39

7 Sequences of real numbers

7.1 Definitions and examples
Definition 7.1.1. A sequence of real numbers is a real function whose domain is the set N of
natural numbers.

Let s : N → R be a sequence. Then the values of s are s(1), s(2), s(3), . . . , s(n), . . . . It is
customary to write sn instead of s(n) in this case. Sometimes a sequence will be specified by
listing its first few terms
s1 , s2 , s3 , s4 , . . . ,
and sometimes by listing of all its terms {sn }n∈N or {sn }+∞
n=1 . One way of specifying a sequence
is to give a formula, or recursion formula for its n−th term sn . Notice that in this notation s is
the “name” of the sequence and n is the variable.
Some examples of sequences follow.

Example 7.1.2. (a) 1, 0, −1, 0, 1, 0, −1, . . . ;

(b) 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, . . . ;

(c) 1, 1, 1, 1, 1, . . . ; (the constant sequence)

1 1 2 1 2 3 1 2 3 4 1 2 3 4 5 1 2 3 4 5
(d) , , , , , , , , , , , , , , , , , , , , . . . ; (What is the range
2 3 3 4 4 4 5 5 5 5 6 6 6 6 6 7 7 7 7 7
of this sequence?)

Recursively defined sequences

xn
Example 7.1.3. (a) x1 = 1, xn+1 = 1 + , n = 1, 2, 3, . . . ;
4
xn 1
(b) x1 = 2, xn+1 = + , n = 1, 2, 3, . . . ;
2 xn
√ √
(c) a1 = 2, an+1 = 2 + an , n = 1, 2, 3, . . . ;
√
(d) s1 = 1, sn+1 = 1 + sn , n = 1, 2, 3, . . . ;

9 + xn
(e) x1 = 0.9, xn+1 = , n = 1, 2, 3, . . . .
10
1 1
(f) b1 = , bn+1 = p , n = 1, 2, 3, . . .
2 2 1 − b2n

(g) f1 = 1, fn+1 = (n + 1) fn , n = 1, 2, 3, . . . .

Some important examples of sequences are listed below.

40

bn = c, c ∈ R. n ∈ N, (7.1.1)
pn = an , a ∈ R, n ∈ N, (7.1.2)
 n
1
xn = 1 + , n ∈ N, (7.1.3)
n
 (n+1)
1
yn = 1 + , n ∈ N, (7.1.4)
n
 a n
zn = 1 + , n ∈ N, (7.1.5)
n
f1 = 1, fn+1 = fn · (n + 1), n ∈ N. (7.1.6)

(The standard notation for the terms of the sequence {fn }+∞
n=1 is fn = n!, n ∈ N.)

an
qn = , a ∈ R, n ∈ N, (7.1.7)
n!
1
t1 = 1, tn+1 = tn + n ∈ N, (7.1.8)
n!
an
v1 = 1, vn+1 = vn + n ∈ N. (7.1.9)
n!
Let {an }+∞ +∞
n=1 be an arbitrary sequence. An important sequence associated with {an }n=1 is the
following sequence

S1 = a1 , Sn+1 = Sn + an+1 , n ∈ N. (7.1.10)

7.2 Convergent sequences

Definition 7.2.1. A sequence {sn }+∞
n=1 of real numbers converges to the real number L if for
each ǫ > 0 there exists a number N(ǫ) such that

n ∈ N, n > N(ǫ) ⇒ |sn − L| < ǫ.

If {sn }+∞
n=1 converges to L we will write

lim sn = L or sn → L (n → +∞).
n→+∞

The number L is called the limit of the sequence {sn }+∞

n=1 . A sequence that does not converge to
a real number is said to diverge.
Example 7.2.2. Let r be a real number such that |r| < 1. Prove that limn→+∞ r n = 0.
Solution. First note that if r = 0, then r n = 0 for all n ∈ N, so the given sequence is a constant
sequence. Therefore it converges. Let ǫ > 0. We need to solve |r n − 0| < ǫ for n. First simplify
|r n − 0| = |r n | = |r|n . Now solve |r|n < ǫ by taking ln of both sides of the inequality (note that
ln is an increasing function)
ln |r|n = n ln |r| < ln ǫ.
 41

ln ǫ
Since |r| < 1, we conclude that ln |r| < 0. Therefore the solution is n> . Thus, with
ln |r|
ln ǫ
N(ǫ) = , the implication
ln |r|

n ∈ N, n > N(ǫ) ⇒ |r n − 0| < ǫ

is valid.
n2 − n − 1 1
Example 7.2.3. Prove that lim 2
= .
n→+∞ 2n − 1 2
n2 − n − 1 1
Solution. Let ǫ > 0 be given. We need to solve − < ǫ for n. First simplify:
2n2 − 1 2

n2 − n − 1 1 2 n2 − n − 1 1 2n2 − 1 −2n − 1 2n + 1
2
− = 2
− 2
= 2
= 2
2n − 1 2 2 2n − 1 2 2n − 1 2 (2n − 1) 4n − 2

Now invent the BIN:

2n + 1 2n + n 3n 3
2
≤ 2 2
= 2 = .
4n − 2 4n − 2n 2n 2n
Therefore the BIN is:
n2 − n − 1 1 3
2
− ≤ valid for n ∈ N.
2n − 1 2 2n
Solving for n is now easy:

3 3
< ǫ. The solution is n > .
2n 2ǫ
3
Thus, with N(ǫ) = , the implication
2ǫ
n2 − n − 1
n > N(ǫ) ⇒ −1 <ǫ
2n2 − 1

is valid. Using the BIN, this implication should be easy to prove.

This procedure is very similar to the procedure for proving limits as x approaches infinity. In
fact the following two theorems are true.

Theorem 7.2.4. Let x 7→ f (x) be a function which is defined for every x ≥ 1. Assume that
lim f (x) = L. If the sequence {an }+∞
n=1 is defined by
x→+∞

an = f (n), n = 1, 2, 3, . . . ,

then lim an = L.
n→+∞
42

Theorem 7.2.5. Let x 7→ f (x) be a function which is defined for every x ∈ [−1, 0) ∪ (0, 1].
Assume that lim f (x) = L. If the sequence {an }+∞
n=1 is defined by
x→0

an = f (1/n), n = 1, 2, 3, . . . ,

then lim an = L.
n→+∞

The above two theorems are useful for proving limits of sequences which are defined by a
formula. For example you can prove the following limits by using these two theorems and what
we proved in previous sections.

Exercise 7.2.6. Find

 the following limits. Provide
  proofs.  
1 1 1
(a) lim sin (b) lim n sin (c) lim ln 1 +
n→+∞
n  n→+∞
 n  n→+∞
 n
1 1 1 1
(d) lim n ln 1 + (e) lim cos (f) lim cos
n→+∞ n n→+∞ n n→+∞ n n
The Algebra of Limits Theorem holds for sequences.

Theorem 7.2.7. Let {an }+∞ +∞ +∞

n=1 , {bn }n=1 and {cn }n=1 , be given sequences. Let K and L be real
numbers. Assume that

(1) lim an = K,
x→+∞

(2) lim bn = L.
x→+∞

Then the following statements hold.

(A) If cn = an + bn , n ∈ N, then lim cn = K + L.

x→+∞

(B) If cn = an bn , n ∈ N, then lim cn = KL.

x→+∞

an K
(C) If L 6= 0 and cn = , n ∈ N, then lim cn = .
bn x→+∞ L
Theorem 7.2.8. Let {an }+∞ +∞
n=1 and {bn }n=1 be given sequences. Let K and L be real numbers.
Assume that

(1) lim an = K.
x→+∞

(2) lim bn = L.
x→+∞

(3) There exists a natural number n0 such that

an ≤ bn for all n ≥ n0 .

Then K ≤ L.
 43

Theorem 7.2.9. Let {an }+∞ +∞ +∞

n=1 , {bn }n=1 and {sn }n=1 be given sequences. Assume the following

1. The sequence {an }+∞

n=1 converges to the limit L.

2. The sequence {bn }+∞

n=1 converges to the limit L.

3. There exists a natural number n0 such that

an ≤ sn ≤ bn for all n > n0 .

Then the sequence {sn }+∞

n=1 converges to the limit L.

Prove this theorem.

7.3 Sufficient conditions for convergence

Many limits of sequences cannot be found using theorems from the previous section. For example,
the recursively defined sequences (a), (b), (c), (d) and (e) in Example 7.1.3 converge but it cannot
be proved using the theorems that we presented so far.
Definition 7.3.1. Let {sn }+∞
n=1 be a sequence of real numbers.

1. If a real number M satisfies

sn ≤ M for all n∈N

then M is called an upper bound of {sn }+∞ +∞

n=1 and the sequence {sn }n=1 is said to be bounded
above.

2. If a real number m satisfies

m ≤ sn for all n ∈ N,
then m is called a lower bound of {sn }+∞ +∞
n=1 and the sequence {sn }n=1 is said to be bounded
below.

3. The sequence {sn }+∞

n=1 is said to be bounded if it is bounded above and bounded below.

Theorem 7.3.2. If a sequence converges, then it is bounded.

Proof. Assume that a sequence {an }+∞
n=1 converges to L. By Definition 7.2.1 this means that for
each ǫ > 0 there exists a number N(ǫ) such that

n ∈ N, n > N(ǫ) ⇒ |an − L| < ǫ.

In particular for ǫ = 1 > 0 there exists a number N(1) such that

n ∈ N, n > N(1) ⇒ |an − L| < 1.

Let n0 be the largest natural number which is ≤ N(1). Then n0 + 1, n0 + 2, . . . are all > N(1).
Therefore
|an − L| < 1 for all n > n0 .
44

This means that

L − 1 < an < L + 1 for all n > n0 .
The numbers L − 1 and L + 1 are not lower and upper bounds for the sequence since we do not
know how they relate to the first n0 terms of the sequence. Put

m = min{a1 , a2 , . . . , an0 , L − 1}
M = max{a1 , a2 , . . . , an0 , L + 1}.

Clearly

m ≤ an for all n = 1, 2, . . . , n0
m ≤ L − 1 < an for all n > n0 .

Thus m is a lower bound for the sequence {an }+∞

n=1 .
Clearly

an ≤M for all n = 1, 2, . . . , n0
an < L + 1 ≤M for all n > n0 .

Thus M is an upper bound for the sequence {an }+∞

n=1 .

Is the converse of Theorem 7.3.2 true? The converse is: If a sequence is bounded, then it
converges. Clearly a counterexample to the last implication is the sequence (−1)n , n ∈ N. This
sequence is bounded but it is not convergent.
The next question is whether boundedness and an additional property of a sequence can
guarantee convergence. It turns out that such an property is monotonicity defined in the following
definition.
Definition 7.3.3. A sequence {sn }+∞
n=1 of real numbers is said to be

non-decreasing if sn ≤ sn+1 for all n ∈ N,

strictly increasing if sn < sn+1 for all n ∈ N,
non-increasing if sn ≥ sn+1 for all n ∈ N.
strictly decreasing if sn > sn+1 for all n ∈ N.
A sequence with either of these four properties is said to be monotonic.
The following two theorems give powerful tools for establishing convergence of a sequence.
Theorem 7.3.4. If {sn }+∞ +∞
n=1 is non-decreasing and bounded above, then {sn }n=1 converges.

Theorem 7.3.5. If {sn }+∞ +∞

n=1 is non-increasing and bounded below, then {sn }n=1 converges.

To prove these theorems we have to resort to the most important property of the set of real
numbers: the Completeness Axiom.
The Completeness Axiom. If A and B are nonempty subsets of R such that for every a ∈ A
and for every b ∈ B we have a ≤ b, then there exists c ∈ R such that a ≤ c ≤ b for all a ∈ A and
all b ∈ B.
 45

Proof of Theorem 7.3.4. Assume that {sn }+∞n=1 is a non-decreasing sequence and that it is bounded
above. Since {sn }+∞
n=1 is non-decreasing we know that

s1 ≤ s2 ≤ s3 ≤ · · · ≤ sn−1 ≤ sn ≤ sn+1 ≤ · · · . (7.3.1)


Let A be the range of the sequence {sn }+∞ n=1 . That is A = sn : n ∈ N . Clearly A 6= ∅. Let
B be the set of all upper bounds of the sequence {sn }+∞ +∞
n=1 . Since the sequence {sn }n=1 is bounded
above, the set B is not empty. Let b ∈ B be arbitrary. Then b is an upper bound for {sn }+∞ n=1 .
Therefore
sn ≤ b for all n ∈ N.
By the definition of A this means

a≤b for all a ∈ A.

Since b ∈ B was arbitrary we have

a ≤ b for all a ∈ A and for all b ∈ B.

By the Completeness Axiom there exists c ∈ R such that

sn ≤ c ≤ b for all n ∈ N and for all b ∈ B. (7.3.2)

Thus c is an upper bound for {sn }+∞ n=1 and also c ≤ b for all upper bounds b of the sequence
+∞
{sn }n=1 . Therefore, for an arbitrary ǫ > 0 the number c − ǫ (which is < c) is not an upper bound
of the sequence {sn }+∞
n=1 . Consequently, there exists a natural number N(ǫ) such that

c − ǫ < sN(ǫ) . (7.3.3)

Let n ∈ N be any natural number which is > N(ǫ). Then the inequalities (7.3.1) imply that

sN(ǫ) ≤ sn . (7.3.4)

By (7.3.2) c is an upper bound of {sn }+∞

n=1 . Hence we have

sn ≤ c for all n ∈ N. (7.3.5)

Putting together the inequalities (7.3.3), (7.3.4) and (7.3.5) we conclude that

c − ǫ < sn ≤ c for all n ∈ N such that n > N(ǫ). (7.3.6)

The relationship (7.3.6) shows that for n ∈ N such that n > N(ǫ) the distance between numbers
sn and c is < ǫ. In other words

n ∈ N, n > N(ǫ) implies |sn − c| < ǫ.

This is exactly the implication in Definition 7.2.1. Thus, we proved that

lim sn = c.
n→+∞
46

Example 7.3.6. Prove that the sequence

1 1 1 1
tn = 1 + + + + · · · + − ln n, n = 1, 2, 3, . . . ,
2 3 4 n
converges.
Solution. Use the definition of ln as the integral to prove that for n > 1
Z n 
1 1
tn > − dx.
1 Floor(x) x
Deduce that tn > 0.
Represent

1
tn − tn+1 = ln(n + 1) − ln n −
n+1
as an area (or a difference of two areas). Conclude that tn − tn+1 > 0.
Now use one of the preceding theorems.

8 Infinite series of real numbers

8.1 Definition and basic examples
The most important application of sequences is the definition of convergence of an infinite series.
From the elementary school you have been dealing with addition of numbers. As you know the
addition gets harder as you add more and more numbers. For example it would take some time
to add
S100 = 1 + 2 + 3 + 4 + 5 + · · · + 98 + 99 + 100
It gets much easier if you add two of these sums, and pair the numbers in a special way:

2 S100 = 1 + 2 + 3 + 4 + · · · + 97 + 98 + 99 + 100
100 + 99 + 98 + 97 + · · · + 4 + 3 + 2 + 1

A straightforward observation that each column on the right adds to 101 and that there are 100
such columns yields that
101 · 100
2 S100 = 101 · 100, that is S100 = = 5050.
2
This can be generalized to any natural number n to get the formula
(n + 1) n
Sn = 1 + 2 + 3 + 4 + 5 + · · · + (n − 1) + n = .
2
This procedure indicates that it would be impossible to find the sum

1+2+3+4+5+···+n +···

where the last set of · · · indicates that we continue to add natural numbers.
 47

The situation is quite different if we consider the sequence

1 1 1 1 1
, , , , ..., n, ...
2 4 8 16 2
and start adding more and more consecutive terms of this sequence:
1 1 1
=1− =
2 2 2
1 1 1 3
+ =1− =
2 4 4 4
1 1 1 1 7
+ + =1− =
2 4 8 8 8
1 1 1 1 1 15
+ + + = 1− =
2 4 8 16 16 16
1 1 1 1 1 1 31
+ + + + = 1− =
2 4 8 16 32 32 32
1 1 1 1 1 1 1 63
+ + + + + = 1− =
2 4 8 16 32 64 64 64

These sums are nicely illustrated by the following pictures

1
€€€€
8

1 1 1
€€€€ €€€€ €€€€
2 2 2

1 1
€€€€ €€€€
4 4

1 1
€€€€€€€ €€€€€€€
32 32
1
€€€€€€€
64
1 1 1
€€€€ €€€€ €€€€
8 8 8

1 1 1
€€€€€€€ €€€€€€€ €€€€€€€
16 16 16

1 1 1
€€€€ €€€€ €€€€
2 2 2

1 1 1
€€€€ €€€€ €€€€
4 4 4

In this example it seems natural to say that the sum of infinitely many numbers 12 , 14 , 81 , . . . equals
1:
1 1 1 1 1
+ + + + ··· + n +··· = 1
2 4 8 16 2
48

Why does this make sense? This makes sense since we have seen above that as we add more
and more terms of the sequence
1 1 1 1 1
, , , , ..., n, ...
2 4 8 16 2
we are getting closer and closer to 1. Indeed,
1 1 1 1 1 1
+ + + + ··· + n = 1− n
2 4 8 16 2 2
and  
1
lim 1 − n = 1.
n→+∞ 2
This reasoning leads to the definition of convergence of an infinite series:
Definition 8.1.1. Let {an }+∞
n=1 be a given sequence. Then the expression

a1 + a2 + a3 + · · · + an + · · ·
is called an infinite series. We often abbreviate it by writing
+∞
X
a1 + a2 + a3 + · · · + an + · · · = an .
n=1

For each natural number n we calculate the (finite) sum of the first n terms of the series
sn = a1 + a2 + a3 + · · · + an .
+∞
X
We call sn a partial sum of the infinite series an . (Notice that {sn }+∞
n=1 is a new sequence.) If
n=1
the sequence {sn }+∞
n=1 converges and if

lim sn = S,
n→+∞
P+∞
then the infinite series n=1 an is called convergent and we write
+∞
X
a1 + a2 + a3 + · · · + an + · · · = S or an = S.
n=1

The number S is called the sum of the series.

If the sequence {sn }+∞
n=1 does not converge, then the series is called divergent.

In the example above we have

 n
1 1
an = n = ,
2 2
1 2n − 1
sn = 1 − n =
2 2n

1
lim 1 − n = 1.
n→+∞ 2
 49

Therefore we say that the series

+∞
1 1 1 1 1 X 1
+ + + + ··· + n +··· =
2 4 8 16 2 n=1
2n
+∞
X 1
converges and its sum is 1. We write = 1.
n=1
2n
In our starting example

an = n,
(n + 1)n
sn = 1 + 2 + 3 + · · · + n =
2
(n + 1)n
lim does not exist.
n→+∞ 2
Therefore we say that the series
+∞
X
1 + 2 + 3 + 4 + ··· + n +··· = n
n=1

diverges.
Example 8.1.2 (Geometric Series). Let a and r be real numbers. The most important infinite
series is
+∞
X
2 3 n
a+ ar + ar + ar + ···+ ar + ··· = a rn (8.1.1)
n=0

This series is called a geometric series. To determine whether this series converges or not we
need to study its partial sums:

s0 = a, s1 = a + a r,
s2 = a + a r + a r 2 , s3 = a + a r + a r 2 + a r 3 ,
s4 = a + a r + a r 2 + a r 3 + a r 4 , s5 = a + a r + a r 2 + a r 3 + a r 4 + a r 5 ,
..
.
sn = a + a r + a r 2 + · · · + a r n−1 + a r n
..
.
1
Notice that we have already studied the special case when a = 1 and r = . In this special
2
case we found a simple formula for sn and then we evaluated lim sn . It turns out that we can
n→+∞
find a simple formula for sn in the general case as well.
First note that the case a = 0 is not interesting, since then all the terms of the geometric
series are equal to 0 and the series clearly converges and its sum is 0. Assume that a 6= 0. If
r = 1 then sn = n a. Since we assume that a 6= 0, lim n a does not exist. Thus for r = 1 the
n→+∞
series diverges.
50

Assume that r 6= 1. To find a simple formula for sn , multiply the long formula for sn above
by r to get:

sn = a + a r + a r 2 + · · · + a r n−1 + a r n ,
r sn = a r + a r 2 + · · · + a r n + a r n+1 ;

now subtract,

sn − r sn = a − a r n+1,

and solve for sn :

1 − r n+1
sn = a .
1−r

We already proved that if |r| < 1, then lim r n+1 = 0. If |r| ≥ 1, then lim r n+1 does not exist.
n→+∞ n→+∞
Therefore we conclude that

1 − r n+1 1
lim sn = lim a =a for |r| < 1,
n→+∞ n→+∞ 1−r 1−r
lim sn does not exist for |r| ≥ 1 .
n→+∞

In conclusion

+∞
X 1
• If |r| < 1, then the geometric series a r n converges and its sum is a .
n=0
1−r

+∞
X
• If |r| ≥ 1, then the geometric series a r n diverges.
n=0

The following picture illustrates the sum of a geometric series with a > 0 and 0 < r < 1. The
width of the rectangle below is 1/(1 − r) and the height is a. The slopes of the lines shown are
(1 − r)a and r(1 − r)a.
 51

ar7
5
ar
3
ar

ar ar6

ar4

ar2

a-ra

0 1

In the picture above the terms of a geometric series are represented as areas. As we can see the
areas of the terms fill in the rectangle whose area is a/(1 − r).
In the picture below we represent the terms of the geometric series by lengths of horizontal
line segments. The picture strongly indicates that the total length of infinitely many horizontal
line segments is a/(1 − r). The reason for this is that by the construction the slope of the
hypothenuse of the right triangle in the picture below is −(1 − r). Since its vertical leg is a, its
horizontal leg must be a(1 − r).

a-ar

a
ar
ar2
ar3

a ar ar2 ar3 ar4 ar5 ar6 ...

Remark 8.1.3. How to recognize whether an infinite series is a geometric series?

+∞
X π n+2 π n+2
Consider for example the infinite series . Here an = .
n=1
e2n−1 e2n−1
Looking at the formula (8.1.1) we note that the first term of the series is a and that the ratio
between any two consecutive terms is r.
52

π n+2
For an = given above we calculate
e2n−1
π n+1+2
an+1 2(n+1)−1 π n+3 e2n−1 π
= e n+2 = 2n+1 n+2 = 2 .
an π e π e
e2n−1

+∞
an+1 X π n+2
Since is constant, we conclude that the series is a geometric series with
an n=1
e2n−1

π2 π
a = a1 = and r= for all n = 1, 2, 3, . . . .
e e2
π
Since r = < 1, we conclude that the sum of this series is
e2
+∞
X π n+2 π2 1 π 2 e2 π2 e
= = = 2 .
n=1
e2n−1 e 1− π e e2 − π e −π
e2
Thus, to verify whether a given infinite series is a geometric series calculate the ratio of the
consecutive terms and see whether it is a constant:
+∞
X an+1
an for which =r for all n = 1, 2, 3, . . . (8.1.2)
n=1
an

is a geometric series. In this case a = a1 (the first term of the series).

Example 8.1.4 (Harmonic Series). Harmonic series is the series
+∞
1 1 1 1 X 1
1 + + + + ··· + + ··· = .
2 3 4 n n=1
n

Again, to explore the convergence of this series we have to study its partial sums:
1
S1 = 1, S2 = 1 + ,
2
1 1 1 1 1
S3 = 1 + + , S4 = 1 + + + ,
2 3 2 3 4
1 1 1 1 1 1 1 1 1
S5 = 1 + + + + , S6 = 1 + + + + + ,
2 3 4 5 2 3 4 5 6
1 1 1 1 1 1 1 1 1 1 1 1 1
S7 = 1 + + + + + + , S8 = 1 + + + + + + + ,
2 3 4 5 6 7 2 3 4 5 6 7 8
..
.
1 1 1 1
Sn == 1 + + +···+ +
2 3 n−1 n
..
.
 53

1
Since Sn+1 − Sn = > 0 the sequence {Sn }+∞
n=1 is increasing.
n+1
+∞
Next we will prove that the sequence {Sn }n=1 is not bounded. We will consider only the
natural numbers which are powers of 2: 2, 4, 8, . . . , 2k , . . . . The following inequalities hold:
1 1 1
S2 = 1 + ≥1+ =1 + 1
2 2 2
1 1 1 1 1 1 1 1 1
S4 = 1 + + + ≥ 1 + + + = 1 + + 2 =1 + 2
2 3 4 2 4 4 2 4 2
1 1 1 1 1 1 1
S8 = 1 + + + + + + +
2 3 4 5 6 7 8
1 1 1 1 1 1 1 1 1 1 1
≥1+ + + + + + + =1+ +2 +4 =1 + 3
2 4 4 8 8 8 8 2 4 8 2
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
S16 = 1 + + + + + + + + + + + + + + +
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 1 1 1 1 1 1
≥1+ + + + + + +
2 4 4 8 8 8 8
1 1 1 1 1
=1+ +2 +4 +8 =1 + 4
2 4 8 16 2

Continuing this reasoning we conclude that for each k = 1, 2, 3, . . . the following formula holds:
1 1 1 1 1 1 1 1
S2k = 1 + + + + + · · · + + · · · + k−1 + k−1 + ···+ k
2 3 4 5 8 2 2 +1 2
1 1 1 1 1 1
≥1+ +2 +4 +8 + · · · + 2k−1 k =1 + k
2 4 8 16 2 2
Thus
1
S2k ≥ 1 + k for all k = 1, 2, 3, . . . (8.1.3)
2
sequence {Sn }+∞
This formula implies that the  n=1 is not bounded. Namely, let M be an arbitrary
real number. We put j = max 2 Floor(M), 1 . Then

j ≥ 2 Floor(M) > 2(M − 1) .

Therefore,
1
1+j >M.
2
Together with the inequality (8.1.3) this implies that

S2j > M .

Thus for an arbitrary real number M there exists a natural number n = 2j such that Sn > M.
This proves that the sequence {Sn }+∞
n=1 is not bounded and therefore it is not convergent.
In conclusion:

• The harmonic series diverges.

54

The next example is an example of a series for which we can find a simple formula for the
sequence of its partial sums and easily explore the convergence of that sequence. Examples of
this kind are called telescoping series.
+∞
X 1
Example 8.1.5. Prove that the series converges and find its sum.
n=1
n(n + 1)
Solution. We need to examine the series of partial sums of this series:
1 1 1 1
sn = + + +···+ , n = 1, 2, 3, . . . .
1·2 2·3 3·4 n(n + 1)
It turns out that it is easy to find the sum sn if we use the partial fraction decomposition for
each of the terms of the series:
1 1 1
= − for all k = 1, 2, 3, . . . .
k(k + 1) k k+1
Now we calculate:
1 1 1 1
sn = + + +···+
1·2 2·3 3·4 n (n + 1)
         
1 1 1 1 1 1 1 1 1 1
= − + − + − +···+ − + −
1 2 2 3 3 4 n−1 n n n+1
1
=1 − .
n+1
1
Thus sn = 1 − for all n = 1, 2, 3, . . .. Using the algebra of limits we conclude that
n+1
 
1
lim sn = lim 1 − = 1.
n→+∞ n→+∞ n+1
+∞
X 1
Therefore the series converges and its sum is 1:
n=1
n (n + 1)
+∞
X 1
= 1.
n=1
n (n + 1)

Exercise 8.1.6. Determine whether the series is convergent or divergent. If it is convergent,

find its sum.  
+∞ n−1 +∞ +∞ √ n +∞ n+3
X 2 X (−2)n+3 X ( 2) X e
(a) 6 (b) n−1
(c) n+1
(d)
n=1
3 n=1
5 n=0
2 n=1
π n−1
+∞ 2n−1 +∞ +∞ +∞
X 2 X 5 X
n
X 2
(e) (f) (g) (sin 1) (h)
n=1
πn n=1
2n n=0 n=0
n2 + 4n + 3
+∞ +∞ +∞ +∞  
X
n
X 2 X
n
X 1
(i) (cos 1) (j) 2 −1
(k) (tan 1) (l) ln 1 +
n=0 n=2
n n=0 n=1
n
 55

A digit is a number from the set {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}. A decimal number with digits
d1 , d2 , d3 , . . . , dn , . . . is in fact an infinite series:
+∞
X dn
0.d1 d2 d3 . . . dn . . . = n
.
n=1
10

Therefore each decimal number with digits that repeat leads to a geometric series. We use the
following abbreviation:

0.d1 d2 d3 . . . dk = 0.d1 d2 d3 . . . dk d1 d2 d3 . . . dk d1 d2 d3 . . . dk d1 d2 d3 . . . dk . . .

Exercise 8.1.7. Express the number as a ratio of integers.

(a) 0.9 = 0.999 . . . (b) 0.7 = 0.777 . . . (c) 0.712 (d) 0.5432

8.2 Basic properties of infinite series

An immediate consequence of the definition of a convergent series is the following theorem
+∞
X
Theorem 8.2.1. If a series an converges, then lim an = 0.
n→+∞
n=1

+∞
X
Proof. Assume that an is a convergent series. By the definition of convergence of a series
n=1
its sequence of partial sums {sn }+∞
n=1 converges to some number S: lim sn = S. Then also
n→+∞
lim sn−1 = S. Now using the formula
n→+∞

an = sn − sn−1 , for all n = 2, 3, 4, . . . ,

and the algebra of limits we conclude that

lim an = lim sn − lim sn−1 = S − S = 0.

n→+∞ n→+∞ n→+∞

Warning: The preceding theorem cannot be used to conclude that a particular series con-
+∞
X
verges. Notice that in this theorem it is assumed that an is a convergent.
n=1
On a positive note: Theorem 8.2.1 can be used to conclude that a given series diverges: If we
X+∞
know that lim an = 0 is not true, then we can conclude that the series an diverges. This is
n→+∞
n=1
a useful test for divergence.
Theorem 8.2.2 (The Test for Divergence). If the sequence {an }+∞
n=1 does not converge to 0, then
+∞
X
the series an diverges.
n=1
56

  +∞
1 X
Example 8.2.3. Determine whether the infinite series cos converges or diverges.
n=1
n
Solution. Just perform the divergence test:
 
1
lim cos = 1 6= 0 .
n→+∞ n
+∞  
X 1
Therefore the series cos diverges.
n=1
n
+∞ (−1)n
X n
Example 8.2.4. Determine whether the infinite series converges or diverges.
n=1
n+1
 (−1)n
+∞
n
Solution. Consider the sequence :
n+1 n=1
1 2 1 4 1 6 1 8 1 10 1 12 1 2k
, , , , , , , , , , , , ..., , , ...
1 · 2 3 3 · 4 5 5 · 6 7 7 · 8 9 9 · 10 11 11 · 12 13 (2k − 1) · 2k 2k + 1
(8.2.1)
Without giving a formal proof we can tell that this sequence diverges. In my informal language
the sequence (8.2.1) is not constantish since it can not decide whether to be close to 0 or 1.
+∞ (−1)n
X n
Therefore the series diverges.
n=1
n + 1
  +∞
1 X
Remark 8.2.5. The divergence test can not be used to answer whether the series sin
n=1
n
 
1
converges or diverges. It is clear that lim sin = 0. Thus we can not use the test for
n→+∞ n
divergence.
+∞
X +∞
X
Theorem 8.2.6 (The Algebra of Convergent Infinite Series). Assume that an and bn are
n=1 n=1
convergent series. Let c be a real number. Then the series
+∞
X +∞
X +∞
X



c an , an + bn , and an − bn ,
n=1 n=1 n=1

are convergent series and the following formulas hold

+∞
X +∞
X
c an = c an ,
n=1 n=1
+∞
X +∞
X +∞
X


an + bn = an + bn , and
n=1 n=1 n=1
+∞
X +∞
X +∞
X


an − bn = an − bn .
n=1 n=1 n=1
 57

+∞
X +∞
X
Remark 8.2.7. The fact that we write bn does not necessarily mean that bn is a genuine
n=1 n=1
infinite series.
For example, let m be a natural number and assume that bn = 0 for all n > m. Then
+∞
X Xm +∞
X +∞
X
bn = bn . In this case the series bn is clearly convergent. If an is a convergent
n=1 n=1 n=1 n=1
+∞
X

(genuine) infinite series, then Theorem 8.2.6 implies that the infinite series an + bn is
n=1
convergent and
+∞
X +∞
X m
X


an + bn = an + bn .
n=1 n=1 n=1

This in particular means that the nature of convergence of an infinite series can not be changed
by changing finitely many terms of the series.
For example, let m be a natural number. Then:
+∞
X +∞
X
The series an converges if and only if the series am+k converges.
n=1 k=1

+∞
P
Moreover, if an converges, then the following formula holds
n=1

+∞
X m
X +∞
X
an = aj + am+k .
n=1 j=1 k=1

+∞  
X π 1
Example 8.2.8. Prove that the series − n converges and find its sum.
n=1
n(n + 1) 2

Exercise 8.2.9. Determine whether the series is convergent or divergent. If a series is convergent
find its sum.
+∞ +∞ +∞ n +∞  
X n X X 3 + 2n X 3 π
(a) (b) arctan n (c) n+1
(d) 2 −1
+ n
n=1
n + 1 n=1 n=0
5 n=2
n e
+∞
X en + π n +∞   +∞
X (n + 1)2 +∞
X 1 X
(e) 2n−1
(f) n sin (g) 2 +1
(h) ((0.9)n + (0.1)n )
n=0
2 n=1
n n=0
n n=0

Exercise 8.2.10. Express the following sums as ratios of integers and as repeating decimal
numbers.
(a) 0.47 + 0.5 (b) 0.499 + 0.47 (c) 0.499 + 0.503

8.3 Comparison Theorems

Warning: All series in the next two sections have positive terms! Do not use the
tests from these sections for series with some negative terms.
58

The convergence of the series in Examples 8.1.2 and 8.1.5 was established by calculating the
limits of their partial sums. This is not possible for most series. For example we will soon prove
that the series
+∞
X 1
n=1
n2
π2
converges. To understand why the sum of this series is exactly you need to take a class about
6
Fourier series, Math 430.
I hope that you have done your homework and that you proved that the series
+∞
X 1
n=2
n2 −1

converges and that you found its sum. If you didn’t here is a way to do it: (It turns out that
this is a telescoping series.)
Let
1 1 1 1
Sn = + + +···+ 2 .
3 8 15 n −1
1
Since Sn+1 − Sn = > 0 the sequence {Sn }+∞
n=2 is increasing.
(n + 1)2 − 1
For each k = 2, 3, 4, . . . we have the following partial fractions decomposition
 
1 1 1 1 1
= = − .
k2 − 1 (k − 1)(k + 1) 2 k−1 k+1
Next we use this formula to simplify the formula for the n-th partial sum
n n   n  
X 1 X 1 1 1 1X 1 1
Sn = = − = −
k2 − 1 2 k−1 k+1 2 k−1 k+1
k=2 k=2 k=2
         
1 1 1 1 1 1 1 1 1 1 1
= − + − + − +···+ − + −
2 1 3 2 4 3 5 n−2 n n−1 n+1
 
1 1 1 1 1
= + − −
2 1 2 n n+1
 
1 3 2n + 1 3 2n + 1
= − = − .
2 2 n(n + 1) 4 2n(n + 1)
Using the algebra of limits we calculate
2n + 1 2 1
2n + 1 + 2
lim = lim n2 = lim n n = 0+0 = 0.
n→+∞ 2n(n + 1) n→+∞ 2n(n + 1) n→+∞ n+1 2·1
2
n2 n
Therefore, using the algebra of limits again, we calculate
3 3
lim Sn = −0= .
n→+∞ 4 4
 59

3
Clearly Sn < for all n = 2, 3, . . ..
4
Now consider the series
+∞
X 1 1 1 1 1
2
=1+ + + +···+ 2 +···
n=1
n 4 9 16 n

Let
1 1 1 1
Tn = 1 + + + +···+ 2 .
4 9 16 n
1
The fact that Tn+1 − Tn = > 0 implies that the sequence {Tn }+∞
n=1 is increasing.
(n + 1)2
Since
1 1 1 1 1 1 1 1
< , < , < , ..., 2
< 2 ,
4 3 9 8 16 15 n n −1
we conclude that
1 1 1 1 1 1 1 1 3
Tn = 1 + + + +···+ 2 < 1+ + + +···+ 2 = 1 + Sn < 1 + .
4 9 16 n 3 8 15 n −1 4
7
Thus Tn < for all n = 2, 3, 4, . . .. Since the sequence {Tn }+∞
n=1 is increasing and bounded above
4
+∞
X 1 7
it converges by Theorem 7.3.4. Thus the series 2
converges and its sum is < .
n=1
n 4
The principle demonstrated in the above example is the core of the following comparison
theorem.
+∞
X +∞
X
Theorem 8.3.1 (The Comparison Test). Let an and bn be infinite series with positive
n=1 n=1
terms. Assume that
an ≤ bn for all n = 1, 2, 3, . . . .
+∞
X +∞
X +∞
X +∞
X
(a) If bn converges, then an converges and an ≤ bn .
n=1 n=1 n=1 n=1

+∞
X +∞
X
(b) If an diverges, then bn diverges.
n=1 n=1

Sometimes the following comparison theorem is easier to use.

+∞
X +∞
X
Theorem 8.3.2 (The Limit Comparison Test). Let an and bn be infinite series with
n=1 n=1
positive terms. Assume that
an
lim = L.
n→+∞ bn
+∞
X +∞
X +∞
X +∞
X
If bn converges, then an converges. Or, equivalently, if an diverges, then bn
n=1 n=1 n=1 n=1
diverges.
60

+∞
X n+1
Example 8.3.3. Determine whether the series √ converges or diverges.
n=1
1 + n6
Solution. The dominant term in the numerator is n and the dominant term in the denominator
+∞
√ X 1
6 3
is n = n . This suggests that this series behaves as the convergent series . Since we are
n=1
n2
trying to prove convergence we will take
n+1 1
an = √ and bn =
1 + n6 n2
in the Limit Comparison Test. Now calculate:
n+1 n2 (n + 1) 1
√ 2 1 +
1+n 6 n (n + 1) 3
lim = lim √ = lim √ n = lim r n = 1 .
n→+∞ 1 n→+∞ 1+n 6 n→+∞ 1+n 6 n→+∞ 1
n2 3
+1
n n6
In the last step we used the algebra of limits and the fact that
r
1
lim +1=1
n→+∞ n6
which needs a proof by definition.
n+1
√ +∞
1 + n6 X 1
Since we proved that lim = 1 and since we know that is convergent, the
n→+∞ 1 n2
n=1
n2
+∞
X n+1
Limit Comparison Test implies that the series √ converges.
1 + n6
n=1

In the next theorem we compare an infinite series with an improper integral of a positive
function. Here it is presumed that we know how to determine the convergence or divergence of
the improper integral involved.
Theorem 8.3.4 (The Integral Test). Suppose that x 7→ f (x) is a continuous positive, decreasing
function defined on the interval (0, +∞). Assume that an = f (n) for all n = 1, 2, . . .. Then the
following statements are equivalent
Z +∞
(a) The integral f (x) dx converges.
1

+∞
X
(b) The series an converges.
n=1

At this point we assume that you are familiar with improper integrals and that you know
how to decide whether an improper integral converges or diverges.
We will use this test in two different forms:
 61

Z +∞ +∞
X
• Prove that the integral f (x) dx converges. Conclude that the series an converges.
1 n=1

Z +∞ +∞
X
• Prove that the integral f (x) dx diverges. Conclude that the series an diverges.
1 n=1

+∞
X 1
Example 8.3.5 (Convergence of p-series). Let p be a real number. The p-series is
n=1
np
convergent if p > 1 and divergent if p ≤ 1.
Solution. Let n > 1. Then the function x 7→ nx is an increasing function. Therefore, if p < 1,
then np < n. Consequently,
1 1
p
> , for all n > 1 and p < 1 .
n n
+∞ +∞
X 1 X 1
Since the series diverges, the Comparison Test implies that the series p
diverges for
n=1
n n=1
n
all p ≤ 1.
1
Now assume that p > 1. Consider the function f (x) = p , x > 0. This function is a
x
continuous, decreasing, positive function. Let me calculate the improper integral involved in the
Integral Test for convergence:
Z Z t
+∞ t
1 1 1 1
dx = lim dx = lim
1 xp t→+∞ 1 x
p t→+∞ 1 − p xp−1
1
!
1 1 1 1
= lim − 1 = (−1) =
1 − p t→+∞ tp−1 1−p p−1

Thus this improper integral converges. Notice that the condition p > 1 was essential to conclude
1 1
that lim p−1 = 0. Since p = f (n) for all n = 1, 2, 3, . . ., the Integral Test implies that the
t→+∞ t n
+∞
X 1
series converges for p > 1.
n=1
np
1
Remark 8.3.6. We have not proved this for all p > 1 the function f (x) = p , x > 0, is continuous.
x
One way to prove that for an arbitrary a ∈ R the function x 7→ xa , x > 0 is continuous is to use
the identity
xa = ea ln x , x > 0.
This identity shows that the function x 7→ xa , x > 0 is a composition of the function exp(x) =
ex , x ∈ R and the function x 7→ a ln x, x > 0. The later function is continuous by the algebra of
continuous functions: It is a product of a constant a and a continuous function ln. We proved
that exp is continuous. By Theorem 6.1.17 a composition of continuous function is continuous.
Consequently x 7→ xa , x > 0 is continuous.
62

One way to prove this is to use the squeeze

1 1 1
1 − p(x − 1) ≤ p
≤ for all x > 1 − .
x 1 + p(x − 1) p
For p = 2, 3, 4, . . ., this squeeze can be proved by induction. For other
Exercise 8.3.7. Determine whether the series is convergent or divergent.
+∞ +∞ +∞ +∞
X 1 X
−n2
X 1 X ln n
(a) √ (b) ne (c) (d) √
n=1
n n n=1 n=1
n ln n n=1
n n
+∞ +∞ +∞ +∞
X 1 X 1 X 1
X 1 1

(e) b
(f) (g) sin (h) sin
n=1
n(ln n) n=1
n! n=1
n n=2
n n
+∞ +∞
X π + en +∞ +∞
X1 1
X n! X n2 + 1
(i) cos (j) (k) (l) √
n=1
n n n=0
e + πn n=1
nn n=0
n7 + n3 + 1
For the series in (e) find all numbers b for which the series converges.
Exercise 8.3.8. A digit is a number from the set {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}. A decimal number
with digits d1 , d2 , d3 , . . . , dn , . . . is in fact an infinite series:
+∞
X dn
0.d1 d2 d3 . . . dn . . . = .
n=1
10n

Use a theorem from this section to prove that the series above always converges.

8.4 Ratio and root tests

Warning: All series in this section have positive terms! Do not use the tests from
this section for series with negative terms.
In Remark 8.1.3 we pointed out (see (8.1.2)) that a series
+∞
X an+1
an for which =r for all n = 1, 2, 3, . . .
n=1
an

is a geometric series. Consequently, if |r| < 1 this series is convergent, and it is divergent if
|r| ≥ 1.
+∞
X 1
Testing the series using this criteria leads to the ratio
n=0
3 − 2n+1
n

  n   n
1 n 2 2
n n+1 3 1−2 1−2
3n+1 − 2n+2 = 3 − 2 =  
3
  =
1 3
n  n+1
1 3n+1 − 2n+2 2 3 2
3n+1 1 − 2 1−2
3n − 2n+1 3 3
which certainly is not constant, but it is “constantish.” I propose that series for which the ratio
an+1 /an is not constant but constantish, should be called “geometrish.” The following theorem
tells that convergence and divergence of these series is determined similarly to geometric series.
 63

+∞
X
Theorem 8.4.1 (The Ratio Test). Assume that an is a series with positive terms and that
n=1
an+1
lim = R.
n→+∞ an
Then
(a) If R < 1, then the series converges.
(b) If R > 1, then the series diverges.
Another way to recognize a geometric series is:
+∞ r
X an+1
A series an for which n =r for all n = 1, 2, 3, . . .
n=1
a1
is a geometric series. Consequently, if |r| < 1 this series is convergent, and it is divergent if
|r| ≥ 1.
+∞  n
X 1+n
Testing the series using this criteria leads to the root
n=0
1 + 2n
s n 1
1+n 1+n +1
n
= = n1
1 + 2n 1 + 2n n
+2
which certainly is not constant, but it is “constantish.”
+∞
X
Theorem 8.4.2 (The Root Test). Assume that an is a series with positive terms and that
n=1
√
lim n
an = R .
n→+∞

Then
(a) If R < 1, then the series converges.
(b) If R > 1, then the series diverges.
Remark 8.4.3. Notice that in both the ratio test and the root test if the limit R = 1 we can
conclude neither divergence nor convergence. In this case the test is inconclusive.
Exercise 8.4.4. Determine whether the series is convergent or divergent.
+∞ +∞  n +∞ +∞
X 1 X n+2 X 4n X n!
(a) n
(b) (c) 2n−1
(d)
n=2
2 −3 n=1
2n − 1 n=1
3 n=1
1 · 3 · 5 · · · (2n − 1)
+∞
X 3n n2 +∞
X +∞
X e1/n +∞
X 2 · 4 · 6 · · · (2n)
(e) (f) e−n n! (g) (h)
n=1
n! n=1 n=1
n2 n=1
1 · 3 · 5 · · · (2n − 1)
+∞
X (n!)2 +∞
X 2 n2n +∞
X 23n +∞
X 1
(i) (j) 2 n
(k) 2n
(l)
n=1
(2n)! n=1
(3n + 1) n=1
3 n=1
(arctan n)n
+∞ 2 +∞ +∞ n +∞
X n X (n + 1)2 X a X 1 · 3 · 5 · · · (2n − 1)
(m) n
(n) n
(o) (p)
n=1
2 n=1
n2 n=1
n! n=1
2 · 4 · 6 · · · (2n)
For some of the problems you might need to use tests from previous sections.
64

8.5 Alternating infinite series

In the previous two sections we considered only series with positive terms. In this section we
consider series with both positive and negative terms which alternate: positive, negative, positive,
etc. Such series are called alternating series. For example
+∞
1 1 1 1 1 1 X 1
1− + − + − + · · · + (−1)n+1 + · · · = (−1)n+1 (8.5.1)
2 3 4 5 6 n n=1
n
+∞
1 1 1 1 1 1 1 1 1 1 X 4(−1)n+1
1−1+ − + − + − + − + − +··· =
(8.5.2)
3 2 5 3 7 4 8 5 9 6 n=1
n 3 + (−1) n+1

+∞
3 4 5 6 7 n+1 X n+1
2− + − + − + · · · + (−1)n+1 +··· = (−1)n+1 (8.5.3)
2 3 4 5 6 n n=1
n

Theorem 8.5.1 (The Alternating Series Test). If the alternating series

+∞
X
n+1
a1 − a2 + a3 − a4 + · · · + (−1) an + · · · = (−1)n+1 an
n=1

satisfies the following two conditions:

(i) 0 < an+1 ≤ an for all n = 1, 2, 3, . . .,

(ii) lim an = 0,
n→+∞

then the series is convergent.

Proof. Assume that {an }+∞

n=1 is a non-increasing sequence (that is assume that (i) is true) and
lim an = 0.
n→+∞
By the definition of convergence for each ǫ > 0 there exists N(ǫ) such that

n ∈ N, n > N(ǫ) implies |an − 0| < ǫ .

Since an > 0, the last implication can be simplified as

n ∈ N, n > N(ǫ) implies an < ǫ . (8.5.4)

We need to show that the sequence of partial sums {sn }+∞

n=1 ,

sn = a1 − a1 − a2 + a3 − a4 + · · · + (−1)n+1 an , n = 1, 2, 3, 4, . . . ,

is convergent.
First consider the sequence {s2n }+∞
n=1 of even partial sums. Then

s2(n+1) − s2n = a2n+1 − a2n+2 ≥ 0, since by (i) a2n+2 ≤ a2n+1 .

Thus the sequence {s2n }+∞

n=1 is non-decreasing.
 65

Next we compare an arbitrary even partial sum s2k with an arbitrary odd partial sum s2j−1 .
Assume j ≤ k, then





s2k − s2j−1 = −a2j + a2j+1 + −a2j+2 + a2j+3 + · · · + −a2k−4 + a2k−3 + −a2k−2 + a2k−1 − a2k .

Each of the numbers in the parenthesis is negative. Therefore the last sum is negative. That is
s2k ≤ s2j−1 for j ≤ k.
Assume now that j > k, then





s2j−1 − s2k = a2k+1 − a2k+2 + a2k+3 − a2k+4 + · · · + a2j−5 − a2j−4 + a2j−3 − a2j−2 + a2j−1 .

Each of the numbers in the parenthesis is positive. Therefore the last sum is positive. That is
s2k ≤ s2j−1 for j > k. Thus we conclude that

s2k ≤ s2j−1 for all j, k = 1, 2, 3, . . . . (8.5.5)

In particular (8.5.5) means that {s2n }+∞

n=1 is bounded above and that each s2j−1 , j = 1, 2, 3, . . . is
an upper bound. Since the sequence {s2n }+∞ n=1 is also non-decreasing, the Monotone Convergence
Theorem, Theorem 7.3.4, implies that {s2n }+∞ n=1 converges to its least upper bound, call it S.
Consequently
s2k ≤ S ≤ s2j−1 for all j, k = 1, 2, 3, . . . . (8.5.6)
For each two consecutive natural numbers n, n − 1 one of them is even and one is odd. Therefore
the inequalities in (8.5.6) imply that

|sn − S| ≤ |sn − sn−1 | = an for all n = 1, 2, 3, . . . . (8.5.7)

Let ǫ > 0 be arbitrary. Let n ∈ N be such that n > N(ǫ). Then by (8.5.4) we conclude that

an < ǫ (8.5.8)

Combining the inequalities (8.5.7) and (8.5.8) we conclude that

|sn − S| < ǫ .

Thus we have proved that for each ǫ > 0 there exists N(ǫ) such that

n ∈ N, n > N(ǫ) implies |sn − S| < ǫ .

This proves that the sequence {sn }+∞

n=1 converges and therefore the alternating series converges.

Example 8.5.2. The series in (8.5.1) is called alternating harmonic series. It converges.
Solution. We verify two conditions of the Alternating Series Test:
1 1
an+1 ≤ an since < , for all n = 1, 2, 3, . . . ,
n+1 n
1
lim = 0 is easy to prove by definition.
n→+∞ n

Thus the Alternating Series Test implies that the alternating harmonic series converges.
66

Remark 8.5.3. The Alternating Series Test does not apply to the series in (8.5.2) since the
sequence of numbers

1 1 1 1 1 1 1 1 1 1 4
1, 1, , , , , , , , , , , ...,
, . . .
3 2 5 3 7 4 8 5 9 6 n 3 + (−1)n+1

is not non-increasing. Further exploration of the series in (8.5.2) would show that it diverges.
The Alternating Series Test does not apply to the series in (8.5.3) since this series does not
satisfy the condition (ii):
n+1
lim = 1 6= 0 .
n→+∞ n
Again this series is divergent by the Test for Divergence.

Exercise 8.5.4. Determine whether the given series converges or diverges.

+∞   +∞  π +∞  
X 1 X X 1
(a) cos nπ + (b) sin n (c) sin nπ −
n=1
n n=0
2 n=1
n
+∞   +∞   +∞
X 1 1 X (−1)n X 1  π
(d) cos nπ + (e) ln 1 − (f) sin n
n=1
n n n=1
n n=1
n 2
+∞   +∞
X (−1)n+1 +∞
X (−1)n+1
X π 1
(g) sin n + (h) (i)
n=1
2 n n=1
n − (−1)n n=1
2n − (−1)n

Many of the exercises in the next section use the Alternating Series Test for convergence. Do
those exercises as well.

8.6 Absolute and Conditional Convergence

In the previous section we proved that the alternating harmonic series
+∞
1 1 1 1 1 1 X 1
1− + − + − + · · · + (−1)n+1 + · · · = (−1)n+1 converges. (8.6.1)
2 3 4 5 6 n n=1
n

Later on we will see that the sum of this series is ln 2.

Talking about infinite series in class I have often used the analogy with an infinite column in
a spreadsheet and finding its sum. A series with positive and negative terms one can interpret
as balancing a checkbook with (infinitely) many deposits and withdrawals. Looking at the
alternating harmonic series (8.6.1) we see a sequence of alternating deposits and withdrawals,
infinitely many of them. What we proved in the previous section tells that under two conditions
on the deposits and withdrawals, although it has infinitely many transactions, this checkbook
can be balanced.
Now comes the first surprising fact! Let’s calculate how much has been deposited to this
account:
+∞
1 1 1 1 X 1
1+ + + +···+ +··· = .
3 5 7 2n − 1 n=1
2n − 1
 67

Applying the Limit Comparison Test with the harmonic series it is easy to conclude this series
diverges. Looking at the withdrawals we see
+∞
1 1 1 1 1X1
− − − −···− −··· = − .
2 4 6 2n − 1 2 n=1 n

Again this is a divergent series. This is certainly a suspicious situation: Dealing with an account
to which an unbounded amount of money has been deposited and an unbounded amount of
money has been withdrawn. A simpler way to look at this is to look at the total amount of
money that went through this account (one can call this amount the total “activity” of the
account):
+∞
X 1 1 1 1 1 1 1
(−1)n+1 = 1 + + + + + + · · · + + · · · (8.6.2)
n=1
n 2 3 4 5 6 n

This is the harmonic series which is divergent.

Since we know that an unbounded amount of money has been deposited to this account we
might want to get in the spending mood sooner and do two withdrawals after each deposit,
keeping the amounts the same:

1 1 1 1 1 1 1 1 1 1 1
1− − + − − + − − + − − +··· (8.6.3)
2 4 3 6 8 5 10 12 7 14 16
In any real life checking account this might result in an occasional low balance but if the deposits
and withdrawals are identical, no mater how you arrange them they should result in the same
final balance. Amazingly this is not always the case with infinite series! (This is the second
surprising fact!) The series in (8.6.3) also converges but to a different number then the series in
(8.6.1). The following calculation indicates that the sum of the series in (8.6.3) is 1/2 of the sum
of the alternating harmonic series in (8.6.1).

1 1 1 1 1 1 1 1 1 1 1
S3k = 1 − − + − − + − − +···+ − −
2 4 3 6 8 5 10 12 2k − 1 4k − 2 4k
1 1 1 1 1 1 1 1
= − + − + − +···+ −
2  4 6 8 10 12 4k − 2 4k 
1 1 1 1 1 1 1 1
= 1− + − + − +···+ −
2 2 3 4 5 6 2k − 1 2k

This is a remarkable observation: a change of order of summation can change the sum of an
infinite series. This feature is closely related to the fact that the total activity of the account
expressed in (8.6.2) is a divergent series. This is a motivation for the following definition.
+∞
X
Definition 8.6.1. A convergent series an is called conditionally convergent if the series
n=1
+∞
X
of the absolute values of its terms |an | is divergent.
n=1
68

+∞
X
Definition 8.6.2. A series an is called absolutely convergent if the series of the absolute
n=1
+∞
X
values of its terms |an | is convergent.
n=1

Example 8.6.3. Prove that the series

+∞
1 1 1 1 1 1 X 1
1− + − + − + · · · + (−1)n+1 2 + · · · = (−1)n+1 2
4 9 16 25 36 n n=1
n

is absolutely convergent.
Solution. By the definition of absolute convergence we need to determine the convergence of the
series
+∞ +∞
X 1 X 1 1 1 1 1 1
(−1)n+1 2 = 2
= 1 + + + + + +···
n=1
n n=1
n 4 9 16 25 36
This is a p-series with p = 2. Therefore this series converges. (Notice that at the beginning of
Section 8.3 we proved that this series converges by comparing it to a telescoping series.)
Remark 8.6.4. One can interpreted the series in Example 8.6.3 as a checking account with in-
finitely many alternating deposits and withdrawals. In this case the total activity of the account
is a convergent series. Consequently the total amount deposited
+∞
1 1 1 X 1
1+ + +···+ 2
+ · · · = (8.6.4)
9 25 (2n − 1) n=1
(2n − 1)2

and the total amount withdrawn

+∞ +∞
1 1 1 1 X 1 1X 1
+ + +···+ + · · · = = (8.6.5)
4 16 36 (2n)2 n=1
(2n)2 4 n=1 n2

are both convergent series. As we can see, the total amount withdrawn is 1/4 of the total activity
of the account. We mentioned before that (we can not prove it in this course)
+∞
X 1 1 1 1 1 1 π2
= 1 + + + + + + · · · = .
n=1
n2 4 9 16 25 36 6

Therefore
+∞
X 1 1 1 1 1 1 3 π2 1 π2 1 π2 π2
(−1)n+1 2 = 1 − + − + − +··· = − = =
n=1
n 4 9 16 25 36 4 6 4 6 2 6 12

+∞
X
Theorem 8.6.5. If a series an is absolutely convergent, then it is convergent.
n=1
 69

+∞
X +∞
X
Proof. Assume that an is absolutely convergent, that is assume that |an | is convergent.
n=1 n=1
+∞
X
Then the algebra of convergent series the series 2 |an | is convergent. Since −|an | ≤ an ≤ |an |,
n=1
we conclude that
0 ≤ an + |an | ≤ 2 |an | for all n = 1, 2, 3, . . .
+∞
X

By the Comparison Test it follows that the series an + |an | is convergent. The algebra of
n=1
convergent series implies that the series
+∞ 
X  X+∞


an + |an | − |an | = an
n=1 n=1

is also convergent.
The following stronger versions of the Ratio and the Root test can be applied to any series
to determine whether a series converges absolutely or it diverges.
+∞
X |an+1 |
Theorem 8.6.6 (The Ratio Test). Let an be a series for which lim = R. Then
n→+∞ |an |
n=1

(a) If R < 1, then the series converges absolutely.

(b) If R > 1, then the series diverges.

+∞
X p
n
Theorem 8.6.7 (The Root Test). Let an be a series for which lim |an | = R. Then
n→+∞
n=1

(a) If R < 1, then the series converges absolutely.

(b) If R > 1, then the series diverges.

Notice that if the root or the ratio test apply to a series, then series either converges absolutely
or diverges. This implies that if a series converges conditionally, then either
|an+1 | |an+1 |
lim =1 or lim does not exist,
n→+∞ |an | n→+∞ |an |

and also
p
n
p
n
lim |an | = 1 or lim |an | does not exist.
n→+∞ n→+∞

In other words, the root and the ratio test cannot lead to a conclusion that a series converges
conditionally.
It turns out that our only tool which can be used to conclude conditional convergence is the
alternating series test.
70

Exercise 8.6.8. Determine whether the given series converges conditionally, converges abso-
lutely or diverges.
+∞ +∞ +∞ +∞
X cos(nπ) X sin(nπ/2) X (−1)n+1 X (−1)n+1
(a) 2+1
(b) (c) √ (d) √
n=0
n n=0
n + 1 n=1
n n=1
n n
+∞ +∞ +∞ +∞ √
X (−1)n+1 X
n+1 e
1/n X
n+1 n
n X
n+1 n
(e) (f) (−1) (g) (−1) (h) (−1)
n=1
np n=1
n n=1
n! n=1
n+1
+∞ +∞ +∞ +∞
X (−1)n+1 X ln n X X n+1
(i) (j) (−1)n+1 (k) (−1)n+1 e1/n (l) (−1)n+1 ln
n=2
ln n n=1
n n=1 n=1
n
In problem (e) determine all the values of p for which the series converges absolutely, converges
conditionally and diverges.
Exercise 8.6.9. Determine whether the given series converges conditionally, converges abso-
lutely or diverges.
+∞ 2 +∞
n+1 (sin n) 4
X X
(a) (−1) 2
(b) (−1)n+1
n=1
n n=1
2n + 3 + (−1)n
+∞   +∞  
X
n+1 1 X
n+1 1
(c) (−1) cos (d) (−1) sin
n=1
n n=1
n

9 Series of functions
9.1 Power Series
The most important series is the geometric series:
+∞
X
2 3 n
a+ ar + ar + ar + ···+ ar + ··· = a rn .
n=0

If −1 < r < 1 the geometric series converges. Moreover, we proved

+∞
X a
a rn = a + a r + a r2 + a r3 + · · · + a rn + · · · = for − 1 < r < 1. (9.1.1)
n=0
1−r

Replacing r by x and letting a = 1 we can rewrite the formula in (9.1.1) as

+∞
X 1
xn = 1 + x + x2 + x3 + · · · + xn + · · · = for − 1 < x < 1. (9.1.2)
n=0
1−x

The formula (9.1.2) can be viewed as a representation of the function

1
f (x) = , −1 < x < 1,
1−x
as an infinite series of powers of x: 1 = x0 , x, x2 , x3 , . . .:
+∞
1 X
= 1 + x + x2 + x3 + · · · + xn + · · · = xn for −1 < x < 1.
1−x n=0
 71

You will agree that the (non-negative) integer powers of x are very simple functions. There-
fore, it is natural to explore the following question:

Which functions can be represented as infinite series of

Q1:
constant multiples of (non-negative) integer powers of x?

In other words: Which functions x 7→ f (x) can be represented as

+∞
X
2 3 n
f (x) = a0 + a1 x + a2 x + a3 x + · · · + an x + · · · = an xn for ? < x < ? .
n=0

The infinite series

+∞
X
a0 + a1 x + a2 x2 + a3 x3 + · · · + an xn + · · · = an xn (9.1.3)
n=0

is called a power series.

The first question to answer about a power series is:

Q2: For which real numbers x does the power series converge?

Since we are working with the powers of x and since there is no restriction on the signs of an
and x, we can use Theorems 8.6.6 and 8.6.7 (the ratio and root test) to determine the absolute
convergence of the power series (9.1.3). To apply Theorem 8.6.6 we calculate

|an+1 | |x|n+1 |an+1 | |x| |an+1 |

lim n
= lim = |x| lim .
n→+∞ |an | |x| n→+∞ |an | n→+∞ |an |

Assume that
|an+1 |
lim = L. (9.1.4)
n→+∞ |an |

If L = 0, then Theorem 8.6.6 implies that the series (9.1.3) converges for all real numbers x. If
L > 0, then Theorem 8.6.6 implies that the series (9.1.3)
1 1
converges absolutely for |x| L < 1, that is for −
<x<
L L
1 1
diverges for |x| L > 1, that is for x < − or x >
L L
If the limit in (9.1.4) does not exist, then no conclusion about the convergence or divergence can
be deduced.
To apply Theorem 8.6.7 we calculate
p p
lim n |an | |x|n = |x| lim n |an | .
n→+∞ n→+∞
72

Assume that p
n
lim |an | = L. (9.1.5)
n→+∞

If L = 0, then Theorem 8.6.7 implies that the series (9.1.3) converges for all real numbers x. If
L > 0, then Theorem 8.6.7 implies that the series (9.1.3)
1 1
converges absolutely for |x| L < 1, that is for −<x<
L L
1 1
diverges for |x| L > 1, that is for x < − or x >
L L
If the limit in (9.1.5) does not exist, then no conclusion about the convergence or divergence can
be deduced.
Example 9.1.1. Consider the power series
∞
1 1 1 1 1 X 1
+ x + x2 + x3 + · · · + xn + · · · = xn .
0! 1! 2! 3! n! n=0
n!

In this example an = 1/n!, n = 0, 1, 2, . . .. We calculate

1
|an+1 | (n+1)! 1
L = lim = lim 1 = lim = 0.
n→+∞ |an | n→+∞ n→+∞ n + 1
n!

Consequently the given power series converges absolutely for every x ∈ R.

Example 9.1.2. Consider the power series
∞
X
2 3 n
1 + 2x + 3 x + 4 x + · · · + (n + 1) x + · · · = (n + 1) xn .
n=0

Here an = n + 1, n = 0, 1, 2, . . . and we calculate

|an+1 | n+2
L = lim = lim = 1.
n→+∞ |an | n→+∞ n + 1

Consequently the given power series converges absolutely for every x ∈ (−1, 1). Clearly the series
diverges for x = −1 and for x = 1.
Example 9.1.3. Consider the power series
∞
1 2 1 3 1 4 1 X 1
x− x + x − x + · · · + (−1)n+1 xn + · · · = (−1)n+1 xn .
2 3 4 n n=0
n

Here a0 = 0 and an = (−1)n+1 1/n, n = 1, 2, . . .. We calculate

1
|an+1 | n+1 n
L = lim = lim 1 = lim = 1.
n→+∞ |an | n→+∞ n→+∞ n + 1
n

Consequently the given power series converges absolutely for every x ∈ (−1, 1). Clearly the series
diverges for x = −1 and converges conditionally for x = 1.
 73

Example 9.1.4. Consider the power series

∞
1 1 1 1 X 1
1 + x + 2 x2 + 3 x3 + · · · + n xn + · · · = n
xn . (9.1.6)
2 2 2 2 n=0
2

Here an = 2−n , n = 0, 1, 2, . . .. We calculate

1
|an+1 | 2n+1 1 1
L = lim = lim 1 = lim = .
n→+∞ |an | n→+∞ n→+∞ 2 2
2n

Consequently the given power series converges absolutely for every x ∈ (−2, 2). Clearly the series
diverges for x = −2 and for x = 2.
Notice that we can actually calculate the sum of this series using the following substitution
(or you can call this a trick). Substitute u = x/2 in (9.1.6). Then (9.1.6) becomes
∞
X
2 3 n
1+u+u +u +···+ u +··· = un . (9.1.7)
n=0

We know that the sum of the series in (9.1.7) is 1/(1 − u) for u ∈ (−1, 1), that is,
∞
X 1
1 + u + u2 + u3 + · · · + un + · · · = un = , u ∈ (−1, 1).
n=0
1−u

Substituting back u = x/2 we get:

∞
1 1 1 1 X 1 2
1 + x + 2 x2 + 3 x3 + · · · + n xn + · · · = n
xn = , x ∈ (−2, 2).
2 2 2 2 n=0
2 2−x

Example 9.1.5. Consider the power series

∞
1 1 1 1 X 1
x + x2 + x3 + · · · + 2 xn + · · · = 2
xn .
1 4 9 n n=1
n

We calculate
1
|an+1 | (n+1)2 n2
L = lim = lim 1 = lim = 1.
n→+∞ |an | n→+∞ n→+∞ (n + 1)2
n2
Consequently P the given power series converges absolutely for every x ∈ (−1, 1). For x = 1 we
get the series ∞ 1
n=1 n2 . Therefore, for x = 1 the given power series converges. For x = −1 we
get the alternating series which converges absolutely. Therefore the given power series converges
absolutely on [−1, 1].
The following theorem answers the question Q2 above.
Theorem 9.1.6. Let
+∞
X
a0 + a1 x + a2 x2 + a3 x3 + · · · + an xn + · · · = an xn
n=0

be a power series. Then one of the following three cases holds.

74

(A) The power series converges absolutely for all x ∈ R.

(B) There exists r > 0 such that the power series converges absolutely for all x ∈ (−r, r) and
diverges for all x such that |x| > r.

(C) The power series diverges for all x 6= 0. For x = 0 it is trivial that the power series
converges.
The set on which a power series converges is called the interval of convergence. The number
r > 0 in Theorem 9.1.6 (B) is called the radius of convergence. In the case (A) in Theorem 9.1.6
we write r = +∞. In the case (C) in Theorem 9.1.6 we write r = 0.
Remark 9.1.7. In the case (B) in Theorem 9.1.6 the convergence of the power series at the points
x = r and x = −r must be determined by studying the infinite series
+∞
X +∞
X
n
an r and an (−r)n .
n=0 n=0

A review of the examples in this section shows that the interval of convergence of a power series
can have any of these four forms (−r, r), (−r, r], [−r, r) and [−r, r].

9.2 Functions Represented as Power Series

The following theorem lists properties of functions defined by a power series.
Theorem 9.2.1. Let I be the interval of convergence of the power series
+∞
X
2 3 n
a0 + a1 x + a2 x + a3 x + · · · + an x + · · · = an xn .
n=0

Assume that I does not consist of a single point. Then the function f defined on I by
+∞
X
2 3 n
f (x) := a0 + a1 x + a2 x + a3 x + · · · + an x + · · · = an xn , x ∈ I,
n=0

has the following three properties.

(a) The function f is continuous on I.

(b) The function f is differentiable at all interior points of I. Moreover,

+∞
X
′ 2 n−1 n
f (x) = a1 + 2a2 x + 3a3 x + · · · +nan x + (n + 1)an+1 x + · · · = (n + 1)an+1 xn ,
n=0
for all x ∈ I excluding the endpoints (if any) of I.

(c) The function f has derivatives of all orders 1, 2, 3, . . ., at all interior points of I. In particular

f (0) = a0 , f ′ (0) = a1 , f ′′ (0) = 2 a2 , f ′′′ (0) = 3 · 2 a3 , . . . , f (n) (0) = n! an , . . . . (9.2.1)

 75

(d) If x ∈ I, then
Z x +∞
a1 a2 an−1 n an X an−1
f (t)dt = a0 x + x2 + x3 + · · · + x + xn+1 + · · · = xn .
0 2 3 n n+1 n=1
n

Example 9.2.2. By (9.1.2) we have

1
= 1 + x + x2 + x3 + · · · + xn + · · · for − 1 < x < 1. (9.2.2)
1−x
Thus the function f (x) = 1/(1 − x) defined for x ∈ (−1, 1) can be represented by a power series.
Applying Theorem 9.2.1 we get
1
2
= 1 + 2 x + 3 x2 + 4 x3 + · · · + n xn−1 + (n + 1) xn + · · · for −1 < x < 1.
(1 − x)

Example 9.2.3. Substituting −x for x in (9.2.2) we get

1
= 1 − x + x2 − x3 + · · · + (−1)n xn + · · · for −1 < x < 1. (9.2.3)
1+x
Thus the function f (x) = 1/(1 + x) defined for x ∈ (−1, 1) can be represented by a power series.
Applying Theorem 9.2.1 (d) we get
Z x
1 1 1 1 1
ln(1 + x) = dt = x − x2 + x3 − x4 + · · · + (−1)n+1 xn + · · · for − 1 < x < 1 .
0 1+t 2 3 4 n

For x = 1 the above series is an alternating harmonic series which converges conditionally. Thus
we found a power series representation for the function ln(1 + x) on the interval (−1, 1]. By
Theorem 9.2.1 (a) this implies that the sum of the alternating harmonic series is ln 2.

Example 9.2.4. Substituting x2 for x in (9.2.3) we get

1
= 1 − x2 + x4 − x6 + · · · + (−1)n x2n + · · · for − 1 < x < 1.
1 + x2

Thus the function f (x) = 1/(1 + x2 ) defined for x ∈ (−1, 1) can be represented by a power series.
Applying Theorem 9.2.1 (d) we get
Z x
1 1 1 1 1
arctan(x) = 2
dt = x− x3 + x5 − x7 +· · ·+(−1)n+1 x2n−1 +· · · for −1 < x < 1 .
0 1 + t 3 5 7 2n − 1

For x = 1 the above series is a conditionally convergent alternating series. Moreover,

π 1 1 1 1
= arctan 1 = 1 − + − + · · · + (−1)n+1 +···
4 3 5 7 2n − 1
Thus we have a power series representation for the function arctan(x) on the interval (−1, 1].
76

9.3 Taylor series at 0 (Maclaurin series)

In the preceding section we found power series representations for several well known functions.
It turns out that all well known functions can be represented as power series. The key step
in finding the power series representation of elementary functions are formulas (9.2.1) which
establish the relationship between the coefficients an , n = 0, 1, 2, . . . , of a power series and the
derivatives of the function f which is represented by that power series. We rewrite formulas
(9.2.1) as
1 ′′ 1 (3) 1
a0 = f (0), a1 = f ′ (0), a2 = f (0), a3 = f (0), . . . , an = f (n) (0), . . . . (9.3.1)
2! 3! n!
Let a > 0 and let f be a function defined on (−a, a). Assume that f has all derivatives on
(−a, a). Then the series power series
+inf ty
′ 1 ′′ 2 1 (3) 3 1 (n) n
X 1
f (0) + f (0)x + f (0)x + f (0)x + · · · + f (0)x + · · · = f (n) (0)xn
2! 3! n! n=0
n!

is called Taylor series at 0 or Maclaurin series of f .

Using formulas (9.3.1) it is not difficult to calculate a Maclaurin series for a given function.
The difficulties arise in proving that the function defined by such power series is identical to the
given function. Fortunately this is true for all well known functions.

Example 9.3.1. Let f (x) = ex = exp(x), x ∈ R. Then f (n) (x) = ex for all n = 0, 1, 2, . . ..
Therefore the coefficients of the Maclaurin series for the function exp are an = 1/n! and it can
be proved that for all x ∈ R we have
1 2 1 3 1
ex = 1 + x + x + x + · · · + xn + · · · .
2! 3! n!
Example 9.3.2. Let f (x) = sin(x), x ∈ R. Then

f ′ (x) = cos(x), f ′′ (x) = − sin(x), f (3) (x) = − cos(x), f (4) (x) = sin(x).

Consequently,
f (2k) (0) = 0, f (2k+1) (0) = (−1)k , k = 0, 1, 2, . . . .
Therefore the coefficients of the Maclaurin series for the function sin are
1
a2k = 0, a2k+1 = (−1)k , k = 0, 1, 2, . . . .
(2k + 1)!
It can be proved that for all x ∈ R we have
1 3 1 5 1 1
sin(x) = x − x + x − x7 + · · · + (−1)k x2k+1 + · · · .
3! 5! 7! (2k + 1)!

Example 9.3.3. Let f (x) = cos(x), x ∈ R. Then

f ′ (x) = − sin(x), f ′′ (x) = − cos(x), f (3) (x) = sin(x), f (4) (x) = cos(x).
 77

Consequently,
f (2k) (0) = (−1)k , f (2k+1) (0) = 0, k = 0, 1, 2, . . . .
Therefore the coefficients of the Maclaurin series for the function cos are
1
a2k = (−1)k , a2k+1 = 0, k = 0, 1, 2, . . . .
(2k)!

It can be proved that for all x ∈ R we have

1 2 1 4 1 1 2k
cos(x) = 1 − x + x − x6 + · · · + (−1)k x +··· .
2! 4! 6! (2k)!

Example 9.3.4 (The Binomial Series). Let α ∈ R. Let f (x) = (1 + x)α , x ∈ (−1, 1). Then

f ′ (x) = α(1 + x)α−1 ,

f ′′ (x) = α(α − 1)(1 + x)α−2 ,
f (3) (x) = α(α − 1)(α − 2)(1 + x)α−3 ,
..
.
f (n) (x) = α(α − 1) · · · (α − n + 1)(1 + x)α−n
..
.

Therefore the coefficients of the Maclaurin series for the function f are

α(α − 1) · · · (α − n + 1)
a0 = 1, an = , n ∈ N.
n!
It can be proved that for all x ∈ (−1, 1) we have

α α(α − 1) 2 α(α − 1)(α − 2) 3 α(α − 1) · · · (α − n + 1) n

(1 + x)α = 1 + x+ x + x +···+ x +··· .
1! 2! 3! n!
This series is called binomial series. The reason for this name is that for α ∈ N the binomial
series becomes a polynomial:

(1 + x)1 = 1 + x
(1 + x)2 = 1 + 2x + x2
(1 + x)3 = 1 + 3x + 3x2 + x3
(1 + x)4 = 1 + 4x + 6x2 + 4x3 + x4
(1 + x)5 = 1 + 5x + 10x2 + 10x3 + 5x4 + x5
(1 + x)6 = 1 + 6x + 15x2 + 20x3 + 15x4 + 6x5 + x6
..
.
m    
m
X m k m m!
(1 + x) = x , were m ∈ N and =
k=0
k k k!(m − k)!
78

The last formula is called the binomial theorem. The coefficients

 
m m! m(m − 1) · · · (m − k + 1)
= = with m, k ∈ N, 0 ≤ k ≤ m,
k k!(m − k)! k!
are called binomial coefficients. With a general α ∈ R and k ∈ N the coefficients
 
α α(α − 1) · · · (α − k + 1)
:=
k k!


are called generalized binomial coefficients. By definition α0 = 1. With this notation the
binomial series can be written as
+∞  
α
X α
(1 + x) = xk for x ∈ (−1, 1). (9.3.2)
k
k=0

Notice that formula (9.2.2) is a special case of (9.3.2), since

 
−1 (−1)(−2) · · · (−1 − k + 1) (−1)k k!
= = = (−1)k .
k k! k!

Notice also that differentiating (9.2.2) leads to

+∞
X
−2
(1 + x) =1+ (−1)k (k + 1) xk for −1 < x < 1.
k=1

This is a binomial series with α = −2. To verify this we calculate

 
−2 (−2)(−3) · · · (−2 − k + 1) (−1)k (k + 1)!
= = = (−1)k (k + 1).
k k! k!

For α = 1/2 the expression

  1




1/2 2
− 12
− 23 · · · 12 − k + 1
=
k k!
1

3


2
− 2 − 2 · · · − 2k−3
1
2
=
k!
k−1
(−1) 1 · 3 · · · · (2k − 3)
=
2k k!
Thus
√ 1 1 1·3 1·3·5 4 1·3·5·7 5
1 + x = 1 + x − 2 x2 + 3 x3 − 4 x + x +··· for −1 < x < 1.
2 2 2! 2 3! 2 4! 25 5!
Example 9.3.5. Let f (x) = arcsin(x), x ∈ [−1, 1]. To calculate the Maclaurin series for arcsin
we notice that
d
1
arcsin(x) = √ , x ∈ (−1, 1).
dx 1 − x2
 79

Now calculate the Maclaurin series for the last function using the binomial series with α = −1/2.
For α = −1/2 and k ∈ N, we calculate
 



−1/2 − 12 − 32 − 52 · · · − 21 − k + 1
=
k k!

5


− 2 − 2 − 2 · · · − 2k−1
1 3
2
=
k!
1 · 3 · · · · · (2k − 1)
= (−1)k
2k k!
Thus
1 1 1·3 1·3·5 3 1·3·5·7 4
√ = 1 − x + 2 x2 + 3 x − x + · · · for − 1 < x < 1 ,
1+x 2 2 2! 2 3! 24 4!
that is,
+∞
1 X 1 · 3 · · · · · (2k − 1) k
√ = 1+ (−1)k k k!
x ,
1+x k=1
2
or using the notation of double factorials
+∞
1 X (2k − 1)!! k
√ =1+ (−1)k x .
1+x k=1
(2k)!!

Substituting −x2 instead of x in the above formula we get

+∞
1 X (2k − 1)!! 2k
√ =1+ x , for − 1 < x < 1.
1 − x2 k=1
(2k)!!
Since Z x
1
√ dt = arcsin(x),
0 1 − t2
integrating the last power series we get
+∞ +∞ 2k


X (2k − 1)!! X
arcsin(x) = x + x2k+1 = k
x2k+1 , for −1<x<1
(2k + 1)(2k)!! 4k (2k + 1)
k=1 k=0

It is interesting to note that the above expansion holds at both endpoints x = −1 and x = 1. To
prove this we need to recall Theorem 9.2.1 (a) and prove that the series
+∞
X (2k − 1)!!
1+
k=1
(2k + 1)(2k)!!

converges. This series converges by The Comparison Test. (Hint: Prove by mathematical
(2k − 1)!! 1
induction that < √ for all k ∈ N.) As a consequence we obtain that
(2k)!! 3
k
+∞ +∞ 2k


X (2k − 1)!! X
k π
1+ = k
= .
k=1
(2k + 1)(2k)!! k=0 4 (2k + 1) 2