Machine-checked mathematics behind the paper: the results listed below, checked by the Lean kernel on every push. What is not formalized is listed in LIMITATIONS.md.
Jeromie Beasley
Memory comes before the clock. A present state is worth remembering only as far as some
future observation can tell it apart. Take a linear continuation map L and a declared
observation R. Two states are equivalent when every future reading R Lⁿ agrees on them,
and the quotient by that equivalence is the smallest state that still predicts everything:
| If you want to… | Open |
|---|---|
| See the main results in plain Mathlib terms | DiracTime/Headline.lean |
| Know exactly what is not proved | LIMITATIONS.md |
| Check where every file came from | PROVENANCE.md |
| See the proofs themselves | DiracTime/PredictiveQuotient.lean |
| See the statements that must be rejected | FalseControls/ |
| # | Statement | In words |
|---|---|---|
| 1 | future_decided_by_first_dim_steps |
Two states agree on every future observation exactly when they agree on the first dim V of them |
| 2 | minimal_predictor_dimension |
Any linear summary that determines all future observations has rank at least dim (V / N∞) |
| 3 | present_observation_closes_iff |
The present reading needs no memory exactly when R L = A R for some linear A |
| Paper | Result | File | Status |
|---|---|---|---|
| §2.1–2.4, item T1 | Predictive quotient: equivalence, descent of the dynamics, Cayley–Hamilton closure, rank formula, minimality, closure of the present observation | PredictiveQuotient |
proved |
| §3.1 | Structural memory filtration: a coarser record cannot create distinctions | MemoryFiltration |
proved |
| Theorem P1, item T2 | Projector tangents: tangent equation ⇔ both diagonal blocks vanish, tangents are off-diagonal, metric positivity, unitary invariance of g and Ω, metric–curvature inequality (in block variables) |
ProjectorTangent |
proved |
| Theorem P1, item 2 | Block formulas: g = Re Tr(XY†), Ω = −2 Im Tr(XY†), Q = Tr(XY†) = g − (i/2) Ω |
ProjectorBlocks |
proved |
| Theorem P2 | Weak-value success-probability bound p₀ |A_w|² ≤ ⟨i|A²|i⟩ |
WeakValue |
proved |
| Theorem P3, item T3 | Path-ledger descent: an endpoint potential exists ⇔ every loop is silent; positive reachability is not antisymmetric | PathLedger |
proved |
| Corollaries P4.1–P4.3 | Common gap scaling µⱼ − µₖ = τ̇ (Eⱼ − Eₖ); rate and shift unique for non-scalar H₀; rescaling H ↦ sH keeps every eigenvector but scales every gap |
CommonClock |
proved |
| Theorem P5 | Finite return: a difference returns to the observed sector at some step ⇔ within the first dim V steps |
FiniteReturn |
proved |
| Theorem P5, continuous time | R e^{tL} Δ = 0 for every real t (past and future) ⇔ R Lᵏ Δ = 0 for every k |
ContinuousReturn |
proved |
| Theorem P6, P6.1 | Entropy/correlation ledger ΔS_S − Σ β_r Q_r = ΔI + Σ ΔD_r, from the two stated operator facts; the integrated second law, from that identity plus nonnegativity of multi-information and of relative entropy (taken as hypotheses) |
EntropyLedger |
proved under the stated hypotheses |
| Theorem P6, unitarity | Von Neumann entropy is unitarily invariant, S(UρU†) = S(ρ), proved with Physlib's quantum-information library; P6 restated with that hypothesis discharged (only the total entropy is a quantum-state entropy; the system and reservoir entropies remain real parameters) |
UnitaryEntropy |
proved |
| §9.2 | Affine phase readout: the angular numerator AD − BC is constant |
PhaseSilence |
proved |
| §9.3, item T4 | Moving-frame identity iħχ̇ = (R†HR − iħR†Ṙ)χ and the common-clock form |
MovingFrame |
proved (instantaneous identity) |
| §11.4 | Equal-endpoint witness: |a − ib e^{iΦ}|² = a² + b² + 2ab sin Φ, ⟨Z⟩ = sin 2θ sin Φ, equal present populations, probe blind to π − Φ |
EqualEndpointWitness |
proved |
| §12.6 | Friction on projector rotations: pair friction = reweighted intrinsic metric, Gibbs sign, two-level tanh(β∆/2), zero-temperature band |
FrictionMetric |
proved (pairwise formulas) |
| Theorem 13.1(3–4) | Torus-knot end indices k ≡ a (mod p), k ≡ ±b (mod q); silent sector ⇔ pq ε_μ + ε_λ + a/2 ∈ ℤ |
PhaseSilence |
proved |
| §15.5 | Sector reduction: (1−P)HP = 0 ⇔ HP = PHP; redistribution vanishes ⇔ [H, P] = 0 |
SectorReduction |
proved |
| T5–T10 | Weak-history readout protocol, thermodynamic repair for a Davies bath, trefoil / 8₁₉ APS family, light-path material realization, Maxwell–Dirac gate, arrow model | — | open |
Every push runs the proof check on GitHub:
- Build: every module compiles against Lean v4.34.1, Mathlib
v4.34.1and Physlib. - Independent replay: every module is re-checked by Lean's separate kernel checker.
- Axiom audit: every named theorem depends only on
propext,Classical.choiceandQuot.sound. Nosorry, no project axioms, nonative_decide. - False controls: thirteen deliberately false statements must fail to compile, and fail for a mathematical reason, not a typo. This shows the checker can say no.
To check it yourself with Lean installed:
lake exe cache get
lake build
python3 scripts/verify.pyThe Dirac Time of the Gigantefermion, Jeromie Beasley. DOI 10.5281/zenodo.22978630.
Citation metadata is in CITATION.cff. The Lean code and scripts are released under the MIT License and the written text under CC BY 4.0; see LICENSING.md. How AI tools were used is stated in AI_USE.md.