An inverse problem for data-driven prediction in quantum mechanics

TL;DR

This paper addresses the inverse problem of determining a quantum system's Hamiltonian from initial and final state data, establishing uniqueness for a class of potentials, which advances data-driven quantum prediction methods.

Contribution

It formulates and proves a uniqueness theorem for recovering the Hamiltonian from finite-time state data in quantum mechanics.

Findings

Uniqueness of Hamiltonian determination from initial and final states.

Applicable to non-compactly supported potentials.

Enables prediction of quantum states without prior Hamiltonian knowledge.

Abstract

Data-driven prediction in quantum mechanics consists in providing an approximative description of the motion of any particles at any given time, from data that have been previously collected for a certain number of particles under the influence of the same Hamiltonian. The difficulty of this problem comes from the ignorance of the exact Hamiltonian ruling the dynamic. In order to address this problem, we formulate an inverse problem consisting in determining the Hamiltonian of a quantum system from the knowledge of the state at some fixed finite time for each initial state. We focus on the simplest case where the Hamiltonian is given by $-\Delta + V$, where the potential $V = V(\mathrm{t}, \mathrm{x})$ is non-compactly supported. Our main result is a uniqueness theorem, which establishes that the Hamiltonian ruling the dynamic of all quantum particles is determined by the prescription of the initial and final states of each particle. As a consequence, one expects to be able to know the state of any particle at any given time, without an a priori knowledge of the Hamiltonian just from the data consisting of the initial and final state of each particle.