Bound Entanglement with Negative Partial Transpose

For a finite-dimensional bipartite system, it suffices to consider equal local dimensions. Let $d\geq 3$, let $\mathcal H_d=\mathbb C^d$, and let $\rho$ be a density operator on $\mathcal H_d\otimes\mathcal H_d$ (that is, $\rho\succeq 0$ and $\operatorname{Tr}\rho=1$). In a fixed product basis, write $\rho^\Gamma=(\operatorname{id}\otimes T)(\rho)$, where $T$ is matrix transposition on the second factor. Call $\rho$ NPT if $\rho^\Gamma$ has a strictly negative eigenvalue. The state is distillable if, for some finite integer $n\geq1$, there are rank-two orthogonal projectors $P,Q$ on the respective spaces $\mathcal H_d^{\otimes n}$ such that

$$(P\otimes Q)(\rho^\Gamma)^{\otimes n}(P\otimes Q)$$

has a negative eigenvalue. An entangled state that is not distillable is called bound entangled; NPT states are entangled. The question is whether there exist $d\geq3$ and such a state $\rho$ for which

$$\rho^\Gamma\not\succeq0,\qquad (P\otimes Q)(\rho^\Gamma)^{\otimes n}(P\otimes Q)\succeq0$$

for every finite $n\geq1$ and every such pair $P,Q$. The equal-dimension formulation covers the general finite-dimensional question by reduction; dimension two cannot supply an example, since every entangled two-qubit state is distillable. Five open problems in theory of quantum information

Background

The distinction between positive and negative partial transpose matters because distillation is possible only for NPT states, whereas PPT entangled states are nondistillable and thus already provide bound entanglement. The question asks whether bound entanglement is restricted to the PPT class, or whether a state whose partial transpose has a negative eigenvalue can nevertheless resist every finite-copy distillation test. This contrasts with the two-qubit result that all entangled states are distillable. Mixed-state entanglement and distillation: is there a “bound” entanglement in nature? Five open problems in theory of quantum information

An early line of attack studied canonical families of NPT states and tested whether projections across multiple copies could extract two-qubit entanglement. The 1999 work established finite-copy nondistillability in parameter ranges that depended on the number of copies, but its proposed NPT bound-entangled states remained conjectural: the crucial issue was whether a positive nondistillability threshold persisted uniformly over all copy numbers. The problem is also connected to the structure of 2-co-positive maps. Evidence for Bound Entangled States with Negative Partial Transpose Five open problems in theory of quantum information

Known results

A reduction reported in the 2020 account makes Werner states a sufficient test family: for local dimension $d>2$, NPT nondistillable entanglement exists if and only if an NPT nondistillable Werner state exists. In the parameterization

$$\rho_\alpha=\frac{I_{d^2}+\alpha F_d}{d^2+\alpha d},\qquad -1\leq\alpha\leq1,\qquad F_d(x\otimes y)=y\otimes x,$$

the NPT range is $\alpha<-1/d$, and one-copy distillability holds exactly for $\alpha<-1/2$. Thus for $d\geq3$, the range $-1/2\leq\alpha<-1/d$ is NPT but not one-copy distillable. This does not establish bound entanglement, which requires nondistillability for every finite number of copies. Five open problems in theory of quantum information Two-copy nondistillability of Werner states: sharp partial-trace inequalities and finite-copy extensions

A 2026 preprint reports that, in every local dimension, a Werner state is two-copy distillable if and only if $\alpha<-1/2$. In particular, the endpoint $\alpha=-1/2$ is two-copy undistillable and, for $d\geq3$, is still NPT. The preprint gives additional finite-copy bounds, but explicitly leaves the endpoint question unresolved for unrestricted $k\geq3$ copies. Its two-copy result therefore sharpens what is known at the Werner boundary without showing that any state is nondistillable for all finite copy numbers. Two-copy nondistillability of Werner states: sharp partial-trace inequalities and finite-copy extensions