increasing the number of iterations in PDFGUI

135 views
Skip to first unread message

Patrick Cottingham

unread,
Apr 6, 2022, 7:40:17 PM4/6/22
to diffpy-users
Hi Everyone,
     Thanks for adding me to the group and thanks to the PDFFit2/PDFGUI team for maintaining this incredibly useful software.
 
     I was wondering if there is a simple way to increase the number of iterations that are performed in PDFGUI per click of the "start fit or calculation button".

     My specific use-case is that I have used the Temperature macro to set up a large linked temperature series and the default number of iterations is not enough to ensure convergence at each temperature point.

Thanks,
Patrick

Mikkel Juelsholt

unread,
Apr 7, 2022, 1:38:59 PM4/7/22
to diffpy-users
Hi Patrick

I believe you could run the macro several times after another. Maybe that works. 

However in my experience, if your refinement cannot converge in PDFgui it is usually because your initial model is not good enough. Make sure that you have a good fit over the whole r-range and you have no misbehaving parameters like very high u-values, delta2 or you are refining the atomic positions of light elements. 
If it is because you have very sudden changes in your data it is worth splitting the refinement into several smaller refinements. 

Cheers 

Mikkel

Simon Billinge

unread,
Apr 7, 2022, 1:59:35 PM4/7/22
to diffpy...@googlegroups.com
Thanks for the response Mikkel, I agree with you. 

 I guess one interesting question is if the sample is changing over the temperature range which could cause this.  I recommend uploading the set of data to the SimilariyMapping and/or nmfMapping apps on PDFitc.org to investigate for any such changes.


S

--
You received this message because you are subscribed to the Google Groups "diffpy-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diffpy-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/diffpy-users/a7f4e5a6-6317-4b4e-8332-3a28033e8787n%40googlegroups.com.

Patrick Cottingham

unread,
Apr 7, 2022, 2:46:00 PM4/7/22
to diffpy...@googlegroups.com
Thank you both for the responses.

In fact, I am using delta2 and it is refining to negative values (I assumed this was OK since I have read that delta1 and delta2 "can be treated as empirical refinement parameters").

Is delta2 not an appropriate parameter to use for a linked temperature-series over the range 253K < T < 413K?

Thanks!,
Patrick


You received this message because you are subscribed to a topic in the Google Groups "diffpy-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/diffpy-users/Aobl4w8FqzE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to diffpy-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/diffpy-users/CANugsUGgyF7%2BQMvX-PVAP%2BsEFMrAi1oJcA5z5mYVOYZYYN70ig%40mail.gmail.com.

Mikkel Juelsholt

unread,
Apr 9, 2022, 9:40:59 AM4/9/22
to diffpy-users
Hi Patrick 

That sounds like it coukd be the issue. Delta2 is fine for that temperature range 
It usually does not make sense to extract any quantitative/real information from the delta2 parameter. However it still have physical meaning and impact on your fit. So the delta2 parameter describes correlated motion which sharpens the PDF peaks at low r. It does this by subtracting a factor of delta2/r^2 from the "average" peak width. Therefore delta2 should always be larger than 0. If your model makes delta2 negative it is trying to broaden your peaks at low r. This broadening must come from something. What it is difficult to say without knowing more about your data. 
I think the first step is to get a positive delta2 and see if this fixes the issue.

Cheers Mikkel 
Reply all
Reply to author
Forward
0 new messages