LECTURE 2.

PROPERTIES OF MULTIPLE
INTEGRALS. FUBINI’S THEOREM

A course on Multivariable calculus

MATH - K72K SP TA
Hanoi National University of Education - 2024
Outlines

Summarizing the previous lecture

Sets with zero content

Integrals over more general subsets of Rn

Properties of multiple integrals

Iterated integrals and multiple integrals. Fubini’s theorem

Integrals over rectangles
Integrals over more general sets
Fubini’s theorem in Rn

Exercises
Summarizing the previous lecture

For a bounded set T the oscillation of function f on T denoted by

ωT (f ) is the following

ωT (f ) = sup f − inf f .
T T
Summarizing the previous lecture

Proposition
Let R be a rectangle in Rn and f be a bounded function
f : R → R. Then f is Riemann integrable on R if and only if
Z Z
f (x)dx = f (x)dx,
R R

or equivalently,

lim S(P) − s(P) = 0,
∥P∥→0
or X
lim ωRk (f )V (Rk ) = 0.
∥P∥→0
Rk ∈P
Sets with zero content
R
The next definition will enable us to establish the existence of R f (x)dx
in cases where f is bounded on the rectangle R, but is not necessarily
continuous for all x in R.
Definition
A subset E of Rn has zero content if for each ε > 0 there is a finite
set of rectangles T1 , T2 , . . . , Tm such that
m
[
E⊂ Tj
j=1

and
m
X
V (Tj ) < ε.
j=1
Example

1. Since the empty set is contained in every rectangle, the empty

set has zero content.
2. If E consists of finitely many points then E has zero content.
3. Any bounded set E with only finitely many limit points has
zero content.
4. If f is continuous on [a, b], then the curve

y = f (x), a≤x ≤b

(that is, the graph {(x, f (x)) a ≤ x ≤ b}), has zero content in
R2 .
The next lemma follows immediately from the definition.
Lemma
The union of finitely many sets with zero content has zero content.
A criterion for integrability

The following theorem will enable us to define multiple integrals over

more general subsets of Rn .
Theorem
Suppose that f is bounded on a rectangle

R = [a1 , b1 ] × [a2 , b2 ] · · · × [an , bn ]

and continuous except on a subset E of R with zero content. Then

f is integrable on R.
Proof

(Sketch) Denote by D the set of discontinuity of f in R. Since D

has zero content, there is a finite set of rectangles T1 , T2 , . . . , Tm
m
such that D ⊂ m
S P
j=1 T j and V (Tj ) < ε.
j=1

Let T be the union m

S
j=1 Tj . We can assume that D, T are strictly
contained in R. Denote the boundary ∂T by ℓ. The content of ℓ
is zero, evidently, therefore we can include ℓ inside another set T 1
which is also a union of rectangles, T 1 strictly ⊂ R with V (T 1 ) < ε.

Let d1 = dist(ℓ, ∂T 1 ) > 0. We note that if R is divided into rectan-

gles by a partition with the norm < d1 then the sum of the volumes
of those rectangle that have non-empty intersection with ℓ is less
than ε.
Proof (continued)
Now if we denote R1 := R \ int(T 1 ), then R1 is a closed set and f
is uniformly continuous on R1 . Therefore there exists d2 > 0 such
that the oscillation of f in every set from R1 with diameter < d2 is
less than ε.

Applying the similar argument as in the 1− dimensional case, we can

show that if P is a partition of R with the partition points generated
by the vertices of rectangles in T and T 1 , and P0 is the refinement
of P with ∥P0 ∥ < min{d1 , d2 } then
k
X
S(P0 ) − s(P0 ) = (Mj − mj )V (Rj ) ≤ [V (R) + 2(M − m)]ε,
j=1

where m and M are infimum and supremum, respectively, of f on

R, including the boundary. Thanks to the arbitrariness of ε, the last
inequality imply that f is integrable on R.
Integrals over more general subsets of Rn

We can now define the integral of a bounded function over more

general subsets of Rn .
Definition
Suppose that f is bounded on a bounded subset of S of Rn , and let
(
f (x), x ∈ S,
fS (x) = (1)
0, x∈/ S.

Let R be a rectangle containing S. Then the integral of f over S is

defined to be Z Z
f (x)dx = fS (x)dx
S R
Z
if fS (x)dx exists.
R
To see that this definition makes sense, we
Z must show that if R1 and
R2 are two rectangles containing S and fS (x)dx exists, then so
Z R1

does fS (x)dx and the two integrals are equal.

R2

Exercise: Prove the above statement.

Definition Z
If S is a bounded subset of Rn and the integral dx (with
Z S
integrand f ≡ 1) exists, we call dx the content (also, area if
S
n = 2 or volume if n = 3) of S, and denote it by V (S); thus,
Z
V (S) = dx.
S
Theorem
Suppose that f is bounded on a bounded set S and continuous
except on a subset E of S with zero content. Suppose also that ∂S
has zero content. Then f is integrable on S.
Proof

Let fS be as in (1). Since a discontinuity of fS is either a discontinuity

of f or a point of ∂S, the set of discontinuities of fS is the union of
two sets of zero content and therefore is of zero content. Therefore,
fS is integrable on any rectangle containing S, and consequently on
S
Jordan measurable sets

The content of the bounded subset S ⊂ Rn is the Jordan measure

of that set defined as Z
V (S) = dx
S
if the last integral exists. If V (S) exists then S is said to be Jordan
measurable.
Set of measure 0

Definition
A subset S of Rn has measure 0 if for each ε > 0 there is a
sequence of rectangles {Tj }∞
j=1 such that

∞
[
S⊂ Tj
j=1

and
∞
X
V (Tj ) < ε.
j=1

We write in that case: meas(S) = 0.

Theorem
Let S be a bounded and Jordan measurable set in Rn . Assume that
f : S → R is a bounded function such that the set of discontinuities
of f has measure 0. Then f is integrable on S.
Sketch of proof

- First we show that meas(∂S) = 0.

- Let R be a rectangle containing S. By considering function fS , we

see that D(fS ) ⊂ D(f ) ∪ ∂S. Therefore meas(D(fS )) = 0 too.

- Thus we can assume that S = R.

- For ε > 0, put Dε = {x ∈ R : ωf (x) ≥ ε}. Then Dε is a compact
subset of D(f ). Thus Dε has zero content.

- Repeating the previous proof for the result with set of discontinuities
of zero content, we can show that f is integrable on R by using the
criterion of integrability.
Proposition
If the bounded set S in Rn satisfies meas(∂S) = 0 then S is Jordan
measurable.
Exercise. Prove the above statement.
Properties of multiple integrals

We now list some theorems on properties of multiple integrals. The

proofs are similar to those of the analogous theorems for 1− dimen-
sional case.
Theorem
If f and g are integrable on S, then so is f + g , and
Z Z Z
(f + g )(x)dx = f (x)dx + g (x)dx.
S S S

Theorem
If f is integrable on S and c is a constant; then cf is integrable on
S, and Z Z
(cf )(x)dx = c f (x)dx.
S S
Properties of multiple integrals

Theorem
If f and g are integrable on S and f (x) ≤ g (x) for all x ∈ S, then
Z Z
f (x)dx ≤ g (x)dx.
S S

Theorem
If f is integrable on S, then so is |f |, and
Z Z
f (x)dx ≤ |f (x)|dx.
S S

Theorem
If f and g are integrable on S, then so is the product fg .
Properties of multiple integrals

Lemma
Suppose that S is contained in a bounded set T and f is integrable
on S. Then fS is integrable on T , and
Z Z
fS (x)dx = f (x)dx.
T S
Proof of the lemma
We have
(
fS (x), x ∈ T,
(fS )T (x) =
0, x∈/ T.

Since S ⊂ T , (fS )T = fS . Suppose that R is a rectangle containing

T . Then R also contains S, so
Z Z
f (x) dx = fS (x) dx
S ZR

= (fS )T (x) dx
ZR
= fS (x) dx.
T
Properties of multiple integrals

Theorem
If f is integrable on disjoint sets S1 and S2 , then f is integrable on
S1 ∪ S2 , and
Z Z Z
f (x)dx = f (x)dx + f (x)dx. (2)
S1 ∪S2 S1 S2
Proof of Theorem 18

For i = 1, 2 let
(
f (x), x ∈ Si ,
fSi (x) =
0, x∈/ Si .

From Lemma 17 with S = Si and T = S1 ∪ S2 , fSi is integrable on

S1 ∪ S2 , and
Z Z
fSi (x)dx = f (x)dx, i = 1, 2.
S1 ∪S2 Si
Proof of Theorem 18

Theorem 12 now implies that fS1 + fS2 is integrable on S1 ∪ S2 and

Z Z Z
(fS1 + fS2 )(x)dx = f (x)dx + f (x)dx. (3)
S1 ∪S2 S1 S2

Since S1 ∩ S2 = ∅,

(fS1 + fS2 )(x) = fS1 (x) + fS2 (x) = f (x), x ∈ S1 ∪ S2 .

Therefore (3) implies (2).

Corollary
Suppose that f is integrable on sets S1 and S2 such that S1 ∩ S2
has zero content: Then f is integrable on S1 ∪ S2 , and
Z Z Z
f (x)dx = f (x)dx + f (x)dx.
S1 ∪S2 S1 S2
Theorem
If f is integrable on B, A ⊂ B such that A is Jordan measurable,
then f is integrable on A.

Proof.
Suppose that f : B → R is integrable, A ⊂ B. If A is Jordan
measurable then the function χA is integrable on B. By Theorem
16 the function fA = χA f is integrable on B. Therefore f is
integrable on A.
Lebesgue’s theorem

Theorem
Let S be a bounded and Jordan measurable set in Rn . Assume that
f : S → R is a bounded function. Then f is integrable on S if and
only if the set of discontinuities of f has measure 0.
Proof - the necessary condition

- Take R ⊃ S. Then χS f is integrable on R and D(f ) ⊆ D(χS f ).

Thus, we need only to show that meas(D(χS f )) = 0.

-Therefore, it is sufficient to consider the case S = R.

- Denote Dk (f ) = {x ∈ R ωf (x) ≥ k1 }. Then we have

∞
[
D(f ) = Dk (f ).
k=1

- We will show that meas(Dk (f )) = 0, ∀k ≥ 1.

But it is clear from the criterion of integrability. For every ε > 0,
there exists δ > 0 s.t
X ε
wT V (T ) < , ∀P ∈ F(R) : ∥P∥ < δ.
k
T ∈P

Thus X ε
wT V (T ) < .
k
T ∈P,T ∩Dk (f )̸=∅

If T ∩ Dk (f ) ̸= ∅, then ∃x ∈ T ∩ Dk (f ). Therefore wf (x) ≥ k1 . It

implies wT ≥ k1 .
P
It follows that T ∈P,T ∩Dk (f )̸=∅ V (T ) < ε. By definition we see
that meas(Dk (f )) = 0, ∀k ≥ 1.
Iterated integrals and multiple integrals. Fubini’s theorem

It is impractical to evaluate multiple integrals directly from the defi-

nitions. Fortunately, this can usually be accomplished by evaluating
n successive ordinary integrals. Let’s begin with the case n = 2.
Integrals over rectangles

Theorem
(Fubini’s theorem) Suppose that f is integrable on
R = [a, b] × [c, d] and
Z b
F (y ) = f (x, y )dx
a

exists for each y in [c, d]. Then F is integrable on [c, d], and
Z d Z
F (y )dy = f (x, y )dxdy ; (4)
c R

that is Z d Z b  Z
f (x, y )dx dy = f (x, y )dxdy . (5)
c a R
Proof of Fubini’s theorem in the case n = 2

Let

P1 : a = x0 < x1 < · · · < xr = b and P2 : c = y0 < y1 < · · · < ys = d

be partitions of [a, b] and [c, d], and P = P1 × P2 . Suppose that

yj−1 < ηj < yj , 1 ≤ j < s, (6)

so
s
X
σ= F (ηj )(yj − yj−1 ) (7)
j=1

is a typical Riemann sum of F over P2 .

Proof of Fubini’s theorem in the case n = 2 (continued)

Since Z b r Z
X xi
F (ηj ) = f (x, ηj )dx = f (x, ηj )dx,
a i=1 xi−1

(6) implies that if

mij = inf{f (x, y ) xi−1 ≤ x ≤ xi , yj−1 ≤ y ≤ yj }

and

Mij = sup{f (x, y ) xi−1 ≤ x ≤ xi , yj−1 ≤ y ≤ yj },

then
r
X r
X
mij (xi − xi−1 ) ≤ F (nj ) ≤ Mij (xi − xi−1 ).
i=1 i=1
Proof of Fubini’s theorem in the case n = 2 (continued)

Multiplying this by yj − yj−1 and summing from j = 1 to j = s yields

s X
X r s
X
mij (xi − xi−1 )(yj − yj−1 ) ≤ F (nj )(yj − yj−1 )
j=1 i=1 j=1
s X
X r
≤ Mij (xi − xi−1 )(yj − yj−1 ),
j=1 i=1

which, from (7), can be rewritten as

sf (P) ≤ σ ≤ Sf (P) (8)

where sf (P) and Sf (P) are the lower and upper sums of f over P.
Proof of Fubini’s theorem in the case n = 2 (continued)

Now let sF (P2 ) and SF (P2 ) be the lower and upper sums of F over
P2 ; since they are respectively the infimum and supremum of the
Riemann sums of F over P2 , (8) implies that

sf (P) ≤ sF (P2 ) ≤ SF (P2 ) ≤ Sf (P). (9)

Proof of Fubini’s theorem in the case n = 2 (continued)

Since f is integrable on R, there is for each ε > 0 a partition P

of R such that Sf (P) − sf (P) < ε. Consequently, from (9) , there
is a partition P2 of [c, d] such that SF (P2 ) − sF (P2 ) < ε, so F
is integrable on [c, d] by the corresponding existence result of 1−
dimensional definite integral. Rd
It remains to verify (4). From (7) and the definition of c F (y )dy ,
there is for each ε > 0 a δ > 0 such that
Z d
F (y )dy − σ < ε if ∥P2 ∥ < δ;
c

that is,
Z d
σ−ε< F (y )dy < σ + ε if ∥P2 ∥ < δ.
c
Proof of Fubini’s theorem in the case n = 2 (continued)

This and (8) imply that

Z d
sf (P) − ε < F (y )dy < Sf (P) + ε if ∥P∥ < δ,
c

and this implies that

Z Z d Z
f (x, y )dxdy − ε ≤ F (y )dy ≤ f (x, y )dxdy + ε. (10)
R c R
Proof of Fubini’s theorem in the case n = 2 (continued)

Since Z Z
f (x, y )dxdy = f (x, y )dxdy
R R

and ε can be made arbitrarily small, (10) implies (4).

If f is continuous on R, then f satisfies the hypotheses of Theorem
22 so (5) is valid in this case.
Corollary
If f is integrable on [a, b] × [c, d], then
Z b Z d  Z d Z b 
f (x, y )dy dx = f (x, y )dx dy ,
a c c a
Rd Rb
provided that c f (x, y )dy exists for a ≤ x ≤ b and a f (x, y )dx
exists for c ≤ y ≤ d. In particular; these hypotheses hold if f is
continuous on [a, b] × [c, d].
 Rd Rb 
The converse statement: "if c a f (x, y )dx dy exists then so
R
does R f (x, y )dxdy " is not necessary true.
Example
If f is defined on R = [0, 1] × [0, 1] by
(
2xy if y is rational,
f (x, y ) =
y if y is irrational,

then Z 1
f (x, y )dx = y , 0 ≤ y ≤ 1,
0
and
1 1 1
1
Z Z Z
dy f (x, y )dx = ydy = .
0 0 0 2
However, f is not integrable on R.
The next theorem generalizes Theorem 22 to Rn .
Theorem
Let I1 , I2 , . . . , In be closed intervals and suppose that f is integrable
on R = I1 × I2 × · · · × In . Suppose that there is an integer
p ∈ {1, 2, . . . , n − 1} such that
Z
Fp (xp+1 , xp+2 , . . . , xn ) = f (x1 , x2 , . . . , xn )d(x1 , x2 , . . . , xp )
I1 ×I2 ×···×Ip

exists for each (xp+1 , xp+2 , . . . , xn ) ∈ Ip+1 × Ip+2 × · · · × In . Then

Z
Fp (xp+1 , xp+2 , . . . , xn )d(xp+1 , xp+2 , . . . , xn )
Ip+1 ×Ip+2 ×···×In
R
exists and equals R f (x)dx.
Corollary
R
If f is continuous on R = I1 × I2 × · · · × In then R f (x)dx can be
evaluated by n! iterated integrals.
The following theorem can be proved by induction on n.
Theorem
Let Ij = [aj , bj ], 1 ≤ j ≤ n and suppose that f is integrable on
R = I1 × I2 × · · · × In . Suppose also that the integrals
Z
Fp (xp+1 , xp+2 , . . . , xn ) = f (x1 , x2 , . . . , xn )d(x1 , x2 , . . . , xp ),
I1 ×I2 ×···×Ip
1 ≤ p ≤ n − 1,

exist for all (xp+1 , xp+2 , . . . , xn ) ∈ Ip+1 × Ip+2 × · · · × In . Then the

iterated integral
Z bn Z bn−1 Z b2 Z b1
dxn dxn−1 · · · dx2 dx1
an an−1 a2 a1
R
exists and equals R f (x)dx.
Integrals over more general sets
We now consider the problem of evaluating multiple integrals over
more general sets.
First, suppose that f is integrable on a set of the form

S = {(x, y ) u(y ) ≤ x ≤ v (y ), c ≤ y ≤ d}, (11)

(Figure 1).

Figure:
If u(y ) ≥ a and v (y ) ≤ b for c ≤ y ≤ d, and fS := χS f then
Z Z
f (x, y )dxdy = fS (x, y )dxdy ,
S R

where R = [a, b] × [c, d]. From Theorem 22,

Z Z d Z b
fS (x, y )dxdy = dy fS (x, y )dx
R c a
Rb
provided that a fS (x, y )dx exists for each y in [c, d].
From (11), this integral can be written as
Z v (y )
f (x, y )dx. (12)
u(y )

Thus, we have proved the following theorem.

Theorem
If f is integrable on the set S in (11) and the integral (12) exists
for c ≤ y ≤ d, then
Z Z d Z v (y )
f (x, y )dxdy = dy f (x, y )dx. (13)
S c u(y )
Interchanging x and y in Theorem 28 shows that if f is integrable
on

Figure:
 S = {(x, y ) u(x) ≤ y ≤ v (x), a ≤ x ≤ b} (14)
(Figure 2) and
Z v (x)
f (x, y )dy
u(x)

exists for a ≤ x ≤ b, then

Z Z b Z v (x)
f (x, y )dxdy = dx f (x, y )dy . (15)
S a u(x)
Fubini’s theorem in Rn

In general dimension we have an analogous result.

Suppose that B is a bounded set in Rn . For each ξ ′ ∈ Rn−1 and
ξn ∈ R put

Bξ′ := {xn ∈ R (ξ ′ , xn ) ∈ B} = (ξ ′ × R) ∩ B

and

Bξn := {x ′ ∈ Rn−1 (x ′ , ξn ) ∈ B} = (Rn−1 × ξn ) ∩ B.

Let
B ′ = {x ′ ∈ Rn−1 Bx ′ ̸= ∅}.
In other words, B ′ is the projection of B on Rn−1 - the space of
(n − 1) first coordinates.
Fubini’s theorem in Rn

Similarly, let
Bn = {xn ∈ R Bxn ̸= ∅}
- the projection of B on R, the space of the n -th coordinate. Obvi-
ously,
Bx ′ ̸= ∅ ⇐⇒ x ′ ∈ B ′
and
Bxn ̸= ∅ ⇐⇒ xn ∈ Bn .
Fubini’s theorem in Rn

Theorem
(Fubini’s theorem) Let B be a bounded set in Rn and suppose that
f is integrable on B. If for each x ′ ∈ B ′ the function f (x ′ , xn ) is
integrable on Bx ′ , then the function
Z
′
F (x ) = f (x ′ , xn )dxn
Bx ′

is integrable on B ′ and
Z Z
′ ′
F (x )dx = f (x)dx.
B′ B
Problem 1

Evaluate
Z
a) (2x + y 2 ) dxdy ; S = {(x, y ) 0 ≤ x ≤ 9 − y 2 , −3 ≤ y ≤ 3}.
ZS
b) 2xy dxdy ; S is bounded by y = x 2 and x = y 2 .
Z S
sin y
c) ex dxdy ; S = {(x, y ) ln y ≤ x ≤ ln 2y , π/2 ≤ y ≤ π}.
S y
Problem 2

Let R := [a1 , b1 ] × [a2 , b2 ] × · · · [an , bn ]. Evaluate

Z
a) (x1 + x2 + · · · + xn ) dV
ZR
b) (x12 + x22 + · · · + xn2 ) dV
ZR
c) x1 x2 · · · xn dV .
R
Problem 3

Assuming that f is continuous, express

Z 1 Z √ 1−y 2
dy √ f (x, y ) dx
1/2 − 1−y 2

as an iterated integral with the order of integration reversed.

Problem 4

√
Z 1Z 1−x 2
dy
Evaluate p .
0 0 x + y2
2
Problem 5
Let Tρ = [0, ρ] × [0, ρ], ρ > 0. By calculating
Z
I (a) = lim e −xy sin ax dxdy
ρ→∞ T
ρ

by two different ways, show that

Z ∞
sin ax π
dx = if a > 0.
0 x 2

Hint: Recall that for a ̸= 0, b ̸= 0

a sin bx − b cos bx ax
Z
e ax sin bx dx = e + C.
a2 + b 2