pattern

yci   s  I ,h (Ti  T ,h )  bi
 {h}

Several phases ( = 1,… n) contributing

to several (p=1,… np) diffraction patterns

y   s
p
ci
p
 I p
 ,h
 (Ti  T ,h )  bi
p p

 
{ h}
 yci   I h (Ti  Th )  bi
{h}

Ih  S L pO ACF 2
h
Integrated intensities are proportional
to the square of the structure factor F.
The factors are:
Scale Factor (S), Lorentz-polarization
(Lp), preferred orientation (O),
absorption (A), other “corrections” (C)
 The Structure Factor contains the
structural parameters
(isotropic case)

 
n
F  h    O j f j  h  T j  exp 2 i h S ts r j 
j 1 s

rj  ( x j , y j , z j ) ( j  1, 2,...n)
sin 
T j  exp( B j )
 
 Structural Parameters
(simplest case)

rj  ( x j , y j , z j ) Atom positions (up to 3n

parameters)

Oj Occupation factors (up to

n-1 parameters)

Bj Isotropic displacement
(temperature) factors (up
to n parameters)
 Structural Parameters
(complex cases)

As in the simplest case plus additional (or alternative)

parameters:
• Anisotropic temperature (displacement) factors
• Anharmonic temperature factors
• Special form-factors (Symmetry adapted spherical
harmonics ), TLS for rigid molecules, etc.
• Magnetic moments, coefficients of Fourier components
of magnetic moments , basis functions, etc.
 The Structure Factor in
complex cases

 
n
F  h    O j f j  h  T j  g j (h s ) exp 2 i h S ts r j 
j 1 s

h h
  T  
h s   k   Ss  k   s  1, 2,...NG 
l l
 s  
Complex form factor of object j
g j (h s ) Anisotropic DPs
Anharmonic DPs
 Example: General 2 dependence of the
instrumental broadening (determined by
a standard sample)
I fG
H 2
 (U f  (1   f ) D
2 2
(α D )) tan  
2
 H gG
2

cos2 
hG fST

[Y f  Ff (α S )]
H hL  ( X f   f D fST (α D )) tan    H gL
cos 
The Gaussian and Lorentzian components of the instrumental Voigt
function are interpolated between empirically determined values.
If needed, axial divergence is convoluted numerically with the
resulting profile.
 The Rietveld Method
The Rietveld Method consists of refining a crystal (and/or
magnetic) structure by minimising the weighted squared
difference between the observed and the calculated
pattern against the parameter vector: 

n
   wi  yi  yci (  )
2 2

i 1

wi  12
i
 2
i: is the variance of the "observation" yi
 Least squares: Gauss-Newton (1)
Minimum necessary condition:   0
2


A Taylor expansion of yic (  ) around  0 allows the application
of an iterative process. The shifts to be applied to the parameters
at each cycle for improving 2 are obtained by solving a linear
system of equations (normal equations)

A 0  b
yic (  0 ) yic (  0 )
Akl   wi
i  k  l
yic (  0 )
bk   wi ( yi  yic )
i  k
 Least squares: Gauss-Newton (2)
The shifts of the parameters obtained by solving the
normal equations are added to the starting parameters
giving rise to a new set
1   0    0

The new parameters are considered as the starting ones in the

next cycle and the process is repeated until a convergence
criterion is satisfied. The variances of the adjusted parameters
are calculated by the expression:
  (  k )  ( A 1 ) kk 

 
N - P+C
 Least squares: a local
optimisation method
• The least squares procedure provides (when it converges)
the value of the parameters constituting the local
minimum closest to the starting point

• A set of good starting values for all parameters is needed

• If the initial model is bad for some reasons the LSQ

procedure will not converge, it may diverge.
The peak shape function of powder diffraction patterns
contains the Profile Parameters
(constant wavelength case)

( xhi , P )  (Ti  Th , P )

 
( x)dx  1
The cell parameters are included, through Th, within the
profile function. They determine the peak positions in
the whole diffraction pattern.
 Gaussian function

G( x)  aG exp(bG x ) 2

2 ln 2 4ln 2
aG  bG  2
H  H

1 H 
Integral breadth: G  
aG 2 ln 2
 Lorentzian function
aL
L( x ) 
1  bL x 2

2 4
aL  bL 
H H 2

1 H
Integral breadth: L  
aL 2
Comparison of Gaussian and Lorentzian peak shapes
of the same peak height “I” and same width “Fwhm”

Fwhm

BG

x0
Convolution properties of Gaussian
and Lorentzian functions

L( x, H1 )  L( x, H 2 )  L( x, H1  H 2 )
G( x, H1 )  G( x, H 2 )  G( x, H  H )
1
2 2
2

L( x, H L )  G( x, HG )  V ( x, H L , HG )
 The Voigt function

V ( x )  L( x )  G ( x )   L( x  u)G(u)du

V ( x)  V ( x, H L , HG )  V ( x,  L , G )

The pseudo-Voigt function

pV ( x)   L( x)  (1 )G( x)
pV ( x)  pV ( x,, H )
 Mapping: Pseudo-Voigt  Voigt
Thomson-Cox-Hasting formulation
( H , )  F ( H G , H L )
H  ( HG5  2.69269HG4 H L  2.42843HG3 H L2  4.47163HG2 H L3  0.07842 HG H L4  H L5 )
2 3
H H  H 
  1.36603 L  0.47719  L   0.11116  L 
H  H   H 

( HG , H L )  F 1 ( H ,)
HL
 0.72928  0.19289 2  0.07783 3
H
HG
 (1  0.74417  0.24781 2  0.00810 3 )1/ 2
H
 Profile Parameters
(simple cases)

Parameters controlling the Full-Width at half maximum

U, V, W, IG, X, Y

IG
H  U tan   V tan   W 
2 2

cos2 
G

Y
H L  X tan  
cos
Modeling the Gaussian and Lorentzian
components of the profile function in
terms of microstructural parameters

IG
H  (U  (1   ) D ) tan   V tan   W 
2 2 2 2

cos2 
G ST

[Y  F ( SZ )]
H L  ( X   DST ) tan  
cos
Profile R-factors used in
Rietveld Refinements
R-factors and Rietveld Refinements (1)

 y y obs ,i calc ,i
Rp  100 i R-pattern
y i
obs ,i

1/ 2
 w y y 

2
i obs ,i calc ,i

Rwp  100  i  R-weighted pattern

2
 wi yobs ,i 
 i 
1/ 2
 
(N  P  C) 
Rexp  100  Expected R-weighted pattern
  w y2 
 i i obs ,i 
R-factors and Rietveld Refinements (2)
2
 Rwp 
  
2
 Reduced Chi-square
 Rexp 

Rwp
S Goodness of Fit indicator
Rexp
R-factors and Rietveld Refinements (3)

Two important things:

• The sums over “i” may be extended only to the
regions where Bragg reflections contribute
• The denominators in RP and RWP may or not contain
the background contribution
 Crystallographic R-factors
used in Rietveld Refinements
 ' I ' I
obs , k calc , k
Bragg R-factor
RB  100 k

 'I '
k
obs , k

 ' F ' F
obs , k calc , k
RF  100k Crystallographic RF-factor.
 'F '
k
obs , k
 Crystallographic R-factors
used in Rietveld Refinements

  Provides ‘observed’
 (Ti  Tk )( yobs ,i  Bi ) 
' I obs ,k '  I calc ,k    integrates intensities for
i 
 ( ycalc ,i  Bi ) 
 calculating Bragg R-factor

' I obs ,k ' In some programs the crystallographic

' Fobs ,k '  RF-factor is calculated using just the
jLp square root of ‘Iobs,k’