no. 17 · Wednesday 18 November 2026
Stochastic Homogenization of Schrödinger Equations with Exponentially Mixing Sub-Gaussian Potentials
Abstract
We study $i\partial_t u_{\varepsilon}=H_{\varepsilon} u_{\varepsilon}$, $H_{\varepsilon} = -Δ+ V\left(\frac{x}{\varepsilon},ω\right)$, where $V$ is a stationary ergodic sub-Gaussian field with exponential mixing. We establish the optimal algebraic error rate $\varepsilon^{\frac{d}2}$ in $L^2(Ω;L^2_x)$ for every $d\ge1$, prove it sharp by a matching lower bound under a non degeneracy condition, and upgrade the mean square convergence of Zhang–Bal [Stoch. Dyn. 14 (2014), 1350013] to almost sure convergence. The results are non perturbative in the amplitude of $V$ and require neither Gaussian marginals nor weak randomness. We further establish a Gaussian fluctuation central limit theorem with explicit covariance. The limiting fluctuation is the solution to the linear stochastic Schrödinger equation with additive spatial white noise \begin{equation*} i\partial_t\mathcal{G} = H_0 \mathcal{G} + \sqrt{\widehat{R}(0)}\,\bar{u}\,ξ, \qquad H_0 = -Δ+ \mathbb{E}[V], \end{equation*} where $ξ$ is a time frozen spatial white noise on $\mathbb{R}^d$ and $\mathcal{G}$ takes values in $H^{-\mathfrak{s}}(\mathbb{R}^d)$ for every $\mathfrak{s}>\frac{d}{2}$. The proofs combine Duhamel iteration with multiscale cluster expansion for the cumulants of smoothed linear functionals of the potential.
- comments
- 31 pages. This version replaces the limiting SPDE and its covariance by the correct time frozen one, and replaces the block martingale argument by a weighted spatial average
- fields
- Analysis of PDEs, Probability