CNOT-Efficient Circuits for Arbitrary Rank Many-Body Fermionic and Qubit Excitations

TL;DR

This paper develops CNOT-efficient quantum circuits for arbitrary excitation ranks in fermionic and qubit systems, significantly reducing gate counts for quantum simulations of many-body problems.

Contribution

It extends previous CNOT-efficient circuit constructions to arbitrary excitation ranks and demonstrates their effectiveness in reducing gate counts in quantum algorithms.

Findings

SPQE with these circuits reduces CNOT gates by up to 15 times compared to traditional methods.

QEB-SPQE generally requires more parameters but results in larger circuits than FEB-SPQE.

SPQE needs fewer residual evaluations than ADAPT-VQE, despite larger circuits.

Abstract

Efficient quantum circuits are necessary for realizing quantum algorithms on noisy intermediate-scale quantum devices. Fermionic excitations entering unitary coupled-cluster (UCC) ans\"atze give rise to quantum circuits containing CNOT "staircases" whose number scales exponentially with the excitation rank. Recently, Yordanov et al. [Phys. Rev. A 102, 062612 (2020); Commun. Phys. 4, 228 (2021)] constructed CNOT-efficient quantum circuits for both fermionic- (FEB) and qubit-excitation-based (QEB) singles and doubles and illustrated their usefulness in adaptive derivative-assembled pseudo-Trotterized variational quantum eigensolver (ADAPT-VQE) simulations. In this work, we extend these CNOT-efficient quantum circuits to arbitrary excitation ranks. To illustrate the benefits of these compact FEB and QEB quantum circuits, we perform numerical simulations using the recently developed selected projective quantum eigensolver (SPQE) approach, which relies on an adaptive UCC ansatz built from arbitrary-order particle-hole excitation operators. We show that both FEB- and QEB-SPQE decrease the number of CNOT gates compared to traditional SPQE by factors as large as 15. At the same time, QEB-SPQE requires, in general, more ansatz parameters than FEB-SPQE, in particular those corresponding to higher-than-double excitations, resulting in quantum circuits with larger CNOT counts. Although ADAPT-VQE generates quantum circuits with fewer CNOTs than SPQE, SPQE requires orders of magnitude less residual element evaluations than gradient element evaluations in ADAPT-VQE.