Hybrid quantum-classical simulations of semiclassical gravity

Abstract: We propose a hybrid quantum-classical algorithm for the simulation of real-time dynamics in interacting quantum field theories coupled to classical fields, focusing on the self-consistent estimation of semiclassical backreaction. By discretizing space and time, we construct an iterative protocol that simulates the Trotterized dynamics of the quantum fields subject to the dynamical classical fields. By estimating certain quantum expectation values through a set of projective measurements, we source the equations of motion of the classical fields, and solve them numerically to feed them forward to the quantum simulation in an iterative self-consistent loop. Semiclassical backreaction is relevant in various fields of physics, particularly in cosmology, where quantum matter fluctuations affect the gravitational field dynamics, and a controlled renormalization must be carefully considered to get a sensible continuum limit. We benchmark our algorithm in this context, focusing on scalar-tensor theories of modified gravity exhibiting a chameleon mechanism, such that a light classical scalar field driving cosmic acceleration becomes massive in high-density regions, effectively screening any possible yet unobserved fifth force. By focusing on numerically tractable regimes, we explicitly show the convergence and robustness of our algorithm when considering the continuum limit and the effect of quantum shot noise. Our work paves the way for future experiments exploring other non-tractable regimes, including non-perturbative interactions of the quantum fields and how these can change backreaction and the gravitational dynamics.

• The paper develops a self-consistent algorithm that evolves quantum fields with Trotterized steps, measures renormalized observables, and feeds them into a classical background update.
• The protocol converges toward continuum semiclassical dynamics in a solvable 1+1-dimensional chameleon model, while measurement shot noise decreases with increasing shot counts and does not destabilize feedback.
• The approach offers a path to interacting, non-Gaussian semiclassical simulations, but continuum costs scale linearly with inverse lattice spacing for both qubit counts and Trotter steps, and higher-dimensional renormalization remains unresolved.

Overview

The paper develops a hybrid quantum-classical algorithm for simulating real-time dynamics in semiclassical field theories, in which interacting quantum fields evolve on dynamical classical backgrounds that are themselves sourced by quantum expectation values. The central contribution is a self-consistent iterative loop: the quantum sector is Trotterized and evolved on a quantum device (or emulated classically), renormalized expectation values of selected observables are extracted via projective measurements, these sources are fed into a numerical integrator for the classical fields, and the updated classical fields are fed forward into the next quantum evolution step. The authors benchmark the protocol on a scalar-tensor theory of modified gravity exhibiting the chameleon screening mechanism, demonstrating convergence to the continuum limit and robustness against shot noise (2603.06511).

Motivation and problem setting

Semiclassical field theories—quantum fields coupled to classical degrees of freedom through equations such as $$G_{\mu\nu} = \frac{8\pi G}{c^4}\langle\hat{T}_{\mu\nu}\rangle^{\rm ren}$$—arise across high-energy physics, strong-field physics, and cosmology. The computational difficulty is that once interactions generate non-Gaussian correlations, the quantum sector becomes classically intractable (sign problems, entanglement barriers), while the backreaction onto the background depends precisely on those non-Gaussian corrections. The paper frames this as "nonlinear semiclassical dynamics" and positions quantum simulation as the natural route beyond the Gaussian regime, where two-point functions suffice and classical methods remain efficient.

The general framework considers $n$ classical fields $\Phi$ and $m$ quantum fields $\hat\Psi$ in $D=d+1$ dimensions, with a Hamiltonian functional $$\hat{H}=\int d^dx\sum_l \lambda_l(\Phi)\hat{O}_l[\hat\Psi,\hat\Pi_\psi]$$ and classical equations of motion sourced by renormalized expectation values $\langle\hat{O}_l\rangle^{\rm ren}$. The authors restrict to homogeneous classical backgrounds $\Phi(t)$, noting the structure generalizes to inhomogeneous cases.

The hybrid algorithm

The protocol discretizes time on a uniform grid $\{t_n\}$ and alternates between sectors. The classical update uses a symplectic implicit Euler integrator, chosen because the sourcing functional is not known a priori—it must be measured at each step—so integrators requiring evaluations at future or intermediate times are inadmissible. The quantum sector uses first-order Trotter-Suzuki steps $n$0, implementable either as native gates on digital processors or as infinitesimal evolutions on analog platforms (trapped ions, Rydberg arrays, optical lattices). After each step, observables are measured, UV divergences are removed by classical post-processing (adiabatic subtraction), and the loop iterates.

A key methodological point concerns the continuum limit. Since $n$1 is not directly tunable on hardware, the bare dimensionless parameters must be retuned via RG flows as $n$2 is reduced; for quadratic Hamiltonians this reduces to mass-dimension scaling, while interacting theories require extracting $n$3-functions from comparisons of lattice-computed and dimensionful quantities. Crucially, the spatial ($n$4) and temporal limits are taken simultaneously by holding $n$5 fixed.

Benchmark: chameleon screening

The testbed is a scalar-tensor theory in $n$6 dimensions with a Dirac matter field in the Jordan frame, metric $n$7, and a Gross-Neveu-type fermion self-interaction. Fixing flat Einstein-frame metric, the conformal coupling reduces to a multiplicative mass renormalization $n$8, and the scalar equation becomes

$n$9

With the runaway potential $\Phi$0 and exponential conformal factor $\Phi$1, a vanishing condensate yields unbounded runaway evolution, whereas a nonzero condensate generates a density-dependent minimum about which $\Phi$2 oscillates—the chameleon mechanism, whereby fifth forces are screened in high-density regions. Because $\Phi$3 and the condensate are mutually determined, a self-consistent treatment is essential, making this an apt stress test of the loop.

Setting $\Phi$4 renders the fermion Hamiltonian quadratic and exactly solvable by Gaussian methods, providing a quantitative benchmark without running quantum hardware. The lattice implementation uses staggered fermions (which remove doublers via a reduced Brillouin zone), periodic boundary conditions, and time-dependent normal ordering—a zeroth-order adiabatic subtraction using instantaneous vacuum mode functions—to render the sourced condensate UV-finite with a well-defined continuum limit. The authors note plainly that higher-order adiabatic subtractions would be required for other theories, e.g., higher-dimensional Dirac fields.

Convergence results

The comparison against exact continuum solutions isolates two error sources: lattice artifacts and finite $\Phi$5. The $\Phi$6 distance $\Phi$7 between algorithm-generated and continuum trajectories exhibits non-monotonic behavior at large $\Phi$8 attributable to lattice effects, but converges monotonically once the UV cutoff $\Phi$9 exceeds the dynamically relevant scale. This substantiates the claim that spatial and temporal continuum limits can be taken simultaneously while preserving correct semiclassical dynamics—an essential consistency check for any hardware deployment.

Shot-noise analysis treats each measurement outcome as a random variable whose variance scales as $m$0. Across $m$1 repetitions, both the mean $m$2 distance to the noiseless trajectory and its statistical spread vanish as $m$3 grows, showing that measurement-induced stochasticity does not destabilize the feedback loop despite the measurements backreacting on the classical evolution.

Resource scaling and limitations

The resource requirements impose concrete constraints. Holding $m$4 fixed forces the number of Trotter steps to grow as $m$5, deepening circuits linearly in $m$6; holding physical size fixed forces qubit counts to scale as $m$7. These combined scalings bound how close to the continuum limit near-term devices can reach. The demonstrated regime is deliberately restricted: the benchmark uses free fermions ($m$8) in $m$9 dimensions with homogeneous backgrounds, so the full promise of the method—interacting, non-Gaussian, non-perturbative regimes inaccessible to classical methods—is asserted rather than demonstrated. The renormalization prescription validated here (zeroth-order adiabatic subtraction) suffices only for the specific operator and dimensionality considered. Extension to genuine general-relativistic backreaction in $\hat\Psi$0 dimensions remains contingent on simulating quantum matter in higher dimensions, which the paper leaves open.

Conclusion

This work establishes a controlled, self-consistent hybrid quantum-classical protocol for nonlinear semiclassical dynamics, validated quantitatively in a solvable chameleon-scalar-tensor model where simultaneous spacetime continuum limits converge and shot noise averages out. The framework's value lies in its modularity—swapping classical equations of motion and lattice Hamiltonians adapts it to inflationary dynamics or modified gravity—and its stated target is the genuinely interacting, non-perturbative regime where classical methods fail. Whether the RG-based parameter tuning and adiabatic subtraction generalize efficiently to interacting theories and higher dimensions is the principal open question the paper identifies.