Collection · Catalogue · Verification · Lean status
Result
The values below 1/6 are exactly (P−1)/(6P) for P ≥ 7 congruent to ±1 modulo 6. Cordella’s nine sporadic sextuples exhaust the exceptions to the three known families.
Proof and sources
Read the proof for the original problem, assumptions, argument, and cited sources. These notes are unrefereed.
Verification
Run from the repository root:
python3 results/six-speed-spectrum/verify.pySee the verification guide for dependencies and the distinction between computational checks and formal proofs.
Formalization
The realization direction is complete in Lean. For every P ≥ 7 congruent to 1 or 5 modulo six, the formal proof constructs six distinct positive integer speeds with gcd one and exact global loneliness . This includes both infinite branches, the progression parameter zero, and the separate P=11 case. Each value is attained, and the corresponding upper bound holds at every real time.
The exact 11-module closure passed fresh strict compilation, an audit of all 392 owned declarations including 30 private declarations, and independent stock-kernel replay of all 11 modules. All 15 required endpoints are theorems. Exact scope · Raw receipt.
The general pair-sum maximum theorem and displayed finite examples also have checked global values. Universal classification of the three families and exhaustiveness outside them remain incomplete in Lean.
Reproduce the realization gate from the repository root with the pinned Lean toolchain and Mathlib cache:
python3 scripts/verify_six_spectrum_realization.pyActual lattice and conditional norm components
The registered components construct the full integer-point lattice of every rational coordinate subspace, prove its span and rank, and identify its squared covolume with the Gram determinant of an actual integer lattice basis. They prove squared covolume at least one. In rank four in six coordinates, the lower bound is six when the subspace contains an integer vector whose coordinates are nonzero and pairwise unequal up to sign.
Suppose an integer matrix has r columns lying in a rational subspace of dimension r, and its Gram determinant is positive. The proof gives an exact factorization of the column Gram determinant as a nonzero integer square times the full lattice's squared covolume. It therefore bounds the full lattice's squared covolume by the column Gram determinant. The endpoints do not identify that integer with the cardinality of a finite quotient.
The rectangular components also prove integer-parameter recovery under an integer left inverse, primitive basis transfer, Cauchy--Binet and integer Gram identities, and the literal map Gram determinants 3, 9 and 8. The six-by-four Gram lower bound assumes positive Gram determinant, nonzero rows and rows that are pairwise unequal up to sign; it derives the bound six internally.
The original norm constants 9075, 54450 and 163350 now use the covolume of the actual full integer-point lattice. The 9075 theorem assumes rank four, containment of the speed vector, and distinct positive primitive speeds. The 54450 and 163350 theorems hold for every rational coordinate subspace, without a rank or containment premise. All three norm theorems retain hfour : CoordinateSubspace.LonelyRunnerBound 4 and hplane, the containing-plane gap hypothesis, as well as positivity, primitivity and loneliness at most 1/6. These are component theorems. The unconditional plane-gap deduction and the remaining exhaustive classification are not supplied here; the six-speed spectrum result remains partially formalized.
Conditional two-parameter radius estimate
For an integer matrix with two columns and a finite nonempty row set, the theorem uses the actual real column span and the actual line generated by a primitive signed integer parameter pair. The primitive premise is Int.gcd x y = 1 on the parameters. For any real upper bound C on every squared row norm, plane_value_le bounds the plane value by the line value plus 1/2 * sqrt(C / (x^2+y^2)).
If the actual plane value is at least 2/11 and the actual line value is strictly below 1/6, the proof gives x^2+y^2 < 1089*C. Taking the exact finite maximum of the squared row norms gives the natural-number bound x.natAbs^2 + y.natAbs^2 < 1089 * maxRowSq A. An explicit coordinate-equality theorem carries this bound to a natural six-speed vector. It does not separately require the natural speed vector to be positive, distinct or primitive, and adds no rank, integer-left-inverse, four-speed or six-speed-classification premise.
The plane gap remains an assumption. This theorem does not classify the 78 maps or complete the six-speed spectrum.
Six-speed norm bounds with the four-speed input discharged
The three new six-speed lattice norm bounds supply the proved four-speed runner theorem internally. They retain the containing-plane gap, positivity and primitivity of the natural speed vector, and loneliness at most 1/6. The bound with constant 9075 additionally assumes a rational subspace of rank four containing the speed vector and pairwise distinct speeds. The bounds with constants 54450 and 163350 apply to every rational coordinate subspace, without a rank, containment or distinctness premise. The lattice is the actual full integer-point lattice.
The new endpoints end in _of_plane_gap. They remove the explicit hfour input and retain hplane : SixSpectrumNormGeometry.ContainingPlaneGap v. The old endpoints keep their original signatures. These results do not prove the containing-plane gap or complete the six-speed classification.