Nonclassicality from Incomplete Fock-State Statistics

AI summary

Overview

Research area: Quantum physics — specifically, certifying nonclassicality of bosonic (harmonic oscillator) states from measured Fock-state occupation probabilities, with connections to quantum optics, cavity/circuit QED, trapped ions, and optomechanics.

Technical level: Advanced. The paper presupposes familiarity with phase-space representations of quantum states, coherent states, and convex geometry, though the core idea (testing whether measured statistics are compatible with a mixture of coherent states) is conceptually accessible.

Scope (one sentence): The paper develops a numerical optimization framework that constructs optimal linear nonclassicality witnesses from arbitrary, possibly incomplete subsets of measured Fock-state occupation probabilities, benchmarks it against existing analytical criteria, and applies it to the resonant Jaynes–Cummings model initialized in thermal states.

What This Paper Is About

A quantum state is called "classical" if its statistics can be explained as a statistical mixture of coherent states; if not, it is nonclassical — a resource required for quantum technologies. Experimentally, the easiest thing to measure in many bosonic platforms is the set of Fock-state occupation probabilities, but usually only a few of them are available, and existing analytical nonclassicality criteria (Klyshko inequalities, majorization criteria, counting-matrix criteria) exploit only a fraction of that information and are hard to derive as more probabilities become available. The paper's goal is to replace hand-derived criteria with an automated optimization that extracts the strongest possible linear witness from whatever probabilities a given experiment provides.

Key Contributions

1. A semi-infinite linear program for optimal witnesses. The authors formulate nonclassicality certification from an arbitrary subset of occupation probabilities as an optimization over linear weight vectors λ, using the Poisson-mixture structure of classical Fock statistics. The problem is solved with an exchange (cutting-plane) procedure that adds only the relevant Poisson parameters μ on demand rather than discretizing μ in advance.

2. A numerical replacement for non-smooth optimization. By introducing an auxiliary variable t bounding the classical witness value, the original non-smooth problem is converted into a sequence of ordinary finite linear programs plus one-dimensional maximizations of g(μ) = e^{−μ} Σ_k λ_k μ^k/k!, whose stationary points are roots of a finite polynomial. The authors report that standard numerical optimization methods were found to be unreliable for the original formulation.

3. Model-independent benchmarking against three analytical families. The numerical boundary is compared with the Klyshko-type family 𝒲^𝒦 (including the closed-form two-probability conditions in Eq. (5), standard Klyshko and complement conditions in Eq. (6), and higher-order conditions in Eq. (7)), majorization witnesses 𝒲^ℳ (Eq. (8)), and counting-matrix witnesses 𝒲^𝒞 (Eq. (9)).

4. Open-source implementation and a physically relevant case study. A Python implementation of the numerical optimization and automatic generation of all analytical criteria is released, and the method is applied to the resonant Jaynes–Cummings model with initially thermal qubit and oscillator, revealing nonclassicality at short interaction times and close to thermal balance.

Main Findings

• Exact agreement in small probability spaces. In the (P₀, Pₙ) and (P₀, P₁, P₂) spaces, the numerical boundary coincides with the analytical Klyshko-type boundary; the authors report finding no counterexample in these spaces. Since the boundaries coincide for N_P ≤ 3, only results for N_P ≥ 4 are shown for the Jaynes–Cummings analysis.

• Strictly larger regions with four or more consecutive probabilities. With four consecutive probabilities (Fig. 1, panel g), the numerical criterion identifies a strictly larger nonclassical region than the analytical witnesses.

• Majorization criteria matter for non-consecutive indices. When all probabilities up to an order N are available, 𝒲^ℳ gives no coverage beyond 𝒲^𝒦; but for non-consecutive probabilities (panels e and h), majorization witnesses substantially enlarge the detected region.

• Counting-matrix criteria need at least five arithmetic-progression indices. A 3×3 counting matrix requires at least five indices forming an arithmetic progression, and in that case 𝒲^𝒞 can be stronger than both 𝒲^ℳ and 𝒲^𝒦 (panel i). Even then, the numerical criterion covers a larger probability space than the combined analytical criteria.

• Detection where no analytical witness can even be constructed. In panels (c) and (f), neither Klyshko, majorization, nor counting-matrix witnesses are constructible from the available probabilities, yet the numerical optimization finds large nonclassical regions in the (P₁, P₂) and (P₁, P₄, P₅ = 0.05) spaces.

• Jaynes–Cummings: extended reach to shorter times and hotter oscillators. For p_e = 0.4, 0.45, 0.48 (corresponding to k_B T_q/ħω = 2.47, 4.98, 12.5), the numerical criterion extends the detected nonclassical region toward shorter interaction times τ = gt and larger initial n̄, both experimentally favorable (less decoherence exposure, less oscillator cooling needed).

• Largest gain near thermal balance. The greatest improvement occurs near the thermal-balance condition n̄ = n̄_e ≡ p_e/(1 − 2p_e) = 2, 4.5, 12. As the number of available probabilities increases, the undetected region around thermal balance shrinks substantially. The state is strictly classical exactly at n̄ = n̄_e, but sizable nonclassical regions appear in its immediate vicinity — the authors identify this as a novel regime suitable for immediate experimental exploration.

• Exponential narrowing confirmed independently. In Fig. 3 (p_e = 0.40, n̄_e = 2), even at N_P = 10 the numerical criterion lies closer to thermal balance than the best analytical criterion 𝒲^𝒞. The counting-matrix criterion at sufficiently large N_P, together with an independent check using characteristic-function criteria, confirms that the classical region around thermal balance narrows exponentially with interaction time, so any finite thermal imbalance in principle leads to detectable nonclassicality with interaction time growing only logarithmically as the imbalance decreases. The authors also state the effect is stable under small pure loss on the oscillator.

• Numerical optimization is the most general witness considered. Every state detected by an analytical witness is also detected numerically, but not conversely — the orange regions of Fig. 1 correspond to states with a valid linear witness not captured by the analytical criteria.

• Numerical and analytical approaches are complementary. As illustrated in Fig. 3, when a large number of probabilities is required, the numerical optimization becomes increasingly unstable, so analytical criteria remain valuable in demanding regimes.

Methodology in Plain English

The starting point is that any classical state's Fock-probability vector must be a weighted average (a convex combination) of Poisson distributions p_k(μ) = e^{−μ} μ^k/k!, since each coherent state gives a Poisson photon-number distribution. A witness is a set of weights λ assigned to the probabilities you actually measured; you compute λ·P (a weighted sum of your data) and compare it to the largest value that same weighted sum could take over all Poisson distributions, max_μ λ·p(μ). If your data exceeds that classical ceiling, the state is nonclassical.

Finding the best λ is hard because the maximum over μ jumps discontinuously. The authors introduce a variable t that must sit above the classical value for every μ, turning the problem into a linear program — except there are infinitely many constraints, one per continuous μ. They solve it iteratively: keep a small list of μ values, solve the finite linear program, then check the candidate witness against the full continuous Poisson family by maximizing a one-dimensional function g(μ) (whose stationary points are polynomial roots, evaluated at μ = 0 plus all non-negative roots, with the μ → ∞ limit contributing zero). If the worst violation exceeds the tolerance δ, that μ is added to the list and the program is resolved. Only a small number of constraints is needed before convergence. For the physics case study, they compute the Jaynes–Cummings evolution of an oscillator and qubit prepared in separate thermal states, trace out the qubit, and test the resulting occupation probabilities against both the numerical and the analytical witnesses.

Why This Matters

Impact on research: The work generalizes and subsumes existing Fock-probability-based nonclassicality criteria into a single optimization, provides a principled answer to "how much nonclassicality information is in this incomplete data set," and identifies which occupation probabilities carry the relevant information — information that can guide both experiment and the derivation of new, simpler analytical witnesses.

Real-world applications (as framed by the paper's platform list):

• Quantum optics and photonic sources — certifying nonclassical light using boson-number-resolving detectors without full state tomography.
• Cavity and circuit QED — characterizing nonclassical microwave field states with limited photon-number resolution.
• Trapped ions and optomechanical systems — detecting nonclassicality from incomplete phonon or mechanical occupation statistics.
• Thermal, noisy experiments — the Jaynes–Cummings results suggest that nonclassicality can be witnessed closer to thermal balance and at higher oscillator temperatures, easing cooling requirements and reducing decoherence exposure.

Industry relevance: Direct relevance to quantum hardware developers who need efficient, low-overhead certification of nonclassical resource states rather than costly tomographic reconstruction, and to builders of number-resolving detector systems.

Future Directions

1. Guiding new analytical witnesses. Using the numerically identified weighting vectors to construct simpler, closed-form criteria that remain robust where the numerical optimization becomes unstable at large N_P (as seen in Fig. 3).

2. Experimental exploration near thermal balance. The authors explicitly flag the sizable nonclassical regions near n̄ = n̄_e at p_e approaching 1/2 as a regime suitable for immediate experimental testing.

3. Adaptive measurement design. Using the optimization to determine which Fock probabilities should be measured to maximize nonclassicality detection power, rather than measuring all available orders.

4. Stability and imperfection analysis. The authors note stability under small pure loss on the oscillator, but a systematic treatment of realistic detector inefficiency, noise, and limited number resolution across platforms is a natural extension.

Target Audience

Experimentalists in quantum optics, cavity and circuit QED, trapped ions, and optomechanics who have partial Fock-state counting data and need a practical nonclassicality test; theorists working on nonclassicality criteria and convex-geometry approaches to quantum resources; and quantum-hardware engineers seeking lightweight state-certification tools. Readers will benefit from background in coherent-state P-representations and basic convex optimization; the released Python code makes the method accessible to those without that background.

Authors’ abstract

Nonclassicality in bosonic systems can be certified directly from occupation probabilities without full state reconstruction. To complete a library of such criteria we introduce a numerical framework that constructs optimal linear nonclassicality witnesses from arbitrary subsets of measured occupation probabilities by solving a semi-infinite linear program. The method contains known analytical criteria as special cases while generally detecting substantially larger nonclassical regions. We provide an open-source implementation of the numerical optimization together with the analytical criteria discussed here, enabling direct application of the framework to measured occupation statistics across bosonic platforms. As a case study, we apply it to the resonant Jaynes--Cummings model with thermal states and reveal nonclassicality at short interaction times and close to thermal balance, where detection of nonclassicality is otherwise challenging.