OpenStax-CNX module: m38159 1

Differentiation under the integral

sign *

Narayan C.R.

This work is produced by OpenStax-CNX and licensed under the

„
Creative Commons Attribution License 3.0

A denite integral f (x, t) dx of a continuous functionf (x, t) has a value that depends on the choice of
Rb
a
t. For example let

f (x, t) = cos (tx) dx (1)

Z Z
1
f (x, t) dx = cos (tx) dx = sin (tx) (2)
t
It can be seen that the indenite integral is a function of x and t. Consider a denite integral .
Z π/2 π/2
1 1
(3)

cos (tx) dx = sin (tx) =
0 t 0 t
It is clear that the denite integral is a function of t alone since the variable x disappears when the values
of the limits are plugged into the equation.
Now we will consider a case where the limits themselves depend on t . ' '

Z t t
1 2
= sin2 t (4)

cos (tx) dx = sin (tx)
−t t −t t
Which is again a function of t. Therefore we can easily conclude that
Z b
f (x, t) dx = F (t) . (5)
a
The above expression is called an integral depending on a parameter and here t is termed the parameter. In
many cases the indenite integral in the above expression cannot be found explicitly . Then the evaluation
is eected considering F (t) and investigating its contiuity and dierentiability. In particular we will try to
nd an expression for the derivative F (t). '

We shall show that in general F (t) can be dierentiated , when f(x,t) has continuous derivative with
respect to t. It is called dierentiating under the integral sign . We can show
Z b
dF (t) ∂
= f (x, t) dx (6)
dt a ∂t
* Version 1.2: Apr 29, 2011 1:22 pm -0500
„ http://creativecommons.org/licenses/by/3.0/

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 2

. That is
Z b Z b
d ∂
f (x, t) dx = f (x, t) dx (7)
dt a a ∂t
In otherwords, dierentiation and integration can be interchanged. The above result is called Leibnitz's
rule, which is formally stated as follows:
Let f(x,t) be a continuous function and have a continuous derivative ∂f
∂t in the domain a ≤ x ≤ b, c ≤ t ≤ d,
then
Z b Z b
d ∂
f (x, t) dx = f (x, t) dx (8)
dt a a ∂t
Proof: Let f (x, t) and ∂f (x,t)
∂t both be continuous functions of x and t, where a ≤ x ≤ b, c ≤ t ≤ d,. Also
let
Z b
F (t) = f (x, t) dx (9)
a
Let ∆t denote a change in t. We then have
Z b
F (t + ∆t) = f (x, t + ∆t) dx (10)
a
Therefore
Z b
F (t + ∆t) − F (t) = f (x, t + ∆t) − f (x, t) dx (11)
a
Using Lagrange's mean value theorem, we obtain
Z b
F (t + ∆t) − F (t) = ∆tft (x, t + θ∆t) dx [0 < θ < 1] (12)
a
Z b
F (t + ∆t) − F (t)
= ft (x, t + θ∆t) dx (13)
∆t a
Z b Z b
F (t + ∆t) − F (t)
=⇒ = ft (x, t) dx + [ft (x, t + θ∆t) − ft (x, t)] dx (14)
∆t a a
Let  , be any positive number. Then, sinceft (x, t) is assumed to be continuous, there exists, a positive
number δ such that

|[ft (x, t + θ∆t) − ft (x, t)| <  (b − a) , (15)

When |∆t| ≤ δ.
Z #
 b

[f (x, t + θ∆t)] − ft (x, t) dx <  (b − a) , (16)
 a t

 Z b 
 F (t + ∆t) − F (t) 
=⇒ 

− ft (x, t) dx <  (b − a) when|∆t| ≤ δ.

(17)
 ∆t a 
Therefore
b
F (t + ∆t) − F (t)
Z
lim = ft (x, t) dx (18)
∆t→0 ∆t a

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 3

That is,
Z b
dF (t)
= ft (x, t) dx (19)
dt a
Rewriting
Z b Z b
d ∂
f (x, t) dx = f (x, t) dx . (20)
dt a a ∂t
In a similar way, the rule for dierentiating an integral with a parameter when the parameter occurs in the
limits can be found. That is we wish to nd the derivative of
Z b(t)
F (t) = f (x, t) dx (21)
a(t)

We have now

F (t) = F (t, a, b) . (22)

Where a (t) and b (t) have continuous derivatives with respect to t throught out the interval, also f(t,x) is
continuously deerentiable in the given region. By the chain rule, we obtain,
dF ∂F ∂F db ∂F da
= + + (23)
dt ∂t ∂b dt ∂a dt
Applying the fundamental theorem of integral calculus we obtain the following formula .
Z b
∂f (x, t) db da
F' (t) = dx + f (b, t) − f (a, t) (24)
a ∂t dt dt
EXAMPLES-1
Obtain the following derivatives in the form of integrals:
1.
2
x2
Z
d
2 dx. (25)
dt 1 (1 − tx)
We have
" #
2 2
x2 x2
Z Z
d ∂
2 dx = 2 dx (26)
dt1 (1 − tx) 1 ∂t (1 − tx)
Z 2 Z 2
2x3
= 3 −3
2 x (1 − tx) dx = 3 dx (27)
1 1 (1 − tx)

2.
Z π
d cos (xt)
dt. (28)
dt π/2 x
Z π Z π  
d cos (xt) ∂ cos (xt)
dt = dt (29)
dt π/2 x π/2 ∂t x
Z π  
−x sin (xt)
= dt (30)
π/2 x

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 4

Z π
= − sin (xt) dx . (31)
π/2

3.
Z 2
d
log (xu) dx. (32)
du 1
Z 2 Z 2
d ∂
log (xu) dx = [log (xu)] dx (33)
du 1 1 ∂u
Z 2 Z 2
x dx
dx = . (34)
1 xu 1 u
4.
2
dn
Z
sin x
dx . (35)
dy n
1 x −y
2 Z 2 n 
dn
Z 
sin x ∂ sin x
dx = dx . (36)
dy n 1 x−y 1 ∂y
n x − y

We have
 
∂ sin x ∂ h i h i
= sin x (x − y)
−1 −2
= sin x −(x − y) (−1) (37)
∂y x − y ∂y

= 1 sin x (x − y)
−2
(38)
Dierentiating the above expression,
∂2 h i
sin x (x − y)
−1
= 1 . 2 sin x (x − y)
−3
(39)
∂y 2
Going by the above pattern we can easily conclude that
∂ n sin x
 
n! sin x
= n+1 . (40)
∂y n x − y (x − y)
Z 2 Z 2
dn sin x sin x
n
dx = n! n+1 dx. (41)
dy 1 x − y 1 (x − y)

5. Find dI
dx if
2x
ext
Z
I = dt (42)
x t
Since both the limits are functions of the parameter x, we have to use the second form of Leibnitz rule.
Applying which we get,
2x
text ex.2x ex x
Z
dt + .2 − .1 (43)
x t 2x x
Simplifying
2x
ext

1h 2 i
x2
(44)
2
e − ex +
x x x

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 5

2  2x2 
(45)
2
= e − ex .
x
EXAMPLES-2
1. Evaluate the following integral.
∞


log 1 + a2 x2
Z
dx (46)
0 (1 + b2 x2 )
Let
∞


log 1 + a2 x2
Z
I= dx (47)
0 (1 + b2 x2 )
Dierentiating I with respect to a we get
∞
2ax2
Z
dI 1
= dx (48)
da 0 (1 + b2 x2 ) (1 + a2 x2 )
Z ∞ 
2a 1 1
= 2 − dx (49)
a − b2 0 1 + b2 x 2 1 + a2 x2
since it is known that
Z
1 1 x
dx = tan−1 (50)
a2 + x2 a a
it follows that
1
b2 dx
Z Z Z
1 1 1 1 1 1
= = dx = 2 tan−1 bx = tan−1 bx. (51)
1 + b2 x2 (1/b2 + x2 ) b2 (1/b2 + x2 ) b 1/b b
similalrly
Z
1 1
2 2
dx = tan−1 ax (52)
1+a x a
Therefore, by these values, equation (1) shall be
 ∞
2a 1 1
= 2 −1 −1
tan bx − tan ax (53)
a − b2 b a 0
π
= (54)
b (a + b)
That is,
dI π
= (55)
da b (a + b)
Therefore,
Z Z
πda
dI = (56)
b (a + b)
Integrating
π
I = loge (a + b) + k (57)
b

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 6

where k is arbitrary constant. When a = 0 then I = 0 So, in equation (2) putting a = 0 , We get,
π
loge b + k = 0 (58)
b
Therefore
π
k = − loge b (59)
b
Using the value of k in equation (2)
π π
I= loge (a + b) − loge b (60)
b b
Therefore,
 
π a+b
I = loge . (61)
b b
2. Starting from
Z ∞
1
e−ax dx = , a>0 (62)
0 a
deduce that
Z ∞
n
xn e−ax dx = (63)
0 an+1
Given,
Z ∞
1
e−ax dx = (64)
0 a
Dierentiating with respect to a, n times ,
∞
∂n dn
Z  
1
−ax
(65)


e dx =
0 ∂an dan a
Now, denoting
d −ax

e by y1 (66)
dx
We have y1 = (−1) xe−ax , dierentiating y1 we get, y2 = (−1)2 x2 e−ax . Proceeding in the same manner,
we can guess yn to be equal to (−1)n xn e−ax . Similalrly, let
1
y= , (67)
a
then y1 = − a12 . Continuing in the same manner we can guess yn , going by the above pattern, to be equal
to
n
(−1) n!
yn = (68)
an+1
Therefore, substituting the values of nth derivative on both side of the integral we have,
Z ∞ n
(−1) n!
n
(−1) xn e−ax = (69)
0 an+1

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 7

That is,
Z ∞
n!
xn e−ax = (70)
0 an+1
3. Dierentiating under integral sign the integral
∞
Z r
1 π
e −ax2
dx = ; a>0 (71)
0 2 a
show that
∞ √
π 1.3... (2n − 1)
Z
(72)
2
x2n e−ax dx =
0 2 2n an+1/2
Dierentiating with respect to a, n times
∞ ∞ √
dn ∂ n  −ax2 
Z Z
π ∂  −1/2 
e −ax2
dx = e dx = a (73)
dan 0 0 ∂an 2 ∂an
Proceeding in a similar manner as in the previous examples we get the result.
4. Evaluate :
Z π/2
log α2 cos2 θ + β 2 sin2 θ dθ (74)


0
Let
Z π/2
log α2 cos2 θ + β 2 sin2 θ dθ (75)


I= (1)
0
Also we presume α and β to be greater than 0. Then
π/2
2α cos2 θ dθ
Z
dI
= (76)
dα 0 α2 cos2 θ + β 2 sin2 θ
π/2
cos2 θ dθ
Z
= 2α (77)
0 α2 cos2 θ + β 2 sin2 θ
Divinding the numerator and denominator of the fraction in the the integral, we get
Z π/2
dθ
2α (78)
0 α2 + β 2 tan2 θ
Substituting tanθ = t [range of integration changes from (0 to π/2) to (0 to ∞)] we get
dt
dθ = (79)
sec2 θ
Z ∞
1 dt
= 2α (80)
0 α2 + β 2 t2 sec2 θ
Since sec2 θ = 1 + tan2 θ = 1 + t2 , we can write the above integral as
Z ∞
dt
= 2α (81)
0 (1 + t2 ) (α2 + β 2 t2 )
Z ∞
β2

2α 1
= − dt (82)
β 2 − α2 0 α 2 + β 2 t2 1 + t2

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 8

   ∞
2α β βt
= tan−1
− tan −1
t (83)
β 2 − α2 α α t=0
  
2α β π π
= −1 = (84)
β 2 − α2 α 2 β+α
That is
dI π
= (85)
dα β+α
π
dl = dα (86)
β+α
Integrating, we get,

I = π loge (β + α) + c (2) (87)

Putting α = β in equation (1)
Z π/2
π
loge α2 dθ = 2 (logα) = π loge α (88)


I=
0 2
So it can be concluded that when α = β , I = π loge α. Sustituting the value of I in equation (2) while
subjecting it the condition that α = β we get,

π loge α = π loge 2α + c (89)

which can be written as

π loge α = π loge α + π loge 2 + c (90)

Which implies that

c = − π loge 2 (91)
Substituting this in equation (2) ,

I = π loge (α + β) − π loge 2 (92)

Therefore
 
α+β
I = π loge (93)
2
5. Evaluate
∞
tan−1 (ax)
Z
dx (94)
0 x (1 + x2 )
Let
∞
tan−1 (ax)
Z
I= dx (1) (95)
0 x (1 + x2 )
Dierentiating
∞
tan−1 (ax)
Z  
dI ∂
= dx (96)
da 0 ∂a x (1 + x2 )

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 9

Z ∞
x dx
= (97)
0 x (1 + x2 ) (1 + a2 x2 )
Z ∞
dx
= (98)
0 (1 + x2 ) (1 + a2 x2 )
Throwing into partial fractions, we obtain
∞
a2
Z  
1 1
− dx (99)
1 − a2 0
2 2
a (1 + x ) 1 + a2 x2
∞
a2

1 1
= tan−1 x − tan−1 ax (100)
1 − a2 a2 a 0
Plugging the values of the limits, we get
dI π
= (101)
da 2 (1 + a)
π
dI = da (102)
2 (1 + a)
Integrating, we get
π
I = loge (1 + a) + c (2) (103)
2
When a = 0, I in eqution (1) will be zero. Putting these values in equation (2) we get,
π
loge (1 + 0) + c = 0, (104)
2
Therefore c = 0. Putting back the value of c in equation (2) gives
π
I = loge (1 + a) . (105)
2
6. Assuming the validity of dierentiation under integral sign, show that
∞
1 √ − 1 a2
Z
(106)
2
e−x cos ax dx = πe 4
0 2
Let
Z ∞
(107)
2
F (a) = e−x cos ax dx (1)
0
Dierentiating
Z ∞
(108)
2
'
F (a) = − x e−x sin ax dx
0
Integrating by parts,
∞
a ∞ −x2
Z
1 −x2
(109)

e sin ax −
 e cos ax dx
2 0 2 0
a
F ' (a) = 0 − F (a) (110)
2

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 10

Which implies
F ' (a) a
= − (111)
F (a) 2
Integrating, we get,
a2
loge F (a) = − + loge c (112)
4

F (a) a2
loge = − (113)
c 4
Therefore
a2
F (a) = c e− 4 (2) (114)
Putting a = 0 in (1) we get,
Z ∞ √
π
(115)
2
F (0) = e−x dx =
0 2
Putting a = 0 in (2) , we get F (0) = c. Therefore

F (a) = F (0) = c (116)

1 √ − 1 a2
F (a) = πe 4 . (117)
2
7.Prove that
a a
F (x) + 2x dFdx(x)
Z Z
d F (x) dx
√ = √ dx (118)
da 0 a−x 0 2a a − x
Under what circumstances will
Z a
F (x) dx
√ (119)
0 a−x
be independent of a ?
Let
Z a
F (x) dx
u= √ (120)
0 a−x
and let x = at then dx = a dt . The given integral will become
Z 1 Z 1 Z 1 √ √ Z 1√
F (x) dx F (at) a dt a aF (at) dt aF (at) dt
√ = √ = √ √ = √ . (121)
0 a−x 0 a − at 0 1−t a 0 1−t
We have
Z 1 √ 
du ∂ aF (at)
= √ dt. (122)
da 0 ∂a 1−t
Z 1 F (at) √
√ + a tF (at)
=
2 a
√ dt (123)
0 1−t

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 11

1
F (at) + 2atF ' (at)
Z
= √ √ dt (124)
0 2 a 1−t
which is equivalent to
a
F (x) + 2xF ' (x)
Z
du
= √ dx (125)
da 0 2a a − x
If u = 0a F√(x)dx is to be independent of a , then its dierential coecient with respect to a must be zero.
R
a−x
Therefore we must have

F (x) + 2xF ' (x) = 0 (126)

F ' (x) 1
= = − (127)
F (x) 2x
integrating the above expression we get
1
logF (x) = − logx + constant (128)
2
Simplifying we get
C
F (x) + √ . (129)
x
8. Find the form of F(x) , so that for all values of a
Ra 2
x[F (x)] dx a
R0 a 2 = (130)
[F (x)] dx n
0
We have
Z a Z a
a
x[F (x)] =
2 2
[F (x)] dx (131)
0 n 0
Dierentiating with respect to a , we have
Z a
1 a
a[F (a)] =
2
[F (x)] +
2
[F (a)]
2
(132)
n 0 n
simplifying
1 a
Z
aih
[F (a)] a −
2
=
2
[F (x)] dx (133)
n n 0
1 a
  Z
1
= a 1−
2
[F (x)] =
2
[F (x)] dx (134)
n n 0
Dierentiating again with respect to a we have
    2
1 1 [F (a)]
1−
2
[F (x)] + 2a 1 − F (a) F ' (a) = (135)
n n n
Simplifying we get
   
2 1
1− F (a) + 2 a 1 − F ' (a) = 0 (136)
n n

http://cnx.org/content/m38159/1.2/
OpenStax-CNX module: m38159 12

Re adjusting the terms we get

F ' (a) 1 2−n
= . (137)
F (a) a 2 (n − 1)
Integrating we get,
2−n
logF (a) = loga + K (138)
2 (n − 1)
Where K a constant. Therefore
2−n
F (a) = K a 2(n−1) . (139)
Or we have
2−n
F (x) = K x 2(n−1) . (140)

http://cnx.org/content/m38159/1.2/