Seems like there is a difference in the energies depending on the use of intel or GNU compilers with the CPCM model. It seems to happen only when casting string to double.

Seems like there is a difference in the energies depending on the use of intel or GNU compilers with the CPCM model. It seems to happen only when casting string to double.

I compiled a fresh version and got a different value for CrCp2 with CPCM and an epsilon value of 7.0, which is now consistent with the results I got from the C-API test. Would be happy to have somebody else checking these values.

Codecov Report

Attention: 35 lines in your changes are missing coverage. Please review.

Comparison is base (3191316) 71.39% compared to head (f70cec5) 71.51%.

@@            Coverage Diff             @@
##             main     #130      +/-   ##
==========================================
+ Coverage   71.39%   71.51%   +0.11%     
==========================================
  Files         134      134              
  Lines       16654    16759     +105     
  Branches     5081     5121      +40     
==========================================
+ Hits        11890    11985      +95     
+ Misses       2514     2484      -30     
- Partials     2250     2290      +40     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Looks good! I only have two small things to add:

1. We could add the newly available interactions to the docstring of the add function in the Calculator class:
   

   tblite/python/tblite/interface.py

   Lines 510 to 519 in 0c28978

   	"""
   	Add an interaction to the calculator instance. Supported interactions are
   	=================== =========================== ===================
   	name description Arguments
   	=================== =========================== ===================
   	electric-field Uniform electric field Field vector (3,)
   	spin-polarization Spin polarization Scaling factor
   	=================== =========================== ===================
   	"""

2. and I found two small typos (cpcm instead of alpb):
   

   tblite/src/tblite/api/container.f90

   Line 250 in 0c28978

   if (debug) print '("[Info]", 1x, a)', "new_cpcm_solvation"

   
   and
   

   tblite/src/tblite/api/container.f90

   Line 292 in 0c28978

   if (debug) print '("[Info]", 1x, a)', "new_cpcm_solvation float input"

I compared the results of the Python API with the command line for Intel compilers with CrCp2 and the results nicely matched!

The GBSA solvation model could also be added in the future but first the full ALPB and GBSA should be implemented I think (see #138)

I changed the suggested lines, thanks for catching the typos.
Once you give approval, I will merge and close the PR.