Quantum matchgate computations and linear threshold gates

Abstract

The theory of matchgates is of interest in various areas in physics and computer science. Matchgates occur, for example, in the study of fermions and spin chains, in the theory of holographic algorithms and in several recent works in quantum computation. In this paper, we completely characterize the class of Boolean functions computable by unitary two-qubit matchgate circuits with some probability of success. We show that this class precisely coincides with that of the linear threshold gates . The latter is a fundamental family that appears in several fields, such as the study of neural networks. Using the above characterization, we further show that the power of matchgate circuits is surprisingly trivial in those cases where the computation is to succeed with high probability. In particular, the only functions that are matchgate-computable with success probability greater than 3/4 are functions depending on only a single bit of the input.