Large deviation theory for a homogenized and ``corrected'' elliptic ODE (Q639504)

The authors consider for a given probability space \((\Omega,{\mathcal F},\operatorname{P})\) the differential equation \[ -{d\over dx} A\Biggl(x,\xi,{x\over \varepsilon},\omega\Biggr)\,{d\over dx} u_\varepsilon= f(x),\quad x\in (0,1),\quad \omega\in\Omega,\quad \varepsilon\gg 1,\tag{1} \] where \(u_\varepsilon\in L^2(\Omega; H^1_0(0,1))\), i.e., \(\int_\Omega\| u_\varepsilon\| u_\varepsilon(.,\omega)\|^2_{H^1_0} d\operatorname{P}< \infty\), \(\xi\) is a random vector. The authors present results on the homogenized convergence \(u_\varepsilon\rightarrow u_0\), where \(u_0\), and the corrector characterizing moderate deviations of \(u_\varepsilon\to u_0\), where \(u_0\) is a homogenized solution. Next, the authors study the large-deviation behavior of the solution \(u_\varepsilon\) of equation (1) for two types of the operator \(A\). They derive an approximate large deviation principle that corresponds to the approximation \(u_\varepsilon\approx u_0+ v_\varepsilon\), where \(v_\varepsilon\) can be given explicitly as an oscillatory integral. The obtained obtained results were illustrated by numerical examples.