TAYLOR SERIES, POWER SERIES

The following represents an (incomplete) collection of things that we covered on the sub-
ject of Taylor series and power series.
Warning. Be prepared to prove any of these things during the exam. Things you should
memorize:
• the formula of the Taylor series of a given function f (x)
1
• geometric series (i.e. the Taylor expansion of 1−x )
x
• the Taylor expansions of the functions e , sin x, cos x, ln(1 + x) and range of va-
lidity.
• the relation f (x) = Pn (x) + Rn (x) and Lagrange formula for Rn (x)
You should also understand the actual proofs of the Taylor series expansions enumerated
above.

1. TAYLOR S ERIES
P∞ (n)
f (x) = n=0 f n!(0) xn ⇔ Rn (x) → 0
In other words, the Taylor expansion takes place only at those values of x for which
Rn (x) → 0. If you want to prove from scratch a Taylor series expansion (as we did in
the case of ex , cos(x), sin(x) and ln(1 + x)) you need to show Rn (x) → 0, and one
usually proves this by
• employing Lagrange formula
• estimating Rn (x) (get rid of c)
See the slides of Nov 24 lecture.
Exponential function.
∞
X xn
ex = , ∀x ∈ R
n=0
n!
Understand why this gives, among others, the following formula
∞ ∞
X 1 √ X (−1)n 1
n
= e, =
n=0
2 n! n=0
n! e
Cosine. Know how to estimate the remainder in this case to prove
∞
X x2n
cos x = (−1)n , ∀x ∈ R
n=0
(2n)!
In particular this gives (set x = 1/2)
∞
X (−1)n
= cos 0.5
n=0
4n (2n)!
Sine.
∞
X x2n+1
sin x = (−1)n , ∀x ∈ R
n=0
(2n + 1)!
1
2 TAYLOR SERIES, POWER SERIES

Hyperbolic Cosine. From

∞
X xn
ex = and
n=0
n!
∞
X xn
e−x = (−1)n
n=0
n!
one derives
∞
ex + e−x X x2n
= , x∈R
2 n=0
(2n)!
If we set x = 1 we obtain form example
∞
X 1 e + 1/e
=
n=0
(2n)! 2
as opposed to
∞
X 1
= e and
n=0
n!
∞
X (−1)n
= cos(1)
n=0
(2n)!
Geometric series.
∞
1 X
= xn holds only for − 1 < x < 1
1 − x n=0
Logarithm. Start with the fake geometric series
∞
X 1
(−1)n xn =
n=0
1+x
Integrate (apply the nice theorem on power series):
∞ ∞
X xn+1 X xn
ln(1 + x) = (−1)n = (−1)n+1 , x ∈ (−1, 1)
n=0
n + 1 n=1 n
If we want to justify this identity in the range S = (−1, 1], we need to appeal to Abel’s
theorem. In particular, for x = 1 we get
∞
X (−1)n+1 1 1 1 1
ln(2) = = 1 − + − + ...
n=1
n 2 3 4 5
Equivalently, we have the following power expansion in x − 1
∞
X (−1)n+1
ln(x) = (x − 1)n valid for 0 < x ≤ 2
n=1
n
Approximate Computations. Starting with f (x) = Pn (x) + Rn (x) for a given f (x),
one can presumably find n such that Rn (x) is smaller than the desired degree of accuracy
(estimate Rn (x)!) in order
√ to know that Pn (x) approximates f (x) well enough. Examples:
computing e0.2 , 1/e, 1/ e, sin 0.5 to three decimal places (i.e. approximate the function
by an appropriate Taylor polynomial, etc.)
 TAYLOR SERIES, POWER SERIES 3

Example not done in class: compute ln(1.4) to 2 decimal places by approximating the
function ln(1 + x) by Taylor polynomial.

2. P OWER S ERIES
P∞
Given a power series n=0 an xn , one can determine:
• The radius of convergence R ≥ 0 with the formula
1
= lim |an |1/n
R n→∞
• The domain of convergence S which consists of all the numbers x for which the
series an xn is convergent: the open interval (−R, R) is for sure included, and
P
then we only have to check the endpoints x = ±R separately. The power series is
divergent outside this range, i.e. for |x| > R.
Example. Find the radius of convergence R and the domain of convergence S for each of
the following power series:
∞ ∞ ∞ ∞ ∞ ∞
X X xn X xn X n n X xn X (−1)n 2n
xn , , , n x , , x
n=0 n=1
n n=0 nn n=0 n=0
n! n=0 n2
P∞ P∞
Hwk problem: if the series k=0 4n an is convergent, then n=0 an (−2)n is also con-
P reduces to understanding the shape of the domain of convergence S
vergent. (the question
of the power series an xn )
2.1. The "Nice Theorem". The nice theorem allows us to differentiate/integrate a Taylor
series expansion inside the radius of convergence, in order to obtain new identities (Taylor
series expansions). If
X∞
f (x) = an xn , x ∈ (−R, R)
n=0
then
∞
X
f 0 (x) = nan xn−1 (differentiate)
n=1
Z x ∞
X an
f (t)dt = xn+1 (integrate)
0 n=0
n + 1
The coefficients of the power series obtained through differentiation are nan .
an
The coefficients of the power series obtained through integration are n+1 .
The above two identities are valid whenever x ∈ (−R, R).
2.1.1. Side Remark. Why is this thing called a theorem? To give a simple example, let
g(x) = 2x + x3 + x4
It is easy to differentiate and integrate g(x):
g 0 (x) = 2 + 3x2 + 4x3
Z x
x4 x5
g(t) = x2 + +
0 4 5
Now, the nice theorem says that i can do the same thing even if g(x) was not a polynomial
(finite sum of powers), but a power series (infinite sum of powers !). However when
dealing with a power series we are facing the issue of convergence, and the process of
4 TAYLOR SERIES, POWER SERIES

differentiation (integration) term-by-term needs justification. The nice theorem takes care
of that.
Also, think of the nice theorem as allowing as to obtain new identities from old ones.
P∞ 1
Example 1. Start with n=0 xn = 1−x , x ∈ (−1, 1). Differentiate/integrate:
∞
X 1
nxn−1 = (differentiation)
n=1
(1 − x)2
∞ x
xn
Z
X dx
= = − ln(1 − x) (integration)
n=1
n 0 1−x
and these identities are valid for x ∈ (−1, 1).
P∞ x
• We can multiply both sides of the first one to obtain n=1 nxn = (1−x) 2,x ∈
1
P ∞ n n 4
(−1, 1). For example, taking x = − 4 gives n=1 (−1) 4n = − 25 .
• We can take x = −1 in the second identity (Abel’s theorem) to obtain
P∞ (−1)n P∞ n+1

n=1 n = − ln(2). In other words, n=1 (−1)n = ln(2)

Example 2. Start with
∞
1 X
= (−1)n x2 n, x ∈ (−1, 1)
1 + x2 n=0
Integrate:
∞
−1
X x2n+1
tan x= (−1)n , x ∈ (−1, 1)
n=0
2n + 1
Extend this identity to x = 1 (ok by Abel’s theorem):
∞
π X (−1)n 1 1 1
= = 1 − + − + ...
4 n=0
2n + 1 3 5 7
Example 3. Put −x2 in the Taylor expansion of the exponential function to obtain the
identity
∞
2 X x2n
e−x = (−1)n , x∈R
n=0
n!
Integrate:
x ∞
x2n+1
Z
2 X
e−t dt = (−1)n , x∈R
0 n=0
n!(2n + 1)
For x = 1 we get
1 ∞
(−1)n
Z
2 X
e−t dt =
0 n=0
n!(2n + 1)
This allows us to compute the integral on the left-hand side (otherwise hard to figure out)
to desired accuracy, as in the Example 7 on page 695 of the textbook. Example 4. Find
the sum of the series
∞
X 1
n=0
n!(n + 2)
Start with the power series expansion
∞
X xn
ex =
n=0
n!
 TAYLOR SERIES, POWER SERIES 5

Multiply both sides by x

∞
X xn+1
xex =
n=0
n!
Integrate
x ∞
xn+2
Z X
tet dt =
0 n=0
n!(n + 2)
Evaluate integral on the left by integration by parts
Z x Z x
x
tet dt = tet 0 − et = xex − ex + 1
0 0

Therefore
∞
X xn+2
xex − ex + 1 =
n=0
n!(n + 2)
Set x = 1 to obtain
∞
X 1
=1
n=0
n!(n + 2)
Set x = 2 to obtain
∞
X 2n e2 + 1
=
n=0
n!(n + 2) 4

3. R ELATION BETWEEN TAYLOR SERIES AND POWER SERIES

A power series = Taylor series of its sum
In other words, every time you obtain an identity
∞
X
an xn = (something)
n=0

then the power series on the left-hand side must be the Taylor series of that something on
the right-hand side.
Example 1. We know that
∞
X xn
ex =
n=0
n!
holds for any x ∈ R (we proved this statement by means of the remainder formula). There-
fore there is no harm in considering x2 instead of x in the above "formula", only to obtain
∞
x2
X x2n
e =
n=0
n!

Looking at the boxed principle (above), we can now see that what we have in fact here is
2
the Taylor expansion of the function ex which we obtained almost for free. (Convince
2
yourselves that it is not so trivial to construct the Taylor series of the function f (x) = ex
from scratch. Not to mention that to justify the Taylor series expansion one usually needs to
2
show that Rn (x) → 0, and in the case of f (x) = ex Lagrange’s formula for the remainder
is really complicated.
6 TAYLOR SERIES, POWER SERIES

Example 2. A simpler example is the identity

∞
1 X
= (−1)n x2n
1 + x2 n=0
which is valid for x ∈ (−1, 1)1 and obtained from the geometric series (simply by replac-
ing x by −x2 ). In view of the boxed principle above, this has to be the Taylor expansion
1 (10)
of the function g(x) = 1+x 2 . Hence, if you need to compute g (0) simply identify the
10
coefficient of x :
g (10) (0) 10
x = (−1)n x2n (for some n) ⇒ 2n = 10, n = 5
10!
Therefore
g (10) (0)
= (−1)5 = −1 ⇒ g (10) (0) = −10!
10!