Section 3.2, Eqs. (3.19) ⟺ (3.21) — for every iff every logical operator has a representation on
ProvedADH2015.correctsErasure_iff_allRepLet be finite-dimensional ( the erased part, the retained part) and let be a code subspace. An operator acts within if and both map into itself. It has a representation on if some satisfies and for all (and analogously on , with ). Then the following are equivalent:
- (Eq. (3.19)) for every operator on there is with for all code states, i.e. the projection of onto the code subspace is proportional to the identity;
- (Eq. (3.21)) every operator acting within has a representation on .
This is the operator form of the erasure-correction condition: every logical operator can be implemented on the surviving subsystem alone exactly when operators on the erased subsystem cannot distinguish code states. It is the special case of the basic theorem of operator-algebra quantum error correction in which the logical algebra is the full operator algebra of the code subspace.
Formalization Note The source cites (3.19) as equivalent to correctability of the erasure of (the existence of a recovery unitary, Eqs. (3.14)–(3.18), from refs. [26], [28]); that equivalence is not formalized here. Condition (3.19), in an orthonormal basis, is stated basis-free as for one constant and all code vectors.
import Mathlib import Definitions.Def_ADH2015_defs open Matrix open scoped Kronecker
namespace ADH2015
theorem correctsErasure_iff_allRep {e ē : Type*} [Fintype e] [Fintype ē] [DecidableEq e]
[DecidableEq ē] (C : Submodule ℂ (e × ē → ℂ)) :
CorrectsErasureOfE C ↔ ∀ O : Matrix (e × ē) (e × ē) ℂ, ActsWithin C O → HasRepOnBar C O := by sorry
end ADH2015Read-back
What the Lean code literally says, in plain math · claude-opus-5-5
Read-back of ADH2015.correctsErasure_iff_allRep.
Setting and binders. Let and be arbitrary finite types, each with decidable equality. Nothing else is assumed: either may be empty, and there is no other size condition. The ambient space is
the complex vectors indexed by pairs with and . The first factor is the one called , and the second factor is the one called . Let be an arbitrary complex linear subspace (the "code subspace"). It may be , all of , or anything in between. Operators are complex square matrices indexed by . is the conjugate transpose. For the pairing is , the standard Hermitian inner product, conjugate-linear in . For matrices (on ) and (on ), is the Kronecker product with entries . So acts only on the factor, and acts only on the factor.
The statement. For every such :
The connective is a genuine "if and only if". Inside the right-hand side, "acts within" implies "has a representation" for every complex matrix .
Definitions, unfolded.
- : for every complex matrix on the first factor (indexed by ), there exists a complex number such that for all and all ,
The quantifier order is . So may depend on but must be the same for all pairs . Here ranges over all complex matrices on , not only Hermitian or unitary ones. Equivalently, if is the orthogonal projection onto , then .
-
: for every , both and . This means is invariant under both and .
-
: there exists a complex matrix on the second factor (indexed by ) such that for every ,
Note that . The order is : one works for all of . Both equalities are equalities of vectors in the full space , required only on vectors of . is any complex matrix, with no unitarity or Hermiticity condition.
So the theorem asserts the following two statements are equivalent:
- Every operator acting only on is compressed to a scalar multiple of the identity on .
- Every operator on that leaves invariant together with agrees on , together with its adjoint, with some operator of the form and that operator's adjoint.
Edge cases.
- Zero subspace : every condition quantified over holds trivially. So the left side is true (any works), every acts within , and every has a representation (any works). Both sides are true.
- Empty types: if or is empty, then is the zero space. So and the zero-subspace case applies.
- One-dimensional code with : the left side always holds, taking . For such the theorem therefore states something unconditional. Take any such that is an eigenvector of both and , say and . Then there is a with
- Full space (nonzero): every acts within . The right side then requires every operator on to equal some . The left side requires every to be a scalar multiple of the identity.
Logical strength. The statement is a biconditional, universally quantified over all finite decidable index types and and all complex subspaces . It has no hypotheses beyond finiteness and decidable equality of the index types. No normalization, orthonormal basis, Hermiticity, or unitarity is assumed anywhere.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.