Comments
Description
Transcript
Confluent hypergeometric functions
18.11 CONFLUENT HYPERGEOMETRIC FUNCTIONS 18.11 Confluent hypergeometric functions The confluent hypergeometric equation has the form xy + (c − x)y − ay = 0; (18.147) it has a regular singularity at x = 0 and an essential singularity at x = ∞. This equation can be obtained by merging two of the singularities of the ordinary hypergeometric equation (18.136). The parameters a and c are given real numbers. Show that setting x = z/b in the hypergeometric equation, and letting b → ∞, yields the confluent hypergeometric equation. Substituting x = z/b into (18.136), with d/dx = bd/dz, and letting u(z) = y(x), we obtain z d2 u du bz 1 − + [bc − (a + b + 1)z] − abu = 0, b dz 2 dz which clearly has regular singular points at z = 0, b and ∞. If we now merge the last two singularities by letting b → ∞, we obtain zu + (c − z)u − au = 0, where the primes denote d/dz. Hence u(z) must satisfy the confluent hypergeometric equation. In our discussion of Bessel, Laguerre and associated Laguerre functions, it was noted that the corresponding second-order differential equation in each case had a single regular singular point at x = 0 and an essential singularity at x = ∞. From table 16.1, we see that this is also true for the confluent hypergeometric equation. Indeed, this equation can be considered as the ‘canonical form’ for secondorder differential equations with this pattern of singularities. Consequently, as we mention below, the Bessel, Laguerre and associated Laguerre functions can all be written in terms of the confluent hypergeometric functions, which are the solutions of (18.147). The solutions of the confluent hypergeometric equation are obtained from those of the ordinary hypergeometric equation by again letting x → x/b and carrying out the limiting process b → ∞. Thus, from (18.141) and (18.143), two linearly independent solutions of (18.147) are (when c is not an integer) a(a + 1) z 2 ax + + · · · ≡ M(a, c; x), c 1! c(c + 1) 2! y2 (x) = x1−c M(a − c + 1, 2 − c; x), y1 (x) = 1 + (18.148) (18.149) where M(a, c; x) is called the confluent hypergeometric function (or Kummer function).§ It is worth noting, however, that y1 (x) is singular when c = 0, −1, −2, . . . and y2 (x) is singular when c = 2, 3, 4, . . . . Thus, it is conventional to take the § We note that an alternative notation for the confluent hypergeometric function is 1 F1 (a, c; x). 633 SPECIAL FUNCTIONS second solution to (18.147) as a linear combination of (18.148) and (18.149) given by M(a − c + 1, 2 − c; x) M(a, c; x) π − x1−c U(a, c; x) ≡ . sin πc Γ(a − c + 1)Γ(c) Γ(a)Γ(2 − c) This has a well behaved limit as c approaches an integer. 18.11.1 Properties of confluent hypergeometric functions The properties of confluent hypergeometric functions can be derived from those of ordinary hypergeometric functions by letting x → x/b and taking the limit b → ∞, in the same way as both the equation and its solution were derived. A general procedure of this sort is called a confluence process. Special cases The general nature of the confluent hypergeometric equation allows one to write a large number of elementary functions in terms of the confluent hypergeometric functions M(a, c; x). Once again, such identifications can be made from the series expansion (18.148) directly, or by transformation of the confluent hypergeometric equation into a more familiar equation for which the solutions are already known. Some particular examples of well known special cases of the confluent hypergeometric function are as follows: M(a, a; x) = ex , M(−n, 1; x) = Ln (x), M(−n, 12 ; x2 ) = ex sinh x , x n!m! Lm (x), M(−n, m + 1; x) = (n + m)! n (−1)n n! H2n+1 (x) , M(−n, 32 ; x2 ), = 2(2n + 1)! x √ π M( 12 , 32 ; −x2 ) = erf(x), 2x M(1, 2; 2x) = (−1)n n! H2n (x), (2n)! x M(ν + 12 , 2ν + 1; 2ix) = ν!eix ( )−ν Jν (x), 2 where n and m are integers, Lm n (x) is an associated Legendre polynomial, Hn (x) is a Hermite polynomial, Jν (x) is a Bessel function and erf(x) is the error function discussed in section 18.12.4. Integral representation Using the integral representation (18.144) of the ordinary hypergeometric function, exchanging a and b and carrying out the process of confluence gives 1 Γ(c) etx ta−1 (1 − t)c−a−1 dt, M(a, c, x) = Γ(a)Γ(c − a) 0 (18.150) 634