The X-ray Spectroscopy Laboratory of Hunter College conducts
research on the structure of various types of materials in solid, liquid, and
gaseous phases. This laboratory studies a phenomenon known as EXAFS (extended
x-ray absorption fine structure) which provides insight into the local
structure of substances (the nature of EXAFS is explained in detail in a later
section). Experiments are conducted at the National Synchrotron Light Source
(NSLS) - a department of Brookhaven National Laboratories, located in Upton,
Long Island. The Light Source contains a linear particle accelerator from which
electrons are sent into two rings - a large X-ray ring and a smaller VUV
(Vacuum Ultra-Violet) ring. Many beams of varying energies come out tangent to
the rings into experimental stations called beamlines, where experiments are
carried out. This lab works mainly on
two beamlines, X23A2 and X23B, and the materials that can be studied are
limited by the energy ranges of these two beamlines.
All design and
preparaton of experiments, as well as all of analysis, is done at the Hunter
College lab, room 1231HN. Here we
prepare samples, analyze data and discuss results. All chemicals are located in the cabinet on the wall and are
stored alphabetically. The cabinet also
contains things like labels and a scale; spatulas, sample plates, mortars and
pestles are in drawers beneath the counter.
Analysis of data is
performed using several different computer programs such as ATOMS, AUTOBK,
FEFF, FEFFIT and ORIGIN. ATOMS and FEFF
are used to create a mathematical model of EXAFS, which will then be fitted to
the data collected at Brookhaven.
AUTOBK and FEFFIT are used to process the data itself. AUTOBK is used to subtract the background
(like noise) from the data and FEFFIT is used to fourier-transform the data and
run new fits. Origin is used to graph
the data and the fits. (Details on how
to use these programs are described in later sections). Once the transformed data has been plotted
in ORIGIN, fitting parameters are varied to get the best least-squares
fit. After each fit, FEFFIT creates a file containing the results of the
fit - how close it is to the data and
what values were assigned to the fitting parameters. The bulk of the research time is spent on getting the best
possible fit. The best fit is obtained
when the fit plot and the data plot (both on the same graph) give the most
optimal overlap, and the values for parameters are reasonable. Once it is obtained, the results are
interpreted and put into the context of the system studied. Some of the results that can be obtained are
distances between neighboring atoms, bond lengths, and the degree of electron
delocalization within specific bonds.
---------------------------------
Do You Yahoo!?
Yahoo! Auctions - Buy the things you want at great prices.
--------------------
Hi Scott, it's Vitaly. Here's the overview for the EXAFS
handbook. You can send comments to my hunter e-mail address
<EM>vkishiny@hejira.hunter.cuny.edu</EM>.
The X-ray Spectroscopy Laboratory of Hunter
College conducts research on the structure of various types of materials in
solid, liquid, and gaseous phases. This laboratory studies a phenomenon known
as EXAFS (extended x-ray absorption fine structure) which provides insight into
the local structure of substances (the nature of EXAFS is explained in detail
in a later section). Experiments are conducted at the National Synchrotron
Light Source (NSLS) - a department of Brookhaven National Laboratories, located
in Upton, Long Island. The Light Source contains a linear particle accelerator
from which electrons are sent into two rings - a large X-ray ring and a smaller
VUV (Vacuum Ultra-Violet) ring. Many beams of varying energies come out tangent
to the rings into experimental stations called beamlines, where experiments are
carried out. This lab works mainly on two beamlines, X23A2 and X23B,
and the materials that can be studied are limited by the energy ranges of these
two beamlines.
All design and preparaton of experiments, as well
as all of analysis, is done at the Hunter College lab, room 1231HN.
Here we prepare samples, analyze data and discuss results. All
chemicals are located in the cabinet on the wall and are stored alphabetically.
The cabinet also contains things like labels and a scale; spatulas,
sample plates, mortars and pestles are in drawers beneath the counter.
Analysis of data is performed using several
different computer programs such as ATOMS, AUTOBK, FEFF, FEFFIT
and ORIGIN. ATOMS and FEFF are used to create a mathematical
model of EXAFS, which will then be fitted to the data collected at
Brookhaven. AUTOBK and FEFFIT are used to process the data
itself. AUTOBK is used to subtract the background (like noise) from
the data and FEFFIT is used to fourier-transform the data and
run new fits. Origin is used to graph the data and the
fits. (Details on how to use these programs are described in later
sections). Once the transformed data has been plotted in ORIGIN,
fitting parameters are varied to get the best least-squares fit.
After each fit, FEFFIT creates a file containing the results of the
fit - how close it is to the data and what values were assigned to
the fitting parameters. The bulk of the research time is spent on
getting the best possible fit. The best fit is obtained
when the fit plot and the data plot (both on the same graph) give the most
optimal overlap, and the values for parameters are reasonable. Once
it is obtained, the results are interpreted and put into the context of the
system studied. Some of the results that can be
obtained are distances between neighboring atoms,
bond lengths, and the degree of electron delocalization within specific bonds.
As Early as Possible:
Any non-citizens who might participate in the run and have never been on one
before should go to nslsweb.nsls.bnl.gov/nsls/users/prcoedures/foreign.html and
fill out the IA-473 form online immediately. The NSLS says some of these forms
have to be in 90 days before the run. In practice, we usually fill them out
later, but that could in theory cause a problem.
30 days before the
run: Find out what samples will be run. Check with both Marten and Steve. At
this point there will probably not be specifics (“I think we’ll have a couple
of vanadates from Russia”) and the list will be added to as the run approaches,
but it helps with planning. Begin considering whether the list contains any of
the following:
Safety hazards:
Solids are generally OK, even if they’re toxic or carcinogenic, unless they
produce a lot of dangerous dust or something like that. People at the NSLS know
enough not to eat random black powders that are sitting around (we hope!). But
volatile materials (solutions and particularly gasses) probably fall in this
category even if you know they’re routinely used in freshman chemistry classes!
If there are any
safety hazards, contact Andrew Ackerman 631-344-5431 as soon as possible to
begin a dialogue as to how to handle them. The basic guideline is to use as
little as possible for the experiment.
Sample chambers: If
the run includes any materials that need a sample cell or instrumentation
designed (e.g. solutions, gasses, crystals on which resistivity measurements
will be performed, etc.), begin immediately, in consultation with Marten and
Rich. Failing to do this early has been the most common reason for frustrating
and unsuccessful runs.
Materials which must
be synthesized: Begin immediately. You never know what might go wrong.
14 days before the
run: Order any samples which must be obtained from commercial sources, such as
Alfa Aesar (www.alfa.com).
If there are any first-time
minors on the run, Andrew Ackerman (631-344-5431 ) should be contacted for the
necessary paperwork and to set up a safety consultation.
10 days before the
run: Try to get a schedule together for who will be at the beamline when. This is
something like a GRE analytical problem, because the following considerations
apply:
At least one of the
most experienced team members should be there on the first day and should plan
to remain through the next morning. “Most experienced team members” means
Marten, postdocs, or doctoral students who have been coming to the NSLS for at
least two years. In certain cases we may be collaborating with a team from
Germany or elsewhere, and, if necessary, one of them can serve. It is
preferable, however, that someone from our team be there as well in case of
safety or logistical issues.
An experienced person
should also be there at the end of the run.
At least two people
should be there at all times.
If the run is going
to have a lot of samples, more people are needed so that we can work in shifts.
If the run uses a
Displex, at least three people should be there during the setup.
People without cars
should not be coming and leaving at too many different times, since then a lot
of time is wasted taking people to and from trains: try to group people into
two’s and three’s for travel.
Be aware of when
drivers are needed.
First-time foreign
students and first-time minors MUST arrive before 3:00 on a weekday. Watch for
lab holidays, such as Veteran’s Day...the offices are closed then.
9 days before the
run: File a safety form online, if necessary. You do not need to file a new
form if there is one filed during the past year which is sufficiently similar,
although you may need to add users to it. The safety form is at:
130.199.76.52/safety. For the safety form you need to know everyone’s birthday.
8 days before the
run: Make housing reservations, particularly for males in the summer (other
times or for females the reservations can be made closer to the date). You can
do this by emailing housing@bnl.gov. Ask for dorm reservaions. Tell them which
beamline we are visiting, the name of the beamline representative (Joe Woicik
for X23A2 or Johnny Kirkland for X23B), that we’re from Hunter, and that we’ll
pay on arrival. Give them your contact information (email, phone number, etc.)
Then give them each person’s arrival and departure dates, full name, gender,
and whether they are a smoker. (Note that arrival date is when they check in,
and departure date is when they check out, so even a one night stay reads as
something like arrives June 18 departs June 19).
For each person, file
an online “gate reservation form” at
nslsweb.nsls.bnl.gov/nsls/dbforms/user-regis.asp. You will need their
citizenship and country of birth.
7 days before the
run: Begin sample preparation. Since Steve gets sent a lot of his samples, some
may arrive the day before the run, or even during it. Be prepared for this
possibility.
6 days before the
run: email Syed Khalid (Khalid@bnl.gov) to reserve Displex, Lytle detector, or
hot and cold sample holder from EXAFS pool. This is essentially a courtesy;
they are rarely checked out.
5 days before the
run: Ship any hazardous samples (such as gasses or nasty solutions) to
Brookhaven. Yes, that includes routine freshman chemicals, such as acids.
3 days before the
run: In consultation with Marten, make up a tentative sequence of samples to
run. Factors to consider:
All samples which use
the same equipment should be run in a row. Example: all Displex samples should
be run sequentially.
Samples should be run
when the people most involved with analyzing them are present, if possible.
Samples of similar
energy should be run sequentially. Large changes in energy require adjustment
of feedback, switching gasses, or, in the case of X23B, switching crystals.
Day of run: make sure
first person to go out has safety form number.
Check-in:
If
it is before 3 on a weekday, all users must go to the User Administration
Office on the second floor of the NSLS to get film badges. First-timers must
also take a safety orientation.
If
it is after hours or a weekend, get film badges at the control room on the
epxerimental floor of the NSLS. It’s possible first-timers who are US Citizens
may be able to get their safety orientation there as well, but don’t count on
it.
Any
first-timer who has not received their safety orientation and film badge,
citizen or not, may come to the beamline as an escorted visitor. This means
they may not participate in the experiment and must remain with their escort at
all times. (This is more a film badge issue than a security issue.) Escorted
visitors must be signed in at the entrance, and may not remain more than eight
hours. Minors under 18 may not come on the floor under this policy; if a minor
arrives after hours, he or she should stay in the dorm until receiving a film
badge.
Once
at the beamline, ALL users must receive BeamLine Operations and Safety Awareness
(BLOSA) training if they have not had it on that beamline in the last two
years. If there is any confusion, the beamlines keep lists of who has current
training somewhere on the yellow board. Anyone with current BLOSA training may
train the others; there should be a form for this somewhere on the yellow
board.
The
next step is to perform the safety checklist. The details of this are covered
as part of your BLOSA training, but mainly consists of checking that all
beamline shielding is in place.
Next,
call the Operations Coordinator (OpCo): dial 5824, wait for the beeps, and then
enter the extension for the beamline you are on. The OpCo will call you back
within a few minutes. Tell them you are ready to begin your experiment. They
will come out, collect your safety checklist, ask a few questions (e.g. has
everyone had BLOSA training?), and then unlock your beamline. If there are any
experimenters who were not included on the SAF, have the OpCo add them.
Open
the safety shutter (not the photon shutter).
Start
the data collecting program, if you haven’t already, and change the directory
to denBoer (it’s often good to make a new subdirectory for the particular run).
Also find an old html log file (they have names like June19), open it in Netscape
Composer, save it using today’s date as its name, and delete all the text (but
leave the nicely formatted table). Use this file as your labbook, saving often.
Record all scans of any type in the table, using the first column for file
name, the second column for military time, the third for scan regions, the
fourth for step sizes, the fifth for integration times, and the last for
comments. Other tips: “Insert Target” lets you put in an entry for later links.
Youshould do this for every new sample, temperature, etc.. “Insert Table Row”
is self-evident.
Users
arriving later:
These
users must undergo the safety and BLOSA training as above. If for some reason
they are still not on the SAF, call an OpCo and add them.
Beginning
an experiment:
In
the data collection program, choose an appropriate edge and then move 500 eV
above that edge. When the safety shutter is first opened, it takes a while for
the crystals (and, on X23B, the mirrors) to warm up. Letting them sit at the
right energy while you do other things helps prevent problems later.
Choose
an appropriate gas for the detectors. Nitrogen is good below about 13keV, argon
above that. This is not a hard-and-fast line; if you’re going to be running a
bunch of samples at 7 keV and one at 14 keV, you might as well use nitrogen the
whole time. But a 30 keV sample really requires argon (the nitrogen does not
absorb enough). Helium is also available, but probably won’t be needed (it
would be for really low energies). A flow rate of about 40 on the venturi meter
is good for all gases.
Mount
the sample on the stage so it looks like it’s aligned roughly correctly.
Interlock the hutch and start the beam. The collection program should show an
increase in I0 and maybe It when the shutter is opened. If It does not increase
the sample is probably misaligned, but focus on I0 first.
Run
through the feedback procedure. X23A2 has a nice step-by-step guide on how to
do this at the beamline. For X23B this means adjusting the mirrors. In both
cases, it’s good to have someone show you how the first time you try it.
Move
the monochromator about 200 eV below the edge and look at I0. If it is less
than 1, the gain should probably be increased. Close the photon shutter and go
into the hutch. Find the Keitheley amplifier corresponding to I0 and increase
the gain by 1. On X23A2, also increase the current suppresion; the current
suppresion and gain should always be at equivalent settings. On X23B, the
current suppression is a bit trickier to set; have someone show you how the first
time.
In
the data collection program, choose the option “measure offsets.” You should do
this when you first start, whenever you change the gain, and whenever you
notice significant nonzero values for I0 and It when the photon shutter is
closed. Set the gains appropriately and measure the offsets.
Move
500 eV above the edge and repeat the feedback procedure.
Now
it’s time to align the sample. There are several methods for doing this:
Motor
scans: If you’re close to the right position, this is a good method. Go to
“move motors” in the data collection program, and choose the vertical sample
motor. Calibrate the current position to 0. Repeat for the horizontal sample
motor. Now exit the move motors section of the program, and choose a 1 motor
scan with the vertical sample motor. The scan should have one region, probably
from something like -10 to 10 (the numbers are millimeters) by 0.5’s, with a
one second integration time. Before you begin the scan, make sure the stage is
free to move that far! Now begin the scan, plotting It. Often, you’ll be able
to find the frame (It will drop to near zero). The motor scan should look like
a big flat-topped mountain. Choose a flat area, move the cursor to that
position, and calibrate that position as zero. Repeat with a horizontal motor
scan.
Burn
paper: Burn paper is available in the desk at each beamline. If you want to use
burn paper, tape a piece over the sample and leave it in there with the beam on
for at least five minutes. Then, after a few more minutes, you’ll see a dark
spot or line where the beam hit the paper. This can help you align.
Phosphorescent
card: Each beamline has a small phosphorescent card that glows when the beam
hits it. This is more useful on X23A2 then X23B because of the camera. Feel free
to put a small piece of tape on the monitor to mark the position of the beam
determined in this way, but realize if the card was significantly in front of
the sample, then the monitor view can be misleading (the camera is at a large
angle to the beam).
Now
that the sample’s aligned, you can see if the gain needs to be adjusted on It.
If it does, you need to remeasure offsets, but you do not need to go through
the feedback tuning again (feedback works off of I0).
Next,
do a rough energy scan to find the edge. One region of -200,200 by 5 eV steps
and one second integration time should do it. Plot ln(I0/It). (If you’re using
a reference, plot ln(It/Ir) instead.) When you find the edge, move the cursor
to the peak in the first derivative spectrum and calibrate that as E0. Now do a
more detailed edge scan, perhaps -20,20 by 0.5 eV steps with a one second
integration time. Recalibrate.
Finally,
the EXAFS scan. Typical settings would be three regions: -300,-30,20,20k with
steps of 5, 0.5, and 0.05k and integration times of 2 or 3 seconds would be
typical. Use 3 seconds and at least 20k for cumulant studies. 2 scans are
generally sufficient, if they look consistent.
Displex:
Definitely
get help from someone who has done this before!
The
Displex must be counterbalanced. The stages are not strong enough to move the
whole displex up and down; if you attempt it, you may burn out the motors.
Setting up the counterweights is at least a two-person job.
The
EXAFS pool has instructions for setting up the Displex. The temperature
controller they use is different from ours, but the rest of the instructions
are essentially accurate. A turbopump is optional; it works OK with a rough
pump attached directly.
Our
temperature controller is a bit odd: to set the temperature, hold down ENTER
for a second or so. Then use SCROLL to set the first digit of the temperature.
Then hold down ENTER again for the next digit, and so on. When you’ve set the
temperature, hit RETURN. RETURN also can be used on its own to disable the heater;
that will cause the Displex to cool to its minimum temperature regardless of
the setting. Also note that the temperature controller overshoots a bit. Near
phase transitions, first choose a temperature 1 K above your desired
temperature, then, once it passes that temperature, choose the temp you desire.
That will keep it from overshooting so much. Depending on the size of your
sample, you should wait a few minutes for the sample to equilibrate.
Things to know how to do before starting.
1)
To
plot in Origin highlight the columns you wanted plotted and the go to plot on
the
toolbar. You want a line
plot to be specific.
A- Start with raw data. Usually will have different
scans of a sample.
1)
Average
different scans. In Origin go to analysis -> average multiple curves.
2)
Plot
this average
3)
Once
averaged, go to file and export. The extension should be .exf . For example if you
are working on copper, the name can be Cu.exf.
4)
Close
Origin and open your .exf file. You need to put a heading and title, so find
someone else’s old .exf file and copy and paste their heading into your file.
You can now make the appropriate changes such as putting the name of your atom.
Be warned tabs and spacing will not allow you to use this file so once you have
copied the heading try to not move space anything.
B- Now you are ready to do the background
subtraction.
There was a gradually change in
absorption due to other things so now you have to subtract this junk out.
1)
Find
someone else’s old file and copy the programs Feff800, Dos4GW.exe, Feffit,
Autobk, Atoms and autobk.inp and paste into your file.
2)
In
autobk.in the data file should have the extension .exf and the output should
have the extension. chi.
3)
Now
you have to pick your parameters.
4) Go
back to your graph of the average scans. Go to analysis on the toolbar->
calculus
->
Differentiate. Once you have a
differentiated graph of the average, go to tools on the toolbar -> pick
peaks -> find peaks. This will find the maximum peak of the graph and that
is Eo
5) r-bkg
is usually picked by trial and error. Typically between .9-1.9. If r-bkg is too
big the background will tend to oscillate. If r-bkg is too small the graph may
not be able to curve enough.
6)
E-min
tells us how far above the edge we should start fitting. Usually 20-30eV above
the edge.
7)
Generally
kweight = 0 and dk= 0
8)
Save
and run the program Autobk.
9) Now
go to Origin and look at background. Open Origin, import (single asci). When
the import window opens up you to your file and type in *.bkg. Import the bkg
in energy space (usually it will be .ebkg). Once it is imported plot the first
two columns. The graph you just created is the background.
10) You have to now compare the background and
your actual data. In order for you to do this you will have to put both graphs
on top of each other. Open your graph of the average scans, now highlight the
first two columns of the bkg worksheet and click on top the average graph. When
you do this graph should appear on the toolbar -> add plot to layer
->line. Now you should have the graphs on top of each other. Click on the
graph and go to plot details and change the color of the background to red.
Now you will generate your
theoretical model using x-ray Diffraction Data.
1) In
the program RETRIEVE, go to Search -> chemical composition. When you see the
periodic table put in the atom you are looking for and let the program find
various entries. While choosing an
entry you want one that correlates with your temperatures and pressures. Also
try to find one that is most recent.
2) In
your lab notebook copy down important information such as a, b, c, x, y, z
coordinates. Also copy the space group (be careful with the way it is spaced –
it makes a difference). Also copy the source, date and entry number.
3) Copy
someone’s old atoms.inp file and ATOMS.exe.
4) Put
in the appropriate title and values. Core atom is the one you’re focusing on.
R-max is usually 6-7 Angstroms.
5) Now
run the program ATOMS.exe. This will in turn create a FEFF.inp file.
6) There
some changes to be made in the FEFF.inp because we are using FEFF800.exe
7) Changes
for FEFF.inp - Change PRINT to PRINT 0 0 0 0 0 3
ADD: SCF 5.0
(this means self-consistent field calculation)
The 5.0 is the distance out to which this calculation is
done. The
documentation suggests "past second coordination shell”
8) Now
run FEFF800.exe. Feff figures out what model should be; now generates all
possible
scattering paths.
9) Last
file you will need is feffit.inp, again copy it from someone else.
10) Make the
appropriate title changes. Your data should have the extension .chi and your output the extension .dat. The first
time you run feffit nofit = true. This will do a fourier transform from k space
to r space
11) Now run
feffit.exe.
12) Open up Origin
-> File -> New -> Worksheet -> ok. Now go to File -> Import
-> single ASCII -> make sure you are in your folder in file name type
*k.dat and click on the k.dat file when it appears. Highlight the first two
columns go to plot on toolbar ->line.
13) Now you should
have a k space graph. You will figure out k min and k max from this graph. K-max is where the wiggles on the k graph
stop. Attached is a k graph for FeS. I chose the kmax = 10. K min is a little
tricky. Look at your graph of the background and average data plotted together.
Zoom in to the area right after the edge. Look at where the background is
compared to the data. Look at the example of FeS. The background goes above the data, then below and then above
before coming together with the data. So I had to remember above (up), below
(down) and above (up). Keeping this in mind go to the K-graph. Follow the graph
from the beginning it goes up, then down and back up. If you followed it with
your finger you should have stopped around 4 on the x-axis, and that was my
K-min. Finding K-min and max is tricky it takes time to get the feel for it.
14) Once you pick
k-min and max put it in your feffit.inp file and change nofit = false.
15) Once you have done
this run feffit again.
16) Open up a new
worksheet in Origin. Go to File and import r- data. File name type is *r.dat.
Once you have imported this highlight the first two columns and plot. This is
your r- data.
Finally time to FIT
1) In
your feffit.inp you can either set or guess parameters. For the ones you guess
the program with generate the best-fit values and the uncertainties. R-range is
the range on your r-graph, which you are trying to fit. You can vary this as
well as eO, SO2, sigma2, third1 and delr.
2) Once
you have finished making changes in the feffit.inp file, run feffit. Feffit
will also generate a log file, which will give an r factor, your best-fit
values and uncertainties for the parameters you had guessed. The r-factor is
how good your fit is. Generally a good fit has an r-factor less than .03.
3) To
plot your fit against your data, open up Origin and make a new worksheet.
Import from your folder your rfit which would have the file name type *.fit.
Now you have to plot this fit on top of your r-data. Highlight the first two
columns on the r-data worksheet and plot ->line. Now highlight the first columns
of the fit worksheet. Now click on r-graph and go to Graph on the toolbar ->
add plot to layer-> line. Now the two graphs should on top of each other.
Click on the graph and a window should open up go to plot details and change
the color of the fit to red.
After collecting and averaging the data over
all scans, one is ready to subtract the background absorption. This is done by
a computer program Autobk which removes the background absorption in such a way
that EXAFS retains all the structural
information about the studied material. In addition to this, in order to find
the normalization constant, Autobk normalizes the pre and post edges
The background subtraction process proceeds as
follows:
1. Export the average data to a file whose name has
to have the suffix .exf. For example if you are analyzing copper, then the name
of the exported file can be Cu.exf. In fact, if you forgot to export the data,
autobk will not work.
2. Copy an old Autobk.input file and make changes
relevant to the studied material such as changing the title, the energy edge
and the temperature.
In order to find the energy edge E0, you
can differentiate the average of the scans.
If you are using Origin, go to the graph option on the tool bar, choose
calculus and then choose derivative. The energy edge corresponds to the maximum
of the derivative (graph1).
Graph1
3.Determine the value of the following parameters
a) Emin
: low energy value relative to E0 of the region over which the
background function
is fitted. Emin corresponds to
where you want to start fitting your background which in turn corresponds to where
you think EXAFS starts.
b) pre1 :
low energy limit of pre-edge range, relative to E0.
c) pre2 :
high energy limit of pre-edge range, relative to E0.
d) rbkg : The highest r value to consider
background. Its value ranges from0.9 to 1.9 A depending on the material
studied. In general, you cannot go below an rmin of 0.9 A simply
because you will then be inside the
absorbing atom. However, an rbkg
larger than 1.9 A is not appropriate because some absorption effects will then
be ignored. It is customary to start
with an rbkg of 1.1.
Graph2
e) The final step is to run the Autobk program and
plot the background function in the same graph as the averaged collected data
in order to see the quality of the background subtraction. A good background subtraction
corresponds to a smooth variation of the
background function with respect to the average data. Graph 2 illustrates a good background
subtraction.
If your background oscillates a lot, then you should
try changing your rbkg to a low value with steps of 0.1 until you get a good
background. On the other hand, if the
background function does not oscillate at all, then you will have to slightly
increase the rbkg until you get a background that varies smoothly with respect
to the data.
Graph3
After subtracting the background absorption, running
atoms.exe and Feff.exe, one is ready to start fitting the data. The procedure
is as follows:
1. Copy an old feffit input file into your folder
and make changes relevant to the studied material such as changing the title
and the temperature at which the material is analyzed.
2.determine the r-range. To do so, it is helpful to
go to the paths.dat file and look at
the distances between the absorbing atom and the various neighboring atoms.
Determine the paths that interest you most and set up the r-range accordingly.
In fact, it is always recommended to first fit the first path or the first two
paths if they both have equal amplitudes. For example, if the first path
corresponds to a distance of 1.9, and the second path with a lower amplitude
corresponds to a distance of 2.4, a
reasonable r-range can be: 1.7 < r < 2.2 if you choose to fit the first
path only.
3. Determine the k-range.
kmax
corresponds to where EXAFS gets very weak or overwhelmed with noise. In this
case kmax can be 16.0. To determine kmin, we have to look first at the
background subtraction graph ( graph 3) and determine where the background
starts fitting the data properly. In graph 3, this can be taken few electron
volts beyond Emin. This corresponds to the background going relative
to the data: up, down, up, down.
The next
step is looking at the k-data. From the k-graph, we see that kmin
corresponds to: 7.8
4.
Set up a kweight value, if the EXAFS is clear, one can use a k-weight of zero,
however, higher kweights can be used to see EXAFS more clearly. It is customary
to use kweight of 1 or 2.
The
fitting parameters are:
|
E0
|
Energy
shift from absorbing edge, it is usually couple of eV’s.
|
|
SO2
|
Amplitude
factor accounting for energy transfer from photoelectrons to valence
electrons. Its maximum value is 1.
|
|
delR
|
Difference
in bond distance between measurement and model. It is usually about 0.01.
|
|
sigma2
|
Debye-Waller
factor; mean square displacement due to thermal vibration and static
disorder. It is always positive and about 0.001.
|
|
third
|
third
cumulant; a measure of the local anharmonicity of the interatomic potential
and is usually about 0.0001 or smaller.
|
a) As previously mentioned, it is recommended to fit
only the first path when first fitting the data because you will have an idea
about the value of the parameters. In addition, the third cumulant is usually
set to zero in first fits because including it might just improve the fit a
little bit but not dramatically.
b) After fitting the first path and getting a decent
fit, you can go ahead and include further direct scattering paths. The number
of the fitted paths depends on the r-range.
In fact, it is very useful to look at where each
path is. To do so, go to the feffit.inp, change the nofit from false to true,
run feffit, then plot several paths in the same graph as the r-data. This will
allow you to see where every path is and accordingly determine which paths to
include in you fit.
c) Multiple scattering paths are usually omitted because
they tend to confuse the fit. However, if these paths have a high amplitude,
one might try fitting them. If the fit is not good, one can try to introduce
the multiple scattering paths in terms of the direct scattering ones. For
example:
In
this case,
delR3=
(delR1+delR2)
sigma23=(
sigma21+sigma22)/2
d)
Changing k-range can improve the
quality of your fit. Try high kmin or low kmax.
Checking
the stability of the fit is the last step in EXAFS analysis. Its purpose is to
check whether you have a false fit or a stable one. Therefore, perturbing the
fit by changing the variables by small values is the best way to check its
stability. For example, you can try changing:
1.the r-range in steps of 0.1 A keeping all other
parameters unchanged. However, when the paths are very close, changing rmax
might result in a different fit. In this case, changing rmin can be
enough.
2. The
k-range by changing kmax by
small steps of value 0.02. It is possible that changing kmin
might result in a totally different fit, therefore changing kmax might be
enough.
OK, so you know how to obtain a rough fit, and how to
improve it. How do you know when to stop?
The answer will, of course, depend on the purpose of the
fit. Some studies demand high quality results, others can be more rough. And as
long as you’re making rapid progress improving a fit, you might as well keep at
it. But if the quality of the fit seems to be stalling out, you want to know
whether a fit is “good enough” to consider done. Following is the minimum
criteria for a “good enough” fit:
Minimum Criteria for Acceptable Fit
·
r factor less than 0.04
·
delR for paths of interest includes values between –0.2
and +0.2. For example, -0.6 +/- 0.5 would be minimally acceptable, since –0.6 +
0.5 = -.1, which is within the range. Likewise, 0.6 +/- 0.5 is also OK, but 0.6
+/- 0.2 is not. “Paths of interest” is defined by the study being performed…it
doesn’t mean every path in the fit, only those for which results will be
reported.
·
Range of s2 for paths of interest includes values
between zero and 0.05 Å2.
·
Range of 
includes values
between 0.25 and 1.00.
·
Range of E0, if fitted, includes values
between –10 and 10 eV.
·
Parameters of interest for paths of interest are stable
relative to small changes in fitting parameters such as k-range, r-range, k-weighting,
window method, constraints on “non-interesting” paths, and constraints on
energy shift
·
E0, delR, and third are often correlated. If
you are interested in the third cumulant, constraining delR to 0 should not
change the third cumulant by much.