Page 73 - 6637
P. 73

Solved Problems


               Since u−v = 0, we conclude that the solution of the Dirichlet problem is unique.


               Exercise 200. Not all combinations of boundary conditions/initial conditions lead
               to so called well-posed problems. Essentially, a well posed problem is one where the
               solutions depend continuously on the boundary data. Otherwise it is considered “ill

               posed”.
                   Consider Laplace’s equation on the unit-square

                                                       u xx  + u yy  = 0,


               with u(0, y) = u(1, y) = 0 and u(x, 0) = 0, u (x, 0) =  sin(nπx).
                                                                     y

                 (a) Show that even as  → 0, you can find n so that the solution can attain any
                     finite value for any y > 0. Use this to then show that this problem is ill posed.

                (b) Contrast this with the case where u(0, y) = u(1, y) = 0 and u(x, 0) = 0,

                     u(x, 1) =  sin(nπx). Is this well posed?
               Solution:


                 (a) We seek a solution of the form u(x, y) = sin(nπx)Y (y). This form satisfies
                     the boundary conditions at x = 0, 1.

                                                          u xx  + u yy  = 0

                                                                00
                                                       2
                                               −(nπ) Y + Y = 0,            Y (0) = 0
                                                         Y = c sinh(nπy)

                     Now we apply the inhomogeneous boundary condition.


                                            u (x, 0) =  sin(nπx) = cnπ sin(nπx)
                                              y

                                               u(x, y) =       sin(nπx) sinh(nπy)
                                                           nπ

                     For  = 0 the solution is u = 0. Now consider any  > 0. For any y > 0 and
                     any finite value M, we can choose a value of n such that the solution along

                     y = 0 takes on all values in the range [−M . . . M]. We merely choose a value
                     of n such that
                                                        sinh(nπy)        M
                                                                      ≥     .
                                                             nπ
                     Since the solution does not depend continuously on boundary data, this prob-
                     lem is ill posed.


                (b) We seek a solution of the form u(x, y) = c sin(nπx) sinh(nπy). This form
                     satisfies the differential equation and the boundary conditions at x = 0, 1 and
                     at y = 0. We apply the inhomogeneous boundary condition at y = 1.


                                         u(x, 1) =  sin(nπx) = c sin(nπx) sinh(nπ)


                                                              69
   68   69   70   71   72   73   74   75   76   77   78