Unextendible Biseparable Basis (UBB)

• Unextendible Biseparable Basis (UBB) is a set of orthogonal biseparable states whose complement forms a genuinely entangled subspace, ensuring no biseparable state exists in the complement.
• The UBB framework converts an orthogonality challenge into a subspace construction task, enabling the systematic creation of genuinely entangled subspaces with bidistillable properties across bipartitions.
• UBBs extend the unextendible product basis (UPB) concept by excluding all biseparable vectors from the complement, thereby enhancing operational criteria such as LOCC-indistinguishability and strong quantum nonlocality.

An unextendible biseparable basis (UBB) is a multipartite orthogonal-state structure defined by a separation property of its complement: it is a set of pairwise orthogonal pure biseparable states spanning a proper subspace of a composite Hilbert space such that the orthogonal complement contains no biseparable state. Consequently, the complement is a genuinely entangled subspace (GES), so every nonzero vector in it is genuinely multipartite entangled. Since its introduction as a multipartite strengthening of the unextendible product basis (UPB) mechanism, the notion has become a focal point for constructing GESs with additional operational properties, including distillability across all bipartitions and strong quantum nonlocality (Agrawal et al., 2018, Bhunia et al., 2024).

1. Definition and basic geometry

For a multipartite Hilbert space

$\mathcal H=\bigotimes_{j=1}^{m}\mathbb C^{d_j},$

a pure state is biseparable if it is product across at least one nontrivial bipartition of the parties. In the tripartite case, this means product in one of the cuts $A|BC$, $B|AC$, or $C|AB$. A UBB is then defined as a set of pairwise orthogonal states $\{|\psi_i\rangle\}$ spanning a proper subspace of $\bigotimes_j \mathbb C^{d_j}$ such that every $|\psi_i\rangle$ is biseparable and the complementary subspace contains no biseparable state (Agrawal et al., 2018, Zhou et al., 10 Mar 2026).

If $$\mathcal H_{\rm UBB}:=\operatorname{span}\{|\psi_i\rangle\}$$, the defining property is that

$\mathcal H_{\rm UBB}^{\perp}$

contains no biseparable pure state. In a tripartite system, that immediately implies that every pure state in $\mathcal H_{\rm UBB}^{\perp}$ is genuinely entangled, so $A|BC$0 is a GES (Bhunia et al., 2024).

A useful technical criterion appears in the tripartite rank formulation. For

$A|BC$1

the state is biseparable iff in at least one bipartition the reshaped coefficient array has matrix rank $A|BC$2; it is genuinely entangled iff in every bipartition the corresponding reshaped matrix has rank $A|BC$3. This criterion is used explicitly in unextendibility proofs for concrete UBBs (Bhunia et al., 2024).

2. Relation to UPBs, CESs, and GESs

UBBs are the biseparable analogue of UPBs. A UPB is a set of orthogonal fully product states spanning a proper subspace whose orthogonal complement contains no fully product state. Its complement is therefore a completely entangled subspace (CES), meaning that it contains no fully product vector. A UBB strengthens this logic by excluding not merely fully product states but all biseparable states from the complement, and therefore yields a GES rather than merely a CES (Agrawal et al., 2018, Bhunia et al., 2024).

This distinction is structurally important. In multipartite systems, a vector may fail to be fully product while remaining separable across some bipartition. Thus a CES need not be genuinely entangled. This was already emphasized in work on UPB-generated GESs: a multipartite UPB has a GES in the orthocomplement of its span iff it is a bipartite UPB across every possible cut (Demianowicz et al., 2017). That criterion identifies the precise gap between ordinary product-state unextendibility and the stronger condition needed to eliminate all biproduct or biseparable structure.

Several precursor lines sharpened this gap without yet formulating UBB directly. One $A|BC$4 UPB in four-qubit space was shown to be orthogonal to an “almost genuinely entangled space,” in the sense that its complement contains no $A|BC$5 product vector in any balanced bipartition, though it is not a full GES because $A|BC$6 product vectors remain (Wang et al., 2019). Related work on tile structures and stopper constructions produced UPBs that are uncompletable or, under sufficient conditions, remain UPBs in every bipartition, thereby providing a combinatorial route toward subspaces with stronger multipartite entanglement properties (Shi et al., 2022).

3. Original UBB constructions and the distillability program

The term UBB was introduced in the three-party setting as a systematic way to construct GESs in $A|BC$7 for all $A|BC$8. Two classes were given: party-symmetric and party-asymmetric UBBs. In the symmetric construction, the resulting GES was shown to be bidistillable, meaning that every supported state has distillable entanglement across every bipartition (Agrawal et al., 2018).

The original three-qutrit construction was formulated through a twisted orthogonal product basis on a $A|BC$9 block-cube. The local superpositions

$B|AC$0

generate six face blocks together with a body-diagonal block $B|AC$1. The UBB mechanism then removes selected states and uses a stopper configuration so that the missing directions cannot be reinserted by any biseparable vector orthogonal to the retained set (Agrawal et al., 2018).

This early framework fixed two themes that continue through later work. First, UBBs are valuable because they convert an orthogonality problem into a subspace-construction problem: once the basis is unextendible against biseparable states, the complement is automatically a GES. Second, the most useful GESs are not merely genuine but distillable across every cut. That “all-encompassing distillable entanglement across every bipartition” became a recurring criterion for evaluating the operational strength of UBB-derived subspaces (Agrawal et al., 2018).

4. Strong quantum nonlocality and the $B|AC$2 construction

A major development was the connection between UBBs and strong quantum nonlocality in the local irreducibility paradigm. In this framework, a local measurement is orthogonality-preserving if the postmeasurement states remain pairwise orthogonal; a set is locally irreducible if no party can eliminate one or more states by any nontrivial orthogonality-preserving local measurement; and in a tripartite system it is strongly nonlocal if it is locally irreducible both in the full tripartition and in every bipartition. This notion is stronger than ordinary LOCC indistinguishability and stronger than the usual “nonlocality without entanglement” criterion (Bhunia et al., 2024).

Earlier proof techniques for strong nonlocality were developed in the UPB setting, where strongly nonlocal UPBs were shown to exist in $B|AC$3 for all $B|AC$4 (Shi et al., 2021). The 2024 UBB advance carried this program beyond UPB-based constructions. It introduced two subclasses: $B|AC$5, which contains a UPB as a subset, and $B|AC$6, which does not contain any UPB as a subset. The main result is a family of tripartite $B|AC$7 constructions in $B|AC$8, $B|AC$9, thereby going beyond UBBs obtained by extending UPBs (Bhunia et al., 2024).

In the explicit $C|AB$0 instance, the basis $C|AB$1 is built from a complete orthogonal biseparable basis, a stopper state

$C|AB$2

and the removal of three states $C|AB$3 together with the three diagonal states $C|AB$4. The resulting set has size $C|AB$5 in a $C|AB$6-dimensional space, so its complement has dimension $C|AB$7. Unextendibility is proved by showing that any vector in the orthogonal complement must be a linear combination of the six missing states and orthogonal to the stopper, after which a bipartition rank analysis rules out biseparability. This construction was shown to be strongly nonlocal and thereby answered positively the open problem raised by Agrawal et al. concerning the existence of a UBB with strong quantum nonlocality in the local irreducibility sense (Bhunia et al., 2024).

5. Structural results, no-go phenomena, and corrected misconceptions

One persistent question has been whether the complement of a product-state construction can already be genuinely entangled. In that language, the relevant object is a genuinely unextendible product basis (GUPB): a multipartite UPB that is also a UPB for every bipartition. This is effectively the product-state analogue of the UBB objective. Broad no-go results show that such constructions are highly constrained. In particular, for equal local dimensions $C|AB$8, all candidate cardinalities

$C|AB$9

are forbidden for GUPBs, so maximal-dimension and near-maximal-dimension GESs cannot in general be obtained from orthogonal UPBs of those smallest sizes (Demianowicz, 2022).

These negative results do not imply nonexistence of UBBs. Rather, they show that product-state unextendibility is much more restrictive when one demands absence of biproduct vectors in the complement. This sharpened the motivation for genuine biseparable-state constructions, as opposed to attempts to realize all GESs as complements of UPBs (Demianowicz, 2022).

A later conceptual consolidation established that every UBB is LOCC-indistinguishable. The same work introduced a sufficient criterion for proving that a subspace is genuinely entangled by using product-forming matrices and symmetrization matrices: if these span the full matrix space $\{|\psi_i\rangle\}$0 in each bipartition, then the subspace contains no product state in that cut. Applied to every bipartition, this yields a practical sufficient condition that a complementary subspace is a GES, and hence that the original orthogonal biseparable set is a UBB (Bera et al., 23 Aug 2025).

That paper also corrected a potential misconception. Because previously known UBBs satisfied strong-nonlocality criteria, it was natural to suspect that strong nonlocality might be inherent to UBBs. The paper proves instead that while every UBB is LOCC-indistinguishable, strong nonlocality is not automatic: it constructs a UBB that violates the stated no-go condition and exhibits locality across certain bipartitions (Bera et al., 23 Aug 2025). This suggests that unextendibility, LOCC indistinguishability, and strong nonlocality form a strict hierarchy rather than a single equivalence class.

6. Low-dimensional and multipartite extensions

Recent work has pushed UBBs both downward in local dimension and upward in party number. In $\{|\psi_i\rangle\}$1, a five-state orthogonal biseparable set

$\{|\psi_i\rangle\}$2

was shown to be a UBB. Here $\{|\psi_i\rangle\}$3 in an $\{|\psi_i\rangle\}$4-dimensional Hilbert space, so $\{|\psi_i\rangle\}$5. The orthogonal complement is proved to be a GES by constructing an explicit three-vector spanning set and excluding product states in each of the cuts $\{|\psi_i\rangle\}$6, $\{|\psi_i\rangle\}$7, and $\{|\psi_i\rangle\}$8 through a product-forming matrix criterion and Gröbner-basis reduction. The same work shows that $\{|\psi_i\rangle\}$9 is a locally indistinguishable subspace and that every state supported on $\bigotimes_j \mathbb C^{d_j}$0 is one-shot distillable in every cut. The construction is presented as the smallest possible UBB in the three-qubit system and as realizing the $\bigotimes_j \mathbb C^{d_j}$1 case of the proposed $\bigotimes_j \mathbb C^{d_j}$2 benchmark (Bera et al., 7 Apr 2026).

At the same time, UBB theory has been extended to four-partite systems $\bigotimes_j \mathbb C^{d_j}$3, $\bigotimes_j \mathbb C^{d_j}$4. An explicit 4-qutrit UBB $\bigotimes_j \mathbb C^{d_j}$5 is obtained by modifying a strongly nonlocal orthogonal product set: selected product states are removed from local blocks and replaced by orthogonal biseparable states spanning the same subblocks, after which a stopper state

$\bigotimes_j \mathbb C^{d_j}$6

enforces unextendibility. The qutrit complement $\bigotimes_j \mathbb C^{d_j}$7 has dimension $\bigotimes_j \mathbb C^{d_j}$8, and the general family $\bigotimes_j \mathbb C^{d_j}$9 in $|\psi_i\rangle$0 yields

$|\psi_i\rangle$1

These four-partite UBBs are proved to be strongly nonlocal, and the paper also isolates proper GESs of dimension

$|\psi_i\rangle$2

that are distillable across every bipartition (Zhou et al., 10 Mar 2026).

A parallel 2025 construction program in $|\psi_i\rangle$3 and $|\psi_i\rangle$4 emphasizes the geometry of the complement rather than the basis itself. Using stopper states

$|\psi_i\rangle$5

it constructs UBBs whose complements include a $|\psi_i\rangle$6-dimensional GES in $|\psi_i\rangle$7 and an $|\psi_i\rangle$8-dimensional GES in $|\psi_i\rangle$9, the latter described there as the largest known GES obtained from a UBB. The same work states that every state supported on that $$\mathcal H_{\rm UBB}:=\operatorname{span}\{|\psi_i\rangle\}$$0-dimensional subspace is $$\mathcal H_{\rm UBB}:=\operatorname{span}\{|\psi_i\rangle\}$$1-distillable across every bipartition (Bera et al., 23 Aug 2025).

Taken together, these developments place UBBs at the intersection of multipartite entanglement geometry, state discrimination, and distillation theory. The established picture is that UBBs provide a direct method for constructing GESs; some such complements are bidistillable or $$\mathcal H_{\rm UBB}:=\operatorname{span}\{|\psi_i\rangle\}$$2-distillable across every cut; every UBB is LOCC-indistinguishable; and only a proper subclass is strongly nonlocal (Agrawal et al., 2018, Bhunia et al., 2024, Bera et al., 23 Aug 2025).