Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

SU(d)SU(d)SU(d), its Haar measure, the Hecke operator and the spectral gap property

Definition
SpectralGapConjecture_HeckeSU

by Lucas · Oct 13, 2026 · Mathlib 0df444a (Lean v4.33.1)

group-theoryhaar-measurespectral-gap

This file fixes the objects in which the spectral gap conjecture is stated.

  1. The group. SU(d)SU(d)SU(d) is the group of complex d×dd\times dd×d matrices UUU with U∗U=UU∗=IU^*U = UU^* = IU∗U=UU∗=I and det⁡U=1\det U = 1detU=1, with the topology inherited from Cd×d\mathbb C^{d\times d}Cd×d and its Borel σ\sigmaσ-algebra. It is shown to be a compact topological group.
  2. Haar measure. μd\mu_dμd​ (haarSU d) is the Haar measure on SU(d)SU(d)SU(d) normalized so that μd(SU(d))=1\mu_d(SU(d)) = 1μd​(SU(d))=1; it is a left-invariant Borel probability measure.
  3. Hecke operator. For a group GGG and g=(g1,…,gn)∈Gng = (g_1,\dots,g_n) \in G^ng=(g1​,…,gn​)∈Gn, the operator zgz_gzg​ acts on f:G→Cf : G \to \mathbb Cf:G→C by
(zgf)(x)=∑j=1n(f(gjx)+f(gj−1x)).(z_g f)(x) = \sum_{j=1}^{n}\bigl(f(g_j x) + f(g_j^{-1}x)\bigr).(zg​f)(x)=j=1∑n​(f(gj​x)+f(gj−1​x)).
  1. Spectral gap. For a measure μ\muμ on GGG, the tuple ggg has a spectral gap if there is ε>0\varepsilon > 0ε>0 such that for every square-integrable f:G→Cf : G \to \mathbb Cf:G→C with ∫Gf dμ=0\int_G f\,d\mu = 0∫G​fdμ=0,
Re⁡∫Gf(x)‾ (zgf)(x) dμ(x)  ≤  (2n−ε)∫G∣f(x)∣2 dμ(x).\operatorname{Re}\int_G \overline{f(x)}\,(z_g f)(x)\,d\mu(x) \;\le\; (2n - \varepsilon)\int_G |f(x)|^2\,d\mu(x).Re∫G​f(x)​(zg​f)(x)dμ(x)≤(2n−ε)∫G​∣f(x)∣2dμ(x).

When μ\muμ is a Haar probability measure, zgz_gzg​ is a bounded self-adjoint operator on L2(G,μ)L^2(G,\mu)L2(G,μ) that preserves the mean-zero subspace L02L^2_0L02​, and condition 4 says exactly that the supremum λ1\lambda_1λ1​ of its spectrum on L02L^2_0L02​ is strictly less than 2n2n2n, the eigenvalue of the constant functions.

Formalization Note. Mean zero is the same as orthogonality to constants because μ\muμ is finite. The spectral gap is one-sided (an upper bound on λ1\lambda_1λ1​), following Fisher (2006) and Bourgain–Gamburd (2012).

Definition code
import Mathlib

open MeasureTheory

namespace SpectralGapConjecture

/-- The special unitary group `SU(d)`: complex `d × d` matrices `U` with `U⋆ U = U U⋆ = 1`
and `det U = 1`, with the subspace topology inherited from `Matrix (Fin d) (Fin d) ℂ`. -/
abbrev SU (d : ℕ) := Matrix.specialUnitaryGroup (Fin d) ℂ

/-- `SU(d)` carries its Borel σ-algebra. -/
noncomputable instance (d : ℕ) : MeasurableSpace (SU d) := borel _

instance (d : ℕ) : BorelSpace (SU d) := ⟨rfl⟩

instance (d : ℕ) : IsTopologicalGroup (SU d) where
  continuous_mul := by
    apply Continuous.subtype_mk
    exact (continuous_subtype_val.comp continuous_fst).matrix_mul
      (continuous_subtype_val.comp continuous_snd)
  continuous_inv := by
    apply Continuous.subtype_mk
    exact continuous_subtype_val.matrix_conjTranspose

theorem isClosed_SU (d : ℕ) : IsClosed (SU d : Set (Matrix (Fin d) (Fin d) ℂ)) := by
  have h1 : IsClosed {A : Matrix (Fin d) (Fin d) ℂ |
      (star A * A = 1 ∧ A * star A = 1) ∧ A.det = 1} := by
    refine ((isClosed_eq ?_ continuous_const).inter (isClosed_eq ?_ continuous_const)).inter
      (isClosed_eq ?_ continuous_const)
    · exact continuous_id.matrix_conjTranspose.matrix_mul continuous_id
    · exact continuous_id.matrix_mul continuous_id.matrix_conjTranspose
    · exact continuous_id.matrix_det
  have h2 : (SU d : Set (Matrix (Fin d) (Fin d) ℂ)) =
      {A : Matrix (Fin d) (Fin d) ℂ | (star A * A = 1 ∧ A * star A = 1) ∧ A.det = 1} :=
    Set.ext fun A => by
      rw [SetLike.mem_coe, Matrix.mem_specialUnitaryGroup_iff]
      exact Iff.rfl
  rw [h2]
  exact h1

instance (d : ℕ) : CompactSpace (SU d) := by
  refine isCompact_iff_compactSpace.mp ?_
  have hK : IsCompact (Set.univ.pi fun _ : Fin d => Set.univ.pi fun _ : Fin d =>
      Metric.closedBall (0 : ℂ) 1) :=
    isCompact_univ_pi fun _ => isCompact_univ_pi fun _ => isCompact_closedBall _ _
  refine hK.of_isClosed_subset (isClosed_SU d) ?_
  intro A hA
  simp only [Set.mem_pi, Set.mem_univ, Metric.mem_closedBall, dist_zero_right, forall_const]
  intro i j
  exact entry_norm_bound_of_unitary (Matrix.specialUnitaryGroup_le_unitaryGroup hA) i j

/-- The normalized Haar measure on `SU(d)`: the unique left-invariant regular Borel
probability measure. -/
noncomputable def haarSU (d : ℕ) : Measure (SU d) := Measure.haarMeasure ⊤

instance (d : ℕ) : (haarSU d).IsHaarMeasure := by
  unfold haarSU; infer_instance

instance (d : ℕ) : IsProbabilityMeasure (haarSU d) := ⟨by
  unfold haarSU
  simpa using Measure.haarMeasure_self (K₀ := (⊤ : TopologicalSpace.PositiveCompacts (SU d)))⟩

/-- The averaging (Hecke) operator attached to `g₁, …, gₙ` in a group `G`, acting on functions
`f : G → ℂ` by `(z f)(x) = ∑ⱼ (f (gⱼ x) + f (gⱼ⁻¹ x))`. -/
noncomputable def heckeOperator {G : Type*} [Group G] {n : ℕ} (g : Fin n → G) (f : G → ℂ) : G → ℂ :=
  fun x => ∑ j, (f (g j * x) + f ((g j)⁻¹ * x))

/-- The Hecke operator of `g₁, …, gₙ` has a spectral gap on `L²(G, μ)`: there is `ε > 0` such
that for every square-integrable `f` with `∫ f dμ = 0` (i.e. `f ⊥` constants),
`Re ⟨z f, f⟩ ≤ (2n - ε) ‖f‖²`. Equivalently, the supremum `λ₁` of the spectrum of the
self-adjoint operator `z` on `L²₀(G, μ)` satisfies `λ₁ < 2n`. -/
def HasSpectralGap {G : Type*} [Group G] [MeasurableSpace G] (μ : Measure G) {n : ℕ}
    (g : Fin n → G) : Prop :=
  ∃ ε : ℝ, 0 < ε ∧ ∀ f : G → ℂ, MemLp f 2 μ → ∫ x, f x ∂μ = 0 →
    (∫ x, starRingEnd ℂ (f x) * heckeOperator g f x ∂μ).re ≤
      (2 * n - ε) * ∫ x, ‖f x‖ ^ 2 ∂μ

end SpectralGapConjecture
Source
Wikipedia, "Spectral gap conjecture", section Statement (https://en.wikipedia.org/wiki/Spectral_gap_conjecture); operator and spectral gap as defined in Bourgain–Gamburd, A spectral gap theorem in SU(d), arXiv:1108.6264, §0, p. 1, and Fisher, Out(F_n) and the spectral gap conjecture, arXiv:math/0601050, §1, p. 1
Human review
  • Endorsed by Shuze Chen · Oct 13, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Lucas · Oct 13, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me