EXAFS Data Analysis Using FEFF

Prof. Anatoly Frenkel

 EXAFS spectroscopy: experimental data

Detector
Detector

Sample

Cryostat & sample

holder

1st detector

2nd detector
3rd detector

11/28/2024 2
 Origins of EXAFS

𝜇 ( 𝐸 )=𝜇 0 ( 𝐸 )(1+ 𝜒 ( 𝐸 ))

( 𝜇 ( 𝐸 ) − 𝜇0 ( 𝐸 ) )
𝜒 ( 𝐸)=
𝜇0 ( 𝐸 )

𝜒 ( 𝐸 ) → 𝜒 (𝑘)

𝑘=
ℏ√
2𝑚
2 ( 𝑝h𝑜𝑡𝑜𝑛
𝐸 − 𝐸0 )

11/28/2024 4
 Origins of EXAFS equation

EXAFS: result of Photoelectron wavenumber:

interference between
orinal photolectron wave
and backscattered wave
Contribution from a single
backsattering atom:

𝑅𝑖
2 𝑅𝑖
−
𝜆 (𝑘)
2 𝐹𝑖 ( 𝑘 ) 𝑒
𝜒 𝑖 ( 𝑘 )=𝑆 0 2
sin ( 2𝑘 𝑅𝑖 + 𝜙 𝑖 (𝑘)
11/28/2024
𝑘𝑅 𝑖 5
 Origins of EXAFS equation

EXAFS: result of Photoelectron wavenumber:

interference between
orinal photolectron wave
and backscattered wave
Contribution from a single
backsattering atom:

Interatomi
c distance
𝑅𝑖
2 𝑅𝑖
−
𝜆 (𝑘)
2 𝐹𝑖 ( 𝑘 ) 𝑒
𝜒 𝑖 ( 𝑘 )=𝑆 0 2
sin ( 2𝑘 𝑅𝑖 + 𝜙 𝑖 (𝑘)
11/28/2024
𝑘𝑅 𝑖 6
 Origins of EXAFS equation

EXAFS: result of Photoelectron wavenumber:

interference between
orinal photolectron wave
and backscattered wave
Contribution from a single
backsattering atom:

Effective
mean-free
𝑅𝑖
path
2 𝑅𝑖
−
𝜆 (𝑘)
2 𝐹𝑖 ( 𝑘 ) 𝑒
𝜒 𝑖 ( 𝑘 )=𝑆 0 2
sin ( 2𝑘 𝑅𝑖 + 𝜙 𝑖 (𝑘)
11/28/2024
𝑘𝑅 𝑖 7
 Origins of EXAFS equation

EXAFS: result of Photoelectron wavenumber:

interference between
orinal photolectron wave
and backscattered wave
Contribution from a single
backsattering atom:

Scattering
amplitude
𝑅𝑖
2 𝑅𝑖
−
𝜆 (𝑘)
2 𝐹𝑖 ( 𝑘 ) 𝑒
𝜒 𝑖 ( 𝑘 )=𝑆 0 2
sin ( 2𝑘 𝑅𝑖 + 𝜙 𝑖 (𝑘)
11/28/2024
𝑘𝑅 𝑖 8
 Origins of EXAFS equation

EXAFS: result of Photoelectron wavenumber:

interference between
orinal photolectron wave
and backscattered wave
Contribution from a single
backsattering atom:

Scattering
phase
𝑅𝑖
2 𝑅𝑖
−
𝜆 (𝑘)
2 𝐹𝑖 ( 𝑘 ) 𝑒
𝜒 𝑖 ( 𝑘 )=𝑆 0 2
sin ( 2𝑘 𝑅𝑖 + 𝜙 𝑖 (𝑘)
11/28/2024
𝑘𝑅 𝑖 9
 Origins of EXAFS equation

EXAFS: result of Photoelectron wavenumber:

interference between
orinal photolectron wave
and backscattered wave
Contribution from a single
backsattering atom:

(Will use
FEFF here!)
𝑅𝑖
2 𝑅𝑖
−
𝜆 (𝑘)
2 𝐹𝑖 ( 𝑘 ) 𝑒
𝜒 𝑖 ( 𝑘 )=𝑆 0 2
sin ( 2𝑘 𝑅𝑖 + 𝜙 𝑖 (𝑘)
11/28/2024
𝑘𝑅 𝑖 10
 Origins of EXAFS equation

EXAFS: result of Photoelectron wavenumber:

interference between
orinal photolectron wave
and backscattered wave
Contribution from a single
backsattering atom:

Amplitude
reduction
𝑅𝑖 factor
(~ 0.7 –
−
2 𝑅𝑖 1.0)
𝜆 (𝑘)
2 𝐹𝑖 ( 𝑘 ) 𝑒
𝜒 𝑖 ( 𝑘 )=𝑆 0 2
sin ( 2𝑘 𝑅𝑖 + 𝜙 𝑖 (𝑘)
11/28/2024
𝑘𝑅 𝑖 11
 Amplitude factors and their effects on EXAFS
T=200 K; Size is varied Bulk Pt; Temperature is varied

2.5
Bulk Pt 200 K
2.5
80 Å 300 K
2.0 45 Å 2.0 473 K
20 Å 673 K
1.5 1.5

1.0 1.0

0.5 0.5
-2

-2
k (k), Å

k (k), Å
0.0 0.0

-0.5
2

-0.5

2
-1.0 -1.0

-1.5 -1.5

-2.0 -2.0

-2.5 -2.5
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
-1
k, Å k, Å
-1

As a function of size, EXAFS amplitude is

scaled uniformly throughout the k-range
 2 2 k 2
 (k ) ~ N e
As a function of temperature, EXAFS
amplitude is scaled nonuniformly
 Fourier transform
Interatomic distances are related to frequencies of EXAFS: one can
use Fourier transform to separate different contributions and to select
those of interest!
Spectrum Frequency content

Fourier transform:
k max FT
 (k )k
n  2 ikr
FT ( r )   ( k )e dk
k min

Cu-Cu

Cu-N

Cu K-edge
in Cu3N

Cu N
11/28/2024 13
 Fourier Transforms in EXAFS:

~ 1
 ( R)   k  ( k ) ( k )
i 2 kR w
dk e
2 

Examples of Window Functions:

Effect on the FT on the sampled k-
interval

Ag foil at RT
 Objective of the analysis

3 Structure
𝜒 (𝑘)

11/28/2024 16
 Parameters in EXAFS Equation
Amplitude

2  2R
  (k ) 
NS eff
kR
0
f (k ) e
2
 2 2 2
k
e 
 3
sin 2kR  C3k   (k )
4
3

FEFF
Same parameterization for single and
E0 E0bkg  E0
multiple-scattering paths.
R Rmodel  R
 (k )    (k )


• Easy and transparent parameterization of

IFEFFIT structural models for EXAFS data fitting
(GUI: • FEFF theory
Artemis) • Error analysis
F F

Ag
Ag

Tb
Tb

Pb
Pb
 FEFF
FEFF6,8, 9:
Spherically symmetric potentials (muffin tin approximation)
Input: (xyz) coordinates and atomic numbers
Mean free path
Imaginary part of interstitial potential and life time broadening
Self Energy
Metals – Hedin-Lundqvist
Insulators – Hedin-Lundqvist or Dirac-Hara
Molecules – Dirac-Hara or ground state
Multiple scattering expansion
Each photoelectron path with its f(k), (k), (k)
is saved as a file

FEFF: John Rehr (U. Washington) and his group members

IFEFFIT: Matthew Newville (U. Chicago) and Bruce Ravel (NIST)
 NiO at RT

O1

O3
Ni2
Ni4
O5
 Fitting EXAFS Theory to the Data:
f ( Ri ) ~ ( Ri )  ~M ( Ri )
2 2
 fi 
N id p
N idp  f i  N
     
2
  
i 1   i  N i 1   i 
2kR
N idp  (Number of relevant independent data points)

(amount of information in the data is limited)
E. A. Stern
  N idp  P (Number of degrees of freedom) Phys. Rev. B 48, 9825-9827
(1993)

N id p 2
 fi  N idp
 Re( f ) 
N
1
     
2
i
2
 Im( f i ) 2
 i 1   i  N 2 i 1
To find uncertainty xi in the best fit result for xi: xi  H  1 ii

Physical meaning:
 ... ... ...
  2 2 
since H  ... ( x) ...
 xi x j 
 ... ... ...

The second derivative: rate of curvature of the curve.

The higher second derivative,

the higher rate of curvature,
the more this variable affects the fit.

When we invert the second derivative for such a variable,

the standard deviation becomes small and vice versa.
 EXAFS data analysis and modeling
I. Processing and visual examination of the data
1.4 Bkg removal 1.0
Fourier Transform
2.0
1.3 Pt/C nanoparticles

1.2 0.5
1.5

-3
FT Magnitude, Å
-2
x

1.1

k (k), Å
0.0
1.0
1.0

2
0.9 -0.5
0.5

0.8
-1.0
0.0
11500 12000 12500 13000 0 1 2 3 4 5 6 7 8 9 10
0 2 4 6 8 10 12 14 16 18 20

E, eV -1 r, Å
k, Å

II. Deciding on the model and

III. Fitting theoretical
refinement parameters
Structural modeling
EXAFS signal to the data
(different for homogeneous Path expansion
2.0
and heterogeneous systems) and FEFF
modeling 1.5
Data

-3
f(k), (k), (k)

FT Magnitude, Å
Fit
1.0

0.5

0.0
0 1 2 3 4 5 6 7 8 9 10

r, Å
 Data analysis by fitting methods
Find the best analysis software that can
implement your strategy.
IFEFFIT has convenient interface and
allows for a large variety of strategies
Implementation -Construct structural model
-Calculate theoretical EXAFS signal
-Perform structural refinement
-Carry out adequate error analysis

-Plan on doing reality checks;

-Reference compounds should be measured and
Analysis analyzed first;
-Try to maximize the number of degrees of
Strategy freedom in the fits
(use constraints, experimental/theoretical info etc.)

-Start with the crude picture first, then refine it;

- Homogeneous or heterogeneous
Conceptual environment?
Modeling (bulk or nano, eq. or ineq. unit cell positions,
solution or separate phases etc.)
 Radial distribution function
Even in perfect crystals atoms at given time-
monent are rarely at their equilibrium
positions: there is a lot of paths with slightly
different lengths.

It makes sense to group them in p=2

p=1
coordination shells and replace summation
over paths belonging to the same
coordination shell with an integral:

(for single-scattering path)

Radial
distribution 26
function
 Gaussian approximation
In commonly used Gaussian
approximation ( 𝑅− 𝑅 𝑝 )
2

−
1 2
2 𝜎𝑝
𝑔 𝑝 ( 𝑅 ) =𝑁 𝑝 e
√2 𝜋 𝜎 2
𝑝

2 𝜎❑
𝑝
and
Rp

11/28/2024 27
Disorder: Debye-Waller Factors
 Gaussian approximation
In commonly used Gaussian
approximation ( 𝑅− 𝑅 𝑝 )
2

−
1 2
2 𝜎𝑝
𝑔 𝑝 ( 𝑅 ) =𝑁 𝑝 e
√2 𝜋 𝜎 2
𝑝

Average 2 𝜎❑
𝑝
and
interatomic Rp
distance

11/28/2024 29
 Gaussian approximation
In commonly used Gaussian
approximation ( 𝑅− 𝑅 𝑝 )
2

−
1 2
2 𝜎𝑝
𝑔 𝑝 ( 𝑅 ) =𝑁 𝑝 e
√2 𝜋 𝜎 2
𝑝

2 𝜎❑
𝑝
and
Rp

Coordination
11/28/2024 number 30
 Gaussian approximation
In commonly used Gaussian
approximation ( 𝑅− 𝑅 𝑝 )
2

−
1 2
2 𝜎𝑝
𝑔 𝑝 ( 𝑅 ) =𝑁 𝑝 e
√2 𝜋 𝜎 2
𝑝

2 𝜎❑
𝑝
and
Rp

Disorder
(Debye-Waller or
11/28/2024 MSRD factor) 31
 Cumulant expansion*

g ( R)
 p ( k )  2
f ( k , R ) sin( 2kR   ( k )) dR
0
kR
 2R
g ( R) 
 2 F ( k , R )e  ( k ) sin( 2kR   ( k )) dR
0
kR

The effective distribution P(R,): P ( R ,  )  g ( R )e  2 R /  ( k ) / R 2

 
The cumulant expansion into a MacLaurin series: ln P ( R,  ) exp( 2ikR )dR  ( 2ik ) Cn / n!
n

0 n 0

1 2 4
 p k   S 02 F k exp(C0  2C2 k 2  C4 k 4  C6 k 6 )
k 3 45
 4 4 
sin 2kC1  C3 k 3  C5 k 5   k 
 3 15 
expC0   N exp 2C1 /  / C12 C1 is the mean value of the distribution (average distance)
C2 is the variance of the distribution (MSRD factor)
C3 is the asymmetry of the distribution (skewness)
C4 describes symmetric deviations from the Gaussian shape
11/28/2024 32
 EXAFS equation: summary

M the number of scattering paths

Np the coordination number for the p-th path
parameters
Structural

Rp the half-length of p-th path (the interatomic distance for SS paths)

2p the mean square radial displacement (a.k.a. Debye-Waller factor)
C3p 3rd cumulant of the distribution, which accounts for anharmonic effects,non-Gaussian disorder

S 02 the scale factor, which accounts for the multielectron effects ( 0.7-1.0)

photoelectron effective mean free path

Fp the backscattering amplitude
p the backscattering phase

And one more parameter:

E0 reference energy value ( )
11/28/2024 33
 EXAFS equation: summary

M the number of scattering paths

Np the coordination number for the p-th path
parameters
Structural

Rp the half-length of p-th path (the interatomic distance for SS paths)

2p the mean square radial displacement (a.k.a. Debye-Waller factor)
C3p 3rd cumulant of the distribution, which accounts for anharmonic effects,non-Gaussian disorder

S 02 the scale factor, which accounts for the multielectron effects ( 0.7-1.0)

photoelectron effective mean free path

Fp the backscattering amplitude
Fitting
p the backscattering phase parameters

And one more parameter:

E0 reference energy value ( )
11/28/2024 34
 EXAFS equation: summary

M the number of scattering paths

Np the coordination number for the p-th path
parameters
Structural

Rp the half-length of p-th path (the interatomic distance for SS paths)

2p the mean square radial displacement (a.k.a. Debye-Waller factor)
C3p 3rd cumulant of the distribution, which accounts for anharmonic effects,non-Gaussian disorder

S 02 the scale factor, which accounts for the multielectron effects ( 0.7-1.0)

photoelectron effective mean free path

Fp the backscattering amplitude
FEFF
p the backscattering phase
calculations
And one more parameter:
E0 reference energy value ( )
11/28/2024 35
 EXAFS analysis: summary
1 Data extraction, normalization, alignment, deglitching [Athena]

( 𝜇 ( 𝐸 ) − 𝜇0 ( 𝐸 ) ) ( 𝜇 ( 𝐸 ) − 𝜇0 ( 𝐸 ) )
𝜒 ( 𝐸)= ≈ 𝜒 ( 𝐸 ) → 𝜒 (𝑘)
𝜇0 ( 𝐸 ) 𝜇0 ( 𝐸 0 )

where is not known and needs to be somehow approximated

2 Fourier transform [Athena/Artemis]
k max

 n  2 ikr
FT ( r )  ( k ) k  ( k ) e dk
k min

3 Initial guess of structure model, calculations of scattering properties [Artemis/FEFF]

4 EXAFS data fitting with EXAFS equation [Artemis]

5 Interpretation of fitting results

11/28/2024 36
 Software for data analysis:

Demeter package (download online)

Programs:
Athena (for data processing and linear combination analysis)
Artemis (for fitting EXAFS equation to the data)

Install and test them

During Nov. 3 class:

Study my online talk on “XAFS for nanoparticles”:
https://youtu.be/w5j1bX2Bssk
 Multiple Scattering Paths

SS1
TR2
SS4 TS DS

SS2

SS3
TR1

SS5 TR3

In-plane atom

Above-plane atom

Absorbing atom
Debye Waller Factors of Collinear Multiple-Scattering Paths

Notation:
A.V.Poiarkova and J.J. Rehr,
Phys. Rev. B 59, 948 (1999).

⃗ =⃗
𝑟 𝑅+ 𝑢
⃗ 1−𝑢
⃗0 ⃗ 𝑖𝑖 +¿ = ⃗
𝑟 𝑅 𝑖𝑖+ ¿ + ⃗
𝑢𝑖 𝑢 ¿
+¿ − ⃗ ¿
¿
𝑖

𝜎 2≡ ⟨ ( 𝑟 − 𝑅) 2⟩
𝑛
1
𝑟 ≡ ∑ 𝑟 𝑖𝑖+¿ ¿
2 𝑖=1
𝑛
1
𝑅≡ ∑ 𝑅𝑖𝑖+¿ . ¿
2 𝑖=1
𝑟 𝑖𝑖 + ¿ 2
=¿ ¿ ¿
𝑟 𝑖𝑖 + ¿ ≈ 𝑅 ¿
√
^
𝑅 𝑖𝑖+ ¿
𝑖𝑖 +¿ 1 −2 ¿¿ ¿
𝑅𝑖𝑖 +¿ ¿ ¿

¿ 𝑅 𝑖𝑖+ ¿ + ¿ ¿
𝑛
1
𝑟 ≈ 𝑅+ ∑ ¿ ¿ a
R̂0
b c
2 𝑖 =1
a b c

a b c

 TS
2
 DS
2
 SS
2

𝜎 2 ≈ ⟨ 𝑢 20 𝑥 ⟩ + ⟨ 𝑢21 𝑥 ⟩ −2 ⟨ 𝑢0 𝑥 𝑢1 𝑥 ⟩