Upload
others
View
17
Download
1
Embed Size (px)
Citation preview
May 2001 Short Rietveld Course, Atlanta
Short Rietveld Course, Atlanta May 2001
The Rietveld Method
What is Rietveld refinement?Structural and profile parameters,standard deviations and R-factors
May 2001 Short Rietveld Course, Atlanta
Juan Rodríguez-CarvajalLaboratoire Léon BrillouinCEA/Saclay FRANCE
May 2001 Short Rietveld Course, Atlanta
Juan Rodríguez-CarvajalService de Physique Statistique, Magnétisme et SupraconductivitéDRFMC-CEA/Grenoble, France
May 2001 Short Rietveld Course, Atlanta
A powder diffraction pattern can be recorded in numerical form for a discrete set of scattering angles, times of flight orenergies. We will refer to this scattering variable as : T. The experimental powder diffraction pattern is usually given as two arrays :
The profile can be modelled using the calculated counts: yciat the ith step by summing the contribution from neighbouring Bragg reflections plus the background.
1,...,,i i i n
T y=
A powder diffraction pattern from the data treatment point of view
May 2001 Short Rietveld Course, Atlanta
yi
Position “i”: Ti
Bragg position Th
yi-ycizero
Powder diffraction profile:scattering variable T: 2θ, TOF, Energy
May 2001 Short Rietveld Course, Atlanta
profileThe model to calculate a powder diffraction pattern is:
( )h hh
ci i iy I T T b= Ω − +∑( ) 1x dx
+∞
−∞Ω =∫
Profile function characterized by its full width at half maximum (FWHM=H)and shape parameters (η, m, ...)
( ) ( ) ( )x g x f x instrumental intrinsic profileΩ = ⊗ = ⊗
The profile of powder diffraction patterns
May 2001 Short Rietveld Course, Atlanta
The profile of powder diffraction patterns
( )h hh
c i iiy I T T b= Ω − +∑Contains structural information: atom positions, magnetic moments, etc( )h h II I= β
( , )h PixΩ = Ω β Contains micro-structural information: instr. resolution, defects, crystallite size, ..
( )Bi ib b= β Background: noise, diffuse scattering, ...
May 2001 Short Rietveld Course, Atlanta
, ,h
( )h hci i iy s I T T bφ φ φφ
= Ω − +∑ ∑
Several phases (φ = 1,nφ) contributing to the diffraction pattern
, ,h
( )h h
p p p p pci i iy s I T T b
φφ φφ
= Ω − +∑ ∑
Several phases (φ = 1,nφ) contributing to several (p=1,np) diffraction patterns
May 2001 Short Rietveld Course, Atlanta
The Rietveld Method consist 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: β
22
1
( )n
i i cii
w y yχ β=
= −∑
21i
iwσ
=2iσ : is the variance of the "observation" yi
May 2001 Short Rietveld Course, Atlanta
Least squares: Gauss-Newton (1)Minimum necessary condition:
A Taylor expansion of around allows the application of an iterative process. The shifts to be applied to the parameters at each cycle for improving χ2 are obtained by solving a linear system of equations (normal equations)
2
0∂=
∂χβ
( )icy β 0β
0
0 0
0
( ) ( )
( )( )
A b
ic ickl i
i k l
ick i i ic
i k
y yA w
yb w y y
β β
β
=
∂ ∂=
∂ ∂∂
= −∂
∑
∑
βδ
β β
β
May 2001 Short Rietveld Course, Atlanta
Least squares: Gauss-Newton (2)
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 variance of the adjusted parameters are calculated by the expression:
The shifts of the parameters obtained by solving the normal equations are added to the starting parameters giving rise to a new set
01 0= + ββ β δ
1( ) ( )Ak kk
N - P+C
2 2ν
22ν
σ β χ
χχ
−=
=
May 2001 Short Rietveld Course, Atlanta
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.
May 2001 Short Rietveld Course, Atlanta
The structural information contained in the integrated
intensities
May 2001 Short Rietveld Course, Atlanta
2h h
I jL pO ACF=Integrated intensities are proportional to the square of the structure factor F. The factors are: Multiplicity (j), Lorentz-polarization (Lp), preferred orientation (O), absorption (A), other “corrections” (C) ...
Integrated intensities
May 2001 Short Rietveld Course, Atlanta
The Structure Factor contains the structural parameters
( ) ( ) 1
2h h h t rn
j j j jsj s
F O f T exp i Sπ=
= ∑ ∑
( , , ) ( 1, 2, ... )r j j j jx y z j n= =
( ) ( ) ( ), , , , 1, 2, ...h Ts s Gs
h k l h k l S s N= = =
sinexp( )j jT B2
2
θλ
= −
May 2001 Short Rietveld Course, Atlanta
Structural Parameters(simplest case)
( , , )r j j j jx y z= Atom positions (up to 3nparameters)
jO Occupation factors (up to n-1 parameters)
jB Isotropic displacement (temperature) factors (up to n parameters)
May 2001 Short Rietveld Course, Atlanta
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.
May 2001 Short Rietveld Course, Atlanta
The peak shape function of powder diffraction patterns contains the Profile Parameters
(constant wavelength case)
( , ) ( , )h P h Pi ix T Tβ βΩ = Ω −
( ) 1x dx+∞
−∞Ω =∫
The cell parameters are included, through Th, within the profile function. They determine the peak positions in the whole diffraction pattern.
May 2001 Short Rietveld Course, Atlanta
2( ) exp( )G GG x a b x= −
2
2 ln2 4ln2G Ga b
H Hπ= =
Gaussian function
Integral breadth:1
2 ln 2GG
Ha
πβ = =
May 2001 Short Rietveld Course, Atlanta
2( )1
L
L
aL xb x
=+
2
2 4L La b
H Hπ= =
Lorentzian function
Integral breadth:1
2LL
Ha
πβ = =
May 2001 Short Rietveld Course, Atlanta
Fwhm
BG
x0
x
I
Comparison of Gaussian and Lorentzian peak shapes of the same peak height “I” and same width “Fwhm”
May 2001 Short Rietveld Course, Atlanta
Convolution properties of Gaussianand Lorentzian functions
1 2 1 2
2 21 2 1 2
( , ) ( , ) ( , )
( , ) ( , ) ( , )
L x H L x H L x H H
G x H G x H G x H H
⊗ = +
⊗ = +
( , ) ( , ) ( , , )L G L GL x H G x H V x H H⊗ =
May 2001 Short Rietveld Course, Atlanta
( ) ( ) (1 ) ( )pV x L x G xη η′ ′= + −The pseudo-Voigt function
The Voigt function
( ) ( ) ( ) ( ) ( )V x L x G x L x u G u du+∞
−∞= ⊗ = −∫( ) ( , , ) ( , , )L G L GV x V x H H V x β β= =
( ) ( , , )pV x pV x Hη=
May 2001 Short Rietveld Course, Atlanta
Mapping: Pseudo-Voigt ⇔ VoigtThomson-Cox-Hasting formulation
( , ) ( , )G LH F H Hη =
1( , ) ( , )G LH H F H η−=
5 4 3 2 2 3 4 5( 2.69269 2.42843 4.47163 0.07842 )G G L G L G L G L LH H H H H H H H H H H= + + + + +2 3
1.36603 0.47719 0.11116L L LH H HH H H
η = − +
2 30.72928 0.19289 0.07783LHH
η η η= + +
2 3 1/ 2(1 0.74417 0.24781 0.00810 )GHH
η η η= − − −
May 2001 Short Rietveld Course, Atlanta
2 22tan tan
cosG
GIH U V Wθ θθ
= + + +
tancosL
YH X θθ
= +
Profile Parameters(simple cases)
Parameters controlling the Full-Width at half maximumU, V, W, IG, X, Y
May 2001 Short Rietveld Course, Atlanta
2 2 2 22( (1 ) ) tan tan
cosG
G STIH U D V Wξ θ θθ
= + − + + +
[ ( )]( ) tancos
ZL ST
Y F SH X Dξ θθ
+= + +
Modeling the Gaussian and Lorentziancomponents of the profile function in terms of microstructural parameters
May 2001 Short Rietveld Course, Atlanta
Profile R-factors used in Rietveld Refinements
May 2001 Short Rietveld Course, Atlanta
R-factors and Rietveld Refinements (1)
, ,
,
100obs i calc i
ip
obs ii
y yR
y
−=
∑∑
1/ 22, ,
2,
100i obs i calc i
iwp
i obs ii
w y yR
w y
−
=
∑
∑
R-pattern
R-weighted pattern
1/ 2
2,
( )100expi obs i
i
N P CRw y
− + =
∑
Expected R-weighted pattern
May 2001 Short Rietveld Course, Atlanta
R-factors and Rietveld Refinements (2)
Reduced Chi-square
Goodness of Fit indicator
2
2 wp
exp
RRνχ
=
wp
exp
RS
R=
May 2001 Short Rietveld Course, Atlanta
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
May 2001 Short Rietveld Course, Atlanta
Crystallographic R-factors used in Rietveld Refinements
, ,
,
' '100
' '
obs k calc kk
Bobs k
k
I IR
I
−=
∑∑
Bragg R-factor
, ,
,
' '100
' '
obs k calc kk
Fobs k
k
F FR
F
−=
∑∑
Crystallographic RF-factor.
May 2001 Short Rietveld Course, Atlanta
Crystallographic R-factors used in Rietveld Refinements
Provides ‘observed’ integrates intensities for calculating Bragg R-factor
In some programs the crystallographic RF-factor is calculated using just the square root of ‘Iobs,k’
,, ,
,
( )( )' '
( )i k obs i i
obs k calc ki calc i i
T T y BI I
y B Ω − − = −
∑
,,
' '' ' obs k
obs k
IF
jLp=
May 2001 Short Rietveld Course, Atlanta
Simulation of systematic errors in Rietveld refinements
•What is the effect of the resolution and the systematic errors in the profile peak shape function in the quality of the refined structural parameters?
•Are the estimated standard deviations a measure of the accuracy of the refined parameters?
•Are the R-factors good indicators of the quality of a structural model?
Important questions:
May 2001 Short Rietveld Course, Atlanta
Complexity of a structural problem:effective number of reflections and solvability
indexIf one is interested in “structural parameters” the number of independent observations is not the number of points in the pattern N.What is the number of “independent” observations? (No rigorous answer …)Points to be considered:
• Signal-to-noise ratio, statistics.• Number of independent Bragg reflections: NB• Number of structural free parameters: NI=Nf• Degree of reflection overlap: resolution versus separation
between consecutive reflections.
• Effective number of observations (resolution weighted): Neff• “Solvability” index: ratio between the effective number of
observations and the number of structural parameters: r = Neff/NI
May 2001 Short Rietveld Course, Atlanta
Effective number of reflections Neff
Two reflections separated by ∆(Q) can be discriminated properly if the following relation holds:
∆(Q) = 2π2j/(Q2Vo) ≥ p DQ
DQ is the FWHM in Q-space, p is of the order of the unity A single reflection at Qo contributes to Neff as 1/(1+Nn), where Nn is the number of reflections in the neighbourhood of Qo , verifying:
Qo - p DQ ≤ Qn ≤ Qo + p DQ
1,
11
B
effi N i
NN=
=+∑The formula for calculating Neff
is:
May 2001 Short Rietveld Course, Atlanta
Plot of ∆Q and DQ versus Q
May 2001 Short Rietveld Course, Atlanta
Simulation of systematic errors in Rietveld refinements (1)
(1) For a hypothetical (or real) compound characterised by a given crystal structure, scale factor (related to a “ counting time ”), resolution function, peak shape, angular range and background level, calculate a theoretical (deterministic) diffraction pattern: yi(theo)
(2) Nr versions of the “ observed ” diffraction pattern are generated by corrupting the deterministic pattern by a noise calculated with a generator of pseudo-random numbers according to the Poisson distribution : yi(obs) = Pyi(theo).
(3) The process is repeated for Nt different values of the “ counting time ” (scale factor), thus producing Nr x Ntobserved diffraction patterns for the same compound.
Method:
May 2001 Short Rietveld Course, Atlanta
Simulation of systematic errors in Rietveld refinements (2)
(4) Each pattern is refined by the RM by using either the “ true ” model or a biased model (e.g. wrong peak shape).
(5) The values of the refined parameters are then compared to the true valued (bias and dispersion) and the computed standard deviations are compared to the empirical ones obtained from the Nr versions of the same diffraction pattern, as a function of the counting time.
(6) Finally we study the behaviour of the R-factors versus counting time for the different conditions of refinement.
(7) Refine the diffraction patterns without structural model to obtain the “ expected ” R-factors for the best structural model by profile-matching iterating the Rietveld formula for the Bragg R-factor (Lebail fit)
May 2001 Short Rietveld Course, Atlanta
Results of simulations: Refinements of simulated
powder diffraction patterns using correct and biased models
May 2001 Short Rietveld Course, Atlanta
0
2000
4000
6000
8000
10000
0.0 25.0 50.0 75.0 100.0 125.0 150.0
Refinement of simulated patternwith biased peak shape and correct
structural model.NIntdp=39, Nprof=5, Nref=1507
Neff=(70, 137, 248)/Solv=(1.8,3.5,6.4)
Inte
nsity
(a.u
.)
2Θ (°)
RM (systematic errors): Biased peak shape+ correct structural model NI=39 + bad resolution
May 2001 Short Rietveld Course, Atlanta
0
4000
8000
12000
0.0 25.0 50.0 75.0 100.0 125.0 150.0
Refinement of simulated patternwith biased peak shape and correct
structural model.NIntdp=39, Nprof=5, Nref=1507
Neff=(185, 352, 611)/Solv=(4.8,9.0,15.7)
Inte
nsity
(a.u
.)
2Θ (°)
RM (systematic errors): Biased peak shape+ correct structural model NI=39 + better resolution
May 2001 Short Rietveld Course, Atlanta
Behaviour of Rietveld R-factors, and other indicators,
versus counting statistics for perfect and biased models
May 2001 Short Rietveld Course, Atlanta
0
5
10
15
20
25
30
35
40
2,0 2,5 3,0 3,5 4,0 4,5 5,0 5,5
R-factors (RP, R
WP, R
Bragg and R
F)
cRp(B)cRwp(B)RbRF
R(%
)
Log(Counting Time)
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)
Correct model
RM (systematic errors): Behaviour of R-factors versus counting time (correct model)
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Behaviour of RWP factors versus counting time (biased peak shape)
0
10
20
30
40
50
2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
R-factors (RWP
) RwpcRwpRwp(B)cRwp(B)R
WP(%
)
Log(Counting Time)
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)Biased peak shape
May 2001 Short Rietveld Course, Atlanta
0
2
4
6
8
10
12
14
2,0 2,5 3,0 3,5 4,0 4,5 5,0 5,5
R-factors (RBragg
and RF)
RbRF
R(%
)
Log(Counting Time)
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)Biased peak shape
RM (systematic errors): Behaviour of RBragg and RFfactors versus counting time (biased peak shape)
May 2001 Short Rietveld Course, Atlanta
0,50
1,00
1,50
2,00
2,50
2,0 2,5 3,0 3,5 4,0 4,5 5,0 5,5
Durbin-Watson DW DWDW(S)
Dur
bin-
Wat
son
stat
istic
s
Log(Counting Time)
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)
DW(S): Biased peak shape
RM (systematic errors): Behaviour of DW indicator versus counting time (biased/unbiased peak shape)
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Behaviour of reduced Chi-square versus counting time (correct model)
0,600
0,800
1,000
1,200
1,400
2,0 2,5 3,0 3,5 4,0 4,5 5,0 5,5
R-factors(χ 2)
Chi2Chi2(B)
Red
uced
χ2
Log(Counting Time)
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)
Correct model
May 2001 Short Rietveld Course, Atlanta
0
10
20
30
40
50
60
2,0 2,5 3,0 3,5 4,0 4,5 5,0 5,5
R-factors(χ2)
Chi2Chi2(B)
Red
uced
χ2
Log(Counting Time)
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)Biased peak shape
RM (systematic errors): Behaviour of reduced Chi-square versus counting time (biased peak shape)
May 2001 Short Rietveld Course, Atlanta
Behaviour of positional parameters (atom co-ordinates), versus counting
statistics, resolution, ... for perfect and biased models
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Behaviour of positional parameters versus counting time (biased peak shape)
0,1380
0,1400
0,1420
0,1440
0,1460
0,1480
2,0 2,5 3,0 3,5 4,0 4,5 5,0 5,5
Coordinate "x" of the first atom x1
Red
uced
coo
rdin
ate
Log(Counting Time)
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)Biased peak shape
May 2001 Short Rietveld Course, Atlanta
Atom co-ordinates versus counting statistics
May 2001 Short Rietveld Course, Atlanta
-0,0300
-0,0200
-0,0100
0,0000
0,0100
0,0200
0,0300
0 2 4 6 8 10 12 14
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)Biased peak shapeLow counting time
Del
ta (Å
)
Positional Parameter
RM (systematic errors): Positional parameters NI=16 Low counting time (biased peak shape)
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters, NI=16 Intermediate counting time (biased peak shape)
-0,0300
-0,0200
-0,0100
0,0000
0,0100
0,0200
0,0300
0 2 4 6 8 10 12 14
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)Biased peak shape
Intermediate counting time
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters, NI=16 High counting time (biased peak shape)
-0,0300
-0,0200
-0,0100
0,0000
0,0100
0,0200
0,0300
0 2 4 6 8 10 12 14
NIntdp=16, Nprof=8, Nref=393Neff=(173, 244, 301)Biased peak shape
High counting time
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
Atom co-ordinates versus resolution (“solvability index”)
r=Neff / NI
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters NI=39, Counting Time: 12500 (correct model, r=3.5)
-0,0400
-0,0200
0,0000
0,0200
0,0400
0 5 10 15 20 25 30 35 40
NIntdp=39, Nprof=5, Nref=1507Neff=(70, 137, 248)/Solv=(1.8,3.5,6.4)
Correct model-(Counting time:12500)
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
-0,0400
-0,0200
0,0000
0,0200
0,0400
0 5 10 15 20 25 30 35 40
NIntdp=39, Nprof=5, Nref=1507Neff=(102, 192, 362)/Solv=(2.6,4.9,9.3)Correct model-(Counting time:12500)
Del
ta (Å
)
Positional Parameter
RM (systematic errors): Positional parameters NI=39, Counting Time: 12500 (correct model, r=4.9)
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters NI=39, Counting Time: 12500 (correct model, r=9)
-0.0400
-0.0200
0.0000
0.0200
0.0400
0 5 10 15 20 25 30 35 40
NIntdp=39, Nprof=5, Nref=1507Neff=(185, 352, 611)/Solv=(4.8,9.0,15.7)Correct model-(Counting time:12500)
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
Atom co-ordinates versus statistics (correct model, NI=72)
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters NI=72, Counting Time: 125 (correct model, r=5.1)
-0,3000
-0,2000
-0,1000
0,0000
0,1000
0,2000
0,3000
0 10 20 30 40 50 60 70 80
NIntdp=72, Nprof=5, Nref=3960Neff=(188, 368, 726)/Solv=(2.6,5.1,10.1)
Correct model-(Counting time:125)
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters NI=72, Counting Time: 1250 (correct model, r=5.1)
-0,1000
-0,0500
0,0000
0,0500
0,1000
0 10 20 30 40 50 60 70 80
NIntdp=72, Nprof=5, Nref=3960Neff=(188, 368, 726)/Solv=(2.6,5.1,10.1)Correct model-(Counting time:1250)
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters NI=72, Counting Time: 12500 (correct model, r=5.1)
-0,1000
-0,0500
0,0000
0,0500
0,1000
0 10 20 30 40 50 60 70 80
NIntdp=72, Nprof=5, Nref=3960Neff=(188, 368, 726)/Solv=(2.6,5.1,10.1)Correct model-(Counting time:12500)
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters NI=72, Counting Time: 125000 (correct model, r=5.1)
-0,1000
-0,0500
0,0000
0,0500
0,1000
0 10 20 30 40 50 60 70 80
NIntdp=72, Nprof=5, Nref=3960Neff=(188, 368, 726)/Solv=(2.6,5.1,10.1)Correct model-(Counting time:125000)
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
Atom co-ordinates versus statistics (biased peak shape, NI=39)
May 2001 Short Rietveld Course, Atlanta
-0,1000
-0,0500
0,0000
0,0500
0,1000
0 5 10 15 20 25 30 35 40
NIntdp=39, Nprof=5, Nref=1507Neff=(185, 352, 611)/Solv=(4.8,9.0,15.7)
Biased peak shape-(Counting time:12500)
Del
ta (Å
)
Positional Parameter
RM (systematic errors): Positional parameters NI=39 Counting Time: 12500 (biased peak shape, r=9)
May 2001 Short Rietveld Course, Atlanta
RM (systematic errors): Positional parameters NI=39 Counting Time: 12500 (biased peak shape, r=4.9)
-0,1000
-0,0500
0,0000
0,0500
0,1000
0 5 10 15 20 25 30 35 40
NIntdp=39, Nprof=5, Nref=1507Neff=(102, 192, 362)/Solv=(2.6,4.9,9.3)
Biased peak shape-(Counting time:12500)
Del
ta (Å
)
Positional Parameter
May 2001 Short Rietveld Course, Atlanta
-0,1000
-0,0500
0,0000
0,0500
0,1000
0 5 10 15 20 25 30 35 40
NIntdp=39, Nprof=5, Nref=1507Neff=(70, 137, 248)/Solv=(1.8,3.5,6.4)
Biased peak shape-(Counting time:12500)
Del
ta (Å
)
Positional Parameter
RM (systematic errors): Positional parameters NI=39 Counting Time: 12500 (biased peak shape, r=3.5)
May 2001 Short Rietveld Course, Atlanta
Conclusions (1)
(1) When the refined model is unbiased (no systematic errors), least squares estimates of the standard deviations represent not only the precision but also the accuracy. For biased models (usual practical situation) the standard deviations do not represent accuracy.
(2) For biased models (a known systematic error in profile is introduced) in simple structures some fitted parameters (profile and thermal parameters) show a bias while others are barely affected(position parameters).
(3) Concerning accuracy, nothing is gained repeating the measurement several times if the systematic error is expected to be the same. The standard deviation obtained from different measurement is not, in such a case, a measure of the accuracy.
May 2001 Short Rietveld Course, Atlanta
Conclusions (2)
(4) For complex structures high statistical accuracy and high resolution is required for getting the true parameter values even is the refined model is unbiased. Suggestion: the solvability index (r = Neff/Nf for p=1/2) should be largely greater than 4-5 to be sure that the structural parameters are accurate enough. More experience is needed to establish precise rules.
(5) The absolute value of the profile R-factors has little significance because their values depend on the quality of the data as well as on the goodness the structural model. The R-factors obtained by a refinement of the whole pattern without structural model provide the “expected” values for the best structural model.
(6) Well behaved peak shape could be more important than resolution in some cases.