Cyclotomic Polynomials

Brett Porter
May 15, 2015

Abstract
If n is a positive integer, then the nth cyclotomic polynomial is de-
fined as the unique monic polynomial having exactly the primitive nth
roots of unity as its zeros. In this paper we start off by examining some
of the properties of cyclotomic polynomials; specifically focusing on their
irreducibility and how they relate to primes. After that we explore some
applications of these polynomials, including proofs of Wedderburn’s The-
orem, and when a regular n-gon is constructible with a straightedge and
compass.

1 Introduction
Cyclotomic polynomials are an important type of polynomial that appears fre-
quently throughout algebra. They are of particular importance because for any
positive integer n, the irreducible factors of xn − 1 over the rationals (and in-
tegers) are cyclotomic polynomials. Furthermore, the minimal polynomial of
any nth root of unity over the rationals is a cyclotomic polynomial. Records
indicate that certain cyclotomic polynomials were studied as early as Euler, but
perhaps their most famous use is due to Gauss. Cyclotomic polynomials appear
in his Disquisitiones Arithmeticae, where they play a role in the proof of when
a regular n-gon is constructible with a straightedge and compass (a result we
will examine in more depth later).
We will start off by developing some concepts behind where the cyclotomic
polynomials come from, taking a look at the nth roots of unity, and then mov-
ing on to a formal definition of the nth cyclotomic polynomial. From here we
will explore the polynomials themselves; first looking at general properties such
as their degree and how they relate to each other. The next section will be
dedicated to proving that the cyclotomic polynomials are irreducible over the
integers. After that we will explore how cyclotomic polynomials relate to prime
numbers; starting with a discussion of the Bunyakovsky Conjecture, and then
examining results that may be useful in proving or disproving a special case of
this. The rest of the paper will explore the applications of cyclotomic polyno-
mials and how they can be used in various proofs. We will focus on two main
results; Wedderburn’s Theorem, and when a regular n-gon is constructible with

1
a straightedge and compass. Except where explicitly noted, the notation and
terminology throughout this paper mimics [5].

2 Preliminaries and Definitions

Before we can formally define a cyclotomic polynomial we must first introduce
some concepts.
Definition 2.1 (nth Root of Unity). Let n be a positive integer. A complex
number ω is an nth root of unity if ω n = 1.

It is a well known result that there are n distinct nth roots of unity, which
are given by
2πi 2πi 2πi 2πi
e n ,e n 2 ,...,e n n = {e n k : 1 ≤ k ≤ n}.
2πi
Note that any complex number of the form e n k , where k ∈ Z, will be an nth
2πi
roots of unity. From here on out when we talk about an nth root of unity, e n k ,
we will assume that 1 ≤ k ≤ n.
Theorem 2.2. Suppose n is a positive integer, then the nth roots of unity form
a group under multiplication.
2πi 2πi
Proof. Suppose e n k and e n j are any nth root of unity. It can be shown
from the division algorithm that there exist q, r ∈ Z such that 1 ≤ r ≤ n, and
j + k = n · q + r. It follows that
2πi 2πi 2πi 2πi 2πi
e n j e n k =e n (n·q+r) = e2πiq e n r =e n r .
2πi
Note that e n r is an nth root of unity, and so the nth roots of unity are closed
under multiplication.
Because this multiplication is the same multiplication defined on C, it is
associative. Note that 1n = 1, so 1 is an nth root of unity. If ω is any other
nth root of unity, then 1 · ω = ω · 1 = ω. Hence 1 is the identity for this set.
2πi 2πi
Suppose e n k is any nth root of unity, then e n (n−k) is an nth root of unity.
2πi 2πi
k (n−k) 2πi
Note that e n e n =e = 1. Therefore every nth root of unity has a
multiplicative inverse that is also an nth root of unity, and so the nth roots of
unity form a group under multiplication.
The main reason why we are interested in the nth roots of unity as a group
is because of the following result; which allows us to derive properties of the nth
roots of unity by looking at a more familiar group. For our purposes, we will
say that Zn = {1, 2, . . . , n}.
Lemma 2.3. The function ψ from Zn to the nth roots of unity given by ψ(k) =
2πi
e n k is a group isomorphism.

2
 2πi 2πi
Proof. Suppose e n k is an nth root of unity; then k ∈ Zn , and ψ(k) = e n k .
2πi 2πi
Hence ψ is onto. Suppose j, k ∈ Zn and ψ(k) = ψ(j). In this case, e n k = e n j ,
which happens only if j = k. Therefore ψ is one-to-one.
Again suppose that j, k ∈ Zn . Say that j + k ≡ r (mod n), then it follows
that j + k = n · q + r for some q ∈ Z. Hence,
2πi 2πi 2πi 2πi
ψ(j + k) = ψ(r) = e n r =e n (n·q+r) =e n j e n k = ψ(j)ψ(k).

Thus ψ is operation preserving, and so it is an isomorphism.

Definition 2.4 (Primitive nth Root of Unity). A primitive nth root of unity is
an nth root of unity whose order is n.
It is worth noting that if ω is a primitive nth root of unity; then hωi contains
n distinct elements, and so ω is a generator of the group of nth roots of unity.
Theorem 2.5. If n is a positive integer, then the primitive nth roots of unity
are
2πi
{e n k : 1 ≤ k ≤ n, gcd(k, n) = 1}.

Proof. By [5, Corollary 4 to Theorem 4.2], an element k of Zn has order n if

and only if gcd(n, k) = 1. It follows from Lemma 2.3 that an nth root of unity,
2πi
e n k , has order n if and only if gcd(n, k) = 1.
We have now developed enough background to give a formal definition, and
derive a more convenient formula for, the nth cyclotomic polynomial.
Definition 2.6 (nth Cyclotomic Polynomial). For any positive integer n the
nth cyclotomic polynomial, Φn (x), is given by

Φn (x) = (x − ω1 )(x − ω2 ) . . . (x − ωs ),

where ω1 , ω2 ,. . . , ωs are the primitive nth roots of unity.

It follows from Theorem 2.5 that we can write the nth cyclotomic polynomial
as

2iπk
Y
Φn (x) = (x − e n ). (1)
1≤k≤n
gcd(n,k)=1

Definition 2.7 (Möbius Function). Suppose n is a positive integer with prime

Qr
factorization k=1 pfkk . The function µ : N → N given by

(−1)r if fk = 1 for all k,


µ(n) =
0 if fk > 1 for some k.

is called the Möbius function.

3
 It is worth noting that µ(n) = 0 if and only if n is divisible by a perfect
square. Hence, we will say that µ(n) = 0 if n is not square free. It is also worth
noting that µ is a multiplicative function.
The Möbius function is of concern to us because we can express the nth
cyclotomic polynomials as
Y
Φn (x) = (xd − 1)µ(n/d) . (2)
d|n

A proof of this can be found in [1].

3 General Properties
Now that we have a formal definition and two formulas for the cyclotomic poly-
nomials, we will explore some of their simpler properties. Notice that some
results are easier to prove with Equation (1), while other are easier to prove
with Equation (2); demonstrating the usefulness of both formulas.

Theorem 3.1. If n is a positive integer, then Φn (x) is monic and its degree is
φ(n), where φ is the Euler phi function.
2iπk
Y
Proof. Since Φn (x) = (x − e n ), when written as a product of linear factors
1≤k≤n
gcd(n,k)=1
every x term in Φn (x) has a coefficient of 1. It follows that when these linear
factors are multiplied out, the x term with the largest exponent will have a
coefficient of 1. Furthermore, it is apparent from this formula that the degree of
Φn (x) will be the number of integers, k, such that 1 ≤ k ≤ n and gcd(k, n) = 1.
By definition this is φ(n).
These next few results give us ways to relate different cyclotomic polynomi-
als.
r
Y
Theorem 3.2. Let n be a positive integer and pfkk be the prime factorization
k=1
r
Y
of n. If m = pgkk where for all k, 1 ≤ gk ≤ fk , then Φn (x) = Φm (xn/m ).
k=1

Proof. Suppose d | n but d - m, then d is not square free, and so µ(d) = 0. This

4
means that (xn/d − 1)µ(d) = (xn/d − 1)0 = 1, and therefore,
Y
Φn (x) = (xd − 1)µ(n/d)
d|n
Y
= (xn/d − 1)µ(d)
d|n
Y
= (xn/d − 1)µ(d)
d|m
Y
= ((xn/m )m/d − 1)µ(d)
d|m
Y
= ((xn/m )d − 1)µ(m/d)
d|m

= Φm (x(m/n) ).

Corollary 3.3. Let p be a prime and m a positive integer. If p divides m, then

Φpm (x) = Φm (xp ).
r
Y
Proof. Let pfkk be the prime factorization of m, and assume without a loss
k=1
of generality that p = p1 . It follows that the prime factorization of pm will be
r
Y
pf11 +1 pfkk . Therefore, by Theorem 3.2,
k=2

Φpm (x) = Φm (xpm/m )

= Φm (xp ).

k−1
Corollary 3.4. If p is prime and k is a positive integer, then Φpk (x) = Φp (xp ).
Proof. This follows immediately from Theorem 3.2.
Theorem 3.5. Let p be a prime and m be a positive integer. If p does not
divide m, then Φpm (x)Φm (x) = Φm (xp ).

Proof. Say that d | pm and p - d. Since p is prime; this means that gcd(d, p) = 1.
Hence, by Euclid’s Lemma, d | m. Now suppose d | m; then since p - m, p - d.
Also since m | pm, d | pm. It follows that d | pm and d - p if and only if d | m.
Now since p is prime its only divisors are 1 and p. Since p is not a divisor
of m, this means that gcd(p, m) = 1. It follows that if d is a divisor of m,
then gcd( md , p) = 1. Because µ is a multiplicative function, this means that
µ( mp
d ) = µ( m m
d )µ(p) = −µ( d ).

5
 Suppose that p | d and d | pm. It follows that dp | m, so d = pn for some
integer n where n | m. Now suppose that n | m, then pn | pm, and if we let
d = pn then p | d and d | pm. Thus d | pm and p | d if and only if n | m where
n = dp .
Using this we can now write
Y
Φpm (x)Φm (x) = (xd − 1)µ(pm/d) Φm (x)
d|pm
Y Y
= (xd − 1)µ(pm/d) (xd − 1)µ(pm/d) Φm (x)
d|pm d|pm
p|d p-d
Y Y
= (xpn − 1)µ(pm/pn) (xd − 1)µ(pm/d) Φm (x)
n|m d|m
Y
p d −µ(m/d)
= Φm (x ) (x − 1) Φm (x)
d|m

= Φm (xp )(Φm (x))−1 Φm (x)

= Φm (xp ).

Lemma 3.6. If n is odd, then ω is a primitive nth root of unity if and only if
(−ω) is a primitive 2nth root of unity.
Proof. Suppose n is odd and that ω is a primitive nth root of unity. Let G be
the group of 2nth roots of unity, and note that set of nth roots of unity will be a
subgroup of G. Also note that because n is odd, −1 is not an nth root of unity.
Now ω is an nth root of unity, and thus since ω has order n, hωi will be the
set of nth roots of unity. This means that (−1) ∈ / hωi, but because 2n is even,
(−1) ∈ G. Since G is closed, (−ω) ∈ G, and so (−ω) has order k, where by [5,
Corollar 2 to Theorem 7.1], k | 2n. We now have that 1 = (−ω)k = (−1)k ω k ,
which means that (−1)k = ω k . Because (−1) ∈ / hωi, k cannot be odd. This
means that k = 2j for some j ∈ Z. Now (−1)2j ω 2j = (−ω)2j = 1, so ω 2j =
(−1)2j = 1. Hence, ω j = ±1, but we know ω j 6= −1, so ω j = 1. Thus by [5,
Theorem 4.1], n | j, and since 2j = k | 2n, j | n. This means that j = n, and so
(−ω) has order k = 2n; meaning that (−ω) is a primitive 2nth root of unity.
Now suppose ω is a primitive 2nth root of unity, so ω 2n = 1 and for all
m < 2n, ω m 6= 1. Because ω 2n = 1, ω n = ±1. If ω n = 1 then this contradicts
that ω is a primitive 2nth root of unity. Thus ω n = −1. It follows that (−ω)n =
(−1)n ω n = 1, and so (−ω) is an nth root of unity. Now suppose there exists an
m < n such that (−ω)m = 1. It follows that 2m < n, and ω 2m = (−1)2m ω 2m =
1, which contradicts that ω is a primitive root of order 2m. Hence there does
not exist an integer m < n such that (−ω)m = 1, and so (−ω) is a primitive
nth root of unity.

6
 r
Y
Lemma 3.7. If pfkk is the prime factorization of n, then
k=1

r
pfkk −1 (pk − 1).
Y
φ(n) =
k=1

Proof. See [4, Theorem 5.22].

Theorem 3.8. If n is an odd integer greater than 1, then Φ2n (x) = Φn (−x).
Proof. Let n be odd and ω1 , ω2 ,. . . , ωs be the primitive nth roots of unity. By
Lemma 3.6, −ω1 , −ω 2nth roots of unity. Hence, by
Q2s,. . . , −ωs are the primitive Q s
definition, Φn (x) = k=1 (x − ωk ) and Φ2n (x) = k=1 (x + ωk ).
Note that the degree of Φn (x) will be s, and so by Theorem 3.1, s = φ(n).
Because n is odd, there will be an odd prime, p, such that p | n. Therefore p − 1
will be even. By Lemma 3.7 (p − 1) | φ(n), so φ(n), and hence s, will be even.
It follows that
s
Y
Φn (−x) = (−x − ωk )
k=1
Ys
= (−1)(x + ωk )
k=1
s
Y
= (−1)s (x + ωk )
k=1
s
Y
= (x + ωk )
k=1
= Φ2n (x).

While the next two results are interesting in their own right, they are of
particular importance because of their use in the proofs of later theorems.
Theorem 3.9. Let n be a positive integer, then
Y
xn − 1 = Φd (x).
d|n

Proof. Suppose that ω is a root of Φd (x), where d | n. It follows that ω is a dth

root of unity. Let q be the integer such that n = d · q; then

ω n = (ω d )q = 1q = 1.

It follows that ω is a root of xn − 1.

7
 Now suppose that ω is a root of xn − 1. It follows that ω is an nth root of
unity. Say that the order of ω is d, and note that ω will be a primitive dth root
of unity. Therefore, ω is a root of Φd (x). Since the nth root of unity form a
group of n elements, by [5, Corollary 2 to Theorem 7.1], d | n, and so ω is a
root of Φd (x) for some d that divides n. Q
We have now shown that xn − 1 and d|n Φd (x) share all their roots. Note
Q
that d|n Φd (x) is a product of a collection of monic polynomials, and so it will
be monic. Hence, xn − 1 and d|n Φd (x) are both monic; which means that
Q
they must be equal.
Theorem 3.10. Let n be a positive integer, then the coefficients of Φn (x) are
integers, i.e. Φn (x) ∈ Z[x].

Proof. See [1, Corollary 3].

4 Irreducibility of the Cyclotomic Polynomials

Over the Integers
We will now show that for all positive integers n, Φn (x) is irreducible over Z.
Before we can prove this though we need a few results.

4.1 Lemmas and Interesting Results

Theorem 4.1. A polynomial f (x) over a field F has a multiple zero in some
extension field of F if and only if f (x) and its derivative, f 0 (x), have a common
factor of positive degree in F [x].

Proof. See [5, Theorem 20.5].

Lemma 4.2. Let g(x) and h(x) belong to Z[x], and let h(x) be monic. If h(x)
divides g(x) in Q[x], then h(x) divides g(x) in Z[x].
Proof. We will prove this by induction on the degree of g(x). Let deg g(x) = m
and deg h(x) = n, and say that a is the leading coefficient of g(x). Assume that
m = n, then g(x) = a · h(x), and since a is an integer, this means that h(x)
divides g(x) in Z[x].
Now assume that m > n and for all q such that n ≤ q < m, if l(x) ∈ Z[x]
is a polynomial of degree q that is divisible by h(x) in Q[x], then h(x) divides
l(x) in Z[x]. Let k(x) = g(x) − axm−n h(x) and note that h(x) divides k(x)
in Q[x], deg k(x) < m, and k(x) has integer coefficients. If k(x) = 0, then
g(x) = axm−n h(x), so h(x) divides g(x) in Z[x]. If k(x) 6= 0 then since h(x)
divides k(x), deg k(x) ≥ n. It follows from the induction hypothesis that h(x)
divides k(x) in Z[x], and thus h(x) divides k(x)+axm−n h(x) = g(x) in Z[x].
Lemma 4.3. If g(x) ∈ Zp [x] where p is prime, then (g(x))p = g(xp ).

8
Proof. Consider the function ψ : Zp [x] → Zp [x] given by ψ(f (x)) = (f (x))p .
Suppose f (x), h(x) ∈ Zp [x]. Note that ψ(f (x)h(x)) = (f (x)h(x))p = (f (x))p p

(h(x)) .
p
Now, by [4, Theorem 3.17], if p is prime and 0 < j < p, then p divides j . Be-
cause Zp [x] has characteristic p, it follows from the binomial theorem that

ψ(f (x) + h(x)) = (f (x) + h(x))p

p  
X p
= (f (x))p (h(x))p
j=0
j
p−1  
p p
X p
= (f (x)) + (g(x)) + (f (x))p (h(x))p
j=1
j
p−1
X
= ψ(f (x)) + ψ(h(x)) + 0 · (f (x))p (h(x))p
j=1
= ψ(f (x)) + ψ(h(x)).

Therefore ψ is operation preserving, and hence a homomorphism.

Suppose a ∈ Zp [x] and a is a constant polynomial. If a 6= 0, then gcd(a, p) =
1, and so by Fermat’s Little Theorem, ap ≡ a (mod p). Also 0p = 0, and
therefore if a is a constant polynomial in Zp [x],
Pthen ψ(a) = ap = a.
n
If g(x) ∈ Zp [x], then we may write g(x) = j=0 aj xj . It follows that

(g(x))p = ψ(g(x))
Xn
= ψ( aj xj )
j=0
n
X
= ψ(aj )ψ(xj )
j=0
n
X
= aj xpj
j=0
= g(xp ).

4.2 The Proof

Theorem 4.4. The cyclotomic polynomials Φn (x) are irreducible over Z.
Proof. Let f (x) ∈ Z[x] be a monic irreducible factor of Φn (x). We will show
that every zero of Φn (x) is a zero of f (x).
Since Φn (x) divides xn − 1 in Z[x], there exists a polynomial g(x) ∈ Z[x]
such that xn − 1 = f (x)g(x). Let ω be a primitive nth root of unity that is
a zero of f (x). Let p be a prime that does not divide n; then gcd(p, n) = 1,
and so by [1, Lemma 3], ω p is also a primitive nth root of unity. It follows

9
that 0 = (ω p )n − 1 = f (ω p )g(ω p ), and so ω p is a zero of either f (x) or g(x).
Suppose f (ω p ) 6= 0, then g(ω p ) = 0, so ω is a zero of g(xp ). Because f (x)
is monic, irreducible, and has ω as a zero, by definition f (x) is the minimal
polynomial for ω over Q. Therefore, by [5, Theorem 21.3], f (x) divides g(xp )
in Q[x], and so because f (x) is monic, by Lemma 4.2, f (x) divides g(xp ) in
Z[x]. Say that g(xp ) = f (x)h(x) where h(x) ∈ Z[x]. Let ḡ(x), f¯(x), and h̄(x)
be the polynomials in Zp [x] formed by reducing the coefficients of g(x), f (x),
and h(x) modulo p respectively; then ḡ(xp ) = f¯(x)h̄(x). Now by Lemma 4.3,
(ḡ(x))p = ḡ(xp ) = f¯(x)h̄(x). Since by [5, Corollary to Theorem 18.3], Zp [x]
is a unique factorization domain, f¯(x) and ḡ(x) share a common irreducible
factor, call it k(x). Thus for some m1 (x), m2 (x) ∈ Zp [x], f¯(x) = k(x)m1 (x) and
ḡ(x) = k(x)m2 (x). It follows that in Zp [x],

xn − 1 = f¯(x)ḡ(x) = (k(x))2 m1 (x)m2 (x).

Hence xn − 1 has a multiple zero in some extension field of Zp , and so by

Theorem 4.1, xn − 1 and its derivative, nxn−1 , must have a common factor of
positive degree. Note that every factor of nxn−1 will be of the form cxq , where
c | n and 0 ≤ q ≤ n − 1. Also note that any element of this form with positive
degree cannot divide xn − 1, which means that xn − 1 and nxn−1 cannot share
a common factor. Thus we have reached a contradiction, and so f (ω p ) = 0.
Therefore, if ω is any primitive nth root of unity that is a zero of f (x) and p is
any prime that does not divide n, then ω p is a zero of f (x).
Still assuming that ω is a primitive nth root of unity that is also a zero of
f (x), let ξ be any other primitive nth root of unity. Because ω generates the
group of nth roots of unity, there exists an integer k such that ω k = ξ. Now
by [1, Lemma 3], gcd(k, n) = 1. It follows that k = p1 p2 . . . pr where for all i,
pi is a prime that does not divide n. Therefore ω, ω p1 , (ω p1 )p2 , (ω p1 p2 )p3 , . . . ,
(ω p1 p2 ...pr−1 )pr = ω k are all zeros of f (x) that are primitive nth roots of unity,
in particular ξ is a zero of f (x). Hence every primitive nth root of unity is a
zero of f (x), and so f (x) and Φn (x) share all their zeros. Since by Theorem
3.1, Φn (x) is monic, this means that Φn (x) = f (x), and so Φn (x) is irreducible
over the integers.
Corollary 4.5. The cyclotomic polynomials Φn (x) are irreducible over Q.

Proof. By [5, Theorem 17.2], if a polynomial with integer coefficients is reducible

over Q then it is reducible over Z. The result follows from the contrapositive of
this.

5 Cyclotomic Polynomials and Primes

5.1 Unsolved Questions and Conjectures
A well known unanswered question about cyclotomic polynomials is if n is a fixed
positive integer, then is Φn (m) prime for an infinite number of integer inputs

10
m? It is worth noting that this question is a special case of the Bunyakovsky
Conjecture, an unproved result which states as follows:
Bunyakovsky Conjecture. Suppose f is a polynomial of one variable with
positive degree and integer coefficients. If
1. the leading coefficient of f is positive;
2. f is irreducible over the integers;
3. as n runs over the positive integers, the numbers f (n) are relatively prime
(i.e. the only integer that divides f (n) for all n ∈ N is 1)
then the polynomial f (x) is prime for infinitely many positive integers m.
We have already seen that every cyclotomic polynomials satisfies the first
two conditions; it is left to the reader to verify that they also satisfy the third. It
follows that if the Bunyakovsky Conjecture is true then for all positive integers
n, Φn (m) is prime for an infinite number of integer inputs m.

5.2 Relations Between Cyclotomic Polynomials and Primes

We will now explore some of the relationships between cyclotomic polynomials
and prime numbers. The main focus of this subsection is results that may be
useful in proving whether or not Φn assumes an infinite number of prime values
over the integers.
Theorem 5.1. If p is a prime and k is any positive integer, then Φpk (1) = p.
Proof. For any prime p,
Y
Φp (x) = (xd − 1)µ(p/d)
d|p
p
x −1
=
x−1
p−1
X
= xk .
k=0

p−1
k−1
X
Therefore Φp (1) = 1k = p, and so by Corollary 3.4, Φpk (1) = Φp (1p )=
k=0
Φp (1) = p.
Lemma 5.2. Let p be a prime. Suppose that the polynomial xn − 1 has a
root of multiplicity greater than 1 modulo p, so there exists an integer b and a
polynomial f (x) ∈ Z[x] such that

xn − 1 ≡ (x − b)2 f (x) (mod p).

Then p divides n.

11
Proof. See [1, Lemma 6].
Theorem 5.3. Suppose n is a positive integer, d is a proper divisor of n, and
b is any integer. If p is a common prime divisor of Φn (b) and Φd (b), then p
divides n.

Proof. By Theorem 3.9,

Y
xn − 1 = Φt (x),
t|n

so xn − 1 is divisible by Φn (x)Φd (x). Since Φn (b) and Φd (b) are both divisible
by p, Φn (b) ≡ 0 ≡ Φd (b) (mod p), so b is a root of both Φn (x) and Φd (x) in
Zp [x]. Because Zp is a field, by [5, Corollay 2 to Theorem 16.2], x − b is a factor
of both Φn (x) and Φd (x) in Zp [x], and so (x − b)2 is a factor of xn − 1 in Zp [x].
It follows from Lemma 5.2 that p divides n.
Theorem 5.4. Let n be a positive integer and let b be any integer. If p is a
prime that divides Φn (b), then either p divides n or p ≡ 1 (mod n).
Proof. Let p be a prime divisor of Φn (b). By Theorem 3.9, Φn (b) | bn − 1, and
so p | bn − 1. Therefore, since gcd(bn − 1, b) = 1, p - b. It follows that b (mod p)
is an element of Up (the group of units modulo p). Let k be the order of b in Up .
Since p | bn − 1, bn ≡ 1 (mod p), which by [5, Theorem 4.1] means that k | n.
If k = n, then since Up has p − 1 elements, by [5, Corollary 2 to Theorem
7.1], n | p − 1. In this case p − 1 ≡ 0 (mod n), and so p ≡ 1 (mod n).
If k < n, then since
Y
0 ≡ bk − 1 = Φd (b) (mod p),
d|k

there exists a divisor d of k such that p | Φd (b). Since k | n, this means that d
is a proper divisor of n. It follows from Theorem 5.3 that p | n.
Lemma 5.5. If b is an integer and m and n are positive integers, then

gcd(bm − 1, bn − 1) = |bgcd(m,n) − 1|.

Proof. See [1, Lemma 7].

Theorem 5.6. Let n and m be positive integers. Suppose that

gcd(Φn (b), Φm (b)) > 1

n
for some integer b; then m = pk for some prime p and integer k.
Proof. Suppose the p is a common prime divisor of Φm (b) and Φn (b). Let
m = pα M and n = pβ N where α, β ≥ 0 are integers and M, N are positive
integers not divisible by p. Note that since p | Φm (b) and Φm (b) | bm − 1, p - b.

12
 We will now show that p | ΦM (b). If α = 0, then Φm (b) = ΦM (b), so
p | ΦM (b). If α ≥ 1 then by Corollary 3.3 and Theorem 3.5,

0 ≡ Φm (b)
= Φpα−1 pM (b)
α−1
= ΦpM (bp )
pα
ΦM (b )
= (mod p).
ΦM (bpα−1 )
α−1 α
Multiplying both sides of this congruence by ΦM (bp ) gives us ΦM (bp ) ≡ 0
(mod p). Now since p - b, gcd(p, b) = 1, so b (mod p) ∈ Up . By Lemma 5.5,
p − 1 | pα − 1, so pα − 1 = c(p − 1) for some integer c. Since Up has p − 1
elements, by [5, Corollary 5 to Theorem 7.1],
α α
−1
bp = bp b = (bp−1 )c b ≡ 1c b = b (mod p).

It follows that
α
0 ≡ ΦM (bp ) ≡ ΦM (b) (mod p),

so p | ΦM (b). Similarly p | ΦN (b).

If we show that M = N , we will have that

n pβ N
= α = pβ−α ,
m p M

which proves the theorem. Suppose that M > N . Let t = gcd(M, N ) and note
that t < M . Now since ΦM (b) | bM − 1 and ΦN (b) | bN − 1, and because p
divides both ΦM (b) and ΦN (b), p is a common divisor of both bM −1 and bN −1.
Therefore p | gcd(bM − 1, bN − 1). By Lemma 5.5, gcd(bM − 1, bN − 1) = |bt − 1|,
so p | bt − 1. Hence
Y
0 ≡ bt − 1 = Φd (b) (mod p),
d|t

so there exists a divisor d of t such that p | Φd (b). Since t | M and t < M , d is

a proper divisor of M , and because p | ΦM (b), by Theorem 5.3, p | M , which
is a contradiction since we assumed that p - M . Thus M ≤ N . A similar proof
gives us N ≤ M , which means that M = N .

6 Wedderburn’s Theorem
In this section we will look at an interesting application of the cyclotomic poly-
nomials, but first some background. A division ring is a ring with unity in
which every nonzero element a has a multiplicative inverse, i.e. an element x
such that ax = xa = 1 (we will use 1 to denote the multiplicative identity of

13
a ring). If R is a ring then we use R∗ to denote the nonzero elements of R;
note that if R is a division ring then R∗ is a group under multiplication. For an
element g of a group G, I(g) denotes the centralizer of g, which is the set of all
element of G that commute with g.
xn − 1
Lemma 6.1. If 0 < r < n and r divides n, then Φn (x) divides in Z[x].
xr − 1
Proof. By Theorem 3.9, for all positive integers m, xm
Q
− 1 = d|m Φd (x).
Therefore
Q
xn − 1 d|n Φd (x)
r
= Q
x −1 d|r Φd (x)
Y
= Φd (x).
d|n
d-r

xn − 1
Because r < n, n divides n but not r, and so Φn (x) divides .
xr − 1
Lemma 6.2. Let G be a group and suppose a is an element of G, then a is in
a conjugacy class with one element if and only if a is in the center of G.
Proof. Suppose a is in a conjugacy class with only itself; then for all g in G,
g −1 ag = a, so ag = ga. By definition a commutes with every element of G, so
a is in the center of G.
Now suppose a is in the center of G, then I(a) = G, and so by [6, Lemma
13.7], the conjugacy class containing a has

|G| |G|
=
|I(a)| |G|
= 1

element.
Lemma 6.3. If b > 1 and n and r are positive integers such that br − 1 divides
bn − 1, then r divides n.
Proof. If br − 1 divides bn − 1 then gcd(br − 1, bn − 1) = br − 1. By Lemma 5.5,
gcd(br − 1, bn − 1) = bgcd(r,n) − 1. Thus r = gcd(r, n), and so r divides n.
Now we have all the information we need to prove the main result of this
section.
Theorem 6.4 (Wedderburn’s Theorem). A finite division ring is a field.

Proof. Let D be a finite division ring and let

K = {a ∈ D : ax = xa for all x ∈ D}

14
be the center of D. Note that K is closed under subtraction and multiplication
(meaning it will be a ring), every nonzero element in K will also have its inverse
in K, and K is commutative, making it a finite field. Thus because every finite
field has prime power order, |K| = q where q = pm for some prime p and non-
negative integer m. Also note that D is a vector space over K, say of dimension
n, so |D| = q n . Assume that n > 1. Now for all a ∈ D∗ , I(a) ∪ {0} is a division
ring, and K ⊆ I(a) ∪ {0}. Therefore I(a) ∪ {0} is a vector space over K, so for
some integer r, |I(a) ∪ {0}| = q r , and thus |I(a)| = q r − 1. Note that for all
a ∈ D∗ , a · 1 = a = 1 · a, so 1 ∈ I(a). Thus |I(a) ∪ {0}| ≥ 2, so q r ≥ 2. It follows
that r > 0, since if r = 0 then q r = 1  2. Also, by [5, Theorem 3.6], I(a) is a
subgroup of D∗ , so by Lagrange’s Theorem q r − 1 divides q n − 1, and therefore
by Lemma 6.3, r divides n.
Let C1 , C2 , . . . , Cs be the conjugacy classes of D∗ ; then the class equation is
s
X
|D∗ | = |Cj |.
j=1

Note that the center of D∗ will be the same as the center of D without the
element 0, and thus the center of D∗ has q − 1 elements. By Lemma 6.2 there
are exactly q − 1 conjugacy classes of D∗ with one element; assume without
a loss of generality that these classes are C1 , C2 , . . . , Cq−1 . The class equation
then becomes
Xs
|D∗ | = (q − 1) + |Cj |.
j=q

For all j such that q ≤ j ≤ s, let aj be an element of Cj , then there exists an

integer rj such that 0 < rj < n, rj divides n, and |I(aj )| = q rj − 1. Therefore
qn − 1
by [6, Lemma 13.7], for all j ≥ q, |Cj | = rj , and so the class equation
q −1
becomes
s
X qn − 1
q n − 1 = (q − 1) + ,
j=q
q rj − 1

or, in another form

s
X qn − 1
(q n − 1) − = q − 1.
j=q
q rj − 1

qn − 1
By Lemma 6.1, for all j such that q ≤ j ≤ s, Φn (q) divides , and thus
q rj − 1
s
X qn − 1
because Φn (q) divides q n − 1, Φn (q) divides (q n − 1) − = q − 1. On
j=q
q rj − 1
the other hand recall that
2iπk
Y
Φn (x) = (x − e n ),
1≤k≤n
gcd(n,k)=1

15
and so (using the reverse triangle inequality),
2iπk
Y
|Φn (q)| = |q − e n |
1≤k≤n
gcd(n,k)=1
2iπ
Y
≥ ||q| − |e n |k |
1≤k≤n
gcd(n,k)=1
Y
= (q − 1)
1≤k≤n
gcd(n,k)=1
> q−1
when n > 2. When n = 2, Φn (q) = q + 1, which is greater than q − 1. Thus
Φn (q) cannot divide q − 1, and so we have a contradiction, meaning that it is
impossible that n > 1. Therefore D is a vector space over K of degree 1, which
means that D = K, and hence D is commutative. Since D is a commutative
ring with unity where every element is a unit, by definition it is a field.

7 Constructible Regular N-Gons

Another application of cyclotomic polynomials that we will explore is when a
regular n-gon (in R2 ) is constructible with a straightedge and compass (from
here on out we will assume that n-gon means regular n-gon). Before we do this
though we need to solidify exactly what we mean by constructible. Suppose
that A is a subset of R2 . We say that a line is constructible from A if it passes
through two distinct points that are either constructible from or in A. We say
that a circle is constructible from A if its center is a point that is constructible
from or in A and some other point that is constructible from or in A lies on the
circle. A point in R2 is constructible from A if it is a point shared by two distinct
lines, two distinct circles, or a line and a circle constructible from A. An n-gon
is constructible from a set A if all the points at its vertices are constructible
from A.
We say that an object (line, circle, point, or n-gon) is constructible if it is
constructible from the set A = {(0, 0), (1, 0)}. It is worth noting that it can be
shown that every point in Q2 is constructible.
From here on out we will identify the point (a, b) ∈ R2 with the complex
number a + bi. Finally, recall that a Fermat prime is any prime number of the
j
form 22 + 1, where j is a nonnegative integer.
Theorem 7.1. The set of constructible complex numbers form a subfield of C.
Proof. See [8, Section 7.7; Theorem 3].
Theorem 7.2. Suppose ω is a complex number with minimal polynomial f (x)
over Q. If F is the splitting field for f (x) over Q, then ω is constructible if and
only if [F : Q] = 2l , where l is a nonnegative integer.

16
Proof. See [7, Theorem 9.1].
Theorem 7.3. Suppose n is a positive integer. Let F be the spitting field for
Φn (x) over Q, then [F : Q] = φ(n).
2πi
Proof. Let ω = e n , and note that ω is a zero of Φn (x). It follows that F
contains ω, so Q(ω) ⊆ F . Let ξ be any zero of Φn (x), then ξ is a primitive nth
root of unity. Hence, by Theorem 2.5, ξ = ω k for some k such that 1 ≤ k ≤ n
and gcd(k, n) = 1. It follows that ξ ∈ Q(ω). Therefore all the zeros of Φn (x) are
contained in Q(ω), and so Φn (x) splits in Q(ω). Because by definition Φn (x)
cannot split in any proper subfield of F , this means that Q(ω) = F . By Theorem
3.1, deg Φn (x) = φ(n), and so by [5, Theorem 20.3], {1, ω, ω 2 , . . . , ω φ(n)−1 } is a
basis for Q(ω) over Q. By definition this means that [F : Q] = [Q(ω) : Q] =
φ(n).
Lemma 7.4. If 2n + 1 is prime for some positive integer n, then n = 2k for
some nonnegative integer k (i.e. 2n + 1 is a Fermat prime).
Proof. We will prove this by the contrapositive. Suppose that n 6= 2k for any
integer k. It follows that n must have an odd factor r > 1, and so n = t · r for
some t ∈ N. One can see by routine verification that
r−1
X
(2t + 1)( (−1)j 2(r−1−j)t ) = 2n + 1.
j=0

Since 0 < t < n we have that 2 < 2t + 1 < 2n + 1. Therefore 2n + 1 has a

nontrivial factor, making it composite.
Theorem 7.5. Suppose n is a positive integer, then φ(n) is a power of 2 if
and only if n has the form 2k p1 p2 . . . pr , where k ≥ 0 and the pj ’s are distinct
Fermat primes.
Qr f
Proof. By [4, Theorem 5.21], φ is a multiplicative function. Thus if j=1 pj j is
Qr f
the prime factorization of n, then φ(n) = j=1 φ(pj j ). Hence φ(n) is a power
of 2 if and only if for all primes p and positive integers f such that pf k n, φ(pf )
is a power of 2.
Suppose k ≥ 1, then by Lemma 3.7, φ(2k ) = 2k−1 , which is clearly a power
of 2.
Now Suppose p is an odd prime and φ(pf ) = 2m for some f, m ∈ N. Again
by Lemma 3.7, φ(pf ) = pf −1 (p − 1). This means that if f > 1, then p | φ(pf ),
which would contradict that φ(pf ) is a power of 2. Hence f = 1, and so

p − 1 = φ(p) = φ(pf ) = 2m .

Therefore p = 2m + 1, which by Lemma 7.4 means that p is a Fermat prime. If

j
we suppose that p is a Fermat prime, then p = 22 + 1 for some j ∈ N. In this
j
case, φ(p) = 22 , which is a power of 2. Thus if f ∈ N and p is an odd prime,
then φ(pf ) is a power of 2 if and only if f = 1 and p is a Fermat prime.

17
 Therefore, φ(n) is a power of 2 if and only if for all primes p and f ∈ N such
that pf k n, either p = 2, or p is a Fermat prime and f = 1.
Theorem 7.6. It is possible to construct the regular n-gon with a straightedge
and compass if and only if n has the form 2k p1 p2 . . . pr , where k ≥ 0 and the
pj ’s are distinct Fermat primes.

Proof. Note that if an n-gon is constructible, then the point in the center of the
n-gon is constructible. Hence we may assume that the n-gon we are trying to
construct is centered at the origin. Also note that if we are able to construct
an n-gon then we can construct an n-gon where every point has distance 1 from
the origin. Thus we may assume that the vertices of the n-gon we are trying to
construct are located on the unit circle. It follows that we may assume that the
vertices of the n-gon will be located at the nth roots of unity.
2πi
Now if an n-gon is constructible, then the point e n must be constructible.
2πi
Furthermore, if e n is constructible, then since every nth root of unity is a
2πi
power of e n , by Theorem 7.1 every nth root of unity will be constructible.
2πi
Hence, an n-gon is constructible if and only if the point e n is constructible.
2πi
Notice that e n is a zero of Φn (x), and thus because Φn (x) is irreducible
2πi
over Q and monic, it is the minimal polynomial for e n over Q. It follows from
2πi
Theorems 7.2 and 7.3 that e n is constructible if and only if φ(n) = 2l , where l
is a nonnegative integer. By Theorem 7.5, this happens if and only if n has the
form 2k p1 p2 . . . pr , where k ≥ 0 and the pj ’s are distinct Fermat primes.

8 Acknowledgments
The author would like to thank Dr. Patrick Keef for his support and guidance
on this project, along with suggesting the topic in the first place. He would also
like to acknowledge Austin Sloane for editing this work and ensuring its quality.

References
[1] Yimin Ge, Elementary Properties of Cyclotomic Polynomials,
http://www.yimin-ge.com/doc/cyclotomic polynomials.pdf
[2] Yves Gallot, Cyclotomic Polynomials and Prime Numbers,
http://yves.gallot.pagesperso-orange.fr/papers/cyclotomic.pdf
[3] Pantelis A. Damianou, On Prime Values of Cyclotomic Polynomials,
http://arxiv.org/pdf/1101.1152.pdf

[4] Robbins, Neville. Beginning Number Theory. 2nd ed. Sudbury, Mass.: Jones
and Bartlett, 2006. Print.
[5] Gallian, Joseph A. Contemporary Abstract Algebra. 7th ed. Belmont, CA:
Brooks/Cole, Cengage Learning, 2010. Print.

18
[6] Stewart, Ian. Galois Theory. 2nd ed. London: Chapman and Hall, 1989.
Print.
[7] Kuh, Devin. “Constructible Regular N-gons.”(2013). Web. 4 May 2015.
https://www.whitman.edu/Documents/Academics/Mathematics/Kuh.pdf.

[8] Goldstein, Larry Joel. Abstract Algebra; a First Course. Englewood Cliffs:
Prentice-Hall, 1973. Print.
[9] Warner, Seth. Modern Algebra I. Englewood Cliffs: Prentice-Hall, 1965.
Print.

19