January 24, 2020

IV.1 MULTIPLE INTEGRALS

The Multiple integral of Riemann is a straightforward extension of the simple integral.

1. Integration in intervals of Rp .

Definition 1.1. A compact interval of Rp is a set I = [a1 , b1 ] × · · · × [ap , bp ], where each

[ai , bi ] is a compact interval of R, for i = 1, . . . , m.

The measure of the compact interval I is µ(I) = (b1 − a1 ) · · · (bp − ap ).

A finite collection P = {I1 , . . . , In } of compact intervals is a partition of I if (i) I =

◦ ◦
I1 ∪ · · · ∪ In and (ii) Ii ∩ Ij = ∅ for i 6= j.

Given two partitions P 0 and P of I, P 0 is finer than P if every interval of P 0 is contained

in some interval of P.

The diameter of the partition P of I is the larger of the lengths bi − ai , for i = 1, . . . , n,

and will denoted by |P|.

Definition 1.2. Let f : I → R be a bounded function defined in the compact interval

I ⊆ Rp . Given a partition P = {I1 , . . . , In }, let

mi = inf f (Ii ), Mi = sup f (Ii )

and let µ(Ii ) be the measure of Ii , for i = 1, . . . , n. The lower and the upper (Darboux)
sums of f in the partition P are
p
X p
X
s(f, P) = mi µ(Ii ), S(f, P) = Mi µ(Ii ),
i=1 i=1
respectively.

Proposition 1.3 (Properties of Darboux sums). Let f : I → R be a bounded function

defined in the compact interval I ⊆ Rp . Let

m = inf f (I), M = sup f (I).

Let P, P 0 two partitions of I.

(1) mµ(I) ≤ s(f, P) ≤ S(f, P) ≤ M µ(I).

(2) If P 0 is finer than P, then

s(f, P) ≤ s(f, P 0 ) ≤ S(f, P 0 ) ≤ S(f, P).

1
2 IV.1 MULTIPLE INTEGRALS

(3) s(f, P 0 ) ≤ S(f, P).

(4) The sets

{s(f, P) : P is a partition of I} and {S(f, P) : P is a partition of I}

are bounded.

Definition 1.4. Let f : I → R be a bounded function defined in the compact interval

I ⊆ Rp .

The lower integral of f in I is defined as the number

Z
L f = sup{s(f, P) : P is a partition of I}.
I

The upper integral of f in [a, b] is defined as the number

Z
U f = inf{S(f, P) : P is a partition of I}.
I

Proposition 1.5. Let f : I → R be a bounded function defined in the compact interval

I ⊆ Rp . Then Z Z
L f ≤U f.
I I

Definition 1.6. Let f : I → R be a bounded function defined in the compact interval

I ⊆ Rp .

We say that the function is Riemann integrable (or simply integrable) in the interval I ⊆ Rp
iff Z Z
L f =U f.
I I
In this case, this number is the integral of f in I (or defined integral of f in I) and is
denoted
Z Z Z Z
p
f, f (x)dx = ··· f (x1 , . . . , xp )dx1 · · · dxp .
I I [a1 ,b1 ]×···×[ap ,bp ]

Not every bounded function is integrable.

Theorem 1.7. Let f : I → R be a bounded function defined in the compact interval I ⊆ Rp .

Thenf is integrable iff for all ε > 0, there exists a partition P of I such that

S(f, P) − s(f, P) < ε.

Proposition 1.8. If f is continuous in I, then f is integrable in I.

This result admits a useful generalization.

 IV.1 MULTIPLE INTEGRALS 3

Proposition 1.9. If f is bounded in I and has a finite number of discontinuities, then f

is integrable in I.

2. Properties of the Riemann Integral

Proposition 2.1 (Properties of the integral). Let f, g : I → R be integrable, where I ⊆ Rp

is a compact interval. Let α ∈ R.

(1) Linearity.
R R R
(a) I (f + g) = I f + I g.
R R
(b) I αf = α I f .

(2) Monotonicity.
Z Z
f (x) ≥ g(x) for all x ∈ I, implies gf ≥ g.
I I
Z  Z
 
(3)  f (x)dx ≤ |f (x)|dx.

I I

(4) Additivity with respect to the interval. Let {I1 , . . . , In } be a partition of I. Then f
is integrable in I if and only if f is integrable in I1 , . . . , In , and in this case
Z Z Z
f= f + ··· + f.
I I1 In

Proposition 2.2 (Theorem of the mean). Let f : I → R be a bounded function defined in

the compact interval I ⊆ Rp .

(1) If f is integrable in I and if m and M are lower and upper bounds of f in I,

respectively (they could be the infimum and the supremum), then there is α ∈ [m, M ]
such that Z
f (x) = αµ(I).
I

(2) If f is continuous in I, then there exists c ∈ I such that

Z
f (x) = f (c)µ(I).
I

Remark 2.3. Let f : I → R be a bounded function defined in the compact interval I ⊆ R2

such that f (x, y) ≥ 0 for all (x, y) ∈ I. Let the hypograph of f

H(f ) = {(x, y, z) ∈ R3 : (x, y) ∈ I, 0 ≤ z ≤ f (x, y)}.

The lower (uuper) Darboux sum s(f, P) (S(f, P)) is the sum of volumes of rectangular
prisms with basis Ii and height mi (Mi ). Hence, the lower (upper) Darboux sum is an
4 IV.1 MULTIPLE INTEGRALS

underestimation (overestimation) of the volume of the set H(F ). Since both Darboux sums
tend to the same number when the diameter of the partitions tend to zero, we say that the
integral of f in I is the volume of H(f ).

3. Iterated integrals

In this section we show how the integral of a function f : I → R on a compact in-

terval I = [a1 , b1 ] × · · · × [ap , bp ] can be expressed as p simple integrals in the intervals
[a1 , b1 ], . . . , [ap , bp ].

Remark 3.1. Given x = (x1 , . . . , xp ), and i ∈ {1, . . . , p}, x−i denotes the vector of di-
mension p − 1 obtained from x after eliminating the coordinate xi . In the same way, for
a compact interval I = [a1 , b1 ] × · · · × [ap , bp ], I−i denotes the Cartesian product of all the
individual intervals that form I, except the interval [ai , bi ]. We will write x = (xi |x−i ) and
I = (Ii |I−i ).

For instance, if x = (0, −1, 5), then x−1 = (−1, 5), x−2 = (0, 5) and x−3 = (0, −1). If
I = [0, 1] × [−2, −1], then I1 = [0, 1] and I−1 = [−2, −1].

Theorem 3.2 (Fubini’s Theorem in compact intervals). Let f : I → R be bounded in I,

and suppose that for all x−i ∈ I−i , the function xi 7→ f (xi |xi ) is integrable in [ai , bi ], for all
i = 1, . . . , p. Then f : I → R is integrable in I and
Z b1 Z b2 !!!
Z Z bp
f (x1 , . . . , xp )dx1 . . . dxp = dx1 dx2 ··· f (x1 , . . . , xp )dxp .
I a1 a2 ap

Remark 3.3. The theorem assert that the order of computation of the iterated integrals
does not matter to find the integral. In the case p = 2, the theorem means that
ZZ Z b1 Z b2 Z b2 Z b1
f (x, y)dxdy = dx f (x, y)dy = dy f (x, y)dx.
I a1 a2 a2 a1

Example 3.4. Calculate

ZZ
e−xy xdxdy, I = [0, 1] × [−1, 0].
I

Solution: Applying Fubini’s Theorem, we can compute the integral by iterated integration.
ZZ Z 1 Z 0 Z 1 Z 1
−xy −xy

0
−xy 
1
e xdxdy = dx e xdy = −e  dx = (ex −1)dx = ex −x = e−2.

I 0 −1 0 −1 0 0

Also, ZZ Z 0 Z 1
−xy
e xdxdy = dy e−xy xdx.
I −1 0
The inner integral is not immediate. It can be calculated by using parts.
 IV.1 MULTIPLE INTEGRALS 5

Theorem 3.5 (Fubini’s Theorem in simple regions of R2 ). (1) (type 1 regions). Let ϕ1 , ϕ2
be two continuous functions in [a, b] such that ϕ1 (x) ≤ ϕ2 (x) for all x ∈ [a, b]. Let
the compact set

S = {(x, y) : a ≤ x ≤ b, ϕ1 (x) ≤ y ≤ ϕ2 (x)}

If f is integrable in S, then
Z b !
Z Z ϕ2 (x)
f (x, y)dxdy = dx f (x, y)dy .
S a ϕ1 (x)

(2) (type 2 regions) Let φ1 , φ2 be two continuous functions in [c, d] such that φ1 (y) ≤
φ2 (y) for all y ∈ [c, d]. Let the compact set

T = {(x, y) : c ≤ y ≤ d, φ1 (y) ≤ x ≤ φ2 (y)}

If f is integrable in T , then
!
Z Z d Z φ2 (y)
f (x, y)dxdy = dy f (x, y)dx .
T c φ1 (y)

Example 3.6. Calculate

ZZ
(x2 + y)dxdy, S = {(x, y) : −1 ≤ x ≤ 1, x2 ≤ y ≤ 1}.
S

Solution:
1 1 1 1
y 2 
ZZ Z Z  Z
2 22
(x + y)dxdy = dx (x + y)dy = x y+ dx
S −1 x2 −1 2  x2
Z 1 
x4
  
2 1 4 16
= x + − x + dx =
−1 2 2 15
RR
Example 3.7. Calculate A f (x, y)dxdy, where A is the triangular region of vertex (0, 0),
x
(1, 0) and (1, 1) and f (x, y) = e y if y > 0, f (x, 0) = 0 for all 0 ≤ x ≤ 1.

Solution:

Note that f is continuous in A − {(0, 0)}. Since that there is only a point of discontinuity
in the region A, f is integrable.

Note that A = {(x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ x} = {(x, y) : 0 ≤ y ≤ 1, y ≤ x ≤ 1}

First method: fix x and integrate firstly in y and then in x.

Z 1 Z x Z 1 Z 1 2
e−1
ZZ
y y y x
 x
e dxdy =
x dx e dy =
x xe  dx =
x (e − 1)dx =
A 0 0 0 0 0 2 2

Second method: fix y and integrate firstly in x and then in y.

6 IV.1 MULTIPLE INTEGRALS

ZZ
y
Z 1 Z 1 y
e dxdy =
x dy e x dx
A 0 y

is impossible to compute

4. Areas and volumes

Definition 4.1. Let ϕ1 , ϕ2 be two continuous functions in [a, b] such that ϕ1 (x) ≤ ϕ2 (x)
for all x ∈ [a, b]. Let the compact set

S = {(x, y) : a ≤ x ≤ b, ϕ1 (x) ≤ y ≤ ϕ2 (x)}.

The area of S is
!
ZZ Z b Z ϕ2 (x) Z b
dxdy = dx dy = (ϕ2 (x) − ϕ1 (x))) dx.
S a ϕ1 (x) a

Definition 4.2. Let φ1 , φ2 be two continuous functions in [c, d] such that φ1 (y) ≤ φ2 (y)
for all y ∈ [c, d]. Let the compact set

T = {(x, y) : c ≤ y ≤ d, φ1 (y) ≤ x ≤ φ2 (y)}.

The area of T is
!
ZZ Z d Z φ2 (y) Z d
dxdy = dy dx = (φ2 (y) − φ1 (y)) dy.
T c φ1 (y) c

Definition 4.3. Let A be a subset of R2 of type 1 or type 2, that is A = S or A = T .

Let f : A → R+ be integrable and nonnegative. The volume of the solid of R3 limited by
the plane z = 0, the cylinder generated by the generatrix parallel to the vertical axis z and
directrix the boundary of the set A, and the surface of equation z = f (x, y) is
ZZ
f (x, y)dxdy.
A

5. Change of variable

The formula of change of variable in simple integrals is an straightforward application of

the chain rule. The generalization of the formula to two or more variables is not easy. We
will establish only the formula for double integrals.
 IV.1 MULTIPLE INTEGRALS 7

Theorem 5.1. Let D be a subset of R2 which is of type 1 or of type 2. Let U be an open

set such that D ⊆ U , and let g : U → R2 be a transformation of class C 1 , one to one and
◦
such that the Jacobian of g is non null for all (x, y) ∈D.

Let f : A → R be a continuous function in A = g(D). Then

ZZ ZZ
f (x, y)dxdy = f (g(u, v))| det Jg(u, v)|dudv.
A D

Remark 5.2. The transformation g is the change of variable. It is often used for g the
notation (x, y) = g(u, v), where

x = x(u, v),
y = y(u, v),

(u, v) ∈ U . The Jacobian matrix of g is the matrix

!
∂x ∂x
∂u ∂v
Jg(u, v) = ∂y ∂y
.
∂u ∂v

Note that the Jacobian determinant enters the integral in absolute value.

Example 5.3. Calculate the integral

ZZ
(2y − x)dxdy,
A

where A is the region of the plane enclosed by the lines y = x + 1, y = x + 5, y = −x + 1

and y = −x + 3.

Solution: Let the change of variables u = y − x and v = y + x. Then x = 21 (v − u) and

y = 21 (u + v). The transformation is thus g(u, v) = ( 21 (v − u), 21 (u + v)). Note that if we let
D = [1, 5] × [1, 3], then g(D) = A. Moreover, g is clearly one-to-one. The Jacobian of g is

 
 −1 1  1
det Jg(u, v) =  21 2
=− .
 
 1  2
2 2
8 IV.1 MULTIPLE INTEGRALS

Finally, note that 2y − x = u + v − 12 (v − u) = 23 u + 12 v, hence

ZZ ZZ  
1 3 1
(2y − x)dxdy = u + v dudv
A D 2 2 2
ZZ ZZ
3 1
= ududv + vdudv
4 D 4 D

3 5 1 3
Z Z
= 2udu + 4vdv
4 1 4 1
3 2 5 1 2 3
= u  + v 
4 1 2 1
3 1
= (25 − 1) + (9 − 1) = 22.
4 2

5.1. Polar coordinates. Let g : (0, ∞)×[0, 2π) → R2 be the transformation (ρ, θ) 7→ (x, y)
be defined by
x = ρ cos θ, y = ρ sin θ,
which is of class C ∞ and the Jacobian is
 
 cos θ −ρ sin θ 
det Jg(u, v) =   = ρ cos2 θ + ρ sin2 θ = ρ(cos2 θ + sin2 θ) = ρ.
 
 sin θ ρ cos θ 

Let two sets A and D such that A = g(D) and g is a bijection. Let f : A → R such that
the conditions of the Theorem of the change of variable are fulfilled. Then
ZZ ZZ
f (x, y)dxdy = f (ρ cos θ, ρ sin θ)ρdρdθ.
A D

Example 5.4. Calculate the volume V of the sphere of R3 of radius r > 0.

Solution. By symmetry, the volume V is twice the volume of the set

p
B = {(x, y, z) : 0 ≤ z ≤ (r2 − x2 − y 2 ), x2 + y 2 ≤ r2 } ⊆ R3 .

The volume of B is ZZ p
(r2 − x2 − y 2 )dxdy,
A
where A = {(x, y) ∈ R2 : x2 + y 2 ≤ r2 }. Change to polar coordinates to see that the set A
in polar coordinates is D = {(ρ, θ) : 0 ≤ ρ ≤ r, 0 ≤ θ < 2π}. Then
ZZ p
V =2 ρ r2 − ρ2 dρdθ
[0,r]×[0,2π]
Z r p
=2 2πρ r2 − ρ2 dρ
0
 r  4
2 2 2 32 
= 2π − (r − ρ )  = πr3 .
3 0 3