Theorem 6.1 — Majorana criterion: surjective -homomorphism iff , , ; then
OpenRelDecoding.Supplement.theorem_6_1operator-algebrasp2o-gran-per-chapterp2o-plan-paperp2o-v1quantum-informationrelative-decoding
For let be the Majorana operators of the full Fock space of modes and (increasing order) the Hermitian Majorana monomials. Let , , , with and nonzero even-weight labels , ; let , , , , , over , and (resp. ), for .
- The assignment extends to a surjective unital -homomorphism if and only if
- For the matched Hamiltonians , with and , every such map satisfies
for every polynomial .
This is the fermionic analogue of the Pauli criterion: it decides when polynomial filters of a parity-preserving source Hamiltonian can be transported to a target Hamiltonian with the same coefficient magnitudes.
Differences from the source:
- encoding: Majoranas γ_{L,j} of the full Fock space of n_L modes, from any f_{L,j} satisfying (136) → the Jordan–Wigner matrices of eq. (167), the published FermiCap.Gaussian.majorana (0-based index p ↦ γ_{p+1}), on Matrix (Fin n → Fin 2) (Fin n → Fin 2) ℂ; labels Fin (2n) → ZMod 2; reason: the source gives these matrices itself (eq. (167)); every CAR family on ℂ^{2^n} is unitarily equivalent to them, and the statement is invariant under a unitary change of basis on either side
- encoding: unital ∗-homomorphism π : 𝒜 → ℬ between A = C∗(I_A, a) and B = C∗(I_B, b), surjective → π : genAlg a →⋆ₐ[ℂ] genAlg b (generated ∗-subalgebras of the matrix algebras) with π (gen a i) = gen b i; reason: in finite dimension the generated ∗-subalgebra is the generated C∗-algebra (App. E, p.85)
- encoding: ζ_A(c) = ζ_B(c) for c ∈ K_A, where W_L(c) = ζ_L(c) I for c ∈ K_L → ∀ c ∈ K_A, ∀ ζ : ℂ, W_A(c) = ζ • 1 → W_B(c) = ζ • 1; reason: for c ∈ K_A the word W_A(c) is scalar, so ζ_A(c) is the unique ζ with W_A(c) = ζI (the matrix algebras are nonzero); the implication then says W_B(c) = ζ_A(c)I, i.e. ζ_B(c) = ζ_A(c)
- encoding: π(H_A) = H_B, π(p(H_A)) = p(H_B) for every polynomial p → H_A as the element E_c • 1 + Σ λᵢ • gen a i of A; equalities of matrices after coercion; p : Polynomial ℂ evaluated by Polynomial.aeval; reason: complex polynomials are the source's polynomials in operators; same identities
Preamble
import Mathlib import Definitions.Def_RelDecoding_markedSpace import Definitions.Def_FermiCap_Gaussian_Majorana import Definitions.Def_RelDecoding_Supplement_labelKer import Definitions.Def_RelDecoding_Supplement_majoranaGen set_option autoImplicit false open scoped Matrix
Formal statement
namespace RelDecoding.Supplement
open FermiCap.Gaussian RelDecoding
/-- **Theorem 6.1 (Homomorphisms between Majorana-generated algebras).** For signed Majorana
generators `aᵢ = ηᵢᴬ Γ_A(wᵢ)`, `bᵢ = ηᵢᴮ Γ_B(vᵢ)` with nonzero even-weight labels `wᵢ ∈ F₂^{2n_A}`,
`vᵢ ∈ F₂^{2n_B}`, label matrices `A = [w₁ ⋯ w_s]`, `B = [v₁ ⋯ v_s]`, kernels `K_A`, `K_B` over `F₂`
and ordered products `W_L(c) = a₁^{c₁}⋯a_s^{c_s}` (`= ζ_L(c) I_L` for `c ∈ K_L`), the assignment
`aᵢ ↦ bᵢ` extends to a surjective unital ∗-homomorphism `π : A → B` if and only if
`AᵀA = BᵀB`, `K_A ⊆ K_B`, `ζ_A(c) = ζ_B(c)` (`c ∈ K_A`) (155). For the matched Hamiltonians
`H_A = E_c I_A + Σᵢ λᵢ aᵢ`, `H_B = E_c I_B + Σᵢ λᵢ bᵢ`, `λᵢ > 0`, `E_c ∈ ℝ` (151), such a map satisfies
`π(H_A) = H_B` and `π(p(H_A)) = p(H_B)` for every polynomial `p` (156).
Shang, Stilck França, *From Simple Sources to Quantum Advantage: Homomorphic Polynomial
Transduction via Relative Decoding*, arXiv:2609.35101 (held version, sha256 d2be7859…),
Theorem 6.1, p.35 (setting eqs. (136), (140)–(141), (151)–(154), pp.33–35; proof Proposition E.3,
pp.88–89).
**Formalization Note.** `Γ_L` is `majoranaMonomial` on the Jordan–Wigner Majoranas (167) (the
published `FermiCap.Gaussian.majorana`); `A = genAlg a`, `B = genAlg b`, and `π : A →⋆ₐ[ℂ] B` with
`π (gen a i) = gen b i` and `Function.Surjective π`. The phase condition is encoded as: for
`c ∈ K_A`, `W_A(c) = ζI_A` implies `W_B(c) = ζI_B`. Polynomials are complex, `p ∈ ℂ[X]`, evaluated
by `Polynomial.aeval` in `A` and in `Matrix`. -/
theorem theorem_6_1 {nA nB s : ℕ} (wA : Fin s → Fin (2 * nA) → ZMod 2) (ηA : Fin s → ℤˣ)
(wB : Fin s → Fin (2 * nB) → ZMod 2) (ηB : Fin s → ℤˣ)
(hwA : ∀ i, wA i ≠ 0 ∧ Even (hammingNorm (wA i)))
(hwB : ∀ i, wB i ≠ 0 ∧ Even (hammingNorm (wB i))) :
((∃ π : genAlg (majoranaGen ηA wA) →⋆ₐ[ℂ] genAlg (majoranaGen ηB wB),
(∀ i, π (gen (majoranaGen ηA wA) i) = gen (majoranaGen ηB wB) i) ∧
Function.Surjective π) ↔
((labelMatrix wA)ᵀ * labelMatrix wA = (labelMatrix wB)ᵀ * labelMatrix wB ∧
labelKer wA ≤ labelKer wB ∧
∀ c ∈ labelKer wA, ∀ ζ : ℂ, orderedProd (majoranaGen ηA wA) c = ζ • 1 →
orderedProd (majoranaGen ηB wB) c = ζ • 1)) ∧
∀ π : genAlg (majoranaGen ηA wA) →⋆ₐ[ℂ] genAlg (majoranaGen ηB wB),
(∀ i, π (gen (majoranaGen ηA wA) i) = gen (majoranaGen ηB wB) i) →
Function.Surjective π →
∀ (Ec : ℝ) (lam : Fin s → ℝ), (∀ i, 0 < lam i) →
((π ((Ec : ℂ) • (1 : genAlg (majoranaGen ηA wA)) +
∑ i, ((lam i : ℝ) : ℂ) • gen (majoranaGen ηA wA) i) : Op nB) =
(Ec : ℂ) • (1 : Op nB) + ∑ i, ((lam i : ℝ) : ℂ) • majoranaGen ηB wB i) ∧
∀ p : Polynomial ℂ,
((π (Polynomial.aeval ((Ec : ℂ) • (1 : genAlg (majoranaGen ηA wA)) +
∑ i, ((lam i : ℝ) : ℂ) • gen (majoranaGen ηA wA) i) p) : Op nB) =
Polynomial.aeval ((Ec : ℂ) • (1 : Op nB) +
∑ i, ((lam i : ℝ) : ℂ) • majoranaGen ηB wB i) p) := by sorry
end RelDecoding.Supplement
Source
Shang, Stilck França, From Simple Sources to Quantum Advantage: Homomorphic Polynomial Transduction via Relative Decoding, arXiv:2609.35101 (held version, sha256 d2be7859…), Theorem 6.1, p.35 (pdf_page 35); setting eqs. (136), (140)–(141), (151)–(154), pp.33–35; proof Proposition E.3, pp.88–89