Contextuality of all optimal quantum cloning

Quantum contextuality is a key nonclassical feature underlying advantages in quantum computation and communication. We introduce a new method to study contextuality in quantum information-processing tasks and protocols, relying solely on observed information-processing statistics. Building on the framework of Shahandeh, Yianni and Doosti in arXiv:2512.10000 and employing rank separation techniques, we prove that contextuality is the necessary resource in both phase-covariant and universal optimal quantum cloning, thereby establishing its role as a fundamental source of nonclassicality in all known optimal cloning scenarios and resolving an open problem on the connection between cloning and contextuality. As a second application, we demonstrate the power of our method by providing a new, streamlined proof of contextuality in minimum-error quantum state discrimination.

💡 Research Summary

The paper introduces a novel, statistics‑only method for detecting contextuality in quantum information‑processing tasks, building on the COPE (Conditional Outcome Probabilities of Events) framework and the rank‑separation criterion recently proposed by Shahandeh, Yianni, and Doosti. In this approach, the observable data of a prepare‑measure scenario are assembled into a COPE matrix C, whose entries are the conditional probabilities p(k|i,j) of obtaining outcome k given preparation i and measurement j. Any non‑negative matrix factorisation (NMF) C = R E yields an ontological model, with R and E representing response functions and epistemic states, respectively. However, a non‑contextual ontological model exists only when the factorisation is an equi‑rank NMF (ENMF), i.e., rank C = rank R = rank E. Failure of this condition—rank separation—signifies contextuality.

Two central theorems bridge task‑level rank separation to theory‑level contextuality. Theorem 1 shows that if a fragment COPE matrix C_F satisfies relative tomographic completeness and lacks an ENMF, then the full COPE matrix C also cannot admit an ENMF. Theorem 2 generalises this: if two fragment matrices C_F and C′_F share the same rank r and C_F has no ENMF, then C′_F cannot have an ENMF either. Together, they allow one to infer contextuality of the underlying operational theory from rank‑separation observed in a specific task’s statistics.

The authors first apply the method to minimum‑error quantum state discrimination (MEQSD). For two non‑orthogonal pure states |ψ⟩ and |ϕ⟩, the optimal discrimination measurement yields a 6 × 6 COPE matrix C_MEQSD. Each row and column contains a unique zero entry, satisfying the sparsity condition of Theorem 3, which forces the rank of either the response‑function matrix or the epistemic‑state matrix to be at least four. Yet the rank of C_MEQSD itself is only three, establishing rank separation. Consequently, C_MEQSD admits no ENMF, and by Theorems 1 and 2 the full qubit operational theory is contextual. This reproduces the known result of Schmid and Spekkens with a much shorter proof. Moreover, imposing the rank‑separation bound reproduces the Helstrom optimal success probability, showing that quantum mechanics attains the maximal performance among all generalized probabilistic theories (GPTs) consistent with contextuality.

The main contribution lies in extending the same reasoning to quantum cloning. In both phase‑covariant cloning (restricted to equatorial qubit states) and universal cloning (all pure qubit states), the cloning task is recast as a prepare‑measure scenario where the fidelity between the actual bipartite output and the ideal double copy determines the relevant statistics. The corresponding COPE matrices again exhibit the required sparsity pattern, leading via Theorem 3 to a lower bound of four on the rank of R or E, while the actual matrix rank is limited by the optimal fidelity expression (e.g., 2d − 1 for d‑dimensional systems). This rank mismatch proves rank separation, implying that optimal cloning machines cannot be described by an ENMF. By invoking Theorems 1 and 2, the authors conclude that any optimal deterministic approximate cloning machine—whether phase‑covariant or universal—necessarily requires contextuality. This resolves the open problem left by earlier work that had only shown contextuality for state‑dependent cloning.

Finally, the paper highlights the broader relevance of rank‑separation: the same technique can be applied to any GPT, and the derived dimensional bounds guarantee that quantum theory provides the optimal success probabilities for the examined tasks. The work thus furnishes a powerful, model‑independent tool for certifying contextuality directly from experimental data, and establishes contextuality as the fundamental non‑classical resource underlying all known optimal quantum cloning protocols.