Page 488 - Electromagnetics
P. 488

Eigenfunction expansion of an arbitrary function.   If L is self-adjoint, then its
                        eigenfunctions form a complete set. This means that any piecewise smooth function may
                        be represented as a weighted series of eigenfunctions. Specifically, if f and f are piece-

                        wise continuous on [a, b], then f may be represented as the generalized Fourier series
                                                             ∞

                                                      f (x) =  c n ψ n (x).                    (A.84)
                                                            n=0
                        Convergence of the series is uniform and gives, at any point of (a, b), the average value
                        [ f (x )+ f (x )]/2 of the one-sided limits f (x ) and f (x ) of f (x). The c n can be found
                                                                         −
                                   −
                            +
                                                               +
                        using orthogonality condition (A.83): multiply (A.84) by ψ m σ and integrate to obtain
                                          b                  ∞      b

                                          f (x)ψ m (x)σ(x) dx =  c n  ψ n (x)ψ m (x)σ(x) dx,
                                        a                   n=0   a
                        hence
                                                        b

                                                        a  f (x)ψ n (x)σ(x) dx
                                                  c n =                  .                     (A.85)
                                                            b  2
                                                             n
                                                          a  ψ (x)σ(x) dx
                        These coefficients ensure that the series converges in mean to f ; i.e., the mean-square
                        error
                                                                      2
                                                           ∞

                                                  b


                                                     f (x) −  c n ψ n (x)  σ(x) dx
                                                 a
                                                          n=0
                        is minimized. Truncation to finitely-many terms generally results in oscillations (Gibb’s
                        phenomena) near points of discontinuity of f . The c n are easier to compute if the ψ n
                        are orthonormal with
                                                        b
                                                          2
                                                        ψ (x)σ(x) dx = 1
                                                         n
                                                      a
                        for each n.
                        Uniqueness of the eigenfunctions.  If both ψ 1 and ψ 2 are associated with the same
                        eigenvalue λ, then
                                     L[ψ 1 (x)] + λσ(x)ψ 1 (x) = 0,  L[ψ 2 (x)] + λσ(x)ψ 2 (x) = 0,
                        hence

                                              ψ 1 (x) L[ψ 2 (x)] − ψ 2 (x) L[ψ 1 (x)] = 0.
                        By (A.79) we have
                                           d            dψ 2 (x)      dψ 1 (x)
                                              p(x) ψ 1 (x)     − ψ 2 (x)       = 0
                                          dx              dx            dx
                        or
                                                       dψ 2 (x)     dψ 1 (x)

                                            p(x) ψ 1 (x)     − ψ 2 (x)      = C
                                                         dx           dx
                        where C is constant. Either of (A.81) implies C = 0, hence
                                                        d    ψ 2 (x)
                                                                  = 0
                                                       dx  ψ 1 (x)



                        © 2001 by CRC Press LLC
   483   484   485   486   487   488   489   490   491   492   493