Quantum Supremacy: Theory and Experiment

• Quantum supremacy is the demonstration that quantum devices perform specialized sampling tasks beyond the reach of classical algorithms, validated by complexity-theoretic proofs and empirical benchmarks.
• It involves experimental protocols such as deep random circuit sampling and cross-entropy benchmarking on noisy intermediate-scale quantum systems, establishing a gap with classical methods.
• The milestone drives progress in both computational theory and hardware innovation, continuously adapting as classical simulation techniques advance.

Quantum supremacy denotes the realization of a quantum device, typically non-error-corrected and operating in the noisy intermediate-scale quantum (NISQ) regime, performing a well-defined computational task with such efficiency and solution quality that no known classical algorithm executed on state-of-the-art supercomputers can match its performance in reasonable time. The canonical instantiation involves sampling from the output distribution of pseudo-random quantum circuits, manifesting behavior associated with quantum chaotic systems and characterized by high entanglement, anti-concentration, and output statistics typical of the Porter–Thomas distribution (Boixo et al., 2016). The eventual goal is a separation between quantum and classical computational resources, validated through complexity-theoretic reductions and empirical benchmarks.

1. Formal Definition and Motivation

Quantum supremacy is defined as the ability of a quantum device—without full error correction—to perform a computational task that is not feasibly achievable by any classical algorithm on contemporary supercomputers. This is operationalized by (1) specifying a computational or sampling task (such as random circuit sampling), (2) defining a quantifiable measure of solution quality or distributional closeness (e.g., total variation distance, Δ, or cross-entropy benchmarking (XEB) scores), and (3) establishing, often through complexity-theoretic arguments, that no classical algorithm with polynomial resources can achieve equivalent performance under plausible assumptions (Boixo et al., 2016, Bouland et al., 2018).

Formally, in sampling settings, a quantum device achieves supremacy if it produces samples from a distribution $D_\text{exp}$ that is within some small total variation distance $\epsilon$ of the ideal quantum distribution $D_\text{ideal}$, and for which there exists no classical scheme to produce samples within comparable $\epsilon$ in polynomial time unless the polynomial hierarchy collapses (Reddy et al., 2021). Supremacy, therefore, is inherently a computational complexity statement about the asymptotic scaling of resources (time, memory) with system size and noise.

2. Complexity-Theoretic Foundation

Quantum supremacy claims rely critically on complexity-theoretic reductions and intractability arguments:

• #P-hardness and Stockmeyer Reduction: For random quantum circuits (RCS) and related models, the central hardness argument is that evaluating an individual output probability is #P-hard in the worst case. Stockmeyer’s approximate counting theorem links this to average-case hardness and demonstrates that if a classical sampler could sample close to the quantum output in total variation, one could then multiplicatively approximate individual probabilities, collapsing the polynomial hierarchy (Boixo et al., 2016, Bouland et al., 2018, Movassagh, 2019).
• Anti-Concentration: Random quantum circuits quickly yield output statistics governed by the Porter–Thomas (exponential) distribution, ensuring that output probabilities are not too concentrated. This prevents classical algorithms from trivially approximating the distribution and is crucial for lifting additive-error hardness to multiplicative hardness (Boixo et al., 2016, Tangpanitanon et al., 2020).
• Average-case Hardness: Robustness to estimation or sampling error is essential. For RCS, average-case hardness is established by showing that computing output probabilities for a typical (random) circuit remains #P-hard up to exponentially small additive error (Movassagh, 2019, Bouland et al., 2018). Similar arguments underlie supremacy in analog Floquet systems and IQP circuits (Tangpanitanon et al., 2020, Lidar, 8 Dec 2025).
• Threshold theorems in noise: It is proved that even in noisy, pre-threshold devices, postselection techniques can simulate ideal outputs up to exponentially small additive error; classical simulation of these outputs with multiplicative error better than $\sqrt{2}$ would collapse the polynomial hierarchy. For example, the surface code threshold for supremacy (i.e., complexity-theoretic hardness) is 2.84%, significantly above the threshold for universal quantum fault tolerance (Fujii, 2016).

3. Experimental Protocols and Benchmarks

Quantum supremacy demonstrations primarily employ sampling tasks—especially random circuit sampling—on quantum processors such as superconducting qubits and photonic platforms. The main experimental steps are:

• Circuit Design: Implementing deep random circuits on 2D qubit arrays, with randomness in both single- and two-qubit gates, designed to rapidly approach the Porter–Thomas output regime and maximize circuit entropy (Boixo et al., 2016, Arute et al., 2019).
• Sample Acquisition: Measuring all qubits in the computational basis after circuit execution, collecting a large number of bit-strings to empirically reconstruct the output distribution.
• Cross-Entropy Benchmarking (XEB): For each sampled bit-string $x_j$, compute $p_U(x_j)$ classically (feasible up to $n \sim 48$), and evaluate the XEB fidelity:

$$\Delta \approx H_0 - \frac{1}{m} \sum_{j=1}^m \log[1/p_U(x_j)]$$

where $H_0 = \log N + \gamma$ and $\gamma \approx 0.577$ is Euler’s constant. For noisy outputs, circuit fidelity $\alpha$ is linearly related to the XEB score (Boixo et al., 2016).

• Verification: For circuit sizes and depths beyond classical reach ($n \gtrsim 50$), direct simulation is infeasible. The experimental protocol extrapolates XEB scores from classically tractable regimes and compares them with theoretical error models; agreement justifies extending the XEB/fidelity estimates to the supremacy regime (Boixo et al., 2016, Arute et al., 2019).

4. Paradigms Beyond Digital Quantum Circuits

Recent research generalizes supremacy beyond standard gate-based systems:

• Floquet-Driven Many-Body Systems: Periodically driven, disordered quantum chains and Bose–Hubbard systems can realize quantum-supreme sampling tasks by evolving under time-dependent Hamiltonians whose effective Floquet operators thermalize to the Circular Orthogonal Ensemble (COE), ensuring Porter–Thomas statistics and anti-concentration. The sampling from stroboscopically evolved states is classically hard for system sizes $L \sim 20-50$, as established via ETH, complexity-theoretic conjectures, and numerical simulation (Tangpanitanon et al., 2020, Thanasilp et al., 2020, Tangpanitanon et al., 2019).
• Digital–Analog–Digital Quantum Computing (DADQC): Hybrid circuits, where analog Ising-type blocks are sandwiched between digital single-qubit layers, can closely approximate Instantaneous Quantum Polynomial-time (IQP) sampling problems believed to be classically intractable. For all-to-all connectivity and certain bounded-degree hardware graphs, rigorous total variation distance bounds and proven anticoncentration are achieved (Lidar, 8 Dec 2025).

5. Limitations, Verification Challenges, and the Evolving Supremacy Frontier

While initial supremacy claims—such as those with the Google Sycamore processor—were supported by large compute-time gaps between quantum and classical sampling, subsequent advances in classical simulation have dramatically narrowed these gaps (Huang et al., 2020, Liu et al., 2021, Wold et al., 8 Dec 2025). For example, classical tensor-network simulators and hybrid CPU/GPU pipelines now reach or exceed quantum sampling performance for certain configuration and fidelity regimes. This necessitates continual recalibration of the supremacy threshold based on current best-in-class classical approaches.

Verification of quantum supremacy is epistemically challenging:

• Verification Beyond Classical Frontier: In the deep supremacy regime, classical computation of ideal probabilities $p_U(x)$ for XEB benchmarking is infeasible. Extrapolation from small-size, low-depth circuits—anchored by detailed noise models—is employed, but this approach is susceptible to unmodeled errors or correlations and undermines absolute confidence (Reddy et al., 2021, Kalai et al., 2023).
• Limitations of Statistical Tests: Existing metrics, such as cross-entropy difference and heavy-output generation, are necessary but insufficient to guarantee closeness in total variation distance, especially in the presence of correlated or adversarial noise (Bouland et al., 2018, Rinott et al., 2020).
• Noise Models and Calibration: Product-form fidelity predictions (e.g., Google’s formula (77)) rely on statistical independence assumptions about errors. Empirical data often corroborate these despite the lack of physical justification; calibration anomalies and unexplained statistical correlations raise concerns about the verifiability and generality of supremacy claims (Kalai et al., 2023, Rinott et al., 2020).

6. Summary Table of Quantum Supremacy Regimes

Regime / Task	Complexity Argument	Platform/Scale	Output Statistic	XEB / Fidelity Example	Classical Barrier (status)
Random Circuit Sampling	#P-hard via Stockmeyer + anti-concentration	Sycamore, Zuchongzhi: 53–66 qubits, depth 20–24	Porter–Thomas, XEB	F ≈ 0.002	10⁴ yrs (2019) → days (2023)
Floquet Analog Chains	COE thermalization + ETH, #P-hard IQP embedding	Driven cold atoms, ions: L ≈ 20–50	Porter–Thomas, KL divergence	-	Polynomial in L for MBL, exponential for COE
DADQC Hybrid Circuits	TV distance to IQP, average-case #P-hardness	Annealer/Ion QPUs, n ≈ 50–100	Anticoncentration prover	-	Conjectural, under PH and Ising hardness

Benchmarks such as XEB fidelity ($\Delta$ or $F_{XEB}$); circuits that yield $\Delta_\text{exp} \gg$ classical best known $\Delta \sim 0$ represent empirical evidence for quantum supremacy (Boixo et al., 2016).

7. Outlook and Conclusions

Quantum supremacy is now better understood as a dynamic threshold, not a static milestone. It demarcates a frontier that shifts with advances in both quantum hardware (qubit count, coherence, gate fidelity), error modelling, and classical simulation methods. Demonstration of supremacy rests on a convergence of complexity-theoretic evidence (#P-hardness, anti-concentration, average-case intractability); experimental metrics (XEB, KL divergence, heavy-output fraction); careful noise calibration and error budgeting; and rigorous, transparent verification protocols (AbuGhanem et al., 2023, Horner et al., 2020). Claims of supremacy must be revised in light of evolving classical capabilities, emphasizing the need for continual reassessment and refined statistical validation. As new quantum and hybrid classical-quantum methodologies emerge, the community must maintain robust, complexity-grounded standards for the operational meaning and verification of quantum supremacy.