CHAPTER 5

The Riemann Integral

1. Riemann Integrability
Definition 1.1. Let a < b. An interval partition of [a, b] is a finite sequence
P = (xi )ni=0 such that
a = x0 < x1 < · · · < xn−1 < xn = b.
We denote by IP[a, b] the collection of all interval partitions of [a, b].
The norm of the interval partition P = (xi )ni=0 is kP k = maxi=1,n (xi − xi−1 ). Given
two partitions P = (xi )ni=0 and Q = (yj )m j=0 of [a, b] we say P is coarser than Q,
equivalently Q is finer than P , denoted P 4 Q, if
{xi | i = 0, . . . , n} ⊆ {yj | j = 0, 1, ..., m}.
n
Example 1.2. (1) The dyadic partition: For n ∈ N, Pnd = (xi )2i=0 , xi = a+ 2in (b−a),
i = 0, 1, ..., 2n .
(2) The uniform partition: For n ∈ N, Pnu = (yi )ni=0 , yi = a + i b−a
n
, i = 0, . . . , n.
Definition 1.3. Let Λ be a nonempty set and ≤ a preorder relation on it: recall
that, this means that ≤ is reflexive and transitive.
A preordered space (Λ; ≤), that is, a nonempty set Λ endowed with a preorder
relation ≤, is called directed if for all α, β ∈ Λ there exists γ ∈ Λ such that α ≤ γ and
β ≤ γ.
For example, if X is a nonempty set then (P(X); ⊆) is a directed ordered space.
Fact 1.4. Let a < b. Then (IP[a, b] ; 4) is a directed ordered space.
Proof. Let P = (xi )ni=0 and Q = (yj )m j=0 be two interval partitions of [a, b]. Con-
sider the set S = {xi | i = 0, . . . , n} ∪ {yj | j = 0, . . . , m}, then order increasingly all
the elements of S, say a = z0 < z1 < · · · < zp = b, let R = (zk )pk=1 and observe that
P 4 R and Q 4 R. 
Definition 1.5. Let a < b, f : [a, b] → R a function, and P = (xk )nk=0 ∈ IP[a, b].
A finite sequence ξ¯ = (ξk )nk=1 is called a choice of intermediate points subordinated
to the partition P if for all k ∈ {1, ..., n} we have ξk ∈ [xk−1 , xk ]. The Riemann sum
corresponding to f, P , and ξ¯ is, by definition,
Xn
¯
σ(f ; P ; ξ) := f (ξk )(xk − xk−1 ).
k=1
117
118 5. THE RIEMANN INTEGRAL

f is Riemann integrable on [a, b] if there exists I(f ) ∈ R such that for all ε > 0,
there exists Pε ∈ IP[a, b] such that, for any P ∈ IP[a, b] with Pε 4 P and for all ξ¯ a
choice of intermediate points subordinated to P , we have
¯ − I(f )| < ε.
|σ(f ; P ; ξ)
Fact 1.6. If the function f is Riemann integrable on [a, b] then the number I(f )
in the definition of Riemann integrability is unique. We then define
Z b
I(f ) := f (x)dx
a

and call it the Riemann Integral of f on [a, b].

Proof. Let I1 and I2 be two real numbers such that for each i = 1, 2 and for all
 > 0, there exist Pε,i ∈ IP[a, b] such that, for all P ∈ IP[a, b] with Pε,i 4 P and for
any choice of intermediate points ξ¯ subordinated to P we have
¯ − Ii | < ε .
|σ(f ; P ; ξ)
2
Then there exists P ∈ IP[a, b] such that, Pε,i 4 P , for i = 1, 2, hence, for any
P ∈ IP[a, b] with P 4 P we have
¯ − I2 | < ε + ε = .
¯ − I1 | + |σ(f ; P ; ξ)
|I1 − I2 | ≤ |σ(f ; P ; ξ)
2 2
Consequently, I1 = I2 . 
Fact 1.7. If the function f is Riemann integrable on [a, b] then it is bounded on
[a, b].

Proof. By definition, letting ε = 1, there exists P = (xi )ni=0 an interval partition

of [a, b] such that for any ξ¯ = (ξi )ni=1 , choice of intermediate points subordinated to P ,
we have
¯ < I(f ) + 1.
I(f ) − 1 < σ(f ; P ; ξ)

Let k ∈ 1, ...n be arbitrary. Choosing ξi = xi , i ∈ N, i 6= k, we have that, for all
x ∈ [xk−1 , xk ]
n
X n
X
I(f ) − 1 − f (xi )(xi − xi−1 ) I(f ) + 1 − f (xi )(xi − xi−1 )
i=1,i6=k i=1,i6=k
< f (x) <
xk − xk−1 xk − xk−1
Hence, for any k ∈ 1, ....n there exists mk ≤ Mk such that for all x ∈ [xk−1 , , xk ] we have
f (x) ∈ [mk , Mk ]. Letting m := min{m1 , .....mn } ∈ R and M := max{M1 , .....Mn } ∈ R,
we have m ≤ f (x) ≤ M for all x ∈ [a, b], that is, f is bounded on [a, b]. 
Remark 1.8. Riemann Integrability is a Limit Concept.
A family (xλ )λ∈Λ of elements in a set X is called a generalised sequence or a a net,
if (Λ; ≤) is a directed preordered set.
 2. DARBOUX INTEGRABILITY 119

A net (xλ )λ∈Λ of real numbers converges to x ∈ R if for any ε > 0, there exists
λε ∈ Λ such that for all λ ∈ Λ if λ ≥ λε then |xλ − x| < ε. We write x = lim xλ .
λ∈Λ
Any sequence is a net and convergence of a sequence is a particular case of conver-
gence of a net.
¯ | P ∈ IP[a, b] and ξ¯ a choice subordinated to P }.
Let a < b and PC[a, b] := {(P, ξ)
¯ ≤ (Q, µ̄) if P 4 Q. Then (PC[a, b]; ≤) is a preordered directed set.
Define (P, ξ)
Let a < b and f : [a, b] → R be a function. Then, (σ(f, P, ξ))¯ (P ;ξ̄)∈PC[a,b] is a net and
f is Riemann integrable if and only if the net of Riemann sums (σ(f, P, ξ)) ¯ (P ;ξ̄)∈PC[a,b]
converges and, in this case,
Z b
lim ¯ :=
σ(f ; P ; ξ) f (x)dx.
(P ;ξ̄)∈PC[a,b] a

In connection with the previous remark, the next fact shows an equivalent charac-
terisation of Riemann integrability in terms of an -δ formulation that reminds us of
the definition of the limit of a function.
Fact 1.9. Let a < b and let f : [a, b] → R be a function. Then f is Riemann
integrable on [a, b] if and only if there exists I(f ) ∈ R such that, for all ε > 0 there
exists δ > 0 such that, for all P ∈ IP[a, b] with kP k < δ and any choice of intermediate
points ξ¯ subordinated to P , we ¯ − I(f )| < ε.
have |σ(f ; P ; ξ)

Proof. Exercise. 

2. Darboux Integrability

Definition 2.1. Let a < b, let f : [a, b] → R be a bounded function and let P =
(xk )nk=0 be an interval partition of [a, b]. For any k = 1, ...., n, let
mk := inf f ∈ R, Mk := sup f ∈ R.
[xk−1 ,xk ] [xk−1 ,xk ]

Then, the lower Darboux sum of f and P is

n
X
L(f ; P ) := mk (xk − xk−1 )
k=1

and the upper Darboux sum of f and P is

n
X
U (f ; P ) := Mk (xk − xk−1 ).
k=1

Fact 2.2. With notation as in the previous definition we have L(f ; P ) ≤ U (f ; P ).

In addition, if P, Q ∈ IP[a, b] and P 4 Q then
L(f ; P ) ≤ L(f ; Q) ≤ U (f ; Q) ≤ L(f ; P ).
120 5. THE RIEMANN INTEGRAL

Proof. Let P = (xk )nk=0 ∈ IP[a, b] and fix k ∈ 1, .....n . For arbitrary x ∈
[xk−1 , xn ], let
mk,l := inf f, mk,r := inf f, Mk,l := sup f, Mk,r := sup f.
[xk−1 ,x] [x,xk ] [xk−1 ,x] [x,xk ]

Clearly,
mk ≤ mk,l ≤ Mk,l ≤ Mk and mk ≤ mk,r ≤ Mk,r ≤ Mk ,
hence
mk (x − xk−1 ) ≤ mk,l (x − xk−1 ) ≤ Mk,l (x − xk−1 ) ≤ Mk (x − xk−1 )
mk (xk − x) ≤ mk,r (xk − x) ≤ Mk,r (xk − x) ≤ Mk (xk − x)
and summing up side by side these inequalities we get
mk (xk −xk−1 ) ≤ mk,l (x−xk−1 )+mk,r (xk −x) ≤ Mk,l (x−xk−1 )+Mk,r (xk −x) ≤ Mk (xk −xk−1 ).
By the definition of Darboux sums, the above inequalities and an induction argu-
ment show that, if P 4 Q then
L(f ; P ) ≤ L(f ; Q) ≤ U (f ; Q) ≤ U (f ; P ). 
Fact 2.3. With notation as in the previous definition, for all P, Q ∈ IP[a, b] we
have L(f ; P ) ≤ U (f ; Q).
Proof. The set IP[a, b] endowed with the order 4 is directed hence, there exists
R ∈ IP[a, b] such that P 4 R and Q 4 R. Then by the previous fact, L(f ; P ) ≤
L(f ; R) ≤ U (f ; R) ≤ U (f ; Q). 
Definition 2.4. Let a < b and let f : [a, b] → R be a bounded function. The upper
Darboux integral of f on [a, b] is
Z b
U (f (x)dx := inf{U (f ; P ) | P ∈ IP[a, b]} ∈ R,
a
and, the lower Darboux integral of f on [a, b] is
Z b
L (f (x)dx = sup{L(f ; P ) | P ∈ IP [a, b]} ∈ R.
a

Remark 2.5. With notation as in the previous definition, note that, by the previous
fact, both Darboux limits are finite and we have:
Z b Z b
L (f (x)dx ≤ U (f (x)dx.
a a

Definition 2.6. Let f : [a, b] → R be a bounded function. f is Darboux integrable

Rb Rb
on [a, b] if L a f (x)dx = U a f (x)dx.
Fact 2.7. Let a < b and f : [a, b] → R be a bounded function. The following
assertions are equivalent:
(1) f is Darboux integrable on [a, b].
(2) For all ε > 0 there exists P ∈ IP[a, b] such that U (f ; P ) − L(f ; P ) < ε.
 2. DARBOUX INTEGRABILITY 121

(3) There exists a sequence (Pn )n∈N of interval partitions of [a, b] such that
U (f ; Pn ) − L(f ; Pn ) → 0 as n → +∞.
Moreover, for such a sequence (Pn )n , we have
Z b Z b
U f (x)dx = lim U (f ; Pn ) = lim L(f : Pn ) = L f (x)dx.
a n→∞ n→∞ a
Rb Rb
Proof. (1) ⇒ (2). Assuming that L a (f (x)dx = U a (f (x)dx =: I ∈ R, by the
Approximation Property of the Supremum, for all ε > 0 there exists P ∈ IP[a, b] such
that I − 2ε < L(f ; P ) ≤ I. Similarly, by the Approximation Property of the infimum,
for all ε > 0 there exists Q ∈ IP[a, b] such that I ≤ U (f ; Q) < I + 2ε . Let R ∈ IP[a, b]
be such that P ≤ R and Q ≤ R. Then
ε ε
I − < L(f ; P ) ≤ L(f ; R) ≤ I ≤ U (f ; R) ≤ U (f ; Q) < I + ,
2 2
hence,
ε ε
U (f ; R) − L(f ; R) < I + − I + = ε.
2 2
(2) ⇐ (1). Observe that, for all P ∈ IP[a, b] we have
Z b Z b
0 ≤ U f (x)dx − L f (x)dx ≤ U (f ; P ) − L(f ; P ),
a a
and apply the assumption and the analytical method for proving inequalities.
We leave the other implications as an exercise. 
Fact 2.8. If f is a bounded and Darboux integrable function on [a, b] then f is
Riemann integrable on [a, b] and
Z b Z b Z b
f (x)dx = L f (x)dx = U f (x)dx.
a a a

Proof. Observe that for all P ∈ IP[a, b], letting

Z b Z b
I(f ) := L f (x)dx = U f (x)dx,
a a
we have
L(f ; P ) ≤ I(f ) ≤ U (f ; P )
¯ n
and that for all ξ = (ξk )k=1 choice of intermediate points subordinated to P we have
¯ ≤ U (f ; P ),
L(f ; P ) ≤ σ(f ; P ; ξ)
hence
¯ − I(f )| ≤ U (f ; P ) − L(f ; P ).
|σ(f ; P ; ξ)
Then apply Fact 2.7 to get that for all ε > 0 there exists Pε ∈ IP[a, b] such that for
all P ∈ IP[a, b] with Pε 4 P and for all ξ¯ = (ξk )nk=1 choice of intermediate points
subordinated to P,
¯ − I(f )| < ε. 
|σ(f ; P ; ξ)
122 5. THE RIEMANN INTEGRAL

Fact 2.9. If f is Riemann integrable on [a, b] then it is Darboux integrable on [a, b]

and
Z b Z b Z b
f (x)dx = L f (x)dx = U f (x)dx.
a a a

Proof. If f is Riemann integrable then f is bounded, hence the Darboux sums

make sense. Let ε > 0 be arbitrary. Then there exists P ∈ IP[a, b] such that, for any
choice ξ¯ of intermediate points subordinated to P ,
¯ − I(f )| < ε .
|σ(f ; P ; ξ)
3
Letting P = (xk )nk=0 and fixing k ∈ {1, ....n}, since
Mk := sup f and mk := inf f,
[xk−1 ,xk ] [xk−1 ,xk ]

by the approximation properties of infimum and supremum, there exists tk , uk ∈

[xk−1 , xk ] such that
ε ε
Mk − < f (tk ) and mk + > f (uk ),
6(b − a) 6(b − a)
hence
ε
Mk − mk − < f (tk ) − f (uk ).
3(b − a)
Then,
n
X
U (f ; P ) − L(f ; P ) = (Mk − mk )(xk − xk−1 )
k=1
n n
X ε X
< (f (tk ) − f (uk ))(xk − xk−1 ) + (xk − xk−1 )
k=1
3(b − a) k=1

n
X
where (xk − xk−1 ) = b − a, hence
k=1

ε
= σ(f ; P ; t̄) − σ(f ; P ; ū) +
3
ε
≤ |σ(f ; P ; t̄) − σ(f ; P ; ū)| +
3
ε
≤ |σ(f ; P ; t̄) − I(f )| + |σ(f ; P ; ū) − I(f )| +
3
ε ε ε
< + + = ε. 
3 3 3
As a conclusion, Riemann integrability is the same with Darboux integrability and
the Riemann integral is the same with the Darboux integral.
 3. EXISTENCE AND PROPERTIES OF THE RIEMANN INTEGRAL 123

Example 2.10. (1) The Dirichlet Function: d : [0, 1] → R, defined by

(
1, x ∈ Q ∩ [0, 1],
d(x) =
0, otherwise ,
is not Riemann integrable. To see this, observe that for any interval partition P =
(xk )nk=0
Mk = sup d = 1, mk = inf d = 0,
[xk−1 ,xk ] [xk−1 ,xk ]

hence U (d; P ) − L(d; P ) = 1 and, consequently, d is not Riemann integrable.

(2) The Riemann Function: r : [0, 1] → R, defined by
(

0, x ∈ (R \ Q) ∩ [0, 1] ∪ {0},
r(x) =
1/q, x = p/q, p, q ∈ N, (p, q) = 1,
has the following properties:
• r is continuous at each irrational number and at 0.
• r is discontinuous at each rational number
R 1 except 0.
• r is Riemann integrable on [0, 1] and 0 r(x)dx = 0.
We leave the proof of these facts as an exercise.

3. Existence and Properties of the Riemann Integral

Theorem 3.1. (1) (Continuous Functions are Integrable) If a < b and f : [a, b] → R
is a continuous function, then f is Riemann integrable on [a, b].
(2) (Monotone Functions are Integrable) If a < b and f : [a, b] → R is monotone
on [a, b], then f is integrable on [a, b].
(3) Let a < b and f : [a, b] → R a function.
(i) (Hereditary) If f is Riemann integrable on [a, b] then, for any c, d with a ≤
c < d ≤ b, f is integrable on [c, d].
(ii) (Interval Additivity) Let a < c < d. If f is Riemann integrable on [a, c] and
on [c, d], then f is Riemann integrable on [a, b] and
Z b Z c Z b
f (x)dx = f (x)dx + f (x)dx.
a a c
(4) (Piecewise Monotone Functions are Integrable) Assume that the function f : [a, b] →
R is bounded and there exists P = (xk )nk=1 ∈ IP[a, b] subject to the following properties:
for each k = 1, 2, ...., n, either f |(xk−1 , xk ) has a continuous extension to [xk−1 , xk ], or
f |[xk−1 , xk ] is monotone. Then f is Riemann integrable on [a, b].
(5) (Linearity of the Riemann Integral) Let a < b, f, g : [a, b] → R two integrable
functions, and α, β ∈ R. Then the function αf + βg is integrable on [a, b] and
Z b Z b Z b


αf (x) + βg(x) dx = α f (x)dx + β g(x)dx.
a a a
124 5. THE RIEMANN INTEGRAL

(6) (Comparison for Riemann Integrals) Let a < b and f, g : [a, b] → R be two
integrable functions such that,
f (x) ≤ g(x), x ∈ [a, b].
Then Z b Z b
f (x)dx ≤ g(x)dx.
a a

(7) (Estimations of Integrals) Let a < b and f : [a, b] → R an integrable function.

Then Z b
inf f · (b − a) ≤ f (x)dx ≤ sup f · (b − a)
[a,b] a [a,b]

(8) (Absolute Value of Integrable Functions) Let a < b and f : [a, b] → R be inte-
grable. Then |f | is integrable on [a, b] and
Z b Z b
f (x)dx ≤ |f (x)|dx.
a a

(9) (Product of Riemann integrable Functions) If a < b and f, g : [a, b] → R are

Riemann integrable on [a, b] then the product function f g is Riemann integrable on
[a, b] as well.

Proof. (1) If f is continuous on [a, b] then it is uniformly continuous on [a, b] and

bounded. Then, for any ε > 0 there exists δ > 0 such that, for any x, y ∈ [a, b] with
ε
|x − y| < δ we have |f (x) − f (y)| < 2(b−a) . Then, for all P ∈ IP[a, b] with kP k < δ,
n
letting P = (xk )k=1 , we have
n
X
U (f ; P ) − L(f ; P ) = (Mk − mk )(xk − xk−1 )
k=1
n
ε X ε ε
≤ (xk − xk−1 ) = (b − a) = < ε.
2(b − a) k=1 2(b − a) 2

(2) To make a choice, assume that f is non-decreasing, hence f is bounded and

for any n ∈ N let Pn = Pnu = (xk )nk=0 , xk = a + k b−a
n
, k = 0, 1, . . . , n, be the uniform
interval partition of order n. Then, for any k = 1, . . . , n, we have Mk = f (xk ) and
mk = f (xk−1 ), hence
n
X
U (f ; Pn ) − L(f ; Pn ) = (Mk − mk )(xk − xk−1 )
k=1
n n
X b−a b−aX
= (Mk − mk ) = (f (xk ) − f (xk−1 ))
k=1
n n k=1
f (b) − f (a)
= (b − a) −−−→ 0.
n n→∞
 3. EXISTENCE AND PROPERTIES OF THE RIEMANN INTEGRAL 125

(3) (i)For any ε > 0 there exists P ∈ IP[a, b] such that U (F ; P )−L(f ; P ) < ε. Let
P = P ∪ c, d and reorder it to become an interval partition of [a, b]. Then P 4 Pe.
e
Let Q = (c = y0 < .... < yk = d) be such that, for all j = 0, 1, .., k, yj ∈ Pe. Then,
U (f ; Q) − L(f ; Q) ≤ U (f ; Pe) − L(f ; Pe) ≤ U (f ; P ) − L(f ; P ) < ε.
(ii) Let P1 ∈ IP[a, c], t̄1 a choice of intermediate points subordinated to P1 , P2 ∈
IP, and t̄2 a choice of intermediate points subordinated to P2 . Then P = P1 ∪
P2 ∈ IP[a, b], t̄ = t̄1 ∪ t̄2 is a choice of intermediate points subordinated to P , and
σ(f ; P ; t̄) = σ(f ; P1 ; t̄1 ) + σ(f ; P2 ; t̄2 ) converges to
Z b Z c Z b
f (x)dx = f (x)dx + f (x)dx,
a a c

while kP k = max kP1 k, kP2 k → 0.
(4) This is a direct consequence of the assertions at the previous items.
(5) If either α = 0 of β = 0 we leave the proof to the reader, so let us assume that
α, β 6= 0. By assumption, for any ε > 0 there exists, Pε ∈ IP[a, b] such that, for any
P > Pε and any choice of intermediate points ξ¯ subordinated to P we have
Z b
¯ ε
σ(f ; P ; ξ) − f (x)dx <
a 2|α|
and
Z b
¯ − ε
σ(g; P ; ξ) g(x)dx < .
a 2|β|
Then, since
¯ = ασ(f ; P ; ξ)
σ(αf + βg; P, ξ) ¯ + βσ(g; P ; ξ),
¯
by means of the previous inequalities we get
Z b Z b
¯
|σ(αf + βg; P, ξ) − α f (x)dx − β g(x)dx|
a a
Z b Z b
¯
< |α| σ(f ; P ; ξ) − ¯
f (x)dx + |β| σ(g; P ; ξ) − g(x)dx
a a
ε ε
< + = ε.
2 2
(6) For any P ∈ IP[a, b] and any choice of intermediate points ξ¯ subordinated
¯ ≤ σ(g; P ; ξ).
to P , we have σ(f ; P ; ξ) ¯ Since f and g are integrable, after a moment
of thought we see that, for any ε > 0 there exists Pε ∈ IP[a, b] such that, for any
P ∈ IP[a, b] with Pε 4 P and for any choice of intermediate points ξ¯ subordinated to
P , we have
Z b Z b Z b
¯ ε ¯ ε ε ε
f (x)dx < σ(f ; P ; ξ) + < σ(g; P ; ξ) + < g(x)dx + + = (g(x)dx + ε.
a 2 2 a 2 2 a
126 5. THE RIEMANN INTEGRAL

This implies
Z b Z b
f (x)dx ≤ g(x)dx.
a a

(7) This is a direct consequence of the assertion at the previous item.

(8) By the Triangle Inequality, for any x, y ∈ [a, b]
|f (x)| − |f (y)| ≤ |f (x) − f (y)|,
hence, letting P = (xk )nk=0 ∈ IP[a, b], for any k = 1, . . . , n, we have
Mk (|f |) − mk (|f |) = sup |f (x)| − inf |f (y)|
x∈[xk−1 ,xk ] y∈[xk−1 ,xk ]

≤| sup f (x) − inf f (y)|

x∈[xk−1 ,xk ] y∈[xk−1,xk ]

= Mk (f ) − mk (f ).
Then,
U (|f |; P ) − L(|f |; P ) ≤ U (f ; P ) − L(f ; P ),
and, consequently, since f is integrable, by Darboux Theorem, it follows that |f | is
integrable on [a, b]. Since −|f | ≤ f ≤ |f | on [a, b] by Comparison Theorem we get
Z b Z b Z b
− |(f (x)|dx ≤ (f (x)dx ≤ |(f (x)|dx.
a a a

(9) We divide the proof in two steps. Firstly, assume that f = g. Since f 2 = |f |2
and |f | is integrable, without loss of generality we can assume f ≥ 0. Let P = (xk )nk=0 ∈
IP[a, b]. Since
f (x)2 − f (y)2 = f (x) − f (y) f (x) + f (y) ,

for all k = 1, . . . , n we get

Mk (f 2 ) − mk (f 2 ) ≤ Mk (f ) − mk (f ) 2M,

where M = sup[a,b] f . This implies that

ε
U (f 2 ; P ) − L(f 2 ; P ) ≤ 2M U (f ; P ) − L(f ; P ) < 2M ·


= ε,
2M
provided that M > 0 and P is chosen such that U (f ; P ) − L(f ; P ) < ε/M . For M = 0
the assertion is clear. Hence the assertion is proven for f = g.
In general,
(f + g)2 − (f − g)2
f ·g =
4
and then we use the algebraic properties of Riemann integrability and the assertion
proven at the first step. 
 4. MORE PROPERTIES OF THE RIEMANN INTEGRAL 127

4. More Properties of the Riemann Integral

Fact 4.1 (Change of Variables in the Riemann Integral). Let a < b, ϕ : [a, b] → R
be a continuously differentiable function such that ϕ0 (x) 6= 0 for all x ∈ [a, b], and let
f : [c, d] = ϕ([a, b]) → R be a function integrable on [c, d]. Then, (f ◦ ϕ)|ϕ0 | is integrable
on [a, b] and
Z d Z b
f (t)dt = f (ϕ(x))|ϕ0 (x)|dx.
c a

Proof. By the Intermediate Value Property for derivative functions it follows that
either φ0 (x) > 0 for all x ∈ [a, b] or φ0 (x) < 0 for all x ∈ [a, b]. Without loss of generality,
in the following we assume the first case, hence φ is increasing on [a, b].
We prove that, for any  > 0 there exists δ > 0 such that for all P ∈ IP[a, b] such
that kP k < δ and for all choices ξ of intermediated points subordinate to P we have
Z φ(b)
0
(4.1) σ((f ◦ φ)|φ |, P, ξ) − f (t)dt < .
φ(a)

For arbitrary P = (xk )nk=0 ∈ IP[a, b] and ξ = (ξk )nk=1 a choice of intermediate points
subordinated to P , we have
n
X
0
σ((f ◦ φ)|φ , P, ξ) = f (φ(ξk ))φ0 (ξk )(xk − xk−1 ).
k=1

Since φ is increasing, φ(P ) = (φ(xk ))nk=0 is an interval partition of [φ(a), φ(b)] and,
since φ is a Rolle function on each interval [xk−1 , xk ], there exists tk ∈ (xk−1 , xk ) such
that
φ(xk ) − φ(xk−1 ) = φ0 (tk )(xk − xk−1 ), k = 1, . . . , n,
hence,
n
X

(4.2) σ(f, φ(P ), φ(ξ)) = f (φ(ξk )) φ(xk ) − φ(xk−1 )
k=1
n
X
(4.3) = f (φ(ξk ))φ0 (tk )(xk − xk−1 ).
k=1

In addition, on the one hand, since φ ∈ C 1 [a, b] it follows that it is a Lipschitz

function hence, there exists N1 > 0 such that,
|φ(x) − φ(y)| ≤ N1 |x − y|, x, y ∈ [a, b],
which implies that kφ(P )k ≤ N1 kP k. On the other hand, by the Inverse Function
Theorem φ−1 ∈ C 1 [φ(a), φ(b)], hence φ−1 is a Lipschitz function, that is, there exists
N2 > 0 such that,
|φ−1 (u) − φ−1 (t)| ≤ N2 |u − t|, u, t ∈ [φ(a), φ(b)],
128 5. THE RIEMANN INTEGRAL

hence kP k ≤ N2 kφ(P )k. In conclusion,

1
kP k ≤ kφ(P )k ≤ N1 kP k,
N2
hence kP k → 0 if and only if kφ(P )k → 0.
Since φ0 is continuous on [a, b], and hence uniformly continuous, it follows that there
exists η > 0 such that for all x, y ∈ [a, b] such that |x − y| < η we have

(4.4) |φ0 (x) − φ0 (y)| < ,
2M (b − a)
where M = sup[φ(a),φ(b)] |f | < ∞, since f is integrable
Finally, since f is integrable on [φ(a), φ(b)], there exists δ with 0 < δ ≤ η such that
for all kP k < δ, we have
Z φ(b)
0
σ((f ◦ φ)|φ |, P, ξ) − f (t)dt ≤ σ((f ◦ φ)|φ0 |, P, ξ) − σ(f, φ(P ), φ(ξ))
φ(a)
Z φ(b)
+ σ(f, φ(P ), φ(ξ)) − f (t)dt
φ(a)

which, in view of (4.1), equals

n
X
f (φ(ξk )) φ0 (ξk ) − φ0 (tk ) (xk − xk−1 )


=
k=1
Z φ(b)
+ σ(f, φ(P ), φ(ξ)) − f (t)dt
φ(a)
n
X
≤ |f (φ(ξk ))| · φ0 (ξk ) − φ0 (tk ) (xk − xk−1 )
k=1
Z φ(b)
+ σ(f, φ(P ), φ(ξ)) − f (t)dt
φ(a)

which, in view of (4.4), is

n
 X
<M (xk − xk−1 )
2M (b − a) k=1
Z φ(b)
+ σ(f, φ(P ), φ(ξ)) − f (t)dt
φ(a)
 
< + = . 
2 2
As an application of the Formula for the Change of Variables in the Riemann
Integral, we present the Weierstrass’ Approximation Theorem.
 4. MORE PROPERTIES OF THE RIEMANN INTEGRAL 129

Theorem 4.2 (Weierstrass’ Approximation Theorem). For any a < b and any
continuous function f : [a, b] → R there exists a sequence of polynomial functions (pn )n
that converges uniformly to f on [a, b].

Proof. Step 1. Without loss of generality we can assume that [a, b] = [0, 1].

Indeed, consider the function h : [0, 1] → [a, b], defined by h(t) = (1 − t)a + tb
and observe that h is a homeomorphism, that is, it is continuous, bijective, and its
inverse, the polynomial function h−1 (s) = s/(b − a) − a/(b − a), is continuous as well.
Let g := f ◦ h : [0, 1] → R and observe that g is a continuous function. If (rn )n is a
sequence of polynomial functions that converges uniformly to g on [0, 1] then, letting
pn := rn ◦ h−1 it follows that the sequence of polynomials (pn )n converges uniformly to
f on [a, b].

Step 2. Without loss of generality, we can assume that f (0) = f (1) = 0.

Indeed, let g(t) := f (t) − f (0) − t(f (1) − f (0)) for all t ∈ [0, 1] and observe that
g is a continuous function on [0, 1] and g(0) = g(1) = 0. Then, f (t) = g(t) + f (0) +
t(f (1) − f (0)) for all g ∈ [0, 1] and, if (rn )n is a sequence of polynomial functions that
converges to f uniformly on [0, 1] then, letting pn (t) := rn (t) + f (0) + t(f (1) − f (0)) for
all t ∈ [0, 1], we have a sequence of polynomial functions (pn )n that converges uniformly
to f .

Step 3. The function f has the extension f (t) = 0 for all t ∈ R \ [0, 1], which is
uniformly continuous on R.

We leave this simple fact to be proven by the reader.

Step 4. Let
qn (t) := cn (1 − t2 )n , t ∈ R, n ∈ N,
where Z 1
cn := 1/ (1 − t2 )n dt, n ∈ N,
−1
and hence Z 1
qn (t)dt = 1, n ∈ N.
−1
√
We have cn < n for all n ∈ N.

Indeed, since t 7→ (1 − t2 )n is an even function we have

Z 1 Z 1 Z 1/√n Z 1
2 n 2 n 2 n
(1 − t ) dt = 2 (1 − t ) dt = 2 (1 − t ) dt + 2 √
(1 − t2 )n dt
−1 0 0 1/ n
√
Z 1/ n
≥2 (1 − t2 )n dt
0
130 5. THE RIEMANN INTEGRAL

which, by the Bernoulli Inequality (1 − t2 )n ≥ 1 − nt2 on [0, 1], is

√
√ 1/ n
Z 1/ n 3

nt 4 1
≥2 (1 − nt2 )dt = 2 t − = √ >√ .
0 3 3 n n
0

Step 5. Define Z 1
pn (x) := f (x + t)qn (t)dt, x ∈ R.
−1
Then pn is a polynomial function for all n ∈ N.
Indeed, we firstly observe that, by the way f was extended to R, for each x ∈ [0, 1]
we have f (x + t) = 0 anytime x + t 6∈ [0, 1], that is, t 6∈ [−x, 1 − x]. Then, performing
the change of variables φ(t) := x + t, φ0 (t) = 1, it follows that
Z 1−x Z 1
pn (x) = f (x + t)qn (t)dt = f (s)qn (s − x)ds, x ∈ [0, 1],
−x 0

and, since x 7→ qn (s − x) is a polynomial function, by the additivity of the Riemann

integral it follows that pn is a polynomial function.
Step 6. The sequence (pn )n converges uniformly to f on [0, 1].
Indeed, since f is uniformly continuous, for any  > 0 there exists δ > 0 such that,
for all x, y ∈ R such that |x − y| < δ we have |f (x) − f (y)| < /2. Then, letting
M := sup[0,1] |f |, we have
Z 1


|pn (x) − f (x)| = f (x + t) − f (x) qn (t)dt
−1
Z 1
≤ f (x + t) − f (x) qn (t)dt
−1
Z δ Z
= f (x + t) − f (x) qn (t)dt + f (x + t) − f (x) qn (t)dt
−δ 1≥|t|≥δ
Z δ √

≤ qn (t)dt + 2M n(1 − δ 2 )n
2 −δ
√
which, taking into account that the sequence n(1 − δ 2 )n converges to 0, for n suffi-
ciently large is
 
< + = . 
2 2
Theorem 4.3 (Integral Form of Jensen’s Inequality). Let f : [0, 1] → [a, b] be an
integrable function, for some a < b, and let φ : [a, b] → R be a convex function such
that φ ◦ f is integrable on [0, 1]. Then
Z 1  Z 1
φ f (x)dx ≤ (φ ◦ f )(x)dx.
0 0
 4. MORE PROPERTIES OF THE RIEMANN INTEGRAL 131

We provide two different proofs for this theorem.

First Proof of Jensen’s Inequality. We observe, by estimation of the inte-

gral, that
Z 1
a ≤ inf f ≤ c := f (x)dx ≤ sup f ≤ b,
[0,1] 0 [0,1]

hence the left hand side of the inequality to be proven makes sense. We firstly prove
that, for all y ∈ [a, b], we have
 
(4.5) φ(c) + s f (y) − c ≤ (φ ◦ f )(y),

where
φ(c) − φ(x)
s := sup .
x∈[a,c) c−x
To this end, we firstly observe that, by Proposition 5.4 in Chapter 3, for any
x ∈ [a, c) and any y ∈ (c, b], we have
φ(c) − φ(x) φ(y) − φ(c)
≤ ,
c−x y−c
hence the sup in the definition of s is finite and
φ(y) − φ(c)
s≤ , y ∈ (c, b],
y−c
equivalently
(4.6) φ(c) + s(y − c) ≤ φ(y), y ∈ (c, b].
Since, by the definition of s we clearly have
φ(c) − φ(y)
s≥ , y ∈ [a, c),
c−y
it follows that actually (4.6) holds true for all y ∈ [a, b]. Letting y = f (x) ∈ [a, b] in
(4.6), we obtain (4.5).
Finally, we apply the Comparison Theorem for Integrals and integrate both sides
of (4.5) and get
Z 1  Z 1 Z 1  Z 1
φ f (x)dx + s f (y)dy − f (x)dx ≤ (φ ◦ f )(y)dy,
0 0 0 0

which is the required inequality. 

Second Proof of Jensen’s Inequality. This proof is based on the Discrete

Form of Jensen’s Inequality, see Theorem 5.7 in Chapte 3. To this end, for each
n ∈ N we consider the uniform interval partition Pn of [0, 1] given by the points
0 < 1/n < 2/n < · · · < (n−1)/n < 1 and for each k = 1, . . . , n let xk ∈ [(k−1)/n, k/n],
132 5. THE RIEMANN INTEGRAL

a choice of intermediate points subordinated to this partition. Then we observe the

following Riemann sums
f (x1 ) f (xn )
σ(f, Pn , (xj )j ) = + ··· +
n n
and
φ(f (x1 )) φ(f (xn ))
σ(φ ◦ f, Pn , (xj )j ) = + ··· + ,
n n
and using Theorem 5.7 in Chapter 3 we get
φ(σ(f, Pn , (xj )j ) ≤ σ(φ ◦ f, Pn , (xj )j ), n ∈ N.
R1
Since f and φ ◦ f are integrable and φ is continuous on (a, b), assuming that 0 f (x)dx
is neither a nor b, it follows let we can let n → +∞ on both sides of the previous
inequality and, by the Comparison Theorem for limits of sequences, we obtain
Z 1  Z 1
φ f (x)dx ≤ (φ ◦ f )(x)dx.
0 0
R1
In case 0 f (x)dx equals either a or b, we leave to the reader to show that the Jensen’s
Inequality still follows from here. 
Remark 4.4. The two forms of the Jensen’s Inequality, the discrete form and
the integral form, are actually logically equivalent, in the sense that each one is a
consequence of the other. In view of the second proof, we only have to prove that
the integral form implies the discrete form. To this end, let φ : [a, b] → R be a convex
function and let y1 , . . . , yn ∈ [a, b] and λ1 > 0, . . . , λn > 0 be such that λ1 +· · ·+λn = 1.
We consider the interval partition P = (xj )nj=0 of [0, 1] given by the points
0 < λ1 < λ1 + λ2 < · · · < λ1 + · · · + λn−1 < λ1 + · · · + λn = 1,
that is, (
0, j = 0,
xj =
λ1 + · · · + λj , j = 1, . . . , n.
Then we define the function f : [0, 1] → [a, b] by
f (x) = y1 χ[x0 ,x1 ) (x) + y2 χ[x1 ,x2 ) + · · · + yn χ[xn−1 ,xn ] , x ∈ [0, 1].
Observing that
(φ ◦ f )(x) = φ(y1 )χ[x0 ,x1 ) (x) + φ(y2 )χ[x1 ,x2 ) + · · · + φ(yn )χ[xn−1 ,xn ] , x ∈ [0, 1],
and that both f and φ ◦ f are integrable on [0, 1], by using the Integral Form of the
Jensen’s Inequality we get
Z 1 
φ(λ1 y1 + · · · + λn yn ) = φ f (x)dx
0
Z 1
≤ (φ ◦ f )(x)dx = λ1 φ(y1 ) + · · · + λn φ(yn ),
0
hence the Discrete Form of the Jensen’s Inequality is proven.
 4. MORE PROPERTIES OF THE RIEMANN INTEGRAL 133

Because the Riemann

Rb integral is not multiplicative, it is of interest to obtain simpler
representations of a f (x)g(x)dx.
Fact 4.5 (First Mean Value Theorem). Let f and g be two integrable functions on
[a, b] and g(x) ≥ 0 for all x ∈ [a, b]. Then, there exists c ∈ R such that
inf f ≤ c ≤ sup f
[a,b] [a,b]

and Z b Z b
f (x)g(x)dx = c g(x)dx.
a a

In particular if, in addition, f is continuous on [a, b] then, there exists x0 ∈ [a, b]

such that Z b Z b
f (x)g(x)dx = f (x0 ) g(x)dx.
a a

Proof. Since m := inf [a,b] f ≤ f (x) ≤ M := sup[a,b] f and g(x) ≥ 0 for all x ∈ [a, b],
it follows that
mg(x) ≤ f (x)g(x) ≤ M g(x), x ∈ [a, b],
hence, by the Comparison Theorem for Integrals, we have
Z b Z b Z b
m g(x)dx ≤ f (x)g(x)dx ≤ M g(x)dx.
a a a
Rb Rb
If a
g(x)dx = 0, it follows that a f (x)g(x)dx = 0 and take any c ∈ [m, M ]. If
Rb Rb
a
g(x)dx 6= 0 then a g(x)dx > 0, hence
Rb
f (x)g(x)dx
m ≤ aR b ≤ M,
a
g(x)dx
and let Rb
a
f (x)g(x)dx
c := Rb .
a
g(x)dx
In case f is continuous, the assertion is an application of the Intermediate Value
Theorem for continuous functions. 
Fact 4.6 (Second Mean Value Theorem). Let f and g be two functions integrable
on [a, b] such that g(x) ≥ 0 for all x ∈ [a, b] and let m, M ∈ R be such that
m ≤ inf f ≤ sup f ≤ M.
[a,b] [a,b]

Then there exists c ∈ [a, b] such that

Z b Z c Z b
f (x)g(x)dx = m g(x)dx + M g(x)dx.
a a c
134 5. THE RIEMANN INTEGRAL

If, in addition, f (x) ≥ 0 for all x ∈ [a, b], then there exists c ∈ [a, b] such that
Z b Z b
f (x)g(x)dx = M g(x)dx.
a c

In order to prove this fact, we need a preliminary result on the continuity of the
indefinite integral.
Definition 4.7. Let a < b and f : [a, b] → R be an integrable function. By the
Interval Additivity Property of the Riemann Integral, for any x ∈ [a, b] the function f
is integrable on [a, x] and hence, we can define the function F : [a, b] → R, the indefinite
integral of f , by Z x
F (x) := f (t)dt, x ∈ [a, b].
a

Lemma 4.8. If f : [a, b] → R is a function integrable on [a, b] then, letting F : [a, b] →

R be the indefinite integral of f , the function F is Lipschitz, hence (uniformly) contin-
uous on [a, b].

Proof. Let a ≤ x0 < x ≤ b. Then, by interval additivity,

Z x Z x0 Z x Z x
|F (x) − F (x0 )| = f (t)dt − f (t)dt = f (t)dt ≤ |f (t)|dt ≤ M (x − x0 ),
a a x0 x0

where M := sup[a,b] |f | ∈ R+ .
Letting a ≤ x < x0 ≤ b, similarly we prove
F (x) − F (x0 ) ≤ M (x0 − x).
These show that, in general, for any x, y ∈ [a, b] we have
F (x) − F (y) ≤ M |x − y|,
hence F is Lipschitz and, consequently, it is uniformly continuous on [a, b]. 

R a Definition 4.9. For a < b and an integrable function f : [a, b] → R, we let

a
f (x)dx := 0 and we let
Z a Z b
f (x)dx := − f (x)dx.
b a

Proof of the Second Mean Value Theorem. Let F : [a, b] → R be defined

by
Z x Z b
F (x) = m g(t)dt + M g(t)dt.
a x
By the previous lemma, F is continuous on [a, b] and we observe that, by the Compar-
ison Theorem,
Z b Z b Z b
F (b) = m g(t)dt ≤ f (t)g(t)dt ≤ M g(t)dt = F (a),
a a a
 5. THE FUNDAMENTAL THEOREM OF CALCULUS AND APPLICATIONS 135

hence, by the Intermediate Value Theorem applied to F , there exists c ∈ [a, b] such
that Z b Z c Z b
f (t)g(t)dt = F (c) = m g(t)dt + M g(t)dt.
a a c
If f (x) ≥ 0 for all x ∈ [a, b], it follows that inf [a,b] f ≥ 0, hence we can take m = 0
and apply the fact just proven. 

5. The Fundamental Theorem of Calculus and Applications

The Fundamental Theorem of Calculus establishes connections between the oper-
ations of derivation and integration. Once again, this a particular case of the Main
Problem in Analysis, which is ”Changing the Order of Taking Two Limits”.
Theorem 5.1 (The Fundamental Theorem of Calculus). Let a < b and f : [a, b] →
R be a function.
R x (i) If f is continuous on [a, b] then, its indefinite integral, F : [a, b] → R, F (x) :=
a
f (t)dt for all x ∈ [a, b], is continuously differentiable on [a, b], i.e. F is differentiable
and F 0 is continuous, and
Z x
0 d
F (x) = f (t)dt = f (x), x ∈ [a, b].
dx a
(ii) If f is differentiable on [a, b] and its derivative f 0 is integrable on [a, b], then
Z x
f 0 (t)dt = f (x) − f (a), x ∈ [a, b]. (Leibniz- Newton)
a

Proof. (i) For arbitrary x0 ∈ [a, b] and h 6= 0, sufficiently small (if x0 = a we

allow only h > 0 and if x0 = b we allow only h < 0) we have
 Z x0 +h Z x0
1 x0 +h

F (x0 + h) − F (x0 )
Z
1
− f (x0 ) = f (t)dt − f (t)dt − f (x0 )dt
h h a a h x0
1 x0 +h
Z


= f (t) − f (x0 ) dt.
h x0
Since f is continuous, for all ε > 0, there exists δ > 0 such that, for all h ∈ R with
|h| < δ then |f (x0 + h) − f (x0 )| < 2ε , hence
ε
sup |f (t) − f (x0 )| ≤ < ε.
|t−x0 |<δ 2
Finally, for all h ∈ R with |h| < δ, h 6= 0 (h > 0 if x0 = a and h < 0 if x0 = b) we
have
Z x0 +h
F (x0 + h) − F (x0 ) 1
| − f (x0 )| = | f (t) − f (x0 )dt|
h |h| x0
Z x0 +h
1
≤ | |f (t) − f (x0 )|dt|
|h| x0
136 5. THE RIEMANN INTEGRAL

1
≤ |x0 + h − x0 | sup |f (t) − f (x0 )|
|h| [t−x0 ]<h

|h|
< ε = ε.
|h|
This shows that
F (x0 + h) − F (x0 )
lim = f (x0 ),
h→0 h
hence F is differentiable at x0 and F 0 (x0 ) = f (x0 ).
(ii) Assume that f is differentiable on [a, b]. Let P = (xj )nj=1 be an interval partition
of [a, b]. Then,
Z b n
X Z b
0
|f (b) − f (a) − f (t)dt| = | f (xj ) − f (xj−1 ) − f 0 (t)dt|
a j=1 a

and, by the Lagrange Mean Value Theorem for f on [xj−1 , xj ], j = 1, . . . , n, there exists
tj ∈ (xj−1 , xj ) such that
f (xj ) − f (xj−1 ) = f 0 (tj ) · (xj − xj−1 ).
As a consequence, t̄ := (tj )nj=1 is a choice of intermediate points subordinated to P and
Z b Z b
0 0
|f (b) − f (a) − f (t)dt| = |σ(f ; P, t̄) − f 0 (t)dt|.
a a

If ε > 0 then there exists Pε ∈ IP[a, b] such that, for all P ∈ IP[a, b] with Pε 4 P and
any choices t̄ of intermediate points subordinated to P ,
Z b
0
|σ(f ; P, t̄) − f 0 (t)dt| < ε,
a
hence,
Z b
f 0 (t)dt = f (b) − f (a).
a
By the Interval Additivity, the same conclusion holds for any a ≤ x ≤ b, hence
Z x
f 0 (t)dt = f (x) − f (a). 
a

Fact 5.2 (Integration by Parts). Let a < b and f, g : [a, b] → R differentiable on

[a, b] and f 0 and g 0 integrable on [a, b]. Then f g 0 and f 0 g are integrable on [a, b] and
Z b Z b
0
f (x)g(x)dx = f (b)g(b) − f (a)g(a) − f (x)g 0 (x)dx.
a a

Proof. f g 0 and f 0 g are integrable, since the product of integrable functions is

integrable. Hence, by algebraic properties, the function
(f (x)g(x))0 = f 0 (x)g(x) + f (x)g 0 (x)
 5. THE FUNDAMENTAL THEOREM OF CALCULUS AND APPLICATIONS 137

is integrable. Then, by the Fundamental Theorem of Calculus,

Z b Z b Z b
0 0
f (x)g(x)dx + f (x)g (x)dx = (f (x)g(x))0 dx = f (b)g(b) − f (a)g(a). 
a a a

Fact 5.3 (Taylor’s Theorem with Integral Form of the Remainder). If I is on open
interval n ∈ N0 , and f : I → R is an (n + 1)-times differentiable function on I such
that f (n+1) is continuous on I then, for all x0 , x ∈ I, we have
1 x
Z
Rn,x0 (x) = (x − t)n f (n+1) (t)dt,
n! x0
where Rn,x0 is the Taylor remainder of f and x0 of order n.
Proof. We apply mathematical induction on n ∈ N. For n = 0 by the Fundamen-
tal Theorem of Calculus we have
Z x
f (x) − f (x0 ) = f 0 (t)dt.
x0

Assuming that the statement holds for n, we have

f (n+1) (x0 )
Rn+1,x0 (x) = Rn,x0 (x) − (x − x0 )n+1
(n + 1)!
1 x f (n+1) (x0 )
Z
= (x − t)n f (n+1) (t)dt − (x − x0 )n+1
n! x0 (n + 1)!
Z x
1 1 d f (n+1) (x0 )
= − (x − t)n+1 · f (n+1) (t)dt − (x − x0 )n+1
n! x0 n + 1 dt (n + 1)!
hence, by integration by parts we get
x Z x
1 1
=− (x − t)n+1 f (n+1) (t) + (x − t)n+1 f (n+2) (t)dt
n+1 x0 (n + 1)! x0
(n+1)
f (x0 )
− (x − x0 )n+1
(n + 1)!
Z x
f (n+1) (x0 ) n+1 1
= (x − x0 ) + (x − t)n+1 f (n+2) (t)dt
(n + 1)! (n + 1)! x0
f (n+1) (x0 )
− (x − x0 )n+1
(n + 1)!
Z x
1
= (x − t)(n+1) f (n+2) (t)dt. 
(n + 1)! x0

Fact 5.4 (Indefinite Integral with Variable Endpoints). Let f : [a, b] → R be a

continuous function and let ϕ, ψ : [c, d] −→ [a, b] be two differentiable functions. Then
Z ψ(x)
[c, d] 3 x → f (t)dt ∈ R
ϕ(x)
138 5. THE RIEMANN INTEGRAL

is a differentiable function and the following Generalised Leibniz-Newton Formula holds:

Z ψ(x)
d
f (t)dt = f (ψ(x))ψ 0 (x) − f (ϕ(x))ϕ0 (x).
dx ϕ(x)
Ry
Proof. Let F : [a, b] → R, F (y) = a f (t)dt. Then, the function F ◦ ϕ : [c, d] → R
defined by
Z ϕ(x)
F (ϕ(x)) = f (t)dt, x ∈ [c, d],
a
is differentiable and, by the Chain Rule and the Fundamental Theorem of Calculus,
we have
Z ϕ(x)
d
f (t)dt = ϕ0 (x)F 0 (ϕ(x)) = ϕ0 (x)f (ϕ(x)).
dx a
Then observe that
Z ψ(x)
f (t)dt = F (ψ(x)) − F (ϕ(x)),
ϕ(x)

hence, this is a differentiable function and

Z ψ(x)
d
f (t)dt = f (ψ(x))ψ 0 (x) − f (ϕ(x))ϕ0 (x). 
dx ϕ(x)
Fact 5.5 (Change of Variables in the Riemann Integral for Continuous Func-
tions). Let a < b, let the function ϕ : [a, b] → R be continuously differentiable, and
let f : ϕ([a, b]) → R be a continuous function. Then,
Z ϕ(b) Z b
f (t) = f (ϕ(x))ϕ0 (x)dx.
ϕ(a) a

Proof. Consider the functions:

Z ϕ(x) Z x
[a, b] 3 x 7→ f (t)dt and [a, b] 3 x 7→ f (ϕ(t))ϕ0 (t)dt.
ϕ(a) a

By the Fundamental Theorem of Calculus and the Indefinite Integral with Variable
Endpoints, both F and G are differentiable and have the same derivatives,
d d
F (x) = f (ϕ(x))ϕ0 (x) = G(x),
dx dx
hence, there exists k ∈ R such that F (x) = G(x) + k for all x ∈ [a, b]. Letting x = a it
follows that k = 0. 

We are now in a position to show an alternative procedure to define and prove the
properties of the transcendental functions.
 5. THE FUNDAMENTAL THEOREM OF CALCULUS AND APPLICATIONS 139

Example 5.6. (1) The Logarithm. Consider the function (0, +∞) 3 x 7−→ x1 which
is continuous, hence integrable on any compact interval [a, b] ⊂ (0, +∞). Define
Z x
dt
log : (0, +∞) −→ R by log x := , x > 0.
1 t
Then:
(i) log is a differentiable function, (log x)0 = 1
x
, x > 0, hence increasing, and
log 1 = 0.
(ii) lim log x = −∞ , lim log x = +∞.
x→0+ x→+∞
(iii) log xq = q log x, for all x > 0 and all q ∈ Q.
(iv) log xy = log x + log y, for all x, y > 0.
1
(v) Letting e = lim (1 + )n , we have log e = 1.
n→∞ n
(2) The Exponential. Since log : (0, +∞) → R is continuous, increasing, limx→0+ log x =
−∞, and limx→+∞ log x = +∞, it follows that it is bijective, hence log−1 : R → (0, +∞)
exists as the inverse function. Define,
exp : R → (0, +∞) by exp(x) := log−1 x, x ∈ R,
the inverse function of log. Then,
d
(i) exp is a differentiable function, dx exp(x) = exp(x), x ∈ R, hence the function
exp is strictly increasing, exp(0) = 1, and exp(1) = e.
(ii) lim exp(x) = 0 and lim exp(x) = +∞.
x→−∞ x→+∞
(iii) exp(x + y) = exp(x) exp(y) for all x, y ∈ R.
(iv) exp(q) = eq for all q ∈ Q.
Because of (iv), we use the notation ex := exp(x).

(3) The Power Function. For any α ∈ R consider the function

pα : (0, +∞) → (0, +∞) defined by pα (x) = eα log x , x > 0.

Then,
d
(i) pα is differentiable, dx pα (x) = αpα−1 (x), for all x > 0, hence
• If α > 0, then pα is increasing.
• If α < 0, then pα is decreasing.
• If α = 0, then pα (x) = 1, for all x > 0.
(ii) pα+β (x) = pα (x) pβ (x), for all x > 0.
(iii) p−α (x) = pα1(x) , for all x > 0.
(iv) pq (x) = xq , for all x > 0 and all q ∈ Q.
Because of (iv), we use the notation pα (x) =: xα , for all x > 0 and all α ∈ R.
140 5. THE RIEMANN INTEGRAL

6. Integrability of Sequences and Series of Functions

Fact 6.1 (Integrability of Sequences of Functions). Let a < b and (fn )n≥1 be a
unif.
sequence of integrable functions on [a, b] such that fn −−→ f , where f : [a, b] → R.
n
Then, f is integrable on [a, b] and
Z b Z b Z b
lim fn (x)dx = f (x)dx = lim fn (x)dx.
n→∞ a a a n→∞

Proof. Let ε > 0 be arbitrary. Since fn →

− f uniformly on [a, b], it follows that
n
there exists N ∈ N such that for all n ∈ N, if n ≥ N and x ∈ [a, b] then |fn (x)−f (x)| <
ε
3(b−a)
. Let P = (xj )kj=0 be an interval partition of [a, b]. Then, for all j = 1, . . . , k we
2ε
have Mj (f − fN ) − mj (f − fN ) ≤ 3(b−a) hence U (f − fN ; P ) − L(f − fN ; P ) ≤ 2ε 3
.
Since fN is integrable on [a, b], there exists P ∈ J P[¸a, b] such that,
ε
U (fN ; P ) − L(fN ; P ) < .
3
Then,
2ε ε
U (f ; P ) − L(f ; P ) ≤ U (f − fN ; P ) − L(f − fN ; P ) + U (fN ; P ) − L(fN ; P ) < + = ε,
3 3
hence, f is integrable on [a, b]. On the other hand, for all n ≥ N,
Z b Z b Z b
fn (t)dt − f (t)dt ≤ (fn (t) − f (t))dt
a a a
Z b
≤ fn (t) − f (t) dt
a
ε ε
≤ (b − a) = < ε,
3(b − a) 3
hence Z b Z b
lim fn (x)dx = f (x)dx. 
n→∞ a a

Fact 6.2 (Integrabililty of Series of Functions). Let a ≤ b and let fn : [a, b] → R be

a sequence of functions such that,
• For all n ∈ N, fn is integrable on [a, b].
X∞
• fn converges uniformly on [a, b] to a function f .
n=1

Then:
• f is integrable on [a, b].
∞ Z b
Rb X
• a f (x)dx = fn (x)dx.
n=1 a
 7. THE IMPROPER RIEMANN INTEGRAL 141

Proof. This is a consequence of the Integrability Theorem for sequences of func-

tions. 
∞
X
Corollary 6.3 (Integrability of Power Series). Let an (x − x0 )n be a power
n=0
series such that its interval of convergence is non degenerate, equivalently its radius of
convergence R > 0. Let
∞
X
f (x) := an (x − x0 )n , x ∈ (x0 − R, x0 + R).
n=0

Then f is integrable on any compact interval contained in (x0 − R, x0 + R) and

Z x ∞
X an
f (t)dx = (x − x0 )n+1 , for all x ∈ (x0 − R, x0 + R),
x0 n=0
n+1

with radius of convergence R as well.

∞
X
Example 6.4. Find a power series ak xk which converges to log(1 + x) for x ∈
k=0
(−ε, ε) and find the largest ε > 0 with this property.
1
= ∞ k
P
For |x| < 1 we have the geometric series 1−x k=0 x . Hence, letting t = −x,
|t| < 1, we have
∞ ∞
1 X X
= (−t)k = (−1)k tk .
1 + t k=0 k=0

By the Integrability Theorem we have  = 1 and

Z x ∞ Z x ∞
dt X
k k
X xk+1
log(1 + x) = = (−1) t dt = (−1)k
0 1+t k=0 0 k=0
k+1

∞
X xk
= (−1)k+1 , |x| < 1 i.e. x ∈ (−1, 1).
k=1
k

7. The Improper Riemann Integral

Definition 7.1. Let a, b ∈ ext(R), a < b, and let f : (a, b) → R be a function.

(i) f is locally integrable on (a, b) if, for all a < c < d < b, f is integrable on [c, d].
(ii) f is improperly Riemann integrable on (a, b) if,
• f is locally integrable on (a, b);
Z d
Rb
• lim lim f (x)dx exists in R and is denoted by a f (x)dx.
c→a+ d→b− c
142 5. THE RIEMANN INTEGRAL

Remark 7.2. If f is improperly integrable on (a, b) then, for all a < x0 < b fixed,
Z b Z d Z x0 Z d
f (x)dx = lim lim f (x)dx = lim f (x)dx + lim f (x)dx.
a d→b− c→a+ c c→a+ a d→b− x0

This shows that, in the definition of the improper Riemann integral, we can take any
order of the limits.
Example 7.3. Let p ∈ R. Then:
• xp is improperly integrable on [1, +∞) if and only if p < −1.
• xp is improperly integrable on (0, 1] if and only if p > −1.
Fact 7.4. (1) (Linearity). Let a, b ∈ ext(R), a < b, and let f, g be two functions im-
properly integrable on (a, b). Then, for all α, β ∈ R, the function αf + βg is improperly
integrable on (a, b) and
Z b Z b Z b


αf (x) + βg(x) dx = α f (x)dx + β g(x)dx.
a a a

(2) (Strong Comparison Theorem). Let a, b ∈ ext(R), a < b, and let f, g : (a, b) → R
be two locally integrable functions such that 0 ≤ f (x) ≤ g(x), for all x ∈ (a, b), and g
is improperly integrable on (a, b). Then f is improperly integrable on (a, b) and
Z b Z b
f (x)dx ≤ g(x)dx.
a a

(3) (Soft Comparison Theorem). Let a, b ∈ ext(R), a < b, and let f, g : (a, b) → R
be two improperly integrable functions such that f (x) ≤ g(x), for all x ∈ (a, b). Then,
Z b Z b
f (x)dx ≤ g(x)dx.
a a

Proof. (1) This is a consequence of the definition and the linearity of the Riemann
integral.
(2) Fix x0 ∈ (a, b).
Rx Rb
• [x0 , b) 3 x → x0 f (t)dt ∈ R+ is nondecreasing and upper bounded by x0 g(t)dt
hence Z x Z b
lim f (t)dt exists and is ≤ g(t)dt.
x→b− x x0
0
Rx
• R(a, x0 ] 3 x → x 0 f (t)dt ∈ R+ is nonincreasing and upper bounded by
x0
a
g(t)dt hence
Z x0 Z x0
lim f (t)dt exists and is ≤ g(t)dt. 
x→a+ x a
(3) Exercise.
log
√ x
Example 7.5. f (x) = x5
is improperly integrable on [1, +∞).
 7. THE IMPROPER RIEMANN INTEGRAL 143

For the previous example, the following lemma is needed.

Lemma 7.6. For any α > 0, there exists M > 0 such that for all x ≥ M , log x ≤ xα .
Definition 7.7. Let a, b ∈ ext(R), a < b, and let f : (a, b) → R be a function that
is locally integrable on (a, b).
• f is absolutely integrable on (a, b) if |f | is improperly integrable on (a, b).
• f is conditionally integrable on (a, b) if f is improperly integrable but not
absolutely integrable on (a, b).
Example 7.8. Define f : [1, +∞) → R by
 
−1 2n + 1
f (x) = x− , n ≤ x < n + 1, n ∈ N.
n 2
Then f is conditionally integrable on [1, +∞).
Fact 7.9. If f is absolutely integrable on (a, b) then f is improperly integrable on
(a, b) and
Z b Z b
f (x)dx ≤ |f (x)|dx.
a a

Proof. This follows from the following observations:

• 0 ≤ |f | + f ≤ 2|f | hence |f | + f is improperly integrable.
• f = (|f | + f ) − |f | hence f is improperly integrable.
The inequality is a consequence of the Comparison Theorem of Improper Integrals. 
Fact 7.10 (Integral Test for Convergence of Numerical Series). Let f : [1, +∞) →
∞
X R +∞
R+ be a nonincreasing function. Then, f (k) < +∞ if and only if 1 f (x)dx
k=1
converges as an improper integral.

Proof. Let g : [1, +∞) → R+ defined by

g(x) = f (k), k ≤ x < k + 1.
Then g is nonincreasing and a step function, hence locally integrable. Also, since f is
nonincreasing it is locally integrable as well. By definition, for all x ∈ [1, +∞) we have
g(x + 1) ≤ f (x) ≤ g(x) hence, for all n ≥ 1, by the Comparison Theorem for Integrals,
we have
X n Z n+1 Z n Z n Z n n−1
X
f (k) = g(x)dx = g(x + 1)dx ≤ f (x)dx ≤ g(x)dx = f (k).
k=2 2 1 1 1 k=1
n
X Rn
Consequently, if f (k) < +∞ then ( 1 f (x)dx)n∈N is nondecreasing and upper
k=1
bounded, hence f is improperly integrable on [1, +∞].
144 5. THE RIEMANN INTEGRAL

n
X
On the other hand, if f is improperly integrable on [1, +∞] then ( f (k))n∈N is
k=1
∞
X
nondecreasing and upper bounded, hence f (k) < +∞. 
k=1

Remark 7.11. The Integral Test allows us to obtain, once more, the p-Harmonic
∞
X 1
Series Test: The series p
converges if and only if p > 1. Indeed, this follows by
k=1
k
1
taking f (x) = xp
x ∈ [1, +∞) and applying the Integral Test.
,
∞
X (−1)k+1
The Alternating p-Series is conditionally convergent for 0 < p ≤ 1 and
k=1
kp
absolutely convergent for p > 1.
∞
X (−1)k+1
In particular, the Alternating Harmonic Series is conditionally conver-
k=1
k
gent.

8. Exercises

1. Let (an )n be a sequence with distinct elements in [0, 1] that converges to 1.

Consider the function
(
1, x = an for some n ∈ N,
f (x) :=
0, otherwise.
Investigate the integrability of the function f on [0, 1] and, in the positive case, find
its integral.

2. Let a < b and let f : [a, b] → R be a bounded function. Prove that, f is Riemann
integrable on [a, b] if and only if there exists a sequence (Pn )n∈N of interval partitions
of [a, b] such that
U (f ; Pn ) − L(f ; Pn ) → 0 as n → +∞.
Moreover, for such a sequence (Pn )n , prove that
Z b
f (x)dx = lim U (f ; Pn ) = lim L(f : Pn ).
a n→∞ n→∞

3. Let a < b and let f : [a, b] → R be a function. Prove that, f is Riemann

integrable on [a, b] if and only if for all ε > 0 there exists δ > 0 such that, for all
P ∈ IP[a, b] with kP k < δ and any choice of intermediate points ξ¯ subordinated to P ,
¯ − I(f )| < ε.
we have |σ(f ; P ; ξ)

4. Let f : [a, b] → [0, +∞), for some a < b, be an integrable function, and α > 0.
Show that the function f α is integrable as well.
 8. EXERCISES 145

5. Let f, g : [a, b] → R be two integrable functions. Show that f ∧ g and f ∨ g are

integrable as well.
6. Let f and g be two real valued functions integrable on [a, b]. Show that
Z b 2 Z b Z b 
2 2
f (x)g(x)dx ≤ f (x)dx g (x)dx .
a a a
This inequality is called the integral version of the Cauchy’s Inequality or the Bun-
yakovsky’s Inequality.
Hint:R A modern proof
2 starts by considering the second order polynomial function
b
h(λ) = a f (t) − λg(t) dt, λ ∈ R.
7. Let p, q > 1 be two real numbers such that p1 + 1q = 1. Let f, g : [a, b] → R be
two integrable functions. Prove the Hölder’s Inequality
Z b Z b 1/p Z b 1/q
p q
f (t)g(t)dt ≤ |f (t)| dt |g(t)| dt .
a a a
Hint: You may use the Young’s Inequality, see the exercise in the previous chapter.
8. Let a < b and C[a, b] denote the collection of all continuous functions f : [a, b] →
R. Show that, letting
Z b
d(f, g) := |f (x) − g(x)|dx, f, g ∈ C[a, b],
a
d is a metric on C[a, b].
9. Let p ≥ 1 and f, g : [a, b] → R be two integrable functions. Prove the Minkowski’s
Inequality:
Z b 1/p Z b 1/p Z b 1/p
p p p
|f (t) + g(t)| dt ≤ |f (t)| dt + |g(t)| dt .
a a a
Hint: For p > 1 you may use the Hölder’s Inequality.
10. Let p ≥ 1 and consider the map dp : C[a, b] → R defined by
Z b 1/p
p
dp (f, g) := |f (t) − g(t)| dt .
a
Prove that dp is metric.
11. Let f : [a, b] → R be a bounded function, that is locally integrable, that is, f is
integrable on any interval [c, d] with a < c < d < b. Show that f is integrable on [a, b]
and that Z b Z d
f (x)dx = lim lim f (x)dx.
a c→a+ d→b− c

12. Let a < b and f : [a, b] → R be a continuously differentiable function such that
f (a) = 0 and 0 ≤ f 0 (x) ≤ 1 for all x ∈ [a, b]. Prove that:
146 5. THE RIEMANN INTEGRAL

Rb
(1) f 2 (x) ≤ 2 f (t)dt for all x ∈ [a, b].
a
2
Rb
b
3
R
(2) f (x)dx ≤ f (x)dx .
a a

13. (a) Prove that for all natural numbers n we have

2n n
X (−1)k+1 X 1
= .
k=1
k k=1
n+k

(b) Prove that

∞
X (−1)n+1
log 2 = .
n=1
n

14. Show that the sequence

n
X 1
an = − log n, n ∈ N,
k=1
k
converges. Hint: You may use the idea of the proof of the Integral Test.

15. Let f : [0, 1] → R be a function integrable on [0, 1] and let β > 0 be a real
number. Calculate
Z 1/nβ
α
lim n f (x)dx,
n→+∞ 0
where α < β is a real number as well.

16. Let a < b and f : [a, b] → R be a continuous function.

(a) Prove that
Z b 1/n
n
lim sup |f (x)| dx ≤ sup{|f (x)| | x ∈ [a, b]}.
n→+∞ a

(b) Prove that

Z b 1/n
n
lim |f (x)| dx = sup{|f (x)| | x ∈ [a, b]}.
n→+∞ a

17. Let a < b and f : [a, b] → R be an integrable function.

(a) Prove that, for any 1 ≤ p < q we have
Z b  p1 Z b  1q
|f (x)|p dx ≤ |f (x)|q dx .
a a

(b) Generalize exercise 16.

 8. EXERCISES 147

18. Consider the function f : (0, +∞) → R defined by

R 2
 x dt , for x 6= 1,
x log t
f (x) =
l, for x = 1.
Determine l such that the function f should be continuous at 1. For the determined
value of l, is the function f differentiable at 1?
19. Let a < b, n ∈ N, and f : [a, b] → R be a continuous function such that for all
Rb
k = 0, 1, . . . , n we have tk f (t)dt = 0. Determine the minimum number of elements
a
of the set {x ∈ (a, b) | f (x) = 0}.
20. Prove that the limit Z 3x
sin t
lim dt
x→0 x t2
exists and calculate it.
21. For a < b let f : [a, b] → R be a function that is continuously differentiable on
[a, b] and injective.
(a) Show that
Z b Z f (b)
f (x)dx + f −1 (y)dy = bf (b) − af (a).
a f (a)

(b) Investigate whether the previous equality holds when assuming that the function
f is continuous and injective on [a, b] only.
22. Let a < b and f : [a, b] → R be an integrable function. Investigate the conver-
gence of the sequence (sn )n , where
Z 1
sn = xn f (x)dx, n ∈ N.
0

23. Let f : [1, +∞) → R be an absolutely integrable function. Show that the
sequence
Z +∞
xn = f (xn )dx, n ≥ 1,
1
converges and find its limit.
24. Let f : [0, 1] → R be an integrable function and α and β two real numbers such
that α < β and β > 0. Show that the sequence
Z 1β
n
α
yn = n f (x)dx, n ≥ 1,
0
converges and find its limit.
148 5. THE RIEMANN INTEGRAL

The next three questions present an alternative for the definitions of the tran-
scendental functions. We work under the assumption that we do not know yet the
exponential, the logarithmic, and the power function for irrational exponents.

25. Define the function L : (0, +∞) → R by

Z x
dt
L(x) := , x > 0.
1 t
Prove that:
(a) L is differentiable, strictly increasing, L(1) = 0, and L0 (x) = 1/x, for all x > 0.
(b) limx→+∞ L(x) = +∞ and limx→0+ L(x) = −∞.
(c) L(xq ) = qL(x) for all q ∈ Q and all x > 0.
(d) L(xy) = L(x) + L(y) for all x, y > 0.
(e) L(e) = 1, where e = limn→+∞ (1 + 1/n)n .
The function L(x) is denoted log x and it is called the natural logarithm.

26. Let E := L−1 : R → (0, +∞), where L = log is the function defined at Question
1. Prove that:
(a) E is differentiable on R, E(0) = 1, and E 0 (x) = E(x) for all x ∈ R.
(b) limx→+∞ E(x) = +∞ and limx→−∞ E(x) = 0.
(c) E(xq) = E(x)q for all x ∈ R and all q ∈ Q.
(d) E(x + y) = E(x)E(y) for all x, y ∈ R.
Letting e = limn→+∞ (1 + 1/n)n , in view of item (c) and L(e) = 1, the function E
is the exponential function with base e and we use the familiar notation E(x) = ex , for
all x ∈ R.
27. With notation as in the previous two questions, for any α ∈ R and x > 0, let
us define xα := eα log x = E(αL(x)). Prove that:
(a) xα has the usual meaning for α ∈ Q, that is, it is the power function of exponent
α.
(b) For any α > 0 and any 0 < x < y it follows xα < y α .
(c) xα+β = xα xβ , for all x > 0 and α, β ∈ R.
(d) x−α = 1/xα for all x > 0 and α ∈ R.
(e) (xα )0 = αxα−1 for all x > 0 and all α ∈ R.
28. Show that the function
(
x2 sin x12 , x 6= 0,
f (x) :=
0, x = 0,
is differentiable on R but the derivative function f 0 is not integrable on any compact
interval that contains 0.