Extracting Fobs and Fcalc for MEM calculations

casperwa

Dear Topas forum

I have been using TA for over a year now, but have run into some problems, when I want to do maximum entropy (MEM) calculations (using BayMEM) based on Topas Rietveld refinements.

I need to extract lists of Fobs, Fcalc, and e.s.d.s for each hkl of a powder X-ray diff. pattern.

Out_FCF does not deliver the data I need. I have tried fiddling with the reserved parameters Fobs and Fcalc, and writing macros that "out"s these parameters, but fail every time because Topas cannot do "pk_to_Fobs".

The reserved parameter A01 holds either Fobs or Fcalc in my case. I think it is Fcalc, though I am not entirely sure. This parameter is easily extracted, but only solves 1/3 of my problem.

I hope someone is able to help, since I otherwise will have to go back to mainly using Jana2006.

Thank you
Casper

mfisch

Check my thread on the Fourier Map and the FCF_Vesta Macro, maybe this helps.

http://topas.dur.ac.uk/unb/forum.php?req=thread&id=303

http://topas.dur.ac.uk/topaswiki/doku.php?id=fcf_vesta

Extracting Fobs is probably easiest with a Pawley refinement as you will get the esds.
Reading section 7.5 of the tech ref (several times ;-) helped a lot, too.

alancoelho

Hi Casper

Iobs_no_scale_pks is calculated using the Rietveld decomposition formula. The error is given by Iobs_no_scale_pks_err. This is calculated without reference to structure factors; in other words these parameters are available for non-structure phases.

Iobs_no_scale_pks to is equal to the square of Fobs. Fobs with a Phase (with a real and imaginary part) for non-centrosymmetric structures is possible if you were to use the Phase information from the a str phase. Is this what you want.

Find out from the Jana manual what you want and I will assist.

cheers
alan

casperwa

Thank you both for the quick replies.

In the attached file, a section of a homemade manual for Jana2006 can be found, made by a postdoc in our group. The weighting and extraction schemes for retrieving Fobs, sigmaFobs, and Fcalc are outlined.

To do apply the same schemes on data from TA I am now performing the calculations in Matlab, after trying and failing to implement it as macros in TA. It should however be possible, and it is my limited programming skills that I blame.

More specifically, what I need is:
Fobs, which seems it should be the same as described by Iobs_no_scale_pks, since it should be the structure factors, without taking any structural details into account, just a peak deconstruction, extracting the structure factor for a Bragg peak. Similar to what I assume can be extracted from a Pawley fit, as correctly suggested by Martin.
Fcalc, as calculated from the structural refinement.
Errors for Fobs (sigmaFobs), calculated from either a Poisson distribution of errors or given esds from diffraction data, combined with a weighting scheme (two such are described in the attached file).

Furthermore, I need to calculate F000 and the corresponding esd. This can be done manually, since it is the sum of electrons in the refined structure.

All of this information is in Jana found in the m80-file:
Fobs: 5th and 6th column.
Fcalc: 7th column.
sigmaFobs: 16th (and 17th - last) column

---

I have looked at Iobs_no_scale_pks and Iobs_no_scale_pks_err in my refinement. The error seems to be a factor 10^3 bigger than the data, when I extract this with phase_out. But should I instead extract these from a structureless phase, similar to when I extract the background? When extracted with xdd_out it give me nothing.

---

Thank you for your generous help.
Casper

alancoelho

Hi Casper

Thanks for the PDF document; first it seems that what you want is what is given except for a weighting contribution; first a quick rundown of what Iobs_no_scale_pks is.

>The error seems to be a factor 10^3 bigger than the data, when I extract this with phase_out.

Iobs_no_scale_pks includes the Rietveld scale parameter; ie.

Io(k) = Sum(Ic(k) Peak(k,i) [Yobs(i) / Yc(i)], i)
= M Get(scale) Iobs_no_scale_pks
Peak(i) = calculated normalized peak
Yc(i) = calculated pattern
Ic(k) = M Get(scale) Fcalc^2
M = multiplicity
Iobs_no_scale_pks_err = Sqrt( Sum( (Sigma( p(j) d_Iobs_no_scale_pks/d_p(i))^2, j))

Where p(j) is the jth parameter that is a function of Iobs_no_scale_pks.

In the document it gives another form of Fobs^2 which includes a weighting. TOPAS does not do that but I can do a modification; what is needed is:

Io(k) = Sum(w(i) Ic(k) Peak(k,i) [Yobs(i) / Yc(i)], i) / Sum(w(i), i)

The PDF went to describe what w(i) is which I don't quite understand; I thought it should be just Sigma(Yobs) which is part of the raw data or if not given then Sqrt(1/Yobs(i)) is used. Also what is b(i), is that background.
Send me an e-mail (alancoelho@bigpond.com) and let me know what version of the program you are using.

cheers
alan