Bi-PPT State in Quantum Systems

• Bi-PPT state is a bipartite quantum state whose positive partial transpose signifies separability in low-dimensional systems and potential bound entanglement in higher dimensions.
• Analysis of rank and birank conditions shows that low global rank guarantees separability for PPT states, while NPT states with matching rank often exhibit distillability.
• The geometric and extremal structure of the PPT set underpins classifications such as checkerboard states, generalized UPBs, and bi-PPT channels which are critical in quantum communication.

A bi-PPT state is, in the standard finite-dimensional bipartite sense, a state $\rho$ on $\mathcal{H}_A \otimes \mathcal{H}_B$ whose partial transpose is positive semidefinite. If $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$, then the partial transpose on $B$ is

$$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$

and $\rho$ is PPT precisely when $\rho^{T_B}\ge 0$; otherwise it is NPT. In $2\otimes 2$ and $2\otimes 3$, PPT is equivalent to separability, whereas in higher dimensions PPT is only necessary, and PPT entangled states occur. The term also acquires more specialized meanings in later literatures, including multipartite bipartition-PPT states, absolutely PPT states, and bi-PPT channels, but the core object is the bipartite positive-partial-transpose state (Chen et al., 2011).

1. Definition, ranks, and the PPT–NPT dichotomy

For a bipartite state $\rho$, the reduced operators

$\mathcal{H}_A \otimes \mathcal{H}_B$0

have ranks $\mathcal{H}_A \otimes \mathcal{H}_B$1 and $\mathcal{H}_A \otimes \mathcal{H}_B$2, while $\mathcal{H}_A \otimes \mathcal{H}_B$3 is the global rank. Chen and Đoković use the notation “$\mathcal{H}_A \otimes \mathcal{H}_B$4 state” for a state with $\mathcal{H}_A \otimes \mathcal{H}_B$5 and $\mathcal{H}_A \otimes \mathcal{H}_B$6; this refers to local ranks, not necessarily the ambient Hilbert-space dimensions (Chen et al., 2011).

Distillability is formulated through the partial transpose. A state is 1-distillable if there exists a pure state $\mathcal{H}_A \otimes \mathcal{H}_B$7 of Schmidt rank 2 such that

$\mathcal{H}_A \otimes \mathcal{H}_B$8

More generally, $\mathcal{H}_A \otimes \mathcal{H}_B$9 is $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$0-distillable if $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$1 is 1-distillable, and PPT states are never distillable by LOCC. A PPT entangled state is therefore a bound entangled state. This makes the PPT/NPT boundary the operational dividing line between guaranteed nondistillability and possible distillability in the bipartite setting (Chen et al., 2011).

The same positivity notion admits refinements. In $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$2, a state can be SPPT (“strong PPT”) if positivity of the partial transpose is recognized directly from a canonical block-Cholesky factorization; for these systems, vanishing discord on the qubit side implies SPPT, and any PPT state that is not SPPT necessarily has nonzero discord (Bylicka et al., 2010). This does not redefine PPT, but it exhibits a structurally stronger subclass.

2. Low-rank bi-PPT states and the distillability threshold

A central structural result for bipartite PPT theory is that low global rank strongly constrains both separability and distillability. Before Chen and Đoković, it was known that if

$$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$3

then $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$4 is distillable. Their main extension shows that if $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$5 and $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$6 is NPT, then $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$7 is still distillable; in the $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$8 formulation with $$\rho = \sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k| \otimes |j\rangle_B\langle l|$$9, every NPT state of rank $B$0 is 1-distillable. In particular, any $B$1 NPT state of rank $B$2 is distillable, resolving the conjectured symmetric case, and all rank-3 entangled states are distillable (Chen et al., 2011).

Combining this theorem with earlier low-rank PPT results yields a sharp classification in the regime

$B$3

In that regime, PPT and separability coincide, while NPT and distillability coincide. Equivalently, there are no PPT entangled states and no NPT bound entangled states at ranks up to the maximal local rank. This is a rank-constrained extension of the familiar $B$4 and $B$5 Peres–Horodecki sufficiency regime, but now expressed in terms of ranks rather than local dimensions alone (Chen et al., 2011).

For qubit–qudit systems, birank language makes this especially explicit. If $B$6 and $B$7, then $B$8 is the birank. In $B$9, the length of a separable state is exactly $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$0, and examples exist for every feasible birank. More generally, any qubit–qudit separable state of birank $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$1 has length $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$2, while any qubit–qudit PPT entangled state of birank $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$3 can be built from edge states (Chen et al., 2012).

3. Rank-4 classification, range criteria, and generalized UPBs

Rank 4 is the first nontrivial rank at which PPT entanglement appears in the bipartite theory. Chen and Đoković prove a complete separability criterion for arbitrary bipartite rank-4 states: $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$4 Thus, for rank 4, the trichotomy is exact: PPT plus a product vector in the range implies separability; PPT with no product vector in the range implies PPT entanglement; NPT implies distillability (Chen et al., 2011).

The decisive object here is the range $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$5. If $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$6 contains no product vector, it is a completely entangled subspace. In $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$7, the canonical rank-4 PPT entangled examples are precisely of this kind, including states derived from unextendible product bases (UPBs), whose ranges are orthogonal complements of UPBs and therefore product-free (Chen et al., 2011).

This range-based picture is refined by the UPB classification of low-rank extremal PPT states. In $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$8, a rank-4 entangled PPT state belongs to a continuous family related by non-singular product transformations; its kernel carries a generalized UPB, meaning a basis of product vectors not necessarily orthogonal, with no product vector in the image. The generalized UPB has the special property that it can be transformed to orthogonal form by a product transformation. The paper gives a complete parametrization of orthogonal $$\rho^{T_B}=\sum_{i,j,k,l}\rho_{ij,kl}\,|i\rangle_A\langle k|\otimes |l\rangle_B\langle j|,$$9 UPBs and presents strong numerical evidence that this parametrizes all rank-4 entangled extremal PPT states in $\rho$0 (Leinaas et al., 2010).

A particularly important structured family is the checkerboard family of $\rho$1 rank-4 states. Chen and Đoković prove that all NPT checkerboard states are 1-distillable. Within that family, NPT and distillability coincide, while the PPT members are PPT entangled rank-4 states with completely entangled range (Chen et al., 2011).

4. Geometric and extremal structure of the PPT set

The set of density operators is a convex compact body, and PPT states define a convex subset within it. A geometric description due to Hilbert–Schmidt distance from the maximally mixed state places PPT, separable, and distillable regions in nested shells. For an $\rho$2-qudit bipartite system with total dimension $\rho$3, every state within distance

$\rho$4

from $\rho$5 is necessarily PPT, whereas all states within distance

$\rho$6

are necessarily separable. Since the separable radius is strictly smaller than the PPT radius, there is a nonempty shell in which states are PPT but not forced to be separable; in higher dimensions this shell contains PPT bound entangled states (Banerjee et al., 2017).

Extremal PPT states organize the boundary geometry of the PPT set. For a bipartite $\rho$7 system, Let $\rho$8 denote the set of extreme points of the compact convex set of PPT states, with $\rho$9 the subset of rank-$\rho^{T_B}\ge 0$0 extreme states with full local ranks $\rho^{T_B}\ge 0$1 and $\rho^{T_B}\ge 0$2. It is known that $\rho^{T_B}\ge 0$3 is the set of pure product states, and that $\rho^{T_B}\ge 0$4 is empty for $\rho^{T_B}\ge 0$5, for $\rho^{T_B}\ge 0$6, and also for $\rho^{T_B}\ge 0$7. The Leinaas–Myrheim–Sollid conjecture predicted that $\rho^{T_B}\ge 0$8 is nonempty and that $\rho^{T_B}\ge 0$9 is empty for $2\otimes 2$0; the first part is proved in full generality, while the second is proved when $2\otimes 2$1 and also when $2\otimes 2$2. For a good state $2\otimes 2$3, the range contains no product vectors and $2\otimes 2$4 has the same rank $2\otimes 2$5 (Chen et al., 2012).

Numerical studies complement this by reporting systematic rank constraints for extremal PPT states. If $2\otimes 2$6, $2\otimes 2$7, and $2\otimes 2$8, then extremality implies

$2\otimes 2$9

Across the studied low-dimensional systems, the lowest-rank extremal full-local-rank PPT states appear at

$2\otimes 3$0

and in the $2\otimes 3$1 case these are exactly the rank-$2\otimes 3$2 states discussed above (Leinaas et al., 2010).

A recent bi-qutrit development pushes this extremal geometry to the maximal-rank edge case. If $2\otimes 3$3 is an eight-dimensional subspace of $2\otimes 3$4 whose orthogonal complement is spanned by a Schmidt-rank-3 vector, then there exist PPT entangled edge states with range $2\otimes 3$5 and partial transpose rank 6, i.e. of type $2\otimes 3$6. This gives a large family of bi-qutrit PPT edge states with the largest possible ranks; their faces in the PPT set also contain boundary edge states of types $2\otimes 3$7, $2\otimes 3$8, $2\otimes 3$9, $\rho$0, $\rho$1, $\rho$2, and $\rho$3 (Han et al., 15 Jun 2026).

5. Extended meanings of “bi-PPT”

In the literature, “bi-PPT” is not completely uniform. The following usages all occur.

Context	Meaning	Consequence
Bipartite state	$\rho$4	PPT may mean separable or bound entangled
Multipartite state	PPT with respect to a chosen bipartition $\rho$5	Used in PPT mixtures and biseparability tests
Symmetric $\rho$6-qubit state	PPT across an $\rho$7 cut	Equivalent to positivity of a tensor-built matrix
Channel theory	Channel and a complementary channel are both PPT	Such channels are entanglement breaking
Absolute PPT	State remains PPT under all global unitaries	A spectral robustness notion

For multipartite systems, Ha and Kye define an $\rho$8-$\rho$9 bi-PPT state as a state that is PPT when regarded as bipartite across that split. In three qubits, they construct genuinely entangled states that are PPT with respect to bipartitions, disproving the conjecture that PPT mixtures are necessary and sufficient for biseparability of three qubits (Ha et al., 2015).

For symmetric multi-qubit states, partial transpose criteria can be rewritten in terms of tensor-representation matrices. For an $\mathcal{H}_A \otimes \mathcal{H}_B$00 symmetric state, the partial transpose across an $\mathcal{H}_A \otimes \mathcal{H}_B$01 bipartition is similar to a matrix $\mathcal{H}_A \otimes \mathcal{H}_B$02 built from the spin tensor, and positivity of this matrix is equivalent to positivity of a correlation matrix constructed from tensor products of Pauli operators. The unitary transformations implementing this similarity generalize the magic basis and Bell-type bases (Bohnet-Waldraff et al., 2016).

In channel theory, Hirche and Leditzky introduced bi-PPT channels as channels such that the channel and one complementary channel are both PPT. The decisive result is that bi-PPT channels are always entanglement breaking. Consequently, their Choi matrices and all output states obtained by acting on one half of a bipartite system are not merely PPT but separable (Müller-Hermes et al., 2022).

An even stronger robustness notion is absolute PPT: a state is absolutely PPT if it remains PPT under every global unitary. In $\mathcal{H}_A \otimes \mathcal{H}_B$03, absolutely separable and absolutely PPT states coincide, and each extreme point has at most three distinct eigenvalues. In $\mathcal{H}_A \otimes \mathcal{H}_B$04, absolutely PPT extreme points are characterized by solvable linear equations and have at most seven distinct eigenvalues (Song et al., 2024).

6. Structured families, algorithms, and operational diagnostics

Several papers isolate special classes where PPT becomes an effectively complete invariant. One $\mathcal{H}_A \otimes \mathcal{H}_B$05 family consists of states

$\mathcal{H}_A \otimes \mathcal{H}_B$06

with indices modulo $\mathcal{H}_A \otimes \mathcal{H}_B$07. For this shifted-diagonal class, PPT is both necessary and sufficient for separability, and the paper explicitly constructs separable pure-state decompositions of all PPT members by solving phase constraints that force certain coefficient matrices to rank 1 (Zhang et al., 2013).

PPT structure can also be analyzed through associated positive maps. Given

$\mathcal{H}_A \otimes \mathcal{H}_B$08

define

$\mathcal{H}_A \otimes \mathcal{H}_B$09

For PPT $\mathcal{H}_A \otimes \mathcal{H}_B$10, the map $\mathcal{H}_A \otimes \mathcal{H}_B$11 admits a Sinkhorn–Knopp-type analysis: under a tensor-rank condition on a vector in $\mathcal{H}_A \otimes \mathcal{H}_B$12, one can algorithmically decide whether $\mathcal{H}_A \otimes \mathcal{H}_B$13 is equivalent to a doubly stochastic map, using Perron eigenvectors and at most $\mathcal{H}_A \otimes \mathcal{H}_B$14 unconstrained quadratic minimization problems. Equivalently, one can decide whether the state can be put into filter normal form (Cariello, 2018).

PPT also controls operational distinguishability. A subspace is strongly PPT-unextendible if no PPT operator is supported on the orthogonal complement of any tensor power of that subspace. If a subspace contains a PPT-definite operator, then it is strongly PPT-unextendible. This gives a criterion for many-copy indistinguishability by PPT operations, and in particular implies that any entangled pure state and its orthogonal complement cannot be distinguished by PPT operations in the many-copy scenario (Li et al., 2017).

Taken together, these results identify bi-PPT states as the locus where separability, bound entanglement, distillability, extremality, and operational convertibility meet. In the low-rank regime, PPT becomes nearly classificatory; at rank 4, range structure decides separability; in extremal geometry, PPT edge states organize faces of the PPT cone; and in variant formulations—SPPT, absolutely PPT, multipartite bi-PPT, and bi-PPT channels—the same positivity constraint is reinterpreted as a structural robustness condition across decompositions, symmetries, or complementary maps (Chen et al., 2011).