Hypergeometric function

Solutions to the hypergeometric differential equation are built out of the hypergeometric series ₂F₁(a,b;c;z). The equation has two linearly independent solutions. At each of the three singular points 0, 1, ∞, there are usually two special solutions of the form x^s times a holomorphic function of x, where s is one of the two roots of the indicial equation and x is a local variable vanishing at a regular singular point. This gives 3 × 2 = 6 special solutions, as follows.

Around the point z = 0, two independent solutions are, if c is not a non-positive integer,

$${\displaystyle \,_{2}F_{1}(a,b;c;z)}$$

and, on condition that c is not an integer,

$${\displaystyle z^{1-c}\,_{2}F_{1}(1+a-c,1+b-c;2-c;z)}$$

If c is a non-positive integer 1−m, then the first of these solutions does not exist and must be replaced by ${\displaystyle z^{m}F(a+m,b+m;1+m;z).}$ The second solution does not exist when c is an integer greater than 1, and is equal to the first solution, or its replacement, when c is any other integer. So when c is an integer, a more complicated expression must be used for a second solution, equal to the first solution multiplied by ln(z), plus another series in powers of z, involving the digamma function. See Olde Daalhuis (2010) for details.

Around z = 1, if c − a − b is not an integer, one has two independent solutions

$${\displaystyle \,_{2}F_{1}(a,b;1+a+b-c;1-z)}$$

and

$${\displaystyle (1-z)^{c-a-b}\;_{2}F_{1}(c-a,c-b;1+c-a-b;1-z)}$$

Around z = ∞, if a − b is not an integer, one has two independent solutions

$${\displaystyle z^{-a}\,_{2}F_{1}\left(a,1+a-c;1+a-b;z^{-1}\right)}$$

and

$${\displaystyle z^{-b}\,_{2}F_{1}\left(b,1+b-c;1+b-a;z^{-1}\right).}$$

Again, when the conditions of non-integrality are not met, there exist other solutions that are more complicated.

Any 3 of the above 6 solutions satisfy a linear relation as the space of solutions is 2-dimensional, giving (⁶
₃) = 20 linear relations between them called connection formulas.

A second order Fuchsian equation with n singular points has a group of symmetries acting (projectively) on its solutions, isomorphic to the Coxeter group W(D_n) of order 2ⁿ⁻¹n!. The hypergeometric equation is the case n = 3, with group of order 24 isomorphic to the symmetric group on 4 points, as first described by Kummer. The appearance of the symmetric group is accidental and has no analogue for more than 3 singular points, and it is sometimes better to think of the group as an extension of the symmetric group on 3 points (acting as permutations of the 3 singular points) by a Klein 4-group (whose elements change the signs of the differences of the exponents at an even number of singular points). Kummer's group of 24 transformations is generated by the three transformations taking a solution F(a,b;c;z) to one of

$${\displaystyle {\begin{aligned}(1-z)^{-a}F\left(a,c-b;c;{\tfrac {z}{z-1}}\right)\\F(a,b;1+a+b-c;1-z)\\(1-z)^{-b}F\left(c-a,b;c;{\tfrac {z}{z-1}}\right)\end{aligned}}}$$

which correspond to the transpositions (12), (23), and (34) under an isomorphism with the symmetric group on 4 points 1, 2, 3, 4. (The first and third of these are actually equal to F(a,b;c;z) whereas the second is an independent solution to the differential equation.)

Applying Kummer's 24 = 6×4 transformations to the hypergeometric function gives the 6 = 2×3 solutions above corresponding to each of the 2 possible exponents at each of the 3 singular points, each of which appears 4 times because of the identities

$${\displaystyle {\begin{aligned}{}_{2}F_{1}(a,b;c;z)&=(1-z)^{c-a-b}\,{}_{2}F_{1}(c-a,c-b;c;z)&&{\text{Euler transformation}}\\{}_{2}F_{1}(a,b;c;z)&=(1-z)^{-a}\,{}_{2}F_{1}(a,c-b;c;{\tfrac {z}{z-1}})&&{\text{Pfaff transformation}}\\{}_{2}F_{1}(a,b;c;z)&=(1-z)^{-b}\,{}_{2}F_{1}(c-a,b;c;{\tfrac {z}{z-1}})&&{\text{Pfaff transformation}}\end{aligned}}}$$

The hypergeometric differential equation may be brought into the Q-form

$${\displaystyle {\frac {d^{2}u}{dz^{2}}}+Q(z)u(z)=0}$$

by making the substitution u = wv and eliminating the first-derivative term. One finds that

$${\displaystyle Q={\frac {z^{2}[1-(a-b)^{2}]+z[2c(a+b-1)-4ab]+c(2-c)}{4z^{2}(1-z)^{2}}}}$$

and v is given by the solution to

$${\displaystyle {\frac {d}{dz}}\log v(z)=-{\frac {c-z(a+b+1)}{2z(1-z)}}=-{\frac {c}{2z}}-{\frac {1+a+b-c}{2(z-1)}}}$$

which is

$${\displaystyle v(z)=z^{-c/2}(1-z)^{(c-a-b-1)/2}.}$$

The Q-form is significant in its relation to the Schwarzian derivative (Hille 1976, pp. 307–401).

The Schwarz triangle maps or Schwarz s-functions are ratios of pairs of solutions.

$${\displaystyle s_{k}(z)={\frac {\phi _{k}^{(1)}(z)}{\phi _{k}^{(0)}(z)}}}$$

where k is one of the points 0, 1, ∞. The notation

$${\displaystyle D_{k}(\lambda ,\mu ,\nu ;z)=s_{k}(z)}$$

is also sometimes used. Note that the connection coefficients become Möbius transformations on the triangle maps.

Note that each triangle map is regular at z ∈ {0, 1, ∞} respectively, with

$${\displaystyle {\begin{aligned}s_{0}(z)&=z^{\lambda }(1+{\mathcal {O}}(z))\\s_{1}(z)&=(1-z)^{\mu }(1+{\mathcal {O}}(1-z))\end{aligned}}}$$

and

$${\displaystyle s_{\infty }(z)=z^{\nu }(1+{\mathcal {O}}({\tfrac {1}{z}})).}$$

In the special case of λ, μ and ν real, with 0 ≤ λ,μ,ν < 1 then the s-maps are conformal maps of the upper half-plane H to triangles on the Riemann sphere, bounded by circular arcs. This mapping is a generalization of the Schwarz–Christoffel mapping to triangles with circular arcs. The singular points 0,1 and ∞ are sent to the triangle vertices. The angles of the triangle are πλ, πμ and πν respectively.

Furthermore, in the case of λ=1/p, μ=1/q and ν=1/r for integers p, q, r, then the triangle tiles the sphere, the complex plane or the upper half plane according to whether λ + μ + ν – 1 is positive, zero or negative; and the s-maps are inverse functions of automorphic functions for the triangle group 〈p, q, r〉 = Δ(p, q, r).

The monodromy of a hypergeometric equation describes how fundamental solutions change when analytically continued around paths in the z plane that return to the same point. That is, when the path winds around a singularity of ₂F₁, the value of the solutions at the endpoint will differ from the starting point.

Two fundamental solutions of the hypergeometric equation are related to each other by a linear transformation; thus the monodromy is a mapping (group homomorphism):

$${\displaystyle \pi _{1}(\mathbf {C} \setminus \{0,1\},z_{0})\to {\text{GL}}(2,\mathbf {C} )}$$

where π₁ is the fundamental group. In other words, the monodromy is a two dimensional linear representation of the fundamental group. The monodromy group of the equation is the image of this map, i.e. the group generated by the monodromy matrices. The monodromy representation of the fundamental group can be computed explicitly in terms of the exponents at the singular points.[1] If (α, α'), (β, β') and (γ,γ') are the exponents at 0, 1 and ∞, then, taking z₀ near 0, the loops around 0 and 1 have monodromy matrices

$${\displaystyle {\begin{aligned}g_{0}&={\begin{pmatrix}e^{2\pi i\alpha }&0\\0&e^{2\pi i\alpha ^{\prime }}\end{pmatrix}}\\g_{1}&={\begin{pmatrix}{\mu e^{2\pi i\beta }-e^{2\pi i\beta ^{\prime }} \over \mu -1}&{\mu (e^{2\pi i\beta }-e^{2\pi i\beta ^{\prime }}) \over (\mu -1)^{2}}\\e^{2\pi i\beta ^{\prime }}-e^{2\pi i\beta }&{\mu e^{2\pi i\beta ^{\prime }}-e^{2\pi i\beta } \over \mu -1}\end{pmatrix}},\end{aligned}}}$$

where

$${\displaystyle \mu ={\sin \pi (\alpha +\beta ^{\prime }+\gamma ^{\prime })\sin \pi (\alpha ^{\prime }+\beta +\gamma ^{\prime }) \over \sin \pi (\alpha ^{\prime }+\beta ^{\prime }+\gamma ^{\prime })\sin \pi (\alpha +\beta +\gamma ^{\prime })}.}$$

If 1−a, c−a−b, a−b are non-integer rational numbers with denominators k,l,m then the monodromy group is finite if and only if ${\displaystyle 1/k+1/l+1/m>1}$, see Schwarz's list or Kovacic's algorithm.

If B is the beta function then

$${\displaystyle \mathrm {B} (b,c-b)\,_{2}F_{1}(a,b;c;z)=\int _{0}^{1}x^{b-1}(1-x)^{c-b-1}(1-zx)^{-a}\,dx\qquad \Re (c)>\Re (b)>0,}$$

provided that z is not a real number such that it is greater than or equal to 1. This can be proved by expanding (1 − zx)^−a using the binomial theorem and then integrating term by term for z with absolute value smaller than 1, and by analytic continuation elsewhere. When z is a real number greater than or equal to 1, analytic continuation must be used, because (1 − zx) is zero at some point in the support of the integral, so the value of the integral may be ill-defined. This was given by Euler in 1748 and implies Euler's and Pfaff's hypergeometric transformations.

Other representations, corresponding to other branches, are given by taking the same integrand, but taking the path of integration to be a closed Pochhammer cycle enclosing the singularities in various orders. Such paths correspond to the monodromy action.

Barnes used the theory of residues to evaluate the Barnes integral

$${\displaystyle {\frac {1}{2\pi i}}\int _{-i\infty }^{i\infty }{\frac {\Gamma (a+s)\Gamma (b+s)\Gamma (-s)}{\Gamma (c+s)}}(-z)^{s}\,ds}$$

as

$${\displaystyle {\frac {\Gamma (a)\Gamma (b)}{\Gamma (c)}}\,_{2}F_{1}(a,b;c;z),}$$

where the contour is drawn to separate the poles 0, 1, 2... from the poles −a, −a − 1, ..., −b, −b − 1, ... . This is valid as long as z is not a nonnegative real number.

The Gauss hypergeometric function can be written as a John transform (Gelfand, Gindikin & Graev 2003, 2.1.2).

Gauss used the contiguous relations to give several ways to write a quotient of two hypergeometric functions as a continued fraction, for example:

$${\displaystyle {\frac {{}_{2}F_{1}(a+1,b;c+1;z)}{{}_{2}F_{1}(a,b;c;z)}}={\cfrac {1}{1+{\cfrac {{\frac {(a-c)b}{c(c+1)}}z}{1+{\cfrac {{\frac {(b-c-1)(a+1)}{(c+1)(c+2)}}z}{1+{\cfrac {{\frac {(a-c-1)(b+1)}{(c+2)(c+3)}}z}{1+{\cfrac {{\frac {(b-c-2)(a+2)}{(c+3)(c+4)}}z}{1+{}\ddots }}}}}}}}}}}$$

Euler's transformation is

$${\displaystyle {}_{2}F_{1}(a,b;c;z)=(1-z)^{c-a-b}{}_{2}F_{1}(c-a,c-b;c;z).}$$

It follows by combining the two Pfaff transformations

$${\displaystyle {\begin{aligned}{}_{2}F_{1}(a,b;c;z)&=(1-z)^{-b}{}_{2}F_{1}\left(b,c-a;c;{\tfrac {z}{z-1}}\right)\\{}_{2}F_{1}(a,b;c;z)&=(1-z)^{-a}{}_{2}F_{1}\left(a,c-b;c;{\tfrac {z}{z-1}}\right)\\\end{aligned}}}$$

which in turn follow from Euler's integral representation. For extension of Euler's first and second transformations, see Rathie & Paris (2007) and Rakha & Rathie (2011). It can also be written as linear combination

$${\displaystyle {\begin{aligned}{}_{2}F_{1}(a,b;c,z)={}&{\frac {\Gamma (c)\Gamma (c-a-b)}{\Gamma (c-a)\Gamma (c-b)}}{}_{2}F_{1}(a,b;a+b+1-c;1-z)\\[6pt]&{}+{\frac {\Gamma (c)\Gamma (a+b-c)}{\Gamma (a)\Gamma (b)}}(1-z)^{c-a-b}{}_{2}F_{1}(c-a,c-b;1+c-a-b;1-z).\end{aligned}}}$$

If two of the numbers 1 − c, c − 1, a − b, b − a, a + b − c, c − a − b are equal or one of them is 1/2 then there is a quadratic transformation of the hypergeometric function, connecting it to a different value of z related by a quadratic equation. The first examples were given by Kummer (1836), and a complete list was given by Goursat (1881). A typical example is

$${\displaystyle {}_{2}F_{1}(a,b;2b;z)=(1-z)^{-{\frac {a}{2}}}{}_{2}F_{1}\left({\tfrac {1}{2}}a,b-{\tfrac {1}{2}}a;b+{\tfrac {1}{2}};{\frac {z^{2}}{4z-4}}\right)}$$

If 1−c, a−b, a+b−c differ by signs or two of them are 1/3 or −1/3 then there is a cubic transformation of the hypergeometric function, connecting it to a different value of z related by a cubic equation. The first examples were given by Goursat (1881). A typical example is

$${\displaystyle {}_{2}F_{1}\left({\tfrac {3}{2}}a,{\tfrac {1}{2}}(3a-1);a+{\tfrac {1}{2}};-{\tfrac {z^{2}}{3}}\right)=(1+z)^{1-3a}\,{}_{2}F_{1}\left(a-{\tfrac {1}{3}},a;2a;2z(3+z^{2})(1+z)^{-3}\right)}$$

There are also some transformations of degree 4 and 6. Transformations of other degrees only exist if a, b, and c are certain rational numbers (Vidunas 2005). For example,

$${\displaystyle {}_{2}F_{1}\left({\tfrac {1}{4}},{\tfrac {3}{8}};{\tfrac {7}{8}};z\right)(z^{4}-60z^{3}+134z^{2}-60z+1)^{1/16}={}_{2}F_{1}\left({\tfrac {1}{48}},{\tfrac {17}{48}};{\tfrac {7}{8}};{\tfrac {-432z(z-1)^{2}(z+1)^{8}}{(z^{4}-60z^{3}+134z^{2}-60z+1)^{3}}}\right).}$$

Gauss's summation theorem, named for Carl Friedrich Gauss, is the identity

$${\displaystyle {}_{2}F_{1}(a,b;c;1)={\frac {\Gamma (c)\Gamma (c-a-b)}{\Gamma (c-a)\Gamma (c-b)}},\qquad \Re (c)>\Re (a+b)}$$

which follows from Euler's integral formula by putting z = 1. It includes the Vandermonde identity as a special case.

For the special case where ${\displaystyle a=-m}$,

$${\displaystyle {}_{2}F_{1}(-m,b;c;1)={\frac {(c-b)_{m}}{(c)_{m}}}}$$

Dougall's formula generalizes this to the bilateral hypergeometric series at z = 1.

There are many cases where hypergeometric functions can be evaluated at z = −1 by using a quadratic transformation to change z = −1 to z = 1 and then using Gauss's theorem to evaluate the result. A typical example is Kummer's theorem, named for Ernst Kummer:

$${\displaystyle {}_{2}F_{1}(a,b;1+a-b;-1)={\frac {\Gamma (1+a-b)\Gamma (1+{\tfrac {1}{2}}a)}{\Gamma (1+a)\Gamma (1+{\tfrac {1}{2}}a-b)}}}$$

which follows from Kummer's quadratic transformations

$${\displaystyle {\begin{aligned}_{2}F_{1}(a,b;1+a-b;z)&=(1-z)^{-a}\;_{2}F_{1}\left({\frac {a}{2}},{\frac {1+a}{2}}-b;1+a-b;-{\frac {4z}{(1-z)^{2}}}\right)\\&=(1+z)^{-a}\,_{2}F_{1}\left({\frac {a}{2}},{\frac {a+1}{2}};1+a-b;{\frac {4z}{(1+z)^{2}}}\right)\end{aligned}}}$$

and Gauss's theorem by putting z = −1 in the first identity. For generalization of Kummer's summation, see Lavoie, Grondin & Rathie (1996).

Gauss's second summation theorem is

$${\displaystyle _{2}F_{1}\left(a,b;{\tfrac {1}{2}}\left(1+a+b\right);{\tfrac {1}{2}}\right)={\frac {\Gamma ({\tfrac {1}{2}})\Gamma ({\tfrac {1}{2}}\left(1+a+b\right))}{\Gamma ({\tfrac {1}{2}}\left(1+a)\right)\Gamma ({\tfrac {1}{2}}\left(1+b\right))}}.}$$

Bailey's theorem is

$${\displaystyle _{2}F_{1}\left(a,1-a;c;{\tfrac {1}{2}}\right)={\frac {\Gamma ({\tfrac {1}{2}}c)\Gamma ({\tfrac {1}{2}}\left(1+c\right))}{\Gamma ({\tfrac {1}{2}}\left(c+a\right))\Gamma ({\tfrac {1}{2}}\left(1+c-a\right))}}.}$$

For generalizations of Gauss's second summation theorem and Bailey's summation theorem, see Lavoie, Grondin & Rathie (1996).

There are many other formulas giving the hypergeometric function as an algebraic number at special rational values of the parameters, some of which are listed in Gessel & Stanton (1982) and Koepf (1995). Some typical examples are given by

$${\displaystyle {}_{2}F_{1}\left(a,-a;{\tfrac {1}{2}};{\tfrac {x^{2}}{4(x-1)}}\right)={\frac {(1-x)^{a}+(1-x)^{-a}}{2}},}$$

which can be restated as

$${\displaystyle T_{a}(\cos x)={}_{2}F_{1}\left(a,-a;{\tfrac {1}{2}};{\tfrac {1}{2}}(1-\cos x)\right)=\cos(ax)}$$

whenever −π < x < π and T is the (generalized) Chebyshev polynomial.