Question: (By Ali Abbas)

(i) Suppose that n ∈ N\{2} is prime and that ω is an arbitrary non-real n-th
root of unity. Prove that the powers of ω from 0 to n − 1 form all n-th roots.
n−1 n−1
k
zk.
Q P
(ii) Hence show that (z − ω ) =
k=1 k=0
n−1
(1 − ω k ) = n.
Q
(iii) Deduce that
k=1

Solution:

(i) We will need to establish a couple of minor results before we can tackle the
question at hand, which we will restate as a corollary before offering its proof.

Lemma 1: If ω is an n-th root of unity, then ω k is also an n-th root of unity

∀k ∈ N.

Proof: Fix k ∈ N and suppose that ω is an n-th root of unity. Then, by

definition, ω n = 1. Now, (ω k )n = (ω n )k = 1k = 1.

∴ ω k is also an n-th root of unity. 2

Lemma 2: All n-th roots of unity are distinct.

Proof: Let P (z) = z n − 1 and assume that ω is an arbitrary zero of P . Then

ω 6= 0 as P (0) = 0n − 1 = −1 6= 0. Furthermore, P 0 (z) = nz n−1 and
P 0 (ω) = nω n−1 6= 0 since ω 6= 0 and n 6= 0. Therefore, ω cannot be a root of
multiplicity and hence all roots are distinct. 2
Corollary: Suppose that n ∈ N such that n is prime and n 6= 2. If ω is any
non-real n-th root of unity, then the remaining roots are the powers of ω from
0 up to n − 1 (inclusive). That is, the roots are ω 0 , ω 1 , ω 2 ,..., ω n−1 .

Proof: Suppose that n 6= 2 is prime and that ω is a non-real n-th root of

unity. From Lemma 1, each of ω 0 , ω 2 ,..., ω n−1 is also a root. We note that this
comprises n roots in total, exactly how many should exist by a consequence of

1
the Fundamental Theorem of Algebra. So by Lemma 2, it suffices to show
that each of these are distinct to conclude that we have them all (and have
not double-counted any of them which would imply some are missing).

So suppose, for sake of contradiction, that two of these roots are equal. Let
these roots be ω i and ω j where 0 ≤ i < j ≤ n − 1. Then ω i = ω j .

=⇒ ω j−i = 1 (dividing by ω i )

=⇒ ω l = 1 where 1 ≤ l ≤ n − 1 (letting l = j − i)

=⇒ ω is also an l-th root of unity.

We note that since ω is strictly non-real, the case that l = 1 and l = 2 are
automatically eliminated as these would force ω = ±1 ∈ R. Similarly, l must
also be at least two values below its upper bound n − 1 as ω n = ω n−1 and
ω n = ω n−2 are equivalent to the linear and quadratic cases, respectively.
Hence 3 ≤ l ≤ n − 3. Carrying on, we then have:

1 = ω l = ω n = ω l · ω n−l = 1 · ω n−l = ω n−l , where 3 ≤ n − l ≤ n − 3.

=⇒ ω is an (n − l)-th root of unity.

From Lemma 1, each of ω 0 , ω 1 , ω 2 ,..., ω n−1 is also an l-th and (n − l)-th root
of unity. But the former has only l distinct roots, while the latter has only
n − l distinct roots, both of which are strictly less than n. So by the Pigeon
Hole Principle, some of these roots will occur with repetition.

Now, suppose that p ∈ N such that 0 ≤ p ≤ l. Then:

ω l+p = ω l · ω p = 1 · ω p = ω p . Similarly:

ω (n−1)+q = ω n−1 · ω q = 1 · ω q = ω q , for 0 ≤ q ≤ n − l.

This means that all roots occurring after and including the l-th and (n − l)-th
powers are repeats. Therefore, min{l, n − l} is the maximum possible number
of roots among ω 0 , ω 1 , ω 2 ,..., ω n−1 that are distinct.

2
Case 1: The powers of ω from 0 to min{l, n − l} − 1 are distinct.

If these powers are distinct and we have min{l, n − l} of them, then these
must be all the solutions to the polynomial equation z min{l,n−l} − 1 = 0. These
roots will be uniformly distributed along the circumference of the unit circle,
forming min{l, n − l} identical sectors.

Now we have already established that these roots are also roots of the
polynomial equation z n − 1 = 0. So since these roots are in common and all
n-th roots of unity will also be spread uniformly along the circumference of the
unit circle, the remaining n-th roots will be evenly distributed between each
consecutive pair of min{l, n − l}-th roots, where each sector will be congruent.

Suppose the number of non-mutual n-th roots occurring on each sector is r,

where r 6= 0 as n 6= min{l, n − l}. Then the total number of roots is:

n = min{l, n − l} + r·min{l, n − l}

= (r + 1)·min{l, n − l}

=⇒ n is composite - contradiction!

Case 2: The powers of ω from 0 to min{l, n − l} − 1 are not all distinct.

If there exists a pair of these roots that are equal, then similar to our original
scenario, ∃m ∈ N such that 3 ≤ m ≤ min{l, n − l} − 3 and:

ω m = ω min{l,n−l}−m = ω n = 1

In other words, we have the same situation we had earlier, just with a reduced
number of roots. So the same two cases arise; either the powers of ω from 0 to
the minimum of m and min{l, n − l} − m are all distinct, or they are not. If
they are, then we can simply run the same argument we did in Case 1 and
show that n must be composite. Otherwise, this will just generate an identical
scenario, again with a reduced number of roots.

3
As long as we can eventually arrive at a set of roots that are distinct, we will
get our contradiction. So there is only an issue if there is no smallest set of
distinct roots (among the consecutive powers of ω). We will show that if n is
prime, there cannot be a smallest number of distinct roots, thus forming
another contradiction (since there are finitely-many roots so there has to be a
minimum number of distinct ones).

So suppose that u is the smallest number of roots (starting from ω 0 up to

ω u−1 ) that are distinct, 3 ≤ u < n − 1. Then these powers of ω form all the
u-th roots of unity and all powers of ω exceeding the (u − 1)-th power are
repeats. Since the roots are repeated cyclically in the same order they first
appear and that n is prime, the roots cannot all occur the same number of
times as this would imply that u | n. Hence some of the initial roots will occur
once more than the rest.

Suppose these are the first k roots, where 1 ≤ k ≤ u − 1, which occur s + 1

times (meaning the remaining roots occur s times). Now, using the formula
for the sum of a geometric series:
n−1
1 + z + z 2 + ... + z n−1 = zz−1

=⇒ z n − 1 = (z − 1)(1 + z + z 2 + ... + z n−1 )

Since ω is an n-th root of unity, subbing it in gives zero on both sides. In

particular, the second factor on the RHS is zero since ω 6= 1.

=⇒ 1 + ω + ω 2 + ... + ω n−1 = 0

=⇒ s(ω 0 + ω 1 + ... + ω k−1 + ... + ω u−1 ) + (ω 0 + ω 1 + ... + ω k−1 ) = 0

=⇒ s(0) + (ω 0 + ω 1 + ... + ω k−1 ) = 0 (since the sum of the u-th roots of unity
is zero)
0 1 k−1
=⇒ ω + ω + ... + ω =0

=⇒ (ω − 1)(ω 0 + ω 1 + ... + ω k−1 ) = 0

=⇒ ω k − 1 = 0

4
=⇒ ω is a k-th root of unity

=⇒ there is a smaller number of distinct roots

=⇒ A smallest u does not exist - contradiction!

Therefore, no two powers of ω (from 0 to n − 1) can be equal. We thus have n

distinct n-th roots of unity, which is what we wanted to prove. 2
(ii) Let P (z) = z n − 1. Then the roots of P are the n-th roots of unity which,
from part (i), are ω 0 , ω 1 , ω 2 , ..., ω n−1 . So P (z) has factorization:

P (z) = (z − ω 0 )(z − ω 1 )(z − ω 2 )...(z − ω n−1 )

n−1
(z − ω k )
Q
= (z − 1) (1)
k=1

But from the formula for the sum of a geometric series,

P (z) = z n − 1 = (z − 1)(1 + z + z 2 + ... + z n−1 )

n−1
zk
P
= (z − 1) (2)
k=0

Comparing (1) and (2) gives:

n−1 n−1
k
z k , as required.
Q P
(z − ω ) =
k=1 k=0

(iii) Let z = 1 in the equation above and the result follows.

QED