Accuracy guarantees and quantum advantage in analogue open quantum simulation with and without noise

TL;DR

This paper demonstrates that noisy analogue quantum simulations of local open quantum systems can be both computationally hard for classical computers and advantageously efficient on quantum devices, even with realistic noise levels.

Contribution

It provides theoretical evidence of quantum advantage in simulating local open quantum systems and analyzes the stability of these simulations under noise.

Findings

Quantum simulations can efficiently approximate local observables and fixed points.

Quantum advantage persists even with realistic noise levels.

Classical algorithms require superpolynomial time as noise decreases.

Abstract

Many-body open quantum systems, described by Lindbladian master equations, are a rich class of physical models that display complex equilibrium and out-of-equilibrium phenomena which remain to be understood. In this paper, we theoretically analyze noisy analogue quantum simulation of geometrically local open quantum systems and provide evidence that this problem is both hard to simulate on classical computers and could be approximately solved on near-term quantum devices. First, given a noiseless quantum simulator, we show that the dynamics of local observables and the fixed-point expectation values of rapidly-mixing local observables in geometrically local Lindbladians can be obtained to a precision of $\varepsilon$ in time that is $\text{poly}(\varepsilon^{-1})$ and uniform in system size. Furthermore, we establish that the quantum simulator would provide a superpolynomial advantage, in run-time scaling with respect to the target precision and either the evolution time (when simulating dynamics) or the Lindbladian's decay rate (when simulating fixed-points), over any classical algorithm for these problems, assuming BQP $\neq$ BPP. We then consider the presence of noise in the quantum simulator in the form of additional geometrically-local Linbdladian terms. We show that the simulation tasks considered in this paper are stable to errors, i.e. they can be solved to a noise-limited, but system-size independent, precision. Finally, we establish that, assuming BQP $\neq$ BPP, there are stable geometrically local Lindbladian simulation problems such that as the noise rate on the simulator is reduced, classical algorithms must take time superpolynomially longer in the inverse noise rate to attain the same precision as the analog quantum simulator.