37th

Iranian
Mathematical
Olympiad

2019-2020
Y oung Scholars Club
Ministry of Education, I.R. Iran
www.mathysc.ir
 th
37
Iranian
Mathematical
Olympiad
 37th Iranian Mathematical Olympiad
Selected Problems with Solutions

This booklet is prepared by Sina Ghaseminejad and Siavash Rahimi.

With special thanks to Negar Babashah, Ali Mirzaei, Alireza Danaei,
Mohammad Moshtaghifar, Ali Partofard, Navid Safaei and Alireza
Dadgarnia.
Copyright Young
c Scholars Club 2019-2020. All rights reserved.
Ministry of Education, Islamic Republic of Iran.

1
 Iranian Team Members at the 61th IMO
(St. Petersburg - Russian Federation)

From left to right:

• Seyyed Reza Hosseini Dowlatabadi

• Alireza Haghi

• Matin Yadollahi
• Kian Shamsaei
• Mohammad Moshtaghifar
• Ali Mirzaei Anari

2
Contents

Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Problems
Second Round . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Third Round . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Team Selection Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Solutions
Second Round . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Third Round . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Team Selection Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3
Preface

The 37th Iranian National Mathematical Olympiad consisted of four rounds.

The First Round was held on 19th of February 2019 nationwide. The exam
consisted of 25 multiple-choice and short answer questions and a time of 3.5
hours. In total, more than 12000 students participated in the exam and 1593
of them were admitted for participation in the next round.
The Second Round was held on May 2th and 3th, 2019. In each day,
participants were given 3 problems and 4.5 hours to solve them. After this
round, the top 80 students were selected to participate in the Third Round.
The selection process of the Third Round consisted of eight separate ex-
ams. At the end of this round, 16 students were awarded a bronze medal, 32
students were awarded a silver medal, and the top 14 students were awarded
a gold medal. The following list represents the names of the gold medalists:

1. Pouya Esmaeil Akhondi

2. Shima Amirbeigi Tafti
3. Negar Babashah
4. Seyyed Reza Hosseini Dowlatabadi
5. Alireza Haghi
6. Alireza Danaei
7. Siavash Rahimi Shateranloo
8. Alireza Rezaei Moghaddam
9. Aryan Zamani
10. Kian Shamsaei
11. Mohammad Moshtaghifar
12. Ali Mirzaei Anari
13. Matin Yadollahi
14. Matin Yousefi

The Team Selection Test was held on 4 days, having the same structure
as the International Mathematical Olympiad (IMO). In the end, the top 6
participants were selected to become members of the Iranian Team at the
61th IMO.
In this booklet, we present the 6 problems of the Second Round, 24 prob-

4
lems of the Third Round, and 12 proposed problems of the Team Selection
Test, together with their solutions.
It’s a pleasure for the authors to offer their grateful appreciation to all the
people who have contributed to the conduction of the 37th Iranian Mathemat-
ical Olympiad, including the National Committee of Mathematics Olympiad,
problem proposers, problem selection groups, exam preparation groups, coor-
dinators, editors, instructors and all those who have shared their knowledge
and effort to increase the Mathematics enthusiasm in our country, and as-
sisted in various ways to the conduction of this scientific event.

The 37th Iranian Mathematical Olympiad Problem Selection Committee:

Erfan Salavati (The head of National Mathematical Olympiad Committee)
Omid Naghshineh Arjmand (Former head of National Mathematical
Olympiad Committe), Mahdi Etesamifard (The supervisor of team selection
test), Morteza Saghafian, Ali Khezeli, Ali Daei Nabi, Amin Bahjati,
Fatemeh Sajjadi, Ali Partofard, Masoud Shafaei, Yahya Motevasel, Seyed
Reza Hosseini, Mohammad Amin Sharifi, Alireza Dadgarnia and Heasam
Rajabzadeh.

5
Problems
Second Round

1. (Morteza Saghafian) We have a rectangle with its sides being a mirror.

A light ray enters from one of the corners of the rectangle and after being
reflected several times, gets to the opposite corner of its starting point. Prove
that the light ray has passed the center (Intersection of diagonals) of the
rectangle.
(→ p.18)
2. (Mahdi Etesami Fard) 4ABC is an isosceles triangle with AB = AC.
Point X is an arbitrary point on side BC. Points Y, Z are on the sides
AB, AC, respectively, such that ∠BXY = ∠ZXC. A line parallel to Y Z and
passing trough B cuts XZ at T . Prove that AT bisects ∠A.
(→ p.18)
3. (Mohsen Jamali) Let n > 2 be a natural number. Prove that the following
equation has no solution x1 , x2 , . . . , xn in positive integers numbers greater
than 1:
(x1 x2 . . . xn )2 = x31 + x32 + · · · + x3n
(→ p.19)

4. (Iman Maghsoudi) ω is a circle with diameter AB. Points C, D lie on ω

such that C, D are on different sides of AB. A line passing through C and
parallel to AD cuts AB at F , and a line passing through D and parallel to
AC cuts AB at E. A, B, C and D are in a way that E, F are inside ω. The
line passing through E and perpendicular to AB cuts BD at X and the line
passing through F and perpendicular to AB cuts BC at Y . Prove that the
perimeter of triangle 4AXY equals to 2CD.
(→ p.20)
5. (Erfan Salavati) Polynomial x1398 + 1 is written on the board. Roozbeh
and Keyvan play the following game in turns. Starting from Roozbeh, Each
player in their turn chooses an integer 0 ≤ k ≤ 1398 and adds up xk with
the polynomial on the board. Each time after Keyvan’s turn, if there exists
a real number x such that the polynomial on the board takes negative value,

7
Roozbeh wins and Keyvan loses. Otherwise, the game continues. Prove that
no matter how Roozbeh plays, Kayvan can play in a way that he never loses.
(→ p.20)

6. (Amin Behjati) Consider lattice points of a 6 × 7 grid. We start with

two points A, B. We say two points X, Y are connected if one can reflect
several times with respect to points A, B and reach from X to Y . What is
the minimum number of connected components, over all choices of A, B?
(→ p.21)

8
Third Round

Algebra
1. (Shayan Talaei) a, b and c are positive real numbers such that
X X
(a + b)2 = 2 a + 6abc.
cyc cyc

Prove that  
X  X 
2
(a − b) ≤ 2 a − 6abc .
 
 
cyc cyc

(→ p.23)
2. (Amir Hossein Zolfaghari) Find all function f : R → R such that for any
three real numbers a, b, c that satisfy a + f (b) + f (f (c)) = 0, the following
equality holds:
f (a)3 + bf (b)2 + c2 f (c) = 3abc.
(→ p.24)
3. (Navid Safaei) We are given a natural number d. Find all open intervals of
maximum length I ⊆ R such that for all real numbers a0 , a1 , . . . , a2d−1 inside
the interval I, the polynomial P (x) = x2d + a2d−1 x2d−1 + · · · + a1 x + a0 has
no real roots.
(→ p.25)
4. (Mohammad Heidary) Let A1 , A2 , . . . Ak be points on the unit circle. Prove
that X
d(Ai , Aj )2 ≤ k 2 ,
1≤i<j≤k

where d(Ai , Aj ) denotes the distance between Ai , Aj .

(→ p.27)

9
5. (Yahya Motevassel) P (x) is a non-constant monic polynomial with integer
coefficients. Assume that P1 (x), P2 (x) . . . , Pn (x) are monic polynomials with
integer coefficients such that for all 1 ≤ i ≤ n, deg(Pi ) ≥ deg(P ). We know
that for any natural number x, there exists a natural number y and an index
i (1 ≤ i ≤ n), such that P (x) = Pi (y). Prove that there exists an index j
(1 ≤ j ≤ n) and a natural number k such that P (x) = Pj (x + k).
(→ p.28)
6. (Navid Safaei) Let a, b, c be non-zero distinct real numbers so that there
exist functions f, g : R+ → R so that
 
x
af (xy) + bf = cf (x) + g(y)
y
for all positive real x and large enough y.
Prove that there exists a function h : R+ → R so that
 
x
f (xy) + f = 2f (x) + h(y)
y

for all positive real x and large enough y.

(→ p.29)

Combinatorics
1. (Amin Behjati) Hossna is playing with an m × n grid of points. She can
draw segments between some of the points with the following conditions:
a) No two segments intersect except in points of the grid.

b) Each segment is drawn between two consecutive rows.

c) There is at most one segment between any two points.
Find the maximum number of the finite regions Hossna can create.
(→ p.30)

2. (Mohammad Heidary) Let n, k be positive integers such that n ≥ k. Find

the maximum number of binary sequences of length n such that by fixing any
arbitrary k bits, the achieved sequences do not produce all binary sequences
of length k. For example if k = 1, we can only have one sequence, otherwise
they will differ in at least one bit which means that fixing that bit produces
all binary sequences of length 1.
(→ p.31)

10
3. (Mohammad Heidary) Cells of an n×n table are filled with positive integers
in a way that the number 2n + 2 − i − j is written at the intersection of the
i−th row and j−th column. In every step, we can choose two non-intersecting
n × k rectangles or two non-intersecting k × n rectangles (1 ≤ k ≤ n2 ), and
swap all the numbers inside these two rectangles with one another (without
reflection or rotation). Find the minimum number of moves one should do to
reach the position that in the intersection of the i−th row and j−th column,
the number 2n + 2 − i − j is written.
(→ p.32)

4. (Seyed Reza Hosseini) A bear is in the center of the left down corner of a
100 × 100 square table. It can move from one cell to another cell if they share
a common edge. We call a cycle in this table a bear cycle if it visits each
cell exactly once and gets back to the place it started. Removing a row or
column will divide the bear cycle into number of paths. Find the minimum
k such that for any bear cycle, there exists a row or column that we ensure
that after its removal, the maximum length among the remaining paths is at
most k.
(→ p.33)
5. (Morteza Saghafian) Let T be a triangulation of a convex 100-gon. We
construct P (T ) by copying the same 100-gon and drawing a diagonal if it
was not drawn in T , and there is a quadrilateral with this diagonal and two
other vertices so that all its sides and the other diagonal are in T . Let f (T )
be the number of intersections of diagonals in P (T ). Find the minimum and
maximum of f (T ).
(→ p.33)

6. (Seyed Reza Hosseini) Let n be a positive integer. There are 2n red lines
and n blue lines in general position given on the plane. Prove that there are at
least (n−1)(n−2)
2 regions with a monochromatic perimeter. (An infinite region
is also counted as a region with rays and segments forming its perimeter).
(→ p.34)

Geometry
1. (Alireza Dadgarnia) Given a cyclic quadrilateral ABCD. There is a point
P on side BC such that ∠P AB = ∠P DC = 90◦ . The medians of vertices A
and D in triangles 4P AB and 4P DC meet at K and the angle bisectors of
∠P AB and ∠P DC meet at L. Prove that KL ⊥ BC.
(→ p.35)
2. (Iman Maghsoudi) Consider an acute-angled triangle 4ABC with AB =
AC and ∠A > 60◦ . Let O be the circumcenter of 4ABC. Point P lies on

11
circumcircle of 4BOC such that BP k AC, and point K lies on segment AP
>
such that BK = BC. Prove that line CK bisects the arc BC of circumcircle
of 4BOC.
(→ p.36)

3. (Alireza Dadgarnia) Consider a triangle 4ABC with circumcenter O and

incenter I. The incircle touches sides BC, CA and AB at D, E and F , respec-
tively. Let K be a point such that KF is tangent to circumcircle of 4BF D
and KE is tangent to circumcircle of 4CED. Prove that BC, OI and AK
are concurrent.
(→ p.36)
4. (Mahdi Etesami Fard) Consider a triangle 4ABC with incenter I. Let D
be the intersection point of BI, AC, and let CI intersects the circumcircle of
4ABC at M . Point K lies on the line M D such that ∠KIA = 90◦ . Let F
be the reflection of B with respect to C. Prove that BIKF is cyclic.
(→ p.38)
5. (Siamak AhmadPour) In acute-angled triangle 4ABC, altitudes BE, CF
meet at H. A perpendicular line is drawn from H to EF and intersects arc
BC of the circumcircle of 4ABC (the one that doesn’t contain A) at K. If
AK, BC meet at P , prove that P K = P H.
(→ p.39)
6. (Mahdi Etesami Fard) Given an inscribed pentagon ABCDE with cir-
cumcircle Γ. Line ` passes through vertex A and is tangent to Γ. Points X, Y
lie on ` so that A lies between X and Y . Circumcircle of triangle 4XED
intersects segment AD at Q and circumcircle of triangle 4Y BC intersects
segment AC at P . Lines XE, Y B intersect at S, and lines XQ, Y P at Z.
Prove that circumcircle of triangles 4XY Z and 4BES are tangent.
(→ p.39)

Number Theory
1. (Yahya Motevassel) Let k ∈ N and {an }n≥0 and {bn }n≥0 be two sequences
of positive integers that ai , bi ∈ {1, 2, . . . , 9}. For all n ≥ 0

an . . . a1 a0 + k  bn . . . b1 b0 + k.

Prove that there is a number t with 1 ≤ t ≤ 9 and N ∈ N such that bn = tan

for all n ≥ N .
(Note that (xn xn−1 . . . x0 ) = 10n × xn + · · · + 10 × x1 + x0 )
(→ p.23)

12
2. (Mohsen Jamali) Prove that for any positive integers m > n, there are
infinitely many positive integers a, b such that set of prime divisors of am + bn
is equal to set of prime divisors of a2019 + b1398 .
(→ p.42)
3. (Mohsen Jamali, Yahya Motevassel) Let S be an infinite set of positive
integers, define:
T = {x + y|x, y ∈ S, x 6= y}.
Suppose that there are only finite primes p such that:
a) p ≡ 1 (mod 4).
b) There exists a positive integer s such that p|s, s ∈ T .
Prove that there are infinity many primes that divide at least one element of
S.
(→ p.42)
4. (Ali PartoFard, Mohsen Jamali) Find all functions f : N → N such that
for any distinct positive integers x, y, z, x + y + z is a perfect square if and
only if f (x) + f (y) + f (z) is a perfect square.
(→ p.42)
5. (Yahya Motevassel) Call a polynomial
P (x) = an xn + an−1 xn−1 + · · · + a1 x + a0
with integer coefficients primitive if and only if gcd(an , an−1 , . . . , a1 , a0 ) = 1.
a) Let P (x) be a primitive polynomial with degree less than 1398 and S
be a subset of primes greater than 1398. Prove that there is a positive
integer n so that P (n) is not divisible by any prime in S.
b) Prove that there exists a primitive polynomial P (x) with degree less
than 1398 such that for each nautral number n, P (n) is divisible by
every prime less than 1398.
(→ p.44)
6. (Yahya Motevassel) Let a, m be positive integers such that ordm (a) is odd
and for any integers x, y such that:
a) xy ≡ a (mod m).
b) ordm (x) ≤ ordm (a).
c) ordm (y) ≤ ordm (a).
We have either ordm (x) | ordm (a) or ordm (y) | ordm (a). Prove that ordm (a)
has at most one prime divisor.
(→ p.44)

13
Team Selection Test

1. (Morteza Saghafian) A weighted complete graph with distinct positive

weights is given such that every triangle is degenerate, i.e, the weight of one
of the edges is equal to the sum of the two others. Prove that one can assign
values to the vertices of this graph such that the weight of each edge is equal
to the difference between the values assigned to its endpoints.
(→ p. 45)
2. (Alireza Dadgarnia) Let O be the circumcenter of the triangle 4ABC. Let
D, E on AC, AB and P, Q, R, S be points such that P, C and R, C lie on the
different sides of AB, Q, B and S, B lie on the different sides of AC and R, S
lie on circumcircle of 4DAP , 4EAQ. We have 4BCE ∼ 4ADQ, 4CBD ∼
4AEP (In that order) and ∠ARE = ∠ASD = ∠BAC. If RS k P Q, prove
that RE, DS intersect on AO.
(→ p. 45)
3. (Mojtaba Zare Bidaki) We call an integer number n > 0 interesting if for
each permutation σ of 1, 2, . . . , n there exist polynomials P1 , P2 , . . . , Pn and
 > 0 such that:
i) P1 (0) = P2 (0) = · · · = Pn (0).
ii) P1 (x) > P2 (x) > · · · > Pn (x) for − < x < 0.

iii) Pσ(1) (x) > Pσ(2) (x) > · · · > Pσ(n) (x) for 0 < x < .
Find all interesting numbers.
(→ p. 46)
4. (Ali Zamani) Given a function g : [0, 1] → R such that for every non empty
partition of the interval [0, 1] to subsets A, B either ∃x ∈ A : g(x) ∈ B or
∃x ∈ B : g(x) ∈ A. Furthermore g(x) > x for all x ∈ [0, 1]. Prove that
g(x) = 1, for infinitely many x in its domain.
(→ p. 47)

14
5. (Mohammad Amin Sharifi) Let k ∈ Z prove that there are infinitely many
pairs of distinct positive integer numbers n, m such that

n + S(2n) = m + S(2m),
kn + S(n2 ) = km + S(m2 ),

where S(n) is the sum the digits of n to base 10. (→ p. 48)

6. (Morteza Saghafian) n positive numbers are given. Is it always possible
to find a convex polygon with n + 3 edges and a triangulation of it such that
the lengths of the diameters adopted in the triangulation are the given n
numbers?
(→ p. 48)
7. (Masud Shafaie) We call a monic polynomial P (x) ∈ Z[x] square-free mod
n if there are no polynomials Q(x), R(x) ∈ Z[x] such that Q being non-
constant and
P (x) ≡ Q(x)2 R(x) (mod n).
Given a prime p and integer m ≥ 2. Find the number of monic square-free
mod p polynomials P (x) with degree m and coefficients in {0, 1, 2, 3, . . . , p−1}.
(→ p. 50)
8. (Seyed Reza Hosseini) Alice and Bob take turns alternatively on a 2020 ×
2020 board with Alice starting the game. In each move every person colors a
cell that has not been colored yet and will be rewarded with as many points
as the colored cells in the same row and column. When the table is colored
completely, the points determine the winner. Who has a winning strategy
and what is the maximum difference he/she can guarantee?
(→ p. 50)
9. (Alireza Dadgarnia, Amir Parsa Hosseini) Given a triangle 4ABC with
circumcircle Γ. Points E and F are the feet of angle bisectors of B and C, let I
be incenter and K be the intersection point of AI and EF . Suppose that T is
>
the midpoint of arc BAC. Circle Γ intersects the A-median and circumcircle
of 4AEF for the second time at X and S. Let S 0 be the reflection of S
with respect to AI and J be the second intersection point of circumcircle of
4AS 0 K and AX. Prove that quadrilateral T JIX is cyclic.
(→ p. 50)
10. (Alireza Dadgarnia) Let 4ABC be an isosceles triangle (AB = AC) with
incenter I. Circle ω passes through C and I and is tangent to AI. The circle
ω intersects AC and circumcircle of 4ABC at Q and D, respectively. Let M
be the midpoint of AB and N be the midpoint of CQ. Prove that AD, M N
and BC are concurrent.
(→ p. 52)

15
11. (Mohammad Amin Sharifi) For every positive integer k > 1 prove that
there exists a real number x such that for every positive integer n < 1398:

{xn } < xn−1 ⇐⇒ k | n.



(→ p. 53)
p−1
12. (Ali Partofard) Let p be an odd prime number. Find all 2 -tuples
  p−1
x1 , x2 , . . . , x p−1 ∈ Zp 2 such that
2

p−1 p−1 p−1

2 2 2
X X X p−1
xi ≡ x2i ≡ ··· ≡ xi 2 (mod p).
i=1 i=1 i=1

(→ p. 54)

16
Solutions
Second Round

1. First note that if the line l has slope a, then the reflection of l with respect
to any line which is parallel to one of the axes has the slope −a.
now assume that at the start the ray has slope a. then the ray always has
the the slope ±a. now at the starting point a line with slope −a lies outside
the rectangle so by symmetry at the opposite corner the line with slope −a
lies outside the rectangle. hence the ray reach the opposite corner with slope
a.
now assume that a light ray enters the rectangle from opposite corner, By
symmetry this two rays meet at the time 2t in the center of the rectangle. 

2. Let us denote by K the intersection point of lines BT, XY . note that

4XY B ∼ 4XZC. Using the fact that BT k Y Z, we get

XK XT
= .
YK ZT
Therefore, K and T are corresponding points in triangles 4XY B and 4XZC.
Which gives us
∠T CX = ∠KBX = ∠T BX,
and T B = T C. So, T lies on the perpendicular bisector of BC and since
AB = AC, T also lies on the angle bisector of ∠A.

18
 A

Y
T
K

B X C

3. Assume that (x1 , x2 , . . . , xn ) is a solution to the equation. Without loss of

generality, let xn ≥ xn−1 ≥ · · · ≥ x1 . Since (x1 x2 . . . xn )2 is divisible by x2n ,
we have

x2n  x31 + x32 + · · · + x3n−1 =⇒ x2n ≤ x31 + x32 + · · · + x3n−1 .



On the other hand,

(x1 x2 . . . xn )2 = x31 + x32 + · · · + x3n

So,
(x1 x2 . . . xn−1 )2
nx3n ≥ (x1 x2 . . . xn )2 =⇒ xn ≥ .
n
Combining these two facts, we get

(x1 x2 . . . xn−1 )4
x31 + x32 + · · · + x3n−1 ≥
n2
4
(x x
1 2 . . . x n−1 )
=⇒ (n − 1)x3n−1 ≥
n2
=⇒ n2 (n − 1) ≥ x41 x42 . . . x4n−2 xn−1 .

Since for every index i, we have xi ≥ 2, We’ll get

n2 (n − 1) ≥ 24n−7 .

Which is obviously wrong for n ≥ 2 and we’re done.



19
4. Let K be the intersection point of CD, AX and L be the intersection point
of CD, AY .

L
E
A B
F

Then, since DAEX is a cyclic quadrilateral, we have

∠KDX = ∠BAC = ∠AED = ∠AXD =⇒ DK = KX = AK.

Similarly, AL = LY = LC. By Thales’s theorem, we have XY = 2KL.

Therefore

AX + XY + AY = AK + KL + LA = KD + KL + LC = CD

Which completes the proof.



5.
Lemma. If 1 ≤ k1 < k2 < · · · < kt ≤ n − 1 be natural numbers, then
t
X
∀x ∈ R : x2n + x2ki +1 + x2ki + 1 > 0.


i=1

Proof. If x > 0, it’s trivial. So let’s assume that x < 0. If −1 ≤ x < 0, we

write
Xt
x2n + x2ki +1 + x2ki + 1


i=1
as
t
X
x2n + 1 + (x + 1) x2ki .
i=1

20
 Pt
Since x + 1 ≥ 0 and i=1 x2ki ≥ 0, we have
t
X
x2n + 1 + (x + 1) x2ki > 0.
i=1


If x < −1, then x + 1 < 0 and x2k+1 < x < 1. So, x2n+1 x2k+1 + 1 > 0.
Therfore, we can re-write
t
X
x1398 + x2ki +1 + x2ki + 1


i=1

as
  t−1
X  
x1398−(2kt +1) + 1 x2kt +1 + x2k1 +1 x2ki+1 −(2ki +1) + 22k1 + 1
i=1

1396
where 1 ≤ k1 ≤ k2 ≤ · · · ≤ 2 . So, it is positive. 

Back to the problem. When Roozbeh chooses xk , if k is an even number,

Keyvan should choose xk as well. But, if k is an odd number, and if coefficient
of xk is odd, Keyvan should choose xk−1 and if coefficient of xk is even,
Keyvan should play xk+1 . Obviously after Keyvan’s move, the polynomial
will be in the following form:
m n l
X
2 X X
x1398 xsi + xsi −1 x2ti + x2ki +1 + x2ki + 1.


+2
i=1 i=1 i=1

Where
1398
1 ≤ s1 ≤ s2 ≤ · · · ≤ sm ≤ ,
2
1398
0 ≤ t1 ≤ t2 ≤ · · · ≤ tn ≤ ,
2
1 ≤ k1 < k2 < · · · < kl ≤ 1398.
According to the lemma we proved, the polynomial should be always positive.


6. We claim the answer is 8. Let us first find the points in a connected

component. Let `P the line passing through a point P and parallel to AB.
Let `0P be the reflection of `P with respect to AB. First note that the reflection
of any point P with respect to each of A and B lies on `0P . Hence, any point
connected to P lie on `P or `0P .
We claim if P is connected to a point Q, then Q can be attained by
−−→
some number of transformation by ±2AB and at most one reflection with

21
respect to A:the combination of any two reflections with respect to A, B is a
transformation.
−−→
Transforming by ±2AB does not change the parity of coordinates, so if we
−−→
define an equivalence relation P ≡ Q if and only if P = Q + 2nAB, n ∈ Z,
points will form at least 14 classes. every connected component has points
from at most two classes. note that A and B are not connected and their
components consists of exactly one class so we have at least 8 components.
As an example, which is easy to verify, take the two points having only
half a unit distance form the center of the table as A, B. 

22
Third Round

Algebra
a2 +
P P P
1. We know that cyc cyc ab = cyc a + 3abc, so
!2 !2
X X X
2
a + ab = a + 3abc
cyc cyc cyc
!2 !2 ! !
X X X X
2 2
=⇒ a + ab +2 a ab
cyc cyc cyc cyc
!2
X X
= a + 9a2 b2 c2 + 6abc a
cyc cyc
!2 !2 !2
X X X
2
=⇒ a + ab − a − 9a2 b2 c2
cyc cyc cyc
! !
X X X
2
= 6abc a−2 a ab .
cyc cyc cyc

By AM-GM we have
! ! !2
X X X X
2
a ab ≥ ab ≥ 3abc a
cyc cyc cyc cyc

So
! ! ! !
X X X X X X
2 2
2 a ab − 6abc a ≥ 6abc a−2 a ab
cyc cyc cyc cyc cyc cyc

23
Then using above facts
! !
X X X
2
2 a ab − 6abc a
cyc cyc cyc
!2 !2 !2
X X X
≥ a2 + ab − a − 9a2 b2 c2
cyc cyc cyc
!2
X X
⇐⇒ a + 9a2 b2 c2 − 6abc a
cyc cyc
!2 ! ! !2
X X X X
2 2
≥ a −2 a ab + ab
cyc cyc cyc cyc
!2 !2
X X X
⇐⇒ a − 3abc ≥ a2 − ab .
cyc cyc cyc

a2 ≥ ab, we have cyc a2 − cyc ab ≥ 0. Therefore

P P P P
Since cyc cyc
 
X  X X
a − 3abc ≥ a2 − ab
 

 cyc  cyc cyc
 
 X  X
⇐⇒ 2 a − 6abc ≥ (a − b)2 .
 
 cyc  cyc

2. The answers are f (x) = x, f (x) = −x and f (x) = 0.

First let’s prove that f (x) is injective at point 0. Assume there exist two
distinct real numbers t1 and t2 such that f (t1 ) = f (t2 ) = 0. Comparing
P (−f (b) − f (0), b, t1 ) and P (−f (b) − f (0), b, t2 ) gives us

(f (b) + f (0))bt1 = (f (b) + f (0))bt2 .

So if f (x) is not injective at point 0, we have

⇒ ∀b 6= 0 : f (b) = −f (0).

Which gives us f (x) = 0 as an answer.

So if f (x) is a non-constant function, then it must be injective at point 0.
Now, P (−f (0) − f (f (0)), 0, 0) gives us f (−f (0) − f (f (0))) = 0 and we have
a real number t such that f (t) = 0. P (−f (f (0)), t, 0) gives us

f (−f (f (0))) = f (−f (0) − f (f (0))) = 0.

24
And according to the injectivity at point 0, it follows that f (0) = 0.
Now, P (−f (f (b)), b, 0) and P (−f (f (c)), 0, c) give us

f (b)f (f (b))2 = b2 f (b).

If b 6= 0 then f (b) 6= 0 therefore

f (f (b)) = ±b ∀b 6= 0.

And since f (f (0)) = f (0) = 0, we have

f (f (b)) = ±b ∀b ∈ R.

If there exists a real number c 6= 0 such that f (f (c)) = −c, P (c, 0, c) gives us

f (c)(f (c)2 + c2 ) = 0

which is a Contradiction.
So
∀c ∈ R : f (f (c)) = c
now P (−a, 0, a) gives us f (−a)2 +a2 f (a) = 0 and P (a, 0, −a) gives us f (a)3 +
a2 f (−a) = 0 and they lead to f (a) = ±a.
If there exist non-zero real numbers b, c such that f (b) = b and f (c) = −c,
P (−b − c, b, c) leads to contradiction. So f (x) = x and f (x) = −x are the
only solutions. 

3. The answer is I = 1, 1 + d1 .

Assume that the desired interval is of the form of (b, c). For some q, r in (b, c),
put ai = q for odd and ai = r for even i, where 0 ≤ i ≤ 2d − 1. Then

P (−1) = 1 − dq + dr.

Since P (x) has no real root, we must have P (−1) > 0. Thus, q − r < d1 .
We
can assume that c − b = d1 . Therefore, consider the interval as b, b + d1 . It
is easy to find that b > 0, since by choosing a0 = b +  should be positive.
Now, assign the following numbers to the coefficients of the polynomial.
1
a2d−1 = a2d−3 = · · · = a1 = b + + , a2d = a2d−2 = · · · = a0 = b + ,
d
for some sufficiently small  > 0.
It is clear that for all positive real x, P (x) > 0. For all negative real x, putting
x = −t, where t > 0, then,
   
1 2d−1 1
P (−t) = t2d − b + t + bt2d−2 − · · · − b + t + b + Q(t),
d d

25
for some polynomial Q(t), deg Q(t) = 2d − 1. As  tends to zero, it remains
to find all b > 0 such that
   
1 2d−1 1
R(t) = t2d − b + t + bt2d−2 − · · · − b + t + b ≥ 0.
d d

Note that R(1) = 0. Therefore, R0 (1) should be zero. That is,

   
1 1
2d − (2d − 1) b + + (2d − 2)b − · · · − b + = 0.
d d

Hence, we claim the interval I = 1, 1 + d1 works! It is obvious that P (x)

has no positive real roots. Now, we prove that P (x) > 0 for all negative real
x. Put x = −t, t > 0.
Then,
   
1 2d−1 1
P (−t) > t2d − 1 + t + t2d−2 − · · · − 1 + t + 1.
d d

Thus in remains to prove that

t2d + t2d−2 + · · · + t2 + 1 t2d−1 + t2d−3 + · · · + t3 + t

≥ .
d+1 d

Now, since t2d + 1 ≥ t2d−2k+1 + t2k−1 and t2k + t2k−2 ≥ 2t2k−1 , we are done.
Thus,

1 dt2d − t2d−1 + t2d−3 + · · · + t
b≤ lim .
d t→1+ t2d−1 − t2d−2 + · · · + t − 1
and

1 dt2d − t2d−1 + t2d−3 + · · · + t
b≥ lim ,
d t→1− t2d−1 − t2d−2 + · · · + t − 1
since the function

dt2d − t2d−1 + t2d−3 + · · · + t
t2d−1 − t2d−2 + · · · + t − 1
has a limit as t approaches 1. Thus, we find that



dt2d − t2d−1 + t2d−3 + · · · + t dt2d − t2d−1 + t2d−3 + · · · + t
lim = lim .
t→1+ t2d−1 − t2d−2 + · · · + t − 1 t→1− t2d−1 − t2d−2 + · · · + t − 1
Hence,

1 dt2d − t2d−1 + t2d−3 + · · · + t
b = lim .
d t→1 t2d−1 − t2d−2 + · · · + t − 1

26
Finally note that


dt2d − t2d−1 + t2d−3 + · · · + t
t2d−1 − t2d−2 + · · · + t − 1




t − t2d−1 + t2d − t2d−3 + · · · + t2d − t
2d
=
(t − 1) (t2d−2 + t2d−4 + · · · + 1)




(t − 1) t2d−1 + t2d−3 t2 + t + 1 + · · · + t t2d−2 + · · · + 1
=
(t − 1) (t2d−2 + t2d−4 + · · · + 1)

Thus,


dt2d − t2d−1 + t2d−3 + · · · + t
lim
t→1 t2d−1 − t2d−2 + · · · + t − 1

t2d−1 + t2d−3 t2 + t + 1 + · · · + t t2d−2 + · · · + 1 d2
= lim 2d−2 2d−4
= = 1.
t→1 t +t + ··· + 1 d
1
That is, b = d and d = 1. 

4. Assume that the circle mentioned in the problem is the unit circle on the
complex plane. Then we can say every vertex Ai is equivalent to a complex
number zi such that |zi | = 1. On the other hand we have d (Ai , Aj ) =
|zi − zj |. So we should prove that
X 2
|zi − zj | ≤ k 2 .
1≤i<j≤k

We know that
2 2 2
|zi − zj | = (zi − zj ) (zi − zj ) = |zi | + |zj | − (zi zj + zj zi ) = 2 − zi zj − zj zi .

So we have
 
X k X
|zi − zj |2 = 2 − (zi zj + zj zi ).
2
1≤i<j≤k 1≤i<j≤k

and
X X X
(zi zj + zj zi ) = zi zj = zi (z1 + . . . + zn − zi )
1≤i<j≤k 1≤i6=j≤k 1≤i≤k

 X 2
  
 
X X  
= zi   zi  − k =  zi  − k.
1≤i≤k 1≤i≤k 1≤i≤k 

27
Therefore,
 2 
 X 2
 
X  X 
|zi − zj |2 = k 2 − k −  zi  − k  = k 2 − 
   
zi  ≤ k.
 
1≤i<j≤k 1≤i≤k  1≤i≤k 
P
Equality holds whenever 1≤i≤k zi = 0, in other words, when the circum-
center and the centroid of A1 A2 . . . Ak coincide. 

5. We firstly prove that there exists an index i such that deg Pi (x) = deg P (x)
and P (x) = Pi (y) for infinitely many x, y ∈ N.
Assume to the contrary. Let the degree of Pt+1 (x), . . . , Pn (x) be equal to
deg P (x) and deg Pi (x) > deg P (x)∀i ∈ {1, 2, . . . , t}.
Consider a sufficiently large N0 such that:

1) P (x) is increasing for all real numbers x > N0 .

2) P (x) 6= Pi (y) for all positive integers t + 1 ≤ i ≤ n and x, y.
3) Pi (x) > P (kx) for all positive integers 1 ≤ i ≤ t, all real numbers
x > N0 , and a fixed number k > t.

4) Pi (x) is increasing for all positive integers 1 ≤ i ≤ t and every x > N0 .

Then, let’s assume that

P (N0 + 1) < · · · < P (kN0 ).

We know every one of them is in the form of Pi (y). Where 1 ≤ i ≤ t and y

is a natural number.
If we have P (N0 + i) = Pj (xi ) for some j, then we have xi ≤ N0 . So, we
at most have N0 numbers in form of Pj (y) between P (N0 + 1), . . . , P (kN0 ).
Therefore, we at most have tN0 numbers between them. And since k > t, it
gives us a contradiction.
So we must have an index i such that the equation P (x) = Pi (y) has infinitely
many solutions and deg Pi (x) = deg P (x). We want to prove that if P (x) =
Pi (y), then |x − y| has a fixed upper bound.
Without loss of generality, assume that the leading coefficient of P (x) − Pi (x)
is positive. Then, assume a sufficiently large N1 such that P (x), Pi (x) and
P (x) − Pi (x) are increasing for all real numbers x > N1 . Then, if x, y > N1 ,
P (x) = Pi (y), we have x ≥ y since P (x) ≥ Pi (x) ≥ Pi (y). If we take a
sufficiently large x we can find out that y becomes large as well. Then if

28
 P (x) = xd + ad−1 xd−1 + · · · + a0 d ≥ 2,

Pi (x) = xd + bd−1 xd−1 + · · · + b0 ,

then,

xd − y d = bd−1 y d−1 + · · · + b0 − ad−1 xd−1 + · · · + a0

d−1
X
=⇒ xd − y d  ≤ c |x|i + |y|i ≤ cd xd−1 − y d−1
 

i=1
 d 
x − y d  xd − y d
=⇒ = ≤ cd.
xd−1 + y d−1 xd−1 + y d−1
And we have
xd − y d
x−y ≤ ≤ cd.
xd−1 + y d−1
So, x − y ≤ cd. Therefore, there exists a number k such that x − y = k for
infinitely many times.
Then we get P (x) = Pi (x − k) for infinitely many x and we’re done.
The case that ≥ P (x) = 1 is trivial and if P (x) − Pi (x) = c, then we’ll have
P (x) = Pi (y) and Pi (y) − Pi (x) = c. Consider sufficiently large x, y such that
Pi (x) is increasing. If x 6= y,

Pi (y) ≥ Pi (x + 1) =⇒ Pi (x + 1) − Pi (x) ≤ c

for infinitely many x. So, Pi (x) should be linear and we’re done.


6. Putting xy, xy instead of x, we find that

af xy 2 + bf (y) = cf (xy) + g(y)

and,    
x x
af (x) + bf = cf + g(y).
y2 y
Multiplying the first equation by a and the second one by b, then, adding and
using the original equation, we shall find that
 
x
a2 f xy 2 + b2 f = (c2 − 2ab)f (x) + (a + b + c)g(y).


y2

29
Further, putting y 2 , instead of y, in the original equation, one can find that
 
2 x
= cf (x) + g y 2



af xy + bf
y2
for all sufficiently large y. Multiplying both sides by b, then subtracting,
yielding to

a2 − ab f xy 2 = c2 − 2ab − ac f (x) + T (y)

For some function T (y). Putting y 2 , instead of y, we find that

a2 − ab f xy 4 = c2 − 2ab − ac f (x) + T y 2 .

Thus,
a2 − ab f xy 4 − f xy 2 = T y 2 − T (y).

Hence, if a 6= b, we find that

4 T y 2 − T (y)
2
= S y2




f xy − f xy = 2
a − ab
 4
for some function S(x). That is, f xy 4 − f xy 2 = S xy




xy 2 = S y2 .
Therefore,  
 
2

 x 
f xy − f (x) − f (x) − f  = R(y).

| {z }  y2 
A(y) | {z }
B(y)

That is,  
2

x
f xy +f = 2f (x) + R(y).
y2
√
Putting y instead of y we will eventually arrive at the desired conclusion.
The remaining cases for c = 2a and c = −a are the same. 

Combinatorics
1. We claim the answer is mn − n.
First note that by Pick’s theorem, the area of a region is at least 2b − 1, where
b is the number of the lattice points on its boundary. Obviously, for any
region, b ≥ 3 and if b = 3, the region must be a triangle. But each of the
edges of this triangle lie between two consecutive rows of the grid, which is
impossible. So by Pick’s theorem, the area of a region is at least 42 − 1 = 1.
Consider the 2n small segments of length 1 in the top and the bottom edges
of the grid. We will prove that for each of these segments, we can assign a

30
triangle inside the grid with the area of 12 , counted in the infinite region, such
that this segment is an edge of the triangle, and we can choose these triangles
in a way that none of them intersect.
Label the points of the first row as a1 , a2 , . . . an , and the points of the second
row as b1 , b2 , . . . , bn . For all 1 ≤ i ≤ n, let ti be the greatest index such that
ai and bti are connected by a segment (if it exists). If ai isn’t connected to
any of the bj s and i > 1, let ti = ti−1 , and if a1 isn’t connected to any of the
bj s, let t1 = 1. Clearly, t1 ≤ t2 ≤ · · · ≤ tn , which means for all i 6= j, ai bti
and aj btj do not intersect. Also the segment ai bti doesn’t intersect with any
of the other drawn segments. Now assign triangle 4ai−1 ai bti to the segment
ai−1 ai . The area of this triangle is 21 and it is counted in the infinite region.
Also none of such triangles intersect. Similarly, we can assign such triangles
to the n segments of the last row of the grid.
So the area of the grid counted in the infinite region is at least 2n × 21 = n.
This means that the sum of the areas of finite regions is at most mn − n.
Because the area of each region is at least 1, there can be at most mn − n
finite regions.
For the equality, consider the grid below. Obviously, the area of the grid
counted in the infinite region is n and the area of each finite region is 1. So
there are mn − n finite regions and we’re done. 

2. Let the answer be f (n, k). We will prove by induction on n + k that

k−1
X n
f (n, k) = . Case n = k = 1 is obvious.
i=0
i
Assume that S is the maximum set of the desired sequences and let T be the
set of all binary sequences of length n − 1. Define the sets A, B, C as follows:
 
A = a ∈ T there exists b ∈ S with b = 0a, but there is no c ∈ S with c = 1a
 
B = a ∈ T there exists b ∈ S with b = 1a, but there is no c ∈ S with c = 0a
 
C = a ∈ T there exists b, c ∈ S with b = 0a, c = 1a .

Clearly, f (n, k) = |A| + |B| + 2|C|. Because we can’t choose k bits from
the last n − 1 bits of the sequences that produce all the binary sequences of
length k, we have |A| + |B| + |C| ≤ f (n − 1, k). Now consider the set C. If
|C| ≥ f (n − 1, k − 1), we can fix k − 1 bits from the last n − 1 bits to produce
all sequences of length k − 1. So because of the definition of C, by fixing these
k − 1 bits and the first bit, all sequences of length k are produced, which is a
contradiction. So we have

f (n, k) = |A| + |B| + 2|C| ≤ f (n − 1, k) + f (n − 1, k − 1).

31
By induction and using Pascal’s identity we get
k−1
X  k−2
X n − 1  k−1
X n − 1 k−2 X n − 1
n−1
f (n, k) ≤ + =1+ +
i=0
i i=0
i i=1
i i=0
i
k−1
X n − 1 n − 1
=1+ +
i=1
i i−1
  k−1X n k−1 X n
n
= + = .
0 i=1
i i=0
i

For the equality example, simply choose all binary sequences with at most
k − 1 zeros. This way, by fixing any k bits, the sequence with k zeros will
never be produced and we’re done. 

3. We claim the answer is 2b n2 c.

First we give the algorithm. At first swap the 1st row and nth row, 2nd row
and (n − 1)th row and so on. Then swap the 1st column and nth column,
2nd column and (n − 1)th column and so on. This way, the number written
in the intersection i−th row and j−th column is (n + 1) − i + (n + 1) − j.
Now we will prove that we should swap at least b n2 c vertical (n×k) rectangles
and b n2 c horizontal (k × n) rectangles.
Consider the main diagonal of the table. The numbers written on this diag-
onal are 2, 4, . . . , 2n and by the end they should be in the reverse order.
By swapping horizontal rectangles, the column of a number doesn’t change.
Consider all the swaps of vertical rectangles. So we have a sequence 2, 4, . . . , 2n
and in each step we swap two non-intersecting equal blocks of consecutive
numbers of this sequence and we should produce the reverse order in the end.
Call a pair of consecutive numbers (a, b) ”good” if a < b, and call it ”bad”
otherwise. We will show that after each step, at most 2 good pairs can turn
to bad pairs. Let the sequence be

a1 , a2 , . . . , ai−1 , ai , . . . , ai+k , . . . , aj−1 aj , . . . , aj+k , . . . , an

| {z } | {z }
A B

and assume that we will swap the blocks A and B in this step. After swapping
these blocks, only four pairs may have been changed from good to bad. We
show it’s impossible that both of the pairs (ai−1 , ai ) and (aj−1 , aj ) change
from good to bad. Assume to contrary that both (ai−1 , ai ) and (aj−1 , aj )
were good but the pairs (aj−1 , ai ) and (ai−1 , aj ) are bad.
So we have ai−1 < ai < aj−1 and aj−1 < aj < ai−1 , which is a contra-
diction. Similarly, at most one of the pairs (ai+k , ai+k+1 ) and (aj+k , aj+k+1 )
can change from good to bad. This means that at each step at most two of
the good pairs can change from good to bad. In the beginning, the number

32
of good pairs was n − 1 and in the final sequence there is no good pair. So
we need at least b n2 c steps, which means we should swap at least b n2 c vertical
rectangles. Similarly, we should swap at least b n2 c horizontal rectangles and
we are done. 

4. The answer is k = 4999.

For any bear cycle if we remove the 50th row, there are at most 5000 cells in
each side of this row. Because a path cannot include two cells from different
sides of this row, the length of any remaining path will be at most 4999.
Now consider the cycle shown in the table below.
100

50

50

If we remove any column of the table, all the cells in each side of that column
form a path. Obviously, there are at least 5000 cells in one of these sides. So
the length of the path is at least 4999.
Suppose that we remove a row. Assume that the row is in the lower half of
the table. Clearly, there exists a path including all the cells in the upper half
of the table. So the length of this path is 4999 and we’re done. 

5. Call two triangles in T adjacent if they share an edge. First note that for
any diagonal d drawn in P (T ), there are two adjacent triangles of T such that
d is drawn in P (T ), because of the quadrilateral formed by these triangles.
We call them d’s triangles.
Assume that two diagonals d1 , d2 intersect in P (T ). Because the diagonals of
the triangulation do not intersect, we can easily find three triangles t1 , t2 , t3
of T such that t1 , t2 are d1 ’s triangles and t2 , t3 are d2 ’s triangles.
On the other hand, for any three triangles t1 , t2 , t3 of T that t2 is adjacent to

33
t1 , t3 , we can find a pair of intersecting diagonals in P (T ).
Now construct a graph G of 98 vertices with each vertex corresponding to one
of the triangles in T and an edge connecting two vertices if and only if the
corresponding triangles are adjacent. Clearly, there are 97 edges in G (equal
to the number of diagonals in T ).
So if we let d1 , d2 , . . . , d98 be the degrees of the vertices of G (where 1 ≤ di ≤ 3
for all 1 ≤ i ≤ 98), then the number of pairs of edges in the graph sharing a
vertex is:
98  
X di
.
i=1
2
Let r, s, t be the number of vertices
P98with degree 1, 2, 3, respectively. Obviously
r + s + t = 98. Also we have i=1 di = 2 × 97 = r + 2s + 3t. Note that
P98 di

i=1 2 = s + 3t. So we can conclude that

98  
X di
f (T ) = = r + 4 × 97 − 3 × 98 = r + 94.
i=1
2

Any vertex of degree 1 in G, corresponds to a triangle of the triangulation

that shares 2 edges with the 100-gon. So r ≤ 50. On the other hand, we can
easily prove by induction that 2 ≤ r. So 96 ≤ r + 94 ≤ 144.
As for the construction, for the minimum, simply draw all diagonals going
from one vertex. For the maximum, simply cut off 50 triangles sharing 2
edges with the 100-gon and then triangulate the resulting 50-gon arbitrarily.


6. First note that there are totally 3n+1

2 + 1 regions formed by these lines.
So we have to prove that there are at most 3n+1 + 1 − (n−1)(n−2)


2 2 = 4n2 + 3n
bi-colored regions.
Consider an arbitrary point X on the plane as ”the infinity point”, such that
X is neither inside of any of the finite regions nor any of the 3n lines. On
each of the 3n lines, choose two points as the line’s endpoints such that all
the intersection points of a line, lie between its two endpoints. Now connect
X to all of the endpoints with curved segments in a way that these segments
don’t intersect at any point except X. Color a segment as the color of its
corresponding line.
Consider a table in which each of the rows corresponds a bi-colored region, one
column corresponds the point X, and each of the other columns corresponds
a point which is the intersection of a red and a blue line. Write 1 in the
intersection of a row and a column, if the column’s point lies on the region
corresponding the row, and write 0 otherwise.
Now we count the sum of the numbers in two ways. Obviously X lies on
6n segments. So it is on at most 6n bi-colored regions. Note that these

34
regions are the infinite regions we initially had. An intersection of a red and
a blue line is exactly on 4 bi-colored regions. Note that there are exactly
2n × n = 2n2 such points. So the sum of the numbers written is at most

4 × 2n2 + 6n = 8n2 + 6n.

On the other hand, it’s easy to find that the sum of the numbers in a row
corresponding to a bi-colored (finite or infinite) region is at least 2. So if there
are k bi-colored regions, we have

8n2 + 6n ≥ 2k ⇒ 4n2 + 3n ≥ k

as desired. 

Geometry
1. Extend AP and DP until they meet DC and AB at X and Y , respectively.

L
X

B
K

P
A Y

C
D

Since ∠P AB = ∠P DC = 90◦ , DAY X is a cyclic quadrilateral. Further,

DAY L and DXLA are both cyclic since

∠XAL = ∠XDL = ∠Y AL = ∠Y DL = 45◦ .

>
Hence, DAY LX is cyclic and L is the midpoint of the small arc XY . More-
over, XY is antiparallel to DA which is parallel to BC. Thus, XY is parallel
to BC. It follows that K is the midpoint of XY . However, KL ⊥ XY =⇒
KL ⊥ BC. 

35
2. Let D be the second intersection point of circumcircle of 4BOC and line
AC and K 0 be the intersection point of lines AP and CM , where M is the
>
midpoint of arc BC.
A

O
D

B C

P
M

K0 ≡ K

We have
1
∠BCK 0 = ∠BOC = ∠A = ∠DAB,
2
∠CDB = 2∠A ⇒ ∠ABD = ∠A ⇒ AD = BD.

So we just need to prove that 4BCK 0 ∼ 4DAB to show that K ≡ K 0 . Since

∠ADP = 180◦ − ∠BP D = 180◦ − ∠C,

1
∠ACM = ∠C + ∠BOC = ∠C + ∠A,
2
we get that ∠ADP = ∠ACM . So, P D k CK 0 . Therefore CK
PD AD AD
0 = AC = AB
BC AD
which is equivalent to CK 0 = AB since P B k AC and P D = BC. This
completes the proof. 

3. Without loss of generality, assume that AC > AB. Let OI meet BC at

X and AK meet BC at Y .

36
 A

F K E

I
O

B D X≡Y C

Note that
XB sin ∠XOB sin ∠IOB
= = .
XC sin ∠XOC sin ∠IOC
Which is equivalent to
 
|∠B−∠A|
∠C sin


sin 2 2
 · ∠C

.
|∠B−∠A| sin
sin 2 2

Using Ceva’s theorem in triangle 4BOC for point I. Now, by the Ratio
lemma, we have
YB c sin ∠Y AB c sin ∠KAF
= · = · .
YC b sin ∠Y AC b sin ∠KAE
By Ceva’s theorem in 4AEF for point K, we have

sin 90◦ − ∠B ∠B



sin ∠KAF sin ∠KEF sin ∠KF A 2 sin 2
= · = ∠C

· ∠C

.
sin ∠KAE sin ∠KEA sin ∠KF E sin 90◦ − 2 sin 2

So we have just to prove that

c sin 90◦ − ∠B ∠C



2 sin 2
· ∠C

= ∠B


b sin 90◦ − 2 sin 2
∠B ∠C



sin ∠C cos 2 sin 2
⇐⇒ · ∠C

= ∠B

.
sin ∠B cos 2 sin 2

x x



Which is obviously true since sin x = 2 sin 2 cos 2 . 

37
4. Let E be the intersection point of line M K and the circumcircle of 4ABC,
then, since ∠BEM = ∠C 2 and ∠DIK = 90 − ∠AID so ∠BEM = ∠DIK, it
follows that the points B, I, K, E lie on a circle. Suppose BC intersects this
circle at point F 0 , other than B. Let N be the intersection point of line BD
and the circumcircle of 4ABC and G be where lines CM and BK meet.
A

N
M
D
E K
I
G

B C F0 ≡ F

IK ID
Since M N ⊥ AI, Noting the parallel lines, we have MN = DN . Now we
have
sin ∠C


ID 2 CI
= ∠B

· ,
DN sin 2 N C
and
CI BI
= ,
NC MI
so we have
sin ∠C


IK 2 CI BI
= ∠B

· = .
MN sin 2 N C M I
On the other hand since M I = M B we have
sin 90◦ − ∠A


MN 2
=
.
MI sin ∠C
2

So we get    
∠B ∠A
◦
KI · sin = BI · sin 90 − ,
2 2
and that easily gives us S4M IB = S4M IK . Then we have BG = GK and
since ∠KF 0 B = ∠KID = ∠C 0 0
2 we have KF k IC so F ≡ F . Thus, BIKF
is cyclic as desired. 

38
5. Without loss of generality, let’s assume that AC ≥ AB. Let D be the
second intersection point of line AH and the circumcircle of 4ABC and let
O be circumcenter of 4ABC. Let the lines OD and BC meet at J. Then
we have ∠JDH = ∠JHD since BC is perpendicular bisector of HD. On
the other hand ∠DCA = 90 − ∠B + ∠C so we have ∠ODH = ∠B − ∠C.
Therefore JH k AO, since AO ⊥ EF so we get that JH ⊥ EF and J lies on
the line KH. Note that
∠P KD = ∠AKD = 90◦ − ∠B + ∠C.
Also
∠OJP = ∠OJC = ∠CDJ + ∠DCJ = ∠C + 90◦ − ∠B.
Therefore ∠OJP = ∠P KD and JDKP is cyclic and ∠P KH = ∠P KJ =
∠P DJ. But we have ∠P DJ = ∠P HJ = ∠P HK since D is the reflection of
H with respect to P J. So ∠P KH = ∠P HK which gives us P K = P H as
desired.
A

F
H O

J P
B C

D
K

6. Assume the circumcircles of 4ABY and 4AEX meet for the second time
at K. Since
∠KEX = ∠KAX = 180◦ − ∠KAY = 180◦ − ∠KBY = ∠KBS,
we find out that KBES is concyclic. We have
∠Y KB = ∠Y AB = ∠BCA = ∠P Y B,

39
 ∠EKX = ∠EAX = ∠ADE = ∠QXE.
Which gives us

∠Y ZX = ∠Y SX − ∠SY Z − ∠SXZ
= 180◦ − (∠BKE + ∠Y KB + ∠EKX)
= 180◦ − ∠Y KX.

Therefore, Y KXZ is inscribed in a circle called ω. Let’s say the tangent line
to ω at K meets XY at T . We have

∠T KB = ∠T KY +∠Y KB = ∠KXT +∠Y AB = ∠KEA+∠AEB = ∠KEB.

As a result, T K is tangent to the circumcircle of 4KEB. So, the two circles

are tangent at K.

K
Y

A B

X
E
S P
Q

Z C
D

Number Theory
An + k
1. Let An = an . . . a1 a0 , Bn = bn . . . b1 b0 , and Cn = . Obviously
Bn + k
Cn < 10, so Cn ∈ {1, 2, . . . , 9}.

40
Claim. If the sequence {ai } doesn’t eventually become constant 9, there
exists N ∈ N such that k + An < 10n+1 (∀n > N ).

Proof. Assume that there exists an index i large enough that k < 10i and
ai 6= 9. So we get

k + Ai = ai . . . a0 + k < 9 × 10i + k < 10i+1 .

Now note that

k + Ai+1 = 10i+1 ai+1 + Ai + k ≤ 9 × 10i+1 + Ai + k < 10i+2 .

So inductively, the claim will be proved.

If the sequence ai becomes constant 9, then the sequence bi also becomes
constant 9 and we’re done. Now consider the case that ai doesn’t become
constant 9.
For all sufficiently large n, An + k < 10n+1 . Also we know that (An + k)Cn =
Bn + k and (An+1 + k)Cn+1 = Bn+1 + k. which implies:

10n+1 bn+1 = Bn+1 − Bn = (An+1 + k)Cn+1 − (An + k)Cn

= (An + k)(Cn+1 − Cn ) + 10n+1 × Cn+1 an+1 .

Therefore,
10n+1  (An + k)(Cn+1 − Cn ) .


Now suppose that Cn+1 6= C
n , which means 1 ≤ |Cn+1 − Cn | ≤ 8. Then,

because gcd 103 , Cn+1 − Cn ≤ 8, we have 10n−2  An + k .
For all n, define Sn such that for some θ ∈ N,

An + k = 10θ × Sn

and 10 - Sn .
Let Ai−1 + k = 10α × Si−1 . Then

Ai + k = 10α × Si−1 + 10i × ai .

If we choose i large enough, by the previous claim we know that Ai−1 +

k < 10i . So α < i. This means that Ai + k = 10α × Si . So we get that
An + k = 10α × Sn for all sufficiently large n.
But if Cn+1 6= Cn , we have 10n−2  An + k this means that α ≥ n − 2,
which is impossible for large enough n. So there exists N ∈ N such that
Cn+1 = Cn for all n > N . This means that 10n+1 bn+1 = C × 10n+1 an+1 . So
bn+1 = C × an+1 and we’re done. 

41
2. We are going to construct infinite pairs like (a, b) that suit the problems
condition. Consider c be a arbitrary natural number then put a = cn+1398
and b = cm+2019 . Then

am + bn = cmn c1398m + c2019n ,

and
a2019 + b1398 = c1398×2019 c1398m + c2019n .

So obviously set of the prime divisors of am + bn and a2019 + b1398 are the
same. Since c has been arbitrarily chosen we have infinite choices for (a, b).


3. Assume contrary that there exist infinite number of prime numbers, that
divides an element t of S. Let’s denote them p1 , p2 , . . . , pt . Take

x1 , x2 , . . . , x2mt +1 ∈ S.

For every x ∈ S can write it as

x = pα1 α2 αt
1 p2 . . . pt .

Pigeonhole principle implies that there exist two index i, j such that if

xi = p α1 α2 αt
1 p2 . . . pt and xj = p1β1 pβ2 2 . . . pβt t ,

then αi ≡ βi (mod 2m ). So we have

 m m

xi + xj = pθ11 pθ22 . . . pθt t a2 + b2 ,

where a 6= b, ab 6= 0 and gcd(a, b) = 1. So if q is an odd prime such that

 m m
q  a2 + b2 ,


then we’ll have q ≡ 1 (mod 2m ) (It is obvious since ordq ab = 2m+1 ). And

if we take m large enough such that 2m > p1 , p2 , . . . , pt , we’ll have q = 1.

Contradiction.
m m
Now it’s enough to show that there exists an odd prime divisor of a2 + b2
where a 6= b, ab 6= 0 and gcd(a, b) = 1 for every large enough m. For this
m m m m
note that a2 + b2 > 2m and a2 + b2 ≡ 1, 2 (mod 8). 

4. Claim i. f (x) is injective.

Proof. Assume there exist two natural numbers x1 and x2 such that f (x1 ) =

42
f (x2 ) and x1 6= x2 . now choose y and x such that y + z = t2 − x1 then we
have
f (y) + f (z) + f (x1 ) = a2
for some natural number a. therefore,

f (y) + f (z) + f (x1 ) = f (y) + f (z) + f (x2 ) = a2 ,

so y + z + x2 must be square of some natural number s.

Hence
|x1 − x2 | = t2 − s2  6= 0
 

and we have 2
t − s2  ≥ 2|t| − 1.


If we choose t large enough it gives us a contradiction.

Claim ii. For any x, y ∈ N we have f (x + 1) − f (x) = f (y + 1) − f (y).

Proof. Assume some arbitrary x0 and y0 with x0 ≥ 2, and choose z =

t2 − x0 − y0 for some large t. then

f (x0 ) + f (y0 ) + f (z) = a2 ,

and
f (x0 − 1) + f (y0 + 1) + f (z) = b2 .
So we have

|f (x0 ) + f (y0 ) − f (x0 − 1) + f (y0 + 1)| = a2 − b2  .

 

Assume that f (x0 ) + f (y0 ) − f (x0 − 1) + f (y0 + 1) 6= 0. If we choose z large

enough, f (z) is a large number too (it’s trivial because of the injectivity).
But we know

|f (x0 ) + f (y0 ) − f (x0 − 1) + f (y0 + 1)| = a2 − b2  ≥ 2|a| − 1.

 

That means f (z) has an upper bound. Contradiction.

Claim ii implies that f (x) is a linear function so f (x) = ax + b and

a(x + y + z) + 3b,

becomes a perfect-square if and only if

x + y + z,

does. That means at2 + 3b is a perfect-square for any t ∈ N. In particular

a(at2 + 3b) + 3b = t02 hence

3ba + a = t02 − t2 ,

43
which is a contradiction unless 3ba + 3b = 0 (because the left hand side is
either 0 or very big for big t). Now obviously a > 0 and we get b = 0, a = n2 .


5. a) We know that for every polynomial P (x) with degree n, the equation
P (x) ≡ 0 (mod p) has at most n distinct roots module p, for every prime
number p. Then for every pi ∈ S we have some ti ∈ {1, 2, . . . , p} where
pi - P (ti ). Now choose t by Chinese Reminder Theorem such that t ≡ ti
(mod pi ). Then for every pi ∈ S we have pi - P (t)

b) Put
P (x) = x(x − 1) . . . (x − 1396) + 1398!.
Note that 1397 and 1398 are composite numbers. then obviously for
any p ≤ 1398 and n ∈ N there exists an ai such that 1 ≤ ai ≤ 1398
and n ≡ ai (mod pi ). Therefore pi | P (n) and since P (x) is monic it
satisfies our desired conditions.


6. Assume for a contradiction that ordm (a) has at least two prime factors.
Then we can write ordm (a) = rs where gcd(r, s) = 1 and r, s are both greater
than 1. Now take a natural solution pair (k, `) of the equation ks + r` ≡ 1
(mod ϕ(m)),
That definitely exists by Bézout’s theorem, and consider (x, y) =
−aks , −ar` . We claim this choice of (x, y) will contradict the condition of
problem. First of all:

xy ≡ aks+r` ≡ a (mod m).

Where the last equality follows from Euler’s Theorem. Now we have to prove
a lemma:

Lemma. Let z, t be natural numbers such that ordz (t) is odd. Then we
have ordz (−t) = 2ordz (t).

Proof. Let ordz (t) = w then (−t)w ≡ −1 (mod z) and (−t)2w ≡ 1 (mod z).
So ordz (−t) = 2q is even. Therefore if q < w we have t2q ≡ 1 (mod z) which
gives w | 2q and since w is odd we have w | q which is clearly a contradiction.

Using the lemma, it is easy to see that ordm (x) = 2r and ordm (y) = 2s
which is clearly a contradiction hence ordm (a) has at most one prime factor.


44
Team Selection Test

1. Let dX,Y be the weight assigned to the edge XY of the graph. Clearly, in
each triangle the maximum weight of an edge is the sum of the two others.
Let dA,B be the maximum weight among all weights assigned to the edges.
Now assign 0 to vertex A and dA,X to any other vertex X.
Obviously, this labeling works for all edges including A. We claim that it
works for all other edges.
Assume to the contrary that the labeling doesn’t work for an edge XY of the
graph. So dX,Y is not the difference between dA,X and dA,Y , which means
dX,Y = dA,X + dA,Y . Note that B cannot be any of X, Y . Since dA,B is the
maximum edge number, we have dB,X = dA,B −dA,X and dB,Y = dA,B −dA,Y .
Now consider triangle BXY . If dX,Y is the maximum weight of an edge in
this triangle, we have
dX,Y = dB,Y + dB,X =⇒ dX,Y = dA,B − dA,Y + dA,B − dA,X
=⇒ 2dX,Y = 2dA,B .
Which is impossible since the weights are distinct. So the maximum edge
weight in triangle BXY is not dX,Y . Without loss of generality, assume that
dB,Y is the maximum one. So we have
dB,Y = dB,X + dX,Y =⇒ dA,B − dA,Y = dA,B − dA,X + dA,X + dA,Y
=⇒ dA,Y = 0.
Which is again a contradiction. This completes the proof. 

2. First we prove that it suffices to show that ASCBR is cyclic. Let ω

be the circumcircle of 4ABC. The fact 4P AE ∼ 4BDC implies that
∠P AB = ∠C. Therefore, P A is tangent to ω. Similarly, 4QAD ∼ 4EBC
implies ∠QAD = ∠B. Therefore, QA is tangent to ω. Hence, points A, P
and Q are collinear and P Q is tangent to the ω at A. Now we have RS k P Q
and since R and S lie on ω, we find out that AR = AS and ∠ARS = ∠ASR.
Let M be the intersection point of lines SD and RE. We have
∠M RA = ∠M SA = ∠BAO =⇒ ∠M RS + ∠ARS = ∠M SR + ∠ASR.

45
And since ∠ARS = ∠ASR, we have ∠M RS = ∠M SR and M S = M R.
Therefore A, M and O lie on the perpendicular bisector of RS. Now let X
be the intersection point of lines P E, BD.
Q S
A

P
D
E
R
M
O

B C

Since 4P AE ∼ 4BDC, angles ∠AP E and ∠BDC are equal and AP XD

is cyclic and since R, X lie on the circumcircle of 4P AD, it follows that
AP RXD is cyclic. Furthermore,

∠XBE = ∠XRA − ∠ARE = ∠XDC − ∠BAC = ∠ABD = ∠XBE.

Therefore, XBRE is cyclic as well. We’ll get

∠BRE = ∠EXD = 180◦ − (∠C + ∠A) = ∠B

=⇒ ∠ARB = ∠ARE + ∠BRE = ∠A + ∠B
=⇒ ∠ARB + ∠C = 180◦ .

Which gives us that R lies on ω. Analogously, S lies on ω. Therefore ASCBR

is cyclic and we’re done. 

3. Only n = 2 and n = 3 are interesting.

Note that if n is not interesting, then any integer greater than n is not inter-
esting as well. So we just need to prove that n = 4 is not interesting.
We claim that there is no such example for the case

P2 (x) > P4 (x) > P1 (x) > P3 (x)

46
for infinitesimal positive number x.
For the sake of contradiction, assume that there exists an example. Since
Pi (x) − P1 (x) also satisfies the inequality, assume that P1 (x) = 0. Then by
contradiction, every Pi (x) has constant term zero.
Now, P3 (x) is always negative for values close to 0. Therefore, the term
with the minimal degree in P3 (x) is of the form −ax2b , for some positive real
number a and positive integer b. Analogously, for P2 (x), P4 (x), the terms of
the minimal degrees are of the form ax2b−1 , for some positive real number a
and positive integer b. Let

P2 (x) = a2 x2b2 −1 + · · ·
P3 (x) = −a3 x2b3 + · · ·
P4 (x) = a4 x2b4 −1 + · · ·

where every ai is a positive real number and every bj is a positive integer.

Since P2 (x) > P3 (x) > P4 (x) for x < 0, therefore

2b2 − 1 > 2b3 > 2b4 − 1 =⇒ b2 > b3 > b4 =⇒ P2 (x) < P4 (x),

for all sufficiently small positive real numbers x, contradicting the assumption
P2 (x) > P4 (x) > 0 > P3 (x).
But we claimed that n = 2, 3 are interesting. The case n = 2 is trivial and
for n = 3, we have
Pi (x) = xQi (x) + c.
If σ(i) = i, we need to find polynomials Qi such that Q1 (x) < Q2 (x) < Q3 (x)
for x ∈ (−, 0) and Q1 (x) > Q2 (x) > Q3 (x) for x ∈ (0, ). It’s clear that
linear polynomials will work.
If σ(1) = 1, σ(2) = 3 and σ(3) = 2, let Q1 (x) = 0 and Q2 (x), Q3 (x) be linear
polynomials passing through point 0 with negative leading coefficient. Other
permutations are the same and we’re done. 

4. Let S1 = 1 and for every i ≥ 1,

Si+1 = {x | g(x) ∈ Si } − {1}.

Also
∞
[
Si = A = {x | ∃n ≥ 0 : g n (x) = 1} .
i=1

Clearly A is non-empty. If [0, 1] − A = B is also non-empty, partitions A, B

of [0, 1] will lead to a contradiction:
If x ∈ A:
∃i : x ∈ Si =⇒ g(x) ∈ Si−1 ⊆ A.

47
And if g(x) ∈ A:
∃i : g(x) ∈ Si =⇒ x ∈ Si+1 ⊆ A.
Therefore, B is non-empty. Now if S2 has a finite number of elements, it has
an element that is largest of all the elements. Let us denote it by z. Since
g(x) > x,
max(x | x ∈ Si+1 ) ≤ max(x | x ∈ Si ).
Therefore, all the elements of A except 1 are less than or equal to z. So, the
non-empty interval (z, 1) doesn’t exist in A, which is a contradiction.


5. Let Pk be the set of solutions of

(
n + S(n) = m + S(m),
kn + S(kn) = km + S(km).

We want to map a single solution (m0 , n0 ) ∈ Pk , where 10 - m0 , n0 , to

infinite solutions like (m1 , n1 ), where (m1 , n1 ) ∈ Pk+1 or (m1 , n1 ) ∈ Pk−1
and 10 - m1 , n1 .
If (m0 , n0 ) be a solution in Pk , then we claim that (m0 + 10α , n0 + 10α ) is a
solution in Pk+1 for every large α. It follows by the same easy calculations
and noting that

S n2 + 2 × 10α n + 102α = S n2 + S(2n) + 1,

for all large α.

And if (m0 , n0 ) is a solution in Pk where 10 - m0 , n0 , then we replace them by
(10α − m0 , 10α − n0 ) for all large α to reach an infinite number of solutions in
Pk−1 . Therefore, it’s enough to find a simple solution. For example we have
(9, 12) ∈ P0 and we’re done.


6. The answer is yes.

We prove a stronger result. If positive numbers x ≤ d1 ≤ · · · ≤ dn are given,
we can find a convex polygon with n + 3 edges and a triangulation of it such
that the lengths of diameters used in the triangulation are d1 , . . . , dn , also the
length of one of the edges is x and one of the angles adjacent to this edge is
acute.
We prove the above statement by induction. Case n = 1 is evident.
Assume that the statement is true for n − 1. So if we are given the positive
numbers d1 ≤ d2 ≤ · · · ≤ dn , we can find a convex polygon P1 . . . Pn+1 Pn+2
and a triangulation of it with diameters d2 , . . . , dn , such that Pn+2 P1 = d1
and also ∠Pn+2 P1 P2 < 90◦ . Now assume that x ≤ d1 . We construct the
desired n + 3-gon and its triangulation. Now there are two cases:

48
Case 1) x = d1 .
Consider the circle ω with center Pn+2 and radius d1 . Assume that
Pn+2 Pn+1 intersects ω at K such that Pn+2 lies between Pn+1 and K.
Let Q be a point on the smaller arc P1 K of ω. Obviously

∠Pn+2 QP1 = ∠Pn+2 P1 Q =⇒ ∠Pn+2 QP1 < 90◦ .

Also
∠Pn+1 Pn+2 Q, ∠QP1 P2 < 180◦ .
So the polygon P1 . . . Pn+1 Pn+2 Q is convex and has the desired prop-
erties.

Pn+1

Pn+2

P2
K
P1 Q

Case 2) x < d1 .
Let ω 0 be the circle with center P1 and radius x. Since

∠Pn+2 P1 P2 < 90◦ ,

there exists a point R on ω 0 such that R, P2 lie on the different sides of

Pn+2 P1 and
∠RP1 P2 = 90◦ .
Because x < Pn+2 P1 , there exists a point Q on ω 0 such that Q and R
are on the same side of Pn+2 P1 , the points Q and P1 are on the same
side of Pn+2 Pn+1 , and Q, P2 lie on the same side of P1 R. Now the
polygon P1 . . . Pn+1 Pn+2 Q is the desired polygon because it is convex,
QP1 = x, and
∠QP1 P2 < 90◦ .

49
7. The answer is 1 for m = 0, p for m = 1 and ϕ(pm ) for m > 1.

Note that Zp [x] is a unique factorization domain. So any monic P (x) ∈ Zp [x]
can be uniquely expressed as P (x) ≡ A(x)2 Q(x) (mod p), where A(x), Q(x)
are monic polynomials and Q(x) is square-free. Let fn be the number of all
square-free monic polynomials of Zp [x] with degree n. Any non-square-free
monic polynomial can be expressed as P (x) = A(x)2 S(x) where S(x) is a
square-free monic polynomial.
If deg A(x) = m, then S(x) ∈ fn−2m . We have fn−2m options for S(x) and
pm options for A(x). So
n
b2c
X
n
fn = p − fn−2m pm .
m=1

And by simple induction on n we get fn = ϕ(pn ).



8. We claim that Bob has a winning strategy and the maximum point differ-
2
ence he can make sure will happen is 20202 .
First we show Bob’s strategy. Let ` be the vertical line that dissects the table
into two equal tables. After Alice colors a cell, Bob can easily color the cell
symmetric to that cell with respect to `. This way, Bob gets the same point
Alice got from the column and gets one more point than Alice got from the
row, because the cell he colored is in the same row as Alice’s cell. So after
each move, Bob earns exactly one more point than Alice, which means at the
2
end he has earned 20202 points more than Alice.
Now we show that Alice can make sure the difference will not be greater
2
than 2020
2 points. In each move, Alice chooses the cell that she can earn the
maximum possible points by coloring. Assume that she earns p points by
this move. Because it is the maximum choice, any cell that Bob chooses in
the next move will be in the same row or column with at most p + 1 colored
2
cells.This means the point difference in the end will be at most 2020 2 and
we’re done. 

9. Let P be the A-mixtilinear touchpoint with Γ.

50
 T
A S0
S

KJ E
F 0
J
I

B D M XA C

P
X0 N X

We know that P, I and T are collinear. First suppose that we have IJ k AP .

We have

∠IT X = ∠P T X = ∠P AX = ∠IJX =⇒ ∠IT X = ∠IJX.

So, IJT X is cyclic and it suffices to show that IJ k AP . We know that

EF, AT and BC are concurrent at some point Q. Let D be the intersection
point of lines AI, BC and N be the intersection point of line AI and circum-
circle of 4ABC. Suppose that QN intersects the circumcircle of 4ABC at
X 0 (X 0 6= N ) and let J 0 be the intersection point of the lines AX 0 and EF .
We have
N A
(AX 0 , BC) = (DQ, BC) = −1.
Therefore, X 0 lies on the A-symmedian of 4ABC. Since S is where the cir-
cumcircles of two triangles 4ABC and 4AEF meet for the second time, we
know that S is the Miquel’s point of the quadrilateral BCEF and quadrilat-
eral SQBF is cyclic. So

∠SQF = ∠SBF = ∠SN A =⇒ ∠SQK = ∠SN K,

and SQN K is cyclic. Therefore

∠AJK = ∠ASK = ∠AS 0 K = ∠ASN − ∠N SK = ∠AX 0 N − ∠N QK

= ∠AJ 0 K.

Which gives us ∠AJK = ∠AJ 0 K. So J and J 0 are symmetrical with respect

to line AI. Now let XA be the touchpoint of A-excircle with side BC. Since

51
J and J 0 are symmetrical, it suffices to show that AXA k IJ 0 to get AP k IJ.
If M is the midpoint of BC, we know that M I k AXA . So we just need to
prove that M, I and J 0 are collinear which is true since SQN K is cyclic and

(EF, J 0 Q) = A(EF, JQ) = (BC, M Q).

>
10. Let P be the midpoint of segment BC and J be the midpoint of arc BC
(J 6= A).
A

Q
M D
N
I

B P R C X

J
We call the circumcircle of triangle 4CID, ω and the intersection point of ω
and segment BC, R. We have

∠QIC = 180◦ − (∠IQC + ∠ICQ) = 180◦ − (∠P IC + ∠ICP ) = 90◦ .

So, ∠QIC = 90◦ and N is the center of ω which gives us ∠QRC = 90◦ and
QR k AP . ∠RDC = ∠RQC gives us ∠JAC = ∠JDC. So, points D, R
and J are collinear. Since ∠RN C = 2∠RQC = 2∠P AC = ∠A, we have
RN k AB. Therefore
AN BR AN RC BM
= =⇒ · · = 1.
NC RC N C BR M A
So, the lines AR, BN and CM are concurrent. Let X be the intersection
point of lines AD and BC. It suffices to show that (XR, CB) = −1. Since
>
D, R and J are collinear and J is the minpoint of arc BC, We have
D
(XR, CB) = (AJ, CB) = −1.

Hence the result. 

52
 1
11. Take a sufficiently large m (m > 23000 ) and put x = m + k−1 . Note that
 n  X n mn−i
1
m+ = .
k−1 i m(k−1)i
ki >n

Because
X n mn−i n  
1 X n 2n+1
(k−1)i
< < < 1,
i m m i=0 i m
ki >n
 n
1
and the remaining terms of m + k−1 are all integers. Now if n = kt + r
such that k − 1 ≥ r ≥ 0, we have two cases:
Case 1 r 6= 0.

 n  X n 1
1 i=t+1 1
m+ = k −n
> k−r
,
k−1 i m i m
ki >n
( n−1 )
1 X n − 1 1 n
m+ = < 2n × k−r+1 ,
k−1 i mki −n+1 m
ki >n−1
i≥t+1
 n  ( n−1 )
1 n n 1 1
>2 × =⇒ m+ > m+ .
mk−r mk−r+1 k−1 k−1

Case 2 r = 0.

 n  X n 1
1 n
m+ = < 2n × k ,
k−1 i mki −n m
ki >n
i≥t+1
( n−1 )
1 X n − 1  1 1
m+ = ≥ ,
k−1 i mki −n+1 m
ki >n−1
i≥t
( n−1 )  n 
1 k≥2 n n 1 1
> 2 × k =⇒ m+ > m+ .
m m k−1 k−1

Therefore,
   
1 n−1 1 n
(m + ) > (m + ) ⇐⇒ k | n,
k−1 k−1
 n−1
x > {xn } ⇐⇒ k | n.

53
12. Assume that p = 3. Then obviously
 x1 can take  any value. Now suppose
p−1
that p ≥ 5. We want to prove that x1 , x2 , . . . , x p−1 ∈ {0, 1} 2 .
2
Note that
p−1
2
X p−1 p−1  p−1

(1 − axi ) 2 ≡ + M (1 − a) 2 − 1 (mod p),
i=1
2

where M is common value mod p of

p−1
2  
X p−1
xji 1≤j≤ .
i=1
2

p−1 p−1
If we choose a such that (1 − a) 2 = 1, then by using (1 − axi ) 2 = ±1 or 0
and the above equation we have for each i,
 
p−1 1 − axk
(1 − axi ) 2 = =1
p
 
So if pt = 1 we get
 
1 − (1 − t)xi
=1
p
Now if xi 6= 0 then the map f (t) = 1 − (1 − t)xi is bijective. Hence if P is
the set of square remainder, f define a bijection from P to itself and
X X
f (t) = t = 0,
t∈P t∈P

which implies that xi = 1. 

54