Status. Partial results only. The note studies, without answering it, the question of Yang, Leung and Tang (Linear Algebra Appl. 503 (2016), Question 3.9): is there an indecomposable 2-positive linear map M_3 → M_4? Dually: can a PPT state of a qutrit and a ququart (C^3 ⊗ C^4) have Schmidt number three? Proved. (1) A 2-block-positive operator on C^3 ⊗ C^4 that vanishes on one nonzero product vector is decomposable, so a witness for Schmidt number three among PPT states would have to be strictly positive on all product vectors. (2) For min(M,N) ≥ 2, the range of every PPT state on C^M ⊗ C^N is spanned by its vectors of Schmidt rank at most min(M,N) − 1; hence no range-only test that is valid for all positive operators can show full Schmidt number for a PPT state. (3) With a lemma of Chen, Yang and Tang: every non-extreme PPT state of birank (6,6) on C^3 ⊗ C^4, and every positive operator on C^3 ⊗ C^N of operator Schmidt rank at most seven, has a separable compression to a two-dimensional local subspace, hence Schmidt number at most two. (4) The maps proposed by C. A. Winda (J. Linear Topol. Algebra 14 (2025) 129–136) as indecomposable 2-positive maps M_3 → M_4 are not 2-positive. Open. A birank-(6,6) counterexample would have to be an extreme edge state with product-free kernels and operator Schmidt rank eight or nine. Numerically, every state tested (42 states covering all ten relevant birank classes) has a separable two-dimensional compression; the note describes where the proof attempts for birank (6,6) stopped. Files. ppt-3x4.pdf is the note. ppt-3x4-supplement.zip contains the statements, proofs and blind reviews of the results as written, the exact computation for the Winda maps, the numerical data and scripts (including exact rational birank-(6,6) edge states with separable compressions), the record of failed proof routes, and the unreviewed working files. "python check.py" (NumPy and SymPy) verifies the hashes and repeats the exact and numerical checks (a few seconds on the laptop used to assemble the archive). The written proofs have not yet been verified line by line by a human reviewer. AI tools. The author chose the question and the statements to attack, set the working rules (blind review of every result by the other model family; fixed stopping rules), had the published Winda maps checked and supplied the paper for that, decided when to stop, and is responsible for the content. The research was carried out on 4–5 November 2026 by an AI research system built on Claude Opus 5.5 (Anthropic) and GPT-6 Astra (OpenAI); the text was revised on 9 November 2026. Each proof was written by one of the two models and accepted by a blind review from the other; the Winda computation was done by both independently; Claude Opus 5.5 drafted the manuscript and GPT-6 Astra reviewed it blind. Section IX of the note has the details.
Status. Partial results only. The note studies, without answering it, the question of Yang, Leung and Tang (Linear Algebra Appl. 503 (2016), Question 3.9): is there an indecomposable 2-positive linear map M_3 → M_4? Dually: can a PPT state of a qutrit and a ququart (C^3 ⊗ C^4) have Schmidt number three? Proved. (1) A 2-block-positive operator on C^3 ⊗ C^4 that vanishes on one nonzero product vector is decomposable, so a witness for Schmidt number three among PPT states would have to be strictly positive on all product vectors. (2) For min(M,N) ≥ 2, the range of every PPT state on C^M ⊗ C^N is spanned by its vectors of Schmidt rank at most min(M,N) − 1; hence no range-only test that is valid for all positive operators can show full Schmidt number for a PPT state. (3) With a lemma of Chen, Yang and Tang: every non-extreme PPT state of birank (6,6) on C^3 ⊗ C^4, and every positive operator on C^3 ⊗ C^N of operator Schmidt rank at most seven, has a separable compression to a two-dimensional local subspace, hence Schmidt number at most two. (4) The maps proposed by C. A. Winda (J. Linear Topol. Algebra 14 (2025) 129–136) as indecomposable 2-positive maps M_3 → M_4 are not 2-positive. Open. A birank-(6,6) counterexample would have to be an extreme edge state with product-free kernels and operator Schmidt rank eight or nine. Numerically, every state tested (42 states covering all ten relevant birank classes) has a separable two-dimensional compression; the note describes where the proof attempts for birank (6,6) stopped. Files. ppt-3x4.pdf is the note. ppt-3x4-supplement.zip contains the statements, proofs and blind reviews of the results as written, the exact computation for the Winda maps, the numerical data and scripts (including exact rational birank-(6,6) edge states with separable compressions), the record of failed proof routes, and the unreviewed working files. "python check.py" (NumPy and SymPy) verifies the hashes and repeats the exact and numerical checks (a few seconds on the laptop used to assemble the archive). The written proofs have not yet been verified line by line by a human reviewer. AI tools. The author chose the question and the statements to attack, set the working rules (blind review of every result by the other model family; fixed stopping rules), had the published Winda maps checked and supplied the paper for that, decided when to stop, and is responsible for the content. The research was carried out on 4–5 November 2026 by an AI research system built on Claude Opus 5.5 (Anthropic) and GPT-6 Astra (OpenAI); the text was revised on 9 November 2026. Each proof was written by one of the two models and accepted by a blind review from the other; the Winda computation was done by both independently; Claude Opus 5.5 drafted the manuscript and GPT-6 Astra reviewed it blind. Section IX of the note has the details.
We show that the length of a qubit-qutrit separable state is equal to max(r,s), where r is the rank of the state and s the rank of its partial transpose. We refer to the ordered pair (r,s) as the birank of this state. We also construct examples of qubit-qutrit separable states of any feasible birank (r,s). We determine the closure of the set of normalized two-qutrit entangled states having positive partial transpose (PPT) of rank 4. The boundary of this set consists of all separable states of length at most 4. We prove that the length of any qubit-qudit separable state of birank (d+1,d+1) is equal to d+1. We also show that all qubit-qudit PPT entangled states of birank (d+1,d+1) can be built in a simple way from edge states. If V is a subspace of dimension k<d in a $2\ensuremath{\bigotimes}d$ space such that V contains no product vectors, we show that the set of all product vectors in ${V}^{\ensuremath{\perp}}$ is a vector bundle of rank $d\ensuremath{-}k$ over the projective line. Finally, we explicitly construct examples of qubit-qudit PPT states (both separable and entangled) of any feasible birank.
In this master thesis I study the extremal positive partial transpose (PPT) states of the three qubit (2times2times2) system using numerical methods. Using two algorithms which locate PPT states of a specified rank and extremal PPT states respectively, I have located numerical examples of extremal PPT states with a variety of ranks. These numerical results confirm the analytical result that all PPT states of rank less than four are separable. I also derive an upper limit on the ranks of extremal PPT states. The extremal PPT states of lowest rank, the rank four states, were studied in more detail. These states were confirmed to be biseparable in accordance with both previous analytical and numerical results. The range and kernel of these states were examined for product vectors, but none were found. In an attempt to parametrize the SLotimesSL equivalence classes of these extremal rank four states I have studied an analytical method to construct such states based on unextendible product bases (UPBs). This method can be used to create PPT states from a single equivalence class where, by design, the kernel of all states contain a UPB and the range contains no product vectors. All states where the range is not spanned by a basis of product vectors are necessarily entangled. I also present a numerical method for creating extremal rank four states that are symmetric under various combinations of partial transposes. Numerical examination of these states reveal no product vectors in neither range nor kernel. The existence of rank four states with and without product vectors in their kernel implies the existence of at least two equivalence classes. To get a better impression of these equivalence classes I construct quantities that are invariant under SLotimesSL transformations and must therefore have the same value for all states in the same equivalence class. Calculating the values of these invariants for all the rank four extremal states I have generated gives a seemingly continuous range of values. This indicates that there is an infinite number of equivalence classes likely described by one or more continuous variables. The invariants also revealed an interesting set of states that may belong to a single equivalence class, where one invariant is zero and the others have identical values. This was the only equivalence class where more than one of my states were included. There is obviously something special about this class, but I do not know what it is.
Positive but not completely positive maps provide one of the most direct ways to detect entanglement beyond the positive-partial-transpose (PPT) criterion. We introduce and analyze an exactly solvable two-parameter family of sparse bistochastic positive maps on qutrits, in which two coherence channels are independently tuned by parameters w and z. The sparse structure makes the full phase diagram analytic: positivity holds exactly on the square 0≤w,z≤2/3, complete positivity on the smaller square 0≤w,z≤1/3, and decomposability is lost precisely outside a quarter circle in the corner w,z≥1/3. The indecomposable region is certified by explicit PPT entangled state adapted to the same witness geometry. At the endpoint W∗=W(2/3,2/3) we construct a four-parameter family of PPT edge states of rank type (5,5), derive their analytic detection region, and show that the corresponding rays are exposed faces of the PPT cone. Finally, although W∗ is not optimal, we give an explicit optimal refinement whose detection region on this family is strictly larger. The result is an analytically tractable qutrit setting in which positivity, indecomposability, PPT entanglement, optimality, and exposed convex geometry can be studied in a single framework.
The PPT2 conjecture asserts that composites of PPT (completely positive and completely copositive) maps are entanglement breaking. For PPT maps Φ: Md1 → M3 and Ψ: M3 → Md2 let W be the Choi matrix of Ψ ∘ Φ. With the theorem that composites of PPT maps through a qubit are entanglement breaking (the qubit theorem), it is known that W + (1/5) Φ*(I)⊤ ⊗ Ψ(I) is separable, while PPT2 through a qutrit is the same statement with 1/5 replaced by 0. We ask how far 1/5 can be lowered. Certificates built from the qubit theorem independently of the pair cannot lower it, and for d1, d2 ≥ 3 with max(d1, d2) ≥ 4 neither can averages of certificates that apply maps separately to the input, middle and output spaces, even with PPT2 on M3 added. The least λ for which J(Φ) + λ Φ*(I)⊤ ⊗ I is a sum of PPT operators supported on ℂd1 ⊗ V for planes V is the one-sided constant λQD(Φ) ≤ 1/5; by the qubit theorem it is valid for every partner. It is below 1/5 when one compression of the middle qutrit to a qubit is strictly PPT of full rank, and its largest value Λ(d) over d1 = d is below 1/5 unless some extreme PPT channel is rigid, that is, has λQD = 1/5. We prove, without the qubit theorem, that no PPT channel M3 → M3 is rigid, so that some c < 1/5 makes ρ + c I ⊗ I separable for the Choi matrix ρ of every PPT channel M3 → M3. With the qubit theorem, W + c Φ*(I)⊤ ⊗ Ψ(I) is then separable for all PPT maps Φ: M3 → M3 and Ψ: M3 → Md2, and exact rational certificates beat 1/5 at explicit factors with d1 = 4. The exact computations used in the proofs are also checked in Lean 4. This paper develops Section 10, Appendix A and supplement S12-S13 of version 3 of "Diagonal-unitary twirls of composites of PPT maps through a qutrit" (doi:10.5281/zenodo.23675449) with a new introduction and preliminaries; its supplement S1-S2 is S12-S13 there. The statements and proofs are those of that version, renumbered, with references to the companion paper, version 4 of that record (doi:10.5281/zenodo.23680469). Programs, certificates and Lean checks are in the deposit, version 4 (doi:10.5281/zenodo.23680464). Dagger k of this paper is dagger k+24 in version 3 and in the deposit's folders, map files, replay record and Lean targets; the deposit's README uses this paper's numbering. Mathematics Subject Classification (2020): 81P40; 15A69, 47L07.
The PPT² conjecture asserts that the composition of two completely positive and completely copositive (PPT) maps is entanglement breaking. A recent counterexample uses two maps through a six-dimensional middle algebra between ten-dimensional outer algebras, and certifies entanglement by the range criterion: the range of the composite Choi matrix contains no product vector. We ask how small such a certificate can be. For PPT maps Φ: Mₘ → Mₖ and Ψ: Mₖ → Mₙ with Ψ ∘ Φ ≠ 0 we prove, with computer-assisted steps, that the range of the Choi matrix of Ψ ∘ Φ contains a product vector whenever max(m, k, n) ≤ 6 and either k ≤ 5 or an outer dimension is at most five. If both outer dimensions are four, this holds for every middle dimension k. In dimension six only the reduced type (6, 6, 6) remains, and we bound its Choi ranks. The proofs use that, after a reduction, a product-free composite range forbids Kraus operators of rank at most two, that the rank-loss loci of a PPT map are finite unions of linear subspaces with generically constant range on each subspace, a rank inequality for PPT maps at product vectors in the kernel of their Choi matrices, top Chern classes at middle dimensions four and five, an excess intersection on Gr(3, 6) at middle dimension six, and, for the type (5, 6, 5), the Euler characteristic of a sheaf on a multiple line in ℙ². For the known example we show that the quadratic Veronese images of the rank-loss directions of its base map and of their kernel vectors are Gale dual by the Cayley–Bacharach theorem, and that the kernel conditions determine the Kraus spaces of both factors. If the rank-loss scheme of the base map is reduced and of the expected dimension zero, then among the lifts through arbitrary Schur functors with nonzero composite only those through exterior powers give an unextendible product set of kernel product vectors, and these only when one of the three dimensions is at least ten. For a general base map, the member of the family with outer dimension six is one product vector short of the range criterion, and its factors are rigid. For such base maps on M₄ that are not entanglement breaking, a compression of the lift to outer dimensions 4 ≤ D ≤ 9 keeps at most 2D − 4 of the twenty kernel product vectors. Finally, symmetry under a compact connected group acting irreducibly on both outer spaces, or under SU(2) acting without multiplicities on one of them, puts a product vector into the range. Version 2 splits the paper into a main paper (ppt2-paper2-v2-main.pdf) and a supplement (ppt2-paper2-v2-supplement.pdf); the mathematics is the same as in version 1. Mathematics Subject Classification (2020): 81P40; 15A69, 14M15, 14C17, 47L07.
Krebs and Gachechiladze [Phys. Rev. A 113, 062428 (2026)] introduced a family of positive-partial-transpose (PPT) generalized grid states in local dimension d = 2k − 1. They determined the Schmidt number for k ≤ 5 and proposed an all-order scaling pattern. For their particular explicitly specified family, we prove SN(R_k) = k = (d + 1)/2 for every k ≥ 2 using a universal signed-minor identity. We also prove that each R_k generates an extreme ray of the complex Hermitian PPT cone, and obtain exact ranks rank(R_k) = k² + k − 1 and rank(R_k^Γ) = k² + 2k − 3. Local filtering yields extremal trace-preserving PPT channels whose Choi matrices have Schmidt number k and whose partially transposed Choi matrices have Schmidt number two. Their threefold iterates are entanglement breaking by a theorem of Park (2026). The underlying states, their PPT decomposition, and other families with linear Schmidt-number growth were established in earlier work; the results here concern the specific 2026 grid-state family.