Hadamard variational formula for the Green's function of the boundary value problem on the Stokes equations (Q2434999)

This interesting paper is devoted to the Hadamard variational formula for Green's functions of elliptic boundary value problems. The authors consider the stationary Stokes equations governing the motion of incompressible fluids. These equations have the form: \[ (1_0) \quad \Delta {\mathbf{v}}-\nabla q = {\mathbf{f}}, \;\;\operatorname{div}{\mathbf{v}}=0 \;\;\text{in} \;\;\Omega , \;\;\text{and} \;\;{\mathbf{v}}=0 \;\;\text{on} \;\;\partial\Omega , \] where \({\mathbf{v}}={\mathbf{v}}(x)\in\mathbb{R}^{d}\) is the unknown velocity, \(q=q(x)\) is the pressure at \(x\in\Omega\subset\mathbb{R}^{d}\) (\(d\geq 2\)), \(\Omega \) is a bounded domain with smooth boundary \(\partial\Omega \), and \({\mathbf{f}}={\mathbf{f}}(x)\in\mathbb{R}^{d}\) is an external force. Furthermore, define a perturbed system with perturbed domain. For this purpose introduce for a small \(\varepsilon \geq 0\) the family of domains \(\{\Omega_{\varepsilon }\}_{\varepsilon\geq 0}\). Then define Green's matrix \(G_{\varepsilon }(x,y)= \{G_{\varepsilon ,n}^{i}(x,y)\}_{1\leq i,n\leq d}\), and the quantity \(R(x,y;\varepsilon )=\{R_{\varepsilon ,n}^{i}(x,y)\}_{1\leq n\leq d}\), and the perturbed system has the form: \[ (1_{\varepsilon }) \;\;\;\;\Delta_{x} {\mathbf G_{\varepsilon ,n}}^{i}(x,y)- \nabla_x R_{\varepsilon ,n}(x,y) = e_n\delta (x-y), \;\;\operatorname{div}{\mathbf{G}}_{\varepsilon ,n}^{i}(x,y)=0, \;\;(x,y)\in\Omega_{\varepsilon }\times\Omega , \] with \({\mathbf{G}}_{\varepsilon ,n}(x,y)=0\), \(x\in\partial\Omega_{\varepsilon }\), \(y\in\Omega_{\varepsilon }\) (\( n=1,\ldots ,d\)). Here, \(e_n\in\mathbb{R}^{d}\) is a canonical basis in \(\mathbb{R}^{d}\), \(\Omega_0=\Omega \), \(G_0=G\), \(R_0=R\). The main result is connected with a representation formula for the variation \[ \delta {\mathbf{G}}(x,y)=\lim\limits_{\varepsilon \to 0} \varepsilon^{-1}[{\mathbf G}_{\varepsilon }(x,y)-{\mathbf G}(x,y)]. \] Define a diffeomorphism \(\Phi_{\varepsilon }: \overline{\Omega }\to\overline{\Omega }_{\varepsilon }\) satisfying the conditions: \(\Phi_{\varepsilon }\in C^{\infty }(\overline{\Omega })^d\), \(\Phi_0(x)=x\in \overline{\Omega }\); there is a vector \(S\in C^{\infty }(\overline{\Omega })^d\) such that the function \(K(x;\varepsilon )=\Phi_{\varepsilon }(x)-x-S(x)\varepsilon \) satisfies \(\displaystyle\sup\limits_{x\in\overline{\Omega }}|K(x;\varepsilon )|+ \sup\limits_{x\in\overline{\Omega }}|\nabla K(x;\varepsilon )|=O(\varepsilon^2)\) as \(\varepsilon\to 0\), and det \(\biggl(\partial\Phi_{\varepsilon }^{i}(x)/\partial x^{j}\biggr)_{i,j=1,\ldots ,d}=1\), \(x\in \overline{\Omega }\), \(\varepsilon\geq 0\). Let \(\{{\mathbf{G}}_{\varepsilon ,m}^{n}(x,y)\}_{m,n=1,2,\ldots ,d}\) be Green's matrix to the problem (\(1_{\varepsilon }\)), and \[ \delta {\mathbf{G}}_m^n(y,z)= \lim\limits_{\varepsilon\to 0} \frac{{\mathbf{G}}_{\varepsilon ,m}^{n}(y,z)- {\mathbf{G}}_{m}^{n}(y,z)}{\varepsilon } \] (\(m,n=1,2,\ldots ,d\)) for any \(y\) and \(z\) in \(\Omega \). Then there exists a variational formula \[ \delta {\mathbf{G}}_m^n(y,z)=\int_{\partial\Omega } \sum\limits_{i=1}^{d}\frac{\partial {\mathbf G}_n^i}{\partial\nu_x}(x,y) \frac{\partial {\mathbf G}_m^i}{\partial\nu_x}(x,z)S(x)\cdot\nu_x d\sigma_x \] for \(m,n=1,\ldots ,d\); \(\nu_x\in\mathbb{R}^d\) is the unit outer normal to \(\partial\Omega \) at \(x\in\partial\Omega \) and \(\sigma_x\) denotes the surface element of \(\partial\Omega \).