Asymptotic evolution of quantum walks with random coin

TL;DR

This paper analyzes the long-term behavior of quantum walks on a lattice, including random coin choices, revealing ballistic and diffusive regimes, with explicit formulas for asymptotic distributions and diffusion matrices.

Contribution

It introduces a method to compute the asymptotic position distribution of quantum walks with random and non-random coins using Fourier analysis and perturbation theory.

Findings

Ballistic scaling with a direct computation of the asymptotic distribution.

Gaussian limiting distribution in diffusive scaling with a momentum-dependent covariance.

Diffusion matrix diverges when coins are nearly identical or control process rates are small.

Abstract

We study the asymptotic position distribution of general quantum walks on a lattice, including walks with a random coin, which is chosen from step to step by a general Markov chain. In the unitary (i.e., non-random) case, we allow any unitary operator, which commutes with translations, and couples only sites at a finite distance from each other. For example, a single step of the walk could be composed of any finite succession of different shift and coin operations in the usual sense, with any lattice dimension and coin dimension. We find ballistic scaling, and establish a direct method for computing the asymptotic distribution of position divided by time, namely as the distribution of the discrete time analog of the group velocity. In the random case, we let a Markov chain (control process) pick in each step one of finitely many unitary walks, in the sense described above. In ballistic order we find a non-random drift, which depends only on the mean of the control process and not on the initial state. In diffusive scaling the limiting distribution is asymptotically Gaussian, with a covariance matrix (diffusion matrix) depending on momentum. The diffusion matrix depends not only on the mean but also on the transition rates of the control process. In the non-random limit, i.e., when the coins chosen are all very close, or the transition rates of the control process are small, leading to long intervals of ballistic evolution, the diffusion matrix diverges. Our method is based on spatial Fourier transforms, and the first and second order perturbation theory of the eigenvalue 1 of the transition operator for each value of the momentum.