CRYSTAL23 capability matrix

This page is the explicit feature and configuration matrix of the PBC CRYSTAL23 parity programme (tracker issue #593). It lists the ground-state capabilities that the CRYSTAL23 User’s Manual, Section 1.2 “CRYSTAL23 Program Features”, and Erba et al., J. Chem. Theory Comput. 19, 6891 (2023), DOI 10.1021/acs.jctc.2c00958, advertise, and records for each one what vibe-qc has established on main. It is an inventory of evidence classes, not a parity claim: a row can only be promoted by the acceptance named in its “remaining gap” column, and a passed bounded example never promotes an entire method or backend.

The machine-readable source is benchmarks/crystal23_capability_matrix.json. The tables below are generated from it by scripts/crystal23_capability_matrix.py --write; --check fails when the page and the JSON disagree, and tests/test_crystal23_capability_matrix.py runs that check together with the schema rules. Edit the JSON, not the tables.

How to read the states

Every row carries exactly one state, the strongest evidence class achieved for the bounded scope named in its evidence column:

• numerically-accepted: a matched reference comparison passed on a recorded source and native runtime within a stated tolerance. The evidence names the system, reference, tolerance and runtime. This is the only state that counts toward parity, and only for the scope it names.

• source-tested: implemented and covered by focused repository tests of source functionality, refusals or internal identities, without a matched external reference.

• implemented: a public route exists and runs, with no focused test or reference comparison claimed here.

• unverified: implemented, but an open numerical defect, a failed or contradictory comparison, or a missing reference comparison is on record. The evidence column states the recorded number and its runtime.

• unsupported: absent or explicitly refused. A precise refusal is an acceptable release boundary; it is not parity.

Each row links the tracker issues that own its remaining gap. Required rows that stay unverified or unsupported are blockers for #593, not completed work. The narrower 0.18 release inventory lives in PBC 0.18 supported configurations and is not widened by this page.

Acceptance criteria shared by the rows

• Matched inputs. Lattice orientation, positions, charge and spin, primitive and contraction data, ECP, functional definition, grid, occupation objective, k coordinates and weights, integral precision, exchange zero-mode convention and SCF residual are matched and recorded (#573, #645, #647, #648). SHRINK dimensions alone do not match a mesh.

• Separated error sources. Basis, fitting or rank, image support, grid, BZ sampling, finite-size exchange and SCF termination errors are reported separately. Tolerances come from convergence ladders and reference evidence, never from one passing fixture.

• Recorded identity. Source commit, native build identity, inputs, measured results, reference provenance, errors and cost are recorded on the issue; operational receipts stay outside the product checkout.

• Performance claims need same-accuracy paired full-SCF timings with setup, final confirmation, memory and uncertainty (#574); kernel speedups do not establish time-to-solution parity.

Matrix

Updated 2026-10-24; 61 rows.

State	Rows	Meaning
numerically-accepted	5	A matched reference comparison passed on a recorded source/native runtime within a stated tolerance, for the bounded scope named in the evidence; not whole-method parity.
source-tested	21	Implemented and covered by focused repository tests of source functionality, refusals or internal identities; no matched external numerical reference.
implemented	7	A public route exists and runs; no focused repository test or reference comparison is claimed for this row.
unverified	15	Implemented, but an open numerical defect, a failed or contradictory comparison, or a missing reference comparison is on record.
unsupported	13	Absent or explicitly refused; a precise refusal is an acceptable release boundary, not a parity claim.

Hamiltonian and reference

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
rhf-3d	Restricted Hartree-Fock (RHF), 3D, SHRINK mesh	BIPOLE run_pbc_bipole_rhf / run_periodic_job(method='RHF', jk_method='bipole'); GDF and GPW routes	numerically-accepted	LiH/STO-3G RSGDF Gamma and 2x2x2 within 1.3e-9 and 4.2e-9 Ha of matched PySCF GDF (#179 notes 50997/51683, runtime 63031b08); MgO/STO-3G BIPOLE 2x2x2 within 0.075 mHa of PySCF exxdiv=’ewald’ at converged cutoff (docs/periodic_jk_routes.md, 2026-06-16, older build)	No CRYSTAL23 row at a matched mesh has passed through the #573 harness; the CRYSTAL comparison is exchange-convention confounded at coarse SHRINK and k-mesh confounded at SHRINK 8 8 (+4.7 mHa)	#573, #593
uhf-3d	Unrestricted Hartree-Fock (UHF)	BIPOLE run_pbc_bipole_uhf; Ewald multi-k UHF; GDF open shell (explicit jk_method='gdf')	numerically-accepted	Ewald UHF Cl2 anchors agree with a matched periodic direct-integral reference within 1.5e-9 Ha (#573 description, 6b5e102e); H2 triplet UKS/PBE0 2x1x1 within 3.2e-9 Ha of PySCF AFTDF	Magnetic-state transport and a CRYSTAL23 open-shell crystal row remain open (#647); no matched CRYSTAL23 UHF total energy yet	#647, #573
rohf-roks	Restricted open-shell HF (ROHF)	Ewald run_rohf_periodic_multi_k_ewald3d, GDF run_krohf_periodic_gdf, GPW Gamma; ROKS Ewald/GPW (maintained preview)	source-tested	Preview routes with pinned fixtures (docs/user_guide/periodic_methods.md section 3.8); Fock-unification migration open	No external reference comparison; GDF/GPW ROHF are previews; little-group and broken-symmetry handling open	#601, #597, #577
ghf-soc	Generalized HF / two-component spinors, spin-orbit coupling	none	unsupported	No two-component periodic route exists in python/vibeqc/	Outside the #593 ground-state goal; relativistic/noncollinear extensions are a separate roadmap milestone	#593
rks-semilocal	Kohn-Sham DFT, semilocal LDA/GGA, collinear spin	BIPOLE run_pbc_bipole_rks/uks; GDF; GPW/GAPW; libxc-routed functionals	unverified	#187: 12 of 12 pure-PBE crystals over-bound against CRYSTAL23 at matched SHRINK 2 2 by 1.5 to 35 mHa (v0.15.114); after the physical-density and XC-domain fixes the LiH13/PBE 2x2x2 fine-grid gap is -8.0 uHa on one fixture (note 43219); diamond Gamma PBE0 XC domain converged to 6.0e-7 Ha (#465, closed)	Re-measure the twelve #187 crystals at production defaults on current main with a >=4-cell grid ladder; #405 diffuse-basis XC image sum (3.8 mHa) open	#187, #405, #575
meta-gga	meta-GGA (tau-dependent): SCAN, r2SCAN, Minnesota families	libxc-routed tau functionals on the periodic KS routes (periodic Becke grids)	unverified	#179 compares LDA/PBE/r2SCAN XC increments against CRYSTAL23 SHRINK 2 and finds the increments consistent while the Coulomb block carries the offset; no dedicated meta-GGA acceptance row	Add an r2SCAN row to the #573 matrix; Minnesota functionals have no periodic acceptance evidence	#573, #179
noncollinear	Non-collinear spin DFT, spin-current DFT	none	unsupported	No noncollinear density or spin-current functional path	Separate roadmap milestone, not in #593	#593
global-hybrids	Global hybrids: B3LYP (VWN5), PBE0, PBESOL0, B1WC, WC1LYP, B97H, PW1PW	BIPOLE/GDF/GPW with functional='b3lyp' / 'pbe0' / 'pw1pw'/...; vq.define_functional for custom mixes	unverified	B3LYP VWN5 flavour verified against CRYSTAL14 to 1e-9 Ha on H2/STO-3G; diamond PBE0/def2-SVP Gamma GDF within 6.0e-6 Ha of PySCF Becke level 7 at converged cutoff (#163 note 40656); but #163 records a 440.8 mHa BIPOLE-versus-GDF Coulomb disagreement on diamond PBE0 Gamma on v0.15.133 that has not been re-measured	Re-run the #163 cross-route case on current main (BIPOLE cost was 35 h); add a matched CRYSTAL23 hybrid row	#163, #573
screened-hybrids	Screened-Coulomb hybrids: HSE06, HSEsol, SC-BLYP	functional='hse06' on BIPOLE, Ewald and slab routes (short-range exact exchange)	source-tested	Gamma image-shell erfc exchange gate re-derived in MR !280 (#501); multi-k short-range exchange pinned in the BIPOLE lanes	No matched CRYSTAL23 or PySCF HSE06 crystal comparison recorded; HSEsol/SC-BLYP untested	#501, #573
lc-hybrids-production	Long-range / middle-range corrected hybrids: LC-wPBE, HISS, wB97X, CAM-B3LYP	Corrected 3D physical-pair BIPOLE RKS/UKS implements single-range CAM/LC at Gamma/multi-k; GDF and multiple ranges remain refused	unverified	tests/test_periodic_screened_exchange.py: signed-kernel admission, independent fixed-density AFT exchange matrices and RKS/UKS limits; matched CRYSTAL crystal total and cost missing	Matched CRYSTAL totals and cost, support convergence, GDF and multiple-range HISS remain open (#657)	#657, #593
lc-hybrids-rsgapw	LC hybrids on the range-separated GAPW route	RSGAPW dedicated entry point, 3D Gamma RHF/RKS	implemented	docs/user_guide/periodic_methods.md section 4.7 lists wB97X / CAM-B3LYP on RSGAPW	Experimental route, no multi-k, no reference comparison	#593
sc-hybrids	Self-consistent global hybrids (sc-hyb)	none	unsupported	No dielectric-dependent exact-exchange fraction	Not a #593 deliverable	none
skala	(no CRYSTAL23 analogue) SKALA machine-learned functional	checkpoint-backed extension on the common periodic infrastructure	unverified	#83 runtime campaign; stub results and common-infrastructure timings are explicitly not checkpoint-backed acceptance (#593)	Complete RKS/UKS Gamma and multi-k calculations with the actual checkpoint, invariance and provider derivatives	#83

Dispersion and composite corrections

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
dft-d3	DFT-D3 (automated) and DFT-D2 London corrections	dispersion='d3' / 'd3bj' with compute_d3bj_periodic (built-in supercell sum or dftd3 backend)	source-tested	python/vibeqc/dispersion_periodic.py tests; D2 is not offered	No CRYSTAL23 D3 crystal comparison recorded; D2 unsupported	none
dft-d4	(CRYSTAL23 offers D3; D4 listed for completeness)	molecular D4 generic; periodic D4 not wired into compute_d3bj_periodic	unverified	#548 propagates generic D4 to molecular optimizers and the FD Hessian; periodic D4 has no entry	Decide whether periodic D4 is a deliverable	#548
gcp-3c	gCP counterpoise and composite 3c methods (HF-3c, PBEh-3c, HSE-3c, B97-3c, sol-3c)	vibeqc.composites registry and vibeqc.gcp (molecular)	unverified	Composite recipes exist for molecules; periodic 3c/sol-3c execution has no acceptance evidence	No periodic 3c row in the matrix; sol-3c variants absent	none

k sampling, occupations and SCF control

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
shrink-mesh	SHRINK Monkhorst-Pack net, anisotropic and shifted meshes	vq.monkhorst_pack(system, [n1,n2,n3]), explicit KPoints, IBZ reduction with full-mesh provenance	source-tested	#645 anisotropic CRYSTAL dimensions and printed-factor admission landed (46103c8b); #129 hexagonal meshes use multiples of 3 (756b5b0a)	Actual printed reciprocal coordinate and weight qualification against CRYSTAL output still open (#645)	#645, #129
gilat-net	SHRINK ISP Gilat net for Fermi energy and DOS	bz_integration='gilat' on the multi-k drivers (self-consistent for insulators, post-SCF for metals)	source-tested	docs/user_guide/from_crystal.md and metallic_bz_integration.md; pinned fixtures	Metallic total energies still come from smearing; no CRYSTAL23 metal comparison	#155
smear	SMEAR Fermi-Dirac smearing	smearing_temperature on BIPOLE/GDF/GPW single points	source-tested	Finite-temperature BIPOLE results documented; NOITER entropy from the evaluated density fixed (#649, 9bc34e25)	Graphene BZ ladder non-convergent (#129); metallic/magnetic rows are explicit #593 coverage rows	#129, #155, #649
spinlock-atomspin	SPINLOCK and ATOMSPIN	spinlock_mode/spinlock_value/spinlock_iterations and atomic_spins on BIPOLE UHF/UKS	source-tested	Spin-schedule recursion pinned in the BIPOLE UHF/UKS lanes	Genuine charge/spin/magnetic-state transport through the comparison matrix is #647	#647
guessp-restart	GUESSP density restart, NOITER evaluation	initial_guess=READ/read_from, QVF/Molden restarts, max_iter=0 zero-iteration evaluation	source-tested	#649 restricted and unrestricted multi-k NOITER corrected (4f541cb6, 9bc34e25); front-door restart refusals pinned (#656)	Restart from accepted source-consistent states and conserved charge (#472, #386) remain acceptance items	#649, #472, #386
convergence-aids	FMIXING, LEVSHIFT, DIIS, Broyden, Anderson	fock_mixing, level_shift, EDIIS+DIIS; Anderson/Broyden+Kerker density mixers on GDF and RIJCOSX only	implemented	Molecular and periodic convergence lanes; BIPOLE has no Anderson/Broyden mixer	Same-state convergence on difficult cases (#469, #497 acceptance); mixers absent on BIPOLE	#469, #497
convergence-thresholds	TOLDEE / TOLDEG / MAXCYCLE	conv_tol_energy, conv_tol_grad, max_iter; terminal exact-Fock density check	source-tested	Exact committed-density confirmation and density-closure gate (#187 note 43205, 184 tests)	none beyond the method rows	none

Electrostatics, exchange and truncation

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
coulomb-truncation	TOLINTEG T1/T2 Coulomb overlap and penetration thresholds	BIPOLE physical AO-pair support (cutoff_bohr), SR pair-interaction reach (sr_image_precision), fold-drift guard, and the #531 omitted-shell Coulomb estimate with opt-in sr_truncation_budget_ha	source-tested	Relabelling-invariant physical domain is the default (#316 migration); 100-bohr Ne molecular limit coefficient gate < 1e-8 at the shipped default; omitted-shell estimate bounds the measured 9.21e-8 Ha residual (this landing)	#316 overlap relabelling <= 1e-12 and #531 two-rung vq acceptance on a recorded runtime remain open; #44 native-family migration partial	#531, #316, #44
exchange-truncation	TOLINTEG T3/T4 exchange overlap and pseudo-overlap thresholds	Direct erfc exchange with Schwarz and charge-pair screening; reciprocal long-range exchange; no quartet bipolar replacement	source-tested	Screening thresholds recorded in the resolved domain; bipolar quartet far field fails closed	Measured direct J/K bottlenecks (#186, #490); no CRYSTAL-identical algorithm	#186, #490
exchange-zero-mode	(CRYSTAL applies no exchange G=0 correction)	exchange_exxdiv='ewald' probe-charge Madelung correction or 'none'	source-tested	Convention identity at fixed density documented; BIPOLE coarse mesh matches PySCF exxdiv=’ewald’ (0.075 mHa MgO)	Comparison with CRYSTAL totals only at k convergence; a CRYSTAL-style uncorrected finite-k mode is out of scope	#573
ewald-nuclear	Ewald treatment of nuclear and electron-nuclear electrostatics	3D Ewald V_ne and E_nn sharing alpha; analytic nuclear source radius on physical support	numerically-accepted	Point-charge limit bound per Ye and Berkelbach (2021); 1/L monopole gate on the Ne molecular limit; #457 face-nuclei defect fixed	#483 2-D/GAPW nuclear images around the origin needs verification; #462/#488 physical 1D/2D Coulomb	#483, #462, #488
gdf-coulomb	(no CRYSTAL analogue) Gaussian density fitting Coulomb tails	jk_method='gdf' RSGDF; fit_source='classic'; MDF	unverified	LiH matched within 1e-7 (#179); dense-core cells still carry a fitted-tail deficit (MgO -0.46 Ha Gamma, #159/#115 historical); #466 classic pairs physical default landed	Dense-cell default-policy accuracy, #275 component interpretation and #172 MDF definiteness open	#179, #115, #159, #466, #275, #172
xc-grid	DFT integration grid and image domain	Periodic Becke partition, product angular grids, XC density image domain (2 r_ao + span)	numerically-accepted	#465 diamond def2-SVP PBE0 within 6.0e-7 Ha of the converged 25-bohr domain on runtime f1ad7714; continuous-grid Becke steps fixed (!230)	#405 diffuse-basis image sum (3.8 mHa), #381 multi-k GDF Becke radius, #268 ROKS grid weights open	#465, #405, #381, #268
symmetry	Point and space-group symmetry exploitation	spglib space groups; SYM3b atom-pair-orbit Fock reduction; IBZ k reduction with full-mesh expansion	source-tested	SYM3b reconstruction bit-identical to the full build (docs/periodic_jk_routes.md); #466 IBZ classic-physical	#171 NaCl GDF reduction exceeds a 1225 GB guard; #470 little-group invariance at wedge representatives; #577 unified-Fock integration	#171, #470, #577, #469

Basis sets and pseudopotentials

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
gtf-am	Gaussian basis functions s, p, d, f, g	libint-backed integrals; derivative orders 0/1/2 built with max_am 5/4/4 by default	source-tested	scripts/build_libint.sh configuration; f functions exercised by def2 bases	g functions depend on the libint build; no periodic g-function fixture	none
pople-bases	Standard Pople STO-nG, 3-21G, 6-21G libraries	255 bundled .g94 files plus BSE overlays (BasisSet(mol, 'sto-3g'))	implemented	docs/user_guide/basis_sets.md	none	none
pob-library	Internal POB-TZVP library	Bundled pob-* family (H-Br) and pob-TZVP-rev2 QVF records	source-tested	Basis overlay receipts and catalog tests	Coverage beyond Br and f-in-valence lanthanide/actinide bases absent	none
basis-optimizer	Internal basis set optimizer	vibe-basis companion package (OPTBASIS-style optimiser)	implemented	docs/roadmap.md: shipped and moved to vibe-basis	Separate repository; not part of this parity matrix	none
ecp-gdf-ewald	Pseudopotentials: Hay-Wadt, Stuttgart-Cologne AREP libraries	ecp_library='lanl2dz' / 'ecpNNmdf' on k-GDF, Gamma RIJCOSX, Ewald open-shell and ROHF/GDF routes (#35)	numerically-accepted	Cl2/LANL2DZ UKS PBE Gamma and 3x1x1: molecular offset 6.0e-6 Ha and bond force versus Richardson FD 6.0e-9 Ha/bohr on runtime f1ad7714 (#557 note 53206); NaCl/LANL2DZ GDF RHF within 4.4e-6 Ha of mapped PySCF after the #652 projector-domain repair (note 52563)	#652 shared UHF driver acceptance and density-grid artifact; general ECP forces/stress (#579); these are same-code controls, not CRYSTAL force parity	#35, #652, #579
ecp-bipole	Pseudopotentials on the BIPOLE (CRYSTAL-family) route	Neutral 3D corrected physical-pair RHF/RKS/UHF/UKS scalar single points, effective ionic charges, valence populations and midpoint projectors	unverified	Cl2/LANL2DZ Gamma and 3x1x1 HF reference and raw Hcore controls; PBE0 convergence and open-shell guess checks (#658); retained operator identity	Broad bulk/reference and support-convergence acceptance, cost, symmetry, external XC and derivatives remain open; #652 failures are not superseded; completed #557 scope is unchanged	#658, #35, #593
sorep	Spin-orbit (SOREP) pseudopotentials	none	unsupported	No two-component route	Outside #593	none

Periodicity and geometry

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
dim-3d	3D crystals, 230 space groups	PeriodicSystem(dim=3), spglib symmetry	source-tested	All periodic lanes	none	none
dim-2d	2D slabs and surfaces, 80 layer groups	vq.slab_2d; AUTO selects SLAB_EWALD_2D; explicit closed-shell slab GDF with analytic gradients; BIPOLE refuses dim=2	source-tested	docs/user_guide/periodic_methods.md section 5.1	Physical 2D Coulomb/exchange on the BIPOLE family (#462/#488); layer groups not exploited	#462, #488
dim-1d	1D polymers, rod and helical symmetry, nanotubes	AUTO and GDF refuse dim=1; model as a dim=3 cell with transverse vacuum	unsupported	docs/user_guide/periodic_methods.md section 3.4; #462 physical wires	Physical 1D route (#462); helical symmetry and nanotube builders absent	#462
geometry-editing	Automated geometry editing: slab cuts, supercells, clusters, nanotubes	Python/ASE tooling (make_supercell, CIF/POSCAR I/O, vq.slab)	implemented	docs/user_guide/periodic_systems.md	No CRYSTAL-keyword equivalents; nanorod/nanotube/Wulff builders absent	none

Derivatives, relaxation and lattice dynamics

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
forces-analytic	Analytic nuclear gradients (HF and DFT, all-electron and ECP)	GDF closed-shell Gamma/multi-k/slab analytic gradients; BIPOLE analytic gradient gated preview; ECP UKS semi-numerical preview	unverified	#459 Gamma GDF RKS(PBE) analytic minus FD 3.8e-5 Ha/bohr; #482 multi-k KUKS 2.4e-6 Ha/bohr; Cl2 ECP preview within 6.0e-9 Ha/bohr of FD (#557)	Per-route derivative capability table (#578); independent force acceptance versus CRYSTAL/ORCA	#578, #579, #459, #482
forces-fd	(production BIPOLE forces)	Finite-difference forces and fixed-cell atomic relaxation on BIPOLE (optimize=True)	source-tested	Route-preserving FD optimizer lanes; freeze masks; periodic NEB	Same-objective references for structural relaxation (#578, #175)	#578, #175
stress-cell	Analytic cell gradients, full cell optimization, constant-pressure optimization	No analytic periodic stress; FD stress on GPW/GAPW and semiempirical; variable-cell BIPOLE/GDF fails closed	unsupported	docs/user_guide/periodic_methods.md section 5.5	Matched stress and cell relaxation is #578; MANTA stress is #637	#578, #637
ts-search	Transition-state search (periodic)	Periodic NEB on the BIPOLE force stack; no periodic TS optimizer	implemented	#564 NEB process workers	No periodic saddle-point optimizer	#564
frequencies	Harmonic Gamma frequencies, IR/Raman intensities, phonon dispersion, ADPs, VDOS	GAPW finite-displacement Gamma dynamical matrix (periodic_gapw_phonon.PhononCalculator); molecular Hessians	unverified	GAPW route only, on analytic GAPW forces; no BIPOLE/GDF periodic Hessian; no intensities	Phonons/QHA are a separate roadmap milestone, not #593	#593
anharmonic	Anharmonic X-H, VSCF/VCI	none	unsupported	none	Separate roadmap scope	none
elastic-eos-qha	Elastic tensor, equation of state, quasi-harmonic thermodynamics, piezoelectric and photoelastic tensors	vibeqc.eos fits an E(V) scan; no elastic, QHA or piezo workflows	unsupported	EOS fitting is tooling, not an automated workflow with qualified strains	Separate roadmap scope	none
cphf-dielectric	CPHF/CPKS dielectric and (hyper)polarizability up to fourth order	Molecular CPHF only	unsupported	vibeqc.cphf is molecular	Response spectroscopy is a separate roadmap milestone	none
solid-solutions	Solid-solution configuration tools	none	unsupported	none	Not planned	none

Properties and wave-function analysis

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
bands	Electronic band structure	vibeqc.bands, band_edges, QVF/plot output	unverified	#176: Si non-converged, 2D h-BN fails closed, negative indirect gap from per-k occupations; ‘ewald’ gauge default and recorded exchange q0 (!103)	Bands from accepted source-consistent states (#472, #176)	#176, #472
dos	Density of states, band- and AO-projected	pdos on GDF properties; rebuilt operator gated on the 1e-10 Hermitian contract (#547)	source-tested	#547 DOS gate and bond orders from the accepted density	Accepted-state consistency (#176/#472)	#547, #176
population	Mulliken population analysis, atomic multipoles	Mulliken and Mayer bond orders on every non-AICCM periodic route	source-tested	#547 bond orders from the accepted density (e3939146, 905f5f1d)	none beyond accepted-state items	#547
density-maps	Charge and spin density maps	Cube output and QVF BvK density artifact	unverified	#386: QVF primitive-cell grid integrates 11.98 instead of 10 electrons	Fix the BvK density artifact normalisation	#386
hirshfeld-periodic	Hirshfeld-I partitioning (periodic)	Molecular Hirshfeld only	unsupported	vibeqc.properties is molecular	Not planned in #593	none
qtaim	Topological analysis (TOPOND / QTAIM)	Molecular vibeqc.qtaim; MANTA QTAIM slice (#618)	unverified	#618 QTAIM capability slice c6b17b10 posted	Periodic QTAIM acceptance	#618
coop-cohp	Crystal orbital overlap/Hamiltonian populations	vibeqc.coop_cohp on GDF properties	source-tested	Pinned fixtures	Accepted-state gate as for DOS	#547
wannier	Localized Wannier functions (Boys)	periodic_localise Boys localisation for local correlation	implemented	Used by periodic DLPNO; not a properties-module LOCALWF workflow	No Wannier-based IR/polarization properties	none
other-properties	Electrostatic potential, EFG, Fermi contact, Mossbauer, Compton profiles, structure factors, transport, spontaneous polarization	none	unsupported	none	Properties suite is a separate roadmap milestone	none

Performance and parallelism

Row	CRYSTAL23 capability	vibe-qc route	State	Evidence	Remaining gap	Issues
openmp	OpenMP and MPI parallel SCF and gradients	OpenMP native kernels; MPI output-cell farming for the BIPOLE direct ERI build; fleet runtimes	unverified	#96: GDF thread efficiency saturates near 17 threads (0.27 at 64 cores)	Measured scaling recommendation (#96); production MPI is a 1.0 milestone	#96
time-to-solution	(parity target) Same-accuracy time-to-solution versus CRYSTAL23	#574 paired full-SCF benchmark manifest; #644 CRYSTAL engine-time extraction	unverified	No paired same-accuracy benchmark has been accepted; BIPOLE diamond PBE0 Gamma cost 35 h on 20 cores versus GDF 1.5 h (#163, old runtime)	Target ratio <= 1 per declared production case with uncertainty (#574)	#574, #644

Material gaps without a dedicated owner

Rows whose remaining gap has no bounded tracker issue of its own as of the update date:

• meta-gga, gcp-3c, dft-d4: no periodic acceptance rows exist for meta-GGA functionals, composite 3c methods or periodic D4.

Gaps that received their own issue on 2026-10-24: long-range corrected hybrids on the production routes (#657, row lc-hybrids-production) and pseudopotentials on the BIPOLE route (#658, row ecp-bipole).

Lattice dynamics, elastic and thermodynamic workflows, response properties, two-component and noncollinear Hamiltonians, and the properties suite are deliberately outside the #593 ground-state goal and carry their own roadmap milestones; their unsupported rows are listed for completeness only.

Updating the matrix

1. Edit benchmarks/crystal23_capability_matrix.json: change the row’s state, evidence, gap and issues, and bump updated.

2. Run python scripts/crystal23_capability_matrix.py --write.

3. Run python -m pytest tests/test_crystal23_capability_matrix.py -q.

4. Link the tracker note that carries the evidence from the issue named in the row. A promotion to numerically-accepted needs the recorded runtime identity and the matched reference in that note.