Unexpected averages of mixing matrices

The (standard) average mixing matrix of a continuous-time quantum walk is computed by taking the expected value of the mixing matrices of the walk under the uniform sampling distribution on the real line. In this paper, we consider alternative probability distributions, either discrete or continuous, and first we show that several algebraic properties that hold for the average mixing matrix still stand for this more general setting. Then, we provide examples of graphs and choices of distributions where the average mixing matrix behaves in an unexpected way: for instance, we show that there are probability distributions for which the average mixing matrices of the paths on three or four vertices have constant entries, opening a significant line of investigation about how to use classical probability distributions to sample quantum walks and obtain desired quantum effects. We present results connecting the trace of the average mixing matrix and quantum walk properties, and we show that the Gram matrix of average states is the average mixing matrix of a certain related distribution. Throughout the text, we employ concepts of classical probability theory not usually seen in texts about quantum walks.