Main theorem of cycle 2.
ProvedDiophantineLattice.frac_shortest_isInhomMinaether-catalognovelty
Main theorem of cycle 2. For a shortest vector v and any r ≥ 2, the spectral gap
of the non-homogeneous form at the r-torsion shift v/r is exactly λ₁/r².
theorem DiophantineLattice.frac_shortest_isInhomMin{B : Matrix (Fin n) (Fin n) ℚ} (hpd : PosDef B) {lam : ℚ}
(h : IsMinEnergy B lam) {v : Fin n → ℤ} (hv : form B (emb v) = lam) {r : ℤ} (hr : 2 ≤ r) :
IsInhomMin B (fracPt v r) (lam / (r : ℚ) ^ 2) := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Novelty/DiophantineLatticeTorsionGap.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.
Preamble
-- Thm stub generated from Novelty/DiophantineLatticeTorsionGap.lean
import Mathlib
import Definitions.Def_Novelty_DiophantineLatticeShiftedTheta
import Definitions.Def_Novelty_DiophantineLatticeSpectralGap
import Definitions.Def_Novelty_DiophantineLatticeTorsionGap
/-!
# Torsion refinement: spectral gaps at `r`-torsion points, and even multiplicities
Cycle 2 of the research thread. `Novelty/DiophantineLatticeSpectralGap.lean` proved that the
non-homogeneous form `x ↦ Q(x - v/2)` attached to a shortest lattice vector `v` has spectral
gap exactly `λ₁/4`. Here the `2` is replaced by an arbitrary `r ≥ 2`, i.e. the shift is an
arbitrary `r`-torsion point of `(L ⊗ ℚ)/L` of the special shape `v/r`:
* `torsion_gap_ge` : if `v ∉ rL` then `Q(v/r - m) ≥ λ₁/r²` for every lattice point `m`;
* `frac_shortest_isInhomMin` : for a shortest `v` and `r ≥ 2` the spectral gap at `v/r` is
*exactly* `λ₁/r²` — the `r = 2` case is the previous cycle's main theorem;
* `torsion_no_integral_solution` : the Diophantine corollary, `Q(x - v/r) = c` is unsolvable
in integers for `c < λ₁/r²`.
A second, independent structural theorem concerns *multiplicities* rather than sizes: the
antipodal involution `m ↦ v - m` acts freely on the solution set of `Q(x - v/2) = c`, hence
* `halfPt_multiplicity_even` : every coefficient of the theta series of the non-homogeneous
form at a half shortest vector is **even**.
Finally the diagonal case is worked out completely: `diagonal_isMinEnergy` computes `λ₁` and
`diagonal_covering_radius_least` / `diagonal_covering_le` compute the covering radius² as
`(Σ aᵢ)/4`, so that the ratio covering/packing is `(Σ aᵢ)/(min aᵢ)`, unbounded.
-- !-- Lab Notes -- !--
Hypothesis (Hypothesizer): the factor `1/4` of cycle 1 is the `r = 2` shadow of a general
`1/r²` law indexed by the torsion order of the shift; moreover multiplicities in the shifted
theta series should always be even.
Experiment (Experimenter): the halving identity `Q(v/r - m) = Q(v - rm)/r²` and the
obstruction `v ∈ rL ⇒ Q(v) = r²Q(v/r) ≥ r²λ₁ > λ₁` both survive verbatim for every `r ≥ 2`;
the involution `m ↦ v - m` is fixed-point free precisely because `v ∉ 2L`, which is the *same*
obstruction. So one lemma (`sub_two_smul_ne_zero` and its `r`-analogue) drives both results.
Analysis (Analyst): the two cycle-1 phenomena are thus explained by a single structural fact —
a shortest vector is primitive modulo `r` for every `r ≥ 2`. What genuinely fails to
generalise is the involution: for `r ≥ 3` the map `m ↦ v - m` is not a symmetry of the shifted
form, so "even multiplicity" is special to `2`-torsion (it becomes "divisible by the order of
the stabiliser-free symmetry group" in general — see FUTURE_DIRECTIONS).
Critique (Critic): `halfPt_multiplicity_even` is stated for an arbitrary `Finset` enumerating
the solutions, so it is not vacuous for `S = ∅` only; `frac_shortest_isInhomMin` includes the
attainment clause, so it is an identity, not a one-sided bound.
Synthesis (PI): spectral gap `λ₁/r²` at `r`-torsion shifts, even theta coefficients at
`2`-torsion shifts, and an exact covering radius `(Σaᵢ)/4` in the diagonal case.
-/
open DiophantineLattice
open Finset
variable {n : ℕ}
/-! ## Gaps at `r`-torsion shifts -/Formal statement
theorem DiophantineLattice.frac_shortest_isInhomMin{B : Matrix (Fin n) (Fin n) ℚ} (hpd : PosDef B) {lam : ℚ}
(h : IsMinEnergy B lam) {v : Fin n → ℤ} (hv : form B (emb v) = lam) {r : ℤ} (hr : 2 ≤ r) :
IsInhomMin B (fracPt v r) (lam / (r : ℚ) ^ 2) := by sorrySource