Current fluctuations of a system of one-dimensional random walks in random environment (Q606633)

The effect of a random environment on the fluctuations of the particle current in a system of many particles is investigated. A model of random walk in a random environment on the one-dimensional integer lattice is considered, where a large number of particles evolve independently of each other but in a common, fixed environment \(\omega\). On the level of the averaged (annealed) distribution particles interact with each other through the environment. The authors set the parameters of the model so that an individual particle has a positive asymptotic speed \(v_P\) and satisfies a central limit theorem around this limiting velocity under the averaged distribution. There is also a quenched central limit theorem that requires an environment-dependent correction \(Z_n(\omega)\) to the asymptotic value \(nv_P\) . Initial particle configurations whose distribution may depend on the environment, but in a manner that respects spatial shifts are considered. Under a fixed environment the initial occupation variables are required to be independent. A two-tier fluctuation picture is found. On the scale \( n^{1/2}\) the quenched mean of the current process behaves like a Brownian motion. Fluctuations are described by the same self-similar Gaussian processes that arise for independent particles performing classical random walks. But the environment-determined correction \(Z_n(\omega)\) appears again, this time as an extra shift in the spatial argument of the limit process of the current. The flux function \(H(\mu)\) gives the average rate of mass flow past a fixed point in space when the system is in a stationary state with mean density \(\mu\). If \(H\) is strictly convex or concave, then the current fluctuations have magnitude \(n^{1/3}\) and the limit distributions are related to Tracy-Widom distributions from random matrix theory. If \(H\) is linear, then the magnitude of the current fluctuations is \(n^{1/4}\) and the limit distributions are Gaussian. The random walk in a random environment model has a linear flux. The results of the paper show that in a sense it confirms the prediction stated above, but with additional features coming from the random environment. Limit processes possess covariances that are similar to those that arise for independent classical random walks. However, when the environment is averaged out, limit distributions can fail to be Gaussian.