Lecture 12

Riemann definite integral

Definition partitions.

Let a , b ∈ R , and suppose a< b. By a partition of the interval [a , b] we mean

a collection of intervals
Ƥ={[ x 0 , x 1 ] , [ x 1 , x 2 ] , … , [ x n−1 , x n ]}

such that a=x 0 < x 1< …< x k−1< x k < …< x n=b and where n ∈ N .

Hence, a partition Ƥ defines a finite collection of non-overlapping intervals

I k =[ x k−1 , x k ] , where k =1 , … ,n . The norm of a partition Ƥ is defined as

‖Ƥ‖=max { x 1−x 0 , x 2−x 1 , … , x n−x n−1 }

In other words, ‖Ƥ‖ is the maximum length of the intervals in Ƥ. To ease

n
our notation, we will denote a partition as Ƥ={[ x k−1 , x k ] }k=1. Let

Ƥ={[ x k−1 , x k ] }k=1 be a partition of [ a , b ]. If t k ∈ I k =[ x k−1 , x k ], then we say that t k

n

Ƥ̇= x ˙, x , t
n
I
is a sample of k and the set of ordered pairs {[ k−1 k ] k }k=1 will be called
a sampled partition.

Now consider a function f : [ a ,b ] → R and let Ƥ̇={[ x k−1 ˙, x k ] , tk }k=1 be a sampled

n

partition of the interval [ a , b ]. The Riemann sum of f corresponding

to Ƥ̇={[ x k−1 ˙, x k ] , t k }k=1 is the number
n

n
S ( f ; Ƥ̇ ) =∑ f ( t k ) ( x k −x k−1 ) .
k=1

When f ( x ) >0 on the interval [ a , b ], the Riemann sum is the sum of the areas
of the rectangles with height f (t k ) and width ( x k −x k−1 ). We now define the
notion of Riemann integrability.

Definition Riemann integrability. The function f : [ a ,b ] → R is said to

be Riemann integrable if there exist a number L ∈ R such that for every ε > 0
there exists δ >0 such that for any sampled partition Ƥ̇ that satisfies ‖Ƥ̇‖<δ it
holds that |S ( f ; Ƥ̇ )−L|< ε.
The set of all Riemann integrable functions on the interval [ a , b ] will be
denoted by R [ a , b ] .
If f ∈ R [ a , b ] then the number L in the definition of Riemann integrability
is unique.
If f ∈ R [ a , b ] , we call the number L the integral of f over [ a , b ] and we
denote it by
b n
L=∫ f ( x ) dx= lim S ( f ; Ƥ̇ )= lim ∑ f ( t k ) ( x k −x k−1 ) .
a ‖Ƥ‖ →0 ‖Ƥ‖→ 0 k=1

In addition, we define
a b a

∫ f ( x ) dx=−∫ f ( x ) dx ;∫ f ( x ) dx=0.
b a a

The next result shows that if f ∈ R [ a , b ] then changing f at a finite number of

points in [ a , b ] does not affect the value of
b

∫ f ( x ) dx
a

Theorem 1. Let f ∈ R [ a , b ] and let g : [ a , b ] → R be a function such

that g ( x )=f (x) for all [ a , b ] except possibly at a finite number of points in [ a , b ].
Then g ∈ R [ a , b ], and in fact
b b

∫ g (x)dx=∫ f ( x ) dx .
a a

We now give a necessary condition for Riemann integrability.

Theorem 2. Integrable functions are bounded.
If f ∈ R [ a , b ] then f is bounded on [ a , b ].
b

Proof. Let f ∈ R [ a , b ] , ε=1and put L=∫ f ( x ) dx .

a

There exists δ >0 such that if ‖Ƥ̇‖<δ then |S ( f ; Ƥ̇ )−L|<1, and therefore
|S ( f ; Ƥ̇ )|<|L|+1. Suppose by contradiction that f is unbounded on [ a , b ].
Let Ƥ be a partition of [ a , b ], with sets I 1 , … , I N , and with ‖Ƥ‖<δ . Then f is
unbounded on some I j, i.e., for any M >0 there exists
x ∈ I j=[ x j −1 , x j ] , such that f ( x ) > M . Choose samples in Ƥ by asking that t k =x k
for k ≠ j and t j is such that

|f ( t j ) ( x j−x j −1 )|>|L|+1+|∑ f ( t k )( x k −x k−1)|.

k≠ j

Therefore,

| k≠ j | |
k≠ j |
|S ( f ; Ƥ̇ )|= f ( x j ) ( x j−x j−1) + ∑ f ( t k ) ( x k −x k−1 ) ≥|f ( x j ) ( x j−x j−1 )|− ∑ f ( t k ) ( x k −x k−1 ) >| L|+ 1.
This is a contradiction and thus f is bounded on [ a , b ] .

Darboux sums and their properties.

Let f : [ a ,b ] → R be a bounded function and let Ƥ={[ x k−1˙ , x k ] }k=1 be a sampled
n

partition of the interval [ a , b ]. For k =1 , … ,n let us denote

mk = inf f ( x ) , M k = [ x k−1 , x k ] f (x).
¿

[ x k−1 , xk ]

We define the upper Darboux sum of 𝑓 with respect to Ƥ by

n
U ( f ; Ƥ )=∑ M k ( x k −x k−1 ) ,

and the lower Darboux sum of 𝑓 with respect to Ƥ by

k=1

n
L ( f ; Ƥ )=∑ mk ( x k −x k−1 ) .
k=1

It's obvious that for any partition Ƥ of the interval [ a , b ]we have

n
L ( f ; Ƥ ) ≤ S ( f ; Ƥ̇ )=∑ f ( t k ) ( x k −x k−1 ) ≤ U ( f ; Ƥ )
k=1

[Link] f : [ a ,b ] → R is bounded function and Ƥ is any partition of [ a , b ],

then
L ( f ; Ƥ ) ≤U ( f ; Ƥ ) .
Proof. Let Ƥ={[ x k−1˙ , x k ] }k=[Link] mk ≤ M k for k =1 , … ,n and since x k −x k−1 >0 for
n

k =1 , … ,n , it follows that
n n
L ( f ; Ƥ )=∑ mk ( x k −x k−1 ) ≤ ∑ M k ( x k −x k−1) =U ( f ; Ƥ ) .
k=1 k=1

Definition. If Ƥ={[ x k−1˙ , x k ] , }k =1and Q= {[ y j−1˙ , y j ] , } j=1 are partitions of [ a , b ],we say
n m

that Q is a refinement of Ƥ if each partition x k ∈ Ƥ also belongs to Q , that is Ƥ ⊆ Q .

Lemma [Link] f : [ a ,b ] → R is bounded function and Ƥ is any partition of [ a , b ],
and if Q is a refinement of Ƥ , then
L ( f ; Ƥ ) ≤ L ( f ; Q ) and U ( f ; Q ) ≤U ( f ; Ƥ ).
Proof. Let Ƥ={[ x k−1˙ , x k ] }k=1 . We first examine the effect of adjoining one point
n

to Ƥ .Let z ∈ [ a , b ] satisfy x k−1 < z< x k and Ƥ ´ be the partition

Ƥ ´ ={ x 0 , x 1 , … , x k−1 , z , x x , … , xn }, obtained from Ƥ by adjoining z to Ƥ.
Let m ´ k =inf {f ( x ) : x ∈ [ x k −1 , z ]} , and m ´ ´ k =inf {f ( x ) : x ∈ [ z , x k ] }.Then mk ≤m ´ k and mk ≤m ´ ´ k
and therefore
mk ( x k −x k−1 )=mk ( z−x k−1 ) +mk ( x k −z ) ≤≤ m´ k ( z−x k−1) + m´ ´ k ( x k −z ) .
If we add the terms m j ( x j −x j−1 )for j ≠ k to the above inequality, we obtain
L ( f ; Ƥ ) ≤ L ( f ; Ƥ´ ) .
Now if Q is any refinement of Ƥ (i.e. if Ƥ ⊆ Q), then Q can be obtained from Ƥ by
adjoining a finite number of points to Ƥ one at a time. Hence, repeating the
preceding argument, we infer that L ( f ; Ƥ ) ≤ L ( f ; Q ) .
Upper sums are handled similarly.
Lemma 3. Let f : [ a ,b ] → R be bounded function. If Ƥ1 , Ƥ2 are any two
partitions of [ a , b ], then
L ( f ; Ƥ 1 ) ≤ U ( f ; Ƥ 2 ).
Proof. Let Q=Ƥ1 ∪ Ƥ 2 be the partition obtained combining the points of Ƥ1 and Ƥ2 .
Then Q is a refinement of both Ƥ1and Ƥ2 .Hence by Lemmas 1 and 2, we conclude
that
L ( f ; Ƥ1 ) ≤ L ( f ; Q ) ≤ U ( f ; Q ) ≤U ( f ; Ƥ2 ) .

Darboux criterion for the integrability of a function.

Theorem . Let 𝑓 ∶ [𝑎, 𝑏] → ℝ be a bounded function. Then 𝑓 is integrable on
[𝑎, 𝑏] if and only if
lim ( U ( f ; Ƥ )−L ( f ; Ƥ ) ) =0 ,

for any partition of the interval [𝑎, 𝑏] .

‖Ƥ‖→ 0

Corollary. Let 𝑓 ∶ [𝑎, 𝑏] → ℝ be a bounded function. Then 𝑓 is integrable

on [𝑎, 𝑏] if and only if ∀𝜀 > 0, ∃ a partition Ƥ of [𝑎, 𝑏],

Proof. Let 𝑓 is integrable on [𝑎, 𝑏].Then by the definition, there exists a

( U ( f ; Ƥ )−L ( f ; Ƥ ) ) < ε .

number L ∈ R such that for every ε > 0 there exists δ >0 such that for any
sampled partition Ƥ̇ that satisfies ‖Ƥ̇‖<δ it holds that
|S ( f ; Ƥ̇ )−L|< ε or L−ε ≤ S ( f ; Ƥ̇ ) ≤ L+ ε .
But L−ε ≤ L ( f ; Ƥ ) ≤ U ( f ; Ƥ ) ≤ L+ ε . Consequently
lim L ( f ; Ƥ ) =L= lim U ( f ; Ƥ )
‖Ƥ‖→ 0 ‖Ƥ‖→ 0

Whence it follows that ‖Ƥlim ( U ( f ; Ƥ )−L ( f ; Ƥ ) ) =0.

Now suppose that ∀𝜀 > 0, ∃ a partition Ƥ of [𝑎, 𝑏] such that satisfy
‖→ 0

( U ( f ; Ƥ )−L ( f ; Ƥ ) ) < ε .
On the other hand we know that for any partition Ƥ of [ a , b ]
L ( f ; Ƥ ) ≤U ( f ; Ƥ ).Hence
¿
L(f ; Ƥ)≤ Ƥ L ( f ; Ƥ ) ≤inf U ( f ; Ƥ ) ≤U ( f ; Ƥ ) .
Ƥ

Then we have
0 ≤ inf U (f ; Ƥ)−¿ Ƥ L(f ; Ƥ)≤ ( U ( f ; Ƥ )−L ( f ; Ƥ )) < ε .
Ƥ
¿
Since ε > 0 is arbitrary, we conclude that inf U (f ; Ƥ)= Ƥ L ( f ; Ƥ ) =L.
Ƥ
Therefore we can write
−ε < L ( f ; Ƥ )−L ≤ S ( f ; Ƥ̇ )−L ≤U ( f ; Ƥ )−L<ε .
Hence |S ( f ; Ƥ̇ )−L|< ε , that is f is integrable.

The Cauchy criterion for integrability

The following theorem gives a criterion for integrability that is analogous to the
Cauchy condition for the convergence of a sequence.

Theorem. A bounded function f : [ a ,b ] → R is Riemann integrable if and only

if for every ε > 0 there exists a partition Ƥ of [ a , b ], which may depend on ε , such
that
U ( f ; Ƥ )−L ( f ; Ƥ ) <ε .

Properties of the definite integrals.

Theorem. i). If f : [ a , b ]→ R is integrable and c ∈ ℝ, then cf is integrable

1. Linearity.

and
b b

∫ cf ( x ) dx=c ∫ f ( x ) dx .
a a

ii) If f , g : [a , b ] → Rare integrable functions, then f +g is integrable and

b b b

∫ [f ( x )+ g ( x ) ]dx=∫ f ( x ) dx +∫ f ( x ) dx .
a a a

[Link] product Theorem. If f ∧g : [a , b ] → R are integrable then

fg : [a , b ] → R is integrable. If, in addition, g ≠ 0and 1/ g is a bounded, then
f /g : [a , b ] → R is integrable.
3. Monotonicity Theorem . Suppose that f , g : [a , b ] → Rare integrable
and f ≤ g. Then
b b

∫ f ( x ) dx ≤∫ g ( x ) dx .
a a

Corollary . Let f : [ a ,b ]→ R be integrable on [ a , b ].If f (x)≥ 0 for all x ∈ [ a , b ] ,

then
b

∫ f ( x ) dx ≥ 0.
a

Moreover, if f (x)>0 for x ∈ [ α , β ] ⊆ [ a , b ], then

 b

∫ f ( x ) dx >0.
a

Theorem. Suppose that f : [ a , b ]→ R is integrable and

¿
M = [ a , b ] f ,m=inf f
[ a , b]

Then
a
m(b−a)≤∫ f ( x ) dx ≤ M ( b−a ) .
a

Theorem. If f : [ a , b ]→ Ris integrable on[ a , b ], then |f | is integrable too on

[ a , b ]and

|∫ | ∫|
a b
f ( x ) dx ≤ f (x )|dx .
a a

In particular, if f : [ a , b ]→ R is integrable on [ a , b ]and if |f ( x )|≤ K for all

x ∈ [ a , b ] ,then

|∫ |
a
f ( x ) dx ≤ K |b−a|.
a

Additivity. This property refers to additivity with respect to the interval

of integration, rather than linearity with respect to the function being
integrated.
Theorem. Suppose that f : [ a , b ]→ R and a< c< b . Then f is Riemann
integrable on [ a , b ] if and only if it is Riemann integrable on both [ a , c ] and
[ c ,b ]. In that case,
b c b

∫ f ( x ) dx=¿∫ f ( x ) dx +∫ f ( x ) dx .¿
a a c

The integrability of Monotone and Continuous Functions

If f : [ a , b ]→ R is a bounded function on [ a , b ] and Ƥ={[ x k−1˙ , x k ] }k=1
n

is a partition of [ a , b ]we use the familiar notation

mk = inf f ( x ) , M k = [ x k−1 , x k ] f (x).
¿

[ x k−1 , xk ]
We also note that
n
U ( f ; Ƥ )−L ( f ; Ƥ )=∑ ( M k −mk ) ( x k −x k−1 ) .
k=1

Theorem (Integrability of monotone functions).

A monotonic function f : [ a , b ]→ R on [ a , b ] interval is Riemann integrable.
Proof. Without loss of generality, suppose that f is monotonic increasing. Let
Ƥn ={ I 1 , … , I n } be a partition of [a , b ] into n intervals I k =[ x k , x k−1 ] , of equal length
 k
( b−a ) /n , with endpoints x k =a+ ( b−a ) , k =0 , 1, … , n. Since f is increasing,
n
M k = [ x k−1 , x k ] f (x)=f ( x k ) , m k = inf f ( x )=f (x k−1). Hence, summing a telescoping series,
¿

[ xk −1 , x k ]
we get
n
U ( f ; Ƥ )−L ( f ; Ƥ )=∑ ( M k −mk ) ( x k −x k−1 )=¿ ¿
k=1
n
b−a b−a b−a
¿ ∑
n k=1 [ f ( x k )−f ( x k−1 ) ]=
n [ f ( x n )−f (x 0) ]=
n
[ f ( b )−f (a)] .
b−a
Now if ε > 0 is given, we choose n ∈ N such that n> ε [ f ( b )−f (a) ] .
For the corresponding partition Ƥn we have
n
U ( f ; Ƥn )−L ( f ; Ƥ n )=∑ ( M k −mk )( x k −x k−1) < ε .
k=1

It follows that f is integrable on [ a , b ] .

Theorem (Integrability of continuous functions).
Let f ∈ C [ a , b ]. Then f ∈ R [ a , b ] , that is the continuous function on close interval [ a , b ]
is Riemann integrable.
Proof. Let ε > 0. By the Cantor theorem, a continuous function on close interval
[ a , b ] is uniformly continuous, so there exists δ > 0 such that
ε
|f ( x ) −f ( y)|< b−a for all x , y ∈ [ a ,b ]such that |x− y|< δ .
Now let nϵ N be such that n> ( b−a ) /δ and let Ƥn ={ I 1 , … , I n } be the partition of [ a , b ]
into n equal parts so that |I k|=x k −x k−1=( b−a ) /n< δ .
Since f is continuous, it attains its maximum and minimum values M k and mk on
the close interval I k =[ x k , x k−1 ] at points z k and y k in I k . These points satisfy
|z k − y k|<δ , so
ε
M k −mk =f ( z k )−f ( y k ) < .
b−a
The upper and lower sums of f therefore satisfy
n
U ( f ; Ƥn )−L ( f ; Ƥ n )=∑ ( M k −mk )( x k −x k−1) < ¿ ¿
k=1
n
ε
∑ ( x −x ) =ε .
b−a k=1 k k−1
¿

Since ε > 0 is arbitrary, it follows that f is integrable on [ a , b ] .

{1/ x , x ∈¿
Example 1. Let f ( x )= 0 , x=0 .
Then
1

∫ 1x dx
0

isn’t defined as a Riemann integral because f is unbounded. In fact, if

Ƥn ={0=x 0< x1 < …< x n−1<1=x n }
is a partition of [ 0 , 1 ], then
¿ fx ¿=+∞ ,¿
[0 , x 1]

so the upper Riemann sums are not well-defined.

Example 2. The Dirichlet function: f : [0, 1] → R is defined by

{
f ( x )= 1 , if x ∈ [ 0 ,1 ] ∩Q ,
0 , if x ∈ [ 0 ,1 ] ¿ .

That is, f is one at every rational number and zero at every irrational number.
This function is not Riemann integrable. If Ƥ = { I 1 , … , I n} is a partition of [0, 1],
then M k =max
I
f =1, mk =min f =0, since every interval of non-zero length contains
k I k

both rational and irrational numbers. It follows that U ( f ; Ƥ )=1 , L ( f ; Ƥ )=0 for every
¿
partition P of [0, 1], so inf U (f ; Ƥ)≠ Ƥ L(f ; Ƥ) are not equal. The Dirichlet function
Ƥ

is discontinuous at every point of [0, 1], and the moral of the last example is that
the Riemann integral of a highly discontinuous function need not exist.