C

x² + y² ≥ 2xy a ≡ a(mod p)
A B

50
Authored
Math Competition
Problems
Slobodan Filipovski, Ph.D.
 Preface
This book is the result of my extensive work in organizing mathematical competitions
in Macedonia and Slovenia. Over the past decade, I have been a member of the Problem Se-
lection Committee for various competitions and Olympiads in Macedonia, where I proposed
numerous original problems for primary and high school contests. During my doctoral stud-
ies, I also contributed to the preparation of the University of Primorska team from Slovenia
for international competitions such as the Vojtěch Jarnı́k Competition and the International
Mathematics Competition (IMC).
As a natural outcome of these activities, I have created numerous original mathematical
problems for competitions.
This book contains 50 original mathematical problems: 24 problems from Algebra (pri-
marily inequalities), 5 problems from Combinatorics, 3 from Geometry, 4 from Linear Alge-
bra, and 14 problems from Number Theory. Many of these problems have been included in
the shortlists of regional and national mathematical competitions in Macedonia, the Mace-
donian Mathematical Olympiad, the Junior Macedonian Mathematical Olympiad, the Ju-
nior Balkan Mathematical Olympiad, the Balkan Mathematical Olympiad, the International
Mathematics Competition for university students, and the Vojtěch Jarnı́k Competition for
university students.

August 2024, Izola Slobodan Filipovski

Contents

1 PROBLEMS 5
1.1 ALGEBRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 COMBINATORICS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 GEOMETRY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 LINEAR ALGEBRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 NUMBER THEORY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 SOLUTIONS 15
2.1 ALGEBRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 COMBINATORICS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3 GEOMETRY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.4 LINEAR ALGEBRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5 NUMBER THEORY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3
4 Contents
Chapter 1

PROBLEMS

1.1 ALGEBRA
Problem 1. (Shortlist JBMO 2011) Let x, y, z be positive real numbers. Prove that
x + 2y y + 2z z + 2x 3
+ + ≤ .
z + 2x + 3y x + 2y + 3z y + 2z + 3x 2

Problem 2. (Shortlist MMO 2011) Let a, b, c be positive real numbers such that
1 1 1
2
+ 2 + 2 = 1. Prove the inequality
a b c
a5 + b 5 + c 5
≥ 3.
a3 + b 3 + c 3

Problem 3. (Regional Competition 2012, Macedonia) Prove the inequality

s   s   s  
n n n n n n
+ + ... + ≤ 2n − 1.
0 1 1 2 n−1 n

Problem 4. (Shortlist JBMO 2012) Solve the following equation for x, y, z ∈ N.

 2  2  2
x y z 27
1+ + 1+ + 1+ = .
y+z z+x x+y 4

Problem 5. (JMMO 2012) Let a, b and c be positive real numbers. Prove the inequality
1 √ √ √ 1 1 1
( a + b + c) + + + ≥ 3.
2 1+a 1+b 1+c

Problem 6. (MMO 2012) If a, b, c and d are positive real numbers such that abcd = 1,
then prove that the following inequality holds
1 1 1 1
+ + + ≤ 2.
bc + cd + da − 1 ab + cd + da − 1 ab + bc + da − 1 ab + bc + cd − 1

5
6 Chapter 1. PROBLEMS

When does equality hold?

Problem 7. (MMO 2013) Let x, y and z be positive real numbers such that x4 +y 4 +z 4 =
3. Prove that
9 9 9
2 4 6
+ 4 6 2
+ 6 ≤ x6 + y 6 + z 6 + 6.
x +y +z x +y +z x + y2 + z4

Problem 8. (JMMO 2013) Let a, b and c be positive real numbers satisfying a+b+c+2 =
abc. Prove that
a b c
+ + ≥ 2.
b+1 c+1 a+1
Problem 9. (Shortlist BMO 2013) Let a, b, c and d be positive real numbers so that
abcd = 41 . Prove that holds
  
a 16c 4 b d 1 81
16ac + 2 + 2 + bd + 2
+ 2 + ≥ .
c b a d ac 256d c b a 64bd 4

Problem 10. (IMIO 2013) Let a, b and c be positive real numbers so that ab + bc + ca =
6abc. Prove that holds  
1 a b c
a+b+c≥1+ + + .
6 c a b
Problem 11. (Shortlist JMMO 2013) Let 0 ≤ a, b, c ≤ 1 and a + b + c = 2. Prove that
a2 b + b2 c + c2 a ≤ 1.

Problem 12. (Shortlist JMMO 2014) Let a, b and c be positive real numbers such that
a + b + c = 3. Prove that
a5 + b5 + c5 + 6 ≥ 3(a2 + b2 + c2 ).

Problem 13. (Shortlist JMMO 2014) Let a, b and c be positive real numbers such that
a + b + c = 1. Prove that holds
(ab + 2ac + 3bc)3 ≥ (2a)5 · (3b)4 · (3c)3 .

Problem 14. (MMO 2014) Let a, b and c be positive real numbers such that a+b+c = 4
and a, b, c > 1. Prove that
 
1 1 1 1 1 1
+ + ≥8 + + .
a−1 b−1 c−1 a+b b+c c+a

Problem 15. (Shortlist JMMO 2014) Let a, b and c be the side lengths of a triangle
with an area P and let a + b + c = 6. Prove the inequality
√ √ √ √
3−a 3−b 3−c 3 3
+ + ≤ .
1 + (3 − b)(3 − c) 1 + (3 − c)(3 − a) 1 + (3 − a)(3 − b) 2P
1.1. ALGEBRA 7

Problem 16. (Shortlist BMO 2015) Let a, b, c be side lengths of a triangle and let
ma , mb , mc be the medians at the corresponding sides. Prove that
   b 
b c  a  c  a
ma −1 − 1 + mb −1 − 1 + mc −1 − 1 ≥ 0.
a a b b c c

Problem 17. (Shortlist MMO 2015) Let a, b, c be positive real numbers greater than
1 so that (a − 1)(b − 1)(c − 1) = 1. Prove that
r r r
5 a + b 5 b + c c+a
+ + 5 ≥ 3.
4 4 4

Problem 18. (Vojtěch Jarnı́k 2016) Let a, b and c be positive real numbers such that
a + b + c = 1. Show that
   
1 1 1 1 1 1
+ + + ≥ 1728.
a bc b ca c ab

Problem 19. (Shortlist Vojtěch Jarnı́k 2017) Let k ≥ 3 be an odd number and let
Hm (x) be the Dickson polynomial defined as follows: H0 (x) = 1, H1 (x) = x, Hi+2 (x) =
xHi+1 (x) − (k − 1)Hi (x), with i ≥ 0. Prove that the polynomial Hm (x) − 2 is irreducible.
Problem 20. (National Competition 2020, Macedonia) Let a1 , a2 , . . . , a2020 be pos-
itive real numbers such that a1 ≤ a2 ≤ . . . ≤ a2020 , a1 + a2 + . . . + a2020 = 2020 and
a21 + a22 + . . . + a22020 = 2021. Prove that
1
a2019 ≥ 1 − √ .
2019 · 2020

Problem 21. (University of Primorska Scholarship Competition in Kosovo,

2021) Let x1 ≥ x2 ≥ . . . ≥ xn > 0 be positive real numbers such that x1 + x2 + . . . + xn =
1
x1
+ x12 + . . . + x1n . Prove that

(x1 − xn )2
x1 + x2 + . . . + xn ≥ n + .
2(x21 + x2n )

Problem 22. (Regional Competition 2021, Macedonia) Let a1 , a2 , . . . , an be positive

real numbers such that a1 a2 · · · an = 1. Prove that

(n − 1)(a21 + a22 + . . . + a2n − 1)2 ≥ (a1 + . . . + an − 1)2 .

Problem 23. (Vojtěch Jarnı́k 2022) Let x1 , . . . , xn be given real numbers with 0 <
m ≤ xi ≤ M for each i ∈ {1, . . . , n}. Let X be the discrete random variable uniformly
distributed on {x1 , . . . , xn }. The mean µ and the variance σ 2 of X are defined as

x1 + . . . + xn (x1 − µ(X))2 + . . . + (xn − µ(X))2

µ(X) = and σ 2 (X) = .
n n
8 Chapter 1. PROBLEMS

By X 2 denote the discrete random variable uniformly distributed on {x21 , . . . , x2n }.

Prove that  m 2
σ 2 (X) ≥ σ 2 (X 2 ).
2M 2
Problem 24. (American Mathematical Monthly 2022) Let n, k be positive integers
with n ≥ 3 and let p(x) = xn + xn−1 + · · · + x − k.

(a) Prove that the roots of p(x) in the complex plane are simple.

(b) Prove that if k ≥ n + 1, then p(x) has at least one root with negative real part and
nonzero imaginary part.
1.2. COMBINATORICS 9

1.2 COMBINATORICS
Problem 25. (National Competition 2011, Macedonia) The interior of a convex 12-
gon is divided into ten triangles so that no two triangles overlap, with four of these being
special triangles. Each triangle has a non-zero real number written inside it, and for each
special triangle, the number written inside it is the product of the numbers written inside
the triangles that enclose it. Can the product of all ten written numbers in the 12-gon be
constant? (A triangle is called a special triangle if a triangle has been constructed over
each of its sides).

Problem 26. (Regional Competition 2012, Macedonia) Ten teams participated in a

football tournament, with each team playing a match against every other team. For every
win, a team earns 3 points; for every draw, each team earns 1 point; and for every defeat, a
team earns 0 points. At the end of the tournament, the total number of points accumulated
across all teams was 120. Determine how many matches ended in a draw.

Problem 27. (National Competition 2021, Macedonia) Among a group of n people,

there are exactly 2021 mutual friendships. Moreover, it is known that no person has more
than 45 friendships, and there is at least one person with exactly 45 friendships. Let xk
denote the number of friendships of the k-th person. Prove that

x21 + x22 + . . . + x2n ≤ 2024n + 2018.

Problem 28. (IMC 2023) Let T be a tree with n vertices; that is, a connected simple
graph on n vertices that contains no cycle. For every pair u, v of vertices, let d(u, v) denote
the distance between u and v, that is, the number of edges in the shortest path in T that
connects u and v. Consider the sums
X X 1
W (T ) = d(u, v) and H(T ) = .
{u,v}⊆V (T ) {u,v}⊆V (T )
d(u, v)
u̸=v u̸=v

Prove that
(n − 1)3 (n + 2)
W (T ) · H(T ) ≥ .
4
Problem 29. (Vojtěch Jarnı́k 2024) Let n be a positive integer and let G be a simple
undirected graph on n vertices. Let di be the degree of its i-th vertex, i = 1, . . . , n. Denote
∆ = max di . Prove that if
Xn
d2i > n∆(n − ∆)
i=1

then G contains a triangle. (A graph is called simple if there are no loops and no multiple
edges between any pair of vertices.)
10 Chapter 1. PROBLEMS

1.3 GEOMETRY
Problem 30. (Shortlist JMMO 2011) A triangle has side lengths a, b and c. Let R be
the radius of the circumscribed circle and let r be the radius of the inscribed circle in the
triangle. Prove that
ab + bc + ca
> 9r.
R
Problem 31. (National Competition 2012, Macedonia) In a triangle ABC, the
bisector of angle ∠A intersects side BC at D so that |BD| = 2|DC|. The height CS from
C intersects the bisector AD at the point M. Prove that

m2c
|CM | = .
2hc

Problem 32. (Shortlist MMO 2013) In a triangle ABC, the bisector of the angle ∠A
intersects the side BC at point M such that |BM | : |M C| = 3 : 1. A line parallel to the side
AB is drawn through the vertex C, which intersects the extension of the bisector of ∠A at
the point D. Let S be the intersection between AC and BD, let R be a midpoint of the
side AD and let P be the intersection point between CR and BD. Prove that |P S| = |SD|.
1.4. LINEAR ALGEBRA 11

1.4 LINEAR ALGEBRA

Problem 33. (Team Selection Test, Famnit, 2016) Let λ1 , λ2 , ..., λn ∈ R+ be eigen-
1
values of a real n × n matrix A. If det(A) = , then prove that
n!
n  k
X 1
1 + λk − ≥ n.
k=1
k

When does equality hold?

Problem 34. (Shortlist IMC 2016) Let A and B be 3 × 3 matrices. Prove that

∥(AB − BA)3 ∥2 ≥ |det(AB − BA)|.

Problem 35. (Shortlist IMC 2023) Let A be a real square matrix such that the sum of
each row is equal to d > 1, d ∈ N, and trace(Ai ) = 0, for each i = 1, . . . , 2023. Prove that

A2023 + A2022 + . . . + A + I ̸= J,

where J is the all-ones matrix.

Problem 36. (Shortlist Vojtěch Jarnı́k 2023) Find the eigenvalues of the matrix
 
3 0 1 1 0 0 1 1 1 1
 0 3 0 1 1 1 0 1 1 1 
 
 1 0 3 0 1 1 1 0 1 1 
 
 1 1 0 3 0 1 1 1 0 1 
 
 0 1 1 0 3 1 1 1 1 0 
A=  0 1 1 1
.
 1 3 1 0 0 1  
 1 0 1 1 1 1 3 1 0 0 
 
 1 1 0 1 1 0 1 3 1 0 
 
 1 1 1 0 1 0 0 1 3 1 
1 1 1 1 0 1 0 0 1 3
12 Chapter 1. PROBLEMS

1.5 NUMBER THEORY

Problem 37. (Shortlist JMMO 2012) Let n be a natural number. In the set of integers
solve the following system 
x + y + z = 2012
x5 + y 5 + z 5 + 1 = 6n

Problem 38. (Shortlist JMMO 2013) Let p1 > 3 and p2 > 3 be two primes such that
6 | p1 + p2 . Prove that the number
1p1 p2 +1 + 2p1 p2 +2 + 3p1 p2 +3 + 4p1 p2 +4 + 5p1 p2 +5 + 6p1 p2 +6
is composite.

Problem 39. (Shortlist JMMO 2014) In the set of prime numbers solve the equation
p144 + q 144 − 1 = 2013r .

Problem 40. (Shortlist JMMO 2014) Prove that the product of six consecutive positive
integers cannot be a perfect cube.
Problem 41. (Team Selection Test, Famnit, 2016) Write/represent the number
20162014 as a sum of four cubic numbers.
Problem 42. (Shortlist Vojtěch Jarnı́k 2017) Let p and q be odd prime numbers such
that p2 > 2q and q | p2 + 1. Prove that at least one of the numbers 4p2 − 4q and 8p2 − 16q
is a sum of three or fewer squares.

1
Problem 43. (Shortlist Vojtěch Jarnı́k 2018) The sequence {an }∞ n=1 satisfies a1 =
4
1 s
and an+1 + = 8, n ≥ 1. Prove that a4k = for some s and t such that 4|s and 32|t + 1.
an t

Problem 44. (National Competition 2020, Macedonia) Let n and k be natural

numbers for which n > k ≥ 4. We suppose that k(k − 1) is not divisible by
n − 1 and
n
k(k − 1)(k − 2) is not divisible by n − 2. Prove that the binomial coefficient k has at least
two prime divisors p and q such that p | n − 1 and q | n − 2.

Problem 45. (Shortlist Vojtěch Jarnı́k 2022) Let p ≥ 3 be a prime number and let
n ≥ 1 be a natural number. Prove that for any k such that 2 ≤ k ≤ pn holds
 n   n 
p +2 p +1
gcd , > 1.
k 2

Problem 46. (Shortlist IMC 2022) Let p > 3 be a prime number. Prove that, if p is a
primitive root of 4p + 1, then 2p + 1 is a composite number.
1.5. NUMBER THEORY 13

Problem 47. (National Competition 2023, Macedonia) The natural number n is

called magic if and only if there are exactly five natural numbers ki for which n+ki | (n−ki3 ),
(1 ≤ i ≤ 5). Prove that 29 is a magic number.
Problem 48. (Shortlist IMC 2023) Let {ai }∞ i=1 be a geometric progression of natural
numbers which quotient has exactly k distinct prime divisors. Prove that the (k − 1)-th
differences of the sequence {τ (ai )}∞
i=1 form an arithmetic progression.

Problem 49. (Vojtěch Jarnı́k 2024) Let p > 2 be a prime and let

A = {n ∈ N : 2p | n and n | 3n − 1}.

Prove that T
|A [1, k]| 2 log 3
lim sup ≤ .
k→∞ k p log p

Problem 50. (Shortlist Vojtěch Jarnı́k 2024) Let p ≥ 3 be a prime number and let
m, n, a and b be integers such that gcd(m, p) = gcd(n, p) = 1. Prove that, if the congruence
x2 ≡ −mn (mod p) is not solvable and if p | ma2 + nb2 , then p2 | ma2 + nb2 .
Chapter 2

SOLUTIONS

2.1 ALGEBRA

Problem 1. (Shortlist JBMO 2011) Let x, y, z be positive real numbers. Prove that

x + 2y y + 2z z + 2x 3
+ + ≤ .
z + 2x + 3y x + 2y + 3z y + 2z + 3x 2

x + 2y y + 2z z + 2x
Solution. Let us denote a = ,b = and c = .
z + 2x + 3y x + 2y + 3z y + 2z + 3x
x+y+z x+y+z x+y+z
We easily conclude that 1 − a = ,1−b = and 1 − c = .
z + 2x + 3y x + 2y + 3z y + 2z + 3x
Thus we get
1 1 1 6(x + y + z)
+ + = = 6.
1−a 1−b 1−c x+y+z
Applying the inequality between arithmetic and harmonic means for the positive numbers
1 − a, 1 − b and 1 − c, we get

9 9
3 − (a + b + c) = (1 − a) + (1 − b) + (1 − c) ≥ = .
1 1 1 6
+ +
1−a 1−b 1−c

3
From the last inequality we have a + b + c ≤ , as desired. □
2

Problem 2. (Shortlist MMO 2011) Let a, b, c be positive real numbers such that
1 1 1
2
+ 2 + 2 = 1. Prove the inequality
a b c

a5 + b 5 + c 5
≥ 3.
a3 + b 3 + c 3

15
16 Chapter 2. SOLUTIONS

Solutionq1. We use the q inequality between power means of order 5 and 3.

5 5 5 3 3 3
It yields 5 a +b3 +c ≥ 3 a +b3 +c . This inequality is equivalent to
r !2
a5 + b 5 + c 5 3 a3 + b 3 + c 3
≥ . (2.1)
a3 + b 3 + c 3 3

Again, from the inequality between power means of order 3 and 2, we get
r r
3 3 3 a2 + b 2 + c 2
3 a + b + c
≥ . (2.2)
3 3
In the end, combining (2.1) and (2.2) and using AM-HM inequality for the numbers a2 , b2
and c2 , we get:

a5 + b 5 + c 5 a2 + b 2 + c 2 3 3
≥ ≥ 1 1 1 = = 3.
a3 + b 3 + c 3 3 a2
+ b2
+ c2
1

Solution 2. Without loss of generality we may assume a ≤ b ≤ c. Then a2 ≤ b2 ≤ c2 and

a3 ≤ b3 ≤ c3 . Based on Chebyshev inequality we get

a2 + b 2 + c 2 a3 + b 3 + c 3 a5 + b 5 + c 5
· ≤ . (2.3)
3 3 3
Therefore
a5 + b 5 + c 5 a2 + b 2 + c 2 3 3
≥ ≥ 1 1 1 = = 3.
a3 + b 3 + c 3 3 a2
+ b2
+ c2
1
□

Problem 3. (Regional Competition 2012, Macedonia) Prove the inequality

s   s   s  
n n n n n n
+ + ... + ≤ 2n − 1.
0 1 1 2 n−1 n

Solution. Let 1 ≤ k ≤ n. From the inequality between arithmetic and geometric means
we have
s
n n


+
  
k−1 k n n
≥ · . (2.4)
2 k−1 k
Setting k = 1, 2, . . . , n in (2.4) gives
s   s   s
n n
 

   
n n n n n n 0
+ n n n
+ +. . .+ ≤ + + ... + .
0 1 1 2 n−1 n 2 1 n−1
(2.5)
2.1. ALGEBRA 17

We use the well-known identity n0 + n1 + . . . + nn = 2n . Substituting in (2.5) we get

s   s   s       
n n n n n n n n n
+ + ... + ≤1+ 2 − − = 2n − 1.
0 1 1 2 n−1 n 0 n

Problem 4. (Shortlist JBMO 2012) Solve the following equation for x, y, z ∈ N.

 2  2  2
x y z 27
1+ + 1+ + 1+ = .
y+z z+x x+y 4

Solution. By Cauchy–Schwarz inequality we have

 2
x y z
27

x
2 
y
2 
z
2 1+ y+z
+1+ z+x
+1+ x+y
= 1+ + 1+ + 1+ ≥ ,
4 y+z z+x x+y 3
that is, x, y and z satisfy the inequality
 2
81 x y z
≥ 3+ + + . (2.6)
4 y+z z+x x+y
On the other hand, due to Nesbitt’s inequality we get
x y z 3
+ + ≥ . (2.7)
y+z z+x x+y 2
Consequently it holds
 2  2
x y z 3 81
3+ + + ≥ 3+ = (2.8)
y+z z+x x+y 2 4
Comparing (2.6) and (2.8) we easily conclude that
 2
x y z 81
3+ + + = .
y+z z+x x+y 4
The equality case in the inequality (2.6) holds if and only if x = y = z.
In the end we briefly check that, if x = y = z, then
 2  2  2  2
x y z x 27
1+ + 1+ + 1+ =3· 1+ = .
y+z z+x x+y x+x 4
The triple (x, x, x), where x ∈ N, is a solution of the given equation. □

Problem 5. (JMMO 2012) Let a, b and c be positive real numbers. Prove the inequality
1 √ √ √ 1 1 1
( a + b + c) + + + ≥ 3.
2 1+a 1+b 1+c
18 Chapter 2. SOLUTIONS

Solution. Since
 
1 1 1 a b c
+ + =3− + + ,
1+a 1+b 1+c 1+a 1+b 1+c
it suffices to show that
1 √ √ √ a b c
( a + b + c) ≥ + + . (2.9)
2 1+a 1+b 1+c
√ √ √
a a a b b c c
Based on AM-GM inequality we get ≤ √ = , ≤ and ≤ .
1+a 2 a 2 1+b 2 1+c 2
Summing the last three inequalities we obtain the desired inequality in (2.9). □

Problem 6. (MMO 2012) If a, b, c and d are positive real numbers such that abcd = 1,
then prove that the following inequality holds
1 1 1 1
+ + + ≤ 2.
bc + cd + da − 1 ab + cd + da − 1 ab + bc + da − 1 ab + bc + cd − 1
When does equality hold?
1 1
Solution 1. Since abcd = 1 it follows bc = . By using x + ≥ 2, for x > 0, we have
ad x
1
bc + cd + da − 1 = + ad + cd − 1 ≥ 2 + cd − 1 = 1 + cd ⇔
ad
1 1
≤ . (2.10)
bc + cd + da − 1 1 + cd
Analogously we get
1 1
≤ (2.11)
ab + cd + da − 1 1 + ad
1 1
≤ (2.12)
ab + bc + da − 1 1 + ab
and
1 1
≤ . (2.13)
ab + bc + cd − 1 1 + bc
By summing the inequalities in (2.10), (2.11), (2.12) and (2.13) we get:
1 1 1 1
+ + + ≤
bc + cd + da − 1 ab + cd + da − 1 ab + bc + da − 1 ab + bc + cd + −1
1 1 1 1 1 1 1 1
≤ + + + = + + + =
1 + cd 1 + ad 1 + ab 1 + bc 1 1 1 + ab 1 + bc
1+ 1+
ab bc
ab 1 bc 1
= + + + = 2.
1 + ab 1 + ab 1 + bc 1 + bc
Equality occurs if and only if a = b = c = d = 1. □
2.1. ALGEBRA 19

Solution 2. By multiplying 1 + bc + cd + da and 1 + ab we get

b a
(1+bc+cd+da)(1+ab) = 1+bc+cd+da+ab+ab2 c+abc+a2 bd = 2+ab+bc+cd+da+ + .
d c
(2.14)
From AM-GM applied to the numbers db and ac we get
√
r
b a ab
+ ≥2· = 2 ab · ab = 2ab. (2.15)
d c cd
Replacing (2.15) in (2.14) we get
1 + ab 1
(1+bc+cd+da)(1+ab) ≥ 2+2ab+ab+bc+cd+da ⇔ ≥
ab + bc + cd + da bc + cd + da − 1
(2.16)
Similarly we get
1 + bc 1
≥ (2.17)
ab + bc + cd + da ab + cd + da − 1
1 + cd 1
≥ (2.18)
ab + bc + cd + da ab + bc + da − 1
1 + da 1
≥ (2.19)
ab + bc + cd + da ab + bc + cd − 1
Summing (2.16), (2.17), (2.18) and (2.19) we obtain
4 + ab + bc + cd + da 1 1
≥ + +
ab + bc + cd + da bc + cd + da − 1 ab + cd + da − 1
1 1
+ + .
ab + bc + da − 1 ab + bc + cd − 1
p
Since ab + bc + cd + da ≥ 4 4 (abcd)2 = 4, we get
1 1 1 1
+ + + ≤
bc + cd + da − 1 ab + cd + da − 1 ab + bc + da − 1 ab + bc + cd − 1
4 4
≤1+ ≤ 1 + = 2.
ab + bc + cd + da 4
□

Problem 7. (MMO 2013) Let x, y and z be positive real numbers such that x4 +y 4 +z 4 =
3. Prove that
9 9 9
2 4 6
+ 4 6 2
+ 6 ≤ x6 + y 6 + z 6 + 6.
x +y +z x +y +z x + y2 + z4

Solution. By using Cauchy-Schwarz inequality for the triplets (x, y 2 , z 3 ) and (x3 , y 2 , z) we
get
1 x6 + y 4 + z 2
9 = (x4 + y 4 + z 4 )2 ≤ (x2 + y 4 + z 6 )(x6 + y 4 + z 2 ), i.e. ≤ . (2.20)
x2 + y 4 + z 6 9
20 Chapter 2. SOLUTIONS

Analogously, by applying the Cauchy-Schwarz inequality to the triplets (x2 , y 3 , z) and

(x2 , y, z 3 ), and respectively to (x3 , y, z 2 ) and (x, y 3 , z 2 ), we obtain the following inequal-
ities:
1 x4 + y 2 + z 6
≤ (2.21)
x4 + y 6 + z 2 9
and
1 x2 + y 6 + z 4
≤ . (2.22)
x6 + y 2 + z 4 9
Summing (2.20), (2.21) and (2.22) gives

1 1 1 x6 + y 6 + z 6 + x4 + y 4 + z 4 + x2 + y 2 + z 2
+ + ≤ .
x 2 + y 4 + z 6 x4 + y 6 + z 2 x6 + y 2 + z 4 9
(2.23)
In the end we apply the inequality between arithmetic and quadratic mean for the
numbers x2 , y 2 , z 2 . We have
r
x4 + y 4 + z 4
x2 + y 2 + z 2 ≤ 3 = 3. (2.24)
3
Substituting (2.24) into (2.23) leads to the desired inequality. □

Problem 8. (JMMO 2013) Let a, b and c be positive real numbers satisfying a+b+c+2 =
abc. Prove that
a b c
+ + ≥ 2.
b+1 c+1 a+1
Solution. We derive the following identity:

(a + 1)(b + 1) + (a + 1)(c + 1) + (b + 1)(c + 1) = a + b + c + (a + b + c + 2) + ab + ac + bc + 1 =

= a + b + c + abc + ab + ac + bc + 1 = (a + 1)(b + 1)(c + 1).

Thus
(a + 1)(b + 1) + (b + 1)(c + 1) + (c + 1)(a + 1)
= 1. (2.25)
(a + 1)(b + 1)(c + 1)
Now, from AM-GM we obtain:
 
a b c a+1 b+1 c+1 1 1 1
+ + = + + − + + ≥
b+1 c+1 a+1 b+1 c+1 a+1 b+1 c+1 a+1
s  
3 (a + 1)(b + 1)(c + 1) 1 1 1
≥3 − + + =
(b + 1)(c + 1)(a + 1) b+1 c+1 a+1
(a + 1)(b + 1)(c + 1)
=3− = 3 − 1 = 2.
(b + 1)(c + 1)(a + 1)
Equality occurs if and only if a = b = c = 2.
□
2.1. ALGEBRA 21

Problem 9. (Shortlist BMO 2013) Let a, b, c and d be positive real numbers so that
abcd = 41 . Prove that holds
  
a 16c 4 b d 1 81
16ac + 2 + 2 + bd + 2
+ 2 + ≥ .
c b a d ac 256d c b a 64bd 4

Solution. First we observe the following two identities:

  
a 16c 4 1 1
16ac + 2 + 2 + = a+ 2 16c + 2 , (2.26)
c b a d ac ad cb
  
b d 1 1 1
bd + + + = b+ 2 d+ . (2.27)
256d2 c b2 a 64bd ba 256d2 c
According to (2.26) and (2.27) it suffices to show that
    
1 1 1 1 81
a+ 2 16c + 2 b+ 2 d+ 2
≥ . (2.28)
ad cb ba 256d c 4

From the inequality between arithmetic and geometric mean for the numbers a2 , a2 , a12 d we
q q
get a + a12 d ≥ 3 3 4d
1
. Similarly we prove that b + b21a ≥ 3 3 4a
1
.
Now, from the inequality between q the arithmetic and geometric mean for
q the numbers 8c, 8c
1 1
and c2 b
we get 16c + c2 b
≥ 12 3 1b . Analogously it holds d + 1
256d2 c
≥ 3
8
3 1
2c
.
Finally, if we multiply the last four inequalities and if we use that abcd = 14 , we get:
    
1 1 1 1
a+ 2 16c + 2 b+ 2 d+ ≥
ad cb ba 256d2 c
r r r r
3 1 3 1 3 1 33 1 81
≥3 · 12 ·3 · = .
4d b 4a 8 2c 4
□

Problem 10. (IMIO 2013) Let a, b and c be positive real numbers so that ab + bc + ca =
6abc. Prove that holds  
1 a b c
a+b+c≥1+ + + .
6 c a b
1 1 1
Solution. From ab + bc + ca = 6abc it follows a
+ b
+ c
= 6. By multiplying the last
identity by a, b and c, respectively, we obtain:
a a b b c c
1+ + = 6a, 1 + + = 6b and 1 + + = 6c. (2.29)
b c a c a b
Summing the identities in (2.29) gives
   
a b c a b c
3+ + + + + + = 6(a + b + c). (2.30)
b c a c a b
22 Chapter 2. SOLUTIONS

Moreover, it holds r
a b c 3 a b c
+ + ≥3 · · = 3. (2.31)
b c a b c a
Combining (2.30) and (2.31) we get
   
a b c 1 a b c
6(a + b + c) ≥ 6 + + + ⇔a+b+c≥1+ + + .
c a b 6 c a b
□

Problem 11. (Shortlist JMMO 2013) Let 0 ≤ a, b, c ≤ 1 and a + b + c = 2. Prove that

a2 b + b2 c + c2 a ≤ 1.

Solution. It obviously holds that

(a − 1)2 (b − 1) ≤ 0, (b − 1)2 (c − 1) ≤ 0 and (c − 1)2 (a − 1) ≤ 0.

Therefore
a2 b ≤ a2 + 2ab − 2a − b + 1, b2 c ≤ b2 + 2bc − 2b − c + 1
and
c2 a ≤ c2 + 2ca − 2c − a + 1.
Summing the last three inequalities we get

a2 b + b2 c + c2 a ≤ a2 + b2 + c2 + 2(ab + bc + ca) − 3(a + b + c) + 3 =

= (a + b + c)2 − 3(a + b + c) + 3 = 1.
□

Problem 12. (Shortlist JMMO 2014) Let a, b and c be positive real numbers such that
a + b + c = 3. Prove that

a5 + b5 + c5 + 6 ≥ 3(a2 + b2 + c2 ).

Solution. We transform the left side of the inequality as follows:

a5 + b5 + c5 + 6 = a5 + b5 + c5 + a + b + c + 3 = (a5 + a + 1) + (b5 + b + 1) + (c5 + c + 1).

From AM-GM for the positive numbers a5 , a and 1 we have

√
3
a5 + a + 1 ≥ 3 a5 · a · 1 = 3a2 (2.32)

Analogously we get
b5 + b + 1 ≥ 3b2 (2.33)
2.1. ALGEBRA 23

and
c5 + c + 1 ≥ 3c2 (2.34)
By summing (2.32), (2.33) and (2.33) we get the required inequality. □

Problem 13. (Shortlist JMMO 2014) Let a, b and c be positive real numbers such that
a + b + c = 1. Prove that holds

(ab + 2ac + 3bc)3 ≥ (2a)5 · (3b)4 · (3c)3 .

Solution. From GM-HM for the positive real numbers 2a, 3b and 6c we obtain that
r
√3 3 (2a)(3b)(6c) 1 3 1 18abc
abc = ≥ √ · = √ · .
36 3
36 1 + 1 + 1 3
36 3bc + 2ac + ab
2a 3b 6c
Thus
162a3 b3 c3
abc ≥ . (2.35)
(3bc + 2ac + ab)3
a a a b b
On other hand, from AM-GM for the positive numbers
, , , , , c we have
3 3 3 2 2
r
3 2
a a a b b 6 a b c
1=a+b+c= + + + + +c≥6 ,
3 3 3 2 2 108
from where we derive that
108
a3 b 2 c ≤ . (2.36)
66
From (2.35) and (2.36) we get:

108 3 2 2 2 162a3 b3 c3 162a5 b4 c3

≥ a b c = a b · (abc) ≥ a b · = . (2.37)
66 (3bc + 2ac + ab)3 (3bc + 2ac + ab)3

From the last inequality it follows that (3bc + 2ac + ab)3 ≥ (2a)5 · (3b)4 · (3c)3 . Equality
1 1 1
occurs if and only if 2a = 3b = 6c, i.e., for a = , b = , c = . □
2 3 6

Problem 14. (MMO 2014) Let a, b and c be positive real numbers such that a+b+c = 4.
Prove that  
1 1 1 1 1 1
+ + ≥8 + + .
a−1 b−1 c−1 a+b b+c c+a

1 8 1 8 3(4 − 3a)
Solution. Since − = − = , it suffices to show the
a−1 b+c a−1 4−a (a − 1)(4 − a)
following inequality
(4 − 3a) (4 − 3b) (4 − 3c)
+ + ≥ 0. (2.38)
(a − 1)(4 − a) (b − 1)(4 − b) (c − 1)(4 − c)
24 Chapter 2. SOLUTIONS

Without loss of generality we assume a ≥ b ≥ c. Hence 4 − 3a ≤ 4 − 3b ≤ 4 − 3c.

1 1 1
From 1 < a, b, c < 4 we get that , and are
(a − 1)(4 − a) (b − 1)(4 − b) (c − 1)(4 − c)
positive real numbers. Now we will show that (a − 1)(4 − a) ≥ (b − 1)(4 − b) ≥ (c − 1)(4 − c).
We have

(a − 1)(4 − a) ≥ (b − 1)(4 − b) ⇔ 5a − a2 ≥ 5b − b2 ⇔ (a − b)(a + b − 5) ≤ 0.

From a ≥ b and a+b−5 = 4−c−5 = −1−c < 0 we conclude (a−1)(4−a) ≥ (b−1)(4−b).

1 1 1
Thus, it holds ≤ ≤ .
(a − 1)(4 − a) (b − 1)(4 − b) (c − 1)(4 − c)
Now from Chebyshev’s inequality we have

(4 − 3a) (4 − 3b) (4 − 3c)

+ + ≥
(a − 1)(4 − a) (b − 1)(4 − b) (c − 1)(4 − c)
   
4 − 3a + 4 − 3b + 4 − 3c 1 1 1
≥ · + + = 0.
3 (a − 1)(4 − a) (b − 1)(4 − b) (c − 1)(4 − c)
Clearly, the inequality in (2.38) is satisfied. □

Problem 15. (Shortlist JMMO 2014) Let a, b and c be the side lengths of a triangle
with an area P and let a + b + c = 6. Prove the inequality
√ √ √ √
3−a 3−b 3−c 3 3
+ + ≤ .
1 + (3 − b)(3 − c) 1 + (3 − c)(3 − a) 1 + (3 − a)(3 − b) 2P

Solution. From a + b + c = 6 we have

r r
√ a + b + c b+c−a
−a
3−a 2
=  =  2  .
1 + (3 − b)(3 − c) a+b+c a+b+c a+c−b a+b−c
1+ −b −c 1+ ·
2 2 2 2
(2.39)
Analogously we obtain
r
√ a+c−b
3−b
=  2  , (2.40)
1 + (3 − a)(3 − c) b+c−a a+b−c
1+ ·
2 2

and r
√ a+b−c
3−c
=  2  . (2.41)
1 + (3 − a)(3 − b) b+c−a a+c−b
1+ ·
2 2
2.1. ALGEBRA 25

We introduce the positive variables x, y and z as follows: a = y + z, b = x + z and c = x + y.

b+c−a a+c−b a+b−c
It is easy to derive that x = ,y= and z = .
2 2 2
Hence, the required inequality is equivalent to
√ √ √ √
x y z 3 3
+ + ≤ . (2.42)
1 + yz 1 + xz 1 + xy 2P
√
From AM-GM for the positive real numbers 1 and yz we get 1 + yz ≥ 2 yz i.e.
1 1 1 1 1 1
≤ √ . Similarly we have ≤ √ and ≤ √ .
1 + yz 2 yz 1 + xz 2 xz 1 + xy 2 xy
Now we bound the left side of (2.42) as follows
√ √ √ √ √ √
x y z x y z x+y+z 3
+ + ≤ √ + √ + √ = √ = √ . (2.43)
1 + yz 1 + xz 1 + xy 2 yz 2 xz 2 xy 2 xyz 2 xyz

By Heron’s formula we have

p p √ P
P = s(s − a)(s − b)(s − c) = 3xyz i.e. xyz = √ . (2.44)
3

The required inequality follows from (2.43) and (2.44). □

Problem 16. (Shortlist BMO 2015) Let a, b, c be side lengths of a triangle and let
ma , mb , mc be the medians at the corresponding sides. Prove that
   b 
b c  a  c  a
ma −1 − 1 + mb −1 − 1 + mc −1 − 1 ≥ 0.
a a b b c c

Solution. Let us suppose that a ≥ b ≥ c. Hence ma ≤ mb ≤ mc . Since a − c ≥ a − b and

mc ≥ mb we get
   b 
b c  a  c  a
ma −1 − 1 + mb −1 − 1 + mc −1 −1 =
a a b b c c

ma (a − b)(a − c) mb (b − a)(b − c) mc (c − a)(c − b)

= + + ≥
a2 b2 c2
mc (a − c)(b − c) mb (a − b)(b − c)
≥ − ≥ 0.
c2 b2
□

Problem 17. (Shortlist MMO 2015) Let a, b, c be positive real numbers greater than
1 so that (a − 1)(b − 1)(c − 1) = 1. Prove that
r r r
5 a + b 5 b + c c+a
+ + 5 ≥ 3.
4 4 4
26 Chapter 2. SOLUTIONS
√ √ √
Solution. Let x, y, z be real numbers such that x = 5 a − 1, y = 5 b − 1 and z = 5 c − 1.
From the given conditions, it follows that x, y, z are positive real numbers so that xyz = 1.
It suffices to prove the equivalent inequality
r r r
5 5 5 5 5 5
5 x + y + 2 5 y + z + 2 5 z + x + 2
+ + ≥ 3. (2.45)
4 4 4
Using the inequality between power means of order 5 and 1 for the positive numbers x, y, 1
and 1 we get r r
5 5 5 5 5 5
5 x + y + 2 5 x + y + 1 + 1 x+y+2
= ≥ . (2.46)
4 4 4
Analogously, it holds r
5 5
5 y + z + 2 y+z+2
≥ (2.47)
4 4
and r
5 5
5 z + x + 2 z+x+2
≥ . (2.48)
4 4
Summing the inequalities in (2.46), (2.47) and (2.48) we get

r r r √
5 x5 + y 5 + 2 5 y5 + z5 + 2 5 z 5 + x5 + 2 x+y+z 3 3 3 xyz 3
+ + ≥ + ≥ + = 3.
4 4 4 2 2 2 2
□

Problem 18. (Vojtěch Jarnı́k 2016) Let a, b and c be positive real numbers such that
a + b + c = 1. Show that
   
1 1 1 1 1 1
+ + + ≥ 1728.
a bc b ca c ab

Solution 1. By using the AM-GM inequality we deduce

 3
1 1 1 1 1 1 1 1 a+b+c
+ = + + + ≥ 4√
4
and = ≥ abc.
a bc a 3bc 3bc 3bc 3
27ab c 3 27 3

Therefore,
   
1 1 1 1 1 1 1 1 1
+ + + ≥ 64 · √
4
√
4
√
4
=
a bc b ca c ab 27ab3 c3 27a3 bc3 27a3 b3 c
64 64 √
4
= p ≥ p = 64 312 = 64 · 27 = 1728.
4 9
3 (abc)7 4 9 −3
3 (3 ) 7

□
2.1. ALGEBRA 27
1 1 1
+ + we get
Solution 2. If we replace 1 with a + b + c and k =
a b c
     
1 a+b+c 1 a+b+c 1 a+b+c 1 1 1 a
+ + + = + + + ·
a bc b ca c ab a b c bc
     
1 1 1 b 1 1 1 c a b  c
· + + + + + + = k+ k+ k+ =
a b c ca a b c ab bc ca ab
   
3 2 c b a 1 1 1 1
=k +k + + +k 2
+ 2+ 2 + . (2.49)
ab ca bc a b c abc
From the inequality between arithmetic and harmonic means for the positive numbers a, b
and c we obtain
1 a+b+c 3 3
= ≥ = ⇔ k ≥ 9.
3 3 1 1 1 k
+ +
a b c
1
From Solution 1 we already know that ≥ 27. Replacing this bound in (2.49) we obtain
abc
3 27 1
L.H.S ≥ 93 + 92 · √
3
+p + ≥ 93 + 93 + 27 · 9 + 27 = 1728.
abc 3
(abc)2 abc

Problem 19. (Shortlist Vojtěch Jarnı́k 2017) Let k ≥ 3 be an odd number and let
Hm (x) be the Dickson polynomial defined as follows: H0 (x) = 1, H1 (x) = x, Hi+2 (x) =
xHi+1 (x) − (k − 1)Hi (x), with i ≥ 0. Prove that the polynomial Hm (x) − 2 is irreducible.
Solution. We prove, using induction on m ≥ 3, that Hm (x) = xm + (k − 1)Pm−2 (x), where
Pm−2 (x) is an integer polynomial of degree m − 2. We calculate H3 (x) = x3 − 2(k − 1)x.
Let us suppose that the above formula holds for Hm−1 (x) and Hm−2 (x). That yields

Hm (x) = x(xm−1 +(k −1)Pm−3 (x))−(k −1)(xm−2 +(k −1)Pm−4 (x)) = xm +(k −1)Pm−2 (x).

Therefore, Hm (x) − 2 = xm + (k − 1)Pm−2 (x) − 2. By the induction hypothesis, it follows

m m
that Hm (0) = (−1) 2 (k − 1) 2 for an even m, and Hm (0) = 0 for an odd m. Hence, for an
m m
even m(≥ 4) |(−1) 2 (k − 1) 2 − 2| is not divisible by 22 , and clearly for an odd m(≥ 3),
−2 is not divisible by 22 . Since k − 1 is even, it follows that every coefficient on Hm (x) − 2
except for the coefficient 1 of xm is divisible by 2. Thus, the conditions of the Eisenstein’s
criterion are satisfied, and Hm (x) − 2 is irreducible. □

Problem 20. (National Competition 2020, Macedonia) Let a1 , a2 , . . . , a2020 be pos-

itive real numbers such that a1 ≤ a2 ≤ . . . ≤ a2020 , a1 + a2 + . . . + a2020 = 2020 and
a21 + a22 + . . . + a22020 = 2021. Prove that
1
a2019 ≥ 1 − √ .
2019 · 2020
28 Chapter 2. SOLUTIONS

Solution. From the inequality between quadratic and arithmetic mean for the numbers
a1 , a2 , . . . , a2019 we get
(a1 + a2 + . . . + a2019 )2 (2020 − a2020 )2
2021 − a22020 = a21 + a22 + . . . + a22019 ≥ = . (2.50)
2019 2019
The inequality in (2.50) is equivalent to the inequality
2019
(a2020 − 1)2 ≤ . (2.51)
2020
From the conditions we easilyqconclude that a2020 > 1. Now, by taking the square root
of (2.51) we get a2020 ≤ 1 + 20192020
. From a1 ≤ a2 ≤ . . . ≤ a2020 , and from the previous
inequality we get
r
2019
2020 = (a1 + a2 + . . . + a2019 ) + a2020 ≤ 2019 · a2019 + 1 + ,
2020
that is,
1
a2019 ≥ 1 − √ .
2019 · 2020
□

Problem 21. (University of Primorska Scholarship Competition in Kosovo,

2021) Let x1 ≥ x2 ≥ . . . ≥ xn > 0 be positive real numbers such that x1 + x2 + . . . + xn =
1
x1
+ x12 + . . . + x1n . Prove that

(x1 − xn )2
x1 + x2 + . . . + xn ≥ n + .
2(x21 + x2n )

Solution. From Cauchy-Schwarz inequality we have

   
2 1 1 1 1 1
(x1 +. . .+xn ) = (x1 +. . .+xn ) + ... + = (x1 +x2 +. . .+xn ) + + ... + ≥
x1 xn xn x2 x1
r r r 2 r r 2
x1 x2 xn x1 xn
≥ + + ... + = + +n−2 .
xn x2 x1 xn x1
q q
(x1 −xn )2
It suffices to show that xxn1 + xxn1 ≥ 2 + 2(x 2 +x2 ) which is equivalent to the inequality
1 n

 2  
x1 x1
r
x1
r
xn xn
−2 xn
+1
+ ≥2+    . (2.52)
xn x1 x1
2
2 xn
+1

x1
Let t2 = xn
,t > 0. The inequality in (2.52) is equivalent to the inequalities

1 t4 − 2t2 + 1
t+ ≥2+ ⇔ 2t6 − 5t5 + 2t4 + 2t3 + 2t2 − 5t + 2 ≥ 0.
t 2(t4 + 1)
2.1. ALGEBRA 29

Now we verify that 2t6 − 5t5 + 2t4 + 2t3 + 2t2 − 5t + 2 = (t − 1)4 (2t2 + 3t + 2) ≥ 0. □
 
1 1
Remark: From AM-HM we get (x1 + . . . + xn ) = (x1 + . . . + xn ) x1 + . . . + xn ≥ n2 ,
2

which leads to x1 + . . . + xn ≥ n. This is a well-known lower bound for x1 + . . . + xn , when

x1 + x2 + . . . + xn = x11 + x12 + . . . + x1n .

Problem 22. (Regional Competition 2021, Macedonia) Let a1 , a2 , . . . , an be positive

real numbers such that a1 a2 · · · an = 1. Prove that

(n − 1)(a21 + a22 + . . . + a2n − 1)2 ≥ (a1 + . . . + an − 1)2 .

Solution. Without loss of generality we assume a1 ≥ a2 ≥ . . . ≥ an . Since a1 a2 · · · an = 1,

it holds a1 ≥ 1 ≥ an . Let a21 + a2n − 1 = b21 . From the inequality between the quadratic and
arithmetic mean for the positive numbers b1 , a2 , . . . , an−1 we get

(n − 1)(a21 + a22 + . . . + a2n − 1) = (n − 1)(b21 + a22 + . . . + a2n−1 ) ≥ (b1 + a2 + . . . + an−1 )2 .

p
It suffices to prove that b1 ≥ a1 + an − 1, which is equivalent to a21 + a2n − 1 ≥ a1 + an − 1.
By squaring the last inequality we get the inequality 2(a1 − 1)(1 − an ) ≥ 0, which occurs
due to the above conditions. □

Problem 23. (Vojtěch Jarnı́k 2022) Let x1 , . . . , xn be given real numbers with 0 <
m ≤ xi ≤ M for each i ∈ {1, . . . , n}. Let X be the discrete random variable uniformly
distributed on {x1 , . . . , xn }. The mean µ and the variance σ 2 of X are defined as

x1 + . . . + xn (x1 − µ(X))2 + . . . + (xn − µ(X))2

µ(X) = and σ 2 (X) = .
n n
By X 2 denote the discrete random variable uniformly distributed on {x21 , . . . , x2n }.
Prove that  m 2
σ 2 (X) ≥ σ 2 (X 2 ).
2M 2
Solution 1. We use the following lemma which is equivalent to inequality (2.52).

Lemma 2.1. If x and y are strictly positive real numbers, then

r
(x − y)2
r
x y
+ ≥2+ .
y x 2(x2 + y 2 )

x2 1
Let ai = x2 +...+x
i
2 and bi = n for i = 1, . . . , n. Applying the above lemma for x = ai and
1 n
y = bi we obtain

x2i (x2i n − (x21 + . . . + x2n ))2

 
1 xi
2 2
+ ≥ 2+ 4 2 2 2 2
p . (2.53)
x1 + . . . + xn n 2(xi n + (x1 + . . . + xn ) ) (n(x21 + . . . + x2n )
30 Chapter 2. SOLUTIONS

Now if we sum up the n obtained inequalities in (2.53) we get

n n 2 2
2 X m 1 X
2 x1 + . . . + x n 2
2≥ p xi + p · · (x i − ) ⇔
n(x21 + . . . + x2n ) i=1 n(x21 + . . . + x2n ) 2(M 4 + µ2 (X 2 )) i=1 n
r Pn
x21 + . . . + x2n xi m · σ 2 (X 2 ) m · σ 2 (X 2 )
≥ i=1 + = µ(X) + ⇔
n n 4(M 4 + µ2 (X 2 )) 4(M 4 + µ2 (X 2 ))
p m · σ 2 (X 2 ) m · σ 2 (X 2 )
µ(X 2 ) ≥ µ(X) + = µ(X) + .
4(M 4 + M 4 ) 8M 4
In the end we get
p p mσ 2 (X 2 )  m 2
σ 2 (X) = ( µ(X 2 ) − µ(X))( µ(X 2 ) + µ(X)) ≥ · 2m = · σ 2 (X 2 ).
8M 4 2M 2
□

Solution 2. We have
(x1 − µ(X))2 + . . . + (xn − µ(X))2 (x2 + . . . + x2n ) − nµ2 (X)
σ 2 (X) = = 1 =
n n
2
P
n(x21 + . . . + x2n ) − (x1 + . . . + xn )2 1≤k<l≤n (xk − xl )
= = .
n2 n2
It suffices to show that for each k ̸= l
 m 2 1 m
(xk − xl )2 ≥ 2
(x2k − x2l )2 ⇔ ≥ .
2M xk + xl 2M 2
Using xk + xl ≤ 2M and m ≤ M we derive
1 1 m
≥ ≥ .
xk + xl 2M 2M 2
□

Problem 24. (American Mathematical Monthly 2022) Let n, k be positive integers

with n ≥ 3 and let p(x) = xn + xn−1 + · · · + x − k.

(a) Prove that the roots of p(x) in the complex plane are simple.

(b) Prove that if k ≥ n + 1, then p(x) has at least one root with negative real part and
nonzero imaginary part.
2.1. ALGEBRA 31

Solution.
a) In the proof we use Descartes’ rule of signs: the number of positive roots of a single-
variable polynomial with real coefficients is equal to the number of sign differences
between consecutive nonzero coefficients, or is less than it by an even number; simi-
larly, the number of negative roots is the number of sign changes after multiplying the
coefficients of odd-power terms by −1, or fewer than it by an even number. According
to Descartes’ rule of signs, the polynomial p(x) has exactly one positive real root. If
θ is a positive real root of p(x) with multiplicity greater than 1, then θ also is a root
of its derivative nxn−1 + (n − 1)xn−2 + . . . + 2x + 1, which is impossible. Thus, the
unique positive real root of the polynomial p(x) is simple. If k ̸= n + 1, then the
polynomial p(x) has the same roots as the equation
xn+1 − kx + k − 1 = 0, (2.54)
except for the extra root of (2.54) x = 1; if k = n + 1, then x = 1 is a root of p(x)
with multiplicity 1 and a root of (2.54) with multiplicity 2. Using Descartes’ rule of
signs we have that the equation xn+1 − kx + k − 1 = 0 has at most two positive real
roots (one of them is x = 1) and at most one negative real root. If we suppose that
θ is a root of (2.54) with multiplicity greater than 1, we deduce that θ also satisfies
the first derivative of (2.54), that is, (n + 1)xn − k. Combining (2.54) and the identity
(n + 1)xn − k = 0 we obtain θ = (k−1)(n+1) nk
≥ 0, which yields that there exist no
negative real root nor complex roots of p(x) with multiplicity greater than 1.
b) Now, let k ≥ n + 1 ≥ 4. Clearly p(1) = n + 1 − k < 0, and therefore, the unique
positive root of p(x) belongs to the interval (1, ∞), and we denote it by θ1 . Descartes’
rule asserts that the polynomial p(x) has no negative real root when n is an odd
number. We will prove the existence of a complex root of p(x) with negative real part
when p(x) has a negative real root θ2 ; in such case n must be an even number. The
case when p(x) has no negative root can be handled similarly.
Let θj = pj +qj i, with 3 ≤ j ≤ n, be the complex roots of p(x). By way of contradiction
we assume that pj ≥ 0, for all 3 ≤ j ≤ n. Using the fact that the complex roots come
in conjugate pairs and applying Vieta’s formulas to the polynomial p(x), we obtain
−1 = θ1 + θ2 + . . . + θn = θ1 + θ2 + (p3 + . . . + pn ) ≥ θ1 + θ2 .
On the other hand, since n is an even number and θ1 is a positive, using the inequality
(1 + θ1 )t > (1 + θ1 )t−1 + θ1t + θ1t−1 , for 2 ≤ t ≤ n, we can deduce
p(−1 − θ1 ) = (−1 − θ1 )n + (−1 − θ1 )n−1 + . . . + (−1 − θ1 ) + 1 − k =
= (1 + θ1 )n − (1 + θ1 )n−1 + . . . + (1 + θ1 )2 − (1 + θ1 ) + 1 − k > θ1n + θ1n−1 + . . . + 1 − k = 0.
Since p(0) < 0 and p(−1 − θ1 ) > 0, it follows that the negative root of p(x) belongs
to the interval (−1 − θ1 , 0), that is, −1 − θ1 < θ2 < 0. Thus θ1 + θ2 > −1, which is in
contradiction to θ1 + θ2 ≤ −1.
□
32 Chapter 2. SOLUTIONS

2.2 COMBINATORICS
Problem 25. (National Competition 2011, Macedonia) The interior of a convex 12-
gon is divided into ten triangles so that no two triangles overlap, with four of these being
special triangles. Each triangle has a non-zero real number written inside it, and for each
special triangle, the number written inside it is the product of the numbers written inside
the triangles that enclose it. Can the product of all ten written numbers in the 12-gon be
constant? (A triangle is called a special triangle if a triangle has been constructed over
each of its sides).

Solution. Let a1 , a2 , . . . , a10 be non-zero numbers written in the ten triangles.

From the conditions we get

a7 = a1 · a2 · a10 , a8 = a3 · a4 · a10 , a9 = a5 · a6 · a10 and a10 = a7 · a8 · a9 .

It holds

a10 = a7 · a8 · a9 = (a1 · a2 · a10 ) · (a3 · a4 · a10 ) · (a5 · a6 · a10 ) = a1 · a2 · a3 · a4 · a5 · a6 · a310 ,

from where
a1 · a2 · a3 · a4 · a5 · a6 · a210 = 1. (2.55)
Using (2.55) we compute

a1 ·a2 ·a3 ·a4 ·a5 ·a6 ·a7 ·a8 ·a9 ·a10 = a1 ·a2 ·a3 ·a4 ·a5 ·a6 ·(a1 ·a2 ·a10 )·(a3 ·a4 ·a10 )·(a5 ·a6 ·a10 )·a10 =

= a21 · a22 · a23 · a24 · a25 · a26 · a410 = (a1 · a2 · a3 · a4 · a5 · a6 · a210 )2 = 12 = 1.

□

Problem 26. (Regional Competition 2012, Macedonia) Ten teams participated in a

football tournament, with each team playing a match against every other team. For every
win, a team earns 3 points; for every draw, each team earns 1 point; and for every defeat, a
2.2. COMBINATORICS 33

team earns 0 points. At the end of the tournament, the total number of points accumulated
across all teams was 120. Determine how many matches ended in a draw.
Solution. The total number of football matches is 9 + 8 + . . . + 2 + 1 = 45. With x we
denote the number of draws. Consequently, 45 − x matches ended with a winner. Therefore
we get the following equation
2x + 3(45 − x) = 120.
Solving this equation we get x = 15, i.e. 15 football matches have ended in a draw. □

Problem 27. (National Competition 2021, Macedonia) Among a group of n people,

there are exactly 2021 mutual friendships. Moreover, it is known that no person has more
than 45 friendships, and there is at least one person with exactly 45 friendships. Let xk
denote the number of friendships of the k-th person. Prove that

x21 + x22 + . . . + x2n ≤ 2024n + 2018.

Solution. Since there is a person with 45 friendships, it follows that n ≥ 46.

From the conditions of the problem we have xk ≤ 45 and x1 +x2 +. . .+xn = 2·2021 = 4042.
Therefore

x2k − xk = xk (xk − 1) ≤ 45 · (45 − 1) = 1980 for k = 1, 2, . . . , n. (2.56)

From x1 + . . . + xn = 4042 and (2.56) we have

x21 + x22 + . . . + x2n ≤ 1980n + (x1 + . . . + xn ) = 1980n + 4042.

It remains to prove that 1980n + 4042 ≤ 2024n + 2018, which is equivalent to n ≥ 46. □

Problem 28. (IMC 2023) Let T be a tree with n vertices; that is, a connected simple
graph on n vertices that contains no cycle. For every pair u, v of vertices, let d(u, v) denote
the distance between u and v, that is, the number of edges in the shortest path in T that
connects u and v. Consider the sums
X X 1
W (T ) = d(u, v) and H(T ) = .
{u,v}⊆V (T ) {u,v}⊆V (T )
d(u, v)
u̸=v u̸=v

Prove that
(n − 1)3 (n + 2)
W (T ) · H(T ) ≥ .
4

Solution. Let k = n2 and let x1 ≤ x2 ≤ . . . ≤ xk be the distances between the pairs of

vertices in the tree Tn . Thus

 
1 1 1
W (Tn ) · H(Tn ) = (x1 + x2 + . . . + xk ) · + + ... + .
x1 x2 xk
34 Chapter 2. SOLUTIONS

Since the tree has exactly n − 1 edges, there are exactly n − 1 pairs of vertices at distance
one, that is, x1 = x2 = . . . = xn−1 = 1. Thus
 
1 1 1
W (Tn ) · H(Tn ) = (n − 1 + xn + xn+1 + . . . + xk ) · n − 1 + + + ... + =
xn xn+1 xk
   
2 1 1
= (n − 1) + (n − 1) xn + + . . . + xk + +
xn xk
 
1 1
+(xn + . . . + xk ) + ... + .
xn xk
From Cauchy inequality we have
(n − 1)2 (n − 2)2
 
1 1
(xn + . . . + xk ) + ... + ≥ (1 + 1 + . . . + 1)2 = (k − n + 1)2 = .
xn xk 4
The equality holds if and only if xn = xn+1= . . . = xk . 
1 1
Now we minimize the expression xn + xn + . . . + xk + xk , where xi ∈ [2, n − 1].
It is clear that the minimal value is achieved for xn = xn+1 = . . . = xk = 2. Therefore we
get
(n − 1)2 (n − 2)2 (n − 1)3 (n + 2)
  
2 1
W (Tn )·H(Tn ) ≥ (n−1) +(n−1) 2+ (k − n + 1) + = .
2 4 4
The equality holds for x1 = . . . = xn−1 = 1 and xn = xn+1 = . . . = xk = 2, that is,
the smallest value is achieved for the tree where n − 1 pairs are at distance one, and the
remaining k − (n − 1) = (n−1)(n−2)
2
pairs are at distance two. The unique tree which satisfies
these conditions is the star graph Sn . In this case it holds
(n − 1)(n + 2) (n − 1)3 (n + 2)
W (Sn ) · H(Sn ) = (n − 1)2 · = .
4 4
□

Problem 29. (Vojtěch Jarnı́k 2024) Let n be a positive integer and let G be a simple
undirected graph on n vertices. Let di be the degree of its i-th vertex, i = 1, . . . , n. Denote
∆ = max di . Prove that if
Xn
d2i > n∆(n − ∆)
i=1

then G contains a triangle. (A graph is called simple if there are no loops and no multiple
edges between any pair of vertices.)
Solution. We prove the claim by contraposition assuming that the obtained graph G does
not contain triangles. If the i-th and the j-th vertex are connected we denote i ∼ j. In this
case holds di + dj ≤ n. Hence
n
X X
d2i = (di + dj ) ≤ mn, (2.57)
i=1 i∼j
2.2. COMBINATORICS 35

where m is the number of edges in the graph.

Let v be a vertex of G with maximum degree ∆. Since G is a triangle-free graph there are no
edges in the neighbourhood of v. Moreover, every vertex which is not in the neighborhood
of v has degree at most ∆. Therefore, the maximum number of edges of G is

m ≤ ∆ + (n − ∆ − 1)∆ = ∆(n − ∆). (2.58)

From (2.57) and (2.58) we get

n
X
d2i ≤ mn ≤ n∆(n − ∆). (2.59)
i=1

□
36 Chapter 2. SOLUTIONS

2.3 GEOMETRY

Problem 30. (Shortlist JMMO 2011) A triangle has side lengths a, b and c. Let R be
the radius of the circumscribed circle and let r be the radius of the inscribed circle in the
triangle. Prove that
ab + bc + ca
> 9r.
R

Solution. The proof relies on the formulas P = abc 4R

and P = rs, where P is the area of
the triangle. We rearrange the given inequality as follows:

P (ab + bc + ca) 9P 1 1 1 9
> i.e. + + > . (2.60)
abc 2(a + b + c) a b c 2(a + b + c)

In order to prove the inequality in (2.60) we apply the triangle inequalities

1 1 1 1 1 1
> , > and > .
a b+c b a+c c a+b

Summing the above inequalities and applying the inequality between arithmetic and har-
monic means we conclude

1 1 1 1 1 1 9
+ + > + + ≥ .
a b c b+c a+c a+b 2(a + b + c)

Problem 31. (National Competition 2012, Macedonia) In a triangle ABC, the

bisector of angle ∠A intersects the side BC at D so that |BD| = 2|DC|. The height CS
intersects the bisector AD at the point M. Prove that

m2c
|CM | = .
2hc

Solution. Since AD is a bisector of the angle ∠A, it follows that |AB|

|AC|
= |BD|
|DC|
= 2, that
is, |AB| = 2 · |AC|. Let N be a midpoint of AB and let P be the intersection between
the bisector AD and the median CN. Hence |AN | = 21 |AB| = |AC|, which implies that
AN C is an isosceles triangle. Since the triangles AN P and ACP are congruent we have
|CP | = |P N | = m2c and ∡AP C = 90◦ . Finally, we note that the right triangles CP M and
CSN are similar, since they share a common vertex C. Due to similarity we get |CN |
|CS|
= |CM
|CP |
|
,
mc |CM | m2c
that is, hc
= mc . Thus we get |CM | = 2hc
, as desired.
2
2.3. GEOMETRY 37

Problem 32. (Shortlist MMO 2013) In a triangle ABC, the bisector of the angle ∠A
intersects the side BC at point M such that |BM | : |M C| = 3 : 1. A line parallel to the side
AB is drawn through the vertex C, which intersects the extension of the bisector of ∠A at
the point D. Let S be the intersection between AC and BD, let R be a midpoint of the
side AD and let P be the intersection point between CR and BD. Prove that |P S| = |SD|.
Solution. Since AM is the bisector of the angle ∠A we get |AB|
|AC|
= |BM |
|M C|
= 3. Thus we set
|AC| = x and |AB| = 3x. Let ∡BAM = ∡M AC = ϕ. From AB∥CD we conclude that
∡CDA = ∡DAB = ϕ, that is, the triangle ADC is isosceles. Consequently |AC| = |CD| =
x. From Thales’s theorem we have
|AB| |AS| x
= ⇒ |SC| = .
|CD| |SC| 2
Since |AR| = |RD| we note that P R is a median in the triangle ADP. In the end, from
|AC| x
= x =2
|CS| 2

we conclude that the vertex C is the centroid of the triangle ADP , that is, AS is its median.
Consequently |DS| = |SP |.

□
38 Chapter 2. SOLUTIONS

2.4 LINEAR ALGEBRA

Problem 33. (Team Selection Test, Famnit, 2016) Let λ1 , λ2 , ..., λn ∈ R+ be eigen-
1
values of a real n × n matrix A. If det(A) = , then prove that
n!
n  k
X 1
1 + λk − ≥ n.
k=1
k

When does equality hold?

 k
1
Solution. Let ak = 1 + λk − . By using the inequality between AM-GM we have:
k
 k  k
(k − 1) + kλk (1 + 1 + · · · + 1) + kλk
ak = = ≥ kλk .
k k
Therefore we get:
n  k X n n
X 1 X p p
1 + λk − = ak ≥ kλk ≥ n · n n!(λ1 · λ2 · · · λn ) = n n n! · det(A) = n.
k=1
k k=1 k=1

1 1
The equality holds when λ1 = 2λ2 = ... = nλn i.e. for λ1 = 1, λ2 = , ..., λn = . □
2 n

Problem 34. (Shortlist IMC 2016) Let A and B be 3 × 3 matrices. Prove that

∥(AB − BA)3 ∥2 ≥ |det(AB − BA)|.

Solution 1. Applying the Cayley-Hamilton theorem to the matrix AB − BA we obtain

(AB − BA)3 − a(AB − BA)2 + b(AB − BA) − cI3 = O3 (2.61)

where a = trace(AB − BA) = 0 and c = det(AB − BA). Now we compute the traces of
the matrices on both sides of (2.61). We get

trace((AB − BA)3 ) = 3 · det(AB − BA).

By using that ∥M ∥2 ≥ λmax (M ) we get

|trace(AB − BA)3 |
∥(AB − BA)3 ∥2 ≥ |λmax ((AB − BA)3 )| ≥ = |det(AB − BA)|.
3
□

Solution 2. First, we find the singular value decomposition of AB − BA:

AB − BA = U ΣV ∗ .
2.4. LINEAR ALGEBRA 39

Since U and V ∗ are unitary matrices hold det(U ) = det(V ∗ ) = 1.

Therefore

|det(AB − BA)| = |det(U ΣV ∗ | = |detU ||detΣ||detV ∗ | = 1 · |detΣ| · 1 =

= |σ1 σ2 σ3 | ≤ (max{|σ1 |, |σ2 |, |σ3 |})3 = max{|σ1 |3 , |σ2 |3 , |σ3 |3 }.

We know that

λi (((AB − BA)∗ (AB − BA))3 ) = (λi ((AB − BA)∗ (AB − BA)))3 = ((σi )2 )3 ,
p
for i = 1, 2, 3. Using the fact that ∥A∥2 = λmax (A∗ A) = σmax (A) we get the required
inequality. □

Problem 35. (Shortlist IMC 2023) Let A be a real square matrix such that the sum of
each row is equal to d > 1, d ∈ N, and trace(Ai ) = 0, for each i = 1, . . . , 2023. Prove that

A2023 + A2022 + . . . + A + I ̸= J,

where J is the all-ones matrix.

Solution 1. Let us suppose that there is a square matrix A of size n for which A2023 +
A2022 + . . . + A + I = J. Clearly, the spectrum of A consists of d and some of the roots of
the equation x2023 + x2022 + . . . + x + 1 = 0. Since trace(Ak ) = 0 for each k = 1, . . . , 2023
we get
Xn−1
k
d + xki = 0. (2.62)
i=1
1 1
From xi = xi
= x2023
and (2.62) we obtain
i

n−1 n−1 n−1

X X 1 X
−d = xi = = x2023
i = −d2023 .
i=1
x
i=1 i i=1

Thus we have d = 0 or d = 1, a contradiction. □

Solution 2. Let us suppose that there is a matrix A for which A2023 +A2022 +. . .+A+I = J.
Clearly, d is an eigenvalue of A and d2023 +d2022 +. . .+d+1 = n is an eigenvalue of J. Thus,
the eigenvalues of J are n (with multiplicity 1) and 0 (with multiplicity n − 1). Moreover,
the eigenvalues of A are d and the roots of the polynomial p(x) = x2023 + . . . + x + 1. The
roots of p(x) are the 2024-th roots of unity, (except x = 1), so they are simple roots. We
use the following lemma derived by Feit and Higman.

Lemma 2.2. Let θ be a simple root of the polynomial f (x) and let fθ (x) = fx−θ (x)
. If M is a
trace(fθ (M ))
matrix satisfying f (M ) = O, then fθ (θ)
is the multiplicity of θ as a characteristic root
of M .
40 Chapter 2. SOLUTIONS

Since f (x) = (x − d)(x2023 + . . . + x + 1) is the minimal polynomial of A we have

f (A) = On .
Let θ be an eigenvalue of A different than d and 1. We use the above lemma to compute
the multiplicity of θ, m(θ). Let g(x) = x2024 − 1. We have

f (x) x − d g(x)
fθ (x) = = · (2.63)
x−θ x−θ x−1
θ−d θ−d
fθ (θ) = · D(g(θ)) = · 2024θ2023 . (2.64)
θ−1 θ−1
From (2.63) we have fθ (0) = dθ . Since trace(Ai ) = 0 for i = 1, . . . , 2023 we get
trace(fθ (A)) = fθ (0)tr(I) = dθ · n. By using the above lemma we have

nd θ − 1
m(θ) = · . (2.65)
2024 θ − d
θ−1
Since θ is a root of p(x), we pick θ to be a complex number. From d > 1, we get that θ−d is a
complex number as well, thus m(θ) is a complex number, which is not possible. Therefore,
for each matrix A which satisfies the given conditions holds A2023 +A2022 +. . .+A+I ̸= J. □

Problem 36. (Shortlist Vojtěch Jarnı́k 2023) Find the eigenvalues of the matrix
 
3 0 1 1 0 0 1 1 1 1
 0 3 0 1 1 1 0 1 1 1 
 
 1 0 3 0 1 1 1 0 1 1 
 
 1 1 0 3 0 1 1 1 0 1 
 
 0 1 1 0 3 1 1 1 1 0 
A=  0 1 1 1
.
 1 3 1 0 0 1  
 1 0 1 1 1 1 3 1 0 0 
 
 1 1 0 1 1 0 1 3 1 0 
 
 1 1 1 0 1 0 0 1 3 1 
1 1 1 1 0 1 0 0 1 3

Solution. Note that

 2
0 1 0 0 1 1 0 0 0 0

 1 0 1 0 0 0 1 0 0 0 


 0 1 0 1 0 0 0 1 0 0 


 0 0 1 0 1 0 0 0 1 0 
  2
 1 0 0 1 0 0 0 0 0 1  A 1 I5
A=  = = B2,

 1 0 0 0 0 0 0 1 1 0 
 I5 A 2

 0 1 0 0 0 0 0 0 1 1 


 0 0 1 0 0 1 0 0 0 1 

 0 0 0 1 0 1 1 0 0 0 
0 0 0 0 1 0 1 1 0 0
2.4. LINEAR ALGEBRA 41
     
0 1 0 0 1 0 0 1 1 0 1 0 0 0 0
 1 0 1 0 0   0 0 0 1 1   0 1 0 0 0 
     
 0 1 0 1 0
where A1 =   , A2 =  1 0 0 0 1  and I5 =  0 0 1 0 0 .
    
 0 0 1 0 1   1 1 0 0 0   0 0 0 1 0 
1 0 0 1 0 0 1 1 0 0 0 0 0 0 1
Lemma 2.3. If λ is an eigenvalue of B, then λ2 is an eigenvalue of B 2 .
Based on this lemma, it suffices to calculate the eigenvalues of the block matrix B. Since
the matrices I5 and A2 − λI5 commute, we have
 
A1 − λI5 I5
det(B − λI10 ) = det = det((A1 − λI5 ) · (A2 − λI5 ) − I5 · I5 ) =
I5 A2 − λI5
= det(A1 · A2 − λ(A1 + A2 ) + (λ2 − 1)I5 ).
It is easy to verify that A1 · A2 = A1 + A2 = J5 − I5 , where J5 is all-ones matrix. Thus
det(B − λI10 ) = 0 ⇔ det((1 − λ)J5 + (λ2 + λ − 2)I5 ) = 0.
Now, it remains to find all values of λ such that
 2 
λ −1 1−λ 1−λ 1−λ 1−λ
 1 − λ λ2 − 1 1 − λ 1−λ 1−λ 
 2

 1−λ 1−λ λ −1
det  1−λ 1−λ  = 0. (2.66)

 1−λ 1−λ 1−λ λ2 − 1 1 − λ 
1−λ 1−λ 1−λ 1 − λ λ2 − 1
Next, we reduce the matrix in (2.66) to an upper triangular matrix.
 2 
λ −1 1−λ 1−λ 1−λ 1−λ
 1 − λ λ2 − 1 1 − λ 1 − λ 1 − λ 
 2

det 
 1 − λ 1 − λ λ − 1 1 − λ 1 − λ =

 1−λ 1−λ 1−λ λ −1 1−λ  2

1 − λ 1 − λ 1 − λ 1 − λ λ2 − 1
 
−1 − λ 1 1 1 1

 1 −1 − λ 1 1 1 

= det  1 1 −1 − λ 1 1  · (1 − λ)5 =
 
 1 1 1 −1 − λ 1 
1 1 1 1 −1 − λ
−1 − λ 1
 
1 1 1
λ+2 λ+2 λ+2 2 +2λ

 0 λ+1 λ+1 λ+1
− λλ+1 

5 5 3 2
= (1−λ) ·det 
 0 0 λ + 2 λ + 2 (λ + 2)(1 − λ)   = (1−λ) (λ+2) (λ −λ−6).
 0 0 0 λ+2 −λ2 + 4 
0 0 0 0 −λ2 + λ + 6
Solving the equation (1 − λ)5 (λ + 2)3 (λ2 − λ − 6) = 0 we get that −2, 1 and 3 are the
eigenvalues of B (with multiplicity 4, 5 and 1, respectively).
From the lemma we conclude that 1, 4 and 9 are eigenvalues of A, with multiplicity 5, 4 and
1, respectively. □
42 Chapter 2. SOLUTIONS

2.5 NUMBER THEORY

Problem 37. (Shortlist JMMO 2012) Let n be a natural number. In the set of integers
solve the following system 
x + y + z = 2012
x5 + y 5 + z 5 + 1 = 6n

Solution. Since 6n ends in 6, we get that x5 + y 5 + z 5 ends in the digit 5, from where we
have 5 | x5 + y 5 + z 5 . From Fermat’s theorem we have a5 ≡ a (mod 5) for any integer a.
Using that x5 ≡ x (mod 5), y 5 ≡ y (mod 5) and z 5 ≡ z (mod 5) we get

x5 + y 5 + z 5 ≡ x + y + z ≡ 2012 ≡ 2 (mod 5). (2.67)

The congruence in (2.67), x5 +y 5 +z 5 ≡ 2 (mod 5), contradicts the fact that 5 | x5 +y 5 +z 5 .

Thus the above system has no solutions. □

Problem 38. (Shortlist JMMO 2013) Let p1 > 3 and p2 > 3 be two primes such that
6 | p1 + p2 . Prove that the number

1p1 p2 +1 + 2p1 p2 +2 + 3p1 p2 +3 + 4p1 p2 +4 + 5p1 p2 +5 + 6p1 p2 +6

is composite.
Solution. Since every prime number greater than three is of the form 6k ± 1 and since
6 |p1 + p2 , we can, without loss of generality, assume p1 = 6m − 1 and p2 = 6n + 1.
Hence p1 p2 ≡ −1 (mod 6), from which we get p1 p2 + 1 = 6s. Note that

1p1 p2 + 6p1 p2 +6 ≡ 1 + (−1)6s+5 ≡ 1 − 1 ≡ 0 (mod 7). (2.68)

We will show that

7 | 2p1 p2 +2 + 3p1 p2 +3 + 4p1 p2 +4 + 5p1 p2 +5 .
From Fermat’s little theorem, for a ∈ {2, 3, 4, 5, 6} we get a6 ≡ a7−1 ≡ 1 (mod 7).
Thus
2p1 p2 +2 + 3p1 p2 +3 + 4p1 p2 +4 + 5p1 p2 +5 ≡ 26s+1 + 36s+2 + 46s+3 + 56s+4 ≡
≡ (26 )s · 2 + (36 )s · 9 + (46 )s · 64 + (56 )s · 625 ≡ 2 + 9 + 64 + 625 ≡ 700 ≡ 0 (mod 7). (2.69)
From (2.68) and (2.69) we obtain that the number 1p1 p2 +1 + 2p1 p2 +2 + 3p1 p2 +3 + 4p1 p2 +4 +
5p1 p2 +5 +6p1 p2 +6 is divisible by 7, that is, 1p1 p2 +1 +2p1 p2 +2 +3p1 p2 +3 +4p1 p2 +4 +5p1 p2 +5 +6p1 p2 +6
is a composite number. □

Problem 39. (Shortlist JMMO 2014) In the set of prime numbers solve the equation

p144 + q 144 − 1 = 2013r .

Solution. Note that r > 2. We transform the given equation as follows:

p144 + q 144 = 2013r + 1 = (2013 + 1)(2013r−1 − 2013r−2 + . . . + 1) = 2014 · s.

2.5. NUMBER THEORY 43

Clearly 2014 | p144 + q 144 . Since 19 | 2014 we have 19 | p144 + q 144 .

If p ̸= 19 and q ̸= 19, then gcd(p, 19) = gcd(q, 19) = 1. Now, from the Fermat’s theorem
we get p18 ≡ q 18 ≡ 1 (mod 19). Therefore

p144 + q 144 ≡ (p18 )8 + (q 18 )8 ≡ 18 + 18 ≡ 2 (mod 19),

which is not possible.

Since p144 + q 144 is an even number, we conclude that p and q have the same parity. If
p = q = 2, then we easily verify that 2144 + 2144 − 1 = 2145 − 1 ̸= 2013r . If p = q = 19,
then p144 + q 144 − 1 = 2 · 19144 − 1 = 2013r . This equation is not solvable since 31 divides
2 · 19144 − 1 but does not divide 2013. Finally, if p = 19 and q ̸= 19 we get

p144 + q 144 ≡ 0 + 1 ≡ 1 (mod 19)

which is a contradiction. Hence, the given equation has no solutions. □

Problem 40. (Shortlist JMMO 2014) Prove that the product of six consecutive positive
integers cannot be a perfect cube.
Solution. Let n, n + 1, n + 2, n + 3, n + 4 and n + 5 be six consecutive positive integers.
We will show that the number n(n + 1)(n + 2)(n + 3)(n + 4)(n + 5) lies between the cubes
of two consecutive positive integers. We have

n(n + 1)(n + 2)(n + 3)(n + 4)(n + 5) = (n2 + 5n)(n2 + 5n + 4)(n2 + 5n + 6).

Let us take A = n2 + 5n + 4. Hence

n(n + 1)(n + 2)(n + 3)(n + 4)(n + 5) = (A − 4)A(A + 2) = A3 − 2A2 − 8A.

It obviously occurs A3 − 2A2 − 8A < A3 . Since n ≥ 1 we have A ≥ 10.

If n = 1, we get n(n + 1)(n + 2)(n + 3)(n + 4)(n + 5) = 720 which is not a perfect cube.
If n ≥ 2, we easily conclude that A ≥ 18. Consequently

(A − 1)3 = A3 − 3A2 + 3A − 1 < A3 − 2A2 − 8A.

Hence, for n ≥ 2, we have

(A − 1)3 < n(n + 1)(n + 2)(n + 3)(n + 4)(n + 5) < A3 ,

from where we conclude that the number n(n + 1)(n + 2)(n + 3)(n + 4)(n + 5) is not a
perfect cube. □

Problem 41. (Team Selection Test, Famnit, 2016) Write/represent the number
20162014 as a sum of four cubic numbers.
Solution. We have 2016 = 1000 + 1000 + 8 + 8 = 103 + 103 + 23 + 23 .
Thus,
20162014 = 2016 · 20162013 = (103 + 103 + 23 + 23 ) · (2016671 )3 =
44 Chapter 2. SOLUTIONS

= (10 · 2016671 )3 + (10 · 2016671 )3 + (2 · 2016671 )3 + (2 · 2016671 )3 .

□

Problem 42. (Shortlist Vojtěch Jarnı́k 2017) Let p and q be odd prime numbers such
that p2 > 2q and q | p2 + 1. Prove that at least one of the numbers 4p2 − 4q and 8p2 − 16q
is a sum of three or fewer squares.
Solution. Note that 4p2 − 4q and 8p2 − 16q are positive integers. Since the odd prime
divisors of x2 + 1 are of the form 4k + 1, we have q ≡ 1 (mod 4). Moreover, since p is an
odd number holds p2 ≡ 1 (mod 4). Let us suppose that 4p2 − 4q is not a sum of three or
fewer squares. Using Legendre’s theorem on sums of three squares we have that 4p2 − 4q is
of form 4a (8b + 7), for some non-negative integers a and b. Since 4p2 − 4q is not a sum of
three or fewer squares, it follows that p2 − q is not a sum of three or fewer squares. Thus
p2 − q = 4α (8β + 7), for some non-negative integers α and β. Since 4 | p2 − q we have that
α ≥ 1. Now, if we suppose that 8p2 − 16q is not a sum of three or fewer squares, then there
exist non-negative integers α1 and β1 such that 8p2 − 16q = 4α1 (8β1 + 7).
From p2 − q = 4α (8β + 7) we obtain

4α1 (8β1 + 7) = 8(p2 − q) − 8q = 8(4α (8β + 7) − q).

Since α ≥ 1, it follows that 4α (8β + 7) − q is odd, which implies that α1 = 1. Considering

the last equation modulo 4 we have 2q ≡ 1 (mod 4), which is not possible because q ≡ 1
(mod 4). □

1
Problem 43. (Shortlist Vojtěch Jarnı́k 2018) The sequence {an }∞ n=1 satisfies a1 =
4
1 s
and an+1 + = 8, n ≥ 1. Prove that a4k = for some s and t such that 4|s and 32|t + 1.
an t
Solution. From the conditions of the problem we get

an+1 an − 8an + 1 = 0 (2.70)

We define sequence {bn }∞ n=1 such that b1 = 2 and bn = an bn−1 , n ≥ 2. If we replace this
in (2.70), we get the recursion bn+1 − 8bn + bn−1 √
= 0. Solving the characteristic
√ equation
2
r − 8r + 1√= 0 we get the √ solution r1 = 4 + 15 and r2 = 4 − 15. Consequently,
n n
bn = c1 (4 + 15) + c2 (4 − 15) for some constants c1 , c2 . Since b1 = 2, b2 = 8, it follows
1 1
that c1 = √ and c2 = √ . Thus,
4 + 15 4 − 15
√ √
b2n = (4 + 15)2n−1 + (4 − 15)2n−1 + 2 = b2n−1 + 2.

Since bn+1 = 8bn − bn−1 , it follows that bn+1 ≡ −bn−1 (mod 8). Now we have b2k ≡ −b2k−2 ≡
· · · ≡ (−1)k−1 b2 ≡ 0(mod 8). Therefore b22k ≡ b4k−1 + 2 ≡ 0(mod 64) i.e. b4k−1 = 64m − 2
for some integer m. Because 8|b4k we get b4k = 8r for some integer r. Therefore,
b4k 8r 4r
a4k = = = .
b4k−1 64m − 2 32m − 1
2.5. NUMBER THEORY 45

Problem 44. (National Competition 2020, Macedonia) Let n and k be natural

numbers for which n > k ≥ 4. We suppose that k(k − 1) is not divisible by
n − 1 and
n
k(k − 1)(k − 2) is not divisible by n − 2. Prove that the binomial coefficient k has at least
two prime divisors p and q such that p | n − 1 and q | n − 2.

Solution. Let gcd(n − 1, k(k − 1)) = a and gcd(n − 2, k(k − 1)(k − 2)) = b. From the
conditions of the problem it follows that a < n − 1 and b < n − 2. Additionally, there exist
integers x1 , y1 , x2 , y2 for which (n−1)x1 +k(k−1)y1 = a and (n−2)x2 +k(k−1)(k−2)y2 = b.
Hence    
n n
a· = ((n − 1)x1 + k(k − 1)y1 ) · =
k k
   
n n n−1 n−2
= (n − 1) · x1 · + k(k − 1)y1 · · · =
k k k−1 k−2
    
n n−2
= (n − 1) x1 · + ny1 · .
k k−2
last identity we get (n − 1) | a · nk , i.e., n−1 | nk . Analogously we conclude that



From the a
n−2
| nk . Since a < n − 1 and b
< n − 2 we get n−1 and
n−2


b a b
are not equal to one. Moreover,
n−1 n n−2 n
it holds a ≤ n − 1 < n ≤ k . Analogously b < k .
It remains to show that n−1 a
̸= n−2
b
. Let us suppose that n−1 a
= n−2b
, that is, b(n − 1) =
a(n−2). From the last equation we get (n−1) | a(n−2). Since n−1 and n−2 are co-primes
and a < n − 1, it follows that this divisibility is not possible.
In the end, let p and q be prime divisors of n−1
a
and n−2
b
, respectively. It is clear that p and
n


q divide the binomial coefficient k and moreover p | (n − 1) and q | (n − 2). From this
and the fact that n − 1 and n − 2 are coprimes, we conclude that p ̸= q. □

Problem 45. (Shortlist Vojtěch Jarnı́k 2022) Let p ≥ 3 be a prime number and let
n ≥ 1 be a natural number. Prove that for any k such that 2 ≤ k ≤ pn holds
 n   n 
p +2 p +1
gcd , > 1.
k 2

Solution. Let k = 2. Then

 n   n
(pn + 1) (pn + 1) n pn + 1
  
p +2 p +1 n
gcd , = gcd (p + 2) · , ·p ≥ > 1.
k 2 2 2 2
n
Now, without loss of generality, we may assume 3 ≤ k ≤ p 2+2 . Let gcd(pn , k(k −1)(k −2)) =
d. There exists integers x and y such that pn x + k(k − 1)(k − 2)y = d. We have
 n   n 
p +2 n p +2
d = (p x + k(k − 1)(k − 2)y) =
k k
46 Chapter 2. SOLUTIONS
 n
pn + 2 pn + 1 pn
  n 
n p +2 p −1
=p x + k(k − 1)(k − 2)y · · · · =
k k k−1 k−2 k−3
  n   n 
n p +2 n n p −1
=p x + (p + 2)(p + 1)y .
k k−3
n n
Thus pd | p k+2 . We consider two cases:

n n n n n
1. If d < pn , then pd is a divisor of p k+2 . On the other hand, p 2+1 = (p +1)p



2
=
pn +1 n pn +1 pn pn +1
n



2
· p . Thus p | 2 , that is, d is a divisor of 2 . In this case the claim holds.

2. Let d = pn . Then from pn |k(k − 1)(k − 2) and gcd(k, k − 1, k − 2) = 1, gcd(k, k − 2) ≤ 2

we have pn |k or pn |k − 1 or pn |k − 2. Clearly, these three divisibilities are not possible
n
since k ≤ p 2+2 .

Problem 46. (Shortlist IMC 2022) Let p > 3 be a prime number. Prove that, if p is a
primitive root of 4p + 1, then 2p + 1 is a composite number.
Solution. Since p is a primitive root of 4p + 1 we have 4p + 1 = q k , where q is odd prime
and k ≥ 1. Let k > 1. We get 4p = q k − 1 = (q − 1)(q k−1 + . . . + q + 1). Thus q = 3 or
q = 5. If q = 3 we get 4p + 1 = 3k . From 3k ≡ 1 (mod 4) we get that k is an even number,
k1 k1 +1)
k = 2k1 . Thus p = (3 −1)(3
4
. Since both of the numbers 3k1 −1 and 3k1 +1 are even, and
since exactly one of them is divisible by 4 we get that p is an even number, a contradiction.
Let q = 5 and 4p + 1 = 5k . From ϕ(5k ) = 4 · 5k−1 we get:

ϕ(5k ) ϕ(5k )
4p ≡ −1 (mod 5k ) ⇔ (4p) 2 ≡ (−1) 2 ≡1 (mod 5k ).

k−1 ϕ(5k )
By induction we can prove that 42·5 ≡ 1 (mod 5k ). Thus we get p 2 ≡ 1 (mod 5k ),
which implies that p is not a primitive root of 4p + 1.
Now let k = 1. Thus 4p + 1 = q is a prime number. Since 4p + 1 is prime we have p ≡ 1
(mod 3). In this case 2p + 1 is divisible by 3, that is, 2p + 1 is a composite number. □

Problem 47. (National Competition 2023, Macedonia) The natural number n is

called magic if and only if there are exactly five natural numbers ki for which n+ki | (n−ki3 ),
(1 ≤ i ≤ 5). Prove that 29 is a magic number.
Solution. Let n be a magic number. Since n + x | n3 + x3 it follows that exactly five
natural numbers x satisfy n + x | (n − x3 ) + (n3 + x3 ), that is, n + x | n + n3 .
We conclude that, the number n is magic if and only if n3 + n has exactly five divisors
greater than n. Since 293 + 29 = 2 · 29 · 421 we get that 293 + 29 has exactly five divisors
greater than 29, they are the numbers 2 · 29, 421, 2 · 421, 29 · 421, 2 · 29 · 421.
The unique five natural numbers ki for which holds 29 + ki | 29 − ki3 are: 29, 392, 813, 12180
and 24389. □
2.5. NUMBER THEORY 47

Problem 48. (Shortlist IMC 2023) Let {ai }∞ i=1 be a geometric progression of natural
numbers which quotient has exactly k distinct prime divisors. Prove that the (k − 1)-th
differences of the sequence {τ (ai )}∞
i=1 form an arithmetic progression.

Solution. The proof relies on the following two well-known identities:

Pk+1 i k+1


Lemma 2.4. 1. i=0 (−1) i
= 0.

2. For each 1 ≤ m ≤ k it holds k+1 i m k+1

P

i=1 (−1) i i
= 0.

Let q = pα1 1 pα2 2 · . . . · pαk k be the canonical form of the quotient of the progression and let
a1 = pβ1 1 pβ2 2 · . . . · pβkk · b, where βi ≥ 0 and gcd(pi , b) = 1 for i = 1, 2, . . . , k.
(i−1)α1 +β1 (i−1)α2 +β2 (i−1)αk +βk
Then for any i ≥ 1 we have ai = a1 · q i−1 = p1 · p2 · . . . · pk · b and

τ (ai ) = ((i − 1)α1 + β1 + 1)((i − 1)α2 + β2 + 1) · . . . · ((i − 1)αk + βk + 1)τ (b). (2.71)

It is easy to prove that the (k − 1)th differences of {τ (ai )}∞

i=1 is the sequence

k−1  
j k−1
X
{si }∞
i=1 = { (−1) τ (ak+i−j−1 )}∞
i=1 . (2.72)
j=0
j

We will prove that sn+1 − 2sn + sn−1 = 0 for any n ≥ 2, which is a sufficient condition to
assert that the sequence {si }∞
i=1 is arithmetic. We have

sn+1 − 2sn + sn−1 = 0 ⇔

k−1   k−1   k−1  
X k−1
j
X
j k−1
X
j k−1
(−1) τ (ak+n−j )−2 (−1) τ (ak+n−j−1 )+ (−1) τ (ak+n−j−2 ) = 0.
j=0
j j=0
j j=0
j
(2.73)
The identity in (2.73) is equivalent to the identity
     
k−1 k−1 k−1
τ (ak+n ) − +2 τ (ak+n−1 )+
0 1 0
k−1     
X k−1j k−1 k−1
+ (−1) +2 + τ (ak+n−j )+
j=2
j j − 1 j − 2
      
k k−1 k−1 k+1 k − 1
+(−1) 2 + τ (an ) + (−1) τ (an−1 ) = 0.
k−1 k−2 k−1
Using k−1 = k+1

k−1

, 1 +2 k−1 = k+1

k−1

, j +2 k−1

k−1

+ j−2 = k+1




0 0 0 1 j−1 j
, for 2 ≤ j ≤ k−1,
k−1 k−1 k+1 k−1 k+1






2 k−1 + k−2 = k and k−1 = k+1 , the above identity is equivalent to

k+1  
j k+1
X
(−1) τ (ak+n−j ) = 0. (2.74)
j=0
j
48 Chapter 2. SOLUTIONS

Let bt = αt (k + n − 1) + βt + 1 for 1 ≤ t ≤ k. From the formula (2.71) and since τ (b) ̸= 0,

the identity in (2.74) is equivalent to
k+1  
j k+1
X
(−1) (b1 − jα1 )(b2 − jα2 ) · . . . · (bk − jαk ) = 0. (2.75)
j=0
j

Now we have (b1 − jα1 )(b2 − jα2 ) · . . . · (bk − jαk ) = C0 − jC1 + j 2 C2 + .... + (−1)k j k Ck where
Ci are constants which depends on α’s and b’s. Finally the identity in (2.75) is equivalent
to
k+1   k
j k+1
X X
(−1) (C0 + (−1)i Ci j i ) = 0. (2.76)
j=0
j i=1

We rearrange the last expression and we get

k+1  ! k+1  !
X k + 1 X k + 1
C0 (−1)i − C1 (−1)i i · +
i=0
i i=1
i

k+1  ! k+1  !
X k+1 X k + 1
+C2 (−1)i i2 · − . . . + (−1)k Ck (−1)i ik · = 0.
i=1
i i=1
i
Based on the above lemma we verify that the last identity is valid, which completes the
proof. □

Problem 49. (Vojtěch Jarnı́k 2024) Let p > 2 be a prime and let

A = {n ∈ N : 2p | n and n | 3n − 1}.

Prove that T
|A [1, k]| 2 log 3
lim sup ≤ .
k→∞ k p log p
n n n n
Solution. Let n ∈ A. Then p | (3 2 − 1)(3 2 + 1), from where 3 2 ≡ 1 (mod p) or 3 2 ≡ −1
(mod p). Since p | n and n is an even number, n = pr, where r is even. Since (p, 3) = 1,
n r r r r
Fermat’s little theorem yields 3 2 ≡ (3p ) 2 ≡ 3 2 (mod p). Hence, 3 2 ≡ 1 (mod p) or 3 2 ≡
−1 (mod p). Recalling (p, 3) = 1 again, let l denote the smallest positive integer satisfying
log p
3l ≡ 1 (mod p). This yields p < 3l , and therefore l > . As shown above, there are two
log 3
r
possible residue classes modulo l that might belong to. Thus, the asymptotic density of
2
the multiples rp for which r satisfies the above conditions within the set of all multiples of
log 3
p is at most 2 · . To determine the asymptotic density of the multiples of p within the
log p
set of all positive integers, we can consider the set Mk = {p, 2p, 3p, . . . , mp} with mp ≤ k,
for a positive integer k. Then |Mk | = m, and therefore
m 1
d(Mk ) = lim sup ≤ .
k→∞ k p
2.5. NUMBER THEORY 49

By these observations we get

T
|A [1, k]| 1 2 2 log 3
d(A) = lim sup < · ≤ .
k→∞ k p l p log p
□

Problem 50. (Shortlist Vojtěch Jarnı́k 2024) Let p ≥ 3 be a prime number and let
m, n, a and b be integers such that gcd(m, p) = gcd(n, p) = 1. Prove that, if the congruence
x2 ≡ −mn (mod p) is not solvable and if p | ma2 + nb2 , then p2 | ma2 + nb2 .
Solution. Let us suppose that gcd(a, p) = 1. From ma2 ≡ −nb2 (mod p) we get
 (ma)
2
≡
2
−mnb2 (mod p). Since the congruence x2 ≡ −mnb2 (mod p) is solvable, we get −mnb
p
=
   2
1. However, this is not possible since −mn
p
= −1 and bp = 1. It remains that p | a.
Now we easily conclude that p | b, from where we get p2 | ma2 + nb2 . □
Abbreviations

BMO Balkan Mathematical Olympiad

IMC International Mathematics Competition
IMIO International Mathematical Internet Olympiad
JBMO Junior Balkan Mathematical Olympiad
JMMO Junior Macedonian Mathematical Olympiad
MMO Macedonian Mathematical Olympiad