MATH 2037 – Mathematical Analysis

Lecture 9

University of Nottingham Ningbo China

Faculty of Science and Engineering
School of Mathematical Sciences

20 November

1 / 28

Chapter 5

Functions on the real line

2 / 28
 5.3 The Riemann integral

3 / 28

Differentiation and Integration

Since a differentiable function must be continuous, a function is

not differentiable at its discontinuities.
For a function f , the process of finding its derivative f ′ is called
differentiation.
The reverse process is called antidifferentiation. The concept of
integration was defined as the inverse process of differentiation. In
the inverse process what we want to find is an F such that F ′ = f .

4 / 28
Differentiation and Integration
Not every function f has a function F such that F ′ = f .
▶ Based on Darboux’s Theorem, we see that, for the function

1 for 0 ≤ x < 1
f (x) =
2 for 1 ≤ x ≤ 2
it is impossible to find a function F such that F ′ = f .
▶ If a function f is continuous in an interval containing point a,
Rx
then the function F (x) = a f (t)dt satisfies F ′ = f . Are there
any other functions of such a property? Yes, for example:
2x sin x1 − cos x1 , for x ̸= 0

f (x) =
0, for x = 0
is the derivative on R of a function F (See Lecture 8), is not
continuous on any interval [a, b] containing 0, and is Riemann
integrable on any interval [a, b] (since it is bounded and
continuous excepted at one point (particular case of
continuous almost everywhere) (result not seen)).
5 / 28

Integration and Riemann sums

We consider the area under the curve of f defined on [a, b] and

bounded by the lines x = a and x = b.
▶ We use the ordered set of points

a = x0 < x1 < · · · < xn = b

to partition [a, b] into n subintervals [xk−1 , xk ]

(k = 1, 2, · · · , n).
▶ For each k, pick ck ∈ [xk−1 , xk ] and use f (ck ) as an
approximation for f on [xk−1 , xk ] so that we obtain a thin
rectangle whose area is f (ck ) (xk − xk−1 ).

6 / 28
 Figure 1. A Riemann sum

7 / 28

▶ The area of all such rectangles is the sum

n
X
f (ck ) (xk − xk−1 )
k=1

called a Riemann sum for f on [a, b].

▶ A Riemann sum is an approximation for the area under the
curve of f on [a, b]. The thinner the rectangles are, the more
approximate is the sum to the area under the curve f .
▶ So the area under the curve of f should be the limit of this
sum:
Xn
lim f (ck ) (xk − xk−1 ) ,
∆x→0
k=1

where ∆x := max {xk − xk−1 : k = 1, 2, · · · , n}

Rb
▶ This limit, if it exists, is Riemann’s definition of a f (x)dx.

8 / 28
Lower Sum and Upper Sum

Definition 1. Let f be bounded on [a, b]. A partition P of

[a, b] is a finite, ordered set

P := {a = x0 < x1 < · · · < xn = b} .

For each subinterval [xk−1 , xk ] from P, let

mk := inf {f (x) : x ∈ [xk−1 , xk ]}

and
Mk := sup {f (x) : x ∈ [xk−1 , xk ]} .

9 / 28

Lower Sum and Upper Sum

Definition 2. The lower sum of f with respect to P is

defined by
n
X
L(P, f ) = mk (xk − xk−1 )
k=1

while the upper sum of f with respect to P is defined by

n
X
U(P, f ) = Mk (xk − xk−1 ) .
k=1

10 / 28
Remark.
▶ L(P, f ) and U(P, f ) are obtained from a Riemann Sum for P
by replacing f (ck ) with mk and Mk for each subinterval
[xk−1 , xk ] .
▶ If f is continuous on [a, b], then mk and Mk are minimum and
maximum of f on [xk−1 , xk ] .
▶ If f is not continuous but bounded on [a, b], mk and Mk may
not be minimum and maximum of f on [xk−1 , xk ] . However
both L(P, f ) and U(P, f ) are well defined.
▶ For each k, since mk ≤ f (ck ) ≤ Mk and xk − xk−1 > 0 for
any ck ∈ [xk−1 , xk ]
n
X
L(P, f ) ≤ f (ck ) (xk − xk−1 ) ≤ U(P, f ).
k=1

11 / 28

Definition 3. A partition Q of [a, b] is a refinement of

another partition P if Q contains all of the points of P. In
the case, we write P ⊆ Q.

Remark:
▶ If P1 and P2 are partitions of [a, b], then P = P1 ∪ P2 is also
a partition of [a, b] and P is a refinement of P1 and P2 .
▶ For any two partitions P1 , P2 for [a, b], P1 ∩ P2 is not empty
since a, b ∈ P1 ∩ P2 .
Example:
▶ Consider the partitions of [0, 1] with endpoints
1 1 2 1 1 3
P1 = {0, , 1}, P2 = {0, , , 1}, P3 = {0, , , , 1}.
2 3 3 4 2 4
Thus, P1 , P2 , and P3 partition [0, 1] into intervals of equal
length 21 , 13 , and 41 , respectively. Then P2 is not a refinement
of P1 but P3 is a refinement of P1 .
12 / 28
 Lemma 1. If P ⊆ Q, then

L(P, f ) ≤ L(Q, f ) ≤ U(Q, f ) ≤ U(P, f ).

In particular, if we denote P0 = {a, b}, then for any bounded

function f and any partitions P of [a, b] we have

L (P0 , f ) ≤ L(P, f ) ≤ U(P, f ) ≤ U (P0 , f ) .

Proof
Let P = {x0 , x1 , ..., xn }. If P ⊂ Q, then Q\P = {z1 , . . . , zm }, with
m ∈ N and zi ∈ [a, b]. Let

P1 = P ∪ {z1 } , P2 = P1 ∪ {z2 } , . . . , Q = Pm = Pm−1 ∪ {zm }

Then, z1 ∈ (xk−1 , xk ), for some k ∈ {1, . . . , n}. Define

mk = inf x∈[xk−1 ,xk ] f (x),
mk ′ = inf x∈[z,xk ] f (x),
and mk ′′ = inf x∈[xk−1 ,z] f (x). 13 / 28

Proof(Cont.)
Then, mk′ ⩾ mk and mk′′ ⩾ mk , and we have

mk (xk − xk−1 ) = mk (xk − z1 ) + mk (z1 − xk−1 )

≤ mk ′ (xk − z1 ) + mk ′′ (z1 − xk−1 ) ,
n
X
L(P, f ) = mj (xj − xj−1 )
j=1
n
X
= mj (xj − xj−1 ) + mk (xk − z) + mk (z1 − xk−1 )
j=1,j̸=k
Xn
⩽ mj (xj − xj−1 ) + mk′ (xk − z) + mk ′′ (z1 − xk−1 )
j=1,j̸=k

= L (P1 , f )

14 / 28
Proof(Cont.)
We have by similar arguments

L(P, f ) ⩽ L (P1 , f ) ⩽ L (P2 , f ) ⩽ . . . ⩽ L (Pn , f ) = L(Q, f ).

Similarly, we prove

U(P, f ) ⩾ U (P1 , f ) ⩾ U(P2 , f ) ⩾ . . . ⩾ U(Pn , f ) = U(Q, f ).

15 / 28

Lemma 2. If P1 and P2 are two partitions of [a, b], then

L (P1 , f ) ≤ U (P2 , f ) .

Proof.
Let
P = P1 ∪ P2 .
Then, P is a refinement of P1 and P is a refinement of P2 .
Therefore, by Lemma 1,

L(P1 , f ) ⩽ L(P, f ) ⩽ U(P, f ) ⩽ U (P2 , f )

16 / 28
Lower Integral and Upper Integral

Definition 4. Let P be the collection of all possible partitions

of the closed interval [a, b]. The upper integral of f is defined
and denoted by
Z b
f (x)dx := inf{U(P, f ) : P ∈ P}
a

Similarly, the lower integral of f is defined and denoted by

Z b
f (x)dx := sup{L(P, f ) : P ∈ P}
a

17 / 28

Lemma 3. For any bounded function f on [a, b], there holds

Z b Z b
f (x)dx ≤ f (x)dx.
a a

Proof.
By Lemma 2,

L(P1 , f ) ≤ U(P2 , f ), ∀P1 , P2 ∈ P.

Then, U(P2 , f ) for any P2 ∈ P is an upper bound for

{L(P1 , f )|P1 ∈ P}. Therefore,

sup L(P, f ) ≤ U(P2 , f ), ∀P2 ∈ P.

P∈P

Then, supP∈P L(P, f ) is a lower bound for {U(P2 , f )|P2 ∈ P}.

Therefore,
sup L(P, f ) ≤ inf U(P, f ) 18 / 28
P∈P P∈P
 Definition 5. A bounded function f defined on [a, b] is said
to be Riemann integrable if
Z b Z b
f (x)dx = f (x)dx.
a a

Rb Rb Rb
In this case, we define a f (x)dx := a f (x)dx = a f (x)dx.

Example:
1. Consider the function f (x) on [0, 1] given by
(
1, if x is rational,
f (x) =
0, if x is irrational.

Is it Riemann integrable?
2. Every constant function f (x) = c on [a, b] is Riemann
integrable.
19 / 28

Example: (Dirichlet function) For any P ∈ P, we have

n
X
L(P, f ) = mk (xk − xk−1 ) = 0
k=1
mk = inf ({f (x)|x ∈ [xk−1 , xk ]}) = 0, for k = 1, . . . , n
Xn Xn
U(P, f ) = Mk (xk − xk−1 ) = (xk − xk−1 ) = b − a = 1
k=1 k=1
Mk = sup {f (x) | x ∈ [xk−1 , xk ]} = 1, for k = 1, . . . , n

So,
Z b Z b̄
f = 0 ̸= f =1
a a

and f is not Riemann integrable

20 / 28
 Example: (constant function) For any P ∈ P, we have
n
X n
X
L(P, f ) = mk (xk − xk−1 ) = c (xk − xk−1 ) = c(b − a)
k=1 k=1
X
U(P, f ) = Mk (xk − xk−1 ) = c(b − a)
Z b
⇒ f = sup({L(P, f )|P ∈ P}) = sup({c(b − a)})
a
Z b̄
= c(b − a) = inf({U(P, f )(P ∈ P} = f
a
So, f is Riemann integrable.

21 / 28

Criteria for Integrability

Theorem 1. A bounded function f is Riemann integrable on

[a, b] if and only if, for each ϵ > 0, there exists a partition Pϵ
of [a, b] such that

U (Pϵ , f ) − L (Pϵ , f ) < ϵ.

Proof
(⇒) Assume f is Riemann integrable on [a, b]. Let ε > 0.
Z b Z b Z b̄
We know that sup (L(P, f )) = f = f = f = inf (U(P, f ))
P∈P a a a P∈P

Rb
So, a f − 2ε is not an upper bound for {L(P, f ) : P ∈ P} and
Rb ε
a f + 2 is not a lower bound for {U(P,
Rb
f ) : P ∈ P}.
Therefore, ∃P1 ∈ P s.t L (P1 , f ) > a f − 2ε and
R b̄
∃P2 ∈ P s.t U (P2 , f ) < a f + 2ε . 22 / 28
Proof.
Let Pϵ = P1 ∪ P2 .

U (Pε , f ) − L (Pε , f ) ⩽ U (P2 , f ) − L (P1 , f )

Z b Z b 
ε ε
< f + − f −
a 2 a 2
< ε.
(⇐) We assume ∀ε > 0, ∃Pε ∈ P s.t. U (Pε , f ) − L (Pε , f ) < ε.
Let ε > 0 and let Pε ∈ P given by the former statement. Then,
Z b̄ Z b
0⩽ f − f ⩽ U (Pε , f ) − L (Pε , f ) < ε.
a a

R b̄ Rb
Therefore a f = a f (since ε > 0 was arbitrary).

23 / 28

Definition 6. A function f : [a, b] → R is uniformly continu-

ous on [a, b] if for every ϵ > 0 there exists a δ > 0 such that
for every x, y ∈ [a, b], |x − y | < δ implies |f (x) − f (y )| < ϵ.

Note that f is not uniformly continuous iff

∃ε0 > 0, ∀δ > 0, ∃x, y ∈ [a, b] : |x − y | < δ and |f (x) − f (y )| > ε0

Proposition 1. A function f : [a, b] → R that is continuous

on [a, b] is uniformly continuous on [a, b].

Proof(Proposition 1.)
Assume f continuous on [a, b]. For the sake of contradiction,
suppose ∃ε0 > 0 ∃ (xn )∞ ∞
n=1 , (yn )n=1 s.t. |xn − yn | → 0, as n → ∞,
and |f (xn ) − f (yn )| > ε0 , ∀n ∈ N.
24 / 28
Proof(Cont.)
The set [a, b] is compact (Heine-Borel). Therefore, ∃ (xnk )∞
k=1
which converges to x ∈ [a, b].

ynk = (ynk − xnk + xnk ) → x, as k → ∞

| {z } |{z}
→0 →x
,
f (xnk ) → f (x), as k → 0
f (ynk ) → f (x), as x → ∞

since f is continuous on [a, b]. Then,

|f (ynk ) − f (xnk )| → 0,

as k → ∞, a contradiction.

25 / 28

Theorem 2. If f is continuous on [a, b], then f is Riemann

integrable.

Theorem 3. If f is monotone on [a, b], then f is Riemann

integrable on [a, b].

Proof(Theorem 2.)
Let ε > 0. Assume f continuous on [a, b]. Thus, f is uniformly
continuous on [a, b]. Then, ∃δ > 0 s.t. x, y ∈ [a, b] and
ε
|x − y | < δ implies |f (x) − f (y )| < b−a . Now, let n ∈ N be s.t.
b−a
n < δ, and define P = {x0 , x1 , . . . , xn } by x0 = a and

b−a
xk = xk−1 + ,
n
for k = 1, . . . , n.
26 / 28
Proof(Cont.)
By EVT, ∃yk , zk ∈ [xk−1 , xk ] s.t.

mk = inf f (x) = f (yk )

x∈[xk−1 ,xk ]
, ∀k = 1, . . . , n.
Mk = sup f (x) = f (zk )
x∈[xk−1 ,xk ]

Therefore, we have
n
X
U (Pε , f ) − L (Pε , f ) = (Mk − mk ) (xk − xk−1 )
k=1
X
= (f (zk ) − f (yk )) (xk − xk−1 )
k
X ε

< (xk − xk−1 ) = ε.
b−a
k

Hence, f is Riemann integrable on [a, b].

27 / 28

Proof(Theorem 3.)
Assume f non decreasing on [a, b] (the case f non-increasing is
similar and will not be treated in this proof). Let ε > 0, and let
Pε = {x0 , x1 , ..., xn } be a partition of [a, b] s.t
ε
xk − xk−1 < , ∀k = 1, . . . , n.
f (b) − f (a)

We have Mk = f (xk ) and mk = f (xk−1 ) , ∀k = 1, . . . , n since f is

non-decreasing. Therefore, we have
n
X
U (Pε , f ) − L (Pε , f ) = (Mk − mk ) (xk − xk−1 )
k=1
Xn
= (f (xk ) − f (xk−1 )) (xk − xk−1 )
k=1
n
X ε
< (f (xk ) − f (xk−1 )) = ε.
f (b) − f (a)
k=1

Hence, f is Riemann integrable on [a, b]. 28 / 28