Universally maximal three-party Mermin correlations need more than local strong stability

A correlation test for three observers

Bell inequalities place limits on correlations that local hidden-variable models can reproduce. Quantum systems can exceed those limits. The Mermin inequality tests a three-party version: each observer chooses between two measurements, and four combinations of their outcomes enter a correlation score.

In the paper’s normalization, local hidden-variable models cannot exceed 1, while quantum theory permits 2. A three-qubit Greenberger–Horne–Zeilinger, or GHZ, state reaches that maximum with suitable measurements. This state is an equal quantum superposition of all three qubits being zero and all three being one.

The authors ask something stronger than whether one carefully prepared state can reach 2. They ask when a configuration of three commuting observable algebras allows the maximum to be approached in every normal state. For the global states considered here, normal states are describable by density operators on the shared Hilbert space, including pure vector states and statistical mixtures.

Why quantum fields require different mathematics

For a few qubits, each observer’s measurements usually act on a separate factor of a tensor-product space. Algebraic quantum field theory instead assigns each spacetime region an algebra: a collection of observables closed under specified mathematical operations and limits.

Observables in spacelike separated regions commute, meaning their order of multiplication does not matter. But that does not require the vacuum representation to divide into independent regional tensor factors.

Earlier two-party results connected maximal CHSH violation in every normal state to strong stability, under assumptions involving an algebra, its commutant, and a cyclic and separating vector. The commutant contains all operators that commute with that algebra. Strong stability means an algebra remains mathematically equivalent to itself after taking its von Neumann tensor product with the hyperfinite type II₁ factor.

Strong stability is an algebraic property, not a measure of how strongly a particular pair of systems is entangled. The new work investigates what replaces the two-party criterion for three commuting algebras.

A shared sequence is the key

For three pairwise commuting von Neumann algebras on a separable Hilbert space, the central theorem equates maximal Mermin scores in every global normal state with the existence of one common sequence of local measurement choices. Along this sequence, the four signed products entering the Mermin expression approach the identity operator in the strong operator topology: their action approaches leaving any fixed vector unchanged.

These are asymptotic versions of the GHZ stabilizer relations—the measurement-product identities responsible for the ideal GHZ correlations. Equivalently, the associated Mermin operators converge strongly to twice the identity.

The proof proceeds through a faithful normal state, one that assigns positive expectation to every nonzero positive operator. A sequence approaching the maximal score in that state yields the required operator convergence through positivity. The resulting common sequence then drives the score toward 2 in every normal state.

Crucially, this is an asymptotic statement. It does not say that one fixed set of measurements gives exactly 2 in every state, nor does it establish a convergence rate uniform over all states.

Individual structure does not settle the question

The authors show that strong stability of each local algebra is necessary for maximal Mermin violation in every normal state when the algebras are factors, meaning their centers contain only scalar multiples of the identity. It is not sufficient.

One counterexample places three strongly stable type III₁ factors on separate tensor factors. A product vector state then has a maximal Mermin score of just 1. The structure of the individual algebras therefore does not guarantee the all-normal-state property.

A second construction retains tripartite duality: each party’s commutant is generated by the other two parties’ algebras. Even this relationship, together with strong stability and type III₁ structure, does not guarantee maximal Mermin violation in every normal state.

To establish that obstruction, the authors derive an upper bound of 4/√5, approximately 1.79, whenever a normal state’s expectations factorize between any one pair of parties for all observables belonging to those two parties. They do not claim this bound is optimal.

For a sufficient condition, they require a joint embedding of an infinite tensor product of GHZ triples, with the corresponding local observables contained in each party’s algebra. Measurements move progressively farther along this GHZ sequence, producing Mermin operators that converge strongly to twice the identity. This condition concerns how the three algebras are jointly represented, not merely their separate classifications.

A conditional result for quantum fields

These are abstract mathematical results, not experimental observations or a demonstration of maximal Mermin violation in every normal state for arbitrary triples of spacetime regions. Applying them to a particular quantum field theory requires establishing the joint embedding or directly constructing the asymptotic GHZ stabilizer sequence. Type III₁ local algebras alone do not establish that conclusion.

The paper also notes an obstruction for strictly separated regions satisfying an appropriate split property: the existence of normal product states rules out maximal violation in every normal state. This does not rule out maximal violation in a particular state.

The preprint extends its structural results to any finite number of parties greater than or equal to three. For even party numbers, however, its real-part Mermin expression is weaker than the standard MABK expression. Its conclusions therefore concern the specific functional studied, not all multipartite Bell inequalities.

The result is scoped but exact: maximal three-party Mermin violation in every normal state is characterized by a common asymptotic GHZ stabilizer sequence. Local strong stability is necessary for factors, but cannot replace that joint condition.