Page 492 - Electromagnetics
P. 492

m
                          For m = 0, P (x) is a polynomial of degree n. Each such Legendre polynomial, denoted
                                     n
                        P n (x), is given by
                                                                n
                                                                   2
                                                            1 d (x − 1) n
                                                   P n (x) =             .
                                                            n
                                                           2 n!   dx n
                        It turns out that
                                                                       m
                                                                      d P n (x)
                                                m
                                                           m
                                                                 2 m/2
                                               P (x) = (−1) (1 − x )         ,
                                                n
                                                                       dx m
                               m
                        giving P (x) = 0 for m > n.
                               n
                          Because the Legendre polynomials form a complete set in the interval [−1, 1],we may
                        expand any sufficiently smooth function in a Fourier–Legendre series
                                                             ∞

                                                      f (x) =  c n P n (x).
                                                            n=0
                        Convergence in mean is guaranteed if
                                                      2n + 1     1
                                                 c n =          f (x)P n (x) dx,
                                                        2    −1
                        found using (A.85) along with (A.93).
                          In practice, the associated Legendre functions appear along with exponential functions
                        in the solutions to spherical boundary value problems. The combined functions are known
                        as spherical harmonics, and form solutions to two-dimensional Sturm–Liouville problems.
                        We consider these next.


                        Higher-dimensional SL problems: Helmholtz’s equation.     Replacing d/dx by ∇,
                        we generalize the Sturm–Liouville equation to higher dimensions:

                                           ∇· [p(r)∇ψ(r)] + q(r)ψ(r) + λσ(r)ψ(r) = 0,

                        where q, p, σ, ψ are real functions. Of particular interest is the case q(r) = 0, p(r) =
                        σ(r) = 1, giving the Helmholtz equation

                                                       2
                                                     ∇ ψ(r) + λψ(r) = 0.                       (A.94)
                        In most boundary value problems, ψ or its normal derivative is specified on the surface
                        of a bounded region. We obtain a three-dimensional analogue to the regular Sturm–
                        Liouville problem by assuming the homogeneous boundary conditions
                                                    αψ(r) + β ˆ n ·∇ψ(r) = 0                   (A.95)

                        on the closed surface, where ˆ n is the outward unit normal.
                          The problem consisting of (A.94) and (A.95) has properties analogous to those of the
                        regular one-dimensional Sturm–Liouville problem. All eigenvalues are real. There are
                        infinitely many eigenvalues. There is a smallest eigenvalue but no largest eigenvalue.
                        However, associated with an eigenvalue there may be many eigenfunctions ψ λ (r). The
                        eigenfunctions are orthogonal with


                                                         (r) dV = 0,   λ 1  = λ 2 .
                                                 ψ λ 1  (r)ψ λ 2
                                               V


                        © 2001 by CRC Press LLC
   487   488   489   490   491   492   493   494   495   496   497