Bug Fix: Fock Matrix Autograd Support in scp_mode="fock"

[MauriceDHanisch]

BUG: When using scp_mode="fock" computing the derivative of the Fock matrix with respect to atomic positions failed. This was due to the Fock matrix being disconnected from the computational graph.

FIX: The issue was resolved by triggering one additional SCF cycle, which reconnects the Fock matrix and the computational graph. Similar to #212.

TEST: A regression test was added using torch.autograd.gradcheck, which verifies that the Fock matrix is differentiable with respect to atomic positions.

• The test fails on main, confirming the original bug.
• It passes after the fix.
• Note: The test required a slightly increased atol due to discrepancies observed for some molecules (due to SCF instability or small numerical gradients?).

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.80%. Comparing base (5170e4d) to head (9a63304).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #214      +/-   ##
==========================================
+ Coverage   85.78%   85.80%   +0.01%     
==========================================
  Files         202      202              
  Lines       10160    10160              
  Branches     1150     1150              
==========================================
+ Hits         8716     8718       +2     
+ Misses       1106     1105       -1     
+ Partials      338      337       -1     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.