EVCont v0.2: Excited states + NAMD + low-rank

[KAtalar]

Major developments for the ab initio eigenvector continuation codebase, mainly around extensions to excited states, nonadiabatic molecular dynamics simulations and low-rank compression of reduced density matrices for memory efficient continuation.

• Excited states with gradients and nonadiabatic coupling vectors
• (Transition) dipole moments and oscillator strengths from continuation
• Newton-X interface for nonadiabatic MD simulations
• Low-rank compression of 2RDMs
• Inference of the subspace Hamiltonian and gradients in the low-rank form
• Updated active learning scheme + HPC support
• Quantel interface for SS-CASSCF within CAS class

[Copilot]

Pull request overview

This PR ("EVCont v0.2") is a major feature release for the evcont ab-initio eigenvector continuation codebase. It extends the framework from ground-state energies/gradients to excited states (with gradients, NACs, and (transition) dipole moments), adds a Newton-X driver for nonadiabatic molecular dynamics, introduces a low-rank compression scheme for 2RDMs (with associated fast inference of the subspace Hamiltonian and gradients), and adds new solver interfaces (Quantel SS-CASSCF) plus an updated active-learning workflow with HPC support. The code touches the core continuation modules, adds many example/benchmark scripts, and ships a Newton-X interface directory.

Changes:

• Excited-state machinery: multistate energies, gradients, NACs, and (transition) dipole/oscillator strengths added across evcont/ab_initio_gradients_loewdin.py, evcont/ab_initio_eigenvector_continuation.py, and the FCI/CAS containers.
• Low-rank 2RDM compression and fast subspace-Hamiltonian/gradient inference (new evcont/low_rank_utils.py, new low-rank paths in solver objects, and benchmarking scripts under scripts/low_rank/ and examples/low_rank/).
• New Newton-X NAMD driver (nx-interface/), Quantel SS-CASSCF interface, updated active-learning + HPC scripts, and many example/benchmark scripts (NAC, dipole, low-rank).

Reviewed changes

Copilot reviewed 41 out of 44 changed files in this pull request and generated 20 comments.

Show a summary per file

File	Description
evcont/ab_initio_eigenvector_continuation.py	New multistate / low-rank approximation routines; contains a duplicate approximate_multistate_otf_OAO definition.
evcont/ab_initio_gradients_loewdin.py	Adds multistate energy/grad/NAC and low-rank gradient inference paths.
evcont/FCI_EVCont.py, evcont/CASCI_EVCont.py	Solver containers extended with nroots, lowrank, vectorized low-rank storage, dipole helpers.
evcont/low_rank_utils.py	New module for low-rank decomposition, vectorization, and reconstruction utilities.
evcont/NAMD_utils.py	NAMD/active-learning helpers, including HPC submission scripts.
nx-interface/run-evcont-driver.py	New Newton-X driver; contains a missing.append('ncas') copy-paste bug for missing nelec.
nx-interface/{geom,grad,grad.all,nad_vectors,oldh,epot}	Runtime artifacts that look accidentally committed.
examples/solvers/02-casscf_continuation.py	SA-CASSCF example with docstring/banner copied from the CASCI example.
examples/low_rank/02-cas_cont_lowrank.py	Uses leaked training mol instead of test_mol for nao**2 print.
examples/low_rank/01,03,04 + examples/excited_states/* + examples/dipole/*	New end-to-end usage examples for low-rank, excited-state, and dipole functionality.
scripts/NAC/H4/{run_NAC_molcas.py,compare.py}	OpenMOLCAS reference NAC generation + plotting/benchmark script; minor variable shadowing in run_NAC_molcas.py.
scripts/low_rank/hydrogen_chain.py & related	Extensive low-rank benchmarking script.
.gitignore	Adds broad *.sh/*.npy patterns that may hide intended files.
README.md	Updated with v0.2 description, citation, and new usage examples.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.