See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/380605710

Riemann-Liouville Integral and Mellin Transform

Preprint · May 2024

DOI: 10.13140/RG.2.2.10746.30401

CITATIONS READS

0 174

1 author:

Masaaki Takane

6 PUBLICATIONS 0 CITATIONS

SEE PROFILE

All content following this page was uploaded by Masaaki Takane on 05 May 2025.

The user has requested enhancement of the downloaded file.

 Riemann-Liouville Integral and Mellin
Transform
Masaaki Takane
May 6, 2025

Part I
Introduction
In this paper I would like to introduce the generalized multiple integral,
Riemann-Liouville integral, and see its application on special functions, such
as Gamma function and Zeta function, by using Mellin Transform.

Part II
Complex Analysis
In this part, we will review some basics of complex analysis, which would be
helpful for the calculation in the following sections.

1 Cauchy Integral Theorem

First of all, let z = x + iy be complex variable and f (z) = u + iv be the com-
plex function. In order to calculate well-defined derivatives of function f , its
value should be the same no matter which direction you used in calculation
for derivation. Thus,
∂f ∂f
= (1)
∂(x) ∂(iy)

by substitution,

i
 ∂(u + iv) ∂(u + iv)
= (2)
∂x ∂(iy)

comaparing real and imaginary part of the both sides,

∂u ∂v ∂v ∂u
= , =− (3)
∂x ∂y ∂x ∂y

we call this pair (3), Cauchy-Riemann equations. When function f satisfies

(3) on certain area A, we say f is holomorphic on A. Now consider complex
integral
I
f (z)dz (4)
C

where the integral is along C counter-clockwise. By substitution,

I I I I
f (z)dz = (u + iv)(dx + idy) = (udx − vdy) + i (vdx + udy) (5)
C C C C

By Stokes theorem:
I ZZ
F · dr = (∇ × F ) · dS, (6)
C S

(5) becomes
I ZZ ZZ
∂v ∂u ∂u ∂v
f (z)dz = − ( + )dxdy + i ( − )dxdy (7)
S ∂x ∂y S ∂x ∂y
C

finally when f is holomorphic on S, by (3)

I
f (z)dz = 0 (8)
C

which is Cauchy integral theorem. Intuitively, where f is holomorphic on

S, you can freely shrink or expand the contour of the integral as long as it
doesn’t violate any point where f is not holomorphic.

ii
2 Residue Theorem
In Cauchy-Riemann equations (3), we have to take in consider whether the
derivatives diverge or not. First we will take a single pole:
I
dz
(9)
z−a
C

let z − a = reiϕ , since dz = ireiϕ dϕ,

I Z 2π
dz
=i dϕ = 2πi (10)
z−a 0
C

For any pole of order n ̸= 1,

I Z 2π
dz i
= n−1 ei(n−1)ϕ dϕ = 0 (11)
(z − a)n r 0
C

Thus it can be concluded that you only have to consider coefficient of single
pole of the function f in calculation, and we call this coefficient, residue.

Part III
Riemann-Liouville Integral
In this part we will derive Riemann-Liouville integral and revisit special func-
tions with new interpretation. Definition of Riemann-Liouville integral is as
below.
Z t
s 1
a It f (x) = (t − z)s−1 f (z)dz (12)
Γ(s) a

Intuitively this equation means s-th order integral of function f (t) from a to
t.

3 Derivation from Cauchy Integral Formula

In this subsection, we will derive (12) from Cauchy-Integral formula. Since
f (z)−f (t)
z−t
doesn’t have any single pole, by residue theorem

iii
 f (z) − f (t)
I
dz = 0 (13)
z−t
C

thus,
I I
f (z) f (t)
dz = dz = 2πif (t) (14)
z−a z−t
C C

and we call this Cauchy-Integral formula. Now when we take derivative of

(14),
I
′ 1! f (z)
f (t) = dz (15)
2πi (z − t)2
C

Similarly,

I
′′ 2! f (z)
f (t) = dz (16)
2πi (z − t)3
C
I
3! f (z)
f ′′′ (t) = dz (17)
2πi (z − t)4
C

More generally,
I
(n) n! f (z)
f (t) = dz (18)
2πi (z − t)n+1
C

Extending from integer n to complex number s by using Gamma function

Γ(s) = (s − 1)!,
I
s Γ(s + 1) f (z)
a Dt f (x) = dz (19)
2πi (z − t)s+1
C

where loop C include points a and t. Splitting loop into branches,

iv
 I
s Γ(s + 1) f (z)
a Dt f (x) = dz (20)
2πi (z − t)s+1
C
Γ(s + 1) t πi(s+1)
Z
f (z)
= (e − e−πi(s+1) ) dz (21)
2πi a (t − z)s+1
Γ(s + 1) sin πs t f (z)
Z
=− s+1
dz (22)
π a (t − z)
Z t
1 f (z)
= dz (23)
Γ(−s) a (t − z)s+1

at the last equation we used well known formula

π
Γ(s)Γ(1 − s) = (24)
sin πs

Changing s → −s, we obtain

Z t
1
s
a It f (x) =a Dt−s f (x) = (t − z)s−1 f (z)dz (25)
Γ(s) a

Which we call Riemann-Liouville integral.

4 Revisiting Special Functions

For convenience we will start by simple transform,

z →t−z (26)

then Riemann-Liouville integral (25) becomes

Z t−a
s 1
a It f (x) = z s−1 f (t − z)dz (27)
Γ(s) 0

4.1 Gamma Function

Definition of Gamma function is
Z ∞
Γ(s) = z s−1 e−z dz (28)
0

v
Gamma function is known as the extension of factorial n! as they correspond
each other when s is an integer.

Γ(n) = (n − 1)! (29)

Gamma function has its single pole where s = −m = 0, −1, −2, −3...
In fact by the property Γ(z + 1) = zΓ(z),

Γ(z + m + 1) (−1)m
lim (z + m)Γ(z) = lim (z + m) · = (30)
z→−m z→−m (z + m)(z + m − 1)...z m!

Dividing (28) by Γ(s),

Z ∞ Z 0−(−∞)
1 s−1 −z 1
1= z e dz = z s−1 e0−z dz =−∞ I0s (ex ) (31)
Γ(s) 0 Γ(s) 0

Thus we can conclude that s-th integral of exponential function ez from −∞

to 0 is 1 for any complex number s.

4.2 Zeta Function

Integral form of well known Zeta function ζ(s) = ∞ 1
P
n=1 ns
is
Z ∞ s−1
1 z
ζ(s) = dz (32)
Γ(s) 0 ez − 1

Zeta function has its single pole at s = 1. In fact for n > 1, s > 1,
Z n+1 Z n
dx 1 dx
< < (33)
n xs ns n−1 x
s

thus by summation,
Z ∞ Z ∞
dx dx
< ζ(s) < 1 + (34)
1 xs 1 xs

thus by calcualtion,

1 < (s − 1)ζ(s) < s (35)

vi
as a result,

lim(s − 1)ζ(s) = 1 (36)

s→1

Let us define a function,

∞
X
∆(x) ≡ enx (37)
n=1

Now consider

s ex
−∞ I0 ∆(x) =−∞ I0s ( ) (38)
− ex
Z1 ∞
1 e−z
= z s−1 ( )dz (39)
Γ(s) 0 1 − e−z
Z ∞ s−1
1 z
= dz (40)
Γ(s) 0 ez − 1

thus we can conclude that Zeta function ζ(s) is s-th order integral of function
∆(x) from −∞ to 0.

Part IV
Mellin/Inverse-Mellin
Transform
In this part we will derive Mellin/inverse Mellin transform:

Z ∞
g(s) = xs−1 f (x)dx (41)
0
Z k+i∞
1 g(s)
f (x) = dx (42)
2πi k−i∞ xs

In (42), we take k so that all the single poles of the function inside are on
the left side of the path of integral. We will also discover some features of
special functions by Laurent expansion through (41) and (42).

vii
5 Derivation from Fourier Integral Formula
First we start from Fourier integral formula:
Z ∞Z ∞
1
f (x) = f (y) exp iu(x − y)dudy (43)
2π −∞ −∞

If we define a function:
Z ∞
1
g(u) ≡ √ f (y) exp(−iuy)dy (44)
2π −∞

Then we obtain
Z ∞
1
f (x) = √ g(u) exp iuxdu (45)
2π −∞

Here we got the well known Fourier transform (44) and inverse Fourier trans-
form (45). Now back to (43), substitute f (x) = e−kx F (e−x ) then we get
Z ∞Z ∞
−kx −x 1
e F (e ) = eiu(x−y) e−ky F (e−y )dudy (46)
2π −∞ −∞

More clearly,
Z ∞ Z ∞
−x 1
F (e ) = e(k+iu)(x−y) F (e−y )dudy (47)
2π −∞ −∞

By transform X = e−x , Y = e−y , s = k + iu,

Z k+i∞ Z 0
1 Y s dY ds
F (X) = ( ) F (Y )(− )( ) (48)
2π k−i∞ ∞ X Y i
Z k+i∞ Z ∞
1
= Y s−1 X −s F (Y )dY ds (49)
2πi k−i∞ 0

Now let us define:

Z ∞
G(s) ≡ Y s−1 F (Y )dY (50)
0

then

viii
 Z k+i∞
1
F (X) = X −s G(s)ds (51)
2πi k−i∞

Here we got Mellin transform (50) and inverse Mellin transform (51).

6 Laurent Expansion and Examples

In this section we will calculate Laurent expansion via Mellin/inverse-Mellin
transform and rediscover some features of zeta function.

6.1 Gamma Function

Let’s get back to Gamma function
Z ∞
Γ(s) = z s−1 e−z dz (52)
0

Since we can say that this is a Mellin transform of exponential function e−z ,
by inverse-Mellin transform,
Z k+i∞
−z 1 Γ(s)
e = ds (53)
2πi k−i∞ z s

Since Γ(s) has residues on single poles at s = −m = 0, −1, −2, ... where,

(−1)m
Res(Γ(s), −m) = (54)
m!

(53) is
∞
X 1 (−1)m
e−z = (55)
m=0
z −m m!

changing z → −z, we obtain the well known Maclaurin expansion of ez

∞
X zm
ez = (56)
m=0
m!

ix
6.2 Zeta Function
Multiplying ζ(s) by Γ(s),
∞
z s−1
Z
Γ(s)ζ(s) = dz (57)
0 ez − 1

By inverse-Mellin transform,
Z k+i∞
1 1 Γ(s)ζ(s)
z
= ds (58)
e −1 2πi k−i∞ zs

Since ζ(s) has a residue on single pole at s = 1 where,

Res(ζ(s), 1) = 1 (59)

(58) is
∞
1 1 X (−1)m ζ(−m)
= + (60)
ez − 1 z m=0 m! z −m

By transform z → −z, we obtain function:

∞
1 ez X zm
f (z) = − z = ζ(−m) · (61)
z e − 1 m=0 m!

with this function, we can calculate specific values of zeta function such as

1 ez 1
ζ(0) = lim f (z) = lim( − z )=− (62)
z→0 z→0 z e −1 2
z
1 e 1
ζ(−1) = lim f ′ (z) = lim(− 2 + z 2
)=− (63)
z→0 z→0 z (e − 1) 12

Also we can prove that ζ(s) is zero for any negative even numbers. Let us
define

∞
X zm 1 1 ez
g(z) = ζ(−m) · = + − z (64)
m=1
m! 2 z e −1
ez (2 − z) − (2 + z)
= (65)
ez (2z) − (2z)

x
changing z → −z,

e−z (2 + z) − (2 − z) ez (2 − z) − (2 + z)
g(−z) = = − = −g(z) (66)
e−z (−2z) − (−2z) ez (2z) − (2z)

since function g(z) is an odd function, its coefficients; the value of zeta func-
tion is zero at negative even number.

6.3 Xi Function
Now we consider Xi function
Z ∞
− 2s s 1 s
ξ(s) = π Γ( )ζ(s) = x 2 −1 (ϑ(x) − 1)dx (67)
2 2 0

where
2x
X
ϑ(x) = e−πn (68)
n∈Z

substitute x = X 2 ,
Z ∞
ξ(s) = X s−1 (ϑ(X 2 ) − 1)dX (69)
0

by inverse Mellin transform,

Z k+i∞
2 1 ξ(s)
ϑ(X ) − 1 = ds (70)
2πi k−i∞ Xs
s
π − 2 Γ( 2s )ζ(s)
Z k+i∞
1
= ds (71)
2πi k−i∞ Xs

since ζ(s) has pole at s = 1 and Γ( 2s ) has poles at s = −2m(m = 0, 1, 2, ...),

1 ∞
2 π − 2 Γ( 12 ) X π m ζ(−2m) (−1)m
ϑ(X ) − 1 = + (72)
X m=0
X −2m m!
1 1
= − (73)
X 2

xi
 thus
1 1
ϑ(X 2 ) = + (74)
X 2

this contradicts with the fact that ϑ(X 2 ) is an even function.

References
[1] Complex Analysis.
Lars Ahlfors.

[2] Fractional Derivatives and Special Functions.

J. L. Lavoie, T. J. Osler, R. Tremblay.

[3] Math Exercises

Kenichi Goto

xii

View publication stats