Equiangular polygons

An algebraic approach
Titu Andreescu Bogdan Enescu
University of Texas at Dallas "B.P. Hasdeu" National College
Buzau, Romania

We call a convex polygon equiangular if its angles are congruent. Thus,

an equiangular triangle is an equilateral one, and an equiangular quadrilateral
is a rectangle (or a square). One interesting algebraic characterisation of the
equiangular polygons is stated in the following lemma.

Lemma 1 Let a1 ; a2 ; : : : ; an be positive real numbers and let " be a primitive

nth root of the unity, e.g. " = cos 2n + i sin 2n : If the sides of an equiangular
polygon have lengths a1 ; a2 ; : : : ; an (in counterclockwise order) then

a1 + a2 " + a3 "2 + : : : + an "n 1

= 0:

Proof. Consider the polygon’s sides as vectors, oriented counterclockwise

(see …gure for the case n = 8). Then the sum of the vectors equals zero. Now,
translate all vectors such that they have the same origin O. If we look at the
complex numbers corresponding to their extremities, choosing a1 on the positive
real axis, we see that these are a1 ; a2 "; a3 "2 ; : : : ; an "n 1 ; respectively. We deduce
that a1 + a2 " + a3 "2 + : : : + an "n 1 = 0:

The converse of the statement is not true. For instance, if a; b; c; d are the side
lengths of a quadrilateral and a+bi+ci2 +di3 = 0; then (a c)+i(b d) = 0; this
equality is ful…lled if the quadrilateral is a parallelogram (and not necessarily an
equiangular quadrilateral, that is, a rectangle). However, from the proof we see

Mathematical reflections 1, (2006) 1

that if a1 ; a2 ; : : : ; an are positive numbers and a1 +a2 "+a3 "2 +: : :+an "n 1 = 0;
then there exists an equiangular polygon with sides of lengths a1 ; a2 ; : : : ; an :
We examine now several contest problems involving equiangular polygons,
presenting in some cases two solutions: a geometric solution and an algebraic
solution. Some of them are discussed in [1] :

Problem 1. Prove that if an equiangular hexagon has side lengths a1 ; a2; : : : ; a6

(in this order) then
a1 a4 = a5 a2 = a3 a6 :
(Romanian Selection Test)
First solution. Expand every second side of the hexagon to obtain a tri-
angle ABC (see …gure below). Since the hexagon is equiangular, triangle ABC
is equilateral. Moreover, triangle ABC is the union of the hexagon and three
other smaller triangles, which are equilateral as well. We see that

AB = a1 + a2 + a6 ;
BC = a2 + a3 + a4 ;
CA = a4 + a5 + a6:

Since AB = BC = CA; we obtain the desired result.

Second solution. Let " = cos 26 + i sin 26 be a primitive sixth root of
unity. Then
a1 + a2 " + a3 "2 + a4 "3 + a5 "4 + a6 "5 = 0:
But "3 = cos + i sin = 1; so "4 = " and "5 = "2 : We deduce that

(a1 a4 ) + (a2 a5 )" + (a3 a6 )"2 = 0:

On the other hand, since "3 = 1 (and " 6= 1) we see that "2 "+1 = 0: Thus,
" is a common root of the equations (a1 a4 ) + (a2 a5 )z + (a3 a6 )z 2 = 0 and

Mathematical reflections 1, (2006) 2

z 2 z + 1 = 0; both with real coe¢ cients. Since " 2
= R; it follows that the two
equations share another common root "; so the coe¢ cients of the two equations
must be proportional; that is,

(a1 a4 ) = (a2 a5 ) = (a3 a6 );

as desired.

Problem 2. The side lengths of an equiangular octagon are rational num-

bers. Prove that the octagon has a symmetry center.
(Russian Olympiad)
First solution. Let ABCDEF GH be the octagon. The angles of an
equiangular octagon are equal to 135 ; thus, the lines containing the segments
AB; CD; EF and GH determine a rectangle.

Because the opposite sides of this rectangle are equal, we obtain

p p
2 2
AB + (AH + BC) = EF + (DE + F G);
2 2
or, equivalently,
p
2
AB EF = (DE + F G AH BC):
2
Because the side lengths of the octagon are rational numbers, the above equality
can hold if and only if

AB EF = DE + F G AH BC = 0:

In a similar way, we obtain

CD GH = F G + AH DE BC = 0:

Mathematical reflections 1, (2006) 3

From these equalities it follows that AB = EF; CD = GH; BC = F G and
DE = AH; so the opposite sides of the octagon are equal and parallel. It
follows that the quadrilaterals ABEF; BCF G; CDGH and DEHA are par-
allelograms, hence the midpoints of the segments AE; BF; CG; DH and DH
coincide. Obviously, this common point is a symmetry center of the octagon.
Second solution. Denote by a1 ; : : : ; a8 the octagon’s side lengths. Then

a1 + a2 " + : : : + a8 "7 = 0;

where " = cos 28 +i sin 28 : Observing that "4 = 1; the above equality becomes

a1 a5 + (a2 a6 ) " + (a3 a7 ) "2 + (a4 a8 ) "3 = 0:

Thus, " is a common root of the polynomials with rational coe¢ cients f (X) =
X 4 + 1 and g (X) = a1 a5 + (a2 a6 ) X + (a3 a7 ) X 2 + (a4 a8 ) X 3 : If g is
a non-constant polynomial, we derive that gcd (f; g) is a non-constant rational
polynomial, which is a contradiction since f is irreducible in Q [X] (to see this,
observe that its roots are nonreal, so the only way it can
p be written as approduct
of polynomials with real coe¢ cients is f = x2 + x 2 + 1 x2 x 2 + 1 ).
Therefore, g must be a constant and this implies that the opposite sides of the
octagon are equal and parallel. The solution ends like the previous one.

Problem 3. Let ABCDE be an equiangular pentagon whose side lengths

are rational numbers. Prove that the pentagon is regular.
(Balkan Mathematical Olympiad)
First solution. Let M and N be the intersection points of the line AE
with BC and CD; respectively.

Because ABCDE is equiangular, triangles AM B and DN E are isosceles,

with \M = \N = 36 ; therefore triangle CM N is also isosceles and CM =
CN: It follows that BC + BM = CD + DN: But AB = 2 cos 72 BM and
DE = 2 cos 72 DN; whence

AB DE = 2 cos 72 (BM DN ) = 2 cos 72 (CD BC) :

Since cos 72 is not a rational number, the above equalities imply both AB =
DE and CD = BC: The conclusion follows easily.
Second solution. We can prove more:

Mathematical reflections 1, (2006) 4

Lemma 2 The positive integer p is a prime number if and only if every equian-
gular polygon with p sides of rational lengths is regular.

Proof. Suppose p is a prime number and let the rational numbers a1 ; a2 ; : : : ; ap

be the side lengths of an equiangular polygon. We have seen that
2 2
" = cos + i sin
p p
is a root of the polynomial

P (X) = a1 + a2 X + : : : ap X p 1
:

On the other hand, " is also a root of the polynomial

Q(X) = 1 + X + X 2 + : : : + X p 1
:

Because the two polynomials share a common root, their greatest common di-
visor must be a non-constant polynomial with rational coe¢ cients.This implies
that Q can be factorized as a product of two non-constant polynomials with
rational coe¢ cients, which is impossible (to prove that, one applies the Eisen-
stein’s criterion to the polynomial Q(X + 1) ). Conversely, suppose p is not a
prime number and let p = mn; for some positive integers m; n > 1: It results
that n is a root of order m of the unity, hence 1 + n + 2n + : : : + (m 1)n = 0:
If we add this equality to 1 + + 2 + 3 + : : : + p 1 = 0; we deduce that
is the root of a polynomial of degree p 1; with some coe¢ cients equal to 1
and the others equal to 2. This means that there exists an equiangular polygon
with p sides, some of length 1 and the rest of length 2. Because such a polygon
is not regular, our claim is proved.

Problem 4. Let a1 ; a2 ; : : : ; an be the side lengths (in order) of an equian-

gular polygon. Prove that if a1 a2 : : : an ; then the polygon is regular.
(International Mathematical Olympiad)
First solution. We examine two cases: n odd and n even. If n is odd, say
n = 2k + 1, consider the angle bisector of \A2k+1 A1 A2 : It is not di¢ cult to
see that it is perpendicular to the side Ak+1 Ak+2 : Project all the sides of the
polygon on this line. If we denote by xi the length of the projection of the side
Ai Ai+1 (with the usual convention A2k+2 = A1 ); then

x1 + x2 + : : : + xk = xk+2 + xk+3 + : : : + x2k+1 = A1 B

(see …gure below). On the other hand, the angle between Ai Ai+1 and A1 B is
equal to the angle between A2k+2 i A2k+3 i and A1 B; thus xi x2k+2 i ; for
all 1 i k: It follows that the above equality can be reached only if the sides
of the polygon are equal.
If n is even, there is a similar argument, but instead of the angle bisector of
\A2k+1 A1 A2 ; one considers the perpendicular on the sides A1 A2 and Ak+1 Ak+2
(it is easy to see that these two sides are in this case parallel).

Mathematical reflections 1, (2006) 5

 Second solution. Let
2 2
" = cos + i sin
n n
be a primitive root of the unity. Then " is a root of the polynomial

P (X) = a1 + a2 X + : : : + an X n 1
:

The conclusion is obtained from the following:

Lemma 3 Let P (X) = a1 +a2 X +: : :+an X n 1 ; where a1 a2 : : : an > 0:

If is a root of P; then j j 1; and j j = 1 only if a1 = a2 = : : : = an :

Proof. We have
n 1
a1 + a2 + : : : + an = 0:
If we multiply this equality with 1; we obtain
2 n 1 n
a1 + (a1 a2 ) + (a2 a3 ) + : : : + (an 1 an ) + an = 0;

or, equivalently,
2 n 1 n
a1 = (a1 a2 ) + (a2 a3 ) + : : : + (an 1 an ) + an :

Now, suppose that j j = 1: It results

2 n 1 n
a1 = j (a1 a2 ) + (a2 a3 ) + : : : + (an 1 an ) + an j

j j(a1 a2 ) + j j2 (a2 a3 ) + : : : + j jn 1
(an 1 an ) + an j jn
(a1 a2 ) + (a2 a3 ) + : : : + (an 1 an ) + an = a1 :
Consequently, all inequalities must be equalities. Because 2
= R; this is possible
only if a1 = a2 = : : : = an :

Mathematical reflections 1, (2006) 6

 We conclude that the polygon is regular.

Problem 5. Let n be a positive integer which is not a power of a prime

number. Prove that there exists an equiangular polygon whose side lengths are
1; 2; : : : ; n in some order.
Solution. All we have to do is to prove that there exists a polynomial
f (x) = a1 + a2 x + : : : + an xn 1
;
such that a1 ; a2 ; : : : ; an is a permutation of the numbers 1; 2; : : : ; n and f (") = 0;
where " = cos 2n + i sin 2n : Equivalently, we show that there is a permutation
(0) ; (1) ; : : : ; (n 1) of the numbers 0; 1; : : : ; n 1 such that
(0) (1) (2) (n 1)
" + 2" + 3" + : : : + n" = 0:
First, observe that since n is not a power of a prime, n can be written as
n = pq; where p and q are relatively prime positive integers.
j k j k
Let k be an integer, 0 k n 1 and denote a = kq ; b = k q kq : We
de…ne (k) = aq + bp (mod n): Since p and q are relatively prime numbers it
follows that is well de…ned.
We then have
f (") = 1 + 2"p + 3"2p + : : : + q"(q 1)p

+ (q + 1) "q + (q + 2) "q+p + : : : + 2q"q+(q 1)p

+ (2q + 1) "2q + (2q + 2) "2q+p + : : : + 3q"2q+(q 1)p

+ :::
+ ((p 1) q + 1) "(p 1)q
+ : : : + pq"(p 1)q+(q 1)p
:
If we denote
= 1 + 2"p + 3"2p + : : : + q"(q 1)p
;
p 2p (q 1)p
=1+" +" + ::: + " ;
we have
f (") = + q"q + "q + 2q"2q + "2q + : : : + (p 1) q (p 1)q
+ "(p 1)q
:
Clearly, = 0; therefore

f (") = 1 + "q + "2q + : : : + "(p 1)q

= 0;

as desired.

References
[1] Mathematical Olympiad Treasures, Titu Andreescu, Bogdan Enescu,
Birkhäuser, Boston, Basel Berlin, 2004
[2] Poligoane echiangulare, Titu Andreescu, Bogdan Enescu, Gazeta Mate-
matic¼a, 107, 2002, 11, pp.422-427 (Romanian)

Mathematical reflections 1, (2006) 7