Mathematica Moravica

Vol. 19-2 (2015), 65–73

Convolutions Involving the Exponential Function

and the Exponential Integral

Brian Fisher and Fatma Al-Sirehy

Abstract. The exponential integral ei(λx) and its associated func-

tions ei+ (λx) and ei− (λx) are defined as locally summable functions
on the real line and their derivatives are found as distributions. The
convolutions xr ei+ (x) ∗ xs ex+ and xr ei+ (x) ∗ xs ex are evaluated.

1. Introduction and Results

The exponential integral ei(x) is defined for x > 0 by
Z ∞
(1) ei(x) = u−1 e−u d u,
x
see Sneddon [8], the integral diverging for x ≤ 0. It was pointed out in [1]
that equation (1) can be rewritten in the form
Z ∞
(2) ei(x) = u−1 [e−u − H(1 − u)] d u − H(1 − x) ln |x|,
x
where H denotes Heaviside’s function. The integral in this equation is
convergent for all x and so we use equation (2) to define ei(x) on the real
line.
More generally, see [1], if λ 6= 0, we define ei(λx) in the obvious way by
Z ∞
(3) ei(λx) = u−1 [e−u − H(1 − u)] d u − H(1 − λx) ln |λx|.
λx
Further, we define the functions ei+ (λx) and ei− (λx) by
ei+ (λx) = H(x) ei(λx), ei− (λx) = H(−x) ei(λx)
so that
(4) ei(λx) = ei+ (λx) + ei− (λx).
In particular, if λ > 0, we have
Z ∞
(5) ei(λx) = u−1 [e−λu − H(1 − λu)] d u − H(1 − λx) ln |λx|,
x

2010 Mathematics Subject Classification. Primary: 33B10, 46F10.

Key words and phrases. exponential integral, convolution.

2015
c Mathematica Moravica
65
66 Convolutions Involving the Exponential Function. . .
Z ∞
(6) ei+ (λx) = u−1 e−λu d u, x > 0,
x
Z 0
(7) ei− (λx) = −γ(λ) + u−1 (e−λu − 1) d u − ln x− , x < 0,
x
where
γ(λ) = γ + ln |λ|
and Z ∞
γ=− u−1 [e−λu − H(1 − λu)] d u
0
is Euler’s constant.
The derivatives of these functions are given by
[ei(λx)]0 = −e−λx x−1 ,
(8) [ei+ (λx)]0 = −e−λx x−1
+ − γ(λ)δ(x),
[ei− (λx)]0 = e−λx x−1
− + γ(λ)δ(x)),
for all λ 6= 0.
In particular, we have
Z ∞
ei(x) = u−1 [e−u − H(1 − u)] d u − H(1 − x) ln |x|,
Zx∞
(9) ei+ (x) = u−1 e−u d u, x > 0,
x
Z 0
ei− (x) = −γ + u−1 (e−u − 1) d u − ln x− , x < 0,
x
where Z ∞
γ=− u−1 [e−u − H(1 − u)] d u
0
is Euler’s constant.
The derivatives of these functions are given by
[ei(x)]0 = −e−x x−1 ,
(10) [ei+ (x)]0 = −e−x x−1
+ − γδ(x),
[ei− (x)]0 = e−x x−1
− .
The classical definition of the convolution of two functions f and g is as
follows:

Definition 1. Let f and g be functions. Then the convolution f ∗ g is

defined by Z ∞
(f ∗ g)(x) = f (t)g(x − t) d t
−∞
for all points x for which the integral exist.
 Brian Fisher and Fatma Al-Sirehy 67

It follows easily from the definition that if f ∗ g exists then g ∗ f exists

and

(11) f ∗g =g∗f

and if (f ∗ g)0 and f ∗ g 0 (or f 0 ∗ g) exists, then

(12) (f ∗ g)0 = f ∗ g 0 ( or f 0 ∗ g).

Definition 1 can be extended to define the convolution f ∗ g of two distri-

butions f and g in D0 with the following definition, see Gel’fand and Shilov
[7].

Definition 2. Let f and g be distributions in D0 . Then the convolution

f ∗ g is defined by the equation

h(f ∗ g)(x), ϕi = hf (y), hg(x), ϕ(x + y)ii

for arbitrary ϕ in D, provided f and g satisfy either of the conditions

(a) either f or g has bounded support,
(b) the supports of f and g are bounded on the same side.

It follows that if the convolution f ∗ g exists by this definition then equa-

tions (10) and (11) are satisfied.
The locally summable functions ex+ and ex− are defined by

ex+ = H(x)ex ex− = H(−x)ex .

In the following we need the following lemma, which is easily proved by

induction.

Lemma 1.
Z u k
X k!
tk e−t d t = − ui e−u + k!,
0 i!
i=0
Z u k
X k! k!
tk e−2t d t = − ui e−2u + ,
0 2k−i+1 i! 2k+1
i=0

for k = 0, 1, 2, . . ..

We now prove the following theorem.

68 Convolutions Involving the Exponential Function. . .

Theorem 1. The convolution xr ei+ (x) ∗ xs ex+ exists and

xr ei+ (x) ∗ xs ex+ =
s Xr+k   i−1
X s hX (i − 1)! j −x (i − 1)! x i
= (−1)k (r + k)!xs−k x e+ − e
k 2i−j i!j! 2i i! +
k=0 i=1 j=0
s  
(13) X s
+ (−1)k (r + k)!xs−k [ex ei+ (2x) − ex ei+ (x) + ln 2 ex+ ]
k
k=0
s   r+k i
X s hX x i
− (−1)k (r + k)!xs−k + (1 − ex ) ei+ (x),
k i!
k=0 i=1
for r, s = 0, 1, 2, . . . and r, s not both zero.
In particular,
r hX
i−1
X (i − 1)! (i − 1)! x i
xr ei+ (x) ∗ ex+ = r! xj e−x
+ − e
2i−j i!j! 2i i! +
i=1 j=0
(14) + r![e ei+ (2x) − ex ei+ (x) + ln 2 ex+ ]
x

r
hX xi i
− r! + (1 − ex ) ei+ (x),
i!
i=1
for r = 1, 2, . . . and
(15) ei+ (x) ∗ ex+ = − ei+ (x) + ex ei+ (2x) + ln 2 ex+ .
Proof. The convolution xr ei+ (x) ∗ xs ex+ = 0 if x < 0 and so when x > 0, we
have
Z x Z ∞
r s x
x ei+ (x) ∗ x e+ = r s x−t
t (x − t) e u−1 e−u d u d t
Z0 x Z u t
= u−1 ex−u tr (x − t)s e−t d t d u
(16) 0 0
Z ∞ Z x
+ u−1 ex−u tr (x − t)s e−t d t d u
x 0
= I1 + I2 ,
where
Z u
tr (x − t)s e−t d t =
0
s   Z u
X s
= k s−k
(−1) x tr+k e−t d t
(17) k 0
k=0
s
X s   r+k i
hX u −u i
=− k
(−1) (r + k)!x s−k
e + (e−u − 1)
k i!
k=0 i=1
 Brian Fisher and Fatma Al-Sirehy 69

and in particular when r = s = 0,

Z u
(18) e−t d t = −e−u + 1,
0
on using the lemma.
Hence
s  
X s
I1 = − (−1)k (r + k)!xs−k ex ·
k
k=0
r+k i−1
Z x hX
u i
· e−2u + u−1 (e−2u − e−u ) d u
0 i!
i=1
(19)
s r+k   Z x i−1
X X s u
=− (r + k)! k s−k x
(−1) x e e−2u d u
k 0 i!
k=0 i=1
s
X s   Z x
− (−1)k (r + k)!xs−k ex u−1 (e−2u − e−u ) d u.
k 0
k=0

Further, we have
Z x i−1 i−1
u −2u
X (i − 1)! j −2x (i − 1)!
(20) e du = − i−j i!j!
x e + ,
0 i! 2 2i i!
j=0

on using the lemma, and

Z x
u−1 [e−u − H(1 − u)] d u =
0
Z ∞
= u−1 [e−u − H(1 − u)] d u
0
(21) Z ∞ Z ∞
−1 −u
− u e du + u−1 H(1 − u) d u
x x
Z ∞
= −γ − ei+ (x) + u−1 H(1 − u) d u.
x
Similarly
Z x Z ∞
−1 −2u
(22) u [e − H(1 − 2u)] d u = −γ − ei+ (2x) + u−1 H[1 − 2u] d u.
0 x
It follows from equations (21) and (22) that
Z x
u−1 (e−u − e−2u ) d u =
0
Z ∞
(23)
= ei+ (2x) − ei+ (x) + u−1 [H(1 − u) − H(1 − 2u)] d u
0
= ei+ (2x) − ei+ (x) + ln 2.
70 Convolutions Involving the Exponential Function. . .

In particular, when r = s = 0, we have

(24) I1 = [ei+ (2x) − ei+ (x) + ln 2]ex+ .
Next, as in equation (17), we have
Z x
tr (x − t)s e−t d t =
0
(25) s   r+k i
X s hX x i
=− (−1)k (r + k)!xs−k e−x + (e−x − 1)
k i!
k=0 i=1
and so
s   r+k i
X s k s−k
hX x i
(26) I2 = − (−1) (r + k)!x + (1 − ex ) ei+ (x).
k i!
k=0 i=1
In particular, when r = s = 0, we have
Z ∞
(27) x
I2 = (e − 1) u−1 e−u d u = (ex − 1) ei+ (x).
x
Equation (13) now follows from equations (20), (21), (22), (25) and (26).
Equation (14) follows on putting s = 0 in equation (13) and equation (15)
follows on putting r = 0 in equation (14). 

In the corollary, the distribution x−2 −2 −1 0

+ is defined by x+ = (x+ ) and not
as in Gel’fand and Shilov.

Corollary 1.1. The convolutions (e−x x−1 x −x −2 x

+ ) ∗ e+ and (e x+ ) ∗ e+ exist and

(28) (e−x x−1 x x x

+ ) ∗ e+ = −e ei+ (2x) − γ(2)e+
(29) (e−x x−2 x x x −x −1
+ ) ∗ e+ = 2e ei+ (2x) + 2γ(2)e+ − e x+ .

Proof. The convolution (e−x x−1 x −x −1

+ ) ∗ e+ exists by Definition 2, since e x+
and ex+ are both bounded on the left. From equation (12), we have
[ei+ (x) ∗ ex+ ]0 = −[e−x x−1 x
+ + γδ(x)] ∗ e+
= −(e−x x−1 x x
+ ) ∗ e+ − γe+
= ei+ (x) ∗ [ex+ + δ(x)]
= ex ei+ (2x) + ln 2 ex+
and equation (28) follows.
From equations (12) and (28), we now have
[(e−x x−1 x 0 −x −1 −x −2 x
+ ) ∗ e+ ] = −(e x+ + e x+ ) ∗ e+
= ex ei+ (2x) + γ(2)ex+ − (e−x x−2 x
+ ) ∗ e+
= (e−x x−1 x
+ ) ∗ [e+ + δ(x)]
= −ex ei+ (2x) − γ(2)ex+ + e−x x−1
+
 Brian Fisher and Fatma Al-Sirehy 71

and equation (29) follows. 

Theorem 2. The convolution xr ei+ (x) ∗ xs ex exists and
s Xr+k  
r s x
X s (−1)k (r + k)! s−k x
x ei+ (x) ∗ x e = − x e
k 2i i!
k=0 i=1
(30) s  
X s
+ (−1)k ln 2(r + k)!xs−k ex ,
k
k=0
for r, s = 0, 1, 2, . . . and r, s not both zero.
In particular
r
r x
X r! x
(31) x ei+ (x) ∗ e = − e + ln 2r!ex ,
2i i!
i=1
for r = 1, 2, . . . and
(32) ei+ (x) ∗ ex = ln 2 ex
1
(33) ei+ (x) ∗ xex = ln 2 xex − ln 2 ex + ex .
2
Proof. We have
Z ∞ Z ∞
r
x ei+ (x) ∗ x e =s x r
t (x − t) es x−t
u−1 e−u d u d t
0 t
Z ∞ Z u
−1 x−u
= u e tr (x − t)s e−t d t d u
0 0
s   r+k Z ∞ i−1
X s X u
=− k
(−1) (r + k)!x es−k x
e−2u d u
k i!
k=0 i=1 0
s
X s   Z ∞
− k s−k x
(−1) (r + k)!x e (u−1 e−2u − u−1 e−u ) d u
k 0
k=0
s r+k  
XX s (−1)k (r + k)! s−k x
=− x e
k 2i i
k=0 i=1
s  
X s
+ (−1)k ln 2(r + k)!xs−k ex
k
k=0
on making use of equation (17), the lemma and noting that
Z ∞ Z ∞
−1 −2u −u
u (e − e )du = ln u(2e−2u − e−u ) d u
0 0
= Γ0 (1) − ln 2 − Γ0 (1) = − ln 2,
proving equation (30).
Equation (31) follows on putting s = 0 in equation (30) and equation (32)
follows on putting r = 0 in equation (31).
72 Convolutions Involving the Exponential Function. . .

Equation (31) follows on putting r = 0 and s = 1 in equation (30). 

Corollary 2.1. The convolution (e−x x−n x
+ ) ∗ e exists and

(−1)n 2n−1
(34) e−x x−n x
+ ∗e = γ(2)ex
(n − 1)!
for n = 1, 2, . . . .
In particular,
1 x
(35) e−x x−1 x x
+ ∗ xe = −γ(2)xe − e .
2
Proof. Differentiating equation (32), we get
[−e−x x−1 x −x −1 x x
+ − γδ(x)] ∗ e = −(e x+ ) ∗ e − γe = ln 2e
x

and we see that equation (34) is true when n = 1.

Now assume that equation (34) is true for some n. Then differentiating
equation (34), we get
(−e−x x−n −x −n−1
+ − ne x+ ) ∗ ex = (e−x x−n x
+ )∗e .
It follows that
ne−x x−n−1
+ ∗ ex = −2(e−x x−n
+ )∗e
x

(−1)n+1 2n
= γ(2)ex
(n − 1)!
and so equation (33) is true for n+1. Equation (34) now follows by induction.
Differentiating equation (33), we get
1 x
[−e−x x−1 x x
+ − γδ(x)] ∗ xe = ln 2 xe + e
2
and equation (35) follows. 

For further results involving the exponential integral, see [2, 3, 4, 5] and
[6].

References
[1] B. Fisher and J.D. Nicholas, The exponential integral and the convolution, SUT J.
Math., 33(2)(1997), 139-148.
[2] B. Fisher and J.D. Nicholas, The neutrix convolution product of ei(λx) and xr , Math.
Montisnigri, 9(1998), 51-64.
[3] B. Fisher and J.D Nicholas, On the exponential integral, Hacet. Bull. Nat. Sci. Eng.
Ser. B, 28(1999), 55-64.
[4] B. Fisher and J.D Nicholas, The exponential integral and the neutrix convolution
product, J. Nat. Sci. Math., 40(1,2)(2000), 23-36.
[5] B. Fisher and J.D. Nicholas, On the exponential integral and the non-commutative
convolution product, Integral Transforms Spec. Funct., 11(4)(2001), 353-362.
 Brian Fisher and Fatma Al-Sirehy 73

[6] B. Fisher, E. Özçaḡ and Ü. Gülen, The exponential integral and the commutative
neutrix convolution product, J. Analysis, 7(1999), 7-20.
[7] I.M. Gel’fand and G.E. Shilov, Generalized functions, Vol. I, Academic Press (1964).
[8] I.N. Sneddon, Special Functions of Mathematical Physics and Chemistry, Oliver and
Boyd, (1961).

Brian Fisher
Department of Mathematics
University of Leicester
Leicester, LE1 7RH
England
E-mail address: fbr@le.ac.uk

Fatma Al-Sirehy
Department of Mathematics
King Abdulaziz University
Jeddah
Saudi Arabia
E-mail address: falserehi@kau.edu.sa