Feature matrix

Molecular

Method	Driver	Open-shell	Gradient	MP2	Validated vs PySCF
RHF	run_rhf	—	✓	✓ (run_mp2)	✓
UHF	run_uhf	✓	✓	✓ (run_ump2)	✓
RKS	run_rks	—	✓	—	✓
UKS	run_uks	✓	✓	—	✓
ROHF	—	—	—	—	(roadmap)

Any functional that libxc supports is available through RKS and UKS (over 500, including LDA, GGAs, hybrids, meta-GGAs). HF analytic gradients and DFT analytic gradients are both available for LDA, GGA, and hybrid functionals.

Periodic

Method	Driver	Dim 1	Dim 2	Dim 3	Notes
RHF, molecular limit	run_rhf_periodic_gamma	✓	✓	✓	large unit cell only
RHF, multi-k	run_rhf_periodic	✓	✓	⚠	3D bulk Coulomb is cutoff-dependent
RKS, multi-k	run_rks_periodic	✓	✓	⚠	pure LDA/GGA validated in molecular limit
Hybrid DFT	run_rks_periodic	wired	wired	⚠	not yet benchmarked
ROHF / UHF / UKS periodic	—	—	—	—	roadmap

⚠ = cutoff-dependent results until the Ewald + Saunders-Dovesi splitting lands in Phase 12e. Molecular-limit tests pass to machine precision in all three dimensions for both HF and DFT.

Ewald summation (Phase 12e, in progress)

See user_guide/ewald.md for the math and API.

Sub-phase	Scope	Status
12e-a	Classical Ewald for the point-charge nuclear lattice sum. CoulombMethod.EWALD_3D on nuclear_repulsion_per_cell.	✅ shipped
12e-b	erfc-screened nuclear attraction compute_nuclear_erfc_lattice (Ewald building block; no SCF dispatch yet).	✅ shipped
12e-c	Reciprocal-space parts of V(g) and ERI Coulomb J, Saunders-Dovesi multipolar splitting, FFTW3 dep. Full EWALD_3D end-to-end.	pending

Literature-Madelung-constant validations (tolerance from the implementation; source uncertainty in parens):

Crystal	M reproduced	Reference
NaCl (rocksalt, ±1 charges)	1.7475645946 to 1e-8	1.7475645946…
CsCl	1.76267477 to 1e-6	1.76267477 (±1e-8)
ZnS (zincblende)	1.63805 to 1e-4	1.63805 (±1e-5)
Simple-cubic jellium	1.4186487 to 1e-6	1.4186487 (Nijboer-De Wette 1957)

Input / output

Capability	API	Notes
High-level “run-a-job” driver	run_job	dispatches to the right SCF driver, writes .out + .molden, runs BFGS if asked
Formatted text log	format_scf_trace, log_scf_trace	banner, iteration trace, energy components (DFT), orbital table, HOMO-LUMO gap
Molden file export	write_molden	RHF / UHF / RKS / UKS; verified by PySCF round-trip
Geometry trajectory	run_job(..., optimize=True)	ASE .traj file, viewable with ase gui
XYZ load	Molecule.from_xyz	picks up charge / multiplicity
POSCAR load / save	read_poscar, write_poscar	VASP 5 format

Basis sets

• Standard — 90 sets from libint, including STO-nG, 6-31G**, cc-pVXZ, def2-TZVP, ANO-RCC, and the Karlsruhe core-valence family.

• Solid-state — pob-TZVP, pob-TZVP-rev2, pob-DZVP-rev2 (H–Br) from the Peintinger-Vilela-Oliveira-Bredow family, designed to avoid small-exponent linear dependencies in crystals.

• CRYSTAL-format parser — imports arbitrary CRYSTAL per-element basis files, converts to libint-compatible .g94.

• ECP basis sets — parsed but not yet applied; requires libecpint integration.

Crystal / lattice infrastructure

• spglib integration: Crystal, analyze_symmetry, to_primitive, irreducible_kpoints.

• POSCAR I/O (VASP 5 format).

• Monkhorst-Pack k-mesh generation with IBZ reduction.

Validation

Test class	Count	Level
Molecular 1e integrals vs PySCF	24	machine precision
RHF / UHF / RKS / UKS energies vs PySCF	59	machine precision (HF), grid-accuracy (DFT)
MP2 / UMP2 vs PySCF	21	1e-8 Ha
Gradients (HF, DFT) vs finite-difference	17	1e-6 Ha/bohr
Periodic machinery (invariants, molecular limit, Bloch folding)	76	machine precision
Crystal / POSCAR / basis parsing	18	—

Full suite: ~280 tests, runs in about a minute.