A Minimum-Cardinality Genuinely Unextendible Product Basis in Three Qutrits

Technical Summary: A Minimum-Cardinality Genuinely Unextendible Product Basis in Three Qutrits

Problem Statement
The paper addresses a long-standing open question in quantum information theory: the existence of a Genuinely Unextendible Product Basis (GUPB). While Unextendible Product Bases (UPBs)—sets of orthogonal product states whose orthogonal complement contains no product state—are well-established, a GUPB is a stricter concept. A GUPB is defined as a set of orthogonal product states whose orthogonal complement contains no biproduct state (a state that is a product state across any bipartition of the system). Equivalently, a GUPB must be a UPB across every possible bipartition.

Prior to this work, the existence of orthogonal GUPBs was unresolved. Although general lower bounds on their cardinality existed, and it was known that a GUPB cannot exist if any local subsystem has dimension 2 (making the three-qutrit space, $(\mathbb{C}^3)^{\otimes 3}$, the smallest candidate tripartite setting), no explicit construction had been found. Furthermore, while a cardinality of 13 was recently excluded for three-qutrit GUPBs, the minimum possible cardinality remained undetermined.

Methodology
The authors employ a graph-theoretic approach combined with computer-assisted search and exact algebraic verification:

1. Graph-Theoretic Formulation: The problem is mapped to finding three graphs ($G_A, G_B, G_C$) on $n$ vertices (representing the product states) such that their union is the complete graph $K_n$. An edge in $G_X$ exists between vertices $i$ and $j$ if the local vectors for party $X$ are orthogonal. The condition that the product states are mutually orthogonal requires that for every pair of distinct states, at least one local inner product vanishes, meaning the union of the three graphs must cover all edges of $K_n$.
2. Computer-Assisted Search: The authors searched for triples of graphs on 14 vertices whose union is $K_{14}$ and which admit compatible faithful orthogonal representations in $\mathbb{C}^3$. To reduce the search space, they imposed the symmetry $G_A \cong G_B$.
3. Exact Verification: Once a candidate construction was identified, the authors performed exact calculations (using integer arithmetic to avoid numerical errors) to verify:
   • Mutual Orthogonality: Confirming the union of the orthogonality graphs is indeed $K_{14}$.
   • Genuine Unextendibility: Using a rank-based criterion (Lemma 1), they verified that for every bipartition ($A|BC$, $B|AC$, $C|AB$), the set of states forms a bipartite UPB. This involved checking that for any subset of states $J$, if the local vectors on one side do not span the full local space, the vectors on the complementary side must span the full space of the remaining parties. This required checking the ranks of submatrices derived from the local vectors.
4. Extension via Padding: To generalize the result to arbitrary local dimensions $d_A, d_B, d_C \geq 3$, the authors used a "padding" procedure. They embedded the 14-state three-qutrit GUPB into the subspace spanned by $\{|0\rangle, |1\rangle, |2\rangle\}$ of the larger Hilbert space and filled the remaining orthogonal complement with computational basis states involving indices $\geq 3$.

Key Contributions and Results

• Existence Proof: The paper constructs an explicit GUPB in the three-qutrit space $(\mathbb{C}^3)^{\otimes 3}$ consisting of 14 product states. This resolves the open question of whether GUPBs exist.
• Minimum Cardinality: By combining their construction with previous results excluding cardinality 13, the authors prove that 14 is the minimum cardinality for a three-qutrit GUPB.
• Explicit Construction: The paper provides the specific unnormalized real local vectors for all 14 states in Table I and the corresponding orthogonality graphs in Figure 1.
• Generalization: The construction is extended to any tripartite system $\mathbb{C}^{d_A} \otimes \mathbb{C}^{d_B} \otimes \mathbb{C}^{d_C}$ where $3 \leq d_A \leq d_B \leq d_C$, yielding a GUPB of cardinality $d_A d_B d_C - 13$.
• Applications:
  • Bound Entanglement: The normalized projector onto the 13-dimensional orthogonal complement of the constructed GUPB is shown to be Positive under Partial Transposition (PPT) and entangled across every bipartition. Consequently, it represents a bound entangled state that is genuinely entangled.
  • Strong Quantum Nonlocality: The constructed GUPB is shown to exhibit strong quantum nonlocality without entanglement. This means the set of states is locally irreducible across every bipartition; no party (or pair of parties) can perform a nontrivial orthogonality-preserving local measurement to eliminate any state from the set. This is verified by showing that the operator space spanned by the relevant local transitions has the maximal possible dimension ($d^2-1$).

Significance
The paper claims to resolve a fundamental open problem in the theory of quantum nonlocality and entanglement by demonstrating the existence of GUPBs. By establishing the minimum cardinality in the smallest possible tripartite Hilbert space, it sets a precise benchmark for the complexity of such structures. The work further provides a concrete resource for generating bound entangled states that are PPT across all cuts and for studying strong nonlocality, a phenomenon where quantum correlations prevent local discrimination of states even without entanglement. The authors note that while independent concurrent work exists constructing GUPBs from MDS codes, their work is distinct in providing the minimum-cardinality instance in the smallest possible tripartite system.