Resolving Structure in Prethermal Floquet Dynamics with Precision Quantum Computation

AI summary

Overview

Research area: Quantum simulation of non-equilibrium quantum many-body physics — specifically Floquet (periodically driven) prethermalization — using error-mitigated superconducting and trapped-ion quantum processors, benchmarked against state-of-the-art classical simulation.

Technical level: Advanced. The paper assumes familiarity with Floquet unitaries, entanglement and operator complexity, tensor-network methods (PEPS, TEBD), sparse Pauli-path representations, and quasi-probability-based error mitigation (PEC and ZNE).

Scope: The paper reports error-mitigated experiments on up to 74 qubits of a Floquet mixed-field Ising magnet on a heavy-hex lattice, resolves subharmonic prethermal oscillations beyond the reach of the classical methods tested, and uses finite-size scaling to argue that the oscillatory response survives in the thermodynamic limit of heavy-hex ladders.

What This Paper Is About

Periodically driven quantum magnets can settle into a "prethermal" regime where local observables keep oscillating in a structured way long before the system eventually heats up to a featureless state. Reaching the large system sizes and long times where this regime becomes physically meaningful is hard: classical simulation loses control, and real quantum hardware accumulates noise that distorts the measured values. The paper's goal is to show that a precision error-mitigated quantum processor can measure these dynamics accurately enough to answer a genuine physics question — whether the observed oscillations persist as the system grows — that neither small exact simulations nor leading classical methods can settle.

Key Contributions

1. A precision measurement of Floquet prethermal dynamics. Using QESEM error mitigation on an IBM Heron r3 processor (ibm_boston), the authors measure magnetization dynamics with percent-level precision on 51- and 74-qubit heavy-hex geometries, resolving long-lived subharmonic oscillations over 30 Floquet cycles.

2. A quantitative demonstration that classical methods fail in this regime. Projected entangled-pair-state simulations with belief propagation (PEPS-BP, bond dimensions D = 512 and D = 700, run on an NVIDIA H100 GPU) and sparse Pauli-path simulations (via the ORQA framework, retaining O(10^12) Pauli strings across 12888 Fugaku nodes with W = 20) are shown to lose convergence or develop truncation-dependent artifacts exactly where the quantum data remains stable.

3. A finite-size scaling result for the oscillation amplitude. Combining exact simulations of 21-, 28- and 35-qubit systems with error-mitigated 51- and 74-qubit data, the authors extract oscillation amplitudes and fit them, finding that small-system data alone cannot determine whether a nonzero asymptotic amplitude exists, while the inclusion of the larger error-mitigated points gives strong evidence for one.

4. A multi-layered validation hierarchy for the quantum results. The paper builds confidence through noise-model validation on the IBM hardware, agreement between the unbiased (PEC-based) and extrapolated (ZNE-based) QESEM estimators over their shared cycle window, internal consistency across different noise-amplification factors, cross-checks against converged classical results where available, and cross-platform corroboration on Quantinuum System Model H2 and Quantinuum Helios trapped-ion hardware at selected Floquet cycles.

Main Findings

• Subharmonic prethermal oscillations persist to 74 qubits. The error-mitigated magnetization shows a slowly decaying background plus oscillations whose period is approximately four times that of the underlying Floquet drive, and the mitigated dynamics approach the prethermal plateau value predicted by the fourth-order Magnus-expansion effective Hamiltonian after roughly ten Floquet cycles.

• The oscillation amplitude decays surprisingly slowly with system size. In the exponential finite-size fit A(N_q) = a·exp(−√N_q/ξ) + c, fitting only the exactly simulated small systems gives c = 7.3 × 10^−3 with a central 68% interval of (−3.7 × 10^−3, 1.2 × 10^−2), which includes zero. Fitting all available sizes gives c = 9.6 × 10^−3 with a central 68% interval of (8.1 × 10^−3, 1.1 × 10^−2), and ξ = 0.7 with a central 68% interval of (0.6, 0.9). The decay scale is comparable to the independently extracted correlation length ξ′ = 1.1 ± 0.1.

• A parameter regime that balances entanglement and signal. Exact 21-qubit parameter scans identified a point near the Pareto front, θ_x ≈ π/6, θ_z ≈ π/27, θ_zz = π/3, equivalently (3θ_x, 3θ_z, θ_zz) = (1.57, 0.34, 1.05). At this point, in the 28-qubit exact simulation the normalized half-system entanglement entropy exceeds S = 1/2 by N_c = 10 Floquet cycles while structured magnetization oscillations survive. At θ_z = 0 the system also shows prethermal behavior but the magnetization relaxes rapidly; a more strongly entangling point (3θ_x, 3θ_z, θ_zz) = (1.65, 1.32, 1.05) thermalizes without structured oscillations.

• QESEM-Unbiased provides a trusted window to about cycle 11–16. The unbiased estimator agrees with converged classical results through approximately 11 Floquet cycles and remains statistically feasible through cycle 16, establishing a quantitatively trusted reference. Beyond that, the sampling overhead becomes prohibitive and QESEM-Extrapolated takes over, agreeing with the unbiased estimator within statistical uncertainties over their common window.

• PEPS-BP loses convergence around cycle 12. The D = 512 and D = 700 calculations agree with each other and with the quantum data through roughly 12 cycles (with D = 700 likely accurate for one to two additional cycles), then separate from each other and from the data, developing a nonphysical upward drift that exceeds the effective-Hamiltonian prethermal prediction at approximately 16 cycles.

• Sparse Pauli-path simulations fail to reproduce the oscillations. Two 51-qubit SPP runs with different truncation settings diverge from each other after approximately seven cycles, and around N_c = 15 show a clear phase mismatch with the mitigated data — a maximum where the quantum data shows a minimum. A 21-qubit diagnostic where all 4^21 Pauli strings can be retained shows the oscillations are recovered only when roughly 10–20% of the full Pauli basis is kept.

• Heuristic-corrected TEBD produces the wrong size dependence. Heuristic-Corrected TEBD with χ = 4096, combined with statevector results for smaller systems, yields a non-monotonic system-size dependence of the oscillation amplitude, unlike the monotonic decrease seen in the combined statevector and QESEM results.

• A substantial circuit-volume boost. In the runtime model of Fig. 1b, reaching 30 Floquet cycles at fixed target accuracy corresponds to a circuit-volume boost of approximately B_V ≃ 30 relative to unmitigated execution. The model assumes an effective two-qubit-gate infidelity I_F = 2.4 × 10^−3 representative of the ibm_boston experiments, with QESEM runtimes normalized to absolute magnetization precisions of ε = 0.02 for QESEM-Unbiased and ε = 0.005 for QESEM-Extrapolated.

• Cross-platform agreement. At the Floquet cycles evaluated on Quantinuum hardware, the Quantinuum and IBM estimates agree within statistical uncertainties, despite using different mitigation heuristics (characterization-free unitary gate folding and extrapolation) on platforms with distinct noise characteristics.

Methodology in Plain English

The authors define a repeating circuit — one "Floquet cycle" — built from layers of single-qubit X and Z rotations applied to every qubit, interleaved with three edge-coloured layers of nearest-neighbour ZZ rotations at a fractional angle. The three-colouring of the heavy-hex lattice lets each ZZ layer run in parallel, so the circuit maps naturally onto hardware connectivity. Starting from a fully ferromagnetic state (all qubits in |0⟩), they track the average magnetization M = (1/N_q) Σ⟨Z_q⟩ and the normalized half-system entanglement entropy.

Before touching hardware, they use exact statevector simulations of 21-, 28- and 35-qubit systems to search parameter space for a point where entanglement grows quickly but a measurable magnetization signal survives 30 cycles — a compromise between having a genuinely complex quantum state and still having something to measure. They also track operator entanglement, a diagnostic of how complex the observable's evolution becomes.

On hardware, they use QESEM, a software framework that first characterizes the noise of each gate layer, then constructs quasi-probability representations to either cancel the noise statistically (QESEM-Unbiased, based on probabilistic error cancellation) or amplify and extrapolate it away at lower cost (QESEM-Extrapolated, based on zero-noise extrapolation). Because the unbiased version is only as trustworthy as the noise model, the authors validate the model independently and compare against exact or converged classical results over the window where those exist.

To establish that the quantum results are not artifacts, they compare against four classical approaches: PEPS-BP (Schrödinger-picture two-dimensional tensor networks), sparse Pauli-path propagation (Heisenberg-picture, run with ORQA on Fugaku), a noisy variant of Pauli paths with zero-noise extrapolation, and a heuristic correction to one-dimensional TEBD. Finally, they fit the late-cycle traces to a decaying background plus a cosine, extract the amplitude A for each system size, and fit A(N_q) to an exponential-plus-offset form using √N_q as the scaling variable appropriate for a two-dimensional geometry.

Why This Matters

Impact on research. The paper reframes error-mitigated quantum processors as quantitative scientific instruments rather than demonstrations. It provides a concrete example where the quantum device resolves a physics question — whether a prethermal oscillatory response survives in the thermodynamic limit — that neither exact small-system simulation nor the leading classical approximate methods tested could answer. It also supplies a template for how to argue that mitigated data is trustworthy: noise-model validation, estimator cross-agreement, exact-simulation anchoring, and cross-platform replication.

Real-world applications (potential, as motivated by the work):

• Characterizing periodically driven quantum materials and Floquet-engineered systems, where the drive period and interaction strength are experimental knobs.
• Benchmarking and stress-testing classical simulation algorithms (tensor networks, Pauli-path methods) by providing trusted reference data in regimes where those algorithms break down.
• Validating error-mitigation stacks for near-term quantum computing workloads in materials and chemistry, where expectation values must be quantitatively accurate, not just qualitatively indicative.
• Using the effective-Hamiltonian prethermal plateau as a calibration and verification target for driven quantum devices more broadly.

Industry relevance. The work is a multi-organization collaboration spanning a quantum software company (Qedma), a quantum cloud/algorithm provider (BlueQubit), IBM Quantum, RIKEN, and academic institutions, and it explicitly quantifies runtime and accuracy tradeoffs against classical hardware including Fugaku and NVIDIA H100 GPUs. It gives hardware and software vendors a concrete accuracy-versus-runtime frontier to argue about, and it demonstrates that cross-vendor corroboration (IBM superconducting, Quantinuum trapped-ion) is now feasible for a nontrivial many-body problem.

Future Directions

• Pushing to later cycles and larger lattices. QESEM-Unbiased becomes statistically demanding beyond cycle 16 and QESEM-Extrapolated carries extrapolation bias; reducing both the sampling overhead and the bias would extend the trusted window and allow larger heavy-hex geometries than 74 qubits.
• A converged classical reference in the same regime. The 21-qubit diagnostic showing that roughly 10–20% of the full Pauli basis must be retained implies a converged SPP calculation at 51 qubits would need resources far beyond present-day classical computation; developing methods that close this gap, or definitively bounding them, remains open.
• Sharpening the finite-size scaling argument. The authors note that exponential decay in √N_q, exponential decay in N_q, and algebraic decay are all plausible phenomenological forms, and that three small-system points with three fit parameters leave the large-size trend essentially unconstrained. Additional system sizes, and a first-principles justification for the scaling ansatz, would strengthen the thermodynamic-limit claim.
• Understanding the mechanism. The paper observes that these circuits have no obvious mechanism for stabilizing a long prethermal window (no high-frequency drive, no small-step Trotter structure, no disorder), yet finds parameters where it happens. Explaining why this parameter regime prethermalizes, and how the period-four subharmonic response arises, is left as a physics question.

Target Audience

Quantum computing researchers and practitioners working on error mitigation and near-term applications; condensed matter and statistical physicists interested in Floquet prethermalization, discrete time crystals, and non-equilibrium phases; developers of classical simulation methods (tensor networks, Pauli-path techniques) who need benchmarks in the regime where their methods fail; and quantum hardware vendors and cloud providers who need to characterize where their devices deliver quantitative scientific value. The paper is not beginner-friendly — it presupposes comfort with quantum many-body formalism, quasi-probability error mitigation, and tensor-network approximations.

Authors’ abstract

Periodically driven interacting quantum many-body systems can exhibit long-lived prethermal dynamics, where local observables retain coherent structure even as entanglement and operator complexity grow. Accessing this regime at the system sizes and times needed to determine physical properties of the prethermal state remains a central challenge: state-of-the-art classical methods become unreliable, while noise in quantum hardware degrades observable expectation values. Here we overcome these limitations for a Floquet Ising magnet realized on a heavy-hex lattice. Using the advanced error mitigation software QESEM on an IBM Heron r3 superconducting quantum processor, we measure magnetization dynamics with percent-level precision and resolve long-lived subharmonic prethermal oscillations in systems of up to 74 qubits. These experiments reach regimes for which leading tensor-network simulations fail to converge, while sparse Pauli-path simulations remain strongly truncation dependent despite extensive computations on advanced GPUs and the Fugaku supercomputer. Leveraging this quantum-accessible regime, we extend finite-size scaling to larger systems and find an unexpectedly slow decrease of the oscillation amplitude with system size, providing strong evidence that this oscillatory response persists in the thermodynamic limit of heavy-hex ladders. A hierarchy of mitigation and validation tests, including unbiased error mitigation, agreement between independent mitigation estimators, noise-model validation on the superconducting hardware, and cross-platform corroboration at selected Floquet cycles on Quantinuum System Model H2 and Quantinuum Helios trapped-ion hardware, supports the reliability of these findings. Our work establishes error-mitigated quantum processors as quantitative scientific instruments for discovering new physics in non-equilibrium quantum matter.