Dan, I too have had problems with positive sigma-squared values. I think it may indicate a deeper problem with your fit. Two things I can see that might be causing it are: 1. How did you come up with the value for amp (0.7) If your amp is too small, your debye-waller factors get smaller, and sometimes even become negative to compensate. One way to check this is fitting in different k-weights. If your ss variables stay relatively constant at different k-weights, your amp value is probably not the problem. If they change much, you might look at http://leonardo.phys.washington.edu/~ravel/course/notes.pdf page 37. 2. How about background subtraction? I don't know if Mo has a large white line, but if it does, k=2 is probably too close to the edge for the background subtraction to be trusted. Trying a larger k will get you away from the range where the background subtraction is questionable. Good luck, Mike On Thu, 4 Nov 2004, Kelly, Shelly D. wrote:
Hi Dan,
Sigma2 can not be negative. It is the mean square displacement of the half path length. You can force it to be positive by using abs(ss) on the definition page for each path in the fit.
Shelly
-----Original Message----- From: dmc@pdx.edu [mailto:dmc@pdx.edu] Sent: Thursday, November 04, 2004 10:19 AM To: XAFS Analysis using Ifeffit Subject: Re: [Ifeffit] Fitting Commercial Molybdenum Disulfide
Hi List,
Here's my most recent data. I was able to look at the individual paths and changed my R-range to 2-4.5 angstroms on the first set of data, I set the second set of data to the same range, although 4 works better as a maximum for that set. The k-range on the second range had to be truncated though.
Six paths are used for each fit, same parameters, except for amp (0.7 for one ~0.8 for the other). The seperate data sets still have a few parameters that are outside of the error bars of the other (ss2,ss3,ss4,alpha, alpha2,alpha5). What can I do to get these beamlines to agree?
delr_# define alpha# * reff all other parameters guessed. ss for each path, only one e0 for all
Thanks for the help, Dan
First Data set -- Full k-range works best, Full range used for fit. Project title : Fitting merge 10_15_04.chi
Guess parameters +/- uncertainties (initial guess): e0 = 0.8686930 +/- 1.1038290 (guessed as 0.868694 (1.103841)) ss = 0.0041150 +/- 0.0004990 (guessed as 0.004115 (0.000499)) ss2 = 0.0044360 +/- 0.0004450 (guessed as 0.004436 (0.000445)) ss3 = 0.0243560 +/- 0.0201400 (guessed as 0.024368 (0.020159)) ss4 = -0.0065470 +/- 0.0047820 (guessed as -0.006545 (0.004786)) ss5 = -0.0016100 +/- 0.0050490 (guessed as -0.001611 (0.005047)) ss6 = 0.0059980 +/- 0.0139390 (guessed as 0.005996 (0.013935)) alpha = -0.0011360 +/- 0.0029380 (guessed as -0.001136 (0.002938)) alpha2 = 0.0071480 +/- 0.0019780 (guessed as 0.007148 (0.001978)) alpha3 = -0.0060900 +/- 0.0273580 (guessed as -0.006078 (0.027378)) alpha4 = -0.0198570 +/- 0.0154200 (guessed as -0.019851 (0.015430)) alpha5 = 0.1399050 +/- 0.0148530 (guessed as 0.139906 (0.014848)) alpha6 = 0.0199630 +/- 0.0268690 (guessed as 0.019961 (0.026862))
Def parameters: delr_1 = -0.0051820 delr_2 = 0.0326180 delr_3 = -0.0277930 delr_4 = -0.0906110 delr_5 = 0.6384270 delr_6 = 0.0910950
Set parameters: amp1 = .7
Correlations between variables: e0 and alpha --> 0.8170 e0 and alpha2 --> 0.6726 ss and ss2 --> 0.6591 alpha and alpha2 --> 0.5516 alpha5 and alpha6 --> -0.4856 e0 and alpha3 --> 0.4083 alpha3 and alpha4 --> 0.3588 ss5 and ss6 --> -0.3581 alpha and alpha3 --> 0.3100 alpha2 and alpha3 --> 0.3045 ss2 and alpha6 --> 0.2745 e0 and ss3 --> 0.2649 ss and ss3 --> 0.2627 ss2 and ss3 --> 0.2626 ss6 and alpha5 --> -0.2510 All other correlations are below 0.25
Second Data Set -- Trucated k-range (2-13.5) Project title : Fitting merge 10_15_04.chi
Independent points = 25.270507813 Number of variables = 14.000000000 Chi-square = 1202.598221935 Reduced Chi-square = 106.703108852 R-factor = 0.057229121 Measurement uncertainty (k) = 0.000827712 Measurement uncertainty (R) = 0.022111003 Number of data sets = 1.000000000
Guess parameters +/- uncertainties (initial guess): amp = 0.7976070 +/- 0.1361170 (guessed as 0.776348 (0.139898)) ss = 0.0037770 +/- 0.0015530 (guessed as 0.003589 (0.001617)) e0 = 0.5608430 +/- 1.8018380 (guessed as 0.642687 (1.861307)) alpha = -0.0061290 +/- 0.0049080 (guessed as -0.005966 (0.005078)) ss2 = 0.0029050 +/- 0.0009970 (guessed as 0.002686 (0.001078)) ss3 = 0.0075410 +/- 0.0094220 (guessed as 0.005992 (0.008858)) ss4 = -0.0110250 +/- 0.0035850 (guessed as -0.011776 (0.004653)) ss5 = -0.0042000 +/- 0.0078070 (guessed as -0.007610 (0.005250)) alpha2 = 0.0038110 +/- 0.0030450 (guessed as 0.003762 (0.003187)) alpha3 = -0.0037820 +/- 0.0145580 (guessed as 0.001200 (0.015976)) alpha4 = -0.0212550 +/- 0.0107080 (guessed as -0.011852 (0.014417)) alpha5 = 0.1151980 +/- 0.0236640 (guessed as 0.113430 (0.014973)) ss6 = 0.0028560 +/- 0.0166050 (guessed as -0.007250 (0.003624)) alpha6 = 0.0193530 +/- 0.0299640 (guessed as 0.001791 (0.008195))
Def parameters: delr_1 = -0.0279670 delr_2 = 0.0173920 delr_3 = -0.0172570 delr_4 = -0.0969910 delr_5 = 0.5256830 delr_6 = 0.0883130
Correlations between variables: amp and ss --> 0.8382 e0 and alpha --> 0.8048 amp and ss2 --> 0.7964 e0 and alpha2 --> 0.7533 ss3 and alpha5 --> 0.7397 ss5 and alpha4 --> -0.6999 ss and ss2 --> 0.6591 alpha and alpha2 --> 0.6159 alpha5 and ss6 --> 0.6116 ss5 and alpha3 --> -0.5790 ss5 and alpha6 --> -0.5023 ss3 and ss6 --> 0.4804 alpha4 and alpha5 --> -0.4375 e0 and alpha3 --> 0.4083 ss4 and ss5 --> -0.4047 ss3 and alpha4 --> -0.3972 alpha5 and alpha6 --> 0.3827 alpha3 and alpha6 --> 0.3498 ss3 and ss5 --> -0.3278 amp and ss3 --> 0.3209 ss4 and ss6 --> 0.3181 alpha and alpha3 --> 0.3100 alpha2 and alpha3 --> 0.3045 ss4 and alpha3 --> 0.2948 amp and ss5 --> 0.2820 ss2 and alpha6 --> 0.2745 ss and ss3 --> 0.2627 ss2 and ss3 --> 0.2626 amp and alpha6 --> 0.2589 alpha2 and alpha5 --> 0.2581 All other correlations are below 0.25
_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit