I am inquiring about the meaning of convergence from you size distribution model.

I recently completed some SANS measurements on a low alloy steel and found the size dist model is not able to converge without increasing the weight factor. However the ChiSq value is quite low, almost one, and upon repeated fittings the resultant distributions are very similar, which is encouraging results and a distribution I would expect from these samples.

Increasing the weight factor to 1.5 allows the fit to converge. However I do notice that visually the curve fit worsens even though the ChiSq value is better. This convergence feels artificially forced to me, as my understanding is that this increases the error bars. I also notice this shifts the dominant size distribution slightly larger. Repeating this fit does not give a unique solution.

Would I be able to get an explanation on why the convergence matters here. I understand that the population is located on the boundary of the trust range, this is expected with this material, we are limited by the neutron wavelength in this case. Of course the reported volume fraction is to be taken with some uncertainty.

Any help is appreciated. The file is attached, as is the raw data.

Best regards,

Jacob Barron,