Dear Armel:
Although I sent the message unintentionally to the whole list, why not have
a public discussion.
> including Gaussian. But the size model is currently only
> Lorentzian in ARIT. There is no way to have a pure Gaussian
> shape for the size effect since it corresponds to unphysical
> size distribution function (negative proportions of
> cell columns for some defined lengths...). But if you
To be consistent: As you know, the Lorentzian function is used in the
size-broadening term only as an approximation (generally with larger
discrepancies for larger Fourier coefficients) for the Laue function, with
the latter being a "more correct" function from a theoretical point of view.
That is the whole point -- one always has to find a reasonable compromise
between "correctness" and "usability". Very often, a simple assumption that
size-broadened profile has both Lorentzian and Gaussian terms yields
satisfactory results. Of course, that doesn't mean that it is applicable to
any specimen with arbitrary accuracy, but until we have a "perfect" solution
that all of us desire, perhaps it is still a reasonable approximation.
> profile shape), Poisson, etc. And you have to consider the
> general case : anisotropic, contrarily to the CeO2 sample.
The main intent of this round robin was to try different methods on a simple
case, where it is still possible to compare these simplified models to more
model-independent approaches, such as the WA. Besides, if some model doesn't
work in simple cases, what can you expect for complicated samples?
Best regards, Davor
