In diffpy-cmi, there is no sratio and rcut functions available.
Sigma_ij = sqrt( sigma_isigma_j + f(delta1) + f(delta2) )
Sigma_ij is the width of the peak jointing atom i and atom j. Sigma_i is the adp of atom i.
If r_ij < rcut:
sigma_ij -> sigma_ijsratio
Ex: if rcut = 4.0, peak 3.95 A then we expand it via sratio so it gets to 4.1A.
Hi @sbillinge , I dig into our previous email threads, and I found this issue. I think CMI does not support sratio and rcut. You and I discussed some temporary solution to wrap it at the pdfgui end, instead of touching the cmi source codes. wrote down some temporary wrap up function.
Hi @sbillinge , I dig into our previous email threads, and I found this issue. I think CMI does not support sratio and rcut. You and I discussed some temporary solution to wrap it at the pdfgui end, instead of touching the cmi source codes. wrote down some temporary wrap up function.
Thanks @dragonyanglong . Let's look at this this summer. I will have some people helping but they will need some mentoring.
the rcut/sratio is kind of brutal. I wonder if we can kind of bootstrap it by changing the rcut into something like a Fermi function that crosses over rapidly at some point in r and simply multiply this by sratio. We could have it go over to 1 at low r and zero at high-r and then maybe add this to our sigma model as an additional term. Could something like that work? Anyway, something along those lines.
Activity
dragonyanglong commented on May 13, 2025
Hi @sbillinge , I dig into our previous email threads, and I found this issue. I think CMI does not support sratio and rcut. You and I discussed some temporary solution to wrap it at the pdfgui end, instead of touching the cmi source codes. wrote down some temporary wrap up function.
sbillinge commented on May 13, 2025
Thanks @dragonyanglong . Let's look at this this summer. I will have some people helping but they will need some mentoring.
sbillinge commented on May 13, 2025
A general thought from @sbillinge:
the rcut/sratio is kind of brutal. I wonder if we can kind of bootstrap it by changing the rcut into something like a Fermi function that crosses over rapidly at some point in r and simply multiply this by sratio. We could have it go over to 1 at low r and zero at high-r and then maybe add this to our sigma model as an additional term. Could something like that work? Anyway, something along those lines.