Basics of Data Processing · an Angstrom. September 30, 2005 2005 NSLS EXAFS Course: Theory,...
Transcript of Basics of Data Processing · an Angstrom. September 30, 2005 2005 NSLS EXAFS Course: Theory,...
1
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
1
Basics of Data Processing
Scott CalvinSarah Lawrence College
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
2
Importing Raw Data in Athena
2
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
3
What This Talk Is
• An overview of some of the generalfeatures of Artemis and Athena
• A highlight of a few areas where novicesoften go astray
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
4
What This Talk Is Not
• An exhaustive documentation of everyfeature in Artemis and Athena
• A detailed tutorial• A discussion of EXAFS theory
3
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
5
“Groups” includeindividual scans
Groups in Athena
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
6
This is the “current” group
Checks indicate“marked” groups
Note that these buttonsplot the current group
These buttons plot allthe marked groups
“Marked” and “Current” Groups
4
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
7
These options affectplots of the current group
These options affect plotsof all marked groups
“Marked” and “Current” Groups
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
8
Vertical translation betweenscans will not affect EXAFS
But a closer look…
Compare Scans
5
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
9
…reveals other problems.
Scans 38 and 39 appearinconsistent with the rest
Energy calibration appearsto have drifted
Compare Scans
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
10
Under the “align”menu, you can haveAthena attempt toshift scans for you,or you can do it “byhand”
Alignment in Athena
6
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
11
The scans are now aligned
If we had not alignedthem, the average wouldshow considerabledifferences:
Note missing shoulderon unaligned version
Alignment Matters
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
12
alignment is very important
7
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
13
This informationapplies to thecurrent scan
Right clicking bringsup a menu that lets
you apply thesesettings to all scans,
or to all marked scans
Background Removal in Athena
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
14
Choose Eo at firstpeak in the
derivative spectrum
Click here to see aplot of the
derivative spectrum
Choosing Eo in Athena
8
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
15
Choosing Eo carefully is of lowimportance for EXAFS, as long as it
is done consistently for all scans
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
16
Choose pre-edge andnormalization ranges
Normalizing in Athena
9
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
17
These pre- and post-edge lines look OK
This post-edge line goestoo high near edge
Good and Bad Normalization
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
18
Merge aligned andnormalized scans here
Merging Data in Athena
10
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
19
Normalization may be very important insome cases, such as determining
coordination number when you think youknow So
2, but if you are fitting So2 it is
relatively unimportant
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
20
Rbkg controls how far up in theFourier transform the background
will try to suppress; i.e. how“wiggly” the background can be
Rbkg
11
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
21
Rbkg of 2.0 _ wiggles toomuch in energy space andthus kills off low-R peak
in Fourier transform
Energy plot k-space plot Magnitude ofFourier transform
Rbkg
Rbkg of 0.2 Å “misses” inenergy space and thus
introduces spurious low-Rpeak in Fourier transform
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
22
Rbkg is fairly important
Ideally, small changes in Rbkg should notsignificantly change the parameters you
find via your fits
12
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
23
Fourier Transforms in Athena
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
24
This data appears consistent for allscans. If there is trouble at high-k,it is systematic and reproducible.
Choosing kmax
13
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
25
Choice of kmax should be of low importanceIdeally, small changes in choice of kmaxshould not change parameters you find
via your fits
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
26
Choose kmin where backgroundstops being strongly dependent onsmall changes in backgroundparameters
Choosing kmin
14
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
27
Choice of kmin should be of low importanceIdeally, small changes in choice of kminshould not change parameters you findvia your fits. In practice, fit is often
more sensitive to kmin than kmax
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
28
Unweighted
k-weight 1
k-weight 2
k-weight 3
k-weight
15
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
29
This affects theFourier transform
This affects thek-space plot
Athena Reminder!
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
30
Effect of k-weight on Fourier Transform
16
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
31
k-weight should be of low importanceIdeally, different k-weights should notchange parameters you find via your fits.
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
32
Entering a Crystal ModelGet here fromthe Theory menuLattice parameters
are given inAngstroms
Artemis givescoordinates of atomsin fractionalcoordinates; i.e.multiply by latticeparameters to getAngstroms
This is the maximumdistance from theabsorbing atom (inAngstroms) that youwant feff to calculate.Make it a few Angstromslarger than the largestpath you plan to use, sothat feff can calculatepotentials properly
17
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
33
What if the Sample Isn’t Crystalline?
• Short-Range Amorphous (e.g. glasses): Use acrystal structure which is locally similar on theAtoms page, but only use the inner paths foranalysis
• Macromolecular: Use the space group P 1 andlattice parameters A=B=C=100. Then entercoordinates of atoms in units of hundredths ofan Angstrom.
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
34
This list of potentials can bechanged, but the absorbermust be 0 and you cannot skipnumbers
This file is generatedautomatically when you runAtoms, but you should look at itto see if it makes sense
Atoms Generates a feff.inp File
These coordinates anddistances are given inAngstroms; you shouldalways check that first-shell distances andcoordination numbers makesense to you
18
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
35
The left part ofthe screenchangesdepending onwhat ishighlighted in…
The Artemis Main Screen
…the “Data &Paths” panel.
The right side ofthe screenmostly controlsthe plots……and the biggreen Fit buttonmakes it go!
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
36
Interpretation shows you a listof paths:
Feff Screen
Black non-italicised paths areincluded in the fit
Paths with the text in brown arenot currently included in the fit
Italicised black paths are notcurrently in the path list at all
Paths with brown backgrounds aredirect scattering
Paths with blue backgrounds arefocused
Right-clicking allows you to addpaths to the list, include them inthe fit, and more…
19
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
37
Clicking on a data set brings upa screen where you can choosek- and R-ranges for fit, k-weights, etc..
Data Screen
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
38
This screen allows you todefine various kinds ofparameters:
Guess, Def, Set Screen
Guess parameters are optimizedduring fittingDef parameters are calculatedduring fittingSet parameters are calculatedbefore fittingAfter parameters are calculatedafter fittingSkip parameters are not used
Restraints force a parameter tobe close to a specified value“reff” is a special parameter equalto half of the nominal path length
20
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
39
Clicking on a path lets youspecify path parameters interms of Guess, Def, Setparameters
Guess, Def, Set Screen
September 30, 2005 2005 NSLS EXAFS Course:Theory, Experiment, and Advanced Applications
40
Clicking on Fit in the Data &Paths screen brings up ascreen that lets you viewresults from previous fits andcompare them graphically
Log Files Screen