A novel boundary integral formulation for the biharmonic wave scattering problem (Q6182313)

The authors present numerical boundary integral methods for the solution of the biharmonic equation in \(\mathbb{R}\setminus \overline{D}\), where \(D\) is a bounded domain. The boundary \(\Gamma=\partial D\) of the domain \(D\) is assumed to be analytic. In Section 2 the authors derive the boundary value problem \[ \Delta^2v-\kappa^4v \: = \: 0,\: \mbox{in }\mathbb{R}^2\setminus \overline{D} \; (1) \] with boundary condition \begin{eqnarray*} v\:=\: -u^{\mbox{inc}}, & & \partial_{\nu}v\:=\: -\partial_{\nu}u^{\mbox{inc}}\; \mbox{on }\Gamma \end{eqnarray*} and Sommerfeld's radiation condition at infinity. Here \(u^{\mbox{inc}}=e^{i\kappa x\cdot d}\), \(d\) a unit vector in \(\mathbb{R}^2\), wave number \(\kappa>0\), is an incoming wave. Then the solution \(v\) of (1) is written as \(v=v_H+v_M\) where the functions \(v_H\) and \(v_M\) are defined by \[ v_H\:=\: -\frac{1}{2\kappa^2}(\Delta v-\kappa^2 v),\: v_M\:=\: \frac{1}{2\kappa^2}(\Delta v+\kappa^2 v). \] The functions \(v_H\) and \(v_M\) solve boundary value problems on \(\mathbb{R}\setminus \overline{D}\) with a coupling condition along \(\Gamma\). The authors prove that the boundary value problem for \(v_H\) and \(v_M\) has at most one solution. In Section 3 the functions \(v_H\) and \(v_M\) are represented by potentials \[ v_H(x) \:=\: \int_\Gamma \frac{\partial G_H(x,y)}{\partial \nu(y)} g_1(y)\; ds(y), \; v_M(x) \:=\: \int_\Gamma G_M(x,y) g_2(y)\; ds(y) \] where \[ G_H(x,y)=\frac{i}{4}H^{(1)}_0(\kappa |x-y|),\; G_M(x,y)=\frac{i}{4}H^{(1)}_0(i\; \kappa |x-y|), \] and \(H^{(1)}_0\) is the Hankel function of the first kind of order 0. The functions \(g_1\) and \(g_2\) are parameterized over \([0,2\pi]\).It is proved that the resulting integral equation system on \([0,2\pi]\) \[ \mathcal{A}\psi\: =\: \eta,\: \eta\in H^p([0,2\pi])^2 \;(2) \] has a unique solution solution in \(H^p([0,2\pi])^2\) for every \(p\geq 0\). In Section 4 a fully discrete collocation method for the solution of (2) is presented with a trigonometric polynomial trial space \[X_{n}=\mbox{span}(1,\cos(t),\ldots,\cos(nt),\sin(t),\ldots,\sin((n-1)t)).\] The authors prove that the collocation method for (2) has a unique solution for \(p>3/2\), and sufficiently large \(n\). An error estimate for the collocation solution is given. In Section 5 several approaches for the solution of the boundary problem using boundary integral equations are compared. Various boundaries \(\Gamma_i\) , \(i=1,\ldots,5\) are given, some smooth, some non-smooth. All methods seem to converge very rapidly. The errors in the \(L^2\)-norm are calculated by comparison with a numerical reference solution that is calculated with a large \(n\).