Incompatibility in restricted operational theories: connecting contextuality and steering

Martin PlávalaJournal of Physics A Mathematical and Theoretical 2022 · DOI

TL;DR

This paper explores the deep connections between fundamental quantum phenomena like contextuality and steering within the broad framework of general probabilistic theories (GPTs). The authors establish a crucial link: the steerability of a bipartite quantum state by a set of measurements is equivalent to the non-existence of a specific type of "preparation noncontextual hidden variable model" for a restricted theory derived from that state and those measurements. This work provides a unified perspective on these concepts, offering new insights into the non-classical nature of quantum mechanics.

What Problem Does This Paper Address?

Quantum theory, despite its predictive power, often lacks intuitive interpretations. One avenue to understanding its peculiarities is to hypothesize a deeper, classical "hidden variable" theory that could explain quantum predictions. However, Bell's theorem and the Kochen-Specker theorem have shown that such classical explanations are incompatible with quantum mechanics under certain assumptions (locality and noncontextuality, respectively).

This paper delves into these questions within the more general framework of General Probabilistic Theories (GPTs), which encompass classical and quantum theories, as well as other exotic possibilities like PR boxes. Specifically, it investigates the relationship between:

  • Steering: A phenomenon where one party (Alice) can influence the state of another party's (Bob's) system, even if Bob's system is spatially separated, by performing measurements on her own system. This is a one-way quantum information task.
  • Contextuality: The idea that the outcome of a measurement can depend on which other compatible measurements are performed alongside it. This is a hallmark of quantum mechanics, distinguishing it from classical probability.
  • Incompatibility: Two measurements are incompatible if they cannot be performed simultaneously. This is a fundamental property of quantum measurements.
  • Hidden Variable Models: Classical theories that aim to explain the statistics of quantum experiments by introducing underlying, unobserved variables. The paper focuses on "preparation noncontextual" models, where the distribution of these hidden variables depends only on the prepared state, not on the specific measurement chosen.

The paper aims to bridge the gap between these concepts by showing how incompatibility in restricted theories (theories where not all possible states or measurements are allowed) directly relates to contextuality and, ultimately, to the steerability of quantum states.

Methods

The paper builds a formal framework for analyzing these concepts within restricted GPTs. Key methodological elements include:

  • Formalism of GPTs: The paper starts by reviewing the standard mathematical framework for GPTs, defining state spaces as convex sets and effects as affine functions representing measurement outcomes.
  • Restricted GPTs: A crucial innovation is the introduction of "restricted GPTs," which model scenarios where only a subset of states and measurements are available. This is achieved by defining a theory by a pair $(K, E)$, where $K$ is the set of preparable states and $E$ is the set of allowed effects (two-outcome measurements).
  • Hidden Variable Models in GPTs: The paper defines three types of hidden variable models: preparation noncontextual, measurement noncontextual, and preparation and measurement noncontextual. These models are framed as classical theories that can reproduce the statistics of experiments in a given GPT.
  • Incompatibility Definitions: Two distinct definitions of measurement compatibility are introduced for restricted theories: "E-compatibility" (where joint measurement effects must be in the allowed set $E$) and "E(K)-compatibility" (where joint measurement effects can be in the larger set $E(K)$ of all possible effects).
  • Connecting Concepts: The core of the paper lies in proving equivalences between these different concepts. For instance, it shows that all measurements in a restricted theory are E-compatible if and only if its state space is a simplex. A key result (Proposition 2) establishes that all measurements in a restricted theory are E(K)-compatible if and only if there exists a preparation noncontextual hidden variable model for that theory.
  • Steering Scenario Construction: The paper then constructs a specific restricted theory corresponding to a given steering scenario. This involves defining a restricted state space $K_\rho$ based on a bipartite state $\rho_{AB}$ and a set of Alice's measurements.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Incompatibility as a Bridge: The paper highlights that incompatibility, particularly in restricted theories, acts as a crucial bridge connecting contextuality and steering.
  • Restricted Theories are Key: The framework of restricted GPTs is essential for establishing the link between steering and noncontextual hidden variable models. It allows for a more nuanced analysis than considering only theories with no restrictions.
  • Steerability $\iff$ Non-existence of Preparation Noncontextual HV Model: The central finding is that for a specific class of bipartite states and measurements, the steerability of the state is precisely equivalent to the absence of a preparation noncontextual hidden variable model for the induced restricted theory. This is a powerful conceptual unification.
  • Hierarchy of Compatibility: The paper establishes a hierarchy: E-compatibility implies E(K)-compatibility, and simplex-embeddability implies E(K)-compatibility. However, the converse implications do not always hold, revealing subtle distinctions.
  • Contextuality and Incompatibility are Deeply Linked: The work demonstrates an inherent connection between incompatibility and contextuality in all GPTs, not just quantum theory.

Key Results

  • Proposition 1: All measurements in a restricted theory $(K, E)$ are E-compatible if and only if $S(E)$ (the state space of $E$) is a simplex.
  • Proposition 2: All measurements in a restricted theory $(K, E)$ are E(K)-compatible if and only if there exists a preparation noncontextual hidden variable model for $(K, E)$.
  • Theorem 1:
    • If all measurements in $(K, E)$ are E-compatible, then $(K, E)$ is simplex-embeddable.
  • If $(K, E)$ is simplex-embeddable, then all measurements in $(K, E)$ are E(K)-compatible.
  • Theorem 2: For a KB-full-dimensional bipartite state $\rho_{AB}$ and a corresponding restricted theory $(K_\rho, E_m)$, $\rho_{AB}$ is not steerable by the set of measurements $\{m_x\}$ if and only if there exists a preparation noncontextual hidden variable model for $(K_\rho, E_m)$.
  • Examples: The paper provides examples (e.g., square-in-square, square-in-tetrahedron) to illustrate the distinctions between different types of compatibility and simplex-embeddability, and to show that the inclusion in Theorem 1(i) is strict.

Limitations and Future Work

  • Focus on Preparation Noncontextuality: The main results primarily concern preparation noncontextual hidden variable models. The paper suggests searching for operational equivalents of measurement noncontextual hidden variable models, similar to the results for preparation noncontextuality.
  • GPTs and Noncontextual Models: It remains an open question whether measurement noncontextual hidden variable models exist for all GPTs, or only for those with simplex state spaces (analogous to Corollaries 1 and 2 for preparation noncontextuality).
  • Full-Dimensionality Requirement: Theorem 2 has a requirement of KB-full-dimensionality for the bipartite state. The paper briefly discusses how this might be relaxed in some cases, but a full treatment is left for future work.
  • Operational Equivalents: The authors explicitly state that finding operational equivalents for measurement noncontextual hidden variable models is a key future direction.

Summary updated

Abstract

Abstract We investigate the connection between steering and contextuality in general probabilistic theories. We show that for a class of bipartite states the steerability of the state by given set of measurements is equivalent to non-existence of preparation noncontextual hidden variable model for certain restricted theory constructed from the given state and measurements. The connection between steering and contextuality is provided by the concept of incompatibility in restricted theories, which we also investigate.
BibTeX
@article{plavala2022incompatibility,
  author = {Martin Plávala},
  title = {{Incompatibility in restricted operational theories: connecting contextuality and steering}},
  journal = {Journal of Physics A Mathematical and Theoretical},
  year = {2022},
  doi = {10.1088/1751-8121/ac5afe},
}
Figure