This booklet is dedicated to

our science and technology martyrs;

the scientists who sacrificed their lives
for the progress of our country.

Majid Shahriari Masoud Alimohammadi

Born: 1966 Born: 1959
Assassinated: 2010 Assassinated: 2010

Mostafa Ahmadi-Roshan Daryoush Rezaeinejad

Born: 1980 Born: 1976
Assassinated: 2012 Assassinated: 2011

1
2
Iranian Team Members in 53rd IMO (Mar Del Plata-Argentina) (from left to right):
 Mina Dalirrooyfard
 Goodarz Mehr
 Alek Bedroya
 Pedram Safaei
 Alireza Fallah
 Amir Ali Moinfar

This booklet is prepared by Ali Khezeli, Hesam Rajabzadeh,

Morteza Saghafian and Masoud Shafaei.
With special thanks to Javad Abedi, Shayan Aziznejad, Ali Babaei, Atieh Khoshnevis, Omid
Naghshineh Arjmand and Sina Rezaie.
Copyright © Young Scholars Club 2011-2012. All rights reserved.

3
4
 29th
Iranian Mathematical Olympiad
2011-2012

Contents

Problems
First Round.…………………………………..………………………………………..........8
Second Round....……….…………………..…………………..……………………….……9
Third Round.…...……………………………..……………………..……………………..11

Solutions
First Round..………………..………….………………..…………….…….……………..16
Second Round..…………….……………………………..……………………...…………20
Third Round…………………………………...……………..……………….……………27

5
6
Problems

7
 First Round

1. (Mohsen Jamali) 1390 ants are near a straight line such that
the distance between the head of each ant and the line is less
than 1 centimeter and the distance between the head of each
two ants is more than 2 centimeters. Prove that there exist
two ants which their heads are at least 10 meters far.
(Assume that the head of each ant is a point!) (→p. 16)

2. (MirSaleh Bahavarnia) In triangle ABC we have BAC  60 . The perpendicular

line to AB at B intersects the bisector of BAC at D and the perpendicular line
to BC at C meets the bisector of ABC at E . Prove that BED  30 . (→p. 16)

3. (Mohsen Jamali) Determine all increasing sequence a1, a2 , a 3,  of positive integers

such that for every i, j   , the number of positive divisors of i  j and ai  a j are
equal (A sequence a1, a2, a3 ,  is increasing if i  j implies ai  a j ). (→p. 17)

4. (Mohammad Mansouri, Shayan Dashmiz) Find the smallest positive integer n such
that there exist n real numbers in the interval ( 1,1) such that their sum is zero
and the sum of their squares equals 20. (→p. 17)

5. (Mohsen Jamali) A beautiful rare bird called n -colored Rainbow

can be seen in one of n different colors each day and its color is
always different from the previous day. Recently, scientists have
discovered a new fact about this bird’s life: “there does not exist
four days like i, j , k , l in its life such that i  j  k  l with
colors namely a, b, c, d respectively, such that a  c  b  d ”. Find the maximum
possible age of an n -colored Rainbow as a function of n . (→p. 18)

6. (Ali Khezeli) We have extended the sides AB and AC of triangle ABC from B
and C respectively to intersects a given line l at D and E respectively. Suppose
the reflection of l with respect to perpendicular bisector of BC intersects mentioned
extensions at D and E respectively. If BD  CE  DE , show that
BD   CE   D E  . (→p. 18)

8
 Second Round

1. (Kasra Alishahi) A regular dodecahedron is a convex

polyhedron such that its faces are regular pentagons. It has 20
vertices and 3 edges connected to each vertex. (As you see in
the picture.)
Suppose that we have marked 10 vertices of a regular
dodecahedron.
a) Prove that we can rotate the dodecahedron in such a way
that the dodecahedron is mapped to itself and at most four marked vertices are
mapped to a marked vertex.
b) Prove that number 4 cannot be replaced with number 3 in the previous part.
(→p. 20)

2. (Mohammad Mansouri) Prove that for every positive integers k and n there exist k
monic polynomials P1(x ), P2 (x ),, Pk (x ) of degree n with integer coefficient such
that each two of them have no common factor and the sum of each arbitrary number
of them has all its roots real. (→p. 20)

3. (Erfan Salavati) Four metal pieces are joined to

each other to form a quadrilateral in the space.
The angle between them can vary freely. In a case
that the quadrilateral is not planar, we mark one
point of each piece such that the points lie in a
plane. Prove that these four points are always
coplanar as the quadrilateral varies. (→p. 21)

4. (Mohammad Ghiasi) The escalator of “Champion Butcher”

metro station has this property that if m persons are on it,
its speed is m  where  is a positive constant number.
Suppose that n persons want to go upstairs by the escalator.
If the length of the escalator is l , what is the least time
required for these persons to go to upstairs? (Suppose the
persons can use the escalator simultaneously at any time). (→p. 21)

9
5. (Mahyar Sefidgaran, Mostafa Eynollahzadeh) Let  be a real number and
a1, a2 , a 3 , a strictly increasing sequence of positive integers such that for every
n   , an  n  . A prime number q is called golden if there is a positive integer m
such that q am . Suppose that q1  q 2  q 3   are all golden prime numbers.
a) Prove that if   1.5 , then qn  1390n .
b) Prove that if   2.4 , then qn  13902n . (→p. 22)

6. (Ali Khezeli) Two circles in the space are called

linking if they intersect at two points or they are
interlocked. Find a necessary and sufficient
condition for four distinct points A, B, A, B  in the
space such that every two different circles passing
through A, B and the other passing through A, B  respectively are linking. (→p. 23)

7. (Sepehr GhaziNezami) For a function f :  ()   and a subset A   , we say f

is A -predictor if the set {x   | x  A , f (A  {x })  x } is finite. Prove that there
exists a function that for every subset A of natural numbers is A -predictor. (→p. 24)

8. (Ali Khezeli) A sequence d1,, dn of not necessarily distinct natural numbers is called
a covering sequence if there exist arithmetic progressions of the form
{ai  kdi : k  0,1, 2,} such that every natural number comes in at least one of
them. We call this sequence minimal if we cannot delete any of d1,, dn such that
the resulting sequence is still covering.
a) Suppose d1,, dn is a minimal covering sequence and suppose we've covered all the
natural numbers with arithmetic progressions {ai  kdi : k  0,1,2,} . Suppose that
p is a prime number that divides d1,, dk but does not divide dk 1,, dn . Prove that
the remainders of a1,, ak modulo p contain all the numbers 0,1,, p  1 .
b) Write anything you can about covering sequences and minimal covering sequences
in the case that each of d1,, dn has only one prime divisor. (→p. 24)

10
 Third Round

1. (Mahdi E’tesamiFard) Find all natural numbers n  2 such that for every pair of
n  n 
integers i, j  [0, n ] , i  j and      have the same parity. (→. 27)
 i   j 

2. (Mahdi E’tesamiFard) Let  be the circumcircle of an acute triangle ABC . Let D


be the midpoint of arc BAC in  and I be the incenter of triangle ABC . Suppose
DI intersects BC at E and  at F for the second time. Suppose the parallel line
to AI from E meets AF at P . Prove that PE is the bisector of BPC . (→p. 28)

3. (Erfan Salavati) Let n be a natural number. A subset S of points in the plane has
following properties:
i) There are not n lines in the plane such that each element of S lies on at least one
of them.
ii) For every X  S there exist n lines in the plane such that each point of S  {X }
lies on at least one of them.
Find the maximum possible number of points in S . (→p. 28)

4. (Ali Khezeli) There are m  1 horizontal and n  1 vertical lines (m, n  4) that
make a m  n table. Consider a closed path that does not intersect itself and passes
through all (m  1)(n  1) interior vertices and none of
the outer vertices. (Each vertex is the intersection
point of two lines!) A is the number of interior
vertices that the path passes through them straight-
forward, B is the number of squares in the table that
only two opposite sides of them are used in the path
and C is the number of squares that none of its sides
are used in the path. Prove that A  B  C  m  n  1. (→p. 29)

5. (Masoud Shafaei) Let f :   0    0 be a function such that for all a, b   0 :

i) f (a )  0  a  0 .
ii) f (ab )  f (a )f (b) .
iii) f (a  b)  2 max{f (a ), f (b )} .

11
 Prove that for every a,b    0 , f (a  b)  f (a )  f (b) . (→p. 31)

6. (Morteza Saghafian, Ali Khezeli) Let ABCDE be a cyclic pentagon with

circumcircle  . Suppose that a , b , c , d , e are the reflections of  with respect to
AB , BC , CD , DE and EA respectively. A is the second intersection of a and
e . B ,C , D , E  are defined similarly. Prove that
S (AB C D E )
2  3,
S (ABCDE )
where S (X ) denotes the area of figure X . (→p. 32)

n 
7. (Morteza Saghafian) Is it possible to write   consecutive natural numbers on the
 2 
edges of a complete graph with n vertices such that for every path (or cycle) of
length 3 with edges a,b, c ( b lies between a, c ) the greatest common divisor of the
numbers of edges a and c divides the number of edge b ? (→p. 33)

8. (Mohammad Ja’fari) Let g be a polynomial of degree at least 2 with nonnegative

coefficients. Find all functions f :      such that for every x, y   
f ( f (x )  g(x )  2y )  f (x )  g (x )  2 f (y ). (→p. 34)

9. (Ali Khezeli) Let ABCD be a parallelogram. Consider circles 1 and 2 such that
1 is tangent to segments AB , AD and 2 is tangent to segments BC ,CD . Suppose
that there exist a circle tangent to lines AD, DC and externally tangent to 1, 2 .
Prove that there exists a circle tangent to lines AB and BC and externally tangent
to 1, 2 . (→p. 35)

10. (Morteza Saghafian) Let a,b, c be positive real numbers such that ab  bc  ca  1.
Show that
a a b b c c (→p. 35)
3( a  b  c )    ·
bc ca ab

11. (Mahdi E’tesamiFard, Ali Khezeli) Let A, B be two different points on a circle 
with center O such that 60  AOB  120 . Let C be the circumcenter of AOB.
l is a line passing through C such that the angle between l and OC is 60.
Tangent lines to  at A, B meets l at M , N respectively. Suppose the circumcircle

12
 of triangles CAM and CBN intersect  for the second time at Q and R
respectively and meets each other in P (different from C ). Prove that OP  QR .
(→p. 36)

12. (Javad Abedi) A subset B of natural numbers is called loyal if there exist positive
integers i  j such that B  {i, i  1, , j } . Q is the collection of loyal subsets of
natural numbers. For every subset A  {a1  a2    ak } of {1,2, , n } we define:
g (A)  max B and f (A)  max a i 1  ai .
B A,B Q 1i k 1

Also, Define
G (n )   g(A) and F (n )   f (A).
A{1,2,,n } A{1,2,,n }

Prove there exists m   such that for every positive integer n greater than m we
have F (n )  G (n ). ( A denotes the number of elements of a set A and if A  1 we
set f (A)  0 ). (→p. 37)

13. (Hesam Rajabzadeh) Consider a regular 2k -gon with center O in the plane and let
l1, l2,, l2k be its sides with the clockwise order. Reflect O with respect to l1 , reflect
the resulting point with respect to l2 and continue this process until the last side.
Prove that the distance between final point and O is less than the perimeter of the
mentioned 2k -gon. (→p. 38)

14. (Morteza Saghafian) Are there 2000 real numbers (not necessarily distinct), not all
zero, such that if we put any 1000 of these numbers as roots of a monic polynomial
of degree 1000, its coefficients (except the coefficient of x 1000 ) are a permutation of
the 1000 remaining numbers? (→p. 40)

15. (Mahyar Sefidgaran) Determine all integers x , y satisfying the equation

(y 3  xy  1)(x 2  x  y )  (x 3  xy  1)(y 2  x  y ). (→p. 41)

16. (Mohyeddin Motevassel) Let p be an odd prime number. We say a polynomial

n
f (x )  ajx j is i -residue if  a j  i (mod p ) .
j 0 p 1| j , j  0

13
 Show that  f (1),..., f (p  1)  is a complete residue system modulo p if and only if
polynomials f (x ) ,…, f (x )p2 are 0-residue and ( f (x ))p1 is 1-residue. (→p. 42)

17. (Ali Khezeli) n is a positive integer. Let A, B be two sets of n points in the plane
such that no three points of them are collinear. Denote by T (A) the number of non-
self-intersecting broken lines containing n  1 segments such that its vertices in A .
Define T (B ) similarly. If the elements of B are the vertices of a convex n -gon but
the elements of A are not, prove that T (B )  T (A) . (→p. 43)

18. (Mahdi E’tesamiFard) Let O be the circumcenter of triangle ABC . Points A, B ,C 
lie on the segments BC ,CA, AB respectively such that the cicumcircles of triangles
AB C  , BC A and CAB  pass through O . Denote by la the radical axis of the
circle with center B and radius BC and circle with center C  and radius C B .
Define lb and lc similarly. Prove that the lines la , lb , lc form a triangle such that its
orthocenter coincides with the orthocenter of triangle ABC . (→p. 44)

14
Solutions

15
 First Round

1. Consider the line in the problem as the x  axis in the coordinate plane and assume
that the y  axis is an arbitrary line perpendicular to the x  axis.
Let Ai  (x i , yi ) , 1  i  1390 be the coordinates of the head of ants. We can
assume that x1  x 2    x1390 .
Denote by C i , 1  i  1390 the circle with center Ai and radius 1. According to the
i j  2 for all 1  i  j  1 39 0 . So the circles are
problem’s condition, we have AA
pairwise disjoint and all of them are in the rectangle
A  {(x , y ) : x 1  1  x  x 1390  1,  2  y  2; 1  i  1390}.
Thus the sum of the areas of the circles is less than the area of the rectangle.
Therefore
1390    4(x1390  x1  2).
So
 1390  3  4(xn  x1  2)  1042.5  xn  x 1  2
 xn  x1  1000(cm )  10m.
Finally by Pythagorean Theorem we have An A1  x n  x1  10m . 

2. Denote by I the intersection point of AD and BE , so I is the incenter of triangle

BAC
ABC . Suppose that   . We have
2
CBA
IBD  90  IBA  90   90  30  60,
2
CBA
CEB  90  CBE  90   90  30  60.
2
Thus in triangles BEC and BED we have CEB  EBD  60 and BE is side
of both of them. Since BEC  30 it suffices to prove BD  CE but we have
BD EC
 tan  and  tan 30 , So
AB BC
AB
BD  CE  AB tan   BC tan 30  . tan   tan 30
BC
sin(120  2)
 tan   tan 30 (Law of Sines)
sin(2)
sin  sin(120  2)
  tan 30
cos  2 sin  .cos 
 sin(120  2)  2 cos2 . tan 30

16
  sin120. cos(2)  sin120.sin(2)  (1  cos(2))tan 30
 (sin120  tan 30). cos(2)  cos120.sin(2)  tan 30.
We know by Cauchy-Schwarz inequality that
(LHS )2  ((sin 120  tan 30)2  (cos120)2 )(cos2 (2)  sin2 (2))
3 3 2 1 3 1 1
(  )  ( )2     (tan 30)2  (RHS )2 . 
2 3 2 36 4 3

3. First it is easy to show that the sequence is strictly increasing. Assume that
ai  ai 1 for some i . Let j  p  i for a large prime p . Now i  j is a prime
number. So ai  a j is prime too. But from ai  ai 1 we get ai 1  a j and i  1  j
are prime numbers. So i  j and i  j  1 are two consecutive large prime numbers,
contradiction.
Now put i  j  2p 2 for a large prime p . So the number of divisors of 2ai  2p1
equals p . Hence 2ai is of the form q p 1 for a prime q . Obviously q should be 2 and
therefore we have a2p2  2p 2 .
Now we have a strictly increasing sequence of integers with infinitely many fixed
points. So, for each n , an  n . 

4. Suppose that a1, a2 ,, an satisfies the conditions. First, we have

20  a12  a22    an2  1
 1 
  1  n.
n times
So 21  n . We want to show that n  22 is the answer. So we prove that there are
not 21 numbers a1, a2 ,, a21 in the interval ( 1,1) such that a1  a 2    a21  0
and a12  a22    a21
2
 20 . Assume that sequence ai is in increasing order so
a1  a2    a21
a1   a21 . Thus a1  0  a21 . But because of minimality of
21
number 21, we have ai  0 for 1  i  21 . So there exist a unique number
1  k  21 such that
1  a1  a2    ak  0  ak 1    a21  1.
We know that numbers a1, a2 ,, a21 satisfies the problem condition, too.
21
Thereby we can assume that k  and since k   we have k  10 .
2
Now for every k  1  i  21 . We have 0  ai  1 , so 0  ai2  ai .

17
 20  a12  a22    a212  (a12    ak2 )  (ak21    a21
2
)
 (a12    ak2 )  (ak 1    a21 )
 (a12    ak2 )  (a1  a2    ak )
 2k  20.
This contradiction shows that n  22 . The following numbers are an example for
n  22 and so the answer is 22.
11 11
ai  (1  i  11) and ai   (12  i  22). 
10 10

5. We prove by induction that the maximum number of days it can live is 2n  1 . The
construction is easy, like 1, 2,, n  1, n, n  1,,2, 1 (These are the labels of colors).
For n  1 the result is evident. Suppose that it is true for numbers smaller than n .
Let its color in the first day be R , and this colors appear k times, at days
R1, R2,, Rk .
Consider the interval of days (R1, R2 ),(R2 , R3 ),,(Rk 1, Rk ),(Rk ,) . If some color
appears in two intervals it contradicts the problem statement. So each color appears
in exactly one interval. Suppose that there are C i colors in the interval (Ri , Ri 1 ) , so
k

C i  n  1 . By the induction hypothesis, the interval (Ri , Ri 1 ) consists of at

i 1
most 2C i  1 days. Only the last interval can be empty. If it is empty then the
k 1
number of days is at most k   (2C i  1)  2n  1 . Otherwise, the number of days
i 1
k
is at most k   (2C i  1)  2n  2 and we’re done. 
i 1

6. Suppose that M is the midpoint of side BC and a its perpendicular bisector. X is

the intersection point of lines a and l . Put MXE  , ABC   and
ACB   . Let Y be a point on segment DE such that EY  CE and Z the
reflection of Y with respect to a , So Z lies on l . It is obvious that BCYZ is a
trapezoid and hence cyclic. Denote by K , the intersection point of l and
circumcircle of BCYZ (The solution is similar when the circle is tangent to l ). Now
we want to prove that D B  D K .

18
CED  360  (MCE  CMX  MXE )
 360  (180    90  )  90     Obviously,
180  CED 180  90      
 CYE    45   . (1)
2 2 2 2
triangle ZXY is isosceles, so XYZ  90  MXE  90  . (2)
   
(1),(2)  CYZ  180  (90  )  (45   )  45   .
2 2 2 2
 
CYKZ is cyclic, so CKZ  CYZ  45   . Therefore
2 2
 
CKE   180  CKZ  135   . (3)
2 2
CE K  360  (MCE   XMC  MXE )
 360  (180    90  180  )      90 . (4)
So
KCE   180  (CE K  CKE )
(3),(4)
 
 KCE   180  (    90  135   )
2 2
 
 135   CKE .
2 2
Thereby triangle CE K is isosceles and hence CE   KE  . Similarly, we have
BD   KD  and so D E   D K  KE   BD   CE  and this is desired assertion.


19
 Second Round

1. a) First we count the number of rotations of a dodecahedron. In 12 ways we can

change a face of the dodecahedron with the down face. Although we can put this face
in 5 distinct ways but one of this 12  5  60 states is the original state, so we have
59 states for rotating the dodecahedron.
Now consider one of the marked vertices named A . We have 10 marked vertices and
each of them can lie on its locations in 3 distinct ways, so in 3  10  1  29
rotations (other than the original state) a special vertex lie on the location of A in
original state, therefore totally 29  10  290 times a marked vertex lie on a marked
vertex of the original state.
290
Now by the Pigeonhole principle less than  5 of this coincidences is in one of
59
the states and so we’re done. 

b) It suffices to give an example that in each rotation at least 4 marked vertices lie
on a marked vertex of original state.
Mark the vertices of two opposite faces. Suppose that there exists a rotation with at
most 3 coincidences. Consider this state is  .
So one of the marked faces in  has at most one marked vertex of the original state,
but each face has at least two marked vertices, contradiction. 

2. For each 1  i  k we define

Pi (x )  (x  i )(x  (k  i ))(x  ((n  1)k  i )).
We claim that these polynomials satisfy the problem condition.
For each 1  i  k and each 0  j  n  1 , Pi (x ) has exactly one simple root in the
1 1
interval ( jk  ,( j  1)k  ) so invoking the mean value theorem we deduce that
2 2
1 1 1
Pi ( jk  ) and Pi (( j  1)k  ) have different signs. Note that Pi (nk  )  0
2 2 2
because Pi is monic and so is positive for large positive values and does not have any
1
root greater than n . Thus for each 1  i  k , Pi ( jk  )  0 if j  n (mod2) and
2
1
Pi ( jk  )  0 if j 
 n (mod2) .
2

20
 Now let Q(x )  Pi (x )  Pi (x )    Pi (x ) where i1, i2 ,, it  {1, 2,, k } are
1 2 t

distinct. Obviously Q(x )  [x ] is a polynomial of degree n .

1 1
For each 0  j  n  1 , numbers Q( jk  ) and Q(( j  1)k  ) have different signs
2 2
because Pi , Pi ,, Pi have this property. So again according to mean value theorem
1 2 t

1 1
we deduce that Q has a root in the interval ( jk  ,( j  1)k  ) and Q (x ) has at
2 2
most n real roots so all its roots are real, hence the claim is proved. 

3. Let quadrilateral that this four pieces form in the space

be ABCD and we marked points M , N , P and Q on
sides AB, BC ,CD and DA respectively, So that the
marked points are on a plane named  . Let a,b, c and
d be the distances from A, B,C and D to 
respectively. So we have
AM a BN b CP c DQ d AM BN CP DQ
 ,  ,  ,       1.
MB b NC c PD d QA a MB NC PD QA
Now ABCD varies, let  be the plane passing through M , N , P . Suppose that
a ,b , c  and d are the distances of A, B,C and D to  respectively. Thus
AM a  BN b  CP c DQ d
 ,  ,    .
MB b  NC c  PD d QA a
So Q must be on  too. 

4. In every moment consider the number of persons that are on the escalator at that
time. Now consider the intervals such that in every time of such intervals the
number of persons on the escalator is equal to a fixed integer. Suppose that we have
k intervals I 1, I 2 ,, I k and for 1  i  k , ai and li denotes the length and number
k
of persons on the escalator in every moment of I i . We claim that  ai1ti  nl .
i 1
Since every person have moved a distance equal l so the sum of travelled distance of
people is nl . On the other hand travelled distance of a person who is on the
escalator on the interval I i equals aiti . So the sum of travelled distance in the
k
interval I i equals ai .aiti  ai1ti , hence the total travelled distance is  ai1ti .
i 1

So the claim is proved.

21
 Now consider the following cases.
Case 1.   1 . If ai  1 since   1 we have ai1  1 , hence ai1  1 . If ai  0
k k
also ai1  0  1 . So nl   ai1ti   ti . So the required time is at least nl . If
i 1 i 1

each person goes on the escalator when the previous one reached the top of the
escalator the required time equals nl . Hence in this case the required time is at least
nl .
Case 2.   1 . Since ai  n and 1    0 we have ai1  n1 so
k k k k
nl   ai1ti   n 1ti  n 1  ti  n l   ti
i 1 i 1 i 1 i 1

So the required time is at least n l . If all n people go on the escalator together the
l
velocity equals n  and so the required time equals  n l . Hence in this case

n

the required time is at least n l . 

5. a) Denote by t the number of golden prime numbers less than or equal to 1390n .
We want to show that t  n . Suppose that S is collection of all natural numbers
less than or equal to 1390n with prime factors from the set q1, q 2,, qt  . Obviously
each element of S can be written in the form a 2b where a,b   and b is out of
n
  
square. So a  1390n  1390 2 and b  q1 1q 2 2 qt t such that i   0,1  .
n n
Therefore a and b have 1390 2 and 2t states respectively, and so S  2t  1390 2 .
2
n
On the other hand for each integer 1  i  k  1390 3 we have
2
n1.5
ai  i 1.5  k 1.5  1390 3  1390n .
2
n
And all prime divisors of ai are in the set q1, q 2,, qt  , so ai  S (1  i  1390 3 ) .
2 n
n
Therefore S has at least  k  elements. So 13903  1   k   S  2t  1390 2 , But
1 2
it is easy to check that 2  13902  1390 3  1 and this implies t  n, because
n 1 2 2 n
n
n
2  1390 2  (2  1390 2 )n  (1390 3 n
 1)  1390 3 1  2 t
 1390 2 . 
b) The proof of this part is very similar to part a. Denote by t the number of golden
prime numbers less than or equal to 13902n . We want to show that t  n . Suppose

22
 that S is collection of all natural numbers less than or equal to 13902n with prime
factors from the set {q1, q2 ,, qt } . Then every element of S can be written in the
form a 4b 2c where a, b, c   and b, c are out of square. (In part a writing a as x 2y
where x, y   and y is out of square implies this claim.). Now
n
4 2n      
a  1390  1390 2 , b  q1 1q 2 2 qt t and c  q1 1q 2 2 qt t such that i , i  {0,1} .
n
Thus we have 1390 2 , 2t and 2t states for a,b and c respectively and so
n
2t
S 2  1390 2 .
5 5
n n2.4
In this case if 1  i  k  1390 6 (i  ) then ai  i 2.4  k 2.4  1390 6  13902n
. Although prime divisors of ai (1  i  k ) are in the set {q1, q 2,, qt } so ai  S ,
5 n
n
hence 1390 6  1   k   S  22t  1390 2 . On the other hand
1 5
4  13902  13906  1 and so
n 1 5 5 n
n
2n
2  1390 2  (4  13902 )n  (13906 n
 1)  1390 6 1  2 2t
 1390 2 ,
which implies t  n . 

6. The points should be on a circle (or line) and A, B should separate A, B  on it. To
prove necessity, first suppose that the points are not coplanar. Then there exist two
parallel planes passing through A, B and A, B  respectively. Any two circles in these
planes are not linking. So the points should be coplanar.
Now suppose B  is not on the circumcircle of ABA (which can be a line). So we can
slightly change the circle to find a circle passing through A, B such that A, B  are
both outside or both inside it. Now, this circle is not linking with the circle with
diameter A B  orthogonal to the plane containing the points.
So the points should be on a circle (or line). Now, suppose A, B do not separate
A, B  on the circle. If we change the circle slightly, still passing through A, B , then
A, B  will be both inside or both outside the new circle and we arrive to a
contradiction like the previous case. So, the necessity of the condition is proved.
To prove sufficiency, Let C ,C  be two different circles passing through A, B and
A, B  respectively. Let P , P  be the planes containing C ,C  respectively. If the
points are collinear, then C   P is consisted of a point inside C and a point outside
C . So C ,C  are interlocked. So, suppose the points are on a circle. Let M be the

23
 intersection of the segments AB and A B  . We have MAM
. B  MA.MB  . Let
l  P  P  which passes through M . M is inside C , so l intersects C at two
points like X ,Y and M is between X ,Y . Similarly, l intersects C  at X ,Y 
namely, and M is between X ,Y  . Suppose X, X  are in one side of M . We have
MX .MY  MA.MB  MA.MB   MX .MY .
So if MX  MX  , then MY  MY  and vice versa. So the points of
C   P   X ,Y   are in different sides of C or both are on C . So C ,C  are linking
and sufficiency of the condition is proved. 

7. Define f (A)  max(A) when A is finite. Evidently, f is A -predicator when A is

finite. We extend f to all subsets of  . We say two subsets A, B are equivalent if
B is derived from A by adding and deleting a finite number of elements; i.e. AB
is finite. This is an equivalence relation. By the Axiom of Choice, we can select an
element from each class of equivalency. For an arbitrary proper subset A , let S A be
the selected element from the class of A . So, AS A is finite. Define
f (A)  max(ASA ) when A  S A and define f (S A ) arbitrarily. We claim this
function is A -predicator for all subsets A   .
 A . A and A   x  are equivalent, so
Let x be a natural number such that x 
SA  SA{x } . So
f (A   x )  f (A  x )  max((A  x )S A )  max((AS A )  x ).
For x  max(AS A ) we have (AS A )  x   (AS A )   x  . Therefore
f (A   x )  x and the claim is proved. 

8. a) Let’s name the arithmetic progressions as S1,, Sn . Suppose the remainders of

a1,, ak modulo p don’t contain r . So no member of S1,, Sk is r (mod p ).
Consider the arithmetic progression S   r  ip : i  0,1,2, which has empty
intersection with S1,, Sk . So S is covered by Sk 1,, Sn .
Lemma. If d, d  are coprime natural numbers, then intersection of the arithmetic
progressions a  kd : k  0,1,2, and a   kd  : k  0,1,2, is an arithmetic
progression whose common difference is dd  .
Proof. The sequences have a nonempty intersection by the Chinese remainder
theorem. The difference of two consecutive elements of the intersection is the least
common multiple of d, d  which is dd  . 

24
According to the lemma, the sets S  Sk 1,, S  Sn are arithmetic progressions
with common differences pdk 1,, pdn . Consider the map f : S   0,1, 2, with
x r
formula f (x ) 
. So, the sets f (S  Sk 1 ),, f (S  S n ) are arithmetic
p
progressions with common differences dk 1,, dn . These sets cover the natural
numbers, because S is covered by Sk 1,, Sn . So dk 1,, dn is also a covering
sequence which contradicts the minimality of d1,, dn . 
b) Let p1,, pk be the prime factors of d1,, dn and let I i   j : pi | d j  (which can
contain multiplicities). Suppose the natural numbers are covered by arithmetic
progressions Si  ai  kdi : k  0,1,2, . We claim that at least one of the sets of
sequences Si   S j : j  I i  covers the natural numbers. As a result, one of the
sequences d j : pi | d j  is a covering sequence.
Suppose for each i the sequences in Si doesn’t cover a number ri . Let Di   dj .
pi |d j

By the Chinese remainder theorem, there is a natural number r such that r  ri

mod Di for each i . So r is not covered by any of the arithmetic progressions, a
contradiction. So the claim is proved.
Now, it is enough to suppose di  pri . We claim it is a covering sequence if and only
1 1
if d  1 and it is minimal if and only if d  1.
i i i i

i) First, suppose we have covered the natural numbers by progressions S i as

N 1
above. For any natural number N , S i covers at most members of
di
N 1 1 N
1,,N  . So we have   N . So d  . So we should have
i di i i N 1
1
d  1.
i i
1
ii) Second, suppose d  1 . We use induction on n to prove that this is a
i i

covering sequence. If n  1 it is evident. If not, let ni be the number of p i .’s

in d1,, dn . We can suppose n 0  0 . There should exist i such that ni  p ,

1 1 1
or else d  (p  1)( 
p p2
 )  1 . Remove p ones of p i ’s and add a
i i
i 1
p to d1,, dn . The sum doesn’t change. So, cover the natural numbers by
progressions according to the induction hypothesis. Now, split a progression

25
 with common difference p i 1 into p progressions with common difference p i .
The induction claim is proved.
1 d
iii) Moreover, suppose d  1 . Suppose d1    dn . We have  dn  dn
i i i i
1 1
and the summands are natural numbers. So d 1
dn
. So, we can
i i
remove dn and the sequence is not minimal. 

26
 Third Round

 n  n  n 
1. Lemma. Suppose n  2 is an integer, then all of the numbers  ,  ,,   are odd
 0   1   n 
if and only if n  2k  1 for an integer k  2 .
Proof. For an integer t , let v (t ) be the greatest integer u such that 2u | n . We
know that
“ If p, q   and 0  p  2q then v(p)  v(2q  p)  v(2q  p) ” (1)
Because if p  2a b with a,b   and b an odd number then a  q and
2q  p  2a (2q a  b) where 2q a  b are odd numbers.
Now there is one and only one m   such that 2m  n  2m 1 . Let n  2m  s
 n 
with 0  s  2m . Now consider the number  m  . We have
 2  1 
 n   2m  s   2m  s  (2m  s )(2m  s  1)(2m  1)(2m )
    
 2m  1   2m  1    s  1   s(s  1)(1)(s  1)
.
     
By (1) we have v(2m  i )  v(i ) where 1  i  s , therefore
v((2m  s )(2m  1))  v(s !),
 n 
and by assumption  m  is odd therefore v(2m )  v(s  1) and consequently
 2  1 

2m | s  1 and s  1  1 . Hence we have 2m  1  s and therefore s  2m  1 and

n  2m  s  2m 1  1 .
Now if n  2k  1 for some natural number k , for each 1  c  n we have
 2k  1  (2k  1)(2k  2) (2k  c)
  and by (1) we know for 1  l  c
 c   (1)(2) (c)
 
 2k  1 
v(2k  l )  v(l ) , so   is odd for 0  c  n . 
 c 
Now we return to main problem:
Positive integer n has the property of the problem if and only if all the numbers
 n 
   i (0  i  n ) have the same parity. It means that for every 0  i  n  1 ,
 i 
 
 n   n   n  1   n   n 
 ,        
 i   i  1  have different parities. So  i  1    i    i  1  (1  i  n  1) is odd
        

27
  n  1 
and as we know    1 is also odd. Therefore by the lemma this is equivalence
 0 
to n  1  2k  1 for some integer k  2 so n  2k  2 where k  2 is an integer.


2. Let T be the point of intersection of

perpendicular bisector of BC and circle  , so
TD is a diagonal of  and DFT  90 .

Since D is the midpoint of arc BAC , FD is
the angle bisector of BFC . Therefore
(JEBC )  1 where J is the intersection
point of line TF and extension of BC .
1   1   1
EJF  (CT  BF )  (BT  BF )  FT  TAF  EPF
2 2 2
So quadrilateral PEFJ is cyclic and consequently JPE  90 . This fact and
(JEBC )  1 implies that PE is the angle bisector of BPC and this is what we
wanted to prove. 

3. For each p  (x p , y p )  S suppose that Li, p : ai, px  bi, py  ci, p  0 (1  i  n ) are

n lines that cover S  {p} . We know that p  L1, p  L2, p    Ln, p . Let
n ai, p x  bi, py  ci, p
Pp (x, y )   ( ) (p  Li, p  ai, px p  bi, py p  ci, p  0)
i 1 ai, p x p  bi, py p  ci, p

Pp is a polynomial of degree n in two variables and we have Pp |S {p}  0 and

Pp (p)  1 so {Pp }p S are linearly independent in Fn , where Fn denotes the vector
space of polynomials in two variables of degree less than or equal to n , therefore
 n  2 
S  dimFn    .
 2 

28
  n  2 
Now we give an example of S with   points.
 2 
 n  2 
Consider U n be   points forming a triangular lattice (The left figure above
 2 
 n  2 
shows U 8 .) Then U n  1  2    (n  1)    . For every point p  U n we
 2 
can cover U n  {p} with n lines (Right figure above shows an example for n  8 ).
Now we use induction on n to prove that we cannot cover U n with n lines. The
base n  1 is obvious. Assume the statement to be true for n  k  1 and suppose
U n is covered with k  1 lines. Now consider line L which points A1, A2,, Ak 2 lie
on it. We must have L in our k  1 lines because if we do not have it then we must
cover A1, A2 ,,Ak 2 with at least k  2 distinct lines. Now U k  U k 1  L and
according to the induction hypothesis U k cannot be covered with k lines so the
statement was proved. 

Comment. The minimum possible number of elements of S is 2n  1 . Obviously we

have S  2n  1 . Also any 2n  1 points such that no three of them are collinear is
an example.

4. Solution 1. Let Ni for 0  i  3 be the set of squares with i sides in the cycle and
let ni be the number of such squares. There are a total number of mn squares, so
 ni  mn (1)
i
Each vertex that is not on the boundary is adjacent to 2 edges and there are
(m  1)(n  1) such vertices. So
 ini  2(m  1)(n  1) (2)
i
For each 90 degree turn in the cycle, consider the square that contains its two
adjacent edges. This square has at least two adjacent edges in the cycle. If we let N 2
be the set of squares with only two opposite sides in the cycle, then each square in
N 2 \ N 2 is counted once, each square in N 3 is counted twice and no other squares is
counted. So
A  (n2  B )  2n 3
Where A is the number of 90 degree turns. By subtracting equation (1) from
equation (2) we get

29
 A  n 0  B  2(m  1)(n  1)  mn
and so
A  (m  1)(n  1)  A
 B  n 0  mn  (m  1)(n  1)
 B  C  m  n  1.


Solution 2.
Lemma 1. Let P be an arbitrary cycle in the graph with length more than 4 , such
that no vertex is inside P . Then the number of vertices of P with no 90 degree turns
(A) is 2B  2C  2 , where B is the number of squares inside P with only two
opposite sides in P and C is the number of squares inside P with no edge in P .
Proof. We use induction on k , the number of squares inside P , which is more than
one by assumption. For k  2 the assertion is trivial. Suppose k  2 . Consider a
new graph whose nodes are the centers of the squares inside P and two nodes are
adjacent if and only if their corresponding squares have a common edge (not
necessarily in P ). This graph is trivially connected and has no cycles, because if it
has a cycle, then there is a vertex inside it which is also inside P . So the graph is a
tree and has a leaf. Consequently, there is a square Q inside P with three sides in
P . Consider the induction hypothesis for the cycle obtained by deleting these three
edges and adding the remaining side of Q :
A  2B   2C   2.
Let xyzt be the path in P with the vertices of Q . Consider the following cases. The
claims can be easily proved.
 There are 90 degree turns at both x and t . In
this case we have A  A  2 . Also
B  B   1,C  C  or B  B ,C  C   1
depending on Q’s adjacent square inside P .

 There is a 90 degree turn at only one of x, t .

In this case we have A  A , B  B  and
C  C.

 There is no 90 degree turn at x and t . In this

case we have A  A  2 , B  B   1 and
C  C.

30
 The induction claim follows trivially in each case. 

Using this lemma, we get

A  2B  2C   2, (3)
where B ,C  are the number of squares inside P with the same properties as B,C .
Let R be the inner (m  2)  (n  2) rectangles and draw its boundary together with
the edges of P . These edges partition the square into some regions. It is easily seen
that the boundary of each region, other than the inside of P and the out-most
region, is a cycle containing one edge in the boundary of R and a path in P . Use
the lemma for each region to get
Ai  2Bi  2C i  2, 1  i  t, (4)
where t is the number of mentioned regions. It is easily seen that
B  B    Bi  1,
i
C  C    C i  t  4,
i
1
A  (A   Ai  A),
2 i
where A is the number of vertices without 90 degree turns that on the boundary of
R. Now the assertion follows easily by substituting equation (3) and (4) in
A  B C . 

5. We claim that for every k   and real numbers a1, a2,..., a2k :
f (a1  a2    a2k )  2k max  f (a1), f (a2 ),, f (a2k ) 
Proof is done by induction on k .Basis is obviously the condition (iii). Suppose the
claim is true for k . For k  1 we have:
f (a1  a2    a2k 1 )  f ((a1  a2    a2k )  (a2k 1    a2k 1 ))
 2 max{f (a1  a2    a2k )  f (a2k 1    a2k 1 )}
 2max  2k max{ f (a1 ), f (a2 ),, f (a2k )},2k max{f (a2k 1 ),, f (a2k 1 )} 
 2k 1 max{f (a1 ), f (a2 ),, f (a2k 1 )}
Now suppose that 2k 1  n  2k :
f (a1    an )  f (a1    an  0
 0 
   0)
2k n
k
 2 max{f (a1 ),, f (an ), f (0),, f (0)}
 2k max{f (a1 ),, f (an )}  2n max{f (a1 ),, f (an )}
2k 1  n So 2k  2n and this implies the last inequality.

31
 If we put a1  a2    an  1 then f (n )  f (1
 1    1)  2nf (1) . Therefore
n
n  n   n 
( f (a  b ))n  f ((a  b)n )  f (   a ib n i )  2(n  1) max {f ( a ib n i )}
 i 
i 0 
0i n  i 
n  n  n  
n
 2(n  1) f ( )f (a i )f (b n i )  4(n  1)f (1)   f (a )i f (b )n i
 i 
i 0   i 
i 0 
 4(n  1)f (1)( f (a )  f (b ))n  f (a  b )  n 4(n  1)f (1)( f (a )  f (b ))
If n tends to infinity, we have n 4(n  1)f (1)  1 and this implies the desired result.


6. , a , e are three equal circles and A, B , E are other intersection points of these
circles. By some angle chasing it is easy to prove that A, A, B, E form an
orthocentric system of points.(each one is the orthocenter of others.) so A is
orthocenter of triangle EAB . Similarly B ,C , D  and E are orthocenter of triangles
ABC , BCD ,CDE and DEA respectively. Note that AE and B C are both
perpendicular to AB so they are parallel. On the other hand AA  BE so
1
AAE=90  AEB  90  AB ,
2
1
Similarly B BC =90  BCA  90  AB . Therefore AAE  B BC . Since
2
b , e have equal radius, AE  B C are of equal length and parallel, consequently
quadrilateral AB CE is parallelogram. Thereby segment AB  is parallel to the
diagonal CE of pentagon and they have equal length. Similarly pairs of segments
(B C , AD ),(C D , BE ),(D E ,CA) and (E A, DB ) are parallel and have equal length.
So the sides of pentagon AB C D E  and diagonals of pentagon ABCDE are of
equal length and angle between sides equals to angle between corresponding
diagonals.
Now consider pentagon A1B1C 1D1E1 , similar to ABCDE such that A1B1  AB and
A1B1  2AB (and similar relations for other sides.). Let A2 , B2 ,C 2, D2 , E2 be
midpoints of sides C 1D1, D1E1, E1A1, A1B1 and B1C 1 respectively. By Thales’ theorem,
1
A2B2  C 1E1 and A2B2  C 1E1 . Other sides have similar situation, so pentagons
2
ABCDE and A2B2C 2D2E2 are equal. Therefore we have
S (ABCDE )  S (A2B2C 2D2E 2 ) 
 S (A1B1C 1D1E1 )  S (A2B2D1 )  S (B2C 2E1 )  S (C 2D2A1 )  S (D2E2B1 )  S (E2A2C 1 ) 
 4S (ABCDE )  S (CDE )  S (DEA)  S (EAB )  S (ABC )  S (BCD )

32
 Let S be area of pentagon ABCDE , S 1 area of pentagon A1B1C 1D1E1 and by S 
sum of area of triangles ABC , BCD,CDE , DEA, EAB so the
assertion is equivalent to
S  S   2S . ()
The right hand inequality in (*) is easy, because if we color
triangles ABC , BCD ,CDE , DEA and EAB each region in
pentagon ABCDE is colored at most two times.

We claim that in every convex pentagon the sum of area of corner triangles is more
than or equal to area of pentagon.
Consider a convex pentagon ABCDE .
Obviously, There exist two adjacent angles
with some more than 180 . Suppose that
these two angles are D and E . Now fix
points B,C , D and E in the plane and move
point A . Since ABCDE is convex, place of
point A must be in the shaded region (Look
at the figure) which is convex itself.
Let f (A)  S (ABC )  S (BCD )  S (CDE )  S (DEA)  S (BCDE ) . We must prove
f (A)  0 for every point A in mentioned convex region. f is a linear function of A
defined in a convex region, so f takes its minimum at corner points B , E or  .
f ()    0,
f (B )  0  S (BCD )  S (CDE )  S (DEB )  S (BCDE )  0,
f (E )  S (EBC )  S (BCD )  S (CDE )  0  S (BCDE )  0.
So f is nonnegative in this region and the assertion is proved. 

7. First we claim that for every positive integer k , the edges which their numbers are
divisible by k form a cluster. Indeed, suppose on the contrary that S is the largest
cluster such that the number of its edges is divisible by k and v 
 S is another such
edge. Applying problem statement on edge v and edges of cluster we get there exists
a larger cluster with number on edges divisible by k .
n 
Now suppose that p   where p is a prime number. Therefore the edges with
 2 
1 n 
numbers divisible by   form a cluster and the number of such edges is p . So
p  2 

33
 t 
p     2p  t (t  1)
 2 
 p | t  t  1 | 2  t  2 or t  3  t  3, p  3
 
 p | t  1  t | 2  t  1 or t  2  Contradiction !

n  n 
Thereby the only prime divisor of   is three, so    3a where a   . If a  1
 2   2 
and so n  3 and numbers 1,2, 3 satisfies problem statement. If a  1 there exist 9
1 n  t 
multiple of   among numbers, but 9 cannot be written in the form   so for
9  2   2 
n  3 such numbers do not exist. 

8. Let h(x )  f (x )  x , So we have

h(h(x )  g(x )  x  2y )  2h(y ). (1)
Hence For every x, y, z    we have
h(h(x )  g(x )  x  2y )  2h(y )  h(h(z )  g(z )  z  2y ). (2)
There exist some x, z    such that T  h (x )  g (x )  x  h (z )  g (z )  z is
positive. Otherwise h (x )  g (x )  x must be constant. So
h (x )  g (x )  x  h (1)  g (1)  1  h (x )  g(x )  x  C ,
where C  h (1)  g(1)  1 (constant). By definition of h (x ) we get
f (x )  g (x )  C .
But coefficients of g (x ) all are positive, therefore g (x )   as x   so
g (x )  C for large values x , Which contradicts the assumption that f (x )    for
every x    . Therefore function h(x ) is periodic because according to equation (2)
for each x, z   , T  h (x )  g (x )  x  h (z )  g (z )  z is a period for h . (We can
choose x, z    such that T  0 .) Therefore
S (x )  h(x  T )  g(x  T )  (x  T )  h (x )  g (x )  x  g (x  T )  g (x )  T .
S (x ) is a period for h for every x    . Since g (x ) is a polynomial of degree at
least two, S (x ) is not constant and is a function of x such that its image contains all
numbers greater than a fixed positive real number A and this implies that h (x ) is
constant for every x    and so h(x )  K for some constant K . Now we replace
function h by K in (1)
h (h (x )  g (x )  x  2y )  2h(y )  K  2K  K  0
 h (x )  0  x     f (x )  x x   .


34
9. Suppose 1 is tangent to AB , AD at P1,Q1 respectively and 2 is tangent to
CD, BC at P2 ,Q2 respectively.
Lemma. Let a circle  be tangent to the half-lines BA, BC at P,Q respectively.  is
tangent to 1 if and only if BP  AB  AP1 for some choice of the sign.
Proof. PP1 is the common external tangent of  and 1 . So , 1 are tangent if
and only if PP1  2 rr1 , where r, r1 are the radii of , 1 respectively. If we let
ABC  2 , then r  BP .tan  and r1  AP .cot  . So 2 rr1  2 BP .AP1 . So
, 1 are tangent if and only if AB  AP1  BP  2 BP .AP1 and the claim
follows. 

By assumption, there is a circle  tangent to the lines DA, DC at Q, P respectively

and externally tangent to 1, 2 . It is easily seen that  should be inside the angel
ADC . So, by the lemma we have:
DQ  AD  AQ1
DP  CD  CP2
for some choice of the signs. We have DP  DQ , AQ1  AP1 , CP2  CQ2 ,
AD  BC and CD  AB . So by the above equations we get
BC  AP1  AB  CQ2  BC  CQ2  AB  AP1 .

Let  be a circle tangent to the half-lines BA, BC at P ,Q  respectively, such that
both sides of this equation are equal to BP (note that the value is positive). So 
is tangent to 1, 2 by the lemma and the assertion is proved. 

10. Solution 1. By Holder Inequality we have

35
 a a a a 1
( )( bc)( 1)  ( a )3    ( a )3 .
cyc bc cyc cyc cyc cyc bc 3 cyc
So it suffices to have
1
( a )3  3( a )  ( a )2  3 3   a  4
27
3 cyc cyc cyc cyc

a a
Now suppose that  a  4
27 . It suffices to prove that  bc
 3.4 27 , but
cyc cyc
1
a a a a 
we have  bc
 33 
bc
 3(abc) 6 () by AM-GM inequality
cyc
2 2 1
 
1  ab  3(abc)3  (abc) 3  3  (abc ) 6 
4
3
cyc
() 1 5 1 3
a a 

4 4
 3(abc) 6  3 3  3 4  32.3 4  3. 27.
cyc bc 

Solution 2. First by Cauchy-Schwarz inequality we have

a2
( )( bc a )  ( a )2 .
cyc bc a cyc cyc

So it suffices to prove that

( a )2 ab  3  abc ( bc )( a ).
cyc cyc cyc cyc

Since  ab  1 .
cyc

But by AM-GM and Cauchy-Schwarz we have

1
 ab  3
 ab (Cauchy–Schwarz) (1)
cyc cyc
1
 a  3 ( a )2 (Cauchy–Schwarz) (2)
cyc cyc
1

 a  3.(abc)6 (AM  GM ) (3)

cyc
1

 a  3.(abc)3 (AM  GM ) (4)

cyc

Multiplying (1),(2),(3) and (4) finishes the proof. 

11. We can suppose the 60 angle between l and half-line CO is in the same side of
CO as B . We take all angles with a plus or minus sign according to their

36
 orientations and consider them modulo 180 . It can be seen that points X ,Y , Z ,T
are on a circle (or line) if and only if XYZ  XTZ .
Let AOC  COB   . We have
CMA  360  (90    120)  150    APC    30
CNB  (60  )  90    30  CPB    30
 APB  2  AOB.
Thus, point P is on the circumcircle of AOB
and so CP  CA  CO .
So CAP  APC    30 . Since
CAO  AOC   then OAP  30
  2OAP  60
OCP  OP
Now . So,
triangle OCP is equilateral and so OC  OP .
So,  and the circumcircles of triangles CAM
and CBN are symmetric with respect to the
perpendicular bisector of CP . By considering
their intersections, we get PQ  CA and PR  CB . So, PQ  PR and OP is the
perpendicular bisector of QR (If D is the intersection of AM and CN , the
ˆ only ensures that B is between D, N to have similar figures). 
condition on AOB

12. We divide subgroups into three groups

1) Subset  such that f ()  g ()  0 .
2) One element subsets. Of course we know that for an arbitrary element of this
group like X , f (X )  0 and g (X )  1 .
So for one element subsets we have  (f (A)  g(A))  n .
A 1

3) Subsets with at least two elements.

For subsets in group 3. For integers 1  i  j  n Let Aij be the collection of such
binary strings of length n that the first 1 lies in the i ‘s place and the last 1 lies in
the j ’s place. Hence, if i  j and X  a1a2 an  A then ai 1,, a j can be 0 or 1
so Aij  2 j i 1 . For a  {0,1} let a  1 if a  0 and a  0 if a  1 .
For every X  a1a2 an  Aij we define X  a1 aiai 1 a j 1a j an . (We change
the numbers between first and last 1.) Now we compare f (X ) and g (X ) . (There is
bijection between binary strings of length n and subsets of {1, 2,,n } so we can
define f and g on binary strings.)
For the largest block of 1’s in X we have three cases:

37
 Case 1. The largest block contains places i and j . So ai  ai 1    a j  1 ,
hence f (X )  j  i and g (X )  j  i  1 . Thus f (X )  g (X )  1 .
Case 2. The largest block contains exactly one of places i or j . So for example for
some k (i  k  j ) , ai  ai 1    ak  1 and ak 1  0 . Therefore we deduce
that g (X )  k  i  1 and f (X )  k  i  1 . Hence f (X )  g (X )  0
Case 3. The largest block contains none of i and j . So for example for some k and
l i k l  j we have ak  0, ak 1    al 1  1 and al  0 . Thus
g (X )  l  k  1 and f (X )  l  k . Hence f (X )  g (X )  1 .
Note that for each 1  i  j  n one of elements of Aij satisfies the condition of
case1 and there are 2n4  1 binary strings X  a1a2  a n such that a1  an  0
and a2  an 1  0 also they have another 1.
 n 
 ( f (X )  g(X ))  (1)    (1)(2n 4  1).
X Aij  2 
1i  j n

Our conclusions are summed up in the following result.

 n 
   n.
 ( f (X )  g (X ))  2n 4
 1   2 
 
X

But it is very easy to prove by induction that for n  10 we have

 n 
2n 4     n  1 and this implies the assertion. 
 2 

13. Lemma 1. Suppose that l1 and l2 are two lines which intersect at O and the angle
between them is  . Then the composition of reflections with respect to l1 and l2 is
rotation with center O and angle 2 . 

Lemma 2. Suppose that R1 and R2 are two rotations with centers O1 and O2 and
directed angles  and  respectively. Then if    
 2k  (k  ) the composition
of R2 and R1 is a rotation with center O and directed angle    where O is a
1 1
point such that OO1O2   and OO2O1   else if     2k  for some
2 2
integer k then this composition is a translation. 

Lemma 3. Consider a 2k -gon with center O and radius of circumcircle r . Let

A1, A2,, A2k be its vertices clockwise and Ri be the rotation with center Ai and angle

38
2 
clockwise. Then O O  2k 1 r sin( ) where O   R2k  R2k 1    R1(O ) . (Note
k
2 2k

that 2k 1 r sin( ) is the perimeter of 2k -gon.)
2k
Proof. Proof is by induction on k . For the base case k  2 it is easy to verify that

O O  4r 2  221r sin( ) . Now suppose that A1A2  A2k 1 is a 2k 1 -gon with
4
2
radius of circumcircle r and let Ri be the rotation with center Ai and angle
2k 1
clockwise ( 1  i  2k ). According to the lemma 2 composition S i  R2i  R 2i 1 is a
2 2
rotation of angle 2   clockwise and center Bi . Where Bi is a point such
k 1
2 2k
2
that Bi A2i A2i 1  Bi A2i 1A2i  . By definition of Bi we have
2k 2
 A2iO r
Bi A2iO  Bi A2i 1O   BiO   .
2 2 2
cos( k 2 ) cos( k 2 )
2 2
And also
2 2 2 2
BiOBi 1  BiOA2i  A2iOA2i 1  A2i 1OBi 1  .   
k 2 k 1 k 2
2 2 2 2k
So B1B2 B2k is a regular 2k -gon with center O and radius of circumcircle
r
. Therefore according to the induction hypothesis we have
2
cos( k 2 )
2
r  
OO   2k 1 sin( k )  2k 2 r sin( k 1 ) .
 2 2
cos( k 1 )
2
Where O   S 2k  S2k 1    S1(O ) , but by definition S i  R2i  R 2i 1 and so
O   S2k  S2k 1    S1(O )  R2k 1  R2k 1 1    R1(O )
This completes the proof of the lemma. 

We proceed the proof in two different ways.

Solution 1. Let A1, A2 ,, A2k be the vertices of 2k -gon such that li  Ai Ai 1 where
A2k 1  A1 and Li be the reflection with respect to li (1  i  2k ) . Invoking the
lemma1 the composition Ri  L2i  L2i 1 (1  i  2k 1 ) is a rotation with center A2i
2
and angle . A2A4  A2k are vertices of a 2k 1 -gon with center O . So by lemma3
k 1
2

if O   R2k 1  R2k 1 1    R1(O ) , We have O O  2k r sin( ) . Since
k 1
2

39
 Ri  L2i  L2i 1 then O   L2k  L2k 1    L1(O ) so it suffices to prove that
 
2k r sin( ) is less than the perimeter of 2k -gon which equals 2k 1 r sin( ) .
k 1
2 2k
    
cos( )  1  2 sin( )cos( )  sin( )  2 sin( )
k k k k 1
2 2 2 2 2k
 
 2k r sin( )  2k 1r sin( )
k 1
2 2k
And this finishes our proof. 
Solution 2. We prove the assertion of problem for regular polygon with even number
of sides. Denote by A1, A2 ,, A2k the vertices of this polygon clockwise and let
li  Ai Ai 1 for 1  i  2k where A2k 1  A1 .
Now suppose that Li is the reflection with respect to li . Invoking the lemma 1 we
2(k  1)
deduce Ri  L2i  L2i 1 (1  i  k ) is a rotation with center A2i and angle
k
2(k  1)
in triangular direction, but we have k   2(k  1) so the composition
k
Rk  Rk 1    R1 is a translation by lemma2. So it suffices to prove that the length
of the vector of translation is not greater than the perimeter of 2k  gon.
Let B1  R1(A1 ), B2  R2 (B1 ),, Bk  Rk (Bk 1 ) and l be the length of sides of
polygon. We must prove that Bk A1  2kl . We have
A2A1  l  A2B1  l (R1 is a rotation with center A2 )
 A4B1  A4A3  A3A2  A2B1  3l
 A4B2  3l (R2 is a rotation with center A4 )
 A6B2  A6A5  A5A4  A4B2  5l

 A2k Bk  (2k  1)l  A1Bk  A1A2k  A2k Bk  l  (2k  1)l  2kl
So it finishes the proof. 
Comment. Another proof can be given by complex numbers and think of rotations as
multiplying by a complex number.

14. First suppose the case that none of the numbers are zero. Note that there exist at
least 1000 positive numbers or at least 1000 negative numbers among these 2000
numbers. If there exist at least 1000 negative numbers and we put these 1000
numbers as roots of a degree 1000 polynomial all its coefficients are positive. So in
every case we have 1000 positive numbers.

40
 Now take 1000 positive numbers and put the 1000 remaining numbers as roots of a
polynomial, all of its coefficients are positive, so the numbers must be negative.
Therefore there is 1000 positive and 1000 negative numbers. If we put 1000
positive numbers as roots of a polynomial its coefficients are alternating positive and
negative. This is a contradiction because remaining 1000 numbers all are negative.
The contradiction shows that there at least one number equal to zero among
numbers.
Denote by k the number of zeros among numbers so k  0 . If k  1000 , then put
these k numbers zero and arbitrary 1000  k numbers among others as roots of a
polynomial. The product of roots is zero so there exists another number equal to zero
and this contradicts the definition of k , thereby k  1000 . Now put 1000 numbers
equal to zero as roots of a polynomial this polynomial is x 1000 , so other numbers are
equal to zero. So there do not exist 2000 numbers with mentioned property. 

15. We have
x 3  xy  1 y 3  xy  1 x2  1 y2  1
 x y
x2  x  y y2  x  y x2  x  y y2  x  y
2 2
x 1 y 1
 x y   . ()
2 2
x x y y  x y
Now we have three cases:
Case1.
x2  1
x y 2  0  x 2  1 or x  0  (x , y )  (0, 0),(1,1),(1, 1)
x
x 2  1 y2  1
Case2. x  y . Let k  x  y  0 so by () we have k   . If x  1
x 2  k y2  k
or y  1 by using main equation we get (x , y )  (1, 0),(1,2),(0,1),(2, 1) . Now we
can assume x , y  1
x2  1 y2  1
x2  1  x2  k  0   1 and similarly 0   1 , therefore
x2  k y2  k
x2  1 y2  1 x2  1 y2  1
0k   2k 11  . ()
x2  k y2  k x2  1 y2  1
On the other hand x  1 so x 2  3 and 2x 2  2  x 2  1 , consequently
x2  1 1 y2  1 1 x 2  1 y2  1
 , similarly  , therefore   1 . Contradicts ()
x2  1 2 y2  1 2 x 2  1 y2  1

41
 x2  1 y2  1
Case3. x  y . Let t  y  x  0 so by () we have t   and by
t  x2 t  y2
t  x2 t  y2 t 1 t 1
adding 2   to both sides we get t  2   .
2 2 2
t x t y t x t  y2
We claim that t  x 2 and t  y 2 are not positive, simultaneously. Assume by
contrary t  x 2  0, t  y 2  0 then
t  x 2, t  y 2  2(y  x )  2t  x 2  y 2  (x  1)2  (y  1)2  2 .
t 1
This contradicts because x , y  1 . Therefore at least one of fractions and
t  x2
t 1
is less than or equal to 0 and the other less than or equal to t  1 . So
t  y2
t 1 t 1
t 2  2
  0  (t  1)  t .
t x t  y2
This contradiction shows that case3 does not have a new solution and these are all
the solutions: (0, 0),(1,1),(1, 1),(1, 0),(1,2),(0,1),(2, 1). 

16. Lemma. Let p be an odd prime number and k a positive integer then we have
 0 p 1
p  1 | k
   ik (mod p).
i 0

 1 p  1 | k
Proof. If p  1 | k the statement is obvious by Fermat Little Theorem. For p  1 | k
consider g a primitive root modulo p then
p 1 p 1 p 1
g k ( p 1)  1
i k
 i k
  g ik 
k
g 1
 0 (mod p ) (p  1 | k  g k 
 1).
i0 i 1 i 1 

m
Now for every polynomial f  [x ] with f (x )   an x n , we have
n 0
p 1 p 1 m m p 1
 f (i )    ani n   an  i n   an (mod p).
i0 i 0 n  0 n 0 i 0 p 1|n,n 0

Therefore if f (0), f (1),, f (p  1) are a complete system of residues modulo p then

p 1
we have  f (i)s  0 (mod p ) for 1  s  p  2 so f (x ),( f (x ))2,, ( f (x ))p 2 are all
i 0
p 1
0-residue and  f (i )p 1  1 (mod p ) so ( f (x ))p1 is 1-residue.
i 0

Now for reverse by (1) it suffices to prove that if for p numbers a0, a1,, a p 1  
we have

42
 a 0i  a1i    a ip 1  0 ( mod p) (1  i  p  2)
a 0p 1  a1p 1    a pp11  1 ( mod p)
Then {a0, a1,, a p 1} is a complete system of residues. Now Suppose that
g (x )  (x  a 0 )(x  a1 )(x  a p 1 )  x p  b1x p 1    bp 1x  bp
and Si  a 0i  a1i    a pi 1 for i   .
If for all 0  i  p  1 we have ai 
 0 (mod p) then invoking the Fermat little
theorem we have 1  a 0p 1  a 0p 1    a 0p 1  1
 1    1  0 (mod p)
p times

Contradiction. So there exists a 0  j  p  1 such that a j  0 (mod p) and

therefore bp  0 ( mod p) . Now by Newton identities we have
 S  b  0
 1 1
 S  b S  2b  0
 2 1 1 2
 3 S  b S
1 2  b2 1  3b3  0
S

 

 S p 1  b1S p 2    b p2S1  (p  1)bp 1  0
Hence
S1  b1  0, S1  0 (mod p)  b1  0 (mod p)
S 2  b1S1  2b2  0, S1  S 2  0 (mod p)  2b2  0 (mod p)  b2  0 (mod p)

S p 2  b1S p 3    b p3S1  (p  2)bp 2  0, S1  S 2    S p2  0 ( mod p)
 (p  2)bp 2  0 (mod p)  bp2  0 (mod p)
and
S p 1  b1S p 2    bp 2S1  (p  1)bp 1  0 

S1  S2    S p 2  0, S p 1  1 ( mod p) 

 (p  1)bp 1  1 (mod p)  bp 1  1 (mod p)
So in  p [x ] , g (x ) is x p  x  x (x  1)(x  (p  1)) and  p is a field so  p [x ] is
UFD and therefore {a0, a1,, a p 1} is a complete system of residues and this finishes
our proof. 

17. We call such a broken line a good path.

Lemma. Let C be a set of n  2 points in the plane, no three of which are collinear
and let x 0 be a vertex of the convex hull of C . The number of good paths with
vertices of C starting at x 0 is at least 2n2 . Equality holds only when C is convex.

43
 Proof. We use induction on n . For n  2 it is trivial. Suppose the claim is true for
n  1 . Let x 0y be a line such that C   x 0  is entirely in one side of it. Sort the
vertices of C   x 0  as x1, x 2 ,, xn 1 such that the angles xi x 0y are increasing. So
C   x 0, x 1  is entirely in one side of x 0x1 and C   x 0, x n 1  is entirely in one side
of x 0x n 1 . There are at least 2  2n3 good paths with vertices of C   x 0  starting
at either x 1 or xn 1 . By joining the segments x 0x1 or x 0x n 1 we obtain at least 2n2
good paths with the vertices of C starting at x 0 .
If C is convex, then in any good path starting at x 0 , x 0 should be joined to x1 or
xn 1 , because in other cases the vertices will be in both sides of the first segment and
the path will intersect the first segment. So equality for convex sets follows by the
induction hypothesis.
Now, suppose C is not convex. Let z be a vertex of C not on the convex hull.
Either z is in triangle x 0x1xn 1 or is inside the convex hull of C   x 0  (depending
on which side of x1xn 1 that z is in). In the first case, if z  is the farthest vertex
from line in triangle x 0x1xn 1 (other than x 0 ), then the segment x 0z  doesn’t
intersect the convex hull of C   x 0  and there is a good path starting with x 0z  by
the induction hypothesis. In the second case, C   x 0  is not convex and the
number of good paths starting with x 0x1 is more than 2n3 by the induction
hypothesis. So the lemma is proved. 
n 3 n 3
According to the lemma, we have T (B )  n 2 . We prove T (A)  n2 . Let x 0 be
a vertex on the convex hull of A and sort the other vertices of A as described in the
lemma. For any 1  i  n  2 , by joining any two good paths starting at x 0 with
vertices of  x 0, x 1,, x i  and  x 0, x i 1,, x n 1  , we get a good path with the vertices
of A , because the two sets can be divided by a line through x 0 . This way we get
n 2

 2i 1  2n i 2  (n  2)2n 3 good vertices not starting at x 0 . There are more than

i 1

2n2 good vertices starting at x 0 and so T (A)  n 2n 3 . So, the assertion is proved.


18. Lemma 1. Triangle A B C  is similar to triangle ABC and O is its orthocenter.

Moreover, the corresponding sides make the same angle.
Proof. We have OB C   OAC   90  C and
OB A  OCA  90  A . So C B A  B . Compute the other angles
similarly. This way it is proved that the triangles are similar and O is the

44
orthocenter of triangle A B C  . Now, the altitudes of triangle A B C  make the
same angle with the corresponding sides of triangle ABC (angles OB A and so
on).So, the corresponding sides of the triangles also make the same angle. 

Lemma 2. Let B ,C  be points on AC , AB

respectively such that AB OC  is cyclic. The
circle w1 with center B  passing through C
intersects the circle 2 with center C  passing
through B at points namely P,Q such that P
is on the circumcircle of ABC and Q is on
BC . Moreover, PQ and the altitude of A
intersect on the circumcircle of ABC .

Proof. Let P be the second intersection of the

circumcircles of triangles ABC and AB C  . We claim P is on both 1, 2 . We have
1 1
PBA  PCA  PA  POA,
2 2
PC A  PB A  POA.
So, triangles C PB and B PC are isosceles and P is on both 1, 2 . Also, these
1
triangles are similar. Let  
POA .
2
Triangles PBC and PC B  are similar (Consider the angles of P and the ratio of
the sides incident with P in these triangles). The ratio of similarity is 2 cos  and
the corresponding sides make angles equal to  . So, if D, D  are the orthogonal
PD
projections of P on BC and C B  respectively, then  2 cos  and
PD 
D PD   . So triangles D PD and C PB are similar. So, D P  D D . Thus, if
PD  intersects BC at Q , then D  is the midpoint of the hypotenuse in the right
angled triangle PDQ . So, B C  is the perpendicular bisector of PQ and Q is the
second intersection of 1, 2 .
Suppose PD and AH intersect the circumcircle of triangle ABC for the second
time at E and A1 respectively. We have
1 1
QPE  D PD    PA  A E  A1PE .
2 2 1
So, PQ passes through A1 and the lemma is proved. 

According to the lemma 2, the angle between PQ and AH is equal to the angle
between B C  and BC . So, by lemma 1 the radical axis’s in the problem are

45
obtained by rotating the altitudes of triangle ABC with a fixed angle (but with
different centers). The orientations of the rotations are the same, because all the
equations in the proof remain valid by considering orientations. So, the triangle
formed by the radical axis’s is similar to triangle ABC . Therefore, it suffices to
prove that the ratios of distances of H from the radical axis’s is the same as the
ratios of distances of H from sides of triangle ABC .
Let F, A2 be the orthogonal projections of H on PQ and BC as in the lemma. We
HF HF
have 2  2 sin  which is the same as the other ratios of distances of H
HA2 HA1
from the sides of the mentioned triangles. So the assertion is proved. 

46