. . . . . . . . . . . . . . . . . . . . . . . . . .

Olympiad Training for Individual Study

Integer Polynomials
Evan Chen《陳誼廷》
3 December 2024
DNY-INTPOLY

》
廷
誼 se
陳 U
n《 al
h e rn
C nt e
n
a , I
Ev I S
y
B O T

OTIS, © Evan Chen, internal use only. Artwork contributed by Aaryan Vaishya.

1
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

§1 Reading
You should read §53.1-§53.3 of Napkin v1.5 (Algebraic Integers). But there is a lot of
overlap with the stuff to follow.

§1.1 Rings/fields/etc
Here is the short version (again, see Napkin for full details). A ring is a set equipped
with a commutative addition/subtraction and multiplication with an identity for both.
Examples include Z/n for composite n, Z, and R[x] (polynomials with coefficients in R).
A field is a ring with division as well, such as Fp , Q, R, C, but not any of the previous
examples. (All fields are also rings by definition though.)
Definition 1.1. A unit is an element which has a multiplicative inverse.

》
Definition 1.2. An irreducible element is a nonzero x ∈ R which can’t be the product

廷
of two non-units.

誼 se
Definition 1.3. A unique factorization domain (UFD) is a ring R satisfying

• unique factorization of elements into irreducibles, up to multiplication by units;

陳 U
• the ring R is also an integral domain, i.e. the product of two nonzero elements of

《 al
R is nonzero. (We won’t ever worry about this condition, and you can basically

n
forget about it.)

h e rn
§1.2 Minimal polynomials and Galois conjugates

C t e
This is the first section of Napkin chapter 46.

n n
In high school precalculus, you’ll often be asked to find the roots of some polynomial

a , I
with integer coefficients. For instance,

Ev I S
x3 − x2 − x − 15 = (x − 3)(x2 + 2x + 5)

y T
has roots 3, −1 + 2i, −1 − 2i. Or as another example,

B O
x3 − 3x2 − 2x + 2 = (x + 1)(x2 − 4x + 2)
√ √
has √
roots −1, 2 + 2, 2 − 2. You’ll notice that the irrational roots, like −1 ± 2i and
2 ± 2, are coming up in pairs. In fact, I think precalculus explicitly tells you that the
complex roots come in conjugate pairs. More generally, it seems like all the roots of the
√
form a + b c come in “conjugate pairs”. And you can see why.
But a polynomial like
x3 − 8x + 4
has no rational roots. (The roots of this are approximately −3.0514, 0.51730, 2.5341.)
Or even simpler,
x3 − 2
√
has only one real root, 3 2. These roots, even though they are irrational, have no
“conjugate” pairs. Or do they?
Let’s try and figure out exactly what’s happening. Let α be any complex number. We
define a minimal polynomial of α over Q to be a polynomial such that

• P (x) has rational coefficients, and leading coefficient 1,

2
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

• P (α) = 0.
• The degree of P is as small as possible. We call deg P the degree of α.

√
Example (a) 2 has minimal polynomial x2 − 2.
√
(b) The imaginary unit i = −1 has minimal polynomial x2 + 1.
2πi
(c) A primitive pth root of unity, ζp = e p , has minimal polynomial xp−1 + xp−2 +
· · · + 1, where p is a prime.
√
Note that 100x2 − 200 is also a polynomial of the same degree which has 2 as a root;
that’s why we want to require the polynomial to be monic. That’s also why we choose to

》
work in the rational numbers; that way, we can divide by leading coefficients without
worrying if we get non-integers.

廷
Why do we care? The point is as follows: suppose we have another polynomial A(x)
such that A(α) = 0.

誼 se
Claim — P (x) actually divides A(x)!

陳 U
That means that all the other roots of P will also be roots of A.

《 al
Proof, by contradiction. If not, by polynomial long division we can find a quotient and

n
remainder Q(x), R(x) such that

h e rn A(x) = Q(x)P (x) + R(x)

C e
and R(x) 6≡ 0. Notice that by plugging in x = α, we find that R(α) = 0. But deg R <

t
deg P , and P (x) was supposed to be the minimal polynomial. That’s impossible!

n
a , I n
It follows from this and the monotonicity of the minimal polynomial that it is unique

v
(when it exists), so actually it is better to refer to the minimal polynomial.

E I S
Exercise 1.4. Can you find an element in C that has no minimal polynomial?

y T
Let’s look at a more concrete example. Consider A(x) = x3 − 3x2 − 2x + 2 from the

B O
√
beginning. The minimal
√ polynomial of 2 + 2 is P (x) = x2 − 4x + 2 (why?). Now we
know that if
√ 2 + 2 is a root, then A(x)
√ is divisible by P (x). And that’s how we know
that if 2 + 2 is a root of A, then 2 − 2 must be√a root too. √
As another example, the √minimal polynomial of 3 2 is x3 − 2.√So 3 2 actually has two
conjugates, namely, α = 3 2 (cos 120◦ +√i sin 120◦ ) and β = 3 2 (cos 240◦ + i sin 240◦ ).
Thus any polynomial which vanishes at 3 2 also has α and β as roots!

Remark (Important but tautological: irreducible ⇐⇒ minimal). Let α be a root of

the polynomial P (x). Show that P (x) is the minimal polynomial of α if and only if it is
irreducible.

In general, for any α which is the root of a monic irreducible P (x) ∈ Q[x], we say
the roots of P are Galois conjugates of α (over Q). For technical simplicity, α is
considered
√ √its own Galois conjugate. So for√ example√ over
√ Q the Galois conjugates of
√ 2 are ± 2 and the Galois conjugates of 3
2 are 3
2, 3
2 (cos 120◦ + i sin 120◦ ), and
3
2 (cos 240 + i sin 240 ).
◦ ◦

We’ve done this whole section over the field Q, but actually you can replace Q with
any other field and things will work exactly the same. You’ll sometimes do this with Fp
replacing Q.

3
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

§1.3 Algebraic closures

Definition 1.5. The algebraic closure of a field k is, informally, the set of possible roots
of a polynomial with coefficients in k, and is denoted k.

Example
The algebraic closure of R is C.

These days teenagers often freak out when they see an algebraic closure other than
C for the first time, because they have no idea what it looks like. So I wrote up this
analogy to make it less unsettling:

• Ancient Greek algebra lived in a world of rational numbers, with no concept of

》
real number. Of course, if you do this, you get polynomials like x2 + 1 or x2 − 2
with no rational roots. If you’re Greek, and need to work with the roots anyways,

廷
what you do is imagine there’s a field C (of complex numbers in our world, but

誼 se
shh) containing Q. You have no idea what its elements are, but you just grit your
teeth, because you don’t need to know anything about e.g. the complex plane to

陳 U
trod through. They’re “imaginary” for a reason.

• In fact, the Greeks can’t even get the whole set C: for example, there’s no polynomial

《 al
with rational coefficients that has π or e as a root (they are transcendental). The

n
set Q of things the Greeks do get are called algebraic numbers. This is quite

e rn
literally defined as the set of possible roots of a rational polynomial. Unlike with C

h
where you could describe it as “a + bi”, there isn’t a good alternate definition of Q.

C t e
But the Greeks don’t notice anything missing from their lives in Q: if you’ll never

n n
run into π anyways (e.g. geometry allergy), it may as well not exist.

va , I
• Anyways, the point of the analogy is that you can do this for any field. For example,

S
sometimes people will take an algebraic closure of Fp to get Fp . And everything’s

E I
still a-OK.

y T
• The reason Fp is scary compared to Q is that
√ you’re used to being able to treat

B O
elements of Q as complex numbers like 3 + 2i, which you can visualize. But since
you’re doing algebra and not geometry, the absence of a good picture is irrelevant.
Just close your eyes.

§2 Lecture Notes
§2.1 Warm-up example
Before doing some fancy algebraic number theory, let’s make sure we have the basics
down.

Example 2.1 (USAMO 1974/1)

Let a, b, c be distinct integers, and let P be a polynomial with integer coefficients.
Prove that it’s impossible for P (a) = b, P (b) = c, P (c) = a.

74AMO1 Walkthrough. Solve this problem by using the fact x − y | P (x) − P (y) for any integer
polynomial P .

4
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

§2.2 About this unit

The title is a bit of a misnomer as we’ll be working also with polynomials with coefficients
in either Fp , Q, R, and C. A better title for this lecture would be “algebra of polynomials”.
We’ll be dealing with the technical aspects of polynomials in this lecture.
Philosophical musing: polynomials feel powerful because there are three ways to think
of them (which can all be interchanged): in terms of the values, the coefficients, or the
roots.
The following lecture notes are meant to be a quick reminder of what you saw in the
reading. They’re pretty brief; if you’re not familiar with some of these things, you should
go see the full details in Napkin again.

§2.3 Fundamental theorem of algebra

》
Facts you probably already know:

廷
• Fundamental theorem of algebra: every polynomial with complex coefficients
decomposes into linear factors (some refer to this as “splits completely”).

誼 se
• Over a field, a polynomial of degree d has at most d roots.

陳 U
• For any field F , the ring of polynomials F [x] has a division algorithm.1

§2.4 UFD

n《 al
h e rn
Theorem 2.2

C e
Let R be a UFD. Then R[x] is a UFD.

n I nt
Hence by induction R[x1 , . . . , xn ] is a UFD.

a ,
So polynomial factorization works as well as possible (up to units) — you can uniquely

v S
decompose them into irreducible polynomial up to multiplication by units. (In particular,

E I
two polynomials always have a GCD.)

y T
§2.5 Irreducible and minimal polynomials

B O
Irreducible and minimal polynomials are the same, in the following sense:

Lemma 2.3
A polynomial f is the Q-minimal polynomial for an α ∈ C if and only if f (α) = 0
and also f is irreducible.

As mentioned earlier, the other roots of f are called the Q-Galois conjugates of α.
Classic examples:
√ √
• 3 + 2 has Galois conjugates 3 ± 2.

• A primitive nth root of unity will have the other primitive nth roots of unity as
Galois conjugates, since Φn is irreducible (proved later).

1
Relevant buzzword: these are called Euclidean domains.

5
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

§2.6 Irreducible polynomials usually don’t have double roots

Proposition 2.4
In any “reasonable” situation, irreducible polynomials can’t have double roots.

This includes polynomials over the following fields:

• Q, R, C, and more generally any field with characteristic zero.
• Fp , and more generally any field with finitely many elements.
• any algebraically closed field, like Fp .
We won’t see a non-example of this proposition in olympiad context. Correct buzzword
though: “perfect fields”, “separable extension”.

》
Proof for Q and fields of characteristic zero. If f is irreducible and has a double root α,

廷
then it’s the minimal polynomial of its double root α. But the derivative f 0 has f 0 (α) = 0,
contradiction to minimality.

誼 se
The problem with Fp is that a nonconstant polynomial f could have derivative equal

陳 U
to zero — for example, f (x) = xp has derivative zero in Fp .

《 al
§2.7 Symmetric polynomials

n
e rn
We may use Vieta’s relations every so often. So let’s say a, b, c, d are the roots of

h
9X 2 + 2X + 1 (or whatever). You may have the intuition that if you wanted
X 4 + 7X 3 + P

e
to calculate sym a2 b, it ought to be possible from Vieta. In fact it is.

C t
Definition 2.5. A symmetric polynomial in variables X1 , . . . , Xn with coefficients in

n n
I
R is a polynomial which is unchanged by permutations. The elementary symmetric

va ,
polynomials are the ones in Vieta relations:

E I S
P1 = X1 + · · · + Xn

y T
X
P2 = Xi Xj

B O
i<j
..
.
Pn = X1 . . . Xn

Theorem 2.6 (Fundamental theorem of elementary symmetric polynomials)

For any commutative ring R, every symmetric polynomial is uniquely written as an
element of R[P1 , . . . , Pn ].

This is quite strong; not even division is necessary. So sym a2 b in the earlier dummy
P
example will actually be an integer, no denominators needed in the process.

§2.8 Algebraic integers and numbers

Definition 2.7. An α ∈ C is an algebraic number if it is the root of some polynomial
with rational coefficients. The set of algebraic numbers is denoted Q.
An α ∈ C is an algebraic integer if it is the root of some monic polynomial with
integer coefficients. The set of algebraic integers is denoted Z.

6
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

Note that in both definitions, we can replace “polynomial” with “irreducible polynomial”.
(Indeed, if f (α) = 0, then some irreducible factor of f has α as a root.)
From the reading you should know the following useful facts:

• Z is a ring.

• Q is a field.

• Z ∩ Q = Z.

The first two facts are actually implied by the fundamental theorem of elementary
symmetric polynomials. We give a proof of one case.

Proof that Z is closed under addition. Let α and β be algebraic integers, and let α1 , . . . ,
αn and β1 , . . . , βm be their Galois conjugates. Set

》
廷
YY
P = (X − (αi + βj )) .
i j

誼 se
Now full expand. For every monomial β1e1 . . . βm
em , its coefficient is a symmetric polynomial

in α• therefore is an integer. Then re-apply the result for β• to conclude all coefficients

陳 U
of P are integers, as desired.

《 al
§2.9 Examples involving algebraic integers

n
e rn
h
Example 2.8 (Niven’s theorem)

e
Prove that if cos(qπ) is rational for some rational number q, then cos(qπ) ∈

C t
{−1, −1/2, 0, 1/2, 1}.

NIVEN

n
a , I n
Walkthrough. This is a classical result.

v S
(a) Show that cos(qπ) ± i sin(qπ) is an algebraic integer.

y E I
(b) Prove that 2 cos(qπ) is a rational algebraic integer, hence in Z.

T
B O
(c) Prove the quoted result.

(There is actually an elementary solution, too, which is less well-known; see https:
//arxiv.org/pdf/1006.2938.pdf.)

Example 2.9 (USA TST 2007/3)

Let θ ∈ (0, π/2) be an angle. Assume cos θ is irrational, but cos kθ and cos [(k + 1)θ]
are rational for some positive integer k. Prove that θ = π/6.

07USATST3 Walkthrough. Let’s dispense first of an edge case:

(a) If θ is a rational multiple of π, use Niven’s theorem to solve the problem.

Now assume that θ is not a rational multiple of π. We will prove that if cos(kθ)
and cos((k + 1)θ) are rational then cos(θ) is rational. Again, the idea is to consider
z = cos(θ) + i sin θ, which gives us a way of dealing with cos(kθ).

(b) Using the fact that cos(kθ) is a rational, write down a monic polynomial P (x) ∈ Q[x]
of degree 2k which has z as a root.

7
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

(c) Do the same using cos((k + 1)θ) ∈ Q[x]. This gives a monic polynomial Q(x) ∈ Q[x]
of degree 2k + 2.

(d) Identify all the roots of P and Q.

(e) Show that gcd(P, Q) = (x − z)(x − z −1 ). (You will need the fact that θ ∈
/ πQ.)

(f) The gcd also lives in Q[x]. Use this to conclude that cos θ is rational, end proof.

Example 2.10 (Kronecker)

Let α be an algebraic integer. Suppose all its Q-Galois conjugates have absolute
value one. Prove that αN = 1 for some positive integer N .

》
ZD499500 Walkthrough. This one’s a bit of a gem.
Let α1 = α, α2 , . . . , αn denote the n Galois conjugates. Across integers e, consider

廷
the polynomial

誼 se
Pe (X) = (X − α1e ) . . . (X − αne ).
Thus P1 (X) is the minimal polynomial of all the αi .

陳 U
(a) Prove that Pe (X) ∈ Z[x] for every e ≥ 1.

《 al
(b) Show that the number of possible polynomials Pe is at most .

n
n n+1


n/2

e rn
(c) Follow through using pigeonhole principle.

C h e
t
§2.10 Two more technical examples

n I n
Example 2.11 (HMMT 2017 A9)

va ,
What is the period of the Fibonacci sequence modulo 127?

E I S
17HMMTA9 Walkthrough. In what follows let p = 127. Our hope is to use the Fibonacci formula

y T
αn − β n

B O
Fn =
α−β
√ √
where α = 1+ 5
2 and β = 2 .
1− 5
Unfortunately:

(a) Show that (5/p) = −1. Thus we cannot find α, β ∈ Fp .

The idea is that we need to instead work in the so-called finite field
√ n √ o
Fp2 := Fp [ 5] = a + b 5 | a, b ∈ Fp .

I’ll explain the reason for the notation Fp2 later, but for not, just think of it as that field
√
a + b 5. Note that this field Fp2 still has a copy of Fp inside it, namely the numbers
√
of the form a = a + 0 5. They still behave as we expect, for example, ap ≡ a (mod p)
since p | ap − a.

(b) Convince yourself this really is a field. How many elements does it have?

(c) Show that xp = 1 for any nonzero x ∈ Fp2 . (This is the analog of Fermat’s little
2 −1

theorem.)

8
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

Now we take advantage of the fact that α, β ∈ Fp2 .

(d) Consider the so-called Frobenius endomorphism

σ : Fp2 → Fp2 σ(t) = tp .

Determine its fixed points. (Hint: you should already know the roots of X p − X.)

(e) Deduce that σ(α) = αp 6= α.

(f) Prove that σ(x + y) = σ(x) + σ(y) by using the binomial theorem, and also that
σ(xy) = σ(x)σ(y). Thus the Frobenius endomorphism deserves its name.

(g) Consider the polynomial

》
P (X) = X 2 − X − 1.
Prove that P (σ(x)) = σ(P (x)) for any x ∈ Fp2 . (Actually, this holds for any

廷
P ∈ Fp [X].)

誼 se
(h) Put together (e) and (g) to deduce that σ(α) = αp = β, by showing σ(α) is a root
of P not equal to α.

陳 U
(i) Find σ(β).

《 al
Now we are in business: we can compute Fibonacci numbers using

n
e rn
αn − β n

h
Fn = .
α−β

C t e
(j) Using (h) and (i), show that Fp ≡ −1 (mod p), Fp+1 ≡ 0 (mod p).

n I n
(k) Prove that F2p+1 ≡ 1 (mod p) and F2p+2 ≡ 0 (mod p).

va ,
(l) Conclude that the period of the Fibonacci sequence modulo p divides 2p + 2 = 256

E I S
but not p + 1 = 128. Deduce the answer 256.

y T
The solution notes contain some extended comments about the higher algebra that this

B O
problem introduces.

Example 2.12
Show that the nth cyclotomic polynomial Φn (x) is irreducible for every integer n.

CYCIRRED Walkthrough. Assume Φn (x) = f (x)g(x), where f is irreducible. The main lemma is
the following: if p is a prime not dividing n and ζ is a root of f , then so is ζ p .
We begin by proving this lemma.

(a) Assume not. Show that f (x) divides g(xp ).

(b) Use the fact that g(xp ) ≡ g(x)p (mod p) to show that Φn (x) (mod p) has double
roots in Fp .

(c) Differentiate xn − 1 to get a contradiction.

Now with the lemma, we are basically done.

(d) Using the lemma together with Dirichlet’s theorem or otherwise, deduce the problem
statement.

9
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

§2.11 Footnote on Hensel’s lemma

Occasionally, you may be working with polynomials modulo pe for e ≥ 2. In this situation,
you can often lift roots from Fp into the non-fields Z/pe Z.

Lemma 2.13 (Hensel’s lemma)

Let n ≥ 1 be integers, p a prime and f (x) ∈ Z[x]. Suppose r satisfies f (r) ≡ 0
(mod pn ) and f 0 (r) 6≡ 0 (mod p). Then for every m > n, there exists a unique s
such that f (s) ≡ 0 (mod pm ) and s ≡ r (mod pn ).

Proof. The idea is to prove the identity

》
f (x + pn ) ≡ f (x) + pn f 0 (x) (mod pn+1 ).

廷
Even when Hensel’s lemma is not used directly, the identity just mentioned in the proof
pops up every so often.

誼 se
Relevant video if you like watching Evan goof off on YouTube: https://www.youtube.
com/watch?v=D0OY2h3MkSQ

陳 U
n《 al
h e rn
C nt e
n
a , I
Ev I S
y
B O T

10
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

§3 Problems
Instructions: Solve [45♣]. If you have time, solve [52♣]. Problems with red weights are mandatory.

I’m a magical princess from another dimension!

Star Butterfly introducing herself to Marco Diaz

14TWNTST1J2
[2♣] Problem 1 (Taiwan TST 2014/1J/2). For a fixed integer k, determine all polyno-
mials f (x) with integer coefficients such that f (n) divides (n!)k for every positive integer
n.
13INMO3
[2♣] Problem 2 (INMO 2013/3). Let a, b, c, d be positive integers such that a ≥ b ≥
c ≥ d. Prove that the polynomial x4 − ax3 − bx2 − cx − d has no integer roots.

》
20INMO5
[2♣] Problem 3 (INMO 2020/5). An infinite family of equidistant parallel lines is

廷
colored black in the plane. Find all integers n ≥ 3 for which one can draw a regular
n-gon all of whose vertices lie on black lines, but no two vertices lie on the same black

誼 se
line.
23GTWO5

陳 U
[3♣] Required Problem 4. Let p ≥ 3 be prime. Find the smallest possible degree of a
polynomial f with integer coefficients for which f (0), f (1), . . . , f (p − 1) leave exactly

《 al
three distinct remainders upon division by p.
18BAMO5

n
[3♣] Problem 5 (BAMO 2018/5). A triangle is Pythagorean if it is similar to a right

e rn
triangle with integer side lengths. Find all integers n ≥ 3 for which it’s possible to dissect

h
a regular n-gon into Pythagorean triangles.

C e
14HMIC4

t
[9♣] Required Problem 6 (HMIC 2014/4). Suppose a1 , . . . , an are integers and ζ1 ,

n n
. . . , ζn are roots of unity. Let α = a1 ζ1 + · · · + an ζn . Prove that if |α| = 1 then α is a

a , I
root of unity.

v
09USATST8
[3♣] Problem 7 (USA TST 2009/8). Fix a prime number p > 5. Let a, b, c be integers

E I S
no two of which have their difference divisible by p. Let i, j, k be nonnegative integers
such that i + j + k is divisible by p − 1. Suppose that for all integers x, the quantity

19HKGTST14
y
B O T (x − a)(x − b)(x − c)[(x − a)i (x − b)j (x − c)k − 1]
is divisible by p. Prove that each of i, j, k must be divisible by p − 1.
[3♣] Problem 8√(Hong Kong TST 2019/1/4). Find the total number of primes p < 100
such that b(2 + 5)p c − 2p+1 is divisible by p.
MOPLAMPS
[3♣] Problem 9 (Black MOP 2010). There are n > 2 lamps arranged (evenly spaced) in
a circle. Initially, one of them is turned on, and the rest are off. It is permitted to choose
any regular polygon whose vertices are lamps and toggle all of their states simultaneously.
For which positive integers n is it possible to turn all the lamps off after a finite number
of such operations?
10TTSSO5
[5♣] Required Problem 10 (Tournament of Towns, Spring 2010). A needle lies on
the plane. One can rotate it 45◦ about either endpoint. Is it possible that after several
rotations the needle returns to initial position with the endpoints interchanged?
22MP4G17
[3♣] Problem 11 (MP4G 2022/17). Let O be the set of odd numbers between 0 and
100. Let T be the set of subsets of O of size 25. For any finite subset of integers S, let
P (S) be the product of the elements of S. Compute the remainder when S∈T P (S) is
P
divided by 17.

11
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

23JMO1
[2♣] Problem 12 (JMO 2023/1). Find all triples of positive integers (x, y, z) satisfying
2(x + y + z + 2xyz)2 = (2xy + 2yz + 2zx + 1)2 + 2023.
14USATST6
[5♣] Problem 13 (USA TST 2014/6). For a prime p, a subset S of residues modulo p
is called a sum-free multiplicative subgroup of Fp if
• there is a nonzero residue α modulo p such that S = 1, α1 , α2 , . . . (all considered


mod p), and

• there are no a, b, c ∈ S (not necessarily distinct) such that a + b ≡ c (mod p).
Prove that for every integer N , there is a prime p and a sum-free multiplicative subgroup
S of Fp such that |S| ≥ N .
20USMCA6

》
[5♣] Required Problem 14 (USMCA 2020/6). Let P (x) ∈ Z[x] be a nonconstant
polynomial such that whenever n is a perfect power, so is P (n). Prove that either P is

廷
the identity function, or P is a perfect power of a polynomial with integer coefficients.
17BRA6
[5♣] Required Problem 15 (Brazil 2017/6). Let a be a positive integer and p a prime

誼 se
divisor of a3 − 3a + 1, with p 6= 3. Prove that p is of the form 9k + 1 or 9k − 1, where k
is integer.

陳 U
20CAMO5
[3♣] Problem 16 (CAMO 2020/5). Let f (x) = x2 − 2. Prove that for all positive

《 al
integers n, the polynomial

n
e rn
P (x) = f (f (. . . f (x) . . . )) − x

h
| {z }
n times

can be factored into two polynomials with integer coefficients and equal degree.

C t e
10ELMO2
[3♣] Problem 17 (ELMO 2010/2). Let r and s be positive integers. Define a0 = 0,

n n
a1 = 1, and an = ran−1 + san−2 for n ≥ 2. Let fn = a1 a2 . . . an . Prove that fk ffn−k is an

I
n

a ,
integer for all integers n and k such that 0 < k < n.

v
18APMO5

S
[5♣] Problem 18 (APMO 2018/5). Find all polynomials P (x) with integer coefficients

E I
such that for all real numbers s and t, if P (s) and P (t) are both integers, then P (st) is

y T
also an integer.
19PTNMA5

B O
[3♣] Problem 19 (Putnam 2019 A5). Let p be an odd prime and define
p−1
X p−1
q(x) = k 2 xk
k=1

in Fp [x]. Find the greatest nonnegative integer n such that (x − 1)n divides q(x) in Fp [x].
21IZHO6
[3♣] Problem 20 (IZhO 2021/6, added by Nurtilek Duyshobaev). Let P (x) ∈ Q[x] be
a nonconstant irreducible polynomial of degree n. Prove that the number of polynomials
Q(x) of degree less than n with rational coefficients such that P (x) divides P (Q(x)) is
16USATST3
at most n.
[5♣] Problem 21 (USA TST 2016/3). Let p be a prime number. Let Fp denote the
integers modulo p, and let Fp [x] be the set of polynomials with coefficients in Fp . Define
Ψ : Fp [x] → Fp [x] by
n n
!
i
X X
i
Ψ ai x = ai xp .
i=0 i=0
Prove that for nonzero polynomials F, G ∈ Fp [x],
Ψ(gcd(F, G)) = gcd(Ψ(F ), Ψ(G)).

12
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

POLYFLT
[5♣] Problem 22 (Fermat’s last theorem for polynomials). Let f , g, h be relatively
prime nonconstant polynomials with complex coefficients. Let n ≥ 3 be an integer. Prove
that
f n + g n 6= hn .
17USATST6
[5♣] Problem 23 (USA TST 2017/6). Prove that there are infinitely many triples
(a, b, p) of integers, with p prime and 0 < a ≤ b < p, for which p5 divides (a + b)p − ap − bp .
24AMO3
[5♣] Problem 24 (USAMO 2024, Krit Boonsiriseth). Let (m, n) be positive integers
with n ≥ 3 and draw a regular n-gon. We wish to triangulate this n-gon into n − 2
triangles, each colored one of m colors, so that each color has an equal sum of areas. For
which (m, n) is such a triangulation and coloring possible?
23RSLA1

》
[5♣] Problem 25 (RMM SL 2023 A1, added by Solal Pivron-Djeddi). Determine all
polynomials P with real coefficients such that whenever x and y are real numbers for

廷
which P (x) and P (y) are both rational, so is P (x + y).

誼 se
[1♣] Mini Survey. Fill out feedback on the OTIS-WEB portal when submitting this
problem set. Any thoughts on problems (e.g. especially nice, instructive, easy, etc.) or

陳 U
overall comments on the unit are welcome.
In addition, if you have any suggestions for problems to add, or want to write hints for

《 al
one problem you really liked, please do so in the ARCH system!

n
e rn
The maximum number of [♣] for this unit is [98♣], including the mini-survey.

C h e
n I nt
va ,
E I S
y
B O T

13
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

§4 Solutions to the walkthroughs

§4.1 Solution 2.1, USAMO 1974/1
We have
a − b | P (a) − P (b) = b − c
and cyclically. If a − b = −(b − c) we get a = c, contradiction. So we must have
a − b = b − c = c − a which is impossible for distinct a, b, c.

§4.2 Solution 2.8, Niven’s theorem

The usual approach is to note that

z = cos(qπ) + i sin(qπ)

》
is an algebraic integer, since it is the root of xN − 1 for some N (i.e. is a root of unity).

廷
Similarly,
z = cos(qπ) − i sin(qπ)

誼 se
is an algebraic integer (actually a Galois conjugate).

陳 U
So the number
z + z = 2 cos(qπ)

《 al
is an algebraic integer. Since it is rational, it must be an integer, and hence contained in

n
the set {−2, −1, 0, 1, 2} as needed.

h e rn
Remark. The link https://arxiv.org/pdf/1006.2938.pdf gives an elementary proof

e
of this result using the double-angle formula. The idea is that for any angle θ with

C t
cos(θ) ∈
/ {−1, −1/2, 0, 1/2, 1}, the number an = cos(2n θ) has unbounded denominators

n n
by the double-angle formula. However, if θ = qπ, the sequence (an )n should be periodic,

I
contradiction.

va ,
E I S
§4.3 Solution 2.9, USA TST 2007/3

y T
Here is Holden Lee’s solution, which proves a more general result.

B OTheorem
If cos(kθ) and cos(`θ) are rational for relatively prime positive integers k and ` and
θ is not a rational multiple of π, then cos θ is rational.

Proof. Consider the two polynomials in Q[x]:

P (x) = x2k − 2 cos(kθ)xk + 1

Q(x) = x2q − 2 cos(`θ)x` + 1.

The 2k roots of P are exactly

 
2πi
exp ±iθ + n n = 0, 1, . . . , k − 1.
k
The 2` roots of Q are exactly
 
2πi
exp ±iθ + n n = 0, 1, . . . , ` − 1.
`

14
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

Since θ is not a rational multiple of π, the common roots are exactly exp(±iθ). So it
follows that

gcd(P, Q) = (x − exp(−iθ)) (x − exp(iθ))

= x2 − 2 cos θx + 1.

This polynomial has rational coefficients so cos θ ∈ Q.

Applying the lemma to the problem, we conclude θ must be a rational multiple of π.

Hence 2 cos(kθ) is a rational algebraic integer, so it must be one of {0, ±1, ±2}. The
same is true for 2 cos ((k + 1)θ). This forces θ = π/3.

§4.4 Solution 2.10, Kronecker

》
Let α1 = α, α2 , . . . , αn be the Galois conjugates. Across integers e, consider the

廷
polynomial
(X − α1e ) . . . (X − αne ).

誼 se
This is a polynomial with integer coefficients by the fundamental theorem
of symmetric
polynomials. But by triangle inequality each coefficient is bounded by n/2
n
, so the total

陳 U
number of possible polynomials is at most n/2 n n+1
. Since we are taking an infinitely

《 al
many values of e, there is some particular polynomial P (X) for which

n
e rn
P (X) = (X − α1e ) . . . (X − αne )

h
holds for infinitely many e.

C e
Fix a complex root z of P (X). Then for any such e, there exists an index i = i(e)

t
(depending on e) for which z = αie . By pigeonhole principle again, eventually there exists

n I n
e and e0 with i(e) = i(e0 ); meaning αie = αie . So αi is a root of unity. Thus so is α1 = α.
0

va ,
S
§4.5 Solution 2.11, HMMT 2017 A9

E I
Recall that the Fibonacci sequence is given by

y T
B O
αn − β n
Fn =
α−β
√ √
where α = 1+2 5 and β = 1−2 5 are the two roots of P (X) := X 2 − X − 1.
Let p = 127 and work modulo p. As
     
5 p 2
= = = −1
p 5 5

we see 5 is not a quadratic residue mod 127. Thus the polynomial P (X), viewed as
a polynomial in Fp [X], is irreducible (intuitively, α and β are not elements of Fp ).
Accordingly we will work in the finite field Fp2 , which is the Fp -splitting field of P (X).
In other words we interpret α and β as elements of Fp2 which do not lie
in Fp .
Let σ : Fp2 → Fp2 by t 7→ t be the nontrivial element of Gal Fp2 /Fp ; in other words,
p

σ is the non-identity automorphism of Fp2 . Since the fixed points of σ are the elements
of Fp , this means σ does not fix either root of P ; thus we must have

αp = σ(α) = β
β p = σ(β) = α.

15
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

Now, compute
αp − β p β−α
Fp = = = −1.
α−β α−β
αp+1 − β p+1 αβ − βα
Fp+1 = = = 0.
α−β α−β
α2p+1 − β 2p+1 β 2 α − α2 β
F2p+1 = = = −αβ = 1.
α−β α−β
α2p+2 − β 2p+2 β 2 α2 − α2 β 2
F2p+2 = = = 0.
α−β α−β
Consequently, the period must divide 2p + 2 but not p + 1.
We now use for the first time the exact numerical value p = 127 to see the period

》
divides 2p + 2 = 256 = 28 , but not p + 1 = 128 = 27 . (Previously we only used the fact
that (5/p) = −1.) Thus the period must be exactly 256.

廷
Remark. This is a first taste of finite fields and Galois theory. The reason for the notation

誼 se
Fp2 is that in fact, all the fields with p2 elements are isomorphic to the set of roots of
X p − X in an algebraic closure of Fp .
2

陳 U
The map σ is an example of a field automorphism: it’s a map Fp2 → Fp2 fixing a base
field Fp . In general, given a base field F and an extension K of it, under certain nice-ness

《 al
conditions, let us define define the Galois group

n
Gal(K/F ) = {automorphisms K → K fixing F } .

h e rn
As we saw, such maps will permute the roots of any polynomial P ∈ F [x], since P (σ(x)) =

e
σ(P (x)). In other words, automorphisms in Gal(K/F ) send elements of K to their Galois

C t
conjugates.

n n
In the case where F = Fp and K = Fpn it turns out Gal(K/F ) = {1, σ, . . . , σ n−1 }, where

I
σ(t) = tp is the Frobenius endomorphism you met, which is why the Frobenius endomorphism

a ,
is so heralded: it generates every such nice automorphism map.

v
You can read about all this and much more in Napkin.

E I S
y T
§4.6 Solution 2.12

B O
Assume for contradiction that Φn (x) = f (x)g(x), where f is nonconstant and also
irreducible (i.e. f is the minimal polynomial of some root of Φn ). The main claim we
prove is that:

Claim — If ζ is a root of f and p is a prime not dividing n, then ζ p is also a root

of f .

Proof. Suppose not. Then ζ p is a root of g instead. Since f is a minimal polynomial of

ζ, and ζ p is a root of g, that means f (x) divides g(xp ). Thus we may write

g(xp ) = f (x)h(x) =⇒ g(xp ) ≡ g(x)p ≡ f (x)h(x) (mod p)

the last part using the Frobenius endomorphism.

If we then consider an irreducible factor k(x) ∈ Fp [x] of f (x) (mod p), then it follows
that k divides g(x)p as well, and therefore that it divides g(x) itself. But f (x)g(x) = Φn (x)
divides xn − 1 even modulo p, so we find that xn − 1 (mod p) has a repeated factor
(namely k). But its derivative nxn−1 (mod p) is vanishes only at x = 0, which is a
contradiction.

16
 Evan Chen《陳誼廷》 (OTIS, updated 2024-12-03) Integer Polynomials

By then taking primes of each residue class modulo n relatively prime to n, we deduce
that f vanishes at every primitive nth root of unity, which is what we wanted to prove.

》
廷
誼 se
陳 U
n《 al
h e rn
C nt e
n
a , I
Ev I S
y
B O T

17