Professional Documents
Culture Documents
Arijit Ganguly
Department of Mathematics and Statistics
Indian Institute of Technology Kanpur
To begin with, let us recall how power functions are defined over R. For x > 0 and a ∈ R,
xa = ea log x .
def
(∗1)
It follows from the definition, given as above in (∗1), that for any fixed a ∈ R, x 7→ xa is a C ∞ function
from (0, ∞) to (0, ∞). We now adapt (∗1) over C to define complex power functions.
Let z ∈ C\{0} and w ∈ C. Unlike the real scenario, here we have too many (branches of) logarithm,
and moreover, there is no canonical choice among them. Hence To make the definition of zw , we first
need to make a choice of α ∈ R, and then we proceed as before:
Thus we can see that, for given z ∈ C \ {0} and w ∈ C, zw does depend upon the choice of the branch
logα . Indeed, varying α, one obtains all possible values of zw . It is now immediate in view of the
definition (∗2) that, given α ∈ R,
C \ Rα −→ C \ {0}, z 7→ zw , (∗3)
is a holomorphic function. From now on, once we have made the choice of α, we shall assume that
the domain of the power function z 7→ zw is C \ Rα , unless otherwise mentioned. It is clear that the
derivative of (∗3) at any point z ∈ C \ Rα is wzw−1 .
As mentioned above, unlike the scenario over R where power of a positive number is a unique
number, here zw has more than one values, until and unless we make a choice of α. In fact, for given
x > 0 and a ∈ R, the value of xa obtained from (∗1) is included in the set of values that (∗2) yields by
varying α. To see this, take α = −π. Then
1
as exp and log−π extends the exponential and logarithm functions over R respectively. In what follows,
when a = 1n , where n ∈ N, we shall see how all n-th roots of a, including the complex ones, can be
obtained by making different choices of α.
√
Let us first consider the simple case where a = 12 . As seen above, taking α = −π, we get the x,
i.e., the positive square root of x. Now observe that
! ! !
1 1 1 √ √
exp logπ x = exp logπ (exp(log x + 2πi)) = exp (log x + 2πi) = x exp(iπ) = − x.
2 2 2
2
Thus α = π gives us the negative square root. As, for any α ∈ R, exp 12 logπ x = exp(logα x) = x,
√ √
so x and − x are precisely all the values that we obtain from (∗2) by varying α. Geometrically the
curve
√ y2 = x has two branches. It is customary to choose the branch in which y > 0 and denote y by
x. We obtain this branch from that of the complex square root function corresponding to α = −π,
while α = π yields us the other branch, i.e., in which y < 0.
Let n ∈ N. Choose k ∈ Z such that 2kπ ≤ α < 2(k + 1)π. If α = 2kπ, we see that
! !
1 1
exp logα x = exp logα exp(log x + 2kπi)
n n
!
1
= exp (log x + 2kπi)
n
!
√n 2kπi
= x · exp .
n
√
Similarly one obtains that exp n1 logα x = n x · exp 2(k+1)πi
n
, when α > 2kπ, as then one has 2(k + 1)π
√
lies in ∈ [α, α + 2π). Thus exp 1n logα x is always of the form n x · exp 2kπi n
, where k ∈ Z. Write
k = nq + r, where q ∈ Z and r ∈ {0, 1, . . . , n − 1}. Then it is easy to see that exp 2kπi
n
= exp 2rπi
n
.
1
This shows that, the values of x n that we obtain from (∗2) by varying α are precisely the following:
! ! !
√n √n 2πi √n 4πi √n 2(n − 1)πi
x, x · exp , x · exp , . . . , x · exp . (∗4)
n n n
Note that, the complex numbers as listed above in (∗4), are precisely all roots of the equation zn = x.