August 2004 - Ifeffit - millenia.cars.aps.anl.gov

Re: High-R data
by Peter Southon 27 Aug '04

27 Aug '04

Bruce, Matt, Thanks for your help, I obviously haven't properly found my way around the Ifeffit manual yet! The rmax_out parameter was just what I needed. My apologies Bruce - I must have been sloppy in checking the k-plots of data. Athena indeed plots all the data beyond 20 Å-1. Thanks also for the advice on epsilon_k. It seems that I won't have to change the rwgt1 and 2 after all, there isn't anything above 15 Å. (In fact has anyone seen exafs at 15 Å?) Regards, Peter

2 1

High-R data
by Peter Southon 26 Aug '04

26 Aug '04

Hello all, Athena and Artemis do not plot chi beyond 20 Å-1 or the Fourier transform beyond 10 Å. Is this a limitation of Athena/Artemis or of Ifeffit? Can it be changed? We have exafs data that may be usable almost to 21 Å-1, and with quite a strong FT peak (from a Pt atom) at 10.5 Å. The latter we would not even have noticed had we been using only Athena to process our data. The 10.5 Å peak raises another issue. I understand that Ifeffit by default assumes that the high-R region is pure noise in order to make its estimates of uncertainty. (I know the actual R-range that is used was mentioned recently but I can't find it in the list archives or any of my manuals.) Can this range be changed? Does anyone have any advice on how to deal with this issue? Thanks, Peter Peter Southon Postdoctoral Fellow School of Chemistry University of Sydney NSW 2006, Australia +61 2 9351 7482

3 3

Export "flattened" normalized spectra
by Norbert Weiher 25 Aug '04

25 Aug '04

Dear Bruce, I was just wondering if you could implement saving the "flattened" Norm(E) spectra from ATHENA. I really like the plotting option for the XANES, but do my final plots always with grace... so I need to put the data in there. Would be great to have a quick option to have these data at hand... Cheers, Norbert -- Dr. rer. nat. Norbert Weiher (weiher(a)chem.ethz.ch) Institute for Chemical and Bioengineering - ETH Hönggerberg HCI E 117 - 8093 Zürich - Phone: +41 1 63 3 48 32

2 3

statistics
by Stefano Ciurli 24 Aug '04

24 Aug '04

Hello folks, I am sure this is an easy question answered many times, but I need some refresh: When it comes down to compare fits using different models, which is the relevant parameter? Chi square (and reduced) or R-factor? I have fits with smaller chi square and larger Rfactor and viceversa... Is it possible that a rationale would be that when the models to compare have the same number of atoms these parameters go in parallel? And that when the number of atoms is different one must rely more on chi square rather than R-factor (or viceversa)? Oh well. Stefano -- ____________________________________________ Stefano Ciurli Professor of Chemistry Department of Agro-Environmental Science and Technology University of Bologna Viale Giuseppe Fanin, 40 I-40127 Bologna Italy Phone: +39-051-209-6204 Fax: +39-051-209-6203 "Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza" Dante Alighieri - Inferno - Canto XXVI "Ihr seid bestimmt, nicht Tieren gleich zu leben, Nein, Tugend zu erringen und Erkenntnis" "Ye were not form'd to live the life of brutes, But virtue to pursue and knowledge high"

2 2

Last few questions
by Scott Calvin 24 Aug '04

24 Aug '04

Dan Carter wrote: > Thats great about the R-factor, I was very impressed to see it so low. > Reduced Chi-square = 382.556284248 > R-factor = 0.015036203 > (fit I started with--two e0's: one for 1st path/1 for others) > > I'm wasn't positive about how to use the third cumulant. So, I made a >variable third to guess in that position for the first path (the nearest >neighbor, right). Here are the results: > Reduced Chi-square = 415.039996182 > R-factor = 0.014768538 > third = 0.000045 > (Can you tell if this is a better fit from the Reduced Chi-square) > > If I separate ss's for the first path and the others (leaving the third >parameter) I get this: > Reduced Chi-square = 462.878645915 > R-factor = 0.014748364 > (not much improvement) > > One more thing, for the mailing list: How can I find out if it is ok to >separate the e0's of the nearest neighbor from the others? > > Thanks, Dan > > >

4 7

RE: [Ifeffit] MoS2 fit
by Kelly, Shelly D. 17 Aug '04

17 Aug '04

Hi Dan, Why are you using only one delr value and two dE values? I would suggest that you try it the other way around. two or more delr values with only one dE value. When you try the different models compare the reduced-chi-square values as you are doing but also look at the Backfourier transfrom of the data and the fit - or the chi(k) data if it is clean. Are they crossing zero at the same places. Do they miss more at low q or high q? Missing at low q/k would suggest separate dE values while missing at high q/k would suggest separate delr values. Would your model be better described by using an overall expansion/contraction term rather than one delr value? Using one delr means that all the paths move by -0.01 angstroms. Using one expansion term would allow all paths to contract by 0.1% which might be more reasonable. So that the long paths shrink in more than the short paths. To use a expansion term use the special "reff" variable which stands for the "effective r" i.e. the effective half path length. set delr1 = alpha*reff set delr2 = alpha*reff guess alpha= -0.01 -yes please drop the third cummulant. HTH Shelly > -----Original Message----- > From: dmc(a)pdx.edu [mailto:dmc@pdx.edu] > Sent: Tuesday, August 17, 2004 1:24 PM > To: XAFS Analysis using Ifeffit > Subject: [Ifeffit] MoS2 fit > > > Hi Mailing list, > > Well, the chief concern I have with my fit now is that > I have some > negative parameters (delr and e0_1): > > Guess parameters +/- uncertainties (initial guess): > > amp = 0.7709790 +/- 0.0631820 > e0 = 1.1165150 +/- 1.0100360 > delr = -0.0106440 +/- 0.0077110 > ss = 0.0032510 +/- 0.0005200 > e0_1 = -4.6848070 +/- 1.7781210 > third = 0.0000370 +/- 0.0001070 > > So, I guess I can drop the third cumulant. > > This is a fit using 6 paths (1,2,3,6 are single scatterers) > These are the results this time around: > > Reduced Chi-square = 410.285252396 > R-factor = 0.014599348 > > As you can see I am still using two e0's, it is still > dramatically improving > the fit. Without: Reduced Chi-square = 1133.485950636 > R-factor = 0.044551155 > > any advice welcomed, Dan > > > > _______________________________________________ > Ifeffit mailing list > Ifeffit(a)millenia.cars.aps.anl.gov > http://millenia.cars.aps.anl.gov/mailman/listi> nfo/ifeffit >

3 3

Re: Artemis Fit Questions
by Scott Calvin 17 Aug '04

17 Aug '04

<excerpt>Hi Dan, At 12:58 PM 8/16/2004 -0700, you wrote: >Scott-- > >Ok, I'll put it on the mailing list from now on, but I used a different model >from the ICSD and it worked muc better. I used the hexagonal model and this is >the results: > >Reduced Chi-square = 382.556284248 >R-factor = 0.015036203 >Do you use these to determine if the fit "doesn't work"? > The improvement in the stats is suggestive that the hexagonal model may be better than the trigonal model. As a rule of thumb, R-factors of below 0.02 are generally indicative of a <italic>statistically </italic>good fit (there may still be other reasons to doubt it!) and between 0.02 and 0.05 is indicative of some problems with the model (perhaps how the parameters are constrained or impurities), but is still on the right track (I've published at that level). The reduced chi-square is nearly worthless as an absolute measure, because of the necessity for having a good estimate of the measurement error. Reduced chi-square <italic>is</italic> useful in the way you're already using it, though...for comparing two fits of the same data. >About separation of the three paths they are here: > ># degen reff amp fs scattering path > 0001 6 2.414 100.00 : [+] S_1 [+] > 0002 6 3.150 59.25 : [+] Mo_1 [+] > 0003 6 3.968 21.22 : [+] S_3 [+] > >What is fitting a third cumulant? You'll see it as a path parameter in Artemis under "third." It is a measure of the asymmetry of the absorber scatterer distance: a positive third cumulant means it is more likely to find the absorber and scatterer a large distance above the mean than a large distance below the mean. It is sometimes helfpul to fit a third cumulant for nearest-neighbors because they are bonded; it is less appropriate for further shells, because the atoms move more independetly. >The two e0's really seems to improve the reduced chi-square. >The two ss's increased the reduced chi-square. > Hmmm...I don't know how you're grouping them between the three paths. I'd try one ss for the nearest-neighbor, and one for the next two paths. You might be doing one for Mo and one for S. > Thanks, Dan > --Scott Calvin Sarah Lawrence College > </excerpt><<<<<<<<

1 0

Re: Artemis Fit Questions
by Scott Calvin 17 Aug '04

17 Aug '04

Dan Carter wrote: >Ok, I'll put it on the mailing list from now on, but I used a different model >from the ICSD and it worked muc better. I used the hexagonal model and this is >the results: > >Reduced Chi-square = 382.556284248 >R-factor = 0.015036203 >Do you use these to determine if the fit "doesn't work"? > >About separation of the three paths they are here: > ># degen reff amp fs scattering path > 0001 6 2.414 100.00 : [+] S_1 [+] > 0002 6 3.150 59.25 : [+] Mo_1 [+] > 0003 6 3.968 21.22 : [+] S_3 [+] > >What is fitting a third cumulant? >The two e0's really seems to improve the reduced chi-square. >The two ss's increased the reduced chi-square. > > Thanks, Dan >

1 0

Re: Artemis Fit Questions
by Scott Calvin 17 Aug '04

17 Aug '04

> >Hi Dan, > >At 12:27 PM 8/16/2004 -0700, you wrote: >> Should I add more paths? > >I always feel more paths are better, until the fit stops working. They will >reduce correlations and uncertainties in general, and also reduce the >incidence of false minima. That does mean you'll have to include some >multiple scattering once you get far enough out, but IMHO poorly >constrained multiple scattering is better than none at all. > >>Is there other parameters that I should add? >>Was it Ok to make 2 separate e0's? > >This is an ongoing debate--it cropped up again on the Ifeffit list >recently. I am not enamored of separate E0's without a very good reason, >but others use them more freely. On the other hand, a separate ss for the >nearest-neighbors from paths further out (I can't tell from your >description if there's a significant separation between the three paths) is >often called for. You might also consider fitting a third cumulant; the >effect on a spectrum is not unlike fitting a separate E0. > >>I think this is a pretty good fit, but I >>feel like there has to be something to improve it. >> > >Seems not unreasonable, although it's hard to tell without uncertainties. >That r-factor is a bit high for a three-path, five parameter fit, though. > >--Scott Calvin >Sarah Lawrence College > >

1 0

Artemis Fit Questions
by Scott Calvin 17 Aug '04

17 Aug '04

Hi all, I've been exchanging emails with Dan Carter, a recent newcomer to EXAFS analysis, and I thought some of the other novices out there might find pieces of it useful. Also, it's always interesting to watch the interplay of the advice from the more experienced users on these basic issues! Therefore, I'm forwarding (with Dan's permission) our exchanges so far to this list (don't worry, there's only a handful!). --Scott Calvin Sarah Lawrence College >Hi Scott, > > Here at Hunter we are trying to make some Li -- MoS2 batteries. >Therefore, I'm trying to do EXAFS on MoS2 and running Artemis on the same. >Basically, I'm trying to create the best fit possible and I was hoping that if >I gave you what I know so far you might be able to give me some suggestions to >improve the fit. > > MoS2 is trigonal or hexagonal depending on who you talk to. >I have 3 paths -- All single scattering below reff = 4 Angstroms. >I made two e0's one for the first path and one for the second, it improved the >fit. All parameters are guesses. > > This is the result I got: >Reduced Chi-square = 809.564600728 >R-factor = 0.039774809 > amp = 0.7511730 > e0 = -1.0221910 > delr = 0.0010090 > ss = 0.0031050 > e0_1 = 2.3828120 > > Should I add more paths? Is there other parameters that I should add? >Was it Ok to make 2 separate e0's? I think this is a pretty good fit, but I >feel like there has to be something to improve it. > > Please help, Dan Carter > >

1 0