⚛️ quantum physics

A Separation between Full-Rank PVM and Assumption-free Self-Testing

This paper constructs a nonlocal game that self-tests a maximally entangled qubit strategy under full-rank projective measurement assumptions while admitting an inequivalent optimal strategy using nonprojective measurements, thereby resolving a conjecture by Baptista et al. and demonstrating that such full-rank PVM self-tests are not robust.

Original authors: Ranyiliu Chen

Published 2026-09-10

📖 8 min read🧠 Deep dive

CC BY 4.0

Original authors: Ranyiliu Chen

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the strange and counterintuitive world of quantum physics, particles can become linked in a way that defies our everyday experience of space and time. When two particles share this deep connection, known as entanglement, measuring one instantly reveals information about the other, no matter how far apart they are. Scientists have long sought a way to verify that these particles are truly entangled and behaving exactly as quantum theory predicts, without needing to trust the equipment used to measure them. This process is called self-testing. It works like a cryptographic lock: by playing a specific game with the particles and checking the results, an observer can be mathematically certain of the internal state of the system, even if the devices themselves are black boxes. For this verification to be useful in the real world, it must be robust, meaning that even if the measurements are slightly imperfect, the conclusion about the quantum state remains valid.

A researcher has now uncovered a subtle but critical flaw in how we might try to certify these quantum systems. They constructed a specific game that reveals a surprising gap between two different ways of looking at quantum measurements. In the idealized world of quantum mechanics, we often assume that measurements are "projective," meaning they act like a perfect filter that cleanly separates possibilities, and that the particles involved are in a "full-rank" state, meaning every possible configuration of the system is active and accessible. The researcher proved that if you insist on both of these conditions simultaneously, you can certify a specific, simple quantum strategy. However, they also showed that if you remove the requirement for perfect projectivity, a completely different, equally successful strategy emerges that looks identical in terms of the game's score but is fundamentally different in its internal mechanics. This discovery resolves a long-standing question about whether these two assumptions could be combined without losing information, and it demonstrates that the standard method for certifying these systems is not as reliable as previously thought when faced with slight imperfections.

The researcher, led by Ranyiliu Chen, designed a complex game to test these ideas. The game involves two players, Alice and Bob, who share a pair of entangled particles. They receive questions and must provide answers based on how they measure their particles. The goal is to win as often as possible. The researcher combined two well-known types of games into one larger challenge. The first part is a classic test of quantum entanglement known as the CHSH game, which is famous for proving that the universe is not governed by local hidden variables. The second part is a newly constructed auxiliary game that behaves differently depending on the nature of the measurements used.

In the ideal scenario, where the players use perfect, projective measurements and their particles are in a fully active state, the only way to win the game at the maximum possible rate is to use a specific, simple strategy. This strategy involves the players always giving a specific "abort" answer when asked certain questions, effectively ignoring the quantum complexity in those moments. Under these strict conditions, the game successfully self-tests the system, confirming that the players are using the intended entangled state and measurements.

However, the researcher discovered that if they relaxed the rule requiring the measurements to be perfect projectors, a different path to the same maximum score opened up. In this alternative strategy, the players use a more complex type of measurement known as a positive operator-valued measure, or POVM. This type of measurement allows for outcomes that are not sharp, clean cuts but rather fuzzy, overlapping possibilities. Using this fuzzy measurement, the players can achieve the exact same winning score as the perfect strategy, but they never choose to abort. Instead, they provide a variety of distinct, non-zero answers.

The crucial finding is that these two strategies, while achieving the same score, are fundamentally incompatible. The first strategy relies on the players effectively turning off their quantum devices for certain questions, while the second strategy keeps them fully active with a different kind of measurement. Because the internal mechanics are so different, it is mathematically impossible to transform the second strategy into the first, even if you allow for extra hidden variables or auxiliary systems. This means that if you only look at the final score of the game, you cannot tell which strategy the players are actually using. The game certifies the entangled state in both cases, but it fails to certify the complete measurement strategy when the strict projective assumption is dropped.

This separation has a profound consequence for the reliability of quantum certification. The researcher showed that the game is not robust. In a robust test, a strategy that is very close to the optimal one should be easily convertible into the ideal strategy. Here, the researcher constructed a family of strategies that are arbitrarily close to the optimal score but remain stuck in the "fuzzy" measurement mode. No matter how close the score gets to the maximum, the internal structure of the measurement remains fundamentally different from the ideal projective version. The error in transforming these near-optimal strategies into the ideal one does not shrink; it stays large. This proves that the self-testing result is fragile and cannot be trusted if there is any possibility of non-projective measurements being involved.

The work also provides a deep algebraic explanation for why this happens. The researcher analyzed the mathematical structure of the optimal strategies and found that the set of all possible winning correlations forms a continuous line. One end of this line represents the simple, deterministic strategy where players always abort. The other end represents the complex, fuzzy strategy where players never abort. The middle of the line contains mixtures of both. When the researcher imposed the condition that measurements must be projective, the mathematical structure collapsed, leaving only the simple, deterministic end. The complex, fuzzy end disappeared because it could not exist under the strict projective rules. This collapse explains why the two strategies are so distinct: they belong to different mathematical worlds that only touch at the very edge of possibility.

The implications of this finding extend beyond the specific game constructed. It challenges the assumption that requiring measurements to be projective is a harmless simplification. In many quantum protocols, scientists assume that if a strategy works well, it can be approximated by a projective one. This paper shows that for certain tasks, this approximation fails completely. The "fuzzy" measurements are not just a slightly imperfect version of the perfect ones; they are a distinct, optimal solution that cannot be reached by simply tweaking the perfect strategy. This suggests that in the design of future quantum networks and cryptographic systems, engineers must be extremely careful about the assumptions they make regarding the nature of their measurements.

The researcher used a specific mathematical framework to prove these results, relying on the properties of operator algebras and the behavior of quantum states under compression. They demonstrated that the "fuzzy" measurement strategy corresponds to a state that is not fully active in the way a standard projective strategy would be. When you try to force this state into a projective mold, you lose essential information about the system's behavior. The paper concludes that while we can still certify the entangled state itself, we cannot certify the full measurement strategy under the joint assumptions of full rank and projectivity. The separation between the two types of strategies is exact and absolute, not a matter of degree.

This work resolves a conjecture made by other researchers who suspected that combining the full-rank and projective assumptions might lead to such obstructions. By providing an explicit example, the author has moved the field from speculation to proof. They have shown that the landscape of quantum strategies is more complex than previously mapped, with hidden valleys that look like peaks from a distance but are actually distinct terrains. For anyone relying on quantum self-testing to guarantee the security or functionality of a system, this is a vital warning: the rules of the game matter as much as the score. If the rules allow for fuzzy measurements, the ideal, sharp strategy might not be the only winner, and the system might not be what it appears to be.

The study does not suggest that quantum mechanics is broken or that entanglement is unreliable. Rather, it highlights the precision required in defining what we mean by a "measurement" and a "state." The researcher has built a tool that can distinguish between two very different ways of playing the quantum game, revealing that our current methods for verifying quantum devices have blind spots. As quantum technology moves from the laboratory to real-world applications, understanding these subtle distinctions will be essential for building systems that are truly secure and reliable. The paper stands as a rigorous demonstration that in the quantum realm, the path to the solution is just as important as the solution itself, and sometimes, the most direct route is not the only one that works.

Technical Summary: Separation between Full-Rank PVM and Assumption-free Self-Testing

Problem Statement
Self-testing allows for the certification of entangled states and measurements from the optimal value of a nonlocal game, up to local isometries and auxiliary states. Standard self-testing theorems often rely on specific assumptions about the strategies, such as the shared state being pure with full Schmidt rank and the measurements being projective (PVMs). While Naimark dilation allows converting positive operator-valued measurements (POVMs) to projective ones, and compression to Schmidt supports restores full rank, these operations do not simultaneously preserve both properties in a way that satisfies the strict requirements of exact self-testing.

Baptista et al. [1] conjectured that this obstruction persists even in exact self-testing problems: a reference strategy satisfying both full-rank and projectivity assumptions might be self-tested within the restricted class of strategies satisfying both, yet fail to be self-tested in the unrestricted class of all pure strategies. This paper addresses the question of whether imposing full rank and projectivity simultaneously creates a genuine separation in self-testing conclusions, distinct from the cases where only one assumption is imposed.

Methodology
The author constructs an explicit nonlocal game, denoted $H$ , to resolve this conjecture. The construction proceeds in two main stages:

Auxiliary Game $G$ : The author first designs a game $G$ with four questions for Alice and three for Bob. The game is designed such that:
- Under the joint assumptions of a pure full-Schmidt-rank state and projective measurements (PVMs), the unique optimal strategy is a deterministic "abort" strategy ( $S_0$ ) where players always output a specific "abort" answer.
- Without these joint assumptions, there exists a second optimal strategy ( $S_1$ ) using a maximally entangled qubit state $|\Phi^+\rangle$ and a specific set of measurements, including a non-projective three-outcome POVM (a "trine" POVM). This strategy never aborts.
- The construction utilizes a three-setting Bell expression combined with penalty terms for abort events. An exact identity is derived where the gap from the optimal score is a sum of non-negative terms. Under full-rank and projective assumptions, the equality conditions force incompatible operator relations on any non-zero non-abort subspace, eliminating the trine strategy.
Main Game $H$ : The game $H$ is formed by a probabilistic mixture of the auxiliary game $G$ and the standard CHSH game. The referee chooses between $G$ and CHSH with equal probability, using disjoint question sets.
- The reference strategy $S_H$ uses the maximally entangled state $|\Phi^+\rangle$ , standard CHSH measurements for the CHSH questions, and the deterministic abort strategy for the $G$ questions.
- The CHSH component ensures that any optimal strategy must share a Bell state and satisfy CHSH measurement relations.
- The $G$ component forces the measurement behavior on the remaining questions.

Key Contributions and Results

Resolution of the Conjecture: The paper proves the conjecture of Baptista et al. by demonstrating that the game $H$ self-tests the reference strategy $S_H$ within the class of pure full-Schmidt-rank PVM strategies, but admits an inequivalent optimal strategy in the unrestricted class.
- Theorem 1.1: There exists a 0/1 game $H$ $H$ with a projective reference strategy $S_H$ $S_{H}$ on $\mathbb{C}^2 \otimes \mathbb{C}^2$ $C^{2} \otimes C^{2}$ such that:
  1. The classical value is $\omega_c(H) = 71/88$ and the quantum value is $\omega_q(H) = (60 + 11\sqrt{2})/88$ .
  2. Every optimal pure full-Schmidt-rank PVM strategy admits a local dilation to $S_H$ .
  3. There exists an optimal pure full-Schmidt-rank strategy (using one non-projective measurement) that admits no local dilation to $S_H$ .
Algebraic Classification: The author provides a complete algebraic analysis of the optimal correlations of the auxiliary game $G$ .
- After compression to the local state supports, the universal unital $C^*$ -algebra generated by the optimality relations is isomorphic to $\mathbb{C} \oplus M_2(\mathbb{C})$ .
- The scalar summand corresponds to the deterministic abort behavior, while the matrix summand corresponds to the trine POVM behavior.
- The set of compatible local states corresponds bijectively to the tracial states of this algebra, parametrized by $\lambda \in [0, 1]$ , where $\lambda$ represents the probability of non-abort.
- Imposing projectivity on the state support eliminates the matrix summand (as the ideal generated by the non-projective effects becomes the entire matrix summand), leaving only the scalar (deterministic) optimum.
Failure of Robustness: The paper demonstrates that the self-test for $H$ within the restricted class is not robust.
- By perturbing the local state of a projective dilation of the trine strategy, the author constructs a family of full-rank PVM strategies with scores arbitrarily close to the optimal value but whose local-dilation error to the reference strategy remains at least 1.
- This indicates that the separation between the restricted and unrestricted optima is exact but not stable under small perturbations of the score.

Significance
The paper establishes that the assumptions of full Schmidt rank and projectivity are not merely technical conveniences that can be removed without consequence; rather, their simultaneous imposition fundamentally alters the set of optimal strategies in a way that prevents the self-testing of certain strategies in the unrestricted class. Specifically, it shows that a game can certify a complete strategy (state and measurements) under joint assumptions while failing to do so when the projectivity constraint is lifted, even if the state remains pure and full-rank.

The work highlights the distinction between state equations (relations holding on the shared state vector) and operator relations (relations holding as operator identities). In the example provided, the state equations allow for a non-projective realization that is optimal, but the operator relations required for a projective realization are incompatible with the non-projective effects. This distinction is formalized through the analysis of zero left ideals in the measurement algebra, showing that for optimal projective realizations with non-zero non-abort probability, the local state is neither faithful nor tracial.

The results clarify the limitations of self-testing when multiple structural assumptions are combined and provide a concrete counterexample to the idea that restricting the strategy class always yields a "stronger" or more complete certification in the absence of those restrictions. The paper concludes by noting that while the Bell state is extracted in all optima, the certification of the complete strategy (including the specific measurement effects) is the property that separates the restricted and unrestricted cases.

Technical Summary: Separation between Full-Rank PVM and Assumption-free Self-Testing

More like this