Boundary layers in smooth curvilinear domains: parabolic problems (Q2379736)

Let \(\Omega\) be a bounded domain in \(\mathbb R^d\), \(d\geq 2\). The boundary \(\partial \Omega\) satisfies the following property: \(\partial \Omega=\bigcup_{i=0}^n\Gamma_i\), where each \(\Gamma_i\) is a connected component of \(\partial \Omega\) which is a smooth Jordan surface in \(\mathbb R^{d-1}\), with \(\Omega\) lying locally on one side of \(\Gamma_i\). The paper deals with the asymptotic behavior of the solution \(u_\varepsilon\) of the heat equation \[ \partial_tu(x,t)- \varepsilon\Delta u(x,t)=f(x,t),\quad (x,t)\in\Omega \times (0,T),\tag{1} \] together with the boundary and initial conditions \[ u(x,t)=0 \quad \text{on }\partial\Omega ,\qquad u|_{t=0}=u_o(x),\tag{2} \] when the thermal diffusivity \(\varepsilon\) tends to zero. Here, \(u_o\) and \(f\) vanish on \(\partial\Omega\). The authors study asymptotic expansions of \(u_\varepsilon\) of the form \[ u_\varepsilon\simeq \sum_{j=0}^{\infty} (\varepsilon^ju_j+ \varepsilon^j\theta_j+ \varepsilon^{j+\frac{1}{2}} \theta_{j+\frac{1}{2}}), \] where the \(u_j\) [resp. the correctors \(\theta_j,\theta_{j+\frac{1}{2}})]\) correspond to the external [resp. inner] expansion. For \(N\geq 0\), set \[ \begin{aligned} w_{\varepsilon,N} &= u_\varepsilon- \sum_{j=0}^N (\varepsilon^ju_j+ \varepsilon^j\theta_j)- \sum_{j=0}^{N-1} \varepsilon^{j+\frac{1}{2}} \theta_{j+\frac{1}{2}},\\ w_{\varepsilon,N+\frac{1}{2}} &= u_\varepsilon- \sum_{j=0}^N (\varepsilon^ju_j+\varepsilon^j\theta_j+ \varepsilon^{j+\frac{1}{2}} \theta_{j+\frac{1}{2}}). \end{aligned} \] (If N=0, the last sum, in \(w_{\varepsilon,N}\), disappears.) If \(\partial\Omega\) is of class \(C^{2N+4}\), \(u_o\in H^{2N+2}(\Omega)\), and \(f\in L^{\infty}(0,T:H^{2N+2}(\Omega))\), they obtain sharp estimates for \(w_{\varepsilon,N}\) and \(w_{\varepsilon,N+\frac{1}{2}}\), in \(L^{\infty}(0,T;L^2(\Omega))\) and \(L^2(0,T;H^1(\Omega))\). The starting point of the proof is the introduction of a curvilinear coordinate system adapted to the boundary. They point out that this method can be generalized when the Laplacian is replaced by a more general elliptic operator.