Essentiality of the Non-stoquastic Hamiltonians and Driver Graph Design in Quantum Optimization Annealing

TL;DR

This paper demonstrates how designing specific driver graphs in non-stoquastic quantum annealing can exponentially speed up solving hard optimization problems with anti-crossings, surpassing classical algorithms.

Contribution

It introduces a method to design non-stoquastic Hamiltonians with tailored driver graphs to achieve exponential speedup in quantum annealing for problems with anti-crossings.

Findings

Exponential speedup over stoquastic QA for certain hard instances.

Proper non-stoquastic Hamiltonians can be engineered without prior problem knowledge.

Large spectral gaps enable efficient diabatic quantum annealing.

Abstract

One of the distinct features of quantum mechanics is that the probability amplitude can have both positive and negative signs, which has no classical counterpart as the classical probability must be positive. Consequently, one possible way to achieve quantum speedup is to explicitly harness this feature. Unlike a stoquastic Hamiltonian whose ground state has only positive amplitudes (with respect to the computational basis), a non-stoquastic Hamiltonian can be eventually stoquastic or properly non-stoquastic when its ground state has both positive and negative amplitudes. In this paper, we describe that, for some hard instances which are characterized by the presence of an anti-crossing (AC) in a transverse-field quantum annealing (QA) algorithm, how to design an appropriate XX-driver graph (without knowing the prior problem structure) with an appropriate XX-coupler strength such that the resulting non-stoquastic QA algorithm is proper-non-stoquastic with two bridged anti-crossings (a double-AC) where the spectral gap between the first and second level is large enough such that the system can be operated diabatically in polynomial time. The speedup is exponential in the original AC-distance, which can be sub-exponential or exponential in the system size, over the stoquastic QA algorithm, and possibly the same order of speedup over the state-of-the-art classical algorithms in optimization. This work is developed based on the novel characterizations of a modified and generalized parametrization definition of an anti-crossing in the context of quantum optimization annealing introduced in [4].