Fitting an amorphous structure model in PDFgui

[Craig Jolley]

Hi...

I'm trying to use PDFgui to analyze data that I've collected on a largely amorphous coordination polymer material.  I've managed to generate a fairly realistic structural model and it agrees well with the experimental PDF.  The part that is bothering me is that, while the peaks are in more or less the right position, the valleys between them are much too shallow in the calculated PDF, relative to the experimental one.  My structural model was derived from finite-temperature molecular dynamics simulations, so some of the bond lengths and angles won't be quite "ideal" in the sense that structural parameters derived from X-ray crystallography often are -- my model incorporates a fair amount of static disorder.  Not surprisingly, the temperature factors for this model converge to much lower values than I'm used to seeing, because a fair amount of disorder is also present.

Is the failure to fit the valleys correctly somehow related to the two sources of peak broadening (static disorder in the structural model and temperature factor-based convolution)?  I've also seen some cases where varying temperature factors for calculated PDFs can alter the positions of peak maxima, which makes me even more concerned about modeling disorder correctly, since those peak positions are a large part of how I'm evaluating the success of my structural model.  Is there a better way to do this?

Thanks,

--craig

--
Craig Jolley
Postdoctoral Research Associate
Astrobiology Biogeocatalysis Research Center
Montana State University
http://capsid.msu.montana.edu/cjolley

[chris farrow]

Hi Craig,

Sorry to have this question go unanswered so long. I'm not quite sure why the valleys in your model PDF are so shallow. It sounds like the refinement is adjusting the global scale factor to fit the peak amplitudes, but is missing the PDF baseline in the process. It is possible that  your MD model may be too "crystalline". In other words, the actual material may have a much shorter structural coherence length than your model. Try refining the spdiameter parameter - this will simulate this effect and give an estimate of the coherence length in your material. This method has been used in the Billinge group to study amorphous pharmaceuticals: 

Simon J. L. Billinge, Timur Dykhne, Pavol Juhás, Emil Božin, Ryan Taylor, Alastair J. Florence and Kenneth Shankland, Characterisation of amorphous and nanocrystalline molecular materials by total scattering, CrystEngComm 12, 1366-1368 (2010)

DOI: 10.1039/b915453a

There are undoubtibly other examples out there, but this is the one that comes to mind.

Good luck. Please let us know how it turns out.

Chris

> --
> You received this message because you are subscribed to the Google Groups
> "diffpy-users" group.
> To post to this group, send email to diffpy...@googlegroups.com.
> To unsubscribe from this group, send email to
> diffpy-users...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/diffpy-users?hl=en.
>

[Simon Billinge]

Hi Craig,

This is interesting; it is usually the other way around: valleys in
the model are deeper than those in the data. I think Chris's
suggestion may not help as it corrects that more common opposite
problem. Basically it is telling you that your calculated model is
too disordered. Without seeing a plot of the fit, I am hard put to
say more.

S

--
Prof. Simon Billinge
Applied Physics & Applied Mathematics
Columbia University
500 West 120th Street
Room 200 Mudd, MC 4701
New York, NY 10027
Tel: (212)-854-2918 (o) 851-7428 (lab)

Condensed Matter Physics and Materials Science Dept.
Brookhaven National Laboratory
P.O. Box 5000
Upton, NY 11973-5000
(631)-344-5661

email: sb2896 at columbia dot edu
home: http://nirt.pa.msu.edu/