Analog quantum simulation of partial differential equations

Shi Jin,Nana Liu

2023-12-06

Abstract:Quantum simulators were originally proposed for simulating one partial differential equation (PDE) in particular - Schrodinger's equation. Can quantum simulators also efficiently simulate other PDEs? While most computational methods for PDEs - both classical and quantum - are digital (PDEs must be discretised first), PDEs have continuous degrees of freedom. This suggests that an analog representation can be more natural. While digital quantum degrees of freedom are usually described by qubits, the analog or continuous quantum degrees of freedom can be captured by qumodes. Based on a method called Schrodingerisation, we show how to directly map D-dimensional linear PDEs onto a (D+1)-qumode quantum system where analog or continuous-variable Hamiltonian simulation on D+1 qumodes can be used. This very simple methodology does not require one to discretise PDEs first, and it is not only applicable to linear PDEs but also to some nonlinear PDEs and systems of nonlinear ODEs. We show some examples using this method, including the Liouville equation, heat equation, Fokker-Planck equation, Black-Scholes equations, wave equation and Maxwell's equations. We also devise new protocols for linear PDEs with random coefficients, important in uncertainty quantification, where it is clear how the analog or continuous-variable framework is most natural. This also raises the possibility that some PDEs may be simulated directly on analog quantum systems by using Hamiltonians natural for those quantum systems.

Quantum Physics

What problem does this paper attempt to address?

This paper discusses how to efficiently simulate types of partial differential equations (PDEs) other than the Schrödinger equation using analog quantum systems. Traditional numerical methods often require the discretization of PDEs, but this approach faces the challenge of "curse of dimensionality" in high-dimensional problems. The paper proposes a new method called "Schrödingerisation," which allows for the direct mapping of a D-dimensional linear PDE onto a (D+1)-dimensional quantum system using continuous variables (qumodes) for simulation without the need for prior discretization. This method is applicable not only to linear PDEs but also to certain nonlinear PDEs and nonlinear ordinary differential equation systems. By utilizing the potential of quantum simulators, this approach reduces the need for large-scale quantum systems, allowing for more efficient handling of various PDEs.

The paper provides examples of important PDEs, including the Liouville equation, heat equation, Fokker-Planck equation, Black-Scholes equation, wave equation, and Maxwell's equations. Additionally, the paper presents a new quantum algorithm for linear PDEs with stochastic coefficients, which is particularly important in the field of uncertainty quantification. The paper emphasizes maintaining the continuity properties of PDEs during the simulation process and notes that further gate decomposition of PDEs in the quantum system may not be necessary in certain cases.

Overall, the goal of this paper is to extend the application scope of quantum simulators and utilize their ability to simulate PDEs in order to overcome limitations of classical computation in handling high-dimensional and complex dynamic systems.