9 Series Expansions

9.1 Series expansions

• If we are given a function f (x), then the series S0 (x), S1 (x), . . . is a series expansion for
f if
– for each n, Sn (x) is a polynomial of degree n in x
– for each x the seqence Sn (x) is convergent and
– limn→∞ Sn (x) = f (x).
• Some series expansions hold true for every value of x, but others are only valid for x in
a specific range.

Example 9.1
We know that
n
X 1 − xn+1
xi = .
1−x
i=0

As long as |x| < 1, this sequence converges as n → ∞ to a limit 1/(1 − x). So it represents a series
expansion for 1/(1 − x) as long as |x| < 1. Outside this range, the sum does not converge. ♦

9.2 Radius of convergence

• A series expansion {Sn (x)} has radius of convergence ρ if
– the series is convergent whenever |x| < ρ
– the series is divergent whenever |x| > ρ
• Every series expansion has a radius of convergence
– ρ = ∞ if the series converges for all x
– ρ = 0 if the series diverges for all x except x = 0
• In the example of 1/(1 − x), the radius of convergence is ρ = 1.

• Outside the radius of convergence, the function may still be well-defined.

• For example, when x = 2, the function 1/(1 − x) takes the value −1.
• But the series ∞ i
P
i=0 2 is not equal to −1 – in fact, it does not converge to any specific
value.
• This shows that the function f and the series expansion for f are not the same thing,
because they behave differently outside the radius of convergence.

Exercise 9.1
Pn 2 i
What is the radius of convergence, ρ, of the series Sn = i=1 (−2x )?
What is the limit of the series when |x| < ρ?

50
9.3 Repeated differentiation
• We have defined the (first) derivative f 0 of a function f
• When looking at maxima, minima and convexity we also looked at the second derivative
f 00 = dx
d
(f 0 ).
• There is no reason to stop at 2. Many functions can be differentiated over and over
again
d3
– The third derivative f 000 = d
dx
(f 00 ) = dx3
(f )
– Once we get beyond the third derivative we use the notation f (n) to denote the
nth derivative of f .

Example 9.2
If f (x) = exp(x) we know that f 0 (x) = exp(x). Therefore f 00 (x) = exp(x), and in fact f (n) (x) = exp(x)
for every x. ♦
Example 9.3
If f (x) = xr then f 0 (x) = rxr−1 and f 00 (x) = r(r − 1)xr−2 . We can show that
(
r! r−n
(r−n)! x if n ≤ r
f (n) (x) =
0 if n > r

♦
Example 9.4
If f (x) = 1/(1 − x) = (1 − x)−1 , then f 0 (x) = (1 − x)−2 .
Taking this one step further, f 00 (x) = 2(1−x)−3 , and in fact we can show that f (n) (x) = n!(1−x)−n−1 .
♦
Exercise 9.2
What is the nth derivative of 1/(1 + 4x) at x = 0?

9.4 Taylor series

• We saw that a good approximation to f (x) in the region around x0 is the tangent,
S1 (x) = f (x0 ) + (x − x0 )f 0 (x0 ), because
– S1 (x0 ) = f (x0 )
– S10 (x0 ) = f 0 (x0 )
• If we wanted a polynomial which also matched the second derivative of f at x0 , it would
be
1
S2 (x) = f (x0 ) + (x − x0 )f 0 (x0 ) + (x − x0 )2 f 00 (x0 ).
2!

• S0 , S1 and S2 are the first three terms in the Taylor series of f about x0 .
• The nth term in the series is
n
X 1
Sn (x) = (x − x0 )i f (i) (x0 ).
i=0
i!

51
 • For the values of x for which this series converges, it is a series expansion for f .

Example 9.5
Consider the expansion of f (x) = exp(x) about x0 = 1.
We know that f (n) (x0 ) = exp(1) = e for all n. Therefore the series is
1 1
Sn (x) = e + (x − 1)e + (x − 1)2 e + · · · + (x − 1)n e
2! n!
Does this converge? Consider the ratio test.
1 n+1
an+1 (n+1)! (x − 1) x−1
= 1 n
= .
an n! (x − 1)
n+1

As soon as n is larger than 2(1 + |x|), this ratio is less than 21 , so the ratio test tells us that the series
converges for all x. ♦

9.5 Maclaurin series

• A Maclaurin series is a Taylor series with x0 = 0.

Example 9.6
We have seen that the nth derivative of 1/(1 − x) is f (n) (x) = n!(1 − x)−n−1 .
Therefore f (n) (0) = n! and the Maclaurin series for 1/(1 − x) is
∞ ∞
X 1 n X
x · n! = xn .
n!
n=0 n=0

This should not come as a surprise! We already know that the radius of convergence is 1. ♦
Example 9.7
Look at the function f (x) = eλx . Then

f 0 (x) = λeλx
f 00 (x) = λ2 eλx
f (n) (x) = λn eλx

So we see that f (n) (0) = λn and therefore the series is ∞ n n

P
n=0 λ x /n!.

As we saw for the Taylor expansion, this is convergent for all x.

This explains why, in probability theory, the probability function of the Poisson distribution sums to 1:
∞
X λn
e−λ = e−λ · eλ = 1.
n!
n=0

♦
Example 9.8
Consider the function f (x) = ln(1 − x), which satisfies f (0) = 0.
The first derivative is f 0 (x) = −(1 − x)−1 . We can use what we know about the derivatives of (1 − x)−1
to deduce that

f (n) (x) = −(n − 1)!(1 − x)−n ⇒ f (n) (0) = −(n − 1)!.

52
Therefore the Maclaurin expansion is
∞ ∞
X 1 n X xn
0+ x × (−(n − 1)!) = − .
n! n
n=1 n=1
For |x| > 1 the series is divergent as the terms do not converge to 0.
For |x| < 1 the ratio test gives us
|an+1 | n
= |x| < |x|.
|an | n+1
Therefore the series is convergent for −1 < x < 1. ♦
xn /n in an exercise in the previous chapter.
P
• We have seen the series
• We saw in that chapter that the sequence is divergent when x = 1.
• Now we can see that ∞ n
P
n=1 (−1) /n = ln(1 − (−1)) = ln(2).

• So the series converges when x = −1 but not when x = +1.

– We say that the series xn /n is conditionally convergent when |x| = 1, because it
may or may not converge depending on the value of x
– By contrast, xn /n is absolutely convergent when |x| < 1
Exercise 9.3
d d
Recall that dx (sin x) = cos x and dx (cos x) = − sin x. Given that sin(0) = 0 and cos(0) = 1,
calculate the Maclaurin expansion of sin x.

Exercise 9.4
Use the first four terms of the Maclaurin expansion of ln(1 − x) to provde an estimate of
ln(1.12).

• The hyperbolic cosine of x and the hyperbolic sine of x are defined by

1 x 1 x
e + e−x , e − e−x .



cosh(x) = sinh(x) =
2 2
• We can see from the definitions that
cosh0 (x) = sinh(x) and sinh0 (x) = cosh(x).

• If we write down the series expansion of ex and e−x and average each term we see that
∞
X x2n
cosh(x) = .
n=0
(2n)!
So the expansion of cosh uses all the terms in the expansion of ex which correspond to
even powers of x.
• Because cosh(x)+sinh(x) = ex , the expansion of sinh uses all the terms in the expansion
of ex which correspond to odd powers of x:
∞
X x2n+1
sinh(x) = .
n=0
(2n + 1)!

Exercise 9.5
Verify that the Maclaurin expansions of cosh and sinh are the same as the series we have just
obtained.

53
 9.6 l’Hôpital’s rule
• L’Hôpital’s rule is a way of finding the limit of a function which is expressed as a
fraction when both the numerator and denominator are equal to 0.

Proposition 9.1 L’Hôpital’s rule If f and g are differentiable functions with f (x0 ) = g(x0 ) =
0 and if g 0 (x0 ) 6= 0 then
f (x) f 0 (x)
lim = lim 0
x→x0 g(x) x→x0 g (x)

Proof. Using o notation,

f (x0 + h) 0 + hf 0 (x0 ) + o(h) f 0 (x0 ) + o(1)
= =
g(x0 + h) 0 + hg 0 (x0 ) + o(h) g 0 (x0 ) + o(1)
As h → 0 the o(1) terms vanish.
Example 9.9
If we want to know the limit as x → 0 of (ex − 1)/ sin x, we note that both numerator and denominator
are equal to 0 at x = 0.
The derivative of the numerator is ex , that of the denominator is cos(x), so
ex − 1 ex 1
lim = lim = = 1.
x→0 sin x x→0 cos x 1
♦

• But what if g 0 (x0 ) = 0?

• If f 0 (x0 ) 6= 0, we can see that the limit is ±∞, depending on the sign of f 0 (x0 ).
• But if f 0 (x0 ) is also 0, we can do it all again:
f (x) f 00 (x)
lim = lim 00
x→x0 g(x) x→x0 g (x)
and keep doing it until either f (n) (x0 ) 6= 0 or g (n) (x0 ) 6= 0.

Example 9.10 2 2
To find the limit as x → 0 of f (x)/g(x) = (ex − 1)/ sin2 x, note that f 0 (x) = 2xex and g 0 (x) =
2 sin(x) cos(x), both of which are equal to 0 when x = 0. So we look at second derivatives.
2 2
f 00 (x) = 2ex + 4x2 ex , so f 00 (0) = 2.
g 00 (x) = 2 cos2 (x) − 2 sin2 (x), so g 00 (0) = 2.
Therefore 2
ex − 1 2
lim 2 = = 1.
x→0 sin x 2
♦
Exercise 9.6

(i) Find the terms up to the term in x3 in the Maclaurin expansion of e2x ln(1 + 4x).
(ii) Calculate
e2x ln(1 + 4x)
lim
x→0 sin(x)

54