INTERNATIONAL JOURNAL OF GEOMETRY

Vol. 1 (2012), No. 2, 46 - 60

A PROBLEM SOLVING APPROACH

TO SOME APPLICATIONS OF THE ROOTS
OF UNITY TO REGULAR POLYGONS

MARCELINA MOCANU

Abstract. The aim of this note is to emphasize the unity between geom-
etry, algebra and analysis, by means of two applications of complex num-
bers to the study of regular polygons. We use an heuristic problem-solving
approach to …nd a complete description of lattice regular polygons and a
formula for the minimum of the sum of squared reciprocals of the distances
from a point on a circle to the vertices of a regular polygon inscribed in
that circle. We also give an example of the use of mathematical software in
arriving at a conjecture for calculating the above-mentioned sum.

1. Introduction

The connection between regular polygons and the roots of unity plays a
fundamental role in algebra and geometry. The images of the n-th roots
of unity in the complex plane are the vertices of a regular n-gon inscribed
in the unit circle. Moreover, a positively oriented n gon in the complex
plane is regular if and only if the a¢ xes of its vertices can be obtained as
the images of the n-th roots of unity under a polynomial function of …rst
degree.
The multiplicative group of nonzero complex numbers contains only one
subgroup with n elements, which is the group Un of the n-th roots of unity.
There are some corresponding isomorphic groups of geometric transforma-
tions, the group of rotations about the origin in plane, respectively its sub-
group preserving each regular n gon centered at the origin.
This connections between regular polygons and roots of unity explain, for
example, why the problem of constructibility of regular polygons by ruler
and compass, as well as the derivation of several trigonometric identities
involving integer multiples of 2n are in essence algebraic problems.

————————————–
Keywords and phrases: regular polygon, lattice polygon, roots of unity
(2010)Mathematics Subject Classi…cation: 97G40, 51M04.
Received: 21.06.2012. In revised form: 3.07.2012. Accepted: 10.07.2012.
 A problem solving approach to some applications of the roots of unity to regular polygons 47

On the other hand, in the process of solving algebraic problems involving

the roots of unity, the geometric representation is meaningful and enlighten-
ing. For example, the property: "The sum of the n distinct n roots of unity
is zero", that usually is proved using the Viète relations for the polynomial
z n 1 or the sum of a geometric progression, shows that the sum of the
vectors from the center of a regular polygon to its vertices is the null vector.
Note that, for every prime p; the sum of k distinct vectors joining the center
of a regular p gon to its vertices cannot be null if 1 k p 1, a fact that
is a consequence of the irreducibility in Z [X] of any cyclotomic polynomial
in X.
In this note we outline a problem solving approach to two applications
of complex numbers to the study of regular polygons. The …rst applica-
tion leads to a complete description of lattice regular polygons. The second
application provides a formula for the minimum of the sum of squared recip-
rocals of the distances from a point on a circle to the vertices of a regular
polygon, which is inscribed in that circle. We prefer to use an heuristic
approach: rather than to write down a short and elegant solution, we try to
explore several paths to a solution. We also give an example of the use of
mathematical software in arriving at a conjecture for calculating the above-
mentioned sum. Our aim is not only to solve two speci…c problems, but also
to give perspectives on other related problems.

2. The existence of lattice regular polygons

Given a Cartesian coordinates system in a plane, it is said that a point

of the plane is latticial if it has integer coordinates. A plane polygon is
called a lattice polygon, with respect to a Cartesian coordinates system, if
its vertices are latticial points.
The following problem, although a classical one, was on the shortlist for
the International Mathematics Olympiad 1985.
Problem 1. For which integers n 3 does there exist a regular n gon
in the plane such that all its vertices have integer coordinates in a Cartesian
coordinates system?
We give three solutions to this problem, some of them opening the way
to other related more general issues.

Necessary condition. Assuming that there exists a lattice regular

n gon, we try to …nd a subset of N, as small as possible, where n belongs.

First solution.
We provide a formula for the a¢ xes of the vertices of a regular n gon, as-
suming that the a¢ xes of two consecutive vertices are known. Let A1 A2 ::An
be a regular n gon in the plane and let zk be the a¢ x of Ak , k = 1; n. We
may assume that A1 A2 ::An is positively oriented, i.e. A1 ! A2 ! ::: !
An ! A1 gives the trigonometric orientation. At a …rst step, we determine
the center C of the given regular n gon, by calculating its a¢ x z0 in terms
of z1 and z2 . At a second step, we take into account that Ak is the image
of A1 under a rotation by the angle 2kn around C.
48 Marcelina Mocanu

First step. Denote by l the length of the polygon’s sides and let
2 2
! := cos + i sin :
n n
1
Then CAk = l 2 sin n . Performing a rotation by the angle 2 n around
1
A1 , then a homotethy of center A1 and ratio 2 sin n , we get
1
z0 z1 = 2 sin cos + i sin (z2 z1 ) :
n 2 n 2 n
1
Denoting a = 2 1 + i cot n , this yields
z0 = (1 a) z1 + az2 :
Second step. We have zk z0 = ! k 1 (z1 z0 ) for k = 1; n, hence zk =
1 ! k 1 z0 + ! k 1 z1 . We obtain the following
Lemma 2.1. Let A1 A2 ::An be a positively oriented regular n gon in the
plane and let zk be the a¢ x of Ak , k = 1; n. Then
zk = 1 a + a! k 1
z1 + a a! k 1
z2 ,

for k = 1; 2; ::; n, where ! := cos 2n + i sin 2n and a = 1

2 1 + i cot n .
We will not use the full power of our assumption that Re zk , Im zk 2 Z for
all k 2 f1; 2; :::; ng. Instead, we will analyze the coordinates of the center
C of the regular polygon. The following well-known result is proved for the
sake of completness.
Lemma 2.2. The a¢ x of the center of a regular polygon is the arithmetic
mean of the a¢ xes of polygon’s vertices.
Proof. With the above notations, zk z0 = ! k 1 (z
1 z0 ) for k = 1; n.
Then we have
Xn Xn
1 !n
(zk z0 ) = (z1 z0 ) ! k 1 = (z1 z0 ) = 0;
1 !
k=1 k=1
n
X n
X
hence 0 = (zk z0 ) = zk nz0 , q.e.d.
k=1 k=1

Assuming that A1 A2 ::An is a lattice polygon, it follows that its center

has rational coordinates. Denote xk := Re zk , yk := Im zk 2 Z for k 2
f0; 1; 2; :::; ng.
We have x0 = 12 (x1 + x2 )+ 12 (y1 y2 ) cot n , y0 = 21 (y1 + y2 ) 12 (x1 x2 ) cot n
and xk , yk 2 Q for k 2 f0; 1; 2g. Taking into account that x1 6= x2 or y1 6= y2 ,
we deduce that cot n 2 Q, which is equivalent to tan n 2 Q.

We arrive to the following auxiliary problem.

Problem 2. Determine the values n 2 N, n 3 for which tan n 2 Q.
Solution. Let us note that tan 4 = 1 2 Q, while tan 3 , tan 6 2 R n Q.
It su¢ ces to assume n 5.
We will make use of the fact that t = n implies tan nt = 0, where n 2 N,
n 3.
 A problem solving approach to some applications of the roots of unity to regular polygons 49

Using de Moivre’s formula and Newton’s binomial Theorem we get

[ n2 ]
X
(1) cos nt = ( 1)k Cn2k cosn 2k
t sin2k t
k=0
and
n 1
[X2 ]

(2) sin nt = ( 1)k Cn2k+1 cosn 2k 1

t sin2k+1 t:
k=0
n o
(2k+1)
For t 2 Rn 2n : k 2 Z , we have sin t = cos t tan t, hence

[ n2 ]
X
cos nt = cosn t ( 1)k Cn2k tan2k t
k=0
and
n 1
[X2 ]

sin nt = cos t n
( 1)k Cn2k+1 tan2k+1 t:
k=0
Both functions tan t and tan nt are well-de…ned for t 2 Dn , where
(2k + 1) (2k + 1)
Dn := Rn :k2Z [ :k2Z :
2 2n

Using the above formulas we get the following

Lemma 2.3. For each n 2 N there exist two polynomials Pn ; Qn with
Pn (tan t)
integer coe¢ cients such that tan nt = Qn (tan t)
for every t 2 Dn . Moreover,
n 1
[X2 ] X[ n2 ]
k 2k+1 2k+1
Pn (X) = ( 1) Cn X and Qn (X) = ( 1)k Cn2k X 2k .
k=0 k=0

Let n 2 N, n 5. Since tan n n = 0, it follows by Lemma 2.3 that

Pn tan n = 0,
Now assume that tan n 2 Q. Write tan n = pq , where p; q 2 N are
relatively prime. Since tan n < tan 4 = 1 for all n 5, we have p q 1.
The constant term of Pn is Cn1 and the leading coe¢ cient of Pn is ( 1)m Cn2m+1 ,
where we denoted m = n 2 1 . We have pj Cn1 and qj ( 1)m Cn2m+1 .
Case 1. If n 5 is odd, then n = 2m + 1 and the leading coe¢ cient
of Pn is ( 1)m . Since q divides the leading coe¢ cient of Pn , it follows that
q = 1, a contradiction with 1 p q 1.
Case 2. If n 8 is even, then we can write n = 2r (2s + 1), where r 2 N
and s 2 N.
Subcase 2.1. Assume s = 0. Then n = 2r , where r 3. According to
Lemma 2.3 or simply due to formula tan 2 = 1 2 tan 2 , tan 2r 2 Q implies
p p tan p p
1
tan 2 r 3
2r = tan 8 2 Q. But sin 8 = 2 2 2 and cos 8 = 12 2 + 2,
hence s p
2 2 p
tan = p = 2 1 2 R n Q:
8 2+ 2
50 Marcelina Mocanu

We obtain a contradiction.
Subcase 2.2. Assume s 1. According to Lemma 2.3, tan 2r (2s+1) 2 Q
implies tan 2r 2r (2s+1) = tan 2s+1 2 Q. For s = 1 it follows that tan 3 2 Q,
which is false. For s 2 the conclusion tan 2s+1 2 Q is false, according to
Case 1.

In all cases, the assumption that tan n 2 Q, where n 2 N, n 5 leads to

a contradiction.
This proves the following
Lemma 2.4. Let n 2 N , n 3. Then tan n 2 Q if and only if n = 4.

We conclude that a necessary condition for the existence of a lattice reg-

ular n gon is that n = 4.

Remark 1. The rational values of the circular trigonometric functions

at the rational multiples of have been well-studied by Underwood (1921),
Lehmer (1933), Olmsted (1945), Niven (1956), Carlitz and Thomas (1962),
see [2]. A thorough study of the rationality of the values of the tangent
function at the rational multiples of has been done in [2], using more
advanced algebraic concepts and methods than in this note.

Second solution. The idea is to use the area of the desired polygonal
region.
By Pick’s Theorem, the area of a lattice polygonal region is given by
B
S=I+ 1;
2
where I and B denote the the number of latticial points belonging to the
interior of the polygon, respectively to the polygon.
An alternative way to express the area S of the region bounded by the
polygon A1 A2 :::An is to use the general formula
X n
1
S = Im zk zk+1 ,
2
k=1
where zk denotes the a¢ x of Ak , k = 1; 2; :::; n and An+1 = A1 .
Note that each of the above formulas shows that the area of a lattice
polygonal region is a rational number, moreover its double is a positive
integer.
But the area of the region bounded by a regular n gon with side length
L is Sn = 41 nL2 cot n .
For the region bounded by a lattice regular n gon the area Sn is rational,
but the squared side length L2 is a positive integer, hence tan n 2 Q. We go
back to Lemma 2.4 and conclude that it is necessary to have n = 4.

Third solution. Here we explain in detail the "o¢ cial solution" from
IMO 1985, see [4]. This solution involves more ingenuity than the previous
ones, since some auxiliary construction and the extremal principle are used.
In addition, this solution has a more geometric character than the above
solutions.
 A problem solving approach to some applications of the roots of unity to regular polygons 51

First suppose that for some n there exists at least one lattice regular
n gon A1 A2 ::An .
Step 1. There exist a family of lattice regular n gons, some of these being
obtained as images of A1 A2 ::An under homotethies of integer ratios, with
center at the origin O.
Since the square of the side length is a positive integer, there exists a
lattice regular n gon S1 S2 ::Sn with a smallest side length a. Let Bj be
! !
the point such that OBj = Aj 1 Aj , for j = 1; 2; :::; n, where we denote
A0 = An . Then each Bj is a latticial point. The n gon B1 B2 :::Bn is
! !
regular, since the angle between OBj and OBj+1 is supplementary to the
angle ]Aj 1 Aj Aj+1 (where j = 1; :::; n, An+1 = A1 ) and the measure of
]Aj 1 Aj Aj+1 is n1 (n 2). Then the side length of B1 B2 :::Bn is 2a sin n .
By the minimality of the side length of S1 S2 ::Sn , we have

2a sin a,
n
hence n 6.
We conclude that there is no lattice regular n gon with n 7.
Step 2. Assume that n 6.
The center of every lattice regular n gon has rational coordinates. Using
a homotethy we may …nd another lattice regular n gon whose center C
has integer coordinates.
Now rotate the vertices of around C with angles 2 , and 32 and take
the union between the set of the vertices of and its images under this
rotations. We obtain the vertices of a regular N gon, where N = 12 for
n = 3 and n = 6, while N = 20 for n = 5. If n = 6 it su¢ ces to rotate the
vertices of around C with 2 , since the set of the vertices of is preserved
under a rotation with . See …gures 1, 2 and 3 corresponding to n = 3,
n = 6 and n = 5, respectively. Obviously, for n = 4 the set of the vertices
of is invariant to the above rotations.

Figure 1 Figure 2
52 Marcelina Mocanu

Figure 3
Step 3. Now we show that rotating a latticial point with 2 around another
latticial point we get a third latticial point, a property that is easy to notice
by graphical representation.
Lemma 2.5. The image of a latticial point under a rotation around an-
other latticial point, with angle 2 , is also a latticial point.
Proof. Let M (x; y) and M0 (x0 ; y0 ) be latticial points and let M (x ; y )
be the point obtained as we rotate M around M0 with 2 . Since

x x0 + i (y y0 ) = cos + i sin [(x x0 ) + i (y y0 )]

2 2
and cos 2 + i sin 2 = i, we get by identifying the real and imaginary parts,
respectively:
x = x0 + y0 y and y = y0 x0 + x.
Now rotate M (x; y) around M0 (x0 ; y0 ) with 2 and denote the point
we obtain by M (x ; y ). Then M0 is the midpoint of the line segment
[M M ], hence
x = 2x0 x = x0 y0 + y
and
y = 2y0 y = x0 + y0 x:
By the above formulas, x0 ; y0 ; x; y 2 Z implies x ; y ; x ,y 2 Z.
Remark 2. We can …nd (x ; y ) without using complex numbers, but
using graphical representations. In this case, we need to discuss the signs
of x x0 and y y0 . If x = x0 , then (x ; y ) = (x0 + y0 y; y0 ) and
if y = y0 , then (x ; y ) = (x0 ; y0 x0 + x). Assume that x 6= x0 and
y 6= y0 . The parallels to Ox through M and M intersect the parallel to Oy
through M0 at N and N , respectively. For all four choices of signs (+; +),
(+; ), ( ; ) and ( ; +) for (x x0 , y y0 ), we can see that the right
triangles N M0 M and N M M0 are congruent and positively oriented,
hence x x0 = (y y0 ) and y y0 = x x0 , q.e.d.
 A problem solving approach to some applications of the roots of unity to regular polygons 53

Assume that our problem has a solution for n = 3 or n = 5 or n = 6.

It follows that the N gon obtained at Step 2, with N = 12, N = 20;
respectively N = 12 is a lattice regular polygon, a contradiction with the
conclusion of Step 1.

We conclude that a lattice regular n gon could exist only if n = 4.

Su¢ cient condition.

Now we check if the condition n = 4 is also su¢ cient for the existence of
a lattice regular n gon.
Assuming that n = 4, Lemma 2.5 shows that for each pair of latticial
points A and B we obtain a latticial point C by rotating A around B with
angle 2 , then another latticial point D by rotating B around C with
2 . Obviously, ABCD is a positively oriented lattice square .
Remark 3. A lattice square region whose sides are not parallel to the
coordinate axes can be decomposed, using parallels through its vertices to
the axes, into a smaller square region and four congruent right triangular
regions, as in a well-known proof of Pythagorean Theorem that makes use
of areas. See Figure 4.

Figure 4

We proved in three di¤erent ways the following known result

Theorem 2.6. There exists a lattice regular n gon if and only if n = 4.
Remark 4. Several authors have considered the problem of determin-
ing all the lattice regular polygons or generalizations of this problem to
higher dimensions. We mention here only I. J. Schoenberg (1937) [10], M.
S. Klamkin and H. E. Chrestenson (1963) [6], D. J. O’Loughlin (2002) [8].
Unfortunately, the above cited articles are not available online, except on
54 Marcelina Mocanu

JSTOR. For more references see [3]. This problem has attracted much in-
terest because it admits various elementary solutions, but is connected to
several non-elementary problems. Note that lattice polygons play an im-
portant role in discrete and computational geometry, a …eld which is at the
cutting-edge frontier of mathematics and computer science.

3. The sum of squared reciprocals of the distances from a

point on a circle to the vertices of a regular polygon

A well-known formula shows that the sum of the squared distances from
a point on a circle of radius R to the vertices of a regular n gon, which is
inscribed in that circle, is nR2 . We will study the sum of squared reciprocals
of the distances from a point on a circle to the vertices of a regular polygon
inscribed in that circle.

Problem 3.
Let A1 A2 :::An be a regular polygon inscribed in a circle C of radius R.
Xn
1
Find out if the sum P A2
attains a minimum value when P moves on
k
k=1
the circle C, avoiding the vertices of the given polygon, and calculate the
minimum value, if this exists.
Solution
Step 1. We may assume, without loss of generality, that the circle C is
centered at the origin and that Ak has the a¢ x
2 (k 1) 2 (k 1)
zk = cos + i sin ; for k = 1; n:
n n
Denote by z = cos t + i sin t the a¢ x of P , where t 2 [0; 2 ). Then P A2k =
jz zk j2 = (z zk )(z zk ), hence
(3) P A2k = 2R2 (zk z + zk z)
2
Using the trigonometric form of complex numbers and denoting := n ,
(3) implies
(4) P A2k = 2R2 (1 cos (t (k 1) )) .
for k = 1; n.
Remark 5. (3) implies
n
X n
X n
X n
X
P A2k = 2R2 (zk z + zk z) = 2nR2 z zk z zk :
k=1 k=1 k=1 k=1
n
X
But zk = 0, as sum of the roots of the equation z n 1 = 0, therefore
k=1
n
X
P A2k = 2nR2 :
k=1
 A problem solving approach to some applications of the roots of unity to regular polygons 55

n
X
Moreover, if OP = , then P A2k = n 2 + R2 (see [9] or [1] for a
k=1
problem solving approach to this formula)
By (4) we have
n
X n
1 1 X 1
= , t 2 [0; 2 ]nf(k 1) : k = 1; :::; ng .
P A2k 2R2 1 cos (t (k 1) )
k=1 k=1

Step 2.
n
X
1
We study the existence of a minimum for the sum P A2k
.
k=1
Let us de…ne a function fn : n ! R, where n := [0; 2 ]nf(k 1) : k = 1; :::; ng,
by
n
X 1
fn (t) = ;t 2 n:
1 cos (t (k 1) )
k=1
The function fn is continuous on n . The set n is not closed, there-
fore it is not compact. It is easy to see that fn has no maximum, since
lim fn (t) = +1 for k = 1; n. We have to prove that fn has a global
t!(k 1)
minimum, but we cannot take advantage by the well-known boundedness
property of real continuous functions on compact topological spaces.

The geometrical symmetry of the problem shows that it su¢ ces to assume
t 2 (0; ). In fact, extending fn to a periodic function ff
n : R n fk : k 2 Zg
with period 2 , we see that ffn has also the period = 2
n .

In Figure 5 and Figure 6 we represent a part of the graph of fn in the

cases n = 3 and n = 5, respectively:

150

100

50

1 2 3 4 5 6 7
Figure 5
Com p ut ed by Wolfram Alp ha
56 Marcelina Mocanu

400

300

200

100

1 2 3 4 5 6 7
Figure 6
Com p ut ed by Wolfram Alp ha

Since fn is continuous on (0; ) and

(5) limfn (t) = lim fn (t) = +1,
t&0 t%

the function fn attains its global minimum at a point in (0; ). In order

to prove this, consider m = inf fn ((0; )). It is clear that m 0, since
fn (t) > 0 for every t 2 (0; ). Let s > m. By (5), we can …nd > 0 with
< 2 , such that fn (t) > s whenever t 2 (0; ) [ ( ; ). Since fn is
continuous on the compact interval [ ; ], there exists tmin 2 [ ; ]
such that fn (tmin ) fn (t) for every t 2 [ ; ]. We conclude that
fn (tmin ) fn (t) for every t 2 (0; ), hence fn attains its minimum on
(0; ) at tmin .
The …rst derivative or fn is

n
X sin (t (k 1) )
fn0 (t) = :
k=1
[1 cos (t (k 1) )]2
It is di¢ cult to solve the equation fn0 (t), t 2 (0; ) in the general case.
Instead of trying to solve this equation, we will study the intervals of con-
vexity/ concavity of fn .
The second derivative of fn is
Xn
2 + cos (t (k 1) )
fn00 (t) = :
k=1
[1 cos (t (k 1) )]2

Since fn00 (t) > 0 for every t 2 (0; ), the function fn is strictly convex
on (0; ), i.e. fn ((1 ) t1 + t2 ) < (1 ) fn (t1 ) + fn (t2 ) whenever
2 (0; 1) and t1 6= t2 are in (0; ). The strict convexity of fn implies
the uniqueness of the minimum point. Indeed, assuming that t1 6= t2 are
two minimum points of fn on (0; ), that is fn (t1 ) = fn (t2 ) = m, we get
fn ((1 ) t1 + t2 ) < m whenever 2 (0; 1), which is a contradiction.
 A problem solving approach to some applications of the roots of unity to regular polygons 57

Again, the geometric symmetry of the problem inspires us to …nd the

n
X
1
unique minimum point. Note that the expression S (P ) := P A2
is in-
k
k=1
variant to a symmetry with respect to the perpendicular bisector b of the
line segment [A1 A2 ]. Assuming that S (P ) attains its minimum at a point
P1 2 C\IntA\ 1 OA2 and that P1 and P2 are symmetric to each other with
respect to b, it follows that S (P ) attains its minimum also at P2 . If P1
and P2 are distinct, this contradicts the uniqueness in (0; )of the minimum
point of fn , otherwise P1 = P2 is the midpoint of the arc C\IntA\ 1 OA2 .
In analytical terms, the graph of fn is symmetric with respect to the line
t = 2 , i.e. fn (t) = fn ( t) for all t 2 (0; ) and from the existence and
uniqueness of the minimum point we get min ffn (t) : t 2 (0; )g = fn 2 .
We should have fn0 2 = 0, according to Fermat’s Theorem. Indeed,
n
X sin (k 1)
fn0 = 2
2
2 1 cos (k 1)
k=1 2

But k + h = n + 1 implies cos 2 (k 1) = cos 2 (h 1) and

sin 2 (k 1) = sin 2 (h 1) , therefore

fn0 = 0.
2
Since fn0 2 = 0, fn00 2 > 0, the function fn has a local minimum at
2 , by the second derivative test. Since fn has a unique point of global
minimum, by the above discussion, it follows that this minimum point is 2 .
The global minimum of fn is

n
X 1
n := fn = ;
2
k=1
1 cos 2k2 3

2
where = n , hence
n
X1 n
X
1 1
n = = :
h=0
1 cos 2hn 1 k=1
1 cos 2kn 1
n
X
1 1
The minimum of P A2k
for P 2 C n Ak : k = 0; n is 2R2 n.
k=1

Step 3. In order to …nd a closed form for n we can use an heuristical

approach, with the aid of mathematical software. In our case, we used
Maple. We calculate n (by using "Evaluate", then "Simplify") for n =
3; 4; 5; 6; 7; 8; 9; 10::We …nd 3 = 92 , 4 = 8, 5 = 25
2 , 6 = 18, 7 = 2 ,
49
81 121
8 = 32, 9 = 2 , 10 = 50, 11 = 2 .
The rough values provided by the computer could look rather compli-
cated, sometimes it is necessary to use a combinationation of commands as
in the following case, where we used "Evaluate"+"Simplify"+"Evaluate nu-
merically", or better, "Evaluate"+Evaluate numerically", as in the following
58 Marcelina Mocanu

example:
9
X 1 2 2 2 9
(2k 1)
= 2 1 + 4 + =
1 cos cos 9 +1 cos 9 1 cos 9 +1 2
k=1 9

1 2 1 4 81
2 1 4 188 cos 188 cos + 188 cos + 81 = :
24 cos 9 24 cos 9 + 24 cos 9 +2 9 9 9 2
Similarly,
10
X 1
=
k=1 1 cos (2k101)

2 2 2 2
1
p p p 1
p p p 1
p pp + p pp
1
+2 = 50:0
4 2 5 5+1 4 2 5 5 1 4 2 5+5 1 4 2 5+5+1
and
11
X 1
=
k=1 1 cos (2k111)

2 2 2 2 2 1
2 1 3 + 4 5 + = 60: 5
cos 11 + 1 cos 11 1 cos 11 1 cos 11 + 1 cos 11 1 2
The "empirical data" lead us to conjecture that
n
X 1 n2
(6) = :
k=1 1 cos (2k n 1) 2

It is not of great value to continue to test this conjecture for the next values
of n, although this may help. Instead of getting "concrete proofs" we need
a rigorous deductive proof.
We have
1 1 1 (2k 1)
(2k 1)
= = 1 + cot2
1 cos n 2 sin2 (2k n 1) 2 n

for k 2 f0; 1; :::; n 1g. The relation (6) is equivalent to

n
X (2k 1)
(7) cot2 = n (n 1) :
n
k=1

A similar identity, namely

n
X k 2n (n + 1)
(8) csc2 =
2n + 1 3
k=1

is proved in the book "Non-elementary problems in an elementary exposi-

tion" by A.M. Iaglom and I.M. Iaglom. The idea for the proof of (8) is to
k
start with sin (2n + 1) 2n+1 = 0 and to apply the formula of sin (2n + 1)
as a polynomial of sin and cos (see (2)).
 A problem solving approach to some applications of the roots of unity to regular polygons 59

Here we use the identity cos n (2k2n1) = 0, k = 1; 2; :::; n. For each

n o
2 (2k2n1) : k = 1; n it follows from (1) that

[ n2 ]
X [ n2 ]
X
k
0 = cos n = ( 1) Cn2k cos n 2k
sin 2k
= sin n
( 1)k Cn2k cotn 2k
:
k=0 k=0
(2k 1)
Since sin 2n 6= 0 for k = 1; 2; :::; n, the equation
[ n2 ]
X
( 1)k Cn2k xn 2k
= 0, x 2 R
k=0
n
X
(2k 1)
admits the distinct roots xk = cot 2n , k = 1; 2; :::; n. Since xk = 0
X k=1
and xk xh = Cn2 , it follows that
1 k<h n
n n
!2
X X X
(xk )2 = xk 2 xk xh = n (n 1) :
k=1 k=1 1 k<h n

So (7) holds, hence (6) is true for every n 3. Then we have the following
n
X (2k 1)
Lemma 3.1. cot2 n = n (n 1) :
k=1
n
X1
1
By Lemma 3.1, we conclude that the required minimum of P A2k
is
k=0
1
4R2
n2 .

We proved the following

Proposition 3.2. Let = A0 A1 A2 :::An be a regular polygon inscribed
in a circle C of radius R. If P moves on the circle C, avoiding the vertices
n
X1
1 1
of the given polygon, then P A2 4R2
n2 , where the inequality becomes
k
k=0
an equality if and only if P is the midpoint of a small arc of C between two
consecutive vertices of .

4. Conclusions

There is a strong interplay between the algebra of complex numbers and

the geometry of regular polygons, in which polynomials occupy a central
place. It is hard to deal with applications like the ones we discussed with-
out a good knowledge of polynomials. Unfortunately, at the time being,
some Mathematics curricula prescribe the teaching and learning of complex
numbers without dealing with polynomials.
60 Marcelina Mocanu

References
[1] Andreescu, T. and Andrica, D., Complex Numbers From A To....Z, Birkäuser, USA,
2005.
[2] Calcut, J.S., Rationality and the tangent function, Preprint, 2006,
http://www.oberlin.edu/faculty/jcalcut/tanpap.pdf
[3] Calcut, J.S., Gaussian integers and arctangent identities for , American Mathemat-
ical Monthly, 116(6)(2009), 515-530.
[4] Djukić, D., Janković, V., Matić, I. and Petrović, N., The IMO Compendium-A Col-
lection of Problems Suggested for The International Mathematical Olympiads: 1959-
2009, Second Edition, Springer Verlag, 2011, 830 pages.
[5] Karpenkov, O., Geometry of Lattice Angles, Polygons and
Cones, Habilitation thesis, Technische Universität Graz, 2009,
http://www.geometrie.tugraz.at/karpenkov/hab/habilitation.pdf
[6] Klamkin, M.S. and Chrestenson, H.E., Polygon imbedded in a lattice, American Math-
ematical Monthly, 70(1963), 447–448.
[7] Iaglom, A.M. and Iaglom, I.M., Non-elementary problems in an elementary exposition
(Romanian), Editura Tehnic¼ a, Bucureşti, 1983.
[8] O’Loughlin, D.J., The scarcity of regular polygons on the integer lattice, Math. Mag.,
75(2002), 47–51.
[9] Oullette, H. and Bennett, G., The Discovery of a Generalization: An Example in
Problem Solving, The College Mathematics Journal, 10(2)(1979), 100-106.
[10] Schoenberg, I.J., Regular simplices and quadratic forms, J. London Math. Soc.,
12(1937), 48–55.

DEPARTMENT OF MATHEMATICS
UNIVERSITY "VASILE ALECSANDRI" OF BACAU ¼
CALEA MAR ¼ AŞEŞTI
¼ ¼ ROMANIA
157, 600115 BACAU,
E-mail address: [email protected]