⎪

TU Delft website Contact Nederlands

Roelof Koekoek

Home Email Route Teaching Research Private

Teaching

Roelof Koekoek's teaching page > Special Functions > Hypergeometric functions > Barnes'
Special Functions
integral representation
Hypergeometric functions
Special Functions – Hypergeometric functions – Barnes' integral representation
Properties

The Mellin transform of a function is given by

Barnes' integral
representation ∞
F(s) := ∫ xs−1 f(x) dx.
0
Differential equations
The inversion formula is given by
1 c+i∞
f (x) = ∫ x−s F(s) ds, c > 0.
2πi c−i∞

Note that the definition of the gamma function,

∞
Γ(s) = ∫ xs−1 e−x dx, Re(s) > 0,
0

implies that Γ(s) is the Mellin transform of e−x . The inversion formula reads
1 c+i∞
e−x = ∫ x−s Γ(s) ds, c > 0.
2πi c−i∞

This can be proved directly by using Cauchy's residue theorem. Consider the rectangular
contour C with vertices c ± iR (c > 0) and −(N + 1/2) ± iR, where N is a positive integer. Then the
poles of Γ(s) inside this contour are at 0, −1, −2, … , −N with residues (−1)j /j! for j = 0, 1, 2, … , N
respectively. Hence, Cauchy's residue theorem implies that
N (−1) j
1
∫ x Γ(s) ds = ∑
−s
xj .
2πi C j=0
j!

Now we let R and N tend to infinity. Then Stirling's asymptotic formula for the gamma
function implies that the integral on C minus the line joining c − iR and c + iR tends to zero.
Hence we have
N (−1) j ∞ (−x) j
1 c+i∞
∫ x Γ(s) ds = lim ∑
−s
x =∑
j
= e−x .
2πi c−i∞ N→∞
j=0
j!
j=0
j!

Another example of a Mellin transform is:

1
B(s, t) = ∫ xs−1 (1 − x)t−1 dx, Re(s) > 0, Re(t) > 0.
0

This implies that the beta function B(s, t) is the Mellin transform of the function
(1 − x)t−1 , 0 < x < 1
f (x) := {
0, x ≥ 1.

The inversion formula then reads

1 c+i∞ Γ(t) c+i∞ −s Γ(s)

f(x) = ∫ x−s B(s, t) ds = ∫ x ds, Re(t) > 0, c > 0.
2πi c−i∞ 2πi c−i∞ Γ(s + t)

Another representation of the beta function is:

∞
xs−1
B(s, t) = ∫ dx, Re(s) > 0, Re(t) > 0.
0 (1 + x)s+t

This implies that the beta function B(s, t − s) is the Mellin transform of the function
1
g(x) = .
(1 + x)t

The inversion formula then reads

1 1 c+i∞
1 1 c+i∞
= g(x) = ∫ x −s
B(s, t − s) ds = ∫ x−s Γ(s)Γ(t − s) ds, 0 < c < Re(t).
(1 + x)t 2πi c−i∞ 2πi Γ(t) c−i∞

These examples suggest that we might obtain a complex integral representation for the
hypergeometric function 2 F1 by finding its Mellin transform:
∞ ∞ Γ(c) 1
∫ xs−1 2 F1 ( ; −x) dx = ∫ xs−1 ∫ tb−1 (1 − t)c−b−1 (1 + xt)−a dt dx
a, b
0 c 0 Γ(b)Γ(c − b) 0

Γ(c) 1 ∞
xs−1
= ∫ t b−1
(1 − )
t c−b−1
∫ a dx dt.
Γ(b)Γ(c − b) 0 0 (1 + x t)

Here we used Euler's integral representation for the 2 F1 . The substitution xt = u now gives
∞
xs−1 ∞
us−1 Γ(s)Γ(a − s)
∫ dx = t −s
∫ du = t −s
B(s, a − s) = t −s
.
0 (1 + xt)a 0 (1 + u)a Γ(a)

Hence we have
∞ Γ(c) Γ(s)Γ(a − s) 1 b−s−1 Γ(c)Γ(s)Γ(a − s)
∫ 2 F1 ( ∫ t
s−1 a, b
x ; −x) dx = (1 − t)c−b−1 dt = B(b − s, c − b)
0 c Γ(b)Γ(c − b) Γ(a) 0 Γ(b)Γ(c − b)Γ(a)

Γ(c)Γ(s)Γ(a − s) Γ(b − s)Γ(c − b) Γ(c) Γ(s)Γ(a − s)Γ(b − s)

= = .
Γ(b)Γ(c − b)Γ(a) Γ(c − s) Γ(a)Γ(b) Γ(c − s)

Here we assumed that min (Re(a), Re(b)) > Re(s) > 0. Applying the inversion formula for the Mellin
transform we might expect that
Γ(a)Γ(b) 1 k+i∞ Γ(s)Γ(a − s)Γ(b − s)
2 F1 ( ∫
a, b
; x) = (−x)−s ds
Γ(c) c 2πi k−i∞ Γ(c − s)

for min (Re(a), Re(b)) > k > 0 and c ≠ 0, −1, −2, ….

In fact, this is Barnes' integral representation for the 2 F1 which is usually written as:

Γ(a)Γ(b) i∞ Γ(a + s)Γ(b + s)Γ(−s)

1
Theorem: 2 F1 ( ∫ | arg(−z)| < π.
a, b
; z) = (−z)s ds,
Γ(c) c 2πi −i∞ Γ(c + s)

The path of integration is curved, if necessary, to separate the poles s = −a − n and s = −b − n

from the poles s = n with n ∈ {0, 1, 2, …}. Such a contour always exists if a and b are not
negative integers.

Proof: Let C be the closed contour formed by a part of the curve used in the theorem
from −(N + 1/2)i to (N + 1/2)i together with the semicircle of radius N + 1/2 to the right of the
imaginary axis with 0 as center. We first show that the above integral defines an analytic
function for |arg(−z)| ≤ π − δ with δ > 0. By using Euler's reflection formula and Stirling's
asymptotic formula for the gamma function we find for the integrand:
Γ(a + s)Γ(b + s)Γ(−s) Γ(a + s)Γ(b + s) π π
(−z)s = − (−z)s ∼ −sa+b−c−1 (−z)s .
Γ(c + s) Γ(c + s)Γ(1 + s) sin πs sin πs

For s = it we have

π eit(ln |z|+i arg(−z))

−s a+b−c−1 s
(−z) = −(it) a+b−c−1
2πi = O(|t|a+b−c−1 e−|t|δ )
sin πs e−πt − eπt

for |arg(−z)| ≤ π − δ. This shows that the integral represents an analytic function in
|arg(−z)| ≤ π − δ for every δ > 0, which implies that it is analytic for |arg(−z)| < π.

On the semicircular part of the contour C the integrand is

a+b−c−1
(−z)s
O(N )
sin πs

for large N . For s = (N + 1/2)eiθ and |z| < 1 we have

(−z)s
= O [e(N+1/2)(cos θ ln |z|−sin θ arg(−z)−π| sin θ|) ].
sin πs

Since −π + δ ≤ arg(−z) ≤ π − δ, the last expression is

O [e(N+1/2)(cos θ ln |z|−δ| sin θ|) ].

Now we have cos θ ≥ 1

2 √2 for 0 ≤ |θ| ≤ π/4 and |sin θ| ≥ 1
2 √2 for π/4 ≤ |θ| ≤ π/2. Hence, since ln |z| < 0, the
integrand is

⎧O (N a+b−c−1 e 2 √2(N+1/2) ln |z| ), 0 ≤ |θ| ≤ π/4

1

⎨
⎩O (N a+b−c−1 e− 2 √2(N+1/2)δ ),
1
π/4 ≤ |θ| ≤ π/2.

This implies that the integral on the semicircle tends to zero for N → ∞. The residue
theorem then implies that the integral tends to the limit of the sums of the residues at s = n
with n ∈ {0, 1, 2, …}, id est
∞ Γ(a + n)Γ(b + n) n Γ(a)Γ(b) ∞ (a)n (b)n zn Γ(a)Γ(b)
∑ ∑ 2 F1 (
a, b
z = = ; z).
+
Γ(c n) n ! Γ(c) (c)n n! Γ(c) c
n=0 n=0

This proves Barnes' integral representation.

The Mellin transform has a convolution property, which can be obtained as follows.
Assume that
∞ ∞
F(s) = ∫ x s−1
f(x) dx and G(s) = ∫ xs−1 g(x) dx,
0 0

then we have
1 c+i∞ 1 c+i∞
f (x) = ∫ −s
F(s)x ds and g(x) = ∫ G(s)x−s ds.
2πi c−i∞ 2πi c−i∞

This implies that

∞ ∞ ∞
1 c+i∞
1 c+i∞
∫ xs−1 f (x)g(x) dx = ∫ xs−1 g(x) (∫ F(t)x−t dt) dx = ∫ F(t) (∫ xs−t−1 g(x) dx) dt
0 2πi 0 c−i∞ 2πi c−i∞ 0

1 c+i∞
= ∫ F(t)G(s − t) dt.
2πi c−i∞

As an application we consider

xb Γ(b + s)Γ(a − b − s)
f (x) = ⟷ F(s) =
(1 + x)a Γ(a)

and

xd Γ(d + s)Γ(c − d − s)
g(x) = ⟷ G(s) = ,
(1 + x)c Γ(c)

which leads to

1 k+i∞ Γ(b + s)Γ(a − b − s)Γ(d + 1 − s)Γ(c − d − 1 + s) 1 k+i∞ ∞

∫ ds = ∫ F(s)G(1 − s) ds = ∫ f(x)g(x) dx
2πi k−i∞ Γ(a)Γ(c) 2πi k−i∞ 0

∞
xb+d Γ(b + d + 1)Γ(a + c − b − d − 1)
=∫ dx = B(b + d + 1, a + c − b − d − 1) =
0 (1 + x)a+c Γ(a + c)

for a suitable k. By renaming the parameters, this can be written as

1 i∞ Γ(a + c)Γ(a + d)Γ(b + c)Γ(b + d)

∫ Γ(a + s)Γ(b + s)Γ(c − s)Γ(d − s) ds = ,
2πi −i∞ Γ(a + b + c + d)

which holds for Re(a) > 0, Re(b) > 0, Re(c) > 0 and Re(d) > 0, shortly for Re(a, b, c, d) > 0. This integral is
called the Mellin-Barnes integral. More general we have

Theorem: If the path of integration is curved to separate the poles of Γ(a + s)Γ(b + s) from the
poles of Γ(c − s)Γ(d − s), then we have

1 i∞ Γ(a + c)Γ(a + d)Γ(b + c)Γ(b + d)

∫ Γ(a + s)Γ(b + s)Γ(c − s)Γ(d − s) ds = .
2πi −i∞ Γ(a + b + c + d)

Proof: Again we use Euler's reflection formula to write the integrand as

Γ(a + s)Γ(b + s) π2
⋅ .
Γ(1 − c + s)Γ(1 − d + s) sin π(c − s) sin π(d − s)

Let C be the closed contour formed by a part of the curve used in the theorem from −iR to
iR together with the semicircle of radius R to the right of the imaginary axis with 0 as
center. By Stirling's asymptotic formula the integrand is

O (sa+b+c+d−2 e−2π|Im s| ) f or |s| → ∞ on C.

Since Im(s) can be arbitrarily small when |s| is large, we assume that Re(a + b + c + d) < 1 to
ensure that the integral on the semicircle tends to zero for R → ∞. Then, by Cauchy's
residue theorem (poles in s = c + n and s = d + n with n ∈ {0, 1, 2, …} with residues (−1)n /n!), the
integral equals
∞ (−1)n ∞ (−1)n
∑ Γ(a + c + n)Γ(b + c + n)Γ(d − c − n) + ∑ Γ(a + d + n)Γ(b + d + n)Γ(c − d − n) .
n! n!
n=0 n=0

Note that for n ∈ {0, 1, 2, …} we have

Γ(d − c) Γ(d − c) Γ(d − c)
Γ(d − c − n) = = = .
(d − c − 1)(d − c − 2) ⋯ (d − c − n) (−1) (1 + c − d)(2 + c − d) ⋯ (n + c − d)
n
(−1)n (1 + c − d)n

This implies that the integral equals

(a + c)n (b + c)n ∞ ∞ (a + d) (b + d)
Γ(a + c)Γ(b + c)Γ(d − c) ∑ + Γ(a + d)Γ(b + d)Γ(c − d) ∑
n n
.
(1 + c − d)n n! (1 + d − c)n n!
n=0 n=0

Finally we use Gauss's summation formula and Euler's reflection formula to find that the
integral equals
a + c, b + c a + d, b + d
Γ(a + c)Γ(b + c)Γ(d − c)2 F1 ( ; 1) + Γ(a + d)Γ(b + d)Γ(c − d)2 F1 ( ; 1)
1+c−d 1+d−c

Γ(1 + c − d)Γ(1 − a − b − c − d)
= Γ(a + c)Γ(b + c)Γ(d − c)
Γ(1 − a − d)Γ(1 − b − d)

Γ(1 + d − c)Γ(1 − a − b − c − d)
+ Γ(a + d)Γ(b + d)Γ(c − d)
Γ(1 − a − c)Γ(1 − b − c)

Γ(a + c)Γ(b + c)Γ(a + d)Γ(b + d) π sin π(a + d) sin π(b + d) π

=
Γ(a + b + c + d) sin π(d − c) π π sin π(a + b + c + d)

Γ(a + d)Γ(b + d)Γ(a + c)Γ(b + c) π sin π(a + c) sin π(b + c) π

+
Γ(a + b + c + d) sin π(c − d) π π sin π(a + b + c + d)

Γ(a + c)Γ(a + d)Γ(b + c)Γ(b + d) sin π(a + c) sin π(b + c) − sin π(a + d) sin π(b + d)
= .
Γ(a + b + c + d) sin π(c − d) sin π(a + b + c + d)

By using trigonometric relations it can be shown that

sin π(a + c) sin π(b + c) − sin π(a + d) sin π(b + d)
= 1.
sin π(c − d) sin π(a + b + c + d)

This proves the theorem for Re(a + b + c + d) < 1. This condition can be removed by using
analytic continuation of the parameters.

We also have a Barnes-type integral representation for the confluent hypergeometric

function. In order to find this representation we evaluate its Mellin transform. By using
Kummer's transformation formula we obtain
∞ ∞ ∞ (c − a) ∞ ∞ (c − a)
c−a
∫ 1 F 1 ( ; −z) dz = ∫ z e 1 F1 ( ; z) dz = ∑ ∫ e z dz = ∑
s−1 a s−1 −z n −z s+n−1 n
z Γ(s + n).
0 c 0 c (c)n n! 0 (c)n n!
n=0 n=0

Now we have Γ(s + n) = Γ(s) (s)n and by using Gauss's summation formula
∞ (c − a)n c − a, s Γ(c) Γ(s)Γ(a − s)
∑ Γ(s + n) = Γ(s) 2 F1 ( ; 1) = .
n=0
(c)n n! c Γ(a) Γ(c − s)

This leads to

Γ(a) i∞ Γ(a + s)
1
Theorem: ( ∫ |arg(−z)| < π/2.
a
F ; z) = Γ(−s)(−z)s ds,
Γ(c) 1 1 c 2πi −i∞ Γ(c + s)

Here the path of integration is curved, if necessary, to separate the poles s = −a − n from
the poles s = n with n ∈ {0, 1, 2, …}.

Proof: The proof is similar to the proof of Barnes' integral representation for the 2 F1
hypergeometric function. Application of Cauchy's residue theorem then gives that the
integral equals the sum of residues
∞ Γ(a + n) (−1)n Γ(a) ∞ (a)n zn Γ(a)
∑ ∑ 1 F 1 ( ; z).
n a
(−z) = =
Γ(c + n) n! Γ(c) (c)n n! Γ(c) c
n=0 n=0

Last modified on May 16, 2021

Author: Roelof Koekoek

Questions, suggestions or comments: Send me an email

© Roelof Koekoek