A Primer on Integration

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 1 / 51

1. The fundamental theorem of calculus
Let f : R ! R be a continuous and non-negative function (f 0). We
want to …nd the area A between the graph of the function and the
horizontal axis on the interval [a, b ] .

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 2 / 51

Let us …x the value a and make the area a function of the upper endpoint
of the interval, A(x ),

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 3 / 51

Let us increase the value x by a small (in…nitessimal) amount ∆x. The
increase in the area is ∆A = A(x + ∆x ) A(x ).

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 4 / 51

 Let us have a closer look to the increase of the area, ∆A,

We see that
f (x ) ∆x A(x + ∆x ) A(x ) f (x + ∆x ) ∆x, if f is increasing,
whereas
f (x ) ∆x A(x + ∆x ) A(x ) f (x + ∆x ) ∆x, if f is decreasing.

Note that for ∆x su¢ ciently small, f is either increasing or decreasing

on the interval [x, x + ∆x ] .
J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 5 / 51
 Divide the three terms in the previous expression by ∆x

A(x + ∆x ) A(x )
f (x ) f (x + ∆x )
∆x

and take the limit when ∆x ! 0,

A(x + ∆x ) A(x )
lim f (x ) lim lim f (x + ∆x ).
∆x !0 ∆x !0
| {z∆x } ∆x !0

A 0 (x )

Since f is continuous it holds that lim f (x + ∆x ) = f (x ).

∆x !0
Therefore, we get
f (x ) A0 (x ) f (x )

=) A0 (x ) = f (x ). (Fundamental theorem of calculus I)

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 6 / 51

 Therefore, the derivative of the area A(x ) at the point x is equal to
the function f evaluated at the point x, and this is true for all x.

This means that the problem of …nding the area on the interval [a, x ]
is equivalent to the problem of …nding a function F such that F 0 = f .

De…nition. F is a "primitive", "antiderivative" or "inde…nite

integral" of the function f if F 0 = f . We write the inde…nite integral
of f as Z
f (x )dx.

Hence, the area A ( ) (as a function of the upper endpoint x ) is a

"primitive", "antiderivative" or "inde…nite integral" of the function f .

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 7 / 51

 Lemma 1. If F is a primitive of f then G = F + C , where C is a
constant (or scalar), is also a primitive of f .

Lemma 2. If F and G are primitives of f then G (x ) F (x ) = C , for

all x.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 8 / 51

 Hence,
A(x ) = F (x ) + C , where F 0 = f .

Note that A(a) = 0 so that

0 = A(a) = F (a) + C =) C = F (a ).

and, thus,

A(x ) = F (x ) F (a ), where F 0 = f .
(Fundamental theorem of calculus II)

In particular, the area A between the graph of the continuous and

non-negative function f and the horizontal axis on the interval [a, b ] is

A = F (b ) F (a ), where F 0 = f .

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 9 / 51

 De…nition. The de…nite integral of the continuous function f on the
interval [a, b ] is
Z b
f (x )dx = F (b ) F (a ) [F (x )]ba F (x )jba , where F 0 = f .
a

Thus, when f is continuous and non-negative on the interval [a, b ] ,

the area A between the graph of f and the horizontal axis on the
interval [a, b ] is
Z b
A= f (x )dx.
a

This is the concept of integral due to Newton and Leibniz. Other

more general concepts of integral are those of Riemann and of
Lebesgue.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 10 / 51

Example:

Z 1 1
2 x3 1
A= x dx = = .
0 3 0 3

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 11 / 51

If the function f were negative,

then the area A of the shaded region will be

Z b Z b
A= [ f (x )] dx = f (x )dx 0.
a
| a {z }
negative

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 12 / 51

If we want to …nd he area A between the graph of a continuous function f
and the horizontal axis on the interval [a, b ] , where the function takes
both positive and negative values on that interval,

then the area of the shaded region will be

Z c1 Z c2 Z c3 Z b
A= f (x )dx + f (x )dx f (x )dx + f (x )dx 0.
|a {z } | c1 {z } |
c2
{z } | {z }
c3
negative positive negative positive

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 13 / 51

2. The inde…nite integral and its properties

Recall that the inde…nite integral of a continuous function f is

Z
f (x )dx = F (x ) + C , where F 0 = f and C is a constant.

Properties of the inde…nite integral of a continuous function:

1. Z Z
cf (x )dx = c f (x )dx, where c is a constant.

2. Z Z Z
[f (x ) + g (x )] dx = f (x )dx + g (x )dx.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 14 / 51

Summing up:
Z
! Z
N N
∑ [cn fn (x )] dx = ∑ cn fn (x )dx , where fcn gN
n =1 are scalars.
n =1 n =1

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 15 / 51

3. The de…nite integral and its properties

Recall that the de…nite integral of the continuous function f is

Z b
f (x )dx = F (b ) F (a), where F 0 = f .
a

Note:
Z b Z b
f (x )dx = f (z )dz = F (b ) F (a), where F 0 = f .
a a

That is, the variable appearing as the argument of the function f is a

"mute" variable.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 16 / 51

Properties of the de…nite integral of a continuous function:
1. Z b Z a
f (x )dx = f (x )dx.
a b

2. Z a
f (x )dx = 0.
a

3. Z b Z b
cf (x )dx = c f (x )dx, where c is a constant.
a a

4. Z b Z b Z b
[f (x ) + g (x )] dx = f (x )dx + g (x )dx.
a a a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 17 / 51

 3 and 4 imply that
Z b
! Z b
N N

a
∑ [cn fn (x )] dx = ∑ cn
a
fn (x )dx ,
n =1 n =1

where fcn gN
n =1 are scalars.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 18 / 51

 5. Z b Z c Z b
f (x )dx = f (x )dx + f (x )dx.
a a c

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 19 / 51

 6. Z x
d
f (z )dz = f (x )
dx a
and Z b
d
f (z )dz = f (x ).
dx x

7. Let b > a. If f (x ) g (x ) for all x 2 [a, b ] , then

Z b Z b
f (x )dx g (x )dx.
a a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 20 / 51

 8. Z b Z b
f (x )dx jf (x )j dx for b > a.
a a

9. Cauchy-Schwarz Inequality:
Z b 2 Z b Z b
2
jf (x )g (x )j dx [f (x )] dx [g (x )]2 dx
a a a

or
Z b Z b 1/2 Z b 1/2
jf (x )g (x )j dx [f (x )]2 dx [g (x )]2 dx for b > a.
a a a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 21 / 51

 If the function f has a countable number of discontinuities on the
interval [a, b ] , then we perform the integral for each subinterval
where the function is continuous and then we sum all the resulting
integrals to obtain the integral on the whole interval [a, b ] .

Z b Z c Z d Z b
f (x )dx = f (x )dx + f (x )dx + f (x )dx
a a c d

Note that it does not matter if we count a point twice since the
integral on a single point is zero by property 2.
J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 22 / 51
4. Di¤erentiation of integrals

On the di¤erentiability and continuity of the integral

Z z
H (z ) f (t )dt, where z 2 [a, b ] .
a

1. If f is continuous at x, then H is di¤erentiable at x and

H 0 (x ) = f (x ).

2. If f is discontinuous at x, then H is not di¤erentiable at x.

3. The function H is continuous on [a, b ] ,

Z z Z x
lim H (z ) = lim f (t )dt = f (t )dt = H (x ), for all x 2 [a, b ] .
z !x z !x a a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 23 / 51

 ∂f (x, y )
Let f (x, y ) be a function such that the partial derivative
∂y
exists and is continuous. Then,
Z b Z b
d ∂f (x, y )
f (x, y )dx = dx.
dy a a ∂y
Therefore, one may interchange the integral and partial di¤erential
operators.
Proof.
Z b Rb Rb
d a
f (x, y + h)dx a
f (x, y )dx
f (x, y )dx = lim =
dy a h !0 h
Rb
[f (x, y + h) f (x, y )] dx
a
lim =
h !0 h
Z b Z b
f (x, y + h) f (x, y ) ∂f (x, y )
lim dx = dx.
a h !0 h a ∂y

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 24 / 51

 Leibniz rule. Let f (x, y ) be a function such that the partial
∂f (x, y )
derivative exists and is continuous, and a(y ) and b (y ) be
∂y
di¤erentiable functions. Then,
Z b (y )
d
f (x, y )dx =
dy a (y )
Z b (y )
∂f (x, y )
dx + f (b (y ), y ) b 0 (y ) f (a (y ), y ) a 0 (y ).
a (y ) ∂y

Proof. Since the variable y appears thrice in the integral

R b (y )
a (y )
f (x, y )dx, we apply the chain rule to obtain
R b (y ) R b (y ) ∂f (x ,y ) h R i
b (y )
d
dy a (y ) f ( x, y ) dx = a (y ) ∂y dx + d
db a (y ) f ( x, y ) dx b 0 (y )
h R i
d b (y )
+ da a (y )
f ( x, y ) dx a 0 (y ) =
Z b (y )
∂f (x ,y )
∂y dx + f (b (y ), y ) b 0 (y ) f (a (y ), y ) a 0 (y ).
a (y )

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 25 / 51

5. Integration rules: immediate integrals and integration
by parts
Immediate integrals:

R
f (x ) f (x )dx = F (x ) + C

0 C (an arbitrary constant)

x n +1
x n , with n 6= 1 +C
n+1

1 = x0 x +C

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 26 / 51

 R
f (x ) f (x )dx = F (x ) + C

1 1
=x , with x 6= 0 ln jx j + C
x

n
f 0 (x ) [f (x )] , with n 6= 1 [f (x )]n +1
+C
n+1

f 0 (x ) 1
= f 0 (x ) [f (x )] ,
f (x ) ln jf (x )j + C
with f (x ) 6= 0

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 27 / 51

 R
f (x ) f (x )dx = F (x ) + C

ex ex + C

f 0 ( x ) e f (x ) e f (x ) + C

ax
ax with a > 0 +C
ln a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 28 / 51

 R
f (x ) f (x )dx = F (x ) + C

sin x cos x + C

cos x sin x + C

tan x, with cos x 6= 0 ln jcos x j + C

ln x, with x > 0 x [(ln x ) 1] + C

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 29 / 51

 Examples:
1. Z Z
2 1/2 1 1/2
x x +4 dx = 2x x 2 + 4 dx
| {z } 2 |{z}| {z }
0 f (x )
f (x ) [f (x )]1/2
Z
" 3/2
#
1 1/2 1 x2 + 4
= 2x
|{z} x2 + 4 dx = +C
2 | {z } 2 3/2
0
f (x )
[f (x )]1/2
1 2 3/2
= x +4 + C.
3
2.
f 0 (x )
Z Z Z Z z }| {
sin x sin x sin x
(tan x ) dx = dx = dx = dx
cos
|{z}x cos x cos
|{z}x
f (x ) f (x )

= ln jcos x j + C , if cos x 6= 0.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 30 / 51

 3.
f (x ) f (x ) f (x )
Z z }| { Z z }| { Z z }| {
+ 1 + 1
e 3x 2 dx = 3 e 3x 2 dx = 3 e + 2 dx
3x
3 |{z} 3 |{z}
f 0 (x ) f 0 (x )

1
= e 3x +2 + C .
3

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 31 / 51

 Integration by parts:

Let F and G be the primitives of the continuous functions f and g ,

respectively. Then,

d [F (x ) G (x )]
= f (x ) G (x ) + F (x ) g (x ).
dx |{z} |{z}
F 0 (x ) G 0 (x )

Therefore, computing the inde…nite integral of both sides we get

Z Z
F (x ) G (x ) + C = f (x ) G (x )dx + F (x ) g (x )dx

or
Z Z
F (x ) g (x )dx = F (x ) G (x ) f (x ) G (x )dx + C .

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 32 / 51

 For de…nite integrals we have
Z b Z b
b
F (x ) g (x )dx = [F (x ) G (x )]a f (x ) G (x )dx.
a a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 33 / 51

 Example: Let x > 0,
2 3
Z Z
6 7
(ln x ) dx = 1 5 dx = ...
4(ln x ) |{z}
| {z }
F (x ) g (x )

1
Make F (x ) = ln x and g (x ) = 1 so that f (x ) = and G (x ) = x.
x
Z Z
1
... = (ln x ) |{z}
x x dx + C = (ln x ) x
|{z} 1dx + C
| {z } x
|{z}
F (x ) G (x ) G (x )
f (x )

= (ln x ) x x + C = x [(ln x ) 1] + C .

Moreover, the corresponding de…nite integral is

Z b
(ln x ) dx = b [(ln b ) 1] a [(ln a) 1] .
a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 34 / 51

6. Improper integrals
So far we have looked at integrals on a closed interval [a, b ] , where a
and b are real numbers.
Let us consider integrals of continuous functions on non-closed
intervals. These integrals are called improper.

Integral on the interval [a, ∞) :

Z ∞ Z b
f (x )dx lim f (x )dx.
a b !∞ a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 35 / 51

 Integral on the interval (∞, b ] :
Z b Z b
f (x )dx lim f (x )dx.
∞ a! ∞ a

Integral on the interval ( ∞, ∞) :

Z ∞ Z 0 Z b
f (x )dx lim f (x )dx + lim f (x )dx.
∞ a! ∞ a b !∞ 0

Integral on the right-semiclosed interval (a, b ] :

Z b Z b
f (x )dx lim+ f (x )dx, where z > a.
a+ z !a z

Integral on the left-semiclosed interval [a, b ) :

Z b Z z
f (x )dx lim f (x )dx, where z < b.
a z !b a

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 36 / 51

 Integral on the open interval (a, b ) :
Z b Z c Z z
f (x )dx lim f (x )dx + lim f (x )dx , with c 2 (a, b ) .
a+ z !a +
| z
{z } z|!b c
{z }
B A

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 37 / 51

 All the previous limits might fail to exist (i.e., they could be equal to
∞ ∞) or be equal to ∞. In the latter case, we say that the
improper integral "diverges".
Examples:
1.
Z 1
1
dx = lim+ [ln jx j]1z = ln 1 lim (ln jx j) = 0 ( ∞) = ∞,
0+ x z !0 z !0 +

so that the previous improper integral diverges.

2.
Z ∞
1
dx = lim [ln jx j]b1 = lim (ln jx j) ln 1 = ∞ 0 = ∞,
1 x b !∞ b !∞

so that the previous improper integral diverges.

3.
Z ∞ b
1 1 1 1
2
dx = lim = lim = 0 + 1 = 1.
1 x b !∞ x 1 b !∞ x 1

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 38 / 51

7. Economic applications of integration

Investment and the stock of capital.

Under discrete time, t = 0, 1, 2, ...

If Kt is the stock of capital at the beginning of period t, and It is the

amount of investment during period t, we have

Kt +1 = Kt + It or Kt +1 Kt = It .

Therefore,
t 1
Kt = K0 + ∑ Is .
s =0

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 39 / 51

 Under continuous time, t 2 [0, ∞) .

If K (t ) is the stock of capital at period t and I (t ) is the

instantaneous amount of investment during the time interval
(t, t + ∆t ), where ∆t is in…nitessimal, we have

K (t + ∆t ) K (t ) = I (t ) ∆t

or
K (t + ∆t ) K (t )
= I (t ),
∆t
so that
K (t + ∆t ) K (t )
lim = K 0 (t ) = I (t ).
∆t !0 ∆t
and, hence,
Z t Z t
K (t ) K (0) = I (s )ds or K (t ) = K (0) + I (s )ds.
0 0

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 40 / 51

 Present value of income ‡ows.

Under discrete time, t = 0, 1, 2, ...

If r 0 is the interest rate per period and Wt > 0 is the wealth at

the beginning of period t, we have

Wt + 1 = ( 1 + r ) W t or Wt + 1 Wt = rWt .

Therefore,
Wt
W t = ( 1 + r ) t W0 or W0 = ,
(1 + r )t

where the latter equality means that the present value (at date t = 0)
Wt
of having Wt euros at the future date t is euros.
(1 + r )t

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 41 / 51

 An implication of the previous formula is that the present value PV of
a stream of income fyt gTt=0 is

T
yt
PV = ∑ (1 + r )t .
t =0

If the time horizon of the stream is in…nite, T ! ∞, then

∞
yt
PV = ∑ (1 + r )t ,
t =0

where the previous sum could diverge.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 42 / 51

 Under continuous time, t 2 [0, ∞) .

If r 0 is the instantaneous interest rate, W (t ) is the wealth at date

t, and ∆t is an in…nitessimal time length, then we have

W (t + ∆t ) W (t ) = rW (t )∆t

or
W (t + ∆t ) W (t )
= rW (t ).
∆t

Thus,
W (t + ∆t ) W (t )
lim = W 0 (t ) = rW (t ),
∆t !0 ∆t

which can be written as

W 0 (t )
= r.
W (t )

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 43 / 51

 Finding the de…nite integral in both sides, we have
Z t Z t
W 0 (s )
ds = rds
0 W (s ) 0

so that
[ln W (s )]t0 = [rs ]t0 .
Thus,
W (t )
ln W (t ) ln W (0) = ln = rt,
W (0)
which implies that

W (t ) W (t )
= e rt or W (t ) = W (0)e rt or W (0) = = W (t )e rt
,
W (0) e rt

where the last equality says that the present value (at date t = 0) of
having W (t ) euros at the future date t is W (t )e rt euros.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 44 / 51

 An implication of the previous formula is that the present value PV of
a stream of income y (t ) for t 2 [0, T ] is
Z T
rt
PV = y (t )e dt.
0

If the time horizon of the stream is in…nite, T ! ∞, then

Z ∞
rt
PV = y (t )e dt,
0

where the previous improper integral may diverge.

Note that, if rc is the instantaneous interest rate in continuous time
and rd is the interest rate per period in discrete time, then

1 + rd = e rc

so that
rc = ln(1 + rd ) or rd = e rc 1.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 45 / 51

8. Integration with respect to several variables

Assume that f : Rn ! R is a continuous function.

The de…nite integral over a rectangle

A = [a1 , b1 ] [a2 , b2 ] ... [an 1 , bn 1] [an , bn ] is
Z
f (x1 , x2 , ..., xn 1 , xn ) d (x1 , x2 , ..., xn 1 , xn ) =
A
Z bn Z bn 1
Z b2 Z b1
.... f (x1 , x2 , ..., xn 1 , xn ) dx1 dx2 ...dxn 1 dxn =
an an 1 a2 a1
Z bn Z bn 1
Z b2 Z b1
.... f (x1 , x2 , ..., xn 1 , xn ) dx1 dx2 ...dxn 1 dxn ,
an an 1 a2 a1

with bi > ai , i = 1, 2, ...n.

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 46 / 51

Example:
y

A
c

a b x

Z
f (x, y ) d (x, y ) , where f (x, y ) = x y and A = [a, b ] [c, d ] .
A

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 47 / 51

 Z Z dZ b Z d Z b
f (x, y ) d (x, y ) = xydxdy = xydx dy
A c a c a
Z d Z b Z d b
x2
= y xdx dy = y dy
c a c 2 a

Z d Z d d
b2 a2 b2 a2 b2 a2 y2
= y dy = ydy =
c 2 2 2 2 c 2 2 2 c

b2 a2 d2 c2 b2 a2 d2 c2
= = .
2 2 2 2 4

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 48 / 51

Properties of the multiple integral (or integral with respect to
several variables) over a rectangle:

1.
Z bn Z b2 Z b1
... f1 (x1 ) f2 (x2 ) ... fn (xn )dx1 dx2 ...dxn =
an a2 a1 | {z }
f (x1 ,x2 ,...,xn )

Z b1 Z b2 Z bn
f1 (x1 ) dx1 f2 (x2 ) dx2 ... fn (xn ) dxn .
a1 a2 an

2. Fubini’s theorem:
Z dZ b Z bZ d
f (x, y )dxdy = f (x, y )dydx.
c a a c

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 49 / 51

 3.
Z xn Z x2 Z x1
∂n
... f (t1 , t2 , ..., tn 1 , tn ) dt1 dt2 ...dtn
∂x1 ∂x2 ...∂xn an a2 a1

= f (x1 , x2 , ..., xn 1 , xn ) .

Integral over a non-rectangular region:

Z
f (x1 , x2 , ..., xn 1 , xn ) d (x1 , x2 , ..., xn 1 , xn ) ,
A

where

A 6= [a1 , b1 ] [a2 , b2 ] ... [ an 1 , bn 1 ] [an , bn ] .

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 50 / 51

 Example:
Consider the following non-rectangular region A :
y

y=h(x)

A
y=g(x)

a b
x
Z Z b Z g (x )
f (x, y ) d (x, y ) = f (x, y ) dydx
a h (x )
A
Z b Z g (x )
= f (x, y ) dy dx.
a h (x )

J. Caballé (UAB - MOVE - BSE) Probability and Statistics IDEA 51 / 51

 Exercises. Probability and Statistics. IDEA.
A Primer on Integration

1. Let f be a continuous function on [a; b]. Prove the following integral version of
the mean value theorem:
Z b
There exists an " 2 [a; b] such that f (x)dx = f (")(b a):
a

2. Let us assume that f is a continuous function on [a; b] with f (x) 0 for all
x 2 [a; b] : Prove that
Z b
f (x)dx = 0
a

if and only if f (x) = 0 for all x 2 [a; b] :

3. Prove the Cauchy-Schwarz inequality: Let f and g continuous functions on

[a; b] ; then it holds that
Z b 2 Z b Z b
2
jf (x)g(x)j dx [f (x)] dx [g(x)]2 dx .
a a a

4. Taking into account that the term being integrated is a square and, thus, it is
positive, the following argument must be erroneous. Explain why.
Z 2 2
1 1
dx = = 1 1 = 2:
0 (x 1)2 x 1 0

5. Let f and g be two continuous and di¤erentiable functions on the real line.
Find the derivative of h(x) and the second derivative of k(x); where
Z x Rx
2
(a) h(x) = x f (t)g(t)dt (b) k(x) = e 0 f (t)g(t)dt .
0

6. Find the derivative of G(x) in the following cases:

Re
(a) G(x) = 1 ln (xt) dt
R 2x
(b) G(x) = 0
t3 dt

R x sin (xt)
(c) G(x) = x2
dt (x > 0) .
t

7. . Compute the following de…nite integrals:

Z ln 2 Z =2 Z 2
2x+3 sin x
(a) 8e dx (b) dx (c) x2 ln x dx.
0 0 1 + cos x 1
 8. Find the following primitives (or inde…nite integrals or antiderivatives):
Z Z Z
1
(a) x ln x dx (b) dx (c) sin3 x dx
tan x
Z p Z Z
x2 + 4 + x 2 2x
(d) p dx (e) x e dx (f) e3x cos 2x dx
x2 + 4

9. Compute the following integrals (some of them could be divergent or fail to

exist):
R1 x R1 2=3
(a) 1
dx (b) 0+
x dx
1 + x2
R0 R5 1
(c) 1
erx dx (analyze the result for all r) (d) 1
dx.
x 2

10. We have seen that the present value of an in…nite incomeRstream (or ‡ow) y(t)
1
discounted at a constant interest rate r > 0 is given by 0 y(t)e rt dt: Let us
assume that the income stream y(t) has the functional form y(t) = A (2t ) ; that
is, the income in t = 0 is A and it doubles in each time unit. For which values
of the interest rate the present value of this income ‡ow is …nite? Find the
present value in this case.

11. Compute the following double integrals on the speci…ed domains:

R
(a) (2x + 3y + 4) d (x; y) ; where D = [0; 1] [0; 2] , that is, 0 x 1 and 0 y 2:
D

R1
(b) d (x; y) ; where C is the region of R2 such that 0 x y; 0 y 1; x + y 1=2:
C y
R
(c) xy 2 d (x; y) ; where B is the region of R2 such that 0 x 1; x y x2 + 1:
B

12. It is known that Z 1

x2
p
e dx =
1

and Z 1
1 2
x
p
e 2 dx = 2 :
1
R1 x2
R1 1 2
x
Find 0
e dx and 0
e 2 dx:
 Probability and Statistics. IDEA. Answers.

A Primer on Integration

1. By the Weierstrass theorem, since f is continuous, f has a maximum M

and a minimum m on [a; b]. Thus,

Z b Z b Z b
mdx = m(b a) f (x)dx M dx = M (b a);
a a a

so that
Z b
1
m f (x)dx M:
b a a
1
Rb
Let K = b a a
f (x)dx. There exists an " 2 [a; b] such that f (") = K. Then,

multiplying by b a one gets the result.

2. If f (x) = 0 for all x, then the integral is obviously 0. In the other direction

we make the proof by contradiction, that is, we suppose that there exists an

y 2 [a; b], such that f (y) > 0. Then, since f is continuous, there exists an

interval [c; d] [a; b] around y such that f is strictly positive on that interval.

Then,
Z b Z d
f (x)dx f (x)dx = f (")(d c) > 0;
a c

by Exercise 1, where " 2 [c; d].

3. For every constant A and any pair of continuous functions f and g on [a; b],

we have
Z b
[A jf (x)j + jg(x)j]2 dx 0
a
and, thus,
Z b Z b Z b
2
2
A jf (x)j dx + 2A jf (x)j jg(x)j dx + jg(x)j2 dx 0
a a a

or
Z b Z b Z b
2
A 2
[f (x)] dx + A 2 jf (x)g(x)j dx + [g(x)]2 dx 0:
a a
| {z } | {z } | a {z }
c d h

This means that the above second order polynomial of A can have 1 or 0

roots. That is, the discriminant D = d2 4ch has to be less than or equal to

0. Therefore,
Z b 2 Z b Z b
2
2 jf (x)g(x)j dx 4 [f (x)] dx [g(x)]2 dx 0:
a a a

Rearranging the above inequality and dividing by 4, one obtains the Cauchy-

Schwarz inequality.

4. The function that is being integrated is discontinuous at x = 1 (in fact, it

is not …nite at x = 1): Then,

Z 2 Z 1 Z 2
1 1 1
dx = dx + dx =
0 (x 1)2 0 (x 1)2 1+ (x 1)2
1 b
lim (x 1) + lim [
0 a!1+
(x 1) 1 ]2a = lim (b 1) 1
1 1+ lim+ (a 1) 1
:
b!1 b!1 a!1

Note that

1 1 1
lim (b 1) = lim (1 b) = 1 and lim+ (a 1) = 1:
b!1 b!1 a!1

Therefore, the integral diverges (it tends to 1).

2
5. (a)
Z x
0
h (x) = 2x f (t)g(t)dt + x2 f (x)g(x):
0
Rx
(b) Let L(x) = 0
f (t)g(t)dt

k 0 (x) = L0 (x)eL(x) ;

and L0 (x) = f (x)g(x). Then,

Rx Rx
k 00 (x) = e 0 f (t)g(t)dt
[f (x)g(x)]2 + e 0 f (t)g(t)dt
[f 0 (x)g(x) + f (x)g 0 (x)]:

6. A straightforward application of the Leibniz’s rule gives the results:

Z e
0 1 e 1
(a) G (x) = dt = ; (b) G0 (x) = (2x)3 2 = 16x3 :
1 x x
sin xt
(c) Denote f (x; t) = . Then,
t
Z x
0 2
G (x) = f (x; x) 1 f (x; x ) 2x + fx (x; t)dt;
x2

where fx (x; t) = cos(xt) is the partial derivative of f with respect to t. Hence,

we get

sin x2 2 sin x3 sin x2 sin x3 2 sin x2 3 sin x3

G0 (x) = + = :
x x x x x x

R ln 2 R ln 2 R ln 2 1 8 R ln 2
7. (a) 0
8e2x+3 dx = 8e3 0
e2x dx = 8e3 0 2
2e2x dx = e3 0 2e2x dx
2
ln 2 2
= 4e3 [e2x ]0 = 4e3 e2 ln 2 e0 = 4e3 eln 2 e0 = 4e eln 4 e0
3

= 4e3 (4 1) = 12e3 :

3
 We have used here the fact that a primitive of f 0 (x)ef (x) is ef (x) :
R sin x
=2
(b) 0
dx = [ ln j1 + cos xj]0 =2 = [ ln (1 + cos x)]0 =2 = ln 2:
1 + cos x
Observe that 1 + cos x > 0 for x 2 [0; =2] :

(c) Integral by parts, with f (x) = x2 and G(x) = ln x:

2 Z 2 3
x3 x 1 8 7
ln x dx = ln 2 :
3 1 1 3 x 3 9

8. (a) Integrating by parts, f (x) = x and G(x) = ln x; to obtain the primitive,

Z
x2 x x2 1
ln x dx = ln x + C:
2 2 2 2
sin x 1 cos x
(b) Given that tan x = so that = , the integrand has the
cos x tan x sin x
f 0 (x)
form and the primitive is thus
f (x)

ln j sin xj + C;

which is well de…ned if sin x 6= 0:

f 0 (x)
We have used here the fact that the primitive of is ln jf (x)j + C:
f (x)
(c) We see that

sin3 x = sin x sin2 x = sin x 1 cos2 x = sin x sin x cos2 x:

Therefore, the primitive is simply

cos3 x
cos x + + C:
3
[f (x)]n+1
We have used here the fact that the primitive of f 0 (x) [f (x)]n is + C:
n+1
(d) Obvious, after dividing both the numerator and the denominator by
p
x2 + 4. The result is
p
x+ x2 + 4 + C;

4
 x 1=2 1 1=2
given that p = x (x2 + 4) = 2x (x2 + 4) is the derivative of
+4x2 2
1=2
p 1=2
2
1 (x + 4)
x2 + 4 = (x2 + 4) = :
2 1=2
We have also used here the fact that the primitive of f 0 (x) [f (x)]n is
[f (x)]n+1
+ C:
n+1
(e) Applying twice the method of integral by parts, that is, …rst G(x) = x2

and f (x) = e2x , and second G(x) = x and f (x) = e2x ; one gets the primitive

e2x 1
x2 x+ + C:
2 2

(f) Integrate by parts with G(x) = cos (2x) and f (x) = e3x . Then,
Z Z
3x e3x 2
e cos (2x) dx = cos (2x) + e3x sin (2x) dx:
3 3
R
Next, integrate by parts e3x sin (2x) dx similarly. Then,
Z Z
3x e3x 2 e3x 2
e cos (2x) dx = cos (2x) + sin (2x) e3x cos (2x) dx :
3 3 3 3
R
Solving the equation for e3x cos (2x) dx one gets
Z
3 3x 2
e3x cos (2x) dx = e cos (2x) + e3x sin (2x) + C:
13 13

9. (a)
b 0
1 1
lim ln 1 + x2 lim ln 1 + x2 = 1;
b!1 2 a! 1 2
0 a

so that the integral diverges.

(b)
h 1 i1
lim+ 3x 3 = 3
a!0 a

(c)
1 era 1
lim = ; if r > 0;
r a! 1 r r

5
and the integral diverges otherwise.

(d) The function is discontinuous at x = 2: In fact, it is not de…ned at

x = 2. Splitting the interval in two sub-intervals, we have:

Z 5 Z 5 Z 2
1 1 1
dx = dx + dx =
1 x 2 2+ x 2 1 x 2

ln 3 lim (ln ja 2j) + lim (ln jb 2j) ln 1 = 1

|{z} 1;
a!2+ b!2
0

which is not well de…ned.

10.
Z 1 h ib
A r t
A(2e r )t dt = lim 2e
0 ln 2 r b!1 0

r
If r > ln 2 (or 2e < 1), then the integral is equal to

A
;
r ln 2

and it diverges otherwise.

11. (a) Calculating the double integral, we get

Z 2 Z 1 Z 2
(2x + 3y + 4) dx dy = (3y + 5) dy = 16:
0 0 0

(b) The integral over region C is the sum of the integrals over regions A

and B:

6
 y

x=y

A
1
2
B
1
4 1
x= −y
2
1 1 x
4 2

Z Z Z1 Zy Z1=2 Zy
1 1 ln 2
f (x; y)d(x; y)+ f (x; y)d(x; y) = dxdy+ dxdy = 1 = 0:65343:
y y 2
A B 1=2 0 1=4 1
y
2

Note that
Z 1 Z y Z 1 y Z 1
1 1 1 1
dxdy = x dy = 1dy = [y]11=2 = 1 =
1
2
0 y 1
2
y 0 1=2 2 2

and
Z 1 Z Z 1 Z 1
" #
y y 1
2 1 2 1 2 y 2
y
dxdy = x dy = dy
1
4
1
2
y y 1
4
y 1
y
1
4
y y
2
Z 1 1=2
2 1 1 1 1
= 1 + 1 dy = 2y ln y = ln 2:
1
4
2y 2 1=4 2 2

(c) The integral can be simply written as

Z 1 Z x2 +1
67
xy 2 dydx = :
0 x 120

7
 x2
12. (a) Since e is a symmetric function with respect to 0, we get
Z 1 p
x2
e dx = :
0 2
1 2
x
Similarly, since e 2 is also a symmetric function with respect to 0, we get

Z 1
p r
1 2
x 2
e 2 dx = = :
0 2 2