Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Theorem 4.1 — four equivalent characterizations of a correctable erasure

Proved
HarlowTASI.erasure_correction_tfae

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

holographyoperator-algebrasquantum-error-correctionquantum-information

Let H=HR⊗HR‾\mathcal H=\mathcal H_R\otimes\mathcal H_{\overline R}H=HR​⊗HR​ be finite-dimensional and let Hcode⊆H\mathcal H_{code}\subseteq\mathcal HHcode​⊆H be a nonzero subspace with orthonormal basis ∣i~⟩|\tilde i\rangle∣i~⟩, i∈Ci\in Ci∈C, and orthogonal projection PPP. The following are equivalent:

  1. For every operator O~\tilde OO~ on Hcode\mathcal H_{code}Hcode​ there is an operator ORO_ROR​ on HR\mathcal H_RHR​ such that for all ∣ψ~⟩∈Hcode|\tilde\psi\rangle\in\mathcal H_{code}∣ψ~​⟩∈Hcode​
OR∣ψ~⟩=O~∣ψ~⟩,OR†∣ψ~⟩=O~†∣ψ~⟩.(4.16)O_R|\tilde\psi\rangle=\tilde O|\tilde\psi\rangle,\qquad O_R^\dagger|\tilde\psi\rangle=\tilde O^\dagger|\tilde\psi\rangle. \tag{4.16}OR​∣ψ~​⟩=O~∣ψ~​⟩,OR†​∣ψ~​⟩=O~†∣ψ~​⟩.(4.16)
  1. For every operator XR‾X_{\overline R}XR​ on HR‾\mathcal H_{\overline R}HR​ there is λ∈C\lambda\in\mathbb Cλ∈C with PXR‾P=λPPX_{\overline R}P=\lambda PPXR​P=λP. (4.17)
  2. With a reference system SSS of dimension ∣S∣=dim⁡Hcode|S|=\dim\mathcal H_{code}∣S∣=dimHcode​ and ∣ϕ⟩=∣S∣−1/2∑i∣i⟩S∣i~⟩RR‾|\phi\rangle=|S|^{-1/2}\sum_i|i\rangle_S|\tilde i\rangle_{R\overline R}∣ϕ⟩=∣S∣−1/2∑i​∣i⟩S​∣i~⟩RR​ (4.18), one has ρSR‾(ϕ)=ρS(ϕ)⊗ρR‾(ϕ)\rho_{S\overline R}(\phi)=\rho_S(\phi)\otimes\rho_{\overline R}(\phi)ρSR​(ϕ)=ρS​(ϕ)⊗ρR​(ϕ). (4.19)
  3. ∣S∣≤∣R∣|S|\le|R|∣S∣≤∣R∣, and for a decomposition HR=(HR1⊗HR2)⊕HR3\mathcal H_R=(\mathcal H_{R_1}\otimes\mathcal H_{R_2})\oplus\mathcal H_{R_3}HR​=(HR1​​⊗HR2​​)⊕HR3​​ with ∣R1∣=∣S∣|R_1|=|S|∣R1​∣=∣S∣ and ∣R3∣<∣S∣|R_3|<|S|∣R3​∣<∣S∣ there are a unitary URU_RUR​ on HR\mathcal H_RHR​ and a state ∣χ⟩R2R‾|\chi\rangle_{R_2\overline R}∣χ⟩R2​R​ such that
∣i~⟩RR‾=UR(∣i⟩R1∣χ⟩R2R‾).(4.20)|\tilde i\rangle_{R\overline R}=U_R\big(|i\rangle_{R_1}|\chi\rangle_{R_2\overline R}\big).\tag{4.20}∣i~⟩RR​=UR​(∣i⟩R1​​∣χ⟩R2​R​).(4.20)

Theorem 4.1 is the basic structure theorem for erasure codes: condition (1) says that the encoded information can be reconstructed on RRR, and (4) exhibits the encoding circuit. It is the subsystem-code special case of the mechanism behind Theorem 5.2, and it is the input of the quantum Singleton bound (4.25).

Formalization Note Hcode\mathcal H_{code}Hcode​ is given by an isometry VVV (columns ∣i~⟩=V∣i⟩|\tilde i\rangle=V|i\rangle∣i~⟩=V∣i⟩), operators by matrices, and ORO_ROR​ acts as OR⊗IR‾O_R\otimes I_{\overline R}OR​⊗IR​. The code subspace is assumed nonzero (the lectures implicitly assume it: condition (4) requires ∣R3∣<∣S∣|R_3|<|S|∣R3​∣<∣S∣ and (4.18) normalizes by ∣S∣−1/2|S|^{-1/2}∣S∣−1/2). In (4), ∣R2∣=⌊∣R∣/∣S∣⌋|R_2|=\lfloor|R|/|S|\rfloor∣R2​∣=⌊∣R∣/∣S∣⌋ and ∣R3∣=∣R∣ mod ∣S∣|R_3|=|R|\bmod|S|∣R3​∣=∣R∣mod∣S∣ are forced, and the statement is required for every identification of the basis of HR\mathcal H_RHR​ with (R1×R2)⊔R3(R_1\times R_2)\sqcup R_3(R1​×R2​)⊔R3​; ∣χ⟩|\chi\rangle∣χ⟩ is required to be a unit vector.

Preamble
import Mathlib
import Definitions.Def_HarlowTASI_QuantumBasics

open Matrix
open scoped Kronecker
Formal statement
namespace HarlowTASI
theorem erasure_correction_tfae {R Rb C : Type} [Fintype R] [DecidableEq R] [Fintype Rb]
    [DecidableEq Rb] [Fintype C] [DecidableEq C] [Nonempty C]
    (V : Matrix (R × Rb) C ℂ) (hV : IsIsometry V) :
    List.TFAE
      [ ∀ O : Matrix C C ℂ, ∃ OR : Matrix R R ℂ,
          (OR ⊗ₖ (1 : Matrix Rb Rb ℂ)) * V = V * O ∧
          (ORᴴ ⊗ₖ (1 : Matrix Rb Rb ℂ)) * V = V * Oᴴ,
        ∀ X : Matrix Rb Rb ℂ, ∃ c : ℂ,
          (V * Vᴴ) * ((1 : Matrix R R ℂ) ⊗ₖ X) * (V * Vᴴ) = c • (V * Vᴴ),
        reducedSB (codeReferenceState V) =
          reducedS (codeReferenceState V) ⊗ₖ reducedB (codeReferenceState V),
        Fintype.card C ≤ Fintype.card R ∧
          ∀ e : R ≃ (C × Fin (Fintype.card R / Fintype.card C)) ⊕
              Fin (Fintype.card R % Fintype.card C),
            ∃ UR : Matrix R R ℂ, IsUnitaryMatrix UR ∧
              ∃ χ : Fin (Fintype.card R / Fintype.card C) × Rb → ℂ,
                ∑ x, ‖χ x‖ ^ 2 = 1 ∧
                ∀ (i : C) (r : R) (b : Rb),
                  V (r, b) i = ∑ r', UR r r' *
                    (match e r' with
                      | Sum.inl (c, k) => if c = i then χ (k, b) else 0
                      | Sum.inr _ => 0) ] := by sorry
end HarlowTASI
Source
Daniel Harlow, TASI Lectures on the Emergence of Bulk Physics in AdS/CFT, PoS(TASI2017)002 (2018), arXiv:1802.01040, §4.3, pp. 28–29, Theorem 4.1, eqs. (4.16)–(4.20).
Read-back

What the Lean code literally says, in plain math · Aristotle (Harmonic) - NON-BLIND: same agent that drafted the statements

Non-blind read-back — not independent testimony. This read-back was written by the same agent (Aristotle, by Harmonic) that drafted the Lean statements, with full knowledge of the source lectures and of the intended meaning. It was not produced blind by an independent auditor, so it must not be mistaken for independent testimony; reviewers should treat it as the drafter's own rendering of the code and check the Lean text directly.

Data. Finite sets RRR, R‾\overline RR, CCC with decidable equality, CCC nonempty; a complex matrix VVV with rows indexed by R×R‾R\times\overline RR×R and columns by CCC, assumed to be an isometry, V†V=ICV^\dagger V=I_CV†V=IC​.

Claim. The following four statements are pairwise equivalent:

  1. For every C×CC\times CC×C complex matrix OOO there is an R×RR\times RR×R matrix ORO_ROR​ with
(OR⊗IR‾)V=VOand(OR†⊗IR‾)V=VO†.(O_R\otimes I_{\overline R})V=VO\quad\text{and}\quad(O_R^\dagger\otimes I_{\overline R})V=VO^\dagger .(OR​⊗IR​)V=VOand(OR†​⊗IR​)V=VO†.
  1. For every R‾×R‾\overline R\times\overline RR×R matrix XXX there is c∈Cc\in\mathbb Cc∈C with (VV†)(IR⊗X)(VV†)=c (VV†)(VV^\dagger)(I_R\otimes X)(VV^\dagger)=c\,(VV^\dagger)(VV†)(IR​⊗X)(VV†)=c(VV†).
  2. Let ϕ(s,r,b)=∣C∣−1/2V(r,b),s\phi(s,r,b)=|C|^{-1/2}V_{(r,b),s}ϕ(s,r,b)=∣C∣−1/2V(r,b),s​ (s∈Cs\in Cs∈C, r∈Rr\in Rr∈R, b∈R‾b\in\overline Rb∈R). Then ρSR‾(ϕ)=ρS(ϕ)⊗ρR‾(ϕ)\rho_{S\overline R}(\phi)=\rho_S(\phi)\otimes\rho_{\overline R}(\phi)ρSR​(ϕ)=ρS​(ϕ)⊗ρR​(ϕ) as matrices indexed by C×R‾C\times\overline RC×R, where ρSR‾((s,b),(s′,b′))=∑rϕ(s,r,b)ϕ(s′,r,b′)‾\rho_{S\overline R}((s,b),(s',b'))=\sum_r\phi(s,r,b)\overline{\phi(s',r,b')}ρSR​((s,b),(s′,b′))=∑r​ϕ(s,r,b)ϕ(s′,r,b′)​, ρS(s,s′)=∑r,bϕ(s,r,b)ϕ(s′,r,b)‾\rho_S(s,s')=\sum_{r,b}\phi(s,r,b)\overline{\phi(s',r,b)}ρS​(s,s′)=∑r,b​ϕ(s,r,b)ϕ(s′,r,b)​, ρR‾(b,b′)=∑s,rϕ(s,r,b)ϕ(s,r,b′)‾\rho_{\overline R}(b,b')=\sum_{s,r}\phi(s,r,b)\overline{\phi(s,r,b')}ρR​(b,b′)=∑s,r​ϕ(s,r,b)ϕ(s,r,b′)​, and ⊗\otimes⊗ is the Kronecker product.
  3. ∣C∣≤∣R∣|C|\le|R|∣C∣≤∣R∣, and, writing q=⌊∣R∣/∣C∣⌋q=\lfloor |R|/|C|\rfloorq=⌊∣R∣/∣C∣⌋ and m=∣R∣ mod ∣C∣m=|R|\bmod |C|m=∣R∣mod∣C∣: for every bijection e:R→(C×{0,..,q−1})⊔{0,..,m−1}e:R\to(C\times\{0,..,q-1\})\sqcup\{0,..,m-1\}e:R→(C×{0,..,q−1})⊔{0,..,m−1} there exist an R×RR\times RR×R matrix URU_RUR​ with UR†UR=IU_R^\dagger U_R=IUR†​UR​=I and URUR†=IU_RU_R^\dagger=IUR​UR†​=I, and a function χ:{0,..,q−1}×R‾→C\chi:\{0,..,q-1\}\times\overline R\to\mathbb Cχ:{0,..,q−1}×R→C with ∑x∣χ(x)∣2=1\sum_x|\chi(x)|^2=1∑x​∣χ(x)∣2=1, such that for all i∈Ci\in Ci∈C, r∈Rr\in Rr∈R, b∈R‾b\in\overline Rb∈R,
V(r,b),i=∑r′∈R(UR)r,r′ Fi(r′,b),Fi(r′,b)={χ(k,b)if e(r′)=(c,k) in the first summand and c=i,0if e(r′)=(c,k) with c≠i,0if e(r′) lies in the second summand.V_{(r,b),i}=\sum_{r'\in R}(U_R)_{r,r'}\,F_i(r',b),\qquad F_i(r',b)=\begin{cases}\chi(k,b)&\text{if }e(r')=(c,k)\text{ in the first summand and }c=i,\\0&\text{if }e(r')=(c,k)\text{ with }c\neq i,\\0&\text{if }e(r')\text{ lies in the second summand.}\end{cases}V(r,b),i​=r′∈R∑​(UR​)r,r′​Fi​(r′,b),Fi​(r′,b)=⎩⎨⎧​χ(k,b)00​if e(r′)=(c,k) in the first summand and c=i,if e(r′)=(c,k) with c=i,if e(r′) lies in the second summand.​

Edge remarks: because CCC is nonempty, ∣C∣≥1|C|\ge1∣C∣≥1, so the division and remainder are ordinary. If no bijection eee exists the second half of (4) would be vacuous, but the cardinalities always match (∣R∣=∣C∣q+m|R|=|C|q+m∣R∣=∣C∣q+m). The isometry hypothesis is the only assumption on VVV.

Human review
  • Endorsed by Shuze Chen · Oct 10, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Lucas · Oct 10, 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