Page 72 - 6099
P. 72

Its restriction,
                                                v(x, t) = u(x, t) for x > 0,                       (7.14)

                   will be the unique solution of our new problem (7.10). There is no difference at all between
                   v and u except that the negative values of x are not considered when discussing v.
                       Why is v(x, t) the solution of (7.10)? Notice first that u(x, t) must also be an odd
                   function of x. That is, u(−x, t) = −u(x, t). Putting x = 0, it is dear that u(0, t) = 0.
                   So the boundary condition v(0, t) = 0 is automatically satisfied! Furthermore, v solves
                   the PDE as well as the initial condition for x > 0, simply because it is equal to u for
                   x > 0 and u satisfies the same PDE for all x and the same initial condition for x > 0.
                   The explicit formula for v(x, t) is easily deduced from (7.13) and (7.14). From (7.13) and
                   (7.11) we have

                                            ∞                        0
                                          Z                        Z
                                u(x, t) =     S(x − y, t)φ(y)dy −       S(x − y, t)φ(−y)dy.
                                           0                        −∞
                   Changing the variable −y to +y in the second integral, we get

                                                 Z  ∞
                                       u(x, t) =     S(x − y, t) − S(x + y, t)φ(y)dy.
                                                  0
                   (Notice the change in the limits of integration.) Hence for 0 < x < ∞, 0 < t < ∞, we
                   have                            Z
                                               1      ∞        2             2
                                  v(x, t) = √           (e −(x−y) /4kt  − e −(x+y) /4kt )φ(y)dy.   (7.15)
                                              4πkt   0

                   This is the complete solution formula for (7.10). We have just carried out the method
                   of odd extensions or reflection method, so called because the graph of φ odd (x) is the
                   reflection of the graph of φ(x) across the origin.

                   Example 7.3 Solve (7.10) with φ(x) ≡ 1. The solution is given by formula (7.15).
                                                                            √
                   This case can be simplified as follows. Let p = (x − y)/ 4kt in the first integral and
                                √
                   q = (x + y)/ 4kt in the second integral. Then
                                                 √
                                             Z  x/ 4kt       √      Z  +∞          √
                                    u(x, t) =         e −p 2 dp/ π −       e −q 2 dq/ π =
                                                                       √
                                               −∞                    x/ 4kt

                                     1    1                 1    1
                                 =     + Erf( √   x  ) −      − Erf( √  x  )  = Erf( √ x  ).
                                     2    2       4kt       2    2      4kt            4kt

                   Now let’s play the same game with the Neumann problem

                                         w t − kw xx = 0 for 0 < x < ∞, 0 < t < ∞,
                                                      w(x, 0) = φ(x),                              (7.16)
                                                        w x (0, t) = 0.

                   In this case the reflection method is to use even, rather than odd, extensions. An even
                   function is a function φ such that φ(−x) = +φ(x). If φ is an even function, then
                   differentiation shows that its derivative is an odd function. So automatically its slope at



                                                             65
   67   68   69   70   71   72   73   74   75   76   77