Peterfalvi (5.4)–(5.6)/(6.6): norm inequalities, -decomposition, coherence, degree gap
DefinitionOddOrder_Peterfalvi_S07_Coherence_NormInequalitiesfeit-thompsonfinite-groupsgroup-theoryishida-odd-order
Part of the coherence development of Peterfalvi §7 (first half of (5.4)–(5.6) and (6.6)): norm inequalities, the -decomposition, the definition of coherence, and the degree gap.
Main definitions: CharacterPsiDecomposition (the setup of (5.4): bundles its hypotheses, with smart constructor ofProjection via integral projection), IsCoherent (Peterfalvi (5.1): is coherent for if it admits an integral extension) and GeneralHypothesis (Peterfalvi Hypothesis (5.2), the book-faithful carrier, p. 25), with lemmas that the difference sublattice contains and .
The bundle also proves 31 supporting lemmas used by later statements.
Definition code
/-
Copyright (c) 2026 Yawara Ishida. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Yawara Ishida
-/
-- From github.com/yawara/odd-order @ 20a6fccbb63c (Apache-2.0; LICENSE at that commit):
-- OddOrder/Peterfalvi/S07_Coherence/NormInequalities.lean
-- Generated for Prove2Me by deleting declarations; the remaining commands are verbatim except:
-- `private` dropped where another file uses the declaration, and a solution's main theorem
-- renamed `_root_.solution` (platform convention).
import Definitions.Def_OddOrder_Peterfalvi_S07_Coherence_DifferenceImage
namespace OddOrder.Peterfalvi.S07
open OddOrder.RepresentationTheory
variable {L G : Type*} [Group L] [Group G]
open OddOrder.RepresentationTheory in
/-- **Peterfalvi (5.4) setup.** Bundles the hypotheses of (5.4): the orthonormal
image family `R(χ)`, the auxiliary isometry `τ₁` agreeing with `τ` on `χ - χ̄`, the
decomposition `(χ - ψ)^{τ₁} = X - Y` with `X ∈ ℤ[R(χ)]` (recorded via integer
coefficients `coeff`) and `Y ⊥ R(χ)`, and the orthogonality relations
`(χ,ψ) = (χ̄,ψ) = (χ,χ̄) = 0`. -/
structure CharacterPsiDecomposition (τ : IntegralCharacterMap L G)
(χ ψ : ClassFunction L ℂ)
[Fintype L] [Fintype G]
[Invertible (Nat.card L : ℂ)] [Invertible (Nat.card G : ℂ)] where
/-- The orthonormal image family `R(χ)` with `(χ - χ̄)^τ = ∑_{α ∈ R(χ)} α`. -/
imageFamily : OrthonormalCharacterImageFamily (L := L) (G := G) τ χ
/-- The auxiliary isometry `τ₁` on `ℤ[χ - ψ, χ - χ̄]`. -/
tau1 : IntegralCharacterMap L G
/-- `τ₁` preserves the inner product on the sponsoring lattice `ℤ[χ, χ̄, ψ]`.
This is **lattice-relative**, not a global `IsIntegralIsometry` on all of `CF(L)`:
it is the same Round-13 weakening already applied to `IsCoherent.extension_inner_eq`.
In Feit–Thompson `dim CF(L) > dim CF(G)`, so a global isometry of the
character-difference lattice into `CF(G)` generically does not exist, but the Dade
isometry / running extension *does* preserve the inner product on the supported
sublattice `ℤ[χ, χ̄, ψ]` — which is all the (5.4.b)/(5.5)/(5.6.2) proofs use (every
access is on `χ - ψ` or `χ - χ̄`, both in `ℤ[χ, χ̄, ψ]`). -/
tau1_inner_eq_on_support :
∀ φ ζ : ClassFunction L ℂ,
φ ∈ zSpan (L := L) {χ - χ.conj, χ - ψ} → ζ ∈ zSpan (L := L) {χ - χ.conj, χ - ψ} →
ClassFunction.inner (tau1 φ) (tau1 ζ) = ClassFunction.inner φ ζ
/-- `τ₁` coincides with `τ` on `χ - χ̄`. -/
tau1_agrees : tau1 (χ - χ.conj) = τ (χ - χ.conj)
/-- The image side `X` of `(χ - ψ)^{τ₁} = X - Y`. -/
X : ClassFunction G ℂ
/-- The orthogonal side `Y` of `(χ - ψ)^{τ₁} = X - Y`. -/
Y : ClassFunction G ℂ
/-- The decomposition `(χ - ψ)^{τ₁} = X - Y`. -/
tau1_image : tau1 (χ - ψ) = X - Y
/-- The integer coefficients of `X ∈ ℤ[R(χ)]`. -/
coeff : ClassFunction G ℂ → ℤ
/-- `X = ∑_{α ∈ R(χ)} (coeff α) • α`, i.e. `X ∈ ℤ[R(χ)]`. -/
X_eq : X = ∑ α ∈ imageFamily.imageSet, (coeff α : ℂ) • α
/-- `Y` is orthogonal to `R(χ)`. -/
Y_orthogonal : ∀ α ∈ imageFamily.imageSet, ClassFunction.inner Y α = 0
/-- `(χ, ψ) = 0`. -/
chi_psi_orthogonal : ClassFunction.inner χ ψ = 0
/-- `(χ̄, ψ) = 0`. -/
chiConj_psi_orthogonal : ClassFunction.inner χ.conj ψ = 0
/-- `(χ, χ̄) = 0` (the distinct elements `χ`, `χ̄ ∈ S` are orthogonal by (5.2.c)). -/
chi_chiConj_orthogonal : ClassFunction.inner χ χ.conj = 0
namespace CharacterPsiDecomposition
open OddOrder.RepresentationTheory
variable {τ : IntegralCharacterMap L G} {χ ψ : ClassFunction L ℂ}
variable [Fintype L] [Fintype G]
variable [Invertible (Nat.card L : ℂ)] [Invertible (Nat.card G : ℂ)]
open scoped Classical in
/-- **Smart constructor for `CharacterPsiDecomposition` via integral projection.**
The hard `X`/`Y`/`coeff` content of the (5.4) decomposition `(χ − ψ)^{τ₁} = X − Y` — the
integral `R(χ)`-projection `X = ∑_{α ∈ R(χ)} (coeff α)•α ∈ ℤ[R(χ)]` and the orthogonal residual
`Y ⊥ R(χ)` — is *computed*, not posited, from the single number-theoretic input
`htau1_mem : (χ − ψ)^{τ₁} ∈ ZIrr G` (the τ₁-image is a virtual character of `G`). This is the
orthogonal projection of `(χ − ψ)^{τ₁}` onto the ZIrr-orthonormal family `R(χ)`
(`exists_intProjection_of_orthonormal_ZIrr`, integer coefficients because `R(χ) ⊆ ZIrr G`); the
residual `Y` is its orthogonal complement.
The remaining inputs are exactly the *structural* data a per-step (5.4)/(5.5)/(5.6.1) construction
must still supply:
* `imageFamily` — the signed-irreducible family `R(χ)` with `(χ − χ̄)^τ = ∑_{α} α` (from the Dade
data / §3 keystone, `OrthonormalCharacterImageFamily`);
* `tau1`, `htau1_inner_eq`, `htau1_agrees` — the (5.4) auxiliary isometry `τ₁`, supplied as
**lattice-relative** inner-preservation on the sponsoring lattice `ℤ[χ, χ̄, ψ]` (the Round-13
weakening: the Dade isometry / running extension preserves `⟨·,·⟩` only on the supported
sublattice, *not* globally on `CF(L)`), agreeing with `τ` on `χ − χ̄`;
* the three (5.2.c)/(5.4) orthogonality scalars `⟨χ, ψ⟩ = ⟨χ̄, ψ⟩ = ⟨χ, χ̄⟩ = 0`.
This isolates the genuinely missing primitives (the Dade `R(χ)` extractor and the `τ₁`
isometry-extension) as the only residual: once those are built, this constructor delivers the full
`CharacterPsiDecomposition` with its `X`/`Y`/`coeff`/`X_eq`/`Y_orthogonal` fields populated by the
projection. -/
noncomputable def ofProjection
(imageFamily : OrthonormalCharacterImageFamily (L := L) (G := G) τ χ)
(tau1 : IntegralCharacterMap L G)
(htau1_inner_eq : ∀ φ ζ : ClassFunction L ℂ,
φ ∈ zSpan (L := L) {χ - χ.conj, χ - ψ} → ζ ∈ zSpan (L := L) {χ - χ.conj, χ - ψ} →
ClassFunction.inner (tau1 φ) (tau1 ζ) = ClassFunction.inner φ ζ)
(htau1_agrees : tau1 (χ - χ.conj) = τ (χ - χ.conj))
(htau1_mem : tau1 (χ - ψ) ∈ ZIrr G)
(hχψ : ClassFunction.inner χ ψ = 0)
(hχbarψ : ClassFunction.inner χ.conj ψ = 0)
(hχχbar : ClassFunction.inner χ χ.conj = 0) :
CharacterPsiDecomposition (L := L) (G := G) τ χ ψ :=
let proj := ClassFunction.exists_intProjection_of_orthonormal_ZIrr htau1_mem
imageFamily.mem_ZIrr imageFamily.orthonormal
{ imageFamily := imageFamily
tau1 := tau1
tau1_inner_eq_on_support := htau1_inner_eq
tau1_agrees := htau1_agrees
X := ∑ α ∈ imageFamily.imageSet, (proj.choose α : ℂ) • α
Y := -(tau1 (χ - ψ) - ∑ α ∈ imageFamily.imageSet, (proj.choose α : ℂ) • α)
tau1_image := by
-- `X − Y = X − (−(τ₁(χ−ψ) − X)) = τ₁(χ−ψ)`, pure algebra.
rw [sub_neg_eq_add]; abel
coeff := proj.choose
X_eq := rfl
Y_orthogonal := by
-- `⟨Y, α⟩ = −⟨τ₁(χ−ψ) − X, α⟩ = −(⟨τ₁(χ−ψ), α⟩ − coeff α) = −(coeff α − coeff α) = 0`.
intro α hα
have hcoeff : ClassFunction.inner (tau1 (χ - ψ)) α = (proj.choose α : ℂ) :=
proj.choose_spec.choose_spec.1 α hα
rw [ClassFunction.inner_neg_left, ClassFunction.inner_sub_left,
inner_orthonormalSum_eq_coeff imageFamily.orthonormal hα, hcoeff, sub_self, neg_zero]
chi_psi_orthogonal := hχψ
chiConj_psi_orthogonal := hχbarψ
chi_chiConj_orthogonal := hχχbar }
omit [Fintype L] in
omit [Invertible (Nat.card L : ℂ)] in
/-- `χ - χ̄ ∈ ℤ[χ−χ̄, χ−ψ]`: the difference sublattice contains the conjugate difference (it is a
generator). -/
theorem chi_sub_conj_mem_zSpan_support :
χ - χ.conj ∈ zSpan (L := L) {χ - χ.conj, χ - ψ} :=
Submodule.subset_span (by simp)
omit [Fintype L] in
omit [Invertible (Nat.card L : ℂ)] in
/-- `χ - ψ ∈ ℤ[χ−χ̄, χ−ψ]`: the difference sublattice contains the `ψ`-difference (it is a
generator). -/
theorem chi_sub_psi_mem_zSpan_support :
χ - ψ ∈ zSpan (L := L) {χ - χ.conj, χ - ψ} :=
Submodule.subset_span (by simp)
/-- `‖X‖² = ∑_{α ∈ R(χ)} (coeff α)²` (Parseval). -/
theorem inner_self_X (D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner D.X D.X = ∑ α ∈ D.imageFamily.imageSet, (D.coeff α : ℂ) ^ 2 := by
rw [D.X_eq]
exact inner_self_orthonormalSum_eq_sum_sq D.imageFamily.orthonormal
/-- `⟨X, ∑_{α ∈ R(χ)} α⟩ = ∑_{α ∈ R(χ)} coeff α`. -/
theorem inner_X_sum (D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner D.X (∑ α ∈ D.imageFamily.imageSet, α) =
∑ α ∈ D.imageFamily.imageSet, (D.coeff α : ℂ) := by
rw [D.X_eq]
exact inner_orthonormalSum_sum_eq_sum_coeff D.imageFamily.orthonormal
/-- `Y ⊥ R(χ)` extends to the sum: `⟨Y, ∑_{α ∈ R(χ)} α⟩ = 0`. -/
theorem inner_Y_sum (D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner D.Y (∑ α ∈ D.imageFamily.imageSet, α) = 0 := by
rw [inner_sum_right]
exact Finset.sum_eq_zero fun α hα => D.Y_orthogonal α hα
/-- **The keystone identity of (5.4.a).** `‖χ‖² = ∑_{α ∈ R(χ)} coeff α`.
`‖χ‖² = ⟨χ - ψ, χ - χ̄⟩` (using the three orthogonality relations), `= ⟨X - Y, ∑ α⟩`
(by the isometry of `τ₁` and `τ₁ = τ` on `χ - χ̄`), `= ⟨X, ∑ α⟩` (since `Y ⊥ R(χ)`),
`= ∑ coeff α`. -/
theorem inner_self_chi_eq_sum_coeff
(D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner χ χ = ∑ α ∈ D.imageFamily.imageSet, (D.coeff α : ℂ) := by
-- Step 1: `⟨χ, χ⟩ = ⟨χ - ψ, χ - χ̄⟩`.
have hpsi_chi : ClassFunction.inner ψ χ = 0 := by
rw [OddOrder.RepresentationTheory.inner_conj_symm, D.chi_psi_orthogonal, star_zero]
have hpsi_chiConj : ClassFunction.inner ψ χ.conj = 0 := by
rw [OddOrder.RepresentationTheory.inner_conj_symm, D.chiConj_psi_orthogonal, star_zero]
have hsrc : ClassFunction.inner χ χ =
ClassFunction.inner (χ - ψ) (χ - χ.conj) := by
rw [ClassFunction.inner_sub_left, ClassFunction.inner_sub_right,
ClassFunction.inner_sub_right, D.chi_chiConj_orthogonal, hpsi_chi, hpsi_chiConj]
ring
-- Step 2: transport across `τ₁` (isometry, agreeing with `τ` on `χ - χ̄`).
have himg : ClassFunction.inner (χ - ψ) (χ - χ.conj) =
ClassFunction.inner (D.X - D.Y) (∑ α ∈ D.imageFamily.imageSet, α) := by
rw [← D.tau1_inner_eq_on_support (χ - ψ) (χ - χ.conj)
chi_sub_psi_mem_zSpan_support chi_sub_conj_mem_zSpan_support,
D.tau1_image, D.tau1_agrees, D.imageFamily.image_eq]
-- Step 3: `Y ⊥ R(χ)` drops `Y`; coefficient recovery finishes.
rw [hsrc, himg, ClassFunction.inner_sub_left, D.inner_Y_sum, sub_zero, D.inner_X_sum]
/-- **Peterfalvi (5.4.a):** `‖X‖² ≥ ‖χ‖²`.
By the keystone identity `‖χ‖² = ∑ coeff α` and Parseval `‖X‖² = ∑ (coeff α)²`, the
integer Cauchy–Schwarz `∑ coeff α ≤ ∑ (coeff α)²` gives the inequality. -/
theorem inner_self_chi_re_le_inner_self_X
(D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
(ClassFunction.inner χ χ).re ≤ (ClassFunction.inner D.X D.X).re := by
classical
have hχ : ClassFunction.inner χ χ =
((∑ α ∈ D.imageFamily.imageSet, D.coeff α : ℤ) : ℂ) := by
rw [D.inner_self_chi_eq_sum_coeff]; push_cast; ring
have hX : ClassFunction.inner D.X D.X =
((∑ α ∈ D.imageFamily.imageSet, (D.coeff α) ^ 2 : ℤ) : ℂ) := by
rw [D.inner_self_X]; push_cast; ring
rw [hχ, hX, Complex.intCast_re, Complex.intCast_re]
exact_mod_cast finset_sum_le_sum_sq D.imageFamily.imageSet D.coeff
/-- `⟨χ, χ⟩` is the integer cast `(∑ coeff α : ℤ)`. -/
theorem inner_self_chi_eq_intCast
(D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner χ χ = ((∑ α ∈ D.imageFamily.imageSet, D.coeff α : ℤ) : ℂ) := by
rw [D.inner_self_chi_eq_sum_coeff]; push_cast; ring
/-- `⟨X, X⟩` is the integer cast `(∑ (coeff α)² : ℤ)`. -/
theorem inner_self_X_eq_intCast
(D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner D.X D.X = ((∑ α ∈ D.imageFamily.imageSet, (D.coeff α) ^ 2 : ℤ) : ℂ) := by
rw [D.inner_self_X]; push_cast; ring
/-- `X ⊥ Y`: `⟨X, Y⟩ = 0` (since `X ∈ ℤ[R(χ)]` and `Y ⊥ R(χ)`). -/
theorem inner_X_Y (D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner D.X D.Y = 0 := by
rw [D.X_eq, inner_sum_left]
refine Finset.sum_eq_zero fun α hα => ?_
rw [ClassFunction.inner_smul_left,
OddOrder.RepresentationTheory.inner_conj_symm, D.Y_orthogonal α hα, star_zero, mul_zero]
/-- `‖X - Y‖² = ‖X‖² + ‖Y‖²` (Pythagoras, since `X ⊥ Y`). -/
theorem inner_self_X_sub_Y (D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner (D.X - D.Y) (D.X - D.Y) =
ClassFunction.inner D.X D.X + ClassFunction.inner D.Y D.Y := by
have hYX : ClassFunction.inner D.Y D.X = 0 := by
rw [OddOrder.RepresentationTheory.inner_conj_symm, D.inner_X_Y, star_zero]
rw [ClassFunction.inner_sub_left, ClassFunction.inner_sub_right,
ClassFunction.inner_sub_right, D.inner_X_Y, hYX]
ring
/-- `‖χ - ψ‖² = ‖χ‖² + ‖ψ‖²` (Pythagoras, since `(χ, ψ) = (ψ, χ) = 0`). -/
theorem inner_self_chi_sub_psi (D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner (χ - ψ) (χ - ψ) =
ClassFunction.inner χ χ + ClassFunction.inner ψ ψ := by
have hpsi_chi : ClassFunction.inner ψ χ = 0 := by
rw [OddOrder.RepresentationTheory.inner_conj_symm, D.chi_psi_orthogonal, star_zero]
rw [ClassFunction.inner_sub_left, ClassFunction.inner_sub_right,
ClassFunction.inner_sub_right, D.chi_psi_orthogonal, hpsi_chi]
ring
/-- The total-norm identity `‖χ‖² + ‖ψ‖² = ‖X‖² + ‖Y‖²` from the isometry of `τ₁`. -/
theorem inner_self_chi_add_psi_eq
(D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
ClassFunction.inner χ χ + ClassFunction.inner ψ ψ =
ClassFunction.inner D.X D.X + ClassFunction.inner D.Y D.Y := by
rw [← D.inner_self_chi_sub_psi, ← D.inner_self_X_sub_Y, ← D.tau1_image,
D.tau1_inner_eq_on_support (χ - ψ) (χ - ψ)
chi_sub_psi_mem_zSpan_support chi_sub_psi_mem_zSpan_support]
/-- **Peterfalvi (5.6.2) opening bound:** `‖Y‖² ≤ ‖ψ‖²`.
This is the first step of (5.6.2) ("We note first that `‖χ‖² + a²‖χ₁‖² = ‖X‖² + ‖Y‖²`;
by (5.4.a) `‖X‖² ≥ ‖χ‖²`, so `‖Y‖² ≤ a²‖χ₁‖²`"). It follows directly from the total-norm
identity `‖χ‖² + ‖ψ‖² = ‖X‖² + ‖Y‖²` and (5.4.a) `‖χ‖² ≤ ‖X‖²`, with `ψ = a·χ₁` in the
application. Stated for the general `ψ` of the (5.4) decomposition. -/
theorem inner_self_Y_re_le_inner_self_psi
(D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ) :
(ClassFunction.inner D.Y D.Y).re ≤ (ClassFunction.inner ψ ψ).re := by
have htotal : (ClassFunction.inner χ χ).re + (ClassFunction.inner ψ ψ).re =
(ClassFunction.inner D.X D.X).re + (ClassFunction.inner D.Y D.Y).re := by
have := congrArg Complex.re D.inner_self_chi_add_psi_eq
simpa [Complex.add_re] using this
have hXge := D.inner_self_chi_re_le_inner_self_X
linarith
/-- **Peterfalvi (5.4.b).** If `‖Y‖² ≥ ‖ψ‖²`, then `‖X‖² = ‖χ‖²`, `‖Y‖² = ‖ψ‖²`
and `X = ∑_{α ∈ E} α` for some `E ⊆ R(χ)` with `|E| = ‖χ‖²`.
The total-norm identity `‖χ‖² + ‖ψ‖² = ‖X‖² + ‖Y‖²` together with `‖X‖² ≥ ‖χ‖²`
(5.4.a) and the hypothesis `‖Y‖² ≥ ‖ψ‖²` forces both inequalities to be equalities.
The norm equality `‖χ‖² = ‖X‖²` reads `∑ coeff α = ∑ (coeff α)²`, so by the tightness
of integer Cauchy–Schwarz each `coeff α ∈ {0, 1}`; then `E = {α | coeff α = 1}` and
`X = ∑_{α ∈ E} α`. The cardinality `|E| = ∑ coeff α = ‖χ‖²` is the form used in
(5.6.3) to compute `‖χ̄^{τ₂}‖² = |R(χ)| - |E| = ‖χ - χ̄‖² - ‖χ‖² = ‖χ̄‖²`. -/
theorem norm_eq_and_X_eq_sum_of_norm_Y_ge
(D : CharacterPsiDecomposition (L := L) (G := G) τ χ ψ)
(hY : (ClassFunction.inner ψ ψ).re ≤ (ClassFunction.inner D.Y D.Y).re) :
(ClassFunction.inner χ χ).re = (ClassFunction.inner D.X D.X).re ∧
(ClassFunction.inner ψ ψ).re = (ClassFunction.inner D.Y D.Y).re ∧
∃ E ⊆ D.imageFamily.imageSet, D.X = ∑ α ∈ E, α ∧
(E.card : ℂ) = ClassFunction.inner χ χ := by
classical
-- The `.re` total-norm identity.
have htotal : (ClassFunction.inner χ χ).re + (ClassFunction.inner ψ ψ).re =
(ClassFunction.inner D.X D.X).re + (ClassFunction.inner D.Y D.Y).re := by
have := congrArg Complex.re D.inner_self_chi_add_psi_eq
simpa [Complex.add_re] using this
have hXge := D.inner_self_chi_re_le_inner_self_X
-- Both inequalities are forced to equalities.
have hχX : (ClassFunction.inner χ χ).re = (ClassFunction.inner D.X D.X).re := by linarith
have hψY : (ClassFunction.inner ψ ψ).re = (ClassFunction.inner D.Y D.Y).re := by linarith
refine ⟨hχX, hψY, ?_⟩
-- Tightness: `∑ coeff = ∑ coeff²` as integers.
have hsum_eq : (∑ α ∈ D.imageFamily.imageSet, D.coeff α) =
∑ α ∈ D.imageFamily.imageSet, (D.coeff α) ^ 2 := by
have h1 := hχX
rw [D.inner_self_chi_eq_intCast, D.inner_self_X_eq_intCast,
Complex.intCast_re, Complex.intCast_re] at h1
exact_mod_cast h1
have hcoeff01 : ∀ α ∈ D.imageFamily.imageSet, D.coeff α = 0 ∨ D.coeff α = 1 :=
(finset_sum_eq_sum_sq_iff D.imageFamily.imageSet D.coeff).mp hsum_eq
-- `E = {α | coeff α = 1}`.
set E := D.imageFamily.imageSet.filter (fun α => D.coeff α = 1) with hE
refine ⟨E, Finset.filter_subset _ _, ?_, ?_⟩
· -- `X = ∑_{α ∈ E} α`.
rw [D.X_eq, hE, Finset.sum_filter]
refine Finset.sum_congr rfl fun α hα => ?_
rcases hcoeff01 α hα with h0 | h1
· rw [h0, if_neg (by norm_num), Int.cast_zero, zero_smul]
· rw [h1, if_pos rfl, Int.cast_one, one_smul]
· -- `|E| = ∑ coeff α = ‖χ‖²`: on `R(χ)`, `coeff α = (if coeff α = 1 then 1 else 0)`.
have hcard : (E.card : ℤ) = ∑ α ∈ D.imageFamily.imageSet, D.coeff α := by
rw [hE, Finset.card_filter, Nat.cast_sum]
refine Finset.sum_congr rfl fun α hα => ?_
rcases hcoeff01 α hα with h0 | h1
· rw [h0, if_neg (by norm_num), Nat.cast_zero]
· rw [h1, if_pos rfl, Nat.cast_one]
rw [D.inner_self_chi_eq_intCast]
exact_mod_cast hcard
/-- **Peterfalvi (5.5).** Applying (5.4) with `ψ = 0`: the hypothesis `‖Y‖² ≥ ‖ψ‖² = 0`
of (5.4.b) holds automatically (the inner product is positive semidefinite), so `Y = 0`
and `χ^{τ₁} = X = ∑_{α ∈ E} α` for some `E ⊆ R(χ)` with `|E| = ‖χ‖²`.
`‖ψ‖² = ⟨0, 0⟩ = 0 ≤ ‖Y‖²` feeds (5.4.b), whose norm equality `‖Y‖² = ‖ψ‖² = 0` forces
`Y = 0` by positive definiteness of `ClassFunction.inner`. Then
`χ^{τ₁} = (χ - 0)^{τ₁} = X - Y = X`. -/
theorem eq_sum_of_psi_eq_zero
(D : CharacterPsiDecomposition (L := L) (G := G) τ χ 0) :
D.Y = 0 ∧ D.tau1 χ = D.X ∧
∃ E ⊆ D.imageFamily.imageSet, D.X = ∑ α ∈ E, α ∧
(E.card : ℂ) = ClassFunction.inner χ χ := by
-- `‖ψ‖² = ‖0‖² = 0 ≤ ‖Y‖²` by positive semidefiniteness.
have hψY : (ClassFunction.inner (0 : ClassFunction L ℂ) 0).re ≤
(ClassFunction.inner D.Y D.Y).re := by
rw [ClassFunction.inner_zero_left, Complex.zero_re]
exact inner_self_re_nonneg D.Y
obtain ⟨_, hYnorm, E, hEsub, hXsum, hEcard⟩ := D.norm_eq_and_X_eq_sum_of_norm_Y_ge hψY
-- `‖Y‖² = ‖0‖² = 0`, so `Y = 0` by positive definiteness.
have hY0 : D.Y = 0 := by
apply eq_zero_of_inner_self_re_eq_zero
rw [← hYnorm, ClassFunction.inner_zero_left, Complex.zero_re]
-- `χ^{τ₁} = (χ - 0)^{τ₁} = X - Y = X`.
have hτ1χ : D.tau1 χ = D.X := by
have := D.tau1_image
rw [sub_zero] at this
rw [this, hY0, sub_zero]
exact ⟨hY0, hτ1χ, E, hEsub, hXsum, hEcard⟩
end CharacterPsiDecomposition
/-- Peterfalvi (5.1): `τ` is coherent for `(S,A)` if it admits an integral
extension on `Z[S]` that is **isometric on the lattice `Z[S]`** and agrees with
`τ` on `Z[S,A]`.
The isometry hypothesis is **lattice-relative** (`extension_inner_eq`, restricted
to `φ, ψ ∈ Z[S] = zSpan S`), not a global `IsIntegralIsometry` on all of `CF(L)`.
This is the object Peterfalvi actually constructs in (5.6): in Feit–Thompson
`dim CF(L) > dim CF(G)`, so a global isometry of the character-difference lattice
into `CF(G)` generically does not exist, but the lattice-relative one
(`retarget_inner_eq_on`) does. Every downstream consumer uses only the
inner-product preservation on lattice elements `ζ ∈ S ⊆ Z[S]`. -/
structure IsCoherent (τ : IntegralCharacterMap L G)
(S : Set (ClassFunction L ℂ)) (A : Set L)
[Fintype L] [Fintype G]
[Invertible (Nat.card L : ℂ)] [Invertible (Nat.card G : ℂ)] where
nonzero : ∃ φ : ClassFunction L ℂ, φ ∈ zSupportedSpan (L := L) S A ∧ φ ≠ 0
extension : IntegralCharacterMap L G
extension_inner_eq :
∀ φ ψ : ClassFunction L ℂ, φ ∈ zSpan (L := L) S → ψ ∈ zSpan (L := L) S →
ClassFunction.inner (extension φ) (extension ψ) = ClassFunction.inner φ ψ
extends_on_supported :
∀ φ : ClassFunction L ℂ, φ ∈ zSupportedSpan (L := L) S A → extension φ = τ φ
/-- The coherent extension sends the **coherent lattice** `ℤ[S]` into `ℤ[Irr G]`.
Peterfalvi's coherent isometries are genuine virtual-character maps *on the coherent set*: the
(5.6.3) running extension is built by `retarget` from virtual-character target pairs
`{X, X̄} ⊂ ℤ[Irr G]`, sending each member of `S` to a virtual character. (It is *not* claimed on
all of `ℤ[Irr L]` — the base Dade map is only an isometry on the supported lattice, not a global
`ℤ[Irr]`-endomorphism.) This is the **ZIrr-codomain** the (6.6)/(6.8) `X`-chain needs: for an
irreducible member `χⱼ ∈ S`, `χⱼ ∈ ℤ[S]` gives `ν χⱼ ∈ ZIrr`, the load-bearing input of
`crux1_of_memberFamily`. -/
extension_mem_ZIrr : ∀ φ : ClassFunction L ℂ, φ ∈ zSpan (L := L) S → extension φ ∈ ZIrr G
/-- **Peterfalvi Hypothesis (5.2), verbatim** (p. 25) — the book-faithful carrier.
Identical to `Hypothesis` except in clause **(5.2.d)**, where the difference image `R(χ)` is an
arbitrary finite orthonormal family (`τ(χ − χ̄) = ∑_{α ∈ R(χ)} α`) rather than the *two-element*
signed pair `ε·(μ − ν)` of `CharacterDifferenceImage`.
That difference is not cosmetic. Taking norms in the two-element form gives
`‖τ(χ − χ̄)‖² = 2`, while the (5.2.b) isometry gives `‖χ − χ̄‖² = 2‖χ‖²` (using `⟨χ, χ̄⟩ = 0` and
`‖χ̄‖² = ‖χ‖²`) — so `Hypothesis` silently forces **every member to be irreducible**. The book
imposes no such thing: a reducible member has `‖χ‖² > 1` and `|R(χ)| = 2‖χ‖²`. This is exactly
what stops (5.7) from being provable at book strength over `Hypothesis` (issue 0157).
`Hypothesis.toGeneralHypothesis` exhibits the two-element carrier as the special case, via
`CharacterDifferenceImage.toOrthonormalImage`. -/
structure GeneralHypothesis (S : Set (ClassFunction L ℂ)) (A : Set L)
[Fintype L] [Fintype G]
[Invertible (Nat.card L : ℂ)] [Invertible (Nat.card G : ℂ)] where
/-- **(5.2.b)**: the linear map `τ`. -/
tau : IntegralCharacterMap L G
/-- **(5.2.b)**: `τ` is an isometry on the `A`-supported sublattice `ℤ[S, A]`. -/
tau_isometry_diff : ∀ ⦃φ ψ : ClassFunction L ℂ⦄,
φ ∈ zSupportedSpan (L := L) S A → ψ ∈ zSupportedSpan (L := L) S A →
ClassFunction.inner (tau φ) (tau ψ) = ClassFunction.inner φ ψ
/-- **(5.2.a)**: `S` is closed under complex conjugation. -/
conjugate_closed : OddOrder.Peterfalvi.S03.ClosedUnderConjugate S
/-- **(5.2.a)**: no member is real. -/
no_real_characters : OddOrder.Peterfalvi.S03.HasNoRealCharacters S
/-- **(5.2.c)**: distinct members are orthogonal. -/
pairwise_orthogonal : OddOrder.Peterfalvi.S03.PairwiseOrthogonal S
/-- **(5.2.d)**: `τ(χ − χ̄) = ∑_{α ∈ R(χ)} α` for an orthonormal `R(χ) ⊆ ℤ[Irr G]` of
**arbitrary size** — the book's clause, with no irreducibility of `χ` implied. -/
difference_image :
∀ ⦃χ : ClassFunction L ℂ⦄, χ ∈ S → OrthonormalCharacterImageFamily (L := L) (G := G) tau χ
/-- **(5.2.e)**: `φ ⊥ {χ, χ̄}` forces `R(φ) ⊥ R(χ)`. -/
difference_images_orthogonal :
∀ ⦃φ χ : ClassFunction L ℂ⦄ (hφ : φ ∈ S) (hχ : χ ∈ S),
ClassFunction.inner φ χ = 0 → ClassFunction.inner φ χ.conj = 0 →
(difference_image hφ).Orthogonal (difference_image hχ)
/-- Peterfalvi (5.2)-style hypotheses for coherence applications. This is a
carrier for later theorems, not a proof that the hypotheses hold in any
particular Dade situation.
⚠ The `difference_image` field is the **two-element** (5.3.a) shape, which forces every member of
`S` to be irreducible (see `GeneralHypothesis`, the book-faithful carrier, for the derivation).
Results that do not need that refinement should be stated over `GeneralHypothesis` and reached
from here through `toGeneralHypothesis`. -/
structure Hypothesis (S : Set (ClassFunction L ℂ)) (A : Set L)
[Fintype L] [Fintype G]
[Invertible (Nat.card L : ℂ)] [Invertible (Nat.card G : ℂ)] where
tau : IntegralCharacterMap L G
/-- **Lattice-relative (5.2)(b) isometry on the `A`-supported sublattice** (issues 9001, 0099):
`τ` preserves inner products on `ℤ[S, A] = {f ∈ ℤ[S] | supp f ⊆ A}` — Coq `subcoherent`
clause (b) `{in 'Z[S, L^#], isometry tau}` (`PFsection5.v:488`). A *global* `IsIntegralIsometry`
does **not** exist for the Feit–Thompson Dade map (`dim CF(L) > dim CF(G)`), and the pre-0099
all-member-difference form is **false** for mixed-degree families (`a − b` with `a(1) ≠ b(1)` is
not `A`-supported) — while every extension mechanism ((5.4)/(5.6)/(5.7)/(5.9a)) only ever
consumes `A`-supported `ℤ`-combinations. The Dade witnesses supply this field *unconditionally*
(any family, mixed degrees included) via `dadeIntegralCharacterMap_inner_eq_of_supported`;
member-difference consumers use the derived `tau_isometry_memberDiff`. -/
tau_isometry_diff : ∀ ⦃φ ψ : ClassFunction L ℂ⦄,
φ ∈ zSupportedSpan (L := L) S A → ψ ∈ zSupportedSpan (L := L) S A →
ClassFunction.inner (tau φ) (tau ψ) = ClassFunction.inner φ ψ
conjugate_closed : OddOrder.Peterfalvi.S03.ClosedUnderConjugate S
no_real_characters : OddOrder.Peterfalvi.S03.HasNoRealCharacters S
pairwise_orthogonal : OddOrder.Peterfalvi.S03.PairwiseOrthogonal S
difference_image :
∀ ⦃χ : ClassFunction L ℂ⦄, χ ∈ S → CharacterDifferenceImage (L := L) (G := G) tau χ
difference_images_orthogonal :
∀ ⦃φ χ : ClassFunction L ℂ⦄ (hφ : φ ∈ S) (hχ : χ ∈ S),
ClassFunction.inner φ χ = 0 → ClassFunction.inner φ χ.conj = 0 →
(difference_image hφ).Orthogonal (difference_image hχ)
/-- **The two-element carrier is the special case of the book's (5.2).**
Every clause of `GeneralHypothesis` is a field of `Hypothesis` except (5.2.d)/(5.2.e), which are
obtained by putting the signed pair `τ(χ − χ̄) = ε·(μ − ν)` into the orthonormal-family shape
`R(χ) = {ε·μ, −ε·ν}` (`CharacterDifferenceImage.toOrthonormalImage`) and transporting the
disjoint-pair orthogonality along it (`toOrthonormalImage_orthogonal`).
So results proved over `GeneralHypothesis` apply verbatim to every existing `Hypothesis` consumer,
while being available for families with reducible members — which the book's (5.2) allows and
`Hypothesis` silently forbids (issue 0157). -/
noncomputable def Hypothesis.toGeneralHypothesis {S : Set (ClassFunction L ℂ)} {A : Set L}
[Fintype L] [Fintype G] [Invertible (Nat.card L : ℂ)] [Invertible (Nat.card G : ℂ)]
(hyp : Hypothesis (L := L) (G := G) S A) :
GeneralHypothesis (L := L) (G := G) S A where
tau := hyp.tau
tau_isometry_diff := hyp.tau_isometry_diff
conjugate_closed := hyp.conjugate_closed
no_real_characters := hyp.no_real_characters
pairwise_orthogonal := hyp.pairwise_orthogonal
difference_image := fun _χ hχ => (hyp.difference_image hχ).toOrthonormalImage
difference_images_orthogonal := by
intro _φ _χ hφ hχ h1 h2
exact CharacterDifferenceImage.toOrthonormalImage_orthogonal _ _
(hyp.difference_images_orthogonal hφ hχ h1 h2)
namespace GeneralHypothesis
variable {S : Set (ClassFunction L ℂ)} {A : Set L}
variable [Fintype L] [Fintype G]
variable [Invertible (Nat.card L : ℂ)] [Invertible (Nat.card G : ℂ)]
/-- (5.2.a): `S` contains the conjugate of each member. -/
theorem conjugate_mem {hyp : GeneralHypothesis (L := L) (G := G) S A}
{χ : ClassFunction L ℂ} (hχ : χ ∈ S) :
χ.conj ∈ S :=
hyp.conjugate_closed hχ
/-- (5.2.a): no member is real. -/
theorem not_isReal {hyp : GeneralHypothesis (L := L) (G := G) S A}
{χ : ClassFunction L ℂ} (hχ : χ ∈ S) :
¬ χ.IsReal :=
hyp.no_real_characters hχ
/-- (5.2.a): a member differs from its conjugate. -/
theorem ne_conj {hyp : GeneralHypothesis (L := L) (G := G) S A}
{χ : ClassFunction L ℂ} (hχ : χ ∈ S) :
χ ≠ χ.conj := by
intro hχ_eq
exact hyp.not_isReal hχ hχ_eq.symm
/-- **Member-difference form of the (5.2.b) lattice isometry** — the `GeneralHypothesis` copy of
`Hypothesis.tau_isometry_memberDiff`. Both read off `tau_isometry_diff`, which is a shared
field; only (5.2.d)/(5.2.e) differ between the two carriers. -/
theorem tau_isometry_memberDiff (hyp : GeneralHypothesis (L := L) (G := G) S A)
⦃a b c d : ClassFunction L ℂ⦄ (ha : a ∈ S) (hb : b ∈ S) (hc : c ∈ S) (hd : d ∈ S)
(hab : ((a - b : ClassFunction L ℂ)).support ⊆ A)
(hcd : ((c - d : ClassFunction L ℂ)).support ⊆ A) :
ClassFunction.inner (hyp.tau (a - b)) (hyp.tau (c - d))
= ClassFunction.inner (a - b) (c - d) :=
hyp.tau_isometry_diff
⟨Submodule.sub_mem _ (Submodule.subset_span ha) (Submodule.subset_span hb), hab⟩
⟨Submodule.sub_mem _ (Submodule.subset_span hc) (Submodule.subset_span hd), hcd⟩
end GeneralHypothesis
@[simp] theorem Hypothesis.toGeneralHypothesis_tau {S : Set (ClassFunction L ℂ)} {A : Set L}
[Fintype L] [Fintype G] [Invertible (Nat.card L : ℂ)] [Invertible (Nat.card G : ℂ)]
(hyp : Hypothesis (L := L) (G := G) S A) :
hyp.toGeneralHypothesis.tau = hyp.tau := rfl
/-- **Peterfalvi (6.6): the common-index p-power degree gap (ℕ).**
This is the same strict gap as `two_mul_lt_sq_of_primePow_gap`, but in the exact
common-index form used in the §8 Sibley X-chain. If
`d₁ = idx * p^m₁` and `dᵢ = idx * p^mᵢ` with `idx > 0`, `p ≥ 3`, and
`d₁ < dᵢ`, then the p-power residual of `dᵢ` is at least one factor of `p`
larger than the residual of `d₁`; hence `dᵢ ≥ 3 * d₁` and
`2 * dᵢ * d₁ < dᵢ^2`.
This removes the need to name a separate quotient `q = dᵢ / d₁` in the
consumer-facing §8 constructors. -/
theorem two_mul_lt_sq_of_commonIndex_primePower_gap
{p idx θ₁ θᵢ d₁ dᵢ m₁ mᵢ : ℕ} (hp : 3 ≤ p) (hidx : 0 < idx)
(hd₁ : d₁ = idx * θ₁) (hdᵢ : dᵢ = idx * θᵢ)
(hθ₁ : θ₁ = p ^ m₁) (hθᵢ : θᵢ = p ^ mᵢ) (hlt : d₁ < dᵢ) :
2 * (dᵢ * d₁) < dᵢ * dᵢ := by
have hθlt : θ₁ < θᵢ := by
rw [hd₁, hdᵢ] at hlt
exact (Nat.mul_lt_mul_left hidx).mp hlt
have hm_lt : m₁ < mᵢ := by
rw [hθ₁, hθᵢ] at hθlt
exact (Nat.pow_lt_pow_iff_right (show 1 < p by omega)).mp hθlt
have hpθ : p * θ₁ ≤ θᵢ := by
rw [hθ₁, hθᵢ]
have hm_succ : m₁ + 1 ≤ mᵢ := Nat.succ_le_of_lt hm_lt
calc
p * p ^ m₁ = p ^ (m₁ + 1) := by
rw [pow_succ']
_ ≤ p ^ mᵢ := Nat.pow_le_pow_right (by omega) hm_succ
have hgap : 3 * d₁ ≤ dᵢ := by
rw [hd₁, hdᵢ]
have h3θ : 3 * θ₁ ≤ θᵢ :=
(Nat.mul_le_mul_right θ₁ hp).trans hpθ
calc
3 * (idx * θ₁) = idx * (3 * θ₁) := by ring
_ ≤ idx * θᵢ := Nat.mul_le_mul_left idx h3θ
have hd₁pos : 0 < d₁ := by
rw [hd₁, hθ₁]
exact Nat.mul_pos hidx (pow_pos (by omega) m₁)
nlinarith [hgap, hd₁pos]
/-- **Peterfalvi (6.6): chaining the gap to the partial degree sum (ℕ).**
The gap `2·dᵢ·d₁ < dᵢ²` together with `dᵢ² ∣ D` (the square-divisibility `χᵢ(1)² ∣ ∑_{j<i}χⱼ(1)²`)
and `0 < D` gives `2·dᵢ·d₁ < D`: positivity upgrades the divisibility to `dᵢ² ≤ D`. -/
theorem two_mul_lt_of_sq_dvd_of_gap
{d₁ dᵢ D : ℕ} (hgap : 2 * (dᵢ * d₁) < dᵢ * dᵢ)
(hdvd : dᵢ * dᵢ ∣ D) (hDpos : 0 < D) :
2 * (dᵢ * d₁) < D := by
have : dᵢ * dᵢ ≤ D := Nat.le_of_dvd hDpos hdvd
omega
/-- **Peterfalvi (6.6): additive complement of a divisor (ℕ).**
If `a` divides both the `tail` and the `total`, and `head + tail = total`, then `a ∣ head`.
This is the combination step of mmd L78+L80: with `head = ∑_{j<i}χⱼ(1)²`,
`tail = ∑_{j≥i}χⱼ(1)²`, `total = |L| - |L:Z|`, the divisibilities `θᵢ(1)² ∣ tail` and
`θᵢ(1)² ∣ total` give `θᵢ(1)² ∣ head`. Stated additively to avoid `ℕ` subtraction. -/
theorem dvd_of_add_eq_of_dvd_dvd
{a head tail total : ℕ} (hsum : head + tail = total)
(htail : a ∣ tail) (htot : a ∣ total) :
a ∣ head := by
have h : a ∣ total - tail := Nat.dvd_sub htot htail
rwa [show total - tail = head from by omega] at h
/-- **Peterfalvi (6.6): coprimality forcing of the full degree square (ℕ).**
The induced degree factors as `χᵢ(1) = idx·θ` (`idx = |L:K|`, `θ = θᵢ(1)`). Given that both
`θ²` and `idx²` divide `D` and that `idx` is coprime to `θ` (mmd L80: `(|L:K|, p) = 1` with
`θ` a power of `p`), the full square `χᵢ(1)² = idx²·θ²` divides `D`.
`Coprime idx θ ⟹ Coprime idx² θ²`, and coprime divisors multiply: `idx²·θ² ∣ D`. -/
theorem sq_dvd_of_factored_coprime
{idx θ chi D : ℕ} (hchi : chi = idx * θ)
(hθ : θ * θ ∣ D) (hidx : idx * idx ∣ D) (hcop : Nat.Coprime idx θ) :
chi * chi ∣ D := by
subst hchi
have hsq : Nat.Coprime (idx * idx) (θ * θ) := by
simpa [sq] using Nat.Coprime.pow 2 2 hcop
have h : (idx * idx) * (θ * θ) ∣ D := hsq.mul_dvd_of_dvd_of_dvd hidx hθ
calc idx * θ * (idx * θ) = (idx * idx) * (θ * θ) := by ring
_ ∣ D := h
/-- **Peterfalvi (6.6): full square-divisibility from the additive complement.**
This is the consumer-facing form of mmd L78-80 for the prefix sum `head = ∑_{j<i}χⱼ(1)²`.
If `θ²` divides the complementary tail and the total degree sum, it divides `head`; combined with
`idx² ∣ head`, `χᵢ(1) = idx·θ`, and `Coprime idx θ`, the full square `χᵢ(1)²` divides `head`.
It packages the two arithmetic steps above so the (6.6) coherence step can pass the square
divisibility `hdvd` to `two_mul_degree_lt_sum_ratCast` directly from the character-theoretic
sum identity and coprimality data. -/
theorem sq_dvd_of_factored_coprime_add_complement
{idx θ chi head tail total : ℕ} (hsum : head + tail = total)
(hθtail : θ * θ ∣ tail) (hθtotal : θ * θ ∣ total)
(hidx : idx * idx ∣ head) (hchi : chi = idx * θ)
(hcop : Nat.Coprime idx θ) :
chi * chi ∣ head :=
sq_dvd_of_factored_coprime hchi
(dvd_of_add_eq_of_dvd_dvd hsum hθtail hθtotal) hidx hcop
/-- **Peterfalvi (6.6): square-divisibility of the complementary tail sum.**
If the anchor p-power degree `θ` divides every tail p-power degree `θdeg j`, then `θ²` divides
the tail sum of the induced degrees `(idx j * θdeg j)²`. This isolates the `Finset.dvd_sum`
part of mmd L80: once the degree sort gives the divisibility of the p-power factors, every
summand carries a factor `θ²`. -/
theorem sq_dvd_sum_sq_mul_of_dvd
{ι : Type*} (s : Finset ι) {θ : ℕ} {idx θdeg : ι → ℕ}
(hθ : ∀ j ∈ s, θ ∣ θdeg j) :
θ * θ ∣ ∑ j ∈ s, (idx j * θdeg j) * (idx j * θdeg j) := by
apply Finset.dvd_sum
intro j hj
obtain ⟨a, ha⟩ := hθ j hj
refine ⟨(idx j * a) * (idx j * a), ?_⟩
rw [ha]
ring
/-- **Peterfalvi (6.6): divisibility of p-power degrees from monotonicity (ℕ).**
If two p-power degree factors use the same base `p ≥ 2`, then the weak degree order
`θ ≤ θ'` implies `θ ∣ θ'`. This is the arithmetic core behind the tail assertion
`θᵢ(1) ∣ θⱼ(1)` after the (6.6) degree sort. -/
theorem dvd_primePow_of_le
{p θ θ' m n : ℕ} (hp : 2 ≤ p)
(hθ : θ = p ^ m) (hθ' : θ' = p ^ n) (hle : θ ≤ θ') :
θ ∣ θ' := by
rw [hθ, hθ'] at hle ⊢
have hexp : m ≤ n :=
(Nat.pow_le_pow_iff_right (show 1 < p by omega)).mp hle
exact Nat.pow_dvd_pow p hexp
/-- **Peterfalvi (6.6): cancelling the common induction index in p-power degrees (ℕ).**
The (6.6) sort compares induced degrees `idx·θ`. Since `idx = |L:K|` is positive and fixed,
`idx·θ ≤ idx·θ'` reduces to `θ ≤ θ'`, and p-power comparison gives `θ ∣ θ'`. -/
theorem dvd_primePow_of_mul_le_mul
{p idx θ θ' m n : ℕ} (hp : 2 ≤ p) (hidx : 0 < idx)
(hθ : θ = p ^ m) (hθ' : θ' = p ^ n) (hle : idx * θ ≤ idx * θ') :
θ ∣ θ' :=
dvd_primePow_of_le hp hθ hθ' (Nat.le_of_mul_le_mul_left hle hidx)
/-- **Peterfalvi (6.6): tail square-divisibility from sorted p-power degrees (ℕ).**
A consumer-facing form of the tail step: if every tail degree factor `θdeg j` is a p-power
and the sorted induced degrees satisfy `idx·θ ≤ idx·θdeg j`, then `θ²` divides the tail sum
`∑ (idx·θdeg j)²`. -/
theorem sq_dvd_sum_sq_mul_const_of_primePow_mul_le
{ι : Type*} (s : Finset ι) {p idx θ : ℕ} {θdeg : ι → ℕ}
{m : ℕ} {n : ι → ℕ} (hp : 2 ≤ p) (hidx : 0 < idx)
(hθ : θ = p ^ m) (hθdeg : ∀ j ∈ s, θdeg j = p ^ n j)
(hle : ∀ j ∈ s, idx * θ ≤ idx * θdeg j) :
θ * θ ∣ ∑ j ∈ s, (idx * θdeg j) * (idx * θdeg j) := by
exact sq_dvd_sum_sq_mul_of_dvd s (θ := θ) (idx := fun _ ↦ idx) (θdeg := θdeg)
(fun j hj ↦ dvd_primePow_of_mul_le_mul hp hidx hθ (hθdeg j hj) (hle j hj))
/-- **Peterfalvi (6.6): p-power square divisibility from a square bound (ℕ).**
If `θ` and `q` are powers of the same base `p ≥ 2`, then `θ² ≤ q` forces
`θ² ∣ q`. In the (6.6) tail/total divisibility chain, this turns the Schur-center
bound `θᵢ(1)² ≤ |K:Z|` into the p-power divisibility input needed for the total sum. -/
theorem sq_dvd_primePow_of_sq_le
{p θ q m n : ℕ} (hp : 2 ≤ p)
(hθ : θ = p ^ m) (hq : q = p ^ n) (hle : θ * θ ≤ q) :
θ * θ ∣ q := by
rw [hθ, hq] at hle ⊢
have hpow : p ^ (m + m) ≤ p ^ n := by
simpa [pow_add] using hle
have hexp : m + m ≤ n :=
(Nat.pow_le_pow_iff_right (show 1 < p by omega)).mp hpow
simpa [pow_add] using Nat.pow_dvd_pow p hexp
/-- **Peterfalvi (6.6): p-power square divisibility through a multiplicative factor (ℕ).**
A factored version of `sq_dvd_primePow_of_sq_le`: once `θ²` divides the p-power factor `q`,
it divides any product `q * c`. -/
theorem sq_dvd_primePow_mul_of_sq_le
{p θ q c m n : ℕ} (hp : 2 ≤ p)
(hθ : θ = p ^ m) (hq : q = p ^ n) (hle : θ * θ ≤ q) :
θ * θ ∣ q * c :=
dvd_mul_of_dvd_left (sq_dvd_primePow_of_sq_le hp hθ hq hle) c
/-- **Peterfalvi (6.6): head square-divisibility from tail and total p-power data (ℕ).**
This packages the full mmd L78-80 arithmetic chain. The complementary tail divisibility is derived
from the sorted p-power degrees, the total divisibility from the p-power square bound, and the
additive identity then moves the `θ²` divisor to the head. Finally the coprime fixed-index factor
upgrades `θ² ∣ head` and `idx² ∣ head` to `(idx·θ)² ∣ head`. -/
theorem sq_dvd_head_of_commonIndex_primePower_sums
{ι : Type*} (tailSet : Finset ι)
{p idx θ chi q c head total : ℕ} {θdeg : ι → ℕ}
{m n : ℕ} {ntail : ι → ℕ} (hp : 2 ≤ p) (hidxpos : 0 < idx)
(hθ : θ = p ^ m) (hθdeg : ∀ j ∈ tailSet, θdeg j = p ^ ntail j)
(htail_le : ∀ j ∈ tailSet, idx * θ ≤ idx * θdeg j)
(hsum : head + (∑ j ∈ tailSet, (idx * θdeg j) * (idx * θdeg j)) = total)
(hq : q = p ^ n) (hθsq_le_q : θ * θ ≤ q) (htotal : total = q * c)
(hidx_head : idx * idx ∣ head) (hchi : chi = idx * θ) (hcop : Nat.Coprime idx θ) :
chi * chi ∣ head := by
have hθtail : θ * θ ∣ ∑ j ∈ tailSet, (idx * θdeg j) * (idx * θdeg j) :=
sq_dvd_sum_sq_mul_const_of_primePow_mul_le tailSet hp hidxpos hθ hθdeg htail_le
have hθtotal : θ * θ ∣ total := by
rw [htotal]
exact sq_dvd_primePow_mul_of_sq_le hp hθ hq hθsq_le_q
exact sq_dvd_of_factored_coprime_add_complement
hsum hθtail hθtotal hidx_head hchi hcop
end OddOrder.Peterfalvi.S07
Source
Yawara Ishida, Lean 4 formalization of the Feit–Thompson (odd order) theorem, github.com/yawara/odd-order @ 20a6fccbb63c (Apache-2.0), https://github.com/yawara/odd-order/blob/20a6fccbb63c0b138acda9f7fafd798c42f9441b/OddOrder/Peterfalvi/S07_Coherence/NormInequalities.lean