Southwest Quantum Information and Technology

Generative Quantum Machine Learning for Quantum Chemistry

Presenting Author: Torin Stetina, University of California Berkeley
Contributing Author(s): Jack Ceroni, Juan Miguel Arrazola, Carlos Ortiz, Mária Kieferová, Nathan Wiebe

The potential energy surface (PES) of molecules with respect to their nuclear positions is a primary tool in understanding chemical reactions from first principles. In short, the PES is a high dimensional landscape that in principle, contains all the information needed to predict reaction rates. Each point on the PES is associated with a parameterized Hamiltonian, defined by its nuclear coordinates. However, obtaining this information is complicated by the fact that at each coordinate on the PES, ground state energies must be computed, which is in general a QMA-Complete problem. With some assumptions, this can be alleviated by the fact that there exist efficiently preparable high-fidelity approximations of the true ground state for many molecular systems, but sampling a large number of ground states over a high dimensional PES can require a vast number of state preparations. In this work, we investigate the utility of a generative quantum machine learning model trained using quantum data associated with the classical nuclear coordinate information, where a subset of ground state wavefunctions are sampled along the PES. In this regime, a successful generative model takes a set of classical nuclear coordinates as an input, and outputs the quantum electronic ground state at the requested nuclear configuration with high efficiency. Theoretical bounds and numerical investigations of a select set of molecular systems are investigated within this work.

(Session 5 : Thursday from 5:00 pm - 7:00 pm)