Contextuality Inequalities in Quantum Systems

• Contextuality inequalities are linear constraints on measurement outcomes that distinguish noncontextual classical models from quantum contextual behaviors through exclusivity relations.
• They are derived using graph-theoretic frameworks where vertices represent measurement events, and independent sets of events define classical bounds, with violations indicating quantum advantages.
• These inequalities underpin resource theories in quantum computation and offer robust criteria for experimental validation of nonclassical phenomena in multi-dimensional systems.

Contextuality inequalities are linear constraints on outcome probabilities that delineate the set of behaviors compatible with noncontextual hidden-variable models. Their violation serves as an operational signature of contextuality: a fundamental deviation from classical realism, closely linked to the nonclassical computational power of quantum theory.

1. Exclusivity-Graph Framework and Formulation of Contextuality Inequalities

A contextuality scenario is defined by a finite set of measurement events, each corresponding to an outcome $e = (a|M)$ ("outcome $a$ for measurement $M$"), which are indexed as vertices in a graph $G$. The exclusivity relation, $e_i \perp e_j$, is encoded by edges between mutually exclusive events—those that cannot both occur in a single run (typically, distinct outcomes of the same measurement). A behavior is a map $p: V(G) \to [0,1]$, assigning to each event-vertex $i$ the probability $p_i = p(i)$; the exclusivity constraint imposes $p_i + p_j \leq 1$ for each edge $(i,j) \in E(G)$.

The set of noncontextual behaviors, $a$0, is the convex hull of all 0–1 deterministic assignments respecting the exclusivity constraints—equivalently, the incidence vectors of independent sets (sets of nonadjacent vertices) in $a$1. Quantum behaviors $a$2, realizable using quantum projectors assigned to vertices and a quantum state, always satisfy $a$3, where $a$4 denotes behaviors satisfying the pairwise exclusivity constraints only (Bharti et al., 2018).

A contextuality inequality is a linear constraint (typically a facet of $a$5 in the probability space) of the form

$a$6

where $a$7 and $a$8 is the weighted independence number, i.e., the maximal sum over weights for any independent set in $a$9 (Wagner et al., 2022). The unweighted case ($M$0) yields

$M$1

where $M$2 is the independence number.

2. Structure and Uniqueness of Fundamental Contextuality Inequalities

The Strong Perfect Graph Theorem, applied to exclusivity scenarios, establishes that contextuality—i.e., $M$3—occurs if and only if $M$4 contains as an induced subgraph either an odd cycle $M$5 with $M$6, or its complement (odd anti-cycle) $M$7 (Bharti et al., 2018). These are the only minimal (fundamental) "obstructions" to noncontextuality; all other exclusivity graphs are perfect and cannot witness contextuality.

Correspondingly, there is a unique facet-defining noncontextuality inequality for each such nontrivial (i.e., $M$8) cycle and anti-cycle:

• Odd cycle $M$9 ($G$0 odd, $G$1):

$G$2

which uses the fact that the maximum number of 1's in a $G$3-independent set is $G$4.

• Odd anti-cycle $G$5:

$G$6

since in the anti-cycle, every vertex is adjacent to all but two others and at most two events can simultaneously be assigned 1.

Uniqueness of these inequalities as the only proper facet-defining contextuality inequalities in these scenarios follows from the convex-geometric argument that these are the sole additional constraints differentiating the stable-set (NC) polytope from the fractional stable-set (E-principle) polytope (Bharti et al., 2018).

3. Examples and Generalizations: KCBS Inequality and High-Dimensional Scenarios

KCBS (Klyachko–Can–Binicioglu–Shumovsky) Scenario ($G$7):

The canonical example is the $G$8 (pentagon). The NC inequality is

$G$9

Quantum theory (e.g., a qutrit measured with pentagon-orthogonal projectors) achieves $e_i \perp e_j$0, thus violating the inequality and saturating the Lovász theta number $e_i \perp e_j$1 (Bharti et al., 2018, Bub et al., 2010).

Generalizations to arbitrary odd $e_i \perp e_j$2:

For $e_i \perp e_j$3, the NC bound is $e_i \perp e_j$4, and quantum theory yields $e_i \perp e_j$5. Realizations require increasing Hilbert space dimension (at least $e_i \perp e_j$6 for odd $e_i \perp e_j$7-cycles), with explicitly constructed projectors as detailed in (Bharti et al., 2018, Sohbi et al., 2016).

High-dimensional (multi-event) contextuality inequalities:

Graph-theoretic generalizations identify families of inequalities, often constructed from multipartite Bell-like operators or through logical-proofs/graph-theoretic constructions, with quantum violations increasing with dimension (contextuality concentration) (Liu et al., 2022). For example, in single-system analogs of MABK inequalities, the quantum-to-classical violation ratio asymptotically approaches 2 as dimension increases, even as the minimal Hilbert space dimension for a single system is $e_i \perp e_j$8.

4. Algorithmic and Graph-Theoretic Approaches for Deriving Inequalities

Systematic derivation and classification of contextuality inequalities in general scenarios leverage the exclusivity (or compatibility) graph and its associated polytopes:

• The stable-set polytope $e_i \perp e_j$9, whose facets correspond to tight noncontextuality inequalities.
• The enumeration of all facet-defining inequalities can be automated, e.g., via convex hull algorithms for the polytope generated by all deterministic 0–1 stable assignments on $p: V(G) \to [0,1]$0 (Wagner et al., 2022, Silva, 2015). Weighted cases, more general event structures, and equivalence with basis-independent coherence witnesses can also be analyzed within this framework.
• Unification with the Abramsky–Brandenburger sheaf-theoretic approach shows that graph invariants—independence number $p: V(G) \to [0,1]$1 (classical bound), Lovász number $p: V(G) \to [0,1]$2 (quantum bound), and fractional packing number $p: V(G) \to [0,1]$3 (E-principle bound)—precisely delineate the possible ranges for contextuality witnesses (Silva, 2015, Wagner et al., 2022).

5. Contextuality Inequalities in Quantum Computation and Resource Theories

The unique fundamental contextuality inequalities for cycles and anti-cycles quantitatively characterize the simplest forms of contextuality, which acts as a monotone in contextuality resource theories. Their degree of violation is directly related to the quantitative resource content:

• Contextuality monotones: The ratio $p: V(G) \to [0,1]$4 or $p: V(G) \to [0,1]$5 (cycle/anti-cycle cases) serves as a monotone, increasing with contextuality resources available in a given scenario (Bharti et al., 2018).
• Quantum computation: Violations of these inequalities are both necessary and sufficient for universal quantum advantage in measurement-based quantum computation. Resource-theoretic perspectives show that contextual scenarios exceeding the classical bound enable measurement scenarios with genuine quantum computational power (Bharti et al., 2018).
• Self-testing: Maximal quantum violation of a fundamental noncontextuality inequality self-tests the involved state and measurement structure (up to local isometries), providing an operational characterization of quantum systems realizing specific exclusivity structures.

6. Robustness, Experimental Realization, and Extensions

Experimentally, contextuality inequalities must be robust to imprecision, signaling, and unsharpness:

• Ontological faithfulness: Explicit criteria quantify how much experimental imperfections (e.g., context variation in a given measurement label) can impact the interpretability of inequality violations. For an observed quantum violation gap $p: V(G) \to [0,1]$6, no $p: V(G) \to [0,1]$7-ontologically faithful noncontextual model can simulate the data if $p: V(G) \to [0,1]$8, where $p: V(G) \to [0,1]$9 is the number of measurement settings (Sohbi et al., 2016).
• Generalized (monogamy) relations: Monogamy of contextuality inequalities arises from the impossibility of simultaneous maximal violations in overlapping (e.g., interlinked) exclusivity structures. Perfectness of the commutation or exclusivity graph yields conditions for the possible joint violations (Ramanathan et al., 2012, Zhu et al., 2015).
• State-independent inequalities: For certain measurement sets (e.g., Yu–Oh's set), there exist optimal and tight state-independent contextuality inequalities (SIC), violated by all quantum states and fully characterizable via linear programming—providing maximal quantum-to-classical margin (Kleinmann et al., 2012, Gonzales-Ureta et al., 2022).

7. Cyclic Systems, Measures, and Extensions beyond No-Signaling

Cyclic systems of dichotomic variables (arising in Bell, KCBS, and Leggett-Garg scenarios) yield a universal family of contextuality inequalities, generally of the form

$i$0

where $i$1 is the two-point correlation (possibly with sign reversals for odd $i$2). The set of all admissible noncontextual assignments forms a polytope, whose facets exactly correspond to the $i$3 contextuality inequalities parameterized by odd-parity sign vectors (Dzhafarov et al., 2019, Khrennikov, 2022).

When signaling or experimental imperfections are present, correction terms proportional to the measured signaling or unsharpness must be included, leading to robustified, "genuine" contextuality inequalities that tolerate experimental noise (Vallée et al., 2023, Dzhafarov et al., 2014, Khrennikov, 2022). The degree of contextuality can then be quantified, e.g., by the $i$4 distance from the observed data to the noncontextual polytope boundary, providing a continuous measure even when violations are not maximal (Dzhafarov et al., 2019).

---

In summary: Contextuality inequalities are the primary operational tool for distinguishing noncontextual behaviors from the broader set permitted by quantum mechanics. Their structure is intimately connected with the graph-theoretic properties of measurement exclusivity, and their violation underpins fundamental nonclassicality, resource-theoretic properties, and quantum computational power (Bharti et al., 2018, Silva, 2015, Sohbi et al., 2016, Dzhafarov et al., 2019, Vallée et al., 2023).