slug: arunachalam-johnston-russo-2015-generalized-choi-absolutely-ppt bibkey: arunachalam2015absolute doi: null url: https://arxiv.org/abs/1405.5853v3 triage: theorem motivation_gids:

• D5/S3/Quantum/Entanglement/AbsolutePPT/GeneralizedChoiNonDetection.result

No generalized Choi map detects absolutely PPT entanglement

Problem

S. Arunachalam, N. Johnston and V. Russo, Is absolute separability determined by the partial transpose?, Quantum Inf. Comput. 15(7–8), 694–720 (2015); arXiv:1405.5853v3. Section 7 (p. 21):

Other open problems in this work include proving that all of the generalized Choi maps ${\Phi }_{b,c}$ in Section 5.3 are incapable of detecting absolutely PPT entanglement (including those in the light gray region of Figure 3) …

Here ${\Phi }_{b,c}(X)=\frac{1}{2}\left(\begin{array}{ccc}a{x}_{11}+b{x}_{22}+c{x}_{33}& -x_{12}& -x_{13}\\ -x_{21}& c{x}_{11}+a{x}_{22}+b{x}_{33}& -x_{23}\\ -x_{31}& -x_{32}& b{x}_{11}+c{x}_{22}+a{x}_{33}\end{array}\right)$ with $a=2-b-c$, positive but not completely positive exactly when $(b,c)\ne (0,0)$, $b,c\ge 0$ and $b+c\le 1$ or $bc\ge (b+c-1{)}^2$ (Section 5.3, p. 14). The verbatim texts are in the literature note. Issue #14738 preregisters the reading and the proof plan.

Motivation

The absolute separability problem asks for a spectral characterization of states that stay separable under every global unitary. The source asks whether absolutely PPT states, those whose every unitary conjugate has positive partial transpose, are absolutely separable, and tests this against the positive maps most likely to detect PPT entanglement. The generalized Choi maps contain the Choi map, the reduction map and indecomposable exposed maps.

Gap

The source’s Theorem 3 proves non-detection on a dark subregion of the parameter set from eigenvalue bounds on $(id\otimes {\Phi }_{b,c}^{†})(|v⟩⟨v|)$; for the light region it has only numerical evidence. The literature check in #14738 found no proof for the whole family; not-found-in-searched-scope.

Route

For a unit vector $v$, $⟨v,(id\otimes {\Phi }_{b,c})(\rho )v⟩=\mathrm{tr}(\rho W_v)$ with $W_v=(id\otimes {\Phi }_{b,c}^{†})(|v⟩⟨v|)$ and $\mathrm{tr}{W}_v=1$.

1. Witness purity. With $v={\sum }_j{x}_j\otimes e_j$, $p_j=\Vert x_j{\Vert }^2$, $S={\sum }_j{p}_j^2\ge 1/3$, $\sigma$ the Gram matrix of the $x_j$, $r=\mathrm{tr}{\sigma }^2\le 1$ and $q=a^2+b^2+c^2$, $\frac{1}{2}-\mathrm{tr}{W}_v^2=\frac{(4-q)(1-r)}{8}+\frac{(2-q)(3S-1)}{8}$, so $\mathrm{tr}{W}_v^2\le \frac{1}{2}$ when $q\le 2$. On the curved part of the region $q-2=2[(b+c-1{)}^2-bc]\le 0$.
2. State purity. The repository’s frozen QutritN3PurityBound.purity_bound gives $\mathrm{tr}{\rho }^2\le 17/121$ for every absolutely PPT state on ${\mathbb{C}}^3\otimes {\mathbb{C}}^3$.
3. Cauchy–Schwarz. $\operatorname{tr}(\rho W_v)=\tfrac19+\operatorname{tr}((\rho-\tfrac I9)(W_v-\tfrac I9)) \ge\tfrac19-\sqrt{\tfrac7{18}\cdot\tfrac{32}{1089}}>0$,since$(\tfrac19)^2-\tfrac7{18}\cdot\tfrac{32}{1089}=\tfrac1{1089}$.
4. Triangle. For $b+c\le 1$, ${\Phi }_{b,c}=(1-b-c){\Phi }_{0,0}+b{\Phi }_{1,0}+c{\Phi }_{0,1}$, where ${\Phi }_{1,0},{\Phi }_{0,1}$ have $q=2$ and ${\Phi }_{0,0}(X)=\frac{1}{2}{\sum }_{D}DXD$ with $D\in \{\mathrm{diag}(1,-1,0),\mathrm{diag}(0,1,-1),\mathrm{diag}(-1,0,1)\}$ is completely positive.

Falsifier

An error in the witness-purity identity, in the hypotheses of the purity bound, or in the reading of ${\Phi }_{b,c}$ or of absolute PPT. Positive controls in #14738: ${\Phi }_{1,1}$, ${\Phi }_{1,0}$ and ${\Phi }_{0.5,1.2}$ detect the maximally entangled state, so the claim is not vacuous; the bound $\mathrm{tr}{W}_v^2\le \frac{1}{2}$ is attained.

Evidence

D5/S3/Quantum/Entanglement/AbsolutePPT/GeneralizedChoiNonDetection.lean defines choiGen (${\Phi }_{b,c}$ entry by entry), idTensor (the map on the second tensor factor, blockwise) and claim, and proves the public witness_purity_bound and result : claim from the frozen purity_bound and APPT of D5/S3/Quantum/Entanglement/AbsolutePPT. The axiom closure of result is exactly propext, Classical.choice and Quot.sound; there is no sorry or native_decide. The module statement is sha256:a6a9d31ad0b16bdf4565d2f33d08939281a73a86eb48d0a01be8e3900feab930; result is sha256:987b9c4a12f2a5f7df6278f4744be224d7385f30daf96dc78f658701fa097a25, claim is sha256:ec3a7c9056ef82a79d549dbc7543bf3f3b795777261aca60d289a647a0bdf9f3 and witness_purity_bound is sha256:ea6338444a379797b6331ad8ffa059ab8a8bbcb96d7b2461aee7d4c37368cde2. The Freeze event is sha256:0c500069bbf6c8fd9d667879a968f2163d59a6619ef81c80d80d0d432342a95e.

Triage

Tier 1: an open problem stated in a 2015 paper, preregistered in #14738 before any Lean. theorem; resolution proved. Utility none; there is no digestion atom.

Every theorem of the module is bind-only by CLAUDE.md §3.2: unfolding of the new definitions with ring normalization, instantiation of frozen and Mathlib lemmas, and linarith/nlinarith. Admission basis open-problem-resolution; the new content is the settlement.

What the proof shows

• Proved by result: for every $(b,c)$ for which ${\Phi }_{b,c}$ is positive and not completely positive, $(i{d}_3\otimes {\Phi }_{b,c})(\rho )\ge 0$ for every absolutely PPT $\rho \in M_3\otimes M_3$. No generalized Choi map, in particular neither the Choi map nor the reduction map nor the indecomposable exposed maps of the family, can certify that an absolutely PPT qutrit pair is entangled.
• Mechanism (proved within the derivation): the witness $W_v$ of every map with $a^2+b^2+c^2\le 2$ has purity at most $1/2$, while an absolutely PPT state has purity at most $17/121$; two trace-one operators whose purities are this small cannot have a negative Hilbert–Schmidt inner product.
• Derived, not formalized as a separate theorem: the argument needs only $(\mathrm{tr}{\rho }^2-\frac{1}{9})(\frac{1}{2}-\frac{1}{9})<\frac{1}{81}$, i.e. $\mathrm{tr}{\rho }^2<\frac{1}{7}$, and $17/121<1/7$. Hence on ${\mathbb{C}}^3\otimes {\mathbb{C}}^3$ no positive map $\Phi$ such that every witness $(id\otimes {\Phi }^{†})(|v⟩⟨v|)$ of a unit vector $v$ has trace one and purity at most $1/2$ detects the entanglement of any state of purity below $1/7$, absolutely PPT or not.
• Open: whether absolutely PPT states on ${\mathbb{C}}^3\otimes {\mathbb{C}}^3$ are absolutely separable (the source’s main question), and the analogous statement for positive maps outside the generalized Choi family.

ASSUMED-UNVERIFIED

The literature check in #14738 is bounded and does not establish worldwide novelty or priority.