Conversion to Cartesian basis sets #259

Activity

1. ARYAMANN7279 commented on Oct 23, 2026

   

   ARYAMANN7279

   on Oct 23, 2026

   More actions

   Hi, I'd like to work on this, as a follow up to #409. Here is the plan, tell me if you want it different.

   1. convert_to_cartesian(obasis) in convert.py, next to convert_to_segmented. It returns the new MolecularBasis (same shells, all kinds 'c') and the transformation matrix, so that coeffs_new = T @ coeffs_old. It builds T from tfs and takes the conventions of the old and new basis into account (ordering and signs), so sign flipped pure functions like the ORCA f ones work too.
   2. In a second PR, a prepare_cartesian in prepare.py that applies T to the orbital coefficients and density matrices, for the WFN and WFX dumpers.
   3. The main test: the overlap of the old basis must equal T.T @ S_cart @ T, with S_cart from compute_overlap on the new basis.

   One question: should T be a dense array or a scipy.sparse matrix? Dense is simpler, sparse scales better for big basis sets. Unless you prefer dense, I would return a sparse matrix.
2. PaulWAyers

   assigned

   PaulWAyers

   and

   ARYAMANN7279

   and unassigned

   PaulWAyers

   on Oct 23, 2026