Professional Documents
Culture Documents
Liskevich, V. - Notas de Analisis Real PDF
Liskevich, V. - Notas de Analisis Real PDF
(1.2.10) (A A) T
(1.2.11) (A A) F
De Morgans laws
(A B) (A B) (1.2.12)
(A B) (A B) (1.2.13)
Contrapositive law
(1.2.14) (A B) (B A)
(1.2.15) (A B) (A B)
6 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
1.3 Sets
The notion of a set is one of the basic notions. We cannot, therefore, give it a precise
mathematical denition. Roughly speaking:
A set is a collection of objects, to which one can assign a size.
It is obviously not a denition (what is collection? what is size?). A rigorous set theory
is constructed in the axiomatic way. We however conne ourselves to the naive set theory,
introducing some axioms only in order to clarify the notion of a set. We accept as the basic
notions set, and the relation to be an element of a set. If A is a set we write a A to
express that a is an element of the set A, or a belongs to A. If a is not an element of A
we write a , A. So the following is true for any x and A:
(x , A) (x A).
One way to dene a set is just by listing its elements.
Example 1.3.1. (i) A = 0, 1. This means that the set A consists of two elements, 0
and 1.
(ii) B = 0, 1, 0, 1. The set B contains three elements: the number 0; the set 1
containing one element: the number 1; and the set containing two elements: the numbers
0 and 1.
The order in which the elements are listed is irrelevant. Thus 0, 1 = 1, 0.
A set can be also specied by an elementhood test.
Example 1.3.2. (i) C = x N [ x is a prime number. The set C contains all primes.
We cannot list them for the reason that there are innitely many primes. (Here N is
the set of natural numbers 1, 2, 3 . . . )
(ii) D = x R [ x
2
x = 0. The set D contains the roots of the equation x
2
x = 0. But
these are 0 and 1, so the set D contains the same elements as the set A in the previous
example. In this case we say that the sets A and D coincide, and write A = D. (Here
R is the set of real numbers.)
Sets satisfy the following fundamental law (axiom)
If the sets A and B contain the same elements, then they coincide, (or are equal).
1.3.1 Subsets. The empty set
Denition 1.3.1. If each element of the set A is also an element of the set B we say that A
is a subset of B and write A B.
This denition can be expressed as a logical proposition as follows:
For every x [(x A) (x B)].
7 January 28, 2006
1.3. SETS
Note that by the denition
(A = B) [(A B) (B A)].
This is the main way we will use to prove equalities for sets.
Since it happens that a set may contain no elements, the following denition is useful.
Denition 1.3.2. The empty set is the set which contains no elements.
Warning: there is only one empty set. It is denoted by .
This in particular means that the set of elephants taking this course of Analysis coincides
with the set of natural numbers solving the equation x
2
2 = 0.
Let us illustrate the notion of a subset by a simple example.
Example 1.3.3. Let S = 0, 1, 2. Then the subsets of S are:
, 0, 1, 2, 0, 1, 0, 2, 1, 2, 0, 1, 2.
Subsets of a set A which do not coincide with A are called proper subsets. In this example
all subsets except for the last one are proper.
Denition 1.3.3. The set of all subsets of a set A is called the power set of A, and is denoted
by P(A) (or 2
A
in some books).
Note that in the last example S contains 3 elements, whereas P(S) has 8 elements (2
3
= 8).
In general, if a set A has n elements then P(A) has 2
n
elements (provided n is nite). Prove
it!
1.3.2 Operations on sets
Now we introduce operations on sets. The main operations are: union, intersection, dierence
and symmetric dierence.
Union of sets
Denition 1.3.4. The union of sets A and B is the set containing the elements of A and the
elements of B, and no other elements.
We denote the union of A and B by A B.
Note: existence of the union for arbitrary A and B is accepted as an axiom.
For arbitrary x and arbitrary A and B the following proposition is true.
(x A B) (x A) (x B).
8 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
Intersection of sets
Denition 1.3.5. The intersection of sets A and B is the set containing the elements which
are elements of both A and B, and no other elements.
We denote the intersection of A and B by A B. Thus for arbitrary x and arbitrary A
and B the following proposition is true.
(x A B) (x A) (x B).
Dierence of sets
Denition 1.3.6. The dierence of sets A and B is the set containing the elements of A
which do not belong to B.
We use the notation A B for the dierence. The following is true for arbitrary x and
arbitrary A and B :
(x AB) [(x A) (x , B)].
By the de Morgan law and by the double negation law
(x AB) [(x A) (x B)].
Symmetric dierence
Denition 1.3.7. The symmetric dierence of the sets A and B is dened by
AB = (AB) (B A).
Let us illustrate the introduced operations by a simple example.
Example 1.3.4. Let A = 0, 1, 2, 3, 4, 5 and B = 1, 3, 5, 7, 9. Then
A B = 0, 1, 2, 3, 4, 5, 7, 9.
A B = 1, 3, 5.
AB = 0, 2, 4, B A = 7, 9.
AB = 0, 2, 4, 7, 9.
Note that
A B = (A B) (AB).
9 January 28, 2006
1.3. SETS
1.3.3 Laws for operations on sets
Commutative laws
A B = B A, (1.3.16)
A B = B A. (1.3.17)
Associative laws
A (B C) = (A B) C, (1.3.18)
A (B C) = (A B) C. (1.3.19)
Distributive laws
A (B C) = (A B) (A C), (1.3.20)
A (B C) = (A B) (A C). (1.3.21)
Idempotent laws
A A = A, A A = A. (1.3.22)
The proofs of the above laws are based on the corresponding laws for conjunction and
disjunction.
Now let us formulate and prove some laws of the dierence.
(1.3.23) A (B A) = A B.
Proof.
[x A (B A)] (x A) [(x B) (x A)]
[(x A) (x B)] [(x A) (x A)]
[(x A) (x B)],
since [(x A) (x A)] is true.
From the last formula it follows that the dierence is not an inverse operation to the union
(which means that in general A (B A) ,= B).
(1.3.24) AB = A(A B).
Proof.
[x A(A B)] (x A) (x A B)
(x A) [(x A) (x B)]
(x A) [(x A) (x B)]
[(x A) (x A)] [(x A) (x B)]
[(x A) (x B)]
(x AB)
10 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
since [(x A) (x A)] is false.
De Morgans laws
A(B C) = (AB) (AC), (1.3.25)
A(B C) = (AB) (AC). (1.3.26)
The proof is based on de Morgans laws for propositions.
1.3.4 Universe. Complement
In many applications of the set theory one considers only sets which are contained in some
xed set. (For example, in plane geometry we study only set containing points from the
plane.) This set is called a universe. We will denote it by |.
Denition 1.3.8. Let | be the universe. The set | / is called the complement to A. It
is denoted by A
c
.
It is easy to see that the following properties hold
(A
c
)
c
= A, (1.3.27)
(A B)
c
= A
c
B
c
, (1.3.28)
(A B)
c
= A
c
B
c
. (1.3.29)
Using the properties of the dierence prove it!
11 January 28, 2006
1.4. PREDICATES AND QUANTIFIERS
1.4 Predicates and quantiers
In mathematics, along with propositions, one deals with statements that depend on one or
more variables, letters denoting elements of sets. In this case we speak of predicates. For
instance, n is a prime number is a predicate. As we see, it may be true or false depending
on the value n. A predicate becomes a proposition after substituting the variable by a
xed element from the set of denition (the set to which the variable belongs). Generally, a
predicate can be written as
A(x)(x S),
where S is the set of denition (which we often omit when it is clear what S is).
A subset of S containing all the elements of S which make A(x) true is called the truth
set for A(x).
For the truth set of A(x) we write
Truth set of A = x S [A(x).
Example 1.4.1. Let A x R[ x
2
x = 0. (The notation R is used to denote the set of
real numbers, which we will discuss in details later on.) Then
x R[A(x) = 0, 1.
We often want to say that some property holds for every element from S. In this case we
use the universal quantier . So for for all x S A(x) we write (x S) A(x). After
applying the universal quantier to a predicate we obtain a proposition (which may be true
or false, as usual). The universal quantier substitutes for the words every, for every,
any, for all.
Example 1.4.2. (i) The proposition Every real number has a non-negative square can
be written as
(x R) [x
2
0].
(ii) The proposition Every real number is non-negative can be written as
(x R) [x 0].
Evidently, (i) is true and (ii) is false.
Note that proposition [(x S) A(x)] means that the truth set of A(x) is the whole set
S. Thus if for an element x of S, A(x) is false, the proposition [(x S) A(x)] is false, hence
in order to state that it is false it is enough to nd one element in S for which A(x) is false.
To express that a property holds for some element of S, or in other words, there exists
an element in S such that A(x) holds, we use
the existential quantier and write
(x S) A(x).
substitutes for the words for some, there exists.
12 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
Note that in order to state that
[(x S) A(x)] is true
it is enough to nd one element in S for which A(x) is true.
Example 1.4.3. The proposition Some real numbers are greater than their squares can
be written as
(x R) [x
2
< x].
It is true, of course.
In propositions (x) P(x) and (x) P(x) P(x) may itself contain quantiers.
Example 1.4.4. (i)
(x N)(y N) [y = x +x].
(ii)
(x R)(y R) [y < x].
(N denotes the set of natural numbers.)
Quantiers negation laws
The following equivalences are laws
[(x S) P(x)] [(x S) P(x)],
[(x S) P(x)] [(x S) P(x)].
This means that we can obtain a proposition with initially placed quantier that is equiv-
alent to the negation of a proposition with initially placed quantier by pushing the negation
to the inside and changing to and to .
Example 1.4.5. Consider the following proposition
(x N) (y N) (2y > x).
The negation of it is the proposition
(x N) (y N) (2y x).
Example 1.4.6.
[(x X)(y Y )(z Z) P(x, y, z)]
[(x X)(y Y )(z Z) P(x, y, z)].
13 January 28, 2006
1.5. ORDERED PAIRS. CARTESIAN PRODUCTS
1.5 Ordered pairs. Cartesian products
Let us talk about sets from a universe |. Recall that a denotes the set containing one
element a. The set a, b contains two elements if a ,= b and one element otherwise. Obviously,
a, b = b, a. In many problems in mathematics we need an object in which the order in a
pair is important. So, we want to dene an ordered pair.
You have already seen ordered pairs studying points in the xy plane. The use of x and y
coordinates to identify points in the plane works by assigning to each point in the plane an
ordered pair of real numbers x and y. The pair must be ordered because, for example, (2, 5)
and (5, 2) correspond to dierent points.
How to dene an ordered pair formally? Whatever we dene the ordered pairs (a, b) and
(c, d) to be, it must come out that
(1.5.30) [(a, b) = (c, d)] [(a = c) (b = d)].
Denition 1.5.1.
(a, b) = a, a, b.
Let us prove that (1.5.30) is fullled.
Proof. 1)(if) [(a = c) (b = d)] [(a, b) = (c, d)].
Suppose a = c and b = d. Then
(a, b) = a, a, b = c, c, d = (c, d).
2)(only if) [(a, b) = (c, d)] [(a = c) (b = d)]. Suppose that (a, b) = (c, d), i.e.
a, a, b = c, c, d.
There are two cases to consider.
Case 1. a = b. In this case
a, b = a so a, a, b = a.
Hence
a = c, c, d, so c = a and c, d = c,
i.e. a = b = c = d as required.
Case 2. a ,= b. In this case we have
[c = a] [c = a, b].
But [c = a, b] is false since a ,= b. Hence c = a, so a = c. Since a ,= b, it follows
that a ,= a, b. Therefore c, d = a, b and as a = c, a, d = a, b. Hence b = d, as
required.
Thus the denition of ordered pair, however articial it may appear, gives to ordered
pairs the crucial property that we require of them; and that is all we can reasonably require
from a mathematical denition.
Now we dene Cartesian product which is an essential tool for further development.
14 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
Denition 1.5.2. Let A and B are sets. The Cartesian product of A and B, denoted by
AB, is the set of all ordered pairs (a, b) in which a A and b B, i.e.
AB = (a, b) [ (a A) (b B).
Thus
(p AB) (a A)(b B) [p = (a, b)].
Example 1.5.1. 1. If A =red, green and B = 1, 2, 3 then
AB = (red, 1), (red, 2), (red, 3), (green, 1), (green, 2), (green, 3).
2. R R = (x, y) [ x and y are real numbers. These are coordinates of all points in the
plane. The notation R
2
is usually used for this set.
X X is called Cartesian square of X.
The following theorem provides some basic properties of the Cartesian product.
Theorem 1.5.2. Let A, B, C, D be sets. Then
A(B C) = (AB) (AC), (1.5.31)
A(B C) = (AB) (AC), (1.5.32)
A = A = . (1.5.33)
Proof of (1.5.31). For arbitrary p
[p A(B C)] (a A)(x B C) [p = (a, x)]
(a A)(x B) [p = (a, x)] (a A)(x C) [p = (a, x)]
p AB p AC
p (AB) (AC)
which proves (1.5.31).
The proof of (1.5.32) and of (1.5.33) is left as an exercise.
15 January 28, 2006
1.6. RELATIONS
1.6 Relations
Denition 1.6.1. Let X, Y be sets. A set R X Y is called a relation from X to Y .
If (x, y) R, we say that x is in the relation R to y. We will also write in this case xRy.
Example 1.6.1. 1. Let A = 1, 2, 3, B = 3, 4, 5. The set R = (1, 3), (1, 5), (3, 3) is
a relation from A to B since R AB.
2. G = (x, y) R R[ x > y is a relation from R to R.
Denition 1.6.2. Let R be a relation from X to Y . The domain of R is the set
D(R) = x X[ y Y [(x, y) R].
The range of R is the set
Ran(R) = y Y [ x X [(x, y) R].
The inverse of R is the relation R
1
from Y to X dened as follows
R
1
= (y, x) Y X[ (x, y) R.
Denition 1.6.3. Let R be a relation from X to Y , S be a relation from Y to Z. The
composition of S and R is a relation from X to Z dened as follows
S R = (x, z) X Z [ y Y [(x, y) R] [(y, z) S].
Theorem 1.6.2. Let R be a relation from X to Y , S be a relation from Y to Z, T be a
relation from Z to V . Then
1. (R
1
)
1
= R.
2. D(R
1
) = Ran(R).
3. Ran(R
1
) = D(R).
4. T (S R) = (T S) R.
5. (S R)
1
= R
1
S
1
.
For the proof see [1], p.170.
Next we take a look at some particular types of relations. Let us consider relations from
X to X, i.e. subsets of X X. In this case we talk about relations on X.
A simple example of such a relation is the identity relation on X which is dened as follows
i
X
= (x, y) X X[ x = y.
Denition 1.6.4. 1. A relation R on X is said to be reexive if
(x X) (x, x) R.
2. R is said to be symmetric if
(x X)(y X) [(x, y) R] [(y, x) R].
3. R is said to be transitive if
(x X)(y X)(z X)[((x, y) R) ((y, z) R)] [(x, z) R].
16 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
Equivalence relations
A particularly important class of relations are equivalence relations.
Denition 1.6.5. A relation R on X is called equivalence relation if it is reexive, symmetric
and transitive.
Example 1.6.3. 1. Let X be a set of students. A relation on X X: to be friends.
It is reexive (I presume that everyone is a friend to himself/herself). It is symmetric.
But its not transitive.
2. Let X = R, a be some positive number. Dene R X X as
R = (x, y) [ [x y[ a.
R is reexive, symmetric, but not transitive.
3. Let X = Z, m N. Dene the congruence mod m on X X as follows:
x y if (k Z [x y = km).
This is an equivalence relation on X.
Denition 1.6.6. Let R be an equivalence relation on X. Let x X. The equivalence class
of x with respect to R is the set
[x]
R
= y X[ (y, x) R.
Let us take a look at several properties of classes of equivalence.
Proposition 1.6.1. Let R be an equivalence relation on X. Then
1.(x X) x [x]
R
.
2.(x X)(y X) [(y [x]
R
) ([y]
R
= [x]
R
)].
Proof. 1. Since R is reexive, (x, x) R, hence x [x]
R
.
2.First, let y [x]
R
. (a) Suppose that z [y]
R
(an arbitrary element of [y]
R
. Then
(x, y) R, (y, z) R and by transitivity (x, z) R. Therefore z [x]
R
, which shows that
[y]
R
[x]
R
. (b) Suppose that z [x]
R
. Similarly show that z [y]
R
.
Therefore [x]
R
= [y]
R
.
The implication ([y]
R
= [x]
R
) (y [x]
R
) is obvious.
From the above proposition it follows that the classes of equivalence are disjoint and every
element of the set X belongs to a class of equivalence (the union of the classes equals to the
set X).
Remark 1. See more on this in [1].
17 January 28, 2006
1.7. FUNCTIONS
1.7 Functions
1.7.1 Function
The notion of a function is of fundamental importance in all branches of mathematics. You
met functions in your previous study of mathematics, but without precise denition. Here
we give a denition and make connections to examples you learned before.
Denition 1.7.1. Let X and Y be sets. Let F be a relation from X to Y . Then F is called
a function if the following properties are satised
(i) (x X)(y Y ) [(x, y) F].
In this case it is customary to write y = F(x).
(ii) (x X)(y Y )(z Y )([(x, y) F] [(x, z) F]) (y = z).
(In other words, for every x X there is only one y Y such that (x, y) F).
X is called the domain of F and Y is called codomain.
Let us consider several examples.
Example 1.7.1. (i) Let X = 1, 2, 3, Y = 4, 5, 6. Dene F X Y as
F = (1, 4), (2, 5), (3, 5).
Then F is a function.
(In contrast to that dene G X Y as
G = (1, 4), (1, 5), (2, 6), (3, 6).
Then G is not a function.)
(ii)Let X = R and Y = R. Dene F X Y as
F = (x, y) R R[ y = x
2
.
Then F is a function from R to R.
(In contrast to that dene G X Y as
G = (x, y) R R[ x
2
+y
2
= 1.
Then G is not a function.)
We often use the notation
F : X Y
for a function F from X to Y .
Note that in order to dene a function F from X to Y we have to dene X, Y and a
subset of X Y satisfying (i) and (ii) of the denition.
Let F : X Y be a function. If x X then we know that there is a unique y such that
(x, y) F. For this y we write y = F(x). This y is called the image of x under F.
Note: one can assign a value of y for each value x X by means of a rule or a formula.
Though these notions is rather vague and cannot be used in a denition of a function.
18 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
Theorem 1.7.2. Let X, Y be sets, F, G be functions from X to Y . Then
[(x X)(F(x) = G(x))] (F = G).
Proof. 1. (). Let x X. Then (y Y )(y = F(x)). But G(x) = F(x), so (x, y) G.
Therefore F G. Analogously one sees that G F.
2. (). Obvious since the sets F and G are equal.
The above theorem says that in order to establish that two functions are equal one has to
establish that they have the same domain and codomain and for every element of the domain
they have equal images. Note that equal function may be dened by dierent rules.
Example 1.7.3. Let f : R R, g : R R, h : R R
+
(By R
+
we denote the set of
non-negative real numbers). Let x R
f(x) = (x + 1)
2
, g(x) = x
2
+ 2x + 1, h(x) = (x + 1)
2
.
Then f and g are equal, but f and h are not since they have dierent codomains.
Denition 1.7.2. Let f : X Y be a function. The set
Ran(f) = y [ (x X)(f(x) = y)
is called the range of f.
Example 1.7.4. Let f : R R, f(x) =
2x
x
2
+1
. Find Ran(f).
Solution. We have to nd a set of y such that the equation y =
2x
x
2
+1
has a solution x R.
The equation is equivalent to yx
2
2x +y = 0. The existence of a real solution is equivalent
to D = 4 4y
2
0. Solving this inequality we obtain that Ran(f) = [1, 1].
The denition of composition of relations can be also applied to functions. If f : X Y
and g : Y Z then
g f = (x, z) X Z [ (y Y )[(x, y) f] [(y, z) g].
Theorem 1.7.5. Let f : X Y and g : Y Z. Then g f : X Z and
(x X) [(g f)(x) = g(f(x))].
Proof. We know that g f is a relation. So we must prove that for every x X there exists
a unique element z Z such that (x, z) g f.
Existence: Let x X be arbitrary. Then y Y such that y = f(x), or in other words,
(x, y) f. Also z Z such that z = g(y), or in other words, (y, z) g. By the denition it
means that (x, z) g f. Moreover, we see that
(g f)(x) = g(f(x)).
Uniqueness: Suppose that (x, z
1
) g f and (x, z
2
) g f. Then by the denition of
composition (y
1
Y ) [(x, y
1
) f] [(y
1
, z
1
) g] and (y
2
Y ) [(x, y
2
) f] [(y
2
, z
2
) g].
But f is a function. Therefore y
1
= y
2
. g is also a function, hence z
1
= z
2
.
19 January 28, 2006
1.7. FUNCTIONS
Example 1.7.6. Let f : R R, g : R R,
f(x) =
1
x
2
+ 2
, g(x) = 2x 1.
Find (f g)(x) and (g f)(x).
Solution.
(f g)(x) = f(g(x)) =
1
[g(x)]
2
+ 2
=
1
(2x 1)
2
+ 2
,
(g f)(x) = g(f(x)) = 2f(x) 1 =
2
x
2
+ 2
1.
Warning: As you clearly see from the above f g ,= g f.
Images and inverse images
Denition 1.7.3. Let f : X Y and A X. The image of A under f is the set
f(A) = f(x) [ x A.
Example 1.7.7. Let f : R R dened by f(x) = x
2
. Let A = x R[ 0 x 2. Then
f(A) = [0, 4].
The following theorem (which we give without proof) establishes some properties of images
of sets.
Theorem 1.7.8. Let f : X Y and A X, B X. Then
(i) f(A B) = f(A) f(B),
(ii) f(A B) f(A) f(B),
(iii) (A B) [f(A) f(B)].
Remark 2. Note that in (ii) there is no equality in general. Consider the following example.
f : R R dened by f(x) = x
2
. Let A = [1,
1
2
] and B = [
1
2
, 1]. Then f(A) = [0, 1],
f(B) = [0, 1], so that f(A) f(B) = [0, 1]. At the same time A B = [
1
2
,
1
2
], and hence
f(A B) = [0,
1
4
].
Denition 1.7.4. Let f : X Y and B Y . The inverse image of B under f is the set
f
1
(B) = x X[ f(x) B.
Example 1.7.9. Let f : R R dened by f(x) = x
2
. Let B = [1, 4]. Then f
1
(B) =
[2, 2].
The following theorem (which we give without proof) establishes some properties of inverse
images of sets.
20 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
Theorem 1.7.10. Let f : X Y and A Y , B Y . Then
(i) f
1
(A B) = f
1
(A) f
1
(B),
(ii) f
1
(A B) = f
1
(A) f
1
(B),
(iii) (A B) [f
1
(A) f
1
(B)].
Remark 3. Note the dierence with the previous theorem.
1.7.2 Injections and surjections. Bijections. Inverse functions
In the last section we saw that the composition of two functions is again a function. If
f : X Y the is a relation from X to Y . One can dene the inverse relation f
1
. Then the
question comes : Is f
1
a function? In general the answer is no. In this section we will
study particular cases of functions and nd out when the answer is yes.
Denition 1.7.5. Let f : X Y . Then f is called an injection if
(x
1
X)(x
2
X)[(f(x
1
) = f(x
2
)) (x
1
= x
2
)].
The above denition means that f is one-to-one correspondence between X and Ran(f).
Using the contrapositive law one can rewrite the above denition as follows.
(x
1
X)(x
2
X)[(x
1
,= x
2
) (f(x
1
) ,= f(x
2
))].
Example 1.7.11. (i) Let X = 1, 2, 3, Y = 4, 5, 6, 7. Dene f : X Y as
f = (1, 5), (2, 6), (3, 7).
Then f is an injection.
In contrast dene g : X Y as
g = (1, 5), (2, 6), (3, 5).
Then g is not an injection since g(1) = g(3).
(ii)Let f : N N dened by f(n) = n
2
. Then f is an injection.
In contrast let g : Z Z dened by g(n) = n
2
. Then g is not an injection since, for instance,
g(1) = g(1).
Denition 1.7.6. Let f : X Y . Then f is called a surjection if
(y Y )(x X)[f(x) = y].
The above denition means that Ran(f) = Y . For this reason surjections are sometimes
called onto.
21 January 28, 2006
1.7. FUNCTIONS
Example 1.7.12. (i) Let X = 1, 2, 3, 4, Y = 5, 6, 7. Dene f : X Y as
f = (1, 5), (2, 6), (3, 7), (4, 6).
Then f is an surjection.
In contrast dene g : X Y as
g = (1, 5), (2, 6), (3, 5), (4, 6).
Then g is not an surjection since 7 is not in its range.
(ii)Let f : Z Z dened by f(n) = n + 2. Then f is a surjection.
In contrast let g : Z Z dened by g(n) = n
2
. Then g is not a surjection since, for instance,
there is no integer such that its square is 2.
Denition 1.7.7. Let f : X Y . Then f is called a bijection if it is an injection an a
surjection.
Example 1.7.13. (i) Let X = 1, 2, 3 and Y = 4, 5, 6. Dene f : X Y by
f = (1, 4), (2, 5), (3, 6).
Then f is a bijection.
(ii) Let X = Y = [0, 1]. Dene g : X Y by g(x) = x
2
. Then g is a bijection.
Now we are ready to answer the question about the inverse to a function. First, recall
that if f : X Y then f
1
is a relation from Y to X with the properties D(f
1
) = Ran(f)
and Ran(f
1
) = D(f). So, if f
1
is a function from Y to X then D(f
1
) = Y . Therefore we
conclude that the condition Ran(f) = Y is a necessary condition for f
1
to be a function.
Which means, that f has to be surjective. Also, from the denition of the inverse relation f
1
it is clear that injectivity of f is also a necessary condition for f
1
to be a function. Hence
bijectivity of f is a necessary condition for f
1
to be a function.
It turns out that this is also a sucient condition.
Theorem 1.7.14. Let f : X Y . Then
(f
1
: Y X) (f is a bijection).
Proof. We have to show only that
(f is a bijection) (f
1
: Y X).
Recall that
f
1
= (y, x) Y X[ (x, y) f.
We have to verify two properties in the denition of a function.
First, existence. We have to show that (y Y )(x X)[(y, x) f
1
], or in other words,
(y Y )(x X)[(x, y) f]. This follows from the surjectivity of f.
Second, Uniqueness. We have to show that (x
1
X)(x
1
X)[((y, x
1
) f
1
)
(((y, x
2
) f
1
)] (x
1
= x
2
), or in other words (x
1
X)(x
1
X)[f(x
1
) = f(x
2
)]
(x
1
= x
2
). But this follows from injectivity of f.
22 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
Denition 1.7.8. Let f : X Y . If f
1
is a function from Y to X we say that f is
invertible. In that case f
1
is called the inverse function.
Theorem 1.7.15. Let f : X Y . Let f
1
be a function from Y to X. Then
f
1
f = i
X
and f f
1
= i
Y
.
Proof. Let x X be arbitrary. Let y = f(x) Y . Then (x, y) f so that (y, x) f
1
.
Therefore f
1
(y) = x. Thus
(f
1
f)(x) = f
1
(f(x)) = f
1
(y) = x.
We have proved that
(x X)[f
1
f(x) = x]
This is the same as to say that f
1
f = i
X
. The second statement is similar and is left
as an exercise.
Example 1.7.16. Let f : R R dened by f(x) =
x+7
5
. Then f is a bijection.
Indeed, Let x
1
, x
2
R be arbitrary and
x
1
+7
5
=
x
2
+7
5
. Then, of course, x
1
+ 7 = x
2
+ 7,
therefore x
1
= x
2
, so f is an injection.
Now let y R be arbitrary and y =
x+7
5
. Then, of course, x + 7 = 5y and x = 5y 7 R.
So we have proved that
(y R)(x R)[y = f(x)].
By denition this means that f is a surjection.
So f is a bijection. Moreover f
1
(y) = 5y 7. This means that g(x) = 5x 7 is the inverse
function to f.
23 January 28, 2006
1.8. SOME METHODS OF PROOF. PROOF BY INDUCTION
1.8 Some methods of proof. Proof by induction
1. First we discuss a couple of widely used methods of proof: contrapositive proof and proof
by contradiction.
The idea of contrapositive proof is the following equivalence
(A B) (B A).
So to prove that A B is true is the same as to prove that B A is true.
Example 1.8.1. For integers m and n, if mn is odd then so are m and n.
Proof. We have to prove that (m, n N)
(mn is odd) [(m is odd) (n is odd)],
which is the same as to prove that
[(m is even) (n is even)] (mn is even).
The latter is evident.
The idea of proof by contradiction is the following equivalence
(A B) (A B) (A B).
So to prove that A B is true is the same as to prove that AB is true or else that AB
is false.
Example 1.8.2. Let x, y R be positive. If x
2
+y
2
= 25 and x ,= 3, then y ,= 4.
Proof. In order to prove by contradiction we assume that
[(x
2
+y
2
= 25) (x ,= 3)] (y = 4).
Then x
2
+y
2
= x
2
+ 16 = 25. Hence (x = 3) (x ,= 3) which is a contradiction.
2. In the remaining part of this short section we will discuss an important property of
natural numbers. The set of natural numbers
N = 0, 1, 2, 3, 4, . . .
which we will always denote by N, is taken for granted.
We shall denote the set of positive natural numbers N
+
= 1, 2, 3, 4, . . . .
24 January 28, 2006
CHAPTER 1. ELEMENTS OF LOGIC AND SET THEORY
The Principle of Mathematical Induction is often used when one needs to prove the state-
ment of the form
(n N
+
) P(n)
or similar types of statements.
Since there are innitely many natural numbers we cannot check one by one that they all
have property P. The idea of mathematical induction is that to list all natural numbers one
has to start from 1 and then repeatedly add 1. Thus one can show that 1 has property P
and that whenever one adds 1 to a number that has property P, the resulting number also
has property P.
Principle of Mathematical Induction. If for a statement P(n)
(i) P(1) is true,
(ii) [P(n) P(n + 1)] is true,
then (n N
+
) P(n) is true.
Part (i) is called base case; (ii) is called induction step.
Example 1.8.3. Prove that n N
+
if n 1 then:
1
2
+ 2
2
+ 3
2
+ +n
2
=
n(n + 1)(2n + 1)
6
.
Solution. Base case: n = 1. 1
2
=
123
6
is true.
Induction step: Suppose that the statement is true for n = k (k 1). We have to prove that
it is true for n = k + 1. So our assumption is
1
2
+ 2
2
+ 3
2
+ +k
2
=
k(k + 1)(2k + 1)
6
.
Therefore we have
1
2
+ 2
2
+ 3
2
+ +k
2
+ (k + 1)
2
=
k(k + 1)(2k + 1)
6
+ (k + 1)
2
=
(k + 1)(k + 2)(2k + 3)
6
,
which proves the statement for n = k + 1. By the principle of mathematical induction the
statement is true for all natural n.
Example 1.8.4. 1 + 3 + 5 + + (2n 1) =?
First we have to work out a conjecture. For this let us try several particular cases for n.
n = 1; 1 =1;
n = 2; 1 + 3 =4 = 2
2
;
n = 3; 1 + 3 + 5 =9 = 3
2
.
So a reasonable conjecture is that n N
+
1 + 3 + 5 + + (2n 1) = n
2
.
25 January 28, 2006
1.8. SOME METHODS OF PROOF. PROOF BY INDUCTION
Proof. We need not make the base case, since it has been done working out the conjecture.
Induction step:
Let
1 + 3 + 5 + + (2k 1) = k
2
(k 1).
Then
1 + 3 + 5 + + (2k 1) + (2k + 1) = k
2
+ 2k + 1 = (k + 1)
2
.
This completes the proof by induction.
The base step can be started not necessarily from 1. It can start from any integer onwards.
Example 1.8.5. Prove that (n N)(3[(n
3
n)).
Solution. Base case: n = 0. (3[0) is true.
Induction step:
Assume that (3[(k
3
k)) for k 0. Then
(k + 1)
3
(k + 1) = k
3
+ 3k
2
+ 3k + 1 k 1 = (k
3
k)
. .
divisible by 3
+ 3(k
2
+k)
. .
divisible by 3
Example 1.8.6. Prove that (n N)[(n 5) (2
n
> n
2
)].
Solution. Base case: n = 5. True.
Induction step:
Suppose that 2
k
> k
2
(k 5). Then
2
k+1
= 2 2
k
> 2k
2
.
Now it is sucient to prove that
2k
2
> (k + 1)
2
.
Consider the dierence:
2k
2
(k + 1)
2
= k
2
2k 1 = (k 1)
2
2.
Since k 5 we have that k 1 4 and (k 1)
2
2 14 > 0 which proves the above
inequality.
26 January 28, 2006
Chapter 2
Numbers
2.1 Various Sorts of Numbers
2.1.1 Integers
We take for granted the system N of natural numbers
N = 0, 1, 2, 3, 4 . . .
stressing only that for N the following properties hold.
(a N)(b N)(c N)(d N)[(a +b = c) (ab = d)].
(closure under addition and multiplication)
(a N)[a 1 = 1 a = a].
(existence of a multiplicative identity)
(a N)(b N)[(a = b) (a < b) (a > b)].
(order)
The rst diculty occurs when we try solving the equation in N
a +x = b,
with b a. In order to make this equation soluble we have to widen the set N by introducing
0 and negative integers as solutions of the equations
a +x = a (existence of additive identity)
and
a +x = 0 (existence of additive inverse)
respectively. Our extended system, which is denoted by Z, now contains all integers and can
be arranged in order
Z = . . . , 3, 2, 1, 0, 1, 2, 3, . . . = N a [ a N.
27
2.1. VARIOUS SORTS OF NUMBERS
2.1.2 Rational Numbers
Let a Z, b Z. The equation
(2.1.1) ax = b
need not have a solution x Z. In order to enable one to solve (2.1.1) (for a ,= 0) we
have to widen our system of numbers again so that it included fractions b/a (existence of
multiplicative inverse in Z 0. This motivates the following
Denition 2.1.1. Rational numbers, or rationals, is the set
r =
p
q
[ p Z, q N.
The set of rational numbers will be denoted by Q. When writing p/q for a rational we
always assume that the numbers p and q have no common factor greater than 1.
All the arithmetical operations in Q are straightforward. Let us introduce a relation of
order for rationals.
Denition 2.1.2. Let b N, d N, with both b, d > 0,. Then
_
a
b
>
c
d
_
_
ad > bc
_
.
The following theorem provides a very important property of rationals.
Theorem 2.1.1. Between any two rational numbers there is another (and, hence, innitely
many others).
Proof. Let b Nd N, both b, d > 0, and
a
b
>
c
d
.
Notice that
(m N) [
a
b
>
a +mc
b +md
>
c
d
] .
Indeed, since b, d and m are positive we have
[a(b +md) > b(a +mc)] [mad > mbc] (ad > bc),
and
[d(a +mc) > c(b +md)] (ad > bc).
2.1.3 Irrational Numbers
Suppose that a Q and consider the equation
(2.1.2) x
2
= a.
In general (2.1.2) does not have rational solutions. For example the following theorem holds.
Theorem 2.1.2. No rational number has square 2.
28 January 28, 2006
CHAPTER 2. NUMBERS
Proof. Suppose for a contradiction that
p
q
with p Z, q N, q ,= 0 and p, q have no
common factors greater than 1, is such that (
p
q
)
2
= 2. Then p
2
= 2q
2
. Hence p
2
is even, and
so is p. Hence, (k Z) [ p = 2k ]. This implies that
2k
2
= q
2
and therefore q is also even. The last statement contradicts our assumption that p and q have
no common factor.
Theorem 2.1.2 provides an example of a number which is not rational, or irrational. Here
are some other examples of irrational numbers.
Theorem 2.1.3. No rational x satises the equation x
3
= x + 7.
Proof. First we show that there are no integers satisfying the equation x
3
= x + 7. For a
contradiction suppose that there is. Then x(x + 1)(x 1) = 7 from which it follows that x
divides 7. Hence x can be only 1, 7. Direct verication shows that these numbers do not
satisfy the equation.
Second, show that there are no fractions satisfying the equation x
3
= x + 7. For a
contradiction suppose that there is. Let
m
n
with m Z, n N, n ,= 0, and m, n have no
common factors greater than 1, is such that (
m
n
)
3
=
m
n
+ 7. Multiplying this equality by n
2
we obtain
m
3
n
= mn
2
+ 7n
2
, which is impossible since the right-hand side is an integer and
the left-hand side is not.
Example 2.1.4. No rational satises the equation x
5
= x + 4.
Prove it!
Algebraic numbers A correspond to all real solutions of polynomial equations with integer
coecients. So x is algebraic if there exists n N, a
0
, a
1
, . . . , a
n
Z such that
a
0
+a
1
x + +a
n
x
n
= 0.
2.1.4 Cuts of the Rationals
In this subsection we discuss how a particular irrational number, say
2 is tted in among
the rationals.
In trying to isolate a number whose square is 2, we rst observe from Theorem 2.1.2 that
the positive rational numbers fall into two classes, those whose squares are greater than 2
and those whose squares are less than 2. Call these classes the left-hand class L and the
right-hand class R, corresponding to their relative positions when represented graphically on
a horizontal line. Examples of numbers l L are 7/5 and 1 41, and of numbers r R are
17/12 and 1 42. At this stage we shall have to convince ourselves that for elements of the
classes L and R the following properties hold.
(i) (l L)(r R)[r > l];
(ii) (l
1
L)(l
2
L)[l
2
> l
1
];
(iii) (r
1
R)(r
2
R)[r
2
< r
1
].
29 January 28, 2006
2.1. VARIOUS SORTS OF NUMBERS
The last three statements become more concrete if we use the arithmetical rule for square
root to nd, to as many decimal places as we please, a set of numbers l L
1, 1 4, 1 41, 1 414, . . . ,
each of which is greater than the preceding (or equal to it if the last digit is 0) and each
having its square less than 2. Moreover, the numbers got by adding 1 to the last digit of these
numbers l L for a set of numbers r R
2, 1 5, 1 42, 1 415, . . . ,
each having its square greater than 2 and each less than (or equal to) the preceding.
If we are now given a particular number a whose square is less than 2, we shall by going
far enough along the set of numbers 1, 1 4, 1 41, 1 414, . . . come to one which is greater
than a.
If, then, we are building a number-system starting with integers and then including the
rational numbers, we see that an irrational number (such as
2) corresponds to and can be
dened by a cutting of the rationals into two classes L, R of which L has no greatest member
and R no least member. This is Dedekinds denition of irrationals by the cut.
There are cuts which correspond to non-algebraic numbers: the transcendental numbers,
denoted by T. For example, the cut corresponding to 2
x
= 3 is transcendental. (Check that
no rational x satises 2
x
= 3). The proof that x , A is more complicated. Summarising:
N Z Q A T A = R.
30 January 28, 2006
CHAPTER 2. NUMBERS
2.2 The Field of Real Numbers
In the previous sections, starting from integers, we have sketched the building-up of the set
of real numbers which is the union of the sets of rationals and irrationals and denoted by R.
Now we are making the list of the basic properties which the real numbers satisfy. Since
we provide no proof of these statements we present them in form of axioms. Naturally they
fall into two groups which concern algebraic operations with real numbers and the relation of
order for them.
Those of you familiar with basic concepts of algebra will nd that the rst group of axioms
denes R to be a eld.
A.1. (a R)(b R) [(a +b) R] .
A.2. (a R)(b R) [a +b = b +a] .
A.3. (a R)(b R)(c R) [(a +b) +c = a + (b +c)] .
A.4. (0 R)(a R) [0 +a = a] .
A.5. (a R)(!x R) [a +x = 0] . We write x = a.
Notation: We use the symbol ! in order to say that there exists a unique... So (!x R)
has to be red : there exists a unique real number x.
The axioms A.6-A.10 that follow are analogues of A.1-A.5 for the operation of multipli-
cation.
A.6. (a R)(b R) [ab R].
A.7. (a R)(b R) [ab = ba].
A.8. (a R)(b R)(c R) [(ab)c = a(bc)].
A.9. (1 R)(a R) [1 a = a].
A.10. (a R 0)(!y R) [ay = 1] . We write y = 1/a.
The last axiom links the operations of summation and multiplication.
A.11. (a R)(b R)(c R) [(a +b)c = ac +bc] .
From the axioms above the familiar rules of manipulation of real numbers can be deduced.
As an illustration we present
Example 2.2.1. (a R)[0 a = 0]. Indeed, by A.11 and A.4 we have
1 a + 0 a = (1 + 0)a = 1 a.
Then by A.2 and A.4 0a = 0.
Let us observe that integers do not form a eld since they do not satisfy axiom A.10.
Now we are adding the relevant axioms of order for real numbers.
31 January 28, 2006
2.2. THE FIELD OF REAL NUMBERS
O.1. (a R)(b R)[(a = b) (a < b) (a > b)].
(a R)(b R)[(a b) (b a) (a = b)].
O.2. (a R)(b R)(c R)[(a > b) (b > c) (a > c)].
O.3. (a R)(b R)(c R)[(a > b) (a +c > b +c)].
O.4. (a R)(b R)(c R)[(a > b) (c > 0) (ac > bc)].
Observe that
(a R)(b R)[a > b] [a b > 0].
This follows from (O.3).
Completeness axiom states that there are no gaps among the reals. We will give it in the
form of
Dedekinds axiom: Suppose that the system of all real numbers is divided into two classes
L and R, such that every number l L is less than every number r R (and R, L ,= ).
Then there is a dividing number with the properties that every number less belongs to L
and every number greater than belongs to R.
The number is either in L or in R. If R then is the least number in R. If L
then is the greatest number in L.
Let us express Dedekinds axiom in the form of a true logical proposition.
[(R = L R) (L ,= ) (R ,= ) (l L)(r R)(l < r)]
( R)(x R)[(x < ) (x L)] [(x > ) (x R)].
Consequences of the Dedekinds axiom:
1) L R = .
2) ( L) ( R) is true.
3) ( L) ( is the greatest number in L).
4) ( R) ( is the least number in R).
Inequalities
The order axioms express the properties of the order relation (inequalities) on the set of real
numbers. Inequalities play an extremely important role in analysis. Here we discuss several
ideas how to prove inequalities. But before we start doing so we give the denition of the
absolute value of a real number and derive a simple consequence from it.
Denition 2.2.1. Let a R. The absolute value [a[ of a is dened by
[a[ =
_
a if a 0,
a if a < 0.
Theorem 2.2.2.
(a R)(b R) [ [a +b[ [a[ +[b[ ] .
32 January 28, 2006
CHAPTER 2. NUMBERS
The proof is left as an exercise. Hint: Consider four cases:
1) a 0, b 0;
2) a 0, b < 0;
3) a < 0, b 0;
4) a < 0, b < 0.
1. First idea of proving inequalities is very simple and is based on the equivalence
(x R)(y R)[(x y) (x y 0)].
Example 2.2.3. Prove that
(a R)(b R) [a
2
+b
2
2ab].
Proof. It equivalent to show that a
2
+b
2
2ab 0. Indeed,
a
2
+b
2
2ab = (a b)
2
0.
The equality holds i a = b.
Example 2.2.4. Prove that
(a R
+
)(b R
+
)
_
a +b
2
ab
_
.
(Reminder: R
+
= x R[ x 0.)
Proof. As above, let us prove that the dierence between the left-hand side (LHS) and the
right-hand side (RHS) is non-negative.
a +b
2
ab =
(
b)
2
2
0.
The equality holds i a = b.
2. The second idea is to use already proved inequalities and axioms to derive new ones.
Example 2.2.5. Prove that
(a R)(b R)(c R) [a
2
+b
2
+c
2
ab +ac +bc].
Proof. Adding the tree inequalities from Example 2.2.3
a
2
+b
2
2ab,
a
2
+c
2
2ac,
b
2
+c
2
2bc,
we obtain 2(a
2
+b
2
+c
2
) 2ab + 2ac + 2bc, which proves the desired. The equality holds i
a = b = c.
33 January 28, 2006
2.2. THE FIELD OF REAL NUMBERS
Example 2.2.6. Prove that
(a R
+
)(b R
+
)(c R
+
)(d R
+
)
_
a +b +c +d
4
4
abcd
_
.
Proof. By Example 2.2.4 and by (O.2) we have
a +b +c +d
4
2
ab + 2
cd
4
4
abcd.
The equality holds i a = b = c = d.
3. The third idea is to use the transitivity property
(a R
+
)(b R
+
)(c R
+
)[(a b) (b c) (a c)].
In general this means that when proving that a c we have to nd b such that a b and
b c.
Example 2.2.7. Let n 2 be a natural number. Prove that
1
n + 1
+
1
n + 2
+ +
1
2n
>
1
2
.
Proof.
1
n + 1
+
1
n + 2
+ +
1
2n
>
1
2n
+
1
2n
+ +
1
2n
. .
n
= n
1
2n
=
1
2
.
4. Inequalities involving integer may be proved by induction as we discussed before. Here
we give an inequality which is of particular importance in analysis.
Theorem 2.2.8. (Bernoullis inequality)
(n N)( > 1)[(1 +)
n
1 +n].
Proof. For n = 0, 1 it is true.
Suppose that it is true for n = k (k 1), i.e. (1 +)
k
1 +k. We have to prove that it is
true for n = k + 1, i.e.
(1 +)
k+1
1 + (k + 1).
Indeed,
(1 +)
k+1
= (1 +)
k
(1 +) (1 +k)(1 +)
= 1 + (k + 1) +k
2
1 + (k + 1).
34 January 28, 2006
CHAPTER 2. NUMBERS
2.3 Bounded sets of numbers
Simplest examples of bounded sets of numbers is nite sets of numbers. In case of a nite set
of real numbers we can always nd the maximal number in the set and the minimal number.
In general we can dene these numbers for any set of real number by the following
Denition 2.3.1. Let S R. Then
(x = max S) (x S) [(y S)(y x)],
(x = minS) (x S) [(y S)(y x)].
Note that the denition says nothing about existence of maxS and minS. And indeed
they not always exist.
Denition 2.3.2. (i) A set S R is called bounded above if
(K R)(x S)(x K).
The number K is called an upper bound of S.
(ii) A set S R is called bounded below if
(k R)(x S)(x k).
The number k is called a lower bound of S.
(iii) A set S R is called bounded if it is bounded above and below.
Note that if S is bounded above with upper bound K then any real number greater than
K is also an upper bound of S. Similarly, if S is bounded below with lower bound k then any
real number smaller than k is also a lower bound of S.
Example 2.3.1. Let a, b R and a < b.
(i) A = x R[ a x b = [a, b] - closed interval. It is bounded above since (x
A)(x b + 1) and bounded below since (x A)(x a 1).
Therefore max A = b, minA = a.
(ii) B = x R[ a < x < b = (a, b) - open interval. It is bounded above and below for the
same reasons as A.
Therefore B has no maximal or minimal number.
Note that if A R is a bounded set and B A then B is also bounded. Prove it!
Example 2.3.2.
A =
_
n
n + 1
[ n N
_
.
A is bounded. Indeed, (n N)
_
n
n+1
0)
_
, so A is bounded below. Moreover,
(n N)
_
n
n + 1
< 1)
_
(Why? Prove this inequality) so A is bounded above, and therefore is bounded.
Therefore minA =
1
2
, max A does not exist.
35 January 28, 2006
2.4. SUPREMUM AND INFIMUM
2.4 Supremum and inmum
If a set S R is bounded above one tries to nd the sharp, or in other words, the least upper
bound (existence of which is non-trivial and will be proved below). The least upper bound of
a set is called supremum of this set. Let us give it a precise denition.
Denition 2.4.1. Let S R be bounded above. Then a R is called supremum of S (the
least upper bound of S) if
(i) a is an upper bound of S, i.e.
(x S)(x a);
(ii) there is no upper bound of S less than a, i.e.
( > 0)(b S)(b > a ).
Analogously one denes the greatest lower bound for a set bounded below.
Denition 2.4.2. Let S R be bounded below. Then a R is called inmum of S (the
greatest lower bound of S) if
(i) a is a lower bound of S, i.e.
(x S)(x a);
(ii) there is no lower bound of S greater than a, i.e.
( > 0)(b S)(b < a +).
The next theorem is one of the corner-stones of analysis.
Theorem 2.4.1. Let S R be non-empty and bounded above. Then supS exists.
Proof. Divide R into two classes L and R as follows.
(x L) (s S)(s > x),
(x R) (s S)(s x).
Then
1. L R = , and L ,= , R ,= .
Indeed, since S ,= , it follows that s S. Then x = s 1 L. Let K be an upper
bound of S. Then K R.
2.
(l L)(r R) (l < r).
Indeed, let l L and r R. The by the denition of L there exists s S such that
s > l. By the denition of R we have that r s, hence l < r.
36 January 28, 2006
CHAPTER 2. NUMBERS
From this it follows that the classes L and R satisfy Dedekinds axiom. Therefore
( R)( > 0) [( L) ( + R)].
From Dedekinds axiom one cannot conclude whether L or R.
We claim that in our circumstances R.
Let us prove this claim by contradiction.
For a contradiction suppose that L. Then by the denition of L there exists s S such
that s > . Fix this s.
Take =
1
2
(s +). Then < < s. Hence R since > . Therefore by the denition of
R we conclude that s which is a contradiction. This proves that R.
So we obtain that satises
(1) (s S) (s );
(2) ( > 0)(s S) (s > ).
By the denition of the supremum we conclude that = supS, which proves the theorem.
We use the following notation
S = x R[ x S.
Theorem 2.4.2. Let S R be non-empty and bounded below. Then inf S exists and is equal
to sup(S).
Proof. Since S is bounded below, it follows that k R such that (x S)(x k). Hence
(x S)(x k) which means that S is bounded above. By Theorem 2.4.1 there exists
= sup(S). We will show that = inf S. First, we have that (x S)(x ), so
that x . Hence is a lower bound of S. Next, let is another lower bound of S, i.e.
(x S)(x ). Then (x S)(x ). Hence is an upper bound of S, and by the
denition of supremum sup(S) = . Hence which proves the required.
Archimedian Principle and its consequences.
Theorem 2.4.3. (The Archimedian Principle) Let x > 0 and y R. Then there exists
n Z such that y < nx.
Proof. Suppose for a contradiction that
(n Z) (nx y).
Then the set A := nx[ n Z is bounded above and y its upper bound. Then by Theorem
2.4.1 there exists supA. The number supAx < supA is not an upper bound of A, so there
exists m Z such that mx > supAx, or otherwise
(2.4.3) (m+ 1)x > supA.
But m+1 Z, hence (m+1)x A, and (2.4.3) contradicts to the denition of supremum.
Corollary 2.4.1. N is unbounded.
37 January 28, 2006
2.4. SUPREMUM AND INFIMUM
Corollary 2.4.2.
(x > 0)(y > 0)(n N) (
y
n
< x).
Proof. By the Archimedian Principle (Theorem 2.4.3)
(n Z) (nx > y).
Since y > 0 it follows that n N. Therefore
y
n
< x.
Corollary 2.4.3.
(y > 0)
_
inf
_
y
n
[ n N, n ,= 0
_
= 0
_
.
Proof. Since all the elements of the set
_
y
n
[ n N
_
are positive, we conclude that 0 is its
lower bound. By Corollary 2.4.2 any positive number x is not its lower bound, which proves
the assertion.
Example 2.4.4.
inf
_
1
n
[ n N, n ,= 0
_
= 0.
Example 2.4.5. Let
A :=
_
n 1
2n
[ n N, n ,= 0
_
.
Then supA =
1
2
and inf A = 0.
Proof. All the element of A are positive, except for the rst one which is 0. Therefore
minA = 0, and hence inf A = 0.
Now notice that
(n N)
_
n ,= 0
n 1
2n
=
1
2
1
2n
<
1
2
_
.
Hence
1
2
is an upper bound of A.
We have to prove that
1
2
is the least upper bound. For this we have to prove that
( > 0)(n N)
_
n 1
2n
>
1
2
_
.
Notice that
_
n 1
2n
>
1
2
_
_
1
2
1
2n
>
1
2
_
_
1
2n
<
_
[n (2) > 1].
By the Archimedian Principle there exists such n N.
Note that for any set A R if max A exists then supA = max A. Also if minA exists
then inf A = minA.
Theorem 2.4.6. For any interval (a, b) there exists a rational r (a, b). In other words,
(a R)(b R)[(b > a) (r Q)(a < r < b)].
38 January 28, 2006
CHAPTER 2. NUMBERS
Proof. Let h = b a > 0. Then by the Archimedian Principle
(n N)
_
1
n
< h
_
.
Fix this n. By the Archimedian Principle (m, m
N)
_
n
< a <
m
n
_
. Indeed,
(m N)
_
m
n
> a
_
and (m
N)
_
m
n
> a
_
.
Now the set of integers between m
n
> a.
Set r =
m
n
. Then we have
a < r =
m
n
=
m
1
n
+
1
n
a +
1
n
< a +h = b.
The next theorem is a multiplicative analogue of the Archimedian Principle.
Theorem 2.4.7. Let x > 1, y > 0. Then
(n N) (x
n
> y).
Prove this theorem repeating the argument from the proof of Theorem 2.4.3.
Theorem 2.4.8. (!x > 0) (x
2
= 2).
Proof. Dene a set
A := y > 0 [ y
2
< 2.
A ,= since 1 A. A is bounded above since (y A) (y < 2). Then by Theorem 2.4.1
there exists supA = x. We will prove that x
2
= 2.
1) First let us suppose that x
2
> 2. Let =
x
2
2
2x
. Then > 0 and
(x )
2
= x
2
2x +
2
> x
2
2x = x
2
2x
x
2
2
2x
= 2.
Hence x is another upper bound for A, so that x is not the least upper bound for A. This
is a contradiction.
2) Now suppose that x
2
< 2. Let =
2x
2
2x+1
. Then by assumption 0 < < 1 so that
(x +)
2
= x
2
+ 2x +
2
< x
2
+ 2x + =
= x
2
+(2x + 1) = x
2
+
2 x
2
2x + 1
(2x + 1) = 2.
Hence x + is also in A, in which case x cannot be an upper bound for A, which is a
contradiction.
In the same way one can prove the following theorem.
Theorem 2.4.9. (a > 0)(n N)[n ,= 0 (!x > 0) (x
n
= a)].
39 January 28, 2006
Chapter 3
Sequences and Limits
3.1 Sequences
In general any function f : N 0 X with domain N
+
and arbitrary codomain X is
called a sequence. This means that a sequence (a
n
)
nN
+ is a subset of X each element of
which has its number, i.e. a
n
= f(n). (We do not call this a series - this term is used for
another concept.)
In this course we conne ourselves to sequences of numbers. In particular, in this chapter
we always assume that X = R. So we adopt the following denition.
Denition 3.1.1. A sequence of real numbers is a function f : N
+
R. We use the
following notation for the general term a
n
= f(n) and for the whole sequence (a
n
)
nN
+.
Example 3.1.1. 1. a
n
=
1
n
. The sequence is 1,
1
2
,
1
3
,
1
4
, . . . .
2. a
n
= n
2
. The sequence is 1, 4, 9, 16, 25, . . . .
3. a
n
=
(1)
n
n
. The sequence is 1,
1
2
,
1
3
,
1
4
, . . . .
3.2 Null sequences
In the example 1. in the previous section nth member of the sequence becomes smaller as n
becomes larger. Making computations we see that a
n
tends to zero. The same refers to the
example 3. Such sequences are called null sequences.
We will give the precise mathematical denition to the notion of a null sequence.
Denition 3.2.1. (a
n
)
nN
+ is a null sequence if
( > 0)(N N
+
)(n N
+
) [(n N) ([a
n
[ < )].
Now we are ready to verify rigorously this denition for some examples.
Example 3.2.1. (a
n
)
nN
+ with a
n
=
1
n
is a null sequence.
40
CHAPTER 3. SEQUENCES AND LIMITS
Proof. We have to prove that
( > 0)(N N
+
)(n N
+
)
_
(n > N)
_
1
n
<
__
.
Otherwise we need n to satisfy the inequality n >
1
_
+ 1.) Then [(n > N) (N >
1
)] (n >
1
n
n + 1
1
< .
An equivalent form of this inequality is
_
1
n + 1
<
_
_
n >
1
1
_
.
Chose n to be any integer greater than or equal to
1
1.
Therefore
_
1
n + 1
<
_
_
n
n + 1
1
<
_
.
If a sequence (a
n
)
n
does not converge, we say that it diverges.
Denition 3.3.2. 1. The sequence (a
n
)
nN
+ diverges to if
(M R)(N N
+
)(n N
+
) [(n > N) (a
n
> M)].
2. The sequence (a
n
)
nN
+ diverges to if
(M R)(N N
+
)(n N
+
) [(n > N) (a
n
< M)].
Example 3.3.2. 1. The sequence a
n
= n
2
diverges to .
2. The sequence a
n
= n
2
diverges to .
3. The sequence a
n
= (1)
n
diverges.
Theorem 3.3.3. A sequence (a
n
)
nN
+ can have at most one limit.
42 January 28, 2006
CHAPTER 3. SEQUENCES AND LIMITS
Proof. We have to prove that the following implication is true for all a, b R
__
lim
n
a
n
= a
_
_
lim
n
a
n
= b
__
(a = b).
By the denition
( > 0)(N
1
N
+
)(n N
+
) [(n > N) ([a
n
a[ < )],
( > 0)(N
2
N
+
)(n N
+
) [(n > N) ([a
n
b[ < )].
Suppose for a contradiction that the statement is not true. Then its negation
__
lim
n
a
n
= a
_
_
lim
n
a
n
= b
__
(a ,= b)
is true.
Fix =
1
3
[a b[ > 0 and take N = maxN
1
, N
2
. Then, if n > N, we have
[a b[ = [(a a
n
) + (a
n
b)[ [a
n
a[ +[a
n
b[ < 2 =
2
3
[a b[,
which is a contradiction.
Theorem 3.3.4. Any convergent sequence is bounded.
Proof. We have to prove that the following implication is true
__
lim
n
a
n
= a
__
[(m, M R)(n N
+
) (m a
n
M)].
Take = 1 in the denition of the limit. Then N N
+
such that for n > N we have
([a
n
a[ < 1) (a 1 < a
n
< a + 1).
Set M = maxa + 1, a
1
, . . . , a
N
and m = mina 1, a
1
, . . . , a
N
. Then
(n N
+
) (m a
n
M).
Note that Theorem 3.3.4 can be expressed by the contrapositive law in the following way
If a sequence is unbounded then it diverges.
So boundedness of a sequence is a necessary condition for its convergence!
Theorem 3.3.5. If lim
n
a
n
= a ,= 0 then
(N N
+
)(n N
+
)
_
(n > N) ([a
n
[ >
[a[
2
)
_
.
Moreover, if a > 0 then for the above n a
n
>
a
2
, and if if a < 0 then for the above n a
n
<
a
2
.
43 January 28, 2006
3.3. SEQUENCE CONVERGING TO A LIMIT
Proof. Fix =
|a|
2
. Then N N
+
such that for n > N
[a[
2
> [a a
n
[ [a[ [a
n
[,
which implies [a
n
[ > [a[
|a|
2
=
|a|
2
, and the rst assertion is proved. On the other hand
(
[a[
2
> [a a
n
[)
_
a
[a[
2
< a
n
< a +
[a[
2
_
.
Thus if a > 0 then for n > N
a
n
> a
[a[
2
=
a
2
.
Also if a < 0 then for n > N
a
n
< a +
[a[
2
=
a
2
.
Theorem 3.3.6. Let a, b R, (a
n
)
n
, (b
n
)
n
be real sequences. Then
__
lim
n
a
n
= a
_
_
lim
n
b
n
= b
_
_
(n N
+
) (a
n
b
n
)
_
(a b).
Proof. For a contradiction assume that b < a. Fix <
ab
2
so that b + < a and choose
N
1
and N
2
such that the following is true
[(n > N
1
)(a
n
> a )] [(n > N
2
)(b
n
< b +)].
If N maxN
1
, N
2
then
b
n
< b + < a < a
n
,
which contradicts to the condition (n N
+
) (a
n
b
n
).
Theorem 3.3.7. Sandwich rule. Let a R and (a
n
)
n
, (b
n
)
n
, (c
n
)
n
be real sequences. Then
[(n N
+
) (a
n
b
n
c
n
)] ( lim
n
a
n
= lim
n
c
n
= a) ( lim
n
b
n
= a).
Proof. Let > 0. Then one can nd N
1
, N
2
N
+
such that
[(n > N
1
) (a < a
n
)] [(n > N
2
) (c
n
< a +)].
Then choosing N = maxN
1
, N
2
we have
(a < a
n
b
n
c
n
< a +) ([b
n
a[ < ) .
The next theorem is a useful tool in computing limits.
Theorem 3.3.8. Let lim
n
a
n
= a and lim
n
b
n
= b. Then
(i) lim
n
(a
n
+b
n
) = a +b;
44 January 28, 2006
CHAPTER 3. SEQUENCES AND LIMITS
(ii) lim
n
(a
n
b
n
) = ab.
(iii) If in addition b ,= 0 and (n N
+
) (b
n
,= 0) then
lim
n
_
a
n
b
n
_
=
a
b
.
Proof. (i) Let > 0. Choose N
1
such that for all n > N
1
[a
n
a[ < /2. Choose N
2
such
that for all n > N
2
[b
n
b[ < /2. Then for all n > maxN
1
, N
2
[a
n
+b
n
(a +b)[ [a
n
a[ +[b
n
b[ < .
(ii) Since (b
n
)
n
is convergent, it is bounded by Theorem 3.3.4. Let K be such that
(n N
+
) ([b
n
[ K) and [a[ K. We have
[a
n
b
n
ab[ = [a
n
b
n
ab
n
+ab
n
a
n
b
n
[
[a
n
a[[b
n
[ +[a[[b
n
b[ K[a
n
a[ +K[b
n
b[.
Choose N
1
such that for n > N
1
K[a
n
a[ /2.
Choose N
2
such that for n > N
2
K[b
n
b[ /2.
Then for n > N = maxN
1
, N
2
[a
n
b
n
ab[ < .
(iii) We have
a
n
b
n
a
b
a
n
b ab
n
b
n
b
[a
n
a[[b[ +[a[[b
n
b[
[b
n
b[
.
Choose N
1
such that for n > N
1
[b
n
[ >
1
2
[b[ (possible by Theorem 3.3.5).
Choose N
2
such that for n > N
2
[a
n
a[ <
1
4
[b[.
Choose N
3
such that for n > N
3
[a[[b
n
b[ <
1
4
b
2
.
Then for n > N := maxN
1
, N
2
, N
3
a
n
b
n
a
b
< .
45 January 28, 2006
3.4. MONOTONE SEQUENCES
3.4 Monotone sequences
In this section we consider a particular class of sequences which often occur in applications.
Denition 3.4.1. (i) A sequence of real numbers (a
n
)
nN
+ is called increasing if
(n N
+
) (a
n
a
n+1
).
(ii) A sequence of real numbers (a
n
)
nN
+ is called decreasing if
(n N
+
) (a
n
a
n+1
).
(iii) A sequence of real numbers (a
n
)
nN
+ is called monotone if it is either increasing or
decreasing.
Example 3.4.1. (i) a
n
= n
2
is increasing.
(ii) a
n
=
1
n
is decreasing.
(iii) a
n
= (1)
n
1
n
is not monotone.
The next theorem is one of the main theorems in the theory of converging sequences.
Theorem 3.4.2. If a sequence of real numbers is bounded above and increasing then it is
convergent.
Proof. Let (a
n
)
nN
+ be a bounded above increasing sequence of real numbers. Since the set
a
n
[ n N
+
is bounded above, there exists supremum
a := sup
n
a
n
.
By the denition of supremum
( > 0)(N N
+
) (a
N
> a ).
Fix this N. Since (a
n
)
n
is increasing, (n > N) (a
n
> a
N
). Therefore
(n N
+
) [(n > N) (a
n
> a ).
Also by the denition of supremum
( > 0)(n N
+
) (a
n
< a +).
Therefore for n > N
(a < a
n
< a +) ([a
n
a[ < ).
In complete analogy to the previous theorem one proves the following one.
46 January 28, 2006
CHAPTER 3. SEQUENCES AND LIMITS
Theorem 3.4.3. If a sequence of real numbers is bounded below and decreasing then it is
convergent.
In many cases before computing the limit of a sequence one has to prove that the limit
exists, i.e. the sequence converges. In particular, this is the case when a sequence is dened
by a recurrent formula, i.e. the formula for the nth member of the sequence via previous
members with numbers n 1 and less.
Let us illustrate this point by an example.
Example 3.4.4. Let (a
n
)
n
be a sequence dened by
(3.4.1) a
n
=
_
a
n1
+ 2, a
1
=
2.
1. (a
n
)
n
is bounded.
Indeed, we prove that
(n N
+
) (0 < a
n
2).
Proof. Positivity is obvious.
For n = 1 a
1
= 2 2, and the statement is true.
Suppose that it is true for n = k (k 1), i.e. a
k
2.
For n = k + 1 we have
a
k+1
=
a
k
+ 2
2 + 2 = 2,
and by the principle of induction the statement is proved.
2. (a
n
)
n
is increasing.
We have to prove that
(n N
+
) (a
n+1
a
n
).
Proof. This is equivalent to prove
[(n N
+
) (
a
n
+ 2 a
n
)] [(n N
+
) (a
n
+ 2 a
2
)].
Indeed
a
n
+ 2 a
2
n
= (2 a
n
)(a
n
+ 1) 0.
By Theorem 3.4.2 we conclude that (a
n
) is convergent. Let lim
n
a
n
= x. Then, of course,
lim
n
a
n1
= x. Passing to the limit in (3.4.1) we obtain that
x =
x + 2,
from which we easily nd that x = 2.
Now we are ready to prove a very important property of real numbers.
Theorem 3.4.5. Let ([a
n
, b
n
])
nN
+ be a sequence of closed intervals on the real axis such
that:
(i) (n N) ([a
n+1
, b
n+1
] [a
n
, b
n
]);
47 January 28, 2006
3.5. CAUCHY SEQUENCES
(ii) lim
n
(b
n
a
n
) = 0
(i.e. the length of the intervals tends to zero).
Then there is a unique point which belongs to all the intervals.
Proof. The sequence (a
n
)
n
is increasing and bounded above, and the sequence (b
n
)
n
is de-
creasing and bounded below. Therefore (a
n
)
n
, (b
n
)
n
are convergent.
Let a := lim
n
a
n
; b := lim
n
b
n
.
Then
a b
by Theorem 3.3.6. Moreover,
b a = lim
n
(b
n
a
n
) = 0.
Hence
a = b
nN
+
[a
n
, b
n
].
3.5 Cauchy sequences
In this short section we are going to derive the intrinsic criterion for convergence of sequences
(i.e. the criterion which does not use the value of the limit and is expressed in terms of a
condition on the members of the sequence).
First we introduce an important notion of a Cauchy sequence.
Denition 3.5.1. A sequence of real numbers (a
n
)
nN
+ is called a Cauchy sequence if
( > o)(N N) [(n > N) (m > N)] ([a
n
a
m
[ < ) .
Proposition 3.5.1. If a sequence (a
n
)
nN
+ is convergent then it is a Cauchy sequence.
Proof. Let a := lim
n
a
n
. Let > 0. Then there exists N N such that for all n > N
[a
n
a[ < /2.
Let m, n > N. Then by the triangle inequality
[a
n
a
m
[ [a
n
a[ +[a
m
a[ < .
It turns out that the inverse to Proposition 3.5.1 is also true. This is expressed in the
next theorem that delivers the Cauchy criterion for convergence of sequences.
Theorem 3.5.1. If (a
n
)
nN
+ is a Cauchy sequence of real numbers then it is convergent.
48 January 28, 2006
CHAPTER 3. SEQUENCES AND LIMITS
Proof. Let (a
n
)
nN
+ be a Cauchy sequence. Take = 1. There exists N N such that for all n, m > N
[a
n
a
m
[ < 1.
Fix m > N. Then
[a
n
[ [a
m
[ [a
n
a
m
[ < 1.
Therefore
(n > N) ([a
n
[ < [a
m
[ + 1).
It follows that (a
n
)
nN
+ is bounded (Why?)
For any n N dene
n
:= inf
k>n
a
k
,
n
:= sup
k>n
a
k
.
Then we have
(n N) ([
n+1
,
n+1
] [
n
,
n
]).
(Justify the above)
The length of the interval [
n
,
n
] converges to zero as since from the Cauchy condition it
follows that for any > 0 there exists N N
+
such that for all k > N
a
N
< a
k
< a
N
+,
from which we conclude that
a
N
N
N
a
N
+,
or else
(n N
+
) [(n > N) (2
N
N
n
n
)].
By Theorem 3.4.5 there exists a
nN
+
[
n
,
n
].
Fix > 0. Find N such that [
N
N
[ < . Clearly (k > N) (a
k
[
N
,
n
]). It follows that
(k > N) ([a
k
a[ < ).
Therefore lim
n
a
n
= a.
49 January 28, 2006
3.6. SERIES
3.6 Series
In this section we discuss innite series. Let us start from an example which is likely to be
familiar from high school. How to make sense of the innite sum
1
2
+
1
4
+
1
8
+
1
16
. . .
This is the known sum of a geometric progression. One proceeds as follows. Dene the sum
of the rst n term
s
n
=
1
2
+
1
4
+
1
8
+ +
1
2
n
= 1
1
2
n
and then dene the innite sum s as lim
n
s
n
, so that s = 1.
This idea is used to dene formally an arbitrary series,
Denition 3.6.1. Let (a
n
)
nN
+ be a sequence of real (complex) numbers. Let
s
n
= a
1
+a
2
+ +a
n
=
n
k=1
a
k
.
We say that the series
k=1
a
k
= a
1
+a
2
+. . .
is convergent if the sequence of the partial sums (s
n
)
nN
+ is convergent. The limit of this
sequence is called the sum of the series.
If the series is not convergent we say that it is divergent (or diverges).
Theorem 3.6.1. If the series
n=1
a
n
is convergent then lim
n
a
n
= 0.
Proof. Let s
n
=
n
k=1
a
k
. Then by the denition the limit lim
n
s
n
exists. Denote it by s.
Then of course lim
n
s
n1
= s. Note that a
n
= s
n
s
n1
for n 2. Hence
lim
n
a
n
= lim
n
s
n
lim
n
s
n1
= s s = 0.
The same idea can be used to prove the following theorem.
Theorem 3.6.2. If the series
n=1
a
n
is convergent then
s
2n
s
n
= a
n+1
+a
n+2
+ +a
2n
0 as n .
50 January 28, 2006
CHAPTER 3. SEQUENCES AND LIMITS
The above theorem expresses the simplest necessary condition for the convergence of a
series.
For example, each of the following series is divergent
1 + 1 + 1 + 1 +. . . ,
1 1 + 1 1 +. . .
since the necessary condition is not satised.
Let us take a look at several important examples.
Example 3.6.3. The series
k=0
x
k
= 1 +x +x
2
+. . .
is convergent if and only if 1 < x < 1.
Indeed,
s
n
=
_
1
1x
x
n
1x
if x ,= 1,
n if x = 1.
Example 3.6.4. Harmonic series The series
1 +
1
2
+
1
3
+
1
4
+ +
1
n
+. . .
diverges.
Indeed,
s
2n
s
n
=
1
n + 1
+
1
n + 2
+ +
1
2n
>
1
2
.
Example 3.6.5. Generalised harmonic series The series
1 +
1
2
p
+
1
3
p
+
1
4
p
+ +
1
n
p
+. . .
converges if p > 1 and diverges if p 1.
3.6.1 Series of positive terms
In this subsection we conne ourselves to considering series which terms are non-negative
numbers, i.e.
k=1
a
k
with a
k
0 for every k N
+
.
The main feature of such a case that the sequence of the partial sums s
n
=
n
k=1
a
k
is
increasing.
Indeed, s
n
s
n1
= a
n
0.
This allows us to formulate a simple criterion of convergence for such a series.
51 January 28, 2006
3.6. SERIES
Theorem 3.6.6. A series of positive terms is convergent if and only if the sequence of partial
sums is bounded above.
Proof. Indeed, if a series is convergent then the sequence of partial sums is convergent by the
denition, and therefore it is bounded.
Conversely, if the sequence of partial sums is bounded above then it is convergent by
Theorem 3.4.2, and therefore the series is convergent.
Comparison tests
Here we establish tests which make possible to infer the convergence or divergence of a series
by means of comparing it with another series convergence or divergence of which is known.
Theorem 3.6.7. Let
k=1
a
k
,
k=1
b
k
be two series of positive terms. Assume that
(3.6.2) (n N
+
) (a
n
b
n
).
Then
(i) If
k=1
b
k
is convergent then
k=1
a
k
is convergent.
(ii) If
k=1
a
k
is divergent then
k=1
b
k
is divergent.
Proof. Let s
n
=
n
k=1
a
k
, s
n
=
n
k=1
b
k
. Then
(n N
+
) (s
n
s
n
).
So if the theorem follows now from Theorem 3.6.6.
Remark 5. Theorem 3.6.7 holds if instead of (3.6.2) one assumes
(K > 0)(n N
+
) (a
n
Kb
n
).
Now since for p 1
(n N
+
)
_
1
n
p
1
n
_
,
we obtain one of the assertions of Example 3.6.5 from the above theorem.
The following observation (which follows immediately from the denition of convergence
of a series) is important for the next theorem.
The convergence or divergence of a series is unaected if a nite number of terms are
inserted, or suppressed, or altered.
52 January 28, 2006
CHAPTER 3. SEQUENCES AND LIMITS
Theorem 3.6.8. Let (a
n
)
n
, (b
n
)
n
be two sequences of positive numbers. Assume that
lim
n
a
n
b
n
= L > 0. (the limit is positive and nite).
Then the series
k=1
a
k
is convergent if and only if the series
k=1
b
k
is convergent.
Proof. By the denition of the limit
(N N)(n > N)
_
1
2
L <
a
n
b
n
<
3
2
L
_
.
Now the assertion follows from the observation before the theorem and from Theorem 3.3.6.
Example 3.6.9. (i) The series
n=1
2n
n
2
+ 1
is divergent since
lim
n
2n
n
2
+1
1
n
= 2 and the series
n=1
1
n
is divergent.
(ii) The series
n=1
n
2n
3
+ 2
is convergent since lim
n
n
2n
3
+2
1
n
2
=
1
2
and the series
n=1
1
n
2
is
convergent.
Other tests of convergence
Theorem 3.6.10. (Cauchys Test or The Root Test) Let (a
n
)
nN
+ be a sequence of positive
numbers. Suppose that
lim
n
n
a
n
= l.
Then if l < 1, the series
n=1
a
n
converges; if l > 1, the series
n=1
a
n
diverges. If l = 1, no
conclusion can be drawn.
Proof. Suppose that l < 1. Choose r such that l < r < 1. Then
(N N)(n N
+
)[(n > N) (
n
a
n
< r)].
Or otherwise a
n
< r
n
. The convergence follows now from the comparison with the convergent
series
n=1
r
n
.
Next suppose that l > 1. Then
(N N)(n N
+
)[(n > N) (
n
a
n
> 1)].
Or otherwise a
n
> 1. Hence the series diverges.
53 January 28, 2006
3.6. SERIES
Theorem 3.6.11. (DAlemberts Test or The Ratio Test) Let (a
n
)
nN
+ be a sequence of
positive numbers. Suppose that
lim
n
a
n+1
a
n
= l.
Then if l < 1, the series
n=1
a
n
converges; if l > 1, the series
n=1
a
n
diverges. If l = 1, no
conclusion can be drawn.
Proof. Suppose that l < 1. Choose r such that l < r < 1. Then
(N N)(n N
+
)[(n > N) (
a
n+1
a
n
< r)].
Therefore
a
n
=
a
n
a
n1
a
n1
a
n2
a
N
+
2
a
N+1
a
N+1
< r
nN1
a
N+1
=
a
N+1
r
N+1
r
n
.
The convergence follows now from the comparison with the convergent series
n=1
r
n
.
Next suppose that l > 1. Then
(N N)(n N
+
)[(n > N) (
a
n+1
a
n
> 1)].
Or otherwise a
n+1
> a
n
. Hence the series diverges.
Example 3.6.12. Investigate the convergence of the series
n
n=1
2
n
+n
3
n
n
.
Let us use the dAlembert test. Compute the limit
lim
n
a
n+1
a
n
= lim
n
2
n+1
+n + 1
2
n
+n
3
n
n
3
n+1
n 1
=
2
3
.
Therefore the series converges.
Example 3.6.13. Investigate the convergence of the series
n
n=1
2
(1)
n
_
1
2
_
n
.
Let us use the Cauchy test. Compute the limit
lim
n
n
a
n
= lim
n
2
(1)
n
n
1
2
=
1
2
.
(We used the fact that lim
n
n
1
x
<
__
It is sucient to choose M such that M > 1 which exists by the Archimedian Principle.
Limits of functions exhibit many properties similar to those of limits of sequences. Let us
prove uniqueness of the limit of a function at a point using the Heine denition.
Theorem 4.1.5. If lim
xa
f(x) = A and lim
xa
f(x) = B then A = B.
57 January 28, 2006
4.1. LIMITS OF FUNCTION
Proof. Let x
n
a as n . Then f(x
n
) A as n and f(x
n
) B as n . From
the uniqueness of the limit of a sequence it follows that A = B.
Sometimes it is more convenient to use the Cauchy denition.
Theorem 4.1.6. Let lim
xa
f(x) = A. Let B > A. Then
( > 0)(x D(f))
_
(0 < [x a[ < )
_
f(x) < B
_
_
.
Proof. Using = B A > 0 in the Cauchy denition of the limit we have
( > 0)(x D(f))
_
(0 < [x a[ < )
_
[f(x) A[ < B A
_
_
which implies that f(x) A < B A, or f(x) < B.
The following two theorems can be proved using the Heine denition of the limit of a
function at a point and the corresponding property of a limit of a sequence. Proofs are left
as exercises.
Theorem 4.1.7. Let lim
xa
f(x) = A and lim
xa
g(x) = B. Then
(i) lim
xa
_
f(x) +g(x)
_
= A+B;
(ii) lim
xa
_
f(x) g(x)
_
= A B;
If in addition B ,= 0 then
(iii) lim
xa
_
f(x)
g(x)
_
=
A
B
.
Theorem 4.1.8. Let lim
xa
f(x) = A and lim
xa
g(x) = B. Suppose that
( > 0)
_
x R[ 0 < [x a[ < D(f) D(g)
_
and
_
0 < [x a[ <
_
_
f(x) g(x)
.
Then A B, i.e.
lim
xa
f(x) lim
xa
g(x).
Using Theorem 4.1.7 one can easily compute limits of some functions.
Example 4.1.9.
lim
x2
x
3
+ 2x
2
7
2x
3
4
=
lim
x2
x
3
+ lim
x2
2x
2
lim
x2
7
lim
x2
2x
3
lim
x2
4
=
8 + 8 7
16 4
=
3
4
.
58 January 28, 2006
CHAPTER 4. LIMITS OF FUNCTIONS AND CONTINUITY
One-sided limits
Denition 4.1.4. (i) Let f be dened on an interval (a, d) R. We say that f(x)
b as x a+ and write lim
xa+
f(x) = b if
( > 0)( > 0)(x (c, d))
_
(0 < x a < )
_
[f(x) b[ <
_
_
.
(ii) Let f be dened on an interval (c, a) R. We say that f(x) b as x a and write
lim
xa
f(x) = b if
( > 0)( > 0)(x (c, d))
_
( < x a < 0)
_
[f(x) b[ <
_
_
.
59 January 28, 2006
4.2. CONTINUOUS FUNCTIONS
4.2 Continuous functions
Denition 4.2.1. Let a R. Let a function f be dened in a neighborhood of a. Then the
function f is called continuous at a if
lim
xa
f(x) = f(a).
Using the denitions of the limit given in the previous section we can formulate the above
denition in the following way.
Denition 4.2.2. A function f is called continuous at a point a R if f is dened on an
interval (c, d) containing a and
( > 0)( > 0)(x (c, d))
_
([x a[ < )
_
[f(x) f(a)[ <
_
_
.
Note the dierence between this denition and the denition of the limit: the function f
has to be dened at a.
Using the above denition it is easy to formulate what it means that a function f is
discontinuous at a point a:
A function f is discontinuous at a point a R if
f is not dened on any neighbourhood (c, d) containing a, or if
( > 0)( > 0)(x D(f))
_
([x a[ < )
_
[f(x) f(a)[
_
_
.
One more equivalent way to dene continuity at a point is to use the Heine denition of
the limit.
Denition 4.2.3. A function f is called continuous at a point a R if f is dened on an
interval (c, d) containing a and for any sequence (x
n
)
nN
such that
(i) (n N)
_
(x
n
(c, d)) (x
n
,= a)
_
,
(ii) x
n
a as n
we have
lim
n
f(x
n
) = f(a).
Example 4.2.1. Let c R. Let f : R R be dened by f(x) = c for any x R. Then f is
continuous at any point in R.
Example 4.2.2. Let f : R R be dened by f(x) = x for any x R. Then f is continuous
at any point in R.
The following theorem easily follows from the denition of continuity and properties of
limits.
60 January 28, 2006
CHAPTER 4. LIMITS OF FUNCTIONS AND CONTINUITY
Theorem 4.2.3. Let f and g be continuous at a R. Then
(i) f +g is continuous at a.
(ii) f g is continuous at a.
Moreover, if g(a) , = 0, then
(iii)
f
g
is continuous at a.
Based on this theorem and the two examples above we conclude that
f(x) =
a
0
x
n
+a
1
x
n1
+ +a
n
b
0
x
m
+b
1
x
m1
+ +b
m
is continuous at every point of its domain of denition.
Theorem 4.2.4. Let g be continuous at a R, f be continuous at b = g(a) R. Then f g
is continuous at a.
Proof. Fix > 0. Since f is continuous at b,
( > 0)(y D(f))
_
([y b[ < )
_
[f(y) f(b)[ <
_
_
.
Fix this > 0. From the continuity of g at a
( > 0)(x D(g))
_
([x a[ < )
_
[g(x) g(a)[ <
_
_
.
From the above it follows that
( > 0)( > 0)(x D(g))
_
([x a[ < )
_
[f(g(x)) f(g(a))[ <
_
_
.
This proves continuity of f g at a.
Another useful characterization of continuity of a function f at a point a is the following.
f is continuous at a point a R if and only if
lim
xa
f(x) = lim
xa+
f(x) = f(a),
which means that the one-sided limits exist, are equal and equal the value of the function at
a.
Theorem 4.2.5. Let f be continuous at a R. Let f(a) < B. Then there exists a neigh-
bourhood of a such that f(x) < B for all the points x from the neighbourhood.
Proof. Take = B f(a) in Denition 4.2.2. Then
( > 0)(x (c, d))
_
([x a[ < )
_
[f(x) f(a)[ <
_
_
.
Remark 6. A similar fact is true for the case f(x) > B.
61 January 28, 2006
4.2. CONTINUOUS FUNCTIONS
Boundedness of functions
Denition 4.2.4. Let f : D(f) R, D(f) R. f is called bounded if Ran(f) is a bounded
subset of R.
Example 4.2.6. Let f : R R be dened by f(x) =
1
1+x
2
. Then Ran(f) = (0, 1], so f is
bounded.
Denition 4.2.5. Let A D(f). f is called bounded above on A if
(K R)(x A)
_
f(x) K
_
.
Remark 7. Boundedness below and boundedness on a set is dened analogously.
Theorem 4.2.7. If f is continuous at a then there exists > 0 such that f is bounded on
the interval (a , a +).
Proof. Since lim
xa
f(x) = f(a)
( > 0)(x D(f))
_
([x a[ < )
_
[f(x) f(a)[ < 1
_
.
So on the interval (a , a +)
f(a) 1 < f(x) < f(a) + 1.
62 January 28, 2006
CHAPTER 4. LIMITS OF FUNCTIONS AND CONTINUITY
4.3 Continuous functions on a closed interval
In the previous section we dealt with functions which were continuous at a point. Here we
consider functions which are continuous at every point of an interval [a, b]. In this case we
speak of functions continuous at [a, b]. In comparison with the previous section we will be
interested here in global behaviour of such functions.
We rst dene this notion formally:
Denition 4.3.1. Let [a, b] R be a closed interval, f a function with f : D(f) R, and
with [a, b] D(f). Then we say that f is a continuous function on [a, b] if:
(i) f is continuous at every point of (a, b) and
(ii) lim
xa
+ f(x) = f(a); lim
xb
f(x) = f(b).
Our rst theorem exhibits the intermediate-value property of continuous function.
Theorem 4.3.1. The Intermediate Value Theorem
Let f be a continuous function on a closed interval [a, b] R. Suppose that f(a) < f(b).
Then
_
[f(a), f(b)]
__
x [a, b]
_
_
f(x) =
_
.
Proof. If = f(a) or = f(b) then there is nothing to prove. Fix (f(a), f(b)). Let us
introduce the set
A := x [a, b] [ f(x) < .
The set A is not empty since a A. The set A is bounded above (by b). Therefore there
exists := supA.
Our aim now is to prove that f() = . We will do that by ruling out two other possibilities:
f() < and f() > .
First, let us assume that f() < . Then by Theorem 4.2.5 we have that
( > 0)(x D(f))
_
_
x ( , +)
_
_
f(x) <
_
_
.
Therefore
(x
1
R)
_
_
x
1
(, +)
_
_
f(x
1
) <
_
_
.
In other words
(x
1
R)
_
_
x
1
>
_
_
x
1
A
_
_
.
This contradicts to the fact that is an upper bound of A.
Next, let us assume that f() > . Then by the remark after Theorem 4.2.5 we have that
( > 0)
_
D(f) ( , +) A
c
(A
c
is the complement to A).
This contradicts to the fact that = supA.
The next theorem establishes boundedness of functions continuous on an interval.
Theorem 4.3.2. Let f be continuous on [a, b]. Then f is bounded on [a, b].
63 January 28, 2006
4.3. CONTINUOUS FUNCTIONS ON A CLOSED INTERVAL
Proof. Let us introduce a set
A := x [a, b]
f is bounded on [a, x] .
Note that
1. A ,= since a A.
2. A is bounded (by b).
Therefore there exists := supA.
Our rst task will be to prove that = b.
First, we prove that > a. Indeed, if we let = a then by continuity of f at a it follows
that
( > 0)(x D(f))
_
(0 x a < )
_
[f(x) f(a)[ < 1
_
.
Hence the function f is bounded on [a, a +) . Therefore
_
x
1
(a, a +)
__
f is bounded on [a, x
1
]
_
,
which proves that > a.
Now suppose that < b. Then by Theorem 4.2.7
(4.3.3) ( > 0)
_
f is bounded on ( , +)
.
Fix as above. By the denition of sup
_
x
1
( , ]
__
f is bounded on [a, x
1
]
.
Also from (4.3.3) it follows that
_
x
2
(, +)
__
f is bounded on [x
1
, x
2
]
.
Therefore f is bounded on [a, x
2
] where x
2
> which contradicts to the fact that is the
supremum of A. This proves that = b.
(Note that this does not complete the proof since supremum may not belong to a set, i.e.
it is possible that , A).
From the continuity of f at b it follows that
(
1
> 0)
_
f is bounded on (b
1
, b]
.
By the denition of supremum
(x
1
(b
1
, b] )
_
f is bounded on [a, x
1
]
.
Therefore f is bounded on [a, b].
The last theorem asserts that the range of a continuous function restricted to a closed
interval is a bounded subset in R, so that there exist supremum and inmum of it. The next
theorem asserts that these values are attained, which means that there are point in [a, b] such
that the values of the function at this points equal to the supremum (inmum).
64 January 28, 2006
CHAPTER 4. LIMITS OF FUNCTIONS AND CONTINUITY
Theorem 4.3.3. Let f be continuous on [a, b] R. Then
(y [a, b])(x [a, b])
_
f(x) f(y)
_
.
(i.e. f(y) = max
x[a,b]
f(x). ) Similarly,
(z [a, b])(x [a, b])
_
f(x) f(z)
_
.
Proof. Let us introduce the set of values of f on [a, b]
F := f(x)
x [a, b]
We know that F ,= and by Theorem 4.3.2 F is bounded. Therefore its supremum exists.
Denote := supF.
Our aim is to prove that there exists y [a, b] such that f(y) = . We prove this by
contradiction.
Suppose that
( x [a, b] )
_
f(x) <
_
.
Dene the following function
g(x) =
1
f(x)
, x [a, b].
Since the denominator is never zero on [a, b] we conclude that g is continuous and by Theorem
4.3.2 is bounded on [a, b].
At the same time by the denition of supremum
( > 0)(x [a, b])
_
f(x) >
_
,
in other words f(x) < , and so g(x) >
1
_
.
Therefore g is unbounded on [a, b] which is a contradiction.
4.4 Uniform continuity
Denition 4.4.1. Let f be a function dened on a set A R. We say that f is uniformly
continuous on A if
( > 0)( > 0)(x
1
, x
2
A)
_
_
[x
1
x
2
[ <
_
_
[f(x
1
) f(x
2
)[ <
_
_
.
Theorem 4.4.1. Let f be dened and continuous on a closed interval [a, b]. Then f is
uniformly continuous on [a, b].
65 January 28, 2006
4.5. INVERSE FUNCTIONS
We leave this theorem without proof. See [4], Theorem 3.82.
The next example shows that the property of uniform continuity is stronger than the
property of continuity.
Example 4.4.2. The function f(x) =
1
x
is continuous on (0, 1) but not uniformly continuous.
Indeed, take x
n
=
1
n
, y
n
=
1
n+1
.
Example 4.4.3. The function f(x) =
x is uniformly continuous on [1, ).
Indeed,
[f(x
1
) f(x
2
)[ = [
x
1
x
2
[ =
[x
1
x
2
[
x
1
+
x
2
[x
1
x
2
[.
4.5 Inverse functions
In this section we discuss the inverse to a continuous function. Recall that for existence of
the inverse of a function, the function must be a bijection. It is advisable to revise Section
1.7.2.
It is not dicult to show (this point is delegated to Exercises) that a continuous bijection
dened on an interval [a, b] is monotone (either increasing or decreasing), the notion which
was discussed for sequences and the intuitive meaning of which is clear. The precise denition
reads as follows.
Denition 4.5.1. A function f dened on [a, b] is called increasing if
(x
1
, x
2
[a, b])
_
(x
1
< x
2
)
_
f(x
1
) < f(x
2
)
_
_
.
Theorem 4.5.1. Let f be continuous on [a, b] and increasing. Let f(a) = c, f(b) = d. Then
there exists a function g : [c, d] [a, b] which is continuous, increasing and such that
(y [c, d])
_
f(g(y)) = y
_
.
Proof. Let [c, d]. By Theorem 4.3.1
( [a, b])
_
f() =
_
.
There is only one such value since f is increasing. (Why?) The inverse function g is dened
by
= g().
It is easy to see that g is increasing.
Indeed, let y
1
< y
2
, y
1
= f(x
1
), y
2
= f(x
2
). Suppose that at the same time x
2
x
1
. Since
f is increasing it follows that y
2
y
1
. Contradiction.
66 January 28, 2006
CHAPTER 4. LIMITS OF FUNCTIONS AND CONTINUITY
Now let us prove that g is continuous.
Let y
0
(c, d). Then
_
x
0
(a, b)
_ _
y
0
= f(x
0
)
,
or in other words
x
0
= g(y
0
).
Let > 0. We assume also that is small enough such that [x
0
, x
0
+ ] (a, b). Let
y
1
= f(x
0
), y
2
= f(x
0
+). Since g is increasing we have that
_
y (y
1
, y
2
)
_
x = g(y) (x
0
, x
0
+)
.
Take = miny
2
y
0
, y
0
y
1
. Then
_
[y y
0
[ <
_
[g(y) g(y
0
)[ <
.
The continuity at the ends of the interval can be established similarly.
67 January 28, 2006
Chapter 5
Dierential Calculus
5.1 Denition of derivative. Elementary properties
Denition 5.1.1. Let f be dened in a neighbourhood of a R. (This mens that
( > 0)
_
(a , a +) D(f)
).
We say that f is dierentiable at a if the limit lim
h0
f(a +h) f(a)
h
exists. This limit, denoted
by f
) =
_
x D(f) [ lim
h0
f(a +h) f(a)
h
exists
_
with the values
f
(x) = lim
h0
f(x +h) f(x)
h
.
Example 5.1.1. Let c R. Let f : R R be dened by f(x) = c. Then f is dierentiable
at any x R and f
(x) = 0.
Proof.
lim
h0
f(a +h) f(a)
h
= lim
h0
c c
h
= 0.
Example 5.1.2. Let f : R R be dened by f(x) = x. Then f is dierentiable at any
x R and f
(x) = 1.
Proof.
lim
h0
f(a +h) f(a)
h
= lim
h0
a +h a
h
= 1.
68
CHAPTER 5. DIFFERENTIAL CALCULUS
Example 5.1.3. Let f : R R be dened by f(x) = x
2
. Then f is dierentiable at any
x R and f
(x) = 2x.
Proof.
lim
h0
f(a +h) f(a)
h
= lim
h0
(a +h)
2
a
2
h
= lim
h0
2ah +h
2
h
= lim
h0
(2a +h) = 2a.
Example 5.1.4. Let n N. Let f : R R be dened by f(x) = x
n
. Then f is dierentiable
at any x R and f
(x) = nx
n1
.
The proof which can be performed by induction, based on the product rule below, is left
as an exercise.
Example 5.1.5. Let f : R R be dened by f(x) = [x[. Then f is dierentiable at any
x R 0. f is not dierentiable at 0.
Proof. If x > 0 then
lim
h0
f(x +h) f(x)
h
= 1.
If x < 0 then
lim
h0
f(x +h) f(x)
h
= 1.
The derivative does not exist at 0 since
lim
h0+
f(x +h) f(x)
h
,= lim
h0
f(x +h) f(x)
h
.
Since in the above example the function f is continuous at 0, it shows that continuity
does not imply dierentiability.
Theorem 5.1.6. If f is dierentiable at a then f is continuous at a.
Proof.
f
(a) = lim
h0
f(a +h) f(a)
h
.
Hence we have
lim
xa
_
f(x) f(a)
_
= lim
h0
_
f(a +h) f(a)
_
= lim
h0
f(a +h) f(a)
h
h = f
(a) lim
h0
h = 0.
Therefore
lim
xa
f(x) = f(a).
69 January 28, 2006
5.1. DEFINITION OF DERIVATIVE. ELEMENTARY PROPERTIES
Remark 8. (Important!) If f is dierentiable at a R then there exists a function (x)
such that lim
xa
(x) = 0 and
f(x) = f(a) +f
(a).
Then (x) 0 as x a and f(x) = f(a) +f
(a) = f
(a) +g
(a).
The proof is left as an exercise.
Theorem 5.1.8. (The product rule). If f and g are dierentiable at a then f g is also
dierentiable at a, and
(f g)
(a) = f
(a).
Proof.
lim
h0
(f g)(a +h) f g(a)
h
= lim
h0
f(a +h)g(a +h) f(a)g(a)
h
= lim
h0
_
f(a +h)[g(a +h) g(a)]
h
+
[f(a +h) f(a)]g(a)
h
_
= lim
h0
f(a +h) lim
h0
g(a +h) g(a)
h
+ lim
h0
f(a +h) f(a)
h
lim
h0
g(a)
=f
(a).
Theorem 5.1.9. If g is dierentiable at a and g(a) ,= 0 then =
1
g
is also dierentiable at
a, and
(a) =
_
1
g
_
(a) =
g
(a)
[g(a)]
2
.
Proof. Follows from
(a +h) (a)
h
=
g(a) g(a +h)
hg(a)g(a +h)
.
Theorem 5.1.10. (The quotient rule). If f and g are dierentiable at a and g(a) ,= 0
then =
f
g
is also dierentiable at a, and
(a) =
_
f
g
_
(a) =
f
(a)
[g(a)]
2
.
70 January 28, 2006
CHAPTER 5. DIFFERENTIAL CALCULUS
Proof. Follows from Theorems 5.1.8 and 5.1.9.
Theorem 5.1.11. (The Chain rule) If g is dierentiable at a R and f is dierentiable
at g(a) then f g is dierentiable at a and
(f g)
(a) = f
(g(a)) g
(a).
Proof. By the denition of the derivative and Remark 8 we have
f(y) f(y
0
) = f
(y
0
)(y y
0
) +(y)(y y
0
),
where (y) 0 as y y
0
. Set in the above equality, for x ,= a, y = g(x), y
0
= g(a) and
divide both sides by x a. We obtain
f(g(x)) f(g(a))
x a
= f
(g(a))
g(x) g(a)
x a
+(g(x))
g(x) g(a)
x a
.
By Theorem 5.1.6 g is continuous at a. Hence y = g(x) g(a) = y
0
as x a, and
(g(x)) 0 as x a. Passing to the limit in the above equality as x a we obtain the
required.
Example 5.1.12. Let f : R R dened by f(x) = (x
2
+ 1)
100
. Then f is dierentiable at
every point in R and f
(x) = 100(x
2
+ 1)
99
2x.
In the next theorem we establish relation between the derivative of an invertible function
and the derivative of the inverse function.
Theorem 5.1.13. Let f be continuous, increasing on (a, b) and given by y = f(x). Suppose
that for some x
0
(a, b) f is dierentiable at x
0
and f
(x
0
) ,= 0. Then the inverse function
g = f
1
given by x = g(y) is dierentiable at y
0
= f(x
0
) and
g
(y
0
) =
1
f
(x
0
)
.
Proof. Remark 8 implies that
y y
0
= f(g(y)) f(g(y
0
))
= f
(g(y
0
))(g(y) g(y
0
)) +(g(y))(g(y) g(y
0
)),
where (g(y)) 0 as g(y) g(y
0
). Since g is continuous at y
0
it follows that g(y) g(y
0
)
as y y
0
, and hence (g(y)) 0 as y y
0
. Therefore we have
g(y) g(y
0
)
y y
0
=
g(y) g(y
0
)
f
(g(y
0
))(g(y) g(y
0
)) +(g(y))(g(y) g(y
0
))
=
1
f
(g(y
0
)) +(g(y))
1
f
(g(y
0
))
as y y
0
.
Example 5.1.14. Let n N. Let f : R
+
R
+
be dened by f(x) = x
n
. Then the inverse
function is g(y) =
n
y = y
1/n
. We have that f
(x) = nx
n1
. Hence by the previous theorem
g
(y) =
1
f
(x)
=
1
nx
n1
=
1
n
1
y
n1
n
=
1
n
y
1
n
1
.
71 January 28, 2006
5.2. THEOREMS ON DIFFERENTIABLE FUNCTIONS
One-sided derivatives
Similar to one-sided limits we dene left and right derivatives of f at a as
f
(a) = lim
h0
f(a +h) f(a)
h
, f
+
(a) = lim
h0+
f(a +h) f(a)
h
.
5.2 Theorems on dierentiable functions
Theorems of this section show how with the help of derivative (which is a local notion) one
can investigate behaviour of functions on intervals.
Theorem 5.2.1. Let f be a function dened on (a, b). If x
0
(a, b) is a maximum (or
minimum) point for f on (a, b) and f is dierentiable at x
0
then f
(x
0
) = 0,
Note that we do not assume dierentiability, or even continuity, of f at any other point.
Proof. We prove the theorem for the case of maximum. The case of minimum is similar.
If h is any number such that x
0
+h (a, b), then
f(x
0
) f(x
0
+h).
Thus for h > 0 we have
f(x
0
+h) f(x
0
)
h
0,
and consequently
lim
h0+
f(x
0
+h) f(x
0
)
h
0.
On the other hand, if h < 0, then
f(x
0
+h) f(x
0
)
h
0,
so that
lim
h0
f(x
0
+h) f(x
0
)
h
0.
By hypothesis f
(x
0
) exists so that f
(x
0
) = f
+
(x
0
). So from the above f
(x
0
) = 0.
Note that the inverse statement is false. A simple example is
f : R R, f(x) = x
3
.
We see that f
(x
0
) = 0
.
72 January 28, 2006
CHAPTER 5. DIFFERENTIAL CALCULUS
Proof. It follows from the continuity of f that f has a maximum and a minimum value on
[a, b].
Suppose that the maximum value occurs at x
0
(a, b). Then by Theorem 5.2.1 f
(x
0
) = 0,
and we are done.
Suppose next that the minimum value occurs at x
0
(a, b). Then again by Theorem 5.2.1
f
(x
0
) = 0.
Finally, suppose that the maximum value and the minimum value both occur at the end
points. Since f(a) = f(b), the maximum value and the minimum value are equal, so that f
is a constant. Hence f
(x
0
) =
f(b) f(a)
b a
_
.
Proof. Let
g(x) = f(x)
_
f(b) f(a)
b a
_
(x a).
Then g is continuous on [a, b] and dierentiable on (a, b), and
g
(x) = f
(x)
f(b) f(a)
b a
.
Moreover, g(a) = f(a) and
g(b) = f(b)
_
f(b) f(a)
b a
_
(b a) = f(a).
Therefore by Rolles Theorem
(x
0
(a, b)) [g
(x
0
) = 0].
Corollary 5.2.1. If f is dened on an interval and f
(x) =
f(b) f(a)
b a
.
But f
(x) = g
(x) > 0. Let a and b be any two points in the interval with a < b.
Then by the Mean Value theorem there is a point x in (a, b) such that
f
(x) =
f(b) f(a)
b a
.
But f
(x
0
) = [g(b) g(a)]f
(x
0
)
_
.
(If g(b) ,= g(a), and g
(x
0
) ,= 0, the above equality can be rewritten as
f(b) f(a)
g(b) g(a)
=
f
(x
0
)
g
(x
0
)
.
Note that if g(x) = x, we obtain the Mean Value Theorem).
Proof. Let h : [a, b] R be dened by
h(x) = [f(b) f(a)]g(x) [g(b) g(a)]f(x).
Then h(a) = f(b)g(a) f(a)g(b) = h(b), so that h satises Rolles theorem. Therefore
_
x
0
(a, b)
_
_
0 = h
(x
0
) = [f(b) f(a)]g
(x
0
) [g(b) g(a)]f
(x
0
)
_
.
Theorem 5.2.5. (LHopitals Rule). Let f and g be dierentiable and g
(x) ,= 0, in a
neighbourhood of a point a R, and such that f(a) = g(a) = 0. Suppose that the limit
lim
xa
f
(x)
g
(x)
exists. Then
lim
xa
f(x)
g(x)
= lim
xa
f
(x)
g
(x)
74 January 28, 2006
CHAPTER 5. DIFFERENTIAL CALCULUS
Proof. By the Cauchy Mean Value Theorem
f(a +h) f(a)
g(a +h) g(a)
=
f
(a +th)
g
(a +th)
for some 0 < t < 1. Passing to the limit as h 0 we get the result.
LHopitals Rule is a useful tool in computing limits.
Example 5.2.6. Let m, n N, 0 ,= a R.
lim
xa
x
m
a
m
x
n
a
n
= lim
xa
mx
m1
nx
n1
=
m
n
a
mn
.
5.3 Approximation by polynomials. Taylors Theorem
Polynomials, i.e. functions which are given
a
0
+a
1
x +a
2
x
2
+ +a
n
x
n
,
where n N, a
0
, a
1
, . . . , a
n
R, is the simplest class of functions. In this section we will
show that functions that are dierentiable suciently many time can be approximated by
polynomials. We use the notation f
(n)
(a) for nth derivative of f at a.
Lemma 1. If f is the polynomial
a
0
+a
1
x +a
2
x
2
+ +a
n
x
n
,
then
a
k
=
1
k!
f
(k)
(0) (0 k n).
Proof. a
k
is computed by k times dierentiation of f and evaluation at 0.
Thus for a polynomial f of degree n we have
f(x) = f(0) +xf
(0) +
x
2
2!
f
(0) + +
x
n
n!
f
(n)
(0).
More generally, if x = a +h, where a is xed, we have
f(a +h) = f(a) +hf
(a) +
h
2
2!
f
(h) + +
h
n
n!
f
(n)
(a).
The next theorem gives approximation of a suciently nice function by a polynomial.
We will call it Taylors theorem (see [4] for comments).
Theorem 5.3.1. Let h > 0, p 1. Suppose that f and its derivatives up to order n 1 be
continuous on [a, a + h] and f
(n)
exists on (a, a + h). Then there exists a number t (0, 1)
such that
f(a +h) = f(a) +hf
(a) +
h
2
2!
f
(h) + +
h
n1
(n 1)!
f
(n1)
(a)
+
h
n
(1 t)
np
p(n 1)!
f
(n)
(a +th). (5.3.1)
75 January 28, 2006
5.3. APPROXIMATION BY POLYNOMIALS. TAYLORS THEOREM
Proof. Set
R
n
= f(a +h) f(a) hf
(a)
h
2
2!
f
(a)
h
n1
(n 1)!
f
(n1)
(a).
Dene g : [a, a +h] R by
g(x) = f(a +h) f(x) (a +h x)f
(x)
(a +h x)
2
2!
f
(x)
(a +h x)
n1
(n 1)!
f
(n1)
(x)
(a +h x)
p
R
n
h
p
.
Clearly g(a + h) = 0. From our denition of R
n
it follows that g(a) = 0. Therefore we
can apply Rolles theorem to g on [a, a +h]. Hence there exists t (0, 1) such that
g
(a +th) = 0.
It is easy to verify that
g
(a +th) =
[h(1 t)]
n1
(n 1)!
f
(n)
(a +th) +
p(1 t)
p1
h
R
n
,
all other terms in the dierentiation canceling in pairs (It is advisable to check this). From
this we nd that
R
n
=
h
n
(1 t)
np
p(n 1)!
f
(n)
(a +th),
which proves the theorem.
Corollary 5.3.1. (Forms of the remainder).
Let the conditions of Theorem 5.3.1 be satised. Then
f(a +h) = f(a) +hf
(a) +
h
2
2!
f
(h) + +
h
n1
(n 1)!
f
(n1)
(a) +R
n
, (5.3.2)
where
(i) (Lagrange) R
n
=
h
n
n!
f
(n)
(a +th) for some t (0, 1).
(ii) (Cauchy) R
n
=
(1 s)
n1
h
n
(n 1)!
f
(n)
(a +sh) for some s (0, 1).
Proof. Putting p = n in (5.3.1) gives (i), an d putting p = 1 gives (ii).
Remark 9. The number in (5.3.2) is called the remainder term. Note that lim
h0
R
n
h
n1
= 0,
which means that R
n
0 as h 0 faster than any other term in (5.3.2).
76 January 28, 2006
CHAPTER 5. DIFFERENTIAL CALCULUS
Example 5.3.2. Let R. Let f : (1, 1] R be given by f(x) = (1 +x)
. Since
f
(n)
(x) = ( 1) . . . ( n + 1)(1 +x)
n
,
f
(n)
(0) = ( 1) . . . ( n + 1),
Taylors expansion of f takes the form
(1 +x)
= 1 +
1!
x +
( 1)
2!
x
2
+ +
( 1) . . . ( n + 1)
n!
x
n
+R
n+1
(x),
where the remainder term in the Lagrange form is
R
n+1
(x) =
( 1) . . . ( n)
(n + 1)!
(1 +tx)
n1
x
n+1
for some t (0, 1).
77 January 28, 2006
Chapter 6
Series
In this chapter we continue to study series. We already studied series of positive terms. (It
is advisable to revise Section 3.6.) In this chapter we will discuss convergence and divergence
of series which have innitely many positive and innitely many negative terms.
6.1 Series of positive and negative terms
6.1.1 Alternating series
Theorem 6.1.1. Let (a
n
)
n
be a sequence of positive numbers such that
(i) (n N)
_
a
n
a
n+1
_
;
(ii) lim
n
a
n
= 0.
Then the series
a
1
a
2
+a
3
a
4
+ + (1)
n1
a
n
+ =
n=1
(1)
n1
a
n
converges.
Proof. For each n N set
s
n
=a
1
a
2
+a
3
a
4
+ + (1)
n1
a
n
=
n
k=1
(1)
k1
a
k
,
t
n
=a
1
a
2
+a
3
a
4
+ +a
2n1
a
2n
= s
2n
,
u
n
=a
1
a
2
+a
3
a
4
+ +a
2n1
= s
2n1
.
Then we have by (i)
t
n+1
t
n
= s
2n+2
s
2n
= a
2n+1
a
2n+2
0.
Thus the sequence (t
n
)
n
is increasing. Similarly
u
n+1
u
n
= a
2n
+a
2n+1
0,
78
CHAPTER 6. SERIES
so that (u
n
)
n
is decreasing. Also for every n N
u
n
t
n
= a
2n
0, i.e. u
n
t
n
.
Thus the sequence (u
n
)
n
is decreasing and bounded below by t
1
. Therefore is converges to
a limit u = lim
n
u
n
. The the sequence (t
n
)
n
is increasing and bounded above by u
1
.
Therefore is converges to a limit t = lim
n
t
n
. Moreover u t = lim
n
(u
n
t
n
) =
lim
n
a
2n
= 0. Denote t = u = s. So we have
s
2n
= t
n
s as n and s
2n+1
= u
n
s.
Therefore s
n
s as n .
Example 6.1.2. The series
1
1
2
+
1
3
1
4
+ +
converges.
Example 6.1.3. The series
n=1
x
n
n
converges if and only if x [1, 1).
Proof. First, let x > 0. Then
lim
n
a
n+1
a
n
= lim
n
xn
n + 1
= x
By the dAlembert test if x < 1 then the series converges, if x > 1 then it diverges. If x = 1
we obtain the harmonic series which is divergent.
Now let x < 0. Denote y = x > 0. Then we have an alternating series
n=1
(1)
n
y
n
n
.
If y 1 then
y
n
n
>
y
n+1
n + 1
and lim
n
y
n
n
= 0, so that the series converges by Theorem 6.1.1. If
y > 1 then lim
n
y
n
n
,= 0, and the series diverges.
6.1.2 Absolute convergence
Denition 6.1.1. The series
n=1
a
n
is said to be absolutely convergent if the series
n=1
[a
n
[
is convergent. A series which is convergent but not absolutely convergent is said to be
conditionally convergent.
Theorem 6.1.4. An absolutely convergent series is convergent.
79 January 28, 2006
6.1. SERIES OF POSITIVE AND NEGATIVE TERMS
Proof. Let
n=1
a
n
be an absolutely convergent series. Dene
b
n
=
_
a
n
if a
n
0,
0 if a
n
< 0.
c
n
=
_
0 if a
n
0,
a
n
if a
n
< 0.
Then (n N)(b
n
0) (c
n
0). Note that
a
n
= b
n
c
n
,
[a
n
[ = b
n
+c
n
.
Since (n N)(b
n
[a
n
[) (c
n
[a
n
[) by the comparison test we conclude that the series
n=1
b
n
and
n=1
c
n
converge, and therefore so does the series
n=1
(b
n
c
n
).
Example 6.1.5. (i) The series
1
1
4
+
1
9
1
6
+. . .
is absolutely convergent since the series
1 +
1
4
+
1
9
+
1
6
+. . .
converges.
(ii) The series
1
1
2
+
1
3
1
4
+. . .
is conditionally convergent since the series
1 +
1
2
+
1
3
+
1
4
+ +
1
n
+. . .
diverges.
Example 6.1.6. The exponential series
1 +x +
x
2
2!
+
x
3
3!
+ +
x
n
n!
+. . .
converges for all x R.
Proof. To see this consider the series of absolute values
n=1
[x[
n
n!
.
Using dAlemberts test prove that it is convergent for every x R. From this conclude that
the exponential series is absolutely convergent. Use Theorem 6.1.4 to conclude that the series
converges. Details are left as an exercise.
80 January 28, 2006
CHAPTER 6. SERIES
6.1.3 Rearranging series
Example 6.1.7. Let s be the sum of the series
1
1
2
+
1
3
1
4
+. . .
Let s
n
be the nth partial sum, i.e.
s
n
= 1
1
2
+
1
3
1
4
+ +
(1)
n1
n
.
Let us now rearrange the series in the following way
1
1
2
1
4
+
1
3
1
6
1
8
+
1
5
1
10
1
12
+. . .
Let t
n
be be the nth partial sum of the new series. Then
t
3n
=
_
1
1
2
1
4
_
+
_
1
3
1
6
1
8
_
+ +
_
1
2n 1
1
4n 2
1
4n
_
=
1
2
_
1
1
2
_
+
1
2
_
1
3
1
4
_
+ +
1
2
_
1
2n 1
1
2n
_
=
1
2
s
2n
.
But s
2n
s as n . Hence t
3n
1
2
s as n . Also t
3n+1
= t
3n
+
1
2n+1
1
2
s and
t
3n+2
= t
3n+1
1
4n+2
1
2
s. Therefore
t
n
1
2
s as n .
Remark 10. The series 6.1.7 is conditionally convergent. We see that by rearranging this
series we change its sum. In fact, one can prove that any conditionally convergent series can
be rearranged in such a way that the sum of the rearranged series is any real number chosen
in advance.
In the next theorem we prove that rearranging an absolutely convergent series does not
alter its sum.
(We say that the series
n=1
b
n
is a rearrangement of the series
n=1
a
n
if there is a bijection
f on the set of natural numbers N such that
(n N)(b
n
= a
f(n)
).)
Theorem 6.1.8. Let
n=1
a
n
be an absolutely convergent series with sum s. Then any rear-
rangement of
n=1
a
n
is also convergent with sum s.
81 January 28, 2006
6.1. SERIES OF POSITIVE AND NEGATIVE TERMS
Proof. First let us consider the case in which a
n
0 for all n. Let b
1
+ b
2
+ b
3
+ . . . is a
rearrangement of a
1
+a
2
+a
3
+. . . . Denote by s
n
and t
n
the corresponding partial sums, i.e.
s
n
= a
1
+a
2
+a
3
+ +a
n
, t
n
= b
1
+b
2
+b
3
+ +b
n
.
The sequences (s
n
)
n
and (t
n
)
n
are increasing and s
n
s as n . So (n N)(s
n
s).
Also
(n N)(k N)
_
b
1
+b
2
+b
3
+ +b
n
s
k
_
.
(Each of the numbers b
1
, b
2
, . . . , b
n
is one of as.) Therefore (n N)(t
n
s). Hence the
sequence (t
n
)
n
is bounded above by s. Thus it converges, and the limit t s. By the same
argument considering a
1
+ a
2
+ a
3
+ . . . as a rearrangement of b
1
+ b
2
+ b
3
+ . . . we obtain
that s t. So we conclude that t = s.
The general case is treated similarly to the proof of Theorem 6.1.4. Dene
d
n
=
_
a
n
if a
n
0,
0 if a
n
< 0.
c
n
=
_
0 if a
n
0,
a
n
if a
n
< 0.
We have
(n N)
_
0 d
n
[a
n
[
_
_
0 c
n
[a
n
[
_
So by the comparison test
n=1
d
n
and
n=1
c
n
converge. Set
n=1
d
n
= d and
n=1
c
n
= c. Then s =
n=1
a
n
= d c. Let b
1
+b
2
+b
3
+. . . is a
rearrangement of a
1
+a
2
+a
3
+. . . . Set
x
n
=
_
b
n
if b
n
0,
0 if b
n
< 0.
y
n
=
_
0 if b
n
0,
b
n
if b
n
< 0.
Then x
1
+x
2
+. . . is a rearrangement of d
1
+d
2
+. . . and from what we proved
x
1
+x
2
+ = d
1
+d
2
+. . .
Similarly
y
1
+y
2
+ = c
1
+c
2
+. . .
But b
n
= x
n
y
n
for all n. Therefore
b
1
+b
2
+ =(x
1
+x
2
+. . . ) (y
1
+y
2
+. . . )
=(d
1
+d
2
+. . . ) (c
1
+c
2
+. . . ) = u v = s.
82 January 28, 2006
CHAPTER 6. SERIES
6.1.4 Multiplication of series
Theorem 6.1.9. Let
n=1
a
n
and
n=1
b
n
be absolutely convergent series with sums s and t
respectively. Then the series a
1
b
1
+ a
2
b
1
+ a
2
b
2
+ a
1
b
3
+ a
2
b
2
+ a
3
b
1
+ . . . is absolutely
convergent with sum st.
Proof. Denote by w
1
, w
2
, w
3
, . . . products a
k
b
l
(k, l N). Let us prove that the series
n=1
[w
n
[
converges. Let S
n
denote its partial sum. S
n
contains terms [a
k
b
l
[. Denote by m the maximal
index from k and l among those that [a
k
b
l
[ are the terms in S
n
. Then
(6.1.1) S
n
([a
1
[ +[a
2
[ + +[a
m
[)([b
1
[ +[b
2
[ + +[b
m
[)
In the right-hand side of (6.1.1) there is the product of mth partial sums of the series
n=1
[a
n
[ and
n=1
[b
n
[. Therefore the sequence (S
n
)
n
is bounded above, which proves the
absolute convergence of the series in question.
It remains to prove that the sum of the series is st. Let S denote the sum of the series.
By Theorem 6.1.8 it does not depend on the order of terms. So we can write S = lim
n
S
n
.
Now notice that
S
n
2 = (a
1
+a
2
+ +a
n
)(b
1
+b
2
+ +b
n
)
which implies that
lim
n
S
n
2 = st.
But of course lim
n
S
n
2 = lim
n
S
n
which completes the proof
Example 6.1.10. Set
e(x) = 1 +
x
1!
+
x
2
2!
+ +
x
n
n!
+. . .
We know from Example 6.1.6 that the series on the right-hand side of the above equality is
absolutely convergent for all x R. Hence any two such series can be multiplied together
in the obvious way and the order of the terms can be changed without aecting the sum.
Therefore for all x, y R we have
e(x)e(y) =(1 +
x
1!
+
x
2
2!
+ +
x
n
n!
+. . . )(1 +
y
1!
+
y
2
2!
+ +
y
n
n!
+. . . )
=1 +x +y +
y
2
2
+xy +
x
2
2
+. . .
=1 +
(x +y)
1!
+
(x +y)
2
2!
+ +
(x +y)
n
n!
+ = e(x +y),
where we used the observation that the terms of degree n in x and y are
x
n
n!
+ +
x
k
k!
y
nk
(n k)!
+ +
y
n
n!
=
(x +y)
n
n!
.
83 January 28, 2006
6.2. POWER SERIES
6.2 Power series
Denition 6.2.1. A power series is a series of the form
a
0
+a
1
x +a
2
x
2
+ +a
n
x
n
+ =
n=0
a
n
x
n
,
where the coecients a
n
and the variable x are real numbers.
Example 6.2.1. The exponential series
1 +
x
1!
+
x
2
2!
+ +
x
n
n!
+. . .
converges for all x R. (see Example 6.1.10).
Example 6.2.2. The series
n=1
n
n
x
n
converges only for x = 0.
Indeed, if x ,= 0 then lim
n
n
n
x
n
,= 0.
Example 6.2.3. The geometric series
1 +x +x
2
+ +x
n
+ =
n=0
x
n
converges if and only if x (1, 1).
Example 6.2.4. The series
x
x
2
2
+
x
3
3
x
4
4
+ + (1)
n1
x
n
n
+ =
n=1
(1)
n1
x
n
n
converges if and only if x (1, 1]. (see Example 6.1.2)
Theorem 6.2.5. Let r R be such that the series
n=0
a
n
r
n
converges. Then the power series
n=0
a
n
x
n
is absolutely convergent for all x ([r[, [r[) (i.e. [x[ < [r[).
Proof. If r = 0 there is nothing to prove, so we can assume that r ,= 0.
From the convergence of the series
n=0
a
n
r
n
it follows that lim
n
a
n
r
n
= 0, therefore the
sequence (a
n
r
n
)
nN
is bounded. Thus
(K R)(n N)(a
n
r
n
K).
84 January 28, 2006
CHAPTER 6. SERIES
Let x R be such that [x[ < [r[, and set y =
[x[
[r[
. Note that 0 y < 1. We have
[a
n
x
n
[ = [a
n
[ [x[
n
= [a
n
[ [r
n
[y
n
= [a
n
r
n
[ y
n
Ky
n
.
Therefore the series
n=0
[a
n
x
n
[ converges by comparison with the convergent geometric series
K(1 +y +y
2
+ +y
n
+. . . ).
Theorem 6.2.6. Let
n=0
a
n
x
n
be a power series. Then one of the following possibilities
occurs:
(i) The series converges only when x = 0;
(ii) The series absolutely converges for all x R;
(iii) There exists r > 0 such that the series is absolutely convergent for all x R such that
[x[ < r and is divergent for all x R such that [x[ > r.
Proof. Let us denote by E the set of all non-negative numbers such that the series
n=0
a
n
x
n
converges. Clearly 0 E. If E = 0 then (i) is true.
If (x R
+
)(x E) (x ,= 0) then the series converges for all y R such that 0 y x (by
the previous theorem).
Suppose that E is not bounded above. Let y be any non-negative real number. Then y is
not an upper bound for E so that (x E)(y < x). By Theorem 6.2.5 y E. Thus E is the
set of all non-negative real numbers. Absolute convergence follows also from Theorem 6.2.5.
Finally let us suppose that E is bounded above and contains at least one positive number.
Then supE exists. Set r = supE. We know that r > 0. Let x (r, r). Then (y ([x[, r))
such that y E.
Indeed, y is not an upper bound for E. Hence (z E)(y < z). By Theorem 6.2.5 y E.
Therefore by Theorem 6.2.5 the series is absolutely convergent at x.
Now suppose that [x[ > r. Then there exists u R such that r < u < [x[ and u , E. The
series is not convergent at u, hence by Theorem 6.2.5 it is not convergent at x
Denition 6.2.2. The radius of convergence, denoted by R, of the power series
a
0
+a
1
x +a
2
x
2
+. . .
is dened as follows: R = if the series converges for all x R; R = 0 if the series converges
for x = 0 only; and R = r if (iii) in Theorem 6.2.6 holds.
Example 6.2.7. The series
1
x
2
2!
+
x
4
4!
x
6
6!
+. . .
has innite radius of convergence.
85 January 28, 2006
6.2. POWER SERIES
Proof. Let us set y = x
2
. Then the series can be written as
a
0
+a
1
y +a
2
y
2
+. . . where a
n
=
(1)
n
(2n)!
.
Using dAlemberts test we compute for any y 0
lim
n
[a
n+1
y
n+1
[
[a
n
y
n
[
= lim
n
y
(2n + 1)(2n + 2)
= 0
Therefore the series is absolutely convergent.
Example 6.2.8. The series
x
x
3
3!
+
x
5
5!
x
7
7!
+. . .
has innite radius of convergence.
The proof is similar to the previous one.
Example 6.2.9. Find all x R such that the series
1
x
3
+
x
2
5
x
3
7
+ =
n=0
(1)
n
x
n
2n + 1
is convergent.
For absolute convergence we use dAlemberts test.
lim
n
[a
n+1
x
n+1
[
[a
n
x
n
[
= lim
n
[x[(2n + 1)
2n + 3
= [x[.
Therefore the series is absolutely convergent if [x[ < 1.
The series diverges if [x[ > 1 since the general term does not converge to zero.
For x = 1 we have the alternating series
1
1
3
+
1
5
1
7
+. . .
which converges.
For x = 1 we have the series
1 +
1
3
+
1
5
+
1
7
+ =
n=0
1
2n + 1
,
which diverges by comparison with the harmonic series.
86 January 28, 2006
Chapter 7
Elementary functions
7.1 Exponential function
Denition 7.1.1. We dene the exponential function exp : R R by the following series
(7.1.1) expx = 1 +
x
1!
+
x
2
2!
+ +
x
n
n!
+. . .
The series on the right is convergent for all x R. (see Example 6.2.1).
Theorem 7.1.1. For all x, y R
expx expy = exp(x +y).
The proof was give in Example 6.1.10.
The following properties easily follow from the denition and the above theorem.
Corollary 7.1.1. (i) exp0 = 1;
(ii) exp(x) =
1
expx
;
(iii) (x R) (expx > 0).
Theorem 7.1.2. The exponential function f(x) = expx is everywhere dierentiable and
f
= expx.
We will also use the notation e
x
= expx for the reason which will become apparent a little
later.
Proof. Using e
x+y
= e
x
e
y
we have
e
a+h
e
a
h
= e
a
e
h
1
h
.
Now
e
h
1
h
= 1 +
h
2!
+
h
2
3!
+ = 1 +g(h).
87
7.1. EXPONENTIAL FUNCTION
If [h[ < 2 we have
[g(h)[
[h[
2
+
[h[
2
4
+ +
[h[
n
2
n
+ =
[
1
2
h[
1 [
1
2
h[
0 as h 0.
Corollary 7.1.2. expx is a continuous function.
Theorem 7.1.3. (i) expx is a strictly increasing function.
(ii) expx + as x +.
(iii) expx 0 as x .
(iv) Ran(exp) = (0, +).
(v) (k N)
_
lim
x+
expx
x
k
= +
.
Denition 7.1.2. e = exp1.
This number, namely
e = 1 +
1
1!
+
1
2!
+ +
1
n!
+. . .
is one of the fundamental constants in mathematics. Its approximate value is
2.718281828459045
It is not dicult to prove that e is irrational.
Note that from the denition and from Theorem 7.1.1 it follows that
(n N) (expn = e
n
).
Theorem 7.1.4. Let r Q. Then
expr = e
r
.
Proof. Let n N and r = n. Then
exp(n) =
1
expn
=
1
e
n
= e
n
.
Let r = p/q, p, q N. Then
exp(p/q)
q
= exp(
p
q
q) = expp = e
p
.
Hence
exp(p/q) = e
p/q
.
Denition 7.1.3. If x is irrational we set
e
x
= expx.
88 January 28, 2006
CHAPTER 7. ELEMENTARY FUNCTIONS
Due to monotonicity and continuity of the function exp() we obtain
e
x
= supe
p
[ p is rational and p < x.
Remark 11. It is useful to redene the function e
x
(after we have studied the properties
above) as a mapping from R to (0, ).
So, by the exponential function we mean the mapping f : R (0, ) dened by f(x) = e
x
.
The above properties show in particular that this function is a bijection.
7.2 Logarithmic function
For the bijection dened in the previous section there exists the inverse function.
Denition 7.2.1. The function from (0, ) to R dened as the inverse to e
x
is called the
logarithmic e function.
In other words, for x > 0 y = log x if e
y
= x.
Theorem 7.2.1. log : (0, ) R is a dierentiable (and therefore continuous) increasing
function satisfying the following properties
(i) (x, y (0, ))
_
log(xy) = log x + log y
;
(ii) (log x)
=
1
x
;
(iii) lim
x
log x = ;
(iv) lim
x0
log x = ;
(v) (k N)
_
lim
x
log x
x
k
= 0
_
.
Proof. The proofs are straightforward from the properties of exp. Let us show (i).
exp(log(xy)) = xy = exp(log x) exp(log y) = exp(log x + log y).
Now (i) follows from the injectivity of exp.
Let y = f(x) = log x. Then x = g(y) = expy. We have
f
(x) =
1
g
(y)
=
1
expy
=
1
x
.
The rest is left as an exercise.
There is a simple representation of log(1 +x) as a power series.
Theorem 7.2.2. For x (1, 1)
log(1 +x) = x
x
2
2
+
x
3
3
x
4
4
+ + (1)
n1
x
n
n
+. . . .
89 January 28, 2006
7.3. TRIGONOMETRIC FUNCTIONS
Now we are able to dene arbitrary powers of nonnegative numbers (until now it was only
clear how to dene rational powers).
Denition 7.2.2. Let a > 0. Then for any x R we set
a
x
:= e
xlog a
.
The standard laws
a
x
a
y
= a
x+y
= a
x+y
, (a
x
)
y
= a
xy
valid for a > 0 and arbitrary x, y R, can be easily veried from the denition and discussed
properties of the exponential and logarithmic functions.
7.3 Trigonometric functions
In this section we sketch an approach to dene trigonometric functions.
Denition 7.3.1. We dene the two main trigonometric functions sin : R R and cos :
R R by the following series
(7.3.2) sinx := x
x
3
3!
+
x
5
5!
x
7
7!
+. . .
(7.3.3) cos x := 1
x
2
2!
+
x
4
4!
x
6
6!
+. . .
The series are absolutely convergent for all x R (see Example 6.2.8 and Example 6.2.7).
The next theorem can be proved using the above denitions by multiplying and adding
the corresponding series. We skip the details of the proof. An easier proof would be to use
the exponential function with complex variables. This will be done in the course Further
topics in Analysis.
Theorem 7.3.1. For all x, y R
sin(x +y) = sinxcos y + cos xsiny; (7.3.4)
sin
2
x + cos
2
x = 1. (7.3.5)
From the above it follows that
(x R)
_
(1 sinx 1) (1 cos x 1)
.
Theorem 7.3.2. The functions sin : R R and cos : R R are everywhere dierentiable
and
(sinx)
= cos x, (cos x)
= sinx.
The proof is similar to that of Theorem 7.1.2. We do not give the details.
90 January 28, 2006
CHAPTER 7. ELEMENTARY FUNCTIONS
Periodicity of trigonometric functions
Theorem 7.3.3. There exists a smalest positive constant
1
2
such that
cos
1
2
= 0 and sin
1
2
= 1.
Proof. If 0 < x < 2 then
sinx =
_
x
x
3
3!
_
+
_
x
5
5!
x
7
7!
_
+ > 0
Therefore for 0 < x < 2 we have that
(cos x)
= sinx < 0,
hence cos x is decreasing.
Notice that
cos x =
_
1
x
2
2!
_
+
_
x
4
4!
x
6
6!
_
+. . .
It is easy to see that cos
2 > 0.
Another rearrangement
cos x = 1
x
2
2!
+
x
4
4!
_
x
6
6!
x
8
8!
_
. . .
shows that cos
3 < 0.
This proves the existence of the smalest number
2 <
1
2
<
i=1
m
i
(x
i
x
i1
).
92
CHAPTER 8. THE RIEMANN INTEGRAL
The upper sum of f for P is dened as
U(f, P) =
n
i=1
M
i
(x
i
x
i1
).
Denition 8.1.3. The integral sum of f for the partition P and
i
[x
i1
, x
i
] [ (x
i
)
n
i=0
P is
(f, P, ) =
n
i=1
f(
i
)(x
i
x
i1
).
Denote
m := inf
x[a,b]
f(x), M := sup
x[a,b]
f(x).
The following inequality is obviously true
(i 1, 2, . . . , n)
_
m m
i
f(
i
) M
i
M
.
Therefore we have
m(b a) =
n
i=1
m(x
i
x
i1
)
n
i=1
m
i
(x
i
x
i1
)
n
i=1
f(
i
)(x
i
x
i1
)
i=1
M
i
(x
i
x
i1
)
n
i=1
M(x
i
x
i1
) = M(b a),
which implies that for every partition and for every choice
i
[x
i1
, x
i
] [ (x
i
)
n
i=0
P the following inequality holds
(8.1.1) m(b a) L(f, P) (f, P, ) U(f, P) M(b a).
In other words, the sets of real numbers
L(f, P) [ P, U(f, P) [ P
are bounded.
Denition 8.1.4. The upper integral of f over [a, b] is
J := infU(f, P) [ P,
the lower integral of f over [a, b] is
j := supL(f, P) [ P.
(The inmum and the supremum are taken over all partitions on [a, b].)
Denition 8.1.5. A function f : [a, b] R is called Riemann integrable if
J = j.
The common value is called integral of f over [a, b] and is denoted by
_
b
a
f(x)dx.
93 January 28, 2006
8.1. DEFINITION OF INTEGRAL
8.1.2 Properties of upper and lower sums
Lemma 2. Let P and Q be two partitions of [a, b] such that P Q. Then
L(f, P) L(f, Q),
U(f, P) U(f, Q).
(The partition Q is called a renement of P.)
Proof. First let us consider a particular case. Let P
k
= inf
x[x
k1
,c]
f(x), m
k
= inf
x[c,x
k
]
f(x).
Then m
k
m
k
, m
k
m
k
, and we have
L(f, P
) =
k1
i=1
m
i
(x
i
x
i1
) +m
k
(c x
k1
) +m
k
(x
k
c) +
n
i=k+1
m
i
(x
i
x
i1
)
k1
i=1
m
i
(x
i
x
i1
) +m
k
(x
k
x
k1
) +
n
i=k+1
m
i
(x
i
x
i1
) = L(f, P).
Similarly one obtains that
U(f, P
) U(f, P).
Now to prove the assertion one has to add to P consequently a nite number of points in
order to form Q.
Proposition 8.1.1. Let P and Q be arbitrary partitions of [a, b]. Then
L(f, P) U(f, Q).
Proof. Consider the partition P Q. The by Lemma 2 we have
L(f, P) L(f, P Q) U(f, P Q) U(f, Q).
Theorem 8.1.1. J j.
Proof. Fix a partition Q. Then by Proposition 8.1.1
(P) L(f, P) U(f, Q).
Therefore
j = supL(f, P) [ P U(f, Q).
And from the above
(Q)j U(f, Q).
Hence
j infU(f, Q) [ Q = J.
Remark. Integrability of f means by denition j = J. If j < J we say that f is not
Riemann integrable.
94 January 28, 2006
CHAPTER 8. THE RIEMANN INTEGRAL
Example 8.1.2. Let f : [a, b] R is dened by f(x) = C. Then
(P)
_
L(f, P) = U(f, P) = C(b a)
_
.
Hence
J = j = C(b a).
Example 8.1.3. The Dirichlet function D : [0, 1] R is dened as D(x) = 1 if x is rational
and D(x) = 0 is x is irrational.
Then
(P)
_
L(D, P) = 0
_
_
U(D, P) = 1
_
.
Hence the Dirichlet function is not Riemann integrable.
8.2 Criterion of integrability
Theorem 8.2.1. A function f : [a, b] R is Riemann integrable if and only if for any > 0
there exists a partition P of [a, b] such that
U(f, P) L(f, P) < .
Proof. 1) Necessity. Let J = j, i.e. let us assume that f is integrable. Fix > 0. Then
(P
1
)
_
L(f, P
1
) > j /2
_
.
Also
(P
2
)
_
U(f, 2
1
) < J +/2
_
.
Let Q = P
1
P
2
. Then
j /2 < L(f, P
1
) L(f, Q) U(f, Q) U(f, P
2
) < J +/2.
Therefore (since J = j)
U(f, Q) L(f, Q) < .
2) Suciency. Fix > 0. Let P be a partition such that
U(f, P) L(f, P) < .
Note that
J j U(f, P) L(f, P) < .
Therefore it follows that
( > 0)
_
J j <
_
.
This implies that J = j.
95 January 28, 2006
8.3. INTEGRABLE FUNCTIONS
8.3 Integrable functions
The following denition is used in the proof of the next theorems and will be also used in the
subsequent sections.
Denition 8.3.1. Let P be a partition of [a, b]. The length of the greatest subinterval of
[a, b] under the partition P is called the norm of the partition P and is denoted by |P|, i.e.
|P| := max
1in
(x
i
x
i1
).
Theorem 8.3.1. Let f : [a, b] R be monotone. Then f is Riemann integrable.
Proof. Without loss of generality assume that f is increasing, so that f(a) < f(b). Fix > 0.
Let us consider a partition P of [a, b] such that
|P| < =
f(b) f(a)
.
For this partition we obtain
U(f, P) L(f, P) =
n
i=1
(M
i
m
i
)(x
i
x
i1
)
=
n
i=1
(f(x
i
) f(x
i1
))(x
i
x
i1
) <
n
i=1
(f(x
i
) f(x
i1
))
= (f(b) f(a)) = .
Theorem 8.3.2. Let f : [a, b] R be continuous. Then f is Riemann integrable.
Proof. Fix > 0. Since f is continuous on a closed interval, it is uniformly continuous (See
section 4.4). Therefore for
ba
there exists > 0 such that
(x
1
, x
2
[a, b])
_
([x
1
x
2
[ < ) ([f(x
1
) f(x
2
)[ <
b a
.
Hence for every partition P with norm |P| < we have
U(f, P) L(f, P) =
n
i=1
(M
i
m
i
)(x
i
x
i1
) <
b a
n
i=1
(x
i
x
i1
) = .
96 January 28, 2006
CHAPTER 8. THE RIEMANN INTEGRAL
8.4 Elementary properties of integral
Theorem 8.4.1. Let a < c < b. Let f be integrable on [a, b]. Then f is integrable on [a, c]
and on [c, b] and
_
b
a
f(x)dx =
_
c
a
f(x)dx +
_
b
c
f(x)dx.
Conversely, if f is integrable on [a, c] and on [c, b] then it is integrable on [a, b].
Proof. Suppose that f is integrable on [a, b]. Fix > 0. Then there exists a partition
P = x
0
, . . . , x
n
of [a, b] such that
U(f, P) L(f, P) < .
We can assume that c P so that c = x
j
for some j 0, 1, . . . , n (otherwise consider
the renement of P adding the point c). Then P
1
= x
0
, . . . , x
j
is a partition of [a, c] and
P
2
= x
j
, . . . , x
n
is a partition of [c, b]. Moreover,
L(f, P) = L(f, P
1
) +L(f, P
2
), U(f, P) = U(f, P
1
) +U(f, P
2
).
Therefore we have
[U(f, P
1
) L(f, P
1
)] + [U(f, P
2
) L(f, P
2
)] = U(f, P) L(f, P) < .
Since each of the terms on the left hand side is non-negative, each one is less than , which
proves that f is integrable on [a, c] and on [c, b]. Note also that
L(f, P
1
)
_
c
a
f(x)dx U(f, P
1
),
L(f, P
2
)
_
b
c
f(x)dx U(f, P
2
),
so that
L(f, P)
_
c
a
f(x)dx +
_
b
c
f(x)dx U(f, P).
This is true for any partition of [a, b]. Therefore
_
b
a
f(x)dx =
_
c
a
f(x)dx +
_
b
c
f(x)dx.
Now suppose that f is integrable on [a, c] and on [c, b]. Fix > 0. Then there exists a
partition P
1
of [a, c] such that
U(f, P
1
) L(f, P
1
) < /2.
Also there exists a partition P
2
of [c, b] such that
U(f, P
2
) L(f, P
2
) < /2.
Let P = P
1
P
2
. Then
U(f, P) L(f, P) = [U(f, P
1
) L(f, P
1
)] + [U(f, P
2
) L(f, P
2
)] < .
97 January 28, 2006
8.4. ELEMENTARY PROPERTIES OF INTEGRAL
The integral
_
b
a
f(x)dx was dened only for a < b. We add by denition that
_
a
a
f(x)dx = 0 and
_
b
a
f(x)dx =
_
a
b
f(x)dx if a > b.
With this convention we always have that
_
b
a
f(x)dx =
_
c
a
f(x)dx +
_
b
c
f(x)dx.
Theorem 8.4.2. Let f and g be integrable on [a, b]. Then f + g is also integrable on [a, b]
and
_
b
a
[f(x) +g(x)]dx =
_
b
a
f(x)dx +
_
b
a
g(x)dx.
Proof. Let P = x
0
, . . . , x
n
be a partition of [a, b]. Let
m
i
=inff(x) +g(x) [ x
i1
x x
i
,
m
i
=inff(x) [ x
i1
x x
i
,
m
i
=infg(x) [ x
i1
x x
i
.
Dene M
i
, M
i
, M
i
similarly. The following inequalities hold
m
i
m
i
+m
i
,
M
i
M
i
+M
i
.
Therefore we have
L(f, P) +L(g, P) L(f +g, P),
U(f +g, P) U(f, P) +U(g, P).
Hence for any partition P
L(f, P) +L(g, P) L(f +g, P) U(f +g, P) U(f, P) +U(g, P),
or otherwise
U(f +g, P) L(f +g, P) [U(f, P) L(f, P)] + [U(g, P) L(g, P)].
Fix > 0. Since f and g are integrable there are partitions P
1
and P
2
such that
U(f, P
1
) L(f, P
1
) < /2,
U(g, P
2
) L(g, P
2
) < /2.
Thus for the partition P = P
1
P
2
we obtain that
U(f +g, P) L(f +g, P) < .
This proves that f +g is integrable on [a, b].
98 January 28, 2006
CHAPTER 8. THE RIEMANN INTEGRAL
Moreover,
L(f, P) +L(g, P) L(f +g, P)
_
b
a
[f(x) +g(x)]dx
U(f +g, P) U(f, P) +U(g, P),
and
L(f, P) +L(g, P)
_
b
a
f(x)dx +
_
b
a
g(x)dx U(f, P) +U(g, P).
Therefore it follows that
_
b
a
[f(x) +g(x)]dx =
_
b
a
f(x)dx +
_
b
a
g(x)dx.
Theorem 8.4.3. Let f be integrable on [a, b]. Then, for any c R, cf is also integrable on
[a, b] and
_
b
a
cf(x)dx = c
_
b
a
f(x)dx.
Proof. The proof is left as an exercise. Consider separately two cases: c 0 and c 0.
Theorem 8.4.4. Let f, g be integrable on [a, b] and
(x [a, b]) (f(x) g(x)).
Then
_
b
a
f(x)dx
_
b
a
g(x)dx.
Proof. For any partition P of [a, b] we have
L(f, P) L(g, P)
_
b
a
g(x)dx.
The assertion follows by taking supremum over all partitions.
Corollary 8.4.1. Let f be integrable on [a, b] and there are M, m R such that
(x [a, b]) (m f(x) M).
Then
m(b a)
_
b
a
f(x)dx M(b a).
Corollary 8.4.2. Let f be continuous on [a, b]. Then there exists [a, b] such that
_
b
a
f(x)dx = f()(b a).
99 January 28, 2006
8.4. ELEMENTARY PROPERTIES OF INTEGRAL
Proof. From Corollary 8.4.1 it follows that
m
1
b a
_
b
a
f(x)dx M,
where m = min
[a,b]
f(x), M = max
[a,b]
f(x). Then by the Intermediate Value Theorem we
conclude that there exists [a, b] such that
f() =
1
b a
_
b
a
f(x)dx.
Theorem 8.4.5. Let f be integrable on [a, b]. Then [f[ is integrable on [a, b] and
_
b
a
f(x)dx
_
b
a
[f(x)[dx.
Proof. Note that for any interval [, ]
(8.4.2) sup
[,]
[f(x)[ inf
[,]
[f(x)[ sup
[,]
f(x) inf
[,]
f(x).
Indeed,
(x, y [, ])
_
f(x) f(y) sup
[,]
f(x) inf
[,]
f(x)
_
, so that
(x, y [, ])
_
[f(x)[ [f(y)[ sup
[,]
f(x) inf
[,]
f(x)
_
,
which proves (8.4.2) by passing to the supremum in x and y.
It follows from (8.4.2) that for any partition of [a, b]
U([f[, P) L([f[, P) U(f, P) L(f, P),
which proves the integrability of [f[ by the criterion of integrability, Theorem 8.2.1. The last
assertion follows from Theorem 8.4.4.
Theorem 8.4.6.
1
Let f : [a, b] R be integrable and (x [a, b])(m f(x) M). Let g : [m, M]
R be continuous. Then h : [a, b] R dened by h(x) = g(f(x)) is integrable.
Proof. Fix > 0. Since g is uniformly continuous on [m, M], there exists > 0 such that < and
(t, s [m, M])
_
([t s[ < ) ([g(t) g(s)[ < )
.
By integrability of f there exists a partition P = x
0
, . . . , x
n
of [a, b] such that
(8.4.3) U(f, P) L(f, P) <
2
.
Let m
i
= inf
[x
i1
,x
i
]
f(x), M
i
= sup
[x
i1
,x
i
]
f(x) and m
i
= inf
[x
i1
,x
i
]
h(x), M
i
= sup
[x
i1
,x
i
]
h(x).
Decompose the set 1, . . . , n into two subset : (i A) (M
i
m
i
< ) and (i B) (M
i
m
i
).
For i A by the choice of we have that M
i
m
i
< .
1
This theorem is is outside the syllabus and will not be included in the examination paper
100 January 28, 2006
CHAPTER 8. THE RIEMANN INTEGRAL
For i B we have that M
i
m
i
2K where K = sup
t[m,M]
[g(t)[. By (8.4.3) we have
iB
(x
i
x
i1
)
iB
(M
i
m
i
)(x
i
x
i1
) <
2
,
so that
iB
(x
i
x
i1
) < . Therefore
U(h, P) L(h, P) =
iA
(M
i
m
i
)(x
i
x
i1
) +
iB
(M
i
m
i
)(x
i
x
i1
)
<(b a) + 2K < [(b a) + 2K],
which proves the assertion since is arbitrary.
Corollary 8.4.3. Let f, g be integrable on [a, b]. Then the product fg is integrable on [a, b].
Proof. Since f +g and f g are integrable on [a, b], (f +g)
2
and (f g)
2
are integrable on [a, b] by
the previous theorem. Therefore
fg =
1
4
[(f +g)
2
(f g)
2
] is integrable on [a, b].
101 January 28, 2006
8.5. INTEGRATION AS THE INVERSE TO DIFFERENTIATION
8.5 Integration as the inverse to dierentiation
Theorem 8.5.1. Let f be integrable on [a, b] and let F be dened on [a, b] by
F(x) =
_
x
a
f(t)dt.
Then F is continuous on [a, b].
Proof. By the denition of integrability f is bounded on [a, b]. Let M = sup
[a,b]
[f(x)[. Then
for x, y [a, b] we have
F(x) F(y)
_
y
x
f(t)dt
M[x y[,
which proves that F is uniformly continuous on [a, b].
If in the previous theorem we in addition assume that f is continuous, we can prove more.
Theorem 8.5.2. Let f be integrable on [a, b] and let F be dened on [a, b] by
F(x) =
_
x
a
f(t)dt.
Let f be continuous at c [a, b]. Then F is dierentiable at c and
F
(c) = f(c).
Proof. Let c (a, b). Let h > 0. Then
F(c +h) F(c)
h
=
1
h
_
c+h
c
f(t)dt.
By Corollary 8.4 there exists [c, c +h] such that
_
c+h
c
f(t)dt = f()h.
Hence we have
F(c +h) F(c)
h
= f().
As h 0, c, and due to continuity of f we conclude that lim
h0
f() = f(c). The
assertion follows. The case h < 0 is similar. The cases c = a and c = b are similar (In these
cases one talks on one-sided derivatives only.)
Theorem 8.5.3. Let f be continuous on [a, b] and f = g
= (F g)
= 0.
Therefore by Corollary 5.2.2 there is a number c such that
F = g +c.
Since F(a) = 0 we have that g(a) = c. Thus for x [a, b]
_
x
a
f(t)dt = F(x) = g(x) g(a).
The next theorem is often called The Fundamental Theorem of Calculus.
Theorem 8.5.4. Let f be integrable on [a, b] and f = g
(t
i
)(x
i
x
i1
) = f(t
i
)(x
i
x
i1
).
Let
m
i
= inf
[x
i1
,x
i
]
f(x), M
i
= sup
[x
i1
,x
i
]
f(x).
Then
m
i
((x
i
x
i1
) f(t
i
)(x
i
x
i1
) M
i
(x
i
x
i1
), that is
m
i
((x
i
x
i1
) g(x
i
) g(x
i1
) M
i
(x
i
x
i1
).
Adding these inequalities for i = 1, . . . , n we obtain
n
i=1
m
i
((x
i
x
i1
) g(b) g(a)
n
i=1
M
i
(x
i
x
i1
),
so that for any partition we have
L(f, P) g(b) g(a) U(f, P),
which means that
g(b) g(a) =
_
b
a
f(x)dx.
103 January 28, 2006
8.6. INTEGRAL AS THE LIMIT OF INTEGRAL SUMS
8.6 Integral as the limit of integral sums
2
Let f : [a, b] R. We dened the integral sums of f in Denition 8.1.3.
Denition 8.6.1. A number A is called the limit of integral sums (f, P, ) if
( > 0)( > 0)(P)((
i
) P)
_
(|P| < )
_
[(f, P, ) A[ <
_
_
.
In this case we write
lim
P0
(f, P, ) = A.
The next theorem shows that the Riemann integral can be equavalently dened via the limit of
the integral sums.
Theorem 8.6.1. Let f : [a, b] R. The f is Riemann integrable if and only if lim
P0
(f, P, )
exists. In this case
lim
P0
(f, P, ) =
_
b
a
f(x)dx.
Proof. First, assume that f is Riemann integrable. Then we know that f is bounded, so that there
is a constant C such that [f(x)[ C for all x [a, b]. Fix > 0. Then there exists a partition P
0
of
[a, b] such that
U(f, P
0
) L(f, P
)
< /2.
Let m be the number of points in the partition P
0
. Choose =
8mC
. Then for any partition P
1
such
that |P
1
| < and P = P
0
P
1
we have
U(f, P
1
) = U(f, P) +
_
U(f, P
1
) U(f, P)
_
U(f, P
0
) +
_
U(f, P
1
) U(f, P)
_
U(f, P
0
) + 2C|P
1
|m < U(f, P
0
) +/4.
Similarly,
L(f, P
1
) > L(f, P
0
) /4.
Therefore we get
L(f, P
0
) /4 < L(f, P
1
) U(f, P
1
) < U(f, P
0
) +/4.
Hence
U(f, P
1
) L(f, P
1
) < ,
which together with the inequalities
L(f, P
1
)
_
b
a
f(x)dx U(f, P
1
),
L(f, P
1
) (f, P
1
, ) U(f, P
1
)
leads to
_
b
a
f(x)dx (f, P
1
, )
< .
Now suppose that lim
P0
(f, P, ) = A. Fix > 0. Then there exists > 0 such that if |P| <
then
A/2 < (f, P, ) < A+/2.
2
The entire section is not included in the examination paper
104 January 28, 2006
CHAPTER 8. THE RIEMANN INTEGRAL
Choose P as above. Varying (
i
) take sup and inf of (f, P, ) in the above inequality. We obtain
A
a
f(x)dx is dened as
_
a
f(x)dx = lim
A
_
A
a
f(x)dx.
We use the same terminology for improper integrals as for series, that is, an improper
integral may converge or diverge.
Example 8.7.1.
_
1
1
x
2
dx = 1.
Indeed,
_
A
1
1
x
2
dx = 1
1
A
1 as A .
Theorem 8.7.2.
_
1
1
x
k
dx converges if and only is k > 1.
Integrals of unbounded functions
Example 8.7.3.
_
1
0
dx
x
= lim
0
_
1
dx
x
= lim
0
(2 2
) = 2.
The notion of the improper integral is useful for investigation of convergence of certain
series. The following theorem is often called the integral test for convergence of series.
Theorem 8.7.4. Let f : [1, ) R be positive and increasing. Then the integral
_
1
f(x)dx
and the series
n=1
f(n) both converge or both diverge.
Proof. Since f is monotone it is integrable on any nite interval (Theorem 8.3.1). For n1
x n we have
f(n) f(x) f(n 1).
Integrating the above inequality from n 1 to n we obtain
f(n)
_
n
n1
f(x)dx f(n 1).
105 January 28, 2006
8.8. CONSTANT
Adding these inequalities on interrvals [1, 2], [2, 3], . . . , [n 1, n] we have
n
k=2
f(k)
_
n
1
f(x)dx
n1
k=1
f(k).
Now the assertion easily follows.
As an application of the above theorem we consider the following
Example 8.7.5. The series
n=1
1
n
converges if > 1.
Proof. By Thoerem 8.7.4 it is enough to prove that
_
1
dx
x
< .
Indeed,
_
1
dx
x
= lim
A
_
A
1
dx
x
=
1
1
_
1
1
A
1
_
=
1
1
< .
Example 8.7.6. The series
n=1
1
n(log n)