Taylor’s Series

The function, f (z) = 1/(1 + z), can be expanded by geometric series around z = 0 as
1
= 1 − z + z2 − z3 + . . . (1)
1+z
From the ratio test1 of convergence, the series above is convergent if |z| < 1. However, the same function
can be also expanded as
1 1
=
1+z z(1 + z1 )
( )
1 1 1 1
= 1 − + ( )2 − ( )3 + . . . . (5)
z z z z

The range of convergence for the series of Eq.(5) is |z| > 1 by the ratio test. It is seen that both of Eqs.(1)
and (5) are partial representation of f (z) = 1/(1 + z).

Expansion of f (z) about z = a

Assuming f (z) does not have any singular point inside the disk shown above, f (z) can be expressed by the
Cauchy integral formula as
I
1 f (ζ)
f (z) = dζ. (6)
2πi C ζ − z
Note that ζ (zeta) is the variable for integration and z is a constant. Noting that |ζ − a| > |z − a|, eq.(6)
can be modified as

I
1 f (ζ)
f (z) = dζ
2πi C ζ −z
1 An infinite series, ∑
an = a0 + a1 + a2 + a3 + . . . , (2)
converges if  
 an+1 
lim
n→∞
 a
 = r, (3)
n
and
r < 1. (4)

1
 I
1 f (ζ)
= dζ
2πi C (ζ − a)(1 − z−a
ζ−a )
I ( )
1 f (ζ) z−a z−a 2 z−a 3
= 1+( )+( ) +( ) + . . . dζ
2πi C ζ − a ζ −a ζ −a ζ −a
( I ) ( I ) ( I )
1 f (ζ) 1 f (ζ) 1 f (ζ)
= dζ + dζ (z − a) + dζ (z − a)2 + . . .
2πi C ζ − a 2πi C (ζ − a)2 2πi C (ζ − a)3
f ′′ (a)
= f (a) + f ′ (a)(z − a) + (z − a)2 + . . . (7)
2!
As expected, this result is identical with the formula of Taylor series expansion in real variable theory.
However, noting that this was derived based on Cauchy’s integral formula, the validity of Taylor series in
complex variable theory is now made clear, i.e. the expansion above is valid until the loop hits the first
nearest singular point(s) of f (z).

Example 1: Expand f (z) = 1/(z − 1) by Taylor’s series about z = i.

Note that the answer must be in the format of

∑
f (z) = cn (z − i)n ,

as z = i is the center of expansion.

It is noted√that the valid range of Taylor’s series expansion of 1/(z − 1) is a circle centered at z = i with a
radius of 2(= |i − 1|). Rather than using the formula of Taylor’s series, a direct method works better, i.e.
isolate the term (z − i) as

1 1
=
z−1 (z − i) − (1 − i)
1
=
−(1 − i)(1 − z−i
1−i )
( ( ) ( )2 ( )3 )
1 z−i z−i z−i
= 1+ + + + ... . (8)
−(1 − i) 1−i 1−i 1−i

More examples
(1) Expand
1
f (z) ≡ ,
z 2 − 3z + 2
by Taylor’s series around z = 0.

2
(Solution) As seen in the figure above, f (z) can be expanded by Taylor’s series inside (but not on) the disk,
|z| < 1. Thus,

1 1 1
=
z 2 − 3z + 2 z−2z−1
1 1
=
(−2)(1 − z/2) (−1)(1 − z)
( )
1 z ( z )2 ( z )3 1 ( )
= 1+ + + + ... 1 + z + z2 + z3 + z4 + . . . . (9)
(−2) 2 2 2 (−1)
(2) Same
z
f (z) ≡ ,
sin z
about z = 0.

(Solution): Note that z = 0 is NOT a singular point. The nearest singular points from z = 0 are z = ±π
so f (z) must be expanded by Taylor series in |z| < π. Knowing that f (z) is Taylor expandable, you can use
the formula directly, i.e.

f ′′ (0) 2 f ′′′ (0) 3

f (z) = f (0) + f ′ (0)z + z + z + ..., (10)
2! 3!
where
f (0) = limz→0 sinz z = 1,
f ′ (0) = limz→0 f ′ (z)
= limz→0 sin z−z cos z
sin2 z
= ...
(11)
= 0,
f ′′ (0) = ...
1
= 3,
...
so

z z2 7z 4
=1+ + + ... (12)
sin z 6 360

3
Remark on Taylor’s series
1
= 1 + z + z2 + z3 + z4 + . . .
1−z
By the ratio test 2 , |z n+1 /z n | = |z| < 1, the range of convergence for the above series is |z| < 1 which makes
sense as 1/(1 − z) diverges if one substitutes z for 1.
1
= 1 − z2 + z4 − z6 + z8 − . . .
1 + z2

By the ratio test, |z 2n+2 /z 2n | = |z 2 | < 1, the range of convergence is |z| < 1 yet 1/(1 + z 2 ) does not diverge
as z → 1. ???? Explanation: The function, f (z) = 1/(1 + z 2 ) has singular points at z = ±i so the valid
region for the Taylor expansion above is in a disk centered at z = 0 with a radius of 1 (the shortest distance
from 0 to ±i). Note that z = 1 and z = ±i are all on |z| = 1. In fact, |z| < 1 is a sufficient condition for
f (z) to be analytic. There may be a point in |z| ≥ 1 at which f (z) is still analytic.

Analytic Continuation
Consider ( )
1 z ( z )2 ( z )3
f (z) = 1− + − + ... .
2 2 2 2
From the ratio test, the range of convergence is |z| < 2 (validate !) which implies that you cannot compute
f (3) (outside the disk) according to the series above. However, it is possible to compute f (1), f ′ (1), f ′′ (1)
. . . by
( ( )2 )
1 1 1
f (1) = 1− + − ...
2 2 2
1 1
= 1
21+ 2
1
= . (13)
3

1( 1 (z ) 1 ( z )2 1 )
f ′ (z) = 0− +2 −3 + ...
2
( 2
)( 2 2 2 2 )
−1 (z ) ( z )2 ( z )3
= 1−2 +3 −4 + ... . (14)
4 2 2 2
( ) ( )2 ( )3
−1 ( 1 1 1 )
f ′ (1) = 1−2 +3 −4 + ...
4 2 2 2
∑
2A series an converges if |an+1 /an | ∼ r and r < 1.

4
 ( )
1 1
= − ( ) 3 (15)
4 1 2
1+ 2
1
= − . (16)
9
1
f ′′ (z) = (. . .)
2
2
f ′′ (1) = ... = . (17)
27
With the values of f (1), f ′ (1), f ′′ (1) . . . , one can define
f ′′ (1)
g(z) = f (1) + f ′ (1)(z − 1) + (z − 1)2 + . . .
2!
1 1 1
− (z − 1) + (z − 1)2 − . . .
= (18)
3 9 27
It is noted that eq.(18) is identical to f (z) in |z| < 1 as it is Taylor’s series of f (z) about z = 1. Yet, the
range of convergence of g(z) is, from the ratio test, computed as
 1 
 27 (z − 1)2  |z − 1|
 
 1 (z − 1)  = 3
<1 (19)
9
i.e.
|z − 1| < 3 (20)

Equation (20) contains |z| < 2 yet g(z) is identical to f (z) in |z| < 2 so we effectively extended the definition
of f (z) ! This process can be repeated until the entire plane is covered except for z = −2 and is called
analytic continuation. Thanks to analytic continuation, we can extend sin x to sin z or ex to ez .
Theorem If F (z) is analytic in a domain, D, and vanishes on some open disk in D, then it vanishes
throughout D.
Proof. If F vanishes in that disk, all the Taylor coefficients are 0 in the disk. Using analytic continuation,
all the Taylor coefficients are 0 in D.
Corollary If f (z) and g(z) are analytic in a domain, D, and f (z) ≡ g(z) in some disk in D, then f (z) = g(z)
throughout D.
Proof. Set F (z) ≡ f (z) − g(z) in the theorem above.
Example Prove that
sin2 z + cos2 z = 1
for all z = x + iy.
Long answer:

1 ( iz )
cos z = e + e−iz
2
1 ( i(x+iy) )
= e + e−i(x+iy)
2
= ...
= cos x cosh y − i sin x sinh y,

5
Similarly,

sin z = cosh y sin x + i cos x sinh y,

so
2
sin2 z + cos2 z = (cos x cosh y − i sin x sinh y) + (cosh y sin x + i cos x sinh y)2
= ...
= 1.

Short answer: Define f (z) ≡ sin2 z + cos2 z and g(z) ≡ 1. Then, f = g for real z. Hence f = g for all z.

1 + 2 + 3 + 4 + . . . = − 12
1
What is going on ?
Let’s begin with f (z) defined as
f (z) ≡ 1 − z + z 2 − z 3 + . . .
This function is convergent for |z| < 1. On the other hand, another function, g(z), defined as
1
g(z) ≡ ,
z+1
converges in the entire plane except for z = −1. However, in |z| < 1,

f (z) ≡ g(z),

hence, g(z) is analytic continuation of f (z). Therefore, if z = −2 is substituted into f (z), it follows

1 + 2 + 4 + 8 + 16 + . . . = −1.

Now, define the Riemann zeta function, ζ(z), as

∑∞
1 1 1 1 1
ζ(z) ≡ + + + + . . . = .
1z 2z 3z 4z n=1
n z

π2
The function, ζ(z), can be shown to be convergent for ℜ(z) > 1. (Example: ζ(2) = 6 ) Let’s denote analytic
continuation of ζ(z) as ζ̄(z). It can be shown that for ℜ(z) > 1,

ζ̄(z) ≡ ζ(z).

while ζ(z) cannot be evaluated for z = −1, ζ̄(−1) can be evaluated using the following functional relationship,

ζ̄(z) = 2z π z−1 Γ(1 − z) sin (πz/2) ζ̄(1 − z),

as
1
ζ̄(−1) = − .
12
Therefore, one can claim that
1
ζ(−1) = 1 + 2 + 3 + 4 + 5 + . . . = − .
12