Bender–Glauberman §3: actions of Frobenius groups and related results
DefinitionOddOrder_BG_Ch1_Preliminary_S03_FrobeniusActionsfeit-thompsonfinite-groupsgroup-theoryishida-odd-order
Starts the Bender–Glauberman Chapter I §3 layer (pp. 17–32), on Lemmas 3.1–3.10, over the existing Isaacs Ch. 6 Frobenius API. The sections are: §3A Frobenius group basics (Lemmas 3.1–3.2), §3B representation-theoretic Frobenius action (Lemma 3.3, Thms 3.4–3.5), §3C -group centralizer and -length (Thm 3.6), §3D Frobenius kernel nilpotence and Fitting control (Thms 3.7–3.8), §3E regular actions and nilpotent targets (Prop. 3.9, Thm 3.10). The first implemented results are BG Lemma 3.1 (the Frobenius structure on is equivalent to for every nonidentity ) and quotient setup lemmas for the branch of Lemma 3.2.
The bundle consists of lemmas only.
The bundle also proves 13 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/BG/Ch1_Preliminary/S03_FrobeniusActions.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_BG_Ch1_Preliminary_S01_Solvable
import Definitions.Def_OddOrder_Isaacs_Ch06_FrobeniusActions_Main
namespace OddOrder.BG.Ch1.S03
open OddOrder.Isaacs.Ch06
/-- **BG Lemma 3.1 (centralizer direction)**: in a Frobenius group `G = K R`,
the `K`-centralizer of every nonidentity `x ∈ R` is trivial.
The BG notation `C_K(x)=1` is represented as
`Subgroup.centralizer ({x} : Set G) ⊓ K = ⊥`. -/
theorem centralizer_complement_inf_kernel_eq_bot
{G : Type*} [Group G] {K R : Subgroup G}
(h : IsFrobeniusGroup G K R) :
∀ x ∈ R, x ≠ 1 → Subgroup.centralizer ({x} : Set G) ⊓ K = ⊥ := by
intro x hxR hx_ne
rw [eq_bot_iff]
intro y hy
rw [Subgroup.mem_inf] at hy
obtain ⟨hy_centralizes, hyK⟩ := hy
rw [Subgroup.mem_bot]
by_contra hy_ne
rw [Subgroup.mem_centralizer_singleton_iff] at hy_centralizes
have h_conj : x * y * x⁻¹ = y := by
calc
x * y * x⁻¹ = (y * x) * x⁻¹ := by rw [← hy_centralizes]
_ = y := mul_inv_cancel_right y x
exact h.conj_frobenius x hxR hx_ne y hyK hy_ne h_conj
/-- **BG Lemma 3.1**: for nonidentity subgroups `K` and `R` with `K ⊴ G`,
`K` complemented by `R`, the Frobenius group structure on `G = K R` is
equivalent to `C_K(x)=1` for every nonidentity `x ∈ R`.
This is the local BG form of Isaacs Thm 6.4, specialized to a named kernel
`K` and complement `R`. -/
theorem isFrobeniusGroup_iff_complement_centralizer_inf_kernel_eq_bot
{G : Type*} [Group G] {K R : Subgroup G}
(hK : K.Normal) (hC : Subgroup.IsComplement' K R)
(hK_ne : K ≠ ⊥) (hR_ne : R ≠ ⊥) :
IsFrobeniusGroup G K R ↔
∀ x ∈ R, x ≠ 1 → Subgroup.centralizer ({x} : Set G) ⊓ K = ⊥ := by
refine ⟨centralizer_complement_inf_kernel_eq_bot, ?_⟩
intro hcentral
exact
{ isNormal := hK
isComplement := hC
ne_bot_kernel := hK_ne
ne_bot_complement := hR_ne
conj_frobenius := by
intro x hxR hx_ne y hyK hy_ne h_conj
have h_comm : y * x = x * y := by
have h_mul := congrArg (fun z => z * x) h_conj
simpa only [mul_assoc, inv_mul_cancel, mul_one] using h_mul.symm
have hy_centralizes : y ∈ Subgroup.centralizer ({x} : Set G) :=
Subgroup.mem_centralizer_singleton_iff.mpr h_comm
have hy_inf : y ∈ Subgroup.centralizer ({x} : Set G) ⊓ K :=
Subgroup.mem_inf.mpr ⟨hy_centralizes, hyK⟩
rw [hcentral x hxR hx_ne, Subgroup.mem_bot] at hy_inf
exact hy_ne hy_inf }
/-- **BG Lemma 3.2 (quotient setup, `N ≤ K`)**: if `K` and `R` complement each
other in `G`, and `N ⊴ G` lies in `K`, then their images in `G/N` still
complement each other.
This is the elementary complement part of the `N ≤ K` branch of BG Lemma 3.2.
The fixed-point-free centralizer condition in the quotient is supplied by
BG Prop. 1.5(d) / Isaacs Cor. 3.28 in the next step. -/
theorem quotient_complement_of_normal_le_kernel
{G : Type*} [Group G] {K R N : Subgroup G} [K.Normal] [N.Normal]
(hC : Subgroup.IsComplement' K R) (hNK : N ≤ K) :
Subgroup.IsComplement'
(K.map (QuotientGroup.mk' N)) (R.map (QuotientGroup.mk' N)) := by
apply Subgroup.isComplement'_of_disjoint_and_mul_eq_univ
· rw [disjoint_iff]
ext q
constructor
· intro hq
rw [Subgroup.mem_inf] at hq
obtain ⟨hqK, hqR⟩ := hq
obtain ⟨k, hkK, hkq⟩ := hqK
obtain ⟨r, hrR, hrq⟩ := hqR
rw [Subgroup.mem_bot]
have h_eq : (QuotientGroup.mk' N) k = (QuotientGroup.mk' N) r := by
rw [hkq, hrq]
have h_div : k / r ∈ N :=
(QuotientGroup.eq_iff_div_mem (N := N)).mp h_eq
have h_divK : k / r ∈ K := hNK h_div
have hrK : r ∈ K := by
have hrinvK : r⁻¹ ∈ K := by
have hmem : k⁻¹ * (k / r) ∈ K :=
K.mul_mem (K.inv_mem hkK) h_divK
have h_eq_inv : k⁻¹ * (k / r) = r⁻¹ := by
rw [div_eq_mul_inv, ← mul_assoc, inv_mul_cancel, one_mul]
rwa [h_eq_inv] at hmem
simpa using K.inv_mem hrinvK
have hr_one : r = 1 := Subgroup.disjoint_def.mp hC.disjoint hrK hrR
rw [← hrq, hr_one]
simp
· intro hq
rw [Subgroup.mem_bot] at hq
rw [hq]
exact Subgroup.one_mem _
· rw [Set.eq_univ_iff_forall]
intro q
obtain ⟨g, rfl⟩ := QuotientGroup.mk_surjective q
have hg_top : g ∈ (K ⊔ R : Subgroup G) := by
rw [hC.sup_eq_top]
trivial
rw [Subgroup.mem_sup_of_normal_left] at hg_top
obtain ⟨k, hkK, r, hrR, hkr⟩ := hg_top
refine ⟨QuotientGroup.mk k, ?_, QuotientGroup.mk r, ?_, ?_⟩
· exact Subgroup.mem_map.mpr ⟨k, hkK, rfl⟩
· exact Subgroup.mem_map.mpr ⟨r, hrR, rfl⟩
· change (QuotientGroup.mk k : G ⧸ N) * QuotientGroup.mk r = QuotientGroup.mk g
rw [← QuotientGroup.mk_mul, hkr]
/-- If `K` is not contained in `N`, then the image of `K` in `G/N` is
nontrivial. -/
theorem quotient_kernel_map_ne_bot_of_not_le
{G : Type*} [Group G] {K N : Subgroup G} [N.Normal]
(hKN : ¬ K ≤ N) :
K.map (QuotientGroup.mk' N) ≠ ⊥ := by
intro hbot
apply hKN
intro k hkK
have hkmap : (QuotientGroup.mk' N) k ∈ K.map (QuotientGroup.mk' N) :=
Subgroup.mem_map.mpr ⟨k, hkK, rfl⟩
rw [hbot, Subgroup.mem_bot] at hkmap
exact (QuotientGroup.eq_one_iff k).mp hkmap
/-- If `R` is nontrivial and complements `K`, then the image of `R` in `G/N`
is still nontrivial whenever `N ≤ K`. -/
theorem quotient_complement_map_ne_bot_of_le_kernel
{G : Type*} [Group G] {K R N : Subgroup G} [N.Normal]
(hC : Subgroup.IsComplement' K R) (hNK : N ≤ K) (hR_ne : R ≠ ⊥) :
R.map (QuotientGroup.mk' N) ≠ ⊥ := by
intro hbot
apply hR_ne
rw [← le_bot_iff]
intro r hrR
have hrmap : (QuotientGroup.mk' N) r ∈ R.map (QuotientGroup.mk' N) :=
Subgroup.mem_map.mpr ⟨r, hrR, rfl⟩
rw [hbot, Subgroup.mem_bot] at hrmap
have hrN : r ∈ N := (QuotientGroup.eq_one_iff r).mp hrmap
have hrK : r ∈ K := hNK hrN
exact Subgroup.disjoint_def.mp hC.disjoint hrK hrR
/-- **BG Lemma 3.2 / Prop. 1.5(d) bridge**: if the cyclic group generated by
`x` fixes the coset `yN` in `K/N`, then coprime action lifts this to an
actual `x`-fixed representative in the same coset.
This is a direct wrapper around Isaacs Cor. 3.28
`OddOrder.Isaacs.Ch04.coprime_fixedPoints_quotient`, specialized to the
conjugation action of `Subgroup.zpowers x` on the normal subgroup `K`. -/
theorem fixedPoint_lift_of_zpowers_quotient_fixed
{G : Type*} [Group G] [Finite G] {K N : Subgroup G} [K.Normal] [N.Normal]
(hNK : N ≤ K) {x y : G}
(hCop : Nat.Coprime (Nat.card ↥(Subgroup.zpowers x)) (Nat.card ↥K))
(hSolv : Group.IsSolvable ↥(Subgroup.zpowers x) ∨ Group.IsSolvable ↥K)
(hyK : y ∈ K)
(hfixed_all :
∀ a ∈ Subgroup.zpowers x,
(QuotientGroup.mk' N) (a * y * a⁻¹) = (QuotientGroup.mk' N) y) :
∃ c : G,
c ∈ K ∧ (QuotientGroup.mk' N) c = (QuotientGroup.mk' N) y ∧
x * c * x⁻¹ = c := by
let A : Subgroup G := Subgroup.zpowers x
let φ : A →* MulAut K := (MulAut.conjNormal : G →* MulAut K).comp A.subtype
let N_K : Subgroup K := N.subgroupOf K
have hN_K_normal : N_K.Normal := (inferInstance : N.Normal).subgroupOf K
have hN_inv : OddOrder.Isaacs.Ch03.IsAInvariant φ N_K := by
rw [OddOrder.Isaacs.Ch03.isAInvariant_iff_smul_mem]
intro a n hn
change ((MulAut.conjNormal (a : G)) n).val ∈ N
rw [MulAut.conjNormal_apply]
exact (inferInstance : N.Normal).conj_mem n.val hn (a : G)
let yK : K := ⟨y, hyK⟩
have hy_fix :
∀ a : A, ∃ n ∈ N_K, (φ a) yK = yK * n := by
intro a
have hq : (QuotientGroup.mk' N) ((a : G) * y * (a : G)⁻¹) =
(QuotientGroup.mk' N) y :=
hfixed_all (a : G) a.property
obtain ⟨z, hzN, hz_eq⟩ := (QuotientGroup.mk'_eq_mk' (N := N)).mp hq
have hz_inv_N : z⁻¹ ∈ N := N.inv_mem hzN
refine ⟨⟨z⁻¹, hNK hz_inv_N⟩, ?_, ?_⟩
· exact hz_inv_N
· apply Subtype.ext
change (a : G) * y * (a : G)⁻¹ = y * z⁻¹
exact eq_mul_inv_of_mul_eq hz_eq
obtain ⟨cK, hc_fixed, ⟨nK, hnK, hc_coset⟩⟩ :=
OddOrder.Isaacs.Ch04.coprime_fixedPoints_quotient
(G := K) (A := A) (φ := φ) hCop hSolv hN_inv (g := yK) hy_fix
refine ⟨cK, cK.property, ?_, ?_⟩
· have hc_val : (cK : G) = y * (nK : G) := congrArg Subtype.val hc_coset
rw [hc_val]
change (y : G ⧸ N) * ((nK : G) : G ⧸ N) = (y : G ⧸ N)
have hn_one : ((nK : G) : G ⧸ N) = 1 :=
(QuotientGroup.eq_one_iff (nK : G)).mpr hnK
rw [hn_one, mul_one]
· have hx_mem : x ∈ A := Subgroup.mem_zpowers x
have hx_fixed := hc_fixed ⟨x, hx_mem⟩
have hx_fixed_val := congrArg Subtype.val hx_fixed
simpa [φ, A, MulAut.conjNormal_apply] using hx_fixed_val
/-- If the generator `x` fixes the quotient coset `yN` by conjugation, then
every element of the cyclic subgroup generated by `x` fixes the same quotient
coset. -/
theorem zpowers_quotient_fixed_of_generator_quotient_fixed
{G : Type*} [Group G] {N : Subgroup G} [N.Normal] {x y : G}
(hgen : (QuotientGroup.mk' N) (x * y * x⁻¹) = (QuotientGroup.mk' N) y) :
∀ a ∈ Subgroup.zpowers x,
(QuotientGroup.mk' N) (a * y * a⁻¹) = (QuotientGroup.mk' N) y := by
intro a ha
obtain ⟨n, ha_eq⟩ := Subgroup.mem_zpowers_iff.mp ha
rw [← ha_eq]
have hgenQ :
(x : G ⧸ N) * (y : G ⧸ N) * (x : G ⧸ N)⁻¹ = (y : G ⧸ N) := by
exact hgen
have hxy_comm : Commute (x : G ⧸ N) (y : G ⧸ N) := by
have hmul := congrArg (fun z => z * (x : G ⧸ N)) hgenQ
simpa only [commute_iff_eq, mul_assoc, inv_mul_cancel, mul_one] using hmul
have hxpow_comm : Commute ((x : G ⧸ N) ^ n) (y : G ⧸ N) :=
hxy_comm.zpow_left n
have hconj :
((x : G ⧸ N) ^ n) * (y : G ⧸ N) * ((x : G ⧸ N) ^ n)⁻¹ =
(y : G ⧸ N) := by
have hmul := congrArg (fun z => z * (((x : G ⧸ N) ^ n)⁻¹)) hxpow_comm.eq
simpa only [mul_assoc, mul_inv_cancel, mul_one] using hmul
have hmap : ((x ^ n : G) : G ⧸ N) = (x : G ⧸ N) ^ n := by
simp
change ((x ^ n : G) : G ⧸ N) * (y : G ⧸ N) *
((x ^ n : G) : G ⧸ N)⁻¹ = (y : G ⧸ N)
rw [hmap]
exact hconj
/-- **BG Lemma 3.2 / Prop. 1.5(d) bridge, generator form**: if `x` fixes
the quotient coset `yN`, then coprime action gives an actual `x`-fixed
representative in the same coset. -/
theorem fixedPoint_lift_of_generator_quotient_fixed
{G : Type*} [Group G] [Finite G] {K N : Subgroup G} [K.Normal] [N.Normal]
(hNK : N ≤ K) {x y : G}
(hCop : Nat.Coprime (Nat.card ↥(Subgroup.zpowers x)) (Nat.card ↥K))
(hSolv : Group.IsSolvable ↥(Subgroup.zpowers x) ∨ Group.IsSolvable ↥K)
(hyK : y ∈ K)
(hgen : (QuotientGroup.mk' N) (x * y * x⁻¹) = (QuotientGroup.mk' N) y) :
∃ c : G,
c ∈ K ∧ (QuotientGroup.mk' N) c = (QuotientGroup.mk' N) y ∧
x * c * x⁻¹ = c :=
fixedPoint_lift_of_zpowers_quotient_fixed hNK hCop hSolv hyK
(zpowers_quotient_fixed_of_generator_quotient_fixed hgen)
/-- **BG Lemma 3.2 (quotient centralizer bridge)**: to prove
`C_{K/N}(\bar x)=1`, it is enough to know that every quotient-fixed coset in
`K/N` contains an actual `x`-fixed element of `K`.
The `hlift` hypothesis is the exact point where BG Prop. 1.5(d) / Isaacs Cor.
3.28 enters: quotient fixed points are images of fixed points upstairs. -/
theorem quotient_centralizer_inf_kernel_eq_bot_of_fixedPoint_lift
{G : Type*} [Group G] {K R N : Subgroup G} [N.Normal]
(hcentral :
∀ x ∈ R, x ≠ 1 → Subgroup.centralizer ({x} : Set G) ⊓ K = ⊥)
(hlift :
∀ x ∈ R, x ≠ 1 → ∀ y ∈ K,
(QuotientGroup.mk' N) (x * y * x⁻¹) = (QuotientGroup.mk' N) y →
∃ c : G,
c ∈ K ∧ (QuotientGroup.mk' N) c = (QuotientGroup.mk' N) y ∧
x * c * x⁻¹ = c) :
∀ qx ∈ R.map (QuotientGroup.mk' N), qx ≠ 1 →
Subgroup.centralizer ({qx} : Set (G ⧸ N)) ⊓ K.map (QuotientGroup.mk' N) = ⊥ := by
intro qx hqxR hqx_ne
rw [eq_bot_iff]
intro qy hqy
rw [Subgroup.mem_inf] at hqy
obtain ⟨hqy_centralizes, hqyK⟩ := hqy
obtain ⟨x, hxR, hxq⟩ := hqxR
obtain ⟨y, hyK, hyq⟩ := hqyK
rw [Subgroup.mem_bot]
have hx_ne : x ≠ 1 := by
intro hx_one
apply hqx_ne
rw [← hxq, hx_one]
simp
rw [Subgroup.mem_centralizer_singleton_iff] at hqy_centralizes
have hq_conj : qx * qy * qx⁻¹ = qy := by
calc
qx * qy * qx⁻¹ = (qy * qx) * qx⁻¹ := by rw [← hqy_centralizes]
_ = qy := mul_inv_cancel_right qy qx
have h_conj_mod :
(QuotientGroup.mk' N) (x * y * x⁻¹) = (QuotientGroup.mk' N) y := by
have hxq_coe : (x : G ⧸ N) = qx := hxq
have hyq_coe : (y : G ⧸ N) = qy := hyq
change (x : G ⧸ N) * (y : G ⧸ N) * (x : G ⧸ N)⁻¹ = (y : G ⧸ N)
rw [hxq_coe, hyq_coe]
exact hq_conj
obtain ⟨c, hcK, hcy, hc_fixed⟩ := hlift x hxR hx_ne y hyK h_conj_mod
have hc_centralizes : c ∈ Subgroup.centralizer ({x} : Set G) := by
have h_comm : c * x = x * c := by
have h_mul := congrArg (fun z => z * x) hc_fixed
simpa only [mul_assoc, inv_mul_cancel, mul_one] using h_mul.symm
exact Subgroup.mem_centralizer_singleton_iff.mpr h_comm
have hc_inf : c ∈ Subgroup.centralizer ({x} : Set G) ⊓ K :=
Subgroup.mem_inf.mpr ⟨hc_centralizes, hcK⟩
rw [hcentral x hxR hx_ne, Subgroup.mem_bot] at hc_inf
rw [← hyq, ← hcy, hc_inf]
simp
/-- **BG Lemma 3.2 (quotient centralizer bridge, `C_K(x) ≤ N` form)**: a relaxation of
`quotient_centralizer_inf_kernel_eq_bot_of_fixedPoint_lift` in which the upstairs centralizers
`C_G(x) ⊓ K` need not be trivial, only land *inside the normal subgroup `N` being quotiented out*.
This still forces `C_{K/N}(\bar x) = 1`, because the lifted `x`-fixed representative `c` lies in
`C_G(x) ⊓ K ≤ N`, hence maps to `1` in `G/N`.
This is exactly the Peterfalvi (6.8)(c2) situation: there `C_K(x) = W₂ ⊆ ⁅K,K⁆ = N` (nontrivial),
so the `= ⊥` hypothesis of the original lemma fails, but the `≤ N` hypothesis holds and the
quotient `K/⁅K,K⁆` still has fixed-point-free `\bar x`-action. -/
theorem quotient_centralizer_inf_kernel_eq_bot_of_fixedPoint_lift_of_le
{G : Type*} [Group G] {K R N : Subgroup G} [N.Normal]
(hcentral :
∀ x ∈ R, x ≠ 1 → Subgroup.centralizer ({x} : Set G) ⊓ K ≤ N)
(hlift :
∀ x ∈ R, x ≠ 1 → ∀ y ∈ K,
(QuotientGroup.mk' N) (x * y * x⁻¹) = (QuotientGroup.mk' N) y →
∃ c : G,
c ∈ K ∧ (QuotientGroup.mk' N) c = (QuotientGroup.mk' N) y ∧
x * c * x⁻¹ = c) :
∀ qx ∈ R.map (QuotientGroup.mk' N), qx ≠ 1 →
Subgroup.centralizer ({qx} : Set (G ⧸ N)) ⊓ K.map (QuotientGroup.mk' N) = ⊥ := by
intro qx hqxR hqx_ne
rw [eq_bot_iff]
intro qy hqy
rw [Subgroup.mem_inf] at hqy
obtain ⟨hqy_centralizes, hqyK⟩ := hqy
obtain ⟨x, hxR, hxq⟩ := hqxR
obtain ⟨y, hyK, hyq⟩ := hqyK
rw [Subgroup.mem_bot]
have hx_ne : x ≠ 1 := by
intro hx_one
apply hqx_ne
rw [← hxq, hx_one]
simp
rw [Subgroup.mem_centralizer_singleton_iff] at hqy_centralizes
have hq_conj : qx * qy * qx⁻¹ = qy := by
calc
qx * qy * qx⁻¹ = (qy * qx) * qx⁻¹ := by rw [← hqy_centralizes]
_ = qy := mul_inv_cancel_right qy qx
have h_conj_mod :
(QuotientGroup.mk' N) (x * y * x⁻¹) = (QuotientGroup.mk' N) y := by
have hxq_coe : (x : G ⧸ N) = qx := hxq
have hyq_coe : (y : G ⧸ N) = qy := hyq
change (x : G ⧸ N) * (y : G ⧸ N) * (x : G ⧸ N)⁻¹ = (y : G ⧸ N)
rw [hxq_coe, hyq_coe]
exact hq_conj
obtain ⟨c, hcK, hcy, hc_fixed⟩ := hlift x hxR hx_ne y hyK h_conj_mod
have hc_centralizes : c ∈ Subgroup.centralizer ({x} : Set G) := by
have h_comm : c * x = x * c := by
have h_mul := congrArg (fun z => z * x) hc_fixed
simpa only [mul_assoc, inv_mul_cancel, mul_one] using h_mul.symm
exact Subgroup.mem_centralizer_singleton_iff.mpr h_comm
have hc_inf : c ∈ Subgroup.centralizer ({x} : Set G) ⊓ K :=
Subgroup.mem_inf.mpr ⟨hc_centralizes, hcK⟩
have hc_N : c ∈ N := hcentral x hxR hx_ne hc_inf
have hc_mk : (QuotientGroup.mk' N) c = 1 := (QuotientGroup.eq_one_iff c).mpr hc_N
rw [← hyq, ← hcy, hc_mk]
/-- **BG Lemma 3.2 (quotient Frobenius assembly, `N ≤ K`)**: in the easy branch
where `N ⊴ G` lies in the Frobenius kernel `K`, the quotient is a Frobenius
group as soon as the quotient centralizer condition is known.
The missing input is exactly the BG Prop. 1.5(d) / Isaacs Cor. 3.28 step:
`C_{K/N}(\bar x)=1` for every nonidentity `\bar x ∈ RN/N`. -/
theorem quotient_isFrobeniusGroup_of_le_kernel_of_centralizer
{G : Type*} [Group G] {K R N : Subgroup G} [N.Normal]
(h : IsFrobeniusGroup G K R) (hNK : N ≤ K) (hKN : ¬ K ≤ N)
(hcentral :
∀ x ∈ R.map (QuotientGroup.mk' N), x ≠ 1 →
Subgroup.centralizer ({x} : Set (G ⧸ N)) ⊓
K.map (QuotientGroup.mk' N) = ⊥) :
IsFrobeniusGroup
(G ⧸ N) (K.map (QuotientGroup.mk' N)) (R.map (QuotientGroup.mk' N)) := by
let : K.Normal := h.isNormal
have hK_normal : (K.map (QuotientGroup.mk' N)).Normal := inferInstance
have hC :
Subgroup.IsComplement'
(K.map (QuotientGroup.mk' N)) (R.map (QuotientGroup.mk' N)) :=
quotient_complement_of_normal_le_kernel h.isComplement hNK
have hK_ne : K.map (QuotientGroup.mk' N) ≠ ⊥ :=
quotient_kernel_map_ne_bot_of_not_le hKN
have hR_ne : R.map (QuotientGroup.mk' N) ≠ ⊥ :=
quotient_complement_map_ne_bot_of_le_kernel h.isComplement hNK h.ne_bot_complement
exact
(isFrobeniusGroup_iff_complement_centralizer_inf_kernel_eq_bot
hK_normal hC hK_ne hR_ne).mpr hcentral
/-- **BG Lemma 3.2 (quotient Frobenius, `N ≤ K`, fixed-point-lift form)**:
the `N ≤ K` branch reduces to the fixed-point lifting input supplied by
BG Prop. 1.5(d) / Isaacs Cor. 3.28.
This is the strongest currently packaged form of the first branch: complement
and nontriviality are proved here, while the remaining mathematical input is
isolated in `hlift`. -/
theorem quotient_isFrobeniusGroup_of_le_kernel_of_fixedPoint_lift
{G : Type*} [Group G] {K R N : Subgroup G} [N.Normal]
(h : IsFrobeniusGroup G K R) (hNK : N ≤ K) (hKN : ¬ K ≤ N)
(hlift :
∀ x ∈ R, x ≠ 1 → ∀ y ∈ K,
(QuotientGroup.mk' N) (x * y * x⁻¹) = (QuotientGroup.mk' N) y →
∃ c : G,
c ∈ K ∧ (QuotientGroup.mk' N) c = (QuotientGroup.mk' N) y ∧
x * c * x⁻¹ = c) :
IsFrobeniusGroup
(G ⧸ N) (K.map (QuotientGroup.mk' N)) (R.map (QuotientGroup.mk' N)) := by
refine quotient_isFrobeniusGroup_of_le_kernel_of_centralizer h hNK hKN ?_
exact
quotient_centralizer_inf_kernel_eq_bot_of_fixedPoint_lift
(centralizer_complement_inf_kernel_eq_bot h) hlift
/-- **BG Lemma 3.2 (quotient Frobenius, `N ≤ K`, coprime-action form)**:
the first branch follows once each nonidentity complement element has cyclic
order coprime to `K` and the kernel side is solvable.
This packages the fixed-point lifting input via Isaacs Cor. 3.28; deriving the
coprimality hypothesis from the Frobenius setup is left as a separate lemma. -/
theorem quotient_isFrobeniusGroup_of_le_kernel_of_coprime_zpowers
{G : Type*} [Group G] [Finite G] {K R N : Subgroup G} [N.Normal]
(h : IsFrobeniusGroup G K R) (hNK : N ≤ K) (hKN : ¬ K ≤ N)
(hCop : ∀ x, x ∈ R → x ≠ 1 →
Nat.Coprime (Nat.card ↥(Subgroup.zpowers x)) (Nat.card ↥K))
(hSolvK : Group.IsSolvable ↥K) :
IsFrobeniusGroup
(G ⧸ N) (K.map (QuotientGroup.mk' N)) (R.map (QuotientGroup.mk' N)) := by
let : K.Normal := h.isNormal
refine quotient_isFrobeniusGroup_of_le_kernel_of_fixedPoint_lift h hNK hKN ?_
intro x hxR hx_ne y hyK hgen
exact
fixedPoint_lift_of_generator_quotient_fixed hNK (hCop x hxR hx_ne)
(Or.inr hSolvK) hyK hgen
end OddOrder.BG.Ch1.S03
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/BG/Ch1_Preliminary/S03_FrobeniusActions.lean