trureturingGitHub MATHEMATICAL ATLAS

RESEARCH / Focused target

The correlation-minor bound fails without filter normal form

Literature status: not recheckedRepository record: refuted

RELEASED FOUNDATIONS / PROPOSED CONNECTION

Research connections

0 released anchors
All source anchors (1)

Problem

The question

B. Y. Peled, A. Te'eni, A. Carmi and E. Cohen, Correlation minor norms, entanglement detection and discord, arXiv:2005.12079v2, Scientific Reports 11 (2021) 2849, states in Theorem 3 that a separable state in filter normal form satisfies the $p=\infty$ correlation-minor bound $\mathcal{M}_{h,p=\infty} \leq \frac{1}{\sqrt{Dd}}\left[\frac{D-1}{D(h-1)}\frac{d-1}{d(h-1)}\right]^{(h-1)/2}$. The source then asks whether the filter-normal-form hypothesis can be removed. In the source TeX file CMN_16_12_20_arXiv.tex, line 281, it says:

As in Theorem \ref{thm_bound_p_1}, we conjecture this theorem still holds without the assumption that $\rho$ is in FNF.

The compiled Lean claim is the universal assertion over natural numbers $da$, $db$, and $h$ that, when $2 \leq da$, $2 \leq db$, $\max(da,db)\min(da,db) \leq h^2$, and $h \leq \min(da,db)^2$, every separable matrix $\rho$ on the product index space and every pair of IsHermBasis families satisfies $\operatorname{cmnInf}(h,\operatorname{corr}(\rho,a,b))\leq\operatorname{bound}(da,db,h)$. Here IsHermBasis, corr, compound, cmnInf, and bound are the definitions in the module, so the formal assertion has no filter-normal-form assumption.

Motivation

Our foothold

The paper presents correlation-minor norms as basis-invariant quantities for entanglement detection: exceeding a separable-state upper bound is intended to certify entanglement. Theorem 3 supplies such a bound for filter-normal-form states, and the conjecture would extend that certification threshold to all separable states in unequal local dimensions. Determining whether that extension is valid matters because a bound that fails for separable states cannot be used as a universal separability threshold.

Gap

Missing bridges

The bounded search recorded in issue #25435 found no earlier repository record and no settlement in the searched literature. The latest arXiv source version retains the conjecture. The recorded citation search found five citing works by OpenAlex retrieval through Semantic Scholar and OpenCitations, including Lenny et al., arXiv:2302.00650, whose Theorem 1 retains the filter-normal-form condition; none of the searched works settles the conjecture. The recorded Zenodo search returned no correlation-minor-norm record.

not-found-in-searched-scope.

Route

Proposed approach

  1. The private lemma projectors_density (j : Fin 3) proves IsDensity (projectors j) for each j : Fin 3. The private lemma rho_separable then proves IsSeparable rho by exhibiting four mixture weights $333/1000$, $333/1000$, $333/1000$, and $1/1000$, with the three projectors paired with the first three diagonal four-dimensional matrix units and the maximally mixed local matrices paired in the fourth term.
  2. The private lemma alice_basis proves IsHermBasis alice for the four matrices $(I,Z,X,Y)/\sqrt{2}$. The private lemmas bob_pairing_0 through bob_pairing_15, each for every j : Fin 16, establish the trace-product pairing for the sixteen Bob matrices, and bob_basis combines those cases with Hermiticity to prove IsHermBasis bob.
  3. For every j : Fin 16, the private lemmas correlation_row_0, correlation_row_1, correlation_row_2, and correlation_row_3 identify the corresponding entry of corr rho alice bob with coefficients. The private lemma correlation_eq combines these four row statements to prove equality of the full correlation matrix and coefficients.
  4. The private lemma selectedCols_injective proves that the map selectedCols : Fin 4 \to \{s : Finset (Fin 16) // s.card = 3\} is injective. For every k : Fin 4, minor_value computes the determinant of the selected row selectedRow and selected column selectedCols k as the corresponding one of four rational values divided by $(\sqrt{2})^3$. The private lemma minor_energy sums the four squared minors and obtains $\frac{559341805929555807}{500000000000000000000}$.
  5. The private lemma row_lower applies to any types m, n, and t with instances [Fintype m], [Fintype n], [Fintype t], and [DecidableEq n], any real matrix c : Matrix m n \mathbb{R}, any row r : m, any map s : t \to n, and any proof hs : Function.Injective s; it proves $\sum_k c,r,(s,k)^2 \leq \lVert c\rVert^2$. Using this lemma with c = compound 3 coefficients, r = selectedRow, and s = selectedCols, witness_lower proves the squared lower bound for cmnInf 3 (corr rho alice bob).
  6. The private lemma bound_value proves bound 2 4 3 = 3/(64\sqrt{2}), and numerical_gap proves $\left(\frac{3}{64\sqrt{2}}\right)^2 < \frac{559341805929555807}{500000000000000000000}$. Finally, result assumes claim, instantiates it at $da=2$, $db=4$, and $h=3$ with the four dimension hypotheses, rho_separable, alice_basis, and bob_basis. Nonnegativity permits squaring the resulting norm inequality, and witness_lower together with numerical_gap gives the contradiction.

Falsifier

What would falsify this route

The formal definitions track the source construction at the relevant level. IsHermBasis requires Hermitian matrices, exactly the squared local dimension many basis elements, and trace-product pairing equal to one on equal indices and zero otherwise. corr takes the real part of the trace against the Kronecker product, compound uses determinants of submatrices with increasing finite-set enumerations, and cmnInf is the norm of that compound matrix. The definition bound uses the real casts of the maximum and minimum local dimensions and the real exponent $((h:\mathbb{R})-1)/2$. The imported IsSeparable is a finite convex combination of product matrices whose local factors are positive semidefinite and have trace one.

The counterexample in the module has $da=2$, $db=4$, and $h=3$, satisfying the formal conditions $8 \leq 9$ and $3 \leq 4$. Its state, Alice basis, Bob basis, correlation coefficients, four selected minors, and numerical comparison are all the objects used by the compiled proof. The result is therefore a refutation of the universal no-FNF claim, rather than a statement about the FNF-restricted Theorem 3.

Evidence

Evidence to collect

The canonical source is D5/S3/Quantum/Entanglement/Witness/CorrelationMinorNormFNFRefutation.lean.

Frozen state: Golden/Frozen/state/D5/S3/Quantum/Entanglement/Witness/CorrelationMinorNormFNFRefutation.lean.json.

Declaration statements:

  • result: sha256:3fe7072d6440018c4e41f6f28de743362cb1f572f503cb2022ce55396ddf40f7;

  • claim: sha256:a2e41b7a623e022e5349330e9f6164c3bae78a071e9b7954cd43841d9057d546.

  • Proof. It proves result : ¬ claim.

  • Axioms. It uses only the standard axioms propext, Classical.choice, and Quot.sound; no sorry, native_decide, or new axiom.

Triage

Scope assessment

Tier 1 conjecture of a 2020 arXiv paper published in Scientific Reports (2021), preregistered in issue #25435 before any Lean. theorem; resolution refuted.

The module is admitted under the open-problem-resolution basis of CLAUDE.md §3.2; the refutation is a certified instance with basis refutes. There is no digestion atom.

What the settlement shows

Proved by result: the formal universal claim asserting the correlation-minor operator-norm bound for all natural da, db, and h satisfying its four dimension and order hypotheses, every separable state at those dimensions, and every pair of IsHermBasis families is false.

Computed, not formalized: None.

Argued, not formalized.

  • Mechanism. Removing FNF allows local expectation components that FNF sets to zero to remain in the correlation matrix. In the displayed construction, the diagonal Bob flags and the three Alice projectors produce nonzero selected minors whose combined norm exceeds the proposed right-hand side; the missing FNF restriction is the structural condition that the counterexample exploits. Whether this is the only failure mechanism for no-FNF bounds is open.
  • Nearby statements. The refutation does not refute the source's Theorem 3 with its FNF hypothesis. It gives no settlement of the source's separate $p=1$ conjecture, its quantum-design saturation statements, or its discord measure.
  • Effect on dependent claims. Any use of the conjectured $p=\infty$ expression as a separability bound for arbitrary separable states must retain FNF or supply a different bound. The source's no-FNF conjectural extension is unavailable as a universal entanglement-detection threshold; the other source statements named above require their own hypotheses and are not settled by this result.

ASSUMED-UNVERIFIED

Unverified assumptions

The literature search is bounded and does not establish exhaustive worldwide novelty or priority. The report that the journal version retains the conjecture was recorded in issue #25435 and was not independently reproduced here. The mechanism explanation, the precise FNF status of the displayed state as a separate calculation, and all consequences beyond the compiled proposition remain unverified. The source's statements outside the formal claim and result declarations are not formalized.

Dossier & anchor history

Loading release observations...