Chapter 7

Series

The final topic we study is infinite sums. These will be key in MATH102 where you will study
Taylor series of functions.

Given a sequence (an ) the associated is the formal infinite sum

∞
X
= an .
n=0

We call this a formal sum because it may not be possible to make sense of it as a real number. The
(finite) sum
X k
= an
n=0

of the first k + 1 terms is said to be the . We say the series

if the sequence S0 , S1 , S2 , . . . of partial sums to a

finite limit, and we write

∞
X
an =
n=0

Otherwise we say the .

P∞
Remark 7.1. If the series converges then n=0 an is a ; if it

diverges it is just a

106
Warning 7.2. The convergence of the is not the same as the convergence

of the . Very roughly, the series converges if the sequence converges

sufficiently quickly to .

Examples 7.3. (a) The series 1 + 1 + 1 + · · · because the kth partial sum is

, which as k → ∞.

(b) The series 1 − 1 + 1 − 1 + · · · because the sequence of partial sums is

which is .

(c) A series of the form 1+r+r2 +· · · is called a .

For instance is a geometric series with ratio 1/2. Let

Sk =

be the kth partial sum. Then

(1 − r)Sk =

so that Sk = . If then Sk → as
k → ∞, i.e. in this situation the series converges, and

∞
X
rn = .
n=0

For instance 1 + 1/2 + 1/4 + · · · = . However, if then

the geometric series diverges: the case is immediate from the explicit formula for
the partial sums, the cases r = ±1 are the first two examples above.

Remark 7.4. The convergence of a series depends only on the ‘tail’ of the series and not on the
values of any finite number of initial terms , i.e.

∞
X
⇐⇒ an converges.
n=N

107
For this reason, we only ever need to consider the of a series when testing.

P∞
Theorem 7.5. If n=0 an converges then limn→∞ an = .

Proof. Let Sk be the kth partial sum. By assumption (Sk ) so limk→∞ Sk =

say. Then limk→∞ Sk−1 = . Moreover Sk −Sk−1 = . Hence,

by the Algebra of Limits limk→∞ = limk→∞ as required.

P∞
Therefore if limn→∞ an or does not exist we can immediately conclude that n=0 an

. Note however, that if limn→∞ an = we

that the associated series .

There are a number of important tests which help us to decide if a given series converges of not.
Using the integral test we will prove

∞
X
np converges ⇐⇒ .
n=1

P∞
The borderline case p = −1 is very important; the series n=0 is called the

, and it is a key fact that it .

The first test compares one series with another, whose behaviour is already known.

Theorem 7.6 (Comparison Test). Suppose that there is some N ∈ N such that

and ∞
P P∞
for all N bn converges. Then n=0 an converges.

PN +k PN +k
Proof. Let Sk = n=N an and Tk = n=N bn . Then

as an for . Also (which ex-

ists by assumption). Hence (Sk ) is an sequence,

108
and therefore converges.

P∞ 2
Example 7.7. We know from above that n=1 1/n (it is the case p = −2).
Hence
∞
X 1
n2 + 1
n=1

too by the Comparison Test because 0 ≤ for

any n.
P∞
A consequence of the Comparison Test is that if 0 ≤ an ≤ bn for n ≥ N and n=0 an diverges
then ∞
P
n=0 bn must (because if it then the above

result would say that ∞

P
n=0 an ).

P∞ ln n
Example 7.8. The series n=1 n by the Comparison Test: we com-

pare with the which is

and use the fact that ln(n) for n ≥ e.

Definition 7.9. The series ∞ if ∞

P P
n=0 an n=0 |an |
converges, i.e. if the series whose terms are the absolute values of the original series converges.
P∞
Theorem 7.10 (Absolute Convergence Test). If n=0 an converges absolutely then it converges.

Pk
Proof. Let Sk = n=0 an and also let

Pk = and Qk =

be respectively the partial sums of the positive and negative terms in the series. Clearly (Pk ) is an
P∞
n=0 |an |
— which exists by the assumption that the series converges absolutely — and (Qk ) is a
− ∞
P
n=0 |an |.

Hence (Pk ) and (Qk ) as k → ∞. Thus (Sk ) =

109
 P∞ (−1)n
Example 7.11. The series n=1 n2
by the Absolute Convergence Test

since .

The next result is really a special case of the Comparison Test in which we compare with a geometric
series, but it is so useful that it is worth stating as a separate result.

Theorem 7.12 (Ratio Test). Suppose there is some N ∈ N such that

 
 an+1 

 an 
 for n ≥ N.

P∞
Then n=0 an .

Proof. Since |an+1 /an | for n ≥ N we have

Therefore, by comparison with the geometric series

P∞ P∞
we see that n=N |an | . Hence n=0 an

Often it is easier to understand the limiting behaviour of the ratio |an+1 /an |. Therefore the next
version of the Ratio Test is the most commonly used in practice.
P∞
Theorem 7.13 (Limit Ratio Test). Suppose that |an+1 /an | → L as n → ∞. Then n=0 an

converges absolutely if and diverges if . (If L = then we cannot conclude

anything about the series from this test.)

Proof. If L < 1 then one can find and such that |an+1 /an | for

. It follows that ∞
P
n=0 an converges absolutely by the Ratio Test I.

If L > 1 then one can find and such that |an+1 /an | for .

Then by comparison with a geometric series of ratio r (which )

we see that ∞
P P∞
n=N an , and hence n=0 an , .

110
 P∞ 1
Examples 7.14. (a) The series n=0 n! by the Ratio Test since

as n → ∞.

P∞ xn
(b) The series n=1 n by the Ratio Test because

as n → ∞. The Ratio Test also shows that the series diverges whenever |x| > 1. However,
it cannot be used to decide the two cases x = ±1. In fact, when x = 1 the series is the
harmonic series, which diverges by the Integral Test, and when x = −1 the series converges by
the Alternating Series Test.

Sometimes it can be difficult to know which convergence test to apply, at least until you have more
experience with series. To help there is a flow-chart in Figure 7.1 which gives you a fixed procedure
for tackling series. The advantage that this will always work, provided the convergence of the series
can be decided using the tests we have seen, but the disadvantage is that following the procedure
may take more time than just ‘seeing’ which test to use. Here are some further examples.

P∞
Examples 7.15. (a) n=1 ln(n) .

P∞ 1
(b) n=0 4n

P∞ 1
(c) n=1 n3

P∞ ln(n)
(d) n=1 2n

as n → ∞. Here we have used the fact that

ln(n + 1) =
in computing the limiting ratio of terms.

111
 P∞ n
(e) n=0 2n

as n → ∞.

P∞ cos(n)
(f) n=0 en

P∞ 3n
(g) n=0 n!

as n → ∞.

P∞ nn
(h) n=0 n!

as n → ∞.

7.1 Power Series

is one of the form ∞
P
A n=0 , i.e. containing a
power of a variable, in this case x. We would like to know for which x ∈ R it converges. The
standard technique is to first use the Ratio Test. Suppose |an+1 |/an | → L as n → ∞. Then

 an+1 xn+1 
 
 an xn  =
 

P∞ n
Hence n=0 an x whenever ,

and whenever . The borderline cases x = must be

checked separately using other convergence tests.

112
 P∞ n
Examples 7.16. (a) n=0 x converges absolutely for , and diverges for .

When we have the series and which each

P∞ n /n
(b) n=1 x converges absolutely for , and diverges for . When

we have the series which ,

and when the series which

P∞ n /n2
(c) n=1 x converges absolutely for , and diverges for . When

we have the series which ,

and when the alternating series which

P∞ n /n!
(d) Applying the Ratio Test to n=0 x we compute

as n → ∞. Hence this series converges absolutely for ; in MATH102

you will see that ∞ n

P
n=0 x /n! = .

P∞ n
(e) Applying the Ratio Test to n=0 n!x we compute

as n → ∞ except when . Hence this series diverges for

To summarise, in the above examples we see that there is a

r such that the power series converges when , diverges when , and can be-

have in any fashion when , by which we mean it may converge at both points x = ,
or diverge at both, or converge at one and diverge at the other. The extreme cases r = 0 and r = ∞
may both occur.

113
7.2 Other tests for convergence of series

Theorem 7.17 (Integral Test). Suppose there is some N ∈ N such that f (x) is ,

and for . Then

i.e. if and only if the

Proof. The idea is that the partial sums can be bounded above by integrals, and vice versa. More
concretely, considering the two diagrams below (which are drawn for the case N = 1 for the sake
of simplicity) we obtain the inequalities

y = f (x) y = f (x)

f (1)
f (2) f (1)
f (3) f (2)
f (3)

1 2 3 1 2 3 4

(Note that the estimates depend crucially on the assumption that f is

R∞
and .) Thus, if N f (x) dx converges then Sk is and

Hence Sk .

114
 R N +k
Conversely, if Sk converges then the integral N f (x) dx with k and is

. It follows that the

R∞
infinite integral N f (x) dx converges.

P∞ p.
Example 7.18. Consider the series n=1 n For p ≥ 0 it is easy to check that the series diverges.
So we assume p < 0 from now on.

Let f (x) = xp . For x > 0 this is , and f 0 (x) = . So we

can apply the Integral Test and the series converges if and only if

(7.1)

For p 6= −1 we have so that the integral in (7.1)

is When p = −1

and the integral in (7.1) is . Putting this together we conclude

∞
X
np converges ⇐⇒ .
n=1

P∞ 1
The borderline case p = is the harmonic series n=0 n , which diverges.

Although it is not true that a series whose terms tend to zero converges (e.g. the harmonic series
P∞ 1
n=1 n diverges), it is true that a series whose terms

Theorem 7.19 (Alternating Series Test). Suppose that an → 0 as n → ∞ and that there is some
N ∈ N such that an is positive and decreasing for n ≥ N . Then

115
Proof. Consider the partial sums Sk = . Since the an are
positive and decreasing

S2k+2 =

and S2k+3 = . Also S2k+3 =

. Putting these inequalities together gives

It follows that the partial sums with index form a

sequence bounded below by and that the partial sums with

index form an sequence bounded above by . Hence

converge, and since |S2k+1 − S2k | = as k → ∞ they

the same limit. Thus (Sk ) converges to that limit too.

P∞ (−1)n
Example 7.20. (a) The series n=1 n converges by the Alternating Series Test since the

P∞ (−1)n
(b) n=0 ln(n) converges by the Alternating Series Test because

P∞ 1
(c) n=2 n ln(n) diverges by the Integral Test because f (x) = 1/x ln(x) is

as R → ∞. (To see that f is decreasing we can either compute the derivative

or note that x 7→ x ln x is increasing.)

116