Time-dependent Hamiltonian Simulation of Highly Oscillatory Dynamics and Superconvergence for Schrödinger Equation

Dong An, Di Fang, Lin Lin

Main idea

How do you simulate an oscillating Hamiltonian? Many existing protocols have error in $$\| H \|$$ and in $$\| \partial H \|$$.

Idea (“qHOP”): Drop the time ordering, and use a Riemann sum to approximate. It only has a log dependence on the norm of the derivative, because of some LCU techniques.

This also achieves a second-order convergence when studying the Schrodinger equation. (This is surprising!) Numerical evidence and a proof using PDEs/pseudo-differential calculus are given.

Can be generalized to higher-order.

tags: hamiltonian-simulation