Feature matrix

The capability matrix for the source revision used to build these docs. For when each feature shipped, see CHANGELOG.md. For what’s coming next, see roadmap.md.

Molecular SCF

Initial guesses share a route-aware AUTO policy: ordinary restricted closed-shell routes use PATOM; molecular unrestricted/ open-shell routes (including singlet UHF/UKS) and transition-metal systems usually use SAD. Isolated spin-polarized atoms use PATOM. Explicit guesses are capability-checked. Supported READ routes preserve complete spin/Bloch restart state and distinguish the physical construction from internal READ transport. Neither convergence nor AUTO alone establishes the global ground state.

Method	Driver	Open-shell	Gradient	Hessian (FD)	Hessian (analytic)	MP2	Validated vs PySCF
RHF	run_rhf	-	✓	✓ (compute_hessian_fd)	✓ (compute_hessian_rhf_analytic)	✓ (run_mp2)	✓ machine precision
UHF	run_uhf	✓	✓	✓	✓ (compute_hessian_uhf_analytic)	✓ (run_ump2)	✓ machine precision
RKS	run_rks	-	✓	✓	✓ (compute_hessian_rks_analytic)	-	✓ grid-accuracy
UKS	run_uks	✓	✓	✓	✓ (compute_hessian_uks_analytic)	-	✓ grid-accuracy
ROHF	run_rohf	✓ (spin-pure)	✓ analytic	✓ (freqs)	(roadmap)	✓ (run_rohf_mp2, semicanonical)	PySCF-parity tests¹
ROKS	run_roks	✓ (spin-pure)	✓ (FD, opt)	(roadmap M5b)	(roadmap M5b)	-	PySCF-parity tests¹

¹ ROHF/ROKS: the Roothaan-coupling + energy math is verified directly (closed-shell→RHF/RKS reduction, SCF stationarity, and ROKS fractional degenerate frontier shells such as OH(2Pi)); the end-to-end PySCF-parity and ROHF-on-singlet==RHF tests are written and run in the full dev-box suite (tests/test_rohf.py, tests/test_roks.py).

Mean-field functionals libxc supports work through RKS / UKS / ROKS (500+ functionals including LDA, GGAs, hybrid GGAs, the τ-dependent meta-GGA family (TPSS, TPSSh, M06-L, M06-2X, SCAN, r²SCAN, r²SCAN01), and the range-separated hybrids ωB97X, ωB97X-D, and the VV10-paired ωB97X-V / ωB97M-V; direct double-hybrid ROKS calls still route through the double-hybrid dispatcher). HF and DFT analytic gradients are available for LDA, GGA, hybrid-GGA, and the τ-dependent meta-GGA family (modulo the f-shell gradient bug on open-shell UHF / UKS). Meta-GGA analytic Hessians and range-separated-hybrid analytic gradients are still queued. Range-separated hybrids run via direct SCF (not yet density-fitting).

Functionals without a complete analytic gradient (GitLab #571). The molecular RKS / UKS gradient kernel differentiates exact exchange through the global fraction only; it omits the long-range erf-attenuated exchange term of every range-separated hybrid (HSE06, ωB97X, ωB97X-D, CAM-B3LYP, ωB97X-V, ωB97M-V, …) and the VV10 nonlocal correlation gradient (VV10, ωB97X-V, ωB97M-V). Measured on a bent O/H/H molecule in STO-3G the partial analytic gradient sits 5.6e-3 (HSE06), 7.0e-4 (VV10) and 5.6e-2 (ωB97X-V) Ha/bohr off the full-energy finite-difference gradient, against a 2.1e-6 noise floor for PBE. Since this fix vibeqc.compute_gradient_rks / compute_gradient_uks refuse such a functional with a NotImplementedError naming the missing terms (pass allow_incomplete=True to receive the partial gradient knowingly), vibeqc.functional_gradient_terms_missing(name) reports them, and every molecular geometry optimizer (run_job(optimize=True) on the ASE and native backends, optimize_molecule, optimize_molecule_brent, the geomopt provider, the VibeQC ASE calculator) falls back to full-energy central finite differences (step 0.005 bohr) and says so in the .out. Routes that finite-difference the analytic gradient (the FD Hessian / frequencies, the analytic-Hessian skeletons, CPCM gradients) and NEB fail closed with the same message. The analytic terms themselves are a separate feature.

Wavefunction methods (v0.9.0-v0.12.0)

Non-mean-field electronic-structure methods that do not conceptually rely on a Hartree-Fock reference. All reachable through run_job(method=…) and as a standalone vibeqc.solvers import. See user_guide/non_hf_solvers.md and the non_hf_solvers tutorial.

Method	run_job(method=…)	Solver entry point	Notes
Full CI	"fci"	(dense eigh on the full determinant Hamiltonian)	Exact within the (active) orbital space; reference for the approximate solvers. Small active spaces only.
Selected CI (CIPSI)	"selected_ci"	solve_selected_ci	Iterative perturbative selection; target_size, pt2_threshold, optional PT2 correction. Spin-unrestricted determinant basis by default (#107, 2026-09-03); spin_restricted=True selects the closed-shell seniority-zero (DOCI) basis, coupled through the pair-move integral ⟨ii|aa⟩ (#639), which cannot represent open-shell configurations.
DMRG	"dmrg"	solve_dmrg	Experimental (the source calls it a toy). Two-site sweep, MPS wavefunction, bond-dimension schedule. Minimal Python implementation; refuses more than 6 spatial orbitals (12 spin orbitals).
Variational 2-RDM	"v2rdm"	solve_v2rdm	Exact small-active-space N-representable 1/2-RDM from the in-tree determinant Hamiltonian; Q/G SDP feedback remains unimplemented.
Transcorrelated CI	"transcorrelated_ci"	build_transcorrelated_hamiltonian + solve_selected_ci	Preview. Similarity-transformed Hamiltonian (Gaussian / Jastrow correlator, research grade) fed to selected CI; same determinant-basis default as selected CI (#107).

• Active-space truncation: pass active_space=(n_orbitals, n_electrons) to run_job to restrict any solver to a frozen-core / CAS-style window. The runner projects the full MO-basis Hamiltonian onto the active block via Hamiltonian.active_space, which applies the standard CAS partition (lowest (nelec − n_elec)/2 orbitals as a doubly-occupied inactive core) with the proper frozen-core dressing, the effective one-electron term h̃_pq = h_pq + Σ_c (2⟨pc|qc⟩ − ⟨pc|cq⟩) and the constant E_core offset folded into the nuclear-repulsion term, so the reported energy is the full CAS total energy. The dressing is shared with the casci path, so fci/selected_ci/dmrg/v2rdm on a given active space agree with casci on the same window to numerical precision (E_FCI ≤ E_CAS ≤ E_HF). See python/vibeqc/solvers/ACTIVE_SPACE.md for the contract and tests/test_solvers_active_space_api.py for the variational-sandwich + cross-method coverage.

• Orbital source is decoupled, the solvers operate on an abstract Hamiltonian in an orthonormal MO basis; get_hf_orbital_provider is a convenience, not a requirement.

• Geometry optimisation, wavefunction methods optimise via numerical gradients (run_job(method="fci", optimize=True)); analytic gradients are roadmap.

• Parity, He / OH FCI parity tests and DMRG-vs-FCI cross-checks ship in tests/test_solvers_*.

Post-HF correlation methods

Correlated wavefunction methods that build on a Hartree-Fock reference. All reachable through run_job(method=…).

Method	run_job(method=…)	Reference	Notes
CISD	"cisd" or method="ci", citype="cisd"	HF orbitals	Fixed-space variational singles+doubles CI via vibeqc.cisd; CISDOptions controls roots and determinant-space guards.
CC2	"cc2" or method="ci", citype="cc2"	RHF	Ground-state second-order approximate coupled cluster. The singles projection is complete and the doubles equation retains the T1-transformed Hamiltonian plus [F,T2]; DF and exact-integral routes are supported. Closed-shell energies only; no perturbative triples. Validated against a spin-orbital oracle and ORCA AUTOCI-CC2.
CCSD	"ccsd"	RHF / UHF / ROHF	Canonical dense CCSD, density-fitted by default or on exact four-index integrals (CCSDOptions(density_fit=False), conventional cross-code parity route); closed-shell and open-shell (UCCSD on a UHF reference, ROHF-reference via the spin-orbital kernel), auto-dispatched from the SCF reference. Also run_ccsd with CCSDOptions.
CC3	"cc3" or method="ci", citype="cc3"	RHF	Ground-state iterative CC3 with converged singles/doubles and approximate connected triples regenerated at every iteration. Exact MO integrals and frozen cores are supported. Energy-only and benchmark-scale because the dense determinant representation grows combinatorially. Validated against out-of-process ORCA 6.1 UHF AUTOCI-CC3 on closed-shell references.
CCSDT	"ccsdt" or method="ci", citype="ccsdt"	RHF / ROHF	Full iterative singles, doubles, and triples from the exact determinant-space similarity projection. Exact MO integrals, frozen cores, and common-orbital open shells are supported. Energy-only and benchmark-scale because the dense determinant representation grows combinatorially.
CCSD(T)	"ccsd(t)"	RHF / UHF / ROHF	Dense small-molecule accuracy reference. Closed-shell DF triples support requested-memory fast/blocked/direct/disk execution; open-shell triples support fast/direct but retain the dense spin-orbital integral floor. The dense CC iteration and the FNO setup remain production-size memory gates. Validated vs PySCF (uHa; conventional route to a few nHa) with ORCA 6.1 cross-checks.
A-CCSD(T)	method="ccsd", triples="A-CCSD(T)"	RHF	Closed-shell asymmetric/Lambda triples correction on CCSD amplitudes. This separate implementation remains dense and rejects bounded triples modes. The effective method and citation route are a-ccsd(t), with Kucharski-Bartlett 1998 and Crawford-Stanton 1998 references surfaced in output.
QCISD / QCISD(T)	"qcisd", "qcisd(t)" (also via citype=)	RHF	Closed-shell quadratic CI singles+doubles via exact monomial selection from the canonical CCSD residual. QCISD(T) uses the original Pople-Head-Gordon-Raghavachari E[T] + 2*E_ST triples convention on QCISD amplitudes. Pinned vs the in-repo spin-orbital oracle and out-of-process ORCA 6.1 RI-QCISD(T).
CCD / LCCD / LCCSD / CEPA(0..3)	"ccd", "lccd", "lccsd", "cepa(0)".."cepa(3)" (also via citype=)	RHF	Coupled-pair variants of the closed-shell DF-CCSD kernel (CCSDOptions field cc_variant): CCD freezes singles; LCCD/LCCSD linearize the amplitude equations (exact polynomial-stencil linearization of the canonical residual); CEPA(1..3) add Meyer’s pair-specific EPV shifts. CEPA validated to sub-uHa vs out-of-process ORCA 6.1 RI-CEPA/n (canonical orbitals). No (T); closed-shell only.
DLPNO-CCSD	"dlpno-ccsd"	RHF / UHF	Domain-based local pair natural orbital CCSD. Closed-shell: reduced-scaling local solver (pair classification, PAO domains, per-pair PNO solver) whose amplitude equations are solved in the compiled PNO-space kernel (every residual term contracted in the PNO space that carries the contracted index, Riplinger-Neese 2013 Sec. II C; exact against the extended-domain oracle and canonical CCSD at full domains). Open-shell (multiplicity > 1) auto-routes to the local spin-orbital DLPNO-UCCSD solver, whose amplitude equations are solved in a compiled spin-orbital PNO-space kernel (spin-pure PNOs, exact against canonical UCCSD at full domains); it applies the shared Liakos threshold triple but does not claim Saitow’s single-spatial-orbital SOMO-pair scaling. The O(N⁶), max_nbf-capped pilot remains opt-in.
DLPNO-CCSD(T)	"dlpno-ccsd(t)"	RHF / UHF	Closed-shell: local solver plus local DLPNO-(T) (per-triple TNO domains), full-domain == canonical CCSD(T). Retained accuracy benchmarks use explicitly pinned pre-#140/#448 all-electron recipes; they are not evidence for the current published-core/NormalPNO default. Open-shell auto-routes to the reduced-scaling UHF-reference spin-orbital DLPNO-UCCSD(T) solver; the dense O(N⁶), max_nbf-capped pilot is selected only by explicitly passing DLPNOUCCSDPilotOptions.
DLPNO-MP2	"dlpno-mp2"	RHF / UHF	Domain-based local pair natural orbital MP2 (Pinski 2015); the open-shell DLPNO-UMP2 extension retains independent UHF alpha/beta orbital and PNO spaces. Boys-localised occupieds, per-pair PNO virtual space; the zero-threshold all-electron limit reproduces canonical RI-MP2 to <= 1 uHa. Historical recovery data use explicitly pinned pre-#140/#448 cutoffs and are not claims about the current named default.
IC-CASPT2	"caspt2"	CASSCF/CASCI	Internally-contracted CASPT2, un-gated at v0.12.0. Large-basis direct active-CI path.
CASCI + CASSCF	"casci" / "casscf"	-	Multi-root CASCI (casci_options=CASCIOptions(nroots=N)) with state-averaged CASSCF (casscf_options=CASSCFOptions(nroots=N)); orbital-optimised. The CASSCF analytic nuclear gradient is FD-tight (~1e-7, default compute_wz=False path), validated in v0.15.0. State-averaged + open-shell CASSCF gradients fall back to FD. CASPT2/NEVPT2 gradients experimental (full-energy FD).
TDDFT (Casida + TDA)	run_tddft_tda / run_tddft_casida / run_job(tddft=True)	RHF/RKS/UHF/UKS; ROHF (TDA only, doubly-occupied to virtual space, labelled TDA-ROHF; Casida on ROHF and any ROKS response are refused before the SCF, #361)	Vertical excitation energies, oscillator strengths, transition dipoles, dominant orbital transitions. Runner integration: `run_job(tddft=True, nto=True, tddft_type=”tda”
CIS / TDA	run_tddft_tda (TDA limit)	HF / MSINDO	Configuration-interaction singles; Tamm-Dancoff approximation. Finite-difference excited-state gradients (state-tracked).
MRCI	"mrci"	CASCI / CASSCF	Uncontracted MR-CISD with the Davidson +Q size-extensivity estimate (mrci+q citation route). Energies; optimisation is finite-difference only through optimizer_backend="native" (#357).
BCCD / BCCD(T)	"bccd", "bccd(t)"	RHF	Brueckner coupled cluster: orbitals rotated until the singles amplitudes vanish (Handy 1989), with the (T) correction on the Brueckner reference. Closed shell only. Also run_bccd.
Green’s function (GF2 / OVGF)	"ovgf"	HF / MSINDO	Quasiparticle IPs and EAs (result .ovgf); renormalised GF2 with full third-order self-energy + geometric screening auto-printed for small systems. Open-shell (UHF) propagator supported. MSINDO reference: vibeqc.semiempirical.methods.msindo.msindo_ovgf (renormalize=True for renormalised GF2).

Density fitting + RIJCOSX (v0.8.0)

JKBuilder polymorphic Fock build via three concrete kernels; see user_guide/density_fitting.md.

Kernel	Driver flags	When
FourIndexJKBuilder	(default)	small (≤250 BFs)
DFJKBuilder (RIJK)	density_fit=True, aux_basis="..."	medium (~250-1000 BFs), hybrid DFT
COSXJKBuilder (RIJCOSX)	density_fit=True, cosx=True, aux_basis="..."	large (>1000 BFs), hybrid DFT

• RIJK gradient: production-ready. The historical ~115 mHa glycine bug closed in v0.8.0; see user_guide/density_fitting.md.

• RIJCOSX SCF energy: validated to 0.26 mHa vs ORCA 6.1.1 on glycine / def2-TZVP. (An earlier revision of this bullet claimed the analytic gradient was “validated” on the strength of that energy comparison; the gradient claim was unproven and, pre-2026-07-29, wrong.)

• RIJCOSX analytic gradient: matches finite differences of the RIJCOSX energy surface to the neglected grid-motion band (~2e-4 Ha/bohr on H2O/def2-tzvp, ~3e-3 worst case OH/def2-svp; gate: tests/test_rijcosx_gradient_fd.py). Against ORCA 6.1.1 EnGrad on glycine/def2-TZVP: |grad| 0.08696 vs 0.08703 Ha/bohr, componentwise RMS 3.3e-3, max 1.1e-2; the residual is the documented fitted-Fock response term plus neglected grid-motion/weight derivatives (see compute_cosx_k_gradient_contribution notes in cpp/include/vibeqc/cosx.hpp).

• default_aux_basis_for(orbital_basis_name, kind="jk"/"ri") helper picks aux basis automatically.

• Experimental second engine (v0.17.16). run_job and run_periodic_job accept integral_engine=IntegralEngineOptions(...), coulomb=CoulombBuildOptions(...) and exchange=ExchangeBuildOptions(...) (from vibeqc.manta): libint or MANTA integrals, exact, RI, BUPO or RI-BUPO Coulomb and exact or RI exchange, chosen independently, with an explicit resource budget and the executed provider recorded in the .out, the manifest and the QVF. MANTA’s exact build reproduces libint to 1e-10 Ha on water/cc-pVDZ. These selectors cannot be combined with the density_fit/aux_basis/cosx keywords above. See user_guide/integral_engines.md.

Periodic SCF (native GDF / GPW / BIPOLE / Γ-CCM / χ-CCM)

Γ-CCM and χ-CCM[1] are distinct constructions: Γ-CCM is the neutral finite-torus construction (producers real-gamma and neutral-bloch), while the four-center variant is the 2014 union-and-weight lineage kept in the same development line, and χ-CCM is the finite-translation-group character construction.

The front door is run_periodic_job(method="aiccm", variant=...), where variant names the formulation and is mandatory; the method name does not imply one. The four variants are "real-gamma", "neutral-bloch", "four-center" and "chi". There is deliberately no variant="gamma": ruling R1 makes Γ-CCM the neutral construction with two admissible producers (real-gamma and neutral-bloch), so the bare word is ambiguous and fails closed naming both. The SCF reference is inferred from the functional and the cell’s multiplicity unless scf_reference= names one: "rhf" or "uhf" for Hartree-Fock (for example a closed-shell UHF), or a restricted open-shell "rohf" / "roks". The legacy jk_method="aiccm2026dev-a" / "aiccm2026dev-b" spellings still resolve, each with a DeprecationWarning naming the variant form. The whole line stays experimental.

KPoints.recommend(classifier="pre-scf", basis=...) runs a three-cycle Gamma GDF SCF, on a compact 3D cell the bulk GDF Hamiltonian (#468), and classifies from the HOMO-LUMO gap of that unconverged result. When no gap can be read, the recommendation warns and takes the SAFE metallic mesh/smearing default. This is not a measured metallic state. Supply a known band_gap, is_metal, or a qualified converged scf_result for electronic classification.

Method	Driver	k-point support	Notes
RHF, Γ-only GDF	run_rhf_periodic_gamma_gdf	Γ, dim=2/3 (a dim=1 wire raises)	native vibe-qc J/K via Gaussian density fitting; no in-process PySCF. A compact dim=3 cell (image cells inside cutoff_bohr) runs run_krhf_periodic_gdf at (1,1,1) with exxdiv='ewald' exchange and says so in a GammaExchangeRouteWarning (#468)
RKS / hybrids, Γ-only GDF	run_rhf_periodic_gamma_gdf(functional=…)	Γ, dim=2/3 (a dim=1 wire raises)	native libxc V_xc; hybrid exact exchange uses the functional’s HF fraction. On a compact dim=3 cell it runs the k-mesh GDF engine at (1,1,1) on PeriodicKSOptions (#468)
KRHF multi-k GDF	run_krhf_periodic_gdf	Γ + multi-k, dim=2/3 (a dim=1 k mesh raises)	native k-dependent GDF. A dim=1 wire with a k mesh raises NotImplementedError (no consistent 1-D Coulomb gauge); run it as a dim=3 cell with transverse vacuum. On dim=2, an explicit full Gamma-centered (n1,n2,1) mesh selects the signed slab-truncated metric and rigorous 2D one-electron gauge; custom/IBZ meshes, smearing, restarts, and open shell stay closed. Closed-shell slab analytic gradients ship. Also reachable via run_periodic_job(jk_method="gdf", kpoints=...)
KRKS multi-k GDF	run_krks_periodic_gdf	Γ + multi-k, dim=2/3 (a dim=1 k mesh raises)	native multi-k RKS / full-range hybrid path; dim=1 with a k mesh raises as for KRHF. The bounded dim=2 route has the same restrictions as KRHF and rejects range-separated functionals; closed-shell slab analytic gradients ship; PySCF / CRYSTAL are out-of-process references only
DLPNO-CCSD(T) on crystals	run_periodic_job(method="dlpno-ccsd(t)") or "dlpno-ccsd"; library vibeqc.periodic_dlpno.run_periodic_dlpno_ccsd	Γ or a Gamma-centred mesh folded to the real BvK supercell, dim=3	Preview. RHF (singlet) or UHF (open shell) reference on the k-mesh GDF engine with its classic RSGDF fit retained as the correlation factors, Pipek-Mezey occupied orbitals, minimum-image pair distances; energies in total and per cell. Full-domain limit equals PySCF KRCCSD(T) to the fit tolerance. Refuses BIPOLE/GPW, functionals, symmetry_reduce_k, smearing, dim < 3, optimize/hessian. See user_guide/periodic_local_correlation.md
Symmetry-reduced exact exchange	run_periodic_job(..., symmetry="attach", symmetry_reduce_k=True) (driver ibz_native=True)	multi-k GDF, dim=3, RHF/UHF/RKS/UKS with exact exchange	Preview. Exchange built at the irreducible-wedge representatives and transported over each star (every operation, glides and screws included, on the SR/LR fit); diagonalisation and energy on the full mesh, so energies equal the unreduced ones to roundoff. Per-star occupation fill and a little-group state gate; refuses pure functionals, Gamma-only meshes, bz_integration="gilat". See tutorial/symmetry_reduced_k_mesh.md
MANTA periodic HF / KS (finite domains)	run_periodic_job(integral_engine=IntegralEngineOptions(backend="manta"), coulomb=..., exchange=..., periodic_integrals=...)	Γ + full regular meshes, neutral all-electron dim=3	Experimental. RHF/UHF/RKS/UKS with exact or RI J/K, pure, global-hybrid and range-separated LDA/GGA, BvK Ewald exchange correction, periodic BUPO for HF; every domain explicit, no automatic accuracy profile. See user_guide/integral_engines.md
RIJCOSX periodic SCF	run_periodic_job(jk_method="rijcosx")	Gamma RHF; vacuum-padded Gamma RKS/UHF/UKS; true multi-k RHF/RKS/UHF/UKS	RI-J from the native GDF loop plus periodic COSX exchange (k_exchange="cosx") on true multi-k meshes (weighted full-BZ quadratures supported); one-cell tight-cell RKS/UHF/UKS fail closed rather than relabeling a GDF-only route
χ-CCM SCF / post-HF	run_aiccm2026dev_b_rhf/rks/uhf/uks, restricted and unrestricted MP2/CCSD(T)/PNO APIs; run_periodic_job(jk_method="aiccm2026dev-b")	real-space cyclic lattice extension, 3D absolute energies only	finite-character (Γ-centred character-mesh) CCM with four-centre, RI, and RIJCOSX restricted/unrestricted SCF plus 3D finite-torus correlation; every 1D/2D B backend fails closed pending a shared wire/slab Coulomb convention; local CC remains an O(N^6) correctness pilot, not production reduced scaling
χ-CCM RI-MP2	run_aiccm2026dev_b_mp2	finite cyclic mesh, 3D	momentum-conserving canonical RI-MP2 on the χ-CCM RI-RHF reference; external KMP2 parity; experimental
variant="four-center" (union-and-weight)	run_periodic_job(method="aiccm", variant="four-center"); library vibeqc.periodic.ccm, run_ccm_rhf, run_ccm_rks, run_ccm_mp2, run_ccm_ump2, run_ccm_ccsd	real-space cyclic cluster extension (nrep), dim=3	union-and-weight/Wigner-Seitz integral-weighting CCM; the 2014 lineage, and under ruling R1 a different construction from the two neutral producers, not a representation of theirs. HF/KS, construction-specific RIJCOSX research, MP2/UMP2, CCSD(T), analytic gradients, properties, localization, symmetry. Experimental (every SCF driver emits AICCM2026DevAExperimentalWarning)
variant="real-gamma" (neutral Γ-CCM producer)	run_periodic_job(method="aiccm", variant="real-gamma"); library run_ccm_rhf_direct, run_ccm_ri_mp2, run_ccm_ri_ccsd, ccm_dlpno_*	one real Γ-supercell (no k-mesh); the Γ-centred mesh IS the BvK torus	The neutral fitted-torus construction evaluated in its real-Γ representation. Carries the exchange-q=0 seam explicitly (exchange_exxdiv). Reports per-unit-cell energies; the library drivers underneath return total cyclic-cluster ones. Experimental
variant="neutral-bloch" (neutral Γ-CCM producer)	run_periodic_job(method="aiccm", variant="neutral-bloch"); library run_ccm_rhf_gdf	finite BvK torus in its Bloch representation	The same neutral Hamiltonian as real-gamma in the other representation (ruling R1: one construction, two admissible producers). Correlation is not available here; the correlation drivers work in real-Γ supercell space; and the refusal names real-gamma as the exact substitute. Experimental
AICCM correlation (correlation=)	run_periodic_job(method="aiccm", variant=..., correlation="mp2" | "ccsd" | "dlpno-mp2" | "dlpno-ccsd")	follows the variant’s SCF	Optional post-HF on top of the variant’s own reference. The driver is chosen to MATCH the construction the SCF ran, never to substitute another, so each arm cites its own lineage and the two never share a citation row. mp2 and ccsd run on four-center and real-gamma; the two DLPNO treatments are real-gamma only, because the bare four-centre operator has no RI decomposition. At their default zero truncations the DLPNO routes reproduce the canonical ones on the same reference to machine precision. Needs a Hartree-Fock reference. Experimental
A-namespace neutral fitted-torus controls (library)	run_ccm_rhf_gdf, run_ccm_rhf_direct, run_ccm_ri_mp2, run_ccm_ri_ccsd, run_ccm_uccsd, ccm_dlpno_*	separately declared finite-torus control operator	The same drivers the two neutral variants use, reachable directly for control work. These APIs are in vibeqc.periodic.ccm for historical reasons; a bare call is not by itself a Γ-CCM or χ-CCM construction result; experimental (emit AICCM2026DevAExperimentalWarning)
Cell-resolved DF integral blocks	compute_2c_eri_lattice_blocks, compute_3c_eri_lattice_blocks, bloch_sum_*_eri_blocks, build_lpq_bloch_native	dim=1/2/3	native translation-resolved 2c/3c storage and Bloch/Lpq assembly for k-dependent GDF; Γ tensors are recovered by summing blocks
2D slab SCF (SLAB_EWALD_2D)	run_periodic_job(jk_method="auto") on a dim=2 system; vibeqc.slab_2d(a1, a2, atoms) to build one	Γ + multi-k, dim=2 only	Rigorous vacuum-free 2D Ewald (Parry 1975; de Leeuw and Perram 1979) for RHF / RKS / UKS. No vacuum gap: the third lattice column is synthesized bookkeeping and the total energy is invariant to it; E_nuclear is k-mesh independent. AUTO selects it for every dim=2 system. Explicit jk_method="gdf" is a bounded closed-shell RHF/RKS alternative with a signed slab-truncated fit and analytic gradients; BIPOLE / GPW / GAPW / RIJCOSX remain closed. Fermi-Dirac smearing and slab-GDF open shell are not shipped yet.
Green’s-function surface embedding	vibeqc.periodic_embedding.run_embedded_surface, EmbeddedSurfaceCalculator	semi-infinite surfaces, 3D Gaussian foundation	Experimental (EmbeddedSurfaceExperimentalWarning); thick-slab GDF parity pending, so absolute energies and densities are qualitative; see surface embedding
BIPOLE RHF / UHF / RKS / UKS	run_pbc_bipole_rhf, ..._uhf, ..._rks, ..._uks	Γ + multi-k, 3D through run_periodic_job	Production exact Ewald-J energy and finite-difference atomic-force route for closed- and open-shell periodic SCF. Fixed-cell atomic relaxation ships; variable-cell work fails closed. Analytic gradients remain a gated preview. The quartet multipole far-field prototypes are driver-unreachable: all four drivers default to False, and explicit True raises before setup
GPW RHF / ROHF / ROKS / UHF / RKS / UKS (v0.12.0)	run_periodic_job(jk_method="gpw")	Gamma supports all six methods; multi-k pure-DFT RKS/ROKS/UKS supports LDA/GGA/meta-GGA in molecular-limit and compact Bloch regimes through the public runner or the matching run_periodic_*_gpw_multi_k API; multi-k UHF/ROHF and hybrids remain gated	Production Gaussian Plane Waves route (serial), unified run_periodic_job user surface; ROHF and ROKS are maintained preview with one restricted orbital set and integer 2/1/0 occupations; GAPW all-electron augmentation; DFT+U for Gamma RHF/RKS/UHF/UKS plus multi-k RKS; D3-BJ dispersion; Gamma-point analytic gradient; FD Hessian; smearing. The MPI z-slab grid overlay (GpwJBuilder(mpi_aware=True)) is experimental: its multi-rank Hartree-J build was broken through v0.12.0, fixed post-release, pending validation on a real multi-rank run
RHF / RKS / UHF / UKS Ewald-3D (legacy)	run_*_periodic_*_ewald3d	Γ + multi-k	native debug path; not the CRYSTAL-parity target
Multi-k UHF / UKS via GDF	run_kuhf_periodic_gdf, run_kuks_periodic_gdf; public runner	Γ + multi-k, dim=3	Public GDF open-shell route. dim=1 has no open-shell route (the Gamma drivers accept only dim=3; a k mesh raises), and slab open shell remains fail-closed.
ROHF periodic, Ewald-3D	run_rohf_periodic_gamma_ewald3d, run_rohf_periodic_multi_k_ewald3d	Γ + multi-k, dim=3	Standalone HF drivers with the shared Roothaan coupling and integer 2/1/0 occupations; not wired through run_periodic_job
ROHF periodic, native GDF	run_periodic_job(method="ROHF", jk_method="gdf"), run_krohf_periodic_gdf	Γ (a (1,1,1) mesh) + full multi-k, dim=3	Public and standalone density-fitted ROHF on the per-(k_i,k_j) cderi cache: same J / per-spin K / exxdiv='ewald' machinery as the PySCF-pinned KRHF/KUHF drivers, with the Roothaan coupling as the only new step. Exchange contracts against the occupied MO blocks. Multiplicity 1 reproduces run_krhf_periodic_gdf exactly; Li/STO-3G Γ is +2.86e-5 Ha from PySCF KROHF/GDF, collapsing to +2.2e-7 Ha with the rsgdf mesh. ROKS, smearing, 2D slabs, gradients, optimization, restarts, COSX and ibz_native remain fail closed
ROHF / ROKS periodic, GPW	run_periodic_job(method="ROHF", jk_method="gpw"), run_periodic_job(method="ROKS", functional=..., jk_method="gpw"), run_periodic_rohf_gpw, run_periodic_roks_gpw, run_periodic_roks_gpw_multi_k	ROHF: Gamma only; ROKS: Gamma plus full multi-k pure-DFT meshes; dim=3	Maintained-preview restricted-open-shell route: GPW Hartree J, per-spin exact K where required, spin-polarised XC for ROKS, the Ewald one-electron/nuclear gauge, and complex-safe Roothaan coupling. Unsupported dimensions, gradients, GAPW, multi-k hybrids, range-separated ROKS, and electronic smearing fail closed; smearing_alpha is nuclear smoothing
Atomic gradients (periodic)	route-specific FD forces; compute_gradient_periodic_* previews	Γ + multi-k by route	BIPOLE production forces use finite differences of the executed energy; analytic coverage is route- and envelope-specific and remains preview where documented
Periodic stress tensor	compute_stress_gpw, compute_stress_gapw; semiempirical routes through vibeqc.semiempirical.periodic	GPW / GAPW Γ, dim=3; semiempirical Γ + full-k	Finite difference, not analytic. Both GPW/GAPW entry points central-difference the fixed-density energy over strain, rebuilding the basis on each strained cell, so Hellmann-Feynman and Pulay contributions arrive together and the density is not relaxed per strain; compute_stress_gapw requires one_centre="block". Semiempirical stress is one native batched FD pass, except Γ-point GFN2-xTB, whose native stress is analytic and FD-validated on all nine components. There is no analytic GPW/GAPW/GDF/BIPOLE stress, so run_periodic_job(optimize_cell=True) fails closed on all four; variable-cell relaxation is reachable only on the full-k semiempirical routes. Periodic ECP stress is unimplemented (#34)
Periodic force-virial diagnostic	compute_stress_tensor	Γ + multi-k	Diagnostic only; not the periodic stress. For the stress itself and for which routes accept optimize_cell, see the row above

Reference benchmark: ΔE = -6.8e-12 Ha for MgO/sto-3g/Γ RHF vs out-of-process PySCF.pbc.GDF. Multi-k KRHF / KRKS GDF is now native, the multi-k Ewald-gauge fix brings a multi-k mesh to within 0.000 mHa of the Γ-point energy in the molecular limit; there is no in-process PySCF backend.

See user_guide/multi_k_scf.md for the full multi-k story including the dense-Lpq RAM ceiling caveat.

Semiempirical methods + MLIP (v0.11.0-v0.12.0)

Status vocabulary: production = validated within the stated scope; experimental = runs, but the route is not validated for published numbers; gated = fails closed with a scientific error message; retirement target = still executable, but the maintainer’s target matrix removes it from the portfolio. “FD” is finite difference. No semiempirical or MLIP method has a nuclear Hessian: run_job(hessian=True) is basis-free and skips vibrational analysis with a note in the .out.

Method	run_job(method=…)	Elements	Molecular: energy / gradient / Hessian	Periodic routes (dimension, k-mesh)	Open-shell	Status and gate
MSINDO	"msindo"; SECCM via "seccm" / "ccm" + ccm_options	INDO H-Xe (Z 1-54); NDDO H, Li-F, Na-Cl closed shell	INDO energy RHF/UHF and closed-shell analytic gradient. NDDO closed-shell energy and analytic gradient for H, Li-F, Na-Cl, including source-parity SPDD terms for Al through Cl. No Hessian	No Bloch Gamma and no full-k route (both fail closed). SECCM cyclic cluster 1-D/2-D/3-D, closed shell; native C++ for H-Kr, deterministic Python reference for Rb-Xe	INDO molecular UHF within validated fixtures; NDDO molecular and SECCM closed-shell only	Production within scope. Independent MSINDO INDO/NDDO implementation. Native C++ production routes cover supported molecular energies and INDO closed-shell analytic gradients; closed-shell NDDO gradients use a mixed native-energy/Python-analytic route. CCM energy/gradients use C++ for H-Kr and the deterministic Python reference for Rb-Xe; a selected native calculation does not fall back after failure. Python also remains for reference/orchestration paths including excited-state/root-tracking workflows
DFTB0 / UDFTB0	"dftb0"	91 in-house parameter records	Energy and analytic gradient, both native; no Hessian	Bloch Gamma energy + analytic gradient (native, 15-bohr image domain). Full-k Bloch energy/bands experimental; gradient and stress by one native batched FD pass; Monkhorst-Pack mesh of three positive integers, default 1x1x1. DFTB0-SECCM 1-D/2-D/3-D with analytic fixed-topology gradients	UDFTB0 molecular and Gamma; full-k and SECCM closed-shell only	Screening / preoptimization; not production-validated. In-house parameters, no DFT-fitted repulsives, no DFTB+ (mio-1-1) parity. Full-k derivatives gated pending analytic derivatives and DFTB+ parity. SECCM is gated experimental and scoped to the explicit repulsive-pair table (H, C, N, O, F, P, S, Cl). Absolute energies are not claimable while the repulsive is an A/R¹² placeholder (issue #306)
SCC-DFTB / USCC-DFTB	"scc_dftb"	91 in-house parameter records	Energy and analytic gradient, both native; no Hessian	Bloch Gamma energy native, gradient by central differences of the converged energy. Full-k Bloch SCC energy/bands experimental with optional Fermi occupations; gradient and stress by one native batched FD pass; Monkhorst-Pack mesh of three positive integers, default 1x1x1. SCC-DFTB-SECCM 1-D/2-D/3-D with analytic fixed-topology gradients; charged cells only through the opt-in Madelung/Ewald embedding	USCC molecular and Gamma; full-k and SECCM closed-shell only	Screening / preoptimization; not production-validated. In-house parameters, no DFTB+ parity. Analytic SCC derivatives reject nonconverged results. Full-k derivatives and SECCM remain gated experimental
GFN2-xTB	"gfn2_xtb"	H-Rn (Z 1-86), fetched on demand	Energy, shell and faithful AES SCC gradients (of the Mermin free energy when smeared); scoped post-SCF native D4; no Hessian	Bloch Gamma energy native; the native gradient and all-nine stress match gapped finite differences. Full-k route fails closed. GFN2-SECCM 1-D/2-D/3-D, energy only, with optional Madelung embedding; Ewald-gamma supports 1-D/2-D and rejects non-molecular 3-D (#444)	None: the route is closed-shell only at every boundary	Gated experimental. Molecular H0, shell SCC, third-order, AES, molecular gradients including atom-resolved analytic faithful AES, scoped post-SCF native D4, and FD-validated Gamma-periodic derivatives are implemented. Faithful periodic AES electrostatics and external xtb parity remain production gates. Native D4 covers H, He, B, C, N, O, F, Ne, Cu and returns zero with GFN2D4UnsupportedWarning elsewhere
PM6 / UPM6	"pm6"	75 chemical-element records from the bundled MOPAC cache	Energy native NDDO; gradient by native batched FD (no analytic gradient); no Hessian	Bloch Gamma energy native, gradient and stress by one native batched FD pass (mixed-native, experimental). Full-k route fails closed. PM6-SECCM 1-D/2-D/3-D energy with FD fixed-topology gradients; s/p AO channels only, so elements whose PM6 record needs d orbitals fail closed	UPM6 molecular only; periodic Gamma and SECCM closed-shell only	Experimental: molecular development route. Exact MOPAC heat-of-formation convention parity is open, and spherical Klopman-Ohno gamma/core-core fallbacks apply where the bundled MOPAC diatomic data is incomplete. Periodic PM6 is experimental
PM6-SECCM CCM Madelung embedding	opt-in madelung=True on vibeqc.semiempirical.seccm.run_pm6_seccm	as PM6	not applicable (SECCM boundary only)	1-D and 2-D only; three cyclic dimensions fail closed because the WS-folded remainder has no thermodynamic limit. Without the embedding a nontrivial cyclic topology needs an explicit allow_truncated_electrostatics=True acknowledgement	closed shell only	Experimental, opt-in, acknowledgement-gated (issue #217). The MSINDO CCM operator: the classical point-charge field enters the Fock diagonal and the potential is the full Ewald lattice sum minus the WS-internal bare 1/r, WS-weighted with the self image excluded. Unembedded cyclic runs are an algorithm-mechanics probe, not a solid-state prediction: the WS-truncated monopole sum reproduces the exact rocksalt Madelung constant only to -43%..+38%
OM2 / OM3	"om2", "om3"	5 published (H, C, N, O, F)	Energy native NDDO; gradient by native batched FD; no Hessian	Bloch Gamma and full-k routes both fail closed: the retired prototype was not the published ORT/ECP/penetration Hamiltonian. OM2/OM3-SECCM 1-D/2-D/3-D energy only, no gradient; cyclic-image execution requires an explicit truncated-electrostatics acknowledgement; published eq-13 three-center image weighting is the default, eq-10 available for comparison	Molecular only; periodic and SECCM closed-shell or gated	Retirement target (maintainer target matrix, 2026-09-02: no molecular, Bloch or CCM target; fail closed, deprecate, remove from the public portfolio). Still executable today as a validated molecular development/prescreening surface; bond minima remain about 0.1-0.3 Å from published references and external implementation parity is open. Nothing has been removed yet, so no release has dropped it
OM1	"om1"	5 published (H, C, N, O, F)	Energy native NDDO with an experimental warning; gradient by native batched FD; no Hessian	Bloch Gamma, full-k, and OM1-SECCM all fail closed	Molecular only	Retirement target (same maintainer decision as OM2/OM3). Experimental: the defining analytic core-valence ECP (Kolb-Thiel 1993) is not implemented, which also closes OM1-SECCM. Nothing has been removed yet, so no release has dropped it
PM7 / UPM7	"pm7", "upm7": recognized spellings that fail closed	75 MOPAC parameter records retained, queryable only	Gated: no energy, no gradient, no Hessian	Gated at every boundary: Bloch Gamma, full-k, and SECCM all fail closed	Gated	Retired as an implementation target (maintainer target matrix, 2026-09-02). Tombstone only: every request raises “PM7 is gated until the published feathered electrostatics and all coupled NDDO terms are implemented and validated.” Stewart’s smooth exact-electrostatic/point-charge feathering is not implemented. PM7 has never been executable in any release, so there is no last version that carried it
MACE MLIP (v0.11.x)	"mace" for molecules; direct periodic drivers	Registered MACE-MP/MPA-0 materials models and MACE-OFF23 organic models	Energy, forces, optimization, and NEB; no Hessian	Direct periodic energy/forces support 1D, true 2D slabs, and 3D; vibeqc.mlip.mace.optimize_periodic_mace_positions provides fixed-cell relaxation with frozen atoms. Stress and vibeqc.mlip.mace.optimize_periodic_mace_cell are 3D-only. run_periodic_job(method="mace") is not supported	model-dependent	Optional mace-torch engine on Python <= 3.13. Model-specific energy scales are not electronic totals

Sources for the statuses above include the ACCEPTANCE-managed families in semiempirical_acceptance_matrix.py, the queryable vibeqc.semiempirical.semiempirical_route_status(route) registry, the boundary gates in vibeqc.semiempirical.routes.SemiempiricalRoutePlan, and the maintainer’s 2026-09-02 target matrix, which is kept in the private operations documentation (handovers are not stored in this repository). The per-route user-guide table is at user_guide/semiempirical.md § Method status.

Effective core potentials

libecpint 1.0.7 vendored. Heavy-element chemistry (d-block, lanthanides, actinides) reachable without all-electron basis.

ECP library	Family	Coverage
ecp10mdf	Stuttgart-Köln MDF, 10-electron core	K-Kr
ecp28mdf	Stuttgart-Köln MDF, 28-electron core	Rb-Xe
ecp46mdf	Stuttgart-Köln MDF, 46-electron core	Cs, Ba
ecp60mdf	Stuttgart-Köln MDF, 60-electron core	Hf-Rn, Ac-U
ecp78mdf	Stuttgart-Köln MDF, 78-electron core	Fr, Ra
lanl2dz	Hay-Wadt LANL	Na-La, Hf-Bi, U-Pu

These are the elements actually present in each bundled XML, read from the files themselves; they are narrower than the family names suggest. In particular no bundled XML library covers Ce-Lu: the lanthanides are reached through a basis’ own .ecp sidecar (pob-TZVP-rev2 carries La-Lu), not through ecp_library.

A bundled basis’ .ecp sidecar is attached inline and per element by the molecular SCF wrappers; the XML libraries above serve explicit ecp_centers + ecp_library requests. Whether an atom needs an ECP in a basis is answered by vibeqc.basis_registry.ecp_requirements, never by the basis name. MP2, CC, DLPNO, the double hybrids, ROHF/ROKS and the determinant solvers consume an ECP reference through its valence count and an ECP-aware frozen core; CASCI/CASSCF/MRCI/NEVPT2/CASPT2 still refuse one, while OVGF runs on an ECP reference (HI/def2-SVP gives a valence IP of 10.09 eV) without a dedicated validation. See user_guide/ecp.md.

Analytic nuclear gradients are ECP-aware on all four molecular drivers when the exact XML-library or inline-primitive ECP provenance is repeated on GradientOptions. The ASE force path copies those fields automatically. The gradient applies Z_eff nuclear pieces plus dV_ECP/dR via libecpint first derivatives, with analytic-vs-FD agreement pinned to <= 1e-5 Ha/bohr. Gas-phase molecular ECP geometry optimization and finite-difference Hessians preserve the same operator while centers follow their atoms. Analytic ECP Hessians, CPCM forces, and top-level ECP NEB currently fail closed rather than differentiating the wrong energy. Note that run_job(hessian=True) builds a gas-phase, dispersion-free Hessian even when solvent= or dispersion= is set, and says nothing about it in the .out, so the frequencies and the thermochemistry rows do not include those terms. See user_guide/ecp.md § Gradients.

Reference: Pt UHF/LANL2DZ = −118.227 Ha (22 BFs, 125 SCF iters).

SCF convergence aids

Aid	API	Coverage	Notes
DIIS extrapolation	*_options.use_diis	molecular + periodic Γ + multi-k	default on, ~10× iter speedup
EDIIS + EDIIS+DIIS hybrid (v0.8.0)	*_options.scf_accelerator=SCFAccelerator.EDIIS_DIIS	molecular + C++ direct-truncated periodic + Python Γ-Ewald (RHF/RKS/UHF/UKS)	flagship for stiff convergence; rollout to Python multi-k Ewald / BIPOLE / GDF in progress
KDIIS (Kollmar)	*_options.scf_accelerator=SCFAccelerator.KDIIS	molecular + C++ Γ-only periodic + Python Γ-Ewald (RHF/RKS/UHF/UKS)	orbital-rotation gradient error; ORCA’s opt-in !KDIIS; multi-k path needs per-k MO-basis design
ADIIS (Augmented Roothaan-Hall)	*_options.scf_accelerator=SCFAccelerator.ADIIS	molecular + C++ direct-truncated periodic + Python Γ-Ewald	EDIIS sibling, no per-iter energy needed
Dynamic damping (Zerner-Hehenberger 1979)	*_options.dynamic_damping, dynamic_damping_min/_max	molecular + C++ direct-truncated periodic + Python Γ-Ewald (RHF/RKS/UHF/UKS)	adaptive density mixing α; composes freely with the accelerator family
Damping (static)	*_options.damping	all	density mixing
CRYSTAL-style FMIXING	*_options.fock_mixing; run_periodic_job(fmixing_percent=30)	molecular + native periodic direct + supported GDF routes	Fock/KS matrix mixing; FMIXING 30 = 0.30, separate from density damping; GDF exceptions are fail-closed or AUTO-filtered as documented in scf_convergence.md
Saunders-Hillier level shift	*_options.level_shift	molecular + periodic Γ + multi-k, with route-dependent GDF coverage	Phase C1a; GDF exceptions are fail-closed or AUTO-filtered as documented in scf_convergence.md
Level-shift warm-up/restart	*_options.level_shift_warmup_cycles; run_*_periodic_gdf(..., level_shift_warmup_cycles=...)	native Γ-GDF + Γ KRHF/KRKS bridge	-1 auto, 0 persistent shift, positive values restart unshifted after that many cycles
Quadratic (“Newton”) SCF	*_options.quadratic_fallback_iter	molecular + periodic Ewald Γ/multi-k	Phase C1c, fallback for small-gap systems; experimental aiccm2026dev-b routes require 0 and fail closed on activation
Fermi-Dirac smearing	*_options.smearing_temperature; `run_periodic_job(smearing_temperature=”auto”	“metal”	…); vq.SmearingOptions, vq.apply_smearing(), vq.resolve_smearing_temperature()`
Periodic density mixers (Anderson / Broyden / Kerker, v0.14)	run_periodic_job(..., density_mixer=...); run_rks_periodic_scf(...) / run_rks_periodic_multi_k_ewald3d(...)	closed-shell explicit-kpoint GDF dispatch + lower-level multi-k RKS Ewald	Kerker preconditions the density residual without moving the SCF fixed point; unsupported BIPOLE, GPW/GAPW, open-shell, and implicit-Gamma routes fail closed
auto_optimize_truncation	PeriodicRHFOptions(auto_optimize_truncation=True)	periodic	default on; jointly optimises lattice cutoffs + Schwarz screening
R-CDIIS / AD-CDIIS (Chupin 2021)	*_options.scf_accelerator=SCFAccelerator.R_CDIIS / AD_CDIIS, diis_restart_tau, diis_adaptive_delta	molecular	restarted and adaptive-depth conjugate DIIS; see adaptive DIIS
OpenTrustRegion orbital optimiser	run_job(orbital_optimizer="opentrustregion", opentrustregion_options=...)	molecular RHF/UHF/RKS/UKS	external trust-region second-order optimiser, build-gated (-DVIBEQC_ENABLE_OPENTRUSTREGION=ON); refuses smearing, solvent and TDDFT; see OpenTrustRegion
Stability analysis	*_options.stability_check and the stability_* fields	molecular RHF/UHF/RKS/UKS	Seeger-Pople internal and external stability of the converged determinant; see SCF convergence
Eigensolver framework	run_job(solver="dense" | "davidson" | "lobpcg"); vibeqc.solvers.eigensolver	molecular (LOBPCG falls back to Davidson on the molecular route) + periodic	Davidson and LOBPCG production; Jacobi-Davidson and GPLHR experimental; see solver framework
Gilat-Raubenheimer BZ integration	run_periodic_job(bz_integration="gilat")	closed-shell multi-k GDF/RIJCOSX and BIPOLE RKS/UHF/UKS; fails closed on GPW/GAPW	linear-interpolation Fermi-surface integration as an alternative to smearing; see metallic BZ integration
SAP initial guess	initial_guess="sap" (also HCORE, SAD, PATOM, HUECKEL, MINAO, READ, FRAGMO)	molecular + periodic by route	superposition of atomic potentials (Lehtola 2019); capability-checked per route; see initial guesses

XC functionals (v0.8.0 expanded)

Resolved through libxc 7.0.0; see user_guide/functionals.md. Short aliases the resolver knows:

Alias	Type	HF-exchange	libxc id
lda / svwn	LDA	0 %	1 + 7
pbe	GGA	0 %	101 + 130
blyp	GGA	0 %	106 + 131
b3lyp (= b3lyp5)	hybrid	20 %	475 (VWN5, the ORCA / TURBOMOLE / ADF / CRYSTAL convention)
b3lyp/g (= b3lypg)	hybrid	20 %	402 (Gaussian/VWN-RPA variant, what Gaussian / PySCF / Psi4 mean by the bare name)
pbe0 / pbeh	hybrid	25 %	406
pw1pw	hybrid	20 %	weighted-sum
hse06	range-sep hybrid	25 % SR	425
vv10 (v0.14)	GGA + nonlocal	0 %	255
wb97x-v (v0.14)	range-sep + VV10	100 % LR	466
wb97m-v (v0.14)	range-sep meta-GGA + VV10	100 % LR	531
r2scan01 (v0.14)	meta-GGA	0 %	642 + 645
revdsd-pbep86 (v0.14)	double hybrid (D4)	69 %	101 + 132
b2plyp, dsd-pbep86, pwpb95, revdsd-pbep86 as run_job(method=…)	double hybrids	per recipe	a double hybrid is a method=, not a functional=: the runner refuses method="rks", functional="b2plyp" because that would run only the SCF half; open shells run ROKS plus RI-ROHF-MP2
skala-1.1 (introduced v0.15.158; periodic expansion v0.15.160; experimental)	neural XC (Microsoft SKALA-1.1, MIT-licensed checkpoint fetched on first use)	0 %	external provider, not libxc; needs the [skala] extra (PyTorch); energies only, fixed geometry; molecular RKS / UKS / ROKS plus selected periodic AO-grid routes (GDF, multi-k RIJCOSX, 3D BIPOLE, GPW/GAPW, and AICCM real-Gamma, literal-WSSC, and four-centre chi routes); forces, Hessians, TDDFT and response paths fail closed; see user_guide/skala.md

Plus everything libxc accepts directly + custom weighted-sum strings (PW1PW pattern: "0.2*HF + 0.8*GGA_X_PW91, GGA_C_PW91").

B3LYP convention: vibe-qc ships the ORCA definition, bare b3lyp is libxc id 475 (VWN5), the same flavor ORCA, TURBOMOLE, ADF, CRYSTAL, and CP2K mean by the name; Gaussian, libxc, PySCF, Psi4, NWChem, and Q-Chem mean the VWN-RPA variant (id 402), which vibe-qc spells b3lyp/g / b3lypg. The flavor gap is ~6.8 mHa on H₂/STO-3G, ~10-15 mHa per heavy atom. v0.12.0 added the explicit b3lyp5 / b3lypg spellings and a measured cross-code audit. See user_guide/functionals.md for the camp table, the measured values, and why there is no b3lyp3 alias.

Ewald summation (Phase 12e, mostly shipped)

See user_guide/ewald.md for the math

• API.

Sub-phase	Scope	Status
12e-a	Classical Ewald nuclear lattice sum (CoulombMethod.EWALD_3D)	✅
12e-b	erfc-screened nuclear attraction compute_nuclear_erfc_lattice	✅
12e-c-1	Gaussian-charge Ewald V(g) via grid integration	✅
12e-c-2	erfc-screened ERIs for Ewald short-range J/K	✅
12e-c-3a	FFTW3 build dep + solve_poisson_*	✅
12e-c-3b	build_j_long_range + auto_grid	✅
12e-c-3c	Quartet multipole far-pair replacement	Fail-closed research prototype
12e-c-4	End-to-end EWALD_3D SCF dispatch (RHF/RKS/UHF/UKS, Γ + multi-k)	✅

Madelung-constant validation:

Crystal	M reproduced	Reference
NaCl rocksalt	1.7475645946 to 1e-8	1.7475645946…
CsCl	1.762674773 to 1e-8	1.762674773…
ZnS zincblende	1.6380550533 to 1e-9	1.6380550533…
Simple-cubic jellium	1.4186487 to 1e-6	1.4186487 (Nijboer-De Wette 1957)

External data fetcher (vqfetch, v0.8.0)

Pulls structures + reference data from public databases on demand. Per-record provenance + license preserved.

Subcommand	Source	Default license
vqfetch optimade --formula MgO	OPTIMADE federation	per-provider
vqfetch mp --id mp-1265	Materials Project	CC-BY 4.0
vqfetch cod --id 1011027	COD	CC0 / public domain
vqfetch canonical <slug>	(5 round-trip-verified slugs)	(per primary)
vqfetch reference --cas 7732-18-5	NIST CCCBDB	US Govt public domain

See user_guide/external_structures.md

• user_guide/reference_data.md. End-to-end compute-reference round-trip: MgO sto-3g RKS-LDA E = −950.4204308512 Ha (13 SCF iters).

Job queue (vq v0.5.6, v0.8.0)

SSH-backed remote job submission, installed separately from the independent vibe-queue project. Cgroup-v2 enforcement, web dashboard, pause/resume, multi-venv --branch routing. See user_guide/queue.md.

Capability	Since
Local queue + CLI + daemon + JobSpec v1 + systemd unit	v0.1.0
Cross-machine SSH submit + ~/.config/vq/config.toml + vq fetch	v0.2.0
Resource watchdog (--mem-mb, --wall-time-seconds, OOM_KILLED / STARVED / TIME_EXCEEDED)	v0.3.0
Cgroup-v2 enforcement (systemd-run scope), pgid-based daemon recovery	v0.4.0
ABORTED_BY_QUEUE terminal state	v0.4.1
Read-only web dashboard (FastAPI + htmx)	v0.5.0
vq pause / vq resume + bearer-token auth + write API	v0.5.1
--all flag on pause/resume + queue-wide POST /api/v1/queue/*	v0.5.2
CRYSTAL14 + PROPERTIES14 reachable through dispatched jobs	v0.5.3
Parallel CRYSTAL14 (Pcrystal/Pproperties default --np 14) + clean systemd PATH drop-in	v0.5.4-5
Multi-venv routing via --branch for vibe-qc dev/release dispatch	v0.5.6

Properties

Property	API	Methods	Notes
Mulliken charges	mulliken_charges	RHF / UHF / RKS / UKS	atomic populations
IAO charges and IBOs	run_job(iao_analysis=True, localize="ibo"); vibeqc.iao.iao_charges, build_iaos, analyse_ibo	RHF / UHF / RKS / UKS	intrinsic atomic orbitals (Knizia 2013), basis-set-stable charges and spin populations, intrinsic bond orbitals with three bonding criteria; see IAO analysis
Löwdin charges	loewdin_charges	all	symmetric-orth populations
Hirshfeld charges	hirshfeld_charges -> HirshfeldResult	all	Classical Hirshfeld (1977) real-space partition against free-atom reference densities, so far less basis-sensitive than the two above. Iterative Hirshfeld is not implemented
Mayer bond orders	mayer_bond_orders	all	per-pair indices
Dipole moment	dipole_moment	all	with optional origin
Natural orbitals	natural_orbitals	all	for post-SCF analysis
Static polarizability α	dipole_polarizability_rhf	RHF (closed-shell)	via CPHF, Phase 17b-1; UHF/KS roadmap
Vibrational frequencies (FD)	compute_hessian_fd + HessianResult.frequencies_cm1	all	FD on analytic gradient; trans/rot projection
Vibrational frequencies (analytic)	compute_hessian_rhf_analytic, compute_hessian_uhf_analytic, compute_hessian_rks_analytic, compute_hessian_uks_analytic	RHF / UHF / RKS / UKS	CPHF/CPKS + libint deriv_order=2, with FD-on-Fock h1ao; same HessianResult shape as compute_hessian_fd, so frequencies, IR intensities and thermochemistry consume either unchanged. The RHF Hessian is validated against PySCF; the RKS/UKS Hessians use the LDA exchange-correlation kernel only for the response, so GGA and hybrid functionals differ from finite differences by up to about 12 cm⁻¹ (PBE 3.4e-3, PBE0 5.0e-3 Ha/bohr² on water) and LDA agrees to 1e-5
Γ-point phonons (periodic)	vibeqc.PhononCalculator, compute_dynamical_matrix_fd, phonon_eigenvalues, phonon_dos	GAPW periodic SCF	Γ only (k = [1,1,1]); multi-k phonons are not wired. Finite displacement on the analytic GAPW atomic force, not on the energy. The q-path helper returns a gamma_flat surface: the cached Γ frequencies repeated at each q, not a force-constant Fourier interpolation
IR intensities	ir_intensities	all	dipole derivatives along normal modes
Thermochemistry (ZPE / U / H / S / G / Cv)	compute_thermochemistry	all	rigid-rotor + harmonic-oscillator + ideal-gas
Relaxed coordinate scans	relaxed_scan, relaxed_scan_2d	HF / DFT / semiempirical	one or two constrained internal coordinates with the rest relaxed; see relaxed scans
Nudged elastic band	run_neb (ASE NEB on VibeQC calculators also works)	RHF / UHF / ROHF / RKS / UKS / ROKS / MACE / semiempirical / SECCM	climbing-image NEB with warm starts; periodic ROHF/ROKS gated; see NEB
Dimer and eigenvector-following transition-state search	run_dimer; run_job(geom_target="transition_state", geom_opt="dimer" | "ef" | "prfo")	molecular mean field / MACE	minimum-mode following without a Hessian (dimer) or with Bofill-updated approximate Hessians (EF, P-RFO); periodic ROHF/ROKS, ECP and +U gated
Molecular dynamics	vibeqc.md.run_md, run_nve, run_nvt (Berendsen, Nosé-Hoover); the top-level vibeqc.run_md family is the GAPW periodic helper	any surface with forces	velocity Verlet; see molecular dynamics
Well-tempered metadynamics	vibeqc.metadynamics.run_metadynamics with DistanceCV / AngleCV	any surface with forces	Laio-Parrinello bias on the MD driver, well-tempered variant
Minimum-energy conical intersection	vibeqc.conical.optimize_conical_intersection, make_cis_meci_fn; vibeqc meci CLI	HF-CIS (finite-difference gradients) / MSINDO CIS	Levine-Coe-Martínez penalty function; needs only two state energies and gradients
Cohesive energies	the energy-engine abstraction in cohesive energies	periodic + molecular references	cohesive energy per formula unit against free-atom references at the same level
Intrinsic reaction coordinate	run_irc	RHF / UHF / ROHF / RKS / UKS / ROKS / MACE	Molecular only; periodic IRC is a follow-up and fails closed. Mass-weighted damped steepest descent (Ishida-Morokuma-Komornicki 1977), not the Gonzalez-Schlegel constrained integrator. ROKS needs an explicit transition_mode until an analytic ROKS gradient exists for the FD Hessian
D3(BJ) dispersion	compute_d3bj	all	Grimme D3 with Becke-Johnson damping
D4 dispersion (v0.8.0)	compute_d4	all	Caldeweyher-Bannwarth-Grimme 2019
Cube file output	write_cube_density, write_cube_mo, write_cube_mos; periodic write_cube_density_periodic, write_cube_mo_periodic	all	Density, single orbital, or a multi-volume MO stack. Write only: vibe-qc has no cube reader
Molden file export	write_molden	RHF / UHF / RKS / UKS	verified by PySCF round-trip
Periodic band structure	band_structure	periodic SCF	k-path sampling
Periodic density of states	density_of_states	periodic SCF	Gaussian broadening on a sampled mesh
Equation-of-state fit	vibeqc.eos.fit_birch_murnaghan, .birch_murnaghan_energy	any energy-vs-volume scan	Third-order Birch-Murnaghan only (BM3_MODEL); no Vinet, Murnaghan or Rose-Vinet form. Fits an energy/volume series the caller supplies; it runs no scan of its own. Quality gates (min_energy_span_hartree, max_rmse_to_span) refuse a nearly flat curve rather than report an interior stationary point as a physical minimum, returning ill_conditioned_fit
Atomisation energy	run_job(..., atomization=True) (vibeqc.atomization.atomization_energy)	HF / DFT	per-element free-atom references at the same level, cached; main-group H-Kr. MSINDO reports its binding energy by default.
Green’s function IP / EA	run_job(method="ovgf") → result.ovgf; MSINDO: msindo_ovgf	HF / MSINDO	quasiparticle ionisation potentials and electron affinities (.ip_homo_ev / .ea_lumo_ev on the MSINDO result)
MPI substrate (v0.12.0)	mpi4py optional extra; vibeqc.mpi helpers	experimental GPW z-slab overlay plus low-level collectives	install with [mpi] extra. Production distributed HF / DFT / MP2 drivers are not wired yet; see MPI Parallelization. The GPW z-slab grid overlay is experimental (multi-rank Hartree-J build broken through v0.12.0, fixed post-release; pending real multi-rank validation)
COOP/COHP (v0.15.0)	compute_coop_cohp	periodic SCF	energy-resolved Crystal Orbital Overlap/Hamilton Population; C++ kernels, QVF dos.coop/dos.cohp, plotters, vibeqc coop CLI
Periodic Mayer bond orders (v0.15.0)	periodic_mayer_bond_orders	periodic SCF	k-space generalisation; QVF bond_orders
QTAIM (v0.15.0)	vibeqc.qtaim.qtaim_analysis	all	critical-point search + bond-path tracing; analytic C++ Hessian; QVF topology.qtaim
Fat bands (v0.15.0)	band_structure_projected	periodic SCF	Mulliken-projected band weights; QVF bands projections
Wiberg bond indices + delocalization index	vibeqc.bond_analysis.wiberg_bond_orders / .delocalization_index / .bond_order_summary	all	Löwdin-basis Wiberg 1968 index; AO-approximated DI; auto-emitted in .population.{txt,json}
NPA charges	vibeqc.nbo.npa_charges / .nbo_charges	not implemented	Raises NotImplementedError until occupancy-weighted Natural Atomic Orbitals are implemented. Population sidecars retain an empty npa field and report unavailable.npa; no NPA calculation is attempted.
Provisional NBO search	vibeqc.nbo.nbo_search / .donor_acceptor_analysis	exploratory Python API	Simplified orbital classification and E(2) donor-acceptor analysis; not a complete NAO/NBO implementation. See the bond analysis guide.
Energy decomposition analysis	vibeqc.eda.eda_lmo / .eda_morokuma	closed-shell RHF	LMO-EDA (Su-Li 2009) and Kitaura-Morokuma two-fragment decomposition of the interaction energy into electrostatic, exchange, Pauli repulsion and polarisation + charge transfer, fragments in the dimer basis through ghost atoms; refuses Kohn-Sham, ECP and open-shell inputs. See the bond analysis guide
Orbital entanglement	vibeqc.entanglement.entanglement_from_density / .single_orbital_entropy / .mutual_information	all	single-orbital entropy, mutual information, entanglement bond orders, correlation clusters

Solvation (CPCM / COSMO v0.9.0, COSMO-RS v0.15.158)

See user_guide/solvation.md.

Capability	API	Methods	Status	Notes
CPCM / COSMO implicit solvation	run_job(..., solvent=...) via vibeqc.solvation.run_cpcm_scf	RHF / UHF / RKS / UKS; MSINDO through its own COSMO driver	production	macro-iteration on a union-of-spheres cavity; presets in vibeqc.solvation.presets (SOLVENT_PRESETS) or a custom SolventModel; epsilon=inf is the exact conductor limit (#548)
Solvent analytic gradient / geometry optimisation in solvent	vibeqc.solvation.cpcm_gradient (finite-difference check: vibeqc.solvation.cpcm_gradient_fd)	RHF / UHF / RKS / UKS	production	the screening factor is carried from the run’s own variant (#546, #548, #549)
COSMO-RS / COSMOSPACE thermodynamics	run_job(..., cosmors=CosmoRSOptions(solvent="water")) or CosmoRSOptions(sigma_potential=...); vibeqc.solvation.cosmors on a vibeqc.solvation.surface.ConductorSurface	rhf / uhf / rks / uks solutes as ideal conductors on the COSMO FINE cavity	experimental (v0.15.158; job route v0.17.14)	sigma profiles and potentials, chemical potentials, activity and partition coefficients, solvation free energies (Klamt 1995; Klamt, Jonas, Bürger, Lohrenz 1998; klamt1998 set); solvents by name from a library of gas-phase BPW91/def2-SVP geometries, computed at the solute’s protocol and cached per user; ring atoms perceived on request (n_ring_atoms="auto"); the shipped cavity differs from the one the 1998 constants were fitted on, validated on pure-liquid vapour pressures only, FreeSolv validation pending
Direct COSMO-RS	SolventModel(variant="dcosmo-rs", sigma_potential=...)	RHF / UHF / RKS / UKS	experimental	the solvent’s sigma potential fed back into the SCF (Sinnecker et al. 2006) on the COSMO FINE cavity at epsilon=inf; analytic nuclear gradient and geometry optimization with a smoothed hydrogen-bond term (Parameterization.hb_smoothing > 0), refused with the published corner
Solvent with correlated, CAS, semiempirical (except MSINDO) or MLIP methods	solvent=	mp2 / cc / dlpno / cas / gfn2_xtb / pm6 / mace, …	gated (refuses with ValueError)	no validated post-SCF-in-solvent composition yet; ROHF / ROKS raise NotImplementedError

Input / output

Capability	API	Notes
High-level “run-a-job” driver	run_job / run_periodic_job	dispatches to right SCF driver, writes .out + .molden, runs BFGS if asked
QCSchema interchange	run_qcschema, read_qcschema, write_qcschema, molecule_from_qcschema, molecule_to_qcschema	MolSSI QCSchema molecular input and atomic-result JSON; see QCSchema
MPI substrate (v0.12.0)	pip install -e '.[mpi]'	optional mpi4py extra; low-level collectives plus experimental GPW grid overlay. Production MPI strategy is tracked in MPI Parallelization
Formatted SCF log	format_scf_trace, log_scf_trace	banner, iteration trace, energy components, orbital table, HOMO-LUMO gap
Molden export	write_molden	verified by PySCF round-trip
TREXIO export / import	run_job(trexio=True), run_periodic_job(trexio=True), write_trexio / read_trexio, field APIs	Optional [trexio] extra; HDF5 and text. Molecular and periodic Gaussian SCF wavefunctions, ECPs, complex k/spin blocks, molecular one-particle RDMs, optional MO operators/AO ERIs, and file-seeded READ. General field APIs transport all installed-library groups, including sparse integrals, determinants, CSFs, higher RDMs and numerical bases. CASCI/CASSCF/FCI determinant exports are available; other correlated payloads use the field API. Native non-Gaussian/Cartesian basis reconstruction is not provided; see TREXIO
Geometry trajectory	run_job(..., optimize=True)	ASE .traj file
XYZ / Extended-XYZ load	Molecule.from_xyz / from_xyz	molecular XYZ by default; periodic=True reads Extended-XYZ Lattice / pbc into PeriodicSystem; symmetrise=True returns SymmetriseResult
POSCAR load / save	read_poscar, write_poscar	VASP 5 format
CIF load / save	read_cif, Crystal.from_cif; writing goes through vibeqc.output (vibeqc.output.formats.cif.write_cif, reached by the geometry output role)	cell parameters + atom-site loops; symmetry operations expand asymmetric-unit sites
Cube save	write_cube_density, write_cube_mo, write_cube_mos (periodic siblings with a _periodic suffix)	volumetric data, write only; there is no cube reader
BXSF (band structure)	write_bxsf	XCrySDen 3D Fermi-surface format

Basis sets

• 255 bundled Gaussian .g94 basis files in the runtime library: 255 QVF-Basis records under basis_library/qvf/, the source of truth, rendered into the .g94 / .ecp text libint and libecpint read; 93 vendor records (libint’s bundle and the pob sets) and 162 Basis Set Exchange records, each with provenance and citation routes.

• vibe-qc’s own sets: pob-TZVP, pob-DZVP-rev2, pob-TZVP-rev2 (Bredow-group periodic-tuned), plus the release-paper reproduction basis 6-311+g3df2p.

• CRYSTAL-format parser, imports arbitrary CRYSTAL per-element basis files; converts to libint-compatible .g94. CRYSTAL INPUT ECP-block parsing is available for basis/ECP migration work.

• ECP basis sets, fully wired through libecpint (see ECP section above).

• Basis-set optimisation (v0.14), re-optimise a basis set’s exponents and coefficients against a molecular energy via the analytic-gradient BDIIS recipe (optimize_molecular_basis); see user_guide/basis_optimization.md.

• On-demand Basis Set Exchange rendering (v0.17.1), vibeqc.basis_fetch.fetch_basis_from_bse renders a BSE record the bundled library does not carry into a per-user cache as .g94 plus an .ecp sidecar, with a provenance file beside it. Opt-in only: run_job(fetch_from_bse=True) or $VIBEQC_FETCH_BSE, and molecular run_job only, run_periodic_job is not wired to it. Needs the optional [bse] extra (the catalogue ships as package data, so a lookup is a local read and works offline; what you get is pinned by the package version, recorded in every provenance file). A bundled name is a no-op, and a bundled basis merely missing one of the molecule’s elements stays a coverage refusal rather than being patched from BSE.

• Basis inventory and attribution documented in docs/license.md, including the libint-inherited files, the Bredow-group pob-* bases, and the BSE-derived custom overlay.

Crystal / lattice infrastructure

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

• CIF, POSCAR, and periodic Extended-XYZ input, with opt-in symmetrise=True standardisation into a calculation-ready PeriodicSystem.

• Monkhorst-Pack k-mesh generation with IBZ reduction, density-based auto-mesh helpers, and on-the-fly generalized regular grids via KPoints.optimal.

• Standardise-then-compress helpers for one-electron lattice integrals: compute_overlap_lattice_symmetrised_with_orbits plus kinetic / nuclear siblings rebuild the basis on the cleaned cell and return the SYM3a orbit-compressed storage view.

• 1D / 2D / 3D PeriodicSystem geometry.

• Hexagonal and other skew-cell geometry support; the native EWALD_3D Hartree path uses the analytical AO-pair FT on the full reciprocal metric.

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 / RI-MP2 / SCS-MP2 / SOS-MP2 / B2PLYP vs PySCF	55	1e-9 Ha (MP2 family) / 1e-7 Ha (B2PLYP)
Gradients (HF, DFT) vs FD / PySCF	17	1e-6 Ha/bohr
FD Hessian skeleton vs FD-on-gradient	8	1e-4 to 1e-3
Analytic RHF Hessian vs PySCF analytic	8	2.5e-9 Ha/bohr², freqs <0.01 cm⁻¹
FD Hessian + IR + thermochemistry vs PySCF	51	1e-7 Ha
CPHF + polarizability vs PySCF FD	11	1e-5 a.u.
Periodic machinery (invariants, molecular limit, Bloch folding)	76	machine precision
Periodic SCF convergence aids	27	per-test scoped
Periodic native GDF parity (Γ and multi-k, v0.11.0)	LiH FCC, MgO	µHa vs out-of-process PySCF (MgO/STO-3G Γ RHF: ΔE = -6.8e-12 Ha); CRYSTAL comparisons out of process
Multi-k KRHF/KRKS native parity (v0.11.0)	LiH FCC	µHa vs out-of-process PySCF; no in-process PySCF backend
JKBuilder + RIJCOSX vs ORCA 6.1.1 (v0.8.0)	curated	RIJCOSX SCF energy max |Δ| = 0.26 mHa on glycine def2-TZVP (energy only; gradient parity tracked separately, see § Density fitting + RIJCOSX)
vqfetch acceptance harness (v0.8.0)	5 structures + 8 molecules	round-trip end-to-end
Crystal / POSCAR / basis parsing	18	-
Bundled basis-library integrity	255 .g94 files + 44 ECP sidecars	generated QVF sidecars, provenance / citation coverage, ECP split, privacy checks

Plus the full regression suite at examples/regression/ (since v0.7.2 Boys’ Crucible) covers the cross-code parity matrix (vibe-qc vs PySCF vs ORCA) on a curated S22 / X23 / pob-TZVP test set.

Full pytest suite: ~1800 test functions across ~190 files covering the molecular, periodic, wavefunction-solver, and output stacks; runs in ~3-4 minutes on an M-class MacBook (faster with pytest -n auto).

Known issues

See troubleshooting and the issue tracker for the live open-bugs list. Highlights:

• DFTB0 and SECCM-DFTB0 absolute energies are not chemistry while the repulsive term is an A/R¹² placeholder (issue #306 on the archived monorepo tracker, which is where the code comments cite it); the semiempirical matrix above therefore labels DFTB0 and SCC-DFTB screening/preoptimization rather than production. Fixed-geometry differences stay valid, absolute energies and EOS fits do not.

• Periodic Γ-point GFN2-xTB derivatives are derivatives of the periodic energy. Since v0.16.0 the native gradient and all nine stress components match gapped central differences, and no analytic stress is gated off; TestPeriodicStressGFN2 pins all nine components at atol=2e-5. Issue #141 stays open for the independent wrong-answer check that a fix to a wrong number needs, not for further implementation. What is still open for the route is faithful periodic AES electrostatics, external xtb parity and multi-k, which is why the matrix above keeps periodic GFN2 experimental.

• Periodic GDF parity, Γ-only and multi-k µHa parity vs PySCF on LiH FCC shipped at v0.11.0 (Sun’s Stingray). Mixed Density Fitting (MDF) shipped at v0.13.0, closing the all-electron GDF accuracy floor (Sun-Berkelbach 2017) for MgO. Dense-crystal periodic-DFT XC error fixed in v0.15.0 (cross-cell density + SCF convergence).

• Multi-k dense Lpq RAM ceiling (multi-TB on production systems). Streaming-Lpq tracks with the periodic GDF work.

• Si-diamond / C-diamond RKS/PBE oscillation on RI path – SCF-settings investigation ongoing (damping/DIIS interplay, not algorithmic). Density mixers (Anderson/Broyden) shipped in v0.14.0 as an alternative convergence path.

• BIPOLE gradients: production forces finite-difference the supported exact SCF energy. Corrected- and legacy-gauge analytic implementations remain gated previews with a narrower maintained regression envelope; they are not evidence that the fail-closed quartet multipole gradient is integrated.

• Open-shell UHF / UKS analytic gradient on f-shells with two or more different second-row elements still disagrees with ORCA. Closed-shell direct gradients now auto-route through the DF gradient path; the open-shell auto-route is queued.

• Initial guesses can reach different stationary states or fail to converge. AUTO uses PATOM for ordinary restricted closed-shell molecules and SAD for ordinary periodic routes. Singlet UHF/UKS follows the unrestricted SAD policy; equal spin counts do not select the restricted PATOM route. Compare supported explicit seeds when diagnosing a difficult case; see the conservative AUTO policy.

• pob-TZVP / pob-TZVP-rev2 missing Ne entry.

• Heavy-atom basis-load test OOMs a 16 GB laptop in batch mode; route via vq submit to a larger box.

• ωB97X / ωB97X-D UKS on orbital-near-degenerate radicals needs a level shift to converge.

• 3c composite methods are calibrated on molecular systems; the periodic flavours need recalibration before being trusted.

The B3LYP local-correlation convention (bare b3lyp = VWN5 / ORCA definition since v0.8.0; explicit b3lyp5 / b3lypg spellings since v0.12.0) is documented in user_guide/functionals.md as informational, not a bug, both flavors are correct B3LYPs; codes differ in which one the bare name means.

---

[1]

Γ-CCM and χ-CCM are distinct CCM constructions. Under ruling R1, Γ-CCM is the neutral finite-torus construction with two admissible producers (real-gamma, neutral-bloch); the four-center variant is the 2014 union-and-weight/Wigner-Seitz integral-weighting lineage, a different construction kept in the same development line; χ-CCM uses finite-translation-group characters. A declared common exchange-q=0 convention is a comparison constraint, not an identification of the constructions, and equality for a specified operator and route must be established rather than inferred from the labels.