polarization in Artemis
Hi, Everything in the discussion between Eugenio and Matt is stuff I agree with, except ... When I rewrote Artemis, one of the motivations was that I wanted to write my own pathfinder rather than to continue using the one in Feff. My main motivation was fuzzy degeneracy (http://bruceravel.github.io/demeter/artug/extended/fuzzy.html) plus I wanted to correct a few other things I saw as shortcomings. When I was writing it years ago, I didn't immediately see an easy way to consider polarization correctly while also doing fuzzy degeneracy correctly. I have since figured that out, but years went by and only one person ever asked me about polarized feff calculations in Artemis. So I never implemented it. It's a funny thing, when you wait for a bus, none comes, then two show up at once. A month ago, I was asked about polarization in Artemis, and now Eugenio brings it up! Polarization works (although it needs testing) in the very latest code in github. But I did that work since the last time I built a Windows installer. Ellipticity does not yet work (in Artemis -- it works just fine in Feff, but Artemis does not deliver the ellipticity-dependent calculation to the user). Anyway, I will try to find a few hours this week to build a new Windows installer, which will included the ability to consider linear polarization.
Artemis used to make this very easy (sum paths without fitting), but I don't recall how to do this in the latest version.
Still is easy. Well, I think it is. YMMV :) http://bruceravel.github.io/demeter/artug/plot/vpaths.html B -- Bruce Ravel ------------------------------------ bravel@bnl.gov National Institute of Standards and Technology Synchrotron Science Group at NSLS-II Building 535A Upton NY, 11973 Homepage: http://bruceravel.github.io/home/ Software: https://github.com/bruceravel Demeter: http://bruceravel.github.io/demeter/
On Sun, May 24, 2015 at 1:50 PM, Bruce Ravel
Hi,
Everything in the discussion between Eugenio and Matt is stuff I agree with, except ...
When I rewrote Artemis, one of the motivations was that I wanted to write my own pathfinder rather than to continue using the one in Feff. My main motivation was fuzzy degeneracy ( http://bruceravel.github.io/demeter/artug/extended/fuzzy.html) plus I wanted to correct a few other things I saw as shortcomings.
When I was writing it years ago, I didn't immediately see an easy way to consider polarization correctly while also doing fuzzy degeneracy correctly. I have since figured that out, but years went by and only one person ever asked me about polarized feff calculations in Artemis. So I never implemented it.
It's a funny thing, when you wait for a bus, none comes, then two show up at once. A month ago, I was asked about polarization in Artemis, and now Eugenio brings it up!
Polarization works (although it needs testing) in the very latest code in github. But I did that work since the last time I built a Windows installer. Ellipticity does not yet work (in Artemis -- it works just fine in Feff, but Artemis does not deliver the ellipticity-dependent calculation to the user).
You mean Artemis wasn't creating a POLARIZATION card? But one can just go in and add that the the Feff.inp anyway, no? The feff.inp file is editable before running Feff with Artemis all the time, right? That should be independent of pathfinder. Or do you mean something else?
Anyway, I will try to find a few hours this week to build a new Windows installer, which will included the ability to consider linear polarization.
Artemis used to make this very easy (sum paths without fitting), but I don't recall how to do this in the latest version.
Still is easy. Well, I think it is. YMMV :)
I understand VPaths, I think. But I meant simply "sum these paths applying the current parameters, without fitting". That used to be very, very simple and I always find it useful and instructive for exploring the contributions from different paths. Now it seems I have to set all values to "set", then change them to "guess" to really do a fit. Am I missing something? --Matt
On 05/24/2015 05:03 PM, Matt Newville wrote:
You mean Artemis wasn't creating a POLARIZATION card? But one can just go in and add that the the Feff.inp anyway, no? The feff.inp file is editable before running Feff with Artemis all the time, right? That should be independent of pathfinder. Or do you mean something else?
I mean something else. The pathfinder did not consider polarization AND when Artemis ran the genfmt part of feff, the polarization card was not passed along. So, regardless of whether you edit the feff.inp window to included a POLARIZATION card, that information was not being passed to feff when the feffNNNN.dat file was generated. There's no convenient way to intervene in that step. It's been years, with not much complaint, so I kind of forgot about it. My apologies to George Strbinsky -- I should have fixed it when he asked me about this a couple years ago. All of that will be fixed soon, with the caveat that ellipticity will not yet be supported by Artemis. (So the "polarization in the plane" trick won't work in Artemis. Again, this is an Artemis shortcoming, not a Feff shortcoming.)
I understand VPaths, I think. But I meant simply "sum these paths applying the current parameters, without fitting". That used to be very, very simple and I always find it useful and instructive for exploring the contributions from different paths. Now it seems I have to set all values to "set", then change them to "guess" to really do a fit. Am I missing something?
Well, I suppose it is possible that things aren't working as intended, but the intent is that you could make a VPath out of /all/ the paths (i.e. select "Mark all" from the Mark menu, then select "Make VPath" from the actions menu, then plot in k, R, or q). The intent also is that the current values for all parameters will be used when the plot is made. If no fit has been performed, the initial values will be used. It should not be necessary to change all guesses to sets just to make that plot. I suppose that could be streamlined... B -- Bruce Ravel ------------------------------------ bravel@bnl.gov National Institute of Standards and Technology Synchrotron Science Group at NSLS-II Building 535A Upton NY, 11973 Homepage: http://bruceravel.github.io/home/ Software: https://github.com/bruceravel Demeter: http://bruceravel.github.io/demeter/
Well, looks like i'm raising a nice discussion :)
Summarizing, if I understood well at the present moment the effect of
polarization is not included in Artemis analysis. However, going back to my
scientific problem i think that the main contribution will come from the 4
in-plane S atoms, so as a first approximation I can use the @S1.1@ path and
try to fit my EXAFS with that neglecting the out-of-plane contributions.
By the way, i've tried the VPaths and I've found it a nice and powerful
tool! Previously i used to do it by exporting the single paths, summing
them up and then import it again as a new X(k)... I don't really read all
the documentation and this is my fault!
Thank you for your help and for clarifying the issue about polarization
that was not really clear to me, I'm looking forward for the next Artemis
version :)
Eugenio
2015-05-24 23:37 GMT+02:00 Bruce Ravel
On 05/24/2015 05:03 PM, Matt Newville wrote:
You mean Artemis wasn't creating a POLARIZATION card? But one can just go in and add that the the Feff.inp anyway, no? The feff.inp file is editable before running Feff with Artemis all the time, right? That should be independent of pathfinder. Or do you mean something else?
I mean something else. The pathfinder did not consider polarization AND when Artemis ran the genfmt part of feff, the polarization card was not passed along. So, regardless of whether you edit the feff.inp window to included a POLARIZATION card, that information was not being passed to feff when the feffNNNN.dat file was generated. There's no convenient way to intervene in that step.
It's been years, with not much complaint, so I kind of forgot about it. My apologies to George Strbinsky -- I should have fixed it when he asked me about this a couple years ago.
All of that will be fixed soon, with the caveat that ellipticity will not yet be supported by Artemis. (So the "polarization in the plane" trick won't work in Artemis. Again, this is an Artemis shortcoming, not a Feff shortcoming.)
I understand VPaths, I think. But I meant simply "sum these paths
applying the current parameters, without fitting". That used to be very, very simple and I always find it useful and instructive for exploring the contributions from different paths. Now it seems I have to set all values to "set", then change them to "guess" to really do a fit. Am I missing something?
Well, I suppose it is possible that things aren't working as intended, but the intent is that you could make a VPath out of /all/ the paths (i.e. select "Mark all" from the Mark menu, then select "Make VPath" from the actions menu, then plot in k, R, or q). The intent also is that the current values for all parameters will be used when the plot is made. If no fit has been performed, the initial values will be used. It should not be necessary to change all guesses to sets just to make that plot.
I suppose that could be streamlined...
B
-- Bruce Ravel ------------------------------------ bravel@bnl.gov
National Institute of Standards and Technology Synchrotron Science Group at NSLS-II Building 535A Upton NY, 11973
Homepage: http://bruceravel.github.io/home/ Software: https://github.com/bruceravel Demeter: http://bruceravel.github.io/demeter/ _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
participants (3)
-
Bruce Ravel -
Eugenio Paris -
Matt Newville