PaperlayerView exact revision on arXiv

Source-linked AI summary

The Sheaf-Theoretic Structure Of Non-Locality and Contextuality

Samson Abramsky, Adam Brandenburger

arXiv:1102.0264v7quant-phcs.LOmath.CT

TL;DR

The paper asks how non-locality and contextuality can be treated within one general framework beyond standard probability tables and tensor-product scenarios. It uses sheaf theory and linear algebra to characterize them through global sections, establishes a hierarchy of no-go theorems and a negative-probability/no-signalling correspondence, and proves generalized no-signalling for quantum mechanics with arbitrary commuting observables. The framework also supports model-independent Kochen-Specker results and a generalized notion of maximal contextuality.

  • Problem

    Non-locality and contextuality require a unified treatment that applies beyond standard non-locality scenarios to arbitrary measurement covers, including Kochen-Specker configurations.

  • Method

    The paper uses sheaf theory and linear algebra over arbitrary measurement covers to analyze global sections, hidden-variable realizations, contextuality, and no-signalling.

  • Results

    The paper identifies non-locality and contextuality with obstructions to global sections, orders Bell, Hardy, and GHZ in a strict hierarchy, and proves negative-probability/no-signalling equivalence plus generalized quantum no-signalling.

  • Takeaways & Limitations

    The framework gives a model-independent account of contextuality and non-locality and extends no-signalling from tensor-product scenarios to arbitrary families of commuting observables.

  • Takeaways & Limitations

    Quantum-representable measurement covers are restricted to those generated by symmetric binary compatibility relations, so some graphs, such as the triangle, cannot arise from quantum observables.

Abstract

from arXiv · show

We use the mathematical language of sheaf theory to give a unified treatment of non-locality and contextuality, in a setting which generalizes the familiar probability tables used in non-locality theory to arbitrary measurement covers; this includes Kochen-Specker configurations and more. We show that contextuality, and non-locality as a special case, correspond exactly to obstructions to the existence of global sections. We describe a linear algebraic approach to computing these obstructions, which allows a systematic treatment of arguments for non-locality and contextuality. We distinguish a proper hierarchy of strengths of no-go theorems, and show that three leading examples --- due to Bell, Hardy, and Greenberger, Horne and Zeilinger, respectively --- occupy successively higher levels of this hierarchy. A general correspondence is shown between the existence of local hidden-variable realizations using negative probabilities, and no-signalling; this is based on a result showing that the linear subspaces generated by the non-contextual and no-signalling models, over an arbitrary measurement cover, coincide. Maximal non-locality is generalized to maximal contextuality, and characterized in purely qualitative terms, as the non-existence of global sections in the support. A general setting is developed for Kochen-Specker type results, as generic, model-independent proofs of maximal contextuality, and a new combinatorial condition is given, which generalizes the `parity proofs' commonly found in the literature. We also show how our abstract setting can be represented in quantum mechanics. This leads to a strengthening of the usual no-signalling theorem, which shows that quantum mechanics obeys no-signalling for arbitrary families of commuting observables, not just those represented on different factors of a tensor product.

1. Introduction

The paper develops a general sheaf-theoretic framework that unifies non-locality and contextuality as structural obstructions to global sections. It establishes a hierarchy of no-go strengths, connects hidden-variable models with no-signalling, and extends the framework to quantum mechanics.

  • 1. Introduction: The paper treats non-locality and contextuality uniformly through a mathematical setting independent of Hilbert space, with standard formulations recovered as special cases.Its measurement-cover framework generalizes probability tables and subsumes both notions.
  • 1. Introduction: Strong contextuality, possibilistic non-locality, and probabilistic non-locality form a strict hierarchy, exemplified successively by GHZ, Hardy, and Bell.The implications run from strong contextuality to possibilistic non-locality to probabilistic non-locality, while converses fail.
  • 1. Introduction: The framework establishes an equivalence between local hidden-variable realizations using negative probabilities and the no-signalling property.The result is based on applying linear algebraic methods to construct global sections.
  • 1. Introduction: Global sections provide the canonical criterion: factorizable hidden-variable realizations are strictly equivalent to compatible global sections of a presheaf.Factorizability includes Bell-locality and distribution-level non-contextuality.
  • 1. Introduction: Quantum mechanics satisfies a generalized no-signalling property for arbitrary families of commuting observables, beyond observables represented on different tensor-product factors.Compatibility of sections on a presheaf corresponds to this generalized no-signalling condition.

2. The Setting

The setting models experiments through measurements, jointly performable contexts, events, and distributions, with sheaf structure formalizing compatible local assignments.

  • A measurement scenario specifies a finite set of measurements X, possible outcomes O, contexts M, and a distribution on events for each context.
  • Events: A section over U assigns an outcome in O to every measurement in U, representing the outcomes observed for those measurements.
  • Events: The event assignment E maps each measurement set U to its sections and acts by restricting assignments to smaller subsets.
  • Events: Compatible local sections agreeing on overlaps glue uniquely into a section over the entire covered set, which is the sheaf condition.
  • Events: The event sheaf condition is straightforward because compatible partial functions can be combined by taking the union of their graphs.

2.3. Distributions

The paper generalizes distributions by parameterizing event weights with a commutative semiring and defines restriction as marginalization.

  • Distributions on events capture statistical rather than deterministic behavior while allowing the algebra of weights to vary.
  • A commutative semiring provides additive and multiplicative commutative monoids, with multiplication distributing over addition.
  • An R-distribution is a finitely supported function whose weights satisfy the distribution normalization condition.Finite support ensures that the defining sum is well-defined.
  • The nonnegative reals give finite-support probability distributions, booleans give nonempty finite subsets, and reals give finite-support signed measures allowing negative probabilities.
  • The presheaf DRE assigns distributions to sections, and restriction maps a distribution to the marginal obtained by summing weights over extensions.

2.4. Measurement Covers

Measurement covers formalize which measurements can be performed jointly, generalizing Bell scenarios to arbitrary compatibility structures and compatible empirical models.

  • Measurement covers: A measurement cover is an antichain of subsets whose union is X, with each maximal subset representing a jointly performable measurement context.
  • Measurement covers: Measurement covers represent both Bell-type non-locality scenarios and Kochen–Specker contextuality configurations.
  • Bell-type covers: For spatially separated parts, contexts contain exactly one measurement from each part, allowing measurements at different sites but not alternatives at the same site.
  • Empirical models: In Bell-type scenarios, compatibility of context distributions is equivalent to standard no-signalling: one party’s marginal is independent of the other party’s measurement choice.
  • Empirical models: When contexts have empty intersection, their compatibility condition is trivially satisfied.
  • Empirical models: The compatible-family definition also applies to arbitrary covers, including empirical models arising from quantum systems with commuting observables.

2.6. Examples

The formalism instantiates familiar non-locality models as empirical models over Bell-type measurement contexts and supports both probabilistic and possibilistic representations.

  • In the bipartite Bell scenario, Alice and Bob each choose between two measurements, with outcomes 0 or 1.
  • The four maximal contexts are {a, b}, {a′, b}, {a, b′}, and {a′, b′}, indexing the table rows.
  • Each table row contains sections for a context, such as the assignment a 7→1 and b 7→0 for {a, b}.
  • The table assigns a weight to each section; probabilistic models use nonnegative real weights summing to 1 along each row.
  • The row distributions collectively form an empirical model, and their compatibility is exactly the no-signalling condition.
  • The formalism also includes a possibilistic non-local Hardy model over the boolean semiring, whose support can represent a standard probabilistic Hardy model.

3. Global Sections

The paper identifies global sections as context-independent assignments whose distributions reproduce empirical observations, making their non-existence the obstruction underlying non-locality and contextuality.

  • Global assignments: A global assignment gives every measurement a definite outcome independent of context, yielding a deterministic hidden variable.Each assignment restricts to an assignment on every measurement context.
  • Global distributions: A distribution over global assignments reproduces empirical probabilities by averaging hidden-variable assignments across contexts.The restrictions of the global distribution must equal the observed distributions for each context.
  • Factorizability: Factorizability requires joint probabilities to factor into individual measurement probabilities independently of context.In Bell scenarios, this is the usual locality condition.
  • Hidden-variable realizations: The existence of a global section implies a local or non-contextual deterministic hidden-variable realization.Conversely, the paper later establishes equivalence for broader factorizable realizations.
  • Obstructions: Locality and non-contextuality are characterized as obstructions to the existence of global sections.This reformulates the central phenomena using sheaf-theoretic structure.

4. Existence of Global Sections

The paper turns global-section existence into a linear or Boolean satisfiability problem, using incidence matrices to test hidden-variable extendability and compare Bell and Hardy obstructions.

  • Problem: The central problem is deciding whether an empirical model has a global section.This is equivalent to asking whether the model admits a local or non-contextual hidden-variable realization.
  • Incidence matrix: The incidence matrix records restrictions from global assignments to context-specific sections and applies to arbitrary measurement covers.Its image consists of the empirical families arising from global sections.
  • Bell scenario: For the (2,2,2) Bell scenario, the incidence matrix is 16 × 16 and has rank 9.The construction generalizes transfer matrices from Bell scenarios to arbitrary covers.
  • Real-valued test: Solutions of the augmented system M′X = V′ correspond bijectively to global sections of the empirical model.The added normalization row ensures that the unknown weights form a distribution.
  • Bell model: The Bell model has no global section, establishing non-locality through infeasibility over non-negative real distributions.The contradiction follows from selected incidence-matrix equations and non-negativity.
  • Hardy model: The possibilistic Hardy model has no Boolean global section, and its probabilistic version consequently has no non-negative-real global section.Replacing positive probabilities by 1 yields the support model used in the Boolean test.
  • Hierarchy: Possibilistic non-extendability is strictly stronger than probabilistic non-extendability: Hardy has both properties, whereas Bell has only the latter.The Bell support has a Boolean global section despite the Bell model lacking a probabilistic one.

5. Negative Probabilities

Signed probabilities provide real-valued extensions precisely for no-signalling models, while linear-algebraic dimension and rank results characterize the shared space generated by non-contextual and no-signalling models.

  • Signed extensions: Signed-probability extendability amounts to solving the incidence linear system over the reals without non-negativity constraints.Negative probabilities may occur globally while context marginals remain standard non-negative probabilities.
  • Scope: Negative probabilities therefore do not characterize quantum mechanics, because they generate the broader class of no-signalling models.The no-signalling class is strictly larger than the empirical models arising from quantum mechanics in this setting.
  • Shared linear space: The linear subspaces generated by non-contextual and no-signalling models coincide for every measurement cover, with dimension D.This follows from matching upper and lower dimension bounds.
  • Real versus constrained solutions: Every probabilistic model yields a real solution of the incidence system, whereas non-negative or Boolean solutions may fail to exist.There is no analogous homomorphism from the reals to the non-negative reals or booleans preserving the relevant structure.
  • Dimension and rank: For homogeneous covers, the dimension is D = (k ·(l −1)+1)^n; for (n, 2, 2) scenarios, the incidence-matrix rank is 3^n.The 18-vector cover has D = 118, while the Peres-Mermin square has D = 34.
  • No-signalling equivalence: Probability models have local hidden-variable realizations with negative probabilities if and only if they satisfy no-signalling.The result applies over arbitrary measurement covers.

6. Strong Contextuality

Strong contextuality requires that no global assignment agrees even with the model’s support, and it is equivalent to maximal contextuality; GHZ models establish a higher obstruction level than Bell and Hardy.

  • Definition: Strong contextuality means that no global assignment is consistent with the support of the empirical model.This is stronger than lacking a probabilistic global section: Bell and Hardy are not strongly contextual.
  • Hierarchy: The obstruction hierarchy is Bell < Hardy < GHZ, with increasing strengths of no-go results.Hardy is possibilistically non-extendable, while GHZ is strongly contextual.
  • GHZ models: GHZ models are strongly contextual for all n ≥3.Their parity constraints force contradictions among possible global assignments.
  • PR boxes: For (2,2,2) no-signalling models, the PR boxes are exactly the strongly contextual models.Thus strong contextuality characterizes the PR boxes in that scenario.
  • Maximal contextuality: Strong contextuality is equivalent to maximal contextuality, defined by a zero non-contextual fraction.The equivalence connects the qualitative support condition with convex decompositions into local and no-signalling models.
  • Constraint formulation: A model is maximally contextual exactly when its support CSP has no solution; for dichotomic measurements, this is equivalent to formula unsatisfiability.In Bell scenarios, the same condition characterizes maximal non-locality.

7. Generic Strong Contextuality and Kochen-Specker Theorems

The paper develops generic, model-independent Kochen-Specker proofs of strong contextuality using combinatorial conditions on measurement covers and graph-theoretic structures. These conditions obstruct global sections and yield explicit quantum-relevant witnesses.

  • Generic Strong Contextuality and Kochen-Specker Theorems: Strong contextuality follows when a measurement cover admits no global section satisfying the exactly-one-outcome formula.The paper reduces Kochen-Specker-type results to defining covers with no such global section, then supplying quantum representations whose supports satisfy the formula.
  • Generic Strong Contextuality and Kochen-Specker Theorems: If every measurement occurs in an even number of contexts while the cover has an odd number of contexts, the exactly-one formula has no global section.This generalizes the parity-proof pattern commonly used for Kochen-Specker results.
  • Kochen-Specker Graphs: The triangle cover cannot arise from quantum observables because quantum-compatible measurements are generated by symmetric pairwise commutation relations.The triangle consists of the contexts {a,b}, {b,c}, and {a,c}.
  • Kochen-Specker Graphs: For a graph, the exactly-one formula has a global section exactly when the graph has a stable transversal.A stable transversal intersects every maximal clique exactly once and is necessarily an independent set.
  • Kochen-Specker Graphs: Kochen-Specker graphs combine faithful orthogonal co-representations, equal-sized maximal cliques, and the absence of a stable transversal.Such graphs provide finite witnesses for generic strong contextuality and can be realized by quantum observables.
  • Kochen-Specker Graphs: Every graph on n nodes whose complementary graph is (n −d)-connected has a faithful orthogonal co-representation in R^d.This supplies a graph-theoretic sufficient condition for the quantum-relevant representation.

8. Global Sections and Hidden Variables

The paper identifies global sections with factorizable hidden-variable realizations of empirical models. This establishes global sections as the precise obstruction underlying non-locality and contextuality.

  • Global Sections and Hidden Variables: Factorizability requires joint outcome probabilities to equal products of individual-measurement probabilities.Compatibility also makes the individual-measurement distributions independent of the contexts in which measurements occur.
  • Global Sections and Hidden Variables: For Bell-type scenarios, factorizability is exactly Bell locality; more generally, it expresses non-contextuality at the distributional level.The hidden-variable distribution is taken to be independent of the measurement context.
  • Global Sections and Hidden Variables: Theorem 8.1 states that an empirical model has a factorizable hidden-variable realization if and only if it has a global section.The two directions construct a hidden-variable realization from a global section and a global section from a factorizable model.
  • Global Sections and Hidden Variables: The equivalence justifies characterizing non-locality and contextuality as obstructions to global sections.The paper presents this as the definitive conceptual consequence of the hidden-variable correspondence.

9. Quantum Representations

The abstract measurement-cover formalism is represented using commuting quantum observables and their induced outcome distributions. Quantum mechanics then satisfies a generalized no-signalling property for arbitrary compatible families.

  • Quantum Representations: A quantum measurement cover is formed from maximal commuting subsets of a set of observables on a Hilbert space.Commuting observables compose into well-defined observables, and the resulting cover is the quantum representation.
  • Quantum Representations: Under Tsirelson’s stated assumptions, tensor-product structure can be recovered from commuting operator families on a single Hilbert space.Thus the tensor-product form used in Bell scenarios is a special case rather than a necessary starting point.
  • Quantum Representations: Quantum-state distributions on families of commuting observables are compatible on overlaps.This is stated as Proposition 9.2, the generalized no-signalling theorem.
  • Quantum Representations: Quantum statistics for a measurement are independent of the context of other compatible measurements performed with it.The generalized result applies beyond observables acting on different tensor-product factors.
  • Quantum Representations: Quantum representations of suitable vector configurations force supported assignments to satisfy the exactly-one condition on each context.This reduces state-independent strong contextuality to finding a Kochen-Specker configuration with no compatible global assignment.
  • Bell-Type Scenarios and Kochen-Specker Theorems: Bell-type scenarios cannot support generic Kochen-Specker theorems because every local assignment can occur in the support of some product state.Their compatibility structure is special: incompatibility is transitive, unlike in general commuting-observable configurations.

10. Postlude

The postlude emphasizes the framework’s quantum-independent structural generality and its use of global sections to express contextuality. It also interprets measurement incompatibility as a theory-independent impossibility of a joint distribution.

  • Postlude: The sheaf-theoretic treatment exposes contextuality without presupposing complex numbers, Hilbert spaces, operator algebras, or projection lattices.The formalism instead uses structures varying over measurement contexts.
  • Postlude: Global sections provide a canonical description that largely replaces explicit hidden-variable language, while empirical models remain closely tied to observation and experiment.Hidden variables reappear explicitly only in the foundational equivalence result.
  • Postlude: Measurement incompatibility can be interpreted as the mathematically provable non-existence of a joint distribution reproducing all observed empirical distributions.This means that all measurements cannot even in principle be performed jointly while remaining consistent with the outcomes.
  • Postlude: The framework’s finite setting could be generalized to measure-theoretic settings because the distribution functor can be defined over general measure spaces.The paper identifies this as a natural direction for future work.
  • Postlude: Unlike the topos approach’s spectral-presheaf focus, this work introduces distribution functors and measurement covers without starting from quantum-mechanical operator algebras.Both approaches use presheaves varying over contexts, but their core structures differ.

Appendix

The appendix introduces set-theoretic, categorical, and functorial notation used throughout the paper.

  • Set-theoretic notation: Set-theoretic notation covers cardinality, function restrictions, function sets, powersets, disjoint families, and unions.These conventions establish the basic language for later constructions.
  • Categories: A category consists of objects and arrows with specified domains and codomains, associative composition, and identity arrows.The identities satisfy f ◦ idA = f and idA ◦ g = g.
  • Categories: The principal category examples are Set and partially ordered sets, together with the opposite category of a poset.Set uses sets and functions, while a poset has an arrow from p to q exactly when p ≤ q.
  • Functors: A functor assigns objects and arrows between categories while preserving composition and identity arrows.For arrows f and g, it satisfies F(g ◦ f) = F(g) ◦ F(f) and F(idA) = idF A.
Loading 1102.0264v7…