ALGEBRAIC INTEGERS AND APPLICATIONS TO

ELEMENTARY PROBLEMS

CEZAR LUPU

Abstract. In this handout we discuss about algebraic integers and their prop-
erties. We provide some applications to elementary problems given at different
olympiads around the world or others taken from the American Mathematical
Monthly or Gazeta Matematică (A-series).

1. Introduction and Preliminaries

The theory of algebraic integers is an important tool in algebraic number theory
for obtaining information about integers and rational numbers.
Definition 1.1. We say that a number α ∈ C is an algebraic integer if it is the root
of a monic polynomial in Z[X],

f (x) = xn + an−1 xn−1 + . . . + a1 x + a0 ,

where n ≥ 1, and a0 , a1 , . . . , an−1 ∈ Z.
In algebraic number theory, a algebraic integer is often called an integer, while
the ordinary integers (the elements of Z) are called rational integers.

Examples. 1. Any integer is an algebraic integer. Indeed, if x = k ∈ Z, then x

satisfies the equation x − k = 0.
2. Gaussian integers. Any number of the form x = a + b · i, with a, b ∈ Z is an
algebraic integer. Indeed, from x = a + b · i, with a, b ∈ Z we have x − a = b · i, and
thus (x − a)2 = −b2 , so x verifies the following equation

x2 − 2ax + a2 + b2 = 0.
3. Roots of unity are algebraic integers. This is quite obvious since x verifies the
equation xn = 1.

We denote by Z the set of all algebraic integers which is a subset of C. Now, we

state and prove two important facts about algebraic integers.

Theorem 1.2. Z ∩ Q = Z; in other words, any algebraic integer which is a rational

number is an integer.
Key words and phrases. algebraic integers, irreducible polynomials, rational integer.
This is a lecture delivered for Mathematical Olympiad Summer Program (MOP), June 7-July
1, 2015 which was held at Carnegie Mellon University, Pittsburgh, PA.
1
2 CEZAR LUPU

Proof. Let x ∈ Q. Assume that x verifies the following equation in Z[X],

xn + an−1 xn−1 + . . . + a1 x + a0 = 0,
where a0 , a1 , . . . , an−1 ∈ Z. Now, if we write x = pq with p, q ∈ Z, q 6= 0, it follows
that p divides a0 and q divides 1, so q = ±1, and therefore x = ±p and we are
done.
In what follows we prove that the set of algebraic integers is a subring of C.

Theorem 1.3. The sum, difference and the product of two algebraic integers is an
algebraic integer.
Proof. Let x and y be two algebraic integers which verify the equations:

(1) xn + an−1 xn−1 + . . . + a1 x + a0 = 0, ai ∈ Z

and

(2) y m + bm−1 y m−1 + . . . + b1 y + b0 = 0, bj ∈ Z

First of all, we treat the on-trivial case when x and y are nonzero. Let us denote
z one of the numbers x + y, x − y or xy. We will show that z is a root of a monic
polynomial in Z[X]. Consider all products of the form xr y s , with r ∈ {0, 1, . . . , n −
1}, s ∈ {0, 1, . . . , m − 1}. There are q = mn products of thus type and denote them
by z1 , z2 , . . . zq . Consider the case z = x − y (similalrly we treat the other cases:
z = x + y and z = xy). We show that for each i ∈ {1, 2, . . . , q} we have

(3) zzi = ci1 z1 + ci2 z2 + . . . + ciq zq , cij ∈ Z

We have the following cases:

Case 1. zi = xr y s , with r < n − 1 and s < m − 1. In this case it follows that

zzi = (x − y)xr y s = xr+1 y s − xr y s+1 ,

where xr+1 y s , xr y s+1 ∈ {z1 , z2 , . . . , zq }, so (3) is satisfied.
Case 2. zi = xn−1 y s , with s < m − 1; from equation (1), we can write

zzi = (x − y)xn−1 y s = xn y s − xn−1 y s+1 =

= −(an−1 xn−1 + . . . + a1 x + a0 )y s − xn−1 y s+1 =
= −an−1 xn−1 y s − an−2 xn−2 y s − . . . − a1 xy s − a0 y s − xn−1 y s+1
and since xn−1 y s , xn−2 y s , . . . , xy s , y s , xn−1 y s+1 ∈ {z1 , z2 , . . . , zq } we have that (3)
is true.
Case 3. zi = xr y m−1 , with r < m − 1. Here we proceed exactly like in case 2.
Case 4. zi = xn−1 y m−1 ; in this case we proceed along the lines in the case 2, but we
will use both equalities (1) and (2). Explicitly, we obtain for every i ∈ {1, 2, . . . , q},
MOP-2015 ALGEBRAIC INTEGERS AND APPLICATIONS 3

(c1,1 − z)z1 + + ··· +


 c1,2 z2 c1,q xq = 0
c2,1 z1 + (c2,2 − z)z2 + · · · + c2,q zq =0


.. .. . .. .
.. ..

 . . .
+ · · · + (cq,q − z)zq = 0

cq,1 z1 + cq,2 z2

This system is linear and homogeneous with unknowns z1 , z2 , . . . , zq , and admits

a nonzero solution, thus its determinant is null,

c11 − z c12 c13 . . . c1q

c21 xc22 − z c23 . . . c2q
= 0.
..................................
cq1 cq2 cq3 . . . cqq − z

By expanding the determinant, we obtain the characteristic polynomial of the

matrix (cij )1≤i,j≤q , i.e.

(c11 − z)(c22 − z) . . . (cqq − z) + . . . = 0.

This implies that z verifies an equation of the form,

z q + dq−1 z q−1 + . . . + d1 z + d0 = 0, dk ∈ Z,
and therefore z is an algebraic integer and we are done.

2. A motivation: Hilbert’s third problem

At the second international congress of mathematicians, held in Paris in 1900,
the German mathematician David Hilbert published a list of 23 problems which he
considered of capital importance for mathematics at that time. The third one was
the following:

Hilbert’s third problem. Is it true that any two polyhedra of the same volume
are scissors-congruent?

Before we start diving more into the above problem we should recall the following
Definition 2.1. Two polyhedra are scissors-congruent if one of them can be cut into
finitely many polyhedral pieces that can be reassembled to give the other.
Three months later, Hilbert’s student, Max Dehn produced a counterexample.
Dehn’s idea was to introduce a numeric invariant associated to every polyhedra
with the property that every two scissors-congruent polyhedra have the same Dehn
invariant.
Here are the details! More precisely, Dehn constructed the following:

• an abelian group G and

4 CEZAR LUPU

• a map (Dehn’s invariant), D : {Polyhedra} → G which satisfies:

(i) Two congruent polyhedra P1 and P2 have the same Dehn invariant, i.e.

D(P1 ) = D(P2 ).
(ii) Additive property:

D(P1 ∪ P2 ) = D(P1 ) + D(P2 ).

It follows that two scissors-congruent polyhedra have the same Dehn invariant.
Now, let us get even more into details! Let us define the group G. In other words,
Dehn’s invariant will compute the sum of all interior dihedral angles of a polyhedra
modulo π, but taking also into account the side lengths that appear. In other words,
the definition of G is
( k )
X
G= (ai , αi ), ai ∈ R, ai ∈ R/πZ / ∼
i=1
where the equivalence relation ∼ is generated by the following two properties:

(a, α) + (a, β) = (a, α + β),

and

(a, α) + (b, α) = (a + b, α).

Basically, we are interested in all finite sums of type (ai , αi ) subject to the equal-
ities above. For example, any pair (a, kπ),
k ∈ Z is equal to (a, 0) in G; the same
π 3π


thing happens if we have terms like a, 2 and a, 2 . They will be replaced by
(a, 0).
Now, let us explore a little bit more Dehn’s invariant. For any polyhedra P ,
denote ai its side lengths, and αi the radian measure of the interior dihedral angles
modulo π. Then, Dehn’s invariant is given by
n
X
D(P ) = (ai , αi ),
i=1
where n is the number of polyhedra’s sides. Clearly, two congruent polyhedra have
the same number of sides which are congruent, and the corresponding dihedral angles
are congruent respectively, and therefore they will have the same Dehn invariant.
Moreover, one can easily see that Dehn’s invariant has the following properties in
G:

(a, α + α0 ) = (a, α) + (a, α0 ),

and

(a + b, α) = (a, α) + (b, α),

MOP-2015 ALGEBRAIC INTEGERS AND APPLICATIONS 5

and

(a, α) + (a, β) = (a, α + β) = (a, π) = (a, 0)

in the case α + β = π.
Last but not least, let us mention that the group structure on G is natural in
the sense that if we add any two elements, we add its components (it is abelian),
the null element is (0, 0), the opposite of (a, α) is (−a, α). Not surprisingly, one can
extend this definition to finite sums. In other words, the definition of G is nothing
else than R ⊗Z R/πZ.
In what follows, we compute Dehn’s invariant of a cube, right parallelepiped, right
triangular prism or tetrahedron.
In
 the case of the cube, we have ai = a and αi = π2 for all i, so D(ai , αi ) =
π 
12 a, = 0. The same will be true for the rectangular parallelepiped and right
2
triangular prism. On the other hand, in the case of a regular tetrahedron, we have

D(ai , αi ) = 6(a, α)
1 1
with cos α = . Unfortunately, problems 2 and 3 imply that arccos is never
3 3
rational multiple of π. One important remark is the following: (a, α) = 0 in G if
and only if a = 0 or α is rational multiple of π.
Finally, one can easily deduce that

D(cube) = 0 6= D(regular tetrahedron),

and therefore Hilbert’s third problem has a negative answer.

3. Problems
In this section we give some applications of the algebraic integers. But first, let
us begin with the following rather simple

Theorem 3.1. Let q ∈ Q. Then the following statements are equivalent:

 
1
(i) cos qπ ∈ Q ⇔ cos qπ ∈ ±1; ± , 0 ;
 2 
1
(ii) sin qπ ∈ Q ⇔ sin qπ ∈ ±1; ± , 0 ;
2
(iii) tan qπ ∈ Q ⇔ tan qπ ∈ {±1, 0};
(iv) cot qπ ∈ Q ⇔ cot qπ ∈ {±1, 0};

Proof. We will prove only the implications from left to right since the other way
around they are trivial.
m
(i) Let q = , with m, n ∈ Z, (m, n) = 1 and n 6= 0. The complex number
n
m m
z = cos qπ + i sin qπ = cos π + i sin π
n n
6 CEZAR LUPU

satisfies z 2n = 1, and thus z is the 2n-root of unity and therefore z is an algebraic

integer. Moreover, the complex conjugate z is also an algebraic integer. It follows
that

z + z = 2 cos qπ
is an algebraic integer, and since it is rational, we obtain that 2 cos qπ ∈ Z. On
the other hand, −2 ≤ 2 cos qπ ≤ 2, so

2 cos qπ ∈ {−2, −1, 0, 1, 2}.

(ii) Since z, z and i are algebraic integers, it follows that i(z − z) = −2 sin qπ is
an algebraic integer and since it is rational, it follows that −2 sin qπ is an integer
and the argument is the same as in (i).
(iii) Here it is just a little bit tricky! Assume that tan qπ ∈ Q. It follows that

1 − tan2 qπ
cos 2qπ = .
1 + tan2 qπ
But, from (i) and solving the above equation for tan qπ, we get our conclusion.
(iv) The same argument as in (iii).

In the same fashion, we are wondering what would happen if we add up two
or three cosines of multiples of π. We leave the reader as exercises the following
theorems:
Theorem 3.2. Let r and q be two rational numbers such that

cos qπ + cos rπ ∈ Q.
Show that
 
3 1
cos qπ + cos rπ ∈ ±2, ± , ± , 0, ±1 .
2 2
What happens for other trigonometric functions?

Theorem 3.3. Let s, m and n be integers, s > 0 and n is even. If

  m  s
cos π ∈ Q,
n
then
  m  s  1

cos π ∈ ±1, ± s .
n 2

Now, let us a present a beautiful consequence of Theorem 3.1, namely

MOP-2015 ALGEBRAIC INTEGERS AND APPLICATIONS 7

Problem 0. Prove that the only regular n-gons all of whose vertices are lattice
points is the square.

Solution. Indeed, consider A1 , A2 , A3 be three consecutive vertices of the polygon

and observe that

1 + cos 4π 2π A1 A22 + A2 A23 − A3 A21

n
= cos2 = ∈ Q.
2 n 4A1 A22 · A2 A23
By Theorem 3.1, it follows that the only possibility is n = 4 and we are done.

Problem 1. Let x be a complex number such that xn and xn+1 are both integers.
Show that x is an integer.

Problem 2. Show that

 
1 1
arccos √ ∈
/ Q,
π n
when n ∈ N, n odd, n ≥ 3.

related to Hilbert’s third problem

Problem 3. Prove that if α is a real number such that cos(απ) = 13 , then α is

irrational.

Putnam, 1974

Problem 4. We say that a triangle is rational provided it has rational side

lengths and its angles are rational multiples of π. Show that any rational triangle
is equilateral.

Problem 5. If k, m and n are integers, then one solution of the equation

π m
= k arctan
4 n
is k = m = n = 1. Find all others.

American Mathematical Monthly, 1972

Problem 6. Consider the sequence (an )n≥1 defined by a1 = 2, and

2 + an
an+1 = , n ≥ 1.
1 − 2an
Show that all the terms of an are nonzero.

Romanian Olympiad, 1996

1 + λi
Problem 7. Find all complex numbers x = , with λ ∈ Q which are the
1 − λi
roots of unity.
8 CEZAR LUPU

Gazeta Matematică, 1987

Problem 8. Let x ∈ C, x 6= 1. Define the sequence of complex numbers (an )n≥1

xn − 1
by an = . Show that if three consecutive terms are integers, then all terms of
x−1
the sequence an are integers.

Problem 9. Prove that the number

√ √ √
N= 10012 + 1 + 10022 + 1 + . . . + 20002 + 1
is irrational.

Chinese Selection Test, 2005

Problem 10. Let x, y ∈ C, x 6= y. Define the following sequence of complex

numbers,
xn − y n
an = .
x−y
Show that if four consecutive terms are integers, then all terms of the sequence
an are integers.

American Mathematical Monthly, 1993

√ √ √
Problem 11. Let a1 , a2 , . . . , ak be positive reals such that n a1 + n a2 + . . . n ak
is rational for all n ≥ 2. Prove that a1 = a2 = . . . = ak = 1.

Problem 12. Find all triplets of positive rational numbers (m, n, p) such that
1 1 1
the numbers m + , n + and p + are integers.
np pm mn
Balkan Mathematical Olympiad, 2006

Problem 13. Show that the acute angles of a right triangle with rational side
lengths are never rational multiples of π.

Problem 14. Show that if the vertices of a triangle belong to Z × Z then the
Brocard angle cannot be a rational multiple of π.

American Mathematical Monthly, 2008

Problem 15. Let ABC be a triangle with vertices of integer coordinates. Show
that if P is an interior point with integer coordinates, then at least one of the angles
M AB, M BC and M CA cannot be rational multiple of π.

American Mathematical Monthly, 2015

Problem 16. Let A, B, C be lattice points such that the angles of triangle ABC
are rational multiples of π. Prove that the triangle ABC is right and isosceles.
MOP-2015 ALGEBRAIC INTEGERS AND APPLICATIONS 9

Problem 17. Let P be a polygon with 2001 sides, inscribed in the unit circle.
Prove that the lenght of all sides and diagonals of P are irrational.
American Mathematical Monthly, 2002
Problem 18. Find all non-isosceles triangles with at least two sides rational
numbers and having all angles rational numbers, when measured in degrees.
American Mathematical Monthly (E 2668), 1977
Problem 19. Determine all rationals r for which 1, cos 2πr, and sin 2πr are
linearly dependent over the rationals.
American Mathematical Monthly, 1974
Problem 20. Let X be the space obtained from the unit square [0, 1]2 ⊂ R2 by
deleting all points with both coordinates rational except (0, 0) and (1, 1). The Baire
category theorem implies the existence of a smooth path in X from (0, 0) to (1, 1).
Find an explicit example of a smooth path in X from (0, 0) to (1, 1).
Problem 21. Let α be a rational number with 0 < α < 1 and

cos(3πα) + 2 cos(2πα) = 0.
2
Show that α = .
3
IMO Shortlist, 1991
√ √
Problem 22. Prove that none of the
numbers n + 1 − n for positive integers
n can be written in the form 2 cos 2kπ
m
for some integers k, m.
Chinese Olympiad
Problem 23. Let θ ∈ 0, π2 be an angle such that cos θ is irrational. Suppose


that cos kθ and cos(k + 1)θ are rational for some positive integer k. Prove that
θ = π6 .
USA Selection Test, 2007
Problem 24. Let p be a prime number. Show that
p q
X p √
p
k+ k
k=1
cannot be rational.
Gazeta Matematică (A-series), 2011
Problem 25. Consider the sequence (xn )n≥0 defined by x0 = 4, x1 = x2 = 0,
x3 = 3 and xn+4 = xn+1 +xn . Prove that for any prime p the number xp is a multiple
of p.
10 CEZAR LUPU

American Mathematical Monthly, 1998

Problem 26. Let a, b, c be integers. Define the sequence (xn )n≥0 by x0 = 4,
x1 = 0, x2 = 2c, x3 = 3b and xn+3 = axn−1 + bxn + cxn+1 . Prove that for any prime
p and any positive integer m, the number xpm is divisible by p.
Romanian Selection Test, 2004
Problem 27. Let n ≥ 1 be a positive integer and n 6= 4. Let p and q be positive
integers less than n and coprime with n. Let
cos(2πp/n)
a= .
cos(2πq/n)
Show that if ak is rational for a positive integer k, then ak is either 1 or −1.
American Mathematical Monthly, 2010
Problem 28. 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 their strates simultaneously.
Show that it is impossible to turn all the lamps off.

References
[1] M. Aigner, G. M. Ziegler, Proofs from THE BOOK, Fourth edition, Springer Verlag, 2010.
[2] D. Benko, A new approach to Hilbert’s third problem, Amer. Math. Mon., 114 (2007),
665–676.
[3] M. Bălună, M. Becheanu, Gh. Eckstein and el., Zece lecţii alese ı̂n matematică, Romanian
Mathematical Society, 1998.
[4] J. S. Calcutt, Gaussian integers and arctangent identities for π, Amer. Math. Mon., 116
(2009), 515–530.
[5] T. Andreescu, G. Dospinescu, Problems from the BOOK, XYZ Press, 2008.
[6] J. M. H. Olmsted, Rational values of trigonometric functions, Amer. Math. Monthly 52
(1945), no. 9, 507508.
[7] L. Ornea, Echidecompozabilitate şi problema a treia a lui Hilbert, Gazeta Matematică (B-
series), 113 (2008), 327–336.

University of Pittsburgh, Department of Mathematics, 301 Thackeray Hall,

Pittsburgh, PA 15260, USA
E–mail address: [email protected], [email protected]