Page 458 - A First Course In Stochastic Models
P. 458

C. GENERATING FUNCTIONS                    453

                region |z| < R in the complex plane for some R > 1. It is essential that the radius
                R is larger than 1. Note that the generating function (C.5) is indeed of the form
                (C.6), where the numerator and denominator are analytic functions on the whole
                complex plane (R = ∞). It is no restriction to assume that N(z) and D(z) have
                no common zeros; otherwise, cancel out common zeros. Let us further assume that
                the following regularity conditions are satisfied:

                 C1 The equation D(z) = 0 has a real root z 0 on the interval (1,R).

                 C2 The function D(z) has no zeros in the domain 1 < |z| < z 0 of the complex
                 plane.
                 C3 The zero z = z 0 of D(z) is of multiplicity 1 and is the only zero of D(z) on
                 the circle |z| = z 0 .

                  The following theorem is of utmost importance. The insightful proof of the
                theorem is included for completeness. Recall that f (x) ∼ g(x) as x → ∞ means
                that f (x)/g(x) → 1 as x → ∞.

                Theorem C.1 Under the conditions C1 to C3,
                                              −j
                                       p j ∼ γ 0 z 0  as j → ∞,                (C.7)

                where the constant γ 0 is given by
                                                 1 N(z 0 )
                                          γ 0 = −       .                      (C.8)
                                                z 0 D (z 0 )
                                                    ′
                      ′
                Here D (z 0 ) denotes the derivative of D(x) at x = z 0 .
                Proof  We first mention the following basic facts from complex function theory.
                The most important fact is that a function f (z) is analytic at a point z = a if
                                                                    ∞
                                                                        a
                                                                  
            n
                and only if f (z) can be expanded in a power series f (z) =  n=0 n (z − a) in
                |z − a| < ρ for some ρ > 0. The coefficient a n of the Taylor series is the nth
                derivative of f (z) at z = a divided by n!. The analytic function f (z) is said to
                have a zero of multiplicity k in z = a if a 0 = · · · = a k−1 = 0 and a k 	= 0. Another
                basic result is the following. The Taylor series  
 ∞  a  n
                                                        n=0 n (z−a) of a function f (z)
                at the point z = a coincides with the function f (z) in the interior of the largest
                circle whose interior lies wholly within the domain on which f (z) is analytic.
                  The proof of (C.7) now proceeds as follows. The conditions C1 to C3 imply
                that there is a circle around z = 0 with radius R 0 larger than z 0 such that P (z) is
                analytic in |z| < R 0 except for the isolated point z = z 0 . Since D(z) has a zero of
                multiplicity 1 at z = z 0 , it follows from the Taylor series that D(z) = (z − z 0 )φ(z)
                in |z| < R 0 , where φ(z) is an analytic function with φ(z 0 ) 	= 0. Thus we can write
                P (z) as P (z) = H(z)/(z − z 0 ) for some analytic function H(z) in |z| < R 0 with
                H(z 0 ) 	= 0. Using a Taylor expansion H(z) = H(z 0 ) + (z − z 0 )U(z), we next find
   453   454   455   456   457   458   459   460   461   462   463