# EMPIRICAL.md — index of the empirical validation suite **This is the reader's jumping-in point for every empirical test in the corpus.** Each row names the Lean file, the headline theorem(s), the claim, and the original physics source. Every theorem listed here is **foundational-triple-only** (`propext, Classical.choice, Quot.sound`) and regression-pinned in [`CsdLean4/Tests/AxiomAudit.lean`](CsdLean4/Tests/AxiomAudit.lean) via `#guard_msgs` against `#print axioms`. Build with `lake build` (library) and `lake build CsdLeanTests` (the pins). The suite splits into the two branches described in [`README.md`](README.md) (Empirical section) and [`specs/qm-empirical-tests.md`](specs/qm-empirical-tests.md) (the roadmap and per-item status). For the axiom posture and the two-strata (operational vs ontic) reading, see [`AXIOMS.md`](AXIOMS.md). ## The two layers - **QM-validity layer** (`CsdLean4/Empirical/QM/`): pure theorems about Hilbert spaces, vectors, and operators. Inputs are abstract (`EuclideanSpace ℂ (Fin n)`, Pauli matrices, isometries); proofs are linear algebra and inner-product geometry. Answers "does the standard QM formalisation produce the predicted number?" - **CSD-ontic layer** (`CsdLean4/Empirical/CSD/`): the same numerical predictions read through the ontic substrate. Two sub-kinds: - **Bridge transports**: each QM-validity statement carried through a `CSDBridge.Context D` bundle (the LF2 `SectorData` + measure-bridge data), providing the structural slot for the ontic interpretation. - **Volume-frequency series**: the Born number is *derived*, not transported, as a Fubini-Study typicality volume on the ontic `Σ`. These are carving-free and Gleason-free (no appeal to `busch_effect_gleason`). Honest residue: the concrete `SectorData` instances behind these entries still carry `Φ = id` (LF5 exercises a genuine `Φ_vN ≠ id` only on the dilated `Σ'`, not in these instances), so these volume readings remain *realisation* not *derivation*; the Born value is derived from the posited sector geometry on `Σ`, the standing assumption these results rest on (see `README.md` headline and `AXIOMS.md`). --- ## QM-validity branch (`CsdLean4/Empirical/QM/`) | Test | File | Headline theorem(s) | Claim | Source | |---|---|---|---|---| | Bell / CHSH | `Bell.lean` | `chsh_singlet_tsirelson_bound`, `chsh_qm_tsirelson_bound`, `chsh_classical_bound_violated`, `chsh_inner_bound` | CHSH on the singlet attains the Tsirelson value `2√2`; the classical bound `2` is violated; Khalfin-Tsirelson upper bound `≤ 2√2` for any four unit vectors (and its QM form) | Bell 1964; CHSH 1969; Tsirelson 1980; Aspect 1982 | | No-signalling | `Bell.lean` | `no_signalling_alice`, `no_signalling_bob`, `singlet_marginal_alice`, `singlet_marginal_bob` | Each wing's marginal is independent of the other's setting; singlet marginals are `1/2` | Aspect 1982; loophole-free Hensen/Giustina/Shalm 2015 | | No-cloning | `NoCloning.lean` | `no_cloning_two_state`, `no_universal_cloner_of_witness` | An isometry cloning two states forces overlap `∈ {0,1}`; no universal cloner exists | Wootters-Zurek 1982; Dieks 1982 | | No-deleting | `NoDeleting.lean` | `no_deleting_two_state`, `no_universal_deleter_of_witness` | Dual of no-cloning: deleting forces overlap `∈ {0,1}`; no universal deleter | Pati-Braunstein 2000 | | No-broadcasting | `NoBroadcasting.lean` | `pure_marginal_confinement` | A bipartite PSD operator with a pure first-factor marginal is confined to that pure sector (the squeeze behind "broadcasting a pure state = cloning it"); the full BCFJS `iff` is the Cat-1 theorem `QuantumInfo.exists_channel_broadcasts_iff_commute` (`Mathlib/QuantumInfo/Broadcasting.lean`) | Barnum-Caves-Fuchs-Jozsa-Schumacher 1996 | | No-communication | `NoCommunication.lean` | `no_communication`, `no_communication_reduced`, `bob_expectation_invariant` | Alice's local unitary leaves every Bob-side expectation, and Bob's reduced state, invariant | Ghirardi-Rimini-Weber 1980; Eberhard 1978 | | Robertson uncertainty | `Uncertainty.lean` | `robertson_uncertainty`, `robertson_core` | `Var(A)·Var(B) ≥ ¼\|⟨[A,B]⟩\|²` for symmetric operators | Robertson 1929 | | Wigner–Araki–Yanase | `WignerArakiYanase.lean` | `arakiYanase_identity`, `wigner_araki_yanase`, `no_exact_record_of_not_commute`, `way_hypotheses_satisfiable`, `swap_exact_record_not_commute`, `sigmaX_no_exact_conserving_record` | A measurement isometry `U` on `HS ⊗ HA` conserving an additive `L = L_S ⊗ 1 + 1 ⊗ L_A` that records the eigenbasis of `A` **exactly** (orthogonal pointers, Yanase condition on the pointers *or* repeatable post-states) forces `[A, L_S] = 0`; hence no exact conserving record of `A` when `[A, L_S] ≠ 0`. Non-vacuous (CNOT records `σ_z` exactly while conserving `σ_z ⊗ 1 + 1 ⊗ σ_x`; its pointer fails Yanase, repeatability saves it) and sharp (SWAP records `σ_x` exactly while conserving `σ_z ⊗ 1 + 1 ⊗ σ_z` with `[σ_x, σ_z] ≠ 0` — both side conditions fail, so the side condition is load-bearing). The record-layer stroke sits *outside* these hypotheses (CSD twin `CSD/WignerArakiYanase.lean`, `no_joint_hilbert_map`); the LF5 coupling `vnUnitary` meets them and WAY's conclusion holds there (`specs/way-theorem-scoping.md`) | Wigner 1952; Araki–Yanase 1960; Yanase 1961; Ozawa 2002 (quantitative form: not formalised) | | Stern-Gerlach | `SternGerlach.lean` | `born_zPlus_zPlus`, `born_xPlus_zPlus`, `born_z_basis_complete`, `born_x_basis_complete` | Spin Born values on `\|+ẑ⟩`: `P(↑\|ẑ)=1`, `P(↑\|x̂)=1/2`; basis completeness | Stern-Gerlach 1922 | | Malus's law | `Malus.lean` | `malus_law`, `malus_basis_complete`, `malus_pi_div_two` | `P(+_z\|θ) = cos²(θ/2)`; recovers the two SG values at `θ=0, π/2` | Malus 1809 (spin-1/2 form) | | Berry phase (spin-½ cone) | `BerryPhase.lean` | `geometricPhase_coneCurve`, `geometricPhase_coneCurve_solidAngle`, `connectionForm_coneCurve` | The state `(cos θ/2, e^{it} sin θ/2)` precessing once around a cone of half-angle `θ` returns with geometric phase `−2π sin²(θ/2) = −π(1 − cos θ)`, minus half the cone's solid angle — Berry's phase for spin-½, computed from the general Aharonov–Anandan theory (`Mathlib/Analysis/InnerProductSpace/GeometricPhase.lean`: gauge invariance, horizontal-lift holonomy). No fibre holonomy is claimed on the CSD side (`specs/berry-phase-scoping.md` §5) | Berry 1984; Aharonov–Anandan 1987; Tomita–Chiao 1986 (measured) | | Hardy non-locality | `Hardy.lean` | `no_lhv_hardy`, `HardyQM.exists_hardy_realisation`, `HardyQMMax.hardyMax_probability_eq` | Hardy's ladder: a paradoxical outcome with no local-hidden-variable assignment; the optimal `≈ 9%` probability `(5√5−11)/2` | Hardy 1992, 1993 | | Unambiguous discrimination | `USD.lean` | `usd_unambiguous_1`, `usd_unambiguous_2`, `usd_success`, `usd_complete`, `usdPOVM` | Zero-error discrimination of two non-orthogonal states via a genuine non-projective POVM; success `1−s`; completeness `E₁+E₂+E?=I` | Ivanovic 1987; Dieks 1988; Peres 1988 | | Syndrome collapse (QEC) | `QEC/SyndromeCollapse.lean` | `errored_pairwise_orthogonal`, `spanError_logical`, `branch_overlap_X1`/`X2`/`X3`, `three_qubit_corrects_span_error` | The half that joins discretization to the four point-checks: the branches are mutually **orthogonal** (disjoint supports = the distinct syndromes), so the overlap with branch `k` extracts exactly `cₖ` and the measurement collapses onto one correctable branch. **The 3-qubit code corrects an arbitrary error in `span ℂ {I,X₁,X₂,X₃}`** — a continuum. Still the bit-flip span; all four Paulis needs Shor-9 | Shor 1995; Nielsen–Chuang §10.1 | | Error discretization (QEC) | `QEC/ErrorDiscretization.lean` | `pauli_decomposition`, `pauli_span_top`, `error_discretization_qubit₁`/`₂`/`₃`, `errored_codeword_eq` | Every single-qubit operator is `c₀I + c₁X + c₂Z + c₃XZ` (explicit coefficients), and `span ℂ {I,X,Z,XZ} = ⊤` — so an arbitrary error carries a codeword nowhere outside the span of the four *correctable* ones. Why correcting four discrete errors corrects a continuum. ⚠️ *Row corrected 2026-08-02: an earlier version said the syndrome-collapse half was open — it landed 2026-07-27 (`QEC/SyndromeCollapse.lean`, the row above).* Discretization + syndrome collapse both done; the concatenated Shor-9 claim remains the open item (`specs/BACKLOG.md`) | Shor 1995; Nielsen–Chuang §10.2 | | Quantum eraser (CSD twin) | `CSD/QuantumEraserVolume.lean` | `eraser_fringe_typicality`, `eraser_dark_typicality_zero`, `eraser_dark_record_null`, `eraser_dark_measurement_zero`, `eraser_dark_basin_null`, `eraserOut_rate_conditional`, `mk_eraserOut_pi` | The marker-conditioned fringe `(1 + c·cos φ)/2` as a fibre-typicality volume at **every** phase — and at `φ=π` the **dark fringe is an ontic impossibility**: the dark cell, the P5 record event, the measurement probability, *and* the v1.0 context-fixed basin are all exactly null (at the dark point the conditioned state *is* the vertex `[e₁]`). Conditioned rates derived as joint/marginal from the QM module. Stated via the **record layer** (like the HOM twin): the DH route's `hpos` excludes exactly these zeros. ⚠️ *Caveat discharged 2026-08-03*: the conditioning **process** is now realised — `CSD/EraserDynamics.lean` derives the conditioned states from the local Lüders strokes themselves (`erased_amp`: the dynamical post-states are `√2·eraserOut`; `marked_no_fringe`; `erased_weight`), and `CSD/EraserSequential.lean` adds `sequential_no_revival` — erasing after the mark record revives nothing: records are statistically irreversible | — | | Quantum eraser | `QuantumEraser.lean` | `eraser_joint`, `eraser_no_interference`, `eraser_fringe_bright`, `eraser_fringe_dark` | Complementarity + erasure: joint `P(a,c) = (1 + ac·cos φ)/4` — full-visibility fringe in the marker-conditioned statistics, yet the unconditioned system marginal is flat `1/2` (which-path information kills interference; erasing it restores fringes only in the conditioned subensembles; delayed choice = same statistics). ⚠️ *Row added 2026-08-02: the module landed earlier, wired and pinned, but this table row was missing* | Scully–Drühl 1982; Kim et al. 2000 | | Hong–Ou–Mandel (CSD twin) | `CSD/HongOuMandelVolume.lean` | `hom_coincidence_typicality_zero`, `hom_coincidence_record_null`, `hom_coincidence_measurement_zero`, `hom_bunch_typicality_half`, `homOut_eq_bsTwo_bosonIn` | The dip as an **ontic impossibility**: the coincidence outcome's typicality measure on `Σ` is exactly `0`, so no microstate yields a coincidence — nothing cancels, there is nothing there to cancel. Stated via the **record layer**, because the Duistermaat–Heckman route needs strictly positive Born weights (see `BACKLOG.md`) | — | | Hong–Ou–Mandel | `HongOuMandel.lean` | `bsTwo_bosonIn`, `hom_coincidence_zero`, `hom_bunching_one`, `distinct_coincidence_half`, `fermion_coincidence_one`, `hom_dip`, `hom_exchange_trichotomy` | Two identical bosons entering opposite ports of a 50:50 beamsplitter never exit separately (coincidence `0`); the same splitter gives `½` for distinguishable particles and `1` for fermions, so the coincidence rate reads out exchange symmetry alone. The effect is the single identity "`H·σₓ·H` is diagonal" | Hong–Ou–Mandel 1987 | | Minimum-error discrimination (Helstrom) | `Mathlib/QuantumInfo/Helstrom.lean` | `successProb_le`, `successProb_helstromTest`, `errorProb_ge`, `errorProb_helstromTest`, `successProbPrior_le`, `helstrom_indistinguishable`, `helstrom_perfect` | The optimal two-outcome test has `P_error = ½(1 − D(ρ₀,ρ₁))` — bound **and** attainment (at the Helstrom projector); general priors give `½(1 − ‖p₀ρ₀ − p₁ρ₁‖₁)`. The operational meaning of the trace distance; complements `USD.lean` (zero error, inconclusive outcome allowed) | Helstrom 1976; Holevo 1973 | | GHZ paradox | `Multipartite/GHZ.lean` | `ghz_expectation_xxx`, `ghz_expectation_xyy`/`yxy`/`yyx`, `no_lhv_assignment_for_ghz` | All-or-nothing 3-qubit correlations `⟨XXX⟩=+1`, `⟨XYY⟩=⟨YXY⟩=⟨YYX⟩=−1`, with no consistent LHV assignment | Greenberger-Horne-Zeilinger 1989; Mermin 1990; Pan et al. 2000 | | Kochen-Specker | `Contextuality/KS18.lean` | `no_value_assignment_18_9`, `ks_no_value_assignment_cabello18`, `cabello_pairwise_orthogonal_in_basis` | No non-contextual value assignment on the Cabello 18-vector configuration in `ℝ⁴` | Kochen-Specker 1967; Cabello 1996 | | Mermin-Peres square | `Contextuality/MerminPeres.lean` | `no_lhv_mermin_peres`, `mermin_peres_R0`..`C2` | The magic-square observables have no consistent local value assignment | Mermin 1990; Peres 1990 | | Superdense coding | `Resources/SuperdenseCoding.lean` | `encode_I`/`X`/`Z`/`XZ`, `bell_basis_orthonormal` | The four local-Pauli images of `\|Φ⁺⟩` are the four orthonormal Bell states | Bennett-Wiesner 1992 | | Teleportation | `Resources/Teleportation.lean` | `teleState_factorises`, `teleportation_bell_expansion`, `teleportation_branch_recovers_input` | The four corrections `{I,Z,X,ZX}` recover the input per Bell-measurement branch (branch-conditional QM-validity layer) | Bennett et al. 1993 | | B92 sequential (CSD twin) | `CSD/Crypto/B92Sequential.lean` | `b92_honest_false_click_null`, `b92_honest_false_click_null_z`, `b92_conclusive_basin_half`, `b92_eve_false_click`, `b92_eve_detectable` | **Unambiguity as a null basin**: an honest `\|+⟩` carrier has a zero-width conclusive-bit-0 arc (no microstate can produce a false click — the eraser-dark-fringe shape); Eve's Z-intercept (calibrated-swap dynamics) raises the false-click basin to exactly `½` whatever she recorded, strictly above the honest zero — eavesdropping announces itself in the conclusive statistics. Instantiates the `BB84Sequential` engine (stated, not hidden) | — | | Wiesner money sequential (CSD twin) | `CSD/Crypto/WiesnerSequential.lean` | `wiesner_honest_x_pass`, `wiesner_honest_z_pass`, `wiesner_forge_x_pass_half`, `wiesner_forge_x_caught_half`, `wiesner_forge_z_invisible`, `wiesner_rate_eq_verifyProb`, `wiesner_forge_pass_avg` | The counterfeit round as a genuine sequential measurement: honest verification's pass basin has **full** measure (`wiesner_rate_eq_verifyProb` ties the rate to the QM `verifyProb`); the forger's measure-resend on a conjugate-basis note leaves the bank's pass/reject basins at exactly `½` each (collapse = pushforward theorem), and on a matching-basis note is repeatability (invisible — the secret basis *is* the security); the `¾` per-position average is the `(3/4)ⁿ` counterfeiting value, **attained** by measure-resend (optimality per Molina–Vidick–Watrous out of scope) | — | | BB84 sequential (CSD twin) | `CSD/Crypto/BB84Sequential.lean` | `bb84_wrong_basis_bob`, `bb84_wrong_basis_error`, `bb84_right_basis_no_disturbance`, `bb84_right_basis_faithful`, `bb84_eve_selector_born`, `bb84_dynamical_matches_marginal`, `xContext_rate_vertex` | Intercept-resend with the **collapse step a theorem**: Eve's measure-and-resend is `csd_sequential_born` (calibrated-swap dynamics), not a classical marginal — dissolving the QM module's own "LF5 gate" scope note. Wrong basis: Bob's basins are exactly `½` whatever Eve saw (Bob reads the **rotated X-context** `basisContext xBasisON` — both measurements are context-field reads, end-to-end dynamical). Right basis: exactly repeatability (error basin null). `hpos` proved from the preparation (`prep_outcome_pos`), not carried. Dual round of the QM module's; the values assemble to the same ¼ QBER | — | | BB84 QKD (intercept-resend) | `Crypto/BB84.lean` | `bb84_qber`, `bb84_intercept_resend_right_basis`, `bb84_intercept_resend_wrong_basis`, `bb84_eavesdropping_detectable`, `bb84_states_nonorthogonal` | The ¼ sifted-key QBER of intercept-resend eavesdropping: matching basis injects `0` error, wrong basis `½`, average `¼ > 0` — detectable; rooted in `⟨0\|+⟩ ≠ 0`. Eve's collapse modelled as a classical marginal (see the CSD twin row above for the dynamical upgrade). ⚠️ *Row added 2026-08-02: the module landed earlier, wired and pinned, but this table row was missing* | Bennett–Brassard 1984 | | B92 QKD (two-state) | `Crypto/B92.lean` | `b92_nonorthogonal`, `b92_unambiguous_one`, `b92_unambiguous_zero`, `b92_conclusive_rate_one` | The minimal QKD protocol: two non-orthogonal states; Bob's unambiguous-discrimination readout (zero-error conclusive clicks at rate `½`); security root is the same non-orthogonality as Wiesner money. ⚠️ *Row added 2026-08-02, as above* | Bennett 1992 | | Quantum money | `Crypto/QuantumMoney.lean` | `quantum_money_unforgeable`, `wiesner_nonorthogonal` | Forgery would clone non-orthogonal Wiesner states, contradicting no-cloning | Wiesner 1983 | | Wiesner money protocol | `Crypto/WiesnerProtocol.lean` | `wiesner_verify_honest`, `wiesner_forge_impossible`, `wiesner_forge_advantage_le` | A 2-state mint/verify protocol on top of unforgeability: honest verification is certain (`verifyProb (mint b) (mint b) = 1`); forgery is impossible (instantiates `quantum_money_unforgeable` on the non-orthogonal pair); `wiesnerSecurity : SecurityBound` is genuinely consumed by the advantage bound. Honest: the security `ε = 1` is the trivial qualitative bound (the quantitative `3/4` cloning fidelity is out of scope); real security is the qualitative no-perfect-forgery | Wiesner 1983 | | E91 device-independent | `Crypto/E91.lean` | `lhvCHSH_abs_le_two`, `lhvCHSH_lt_tsirelson`, `e91_no_lhv_reproduces_singlet` | A measure-theoretic LHV model obeys the CHSH bound `\|S\|≤2`; the singlet attains `2√2`, so no LHV reproduces it (eavesdropper-detectable) | Ekert 1991; CHSH 1969 | | E91 key rate (asymptotic DI) | `Crypto/E91KeyRate.lean` | `e91_key_rate_pos_of_chsh`, `e91_chsh_certifies_secure_key`, `e91_eavesdropper_advantage` | The DI asymptotic secret-key rate `e91KeyRate S = 1 − h((1+√((S/2)²−1))/2)/log2`: a CHSH violation `S>2` yields a **positive** key rate (unconditional), `r(2)=0`, `r(2√2)=1`. The capstone conjoins the proved LHV ceiling `\|S\|≤2` with `secure`; `e91_eavesdropper_advantage` bounds the LHV/eavesdropper CHSH fraction by the protocol `SecurityBound.ε = 1/√2`. Asymptotic/collective-attack; finite-key is a separate tranche | Acín-Brunner-Gisin-Massar-Pironio-Scarani 2007 | | E91 finite-key (finite-sample) | `Crypto/E91FiniteKey.lean` | `e91_chsh_concentration`, `e91_finite_key_confidence` | Hoeffding concentration of the empirical CHSH estimator: `μ{Ŝₙ ≤ S−ε} ≤ exp(−n·ε²/(2c))`, `c=((b−a)/2)²`, proved through Mathlib's sub-Gaussian API (Hoeffding's lemma + `measure_sum_range_ge_le_of_iIndepFun`); the bridge conjoins `e91_key_rate_pos_of_chsh` with an exponentially-small failure-to-certify probability. Honest: finite-**sample** confidence, NOT composable finite-key security (no smooth min-entropy / leftover hash); per-round independence + boundedness are modelling hypotheses | Hoeffding 1963; ABGMPS 2007 | | Protocols interface | `Protocols/Basic.lean` | `SecurityBound`, `IdealQKD`, `RealProtocol`, `RealProtocol.secure`, `secure_emulates` | Minimal reusable security stand-ins (security bound ε, ideal/real protocol, key rate, emulation); `secure_emulates` binds them. Honest: formal stand-ins, not a composable (UC/AVS) security model | Standard | | Gate library | `Gates/{SingleQubit,TwoQubit,MultiQubit,BellPrep}.lean` | `qmH_mul_self`, `qmS_sq`, `qmT_sq`, `qmCNOT_mul_self`, `qmSWAP_mul_self`, `qmCZ_mul_self`, `qmToffoli_mul_self`, `qmFredkin_mul_self`, `qmBellPrep_yields_phiplus` | Hadamard/phase/Pauli involutions and orders; CNOT/SWAP/CZ/Toffoli/Fredkin self-inverse; Bell preparation yields `\|Φ⁺⟩` | Standard | | QEC: 3-qubit bit-flip code | `QEC/ThreeQubit.lean` | `three_qubit_corrects_single_bitflip`, `syndrome_X1`/`X2`/`X3`, `stab_*_fixes_logical` | `\|ψ⟩↦a\|000⟩+b\|111⟩` corrects any single `X` error: stabilisers `Z₁Z₂,Z₂Z₃` fix the codespace, the four errors give distinct syndromes `(±,±)`, and each `X` is self-inverse (recovery). The first QEC theorem. | Shor 1995 | | QEC: 3-qubit phase-flip code | `QEC/PhaseFlip.lean` | `three_qubit_corrects_single_phaseflip`, `syndromePF_*`, `stab_*_fixes_logicalPF` | The Hadamard dual: `\|ψ⟩↦a\|+++⟩+b\|---⟩` corrects any single `Z` error (stabilisers `X₁X₂,X₂X₃`, errors `{I,Z₁,Z₂,Z₃}`, same syndrome/recovery structure). | Shor 1995 | ### Quantum algorithms (`CsdLean4/Empirical/QM/Algorithms/` + `Mathlib/QuantumInfo/`) The n-qubit register `QReg n = EuclideanSpace ℂ (Fin n → Fin 2)` (Cat-1 in `Mathlib/QuantumInfo/Register.lean`, with the genuine Born `prob ψ z = ‖ψ z‖²`) and the algorithm tier built on it. Finite-dimensional QM throughout (no field theory); the Quantum Fourier Transform is a finite unitary. All foundational-triple-only, AxiomAudit-pinned, and Tier-A adversarially audited SOUND (2026-06-08). | Test | File | Headline theorem(s) | Claim | Source | |---|---|---|---|---| | Hadamard transform | `Mathlib/QuantumInfo/Hadamard.lean` | `Hn_unitary`, `Hn_mul_self`, `Hn_apply_zero` | The `n`-fold Hadamard `Hⁿ` is genuinely unitary (`Hⁿᴴ·Hⁿ = 1`, `1/√2` normalised) and an involution; `Hⁿ\|0ⁿ⟩` is the uniform superposition `(1/√2)ⁿ` | Standard | | Deutsch-Jozsa | `Algorithms/DeutschJozsa.lean` | `deutsch_jozsa_constant`, `deutsch_jozsa_balanced` | One query decides constant vs balanced: a constant oracle returns the all-zeros string with Born prob `1`, a balanced oracle with prob `0` | Deutsch-Jozsa 1992 | | Simon | `Algorithms/Simon.lean` | `simon_orthogonal`, `simon_uniform`, `simon_amplitude` | Simon orthogonality: after `Hⁿ` on the coset superposition `(\|x₀⟩+\|x₀⊕s⟩)/√2`, every measured `y` is orthogonal to the hidden period `s` (Born prob `0` when `⟨s,y⟩` odd; equal value `2/2ⁿ` on `s^⊥`). Single-register reduced analysis | Simon 1994 | | Bernstein-Vazirani | `Algorithms/BernsteinVazirani.lean` | `bv_certain`, `bv_zero`, `bv_amplitude`, `bitInner_char_sum` | One query recovers the hidden linear string `a`: the full circuit `Hⁿ ∘ U_f ∘ Hⁿ` on `\|0ⁿ⟩` (phase oracle `\|x⟩↦(-1)^{⟨a,x⟩}\|x⟩`) outputs `\|a⟩` deterministically (`prob = δ_{y,a}`), via the character sum `∑ₓ(-1)^{⟨z,x⟩} = 2ⁿ·[z=0]` | Bernstein-Vazirani 1993 | | Swap test | `Algorithms/SwapTest.lean` | `swap_test_prob`, `swap_test_equal`, `swap_test_orthogonal` | The overlap estimator (ancilla interferometry): `Hₐ ∘ cSWAP ∘ Hₐ` on `\|0⟩⊗ψ⊗φ` gives ancilla-`0` probability `(1 + \|⟨ψ,φ⟩\|²)/2` (general finite system `ι`) — `1` at `ψ=φ`, `1/2` at orthogonal | Buhrman-Cleve-Watrous-de Wolf 2001 | | Hadamard test | `Algorithms/HadamardTest.lean` | `hadamard_test_prob`, `hadamard_test_prob1`, `hadamard_test_prob_diff`, `hadamard_test_eq_one`, `swap_test_via_hadamard`, `hadamard_test_swap_closed` | The expectation-value estimator (parent of the swap test): `Hₐ ∘ cU ∘ Hₐ` on `\|0⟩⊗ψ` gives ancilla-`0` probability `(1 + Re⟨ψ,Uψ⟩)/2`, `P(1) = (1 − Re⟨ψ,Uψ⟩)/2`, so `P(0) − P(1) = Re⟨ψ,Uψ⟩` (general finite system `ι`, plain linear `U` with `\|Uψ\|=1`); `1` at `Uψ=ψ`. The swap test is this at `U=SWAP` on the doubled register (`swap_test_via_hadamard`), value `(1 + \|⟨ψ,φ⟩\|²)/2` | Aharonov-Jones-Landau 2006 | | Quantum Fourier Transform | `Mathlib/QuantumInfo/Fourier.lean` | `qft_unitary` | The QFT matrix `F j k = (1/√N)·ω^{jk}` (`ω = e^{2πi/N}` a primitive root) is genuinely unitary `Fᴴ·F = 1`, via roots-of-unity character orthogonality | Coppersmith 1994 | | Grover search | `Algorithms/Grover.lean` | `grover_success`, `grover_certain` | Amplitude amplification: after `k` Grover steps the marked-item Born prob is `sin²((2k+1)θ)` (`sin θ = 1/√N`); exact certainty at the resonant `k` (e.g. 1-of-4 in one step) | Grover 1996 | | Shor: quantum core | `Algorithms/ShorCore.lean` | `shor_order_readout`, `shor_order_distribution`, `phase_estimation_lower_bound`, `shor_phase_estimation_lower_bound` | Order-finding by phase estimation: ideal `r∣T` reads `s·(T/r)` with prob `1`; the two-register modexp state gives the uniform `1/r` distribution; general `r∤T` has the Dirichlet `≥ 4/π²` lower bound | Shor 1994 | | Shor: period recovery | `Algorithms/ShorRecovery.lean` | `shor_period_determined`, `abs_sub_rat_ge`, `approx_unique` | The measured count determines the order `r` uniquely (distinct lowest-terms fractions are `≥ 1/(b·d)` apart); information-theoretic determination | Shor 1994 | | Shor: factoring + bridge | `Algorithms/ShorRecovery.lean` | `nontrivial_factor`, `even_order_sqrt_unity`, `shor_factor_of_even_order` | A nontrivial `√1 mod N` gives a proper divisor `gcd(x−1,N)`; an even-order unit `a` with `a^(r/2) ≢ ±1` yields a factor — the classical reduction order-finding ⟹ factoring | Shor 1994 | | Shor: random-`a` success | `Algorithms/ShorRandomA.lean` | `shor_random_a_success`, `shor_random_a_success_general`, `shor_success_prob_ge_general` | A random unit `a` is "good" (even order, `a^(r/2) ≠ −1`) with prob `≥ 1/2`, for **arbitrary odd composite `N`** — via the indexed-CRT 2-adic-valuation group-counting argument | Shor 1994 | | Shor: factoring capstone | `Algorithms/ShorCapstone.lean` | `shor_random_a_yields_factor`, `shor_factor_prob_ge` | End to end: a good `a` yields the explicit proper divisor `gcd(rep(a^(r/2))−1, N)`; a random `a` factors any odd composite `N` with prob `≥ 1/2` | Shor 1994 | Namespace note: the earliest files use `CSD.Empirical.` (Bell, NoCloning, NoDeleting, Uncertainty, SternGerlach, Hardy, GHZ, KochenSpecker, MerminPeres, QuantumMoney); later files use `CSD.Empirical.QM.` (NoBroadcasting, NoCommunication, USD, E91, Resources/*, Gates/*). The fully-qualified names above (taken from the AxiomAudit pins) are canonical. --- ## CSD-ontic branch (`CsdLean4/Empirical/CSD/`) > **★ NEW 2026-08-02 — the first entry consuming the dynamical measurement layer** > (`SequentialMeasurement.lean`): **repeatability** (`csd_repeatability` — a repeated > computational-basis measurement gives the same outcome with probability 1, others 0) and > **sequential Born** (`csd_sequential_born` — after outcome `i`, follow-up statistics for *any* > context are the collapsed state's Born weights `c'.rate [eᵢ]`; the preparation has left the > statistics). Both **derived from the calibrated-swap dynamics** (`swap_luders_born` + > `momentMap_vertex`), not posited. Rank-one first measurement; inherits the witness's scope > notes. **Coverage relative to the QM branch.** Every QM-validity test now has a CSD-branch counterpart (a bridge transport, a derived volume reading, or both). USD has no bridge transport but is covered by `USDVolume.lean` (the derived volume reading). The trine and SIC POVMs exist only in the CSD branch (their POVM content lives inside the volume files), with no separate `QM/` file. The bridge transports reduce to their QM-side theorem by field extraction; the genuinely-ontic realisability is the disclosed LF4 obligation in [`BRIDGE-OBLIGATIONS.md`](BRIDGE-OBLIGATIONS.md) (E91 carries no such obligation, by design — see its row). ### Bridge transports (QM-validity statements carried through `CSDBridge.Context D`) | Test | File | Headline theorem | Note | |---|---|---|---| | Bell singlet chain | `Bell.lean` | `bell_singlet_frequency_convergence` | Singlet kernel `(1 − st a·b)/4` via the LF1↔LF2↔LF3 chain | | No-cloning bundle | `NoCloning.lean` | `no_csd_cloning_bundle` | Ontic-bundle reading of no-cloning | | No-deleting bundle | `NoDeleting.lean` | `no_csd_deleting_bundle` | | | No-broadcasting | `NoBroadcasting.lean` | `csd_no_broadcasting` | Pure-marginal confinement on the CSD substrate | | No-communication | `NoCommunication.lean` | `csd_no_communication` | Bob's expectation invariant under Alice's local unitary | | Teleportation | `Resources/Teleportation.lean` | `csd_teleportation_branch_recovers_input` | Branch-conditional input recovery (LF5 for collapse) | | E91 security | `Crypto/E91.lean` | `csd_lhv_chsh_bound` | Any local-realist reading of the source obeys `\|S\|≤2`; CSD reaches `2√2` (no LF4 obligation) | | Robertson | `Uncertainty.lean` | `csd_robertson_uncertainty` | | | Stern-Gerlach | `SternGerlach.lean` | `csd_sg_born_xPlus_zPlus`, `csd_sg_born_x_basis_complete` | Transport tag (cf. the derived form below) | | Superdense coding | `Resources/SuperdenseCoding.lean` | `csd_sdc_encode_X`, `csd_sdc_bell_basis_orthonormal` | | | Quantum money | `Crypto/QuantumMoney.lean` | `no_csd_quantum_money_forger` | | | Mermin-Peres | `Contextuality/MerminPeres.lean` | `no_csd_mermin_peres_assignment` | impossibility transport; the **volume companion** is `Contextuality/MerminPeresVolume.lean` (`mp_xx_born_frequency_volume`, the rank-2 `X⊗X` reading) | | Hardy | `Hardy.lean` | `no_csd_hardy_assignment` | | | Kochen-Specker | `Contextuality/KS18.lean` | `no_csd_ks_assignment_bundle` | impossibility transport; the **volume companion** is `Contextuality/KS18Volume.lean` (`ks18_context_born_frequency_volume`) | | GHZ | `Multipartite/GHZ.lean` | `no_csd_ghz_lhv_bundle` | | | QEC: 3-qubit code | `QEC/ThreeQubit.lean` | `csd_three_qubit_corrects_single_bitflip`, `bitFlipFlow_traceRight_barycenter`, `errorRegion_disjoint` | First QEC reading: codespace = sub-surface of `Σ`; error = decoherence (system→environment volume flow, Liouville-conserved jointly); syndrome measurement = the entropy-extraction that re-concentrates the system; weights = `Σ`-volumes. **Proved 2026-09-11 (W11):** the bit-flip channel is the environment marginal of a `Σ`-flow (concretely `bitFlipFlow` on `cpSectorData` over `ℂℙ³`, no lift hypothesis), the code space is a region of `Σ` whose four error images are pairwise disjoint and returned by recovery; the former unused bundle binder is gone. **2026-09-12:** the syndrome projectors are the indicator of the error regions (`syndromeProj_fixes_errorRegion`), the syndrome-conditioned recovery is one channel on the mixed state (row below), and the **register ⊗ environment joint flow is a theorem** (`QEC/RegisterFlow.lean`: `registerFlow` on `ℂℙ³¹` produces the register's single-error channel as its environment marginal, `registerFlow_traceRight_barycenter`; ★★★ `registerFlow_recovery`, QEC on `Σ` end to end — the recovery returns every code-region preparation's density operator from the flowed preparation's marginal). **2026-09-21 (#52):** the independent-noise channel `(bit-flip_p)^{⊗3}` itself is the environment marginal of `indepFlow` on `ℂℙ⁶³` (`QEC/IndependentNoiseFlow.lean`), and ★★★ `indepFlow_recovery` returns `(1 − p_fail) σ + p_fail X̄ σ X̄`, `p_fail = 3p² − 2p³` — the residual failure of the code on `Σ`. | | A circuit is a flow on `Σ` | `Algorithms/CircuitFlow.lean` | `Circuit.isUnitaryLift_flow`, `Circuit.barycenter_flow_iterate`, `Circuit.epistemicMeasure_globalBasin_flow_iterate` | **2026-09-19 (#34):** any unitary on a register is the projective action on the sector `ℂℙ^{card−1}`, the flow lifts it with no hypothesis (canonical measurable unit section), `k` runs are the flow iterated `k` times with density operator `U^k ρ (U^k)ᴴ` derived from the flow, and the readout is the context-fixed record basin of the outcome: at the flowed ready point the basin of `i` has Born weight `‖(U^k ψ) i‖²`. The `RegisterFlow` pattern, factored out; the readout channel form `readout` is the de-isolation channel's diagonal. | | Grover on `Σ` | `Algorithms/GroverFlow.lean` | `grover_flow_born`, `grover_flow_certain`, `toEuclideanLin_groverMatrix` | **2026-09-19 (#34):** the Grover step is a product of two reflections `1 − 2|v⟩⟨v|`, hence a unitary and a `Σ`-flow; at the point reached from the uniform superposition by `k` runs the marked item's basin has weight `sin²((2k+1)θ)`, and weight `1` at the optimal count. Twin of the QM-side `grover_success` / `grover_certain`; the QM theorem carries the mathematics, the twin the ontic reading. | | Shor on `Σ` | `Algorithms/ShorFlow.lean` | `shor_flow_born_count`, `shor_flow_born_count_zero`, `toEuclideanLin_shorMatrix` | **2026-09-19 (#34):** the modular-exponentiation oracle is a permutation matrix (unitary), the inverse QFT on the counting register is the Kronecker factor `(qftMatrix T)ᴴ ⊗ 1`, their product is a `Σ`-flow on the joint register's sector; at the flowed ready point `uniformCount ⊗ |1⟩` the counting readout `s·(T/r)` — the union over the work register of the joint outcomes' basins — has weight `1/r` and the other counting outcomes weight `0` (ideal case `r ∣ T`). Twin of `shor_order_distribution` / `_zero`. | | QEC-as-decoherence (15b) | `QECDecoherence.lean` | `qec_corrects_decoherence`, `csd_qec_decoherence_corrected` | Discharges the **CPTP-channel** half of the QEC obligation above: the single-qubit error IS the K2 bit-flip CHANNEL with a genuine Stinespring/partial-trace origin (`bitflip_error_is_decoherence`); the correctable branch lifted to the code as `unitaryChannel Xⱼ` and corrected on the encoded density `recoverⱼ∘errorⱼ=id` (`recover_channel_compose_error_on_code`, one Hilbert space, non-vacuous via `error_moves_encoded_density`); bare-qubit-decoheres contrast. **2026-09-12:** the full mixed-channel syndrome-conditioned recovery is a theorem — `QM/QEC/SyndromeRecovery.lean` builds the syndrome projectors (a projective measurement on the register) and the recovery channel `∑ₖ Eₖ Pₖ · Pₖ Eₖ`, and `syndrome_recovery_corrects_mixed` (now the last conjunct of both capstones) returns the encoded density from the whole single-error mixture `∑ₖ qₖ Eₖ · Eₖ`, any weights. **Gated (not discharged):** the ontic Σ-volume / partial-trace-volume-loss origin (`Φ≠id`, entangled tier / D1); **2026-09-21:** the independent-noise channel `(bit-flip_p)^{⊗3}` itself is a theorem (`QM/QEC/IndependentNoise.lean`, `recoveryChannel_apply_indepFlipChannel_apply`: `3p² − 2p³` residual). | | Einselection follow-up (15a+) | `Einselection.lean` | `einselection_degenerate_boundary`, `einselectionN`, `einselectionN_degenerate` | The pointer-basis **selects-vs-degenerates** boundary: einselection selects the computational basis iff `p₀≠p₁` (`decohere_hadamard_offDiag_ne_zero_iff`), and DEGENERATES at `p₀=p₁` (reduced state `½I`, basis-invariant under every unitary). General-N dephasing (`decohereReducedN`, non-vacuous via `decohereReducedN_acts_nontrivial`), degenerate locus all-equal `⇒ (1/N)I`. Qubit channel derived (partial trace of the LF5 dilation); general-N posited-and-bridged. Ontic einselection-from-Σ-dynamics origin D1-gated. | | Weak / unsharp measurement (15c) | `WeakMeasurement.lean` | `weak_born_unsharp_interpolation`, `weak_born_frequency_volume` | One-parameter unsharp qubit POVM `weakEffect±(η) = ½(I±ησ)` interpolating no-measurement (η=0, `½I`, no info — the 15a degeneracy) ↔ projective carve (η=1, sharp `|0⟩⟨0|`, full Born `‖ψ₀‖²`); η-bound load-bearing (PSD fails η>1), genuine partial information (`weak_partial_information_witness`, ¾ at η=½). **Volume reading** (full theorem, Gleason-free, unconditional): the weak Born weight = a sum of FS volumes on the ℂℙ³ Naimark dilation (`weak_born_frequency_volume` via `povm_born_frequency_volume_uncond`) — the "partial Σ-volume nudge". **Gated:** continuous measurement as a dynamical `Φ≠id` weak flow (D1 / entangled tier). | | Quantum Zeno (15d) | `QuantumZeno.lean` | `zeno_survival_quadratic`, `zeno_survival_slope_zero`, `zeno_freezing`, `zeno_nonvacuous` | A watched system does not evolve: frequent projective re-measurement onto `|0⟩` freezes the state. **Mechanism DERIVED** (concrete `σx`/`|0⟩` witness): variance `(ΔH)²=1` computed from the matrices (`varH_eq`), the **quadratic short-time bound** `P(s)≥1−(ΔH)²s²` (`zeno_survival_quadratic`, from `cos²=1−sin²≥1−s²`; the coefficient is the computed variance, not a hypothesis), and the **zero initial slope** `P'(0)=0` (`zeno_survival_slope_zero`, via `HasDerivAt`). **Limit:** `P_n=P(t/n)^n≥1−(ΔH)²t²/n` (Bernoulli, `zeno_survival_lower_bound`) and `P_n→1` (`zeno_freezing`, full squeeze). **Non-vacuity:** `(ΔH)²>0` with full free decay `P(π/2)=0` (`zeno_nonvacuous`), so the freezing is non-trivial. The closed-form `exp(-isσx)=cos s·I − i sin s·σx` is the asserted standard qubit rotation; all else derived. **CSD reading:** each measurement re-carves the Σ-volume onto the initial sector. **Gated:** the dynamical measurement-interspersed Σ-flow `Φ≠id` (D1 / LF6). | | Channel capacities (15e) | `ChannelCapacity.lean` | `dephasing_classical_vs_quantum`, `holevo_classical_eq_log_two`, `dephasing_destroys_coherence` | The de-isolation / dephasing channel `Φ_deph = decohereReducedN` (15a) on the K1-A von Neumann entropy: CLASSICAL info survives, QUANTUM coherence destroyed. **Classical-yes:** computational-basis states are FIXED POINTS `Φ_deph(|i⟩⟨i|)=|i⟩⟨i|` (`dephasing_fixes_basis_state`, general N); single-letter Holevo χ of `{(½,|0⟩⟨0|),(½,|1⟩⟨1|)}` = `log 2` (`holevo_classical_eq_log_two`, `S(½I)−½·0−½·0`), a full classical bit. **Quantum-no:** `Φ_deph(|+⟩⟨+|)=½I` (`dephasing_plus_eq_half_one`), entropy jump `0→log 2` (`dephasing_destroys_coherence`). `S(½I)=log 2` DERIVED via `vonNeumannEntropy_const_smul_one` (charpoly route, `spectral_sum_eq_of_charpoly_prod`); `S(pure)=0` from `vonNeumannEntropy_eq_zero_of_pure`. **Scope:** single-shot Holevo / coherent-information, NOT the regularized capacity; the general `χ≥0` bound (entropy concavity) is gated on the open SSA fork; the ontic Σ-volume capacity (`Φ≠id` throughput) D1-gated to LF6. CLOSES the open-system 15-series (15a-e). | | Wigner–Araki–Yanase (record-layer twin) | `WignerArakiYanase.lean` | `record_not_factor_hilbert`, `no_joint_hilbert_map`, `record_mem_recordRegion_of_register`, `pointerEvolve_record_not_factor_hilbert`, `jointLift_record_not_factor_hilbert` | At the `IsJointLift` level the pointer image is not a function of the Hilbert data `(base ray, pointer)`: from identical data the fibre register coordinate selects any outcome of rate `≥ 2ε` (`IsJointLift.landing`), and distinct outcomes land in disjoint record regions — so no map on any joint Hilbert space `HS ⊗ HA` reproduces the stroke's pointer image from the Hilbert data, and WAY's hypothesis is met by no joint lift (the quantifier is a robustness one — `pointer_eq` pins every lift's pointer image to the witness's — not a WAY-strength one). A **scope** statement, not an escape: no conservation law is modelled at the record layer (`R-015`); the `ε`-Born statistics are untouched. Non-vacuous at `[(1,1)]` for `0 < ε ≤ ¼`; instances for the fibrewise witness and the back-reacting `jointLift` (every shift) | | Gates | `Gates/{Framework,SingleQubit,TwoQubit,MultiQubit,BellPrep}.lean` | (transport) | Bridge readings of the QM gate library | | Bridge scaffolding | `Framework.lean` | `CSDBridge.Context` | The bundle type itself (infrastructure, not a test) | Several bundles carry load-bearing, externally-supplied realisability fields (the ledger is [`BRIDGE-OBLIGATIONS.md`](BRIDGE-OBLIGATIONS.md)). ### Volume-frequency series (Born numbers *derived* as Fubini-Study volumes) Carving-free and Gleason-free: each Born value is realised as a Fubini-Study typicality volume on the ontic `Σ`, and empirical frequencies (LF1 strong law) converge to it. The first five are projective qubit measurements; the next three are non-projective qubit POVMs via canonical Naimark dilation; the last is the first **non-qubit (`N = 3`)** entry, a non-projective qutrit POVM. Since the 2026-06-11 hpos call-site migration, every volume-frequency capstone in this table is **genericity-free at the statement level** (the bornRegion-engine entries route through the hpos-free engine `LF4/BornRegionUncond.lean`; SternGerlach/Malus run on the qubit moment-sublevel chain, which never carried genericity; Hardy still routes through the conditional engine, discharging `hpos` internally for its concrete generic state): boundary preparations with vanishing amplitudes — eigenstates of the measured context, the Mermin GHZ points `Φ = 0, π`, aligned Bell analysers `θ = 0, π` — are covered (their cells are FS-null with frequencies converging to `0`). Since 2026-06-15, **every** volume-frequency headline in this table has a `_canonical` corollary (`Empirical/CSD/VolumeCanonical.lean`, plus `povm_born_frequency_volume_canonical` in `LF4/TrialWitness.lean` for the POVM engine) that discharges the abstract i.i.d. trial bundle `(Ω, Pr, X, hX, hlaw, hindep)` at the in-tree Fubini-Study coordinate process (`fsTrialMeasure p₀ = Measure.infinitePi (fun _ => fsMeasure p₀)`, `fsTrial N n = (· n)`). The hypothesis sets are therefore **Lean-inhabited**, not merely classically satisfiable; each `_canonical` conclusion is its parent's verbatim under `Pr := fsTrialMeasure p₀`, `X := fsTrial _`. This is coverage/completeness (the witness already existed): measure-theoretic existence of the sampling law only; the physical FS-typical repeated-preparation reading remains the LF1 typicality / A5 posit. | Test | File | Headline theorem(s) | Derived value | |---|---|---|---| | Stern-Gerlach | `SternGerlachVolume.lean` | `csd_sg_volume_certain`, `csd_sg_volume_half` | `P(↑\|ẑ)→1`, `P(↑\|x̂)→1/2` | | Malus's law | `MalusVolume.lean` | `csd_malus_law` | `P(↑\|θ)→cos²(θ/2)` | | Ramsey interferometry (metrology A1) | `Empirical/Metrology/Ramsey.lean` | `ramsey_fringe_volume`, `ramseyPhaseFlow_measurePreserving`, `ramseyPhaseFlow_ne_id`, `ramseyVec_eq_circuit`, `ramsey_fringe_hasDerivAt` | The first quantum-sensing validation: the free-precession phase `diag(1,e^{iφ})` is a deterministic **FS-measure-preserving flow** on `Σ = ℂℙ¹` (`Φ ≠ id`), the Ramsey output state is machine-checked as the genuine circuit `qmH·diag·qmH·\|0⟩`, and the end-carving FS volume reproduces the fringe `cos²(φ/2)` (via the qubit Born-volume engine, Gleason-free); sensitivity `dP/dφ = −sin(φ)/2`, maximal at quadrature `φ=π/2` (the QFI precursor). Single-qubit; Born = FS volume imported | Ramsey 1950 | | Quantum Fisher info = FS metric (metrology A2) | `Empirical/Metrology/QuantumFisher.lean` | `ramseyVec_hasDerivAt`, `ramsey_fs_metric`, `ramsey_qfi`, `ramsey_classical_fisher`, `ramsey_qcrb_saturation` | The QFI as the pure-state Fubini-Study line element `g = ‖∂ψ‖²−‖⟨ψ,∂ψ⟩‖²`, `F_Q = 4g`: the Ramsey-family derivative is machine-checked (`ramseyVec_hasDerivAt`, genuine `HasDerivAt`, not asserted), giving `g = 1/4` and `F_Q = 1` (the SQL per-shot information), and the computational-basis readout's classical Fisher info `F_C = 1` (for `sin φ ≠ 0`), so `F_C = F_Q` — the Ramsey measurement **saturates the quantum Cramér-Rao bound** (Fisher-optimal at every working point). Single-qubit; the trajectory-pullback metric on (state + derivative vector), NOT the intrinsic ℂℙⁿ Riemannian tensor (deferred) | Helstrom 1976; Braunstein-Caves 1994 | | Heisenberg limit (metrology A3) | `Empirical/Metrology/Heisenberg.lean` | `ghzPhaseVec_norm`, `ghzPhaseVec_hasDerivAt`, `ghz_qfi`, `heisenberg_advantage`, `ghz_qfi_div_sql` | The `1/N` (Heisenberg) scaling via the entangled GHZ probe, reusing the A2 QFI engine: the phase-accumulated GHZ state `(1/√2)(|0…0⟩+e^{iNφ}|1…1⟩)` on a genuine `2^N`-dim N-qubit carrier has `F_Q = N²` (`ghz_qfi`, the collective phase `Nφ` giving the quadratic enhancement, machine-checked derivative), versus the SQL `F_Q = N` of N independent probes (`sqlQFI`, A2's per-probe `F_Q=1` + Fisher additivity), an N-fold advantage (`heisenberg_advantage`, `ghz_qfi_div_sql`), i.e. precision `1/N` vs `1/√N`. Geometric QFI of the GHZ-phase family (the A2 trajectory pullback); the GHZ-preparation dynamics is not modelled | Giovannetti-Lloyd-Maccone 2006 | | Einselection / pointer-basis (open-system 15a) | `Einselection.lean` | `decohere_hadamard_offDiag`, `decohere_not_diagonal_in_rotated_basis`, `einselection` | Decoherence is **basis-selective**: the LF6-B reduced state `decohereReduced ψ` is diagonal in the measurement (pointer) basis `{e_j}` but its coherences **persist** in the Hadamard-rotated basis — `(qmH·decohereReduced ψ·qmH)₀₁ = (p₀−p₁)/2 ≠ 0` for `p₀≠p₁` (witness off-diagonal `3/2`). So the pointer basis is genuinely selected (Zurek einselection, "why a preferred basis"). HONEST: the basis is the de-isolation's context basis, NOT derived from an environment Hamiltonian; contrasts with the basis-**covariant** FS typicality (`LF4.fubiniStudy_forced_by_symmetry`, unique U(N)-invariant). The basis selection is the symmetry-breaking-by-context, distinct from the symmetric sector/typicality | Zurek 1981/2003 | | Uncertainty (variance as volume) | `UncertaintyVolume.lean` | `variance_eq_four_vol_product`, `born_variance_eq_vol_product`, `uncertainty_volume_frequency` | A `±1` qubit observable's variance is a **product of two FS typicality volumes** `Var = 4·vol₊·vol₋` on `Σ = ℂℙ¹` (composes `context_born_frequency_volume` at M=1 with the `4p(1−p)` algebra). The Robertson *inequality* stays at the QM layer (`robertson_uncertainty`, it is Cauchy-Schwarz not a volume identity); this grounds the variances entering it | | Bell singlet (joint) | `BellVolume.lean` | `bell_singlet_born_frequency_volume`, `bell_singlet_volume_correlation` | joint weights `(1∓cos θ)/4` on `ℂℙ³` at **every** relative angle `θ` (aligned/anti-aligned `θ = 0, π` included since 2026-06-11); `⟨σ_a σ_b⟩=−cos θ` | | GHZ | `GHZVolume.lean` | `ghz_born_frequency_volume`, `ghz_volume_correlation` | GHZ stabiliser Born weights and correlations, at **every** angle-sum `Φ` (the Mermin all-or-nothing points `Φ = 0, π` included since 2026-06-11) | | Hardy | `HardyVolume.lean` | `hardy_max_born_frequency_volume`, `hardy_max_volume_probability` | the optimal Hardy probability `(5√5−11)/2` | | Trine POVM | `TrineVolume.lean` | `trine_complete`, `trine_weight_eq`, `trine_born_frequency_volume` | `pₖ=(2/3)‖⟨ψₖ,ψ⟩‖²` on dilated `ℂℙ⁵` (first non-projective entry) | | USD POVM | `USDVolume.lean` | `usd_weight_e1`, `usd_weight_e2`, `usd_born_frequency_volume` | conclusive weights `a‖⟨ψₖ^⊥,ψ⟩‖²` on dilated `ℂℙ⁵` | | SIC POVM | `SICVolume.lean` | `sic_outer_sum`, `sic_inner_normSq`, `sic_weight_eq`, `sic_born_frequency_volume` | tetrahedral `pₖ=(1/2)‖⟨ψₖ,ψ⟩‖²`, equiangular `\|⟨ψⱼ,ψₖ⟩\|²=1/3`, on dilated `ℂℙ⁷` | | Unsharp qutrit POVM (**N=3**) | `QutritPOVMVolume.lean` | `noisy_complete`, `noisy_weight_eq`, `noisy_born_frequency_volume` | `Eₖ=(1−ε)\|k⟩⟨k\|+(ε/3)I₃`; `pₖ=(1−ε)‖⟨k,ψ⟩‖²+ε/3`, on dilated `ℂℙ⁸` (first non-qubit entry) | | d=3 SIC / Hesse POVM (**N=3**) | `SIC3Volume.lean` | `sic3_complete`, `sic3_inner_normSq`, `sic3_weight_eq`, `sic3_born_frequency_volume` | 9 Weyl–Heisenberg states `Eₖ=(1/3)\|ψₖ⟩⟨ψₖ\|`, equiangular `\|⟨ψⱼ,ψₖ⟩\|²=1/4`; `pₖ=(1/3)‖⟨ψₖ,ψ⟩‖²`, on dilated `ℂℙ²⁶` (first *symmetric* non-qubit entry) | | d=3 complete-MUB POVM (**N=3**) | `MUB3Volume.lean` | `mub3_complete`, `mub3_unbiased`, `mub3_weight_eq`, `mub3_born_frequency_volume` | 4 mutually unbiased bases (12 vectors) `Eₖ=(1/4)\|vₖ⟩⟨vₖ\|`, unbiased `\|⟨v,w⟩\|²=1/3` across bases; `pₖ=(1/4)‖⟨vₖ,ψ⟩‖²`, on dilated `ℂℙ³⁵` | | Any projective context (**contextuality**) | `ContextVolume.lean` | `context_born_frequency_volume`, `context_born_eq_rotated`, `block_born_frequency_volume`, `block_born_frequency_volume_event` (single union event, LF5-F), `block_born_eq_blockSum`, `zz_parity_born_frequency_volume` (concrete Z⊗Z rank-2 witness) | an **arbitrary** orthonormal measurement context `B` at an **arbitrary** unit preparation — eigenstates of the context (`ψ = Bᵢ`, the Kochen-Specker-interesting ones) included since the 2026-06-11 hpos migration: the rank-1 outcome Born weights `‖⟨Bᵢ,ψ⟩‖²` are FS typicality volumes (`context_…`, via `born_frequency_convergence_N_uncond` on the `B.repr`-rotated state), and **degenerate (rank-≥1) outcomes** `⟨ψ,Pₐψ⟩ = ∑_{i∈block a} ‖⟨Bᵢ,ψ⟩‖²` are **block sums** of FS volumes (`block_…`, covers e.g. Mermin-Peres rank-2 eigenspaces). The block frequency is now also available as the frequency of a **single union event** `⋃_{blk i = a} bornRegion` (`block_born_frequency_volume_event`, 2026-06-14, LF5-F), since the per-ray cells are pairwise disjoint (`CSD.LF4.bornRegion_pairwiseDisjoint`) — the `aeece86`-owed union restatement, sum form untouched. The general grounding for projective contextuality (Kochen-Specker, Mermin-Peres). Honest scope: realisation not derivation — context-dependence is frame-dependence of the carved regions (`Φ = id`); the KS/MP no-go stays at the QM-validity layer | | Kochen-Specker volume (**Cabello-18 instance**) | `Contextuality/KS18Volume.lean` | `ks18_context_born_frequency_volume`, `ks18_context_born_frequency_volume_canonical`, `ksCtxVec_orthonormal` | The concrete instantiation of `context_born_frequency_volume` at a Cabello-18 context: `ksContextBasis` (the normalised, complexified context-0 rays, orthonormal via the QM-side `cabello_pairwise_orthogonal_in_basis`) grounds each ray's context Born weight `‖⟨Bᵢ,ψ⟩‖²` as an FS typicality volume on `Σ = ℂℙ³`, every unit `ψ`. Gleason-free. Honest: one representative context (other 8 identical); the no-go itself stays at the QM layer (`ks_no_value_assignment_cabello18`) | | KCBS pentagon volume (`Contextuality/KCBSVolume.lean`, **2026-08-02**) | `kcbs_context_born_frequency_volume` (+ `_canonical`), `kcbs_pentagon_weight` | The representative pentagon context `{kv 0, kv 1}` completed by the cross product (Lagrange identity for unit norm), complexified via the `c3_inner` transport — orthonormality pulled from the QM side (`kv_orth`, `kv_unit`), nothing re-proved. Every ray's context Born weight = FS typicality volume on `Σ = ℂℙ²`; at the apex the ray-0 weight is the pentagon number `1/√5`, whose five-fold sum `√5` violates the noncontextual bound `2` (`kcbs_quantum_violation`). One representative context (KS18 discipline); the other four are identical instantiations under `kv_orth` | Klyachko–Can–Binicioğlu–Shumovsky 2008 | | Mermin-Peres volume (**full 9-observable square**) | `Contextuality/MerminPeresVolume.lean` | `mp_{xx,xi,ix,zi,iz,xz,zx,yy}_born_frequency_volume` (+ `_canonical`), `mp{XX,XI,IX,ZI,IZ,XZ,ZX,YY}Vec_eigenvector`, `…Blk_eq_zero_iff_eigval_one`, `+ zz_parity` (Z⊗Z) | All nine magic-square observables `{X⊗I, I⊗X, X⊗X, I⊗Z, Z⊗I, Z⊗Z, X⊗Z, Z⊗X, Y⊗Y}` grounded as `block_born_frequency_volume` readings on `Σ = ℂℙ³`: each cell's named basis is **machine-checked** as the genuine `σ_a⊗σ_b` eigenbasis (`*Vec_eigenvector` vs the real `sigmaX/Y/Z` via `reindex_sigma_*`) AND its collapsed block as the `+1` eigenspace (`*Blk_eq_zero_iff_eigval_one`) — both faithfulness halves axiom-audited per cell, the label earned not asserted. Four frames (`H⊗I`, `I⊗H`, computational, and the complex `U_Y⊗U_Y` for `Y⊗Y`); shared-frame cells carry distinct correct eigenvalues/blocks (`X⊗I` blocks `{0,1}` vs `X⊗Z` blocks `{0,3}`). Each `σ_a⊗σ_b = +1` outcome Born weight is a block sum of FS volumes. Gleason-free; the no-go stays at the QM layer (`no_lhv_mermin_peres`) | The trine, USD, and SIC entries span the canonical minimal qubit POVM family: minimal symmetric (trine), unambiguous discrimination (USD), and symmetric informationally complete (SIC). The unsharp and Hesse-SIC qutrit POVMs are the first entries exercising the machinery past the qubit (`N = 3`, dilated `ℂℙ⁸` and `ℂℙ²⁶`); the d=3 SIC is the genuine symmetric informationally-complete qutrit measurement (the Weyl–Heisenberg orbit of the Hesse fiducial, equiangular at `1/(d+1) = 1/4`). The general (non-projective) ontic Born = Kähler-volume machinery they instantiate lives in `CsdLean4/LF4/POVM*.lean` (see `README.md` LF4 POVM row). --- ## See also - [`README.md`](README.md) — authoritative status overview (layer tables, axiom posture). - [`specs/qm-empirical-tests.md`](specs/qm-empirical-tests.md) — the roadmap, per-item status tags, and the planned-but-not-yet-done items (BB84/B92, weak-measurement paradoxes, algorithms). - [`AXIOMS.md`](AXIOMS.md) — the axiom ledger. ⚠️ *Line corrected 2026-08-02: an earlier version said `busch_effect_gleason` was the one standing axiom — it was **discharged 2026-07-21** (proved as `OperationalPackage.effect_gleason_representation`), and the corpus imports **zero** axioms; `check-claims.sh` enforces this.* The operational/ontic two-strata reading is unchanged. - [`CLAUDE.md`](CLAUDE.md) — architecture and module-chain guide.