Fitting XRD data for thin films and superlattices

[JET-Sci]

It would be helpful if the kinematical diffraction model could be used to do a rough fit of formatted XRD data for thin films and superlattices in the same way that data for X-ray reflection can be fit. Being able to choose which factors are varied would be helpful as well.

[dkriegner]

I agree that such functionality would be helpful.

A quick fix for one particular model would be to copy the fit_xrr function for the KinematicalModel. However, in the long run this should be possible not only for the kinematical but also for the other diffraction models.

As a note to myself: This is most likely best implemented in a way that all the Models become lmfit models and somehow return a list of their parameters which can then be fitted by a generic wrapper around some lmfit minimize function. The layer stack could also be such a model which has its own parameters which can be easily combined with the ones of the model in a CompositeModel. In this way even the XRR could be fitted using this generic approach.

[dkriegner]

As a note to myself: This is most likely best implemented in a way that all the Models become lmfit models and somehow return a list of their parameters which can then be fitted by a generic wrapper around some lmfit minimize function. The layer stack could also be such a model which has its own parameters which can be easily combined with the ones of the model in a CompositeModel. In this way even the XRR could be fitted using this generic approach.

Unfortunately I was a bit naive with my comment. CompositeModel is not appropriate for this use-case since the parameters of the LayerStack are needed for the XRR/XRD models and can't be separated out.

I can imagine a solution by dynamic generation of the model function used for lmfit.Model, but this seems not a very clean solution.

[dkriegner]

The mentioned commits potentially fix this issue.
Testing and documentation updates are still needed. The issue remains open until this is fixed.

[dkriegner]

Missing things which need some more work for XRD models to fit using the new approach are:

• adding structural parameters (which influence the model) to the list of fit parameters
• select appropriate fit parameters based on the model
• rethink how to treat lattice parameter changes for Pseudomorphic stacks: I am afraid one will have to ignore this and fit only the lattice parameters of individual layers independently which might, however, not be what every user naively expects... This is more complex than I thought. I reported a separate issue Simulation of Pseudomorphic stacks (relaxation) #65
• documentation updates: including an example script

[dkriegner]

I consider the main points of this issue fixed. If problems with the implementation arise or specific improvements are needed (see for example issue #65) a separate issue should be opened