ADR suggestion: Improving DiffusionModels

[henrikjacobsenfys]

General

This is a work in progress, initiated by users wanting the following model, tentatively named DeltaLorentz:

$I= K \exp \left( \frac{-\langle u^2 \rangle Q^2}{3} \right)[A_0 \delta(E) + (A_1) L(E, \Gamma)]$,

where $K$ is the scale factor, $\langle u^2 \rangle$ is the mean square displacement, $Q$ is the scattering vector, $A_0$ and $A_1$ are the relative amplitudes of the delta function and Lorentzian, respectively, with the constraint that $A_0+A_1=1$, and $L(E, \Gamma)$ is the Lorentzian function with width $\Gamma$.

$A_0$, $A_1$ and the width of the Lorentzian can be Q-dependent or not. If they are allowed to be different for each Q, a list of Parameters will be created for each Q.

This means the model needs to know and store what Q is, which is a change from how the other models work. This is already an old issue #77. Since we're anyway rewriting the models, we may as well solve #76 as well (i.e. extend the test coverage).

Files that will be impacted

This will require changes in all diffusion models (DiffusionModelBase, BrownianTranslationalDiffusion, JumpDiffusion), as well as some (minor) changes to SampleModel`.

ParameterAnalysis will also need to be updated. Until now, a DiffusionModel generated a single Lorentzian, which ParameterAnalysis could then handle. With the new DeltaLorentz, this will need to be generalized by updating ParameterAnalysis and FitBinding. More advanced models of diffusion, general Q dependence and general coordinate dependence (#177 ) will alsos need to be supported. A minimal working model will be implemented at first.

Storing Q

I will use the same approach as SampleModel uses, which is to only allow setting Q once. This is because changing Q changes the parameters inside which may lead to loss of information if e.g. a fit has been performed

`ComponentCollection's

With DiffusionModels owning Q, they will also store their own ComponentCollections. In this way, they look more and more like ModelBase - perhaps some refactoring can/should be done there, but it will have to wait a bit since I need to get this working asap (there's a school on 12th of June where it needs to work)

Storing the parameters

I'm a bit unsure what the best way to store the parameters is. A list that matches Q is probably the simplest, but is it the best? What should the parameters be named (by Q index or Q value or something else?)

get_all_variables()

This method needs to be extended to take a Q index, which is something SampleModel also already does. The other get_all_... methods may also need to be extended.

Name change

This model not necessarily a model of diffusion. However, what it and all the present and planned DiffusionModels have in common is their Q dependence. Therefore, they would be more aptly named Q_dependent_models or something similar. Ideas are welcome. (This also generalises to later when people may want T_dependent_models or similar, #177 )