⚛️ quantum physics

An Unbounded Separation of Almost-Quantum and Commuting-Operator Correlations in Zero-Error Communication

This paper demonstrates that almost-quantum correlations, despite being characterized by a single semidefinite program and satisfying various physical principles, are unboundedly stronger than both finite-dimensional and commuting-operator quantum correlations in zero-error equality testing tasks, achieving significantly higher success probabilities and drastically reducing communication costs.

Original authors: Rafiuddin Gazi, Biswadeep Chatterjee, Jayashree Karmakar, Ram Krishna Patra, Manik Banik, Anandamay Das Bhowmik

Published 2026-11-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Rafiuddin Gazi, Biswadeep Chatterjee, Jayashree Karmakar, Ram Krishna Patra, Manik Banik, Anandamay Das Bhowmik

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 realm of physics, there is a fundamental limit to how much two distant people can coordinate their actions without talking to each other. Imagine two friends, Alice and Bob, who are far apart. They share a mysterious connection, perhaps a pair of particles that were created together and then separated. When they each make a choice about what to measure, their results are linked in a way that seems impossible if they were just following a shared script written before they left. This phenomenon, known as quantum entanglement, allows them to share information in a way that classical physics cannot explain. For decades, scientists have wondered if the rules of quantum mechanics are the absolute limit of this coordination, or if there are even stranger, more powerful connections that nature might allow but simply hasn't been observed yet. To test this, researchers often look for tasks where these hypothetical, stronger-than-quantum connections would give a massive advantage over standard quantum ones. If such an advantage exists, it would prove that our current understanding of the universe is incomplete.

A team of researchers has now found a specific task where these hypothetical connections, called "almost-quantum" correlations, vastly outperform everything we know about quantum mechanics. The task is a simple game of checking for equality. Alice and Bob each receive a secret number from a large list. Alice is allowed to send Bob a single bit of information—a simple "yes" or "no" message. Bob must then decide if their numbers are the same. He is not allowed to make a mistake; if he says they are the same, they must be. However, he is allowed to say "I don't know" if he is not sure. The goal is to maximize the chance that he gives a definite answer without ever being wrong.

The researchers proved that for any standard quantum strategy, whether using simple particles or the complex mathematics of quantum field theory, the best Bob can do is to give a correct answer with a probability that shrinks as the list of numbers gets longer. Specifically, if there are $N$ possible numbers, the best chance of a correct, definite answer is exactly one in $N$ . This is the same result Alice and Bob would get if they just shared a random number beforehand and used that to guess. Quantum entanglement, no matter how strong or how many dimensions it involves, offers no help here. It cannot beat the simple strategy of shared randomness.

However, when the researchers allowed Alice and Bob to use these "almost-quantum" connections, the result changed dramatically. They constructed a specific set of rules for these connections that allowed Bob to give a correct, definite answer with a probability of at least one in nineteen, regardless of how large the list of numbers became. Furthermore, they found that the optimal probability for these connections actually improves slightly as the list grows, eventually settling at a limit of one in seventeen. While one in nineteen might sound small, the difference is profound because the quantum limit gets worse and worse as the list grows, while the almost-quantum limit stays fixed at a much higher value. If the list has a thousand numbers, the quantum chance is one in a thousand, but the almost-quantum chance remains near one in seventeen. The gap between the two grows without bound. This means that for large enough lists, the almost-quantum strategy is infinitely better than any quantum strategy could ever be.

The study also explored what happens if Alice is allowed to send more than just one bit. Even with longer messages, quantum entanglement saves at most one bit of communication compared to shared randomness. In contrast, the almost-quantum connections can reduce the communication cost of checking if two long strings of data are equal from many bits down to a fixed, small number. This suggests that if nature allows these stronger correlations, the rules of information processing would be fundamentally different from what we currently observe.

Crucially, the researchers did not just guess that these stronger connections exist; they built a mathematical model that proves they satisfy many of the same basic rules that quantum mechanics follows. These almost-quantum correlations obey principles like "macroscopic locality," which says that large groups of particles should behave classically, and they pass tests that were thought to rule out anything stronger than quantum mechanics. Yet, in this specific game of equality, they break the quantum limit. This finding challenges the idea that a single, simple principle can explain why the universe is limited to quantum mechanics. It shows that even if a theory passes many standard tests, it might still allow for capabilities that are far beyond what we see in our world.

The researchers also identified a clear mathematical signature that distinguishes these almost-quantum connections from standard quantum ones. They found a specific inequality that any quantum system must obey, but which these almost-quantum systems violate. This violation grows larger as the number of possible inputs increases, providing a clear, measurable way to tell the two apart. This work does not claim that such stronger-than-quantum connections actually exist in nature, but it proves that they are logically consistent with many of the principles we use to describe the physical world. It leaves open the possibility that the universe is more powerful than quantum mechanics suggests, or that there is some deeper principle yet to be discovered that prevents these stronger connections from appearing in reality. For now, the result stands as a sharp boundary: in the game of zero-error equality, quantum mechanics is strictly weaker than the almost-quantum possibilities it was thought to nearly exhaust.

Technical Summary: An Unbounded Separation of Almost-Quantum and Commuting-Operator Correlations in Zero-Error Communication

Problem Statement
The paper investigates the operational gap between quantum correlations and "almost-quantum" (AQ) correlations within the framework of zero-error communication complexity. While AQ correlations form the $1+AB$ level of the Navascués–Pironio–Acín (NPA) hierarchy and satisfy several physical principles proposed to single out quantum theory (such as macroscopic locality and local orthogonality), it remains an open question whether they are operationally indistinguishable from quantum correlations in specific tasks. The authors focus on a zero-error certification of equality task: Alice and Bob receive labels $s, t \in [N]$ and must decide if $s=t$ . Alice sends a single classical bit to Bob, who must output "equal," "different," or "abstain." The protocol must never err (zero error), and the goal is to maximize the worst-case probability of a conclusive answer.

Methodology
The authors analyze this task across a hierarchy of correlation sets:

Local Hidden Variables (CL)
Finite-dimensional Quantum (Cq)
Spatial Quantum (Cqs)
Approximately Quantum (Cqa)
Commuting-Operator Quantum (Cqc)
Almost-Quantum (CAQ)
Vector Correlations (CVec, Level 1 of NPA)
Nonsignalling (CNS)

The methodology involves:

Quantum Bounds: Proving that for all quantum models (including infinite-dimensional and commuting-operator frameworks), the optimal worst-case probability of a conclusive answer is exactly $1/N$ . This is achieved using a "packing lemma" based on sequential measurements, which demonstrates that Bob's acceptance events for different inputs must occupy orthogonal subspaces of Alice's reduced state (or the shared Hilbert space).
Almost-Quantum Constructions: Constructing explicit families of AQ boxes ( $p^\star$ ) that violate the quantum bound. The authors reduce the membership of these boxes in the AQ set to a semidefinite program (SDP) involving vectors with prescribed inner products. They solve these SDPs in closed form to determine the optimal AQ performance.
Vector Correlation Analysis: Solving the corresponding optimization problem for Level 1 of the NPA hierarchy (CVec) to establish a separation between CVec and CAQ.
Witness Construction: Deriving a linear inequality (a "face witness") that holds for all commuting-operator correlations on a specific face of the nonsignalling polytope but is violated by AQ correlations.

Key Contributions and Results

Unbounded Separation between AQ and Quantum:
- For every $N \ge 3$ , the optimal efficiency for all quantum models (from finite-dimensional tensor products to commuting operators) is exactly $\eta_Q(N) = 1/N$ . This value is already attained by shared randomness, implying entanglement offers no advantage for this specific zero-error task.
- In contrast, an explicit family of AQ correlations achieves a constant efficiency of at least $1/19$ for all $N$ .
- The ratio of AQ performance to quantum performance, $\eta_{AQ}/\eta_Q$ , grows linearly with $N$ (specifically $\ge N/19$ ), demonstrating an unbounded separation.
- For binary-outcome boxes certifying equality, the optimal AQ efficiency is exactly $1/N$ for $N \le 8$ and $N^2/(17N^2 - 80N + 64)$ for $N \ge 8$ , tending to $1/17$ as $N \to \infty$ . This exceeds the quantum bound precisely when $N \ge 9$ .
Communication Cost Reduction:
- For longer messages (deciding equality of $n$ -bit strings, $N=2^n$ ), entanglement-assisted protocols (even with commuting operators) save at most one bit of communication compared to shared randomness.
- AQ correlations reduce the one-way zero-error cost of deciding equality from $n$ bits to at most 13 bits, regardless of $n$ . This represents an exponential gap in communication complexity.
Separation of NPA Hierarchy Levels:
- The paper separates the AQ set (Level $1+AB$ ) from the Level 1 set (CVec). The optimal efficiency for Level 1 is $N/[8(N-2)]$ , tending to $1/8$ .
- Since $1/8 > 1/17$ , the optimal Level 1 boxes lie outside the AQ set for all $N \ge 5$ , providing an analytic separation of these two sets in an operational task.
Post-Quantumness Witness:
- The authors derive a linear inequality (Eq. 21): $\sum_{x=1}^N p(0,0|x,x) \le 1$ . This holds for all commuting-operator correlations on the face where off-diagonal coincidences vanish ( $F_0$ ).
- The constructed AQ boxes violate this inequality, with the violation growing linearly with $N$ . This serves as an exact-zero (face) witness of post-quantumness.

Significance and Claims
The paper claims to demonstrate that the difference between almost-quantum and quantum correlations is not negligible in operational terms, despite AQ correlations satisfying many principles intended to characterize quantum theory.

Operational Distinction: The separation arises specifically at the boundary between commuting-operator correlations (where operators commute as identities) and AQ correlations (where commutation is only required on the shared state). The quantum bound relies on a sequential-measurement argument requiring commutation with products of operators, a condition satisfied by Cqc but not by CAQ.
Independence from Tsirelson's Problem: The result holds regardless of the resolution of Tsirelson's problem (whether $C_{qa} = C_{qc}$ ), as the quantum bound applies to the entire commuting-operator set.
Limitations of Principles: The findings suggest that principles like macroscopic locality and local orthogonality, which are satisfied by AQ correlations, are insufficient to rule out post-quantum behaviors in zero-error communication tasks.
Information Causality: The authors note in a "Note added" that a companion work [31] uses the boxes constructed here to show that AQ correlations violate information causality without bound, further distinguishing them from quantum theory.

The paper concludes that while AQ correlations are a powerful relaxation of quantum theory, they possess a distinct operational power in zero-error communication that quantum correlations cannot match, highlighting a structural obstruction in the commutation properties of the underlying operators.

More like this