Commonly encountered issues

From the Maud website (http://maud.radiographema.com/): MAUD = Materials Analysis Using Diffraction - A Rietveld extended program to perform the combined analysis. It can be used to fit diffraction, fluorescence and reflectivity data using X-ray, neutron, TOF or electrons.

We love the logo, we want to use the logo here - we want Maud people to come join, be moderators, and share in our love for x-rays, diffraction and materials.
User avatar
MortenBNielsen
Sealed Tube
Posts: 12
Joined: 11 Sep 2015, 10:52

Commonly encountered issues

Post by MortenBNielsen » 30 Oct 2015, 13:35

Below follows a list of common problems that one can have in MAUD and how to fix them. The list will be expanded as Q&A's appear in this forum or I run into issues myself. Each problem/answer is posted in a separate post so I can link to them easier.

User avatar
MortenBNielsen
Sealed Tube
Posts: 12
Joined: 11 Sep 2015, 10:52

Re: Commonly encountered issues

Post by MortenBNielsen » 30 Oct 2015, 13:42

Problem: Cholesky negative diagonal
Sometimes after the refinement a message “Cholesky negative diagonal” appears in the MAUD output panel. This means the problem that MAUD was trying to solve was ill-conditioned, and some of the parameters are not refinable at the given conditions. The first of the three displayed numbers refers to the “wrong” parameter. You can view it on the MAUD output file (with .lst extension) or you may review the parameters list window and find a refined parameter with value “-1” in the “Error” column. After determining the parameter you should adjust it manually, fix it and refrain from refining it.

User avatar
MortenBNielsen
Sealed Tube
Posts: 12
Joined: 11 Sep 2015, 10:52

Re: Commonly encountered issues

Post by MortenBNielsen » 30 Oct 2015, 13:44

Problem: The refinement claims it is converged, but it looks like it isn't
When you have complex data, such a multi-phase powders with many peaks, the relative changes to the goodness of fit for each individual parameter can become so small that it triggers the program's convergence criterium. The first fix is to go to Analysis -> Options -> Algorithm options -> Enable "Use numerical double derivative". This slows down each cycle (because more numbers are being calculated), but enables the program to realise that even though things are changing slowly, it still has not converged.

Return to “Maud”