2 Basic facts about functions and sequences

2.1 Functions

Our main tool to study real numbers, and our primary application
of these numbers, are functions. Before giving a formal definition of
functions, we give an intuitive one. If you give a function a number, it
gives you back a number.

Examples. The following are some examples of functions.

1. The function that always gives you 21.

2. The function that gives you the number you gave it.

3. The function that squares the number you gave it.

4. The function that gives you 1 if you gave it 10, and gives you 2
otherwise.

Describing functions with words is usually not the most efficient

way to express them.

Examples. The same functions written more efficiently.

1. f ( x ) = 21

2. f ( x ) = x

3. f ( x ) = x2
8
<1 if x = 10
4. f ( x ) =
:2 if x 6= 10

Also, we can draw graphs that express these functions.

Polynomial functions
One type of function we will meet regularly is the polynomial function
(three of the four examples above are polynomials). A polynomial is
an expression
n 1
an x n + an 1x + · · · + a2 x2 + a1 x + a0 where ai 2 R and n 2 N.
6

All polynomials are functions. We call the numbers ai the coefficients

of the polynomials. Note that that the coefficients can be any real
number, not only integers. For example;
7 4 p
f (x) = x + 2x + p
3
is a polynomial function.

Domains and natural domains

The proper notation for a function is as follows:

f :D ! R
x 7! f (x)

Here D is a subset of R where f ( x ) makes sense for all x 2 R. This

set D is called the domain of f . The codomain is the set of numbers
that the function can give back. In our class, we can always take the
codomain to be R, but we may choose to be more specific.
Here is an example:

f : [ 1, 1] ! R
1
x 7!
x2 +1
So in this example, although it is perfectly ok to evaluate the quantity
1
x 2 +1
when x = 2, the quantity f (2) is not defined. If you change the
domain of the function, you obtain a new function.
In the first examples, we didn’t define a domain. When the do-
main is not specified, we will use the convention that we are using
the natural domain. The natural domain of a function f is the subset
of all x 2 R for which f ( x ) is well-defined. If needed, we will write
dom( f ) for the natural domain of a function f . Here are some new
examples and their natural domains.

Examples. The same functions written more efficiently.

1. f ( x ) = x2 3x + 9, dom( f ) = R.

2. f ( x ) = 1x , dom( f ) = R ⇤ .
p
3. f ( x ) = x2 1, dom( f ) = R \] 1, 1[.

The first example is a polynomial - these are always defined over

all of R. The second example is a rational function - these are defined
as long as the denominator is not 0. The third example involves
a square root - the square root function is not defined in negative
numbers so we have to exclude the values of x for which x2 1 < 0,
that is when x 2] 1, 1[.
 analysis for applications 7

The codomain of a function is the set of allowed values for the func-
tion. For instance, for a real function, the codomain can always be
taken to be R. But we could choose to be more precise. For instance,
the function sin (that we will see below) only takes values between
1 and 1 so [ 1, 1] is a more precise codomain. More generally, the
image of a function f : D ! R is the set of values it takes. We denote
it f ( D ). More formally it is defined as folows:

f ( D ) = { y 2 R | y = f ( x ), x 2 D }.

Trigonometry
We will define the trigonometric functions in two related ways.
Consider the right-angled triangle with hypotenuse equal to 1;
We can define the Sine and Cosine functions in relation to this
triangle.

sin :]0, p/2[ !]0, 1[,

q 7! Opposite.

cos :]0, p/2[ !]0, 1[,

q 7! Adjacent.

The domain is ]0, p/2[ as this is required for the triangle to have a
right-angle. If q is very close to zero then the adjacent side is very
close to length 1, and if q is very close to p/2 then the adjacent side
has length very close to 0. It follows that the codomains are ]0, 1[, as
above.

The circle
However, we can extend the domains of Sine and Cosine functions to
all of R. Consider the unit circle.
Let Pq be the point of intersection of the circle and a line starting
at the origin with counter-clockwise angle q from the x-axis. We
redefine the sine and cosine functions.

sin : R ! [ 1, 1],
q 7! y co-ordinate of Pq .

cos : R ! [ 1, 1],
q 7! x co-ordinate of Pq .
8

Combining functions
As with numbers, we can combine two functions in multiple ways to
create a new ones. Let f and g be two functions.

( f + g)( x ) = f ( x ) + g( x ) [Both f and g must have the same domain]

(f g)( x ) = f ( x ) g( x ) [Both f and g must have the same domain]
( f g)( x ) = f ( x ) g( x ) [Both f and g must have the same domain]
f (x)
( f /g)( x ) = [Both f and g must have the same domain,
g( x )
and the image of g must not contain zero]
(f g)( x ) = f ( g( x )) [The domain of f must contain the image of g]

If you are given the domains of f and g, so for instance if f : D f !

R, and g : Dg ! R, then the natural domain of one of the combined
function is D f \ Dg (from which we have removed any problematic
points if necessary for the quotient or the composition).

Examples. Let f ( x ) = x2 and g( x ) = x 1

1. ( f + g)( x ) = f ( x ) + g( x ) = x2 + x 1

2. ( f g)( x ) = f ( x ) + g( x ) = x2 x+1

3. ( f g)( x ) = f ( x ) g( x ) = x2 ( x 1) = x 3 x

4. ( f /g)( x ) = f ( x )/g( x ) = x2 /( x 1)

5. ( f g)( x ) = f ( g( x )) = f ( x 1) = ( x 1)2 = x 2 2x + 1

Finally, we can complete our discussion on trigonometric functions

by defining the tangent function.
⇣ sin ⌘
tan(q ) = ( q ).
cos
Going back to our triangle definitions, we can divide the length of
each side by cos(q ), arriving at a triangle whose adjacent side is
cos(q )/ cos(q ) = 1 and opposite side is sin(q )/ cos(q ) = tan(q ). This
then can act as a geometric definition for the tangent function. It is
the the length of the opposite side of a right-angled triangle with
adjacent side length 1 and angle q.
As q gets closer to p/2, this value gets extremely large, and when
q = p/2, the definition doesn’t make sense. This is compatible with
our original definition of tan which is not defined when cos(q ) = 0,
that is, when q = kp/2, for some k 2 Z. In short, the domain of the
tan function is
p
R \ { + kp | k 2 Z }.
2
 analysis for applications 9

This behavior of tan at these special points is beyond our under-

standing at the minute, except to say that it gets very large near p/2,
p/2 etc. To talk about this more accurately we will need to define
limits, which is the topic of the next few classes.

2.2 Sequences

A sequence is a collection of real numbers indexed by the natural

numbers
a0 , a1 , a2 , a3 , · · · 2 R
As there are infinitely many natural numbers, each sequence must
have infinitely many numbers.
Perhaps a more precise way of defining a sequence is as follows.

A sequence is a function f : N ! R.

But sometimes we won’t want the sequence to start for n = 0. In

general we write a sequence as

{ ak }•
k=k0

where k0 2 Z is the first index. Sometimes we write { ak }k2N if the

indices begin at 0 and { ak }k2N⇤ if the indices begin at 1.

Example 2.2.1. Here are some examples of sequences:

1. 1, 2, 3, 4, . . .

2. 2, 4, 6, 8, . . .

3. 0,1, 4, 9, 16, . . .

4. 2, 3/2, 4/3, 5/4, 6/5, . . .

5. 2, 0, 2, 0, 2, 0, 2, 0, 2, . . .

Each of these sequence follows a pattern and sometimes we use

can write formulas for these patterns.

Example 2.2.2. Formulas corresponding to the above sequences.

1. an = n or {n}•
n =1

2. an = 2n or {2n}•
n =1

3. an = n2 or { n2 } •
n =0

n+1 n n + 1 o•
4. an = or
n n n =1

5. an = 1 + ( 1) n
or {1 + ( 1) n } •
n =0
10

A few simple observations about these sequences are; as n gets big-

ger (1), (2), and (3) also get bigger. As n gets bigger (4) gets smaller,
and seems to get closer to 1. Finally, as n gets bigger, (5) alternates
between 0 and 2.
Another way to define sequences is through a recursive formula.
This uses a rule (called a recurrence relation) to construct each each
element in terms of the ones before it. This requires knowing an
initial (or several) value(s), and then applying the rule.

Example 2.2.3. The Fibonnaci sequence. The first elements of this

sequence are a0 = 1 and a1 = 1. The recurrence relation is given by

an = an 1 + an 2

The sequence becomes 1, 1, 2, 3, 5, 8, . . ..

Converging sequences
Definition 2.2.4. A sequence { an }•
n=1 converges to a real number
A 2 R when for each e > 0 there exists N > 0 such that for all n N
| an A| < e.

The idea of convergence should feel quite natural. Intuitively, we

would say that a sequence converges to a number A if all the terms
in the sequence get progressively closer to A. This is exactly what
the definition says, but it may be hard to see that at first. So let’s go
through each part more carefully.
The definition of convergence begins with “for each e > 0”. This
means that whatever follows should be true for any possible positive
value of e imaginable. The next part is “there exists N > 0”. Here, all
we need is some value of N, but remember; we need such a value for
all possible e. In other words, we have no control over the value of
e, but nevertheless, we need to find an N that satisfies the rest of the
definition.
Now, the value | an A| is the difference (or distance) between an
and A. The last part of the definition therefore says that this distance
must be less that e if n N. That is, at some point in our sequence
we reach a number a N , and every term after a N has distance less than
e from A.
The crucial part to understand is that as the statement must be
true for all e > 0, it is necessarily true for extremely small values of e.
In this sense, we like to think of e as being a tiny distance, and so the
definition says that the sequence { an }• n=1 converges to A if all terms
after a N are extremely close (within e) to A.

Example 2.2.5. The sequence with terms an = 1 + 1/n converges to 1.

 analysis for applications 11

2.pdf 2.pdf

Before giving the proof, lets see what happens if we choose e =

1/100. If we believe the statement, then we should be able to find a
value of N such that | an 1| < 1/100 for all n N. But | an 1| =
|1 + 1/n 1| = |1/n| = 1/n, so we need to find a value of N such
that 1/n < 1/100 for all n N. Equivalently, we need to find a value
of N such that if n N then n > 100. Suddenly, this is not a hard
problem to solve. For example, N = 101 would be a good choice (in
fact any number bigger than 100).
To summarize, if e = 1/100 then choose N to be a number bigger
than 100. Then for all n N we see that | an 1| = 1/n < 1/100 = e.
The proof below uses the same argument, but generalizes it to any
value of e > 0 as the definition of convergence requires.

Proof. Let e > 0 and note that | an 1| = |1 + 1/n 1| = |1/n| = 1/n.

It is clear that 1/n < e if n > 1/e, so choose N to be a number
greater than 1/e. Then for all n N we see that | an 1| = 1/n 
1/N < e as required.

Example 2.2.6. The sequence ( an )• n 2

n=n0 with an = ( 1) /n converges
to 0.

The first few terms of the sequence are 1, 1/4, 1/9, 1/16, . . . . So
the numbers look to be converging to 0, but the sign keeps switching.
The idea behind the proof remains the same.

Proof. Let e > 0 and note that | an 0| = |( 1)n /n2 | = 1/n2 . It is

p
clear that 1/n2 < e if n > 1/ e, so choose N to be a number greater
p
than 1/ e. Then for all n N we see that | an 0| = 1/n2 < e.

Notation 2.2.7. If a sequence ( an )•

n=n0 converges to a limit A, we write

lim an = A.
n!•

Example 2.2.8. We just showed that

lim ( 1)n /n2 = 0.

n!•

If, on the contrary, a sequence does not converge to any real

number, we say that it diverges.
12

Example 2.2.9. The sequence of term an = 1 + ( 1)n for n 2 N

diverges.
As we discussed earlier, the terms in this sequence alternate be-
tween 2 and 0. It might be natural to think then that the sequence
converges to 1, as it stays very close to 1. But this does not satisfy
our definition, simply because if e = 1/2 then no term in the se-
quence gets within this distance. A formal proof involves the triangle
inequality that we quickly recall:

for all a, b 2 R | a + b |  | a | + | b |.

Or that the size of the sum is no bigger than the sum of the sizes.

Proof. Suppose that lim an = a for some a 2 R.

n!•
Take e = 1. The definition of a limit means that there exists a
N > 0 such that for all n > N we have |( 1)n a| < 1.
By taking an odd n and an even n we get

| 1 a| < 1 et |1 a| < 1.

Now we’ll just apply the triangle inequality:

2=| 1 1|
= |( 1 a) (1 a)|
| 1 a | + |1 a|
< 1+1
=2

Thus 2 < 2, which is a contradiction. Thus our hypothesis that

lim an = a was wrong and so the sequence cannot converge.
n!•

Definition 2.2.10. If a sequence does not converge to any real number,

we say that it diverges.
So we can say that the previous example is a divergent sequence.
Another example of a divergent sequence is an = n as the numbers
grow and grow, and never approach a real number. However this
divergent behavior is very different from the one studied in above
example.
Definition 2.2.11. A sequence { an }• n=1 tends to infinity if for any
B > 0, there exists an N > 0 such that for all n N an > B.
p
Example 2.2.12. The sequence an = 1 + n tends to infinity.
p
Proof. Let B > 0 and notice that 1 + n > B precisely when n >
( B 1)2 . So let N be a number greater than ( B 1)2 then for any
p
n N we see that an 1 + N > 1 + ( B 1) = B. So this value of N
satisfies the definition, and so the sequence diverges.
 analysis for applications 13

This proof looks very similar to our previous proofs showing that
sequences converge. The key difference is that here we are showing
that an is bigger than something, and not that | an A| is less than
something.
Notice that the term “infinity” or the symbol “•” did not ap-
pear in our proof. This because the concept of “tending to infinity”
does not require us to enter into the messy business of calculations
involving •.

Definition 2.2.13. A sequence { an }•

n=1 tends to negative infinity if
for any B < 0, there exists an N > 0 such that for all n N an < B.

The definitions of tending to infinity and negative infinity are

similar. The only difference is that B is either an arbitrarily large
positive number, or an arbitrarily large negative number.

Example 2.2.14. The an = 2n n2 tends to negative infinity.

Proof. Let B < 0 and notice that if 2n n2 < B then n2 2n + B > 0.

p p
This occurs precisely when n < 1 1 B or n > 1 + 1 B, but
p
n 2 N, so choose N > 1 + 1 B. Now, if n N then

an = 2n n2 = (n 1)2 + 1
 (N 1)2 + 1
< (1 B) + 1
= B.

As required, we have shown that an < B for any B < 0. So the

seqence tends to negative infinity.

Limit notation
We have given the definition of a sequence { an }•n=n0 that converges
to a number A, or tends to infinity, or tends to negative infinity.
The quicker way to write this is an ! A, an ! •, and an ! •
respectively. Another way to write these concepts is as follows

lim an = A, lim an = •, lim an = •.

n!• n!• n!•

The subtlety here is that, despite being about to write limn!• an = •,

the limit does not exist, so the sequence is divergent.
Something we brushed under the rug earlier about limit notation
is that there is a small issue that we must deal with. If a sequence an
converges to two real numbers at the same time, for example 1 and 2,
then we would be able to write

1 = lim an = 2,
n!•
14

which would be extremely foolish.

Luckily, mathematics tells us that this cannot happen.

Theorem 2.2.15. Suppose { an }•

n=n0 converges to A and also converges to
0 0
A . Then A = A .

Proof. We have that the definition of convergence is satisfied for both

A and A0 , so for every z > 0 there exists some M > 0 and some M0
such that for all n M and all n0 M0

| an A| < z and | an0 A0 | < z.

We have changed the letters used in the definition, it will become

clear why we did this soon. Now, if N is bigger than both M and M0
then all n N will also be greater than M and M0 , so our statement
now reads; for every z > 0 there exists some N > 0 such that for all
n N
| an A| < z and | an A0 | < z.
By adding these terms together, we get

| an A| + | an A0 | < 2z.

Now, for any e > 0, let z = e/2. In particular z > 0 so our statement
still holds.
For every e > 0 there exists some N > 0 such that for all n N

| an A| + | an A0 | < 2z = e.

From the triangle inequality we conclude that

|A A0 | = | ( an A) + ( an A0 )|  | an A| + | an A0 | < e.

So we’ve shown that | A A0 | < e for every possible value of e. In

other words, the difference between A and A0 is smaller than every
possible distance. This clearly implies that A = A0 , as otherwise the
| A A0 |
statement would fail for e = 2 .

Combining limits of sequences

Once you able to determine that some sequences have limits, you
can deduce others by multiplying a sequence by a real number or
by adding them. That’s the goal of the following theorems. We’ll
only prove a few of them because the proofs all work more or less
the same way. As convergence is only about what happens towards
infinity, the first terms of the sequence don’t matter. For that reason,
sometimes we’ll use the simplified notation of ( an ) for a sequence
instead of the more mathematically correct ( an )•
n = n0 .
 analysis for applications 15

Theorem 2.2.1. If ( an ) converges to A and (bn ) converges to B, then

( an + bn ) converges to A + B.

Proof. Let e > 0. We want to show that

| a n + bn ( A + B)| < e

for all large enough n.

Let’s use the convergence of ( an ) et (bn ). We know that there exist
N1 and N2 such that
e e
8n > N1 , | an A| < and 8n > N2 , |bn B| <
2 2
Let N = max{ N1 , N2 }.
Now we can compute:

n > N ) | a n + bn ( A + B)|
= |( an A ) + ( bn B)|
 | a n A | + | bn B|
e e
< +
2 2
=e

as required.

The same style of argument shows:

Theorem 2.2.2. If ( an ) converges to A and l 2 R, then (lan ) converges

to lA. More generally, if ( an ) converges to A and (bn ) converges to B, then
( an · bn ) converges to AB.

We also have a quotient rule, but in order to avoid dividing by 0,

we have to be a little more careful.

Theorem 2.2.3. If ( an )•n=n0 converges⇣to A and (bn )•

n=n0 converges to
⌘•
an
B 6= 0 and bn 6= 0 for all n n0 , then bn converges to A
B.
n = n0