Structure-Preserving Quantum Simulation of Wave Equations on a Trapped-Ion Processor

AI summary

Overview

Research area: Quantum simulation of partial differential equations (PDEs) on near-term quantum hardware — specifically structure-preserving, Fourier-based circuits for acoustic wave equations and variable-mass Dirac dynamics, benchmarked on a trapped-ion processor.

Technical level: Intermediate. The paper is written for readers comfortable with finite-difference discretization, the quantum Fourier transform, and Trotter/Strang product formulas; the conceptual core (rotation-based mode evolution, exact removal of directional splitting error) is accessible without the full math.

Scope (1 sentence): The authors build and execute quantum circuits for one- and two-dimensional acoustic waves and one-dimensional Dirac dynamics with a spatially varying mass on Quantinuum's H2-2 trapped-ion device, measuring subdomain kinetic energies rather than reconstructing full fields, and report both logical and hardware-native gate resources.

What This Paper Is About

Wave equations are a natural early target for quantum simulation of PDEs, but hardware demonstrations have so far been limited in spatial dimension, equation class, system size, and in whether the output is physically meaningful. This paper designs a common, structure-preserving discretization-plus-QFT circuit framework for three model problems and runs it on a real trapped-ion processor at grid sizes up to 1024 points in one dimension and 32 × 32 in two dimensions. The goal is not to claim a quantum advantage but to measure empirically how much accurate, observable-level dynamics survives on a present-day NISQ device.

Key Contributions

1. A common structure-preserving framework: A Hermitian first-order finite-difference formulation combined with the quantum Fourier transform gives one compilation and hardware pipeline for one-dimensional acoustic waves, two-dimensional acoustic waves, and one-dimensional Dirac dynamics with a spatially varying mass.

2. An exact bright–dark decomposition for the 2D acoustic problem: Each two-dimensional Fourier block is rotated into a longitudinal "bright" velocity component that couples to pressure and a transverse "dark" component that decouples. Within the constant-coefficient, periodic semi-discrete model this reduction is exact and eliminates the directional product-formula (splitting) error used in prior Fourier-space constructions.

3. A resource benchmark under two independent scans: Logical and H2-2-native total and two-qubit gate counts and depths are reported as functions of evolution time at fixed register size, and as functions of grid-qubit count n_h at fixed time t = 0.1, with the n_h scan extending to n_h = 50.

4. Physical H2-2 executions for all three model problems: Results are compared against a classical reference, ideal finite-shot sampling (8192 shots), the H2-2E noise-modeled emulator where available, and H2-2 measurements (1024 shots) of half-domain marginal observables — kinetic-energy fractions for the acoustic problems and a spinor-component population for the Dirac problem.

Main Findings

• Hardware agreement with classical references: Across all tested acoustic and Dirac dynamics problems, H2-2 results track the classical kinetic-energy dynamics with mean absolute errors between 5.9 × 10⁻³ and 2.4 × 10⁻².

• Largest error comes from the deepest circuits: The maximum error in that range occurs for the deeper split-operator Dirac circuits, which alternate QFT-based kinetic propagators with position-space mass rotations.

• Largest one-dimensional hardware instance: The experiments include one-dimensional grids with up to 1024 points (n_h = 10), exceeding the n_h = 6 (N_h = 64) physical-device instance of the earlier low-mode QFT demonstration (Wright et al., on Quantinuum H1-1) and reaching the size that prior work only studied numerically.

• Two-dimensional scale: The 32 × 32 two-dimensional grids correspond to an encoded state-space dimension of up to 4096.

• Gate scaling with grid qubits: At fixed retained bandwidth, compiled gate counts grow approximately quadratically with the number of grid qubits; for the one-dimensional acoustic circuit, compiled depth grows approximately linearly in n_h while both quantities are approximately quadratic for the two-dimensional circuit.

• Time independence of the acoustic circuit: Acoustic circuit sizes are essentially independent of evolution time, whereas cost also grows with the number of product-formula steps (relevant to the split-operator Dirac case).

• No advantage claimed: The results benchmark structured inputs and observable-level outputs; the authors explicitly state they do not establish a quantum advantage over classical PDE solvers.

Methodology in Plain English

The researchers first rewrite each wave equation as a first-order system (pressure and velocity, or the two spinor components of the Dirac equation) and discretize space with finite differences arranged so that the resulting generator stays Hermitian — meaning the discrete evolution is exactly unitary, like a genuine quantum system. Because the difference operators are circulant under periodic boundary conditions, the quantum Fourier transform diagonalizes them, so time evolution reduces to a set of mode-dependent rotations on an ancilla or field qubit, controlled by the Fourier-mode register.

For one dimension, the nonlinear mode dependence is linearized for the low-frequency modes occupied by smooth inputs (following Wright et al.), turning the propagator into controlled rotations. For two dimensions, instead of approximating the two noncommuting directional terms with a product formula, they rotate each Fourier block into a bright/dark basis so the block can be handled exactly. For the Dirac problem with a mass that jumps from 0 to 2 at the midpoint of the domain, they use Strang splitting: alternating position-space mass rotations (implementable with a single controlled rotation because the step aligns with the most-significant grid qubit) with Fourier-space kinetic propagation.

The evaluation pipeline has four stages: an exact classical reference solution in NumPy/SciPy; Qiskit assembly and ideal AerSimulator evaluation with 8192-shot sampling to get logical resources and noiseless observable curves; retargeting to the H2-2 native gate set via pytket-qiskit and the qnexus cloud interface to obtain compiled resources; and finally execution on H2-2E and H2-2 with 1024 shots, a count kept low because hardware quantum credits scale with shots. Rather than tomographically reconstructing the fields, they estimate subdomain kinetic energies directly from measurement samples.

Why This Matters

This is hardware-level evidence that structured wave problems with thousands of encoded degrees of freedom can still yield resolvable observable dynamics on a present-day trapped-ion machine — a useful datapoint separating what survives NISQ noise from what requires error correction. It also shows that choosing the right algebraic structure (exact bright–dark block diagonalization instead of Trotterizing noncommuting directions) can remove an error source before compilation rather than paying for it in circuit depth.

Real-world applications cited in the paper:

• Seismic modeling and imaging, where the acoustic wave equation is the canonical model.
• Acoustic and electromagnetic wave propagation more broadly.
• Fluid flow and heat transport, which the authors list among the PDE-modeled phenomena motivating quantum PDE solvers.
• Quantum dynamics itself, via the variable-mass Dirac benchmark with an interface between gapped and gapless regions.

Industry relevance: The work is directly relevant to trapped-ion hardware vendors and cloud quantum-computing providers (Quantinuum H2-2, qnexus, pytket), to compiler and resource-estimation toolchains (Qiskit, pytket), and to computational geophysics and scientific-computing groups evaluating whether near-term quantum devices can contribute to wave-based simulation workloads.

Future Directions

• Moving beyond the deliberately limited scope: The authors restrict themselves to periodic problems, structured low-bandwidth inputs, and coarse marginal observables; extending to non-periodic boundaries, general initial data, and richer output observables is the obvious next step.
• Adaptive rather than fixed low-frequency angle compilation: In the present hardware experiments the mode-dependent angles are compiled for a fixed low-frequency set, so generalizing to arbitrary retained bandwidths and inputs is an open circuit-design question.
• Pushing on product-formula cost: The largest observed error came from the deeper Strang-split Dirac circuits; reducing the number of splitting steps or finding exact decompositions for heterogeneous media is a clear target.
• Scaling the two-dimensional case: The exact bright–dark reduction is established for the constant-coefficient, periodic semi-discrete model; whether analogous exact block reductions exist for variable-coefficient or heterogeneous 2D media remains open.

Target Audience

Quantum-algorithm researchers working on PDEs and Hamiltonian simulation; experimentalists and compiler engineers benchmarking application-level circuits on trapped-ion or other NISQ hardware; computational scientists in geophysics and wave propagation curious about realistic near-term quantum resources; and graduate students who want a concrete, end-to-end example of going from a discretized PDE to compiled native gate counts and device measurements.

Authors’ abstract

Wave equations provide a natural testbed for near-term quantum simulation of partial differential equations, but hardware demonstrations have remained limited in spatial dimension, equation class, system size, and physically meaningful output. We develop and benchmark structure-preserving, Fourier-based quantum circuits for the one- and two dimensional acoustic wave equations and Dirac dynamics with variable mass on the Quantinuum H2-2 trapped-ion processor. The experiments include one-dimensional grids with up to \(1024\) points and \(32\times32\) two-dimensional grids, corresponding to an encoded state-space dimension of up to \(4096\). Rather than reconstructing the full fields, we estimate subdomain kinetic energies directly from measurement samples. Across all tested acoustic and Dirac dynamics problems, the H2-2 results track the classical kinetic-energy dynamics with mean absolute errors between \(5.9\times10^{-3}\) and \(2.4\times10^{-2}\). At fixed retained bandwidth, the compiled gate counts grow approximately quadratically with the number of grid qubits; the acoustic circuit sizes are essentially independent of evolution time, whereas the cost also grows with the number of product-formula steps. These results provide hardware-level evidence that accurate observable dynamics can remain resolvable for structured wave problems with thousands of encoded degrees of freedom on a present-day trapped-ion processor.