Foundations of quantum mechanics · Theory
Plenitude of ideal correlations explains quantum theory
arXiv:2611.21448v1 ↗PDF ↗Submitted 16 November 2026
- Question
- Every Bell scenario has its own set of quantum correlations, with a curved boundary. Principles such as information causality, macroscopic locality and local orthogonality each bound that set in some scenarios, but none reproduces it everywhere. Is there a single physical principle that selects the quantum correlations in every scenario at once, and with them the Hilbert-space formalism and the Born rule?
- Result
- Cabello proposes the plenitude of ideal correlations: every correlation possible in a theory in which every observable admits an ideal measurement (repeatable, and not disturbing any compatible observable) is physically realised. He proves that ideal measurements imply the exclusivity principle and Specker's principle. Combining independent experiments on a graph and on its complement then restricts the correlations on every exclusivity graph \(G\) to the theta body \(\mathrm{TH}(G)\), the set reachable by quantum ideal measurements. Plenitude makes the whole theta body realisable, and the paper argues that this forces projective measurements, commuting compatible observables, the Born rule and, through Wigner's and Stone's theorems, unitary dynamics.
The paper in detail
Cabello changes the object of study from Bell scenarios to exclusivity graphs. The vertices of such a graph are events, an outcome assigned to each observable in a context, and two vertices are joined when the events are distinct outcomes of one observable, so they cannot both occur. A correlation assigns a probability to each vertex. Every Bell or contextuality scenario produces such a graph, and graphs embed into one another simply, which is the compositional structure the paper wants a principle to explain. For quantum ideal measurements the achievable assignments on a graph \(G\) form the theta body \(\mathrm{TH}(G)\): there are unit vectors \(\ket{u_i}\), orthogonal whenever vertices are adjacent, and a unit vector \(\ket\psi\) with \(p_i=|\langle u_i|\psi\rangle|^2\), a set described by a semidefinite program.
The first step derives two familiar principles from one assumption. If every observable admits an ideal measurement, pairwise compatible observables are jointly measurable (Specker's principle), and the probabilities of pairwise exclusive events sum to at most 1 (the exclusivity principle). The proof builds one measurement whose outcomes include all the events by measuring their two-outcome coarse-grainings in sequence; non-disturbance guarantees that each step reproduces the original probability.
Applied to one experiment, the exclusivity principle only imposes clique constraints, which are too weak: assigning \(\tfrac12\) to each vertex of a pentagon violates no clique. Independent experiments combine through the OR product of their graphs, and in two copies of the pentagon the five vertices \((1,1),(2,4),(3,2),(4,5),(5,3)\) form a clique carrying total probability \(\tfrac54\). More generally, pairing a graph with its complement shows that \(\sum_ip_iq_i\le1\) for every realisable \(q\) on the complement, which by theta-body duality means \(p\in\mathrm{TH}(G)\) once every theta-body correlation on the complement is available. For self-complementary graphs this closes the circle; every graph is an induced subgraph of a self-complementary graph four times its size, and theta bodies project onto induced subgraphs, so the conclusion holds for all graphs.
The paper then moves to scenarios and to the formalism. One observable gives a complete graph plus normalisation, which forces its vectors to form a resolution of the identity; a joint context of two observables forces commuting projection-valued measures. Cabello concludes that correlations take the form \(\langle\psi|\Pi_{a|x}\cdots\Pi_{c|z}|\psi\rangle\) with commuting projectors, the commuting-operator set, and adds Wigner's theorem and Stone's theorem to obtain unitary evolution and the Schrödinger equation. The discussion reads quantum theory as what remains when no hidden properties restrict the allowed correlations, and argues that a separate interpretation becomes unnecessary. What an experiment must do to realise all the allowed correlations is left for future work.
Transferable methods
Sequential ideal measurements yield the exclusivity bound
In a general probabilistic theory, events that are pairwise exclusive need not be outcomes of a single measurement, so their probabilities need not sum to at most 1. Ideal measurements supply the missing single measurement: measuring coarse-grainings one after another produces a procedure whose outcomes include every event, and non-disturbance keeps each probability unchanged.
Take pairwise exclusive events \(a,b,c\): observables \(AB\), \(AC\) and \(BC\) contain them as distinct outcomes. Let \(A=\{a,\bar a\}\) be the coarse-graining of \(AB\) (equivalently of \(AC\)), and define \(B\) and \(C\) likewise. Each pair is jointly measurable through a common parent, so \(A\), \(B\) and \(C\) are pairwise compatible.
Define \(D\): measure \(A\) ideally; on \(\bar a\), measure \(B\) ideally; on \(\bar b\), measure \(C\) ideally; the outcomes are \(a,b,c\) and the remainder \(d\). Normalisation gives \(p(D=a)+p(D=b)+p(D=c)+p(D=d)=1\), and because an ideal measurement of \(A\) does not disturb the compatible \(B\) and \(C\), \(p(D=b)=p(b)\) and \(p(D=c)=p(c)\). Hence \[p(a)+p(b)+p(c)\le1,\] and \(D\) refines \(A\), \(B\) and \(C\), so they are jointly measurable.
Editorial illustration: in quantum theory exclusive events of ideal measurements are orthogonal projectors, \(\Pi_a\Pi_b=0\), so \((1-\Pi_a)\Pi_b(1-\Pi_a)=\Pi_b\): measuring \(A\) first leaves the probability of \(b\) unchanged, which is the non-disturbance used above. For unsharp effects that relation fails, and the sum of probabilities of pairwise exclusive outcomes is no longer bounded by 1 in general, which is why the assumption of ideal measurements matters.
How to reuse it
To turn operational assumptions into an inequality on probabilities, look for one composite procedure whose outcomes include all the events concerned; its normalisation supplies the bound. The same strategy appears in proofs of Specker's principle and in arguments that sequential compatible measurements have order-independent statistics.
Theta-body duality from a graph and its complement
The clique constraints of one experiment define a polytope that is too large. Running an independent experiment on the complement graph alongside creates large cliques in the product, and the exclusivity bound on those cliques, with the assumption that every theta-body correlation is available, cuts the allowed set down to exactly the theta body.
For independent experiments the joint probabilities factor, \(r_i=p_iq_i\). The diagonal pairs \((i,i)\) of \(G\) and \(\overline G\) form a clique in the product exclusivity graph, since any two distinct vertices are adjacent in one of the two graphs. The exclusivity principle then gives \[\sum_ip_iq_i\le1.\]
The theta body has the dual description \(\mathrm{TH}(G)=\{p\ge0:\vartheta(\overline G,p)\le1\}\), where \(\vartheta(\overline G,p)=\max_{q\in\mathrm{TH}(\overline G)}\sum_ip_iq_i\). So the inequality holds for all \(q\in\mathrm{TH}(\overline G)\) exactly when \(p\in\mathrm{TH}(G)\). For the pentagon, which is self-complementary, the uniform vector \(p_i=\tfrac12\) gives \(\vartheta(\overline{C_5},p)=\sqrt5/2>1\), so it is excluded, while \(p_i=1/\sqrt5\) attains the quantum maximum \(\sum_ip_i=\sqrt5\). Editorial check: a semidefinite program reproduces both values.
To reach every graph, place copies of \(G\) on blocks \(A\) and \(D\), copies of \(\overline G\) on \(B\) and \(C\), and join all of \(A\) to \(B\), \(B\) to \(C\) and \(C\) to \(D\). The block permutation \(A\to B\), \(B\to D\), \(C\to A\), \(D\to C\) maps this graph \(H\) onto its complement, and the projection of \(\mathrm{TH}(H)\) onto the induced copy of \(G\) is \(\mathrm{TH}(G)\), because adding vertices of zero weight does not change the weighted theta function. Editorial check: the permutation is an isomorphism onto the complement for 200 random graphs.
How to reuse it
Product constructions turn a weak one-copy constraint into a strong one; the same move bounds the Shannon capacity of a graph by its theta function and underlies multi-copy tests of local orthogonality. When a constraint is linear in the probabilities, testing it against products with a suitable partner experiment exposes its dual description.
Context & limitations
Why it matters. It gathers into one argument a programme developed over a decade. The exclusivity principle bounds correlations on self-complementary graphs by the theta body, every graph sits inside a self-complementary one, and the theta body is exactly the quantum set for ideal measurements. If the step from graphs to Bell scenarios holds, the quantum correlation sets of all scenarios follow from one operational requirement plus a maximality postulate, rather than being bounded one scenario at a time.
Limitations. The paper describes itself as partly a review, with refined proofs of the author's earlier results. The upper bound on correlations is derived from ideal measurements; the lower bound, that every theta-body correlation exists, is the plenitude postulate itself. Editorial note: the passage from exclusivity graphs to Bell scenarios (Section III) carries the most weight. Theorem 10 obtains commuting projectors from one joint context at a time and does not show that an observable's projectors are the same in every context it belongs to. That consistency separates quantum correlations from almost-quantum correlations, which form a strictly larger set and already satisfy local orthogonality, the Bell-scenario form of the exclusivity principle (Navascués, Guryanova, Hoban and Acín). The claim for Bell scenarios is therefore asserted rather than established in this text. The paper contains no statement about AI use.