Kinetic limit for wave propagation in a random medium (Q1006798): Difference between revisions

The authors study the kinetic limit of the discrete wave equation \[ \begin{aligned} \frac{d}{dt} q_y(t)&=v_y(t),\\ (1+\sqrt{\varepsilon} \xi_y)^{-2} \frac{d}{dt} v_y(t)&= - \sum_{y'\in\mathbb R^3} \alpha(y-y') q_{y'}(t), \end{aligned} \] where \(y\in\mathbb R^3\) and \(q_y(t), v_y(t)\in\mathbb R\). This corresponds to study dielectric crystals in the harmonic approximation. The term \((1+\sqrt{\varepsilon} \xi_y)^{-2}\) is the mass of the atom at \(y\), where \((\xi_y\), \(y\in\mathbb R^3)\) is a certain i.i.d family of random variables. \(\alpha(y)\) stands for the coupling between atoms and \(\sqrt{\varepsilon}\) represents the strength's disorder. Then, sufficient conditions on the initial conditions and the dispersion relation are provided so that the disorder-averaged Wigner function on the kinetic scale (space and time of order \(\varepsilon^{-1}\)) is governed by a linear Boltzmann equation. Some of the techniques used to prove the main result are closely related to the results in [\textit{L. Erdős} and \textit{H.-T. Yau}, Commun. Pure Appl. Math. 53, No. 6, 667--735 (2000; Zbl 1028.82010)], where a similar problem has been considered for a random Schrödinger equation. Concerning the structure of the proof of the main result, let us quote form the authors: ``\dots we exploit general properties about the weak limits of lattice Wigner transforms to reduce the proof of the main theorem into a Proposition stating that the Fourier transforms of the Wigner transforms converge to the characteristic functions of \((\mu_t)\).'' Here, \(\mu_t\) denotes the joint distribution of a Markov jump process corresponding to the Boltzmann equation, and ``The core of the paper is the graphical expansion of Section 4 where the proof of the above Proposition is carried out by dividing it into several layers with even more detailed lemmas acting as links between the different layers. In particular, we have separated the analysis of the non-vanishing parts of the graph expansion\dots''