Amplitude-Phase Separation toward Optimal and Fast-Forwardable Simulation of Non-Unitary Dynamics

Abstract: Quantum simulation of the linear non-unitary dynamics is crucial in scientific computing. In this work, we establish a generic framework, referred to as the Amplitude-Phase Separation (APS) methods, which formulates any non-unitary evolution into separate simulation of a unitary operator and a Hermitian operator, thus allow one to take best advantage of, and to even improve existing algorithms, developed for unitary or Hermitian evolution respectively. We utilize two techniques: the first achieves a provably optimal query complexity via a shifted Dyson series; the second breaks the conventional linear dependency, achieving fast-forwarding by exhibiting a square-root dependence on the norm of the dissipative part. Furthermore, one can derive existing methods such as the LCHS (Linear Combination of Hamiltonian Simulation) and the NDME (Non-Diagonal Density Matrix Encoding) methods from APS. The APS provides an effective and generic pathway for developing efficient quantum algorithms for general non-unitary dynamics to achieve either optimal query complexity or fast-forwarding property, outperforming the existing algorithms for the same problems.