Lecture 10

We begin today’s lecture with a simple claim.

Claim. Let Q ⊆ Rn be a rectangle and f, g : Q → R be bounded functions such that
f ≤ g. Then � �
f≤ g. (3.49)
Q Q

Proof. Let P be a partition of Q, and let R be a rectangle belonging to P . Clearly,

mR (f ) ≤ mR (g), so
�
L(f, P ) = mR (f )v(R) (3.50)
R
�
L(g, P ) = mR (g)v(R) (3.51)
R
�
=⇒ L(f, P ) ≤ L(g, P ) ≤ g, (3.52)
Q

for all partitions P . The lower integral

�
f (3.53)
Q

is the l.u.b. of L(f, P ), so � �

f≤ g. (3.54)
Q Q

Similarly, � �
f≤ g. (3.55)
Q Q

It follows that if f ≤ g, then � �

f≤ g. (3.56)
Q Q

This is the monotonicity property of the R. integral.

3.4 Fubini Theorem

In one­dimensional calculus, when we have a continuous function f : [a, b] → R, then
we can calculate the R. integral
� b
f (x)dx = F (b) − F (a), (3.57)
a

1
where F is the anti­derivative of f .
When we integrate a continuous function f : Q → R over a two­dimensional
region, say Q = [a1 , b1 ] × [a2 , b2 ], we can calculate the R. integral
� � b1 � b2 � b1 �� b2 �
f= f (x, y)dxdy = f (x, y)dxdy (3.58)
Q a1 a2 a1 a2

That is, we can break up Q into components and integrate separately over those
components. We make this more precise in the following Fubini Theorem.
First, we define some notation that will be used.
Let n = k + � so that Rn = Rl × R� . Let c = (c1 , . . . , cn ) ∈ Rn . We can write
c = (a, b), where a = (c1 , . . . , c) ∈ Rk and b = (ck+1 , . . . , ck+� ) ∈ R� . Similarly,
let Q = I1 × · · · In be a rectangle in Rn . Then we can write Q = A × B, where
A = I1 × · · · × Ik ∈ Rk and B = Ik+1 × · · · × Ik+� ∈ R� . Along the same lines, we
can write a partition P = (P1 , . . . , Pn ) as P = (PA , PB ), where PA = (P1 , . . . , Pk ) and
PB = (Pk+1 , . . . , Pk+� ).
Fubini Theorem. Let f : Q → R be a bounded function and Q = A × B a rectangle
as defined above. We write f = f (x, y), where x ∈ A, and y ∈ B. Fixing x ∈ A, we
can define a function fx : B → R by fx (y) = f (x, y). Since this function is bounded,
we can define new functions g, h : A → R by
�
g(x) = fx , (3.59)
B
�
h(x) = fx . (3.60)
B

Note that g ≤ h. The Fubini Theorem concludes the following: If f is integrable over
Q, then g and h are integrable over A and
� � �
g= h= f. (3.61)
A A Q

Proof. Let P = (PA , PB ) be a partition of Q, and let R = RA × RB be a rectangle

belonging to P (so RA belongs to PA and RB belongs to PB ). Fix x0 ∈ A.
First, we claim that
mRA ×RB (f ) ≤ mRb (fx0 ), (3.62)
the proof of which is straightforward.
Next,
� �
mRA ×RB (f )v(RB ) ≤ mRB (fx0 )v(RB )
RB RB

= L(fx0 , PB ) (3.63)
�
≤ fx0 = g(x0 ).
B

2
So, �
mRA ×RB (f )v(RB ) ≤ g(x0 ) (3.64)
RB

for all x0 ∈ RA . The above equation must hold for the infimum of the r.h.s, so
�
mRA ×RB (f )v(RB ) ≤ mRA (g). (3.65)
RB

Observe that v(RA × RB ) = v(RA )v(RB ), so

�
L(f, P ) = mRA ×RB (f )v(RA × RB )
RA ×RB
�
≤ mRA (g)v(RA ) (3.66)
RA
�
≤ g.
A

We have just shown that for any partition P = (PA , PB ),

�
L(f, P ) ≤ L(g, PA ) ≤ g, (3.67)
A

so � �
f≤ g. (3.68)
Q A

By a similar argument, we can show that

� �
h≤ f. (3.69)
A Q

Summarizing, we have shown that

� � � �
f≤ g≤ h≤ f, (3.70)
Q A A Q

where we used monotonicity for the middle inequality. Since f is R. integrable,

� �
f= f, (3.71)
Q Q

so all of the inequalities are in fact equalities.

3
Remark. Suppose that for every x ∈ A, that fx : B → R is R. integrable. That’s
the same as saying g(x) = h(x). Then
� �� � � �� �
fx = dx f (x, y)dy
A B
�A B
(3.72)
= f (x, y)dxdy,
A×B

using standard notation from calculus.

Remark. In particular, if f is continuous, then fx is continuous. Hence, the above
remark holds for all continuous functions.

3.5 Properties of Riemann Integrals

We now prove some standard calculus results.
Theorem 3.13. Let Q ⊆ Rn be a rectangle, and let f, g : Q → R be R. integrable
functions. Then, for all a, b ∈ R, the function af + bg is R. integrable and
� � �
af + bg = a f + b g. (3.73)
Q Q Q

Proof. Let’s first assume that a, b ≤ 0. Let P be a partition of Q and R a rectangle

belonging to P . Then
amR (f ) + bmR (g) ≤ mR (af + bg), (3.74)
so
aL(f, P ) + bL(g, P ) ≤ L(af + bg, P )
�
(3.75)
≤ af + bg.
Q

Claim. For any pair of partitions P � and P �� ,

�
� ��
aL(f, P ) + bL(g, P ) ≤ af + bg. (3.76)
Q

To see that the claim is true, take P to be a refinement of P � and P �� , and apply
Equation 3.75. Thus, � � �
a f +b g≤ af + bg. (3.77)
Q Q Q

Similarly, we can show that

� � �
af + bg ≤ a f +b g. (3.78)
Q Q Q

4
Since f and g are R. integrable, we know that
� � � �
f= f, g= g. (3.79)
Q Q Q Q

These equalities show that the previous inequalities were in fact equalities, so
� � �
af + bg = a f + b g. (3.80)
Q Q Q

However, remember that we assumed that a, b ≥ 0. To deal with the case of

arbitrary a, b, it suffices to check what happens when we change the sign of a or b.
Claim. � �
−f = − f. (3.81)
Q Q

Proof Hint. Let P be any partition of Q. Then L(f, P ) = −U (−f, P ).

You should check this claim, and then use it to complete the proof.