AFTD Auto-Formalizing Theoretical Domains

not_is_bipartite_locally_encodable_zero

theoremverifiedformalization

For nonempty finite types m and n, the zero vector in EuclideanSpace ℂ (m × n) is not locally encodable.

Statement

theorem not_is_bipartite_locally_encodable_zero {m n : Type*} [Fintype m] [DecidableEq m]
    [Fintype n] [DecidableEq n] [Nonempty m] [Nonempty n] :
    ¬ is_bipartite_locally_encodable (0 : EuclideanSpace ℂ (m × n))
Formalizes
arXiv:2611.16921 Sec. II.1
Verified
14 Nov 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 1
  • is_bipartite_locally_encodableA bipartite quantum state ψ in EuclideanSpace ℂ (m × n) is locally encodable if there exists an orthonormal basis B of EuclideanSpace ℂ (m…

Read back from the Lean

For all types m and n that are finite, nonempty, and have decidable equality, the zero vector in EuclideanSpace ℂ (m × n) is not bipartite locally encodable (i.e., ¬ is_bipartite_locally_encodable 0).

Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.

The proof, explained

The proof proceeds by contradiction. If the zero vector were bipartite locally encodable, every element of an orthonormal basis for the space would be the image of zero under some local unitary operator. But because unitary operators act linearly, they map zero to zero, which forces every basis vector to be zero. This contradicts the fact that elements of an orthonormal basis must have norm 1.

  1. Assuming 0 is bipartite locally encodable, there exists an orthonormal basis B whose elements are all images of 0 under local unitary transformations, and we fix one such basis vector B(j₀).

    We proceed by contradiction (`rintro`). By the definition of bipartite local encodability, there must exist an orthonormal basis $B$ such that every basis element is the image of $0$ under some local unitary operator $u_A \otimes u_B$. Since $m$ and $n$ are nonempty, the product $m \times n$ is nonempty (`inhabit`), so we can choose an arbitrary index $j_0 \in m \times n$ (`default`). Specializing the hypothesis to $j_0$ provides local unitary matrices $u_A$ and $u_B$ satisfying $B(j_0) = (u_A \otimes u_B)(0)$ (`obtain`).

    lines 5–7

      rintro ⟨B, hB⟩
      inhabit (m × n)
      obtain ⟨uA, uB, h⟩ := hB default
  2. The basis vector B(j₀) is equal to 0.

    Because the transformation induced by $u_A \otimes u_B$ is linear, it preserves zero (`map_zero`). Rewriting $B(j_0)$ with this identity shows that $B(j_0) = 0$.

    lines 8–9

    uses map_zero

      have hz : B default = 0 := by
        rw [h, map_zero]
  3. This contradicts the requirement that basis vectors have norm 1.

    Because $B$ is an orthonormal basis, every basis vector has unit norm, so $\|B(j_0)\| = 1$. Substituting $B(j_0) = 0$ and evaluating $\|0\| = 0$ (`norm_zero`) yields $0 = 1$, an arithmetic contradiction closed by `norm_num`.

    lines 10–12

    uses norm_zero

      have h1 : ‖B default‖ = 1 := B.orthonormal.1 default
      rw [hz, norm_zero] at h1
      norm_num at h1
map_zero
A zero-preserving map (such as a linear map) sends 0 to 0.
norm_zero
The norm of the zero vector in any seminormed group is 0.

Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.

Lean source view module on GitHub

/-- The zero vector is never locally encodable in a nonempty bipartite space. -/theorem not_is_bipartite_locally_encodable_zero {m n : Type*} [Fintype m] [DecidableEq m]    [Fintype n] [DecidableEq n] [Nonempty m] [Nonempty n] :    ¬ is_bipartite_locally_encodable (0 : EuclideanSpace ℂ (m × n)) := by  rintro ⟨B, hB⟩  inhabit (m × n)  obtain ⟨uA, uB, h⟩ := hB default  have hz : B default = 0 := by    rw [h, map_zero]  have h1 : ‖B default‖ = 1 := B.orthonormal.1 default  rw [hz, norm_zero] at h1  norm_num at h1
See it in the graphDiscuss this resultChallenge it