sayred1
commented
4 years ago @qzhu2017 @sabaronett, I've just added another profiling function to profile.py, the original pseudo-voigt function. This required me to rename the previous pseudo-voigt function to "mod_pseudo-voigt", let me know if you think of a more compact name for this modified version.
I've also updated the profiling notebook. Since the pseudo voigt function is the combination of the gaussian and lorentzian profiles the user may specify either one or two FWHM, corresponding to each profile. By default, one FWHM is specified.
To differentiate the two options, here's an example: user_kwargs = {'FWHM-G': 0.02, 'FWHM-L': 0.03} or user_kwargs = {'FWHM':0.05}
@qzhu2017 I've added numbt.jit to both the gaussian and lorentzian functions, to speed up the original P-V function. Here are the effects:
- gaussian: 22.7 ms --> 1.08 ms
- lorentzian: 897 µs --> 670 µs
- original P-V : 24.1 ms --> 1.71 ms
Deadline 2/19
Additional Profiling
Structure Factor
Notes
~The possible number of parameters will depend on the user. If the user wants, they may specify a FWHM that is different for the composite G and L functions, along with the eta parameter. Else, the user can specify a FWHM that is the same for G and L, along with eta.~
~The two PV functions must have different names.~
~Ignore Debye-Waller corrections.~
The dispersion corrections ( f' and f'') are dependent on the radiation. (An example of Li is below)
VESTA uses scattering factors that aren't tabulated in IUCr. Per a meeting with Keith, either gut GSAS (a more commonly used refinement software) and use their parameters, or look into literature for updated parameters with 11 scattering terms.