ELEMENTARY FUNCTIONS

TSOGTGEREL GANTUMUR

Contents
1. Constant functions 1
2. The exponential 2
3. The argument 6
4. Logarithms 8
5. Powers 11
6. Circular functions 12

1. Constant functions
Before delving into the study of elementary functions, we prove here a simple preliminary
lemma on constant functions. Recall that Dr (c) = {z ∈ C : |z − c| < r}.

Lemma 1. Let f ∈ O(Dr (c)) and f 0 = 0 in Dr (c), where r > 0 and c ∈ C. Then f is
constant in Dr (c).

Proof. Let f (x + iy) = u(x, y) + iv(x, y). Since f is holomorphic in Dr (c), the partial deriva-
tives of u and v exist in Dr (c), and the (complex) derivative f 0 (x + iy) is represented by
multiplication by the Jacobian matrix
!
∂u ∂u
∂x (x, y) ∂y (x, y)
J(x, y) = ∂v ∂v . (1)
∂x (x, y) ∂y (x, y)

Now f 0 = 0 implies that the Jacobian matrix must vanish everywhere in Dr (c). This means
that u and v are constant along every horizontal line and along every vertical line. Since every
point in Dr (c) can be joined to the centre c by a polygonal path consisting of only horizontal
and vertical line segments, we conclude that u and v are equal to their values at c, and hence
they must be constant in Dr (c). 

Remark 2. In the preceding proof, it was not necessary to consider paths consisting of only
horizontal and vertical line segments. We could have joined the centre c with any point in the
disk by a straight line segment, and used directional derivatives instead of partial derivatives.
Moreover, the region in which the result holds can be vastly generalized; The result holds in an
open set Ω if any two points in Ω can be joined by a polygonal path, i.e., if Ω is path-connected.

Date: March 9, 2015.

1
2 TSOGTGEREL GANTUMUR

Figure 1. Any point in a disk can be connected to the centre by a polygonal

path consisting of only horizontal and vertical line segments.

2. The exponential
We look for the complex exponential as a solution of the problem
f 0 = f,f (0) = 1,
an z n . Formally differentiating the
P
and we look for it in the form of a power series f (z) =
power series we find an = an−1 /n = . . . = a0 /n!, and the condition f (0) = 1 gives a0 = 1.
Thus the complex exponential is given by the power series
∞
X zn
exp(z) = , (2)
n!
n=0
whose convergence radius (e.g. by the ratio test) is ∞, and so in particular exp ∈ O(C).
Let us make some simple observations.
• By construction, we have
d
exp = exp in C, and exp(0) = 1. (3)
dz
• For a ∈ C, let g(z) = exp(z) exp(a − z). Then we have
g 0 (z) = exp(z) exp(a − z) − exp(z) exp(a − z) = 0, (4)
for all z ∈ C, which, by g(0) = exp(a) and by Lemma 1, implies that
exp(z) exp(a − z) = exp(a) for a, z ∈ C. (5)
• Putting a = w + z, we get the law of addition
exp(z + w) = exp(z) exp(w) for z, w ∈ C. (6)
• Putting a = 0, we infer
exp(−z) exp(z) = 1 and so exp(z) 6= 0 ∀z ∈ C.
• Therefore exp : C → C× is a group homomorphism, where C× = C \ {0} is the
multiplicative group of C.
• By considering g(z) = f (z) exp(−z), one can show that the only function satisfying
f 0 = f in C with f (0) = 1 is the complex exponential.
In the following theorem, we construct a holomorphic (right) inverse of the exponential
in the open disk D1 (1). By using this inverse, we will also show that given a ∈ C× with
a = exp α for some α ∈ C, a holomorphic inverse of the exponential exists in the open disk
D|a| (a). Recall here that Dr (c) = {z ∈ C : |z − c| < r}.
 ELEMENTARY FUNCTIONS 3

Theorem 3. (a) The power series

∞
X (−1)n−1
λ(z) = (z − 1)n , (7)
n
n=1

converges in D1 (1), and hence defines a function λ ∈ O(D1 (1)). Moreover, we have
1
exp λ(z) = z and λ0 (z) = for z ∈ D1 (1). (8)
z
(b) If a ∈ C× and exp α = a, then λa (z) = λ( az ) + α satisfies
1
exp λa (z) = z and λ0a (z) = for z ∈ D|a| (a). (9)
z
In particular, λa is holomorphic in D|a| (a).
Proof. (a) By the ratio test, the convergence radius of (7) is 1, so (7) converges in D1 (1).
Then a termwise differentiation gives
∞ ∞
0
X
n−1 n−1
X 1 1
λ (z) = (−1) (z − 1) = (1 − z)n−1 = = ,
1 − (1 − z) z
n=1 n=1

provided that |1 − z| < 1, that is, z ∈ D1 (1).

Now let g(z) = z exp(−λ(z)). Then for z ∈ D1 (1) we have
g 0 (z) = exp(−λ(z)) − z exp(−λ(z))λ0 (z) = 0,
meaning that g(z) = g(1) = exp(−λ(1)) = 1 in D1 (1).
(b) We have
exp(λa (z)) = exp λ( az ) exp(α) = z


a · a = z, (10)
and
z
1 a 1 1
λ0a (z) = λ0 · = · = , (11)
a a z a z
as long as | az − 1| < 1, that is, if |z − a| < |a|. This completes the proof. 
Corollary 4. The map exp : C → C× is surjective.
Proof. Let A = {exp z : z ∈ C}. By Theorem 3, we know that D1 (1) ⊂ A, and that a ∈ A
implies D|a| (a) ⊂ A. Hence the proof is reduced to the following “game.” Initially, the disk
D1 (1) is coloured blue, and the rest of the complex plane is white. At any stage in the game,
we can choose a point a in the blue region, and colour all the points in the disk D|a| (a) blue.
The question is, by repeating this procedure, can we colour the entire set C× blue? It is not
difficult to see that it is possible to do so.
One possibility is as follows (Figure 2). Let 1 < r < 2. Then r ∈ D1 (1), and hence
Dr (r) ⊂ A. In particular, r2 ∈ A, and hence Dr2 (r2 ) ⊂ A. By induction, this shows that
Drn (rn ) ⊂ A for any n, and since any z ∈ C with Re z > 0 can be contained in Drn (rn ) for
sufficiently large n, we conclude that {z : Re z > 0} ⊂ A.
Let a = i + ε, where ε > 0 is a small real number. Then i ∈ D|a| (a), and hence i ∈ A,
or D1 (i) ⊂ A. By considering the succession of points {irn } with a constant 1 < r < 2, we
conclude that Drn (irn ) ⊂ A for all n, or {z : Im z > 0} ⊂ A.
Finally, by considering the points a = −1 + iε and a = −i + ε, we get −1 ∈ A and −i ∈ A.
Then we repeat the same procedure with the sequences {−rn } and {−irn } as the disk centres,
to conclude that {z : Re z < 0} ⊂ A and {z : Im z < 0} ⊂ A. 
Definition 5 (Euler 1748). We define the Euler number by e = exp 1.
4 TSOGTGEREL GANTUMUR

1 r r2 i ir2
ir
i

(a) Starting with D1 (1), and (b) Once we coloured the (c) Once i is coloured,
painting the disks Drn (rn ) for right half plane, it is easy we repeat the same proce-
a fixed r ∈ (1, 2) and for all to paint the point i. dure to paint the upper half
n = 1, 2, . . ., we can colour the plane {Imz > 0}.
right half plane {Rez > 0}.

Figure 2. Illustration of the proof of Corollary 4.

Remark 6 (Real exponential). (a) From the definition (2) it is clear that if x ∈ R then
exp x ∈ R. In particular, e is real. We also have
x2
exp x = 1 + x + + ... ≥ 1 + x for x ≥ 0, (12)
2!
1
which implies that exp x → ∞ as x → ∞. Moreover, since exp(−x) = exp x , we have
1
0 < exp(−x) ≤ for x ≥ 0, (13)
1+x
and so in particular, exp x → 0 as x → −∞. We conclude that exp : R → (0, ∞) is surjective.
(b) Let us compute the derivative of exp x with respect to x ∈ R. Suppose that f (x + iy) =
u(x, y) + iv(x, y) is a complex differentiable function, with u and v real. Then we have
∂u ∂v
f 0 (x + iy) = (x, y) + i (x, y), (14)
∂x ∂x
where f 0 denotes the complex derivative of f . If f (x) ∈ R for x ∈ R, then f (x) = u(x, 0) and
v(x, 0) = 0 for all x, implying that
∂u d
f 0 (x) =
(x, 0) = f (x). (15)
∂x dx
Returning back to the exponential function, we infer
d
exp x = exp0 (x) = exp x > 0 for all x ∈ R. (16)
dx
Therefore, exp : R → (0, ∞) is strictly increasing, and hence a bijection. The inverse function
log : (0, ∞) → R is called the real logarithm.
(c) For any z ∈ C and n ∈ N, we have
exp(nz) = exp(z + . . . + z) = exp(z) · · · exp(z) = (exp z)n , (17)
and
1 1
exp(−nz) = = = (exp z)−n , (18)
exp(nz) (exp z)n
 ELEMENTARY FUNCTIONS 5

1
showing that exp(nz) = (exp z)n for all n ∈ Z. Putting z = n, we get exp 1 = (exp n1 )n , or
1
exp n1 = e n . This implies that
1 n
n 1 n
exp m = (exp m ) = (e m )n = e m for n ∈ Z, m ∈ N. (19)
Finally, by continuity of exp x and of ex , we conclude that exp x = ex for all x ∈ R.
Exercise 7. With log : (0, ∞) → R denoting the real logarithm, prove the following.
• log(ab) = log a + log b, for 0 < a, b < ∞.
• log(ax ) = x log a, for 0 < a < ∞ and x ∈ R.
Our next task is to identify the kernel ker(exp) = {z ∈ C : exp z = 1} of exp : C → C× .
Lemma 8. We have | exp z| = exp(Re z) for z ∈ C.
Proof. For y ∈ R, we have (iy)n = (−iy)n , and so
m m
X (iy)n X (−iy)n
= for any m ∈ N. (20)
n! n!
n=0 n=0
The right hand side of this equality converges to exp(−iy) as m → ∞. Moreover, the left
hand side converges to exp(iy), because of the general rule lim z̄m = lim zm . To conclude, we
have exp(iy) = exp(−iy), and hence |exp(iy)|2 = exp(iy)exp(iy) = exp(iy) exp(−iy) = 1.
Finally, if z = x + iy with x, y ∈ R, then we have
|exp z| = |exp(x) exp(iy)| = |exp x||exp(iy)| = exp x, (21)
completing the proof. 
By this lemma, | exp z| = 1 is equivalent to exp(Re z) = 1. Since exp : R → (0, ∞) is a
bijection, exp(Re z) = 1 if and only if Re z = 0.
Corollary 9. We have {z ∈ C : | exp z| = 1} = iR, where iR = {ix : x ∈ R}.
Theorem 10. We have ker(exp) = iT Z for some constant T > 0.
Proof. If exp z = 1, then obviously | exp z| = 1, and so by the preceding corollary, we have
K = ker(exp) ⊂ iR. It is clear that 0 ∈ K, and so a natural question is whether or not K has
any nonzero element. To answer this question, we note that the surjectivity of exp : C → C× ,
combined with the preceding corollary, implies that there is s ∈ R with exp(is) = −1, and so
with exp(2is) = 1. Obviously s 6= 0, and thus K 6= {0}.
1
We have the symmetry K = −K, because exp z = 1 implies that exp(−z) = exp z = 1.
Since K contains numbers with positive imaginary parts, the number
T = inf{t > 0 : it ∈ K}, (22)
is well defined. In order to show that iT ∈ K, let {tk } be a sequence such that {itk } ⊂ K and
tk → T as k → ∞. Then by continuity of the exponential, exp(itk ) → exp(iT ) as k → ∞.
On the other hand, exp(itk ) = 1 for all k, which implies that exp(iT ) = 1, that is, iT ∈ K.
Furthermore, since exp(inT ) = exp(iT )n = 1 for n ∈ Z, we conclude that iT Z ⊂ K.
Now we wish to show that T > 0. To this end, we write
z2 z z2

exp z − 1 = z + + ... = z 1 + + + . . . = zg(z), (23)
2! 2! 3!
where we have introduced the function
∞
z z2 X zn
g(z) = 1 + + + ... = . (24)
2! 3! (n + 1)!
n=0
6 TSOGTGEREL GANTUMUR

The convergence radius of the latter series is ∞, and so in particular we have g ∈ O(C). Since
g(0) = 1, by continuity, g has no zeroes in a small open disk centred at 0. Hence in that disk,
the only solution to exp z = 1 is z = 0, which means that T > 0.
Finally, suppose that r ∈ R with ir ∈ K. Then there is n ∈ Z such that nT ≤ r < (n + 1)T ,
or 0 ≤ r − nT < T . But exp(ir − inT ) = exp(ir) exp(−inT ) = 1, hence r − nT = 0 by the
minimal property of T . This proves that K ⊂ iT Z. 
Remark 11. Suppose that w, z ∈ C satisfy exp(z + w) = exp(z). Then exp(z) exp(w) =
exp(z), that is, we have exp(w) = 1, meaning that w ∈ iT Z. One implication of this is that
the periods of the exponential function are precisely the numbers inT with n ∈ Z. Recall
that w ∈ C is called a period of a function f if f (z + w) = f (z) for all z ∈ C. Another,
quite strong implication is that if I is any half-open interval of length T , such as I = [0, T ),
then the complex exponential restricted to the horizontal strip R + iI is bijective, where
R + iI = {x + iy : x ∈ R, y ∈ I}.

3T i
exp
2T i
log

Ti
1
0

Figure 3. Mapping properties of the complex exponential.

3. The argument
In Corollary 9 we have established that exp−1 (S 1 ) = {z ∈ C : | exp z| = 1} is equal to the
imaginary axis iR. In view of Remark 11, this implies that the map p : R → S 1 defined by
p(t) = exp(it) is surjective with the periods T Z, where S 1 = {z ∈ C : |z| = 1}. Moreover,
for any half-open interval I of length T , the function p : I → S 1 is bijective. In other words,
every z ∈ C with |z| = 1 can be written uniquely as z = p(t) with t ∈ I.
Supposing that exp(x + iy) = u(x, y) + iv(x, y), we have p(t) = u(0, t) + iv(0, t), and hence
∂u ∂v ∂v ∂u
p0 (t) = (0, t) + i (0, t) = − (0, t) + i (0, t)
∂y ∂y ∂x ∂x
 ∂u (25)
∂v 
=i (0, t) + i (0, t) = i exp0 (it) = i exp(it) = ip(t),
∂x ∂x
 ELEMENTARY FUNCTIONS 7

where we have used the Cauchy-Riemann equations in the second step, and exp0 denotes
the complex derivative of exp. From this we infer that |p0 (t)| = | exp(it)| = 1, meaning
that p : [0, T ) → S 1 is an arc length parameterization of the unit circle S 1 . Therefore, the
arclength of the unit circle is equal to T . Moreover, since p0 (0) = i, we have Im p(t) > 0 for
all sufficiently small t > 0, see Figure 4(a).
T
Definition 12. We define the number π by π = .
2
z
We will also use the shorthand notation e = exp(z) for the exponential function. Thus
every z ∈ C× can be written as z = |z|eiθ with θ ∈ R, and moreover θ is unique if one requires
θ ∈ I with any fixed half-open interval I of length 2π.
Definition 13. We define the set-valued function, called the argument, by
z
arg z = t ∈ R : eit = z ∈ C× .

, (26)
|z|
It is customary to write
arg z = θ + 2πn, n ∈ Z, (27)
for a fixed θ ∈ R which depends on z, and say that arg z is a multi-valued function.
Remark 14. In the multi-valued formalism (27), it is understood that arg z returns infinitely
many values θ + 2πn, n ∈ Z, all at once, while in the set-valued formalism (26), what arg z
returns is a single object, that is the set of all the numbers θ + 2πn for n ∈ Z. We will use
these two formalisms interchangeably.

Example 15. Since e0 = 1, we have

arg 1 = 2πZ = {2πn : n ∈ Z}. (28)
We have (eπi )2 = e2πi = 1 and eπi 6= 1, hence eπi = −1, implying that
arg(−1) = π + 2πZ = {π + 2πn : n ∈ Z}. (29)
πi πi πi
Moreover, from (e )2 = eπi = −1 we infer that either e = i or e = −i. Suppose that
2 2 2
πi
e 2 = −i. Then since Im eit > 0 for small t > 0 and Im(−i) < 0, the intermediate value
theorem implies that the real valued function f (t) = Im eit has a zero at some t satisfying
0 < t < π2 . In other words, eit ∈ R for some 0 < t < π2 , and in light of the fact that |eit | = 1,
we have eit = ±1 for some 0 < t < π2 . In either case, we get e2it = 1, which impossible, since
0 < 2t < π, and s = 2π is the smallest positive solution to eis = 1. To conclude, we have
π π
arg i = + 2πZ = + 2πn : n ∈ Z . (30)
2 2
Exercise 16. Compute arg(−i), arg(1 + i), and arg(1 − i).
Remark 17. If arg z = θ + 2πn, n ∈ Z, we have
|z|ei arg z = |z|eiθ+2πin = |z|eiθ e2πin = ze2πin = z, n ∈ Z. (31)
Now recall that for A, B ⊂ C, the sum set is defined as
A + B = {a + b : a ∈ A, b ∈ B}. (32)
Then noting that
arg w + arg z = {τ + 2πn : n ∈ Z} + {θ + 2πn : n ∈ Z} = {τ + θ + 2πn : n ∈ Z}, (33)
we can write
wz = |w|ei arg w · |z|ei arg z = |w||z| · ei(arg w+arg z) . (34)
In other words, when we multiply two complex numbers, the magnitudes are multiplied and
the arguments are added.
8 TSOGTGEREL GANTUMUR

i eit

t
0 1

(a) The map t 7→ eit is an arclength parame-

terization of the unit circle S 1 . The inverse of (b) The graph of arg z. The blue helix repre-
this parameterization is arg z for z ∈ S 1 . sents the graph of arg z restricted to |z| = 1.

Figure 4. The parameterization eit of S 1 , and the multi-valued function arg z.

Definition 18. Let F be a set-valued function, defined on some A ⊂ C, and let Ω ⊂ A. Then
a single-valued function f ∈ C (Ω) is called a (continuous) branch of F in Ω, if f (z) ∈ F (z)
for all z ∈ Ω.
Definition 19. The principal branch of arg z is defined as
z
Arg z = θ, where eiθ = and − π < t ≤ π. (35)
|z|
π
Example 20. We have Arg 1 = 0, Arg(−1) = π, and Arg i = .
2
Exercise 21. Compute Arg(−i), Arg(1 + i), and Arg(1 − i).

4. Logarithms
For z = ρeiθ ∈ C× and ζ = log ρ + iθ, it is clear that
exp(ζ) = exp(log ρ) exp(iθ) = ρ exp(iθ) = z. (36)
Now, if exp(w) = z, then w = ζ + 2πin for some n ∈ Z. Therefore, the solutions of the
equation exp(ζ) = z are precisely the numbers log ρ + iθ + 2πiZ = log |z| + i arg z.
Definition 22. We define the set-valued complex logarithm, by
log z = exp−1 (z) = {w ∈ C : exp w = z}, z ∈ C× . (37)
It is customary to regard it as a multi-valued function, and simply write
log z = log |z| + i arg z, (38)
 ELEMENTARY FUNCTIONS 9

where “log” in the right hand side of course denotes the real logarithm.
Remark 23. The notation “log” is an overloaded notation:
• log x may mean the (single-valued) real logarithm, for x ∈ R with x > 0.
• log z is the multi-valued complex logarithm, as we have just defined.
• Any branch of the complex logarithm is often denoted by log z as well.
Example 24. We have
log(−1) = πi + 2πin, n ∈ Z,
πi (39)
log i = + 2πin, n ∈ Z,
2
and
complex log real log
z }| { z }| {
log x = log x +2πin, n ∈ Z, (40)
for x ∈ R with x > 0.
Definition 25. The principal branch of log z is defined as
Log z = log |z| + iArgz, z ∈ C− , (41)
where C− = C \ (−∞, 0] is called the slit-plane.
Example 26. We have
πi
Log(−1) = πi, Log i = , (42)
2
and
Log x = log x, (43)
for x ∈ R with x > 0, where “log” in the right hand side is the real logarithm.
Remark 27. (a) For z ∈ C× , by construction, we have
exp(log z) = exp(log |z| + i arg z) = |z|earg z = z. (44)
(b) However, the property log ex = x of the real logarithm does not extend to the complex
logarithm. What we have instead is
log(exp z) = log | exp z| + i arg exp z = log eRe z + i(Im z + 2πn)
(45)
= Re z + i Im z + 2πin = z + 2πin, n ∈ Z.
This is not surprising, since log z is now a multi-valued function.
(c) We can ask if the property log ex = x holds for individual branches of logarithm.
Let us consider the principal branch here as an example. First of all, Log(exp z) is not
defined if exp z ∈ R and exp z < 0, that is, if Im z = π + 2πn for some n ∈ Z. So if
π + 2π(n − 1) < Im z < π + 2πn for some n ∈ Z, then Arg exp z = Im z − 2πn, and hence
Log(exp z) = log | exp z| + iArg exp z = log eRe z + i(Im z − 2πn) = z − 2πin. (46)
This means that
Log(exp z) = z, if − π < Im z < π. (47)
(d) For w, z ∈ C× , we have
log(wz) = log |wz| + i arg(wz) = log |w| + log |z| + i(arg w + arg z)
(48)
= log w + log z,
where the equality should be understood as an equality between sets.
10 TSOGTGEREL GANTUMUR

(e) Note that Arg(wz) = Arg w + Arg z is not true in general. For example, Arg 1 = 0, but
Arg(−1) + Arg(−1) = 2π. If Arg(wz) = Arg w + Arg z for some particular numbers w and z,
then we would have
Log(wz) = Log w + Log z. (49)
A sufficient condition for Arg(wz) = Arg w + Arg z to hold would be Arg w, Arg z ∈ (− π2 , π2 ),
that is, Re w > 0 and Re z > 0.
Theorem 28. Let ` ∈ C (Dr (a)), with Dr (a) ⊂ C× . Then the following are equivalent.
(a) ` is a branch of logarithm in Dr (a).
(b) exp(`(a)) = a and ` ∈ O(Dr (a)) with
1
`0 (z) = for z ∈ Dr (a). (50)
z
(c) There is some n ∈ Z and α ∈ C with exp α = a, such that
`(z) = λ( az ) + α + 2πin for z ∈ Dr (a), (51)
where λ is the logarithmic series introduced in Theorem 3.
Proof. Suppose that ` is a branch of logarithm in Dr (a), and let λa (z) = λ( az ) + α with some
α ∈ C satisfying exp α = a. Then by Theorem 3(b) the function λa is a branch of logarithm
in Dr (a), and so we have
exp(`(z)) = z = exp(λa (z)), z ∈ Dr (a), (52)
which implies that
exp(`(z) − λa (z)) = 1, z ∈ Dr (a), (53)
or
`(z) − λa (z) = 2πig(z), z ∈ Dr (a), (54)
for some function g : Dr (a) → Z. Since ` − λa is continuous, so is g, and hence g = const in
Dr (a). This means that
`(z) = λa (z) + 2πin, z ∈ Dr (a), (55)
for some constant n ∈ Z, establishing the implication (a) =⇒ (c).
The implication (c) =⇒ (b) is immediate from Theorem 3. To prove (b) =⇒ (a), we consider
the function
f (z) = z exp(−`(z)), (56)
and compute
1
f 0 (z) = exp(−`(z)) − z exp(−`(z))`0 (z) = exp(−`(z)) − z exp(−`(z)) · = 0, (57)
z
which implies that f = const in Dr (a). This means that f (z) = 1 for z ∈ Dr (a), because
a
f (a) = a exp(−`(a)) = = 1, (58)
exp `(a)
and thus
1 z z
exp `(z) = = = =z for z ∈ Dr (a), (59)
exp(−`(z)) f (z) f (a)
completing the proof. 
 ELEMENTARY FUNCTIONS 11

5. Powers
For x, s ∈ R with x > 0, we can define the real positive power xs by xs = es log x , where
log x is the real logarithm. This can be extended to the complex setting as follows.
Definition 29. For σ ∈ C, we define the multi-valued complex power z 7→ z σ by
z σ = exp(σ log z), z ∈ C× . (60)
The principal value of the power zσ is defined as
p.v. z σ = exp(σLog z), z ∈ C− . (61)
Remark 30. The notation z σ is even more overloaded than “log”:
• xs may mean the (single-valued) real power, for x, s ∈ R with x > 0.
• z σ is the multi-valued complex power, as we have just defined.
• Any branch of the complex power is often denoted by z σ as well.
• The notation “p.v.” in the principal value we have introduced is not completely stan-
dard, and therefore often omitted, resulting in the notation z σ also for p.v. z σ .
Example 31. We have
1σ = exp(2πσin), n ∈ Z,
σ
p.v. 1 = exp(σ · 0) = 1, (62)
i πi − π2
p.v. i = exp(i · 2) =e .
Remark 32. The following simple properties can be derived.
• p.v. xs = exp(sLog x) = xs for x, s ∈ R with x > 0, where xs in the right hand side is
the real positive power.
• z 0 = exp(0) = 1.
• z 1 = exp(log z) = z.
• z −1 = exp(− log z) = z1 .
• z n = exp(n log z) = exp(log z) · · · exp(log z) = |z · z{z· · · z}.
| {z }
n times n times

Exercise 33 (Abel 1826). Show that

p.v. z s+it = |z|s e−tArgz cos sArgz + t log |z| + i sin sArgz + t log |z| ,




(63)
for z ∈ C− and s, t ∈ R.
Remark 34 (Powers with real exponents). For z = |z|eiθ ∈ C× and s ∈ R, we have
z s = exp(s log |z|) exp(is arg z) = |z|s eisθ e2πsik , k ∈ Z, (64)
where |z|s is the real positive power. If s is irrational, then there is no n ∈ Z and k ∈ Z that
satisfy 2πsk = 2πn except n = k = 0, and hence z s has infinitely many distinct values. On
the other hand, if s is rational, and say, s = pq with p ∈ Z and q ∈ N minimal, then
e2πsik = e2πsi(mq+r) = e2πipm e2πsir = e2πsir , (65)
where k = mq + r with m ∈ Z and 0 ≤ r < q, implying that z s has exactly q distinct values.
Remark 35 (Roots). For z = |z|eiθ ∈ C× and n ∈ N, we define the n-th root of z to be
√ 1
z = z n = n |z| eiθ/n e2πik/n ,
p
n
k ∈ Z, (66)
p √
where n |z| is the real positive n-th root. It is easily checked that n z has exactly n distinct
values, corresponding to, e.g., k = 0, 1, . . . , n − 1. We have
√
( n z)n = ( n |z|)n (eiθ/n )n (e2πik/n )n = |z|eiθ e2πik = z,
p
(67)
12 TSOGTGEREL GANTUMUR

so (66) with any fixed k is a right inverse of the function

√ z 7→ z n . We claim that all right
inverses are obtained in this manner, in the sense that z = {w ∈ C : wn = z}. To prove
n

this claim, let w = ρeiτ , and require

wn = ρn eiτ n = |z|eiθ . (68)
|wn | ρn , ρn iτ n iθ
p
n
Since = we infer that = |z|, and hence ρ = |z|. Furthermore, e = e is
equivalent to the condition that τ n = θ + 2πik for some k ∈ Z, or
w = ρeiθ/n e2πik/n , k ∈ Z, (69)
establishing the claim.

6. Circular functions
Definition 36. We define the (circular) cosine and the (circular) sine, respectively, by
exp(iz) + exp(−iz) exp(iz) − exp(−iz)
cos z = and sin z = , (70)
2 2i
for z ∈ C.
Remark 37. For t ∈ R, we have
exp(it) + exp(−it)
cos t = = Re eit , (71)
2
and
exp(it) − exp(−it)
sin t = = Im eit , (72)
2i
which, in light of the fact that t 7→ eit is an arclength parameterization of the unit circle S 1 ,
shows that Definition 36 extends the usual geometric definition of cos t and sin t to complex
values of t, cf. Figure 5.

i
eit
i sin t

cos t 1

Figure 5. The functions cos t and sin t for real values of t.

Remark 38. The following properties can easily be derived from Definition 36.
(a) exp(iz) = cos z + i sin z for z ∈ C (Euler 1748).
(b) cos(−z) = cos z and sin(−z) = − sin z for z ∈ C.
(c) cos2 z + sin2 z = 1 for z ∈ C.
 ELEMENTARY FUNCTIONS 13

(d) (cos z)0 = − sin z and (sin z)0 = cos z for z ∈ C.

(e) cos(w + z) = cos w cos z − sin w sin z for w, z ∈ C.
(f) sin(w + z) = sin w cos z + cos w sin z for w, z ∈ C.
Theorem 39. The maps cos : C → C and sin : C → C are surjective.
Proof. Given w ∈ C, consider the equation cos z = w, that is,
exp(iz) + exp(−iz) = 2w. (73)
1
With u = exp(iz), this is equivalent to u + u = 2w, or
2
u − 2wu + 1 = 0. (74)
We see that u = 0 is not a solution. By writing
(u − w)2 = u2 − 2wu + w2 = w2 − 1, (75)
we find the solution u in terms of the multi-valued square root as
p
u = w + w2 − 1. (76)
Since u 6= 0, we can solve u = exp(iz) as z = −ilog u. Note that to each w ∈ C, in general
there corresponds two distinct values of u, and to each value of u, there corresponds infinitely
many distinct values of z. This can be written as
p
z = cos−1 (w) = −ilog(w + w2 − 1), (77)
where the composition of multi-valued functions is interpreted in the obvious way.
The equation sin z = w can similarly be solved by
p
z = sin−1 (w) = −ilog(w + w2 + 1), (78)
completing the proof. 
Exercise 40. Prove the following.
(a) sin z = 0 if and only if z = πn for some n ∈ Z.
(b) cos z = 0 if and only if z = π2 + πn for some n ∈ Z.
(c) The periods of sin are precisely the numbers 2πn, n ∈ Z.
(d) The periods of cos are precisely the numbers 2πn, n ∈ Z.
Exercise 41. Show that cos z = cos w if and only if either z + w = 2πn for some n ∈ Z, or
z − w = 2πn for some n ∈ Z.
Exercise 42. Let cot : C \ πZ → C and tan : C \ ( π2 + πZ) → C be defined by
cos z sin z
cot z = , and tan z = ,
sin z cos z
respectively. Prove the following.
(a) The functions cot and tan are holomorphic in their respective domains, with
1 1
(cot z)0 = − 2 , and (tan z)0 = .
sin z cos2 z
(b) The periods of cot and tan are the numbers πn, n ∈ Z.
(c) With
z3 z5 z7
arctan z = z − + − + ..., |z| < 1,
3 5 7
we have
1
(arctan z)0 = and arctan(tan z) = z for |z| < 1.
1 + z2