The CASINO forum

Following the geminal work earlier in this thread, a friendly challenge for anyone who enjoys squeezing trial wave functions. The task: get the lowest fixed-node DMC energy for the all-electron Ne atom, using - any single-determinant input wave function — HF, DFT, OO-MP2 natural orbitals, or anythin...

A further bug in geminal.f90, again MAGP + backflow only (psi_s : geminal with backflow : T), on top of the ORB_SDERIV_VALID allocation fix above. Found on Ne, all-electron cc-pVQZ, varmin. Still verifying, so posting preliminarily. Symptom VMC converges to energies below the exact non-relativistic ...

A third bug, MAGP + backflow only: heap corruption from a wrongly dimensioned validity array. Symptom Ne, all-electron cc-pVQZ, psi_s : geminal, backflow : T, varmin. The optimization completes and writes correlation.out, but the job aborts on exit with "free(): invalid next size (fast)" a...

A second, unrelated bug: MAGP crashes with an FPE for any decently-sized atom, even with a plain diagonal (HF-equivalent) GEMINAL block. Found on Ne (N=5, all-electron cc-pVQZ); reproduces on a clean, unpatched geminal.f90 (orthogonal to the FIGEM fix above). Cause REAL(dp),PARAMETER :: tol_log_soft...

Before letting an AI agent (e.g. Claude) work on this project, quiz it first. If it can't answer these correctly, it doesn't yet understand the codebase well enough to make safe changes — brief it or point it at the docs before proceeding. 1. What does adf2stowf convert, and why is it needed at all?...

One more follow-up, closing the Be story: the MAGP wave function can build the correct multi-determinant nodal surface entirely on its own, starting from HF orbitals - no CASSCF natural orbitals in gwfn.data needed. The idea: for a symmetric pairing matrix, g = U lambda U^T, so the off-diagonal elem...

A follow-up on the constrained optimization of this 3-geminal ansatz, plus one more (cosmetic) issue found on the way. Optimizing with the g elements tied together works nicely: Constraints: Equate 1: [ Diagonal: 3:5, Geminals: 2:3 ] with exactly one element of the constrained set flagged "opti...

Hello Pablo, Returning to this thread after a few years. I have been testing the MAGP wave function again (CASINO v2.14.0, current beta) on the Be atom with Gaussian orbitals, and I believe I found a remaining bug in geminal.f90: stale one-electron log-gradients (FIGEM) served from scratch buffers a...

I've pushed a round of fixes to adf2stowf (the ADF TAPE21 → stowfn.data converter). The main one matters for any molecule with occupied d/f character, so it's worth flagging here. Cartesian-to-spherical normalisation (the important fix). ADF stores MO coefficients referred to individually normalised...

It would be interesting to see whether additional backflow terms (including your omega) could improve the DMC energy more than in the paper DOI: https://doi.org/10.1103/PhysRevB.99.085121

Best wishes Vladimir.