Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Section 3.5 / Appendix B — the basic theorem of operator-algebra quantum error correction

Proved
ADH2015.oaqec_basic_theorem

by Alien60 · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

holographyoperator-algebra-qecquantum-error-correctionquantum-gravity

Let H=HE⊗HEˉ\mathcal H=\mathcal H_E\otimes\mathcal H_{\bar E}H=HE​⊗HEˉ​ be finite-dimensional (EEE the erased part, Eˉ\bar EEˉ the retained part) and let HC⊆H\mathcal H_{\mathcal C}\subseteq\mathcal HHC​⊆H be a code subspace. An operator OOO acts within HC\mathcal H_{\mathcal C}HC​ if OOO and O†O^\daggerO† both map HC\mathcal H_{\mathcal C}HC​ into itself. It has a representation on Eˉ\bar EEˉ if some OEˉ=1E⊗YO_{\bar E}=\mathbb 1_E\otimes YOEˉ​=1E​⊗Y satisfies OEˉ∣ψ~⟩=O∣ψ~⟩O_{\bar E}|\tilde\psi\rangle=O|\tilde\psi\rangleOEˉ​∣ψ~​⟩=O∣ψ~​⟩ and OEˉ†∣ψ~⟩=O†∣ψ~⟩O_{\bar E}^\dagger|\tilde\psi\rangle=O^\dagger|\tilde\psi\rangleOEˉ†​∣ψ~​⟩=O†∣ψ~​⟩ for all ∣ψ~⟩∈HC|\tilde\psi\rangle\in\mathcal H_{\mathcal C}∣ψ~​⟩∈HC​ (and analogously on EEE, with X⊗1EˉX\otimes\mathbb 1_{\bar E}X⊗1Eˉ​). Let OOO act within HC\mathcal H_{\mathcal C}HC​ (Eq. (3.26)). Then OOO has a representation OEˉO_{\bar E}OEˉ​ on Eˉ\bar EEˉ (Eq. (3.27)) if and only if OOO commutes, on the code subspace, with every operator on EEE (Eq. (3.28)):

⟨ψ~′∣ [O,  XE⊗1Eˉ] ∣ψ~⟩=0for all operators XE on E and all ∣ψ~⟩,∣ψ~′⟩∈HC.\langle\tilde\psi'|\,[O,\;X_E\otimes\mathbb 1_{\bar E}]\,|\tilde\psi\rangle=0\qquad\text{for all operators }X_E\text{ on }E\text{ and all }|\tilde\psi\rangle,|\tilde\psi'\rangle\in\mathcal H_{\mathcal C}.⟨ψ~​′∣[O,XE​⊗1Eˉ​]∣ψ~​⟩=0for all operators XE​ on E and all ∣ψ~​⟩,∣ψ~​′⟩∈HC​.

The theorem characterizes, by a checkable commutation condition, which logical operators survive the erasure of EEE. Applied to all operators acting within the code subspace it recovers the erasure-correction criterion of Section 3.2, and it underlies the algebraic no-cloning property and the holographic interpretation of bulk reconstruction.

Formalization Note The source states (3.28) as ⟨ı~∣[O,XE]∣ȷ~⟩=0\langle\tilde\imath|[O,X_E]|\tilde\jmath\rangle=0⟨~∣[O,XE​]∣~​⟩=0 for an orthonormal basis ∣ı~⟩|\tilde\imath\rangle∣~⟩ of the code subspace; by sesquilinearity this is the same as vanishing on all pairs of code vectors, which is how it is stated here (no basis is fixed).

Preamble
import Mathlib
import Definitions.Def_ADH2015_defs

open Matrix
open scoped Kronecker
Formal statement
namespace ADH2015

theorem oaqec_basic_theorem {e ē : Type*} [Fintype e] [Fintype ē] [DecidableEq e]
    [DecidableEq ē] (C : Submodule ℂ (e × ē → ℂ)) (O : Matrix (e × ē) (e × ē) ℂ)
    (hO : ActsWithin C O) :
    HasRepOnBar C O ↔
      ∀ X : Matrix e e ℂ, ∀ v ∈ C, ∀ w ∈ C,
        star w ⬝ᵥ ((O * (X ⊗ₖ (1 : Matrix ē ē ℂ)) - (X ⊗ₖ (1 : Matrix ē ē ℂ)) * O) *ᵥ v) = 0 := by sorry

end ADH2015
Source
A. Almheiri, X. Dong, D. Harlow, Bulk Locality and Quantum Error Correction in AdS/CFT, JHEP 04 (2015) 163, https://arxiv.org/abs/1411.7041v3, Section 3.5, Theorem (Eqs. (3.26)–(3.28)); proof in Appendix B, Eqs. (B.1)–(B.17)
Read-back

What the Lean code literally says, in plain math · claude-opus-5-5

Setting. Let EEE and Eˉ\bar EEˉ be arbitrary finite types (index sets, possibly in different universes, possibly empty), each with decidable equality. The ambient space is H=CE×Eˉ\mathcal H = \mathbb C^{E\times\bar E}H=CE×Eˉ, i.e. all functions E×Eˉ→CE\times\bar E\to\mathbb CE×Eˉ→C, viewed as CE⊗CEˉ\mathbb C^{E}\otimes\mathbb C^{\bar E}CE⊗CEˉ with the first factor indexed by EEE and the second by Eˉ\bar EEˉ. Let C⊆HC\subseteq\mathcal HC⊆H be an arbitrary complex linear subspace (no orthonormal basis, dimension, or other structure assumed), and let OOO be an arbitrary complex (E×Eˉ)×(E×Eˉ)(E\times\bar E)\times(E\times\bar E)(E×Eˉ)×(E×Eˉ) matrix (not assumed Hermitian, unitary, or anything else). The Kronecker product is (A⊗B)(i,j),(k,l)=AikBjl(A\otimes B)_{(i,j),(k,l)} = A_{ik}B_{jl}(A⊗B)(i,j),(k,l)​=Aik​Bjl​, so X⊗1EˉX\otimes \mathbb 1_{\bar E}X⊗1Eˉ​ acts only on the EEE-factor and 1E⊗Y\mathbb 1_E\otimes Y1E​⊗Y acts only on the Eˉ\bar EEˉ-factor. O†O^\daggerO† denotes the conjugate transpose. The pairing used is

⟨w,u⟩=∑x∈E×Eˉwx‾ ux,\langle w, u\rangle = \sum_{x\in E\times\bar E}\overline{w_x}\,u_x,⟨w,u⟩=x∈E×Eˉ∑​wx​​ux​,

conjugate-linear in the first argument www, linear in the second.

Hypothesis ("OOO acts within CCC"). For every v∈Cv\in Cv∈C, both Ov∈COv\in COv∈C and O†v∈CO^\dagger v\in CO†v∈C; i.e. OC⊆COC\subseteq COC⊆C and O†C⊆CO^\dagger C\subseteq CO†C⊆C.

Left-hand side ("OOO has a representation on Eˉ\bar EEˉ"). There exists a complex Eˉ×Eˉ\bar E\times\bar EEˉ×Eˉ matrix YYY (no further condition on YYY: not required Hermitian, unitary, etc.) such that for every v∈Cv\in Cv∈C,

(1E⊗Y) v=O vand(1E⊗Y)† v=O†v.(\mathbb 1_E\otimes Y)\,v = O\,v \quad\text{and}\quad (\mathbb 1_E\otimes Y)^\dagger\, v = O^\dagger v .(1E​⊗Y)v=Ovand(1E​⊗Y)†v=O†v.

(Note (1E⊗Y)†=1E⊗Y†(\mathbb 1_E\otimes Y)^\dagger=\mathbb 1_E\otimes Y^\dagger(1E​⊗Y)†=1E​⊗Y†.) The equalities are required only on vectors of CCC, not on all of H\mathcal HH; a single YYY must serve for all v∈Cv\in Cv∈C and for both equations.

Right-hand side (commutator condition). For every complex E×EE\times EE×E matrix XXX (arbitrary: not restricted to Hermitian or unitary ones), and for all v∈Cv\in Cv∈C and all w∈Cw\in Cw∈C,

⟨w, (O (X⊗1Eˉ)−(X⊗1Eˉ) O) v⟩=0.\big\langle w,\ \big(O\,(X\otimes\mathbb 1_{\bar E}) - (X\otimes\mathbb 1_{\bar E})\,O\big)\,v\big\rangle = 0 .⟨w, (O(X⊗1Eˉ​)−(X⊗1Eˉ​)O)v⟩=0.

That is, the commutator [O, X⊗1Eˉ][O,\,X\otimes\mathbb 1_{\bar E}][O,X⊗1Eˉ​] has vanishing matrix elements between any two vectors of CCC (equivalently, its compression to CCC by the orthogonal projection onto CCC is zero). The quantifier order is ∀X ∀v∈C ∀w∈C\forall X\ \forall v\in C\ \forall w\in C∀X ∀v∈C ∀w∈C.

Assertion. Under the hypothesis that OOO and O†O^\daggerO† both map CCC into CCC, the following are equivalent (both directions are claimed):

  • (⇒\Rightarrow⇒) if some YYY on Eˉ\bar EEˉ satisfies (1E⊗Y)v=Ov(\mathbb 1_E\otimes Y)v = Ov(1E​⊗Y)v=Ov and (1E⊗Y†)v=O†v(\mathbb 1_E\otimes Y^\dagger)v = O^\dagger v(1E​⊗Y†)v=O†v for all v∈Cv\in Cv∈C, then ⟨w,[O,X⊗1Eˉ]v⟩=0\langle w,[O,X\otimes\mathbb 1_{\bar E}]v\rangle=0⟨w,[O,X⊗1Eˉ​]v⟩=0 for every XXX on EEE and all v,w∈Cv,w\in Cv,w∈C;
  • (⇐\Leftarrow⇐) if ⟨w,[O,X⊗1Eˉ]v⟩=0\langle w,[O,X\otimes\mathbb 1_{\bar E}]v\rangle=0⟨w,[O,X⊗1Eˉ​]v⟩=0 for every XXX on EEE and all v,w∈Cv,w\in Cv,w∈C, then such a YYY exists.

No condition on CCC about correctability of erasure of EEE (or anything else) is assumed; the only hypothesis beyond finiteness/decidable equality of E,EˉE,\bar EE,Eˉ is the invariance of CCC under OOO and O†O^\daggerO†.

Edge cases. If C={0}C=\{0\}C={0}, the hypothesis holds for every OOO, the left side holds (any YYY, e.g. Y=0Y=0Y=0), and the right side holds vacuously, so the equivalence is trivially true. If EEE or Eˉ\bar EEˉ is empty, then E×EˉE\times\bar EE×Eˉ is empty, H\mathcal HH is the zero space, C={0}C=\{0\}C={0} necessarily, and the statement reduces to the trivial case above. If Eˉ\bar EEˉ is empty but EEE is not (or vice versa), the same applies. When EEE is a single point, X⊗1EˉX\otimes\mathbb 1_{\bar E}X⊗1Eˉ​ is a scalar multiple of the identity, so the right side holds automatically for every OOO, and the theorem then asserts that every OOO for which CCC is invariant under both OOO and O†O^\daggerO† has such a YYY.

Human review
  • Endorsed by Shuze Chen · Oct 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Alien60 · Oct 9, 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