Exact Quantum Maxima of the n-Cycle Overlap Inequalities

This paper derives and establishes exact quantum maximums for overlap inequalities involving cycles of arbitrary length, providing a rigorous benchmark for testing basis-independent coherence and preparation contextuality. The central finding is that the global optimum over all finite-dimensional quantum realizations is already achieved in dimension two, demonstrating dimensional saturation of the overlap-cycle hierarchy. Furthermore, it shows that standard multi-path interferometry offers an experimentally accessible method to probe these nonclassical properties using only pairwise visibility measurements.

Derivation of the Exact Quantum Maximum

The paper addresses the problem of finding the exact quantum maximum for a general n-cycle overlap expression, defined as:

(16) Sn ≡ nX−1 i=1 ri,i+1 − r1n,

where ri,i+1 = ⟨didj⟩2.

The classical bound for jointly diagonalizable (classical) states is established by Proposition 2:

(Proposition 2 (Classical bound for the n-cycle). For any jointly diagonalizable family of states, Sn:= nX−1 i=1 ri,i+1 − r1n ≤ n − 2.

The exact quantum maximum is derived using geometric arguments based on the Fubini–Study metric. The optimal configurations correspond to:

(The optimal configurations correspond to pure qubit states equally spaced along a Fubini–Study geodesic of length (n − 1)π/(2n), where distance is measured with respect to the Fubini–Study metric [14, 15]. Equivalently, they are realized by equally spaced states on a great circle of the Bloch sphere.)

The resulting exact quantum maximum is given by:

(Theorem 5 (Exact quantum maximum of the n-cycle). For every integer n ≥ 3 and every finite-dimensional Hilbert-space realization, Sn = n cos2 π/2n − 1.

Connection to Experimental Visibility

The paper demonstrates that the overlap inequalities can be tested operationally using a standard multi-path interferometer based solely on pairwise visibility measurements. The key connections are:

(For equal path amplitudes, ci = 1/√n (i = 1,..., n), Eq. (7) simplifies to Vij = ⟨didj⟩, V2ij = rij.)

The three-path visibility inequality is derived from the classical polytope constraints:

(V212 + V223 − V213 ≤ 1, (15))

This visibility inequality is shown to be equivalent to the overlap constraint for symmetric interferometers:

(For the symmetric interferometer (ci = 1/√3), this reduces to V212 + V223 − V213 ≤ 1. (13))

Witnessing Noncontextuality and Coherence

A violation of the visibility inequality simultaneously witnesses two fundamental quantum features:

(Thus a violation of V212 + V223 − V213 ≤ 1 simultaneously witnesses basis-independent coherence (by ruling out joint diagonalizability [2]) and preparation contextuality (by ruling out a noncontextual ontological model).)

This operational test avoids the need for state tomography or dedicated overlap-estimation protocols like SWAP tests, as the required overlap statistics are obtained directly from measured visibilities. The required threshold for observing violations in realistic experiments is determined by an efficiency factor η:

(Since each term in the cycle expression is quadratic in the visibility, the experimentally observed value becomes Sexp = η 2S. A violation occurs whenever η 2S > 1, or equivalently, η > 1/√S.)

Dimensional Saturation and Asymptotic Behavior

The results establish dimensional saturation for the overlap-cycle hierarchy:

(The global optimum over all finite-dimensional Hilbert spaces is already achieved in dimension two; higher-dimensional systems offer no advantage.)

For large cycle lengths, the quantum advantage over the classical bound approaches a constant value:

(For large n, expansion of the quantum maximum gives Smax n = n cos2 π/2n − 1 = n − 1 − π/4n + O(n−3). Thus, the quantum advantage over the classical bound Sn - (n−2) approaches a constant: Sn - (n−2) = 1 − π/4n + O(n−3) as n→∞.)

The tightness of the bound is confirmed by showing that for large n, the visibility threshold required scales as:

**(the visibility threshold scales as ηmin = s/(n − 2)Smax n = 1 − 1/2n + O(n−2), so near-perfect visibility is required to observe violations for large n.

Here are specific improvements to AI systems derived from the scientific findings presented in this paper:

1. Enhanced Quantum State Verification and Non-Contextuality Testing:

2. Development of Basis-Independent Coherence Measures for Machine Learning Models:

3. Creation of Robust Adversarial Attacks and Contextuality Probes in Neural Networks:

4. Optimization of Quantum Machine Learning (QML) Architectures via Overlap Constraints:

1. Enhanced Quantum State Verification and Non-Contextuality Testing:

The paper establishes that overlap inequalities are basis-independent coherence witnesses, implying preparation contextuality when violated.

Improvement: AI systems can incorporate overlap inequality checks during training or inference to verify if the underlying quantum state representation adheres to classical (jointly diagonalizable) constraints.

What the improved AI can do: It can act as a detector for non-classicality in quantum computations or simulations, flagging states that violate the expected classical bounds (e.g., flagging states where the measured visibility inequality is violated), thereby identifying genuine quantum coherence or preparation contextuality in complex quantum circuits without relying on full state tomography.

2. Development of Basis-Independent Coherence Measures for Machine Learning Models:

The paper connects pairwise visibility measurements in multi-path interferometers to state overlaps, which are linked to coherence and nonclassicality.

Improvement: AI can be trained to extract operational visibility features from complex data or internal representations of a quantum process (e.g., using techniques analogous to measuring path interference).

What the improved AI can do: It can derive basis-independent measures of coherence for quantum models (like Variational Quantum Eigensolvers or Quantum Neural Networks) that are robust against choice of measurement basis, allowing it to quantify the coherence resource inherent in the model's learned parameters, moving beyond standard fidelity measures.

3. Creation of Robust Adversarial Attacks and Contextuality Probes in Neural Networks:

The framework shows how preparation contextuality manifests through overlap constraints derived from cycle inequalities (e.g., the three-path inequality).

Improvement: Use these overlap constraints as a basis for designing adversarial perturbations or contextuality probes tailored to exploit non-contextual assumptions in classical or semi-classical ML models attempting to mimic quantum behavior.

What the improved AI can do: It can develop noncontextuality awareness modules that actively seek out inputs or internal representations that violate expected classical constraints, potentially leading to more robust defenses against adversarial examples that rely on hidden variable assumptions (i.e., exploiting preparation contextuality).

4. Optimization of Quantum Machine Learning (QML) Architectures via Overlap Constraints:

The exact quantum maximum for the n-cycle overlap inequality is derived and saturated by specific coplanar qubit states equally spaced along a Fubini–Study geodesic.

Improvement: QML architectures can be designed or trained using optimization algorithms that explicitly incorporate the geometric constraints defined by these optimal state configurations (i.e., optimizing parameters to stay on or near the geodesic of maximum overlap).

What the improved AI can do: It can perform more efficient and geometrically constrained optimization in quantum hardware, potentially leading to faster convergence or finding global optima for quantum circuits by leveraging the known geometric structure of maximally non-classical states.

We extend the three-state overlap analysis to determine the exact quantum maximum over finite-dimensional pure-state realizations of the n-cycle overlap inequalities, S n=n 2(π/(2n))-1, for arbitrary cycle length n 3. The bound is saturated by an explicit family of coplanar qubit states equally spaced along a Fubini--Study geodesic, establishing dimensional saturation of the overlap-cycle hierarchy. Thus, the global optimum over all finite-dimensional pure-state realizations is already achieved in dimension two. We further show how, under ideal symmetric interferometric conditions, the overlap quantities can be inferred from pairwise fringe visibilities. The three-state case recovers the known maximum 5/4, while the exact n-cycle result shows that the quantum--classical gap approaches one as n to infinity, with corresponding visibility thresholds. Within generalized noncontextuality frameworks, and subject to the required operational equivalences, such violations can witness preparation contextuality.