Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Finite split reflection positivity and infrared-spectrum interface

Definition
FiniteReflectionPositivityMethodsI

by lisamegawatts · Sep 8, 2026 · Mathlib c5ea003 (Lean v4.30.0)

discrete-fourier-transformfinite-dimensionalinfrared-boundreflection-positivitystatistical-mechanics

This finite interface fixes the domains needed by the reflection-positivity and infrared-bound algebra. For a finite index set, it defines real quadratic forms, positive-semidefinite kernels, full and zero-mode-restricted kernel domination, and the diagonal free covariance. A reflected configuration is a pair of half-configurations whose coordinates are exchanged; observables are explicitly lifted from the plus half. A split weight has the finite feature expansion

W(x,y)=∑a∈Acaϕa(x)ϕa(y).W(x,y)=\sum_{a\in A}c_a\phi_a(x)\phi_a(y).W(x,y)=a∈A∑​ca​ϕa​(x)ϕa​(y).

The bundle also defines the exact four-point discrete Fourier modes, the nearest-neighbor Laplacian on the 4×44\times44×4 torus, and its eigenvalue function. The distinguished constant mode is built into InfraredSpectrum, and admissible infrared sources satisfy f(0)=0f(0)=0f(0)=0. Two finite controls record why full-vertex reflection positivity and unrestricted domination by a zero-mode-removed covariance are not valid replacements. The Prop-valued gates are the precise theorem interfaces used by the mission; the bundle itself makes no claim about a physical Gibbs measure, Gaussian domination, a thermodynamic limit, or a Kosterlitz--Thouless transition.

Definition code
import Mathlib.Algebra.BigOperators.Fin
import Mathlib.Data.Complex.Basic
import Mathlib.Data.Matrix.Basic
import Mathlib.Data.Real.Basic

noncomputable section

open scoped BigOperators ComplexConjugate

namespace FiniteReflectionPositivityMethodsI

/-! ## Finite positive-semidefinite kernels -/

def quadraticForm {I : Type*} [Fintype I]
    (K : Matrix I I ℝ) (f : I → ℝ) : ℝ :=
  ∑ i, ∑ j, f i * K i j * f j

def PositiveSemidefinite {I : Type*} [Fintype I]
    (K : Matrix I I ℝ) : Prop :=
  (∀ i j, K i j = K j i) ∧ ∀ f, 0 ≤ quadraticForm K f

def KernelDominated {I : Type*} [Fintype I]
    (C G : Matrix I I ℝ) : Prop :=
  (∀ i j, C i j = C j i) ∧
    (∀ i j, G i j = G j i) ∧
    ∀ f, quadraticForm C f ≤ quadraticForm G f

/-! ## Reflection on a split finite configuration -/

abbrev FullConfiguration (X : Type*) := X × X

def reflectConfiguration {X : Type*}
    (omega : FullConfiguration X) : FullConfiguration X :=
  (omega.2, omega.1)

def plusLift {X : Type*} (observable : X → ℝ) : FullConfiguration X → ℝ :=
  fun omega => observable omega.1

def reflectObservable {X : Type*}
    (observable : FullConfiguration X → ℝ) : FullConfiguration X → ℝ :=
  fun omega => observable (reflectConfiguration omega)

def reflectedPairing {X : Type*} [Fintype X]
    (weight : X → X → ℝ) (left right : X → ℝ) : ℝ :=
  ∑ omega : FullConfiguration X,
    weight omega.1 omega.2 * plusLift left omega *
      reflectObservable (plusLift right) omega

def reflectedKernel {X I : Type*} [Fintype X]
    (weight : X → X → ℝ) (observable : I → X → ℝ) : Matrix I I ℝ :=
  fun i j => reflectedPairing weight (observable i) (observable j)

structure SplitWeightData (X A : Type*) where
  coefficient : A → ℝ
  feature : A → X → ℝ

def SplitWeightData.weight {X A : Type*} [Fintype A]
    (data : SplitWeightData X A) (left right : X) : ℝ :=
  ∑ a, data.coefficient a * data.feature a left * data.feature a right

def HalfObservableBindingGate : Prop :=
  ∀ (X : Type) [Fintype X] (weight : X → X → ℝ)
      (left right : X → ℝ),
    reflectedPairing weight left right =
      ∑ x, ∑ y, weight x y * left x * right y

def SplitWeightReflectionPositivityGate : Prop :=
  ∀ (X A I : Type) [Fintype X] [Fintype A] [Fintype I]
      (data : SplitWeightData X A),
    (∀ a, 0 ≤ data.coefficient a) →
      ∀ observable : I → X → ℝ,
        PositiveSemidefinite
          (reflectedKernel data.weight observable)

def ChessboardGate : Prop :=
  ∀ (I : Type) [Fintype I] (K : Matrix I I ℝ),
    PositiveSemidefinite K →
      ∀ i j, K i j ^ 2 ≤ K i i * K j j

/-! ## Zero-mode-restricted infrared domination -/

structure InfraredSpectrum (I : Type*) [Zero I] where
  eigenvalue : I → ℝ
  nonnegative : ∀ k, 0 ≤ eigenvalue k
  eq_zero_iff : ∀ k, eigenvalue k = 0 ↔ k = 0

def zeroModeFree {I : Type*} [Zero I] (f : I → ℝ) : Prop :=
  f 0 = 0

def KernelDominatedOffZero {I : Type*} [Fintype I] [Zero I]
    (C G : Matrix I I ℝ) : Prop :=
  (∀ i j, C i j = C j i) ∧
    (∀ i j, G i j = G j i) ∧
    ∀ f, zeroModeFree f → quadraticForm C f ≤ quadraticForm G f

def basisVector {I : Type*} [DecidableEq I] (k : I) : I → ℝ :=
  fun j => if j = k then 1 else 0

def freeFieldModeCovariance {I : Type*} [Fintype I] [DecidableEq I] [Zero I]
    (beta : ℝ) (spectrum : InfraredSpectrum I) : Matrix I I ℝ :=
  fun k q =>
    if k = q then
      if k = 0 then 0 else 1 / (beta * spectrum.eigenvalue k)
    else 0

def FreeCovariancePositiveSemidefiniteGate : Prop :=
  ∀ (I : Type) [Fintype I] [DecidableEq I] [Zero I]
      (beta : ℝ) (spectrum : InfraredSpectrum I),
    0 < beta →
      PositiveSemidefinite (freeFieldModeCovariance beta spectrum)

def OffZeroModeInfraredGate : Prop :=
  ∀ (I : Type) [Fintype I] [DecidableEq I] [Zero I]
      (beta : ℝ) (spectrum : InfraredSpectrum I)
      (C : Matrix I I ℝ),
    0 < beta →
      KernelDominatedOffZero C (freeFieldModeCovariance beta spectrum) →
      ∀ k, k ≠ 0 →
        C k k ≤ 1 / (beta * spectrum.eigenvalue k)

/-! ## Concrete four-cycle and four-by-four torus spectrum -/

abbrev Cycle4 := Fin 4
abbrev Torus4 := Cycle4 × Cycle4

def cycle4Next (x : Cycle4) : Cycle4 :=
  ⟨(x.val + 1) % 4, Nat.mod_lt _ (by norm_num)⟩

def cycle4Prev (x : Cycle4) : Cycle4 :=
  ⟨(x.val + 3) % 4, Nat.mod_lt _ (by norm_num)⟩

def dft4Mode (k x : Cycle4) : ℂ :=
  match k.val, x.val with
  | 0, _ => 1
  | 1, 0 => 1
  | 1, 1 => Complex.I
  | 1, 2 => -1
  | 1, _ => -Complex.I
  | 2, 0 => 1
  | 2, 1 => -1
  | 2, 2 => 1
  | 2, _ => -1
  | _, 0 => 1
  | _, 1 => -Complex.I
  | _, 2 => -1
  | _, _ => Complex.I

def cycle4Eigenvalue (k : Cycle4) : ℝ :=
  if k.val = 0 then 0 else if k.val = 2 then 4 else 2

def cycle4Laplacian (f : Cycle4 → ℂ) (x : Cycle4) : ℂ :=
  2 * f x - f (cycle4Next x) - f (cycle4Prev x)

def torus4Mode (k x : Torus4) : ℂ :=
  dft4Mode k.1 x.1 * dft4Mode k.2 x.2

def torus4Laplacian (f : Torus4 → ℂ) (x : Torus4) : ℂ :=
  4 * f x - f (cycle4Next x.1, x.2) - f (cycle4Prev x.1, x.2) -
    f (x.1, cycle4Next x.2) - f (x.1, cycle4Prev x.2)

def torus4Eigenvalue (k : Torus4) : ℝ :=
  cycle4Eigenvalue k.1 + cycle4Eigenvalue k.2

def Torus4EigenmodeGate : Prop :=
  ∀ k x : Torus4,
    torus4Laplacian (torus4Mode k) x =
      torus4Eigenvalue k * torus4Mode k x

def Torus4SpectrumGate : Prop :=
  (∀ k, 0 ≤ torus4Eigenvalue k) ∧
    (∀ k, torus4Eigenvalue k = 0 ↔ k = 0) ∧
    ∃ spectrum : InfraredSpectrum Torus4,
      spectrum.eigenvalue = torus4Eigenvalue

/-! ## Load-bearing negative controls -/

def swapFinTwo (i : Fin 2) : Fin 2 :=
  if i = 0 then 1 else 0

def diagonalFinTwo : Matrix (Fin 2) (Fin 2) ℝ :=
  fun i j => if i = j then 1 else 0

def reflectedFullVertexKernel : Matrix (Fin 2) (Fin 2) ℝ :=
  fun i j => diagonalFinTwo i (swapFinTwo j)

def FullVertexReflectionNoGo : Prop :=
  ¬ PositiveSemidefinite reflectedFullVertexKernel

def zeroRemovedFinTwo : Matrix (Fin 2) (Fin 2) ℝ :=
  fun i j => if i = j ∧ i ≠ 0 then 1 else 0

def ZeroModeRestrictionControl : Prop :=
  KernelDominatedOffZero diagonalFinTwo zeroRemovedFinTwo ∧
    ¬ KernelDominated diagonalFinTwo zeroRemovedFinTwo

def ClaimBoundary : Prop :=
  HalfObservableBindingGate ∧
    SplitWeightReflectionPositivityGate ∧
    ChessboardGate ∧
    Torus4EigenmodeGate ∧
    Torus4SpectrumGate ∧
    FreeCovariancePositiveSemidefiniteGate ∧
    OffZeroModeInfraredGate ∧
    FullVertexReflectionNoGo ∧
    ZeroModeRestrictionControl

end FiniteReflectionPositivityMethodsI
Source
LeanProofs, LeanProofs/StatMech/ReflectionPositivityInfraredBound.lean, commit dbf503b2909cc17787d40a21eb75a0c9354cc6ef, lines 40-56 (finite PSD/RP interface), 83-137 (chessboard inequality), 141-212 (infrared covariance), and 216-305 (Torus-4 DFT spectrum): https://github.com/MonumentalSystems/LeanProofs/blob/dbf503b2909cc17787d40a21eb75a0c9354cc6ef/LeanProofs/StatMech/ReflectionPositivityInfraredBound.lean . The explicit split-half and zero-mode controls are mission-specific corrections of the domains left abstract in that source.
Read-back

What the Lean code literally says, in plain math · GPT-5 Codex (exact deployment identifier not exposed)

This item introduces the following definitions and propositions. For any possibly empty finite type III, any real matrix K:I×I→RK:I\times I\to\mathbb RK:I×I→R, and any function f:I→Rf:I\to\mathbb Rf:I→R, its quadratic form is

QK(f)=∑i∈I∑j∈If(i)Kijf(j).Q_K(f)=\sum_{i\in I}\sum_{j\in I}f(i)K_{ij}f(j).QK​(f)=i∈I∑​j∈I∑​f(i)Kij​f(j).

The predicate PositiveSemidefinite⁡(K)\operatorname{PositiveSemidefinite}(K)PositiveSemidefinite(K) means both that Kij=KjiK_{ij}=K_{ji}Kij​=Kji​ for every i,ji,ji,j and that QK(f)≥0Q_K(f)\ge 0QK​(f)≥0 for every real-valued fff on III. The predicate KernelDominated⁡(C,G)\operatorname{KernelDominated}(C,G)KernelDominated(C,G) means that both CCC and GGG are symmetric and QC(f)≤QG(f)Q_C(f)\le Q_G(f)QC​(f)≤QG​(f) for every f:I→Rf:I\to\mathbb Rf:I→R. Empty index types are included, in which case all displayed sums are zero.

For any type XXX, a full configuration is an ordered pair (x,y)∈X×X(x,y)\in X\times X(x,y)∈X×X; reflection sends (x,y)(x,y)(x,y) to (y,x)(y,x)(y,x). The plus lift of h:X→Rh:X\to\mathbb Rh:X→R is the function (x,y)↦h(x)(x,y)\mapsto h(x)(x,y)↦h(x), and reflection of a function F:X×X→RF:X\times X\to\mathbb RF:X×X→R is (x,y)↦F(y,x)(x,y)\mapsto F(y,x)(x,y)↦F(y,x). For finite XXX, the reflected pairing of a weight w:X→X→Rw:X\to X\to\mathbb Rw:X→X→R and functions ℓ,r:X→R\ell,r:X\to\mathbb Rℓ,r:X→R is therefore

⟨ℓ,r⟩wref=∑(x,y)∈X×Xw(x,y)ℓ(x)r(y).\langle \ell,r\rangle_w^{\mathrm{ref}}=\sum_{(x,y)\in X\times X}w(x,y)\ell(x)r(y).⟨ℓ,r⟩wref​=(x,y)∈X×X∑​w(x,y)ℓ(x)r(y).

Given any index type III and observables Oi:X→RO_i:X\to\mathbb ROi​:X→R, the reflected kernel is the matrix Kij=⟨Oi,Oj⟩wrefK_{ij}=\langle O_i,O_j\rangle_w^{\mathrm{ref}}Kij​=⟨Oi​,Oj​⟩wref​. Split-weight data on types X,AX,AX,A consists of an arbitrary coefficient function c:A→Rc:A\to\mathbb Rc:A→R and arbitrary features ϕa:X→R\phi_a:X\to\mathbb Rϕa​:X→R; when AAA is finite, its associated weight is

w(x,y)=∑a∈Ac(a)ϕa(x)ϕa(y).w(x,y)=\sum_{a\in A}c(a)\phi_a(x)\phi_a(y).w(x,y)=a∈A∑​c(a)ϕa​(x)ϕa​(y).

The half-observable binding gate is the proposition that, for every possibly empty finite base-universe type XXX, every weight www, and every ℓ,r:X→R\ell,r:X\to\mathbb Rℓ,r:X→R, the reflected pairing equals the iterated sum ∑x∑yw(x,y)ℓ(x)r(y)\sum_x\sum_y w(x,y)\ell(x)r(y)∑x​∑y​w(x,y)ℓ(x)r(y). The split-weight reflection-positivity gate is the proposition that, for every possibly empty finite base-universe types X,A,IX,A,IX,A,I and every split-weight datum, if c(a)≥0c(a)\ge0c(a)≥0 for every aaa, then for every observable family O:I→(X→R)O:I\to(X\to\mathbb R)O:I→(X→R) the matrix

Kij=∑(x,y)∈X×X(∑a∈Ac(a)ϕa(x)ϕa(y))Oi(x)Oj(y)K_{ij}=\sum_{(x,y)\in X\times X}\left(\sum_{a\in A}c(a)\phi_a(x)\phi_a(y)\right)O_i(x)O_j(y)Kij​=(x,y)∈X×X∑​(a∈A∑​c(a)ϕa​(x)ϕa​(y))Oi​(x)Oj​(y)

is symmetric and satisfies ∑i∑jf(i)Kijf(j)≥0\sum_i\sum_j f(i)K_{ij}f(j)\ge0∑i​∑j​f(i)Kij​f(j)≥0 for every f:I→Rf:I\to\mathbb Rf:I→R. No boundedness, normalization, or nonzero condition is imposed on the features or observables; when AAA, XXX, or III is empty, the relevant sums and universal conditions have their usual empty meanings. The chessboard gate is the proposition that every symmetric real matrix KKK on every possibly empty finite base-universe type III whose quadratic form is nonnegative on every f:I→Rf:I\to\mathbb Rf:I→R satisfies Kij2≤KiiKjjK_{ij}^2\le K_{ii}K_{jj}Kij2​≤Kii​Kjj​ for every i,ji,ji,j; the conclusion is vacuous if III is empty and also includes i=ji=ji=j.

For any type III equipped with a distinguished element denoted 000, an infrared spectrum consists of a function λ:I→R\lambda:I\to\mathbb Rλ:I→R such that λ(k)≥0\lambda(k)\ge0λ(k)≥0 for every kkk and λ(k)=0\lambda(k)=0λ(k)=0 exactly when k=0k=0k=0. The distinguished-zero structure supplies an element of III, so every type occurring in this definition is nonempty. A function f:I→Rf:I\to\mathbb Rf:I→R is zero-mode-free exactly when f(0)=0f(0)=0f(0)=0; here too the distinguished-zero structure makes III nonempty. For a finite, necessarily nonempty type III equipped with a distinguished zero, KernelDominatedOffZero⁡(C,G)\operatorname{KernelDominatedOffZero}(C,G)KernelDominatedOffZero(C,G) means that CCC and GGG are symmetric and QC(f)≤QG(f)Q_C(f)\le Q_G(f)QC​(f)≤QG​(f) for every fff satisfying f(0)=0f(0)=0f(0)=0. For a type with decidable equality and a chosen k∈Ik\in Ik∈I, the basis vector at kkk is ek(j)=1e_k(j)=1ek​(j)=1 if j=kj=kj=k and 000 otherwise. Given a finite, necessarily nonempty type III with decidable equality and a distinguished zero, a real number β\betaβ, and an infrared spectrum λ\lambdaλ, the free-field mode covariance is the diagonal matrix

Gkq={0,k≠q,0,k=q=0,1βλ(k),k=q≠0.G_{kq}=\begin{cases}0,&k\ne q,\\0,&k=q=0,\\[2pt]\dfrac{1}{\beta\lambda(k)},&k=q\ne0.\end{cases}Gkq​=⎩⎨⎧​0,0,βλ(k)1​,​k=q,k=q=0,k=q=0.​

The definition itself permits every real β\betaβ; real division is total, so a zero denominator yields the ambient field's totalized value. The free-covariance positive-semidefinite gate says that for every finite, necessarily nonempty base-universe type III with decidable equality and a distinguished zero, every β>0\beta>0β>0, and every infrared spectrum, this GGG is symmetric and has nonnegative quadratic form on every real fff. The off-zero-mode infrared gate says that for every finite, necessarily nonempty base-universe type III with decidable equality and a distinguished zero, every β\betaβ, every infrared spectrum λ\lambdaλ, and every real matrix CCC, if β>0\beta>0β>0 and CCC is dominated by GGG on all fff with f(0)=0f(0)=0f(0)=0 in the preceding symmetric-matrix sense, then for every k≠0k\ne0k=0,

Ckk≤1βλ(k).C_{kk}\le\frac{1}{\beta\lambda(k)}.Ckk​≤βλ(k)1​.

No positive-semidefiniteness assumption on CCC is separately included. Although III is nonempty because it contains 000, it may have no element distinct from 000, in which case the final conclusion over k≠0k\ne0k=0 is vacuous.

The four-cycle is Fin⁡(4)={0,1,2,3}\operatorname{Fin}(4)=\{0,1,2,3\}Fin(4)={0,1,2,3}. Its next and previous maps send xxx respectively to (x+1) mod 4(x+1)\bmod4(x+1)mod4 and (x+3) mod 4(x+3)\bmod4(x+3)mod4. The complex Fourier mode dk(x)d_k(x)dk​(x) has rows, in the order x=0,1,2,3x=0,1,2,3x=0,1,2,3,

d0=(1,1,1,1),d1=(1,i,−1,−i),d2=(1,−1,1,−1),d3=(1,−i,−1,i).d_0=(1,1,1,1),\quad d_1=(1,i,-1,-i),\quad d_2=(1,-1,1,-1),\quad d_3=(1,-i,-1,i).d0​=(1,1,1,1),d1​=(1,i,−1,−i),d2​=(1,−1,1,−1),d3​=(1,−i,−1,i).

Its declared eigenvalue is λ4(0)=0\lambda_4(0)=0λ4​(0)=0, λ4(2)=4\lambda_4(2)=4λ4​(2)=4, and λ4(1)=λ4(3)=2\lambda_4(1)=\lambda_4(3)=2λ4​(1)=λ4​(3)=2. The cycle Laplacian is (Δ4f)(x)=2f(x)−f(x+1)−f(x−1)(\Delta_4f)(x)=2f(x)-f(x+1)-f(x-1)(Δ4​f)(x)=2f(x)−f(x+1)−f(x−1) with the preceding modular maps. The four-by-four torus is T=Fin⁡(4)×Fin⁡(4)T=\operatorname{Fin}(4)\times\operatorname{Fin}(4)T=Fin(4)×Fin(4); for k=(k1,k2)k=(k_1,k_2)k=(k1​,k2​) and x=(x1,x2)x=(x_1,x_2)x=(x1​,x2​), its mode is Dk(x)=dk1(x1)dk2(x2)D_k(x)=d_{k_1}(x_1)d_{k_2}(x_2)Dk​(x)=dk1​​(x1​)dk2​​(x2​), its Laplacian is

(ΔTf)(x)=4f(x)−f(x1+1,x2)−f(x1−1,x2)−f(x1,x2+1)−f(x1,x2−1),(\Delta_Tf)(x)=4f(x)-f(x_1+1,x_2)-f(x_1-1,x_2)-f(x_1,x_2+1)-f(x_1,x_2-1),(ΔT​f)(x)=4f(x)−f(x1​+1,x2​)−f(x1​−1,x2​)−f(x1​,x2​+1)−f(x1​,x2​−1),

with all coordinates reduced modulo 444, and its declared eigenvalue is Λ(k)=λ4(k1)+λ4(k2)\Lambda(k)=\lambda_4(k_1)+\lambda_4(k_2)Λ(k)=λ4​(k1​)+λ4​(k2​). The torus eigenmode gate says that ΔTDk(x)=Λ(k)Dk(x)\Delta_TD_k(x)=\Lambda(k)D_k(x)ΔT​Dk​(x)=Λ(k)Dk​(x) for every k,x∈Tk,x\in Tk,x∈T, with the real scalar viewed as complex. The torus spectrum gate says simultaneously that Λ(k)≥0\Lambda(k)\ge0Λ(k)≥0 for all kkk, that Λ(k)=0\Lambda(k)=0Λ(k)=0 exactly when k=(0,0)k=(0,0)k=(0,0), and that there exists an infrared-spectrum structure on the nonempty type TTT whose eigenvalue function is exactly Λ\LambdaΛ.

On Fin⁡(2)={0,1}\operatorname{Fin}(2)=\{0,1\}Fin(2)={0,1}, the swap map exchanges 000 and 111, the matrix DDD is Dij=1D_{ij}=1Dij​=1 when i=ji=ji=j and 000 otherwise, and the reflected full-vertex kernel is Rij=Di,swap⁡(j)R_{ij}=D_{i,\operatorname{swap}(j)}Rij​=Di,swap(j)​, namely R=(0110)R=\begin{pmatrix}0&1\\1&0\end{pmatrix}R=(01​10​). The full-vertex reflection no-go proposition is ¬PositiveSemidefinite⁡(R)\neg\operatorname{PositiveSemidefinite}(R)¬PositiveSemidefinite(R), where positive semidefiniteness has the symmetry-and-quadratic-form meaning above. The zero-removed matrix is Zij=1Z_{ij}=1Zij​=1 exactly when i=ji=ji=j and i≠0i\ne0i=0, hence Z=(0001)Z=\begin{pmatrix}0&0\\0&1\end{pmatrix}Z=(00​01​). The zero-mode restriction control is the conjunction that DDD and ZZZ are symmetric and QD(f)≤QZ(f)Q_D(f)\le Q_Z(f)QD​(f)≤QZ​(f) for every fff with f(0)=0f(0)=0f(0)=0, and that it is not the case that both matrices are symmetric and QD(f)≤QZ(f)Q_D(f)\le Q_Z(f)QD​(f)≤QZ​(f) for every unrestricted fff. Finally, the claim-boundary proposition is the conjunction, in order, of the half-observable binding gate, split-weight reflection-positivity gate, chessboard gate, torus eigenmode gate, torus spectrum gate, free-covariance positive-semidefinite gate, off-zero-mode infrared gate, full-vertex reflection no-go proposition, and zero-mode restriction control. These declarations define the named propositions; this definition item does not itself supply proofs that the gates or their conjunction hold.

Human review
  • Endorsed by Shuze Chen · Sep 18, 2026

  • Endorsed by lisamegawatts · Sep 18, 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 paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me