Recurrence Relations for Hypergeometric Functions
Abstract Generated abstract
This paper derives recurrence relations for selected hypergeometric functions from known continued fraction identities. It treats the Gauss hypergeometric function, the confluent hypergeometric function, and the generalized series 2F0, using transformations of ratios of contiguous functions to express each function in terms of shifted-parameter functions. The resulting elementary relations involve polynomial coefficients, for which explicit finite sum formulas are provided.
Full Text
RECURRENCE RELATIONS
FOR HYPERGEOMETRIC FUNCTIONS
V. E. Kornilov
(Presented by the Department of Higher Mathematics)
In the present paper, known identical transformations of two particular hypergeometric series are used to obtain recurrence relations between these series.
Elementary relations are obtained in the general case for the following functions:
\[ F(a,b;c;z)=\sum_{m=0}^{\infty}\frac{(a)_m(b)_m}{m!(c)_m}z^m,\quad |z|<1; \]
\[ \Phi(a;c;z)=\sum_{m=0}^{\infty}\frac{(a)_m}{m!(c)_m}z^m,\quad |z|<+\infty; \]
\[ {}_2F_0(a,b;;z)=\sum_{m=0}^{\infty}\frac{(a)_m(b)_m}{m!}z^m,\quad |z|\ll 1, \]
where
\[ (a)_m=a(a+1)\cdots(a+m-1),\quad (a)_0=1. \]
- Let us study the following continued fraction (identity) ([1], p. 98; [2], p. 132)
\[ \left\{ \begin{aligned} \frac{F(a,b+1;c+1;z)}{F(a,b;c;z)} &= \cfrac{1}{1-\cfrac{u_1z}{1}-\cfrac{v_1z}{1}-\cdots- \cfrac{v_nz\,F(a+n,b+n+1;c+2n+1;z)} {F(a+n,b+n;c+2n;z)}} , \\[6pt] u_j&=\frac{(a+j-1)(c-b+j-1)}{(c+2j-2)_2},\quad v_j=\frac{(b+j)(c-a+j)}{(c+2j-1)_2},\quad j=1,\cdots,n. \end{aligned} \right. \tag{1} \]
Let us represent the right-hand side of identity (1) by a suitable fraction
\[ \frac{F(a,b+1;c+1;z)}{F(a,b;c;z)} = \frac{p_{2n}F_1-v_nzp_{2n-1}F_2} {q_{2n}F_1-v_nzq_{2n-1}F_2}. \]
From the last identity it is not difficult to obtain the following elementary relation
\[ \left\{ \begin{aligned} F(a,b;c;z)&=q_{2n}F(a+n,b+n;c+2n;z)-\\ &\quad -v_n z q_{2n-1}F(a+n,b+n+1;c+2n+1;z). \end{aligned} \right. \tag{2} \]
2. The following are two limiting cases of identity (1) ([2], pp. 134–137):
\[ \left\{ \begin{aligned} \frac{\Phi(a+1;c+1;z)}{\Phi(a;c;z)} &= \frac{1}{1} -\frac{a_1z}{1} -\frac{b_1z}{1} -\cdots- \frac{b_nz\,\Phi(a+n+1;c+2n+1;z)\,[\Phi(a+n;c+2n;z)]^{-1}}{1},\\ a_j&=\frac{c-a+j-1}{(c+2j-2)_2},\qquad b_j=-\frac{a+j}{(c+2j-1)_2}. \end{aligned} \right. \tag{3} \]
\[ \left\{ \begin{aligned} \frac{{}_2F_0(a,b+1;;z)}{{}_2F_0(a,b;;z)} &= \frac{1}{1} -\frac{c_1z}{1} -\frac{d_1z}{1} -\cdots-\\ &\quad -\frac{d_nz\,{}_2F_0(a+n,b+n+1;;z)\,[{}_2F_0(a+n,b+n;;z)]^{-1}}{1},\\ c_j&=a+j-1,\qquad d_j=b+j,\qquad j=1,\ldots,n. \end{aligned} \right. \tag{4} \]
3. Similarly to item 1, from equalities (3) and (4) we obtain elementary relations for the functions:
\[ \left\{ \begin{aligned} \Phi(a;c;z)&=B_{2n}\Phi(a+n;c+2n;z)-\\ &\quad -b_nz\,B_{2n-1}\Phi(a+n+1;c+2n+1;z). \end{aligned} \right. \tag{5} \]
\[ \left\{ \begin{aligned} {}_2F_0(a,b;;z)&=D_{2n}{}_2F_0(a+n,b+n;;z)-\\ &\quad -d_nz\,D_{2n-1}{}_2F_0(a+n,b+n+1;;z). \end{aligned} \right. \tag{6} \]
4. The values of the polynomials in equalities (2), (5), (6) are as follows:
\[ \left\{ \begin{aligned} q_{2n}&=q_{2n+1}+u_{n+1}zq_{2n-1},\qquad q_{2n-1}=\\ &=\sum_{m=0}^{n-1}\frac{(-1)^m C_{n-1}^m z^m}{(c+2n-1-m)_m} \left[ \sum_{k=0}^{m} \frac{(-1)^k C_m^k(m+1-2n)_k\,\lambda_m} {(1-n)_k(c)_k[(a)_k(b)_k]^{-1}} \right]. \end{aligned} \right. \tag{7} \]
\[ \left\{ \begin{aligned} B_{2n-1} &=\sum_{m=0}^{n-1} \frac{(-1)^m C_{n-1}^m z^m}{(c+2n-1-m)_m} \left[ \sum_{k=0}^{m} \frac{(-1)^k C_m^k(m+1-2n)_k} {(1-n)_k(c)_k[(a)_k]^{-1}} \right],\\ B_{2n} &=\sum_{m=0}^{n} \frac{(-1)^m C_n^m z^m}{(c+2n-m)_m} \left[ \sum_{k=0}^{m} \frac{(-1)^k C_m^k(m-2n)_k} {(-n)_k(c)_k[(a)_k]^{-1}} \right]. \end{aligned} \right. \tag{8} \]
\[ \left\{ \begin{aligned} D_{2n}&=D_{2n+1}+a_{n+1}zD_{2n-1},\qquad D_{2n-1}=\\ &=\sum_{m=0}^{n-1}\frac{C_{n-1}^m}{z^{-m}} \left[ \sum_{k=0}^{\left[\frac{m}{2}\right]} \frac{C_m^k(-1)^k(k-m)_k(a)_k(b)_k} {(1-n)_k[(a+b+n+2k-m)_{m-2k}]^{-1}} \right]. \end{aligned} \right. \tag{9} \]
In equality (7) the coefficient \(\lambda_m\) is the following:
\[ \lambda_m= \sum_{p=0}^{\left[\frac{m-k}{2}\right]} \frac{C_{m-k}^p(-1)^p(m-k-2p+1)_p(a+k)_p(b+k)_p} {(n-k-p)_p[(a+b+n+2k+2p-m)_{m-2p-2k}]^{-1}}. \tag{10} \]
References
-
H. Bateman and A. Erdélyi. Higher Transcendental Functions. Moscow, Fizmatgiz, 1965.
-
A. N. Khovanskii. The Application of Continued Fractions and Their Generalizations to Problems of Approximate Analysis. Moscow, Gostekhizdat, 1956.