Page 107 -
P. 107

2.5 Remarks    91

                           2.5 Remarks

                           Very often in applications, on different parts of the boundary, different bound-
                           ary conditions are required or, as in classical crack mechanics (Cruse [58]),
                           the boundaries are given as transmission conditions on some bounded mani-
                           fold, the crack surface, in the interior of the domain. A similar situation can
                           be found for screen problems (see also Costabel and Dauge [52] and Stephan
                           [293]).
                              As an example of mixed boundary conditions let us consider the Lam´e
                           system with given Dirichlet data on Γ D ⊂ Γ and given Neumann data on
                           Γ N ⊂ Γ where Γ = Γ D ∪ Γ N ∪ γ with the set of collision points γ of the two
                           boundary conditions (which might also be empty if Γ D and Γ N are separated
                           components of Γ) (see e.g. Fichera [76], [145], Kohr et al [164], Maz‘ya [202],
                           Stephan [295]) where meas (Γ D ) > 0.
                              For n = 2, where Γ is a closed curve, we assume that either γ = ∅ or
                           consists of finitely many points; for n = 3, the set γ is either empty or a
                           closed curve and as smooth as Γ. For the Lam´e system (2.2.1) the classical
                           mixed boundary value problem reads:
                                        2
                                                α
                              Find u ∈ C (Ω) ∩ C (Ω) , 0 <α< 1, satisfying
                                        −∆ u = f     in Ω   with
                                           ∗
                                                                                        (2.5.1)
                                          γ 0 u = ϕ D on Γ D and Tu = ψ N  on Γ N .
                           For reformulating this problem with boundary integral equations we first ex-
                           tend ϕ D  from Γ D and ψ N  from Γ N onto the complete boundary Γ such that

                                               ϕ D  = ϕ| Γ D  and ψ N  = ψ| Γ N         (2.5.2)
                                     α
                           with ϕ ∈ C (Γ) , 0 <α< 1 and appropriate ψ. Then
                                                γ 0 u = ϕ +   ϕ ,Tu = ψ + ψ             (2.5.3)

                           where now
                                               α              α
                                            ϕ ∈ C (Γ N )= {ϕ ∈ C (Γ) | supp ϕ ⊂ Γ N }   (2.5.4)
                                               0
                           and ψ with supp ψ ⊂ Γ D are the yet unknown Cauchy data to be determined.


                           With (2.5.3), the representation formula (2.2.4) reads


                                   v(x)=    E(x, y)ψ(y)ds y −  T y E(x, y)  ϕ(y)ds y
                                         Γ                 Γ


                                        +   E(x, y)ψ(y)ds y −  T y E(x, y)    ϕ(y)ds y  (2.5.5)

                                          Γ                 Γ

                                        +   E(x, y)f(y)dy for x ∈ Ω.
                                          Ω
   102   103   104   105   106   107   108   109   110   111   112