15.

Hypergeometric Functions
F m z OBERHETTINGEB
Contents
page
Mathematical properties . . . . . . . . . . . . . . . . . . . . 556
15.1. G a w Series, Special Elementary Cases, Special Values of
the.Argument. . . . . . . . . . . . . . . . . . . . 556
15.2. DifferentiationFormdaa and GaussRelations for contiguous
Functions. . . . . . . . . . . . . . . . . . . . . . 557
15.3. Integral Representationsand Transformation F6rmulas . . . 558
15.4. Special cases of F(u, b; c; z), Polynomials and Legendre
Functions. . . . . . . . . . . . . . . . . . . . . . 561
15.5. The Hypergeometric Differential Equation . . . . . . . . 562
15.6. RiemannsDifferential Equation . . . . . . . . . . . . 564
15.7. Asymptotic Expansions . . . . . . . . . . . . . . . . 565
References . . . . . . . . . . . . . . . . . . . . . . . . . . 565

1 National Bureau of Stcmdarda. (Presently, Oregon StateUnivemity, Corvallis, Oregon.)

666
 15. Hypergeometric Functions

{{
Mathematical Properties
15.1. Gauss Series, Special Elementary Cases, 15.1.7
Special Values of the Argument F(3, 3; 8; -z')=(l+zZ)*F(1, 1; 8; -z2)
Gauss Series
=z-l In [z+(1+z2)f]
The circle of convergence of the Gauss hyper-
geometric series 15.1.8 F(a, b; b; Z)=(l-z)-"
15.1.1 15.1.9 F(u,3 + ~ f;
; ~'>=3[(1+2)-*+(1-~)-~]
F(a, b;c; 2) =zF1(a, b; c; 2) 15.1.10
z2)=
F(a, #+a;8;
32- (1 -2a)- [(1 +z)l--*- (1 -z)'-"]
15.1.11
F(--a,a; 4; -2') =3 [(1 +z2)*+z]~+[(1 + z y - 21")

is the unit circle 1z1=1. The behavior of this 15.1.12

series on its circle of convergence is:
(a) Divergence when 92 (c-a-b) 6 -1. F(a, 1-a; 3; -2)=
(b) Absolute convergence when 9 (c-a-b)>O. 3(1+2) -* [(1 +z2)*+z]s-'+[ (1 +z2))f--Z]--l}
(c) Conditional convergence when -1 <9? (c-a
-b) 60; the point z=1 is excluded. The Gauss 15.1.13
series reduces to a polynomial of degree n in z F(a, $+a; 1+%; 2)=2"[1+(1-2)*]-*
when a or b isequal to -n, (n=O, 1, 2, . . . ).
(For these cases see also 15.4.) The series 15.1.1 =(l-Z)*F(l+a, ++a;1f2a; 2)

is not defined when c is equal to -m, (m=O, 1, 15.1.14

2, . . .), provided a or b is not a negative integer
n with n<m. For c=-m F(a, i+u; 2a; 2) =23"-'(1-2) -q1+ (1 -z)']'-"
sin [(%-l)z]
15.1.2 15.1.15 F (a, 1--a; 6; sin2 z)=-.-(2a-1) sin2
1
lim -F(u, b; e; 2)'
e+-m r(c)

15.1.17 F(-U, a; 3; sin2Z)=COS (%z)

Special Elementary Cases of Gaurre Seriea 15.1.18 F(u,1--a; 4; sin2z)=cos [cos
(2u-1)zl
(For cases involving higher functions see. 15.4.)
15.1.3 F(1, 1; 2; z)=-z-lln (1-2) *
15.1.19 F(a, ++-a; 3; -tan2 z)=cosr z cos (%z)

1; Q; z2)=+z-'h (r=i>
15.1.4 l+z Special Values of the Argument
F(&,
15.1.20
15.1.5 F(3, 1; Q; -z*)=z-' arctan z
15.1.6
F(3, 3; 3; z2)=(1-z2)*F(1, 1; +; z*)=z-l arcsin z (CZO,-1, -2, . . ., g(c--a-b)>O)
 SYPERGEOMETRIC F"CI!IONS 557
15.1.21 15.2. Werentiation Formulas and Gauss'
Relations for Contiguous Functions
Differentiation Formulae

d ab
15.2.1 - F(a, b; C;z)=-C F(a+!, b+l; c+l; Z)
dz
15.2.2
-d" F(a, b; e; z)=- F(a+n, b+n; c+n; z)
dz" (4 11

15.2.3
d"
-&p [z"+"-'F(a,b; c; z)]= (a),z"-'F(a+n, b; c; 2)

15.2.4
d"
-
dz"
[zC-'F(a,b; c; z)]=(c-n).~~--"-~F(a,b; c-n; z)

15.2.5
-
d" [zc-a+'-l
(l-Z)"+"-"F(a, b; c; z)]
dz"
= (c-a)"Zc--cr-'(l-Z)a+b--c-"F(a-n, b; c; 2)
15.2.6
d"
- [(l-z)a+b-CF(a, b;c; z)]
dz"
3 (l-Z)d+b-C-RF(a, b; c+n; z)
(C)"

15.2.7
-d" [(l-Z>"+"-'F(a, b; c; 2))
dz"
a (-')"(a)n(c--b)~ (1- z)a-lF(a+n, 6; c+n; z)
(4"
15.2.8
d* [Zc-l(l-z)b--c+"F(a, b;c; z)]
-
dz"
(#-%#O, -1, -2,. . .) =(c-n)Rz"-"-'(l-z)b-cF(a-n, b; c-n; z)
15.2.9
d" [r-'(l-z)a+"-"F(a, b; c; z)]
dz"
(ii+@#O, -1, -2,. . .)
= (~-n),,z~-~-~(l b-n; e-n; z)
-~)~+~-~-"F(a-n,

G a u d Relations for Contiguous Functions

ThesixfunctionsF(af1,b;c; z),F(a,bfl;c; z),
F(a, b; c f l ; z) are c d e d contiguous to
#'(a,b; c; 2). Relatiom between P(a, b; c; z) and
 558 HYPERGEOMETRIC FUNCTIONS

any two contiguous functions have been given by 15.2.21

Gauss. By repeated application of these rela- [a-1-(c-b-l)z]F(a, b; e; 2)
tions the function F(a+m, b+n; c+Z; z) with
integral rn, n, I(c+Z+O, -1, -2, . . .) can be +(c-a)P(a-1, b; c; 2)
expressed as a linear combination of F(a, b; c; z) -(c-l)(l-z)F(a, 6; c-1; z)=O
and one of its contiguous functions with coefficients 15.2.22
which are rational functions of a, b, C, Z.
[c--%b+(b-u)z]F(a, b; c; 2)
15.2.10 +b(l-z)F(a, b+l; c; 2)
(c-u)F(a-l, b; e; z)+(2a-c--az+bz)F(a,b; c; 2) -(c-b)F(a, b- 1; c; 2) =o
+a@-l)F(a+l, b; c; z)=O 15.2.23
15.2.11 c[b-(c-a)z]F(a, b; c; 2)-bc(1-z)F(a, b+l; c; 2)
(c-b)F(a, b-1; c; 2)+(2b-c-bz+az)F(a, b; c; 2) + (c-a) (c-b)zF(a, b; e+ 1; 2) =o
+b(z-l)F(a, b+l; e; z)=O 15.2.24
15.2.12 (c-b-l)F(a, b; c; z)+bF(a, b+l; e; 2)
c(c-l)(z-l)F(a, b; E-1; 2) -(c-l)F(a, b; c-1; z)=O
+c[c-1-(2c-a-b-l)z]F(a, b; c; 2) 15.2.25
+(c-a)(c-b)zF(a, b; c+l; z)=O c(l-z)F(a, 6; c; 2)-cF(a, b-1; c; 2)
15.2.13 * + (c-u)zP(a, b; c+l; z)=O
[c-2a-(b-a)z]F(a, b; e; 2) 15.2.26
+a(l-z)F(a+l, b; c; 2)
[b-1-(c-a-l)z]F(a, b; c; 2)
-(c-a)F(a-1, b; e; z)=O
+(c-b)F(a, b-1; c; 2)
15.2.14
-(c-l)(l-z)F(a, b; c-1; z)=O
(Q-a)F(a, b; c; z)+aF(a+l, b; c; 2)
15.2.27
-bF(a, b+ 1; c; 2) =o
15.2.15 c[c-1- (2c-u-b- l)z]F(a, b; c; 2)
(c-a-b)F(a,.b; c; z)+a(l-z)F(a+l, b; c; 2) +(c-a)(c-b)zF(a, b; c+l; 2)
-(c-b)F(a, b-1; c; z)=O -c(c-l)(l-z)F(u, b; c-1; z)=O
15.2.16
c[a-(c-b)z]F(a, b; c; 2)-m(l-z)F(a+l, b; c; 2)
15.3. Integral Repmentations and Transfor-
+(c-a)(c-b)zF(a, b; c+l; z)=O
mation Formulae
15.2.17
(c-a-l)F(a, b; c; z)+aF(a+l, b; e; 2)
-(c-1)F(a, b; c-1; z)=O 15.3.1
15.2.18 F(a, b; c; z)=
(c-a-b)F(a, b; c; 2)-(c-a)F(a-1, b; e; 2)
+b(l-z)F(a, b+l; c; z)=O
15.2.19 (Wc>9b>O)
(b-a)(l-z)F(u, b; e; 2)-(c-a)F(a-1, b; c; 2) The integral represents a one valued analytic func-
+ (c-b)F(u, b- 1; c; 2) =o tion in the zplane cut along the real axis from 1 to
15.2.20 w and hence 15.3.1 gives the analytic continua-
tion of 15.1.1, F(u, b; e; z). Another integral
c(1-z)F(a, b; c; 2)-cF(a-1, b; c; 2) representation is in the form of a Mellin-Barnes
+(c-b)zF(u, b; c+l; z)=O integral
 HYPERGEOMETRIC FWNCl'IONS 559

Here -r<arg(-z)<r and the path of integration is chosen such that the pol- of r(a+s) and
r(b+s) i.e. the points s=-a-n and s=-b-m(n, m=O, 1, 2, . . .) respectively, are at its left
side and the poles of csc(rs8)or I'(-s) i.e. s=O, 1,2, are at ita right side. The cases in which -a, -b
or -c are non-negative integers or a-b equal to an integer are excluded.
Linear Tramformation F o r m u b
From 15.3.1 and 15.3.2 a number of transformation formulas for F(a, b; c; z) can be derived.
15.3.3 F(a, b; C; z)=(l-z)c-'"-bF(c-a, C-b; C; 2)

15.3.4 =(l-z)-T
(a, c-b; c; -
2-1 ">
15.3.5
2-1

15.3.7

15.3.8

15.3.9

Each term of 15.3.6 has a pole when c=a+bfm, (m=O, 1, 2, . . .); this case is covered by

15.3.10 F(a, b; a+b; z)= (a+b,

r(a)r(b)30
- [W(n+
1) -+(a+$ -+(b+n) --In (1-2)](1-2)"
<la% (1--z)l<*, l1-4<1)
Furthermore for m=1, 2, 3, . . .
560 HYPERGEOMETRIC FWNCMONS

(lw(1-4l<~,l1-4<1)
Similarly each term of 15.3.7 has a pole when b=afm or b--a= fm and the case is covered by

The case b-u=m, (m=1,2, 3, . . .) is covered by

15.3.14 F(a,a+m; C; ~)=F(u+~,u;c;

2)

(I&rg(-z)l<rJ 121>1, (c--a)#O, f l J f2,.*')

The case c - - a d , -1, -2, . . . becomes elementary, 15.3.3, and the case c-u=lJ 2, 3, . . . can be
obtained from 15.3.14, by a limiting proceas (see [15.2]).

Quadratic "ranoformadon Formulas

If, and only if the numbers &(l-c), &(a-b), f(-a+b-c) are such, that two of them are equal
or one of them is equal to 3, then there exists a quadratic transformation. The basic formulas
are due to Kummer [15.7Jand a complete list is due to Gomat [15.3]. See also [15.2].
15.3.15 F(-aJb;2b;2)=(l-2)-@F(&.Z,b-&.Z; b+a; 4x)
2'

15.3.16 =(l-a2)-"F(&.Z,3+3a;b+);d(2-2)-2)

15.3.17

15.3.18 =(1-2)-@F a, 2b--a; b+3; -

( -7
4 6

15.3.21 =(1-2)-=
< 2u, 2c-2u-1; E;
dl-z-l)
2 6

15.3.23
 HYPERGEOMETRIC FIJN<;TIONB 561

15.3.25 = (1-z)-#

15.3.26 F(a, 6; a-bi-1; z)=(l+~)-'F(h,

(++*JT_z)'-h F(2a-1,
a-N-1; 42(1+z)-')
a-b++; a+&+;
-9
fi+l

15.3.27 =(lf@-&F(a, a-b+); k-2b4-1; *4&(1&&)-')

15.3.28 = (1 -Z) -"F( h,&- b+); a-b+ 1;-k( 1-Z) -*)

15.3.29 F (a, b; &-lib+*; 2)=(1-2z)-"F(&, W+;; &+*b+*; )-

15.3.30 =F(h,3b; h+)b+k
42-43
15.3.31 F(a, 1-a; C; 3C-kh-3; C; 42-427
z)=(~-z)~-'F(~C-$Z,
15.3.32 = ( l - ~ ) " - ' ( l - 2 ~ ) ~ - ~ (k-h,
F &-h+);C; (42-4Z)(1-2Z)"3
Cubic transformations are listed in [15.2] and [15.3].
In the formulas above, the square roots are defined 80 that their value is real and positive when
OSz<l. All formulas are valid in the neighborhood of z=O.
15.4. Special Cases of F(a, b; c; z)
Polynomidm

When a or b is equal to a negative integer, then

This formula is also valid when c=-m-1; m, Z=O, 1,2, . . .

Some particular cams are

15.4.3 F(-n, n; 3; ~)=Tn(1-2~)
15.4.4 F(-n, n+l; 1; 2)=Pn(l-22)

15.4.5 F(-n,n+%;
1
1 al(:
a+-; 2 =-
2
C$')(1-22)

15.4.6 F(-n, a+l+P+n; a+l; x)=(a+l),,

nl P$ b)(1 -22)

Here T,,, p,,, t?$'), p$#)denote Chebyshev, Legendre's, Geger,auer and Jacobi's polynomials re-
spectively (see chapter 22).
Legendre Functions

Legendre functions are connected with those special casea of the hypergeometric function for which
a quadratic transformation exists (see 15.3).
15.4.7 F(a, b; 2b; ~ ) = 2 ~ - ~ r ( + + b(~-Z)*(~-~-*)P-*
)d-~ .-s-, 1 (1 -21-q [( -;)
15.4.8
 ~~~ ~~~

15.5. The Hypergeometric Merentid Equation

The hypergeometric differential equation

15.5.1 z(1-z) dlW dw

~+[c-(a+b+l)z] z-abw=O
dz
 HYPERGEOMETRIC F"CM0NS 563
has three (regular) singular points z=O, 1, a. The pairs of exponents at these points are

15.5.2 pi!'i=O, 1-C, pit!=O, C - ~ - b , pL?=a,b

resp&ively. The general theory of differential equations of the Fuchsian type distinguishea between
the following cases.
A. None of the numbers c, e-a-b; a-b is e@ to an integer. Then two linearly independent solutions
of 15.5.1 in the neighborhood of the singular points 0, 1, are respectively Q)

15.5.3 Wl(o)=F(a, b; C; Z)=(1-2)c-a-bF(C-a, c-b; C;Z)

15.5.4 m(0)= z1-'F(a- c + 1, b -c+ 1; 2 -c; Z) = z*-'( 1 -Z) '-a-bF(
1 -a, 1 -b; 2 -c; 2)

15.5.5 w~(~)=F(u,
b; a+b+l-C; 1-2)=~~-~F(l+b-c, 1-ta-c; a+b+l.-c; 1-2)

15.5.6 WZ,,,= (1 -~)'-'-~F(c-b, c-U; c-a-b+ 1; 1-2) =zl-'(l -z)~-"-'F(~

-U;l -b; C-U-b+ 1; 1-2)
15.5.7 UII(~)=Z-'F(U,
a-c+l; a-b+l; Z-1)=2'-C(2-l)C-a-bF(1-b, c-b; a-b+l; Z-')

15.5.8 m(-)=z-*F(b,b - ~ + l ; b-a+I; z-~)=~"-'(z-~)'-'-~F(~-u,

C-U; b - ~ + l ; 2-l)

The second set of the above expressions is obtained by applying 15.3.3 to the first set.
Another set of representations is obtained by applying 15.3.4 to -15.5.3 through 15.5.8. This
gives 15.5.9-15.5.14.
15.5.9 ~l(O)=(l-z)-"F 2-1

(
a-C+l, 1-b; 2-C; - =z~-'(~-z)'-~-'F
15.5.10 m(0)=~~-~(1--~)~-'-'F
">
b - ~ + l , 1-U; 2-c;-
2-1 ( ">
2-1

15.5.11 Wy(l)=z-"F(a,a-C+l; a+b-C+l; l-~-l)=z-'F(b, b - ~ + l ; a+b-C+l; 1-z-l)

15.5.12
m(,)=z"-C(l-z)'-'-bF(c-u, 1 --a; c-a-b+ 1 ; 1 -2-1) = i - C ( l -z)C-a-bF(c-b, 1 4 ; c-a-b+ 1; 1-2-1)

(
15.5.13 wl(=)=(z-l)-"F a, c-b; a-b+l; -)=(z-l)-bF(b,
1
1-2
e-a; b-u+l; -
1-2 '>
15.5.14
Wz(=)= 21-' (Z--I)"-"-'F
(
a-c+l, 1-b; a-b+l; - =zl-' (Z-I)"-~-'F b-c+l, 1-a; b-a+l; -
'>
1-2 ( '>
1-2
15.5.3 to 15.5.14 constitute Kummer's 24 solutions of the hypergeometric equation. The analytic con-
tinuation of w1,l(o)(z)can then be obtained by means of 15.3.3 to 15.3.9.
B. One of the numbers a, b, c-a, c-b is an integer. Then one of the hypergeometric series for
instance w1$@), 15.5.3, 15.5.4 terminates and the corresponding solution is of the form
15.5.15 w = ZQ(1 -~)@p,,
(Z)

where p.(z) is a polynomial in z of degree n. This case is referred to as the degenerate case of the
hypergeometric differential equation and its solutions are listed and discussed in great detail in [15.2].
C. The number c-a-b is an integer, c nonintegrd. Then 15.3.10 to 15.3.12 give the analytic continu-
ation of wl,l(o)into the neighborhood of z=1. Similarly 15.3.13 and 15.3.14 give the analytic continu-
ation of w12(o,into the neighborhood of z= m in case a-b is an integer but not c, subject of
course to the further restrictions c--a=O, f l , f 2 . . . (For a detailed discussion of all possible
cmes, see [15.2]).
D. The number c= 1. Then 15.5.3, 15.5.4 are replaced by
15.5.16 w~(O)=F(U,
b; 1; Z)
 ((zl<l and a, b f O , -1, -2, . . . -(m-1))

15.6. Riemann's Differential Equation Special Cases of Riemann's P Function

The hypergeometric differential equation 15.5.1 (a) The generalized hypergeometric function
with the (regular) singular points 0, 1, w is a
15.6.4
special caae of Riemann's differential equation
with three (regular) singular points a, b, e

15.6.1
#W I--oL-u' l-~-Br+l-~-~' dw (b) The hypergeometric function F(a, b; c; z)
=+[ z-a + z-b 2-c IZ
15.6.5
CYCY' (a- b)(a-c) j3Br(b-c)( b-a)
+[ 2-a + z-b
YYr(c--a) (C-b 1 W
+ 2-c I (z-a)(z-b)(z-c)=O 11-c b c-a-b J
(c) The Legendre functions P:(z), @(z)
The pairs of the exponents with respect to the
15.6.6
singular points a; b; c are a, a'; 0, 8'; 7 , 'y' respec-
tively subject to the condition

15.6.2 a+a'+8+8' +r+r'=1 13+4v -4P 3 J

(d) The confluent hypergeometric function
The complete set of solutions of 15.6.1 is denoted
by the symbol 15.6.7
o m C

15.6.3 b c w=P{$+u -c c-k z]

3-u O k
provided lim c+m.
where

15.6.11 P { :, i, ;,
HYPERGEOMETRIC FUNCl'IONS

Transformation Formulas for Riemann's P Function

and A, B, C, D are arbitrary constants such that AD-BOZO.

Riemann's P function reduced to the hypergeometric function is
a b c

~ } = (2-b
~ , . ( - 2-b
)~p{
0

0
a'-a
Q)

d-B+y
a+B'+y 7'-y

by Kummer's 24 solutions 15.5.3 to 15.5.14 the complete set of 24 solutions for Riemann's differential
equation 15.6.1 is obtained. The first of these solutions is for instance by 15.5.3 and 15.6.5

15.7. Asymptotic Expamione

The behavior of F(a, b; c; z) for large (21 is
described by the transformation formulas of 15.3.
For fixed a, b, z and large IcI one has [15.8]

15.7.1

For fixeda, c, z, (c#O, -1, -2,

and large Ibl one has [15.2]
. . . , O<lzl<)

115.11 P. Appell and J. Kampe de FBriet, Fonctions hyper-

g6omBtriques et hyperspheriques (Gauthiers-
Villars, Paris, France, 1926).
115.21 A. ErdBlyi et a l . ,Higher transcendental functions,
vol. 1 (McGraw-Hill Book Co., Inc., New York,.
N.Y., 1953).
i15.31 E. Goursat, Ann. Sci. &ole
3-142(1881).
Norm. Sup(2)10,
References
1

The P function on the right hand side is Gauss' hypergeometric function (see 15.6.5). If it is replaced
4

(z-a)(c-b)
0 (2- b) (c-U)

[15.4] E. Goursat, PropriBtA generales de 1'Bquation

565

d'Euler et de Gauss (Actualitb scientiflquea et

industriellea 333, Paris, France, 1936).
(15.5) J. KampB de FBriet, La fonction hyperg6omBtnque
(Gauthiers-Villars, Paris, France, 1937).
[15.6] E. Klein, Vorlesungen iiber die hypergeometriache
Funktion (B. G. Teubner, Berlin, Germany,
1933).
566 HYPERGEOMETRIC FUNCTIONS
(15.71 E. E. Kummer, uber die hypergeometrische Reihe,
J. Reine Angew. Math. 15,39-83, 127-172(1836).
[15.8] T. M. MacRobert, Proc. Edinburgh Math. Soc. 42,
84-88( 1923).
[15.9] T. M. MacRobert, Functions of a complex variable,
4th ed. (Macmillan and Co., Ltd., London,
England, 1954).
[15.10] E. G. C. Poole, Introduction to the theory of linear
ditrerential equations (Clarendon Press, Oxford,
England, 1936).
[15.11] C. Snow, The hypergeometric and Legendre func-
tions with applications to integral equations of
potential theory, Applied Math. Series 19 (U.S.
Government Printing Office, Washington, D.C.,
1952).
[15.121 E. T.Whittaker and G. N. Watson, A course of
modem analysis, 4th ed. (Cambridge Univ. Press,
Cambridge, England, 1952).