Re: [Ifeffit] Why different installations of Atoms generate different results?
Dear Matt,
Thanks for your positive reply. You are opening a door to solving my
problem. Attached are the fpj file generated by (D)Artemis and the apj
file by the old Artemis, respectively. They cannot be opened by Note Pad
nor by Word Pad, so I do not know their contents.
I actually generated these projects from scratch solely by File -> open
project or data (Au.inp) in (D)Artemis (or alternatively from Feff
calculations -> Add) or by File -> Open file in the old Artemis without
inputting any data. I just wish to look at what kind of EXAFS spectra
generated purely by calculation before using them for model fitting.
For (D)Artemis I went through Run Atoms -> Run Feff and could select
any path from the path list, click chi(k) and then Plot selection. Of
course now I know how to change k-weighting in the Plot window. I know
which path is plotted because each path corresponds to a single curve
with its own color and the path names are included in the legend.
For the old Artemis I went through Run Atoms -> Run Feff and clicked
import All paths. Again I could select any path in the list and Plot k,
R and q with a choice of k-wieghting. Similarly each path corresponds to
a single curve with its own color and the path names are included in the
legend.
These steps illustrate how I copied these plots into the pdf file I sent
to the mail list.
The path page in the old Artemis contains some more information as Path
parameter math expression side by side to the Data & Paths list. They
are N:6, S02:amp, delE0:enot, delR:delr and sigma^2:ss. Indeed you are
right that changing them to numbers results in different plots. But I
thought I should keep the default path parameters intact because they
are supposed to be from calculation . Moreover, many of these path
parameters do not have similar math expressions in (D)Artemis. See the
next paragraph.
Yes, I followed Bruce's suggestion to go through the on line manual of
(D)Artemis with loading the Au foil data first then up to The Path Page,
aiming for fitting real data. I could generate paths from the Atoms and
Feff window and plot their chi(k) and so on, identical to the previous
project without loading in Au foil data. I could drag the selected paths
from the Atoms window into the Data window. By selecting each path the
associated parameters are displayed, but only N and (blank)=(S02)(?)=1,
the rest fields are empty. Are they supposed to be filled like in the
old Artemis? I then copied in math expression contents exactly the same
as in the old Artemis. The plotting shows only the Au foil data. Then I
pressed the yellow Fit button. A message box pops up: Fit cannot
continue: You have not defined any fitting parameters. Could you offer
more clues to continue? The (fit) project up to now is saved and attached.
There are more problems running (D)Artemis on my Windows 7 Professional
SP1. The program can load in the saved (fit) project described above but
always crashes on loading the first project without Au foil data, no
matter which compatibility mode I choose before starting. It may be
worth noting that only the (D)Artemis DOS window appears when choosing
None, Windows 7, Windows Vista SP2, Windows Vista SP1, Windows Vista,
Windows 2000, and Windows NT 4.0 SP5. On choosing Windows XP SP3,
Windows XP SP2, Windows 98 / Windows Me, and Windows 95 there appears
another DOS window on startup, D:\strawberry\perl\bin\perl.exe. I am not
sure if these information may help solving the cash problem of
(D)Artemis. On the other hand, the old Artemis can load in the project
file with no Au foil data without crashing.
Your help is greatly appreciated.
Ku-Ding
Date: Mon, 24 Sep 2012 09:10:48 -0500
From: Matt Newville
Hi Bruce,
Thanks a lot for your reply and comments particularly on how to change k-weight in (D)Artemis-Atoms. Sorry for not being able to provide clear results stating exactly where my confusion came from. I was not sure if an attachment could be readable to this email list but let me try this time. Certainly I do not know how to retrieve data values from these plots yet.
The attached pdf file compares the output plots of k^n*chi(k) with n=0, 1, 2 and 3 of the Au L3 EXAFS of the shortest path calculated from Artemis-Atoms in Demeter and old Artemis. I suppose to read identical plots of k^n*chi(k) from different programs but they are not. That is why I feel confused.
Your help on clearing my confusion is appreciated. Then I can move on to do model fit on real data.
Ku-Ding
The two results you show are quite different, but you don't tell us how you got them, expect to say the they were calculated by Artemis-Atom and D(Artemis)-Atoms. This is an incomplete description, allowing many possible differences. To be clear, Atoms only generates the feff.inp file. These should be the same between the two versions, especially for as simple a case as fcc Au. Next, Feff has to be run. This will be the same, assuming the input files are the same. Next, Artemis reads in the output files from Feff (you say you're using the first shell only, but I encourage you to verify that), applies Path Parameters to these, and generates chi(k). Are the number of Paths used, and the Path Parameters (N*S02, sigma2, e0, deltaR) the same in these two calculations? I'd guess that they are not. Providing plots of outputs is helpful, but as is said many times here, providing the project file is much more helpful as it shows not only the final result, but how you got there. Cheers, --Matt ------------------------------
The APJ files can be opened by changing their extensions to .zip and using an unzip utility.
----- Original Message -----
From: "Tsuei, Ku-Ding"
Dear Matt,
Thanks for your positive reply. You are opening a door to solving my problem. Attached are the fpj file generated by (D)Artemis and the apj file by the old Artemis, respectively. They cannot be opened by Note Pad nor by Word Pad, so I do not know their contents.
I actually generated these projects from scratch solely by File -> open project or data (Au.inp) in (D)Artemis (or alternatively from Feff calculations -> Add) or by File -> Open file in the old Artemis without inputting any data. I just wish to look at what kind of EXAFS spectra generated purely by calculation before using them for model fitting.
For (D)Artemis I went through Run Atoms -> Run Feff and could select any path from the path list, click chi(k) and then Plot selection. Of course now I know how to change k-weighting in the Plot window. I know which path is plotted because each path corresponds to a single curve with its own color and the path names are included in the legend.
For the old Artemis I went through Run Atoms -> Run Feff and clicked import All paths. Again I could select any path in the list and Plot k, R and q with a choice of k-wieghting. Similarly each path corresponds to a single curve with its own color and the path names are included in the legend.
These steps illustrate how I copied these plots into the pdf file I sent to the mail list.
The path page in the old Artemis contains some more information as Path parameter math expression side by side to the Data & Paths list. They are N:6, S02:amp, delE0:enot, delR:delr and sigma^2:ss. Indeed you are right that changing them to numbers results in different plots. But I thought I should keep the default path parameters intact because they are supposed to be from calculation . Moreover, many of these path parameters do not have similar math expressions in (D)Artemis. See the next paragraph.
Yes, I followed Bruce's suggestion to go through the on line manual of (D)Artemis with loading the Au foil data first then up to The Path Page, aiming for fitting real data. I could generate paths from the Atoms and Feff window and plot their chi(k) and so on, identical to the previous project without loading in Au foil data. I could drag the selected paths from the Atoms window into the Data window. By selecting each path the associated parameters are displayed, but only N and (blank)=(S02)(?)=1, the rest fields are empty. Are they supposed to be filled like in the old Artemis? I then copied in math expression contents exactly the same as in the old Artemis. The plotting shows only the Au foil data. Then I pressed the yellow Fit button. A message box pops up: Fit cannot continue: You have not defined any fitting parameters. Could you offer more clues to continue? The (fit) project up to now is saved and attached.
There are more problems running (D)Artemis on my Windows 7 Professional SP1. The program can load in the saved (fit) project described above but always crashes on loading the first project without Au foil data, no matter which compatibility mode I choose before starting. It may be worth noting that only the (D)Artemis DOS window appears when choosing None, Windows 7, Windows Vista SP2, Windows Vista SP1, Windows Vista, Windows 2000, and Windows NT 4.0 SP5. On choosing Windows XP SP3, Windows XP SP2, Windows 98 / Windows Me, and Windows 95 there appears another DOS window on startup, D:\strawberry\perl\bin\perl.exe. I am not sure if these information may help solving the cash problem of (D)Artemis. On the other hand, the old Artemis can load in the project file with no Au foil data without crashing.
Your help is greatly appreciated.
Ku-Ding
Date: Mon, 24 Sep 2012 09:10:48 -0500 From: Matt Newville
To: XAFS Analysis using Ifeffit Cc: bravel@bnl.gov Subject: Re: [Ifeffit] Why different installations of Atoms generate different results? Dear Ku-Ding,
On Sat, Sep 22, 2012 at 8:13 AM, Tsuei, Ku-Ding
wrote: Hi Bruce,
Thanks a lot for your reply and comments particularly on how to change k-weight in (D)Artemis-Atoms. Sorry for not being able to provide clear results stating exactly where my confusion came from. I was not sure if an attachment could be readable to this email list but let me try this time. Certainly I do not know how to retrieve data values from these plots yet.
The attached pdf file compares the output plots of k^n*chi(k) with n=0, 1, 2 and 3 of the Au L3 EXAFS of the shortest path calculated from Artemis-Atoms in Demeter and old Artemis. I suppose to read identical plots of k^n*chi(k) from different programs but they are not. That is why I feel confused.
Your help on clearing my confusion is appreciated. Then I can move on to do model fit on real data.
Ku-Ding
The two results you show are quite different, but you don't tell us how you got them, expect to say the they were calculated by Artemis-Atom and D(Artemis)-Atoms. This is an incomplete description, allowing many possible differences. To be clear, Atoms only generates the feff.inp file. These should be the same between the two versions, especially for as simple a case as fcc Au. Next, Feff has to be run. This will be the same, assuming the input files are the same. Next, Artemis reads in the output files from Feff (you say you're using the first shell only, but I encourage you to verify that), applies Path Parameters to these, and generates chi(k). Are the number of Paths used, and the Path Parameters (N*S02, sigma2, e0, deltaR) the same in these two calculations? I'd guess that they are not.
Providing plots of outputs is helpful, but as is said many times here, providing the project file is much more helpful as it shows not only the final result, but how you got there.
Cheers,
--Matt
------------------------------
--------------------------------------------------------------------------------
_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
On Tuesday, September 25, 2012 05:21:21 AM you wrote:
There are more problems running (D)Artemis on my Windows 7 Professional SP1. The program can load in the saved (fit) project described above but always crashes on loading the first project without Au foil data, no matter which compatibility mode I choose before starting.
I noticed this when looking at the project file you sent. Thanks for uncovering this bug. It is already fixed at GitHub and will be fixed in the next release of the windows installer. B -- Bruce Ravel ------------------------------------ bravel@bnl.gov National Institute of Standards and Technology Synchrotron Methods Group at NSLS --- Beamlines U7A, X24A, X23A2 Building 535A Upton NY, 11973 Homepage: http://xafs.org/BruceRavel Software: https://github.com/bruceravel
participants (3)
-
Bruce Ravel
-
Matthew Marcus
-
Tsuei, Ku-Ding