Pick a system class. Pick a molecule. The compiler returns an active-space rationale, an ansatz, a Z2-tapered qubit count, and a CASSCF or NEVPT2 baseline within stated tolerance. The output is a signed manifest your chemist, your quantum team, and your auditor can each read in their own register.
Single-reference methods near exact. Hartree-Fock plus CCSD(T) for absolute energies.
Highlighted cells are elements in DQ tested systems. Click a system-class chip to filter.
Each class has a discriminator the chemistry team would actually pick. The active space, basis, and classical baseline are recorded in the manifest as the contract between the compiler and the buyer.
Discriminator at the static-correlation regime where single-reference CCSD(T) accuracy degrades. CASSCF and NEVPT2 are essentially exact at this size. The run exercises the regime; it does not claim quantum advantage on H4.
Open-shell, transition-metal, broken-symmetry. AVAS recovers the canonical 10-orbital / 14-electron active space per the standard literature reference. Stretch goal at higher hardware budget.
Hamiltonian-Variational ansatz preserves particle-number and spin symmetry. The reference is DMRG, the gold standard for one-dimensional strongly-correlated systems.
Bravyi-Kitaev mapping over the transition-metal oxide, AVAS over the 3d manifold, k-point grid 4 by 4 by 2, supercell 2 by 2 by 1, frozen core for the transition-metal core electrons.
A molecular-orbital diagram for the active class. Each rung is one spatial orbital. Electrons drop into the rungs by occupancy. The boxed window is the active space the compiler projects the full Hamiltonian onto before fermion-to-qubit mapping.
Selection method recorded in the manifest. AVAS for transition metals, natural-orbital occupancy for closed-shell, DMRG-NO seeds for strongly-correlated fragments, HOMO-LUMO window for triage.
Each spatial orbital maps to two spin orbitals under the standard fermion-to-qubit transformation. Z2 symmetry tapering removes redundant qubits along particle-number, spin-parity, and point-group operations. The budget below updates with the active class.
Closed-shell baseline. Stretched H4 at R = 2.0 angstrom, cc-pVDZ basis, 4-orbital / 4-electron active space. 8 mapped qubits reduce to 6 after Z2 tapering against particle number, spin parity, and the D2h point group.
The full pipeline on the molecular discriminator. Every field below is what the manifest records.
The first ship-gate discriminator is the static-correlation regime where single-reference CCSD(T) accuracy degrades, not H2 where the classical method is exact. The compiler does not claim quantum advantage on stretched H4. CASSCF plus NEVPT2 on H4 at R = 2.0 angstrom is essentially exact at this size. The discriminator exists to exercise the regime, not to claim a win. The H2 and LiH trivial pass is explicitly rejected as ship evidence: passing on H2 demonstrates only that the compiler emits a syntactically valid circuit, not the chemistry-tool value proposition.
Eight blocks, one signed .qapp archive. The customer card translates the technical record into language a procurement lead reads without an internal interpreter; the manifest is the audit-trail layer.
Selection method recorded with citation. AVAS for transition metals, natural-orbital occupancy for closed-shell molecules, DMRG-NO seeds for strongly-correlated fragments, HOMO-LUMO window for triage. The buyer chemistry team reads each line and dissents on any field.
Fermion-to-qubit map chosen by symmetry-cost analysis (Jordan-Wigner, Bravyi-Kitaev, parity, or ternary tree). Z2 symmetry sectors fixed against particle number, spin parity, and point-group operations. Reduction recorded.
UCCSD, k-UpCCGSD, ADAPT-VQE seed pool, or Hamiltonian-Variational for periodic targets. Particle number, spin S squared, and point group preserved through optimisation. OpenQASM 3 emitted.
Dispatcher keyed off system_class. CCSD(T) for weakly-correlated, CASSCF plus NEVPT2 for static correlation or multireference, DMRG for quasi-1D, AFQMC plus DMET against DFT-PBE or HSE06 for periodic-3D. Wrong baselines cannot be selected silently.
Vendor, backend, job ID, vendor log URL, shot count, calibration timestamp, T1 and T2 at run time, transpiler passes, basis gates, dynamic decoupling sequence, mitigation primitive, measured estimator with bias and variance bounds, statistical equivalence test result, RB certificate hash.
For battery cathodes, the predicted Li-intercalation voltage with 95% confidence interval, delta to the in-house DFT-PBE or HSE06 reference in meV, recommended next experiment. Pass criterion 25-50 meV against the customer reference.
Did_demonstrate and did_not_demonstrate written verbatim inside the manifest. Marketing alignment recorded in the same block. No quantum-advantage claim on the discriminator. No H2 or LiH trivial pass.
Canonical SHA-256 over the full manifest. Vendor-API-verified job ID an auditor can replay. The chemistry pipeline replays end-to-end from a signed .qapp archive.
Stretched H4 at R = 2.0 angstrom exercises the static-correlation regime where single-reference CCSD(T) accuracy degrades. CASSCF and NEVPT2 are essentially exact at this size. The discriminator tests the chemistry compilation pipeline against a regime the buyer cares about, with a classical reference that is genuinely informative. Battery cathode runs follow the same compilation path with the periodic-3D handler and the in-house DFT-PBE or HSE06 reference.
CCSD(T) is exact for H2 and near-exact for LiH. A VQE pass on H2 demonstrates only that the compiler emits a syntactically valid circuit. It does not exercise active-space construction, symmetry-aware ansatz building, fermion-to-qubit mapping under symmetry constraints, or classical-baseline dispatch. The chemistry-tool value proposition is not tested.
The chemistry_config block records system_class, active_space_orbitals, active_space_electrons, frozen_core flag, selection_method, basis_set, point_group, symmetry_sectors_kept, fermion_qubit_map, classical_baseline_method, and classical_baseline_rationale. The hardware_run_evidence block records vendor, backend, job ID, shots, calibration timestamp, T1 and T2 at run time, transpiler passes, basis gates, mitigation primitive, and a statistical equivalence test. The buyer chemistry team reads every field and dissents on any line.
The pipeline runs at the cadence of a computational chemistry workflow, not a one-day turnaround. The discriminator completes inside a single ship cycle. A confidential cathode run takes longer because the active-space override, basis-set override, and hardware-budget cap are negotiated before submission. The point is reproducibility and a manifest a procurement team can sign on.