Open mesonepigreco opened 2 years ago
Hi Lorenzo, due to large size of d3 tensor file, I failed to attach here, rather I sent them in email. Pls use those files to reproduce the error message.
Regards, Prasad
The issue seems not related to the d3 part, but to the interpolation using support dynamical matrices. I created a new branch of cellconstructor which fixes this error. https://github.com/SSCHAcode/CellConstructor/pull/48
However, the code seems not to be able to reproduce the same output as the previous interpolation and the testsuite fails (up to today), therefore more testing is needed.
Similar behaviour is observed at little lower temperature (T=20K) as well:
Error in routine frc_blk (1): wrong total_weight
Regards, Prasad
Thanks for the report. I'm working on it. It seems a problem with the interpolation using a support matrix. We will fix this soon, however, if you want to proceed with the calculation, you can try to interpolate without the support matrix and use the ForcePositiveDefinite after the interpolation and before passing it to the get_static_correction_along_path (as you are doing), it should avoid errors in the bubble calculation due to imaginary frequencies.
dyn.ForcePositiveDefinite() before passing it to the get_static_correction_along_path() didn't help: q= [ 2.76026848e-03 2.76026848e-03 -1.06370064e-19] (2pi/A) w(q)= [ -4.03429147 -3.46423257 8.83285856 44.34200573 47.2531449 110.37484302 111.83469055 136.15802509 140.4898301 140.75631651 146.07548221 171.62276582 173.40597864 175.80646154 249.52104453 249.74820731 261.18240083 425.39532056 427.77205168 428.40073337 429.5592279 432.71955735 433.20384253 455.8704559 488.9355318 497.85667105 532.92236704 532.94816711 772.08949798 825.88642211] (cm-1) Cannot continue with SSCHA negative frequencies
I thought by doing so it only fixes at high symmetry points.
Discussed in https://github.com/SSCHAcode/python-sscha/discussions/74