Southwest Quantum Information and Technology

The promise of variational quantum algorithms

Presenting Author: Ryan Babbush, Martinis group (Google)
Contributing Author(s): Pedram Roushan, Peter O'Malley, Hartmut Neven, John Martinis

Recent work has shown that parameterized short quantum circuits can generate powerful variational ansätze for ground states of classically intractable fermionic models. This observation inspires hope that even without error correction, quantum computers may provide insight into problems of industrial importance such as quantum chemistry and superconductivity. As the number of qubits in superconducting devices keeps increasing, their dynamics are becoming prohibitively expensive to simulate classically. In anticipation of such a platform, we use devices with up to nine superconducting qubits to explore the viability of variational approaches. We discuss experiments showing surprising robustness to control errors, including the first quantum simulation of molecular energies to obtain chemical precision without precompilation. Nevertheless, with current gate fidelities it seems unlikely that variational approaches parameterized in terms of conventional logic gates will scale to the classically intractable regime. Instead, we propose to form variational ansätze at the level of hardware by parameterizing quantum circuits in terms of the experimentalist’s control knobs. We discuss an ongoing experiment to simulate Fermi-Hubbard models in this way. We conclude by asking how experiments can guide the design and analysis of quantum variational algorithms.

(Session 11 : Saturday from 11:30 am - 12:00 pm)