INTRODUCTION

Robert Hjalmar Mellin was a Finnish mathematician who studied under Karl Weier-
strass. He is accredited as the developer of the integral transform

∞
ℳ [𝑓 (𝑥 ); 𝑠] = 𝑓 ∗ (𝑠) = ∫ 𝑓(𝑥 )𝑥 𝑠−1 𝑑𝑥
0

known as the Mellin transform.

We will first look at some examples and basic properties of the Mellin transform.
Then we proceed to look at the correspondence between the asymptomatic expansion
of a function and singularities of the transformed function.

Finally we use the Mellin transform in asymptotic analysis for asymptotically

harmonic sums.

1
BASIC PROPERTIES OF THE MELLIN TRANSFORM

Definition
Let f(x) be absolutely integrable on any interval 0 < x < a

𝑎
i.e. ∫0 |𝑓(𝑥)|𝑑𝑥 < ∞

The largest open strip (α, β) in which the integral converges is called the fundamental
strip, where (α, β) is the open strip of complex numbers s = σ + it such that α < σ < β.

ℑ(𝑠)

ℜ(𝑠)
α β

Before we can say more about this fundamental strip, we turn our attention to the
Laplace transform. Suppose for finite a that
𝑎
∫ |𝑓(𝑥)|𝑑𝑥 < ∞
0

and 𝑔(𝑡) = 𝑂(𝑒 α𝑡 ) as t → ∞ for some real constant α. Then the one-sided Laplace
transform
∞
ℒ+ [𝑔(𝑡); 𝑠] = ∫ 𝑔(𝑡)𝑒 −𝑠𝑡 𝑑𝑡
0

converges absolutely and is analytic in the right-half plane R(s) > α. Similarly suppose
𝑎
∫ |𝑔(−𝑡)|𝑑𝑡 < ∞
0

and 𝑔(𝑡) = 𝑂(𝑒 β𝑡 ) as t → ∞ for some real constant β.

Then the one-sided Laplace transform

2
 0
ℒ 𝑔(𝑡); 𝑠] = ∫ 𝑔(𝑡)𝑒 −𝑠𝑡 𝑑𝑡
−[
−∞

∞
= ∫ 𝑔(−𝑡)𝑒 𝑠𝑡 𝑑𝑡
0

converges absolutely and is analytic in the left half plane Re(s) < β.

Under the same condition on g(t) and if β > α the two-sided Laplace transform
∞
ℒ [𝑔(𝑡); 𝑠] = ∫ 𝑔(𝑡)𝑒 −𝑠𝑡 𝑑𝑡
−∞

converges absolutely and is analytic in the vertical strip α < R(s) < β.

Now if we let 𝑡 = − log 𝑥 and g( − log 𝑥) = 𝑓(𝑥) then 𝑒 −𝑠𝑡 = 𝑒 𝑠 log 𝑥 = 𝑥 𝑠 .

Hence
∞
ℒ [𝑔(𝑡); 𝑠] = ∫ 𝑔(𝑡)𝑒 −𝑠𝑡 𝑑𝑡
−∞

0
= − ∫ 𝑥 𝑠−1 𝑓(𝑥)𝑑𝑥
∞

= ℳ[𝑓 (𝑥); 𝑠]

So the Mellin transform of f (x) is the two-sided Laplace transform of g(t) where 𝑡 =
− log 𝑥 and it converges absolutely and analytic in the vertical strip α < R(s) < β.

Now since 𝑔(𝑡) = 𝑂(𝑒 ∝𝑡 ) as t → ∞ we obtain 𝑓(𝑥) = 𝑂(𝑥 −∝ ) as x →0+.

Also 𝑔(𝑡) = 𝑂(𝑒 𝛽𝑡 ) as t → -∞ we obtain 𝑓(𝑥) = 𝑂(𝑥 −𝛽 ) as x → ∞.

Summing up we have proved the following lemma:

Lemma

The conditions 𝑓 (𝑥 ) = 𝑂(𝑥 −𝛼 ) as x → 0+ and 𝑓 (𝑥 ) = 𝑂(𝑥 −𝛽 ) as x → ∞ where α < β

guarantee that 𝑓 ∗ (𝑠) exists in the strip (α, β).

Hence monomials 𝑥 𝑐 , including constant do not have Mellin transforms.

Example

3
The function 𝑓 (𝑥 ) = 𝑒 −𝑥 satisfies 𝑒 −𝑥 = 𝑂(𝑥 0 ) as x → 0+ and 𝑒 −𝑠 = 𝑂(𝑥 −𝑏 ) as
x → ∞ for any b > 0 so that its transform (the gamma function)
∞
𝑓 ∗ (𝑠) = ∫ 𝑒 −𝑥 𝑥 𝑠−1 𝑑𝑥 = 𝛤(𝑠)
0

is defined and analytic on (0, ∞).

Example

The function 𝑓 (𝑥 ) = (𝑒 𝑥 − 1)−1 satisfies 𝑓 (𝑥 ) = 𝑂(𝑥 −1 ) as x → 0+ and 𝑓 (𝑥 ) =

𝑂(𝑥 −𝑏 ) for all b > 0 as x → ∞. Hence 𝑓(𝑥) is analytic and defined on (1, ∞). We find
∞
1
𝑓 ∗ (𝑠 ) = ∫ 𝑥 𝑠−1 𝑑𝑥
0 1 − 𝑒 −𝑥
∞
𝑒 −𝑥
=∫ 𝑥 𝑠−1 𝑑𝑥
0 1 − 𝑒 −𝑥

∞ ∞
= ∫ ∑ 𝑒 −𝑛𝑥 𝑥 𝑠−1 𝑑𝑥
0 𝑛=1

∞ ∞
= ∑ ∫ 𝑒 −𝑛𝑥 𝑥 𝑠−1 𝑑𝑥
𝑛=1 0

∞
𝛤(𝑠)
=∑
𝑛𝑠
𝑛=1

= 𝛤 (𝑠)𝜁(𝑠)

We require that R(s) > 1 for convergence of the Riemann-zeta function and we see
that this validates the strip (1, ∞) on which 𝑓 ∗ (𝑠) is defined and analytic.

Example

The function 𝑓 (𝑥 ) = (1 + 𝑥)−1 is 𝑂(𝑥 0 ) as x → 0+ and 𝑂(𝑥 −1 ) as x → ∞. Hence a

𝑡
guaranteed strip of existence for 𝑓 ∗ (𝑠) is (0, 1). Set 𝑥 = . Then
1−𝑡

1
∗(
𝑡 𝑠−1 1
𝑓 𝑠) = ∫ ( ) (1 − 𝑡)−2 𝑑𝑡
1 − 𝑡 1
0 1+1−𝑡

1
𝑡 𝑠−1
=∫ ( ) (1 − 𝑡)−1 𝑑𝑡
0 1 − 𝑡
1
= ∫ 𝑡 𝑠−1 (1 − 𝑡)−𝑠 𝑑𝑡
0

4
 = 𝛽(𝑠, 1 − 𝑠)

= 𝛤 (𝑠)𝛤(1 − 𝑠).

We can also evaluate the Mellin transform of 𝑓(𝑥) using complex analysis. Consider
𝑧 𝑠−1
the branch of the function defined on the slit plane C\[0, ∞) by
𝑧+1

𝑟 𝑠−1 𝑒 𝑖(𝑠−1)𝜃
𝑓 (𝑧 ) = ; 𝑧 = 𝑟𝑒 𝑖𝜃 , 0 < 𝜃 < 2𝜋
𝑧+1
The values on the bottom edge are obtained from those on the top edge by multiplying
by the phase factor 𝑒 2𝜋𝑖(𝑠−1) .

For ∈ > 0 small and R > 0 large, we consider the keyhole domain D consisting of z in
the slit plane C\ [0, ∞) satisfying ∈ < |𝑧| < 𝑅. 𝑓(𝑧) has a pole in D, a simple pole at
𝑧 = −1, with residue

𝑧 𝑠−1 𝑧 𝑠−1
𝑅𝑒𝑠 | , −1| = lim (𝑧 + 1)
𝑧+1 𝑧→−1 𝑧+1

𝑧𝑠
= lim
𝑧→−1 𝑧

= −𝑒 𝜋𝑖𝑠

Hence the residue theorem yields

∫ 𝑓 (𝑧)𝑑𝑧 = −2𝜋𝑖𝑒 𝜋𝑖𝑠

𝜕𝐷

The integral around δD breaks into the sum of 4 integrals.

𝑅 𝜖 2𝜋𝑖(𝑠−1) 𝑠−1
𝜋𝑖𝑠
𝑥 𝑠−1 𝑧 𝑠−1 𝑒 𝑥 𝑧 𝑠−1
⇒ −2𝜋𝑖𝑒 =∫ 𝑑𝑥 + ∫ 𝑑𝑧 + ∫ 𝑑𝑥 + ∫ 𝑑𝑧
𝜖 1+𝑥 𝛤𝑅 𝑧 + 1 𝑅 1+𝑥 𝛤𝜖 𝑧 + 1

For the integrals over 𝛤𝑅 and 𝛤∈ we have

𝑧 𝑠−1 𝑅 𝑠−1
|∫ 𝑑𝑧| ≤ 2𝜋𝑅 = 𝑂(𝑅 𝑠−1 )
𝛤𝑅 𝑧+1 𝑅−1

𝑧 𝑠−1 ∈𝑠−1
|∫ 𝑑𝑧| ≤ 2𝜋 ∈= 𝑂(∈𝑠 )
𝛤∈ 𝑧 + 1 ∈ −1

Since 0 < s < 1 both these integrals vanish as R → ∞ and ∈ → 0.

∞
𝜋𝑖𝑠 2𝜋𝑖(𝑠−1) ) ∫
𝑥 𝑠−1
⇒ −2𝜋𝑖𝑒 = (1 − 𝑒 𝑑𝑥
0 1+𝑥

5
 ∞
𝑥 𝑠−1 −2𝜋𝑖𝑒 𝜋𝑖𝑠
⇒∫ 𝑑𝑥 =
0 1+𝑥 1 − 𝑒 2𝜋𝑖(𝑠−1)

2𝜋𝑖
=
𝑒 𝜋𝑖𝑠−2𝜋𝑖 − 𝑒 −𝜋𝑖𝑠
𝜋
=
sin(𝜋𝑠)
𝜋
Hence 𝛤 (𝑠)𝛤 (1 − 𝑠) = .
sin(𝜋𝑠)

We proceed to look at some functional properties of the Mellin transform.

Theorem

Let 𝑓(𝑥) be a function whose transform admits the fundamental strip (α, β), Let ρ, μ,
and v be positive real numbers. Then the following relation hold:

(𝑖 ) 𝑀[𝑓(𝜇𝑥 ); 𝑠] = 𝜇−𝑠 𝑓 ∗ (𝑠) 𝑠 𝜖 (𝛼, 𝛽)

(𝑖𝑖 ) 𝑀 |∑ 𝜆𝑘 𝑓 (𝜇𝑘 𝑥 ); 𝑠| = (∑ 𝜆𝑘 𝜇𝑘−𝑠 ) 𝑓 ∗ (𝑠) , 𝜇𝑘 > 0

𝑘∈𝐴 𝑘∈𝐴

(𝑖𝑖𝑖 ) 𝑀[𝑥 𝑣 𝑓(𝑥0; 𝑠] = 𝑓 ∗ (𝑠 + 𝑣) 𝑠 ∈ 〈𝛼 − 𝑣, 𝛽 − 𝑣〉

1 ∗ 𝑠
(𝑖𝑣) 𝑀[𝑓(𝑥 𝜌 ); 𝑠] = 𝑓 ( ) 𝑠 ∈ 〈𝜌𝛼, 𝜌𝛽 〉
𝜌 𝜌

Proof:

(i) 𝑀 [𝑓(𝜇𝑥 ); 𝑠]
∞
= ∫ 𝑥 𝑠−1 𝑓 (𝜇𝑥 )𝑑𝑥
0

∞
1 𝑠−1 𝑑𝑡
= ∫ 𝑓(𝑡) ( )
0 𝜇 𝜇

= 𝜇 −𝑠 𝑓 ∗ (𝑠)

(ii) Since the Mellin transform is linear this follows immediately from (i).

(iii) 𝑀[𝑥 𝑣 𝑓(𝑥0; 𝑠]

6
 ∞
= ∫ 𝑓(𝑥)𝑥 𝑠+𝑣−1 𝑑𝑥
0

= 𝑓 ∗ (𝑠 + 𝑣).

(iv) 𝑀 [ 𝑓 (𝑥 𝜌 ); 𝑠 ]
∞
= ∫ 𝑓(𝑥 𝜌 )𝑥 𝑠−1 𝑑𝑥
0

∞ 𝑠
−1 𝑑𝑡
= ∫ 𝑓(𝑡)𝑡 𝜌
0 𝜌
1 𝑠
= 𝜌 𝑓 ∗ (𝜌).

2 1 𝑠
We have for example from Theorem 2.1(iv) 𝑀[𝑒 −𝑥 ; 𝑠] = 2 𝛤(2) on 〈0, ∞〉 with
𝑓 (𝑥 ) = 𝑒 −𝑥 and 𝜌 = 2. Wanting to expand our range of Mellin transforms we find by
differentiation under the integral sign:

𝑑 ∗
𝑓 (𝑠) = 𝑀[𝑓(𝑥 ) log 𝑥; 𝑠]
𝑑𝑠
For instance the transform of 𝑒 −𝑥 log 𝑥 is Γ’(s). If we want to transform the derivative
of function integration by part yields:
∞ ∞
∫ 𝑓 ′ (𝑥 )𝑥 𝑠−1 𝑑𝑥 = |𝑓(𝑥)𝑥 𝑠−1 |∞
0 − (𝑠 − 1) ∫ 𝑓(𝑥)𝑥
𝑠−2
𝑑𝑥
0 0

The term |𝑓(𝑥)𝑥 𝑠−1 |∞ ∗

0 equals 0 since we assume the reason why 𝑓 (𝑠) exists for 𝑠 ∈
〈𝛼, 𝛽 〉 is because lim 𝑥 𝜌 𝑓(𝑥 ) = 0 for ℜ(𝑠) > 𝛼 and lim 𝑥 𝜌 𝑓(𝑥 ) = 0 for ℜ(𝑠) > 𝛽.
𝑥→0 𝑥→∞
Thus 𝑀[𝑓 ′ (𝑥 ); 𝑠] = −(𝑠 − 1)𝑓 ∗ (𝑠 − 1). For instance from Example 2.3 we can
derive the Mellin transform of 𝑓 (𝑥 ) = log(1 + 𝑥)
𝜋
= 𝑀[𝑓 ′ (𝑥 ); 𝑠] = −(𝑠 − 1)𝑀[log(1 + 𝑥 ) ; 𝑠 − 1]
sin(𝜋𝑠)
𝜋
⟹ 𝑀 [ 𝑓 (𝑥 ); 𝑠 − 1] = −
(𝑠 − 1) sin(𝜋𝑠)
𝜋 𝜋
⟹ 𝑀 [ 𝑓 (𝑥 ); 𝑠 ] = − =
𝑠 sin(𝜋(𝑠 + 1)) 𝑠 sin(𝜋𝑠)

for 𝑠 ∈ 〈−1, 0〉.

The Inversion Formula

7
The inversion formula for the two-sided Laplace transform of g(t) is given by 𝑔(𝑡) =
1 𝑐+𝑖∞
∫ ℒ[𝑔(𝑡); 𝑠]𝑒 𝑠𝑡 𝑑𝑠 for any c inside the region of convergence, provided that g
2𝜋 𝑐−𝑖∞
is continuous at t. Setting 𝑡 = −𝑙𝑜𝑔𝑥 and 𝑓(𝑥 ) = 𝑔(− log 𝑥) we obtain the inversion
formula for the Mellin transform:

1 𝑐+𝑖∞
𝑓(𝑥 ) = 𝑔(−𝑙𝑜𝑔 𝑥 ) = ∫ 𝑀[𝑓(𝑥 ); 𝑠]𝑥 −𝑠 𝑑𝑠
2𝜋𝑖 𝑐−𝑖∞

We will prove the inversion formula for the Laplace transform, from which the
inversion formula of the Mellin transform follows. Before we can do this we need to
prove the following lemma (Riemann Lebesgue):

Lemma 2.2
𝑏
𝐼 (𝜔) = ∫𝑎 𝑓(𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡 → 0 as 𝜔 → ∞ if 𝑓(𝑡) is continuous on [a,b].

Proof:
𝜋
𝑎+ 𝑏
𝜔
𝐼 (𝜔 ) = ∫ 𝑓(𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡 + ∫ 𝑓(𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡
𝜋
𝑎 𝑎+
𝜔

𝜋 𝜋
𝑎+ 𝑏−
𝜔
𝑖𝜔𝑡
𝜔 𝜋
=∫ 𝑓 (𝑡 ) 𝑒 𝑑𝑡 − ∫ 𝑓(𝑡 + )𝑒 𝑖𝜔𝑡 𝑑𝑡
𝑎 𝑎 𝜔

and
𝜋
𝑏− 𝑏
𝜔
𝐼 (𝜔 ) = ∫ 𝑓(𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡 + ∫ 𝑓(𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡
𝜋
𝑎 𝑏−
𝜔

𝜋 𝜋
1 𝑎+𝜔 𝑖𝜔𝑡
𝑏
𝑖𝜔𝑡
1 𝑏−𝜔 𝜋
( )
⟹𝐼 𝜔 = ∫ ( )
𝑓 𝑡 𝑒 𝑑𝑡 + ∫ 𝑓(𝑡)𝑒 𝑑𝑡 + ∫ (𝑓(𝑡) − 𝑓 (𝑡 + ))𝑒 𝑖𝜔𝑡 𝑑𝑡
2 𝑎 𝑏−
𝜋 2 𝑎 𝜔
𝜔

In the first two integral 𝑓(𝑡)𝑒 𝑖𝜔𝑡 is bounded, since it is a continuous function on a
closed interval. Since the length of these integrals goes to 0 as 𝜔 → ∞ they both
𝜋
vanish. Similarly for the last integral we fine (𝑓(𝑡) − 𝑓 (𝑡 + 𝜔 )) → 0 as 𝜔 → ∞ and
since the length of this integral is bounded it also vanishes so that lim 𝐼(𝜔) = 0.
𝜔→∞

Note that the proof also hold for an infinite interval if

∞
∫ |𝑓(𝑡)|𝑑𝑡 < ∞:
0

Choose T large enough so that for all 𝜖 > 0

8
 ∞
∫ |𝑓(𝑡)| 𝑑𝑡 ≤ 𝜖
𝑇

∞ 𝑇 ∞
⟹ ∫ 𝑓 (𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡 = ∫ 𝑓(𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡 + ∫ 𝑓(𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡
0 0 𝑇

∞
⟹ lim ∫ 𝑓(𝑡)𝑒 𝑖𝜔𝑡 𝑑𝑡 = 0
𝜔→∞ 0

since 𝜖 is arbitrary.

Dirichlet Conditions and Dirichlet Integrals

𝑓(𝑡) satisfies the Dirichlet conditions if only has a finite number of discontinuities
with finite jumps and a finite number of minima/maxima. (i.e. f is piecewise
monotone)

The second mean value theorem for integral states:

If 𝑓(𝑡) is monotone [a, b], then any continuous 𝑔(𝑡)

𝑏 𝑐 𝑏
∫ 𝑓 (𝑡)𝑔(𝑡)𝑑𝑡 = 𝑓(𝑎) ∫ 𝑔(𝑡)𝑑𝑡 + 𝑓(𝑏) ∫ 𝑔(𝑡)𝑑𝑡
𝑎 𝑎 𝑐

for same c 𝜖 [a, b].

Assume 𝑓(𝑡) satisfies the Dirichlet conditions. We should like to determine

𝑏 𝑏 𝑖𝜔𝑡
sin 𝜔𝑡 𝑒
lim ∫ 𝑑𝑡 = lim 𝐼𝑚 ∫ 𝑑𝑡
𝜔→∞ 0 𝑡 𝜔→∞ 0 𝑡

Choose c such that 𝑓(𝑡) is continuous and monotone on [0, c].

𝑏 𝑐 𝑏
sin 𝜔𝑡 sin 𝜔𝑡 sin 𝜔𝑡
∫ 𝑓 (𝑡 ) 𝑑𝑡 = ∫ 𝑓(𝑡) 𝑑𝑡 + ∫ 𝑓(𝑡) 𝑑𝑡
0 𝑡 0 𝑡 𝑐 𝑡
𝑏
sin 𝜔𝑡
lim ∫ 𝑑𝑡 = 0
𝜔→∞ 𝑐 𝑡

by Riemann-Lebesgue lemma.

Apply the mean value theorem:

𝑐 ℎ 𝑐
sin 𝜔𝑡 sin 𝜔𝑡 sin 𝜔𝑡
∫ 𝑓 (𝑡 ) 𝑑𝑡 = 𝑓(0+ ) ∫ 𝑑𝑡 + 𝑓(𝑐) ∫ 𝑑𝑡
0 𝑡 0 𝑡 ℎ 𝑡
𝑐 𝑐
sin 𝜔𝑡
+ +
sin 𝜔𝑡
= 𝑓(0 ) ∫ 𝑑𝑡 + (𝑓(𝑐 ) − 𝑓(0 )) ∫ 𝑑𝑡
0 𝑡 ℎ 𝑡

9
 𝑐 𝜔𝑐
sin 𝜔𝑡 𝜔 sin 𝑥 𝑑𝑥
∫ 𝑓 (𝑡 ) 𝑑𝑡 = ∫
0 𝑡 0 𝑥 𝜔
𝜔𝑐
sin 𝑥
=∫ 𝑑𝑥
0 𝑥
𝜔𝑐
sin 𝑥
= lim ∫ 𝑑𝑥
𝜔→∞ 0 𝑥
∞ sin 𝑥
∫0 𝑑𝑥 can not be evaluated by elementary analytic methods. To evaluate it we
𝑥
use complex analysis and the following lemma (Jordan’s lemma):

Lemma 2.3

If 𝛤𝑅 is the semicircular contour 𝑧(𝜃) = 𝑅𝑒 𝑖𝜃 where 0 ≤ 𝜃 ≤ 𝜋, in the upper half-

plane, then ∫𝛤 |𝑒 𝑖𝑧 ||𝑑𝑧| < 𝜋.
𝑅

𝑅 sin 𝑥 𝜋
Now we show that lim ∫0 𝑑𝑥 = 2 .
𝑅→∞ 𝑥

𝑅 𝑆𝑖𝑛 𝑥
Since the integrand is an even function it is equivalent to lim ∫−𝑅 𝑑𝑥 = 𝜋
𝑅→∞ 𝑥

To rvaluate the limit we consider 𝑓(𝑧) = 𝑒 𝑖𝑧 which only gas one pole, a simple pole
at 𝑧 = 0.

𝑒 𝑖𝑧
𝑅𝑒𝑠 | , 0| = 1
𝑧

Consider Figure 3 in the appendix:

The integrand around 𝜕𝐷 breaks into 4:

−𝜖 𝑅
∫ 𝑓 (𝑧)𝑑𝑧 = (∫ + ∫ + ∫ + ∫ ) 𝑓 (𝑧)𝑑𝑧
𝜕𝐷 −𝑅 𝐶𝜖 𝜖 𝛤𝑅

Since 𝑓 is analysis on D, Cauchy’s theorem gives:

0 = ∫ 𝑓(𝑧)𝑑𝑧 (∗)
𝜕𝐷

The integrand 𝐶𝜖 is handled by the fractional residue theorem with angle – 𝜋:

𝑒 𝑖𝑧 𝑒 𝑖𝑧
lim ∫ 𝑑𝑧 = −𝜋𝑖𝑅𝑒𝑠 | , 0| = −𝜋𝑖
𝑧→0 𝐶
𝜖
𝑧 𝑧

Taking the limit as 𝜖 → 0 in (∗) we obtain:

10
 𝑅
𝑒 𝑖𝑥 𝑒 𝑖𝑧
0=∫ 𝑑𝑥 − 𝜋𝑖 + ∫ 𝑑𝑧
−𝑅 𝑥 𝛤𝑅 𝑧

Taking imaginary parts gives:

𝑅
𝑆𝑖𝑛 𝑥 𝑒 𝑖𝑧
𝜋=∫ 𝑑𝑥 + 𝐼𝑚 ∫ 𝑑𝑧
−𝑅 𝑥 𝛤𝑅 𝑧

Now the integral over 𝛤𝑅 us handled by Jordon’s lemma:

𝑒 𝑖𝑧 𝑒 𝑖𝑧 𝜋
|∫ 𝑑𝑧| ≤ ∫ | | |𝑑𝑧| < → 0
𝛤𝑅 𝑧 𝛤𝑅 𝑧 𝑅

𝑅 sin 𝑥
As 𝑅 → ∞. Thus we have lim ∫−𝑅 𝑑𝑥 = 𝜋. Hence
𝑅→∞ 𝑥

𝑅
sin 𝑥 𝜋
lim ∫ 𝑑𝑥 =
𝑅→∞ 0 𝑥 2

Returning to where we were just before Jordan’s lemma we have for the other part:
𝑐
+)
sin 𝜔𝑡
(𝑓 (𝑐 ) − 𝑓 (0 ) ∫ 𝑑𝑡 → 0
ℎ 𝑡

if 𝜔 → ∞, since (𝑓(𝑐) − 𝑓 (0+ )) can be made arbitrarily close to 0 by the choice of c

𝑐 sin 𝜔𝑡 𝑏 sin 𝜔𝑡 𝜋
and since ∫ℎ is bounded. It follows that lim ∫0 𝑓 (𝑡) 𝑑𝑡 = 2 𝑓(0+ ).
𝑡 𝜔→∞ 𝑡

0 sin 𝜔𝑡 𝜋
Likewise lim ∫−𝑏 𝑓 (𝑡) 𝑑𝑡 = 2 𝑓(0− ). By the Riemann-Lebesgue lemma we
𝜔→∞ 𝑡

Have
∞
sin 𝜔𝑡
lim ∫ 𝑓(𝑡) 𝑑𝑡 = 0
𝜔→∞ 𝑏 𝑡

Hence:
∞
sin 𝜔𝑡 𝜋
lim ∫ 𝑓(𝑡) 𝑑𝑡 = 𝑓(0− )
𝜔→∞ 0 𝑡 2

0
sin 𝜔𝑡 𝜋
lim ∫ 𝑓(𝑡) 𝑑𝑡 = 𝑓(0− )
𝜔→∞ −∞ 𝑡 2

So that

11
 ∞
sin 𝜔𝑡 𝜋
lim ∫ 𝑓(𝑡) 𝑑𝑡 = (𝑓(0+ ) + 𝑓(0− ))
𝜔→∞ −∞ 𝑡 2

Now we can finally prove the inversion formula of the one-sided Laplace transform.
Assume 𝑓 is continuous at 𝑡. Since the integral of the Laplace transform is uniformly
convergent we can switch integrals in the following expression:

1 𝑐+𝑖𝑅 1 𝑐+𝑖𝑅 ∞
∫ 𝐹(𝑠)𝑒 𝑠𝑡 𝑑𝑠 = ∫ (∫ 𝑓(𝑢)𝑒 −𝑠𝑢 𝑑𝑢) 𝑒 𝑠𝑡 𝑑𝑠
2𝜋𝑖 𝑐−𝑖𝑅 2𝜋𝑖 𝑐−𝑖𝑅 0

1 ∞ 𝑐+𝑖𝑅
= ∫ 𝑓(𝑢) ∫ 𝑒 𝑠(𝑡−𝑢) 𝑑𝑠𝑑𝑢
2𝜋𝑖 0 𝑐−𝑖𝑅

1 ∞ 𝑅
= ∫ 𝑓(𝑢) ∫ 𝑒 (𝑐+𝑖𝑟)(𝑡−𝑢) 𝑖𝑑𝑟𝑑𝑢 , 𝑠 = 𝑐 + 𝑖𝑟
2𝜋𝑖 0 −𝑅

1 ∞ 𝑐(𝑡−𝑢) ∫
𝑅
= ∫ 𝑓(𝑢) 𝑒 𝑒 𝑖𝑟(𝑡−𝑢) 𝑑𝑟𝑑𝑢
2𝜋 0 −𝑅

𝑅
𝑖𝑟(𝑡−𝑢)
𝑒 (𝑡−𝑢)𝑖𝑅 − 𝑟 −(𝑡−𝑢)𝑖𝑅
∫ 𝑒 𝑑𝑟 =
−𝑅 𝑖(𝑡 − 𝑢)

2 sin((𝑡 − 𝑢)𝑅)
=
𝑡−𝑢
Now

1 𝑐+𝑖𝑅 1 ∞ sin((𝑢 − 𝑡)𝑅)

∫ 𝐹 (𝑠)𝑒 𝑠𝑡 𝑑𝑠 = ∫ 𝑓(𝑢)𝑒 𝑐(𝑡−𝑢) 𝑑𝑢
2𝜋𝑖 𝑐−𝑖𝑅 𝜋 0 𝑢−𝑡

1 ∞ sin(𝑣𝑅)
= ∫ 𝑓(𝑣 + 𝑡)𝑒 −𝑐𝑣 𝑑𝑣
𝜋 −𝑡 𝑣

Since 𝑓 is continuous if we let 𝑅 → ∞ the integral converges to

𝜋
(𝑓(𝑡 + ) + 𝑓(𝑡 − )) = 𝜋𝑓(𝑡)
2
1 𝑐+𝑖∞
So 𝑓(𝑡) = ∫ 𝐹(𝑠)𝑒 𝑠𝑡 𝑑𝑠 if 𝑓 satisfies the Dirichlet conditions and is
2𝜋𝑖 𝑐−𝑖∞
continuous at 𝑡. By the change of variable previously mentioned we obtain the
inversion formula for the Mellin transform from two-sided Laplace transform which
is a sum of two one-sided Laplace transforms.

Theorem 2.2 (Inversion Formula)

12
Let 𝑓(𝑥) be absolutely integral with fundamental strip 〈𝛼, 𝛽 〉. If 𝑓 sarisfies the
Dirichlet conditions and is continuous at 𝑥 then for any 𝑐 in the interval 〈𝛼, 𝛽 〉

1 𝑐+𝑖∞ ∗
𝑓 (𝑥 ) = ∫ 𝑓 (𝑠)𝑥 −𝑠 𝑑𝑠.
2𝜋𝑖 𝑐−𝑖∞

THE FUNDAMENTAL CORRESPONDENCE

There is a correspondence between the asymptotic expansion of a function at 0 and ∞

and poles of its Mellin transform in a left and respectively right half-plane. The
correspondence fares both ways and forms the basis of an asymptotic process. For
estimating asymptotically a function 𝑓(𝑥), determine its Mellin transform and
translate back its singularities into asymptotic terms in the expansion of 𝑓(𝑥 ).

Singularities

If 𝜙(𝑠)is meromorphic at 𝑠 = 𝑠0 it admits a Laurent expansion near 𝑠0 :

𝜙(𝑠) = ∑ 𝑐𝑘 (𝑠 − 𝑠0 )𝑘 .
𝑘≥−𝑟

The function 𝜙(𝑠) has a pole of order 𝑟 if 𝑟 > 0 and 𝑐−𝑟 ≠ 0.

Definition 3.1

A singular element of 𝜙(𝑠) at 𝑠0 is an initial sum of the Laurent expansion truncated

at terms of order 𝑂(1) or smaller.

Example 3.1
1
Consider the function 𝜙(𝑠) = 𝑠 2(𝑠+1), which has two poles in the complex plane, a
double pole at 𝑠0 = 0 and a simple pole at 𝑠0 = −1.

1 1 1
=
𝑠 2 (𝑠 + 1) 𝑠 + 1 (1 − (𝑠 + 1))2
∞
1 −2
= ∑ ( ) [−(𝑠 + 1)]𝑛
𝑠+1 𝑛
𝑛=0

1
= + 2 + 3 (𝑠 + 1) + ⋯
𝑠+1
1 1 1
=
𝑠 2 (𝑠 + 1) 𝑠 2 1 − (−𝑠)

1 1
= − +1−𝑠+⋯
𝑠2 𝑠

13
Singular elements at 𝑠0 = 0 are:

1 1 1 1 1 1
[ − ] , [ − + 1] , [ − + 1 − 𝑠] , ….
𝑠2 𝑠 𝑠2 𝑠 𝑠2 𝑠
We can truncate the Laurent expansion wherever we want as we include all the terms
with negative degree of 𝑠. Similarly we have the following singular elements for 𝑠0 =
−1:

1 1
[ ],[ + 2] , ….
𝑠+1 𝑠+1

Definition 3.2

Let 𝜙(𝑠) be meromorphic in Ω with 𝑃 including all the poles of 𝜙(𝑠) in 𝛺. A singular
expansion of 𝜙(𝑠) in 𝛺 is a formal sum of singular elements of 𝜙(𝑠) at all points of
𝑃. When 𝐸 is a singular expansion of 𝜙(𝑠) in 𝛺 we write 𝜙(𝑠) ≍ 𝐸 (𝑠 ∈ 𝛺).

Example 3.2

For instance

1 1 1 1 1
≍[ ] +[ 2− ] +[ ]
𝑠 2 (𝑠 + 1) 𝑠 + 1 𝑠=−1 𝑠 𝑠 𝑠=0 2 𝑠=1

This expansion is a concise way of combining information contained in the Laurent

expansions of the function 𝜙(𝑠) = 𝑠 2 (𝑠 + 1)−1 at the three points of 𝑃 = {−1,0,1}.
Note that there is no pole at 𝑠0 = 1, but we can include an extra singular element if
we want. 𝑃 is to include all poles of the function but may also include other points.
Singular expansion is a formal sum which we do not want to evaluate. It only shows
us which poles a function have.

Example 3.3

Recall the gamma function

∞
𝛤 (𝑠) = ∫ 𝑒 −𝑥 𝑥 𝑠−1 𝑑𝑥
0

for 𝑠 ∈ 〈0, ∞〉. The well-known functional equation𝛤(𝑠 + 1) = 𝑠𝛤(𝑠) follows if we

integrate by parts. This allows us to extend 𝛤(𝑠) to a meromorphic function in the
whole of 𝐶. We have from the functional equation:

𝛤(𝑠 + 𝑚 + 1)
𝛤 (𝑠 ) =
𝑠(𝑠 + 1)(𝑠 + 2) … (𝑠 + 𝑚)
(−1)𝑚 1
Thus 𝛤(𝑠) has poles at the points 𝑠 = −𝑚, 𝑚 ∈ ℕ ∪ {0} and 𝛤(𝑠) ∼ as
𝑚! 𝑠+𝑚
𝑠 → −𝑚 . This means the right side is a singular element of 𝛤(𝑠) at 𝑠 = −𝑚. Hence
we obtain the following singular expansion in the whole of ℂ
14
 ∞
(−1)𝑘 1
𝛤 (𝑠 ) ≍ ∑ .
𝑘! 𝑠 + 𝑘
𝑘=0

Direct Mapping

𝑒 −𝑥 has a Taylor expansion at 𝑥 = 0:

∞
−𝑥=
(−1)𝑘 𝑘
𝑒 ∑ 𝑥
𝑘!
𝑘=0

Comparing this with the singular expansion of 𝛤(𝑠) we see there is a striking
1
coincidence of coefficients expressed by the rule 𝑥 𝑘 → 𝑠+𝑘 . We will see that this is
general phenomenon.

Some information about the connection between the asymptotic expansion of our
initial function and properties of its Mellin transform are summed up in the following
table:

𝐟(𝐱) 𝐟 ∗ (𝐬)
Order at 0: 𝑶(𝒙−𝜶 ) Leftmost boundary of f.s. at 𝑅𝑒(𝑠) = 𝛼
Order at ∞: 𝑶(𝒙−𝜷 ) Rightmost boundary of f.s. at ℜ(𝑠) = 𝛽
Expansion till 𝑶(𝒙𝜸 ) at 0 Meromorphic continuation till ℜ(𝑠) = −𝛾
Expansion till 𝑶(𝒙𝜹 ) at ∞ Meromorphic continuation till ℜ(𝑠) = −𝛿

The following table will follow from the Direct mapping theorem:

𝐟(𝐱) 𝐟 ∗ (𝐬)
Term 𝒙𝒂 (𝐥𝐨𝐠 𝒙)𝒌 at 0 Pole with singular element (𝑠+𝑎)𝑘+1
(−1)𝑘 𝑘!

Term 𝒙𝒂 (𝐥𝐨𝐠 𝒙)𝒌 at ∞ Pole with singular element − (𝑠+𝑎)𝑘+1

(−1)𝑘 𝑘!

Visualising these two tables:

-γ α β α β -δ

f.s. f.s.
Theorem 3.1

Let 𝑓(𝑥) have Mellin transform 𝑓 ∗ (𝑠) with non-empty fundamental strip 〈𝛼, 𝛽 〉.

15
(i) Assume that 𝑓(𝑥) admits as 𝑥 → 0+ a finite asymptotic expansion of the form

𝑓(𝑥) = ∑ 𝑐𝑎,𝑘 𝑥 𝑎 (log 𝑥)𝑘 + 𝑂(𝑥 𝛾 )

(𝑎,𝑘)∈𝐴

where – 𝛾 < −𝑎 ≤ 𝛼 and the k are nonnegative. Then 𝑓 ∗ (𝑠) is continuable to a

meromorphic function in the strip 〈– 𝛾, 𝛽 〉 where it admits the singular expansion

(−1)𝑘 𝑘!
𝑓 ∗ (𝑠) ≍ ∑ 𝑐𝑎,𝑘 , 𝑠 ∈ 〈−𝛾, 𝛽 〉.
(𝑠 + 𝑎)𝑘+1
(𝑎,𝑘)∈𝐴

(ii) Similarly assume that 𝑓(𝑥) admits as 𝑥 → ∞ a finite asymptotic expansion of the
form

𝑓(𝑥) = ∑ 𝑐𝑎,𝑘 𝑥 𝑎 (log 𝑥)𝑘 + 𝑂(𝑥 𝛾 )

(𝑎,𝑘)∈𝐴

where now 𝛽 ≤ −𝑎 ≤ −𝛾. Then 𝑓 ∗ (𝑠) is continuable to a meromorphic function in

the strip 〈𝛼, −𝛾〉 where

(−1)𝑘 𝑘!
𝑓 ∗ (𝑠) ≍ − ∑ 𝑐𝑎,𝑘 , 𝑠 ∈ 〈𝛼, −𝛾〉.
(𝑠 + 𝑎)𝑘+1
(𝑎,𝑘)∈𝐴

Thus terms in the asymptotic expansion of 𝑓(𝑥) at 0 induce poles of 𝑓 ∗ (𝑠) in a strip
to the left of the fundamental strip. Terms in the expansion at ∞ induce poles in a strip
to the right.

Proof:
1
Since 𝑀 [𝑓 (𝑥) ; 𝑠] = 𝑀[(𝑓 (𝑥 )); −𝑠] we only need to treat case (i) corresponding to
𝑥 → 0+ .

By assumption 𝑔(𝑥) = 𝑓(𝑥 ) − ∑(𝑎,𝑘)∈𝐴 𝑐𝑎,𝑘 𝑥 𝑎 (log 𝑥 )𝑘 satisfies 𝑔(𝑥 ) = 𝑂(𝑥 𝛾 ). For
𝑠 ∈ 〈𝛼, 𝛽 〉 we have:
1 ∞
𝑓 𝑠) = ∫ 𝑓(𝑥 )𝑥 𝑠−1 𝑑𝑥 + ∫ 𝑓(𝑥)𝑥 𝑠−1 𝑑𝑥
∗(
0 1

1 1 ∞
= ∫ 𝑔(𝑥 )𝑥 𝑠−1 𝑑𝑥 + ∫ ( ∑ 𝑐𝑎,𝑘 𝑥 𝑎 (log 𝑥)𝑘 ) 𝑥 𝑠−1 𝑑𝑥 + ∫ 𝑓(𝑥)𝑥 𝑠−1 𝑑𝑥
0 0 (𝑎,𝑘)∈𝐴 1

In the last line the first integral defines an analytic function of 𝑠 in the strip 〈– 𝛾, 𝛽 〉
since 𝑔(𝑥 ) = 𝑂(𝑥 𝛾 ) as 𝑥 → 0, the third integral is analytic in 〈−∞, 𝛽 〉 so that the sum
of these two are analytic in 〈– 𝛾, 𝛽 〉. Finally

16
 1
(−1)𝑘 𝑘!
∫ ( ∑ 𝑐𝑎,𝑘 𝑥 𝑎 (log 𝑥)𝑘 ) 𝑥 𝑠−1 𝑑𝑥 = ∑ 𝑐𝑎,𝑘
0 (𝑠 + 𝑎)𝑘+1
(𝑎,𝑘)∈𝐴 (𝑎,𝑘)∈𝐴

This result follows since

1 1
𝑘
∫ (log 𝑥 𝑥 )𝑘 𝑠+𝑎−1 𝑑𝑥 = − ∫ (log 𝑥 )𝑘−1 𝑥 𝑠+𝑎−1 𝑑𝑥
0 𝑠+𝑎 0

from integration by parts. Integrating by parts 𝑘 times yields the above result. Now
(−1)𝑘 𝑘!
∑(𝑎,𝑘)∈𝐴 𝑐𝑎,𝑘 is meromorphic in all of ℂ and provides the singular expansion
(𝑠+𝑎)𝑘+1
of 𝑓 ∗ (𝑠) in the extended strip. This completes the proof of the Direct mapping
theorem.

Note in the case where there exists a complete expansion of 𝑓(𝑥) at 0 (or ∞), the
transform 𝑓 ∗ (𝑠) becomes meromorphic in a complete left (or right) half-plane. This
situation always occurs for functions that are analytic at 0 (or ∞).

Example 3.4

The Riemann zeta function plays a pivotal role in analytic number theory and has
applications in physics, probability theory and applied statistics. Using singular
expansion and the Direct mapping theorem we can easily derive some of its
properties.
𝑒 −𝑥
Recall example 2.2 where we showed if 𝑓(𝑥 ) = then 𝑓 ∗ (𝑠) = 𝛤(𝑠)𝜁(𝑠) with
1−𝑒 −𝑥
fundamental strip 〈1, ∞〉. 𝑓(𝑥 ) is exponentially small at infinity and it admits a
complete expansion near 𝑥 = 0;
∞
𝑒 −𝑥 𝑥𝑘
= ∑ 𝐵𝑘+1
1 − 𝑒 −𝑥 (𝑘 + 1) !
𝑘=−1

1 1
which defines the Bernoulli numbers 𝐵𝑘 : 𝐵0 = 1, 𝐵1 = − 2 , 𝐵2 = 6 , ….

Thus 𝑓 ∗ (𝑠) = 𝛤 (𝑠)𝜁(𝑠) is meromorphic on the whole of ℂ with singular expansion

∞
𝐵𝑘+1 1
𝛤 (𝑠 )𝜁 ( 𝑠 ) ≍ ∑ ,
( 𝑘 + 1)! 𝑠 + 𝑘
𝑘=−1

if we compare it with the regular expansion of the gamma function

∞
(−1)𝑘 1
𝛤 (𝑠 ) ≍ ∑
𝑘! 𝑠 + 𝑘
𝑘=−1

17
we can extract the singular expansion of the zeta function
∞
1 𝐵𝑗+1
𝜁 (𝑠 ) ≍ [ ] + ∑ [(−1)𝑗 ]
𝑠 − 1 𝑠=1 𝑗 + 1 𝑠=−𝑗
𝑗=0

Hence the zeta function is meromorphic in the whole of ℂ with only a simple pole at
𝑠 = 1. It follows from the singular expansion that

𝐵𝑚+1
𝜁 (−𝑚) = (−1)𝑚 , 𝑚 ∈ ℕ ∪ {0}.
𝑚+1
1
Hence 𝜁 (0) = − 2. Since 𝐵2𝑘+1 = 0 for 𝑘 ∈ ℕ it follows that 𝜁(−2𝑚) = 0 for all
𝑚 ∈ ℕ. These zeros are trivial zeros of the zeta function. The nontrivial zeros of the
zeta function lie in the strip {0 ≤ ℜ(𝑠) ≤ 1} which is called the critical strip.

The most famous unsolved problem in complex analysis is the Riemann hypothesis
1
which states that the nontrivial zeros of the zeta function lie on the line {ℜ(𝑠) = 2}. It
was also one of the Clay Mathematics Institute Millennium Prize Problems.

Computer calculation have shown that the first 10 trillion zeros lie on the critical line.
Whether the Riemann hypothesis is true remains a famous unresolved problem.

Example 3.5
𝜋
Recall from Example 2.3 the Mellin pair 𝑓 (𝑥 ) = (1 + 𝑥)−1 , 𝑓 ∗ (𝑠) = sin 𝜋𝑠 with
fundamental strip 〈0,1〉. As 𝑥 → 0+ ;
∞
1
= ∑(−1)𝑛 𝑥 𝑛
1+𝑥
𝑛=0

and as 𝑥 → ∞:
∞
1 𝑥 −1
= ∑(−1)𝑛−1 𝑥 −𝑛
1 + 𝑥 1 + 𝑥 −1
𝑛=1

These expansions translate into:

∞
∗
(−1)𝑛
𝑓 (𝑠) ≍ ∑ 𝑠 ∈ 〈−∞, 1〉
𝑠+𝑛
𝑛=0

which is the continuation of the transformed function to the left of the fundamental
strip and
∞
∗(
(−1)𝑛−1
𝑓 𝑠) ≍ − ∑ 𝑠 ∈ 〈0, ∞〉
𝑠−𝑛
𝑛=1

18
which is the continuation of the transformed function to the right of the fundamental
strip. We can combine the two singular expansions into one, giving us a singular
expansion in the whole of ℂ:

∗(
𝜋 (−1)𝑛
𝑓 𝑠) = ≍ ∑ , 𝑠 ∈ ℂ.
sin 𝜋 𝑠 𝑠+𝑛
𝑛∈ℤ

3.3 Converse Mapping:

Under a set of mild conditions, a converse to the Direct mapping theorem also holds:
The singularities of a Mellin transform which is small enough towards ±𝑖∞ encode
the asymptotic properties of the original function.

Theorem 3.2

Let 𝑓(𝑥) be continuous with Mellin transform 𝑓 ∗ (𝑠) having a non-empty fundamental
strip 〈𝛼, 𝛽 〉.

(i) Assuming that 𝑓 ∗ (𝑠) admits a meromorphic continuation to the strip 〈𝛾, 𝛽 〉 for
some 𝛾 < 𝛼 with a finite number of poles there, and is analytic on ℜ(𝑠) = 𝛾. Assume
a;sp that there exists a real number 𝜂 ∈ (𝛼, 𝛽 ) such that 𝑓 ∗ (𝑠) = 𝑂(|𝑠|−𝑟 ) with 𝑟 > 1
when |𝑠| → ∞ in 𝛾 ≤ ℜ(𝑠) ≤ 𝜂. If 𝑓 ∗ (𝑠) admits the singular expansion for 𝑠 ∈
〈𝛾, 𝛼 〉,

1
𝑓 ∗ (𝑠) = ∑ 𝑑𝑎,𝑘
(𝑠 − 𝑎)𝑘
(𝑎,𝑘)∈𝐴

then an asymptotic expansion of 𝑓(𝑥) at 0 is

(−1)𝑘−1 −𝑎
𝑓 (𝑥 ) = ∑ 𝑑𝑎,𝑘 ( 𝑥 (𝑙𝑜𝑔 𝑥)𝑘−1 ) + 𝑂(𝑥 −𝛾 ),
(𝑘 − 1)!
(𝑎,𝑘)∈𝐴

(ii) Similarly assume that 𝑓 ∗ (𝑠) admits a meromorphic continuation to 〈𝛼, 𝛾〉 for some
𝛾 > 𝛽 and is analytic on ℜ(𝑠) = 𝛾. Assume again that 𝑓 ∗ (𝑠) = 𝑂(|𝑠|−𝑟 ) with 𝑟 > 1
when |𝑠| → ∞ in 〈𝜂, 𝛾〉 for some 𝜂 ∈ (𝛼, 𝛽). If

1
𝑓 ∗ (𝑠) ≍ ∑ 𝑑𝑎,𝑘
(𝑠 − 𝑎)𝑘
(𝑎,𝑘)∈𝐴

for 𝑠 ∈ (𝜂, 𝛾), then the asymptotic expansion of 𝑓(𝑥) at ∞ is

(−1)𝑘−1 −𝑎
𝑓 (𝑥 ) = − ∑ 𝑑𝑎,𝑘 ( 𝑥 (𝑙𝑜𝑔 𝑥)𝑘−1 ) + 𝑂(𝑥 −𝛾 ).
(𝑘 − 1)!
(𝑎,𝑘)∈𝐴

Proof

19
As in the direct mapping theorem, it suffices to consider case (i0 corresponding to
continuation to the left. Consider the integral:

1
𝐼 (𝑇 ) = ∫ 𝑓 ∗ (𝑠)𝑥 −𝑠 𝑑𝑠
2𝜋𝑖 𝜕𝐷

where D denotes the rectangular contour defined by the segments

[𝜂 − 𝑖𝑇, 𝜂 + 𝑖𝑇], [𝜂 + 𝑖𝑇, 𝛾 + 𝑖𝑇], [𝛾 + 𝑖𝑇, 𝛾 − 𝑖𝑇], [𝛾 − 𝑖𝑇, 𝜂 − 𝑖𝑇].

By the residue theorem 𝐼(𝑇) is equal to the sum of residues, which are:

𝑥 −𝑠
𝑅 = ∑ 𝑑𝑎,𝑘 𝑅𝑒𝑠 ( )
(𝑠 − 𝑎)𝑘 𝑠=𝑎
(𝑎,𝑘)∈𝐴

(−1)𝑘−1 −𝑎
= ∑ 𝑑𝑎,𝑘 ( 𝑥 (log 𝑥)𝑘−1 )
(𝑘 − 1) !
(𝑎,𝑘)∈𝐴

Let 𝑇 → ∞. The integrals along the two horizontal segments vanish, since they are
𝑂(𝑇 −𝑟 ) and tend to 0 as 𝑇 → ∞. The integral along the vertical line ℜ(𝑠) = 𝜂 lies
inside the fundamental strip and tends to the inverse Mellin integral 𝑓(𝑥) by the
inversion theorem since 𝑓 (𝑥 ) is continuous. For integral along the vertical line
ℜ(𝑠) = 𝛾 we have:

1 𝛾+𝑖∞ ∗ 1 𝛾+𝑖∞ ∗
| ∫ 𝑓 (𝑠) 𝑥 −𝑠 𝑑𝑠| ≤ ∫ |𝑓 (𝑠)|𝑥 −ℜ(𝑠) |𝑑𝑠|
2𝜋𝑖 𝛾−𝑖∞ 2𝜋 𝛾−𝑖∞

∞
𝑥 −𝛾
= 𝑂(1) ∫ 𝑑𝑡
0 (1 + 𝑡)𝛾

= 𝑂(𝑥 −𝛾 )

Hence for 𝑇 → ∞ we have 𝑅 = 𝑓 (𝑥 ) + 𝑂(𝑥 −𝛾 ). This completes the proof of the

Converse mapping theorem.

The following table sums up aspects of the Converse mapping theorem:

𝑓 ∗ (𝑠) 𝑓(𝑥)
Pole at 𝑎 Term in asymptotic expansion ≈ 𝑥 −𝑎
Left of fundamental strip expansion at 0
Right of fundamental strip expansion at ∞

20
Simple pole
1
Left: 𝑠−𝑎 𝑥 −𝑎 at 0
1
Right: −𝑥 −𝑎 at ∞
𝑠−𝑎

Multiple pole Logarithmic factor

1 (−1)𝑘
Left:(𝑠−𝑎)𝑘+1 𝑥 −𝑎 (log 𝑥)𝑘 at 0
𝑘!
1 (−1)𝑘
Right: (𝑠−𝑎)𝑘+1 − 𝑘! 𝑥 −𝑎 (log 𝑥)𝑘 at ∞

Example 3.6
𝛤(𝑠)𝛤(𝑣−𝑠)
In example 2.4 we showed that (1 + 𝑥)−𝑣 has Mellin transform . Now
𝛤(𝑣)
𝛤(𝑠)𝛤(𝑣−𝑠)
consider 𝜙(𝑠) = that is analytic in the strip 〈0, 𝑣〉. We use the singular
𝛤(𝑣)
expansion of the gamma function to obtain the singular expansion to the left of
ℜ(𝑠) = 0 (i.e. for 𝑠 ∈ 〈−∞, 𝑣〉):

(−1)𝑘 𝛤 (𝑣 + 𝑘) 1
𝜙 (𝑠 ) ≍ .
𝑘 𝛤 (𝑣 ) 𝑠 + 𝑘

Since the gamma function devreases along vertical lines it follows that 𝜙(𝑠)
decreases along vertical lines, so that the conditions of the converse mapping theorem
is satisfied.

𝜙(𝑠) is the Mellin transform of some function 𝑓(𝑥), which can be obtained by the
inversion formula:
𝑣
1 2+𝑖∞
𝑓 (𝑥 ) = ∫ 𝜙 (𝑠)𝑥 −𝑠 𝑑𝑠
2𝜋𝑖 𝑣−𝑖∞
2

Singularities of 𝜙(𝑠) in 〈−∞, 𝑣〉 encode asymptotics for 𝑓(𝑥) at 0;

∞
(−1)𝑘 𝛤 (𝑣 + 𝑘) 𝑘
𝑓 (𝑥 ) = ∑ 𝑥
𝑘! 𝛤 (𝑣 )
𝑘=0

Furthermore
∞
−𝑣
−𝑣 𝑘
(1 + 𝑥) = ∑( )𝑥
𝑘
𝑘=0

21
 ∞
(−1)𝑘
=∑ 𝑣 (𝑣 + 1)(𝑣 + 2) … (𝑣 + 𝑘 − 1)𝑥 𝑘
𝑘!
𝑘=0

∞
(−1)𝑘 𝛤 (𝑣 + 𝑘) 𝑘
=∑ 𝑥
𝑘! 𝛤 (𝑣 )
𝑘=0

1 𝛤(𝑠)𝛤(𝑣−𝑠)
Hence we again find the Mellin pair 𝑓 (𝑥 ) = (1+𝑥)𝑣 , 𝑓 ∗ (𝑠) = , this time
𝛤(𝑣)
using the Inversion formula and the Converse mapping theorem.

22