Optimization of the Qubit Coupled Cluster Ansatz on classical computers

TL;DR

This paper introduces two novel schemes to improve amplitude optimization in the iterative qubit coupled cluster method on classical computers, enabling more efficient and accurate simulations of molecular systems.

Contribution

It presents two innovative approximation schemes for the QCC ansatz that enhance optimization efficiency and scalability on classical hardware.

Findings

Both schemes allow inclusion of more generators in QCC.

They reduce iteration counts and improve accuracy.

The methods are demonstrated on molecular systems with up to 80 qubits.

Abstract

Immense interest in quantum computing has prompted development of electronic structure methods that are suitable for quantum hardware. However, the slow pace at which quantum hardware progresses, forces researchers to implement their ideas on classical computers despite the obvious loss of any "quantum advantage." As a result, the so-called quantum inspired methods emerge. They allow one to look at the electronic structure problem from a different angle; yet, to fully exploit their capacity, efficient implementations are highly desirable. Here we report two schemes for improving the amplitude optimisation in the iterative qubit coupled cluster (iQCC) method -- a variational quantum eigensolver-type approach which is based on the qubit coupled cluster (QCC) Ansatz. Our first scheme approximates the QCC unitary as a sum of symmetrical polynomials of generators up to a given order. The resulting energy expression allows for a flexible control of computational complexity via the order parameter. It also guaranties smoothness of trial energies and their derivatives, which is important for gradient-based optimization strategies. The second scheme limits the size of the expansion space in which the QCC unitary is generated. It provides better control of memory requirements, but in general may lead to the non-smooth variation of energy estimates upon changes in amplitudes. It can be used to extrapolate energies for a given set of amplitudes towards the exact QCC value. Both schemes allow for a larger number of generators to be included into the QCC form compared to the exact formulation. This reduces the number of iterations in the iQCC method and/or leads to higher accuracy. We assess capabilities of the new schemes to perform QCC amplitudes optimization for a few molecular systems: N$_2$ (16 qubits), H$_2$O (36 qubits), and tris(2-(2,4-difluorophenyl)pyridine) iridium(III), (80 qubits).