Page 105 -
P. 105

88  2. Introductory applications



          (2.85) when expressed in terms of x (as will be necessary for any matching); this gives





          But we are seeking solutions that remain bounded as   and this is possible only
          if the  exponent in  (2.87) is non-positive for   Thus we require





          and this means that          for the same xs i.e.            (for otherwise x < C can occur
          in the domain, and the integral will change sign). Then, for x sufficiently close to
          we have




          and we have a choice for the boundary-layer variable. Hence, for a (x)  > 0, the bound-
          ary layer must sit near the left-hand edge of the domain. Conversely, the same argument
          in the case a (x) < 0 requires that the boundary layer be situated near   (the right-
          hand edge of the domain). When we apply this rule to equation (2.63):
                            we see  that             and  so the boundary  layer is
          in the  neighbourhood of x = 0  (and  we  introduced  for  this example).
          Similarly, equation (2.77):                                       has a(x) = – 1 < 0,
          and so the boundary layer is now near x = 1 (and we used
            It is rarely necessary to incorporate the formal definition of g(x) to generate the
          appropriate variable that  is to be used to  represent the  boundary layer  (although it
          will always produce the simplest form of the solution). For example, the equation of
          this class:                              has  a  boundary layer near x = 2
          (because a (x) =  –1/(2 + x) < 0 for   Now an appropriate scaled variable is
          simply          giving





          and this choice will suffice, even though higher-order terms will require the expansion
          of          (but  this is  usually a small price to  pay—and we  already know that
          this  asymptotic expansion  will  breakdown for   so  retaining   in  the
          coefficient has no unforeseen complications). It should also be noted that, exceptionally,
          a boundary-layer-type  problem may not require  a  boundary  layer at  all, in  order to
          accommodate the given boundary value  (to leading order or, possibly, to  all orders).
          This is evident for the equation given in (2.63) (and see also the exact solution, (1.22));
          in this example, if the boundary values satisfy the special  condition   then no
          boundary  layer whatsoever is  required. Note,  however,  that if   then
   100   101   102   103   104   105   106   107   108   109   110