Power Series

In this section, we consider a special type of series of functions, called the power series, and
study its properties.
P
Definition 1. A series of real functions fn is said to be a power series around x = c if the
function fn has the form
fn (x) = an (x − c)n ,
where an , c ∈ R and n = 0, 1, 2, . . . .

Since the translation x0 = x − c reduces a power series around c to a power series around 0, we
only treat the case when c = 0 to simplify notation (wlog). Thus whenever we refer to a power
series, we shall mean a series of the form
∞
X
an xn = a0 + a1 x + · · · + an xn + · · · . (1)
n=0

Remark 0.1. The set on which a power series converges may be small, medium or large, that is,
even if the functions appearing in (1) are defined for all reals, it should not be expected that the
series in (1) converges for allx ∈ R.
∞
X
Example 1. Consider the series n!xn and let an = n!xn . Then, one has
n=0

an+1 (n + 1)!xn+1
= = |(n + 1)x|,
an n!xn
which increases without bound as n → ∞. That is, the limit of the terms of the series does not
∞
X ∞
X
n
converge to 0. Thus, the series n!x diverges, except when x = 0. If x = 0, then n! · 0 = 0.
n=0 n=0
∞
X
Example 2. Let an = xn and take the series xn . We have
n=0

an+1 xn+1
= = |x|,
an xn
and thus,
an+1 xn+1
lim = lim = lim |x| = |x|.
n→∞ an n→∞ xn n→∞

Hence, by the corollary to the Ratio Test, the series converges when
an+1
lim = |x| < 1.
n→∞ an
∞
X xn xn
Example 3. Let us examine the series . If an = , then we obtain
n! n!
n=0

an+1 |x|
= →0 as n → ∞.
an n+1
This implies that the series is absolutely convergent for all real number. Therefore, it is convergent
for any x ∈ R.
In general, we have the following property.

1
 an xn satisfies exactly one of the following cases:
P
Theorem 1. A power series

1. the series converges for all x;

2. the series converges only for x = 0;
3. there exists R > 0 such that the series converges for all x in (−R, R) (or including the
endpoints) and diverges outside this interval.

The number R mentioned in this theorem is called the radius of convergence. Before we take
its formal definition, let us first have the following.
Definition 2. Let {cn } be a bounded sequence of real numbers. Let

an = inf{ck : k ≥ n}, and bn = sup{ck : k ≥ n}.

We define the limit inferior, denoted by lim inf or lim inf, of the sequence {cn } by
n

lim inf {cn } = lim an = lim inf{ck : k ≥ n},

n n→∞ n→∞

and the limit superior, denoted by lim sup or lim sup, of the sequence {cn } by
n

lim sup{cn } = lim bn = lim sup{ck : k ≥ n}.

n n→∞ n→∞
 ∞
1
Example 4. Consider the sequence . Then,
n n=1
a1 = inf{1, 1/2, 1/3, 1/4, . . .} = 0
a2 = inf{1/2, 1/3, 1/4, 1/5, . . .} = 0
a3 = inf{1/3, 1/4, 1/5, 1/6, . . .} = 0,
and so on. Thus,  
1
lim inf = 0.
n n
Similarly,
b1 = sup{1, 1/2, 1/3, 1/4, . . .} = 1
b2 = sup{1/2, 1/3, 1/4, 1/5, . . .} = 1/2
b3 = sup{1/3, 1/4, 1/5, 1/6, . . .} = 1/3,
and so on. Thus,  
1
lim sup = 0.
n n

Other notations for the limsup and liminf are lim and lim, respectively. We have the following
property for these numbers.
Proposition 1. If {an } is a bounded sequence of real numbers, then lim inf{an } ≤ lim sup{an }.
Moreover, the sequence is convergent if and only if lim sup{an } = lim inf{an } = L and for this
case, lim an = L.
Proposition 2. Let {an } and {bn } be bounded sequences of real numbers such that bn → b for
some b ≥ 0. Then
lim sup{an bn } = b lim sup{an },
and
lim inf{an bn } = b lim inf{an }

2
 an xn .
P
Definition 3. Consider the power series
1. The radius of convergence R is defined by
1 .
n 1
o
R= where ρ = lim sup |an | n . (2)
ρ
We take 

 0 if ρ = +∞,
. 1

R= if 0 < ρ < +∞, (3)


 ρ
+∞ if ρ = 0.
The radius of convergence R is also given by
|an |
R = lim .
|an+1 |
2. The interval of convergence is the largest interval in which the series converges.

Let us revisit the series given in Examples 1-3, and prove the obtained results by finding the
radii of convergences.
Example 5. Let us take the series given in Example 1. We have an = n! and so
|an | 1
R = lim = lim = 0.
|an+1 | n+1
Thus, the series converges only when x = 0.
Example 6. From Example 2, an = 1 so that
|an |
R = lim = 1.
|an+1 |
Thus, the series converges when |x| < 1.
Now when x = ±1, the resulting series diverges by the simple test for divergence. Therefore,
the interval of convergence is (−1, 1).
1
Example 7. In Example 3, one has an = . This yields
n!
|an |
R = lim = lim(n + 1) = ∞,
|an+1 |
which means that the series converges for all real numbers. That is, the interval of convergence is
R.

Before we continue to the next theorem, let us have the facts that,
if lim sup{bn } < v, then bn ≤ v for all sufficiently large n ∈ N, (4)
and
if lim sup{bn } > v, then bn ≥ v for infinitely many n ∈ N. (5)
Note that infinitely many n means there is always a bigger number satisfying the same property.
Meanwhile, for all large enough n implies that there is some number such that any larger number,
will satisfy the property. Clearly, “for all large enough n” covers “infinitely many n” but not the
other direction.
Take for example |an xn | ≤ c for sufficiently large n. This implies the existence of some number
N such that for any n ≥ N , the property holds. That is, there is a starting number in which the
property is true. On the other hand, |an xn | ≥ 1 for infinitely many n means there is no particular
number in which the property is attained first.
The next theorems provide more information on the absolute and uniform convergence of the
power series.

3
Theorem 2 (Cauchy-Hadamard Theorem). If R is the radius of convergence of the power series
n
P
an x , then the series is absolutely convergent if |x| < R and divergent of |x| > R.
Proof:
Let us consider the case R < ∞.
Suppose that 0 < |x| < R. By the Density Theorem, there exists R1 ∈ R such that |x| < R1 <
c
R. In particular, one can find a c ∈ (0, 1) such that |x| < cR < R. It follows from (2) that ρ <
|x|
so that if n is sufficiently large, (4) gives us
1 c
|an | n ≤ .
|x|
That is,
|an xn | ≤ cn ,
for all sufficiently large n. The absolute convergence of the series, then, follows from the Comparison
Test.
1
If |x| > R = ρ1 , then we have ρ > . In view of (2) and (5), there are infinitely many n ∈ N
|x|
for which one has
1 1
|an | n > .
|x|
Thus, |an xn | > 1 for infinitely many n, which makes the terms of the series not converging to 0.
Therefore, the series diverges by the simple test for divergence.
The cases when R = 0 and R = +∞ are left as exercises. 
Remark 0.2. The Cauchy-Hadamard Theorem does not give a conclusion when |x| = R. Indeed,
for this case, anything can happen.
Example 8. Take the series
X X1 X 1
xn , xn , and xn .
n n2
1
Since lim n n = 1, observe that the radius of convergence of each of these power series is 1.
P n
(i) The series x converges at neither of the points x = −1 and x = +1.
P1 n
(ii) The series n x converges at x = −1 but diverges at x = +1.

(iii) The third power series converges at both x = −1 and x = +1.

an xn and let K be a closed and bounded in-
P
Theorem 3. Let R be the radius of convergence of
terval contained in the interval of convergence (−R, R). Then, the power series converges uniformly
on K.
Proof:
The hypothesis on K ⊆ (−R, R) implies that there is a positive constant c < 1 such that
|x| < cR for all x ∈ K. Following the arguments used in the proof of Theorem 2, we thus have

|an xn | ≤ cn ,

for all sufficiently large n and for all x ∈ K. Since c < 1, the uniform convergence of the power
.
series on K follows then, from the Weierstrass M-test with Mn = cn . 
We also have the following properties for the integration and differentiation of a power series.
Theorem 4. The limit of a power series is continuous on the interval of convergence. Moreover,
the power series can be integrated term by term over any closed and bounded interval contained in
the interval of convergence.

4
Proof:
If |x0 | < R, the previous theorem asserts that the power series converges uniformly on any closed
and bounded neighborhood of x0 contained in (−R, R). Thus, the limit of the power series is also
continuous at x0 . The term by term integration follows from one of the properties of uniformly
converging series discussed in the previous section. 

Theorem 5 (Differentiation Theorem). A power series can be differentiated term by term within
the interval of convergence. In fact, if
∞
X
f (x) = an xn ,
n=0

then
∞
X
f 0 (x) = (nan xn−1 ),
n=1

for |x| < R. Both series have the same radius of convergence.

Proof: n o
1 1
Consider the sequence |nan | n . Since lim n n = 1, the boundedness of this sequence depends
n 1
o n 1
o 1
on the boundedness of |an | n . In the case that |an | n is bounded, then lim sup{|an | n } exists.
By Propositions 1 and 2, we therefore obtain
n 1
o 1
n 1
o n 1
o
lim sup |nan | n = lim n n lim sup |an | n = lim sup |an | n ,

so that the radius of convergence of the two power series is the same. Thus, the differentiated
series is uniformly convergent on each closed and bounded interval contained in the interval of
convergence.
Hence, applying the property of uniformly convergent series of functions regarding the deriva-
tives, we conclude that the formally differentiated series converges to the derivative of the given
series. 

Remark 0.3. The theorem makes no assertion about the endpoints of the interval of convergence.
If a series is convergent at an endpoint, then the differentiated series may or may not converge at
this point.

Example 9. Consider the power series

∞
X xn
.
n2
n=0

This series is convergent at both endpoints x = −1 and x = +1. However, the differentiated series
given by
∞
X xn−1
,
n
n=1

converges at x = −1 but diverges at x = +1.

Now, in repeating the differentiation, we can conclude that if k is any natural number, the
∞
X
power series an xn can be differentiated term by term k times to get
n=0

∞
X n!
an xn−k . (6)
(n − k)!
n=k

5
Moreover, this series converges absolutely to f (k) (x) for |x| < R and uniformly over any closed and
bounded interval in the interval of convergence If we let x = 0 in (6), one obtains the formula

f (k) (0) = k!ak ,

and have the following result.

an xn and bn xn converge on some interval (−r, r), r >

P P
Theorem 6 (Uniqueness Theorem). If
0, to the same function f , then
an = bn ,
for all n ∈ N.

Proof:
The preceding remark shows that

n!an = f (n) (0) = n!bn ,

for all n ∈ N. 

Let us now consider few examples.

Example 10. Find the radius of convergence for the series
∞
X x x2 x3 x4 x5
an xn = 1 + + 2 + 3 + 4 + 5 + ··· .
3 4 3 4 3
n=0

S olution: Since
1



n
n is even,
an = 4
1
n is odd,


3n
then,
1

1

 n is even,
|an | n = 4
 1

n is odd.
3
This gives
1 1
ρ = lim sup |an | n = ,
3
and therefore the radius of convergence from (2) is R = 3.
∞
X (4x + 3)n
Example 11. Find the radius and interval of convergence of the series .
n3
n=1
S olution: Observe that
∞ ∞ n
(4x + 3)n 4n

X X 3
= x+ ,
n3 n3 4
n=1 n=1

4n
so we have an = . We obtain,
n3
3
4n (n + 1)3

|an | 1 n+1 1
R = lim = lim 3 · n+1
= = .
|an+1 | n 4 4 n 4

Now, by Theorem 2, the series converges absolutely when

3 1 1
x+ < , or i.e., −1<x<− ,
4 4 2

6
 ∞
X (−1)n
and so the series converges for these values of x. Moreover, when x = −1, we get which
n3
n=1
1
converges by the Alternating Series Test (exercise). On the other endpoint x = − we obtain the
2
∞
X 1
series (−1)n n 3 which is again, a convergent series via the Alternating Series Test (another
2 n
n=1
exercise).
∞
X (4x + 3)n 1
Therefore, the radius of convergence of the series 3
is R = while the interval of
n 4
  n=1
1
convergence is −1, − .
2
∞
X 3n + (−2)n n
Example 12. Determine the intervval of convergence of the series x .
n
n=1
S olution: Let us find the radius of convergence first. We note that
  n  1
1 −2 n
lim n = 1 and lim 1 +
n =1 (7)
n
 n
−2
since lim = 0.
n
3n + (−2)n
The given series provides us an = . Computing now the value of ρ in view of (2)
n
and (7), we get
1
ρ = lim |an | n
1
3n + (−2)n n
= lim
n
   n  1
−2 n
3n 1 +
n
= lim 1
nn
  n  1
−2 n
3 1+
n
= lim 1
nn
= 3.
1
Thus, the radius of convergence by (2) is R = . This means, the series converges for all x ∈
  3
1 1
− , .
3 3
1
Let us investigate the convergence of the series at the endpoints. When x = , the resulting
3
∞   n 
X 1 1 2
series is + − . By noting that the sum of a convergent and a divergent series
n n 3
n=1
∞   n 
X 1 1 2
results to a divergent series (exercise), then the series + − is divergent.
n n 3
n=1
n
∞
1 X (−1)n + 23
On the other hand, when x = − , the resulting series is . We observe that the
3 n
n=1
∞ ∞
(−1)n 1 2 n
X X  
series is convergent by applying the Alternating Series Test (exercise), while
n n 3
n=1 n=1
∞ n
X (−1)n + 2


3
converges by Abel’s Test (exercise), then the series is convergent by the recent
n
n=1

7
assignment. 

1 1
Therefore, the interval of convergence of the series is − , .
3 3
Example 13. Find a power series in x converging to ln(1 + x).
∞
X 1
S olution: Note that xn = for |x| < 1. If x = −t, then for | − t| = |t| < 1, one has
1−x
n=0

∞
X 1
(−t)n = .
1+t
n=0

Now, since there is a uniform convergence within the interval of convergence by Theorem 2, we
can apply the properties discussed in the series of functions and obtain
Z x
dt
ln(1 + x) =
0 1+t
Z xX ∞
= (−t)n dt
0 n=0
∞
X Z x
n
= (−1) (t)n dt
n=0 0
∞
X xn+1
= (−1)n ,
n+1
n=0

for |x| < 1.