EXAFS in theory: an experimentalist’s guide to what it is and how it works

EXAFS in theory: an experimentalist’s guide to what it is and how it works

Presented at the SSRL School on Synchrotron X-ray Absorption Spectroscopy, May 20, 2008

Corwin H. BoothChemical Sciences Division

Glenn T. Seaborg CenterLawrence Berkeley National

Laboratory

Topics

1. Overview2. Theory

A. Simple “heuristic” derivationB. A real derivation (just for show)C. polarization: oriented vs spherically averaged

3. Experiment: Corrections and Other ProblemsA. Dead time (won’t cover)B. Self-absorptionC. Sample issues (size effect, thickness effect, glitches) (won’t cover)D. Energy resolution (won’t cover)

4. Data AnalysisA. Fitting procedures (Wednesday)B. Fourier conceptsC. Systematic errorsD. “Random” errorsE. F-tests

X-ray absorption spectroscopy (XAS) experimental setup

“white” x-rays from

synchrotron

double-crystal monochromator

collimating slits

ionization detectors

I0I1

I2

beam-stop

LHe cryostatsample

reference sample

• sample absorption is given by

t = loge(I0/I1)

• reference absorption is

REF t = loge(I1/I2)

• NOTE: because we are always taking relative-change ratios, detector gains don’t matter!

X-ray absorption spectroscopy

• Main features are single-electron excitations.

• Away from edges, energy dependence fits a power law: AE-3+BE-4 (Victoreen).

• Threshold energies E0~Z2, absorption coefficient ~Z4.

1 10 100

0.01

0.1

1

10

M

LIII

, LII, L

I

K

Xenon

(cm

-1)

E (keV)From McMaster Tables 1s

filled 3d

continuum

EF

core hole

unoccupiedstates

X-ray absorption fine-structure (XAFS) spectroscopy

occupied states

2p

filled 3d

continuum

EF

core hole

unoccupiedstates

16800 17000 17200 17400 17600 17800 18000 182001.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

2.8

log(

I 1/I0)

E (eV)

U L3 edge

UPdCu4

pre

-ed

ge

“edge region”: x-ray absorption near-edge structure (XANES) near-edge x-ray absorption fine-structure (NEXAFS)

EXAFS region: extended x-ray absorption fine-structure

~E0

E0: photoelectron threshold energy

e- : ℓ=±1

17000 17200 17400 17600 17800 180000.0

0.2

0.4

0.6

0.8

1.0

1.2

a

E(eV)

a=-

pre

0

17000 17200 17400 17600 17800 180001.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

2.8

t+

cons

t

E(eV)

data Victoreen-style background

17100 17200 17300

0.0

0.7

1.4

a

E(eV)

data 10*derivative

“pre-edge” subtraction

“post-edge” subtraction

determine E0

• Peak width depends on back-scattering amplitude F(k,r) , the Fourier transform (FT) range, and the distribution width of g(r), a.k.a. the Debye-Waller s.

• Do NOT read this strictly as a radial-distribution function! Must do detailed FITS!

1 2 3 4

-40

-20

0

20

40

16 U-Cu4 U-Pd12 U-Cu

r (Å)

FT o

f k3 (

k)

U LIII

edge

data fit

Amplitude envelope

[Re2+Im2]1/2

Real part of the complex

transformfunctionon distributi-pair radial a is

)2sin(),()()(

g

drkrrkFrgNki

cii

How to read an XAFS spectrum

3.06 Å

2.93 Å

“Heuristic” derivation

• In quantum mechanics, absorption is given by “Fermi’s Golden Rule”:

2ˆˆ~ irf

fff 0

.....

ˆˆˆˆˆˆ

ˆˆˆˆ~

0

2

0

2

0

2

tohcc

irffriirf

irffirf

00

0

Note, this is the same as saying

this is the change in the absorption per photoelectron

How is final state wave function modulated?

• Assume photoelectron reaches the continuum within dipole approximation:

2)ˆˆ( r



kr

er

ikr

2)ˆˆ(



central atom phase shift c(k)

kr

er

ikr

2)ˆˆ(

kr

er

kiikr c )(2)ˆˆ(



electronic mean-free path (k)


kr

er

ikr

2)ˆˆ(

kr

er

kiikr c )(2)ˆˆ(

)(/)(

2)ˆˆ( kRkiikR

ekR

er

c





complex backscattering probability f(,k)

kr

er

ikr

2)ˆˆ(

kr

er

kiikr c )(2)ˆˆ(

)(/)(

2)ˆˆ( kRkiikR

ekR

er

c

)(/)(

2 ),()ˆˆ( kRkiikR

ekkfkR

er

c





complex backscattering probability f(,k)

complex=magnitude and phase: backscattering atom phase shift a(k)

kr

er

ikr

2)ˆˆ(

kr

er

kiikr c )(2)ˆˆ(

)(/)(

2)ˆˆ( kRkiikR

ekR

er

c

)(/)(

2 ),()ˆˆ( kRkiikR

ekkfkR

er

c

kR

eekfk

kR

er

kikirRikkR

kiikR acc )()()()(/

)(2 ),()ˆˆ(



kr

er

ikr

2)ˆˆ(

kr

er

kiikr c )(2)ˆˆ(

)(/)(

2)ˆˆ( kRkiikR

ekR

er

c

)(/)(

2 ),()ˆˆ( kRkiikR

ekkfkR

er

c

kR

eekfk

kR

er

kikirRikkR

kiikR acc )()()()(/

)(2 ),()ˆˆ(

)(/22

)()(222 ),()ˆˆIm( kR

kikikRi

ekfkR

er

ac



complex backscattering probability kf(,k)

complex=magnitude and phase: backscattering atom phase shift a(k)

final interference modulation per point atom!

Assumed both harmonic potential AND k<<1: problem at high k and

Assumed plane wave scattering, curved wave has r-dependence

solution: substitute Feff full curved wave theory

Other factors

• Allow for multiple atoms Ni in a shell i and a distribution function function of bondlengths within the shell g(r)

rdkr

kkkrrgekfrNSk

i

ackri

2)(/222

0

)()(22sin)(),()ˆˆ()(

2

2

2

)(

2

1)(

iRr

erg

i

ackkri kr

kkkreekfrNSk

22)(/222

0

)()(22sin),()ˆˆ()(

22

where and S02 is an inelastic loss factor

EXAFS equation derivation

• This “simple” version is from the Ph.D. thesis of Guoguong Li, UC Santa Cruz 1994, adapted from Teo, adapted from Lee 1974. See also, Ashley and Doniach 1975.

222

)cos1( |ˆˆ),(),(

ˆˆ),(ˆˆ|)ˆ( re

r

kfkfre

r

kfkDkP ikr

cikr

)1

(]ˆˆ

)Re[(2)(ˆ)ˆ(4

1)(

33210 rODIe

r

rIIDkkdkPk ikr

The probability of emitting an electron in the direction is given by:k

The absorption is therefore:

kdkfr

I

kdr

kfkfekI

kdkfekI

kdkD

k

cikr

ikr

ˆ|),(|4

)ˆˆ(

ˆ),(),()ˆˆ(

4

1

ˆ),()ˆˆ(4

1

ˆ)ˆˆ(4

)(

22

3

2

cos1

20

where:

-

derivation continued…

• Some, er, “simplifications”:

)(cos|),(|)],0(Im[

12

2

12

1

12

2

12

1)()(

)1)(12(2

1)(

)(cos)(),(

)(cos)()1)(12(

),(),(3

4),(ˆ),(ˆ

2

,1

1

1 ,1

2

cos

,,

ddkfkf

ll

l

lldxxPxxP

elki

kf

Pkfkf

Pkrjle

YYk

llllll

il

lll

lll

lik

ml

lm

lmml

l

• Rewrite I1, I2 and I3

]),0(),([2

ˆˆ

])()()1[(2

ˆˆ

)]())(1([2

)(

12

)(2

2

ˆˆ

1for )sin(1

)( step,key for the now and

])()()())(1[(12

)(2

2

ˆˆ

sin),(cos4

ˆˆ

2

1

2

1

2

1

2

11

2

11

11

0

2

0

cos1

ikrl

ikr

l

ikrl

ikrl

l

il

ikril

ikril

ikril

ikr

l

ll

ll

ll

ll

ll

ikr

ekfekfkr

ri

ekfekfkr

ri

eeleelkri

i

l

kfr

krkrkr

krj

krjilkrjill

kfr

dkfedr

I


ikri

ll

lcl

ikr

cikr

ekfer

ri

dxxPxilkir

kfer

dkfer

kfI

cl ),()1(

2

ˆˆ

)()1()12)(12(2

1

2

),(ˆˆ

sin),(coscos2

),(

2

1

1

02

),0(Im)ˆˆ(

2

2

3 kfkr

rI


Finishing derivation, beginning polarization

• Notice (angle w.r.t. polarization): can eliminate certain peaks!

)],(Im[)ˆˆ(

3)( 22

2

2

kfekr

rDDk

cliikr

edge-K afor )](22sin[),(cos3

)],(Im[)ˆˆ(3

)(

2

2

222

2

0

0

kiikrkfkr

kfekr

r

k

cl

iikr cl

L2 and L3 edges appear more complicated

• 2p1/2 or 2p3/2 core hole and a mixed s and d final state

)]}(2sin[2

10

)cos31)]((2sin[2

12

)cos31)]((2sin[1021{|),(|

1021

12

1)(

0

2

22

2

202222

kkr

kkr

kkrkr

kfk

Heald and Stern 1977

polarization vs. spherically averaged

• L2 and L3 mostly d final states (yeah!)

edges L and Lfor )cos31(2

1-

edges L andK for cos3)(cos

)](2sin[),(

)(cos)(

5010

12)cos31(5

10

12

IIIII2

12

2

2

2

2

P

kkrkr

kfPk

Stern 1974Heald and Stern 1977

Corrections and Concerns

• “Normal” EXAFS performed on powder samples in transmission

— can tune the thickness• Want t~<1 and t < 3

• We like stacking strips of scotch tape

— can make a flat sample

— diffraction off the sample not a problem

• Working with oriented materials: single crystals, films

— usually cannot get the perfect thickness: too thick• fluorescence mode data collection

• self-absorption can be substantial!

• dead-time of the detector

Fluorescence mode

dxy z

Sample

point of absorption and fluorescence

incident photon trajectoryflu

ores

cing p

hoton

traje

ctory

EKE

1s

filled 3d

continuum

EF

core hole

unoccupiedstates

0I

It F

2p

8800 9000 9200 9400 9600 98000.0

0.2

0.4

0.6

0.8

1.0

1.2

Cu foil, Cu K edgeN

orm

aliz

ed a

bsor

ptio

n

E(eV)

transmission fluorescence

t=1.4

10

exp I

IF

• Fluorescing photon can be absorbed on the way out

• Competing effects:

— glancing angle, sample acts very thick, always get a photon, XAFS damped

— normal-incidence: escaping photon depth fixedd

xy z

Sample

point of absorption and fluorescence

incident photon trajectoryflu

ores

cing p

hoton

traje

ctory

EKE

L. Tröger , D. Arvanitis, K. Baberschke, H. Michaelis, U. Grimm, and E. Zschech, Phys.

Rev. B 46, 3283 (1992).

The full correction

• With the above approximation, we can finally write the full correction:

}4]))1([

]))1(([{2

1

expexp

exp

2

a

a

sin

sin

d

a ed

where

sin1d

e

• In the thick limit (d), this treatment gives:

aa

exp

exp

1

FT

sin

sin

Booth and Bridges, Physica Scripta T115, 202 (2005)

L. Tröger , D. Arvanitis, K. Baberschke, H. Michaelis, U. Grimm, and E. Zschech, Phys.

Rev. B 46, 3283 (1992).

Correction applied to a 4.6 m Cu foil

• Data collected on BL 11-2 at SSRL in transmission and fluorescence using a 32-element Canberra germanium detector, corrected for dead time.

Fitting the data to extract structural information

• Fit is to the standard EXAFS equation using either a theoretical calculation or an experimental measurement of Feff

• Typically, polarization is spherically averaged, doesn’t have to be

• Typical fit parameters include: Ri, Ni, i, E0

• Many codes are available for performing this fits:

— EXAFSPAK

— IFEFFIT• SIXPACK

• ATHENA

— GNXAS

— RSXAP

rdkr

kkkrrgekfrNSk

i

ici

kriii

2)(/222

0

)()(22sin)(),()ˆˆ()(

FEFF: a curved-wave, multiple scattering EXAFS and XANES calculator

• The FEFF Project is lead by John Rehr and is very widely used and trusted

• Calculates the complex scattering function Feff(k) and the mean-free path

TITLE CaMnO3 from Poeppelmeier 1982 HOLE 1 1.0 Mn K edge ( 6.540 keV), s0^2=1.0 POTENTIALS * ipot z label 0 22 Mn 1 8 O 2 20 Ca 3 22 Mn ATOMS 0.00000 0.00000 0.00000 0 Mn 0.00000 0.00000 -1.85615 0.00000 1 O(1) 1.85615 0.00000 1.85615 0.00000 1 O(1) 1.85615 -1.31250 0.00000 1.31250 1 O(2) 1.85616 1.31250 0.00000 -1.31250 1 O(2) 1.85616 1.31250 0.00000 1.31250 1 O(2) 1.85616 -1.31250 0.00000 -1.31250 1 O(2) 1.85616 0.00000 1.85615 -2.62500 2 Ca 3.21495 -2.62500 1.85615 0.00000 2 Ca 3.21495 -2.62500 -1.85615 0.00000 2 Ca 3.21495 0.00000 1.85615 2.62500 2 Ca 3.21495

0 5 10 15 20-5

0

5

10

Phas

e sh

ift (

rad)

k(Å-1)

2c

a

Co-O, Co K edge

Phase shifts: functions of k

• sin(2kr+tot(k))): linear part of (k) will look like a shift in r slope is about -2x0.35 rad Å, so peak in r will be shifted by about 0.35 Å

• Both central atom and backscattering atom phase shifts are important

• Can cause CONFUSION: sometimes possible to fit the wrong atomic species at the wrong distance!

• Luckily, different species have reasonably unique phase and scattering functions (next slide)

0 1 2 3 40

10

20

30

40

50

60

Mag

nitu

de o

f F

T o

f k3

(k)

r (Å)

"CoCaO3" R=1.85 Å

R=3.71 Å

Species identification: phase and magnitude signatures

0 5 10 15 20

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Mag

nitu

de o

f F

T o

f k3

(k)

r (Å)

Co-Mn Co-Co Co-Ba

0 1 2 3 4

-60

-50

-40

-30

-20

-10

0

10

20

30

40

50

60

FT

of

k3 (k

)

r (Å)0 5 10 15 20

-17

-16

-15

-14

-13

-12

-11

-10

-9

-8

-7

-6

-5

-4

-3

a

k (Å-1)

Co-O Co-C

• First example: same structure, first neighbor different, distance between Re and Ampmax shifts

• Note Ca (peak at 2.8 Å) and C have nearly the same profile

• Magnitude signatures then take over• Rule of thumb is you can tell difference in species

within Z~2, but maintain constant vigilance!

More phase stuff: r and E0 are correlated

• When fitting,E0 generally is allowed to float (vary)• In theory, a single E0 is needed for a monovalent absorbing

species• Errors in E0 act like a phase shift and correlate to errors in

R!consider error in E0: ktrue=0.512[E-(E0+)]1/2

for small , k=k0-[(0.512)2/(2k0)] eg. at k=10Å-1 and =1 eV, r~0.013 Å

• This correlation is not a problem if kmax is reasonably large• Correlation between N, S0

2 and is a much bigger problem!

1 2 3 4

-40

-20

0

20

40

16 U-Cu4 U-Pd12 U-Cu

r (Å)

FT o

f k3 (

k)

U LIII

edge

data fit

Information content in EXAFS

• k-space vs. r-space fitting are equivalent if done correctly!

• r-range in k-space fits is determined by scattering shell with highest R

• k-space direct comparisons with raw data (i.e. residual calculations) are incorrect: must Fourier filter data over r-range

• All knowledge from spectral theory applies! Especially, discrete sampling Fourier theory…

Fourier concepts

• highest “frequency” rmax=(2k)-1 (Nyquist frequency)

eg. for sampling interval k=0.05 Å-1, rmax=31 Å

• for Ndata, discrete Fourier transform has Ndata, too! Therefore…

FT resolution is R=rmax/Ndata=/(2kmax), eg. kmax=15 Å-1, R=0.1 Å

• This is the ultimate limit, corresponds to when a beat is observed in two sine wave R apart. IF YOU DON’T SEE A BEAT, DON’T RELY ON THIS EQUATION!!

0 5 10 15 20-0.2

-0.1

0.0

0.1

0.2

exp(

-2k2

2 )sin

(2kr

)

k

r=2.0, =0.1 r=2.1, =0.1 average

0 5 10 15 20

-1.0

-0.5

0.0

0.5

1.0

sin(

2kr)

k

sin(2k2) sin(2k2.1) average

More Fourier concepts

• Assuming Ndata are independent data points, and a fit range over k (and r!):

result EXAFS rule" sStern'" 22

resultFourier 2

2

ind

maxmaxdataind

maxmaxdatamax

1212

kr

krkr

rkk

kr

N

krNN

rkNr

kkrr

• Fit degrees of freedom =Nind-Nfit

• Generally should never have Nfit>=Nind (<1)

• But what does this mean? It means that

For every fit parameter exceeding Nind, there is another linear combination of the same Nfit parameters that produces EXACTLY the same fit function

Systematic errors: calculations are not perfect!

Kvitky, Bridges and van Dorssen, Phys. Rev. B 64,

214108 (2001).

• Systematic errors for nearest-neighbor shells are about 0.005 Å in R, 5% in N, 10% in (Li, Bridges, Booth 1995)

• Systematic error sources:

— sample problems (pin holes, glitches, etc.)

— correction errors: self-absorption, dead time, etc.

— backscattering amplitudes

— overfitting (too many peaks, strong correlations between parameters)

• Random error sources:

— some sample problems (roughly, small sample and moving beam)

— low counts (dilute samples)

Systematic vs. Random error

Systematic vs. Random error

0 1 2 3 4

-6

0

6

data fit

FT o

f k3

(k)

r (Å)

Cp''2Ti(bipy)

Statistics:

freedom) of (degrees

)(

12

22

N

i i

ii dd

constant

2syst

2true

2meas

Error analysis options

• Use error analysis in fitting code (generally from the covariance matrix)

— Always requires assumptions• a single error at all r or k is assumed

• systematic errors ignored

• can be useful in conjunction with other methods

• Collect several scans, make individual fits to each scan, calculate standard deviation in parameters pi

— Fewer assumptions• random errors treated correctly as long as no nearby minima in 2(pi) exist

• systematic errors lumped into an unaccounted shift in <pi>

• Best method(?): Monte Carlo

• EXAFS as a technique is not count-rate limited: It is limited by the accuracy of the backscattering functions

• This does NOT mean that you should ignore the quality of the fit!

• DO a Chi2 test, observe whether Chi2=degrees of freedom

— one limit: random noise is large, and you have a statistically sound fit

— other limit: random noise is small, and you will then know how large the problem with the fit is

not so Advanced Topic: F-test

• F-test, commonly used in crystallography to test one fitting model versus another

F=(12/1)/(0

2/0)0/1R12/R0

2

(if errors approximately cancel)

alternatively: F=[(R12-R0

2)/(1-0)]/(R02/0)

• Like 2 , F-function is tabulated, is given by incomplete beta function

• Advantages of a 2-type test:

—don’t need to know the errors!

F-test how-to

function beta incompletean is ][ and level confidence theis where

]2

,2

[1)(

parametersfit points, datat independen ,with

)(1

)/(

/)(

residual a of definitionproper )(

2,,

0

1

220

20

21

1

22

y,zI

bmnIFFP

mnR

R

b

mn

mnR

bRRF

ddR

x

mnb

N

iii

Example and limitations

• consider 4 different samples with various amount of species TcSx: Are they interconnected?

— r=4.5-1 Å k=13.3-2 Å-1, n=26.8

— model 0 has Tc neighbors and m=14 parameters, R0=0.078 to 0.096

—model 1 has only S neighbors and m=10, R0=0.088 to 0.11

—dimension of the hypothesis b=14-10=4

—each data set, between 35 and 82%, all together 99.9%

• Effect of systematic error: increases R0 and R1 same amount

• This will decrease the % improvement, making it harder to pass the F-test (right direction!)

• Failure mode: fitting a peak due to systematic errors in Feff

Finishing up

• Never report two bond lengths that break the rule

• Break Stern’s rule only with extreme caution

• Pay attention to the statistics

Further reading

• Overviews:—B. K. Teo, “EXAFS: Basic Principles and Data Analysis” (Springer,

New York, 1986).—Hayes and Boyce, Solid State Physics 37, 173 (192).

• Historically important: —Sayers, Stern, Lytle, Phys. Rev. Lett. 71, 1204 (1971).

• History—Lytle, J. Synch. Rad. 6, 123 (1999).

(http://www.exafsco.com/techpapers/index.html)—Stumm von Bordwehr, Ann. Phys. Fr. 14, 377 (1989).

• Theory papers of note:—Lee, Phys. Rev. B 13, 5261 (1976).—Rehr and Albers, Rev. Mod. Phys. 72, 621 (2000).

• Useful links—xafs.org (especially see Tutorials section)—http://www.i-x-s.org/ (International XAS society)—http://www.csrri.iit.edu/periodic-table.html (absorption calculator)

Further reading

• Thickness effect: Stern and Kim, Phys. Rev. B 23, 3781 (1981).• Particle size effect: Lu and Stern, Nucl. Inst. Meth. 212, 475 (1983).• Glitches:

—Bridges, Wang, Boyce, Nucl. Instr. Meth. A 307, 316 (1991); Bridges, Li, Wang, Nucl. Instr. Meth. A 320, 548 (1992);Li, Bridges, Wang, Nucl. Instr. Meth. A 340, 420 (1994).

• Number of independent data points: Stern, Phys. Rev. B 48, 9825 (1993).• Theory vs. experiment:

—Li, Bridges and Booth, Phys. Rev. B 52, 6332 (1995).—Kvitky, Bridges, van Dorssen, Phys. Rev. B 64, 214108 (2001).

• Polarized EXAFS:—Heald and Stern, Phys. Rev. B 16, 5549 (1977).—Booth and Bridges, Physica Scripta T115, 202 (2005). (Self-absorption)

• Hamilton (F-)test:—Hamilton, Acta Cryst. 18, 502 (1965).—Downward, Booth, Lukens and Bridges, AIP Conf. Proc. 882, 129

(2007). http://lise.lbl.gov/chbooth/papers/Hamilton_XAFS13.pdf

Documents

EXAFS in theory: an experimentalist’s guide to what it is and how it works