MIT OpenCourseWare

http://ocw.mit.edu

18.112 Functions of a Complex Variable

Fall 2008

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
 Lecture 2: Exponential function & Logarithm
for a complex argument
(Replacing Text p.10 - 20)

For b > 1, x ∈ R, we defined in 18.100B,

bx = sup bt

t∈Q, t≤x

(where bt was easy to define for t ∈ Q) . Then the formula

bx+y = bx by

was hard to prove directly. We shall obtain another expression for bx making proof
easy.
Let � x
dt
L(x) = , x > 0.
1 t
Then
L(xy) = L(x) + L(y)
and
1
L′ (x) = > 0.
x
So L(x) has an inverse E(x) satisfying

E(L(x)) = x.

By 18.100B,
E ′ (L(x))L′ (x) = 1,
so
E ′ (L(x)) = x.
If y = L(x), so x = E(y), we thus have

E ′ (y) = E(y),

1
It is easy to see E(0) = 1, so by uniqueness,

x2 xn
E(x) = 1 + x + +···+ +··· and E(1) = e.
2 n!

Theorem 1 bx = E(xL(b)), ∀x ∈ R.

Proof: Let u = L(x), v = L(y), then

E(u + v) = E(L(x) + L(y)) = E(L(xy)) = xy = E(u)E(v),

E(n) = E(1)n = en ,
n
and if t = m
,
E(t)m = E(mt) = E(n) = en .
so
E(t) = et , t ∈ Q, t > 0.
Since
E(t)E(−t) = 1,
So
E(t) = et , t ∈ Q.
Now
bn = E(nL(b))
and � �
1 1
b m =E L(b)
m
since both have same mth power.
� �n
� 1
�n n 1 �n �
b m =b m =E L(b) =E L(b) ,
m m
so
bt = E(tL(b)), t ∈ Q.
Now for x ∈ R,

bx = sup (bt ) = sup E(tL(b)) = E(xL(b))

t≤x, t∈Q t≤x, t∈Q

since E(x) is continuous. Q.E.D.

Corollary 1 For any b > 0, x, y ∈ R, we have bx+y = bx by .

2
 In particular ex = E(x), so we have the amazing formula
�x
x2 xn
�
1 1
1+1+ +···+ +··· =1+x+ +···+ +··· .
2! n! 2! n!

The formula for ex suggests defining ez for z ∈ C by

z2 zn
ez = 1 + z + +···+ +··· .
2! n!
the convergence being obvious.

Proposition 1 ez+w = ez ew for all z, w ∈ C.

Proof: Look at the functions

f (t) = etz+w , g(t) = etz ew

for t ∈ R. Differentiating the series for etz+w and etz with respect to t, term-by-term,
we see that
df dg
= zf (t), = zg(t)
dt dt
and
f (0) = ew , g(0) = ew .
By the uniqueness for these equations, we deduce f ≡ g. Thus f (1) = g(1). Q.E.D.
Note that if t ∈ R,

eit e−it = 1, and(eit )−1 = e−it .

Thus
|eit | = 1.
So eit lies on the unit circle.
Put
eit + e−it t2

cos t = = 1 − + · · · ,

2 2
it
e −e −it
t3

sin t = = t − + · · · .

2 3!
Thus we verify the old geometric meaning e = cos t+i sin t. Note that the eit (t ∈ R)
it

fill up the unit circle. In fact by the intermediate value theorem, {cos t | t ∈ R} fills
up the interval [−1, 1], so eit = cos t + i sin t is for a suitable t an arbitrary point on
the circle.

3
 Note that z 7→ ez takes all values w ∈ C except 0. For this note
ez = ex · eiy , z = x + iy.
Choose x with
ex = |w|
and then y so that
w
eiy = ,
|w|
then ez = w.

If
z z = |z|eiϕ , w = |w|eiψ ,
w then
a
zw = |z||w|ei(ϕ+ψ)
Fig.2-1 = |z||w|(cos (ϕ + ψ) + i sin (ϕ + ψ)),

which gives a geometric interpretation

of the multiplication.
From this we also have the following
a 2 /n very useful formula
(cos ϕ + i sin ϕ)n = einϕ = cos nϕ + i sin nϕ.
Thus
Fig. 2-2

Theorem 2 The roots of z n = 1 are 1, ω, ω 2, · · · , ω n−1, where

2π 2π
ω = cos + i sin .
n n
Geometric meanings for some useful complex number sets:
|z − a| = r ←→ circle
|z − a| + |z − b| = r, (|a − b| < r) ←→ ellipse
|z − a| = |z − b| ←→ perpendicular bisector
{z | z = a + tb, t ∈ R} ←→ line
{z | Imz < 0} ←→ lower half plane
� �
z−a
{z | Im < 0} ←→ general half plane
b

 For x real, x 7→ ex has an inverse. This is NOT the case for z 7→ ez , because

ez+2πi = ez ,

thus ez does not have an inverse. Moreover, for w =

6 0,

ez = w

has infinitely many solutions:

w
ex = |w|, eiy = =⇒ x = log |w|, y = arg(w).
|w|
So
log w = log |w| + iarg(w)
takes infinitely many values, thus not a function.
Define

Arg(w) � principal argument of w in interval − π < Arg(w) < π

and define the principal value of logarithm to be

Log(w) � log |w| + iArg(w),

which is defined in slit plane (removing the negative real axis).

We still have
log z1 z2 = log z1 + log z2
in the sense that both sides take the same infinitely many values. We can be more
specific:

Theorem 3 In slit plane,

Log(z1 z2 ) = Log(z1 ) + Log(z2 ) + n · 2πi, n = 0 or ± 1

and n = 0 if
−π < Arg(z1 ) + Arg(z2 ) < π.
In particular, n = 0 if z1 > 0.

Proof: In fact, Arg(z1 ), Arg(z2 ) and Arg(z1 z2 ) are all in (−π, π), thus

−π − π − π < Arg(z1 ) + Arg(z2 ) − Arg(z1 z2 ) < π + π + π,

but
Arg(z1 ) + Arg(z2 ) − Arg(z1 z2 ) = n · 2πi,

5
thus

|n| ≤ 1.
If
|Arg(z1 ) + Arg(z2 )| < π,
since
|Arg(z1 z2 )| < π,
they must agree since difference is a multiple of 2π. Q.E.D.