is_locally_encodable
For a finite set of parties with finite dimension types d p, a state ψ in EuclideanSpace ℂ ((p : Parties) → d p) is locally encodable if there exists an orthonormal basis B of EuclideanSpace ℂ ((p : Parties) → d p) indexed by (p : Parties) → d p such that every basis element B j equals Matrix.toEuclideanLin V ψ for some unitary matrix V that decomposes as a tensor product of local unitaries u p.
Statement
def is_locally_encodable {Parties : Type*} [Fintype Parties] [DecidableEq Parties] {d : Parties → Type*} [∀ p, Fintype (d p)] [∀ p, DecidableEq (d p)] (ψ : EuclideanSpace ℂ ((p : Parties) → d p)) : Prop
- Defined in
- arXiv:2611.16921 Sec II.1
- Verified
- 13 Nov 2026
- Axioms
- none listed
Read back from the Lean
Given a finite type Parties with decidable equality, a family of finite types d : Parties → Type* with decidable equality for each p : Parties, and a state ψ : EuclideanSpace ℂ ((p : Parties) → d p), is_locally_encodable ψ asserts that:
There exists an orthonormal basis B of EuclideanSpace ℂ ((p : Parties) → d p) indexed by (p : Parties) → d p such that for every index j : (p : Parties) → d p, there exists a unitary matrix V of size ((p : Parties) → d p) × ((p : Parties) → d p) over ℂ such that:
1. There exists a family of unitary matrices u : ∀ p, Matrix.unitaryGroup (d p) ℂ such that V is the tensor product of the local unitaries u p (specifically, the matrix entry V x y equals $\prod_{p} (u(p))_{x(p), y(p)}$ for all $x, y : (p : Parties) \to d\ p$); and
2. The basis vector B j equals the image of ψ under the linear map represented by V (i.e., B j = Matrix.toEuclideanLin V.1 ψ).
Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.
Lean source view module on GitHub
/-- A state in the multipartite Euclidean space is locally encodable if its local unitary orbit contains an orthonormal basis. -/ def is_locally_encodable {Parties : Type*} [Fintype Parties] [DecidableEq Parties] {d : Parties → Type*} [∀ p, Fintype (d p)] [∀ p, DecidableEq (d p)] (ψ : EuclideanSpace ℂ ((p : Parties) → d p)) : Prop := ∃ B : OrthonormalBasis ((p : Parties) → d p) ℂ (EuclideanSpace ℂ ((p : Parties) → d p)), ∀ j, ∃ V : Matrix.unitaryGroup ((p : Parties) → d p) ℂ, (∃ u : ∀ p, Matrix.unitaryGroup (d p) ℂ, V.1 = fun x y => ∏ p, (u p).1 (x p) (y p)) ∧ B j = Matrix.toEuclideanLin V.1 ψ