Limiting behavior of solutions of subelliptic heat equations (Q2474183)

The author studies the behavior as \(\varepsilon\to 0+\) of \(\varepsilon\log\,w^\varepsilon(t,x)\), where the \(w^\varepsilon\) are solutions to the family of subelliptic heat equations \[ \partial_tw^\varepsilon-\varepsilon\sum_{i,j=1}^na_{ik}(x)\partial^2_{x_ix_j} w^\varepsilon=0,\quad w^\varepsilon(0,x)=\exp(-g(x)/2\varepsilon),\,\,x\in{\mathbb R}^n,\,\,t>0, \] where the matrix \(A(x)=(a_{ij}(x))_{i,j=1,\dots,n}\) is of the form \(A(x)=\,^tS(x)S(x),\) where \(S(x)\) is an \(m\times n\) matrix (\(m\leq n\)) whose rows, thought of as vector-fields in \({\mathbb R}^n\), satisfy the Hörmander condition. If \(d\) denotes the Carnot-Carathéodory distance associated with \(S(x)\), and \[ g_t(x):=\inf_{y\in{\mathbb R}^n}\Bigl[g(y)+{d(x,y)^2\over{2t}}\Bigr] \] denotes the \textit{inf-convolution} of \(g\) with metric \(d\) as kernel, the author proves, by using the large deviation principle, that for \(g\) continuous and bounded in \({\mathbb R}^n\) \[ \lim_{\varepsilon\to 0+} 2\varepsilon\log w^\varepsilon(t,x)=-g_t(x) \] locally uniformly on \([0,+\infty)\times{\mathbb R}^n.\) In the special case of \(A(x)=I\) her result recovers the classical result.