Exponential quantum advantage for learning signals with a single qubit

AI summary

Overview

Research area: Quantum sensing, quantum learning theory, and quantum metrology, with superconducting-circuit experiments and numerical simulations of applications.

Technical level: Advanced. The paper combines phase-space (Wigner-function) mathematics, statistical learning theory, complexity lower bounds, and a superconducting cavity–qubit experiment. The high-level message is accessible, but the theoretical machinery (Quantum Phase-Space Inference, accessible feature information) assumes graduate-level quantum physics.

Scope (one sentence): The paper proves and experimentally demonstrates that coupling a single controllable qubit to an otherwise conventional sensor allows classical signals (Fourier coefficients, temporal correlations, nonlinear functionals) to be learned exponentially faster than any conventional sensing strategy, and packages this result in a general framework (QΨ) for discovering and certifying such advantages.

What This Paper Is About

Most demonstrated "quantum advantages" need resources far beyond what near-term hardware provides — many entangled qubits, fast high-fidelity local control, long coherence. The authors ask whether a much more modest ingredient, one controllable qubit added to an ordinary sensor, can still deliver provable, rigorous exponential speedups for practically relevant tasks: learning properties of classical signals such as microwave or electromagnetic fields. They answer yes, prove it, and demonstrate it on a superconducting transmon–cavity device.

Key Contributions

1. A single-qubit exponential separation (Theorem 1). A protocol with probe energy O(k), one ancilla qubit, and one control operation learns the k-th Fourier coefficient of a signal using O(k) signal queries; any conventional protocol with the same energy scaling requires exp(Ω(k)) queries. The advantage grows exponentially in the Fourier index k rather than improving precision by a constant factor.

2. A hierarchy of sensing architectures with distinct exponential advantages. Beyond single-qubit control, the authors prove: conventional Gaussian probes (e.g. squeezed states) beat all classical coherent-state strategies exponentially (Theorem 2, O(1) vs. exp(Ω(k)) queries, shown for angular Fourier coefficients / moments); and a single qubit of long-lived quantum memory beats any memoryless strategy exponentially for m-point temporal correlators (Theorem 3, O(1) vs. exp(Ω(m)) queries) — claimed as the first exponential advantage obtained solely by adding one qubit of quantum memory.

3. Quantum Phase-Space Inference (QΨ) and the accessible feature information (AFI). A unifying framework that takes an experimental objective plus platform constraints, and from a phase-space overlap quantity simultaneously produces a tight lower bound on measurement complexity, an optimal quantum-enhanced algorithm attaining it, and a certificate of quantum advantage — going beyond the local parameter-estimation regime described by quantum Fisher information.

4. Experimental and numerical validation. A transmon qubit coupled to the fundamental mode of a high-purity aluminum cavity demonstrates up to k = 20, with 10^7-fold reductions in measurement counts, plus simulation studies of axion dark matter detection and 8-QAM/64-QAM wireless decoding.

Main Findings

• Single-qubit advantage is exponential and experimentally observed. Using 10^4 shots, the experimental single-qubit quantum feature sensing (QFS) protocol accurately learned Fourier amplitudes across all frequencies k, while an idealized noiseless simulation of squeezed homodyne sensing produced erratic estimates. For the depicted signals, single-qubit QFS distinguished signals with 70% prediction accuracy using only ~100 shots — a factor of 10^7 improvement over any conventional Gaussian approach with equal energy. The 70% convention is used because success probability ≥ 1−δ costs only logarithmic shot overhead ~log(δ⁻¹).

• Decoherence does not erase the advantage at this scale. Even with decoherence present, the quantum-enhanced sensor required seven orders of magnitude fewer signal queries than a decoherence-free conventional Gaussian protocol. A Gaussian protocol could match the query complexity only by increasing probe energy by a polynomial factor in k.

• Squeezing alone gives an exponential advantage over classical sensing. A conventional Gaussian sensor with energy O(k) can learn the k-th angular Fourier coefficient in O(1) queries, while any classical protocol — regardless of energy — needs exp(Ω(k)). Experimentally the authors implemented a modified protocol with ideal query complexity O(k^{2/3}).

• One qubit of quantum memory suffices for time-varying signals. For times t₁ < … < t_m separated by at most Δ, a sensor that decoheres much faster than Δ, coupled to a single qubit coherent up to t_m, estimates the m-point temporal correlator in O(1) queries; any memoryless protocol of the same energy needs exp(Ω(m)) queries even with arbitrarily many qubits or modes. Benchmarks in simulation included the Helstrom measurement (optimal binary non-Gaussian measurement at each step) and squeezed homodyne sensing, each given five times the energy of the memory-enabled protocol.

• Dark matter application (simulation). Using velocity parameters estimated for the Sagittarius stream, QFS accesses the Gaussian covariance between spatially separated receivers and resolves the incident stream angle to within 1° quadratically faster than even highly non-Gaussian photon counting or entangled two-mode squeezing. The protocol needs only one qubit coupled separately to each detector at different times (realizable via ancilla state teleportation) and no long-range-entangled probe.

• Wireless communication application (simulation). For 8-QAM and 64-QAM symbol identification, a QFS receiver with three qubits (enabling at least 8-class classification per readout) requires ~100× fewer shots than a two-mode-entangled quantum receiver, which in turn requires ~100× fewer shots than a classical heterodyne receiver (~10^4× fewer than classical overall) in the weak-signal regime. For single-shot decoding, QFS achieves large success probabilities at far smaller field strengths.

• No exotic physics needed. The signal is purely classical and induces Gaussian displacements; no engineered non-Gaussianity is required. Noise in the signal can be absorbed into the distribution P, so distribution-agnostic protocols retain exponential advantage under Markovian, white, or thermal noise backgrounds.

• Simplification of prior entanglement advantages. In many cases the entanglement-enabled advantages of earlier work (including full tomography of a worst-case multimode displacement channel) reduce to entanglement-free QFS advantages between Gaussian and non-Gaussian strategies; QΨ also simplifies the corresponding lower-bound proof.

Methodology in Plain English

The signal is modeled as a random displacement of a continuous-variable sensor mode — a distribution P(α) over phase-space points α = x + ip, which is exactly the kind of thing a microwave cavity or interferometer sees. Fourier coefficients of the signal correspond to phase-space modes of that channel, with higher k resolving finer structure.

The theoretical engine, QΨ, works by writing both the target observable and the response functions available to a given hardware class (classical coherent states; Gaussian probes with squeezers; sensor plus qubit; sensor plus memory qubit) as functions on phase space. The accessible feature information is the overlap between them. The authors prove that any strategy in that class — including adaptive measurements and arbitrary classical postprocessing — needs Ω(1/AFI) measurements, and that the same construction gives a protocol achieving that scaling, so the bound is tight. Exponential advantage is certified when the AFI ratio between the enhanced and conventional families grows as exp(Ω(k)).

Experimentally, a transmon qubit is capacitively coupled to a microwave cavity; the classical signal is a calibrated resonant microwave pulse on the cavity, and qubit–cavity entanglement comes from the cross-Kerr interaction. Control unitaries combine single-qubit rotations with echoed conditional displacements, ECD(β) = D(β/2)|g⟩⟨e| + D(−β/2)|e⟩⟨g|, giving programmable non-Gaussian control and qubit readout. The qubit is read out via an on-chip resonator. Simulations for the dark matter and communications cases were run against leading conventional baselines at equal energy.

Why This Matters

The paper shifts the narrative for near-term quantum advantage: instead of demanding large entangled states and many-qubit control, it shows that a single controllable qubit or a single qubit of memory can yield rigorous, certifiable exponential reductions in measurement count. It also supplies a general design theory (QΨ) that turns "find a quantum advantage" into an optimization over classical statistical overlaps, and it identifies single-photon nonlinearity as a sensing resource in its own right.

Real-world applications the paper discusses:

• Axionic dark matter searches: detecting daily periodicity in the incident angle of a cold axion stream as smoking-gun evidence, using covariance between spatially separated receivers.
• Wireless and satellite communications: quantum-enhanced receivers for QAM symbol decoding in the weak-signal regime, relevant to digital cable, cellular, and satellite links.
• Radar and noise spectroscopy: enabled by efficient estimation of multi-time temporal correlations of time-varying signals.
• Biomedical sensing and polyspectrum learning: listed among promising domains for QFS protocols.

Industry relevance: direct for microwave quantum sensing and superconducting-circuit hardware, quantum-enhanced receivers for communications, precision metrology and interferometry, and any setting where squeezing is already used as a resource. The low control overhead (one qubit, one control operation) is the selling point for near-term deployment.

Future Directions

• When do these gains beat mature Gaussian technology? The authors explicitly flag the open question of identifying settings where QFS's non-Gaussian control or quantum memory outperforms highly squeezed Gaussian architectures, accounting for state-preparation, control, and experimental overhead.
• Automating the discovery of quantum advantages. Because QΨ recasts architecture-and-protocol search as optimization of classical statistical overlaps, the authors propose connecting it to Bayesian experimental design, multiparameter optimization, and semidefinite programming for computational design of certifiable experiments.
• Beyond the demonstrated primitives. Appendix H lays out programs in fundamental particle detection, cosmology, noise spectroscopy, polyspectrum learning, device characterization, and signal detection, plus finer-grained separations for sensing architectures of increasing circuit depth.
• Scaling the hardware. The experiments are proof-of-principle; realizing logical ancillas that serve as both controls and long-lived quantum memory, and extending to multimode and longer-time sensing problems, remain open engineering and architectural questions.

Target Audience

Quantum sensing and metrology researchers; quantum learning theorists interested in rigorous separations with practical relevance; superconducting-circuit experimentalists; and applied physicists in communications, dark matter detection, and precision measurement who want to know whether modest quantum hardware can measurably cut their data-acquisition budgets. Readers seeking a first-principles derivation should expect to work through the appendices (Sections C–G), while the main text and Figure 1 give an accessible overview.

Authors’ abstract

Quantum technology has the potential to transform scientific discovery, but quantum advantages often require processing capabilities well beyond the reach of experimental platforms. We show that coupling a single controllable qubit to an otherwise conventional sensor can exponentially reduce the number of measurements required to learn classical signals. These rigorous quantum advantages apply to fundamental sensing tasks, including learning Fourier coefficients, extracting temporal correlations from time-varying signals, and estimating transformations of physical observables. Using a superconducting cavity--qubit architecture, we experimentally demonstrate $10^7$-fold reductions in the number of measurements required for Fourier-amplitude and time-varying signal learning. Our $\textit{quantum feature sensing}$ algorithms further enable orders-of-magnitude improvements in simulations of weak-signal dark matter detection and wireless communication applications. These quantum advantages are derived from Quantum Phase-Space Inference (Q$Ψ$), a unifying theory of quantum-enhanced experiments that simultaneously converts a set of experimental objectives and constraints into tight lower bounds and optimal quantum-enhanced learning algorithms while producing a certificate of quantum advantage. Q$Ψ$ extends beyond the regimes captured by quantum Fisher information and provides a framework for systematically identifying rigorous quantum advantages in practical experimental tasks. Together, our results establish that near-term quantum technology can exponentially enhance our ability to learn from classical signals.