DEMONSTRATIO MATHEMATICA 10.

1515/dema-2013-0464
Vol. XLVI No 3 2013

M. A. Pathan, O. A. Daman

LAPLACE TRANSFORMS OF THE LOGARITHMIC

FUNCTIONS AND THEIR APPLICATIONS

Abstract. This paper deals with theorems and formulas using the technique of
Laplace and Steiltjes transforms expressed in terms of interesting alternative logarith-
mic and related integral representations. The advantage of the proposed technique is
illustrated by logarithms of integrals of importance in certain physical and statistical
problems.

1. Introduction
The aim of this paper is to obtain some theorems and formulas for the
evaluation of finite and infinite integrals for logarithmic and related func-
tions using technique of Laplace transform. Basic properties of Laplace and
Steiltjes transforms and Parseval type relations are explicitly used in combi-
nation with rules and theorems of operational calculus. Some of the integrals
obtained here are related to stochastic calculus [6] and common mathemati-
cal objects, such as the logarithmic potential [3], logarithmic growth [2] and
Whittaker functions [2, 3, 4, 6, 7] which are of importance in certain physical
and statistical applications, in particular in energies, entropies [3, 5, 7 (22)],
intermediate moment problem [2] and quantum electrodynamics. The ad-
vantage of the proposed technique is illustrated by the explicit computation
of a number of different types of logarithmic integrals.
We recall here the definition of the Laplace transform

∞
(1) L {f (t)} = L[f (t); s] = e−st f (t) dt.
0

Closely related to the Laplace transform is the generalized Stieltjes trans-

form
2010 Mathematics Subject Classification: 33C05, 33C90, 44A10.
Key words and phrases: Laplace and Steiltjes transforms, logarithmic functions and
integrals, stochastic calculus and Whittaker functions.
534 M. A. Pathan, O. A. Daman


∞
f (t)dt
(2) Sρ {f (t)} = = G(a, ρ)
0
(a + t)ρ
which, for ρ = 1, gives the Stieltjes transform

∞
f (t)dt
(3) S{f (t)} = = G(a).
0
(a + t)
After a change of integration variable, (2) is transfomed into
1
f (−a ln x)
(4) ρ
dx = aρ−1 G(a, ρ).
0
x(1 − ln x)
Special cases of (2) and (4), when f (t) = tλ−1 e−t , are generalizations of
gamma function given by Kobayashi [7]
1
(5) Γρ (λ, a) = (− ln x)λ−1 (a − ln x)−ρ dx, R(λ) > 0.
0
Kobayashi [7] applied this generalized gamma function integral in diffraction
theory.

2. Theorems
In this section, we state and prove some theorems in the study of integral
transforms, and briefly discuss some apparent, known and new special cases
of these theorems. We will apply systematically the rules and theorems of
the operational calculus assuming the existence of the Laplace transforms
of the functions involved and the permissibility of performed mathematical
operations.
Theorem 1. If
(6) L{f (t)} = ϕ(s)
and
(7) L{h(t)} = g(s)
then
1  n    
− ln x − ln x − ln x
g f dx
0
s s s

∞
(8) =s ϕ(t)hn (t − s)dt
s

∞
(9) =s ϕ(s + x)hn (x)dx
0
 Laplace transforms of the logarithmic functions and their applications 535


∞ 
∞
hn (t)
(10) L[{hn (t − a)H(t − a)}; s]ds = Γ(n + 1) dt
0 0
(t + a)n+1
and
1 
∞
φ( − ln
u )
x
(11) dx = L[φ(t − a)H(t − a); x]dx
0
(au − ln x) u

provided that, f (t) ∈ L2 (0, ∞), e−st tn g(t) ∈ L2 (0, ∞), h(t) ∈ L2 (0, ∞) and
hn (t) denotes the nth differential coefficient of h(t) such that h0 (0) = h00 (0) =
· · · = hn−1 (0) = 0. H(t) is the Heaviside’s unit function and integrals in (8)
to (11) are convergent.
Proof. Consider the Laplace transform

∞
(12) L{f (t)} = L[f (t); s] = e−st f (t)dt = ϕ(s).
0

Recall a well known property of the Laplace transform [4, p. 129] that is, if
(13) L{h(t)} = g(s)
then
(14) L{hn (t)} = sn g(s), h0 (0) = h00 (0) = · · · = hn−1 (0) = 0
and
(15) L{hn (t − a)H(t − a)} = e−as sn g(s),
where H(t) is a Heaviside’s unit function. To prove (8) and (9), we use
(12) and (15) in the Parseval theorem and then, by changing the integration
variable x = e−st , we find
1 
ln x n
    
ln x ln x
(16) − g − f − dx
0
s s s

∞ 
∞
=s ϕ(t)hn (t − s)dt = s ϕ(s + x)hn (x)dx,
s 0

where a is replaced by s. Now, we integrate both sides of (15) and use (14)
to obtain

∞
(17) L{hn (t − a)H(t − a)}ds
0
   
" #
∞ ∞ ∞ ∞
−as n −st n n n −as−st
= e s e h (t)dtds = h (t) s e ds dt.
0 0 0 0
536 M. A. Pathan, O. A. Daman

By evaluating the integral on the right hand side of (17), we obtain a Parseval
relation for (13). Thus

∞ 
∞
hn (t)
(18) L{hn (t − a)}H(t − a)}dt = Γ(n + 1) dt.
0 0
(t + a)n+1
∞ ∞ h(t)
For n = 0, (18) gives 0 L{h(t − a)H(t − a)}dt
= 0 (t+a) dt. For n = 0 and
a = 0, (18) gives a known result [5, p. 110 (2.4)]

∞ 
∞
h(t)
(19) L{h(t)}ds = dt.
0 0
t
Now, set h(t) = e−ut φ(t) and n = 0 in (19) and use shift property

∞
−ut
(20) L{h(t)} = L{e φ(t)} = e−ut−st φ(t)dt = L[φ(t); u + s],
0
to get another Parseval-type relation

∞
φ(t)
 
(21) L[φ(t − a)H(t − a); x]dx = L ;u ,
u
a+t
by making the change of variable u + s = x. On changing the integration
variable x = e−ut , we get (11). For a = 0, (21) gives a known result [5,
p. 110 (2.8)]

∞ 
φ(t)

(22) L[φ(t); x]dx = L ;u .
u
t
If we take h(t) = tλ e−at and use [4, p. 129]
dn
(23) L{tn f (t)} = (−1)n
F (s), n = 1, 2, ...
dsn
and binomial theorem in Theorem 1, we obtain
Theorem 2. If L[f (t); s] = F (s) then
1
dn
 
a n − ln x
(24) x s (ln x) f dx = sn+1 n F (s + a)
0
s ds
and
1 
ln x −λ−1
n    
ln x ln x
(25) − a− f − dx
0
s s s
n
!
X (−1)n−r an−r n
=s φ(a; s)
Γ(λ − r + 1) r
r=0
 Laplace transforms of the logarithmic functions and their applications 537

where

∞
(26) φ(a; s) = xλ−r e−ax F (s + x)dx
0
provided that Laplace transform of |f (t)| exists, λ > n − 1, R(s + a) > 0 and
the integrals in (25) and (26) are convergent.
1
In the simplest case f (t) = 1 and F (s) = , we have immediately from
s
(25) and [4, p. 294 (6)]
1 
ln x n ln x −λ−1
  
(27) − a− dx
0
s s
n
!
as X (−1)n−r an−(λ+r+1)/2 n
=e a Wk,m (as),
r=0
s(λ−r−1)/2 r
where k = (r−λ−1)/2, m = (λ−r)/2 and Wk,m (x) is Whittaker function [4].
Evidently, if we set f (t) = 1 in (24), we get
1
s + a −n−1
 
(28) xa/s (ln x)n dx = (−1)n n! .
0
s
Another example is
      
1 1 s+1 s
L[f (t)] = L = ψ −ψ , R(s) > 0
1 + e−t 2 s 2
which leads to
1
xa/s (ln x)n sn+1
    
(n) a + s + 1 (n) a + s
(29) = n+1 ψ −ψ ,
0
1 + x1/s 2 2 2

n = 1, 2, . . . , R(s) > 0, where ψ(ζ) is a psi-function and ψ (n) (ζ) means the
nth derivatives of psi-function [4]. On the other hand, the special cases of
(24) and (25), for n = 0 and a = 0, yield known results [1, p. 241, equations
(23), (26) and (28)].
Theorem 3. Let α > 0, β > 0, then
1f ( −α+β
ln x
) p
(30) p dx = πα(α + β)L[g(θ, t); α]
0 ln(1/x)
where
t e−βθ f (θ)
(31) g(θ, t) = p dθ.
0 π θ(t − θ)
538 M. A. Pathan, O. A. Daman

Proof. Let g(θ, t) be defined by (31). Then


∞  t −βθ
e f (θ)

−αt
L[g(θ, t); α] = e p dθ dt
0 0 π θ(t − θ)
 e−βθ  e−αt
" #
∞ ∞
= √ f (θ) √ dt dθ
0 π θ θ
t−θ
1  e−(α+β)θ  e−αs
" #
∞ ∞
= √ f (θ) √ ds dθ
π 0 θ 0
s
where in the inner integral, we have changed the variable of integration by
setting t − θ = s. It follows from
 e−τ t
∞ r
π
(32) √ dt = , R(τ ) > 0
0 t τ

√1 ∞ e−(α+β)θ
that L[g(θ, t); α] = πα 0
√
θ
f (θ)dθ. The uniqueness of Laplace trans-
forms and the substitution e−(α+β)θ = x implies the required result.
It will be shown that, if we set f (θ) = 1 in the integral (31) and use
(32), then Theorem 3 reduces to the following P. Levy’s Arc-Sine Law for
occupation time of (0, ∞) [6, p. 273, Art 4.11].
Let α > 0, β > 0. Then
1
(33) L[h(θ, t); α] = p
α(α + β)
where
1
t
e−βθ
(34) h(θ, t) = p dθ.
π 0 θ(t − θ)

Assuming the existence of the Laplace transforms of f (t), we consider

L{f (t)} = F (s)
and then using the rules of Laplace transform
a
L{f (t + a)} = eas [F (s) − e−su f (u)du], a≥0
0
and
a
−bt
(35) L{e f (t + a)} = e a(s+b)
[F (s + b) − e−(s+b)u f (u)du],
0

we get the following theorem.

 Laplace transforms of the logarithmic functions and their applications 539

Theorem 4. If L{f (t)} = F (s), then

1    a 
(b ln x)/s ln x a(s+b) −(s+b)u
(36) e f a− dx = se F (s + b) − e f (u)du
0
s 0

provided that a ≥ 0, R(s) > 0, Laplace transform of |f (t)| exist and integrals
in (36) are convergent.

Equation (36) is a generalization of the result [1, p. 239 (10)] which

follows for b = 0.

Theorem 5. If
Sρ {g(t)} = G(a; ρ)
then
1
xas−1 g(−a ln x)
(37) ρ
dx = e−s aρ−1 G(a; ρ).
0
(1 − ln x)

Proof. The above theorem can be proved easily if the definition integral (2)
is applied in the form

∞
e−st g(t)
Sρ {e−st g(t)} = dt
0
(a + t)ρ

which, after a change of integration variable and mere integration by parts

together with (36), leads to
1 
∞
xas−1 g(−a ln x) −s
ρ
dx = a ρ−1
[(e − 1)G(a; ρ) + s e−st G(a; ρ)dt]
0
(1 − ln x) 0
= aρ−1 [(e−s − 1)G(a; ρ) + G(a; ρ)].
Particularly, for s = 0, (37) reduces to [1, p. 250, equation 81(b)].

3. Applications
A number of applications of the formulas for the evaluation of finite
and infinite logarithmic integrals, using the operational calculus technique
of Section 2, can be given. We list some of them.
√ √
As an example of (36), we take f (t) = 1/ t so that L{1/ t} = πs =
p
F (s) and (36) gives
1 r 
(b ln x)/s dx (s+b) π p
(38) e p = se (1 − erf a(s + b))
0 a − ln x/s s+b
540 M. A. Pathan, O. A. Daman

where erf ξ is the error function [4]. For b = 0, (38) becomes a known result
[1, p. 240 (14)].
√ √
Put α = as and β = b/s in (36) and use 1 − erf α = erf c α, where
α > 0, to get
1 r
β ln x dx α(1+β) π p
(39) e √ =e erf c α(1 + β)
0 α − ln x 1+β
because erf ξ then becomes the complementary error function erf cξ (see [4]).
In the second case, we start with f (t) = ln t and we use (36) to get
1
sea(s+b) h −a(s+b)
 
(b ln x)/s ln x i
(40) e ln a − dx = e ln a − Ei(−as − ab) ,
0
s s+b
where τ is the Euler’s constant [4] and Ei(ξ) is the exponential integral [4].
Using the Stieltjes transform Sρ {e−t } = ea Γ(1 − ρ, a), R(ρ) > 0 where
Γ(∗.∗) is an incomplete gamma function in Theorem 5, we get
1
xas+a−1 dx
(41) ρ
= aρ−1 Sρ {e−t(s+1) }
0
(1 − ln x)
= [a(s + 1)]ρ−1 ea(s+1) Γ(1 − ρ, a(s + 1)), a > 0, ρ > 0.
For s = 0, we get
1
xa−1 dx
(42) ρ
= aρ−1 ea Γ(1 − ρ, a), a > 0, ρ > 0
0
(1 − ln x)
which is a correct form of the result [1, p. 252, 97(b)]. (42) becomes a known
result [1, p. 252, 97(a)] when ρ = 1. √
Next, we will turn our attention to the case when g(t) = te−st in (37).
Thus, we have from [8, p. 233 (14.30)]
1
ln(x−1/s )dx √ 1
(43) = πaeas Γ(− , as).
0
(as − ln x) 2
Formulas (41) to (42) show that a more general case can be considered by
using g(t) = tλ−1 e−st in Theorem 5. By defining a generalization of gamma
function (see Kobayashi [7])

∞
(44) Γρ,s (λ, a) = tλ−1 e−st (t + a)−ρ dt, R(λ), R(s) > 0
0
and changing the integration variable, we have
1 
ln x −ρ

−λ
(45) Γρ,s (λ, a) = s a− (− ln x)λ−1 dx, R(λ), R(s) > 0
0
s
 Laplace transforms of the logarithmic functions and their applications 541

which, for s = 1, reduces to the known generalization of gamma function

(5) given by Kobayashi [7]. Note that Γρ,1 (λ, a) = Γρ (λ, a). In view of the
result [4, p. 294 (6)], it is more natural to work with (37) and (45) to obtain
(46) Γρ,s (λ, a) = Γ(λ)(sa)(ρ−λ−1)/2 eas/2 Wk,m (a, s),
where k = 1−λ−ρ2 , m = λ−ρ2 and R(λ) > 0. Thus, we have provided an
integral representation for Γρ,s (λ, a) in terms of Whittaker functions.
We now use Theorem 2 and result (24) to obtain a generalization of the
result of Apelbalt [1, p. 238, (4)] involving the nth derivatives of psi-func-
tion
1 b/s  n+1
x (ln x)n
 
s (n) s + b
(47) dx = − ϕ .
0
1 − xa/s a a

For b = 0, (47) gives a corrected form of the result [1, p. 238 (4)]
1
(ln x)n
 
1 (n) 1
(48) dx = − ϕ
0
1 − xα (α)n+1 α
where
α = a/s.
To prove (47), notice that for n = 1, 2, . . . ,
tn
   
−n−1 (n) s
(49) L{f (t)} = L = (−a) ϕ = F (s)
1 − e−at a
which is a well-known result (see [4] and [1, p. 238 (3)]). From the opera-
tional relations (24) and (49), result (47) follows.

References

[1] A. Apelblat, Application of Laplace transformation to evaluation of integrals, J. Math.

Anal. Appl. 186 (1994), 237–253.
[2] C. Berg, H. L. Pedersen, Logarithmic order and type of intermediate moment prob-
lems, Proceedings of the International Conference on Difference Equations, Special
Functions and Orthogonal Polynomials, Munich, Germany, 25–30 July 2005, 51–59.
[3] J. S. Dehesa, A. Matinerz-Finkelshtein, J. Sanchez-Ruiz, Quantum information en-
tropies and orthogonal polynomials, J. Comput. Appl. Math. 133 (2001), 23–46.
[4] A. Erdelyi et al., Tables of Integral Transforms, Vol. 1 , McGraw Hill, New York, 1954.
[5] S. Herman, J. Maceli, S. Rogala, O. Yurekli, Parseval type relations for Laplace trans-
forms and their applications, Internat. J. Math. Ed. Sci. Tech. 39(1) (2008), 109–115.
[6] I. Karatzas, S. E. Shreve, Brownian Motion and Stochastic Calculus, Springer, New
York, 1991.
542 M. A. Pathan, O. A. Daman

[7] K. Kobayashi, On a generalized gamma function occurring in diffraction theory,

J. Phys. Soc. Japan 60 (1991), 1501–1512.
[8] J. S. Lomont, J. Brillhart, Elliptic Polynomials, Chapman and Hall, New York, 1981.

O. A. Daman
DEPARTMENT OF MATHEMATICS
UNIVERSITY OF BOTSWANA
P/BAG 0022, GABORONE, BOTSWANA
E-mail: [email protected]

M. A. Pathan
CENTRE FOR MATHEMATICAL SCIENCES
ARUNAPURAM P.O., PALA
KERALA - 686574, INDIA
E-mail: [email protected]

Received August 11, 2011.