STAR CIRCLE PROGRAM. Assignment 2. Solutions.

Problem 1. Positive integers are painted in 3 colors. Prove that there is a color C such that for any positive
integer n there are infinitely many numbers of color C that are divisible by n.
Wrong solution. Assume, for the sake of contradiction, that no such color C exists. This would mean that for
each color, the numbers divisible by any positive integer n and painted in that color form a finite set. Since there
are only three colors, each color would cover only finitely many of the numbers divisible by n.
However, we know that there are infinitely many numbers divisible by n. Therefore, if each of the three colors
only paints finitely many of these numbers, the total number of colored numbers divisible by n would be finite. This
contradicts the fact that the numbers divisible by n are infinite. Hence, there must exist at least one color C such
that there are infinitely many numbers divisible by n that are painted with C.
Correct Solution.

1. Suppose the opposite, that for each color Ci there is such number ni that there are only finitely many numbers
of color Ci divisible by ni .
2. Consider numbers divisible by n1 n2 n3 . There are finitely many such numbers among every color.
3. But there are infinitely many numbers divisible by n1 n2 n3 in total. So we get a contradiction.

Problem 2. On a board, the numbers 1, 2, 3, . . . , 2018 are written in some order. The position number of each
number is added to it. Prove that either two equal numbers will be obtained, or two numbers will differ by 2018.
Proof: We are given that the numbers 1, 2, 3, . . . , 2018 are written in some order on a board. Each number ai
is located in position bi , and we define ci = ai + bi , where ai is the number in position bi . We want to prove that
either two equal numbers will be obtained, or two numbers will differ by 2018.
Assume, for the sake of contradiction, that neither of these conditions hold. We know that 2 ≤ ci ≤ 2 × 2018
because ci is the sum of ai (which is between 1 and 2018) and bi (its position, also between 1 and 2018).
The condition that "either two numbers will be equal, or two numbers will differ by 2018"can be reformulated
as follows: if two ci ’s are equal modulo 2018, then either the two ci ’s are equal, or one is equal to the other plus
2018.
Let oo be the number of odd numbers in odd positions, ee the number of even numbers in even positions, oe
the number of odd numbers in even positions, and eo the number of even numbers in odd positions. We know that:
oo + ee = 2018/2, eo + oe = 2018/2.
Additionally, from the arrangement of numbers and positions, we have: oo + oe = 2018/2, eo + ee = 2018/2,
eo + oo = 2018/2, oe + ee = 2018/2.
Since there are 2018/2 odd numbers and the same goes for odd positions, even positions, and even numbers,
this implies that oo = eo = oe = ee = 2018/4. However, this is impossible because 2018 is not divisible by 4.
Therefore, our assumption leads to a contradiction, and we conclude that either two equal numbers will be
obtained, or two numbers will differ by 2018.

Problem 3. Given cyclic quadrilateral ABCD. Incenters of ABC, BCD, CDA, ADB, are K, L, M, N . Prove
that KLM N is a rectangular.
Solution. If M is the incenter of △ABC, then
∠ACB
∠AM B = 90◦ + .
2
Proof In △BM C, we have
∠AM B = 180◦ − ∠M AB − ∠M BA
∠BAC ∠ABC
= 180◦ − −
2 2
 ∠BAC + ∠ABC
= 180◦ −
2
◦
180 − ∠ACB
= 180◦ −
2
∠ACB
= 90◦ + .
2
- N is the incenter of △DAB ⇒ ∠AN B = 90◦ + ∠ADB2
- K is the incenter of △ABC ⇒ ∠AKB = 90◦ + ∠ACB 2
- Since ABCD is cyclic ⇒ ∠ACB = ∠ADB
Combining the above equations, we find ∠AN B = ∠AKB, so quadrilateral ABN K is cyclic. It follows that
∠BAD
∠BKN = 180◦ − ∠BAN = 180◦ −
2
Similarly, since BCLK is cyclic, we have:
∠BCD
∠BKL = 180◦ − ∠BCL = 180◦ −
2
From the previous equations, we now deduce:
∠BAD ∠BCD
∠N KL = 360◦ − (∠BKL + ∠BKN ) = + = 90◦ .
2 2
In the same way, we obtain that all angles of the quadrilateral N KLM have a measure of 90◦ .

Problem 4. Prove that there exists a natural number n that has more than 2020 divisors d satisfying the
condition √ √
n < d < 1.01 n.
√
Proof: Consider n of the form n = p2a q 2b , where p and q √are primes, and p > q. Notice that pa q b = n. We
b−k a+k
define the divisors
√ of n of interest
√ as ck = q p , with c0 = n. Our goal is to find at least 2021 divisors of this
form such that n < ck < 1.01 n.
We can express ck in terms of ck−1 :
p
ck = ck−1 · .
q
Since p > √q, it follows that ck > ck−1 > ck−2 > · · · > c0 . We are interested in finding the largest k such that
ck < 1.01 · n. This condition is equivalent to:
√
q b−k pa+k < 1.01 · n.
√
Substituting c0 = n, we get the inequality:
 k
p
< 1.01.
q
To satisfy this inequality for k = 2021, we need to find primes p and q such that:
 2021
p
< 1.01.
q

This simplifies to:

p 1
< (1.01) 2021 .
q
1/2
Nicely’s Theorem (1999) says that there are infinite primes such that pn+1 − pn < pn , we can choose p = pn+1
and q = pn such that:
pn+1 < pn + p1/2
n .
Thus, we have:
pn+1 1
< 1 + 1/2 .
pn pn
Since this happens infinitely often, we can find a sufficiently large pn such that:
!2021
1
1+ 1/2
< 1.01.
pn
!2
1
For example, choosing any pn >  1 and a, b > 2020, we ensure that the condition holds. Therefore,

1.01 2021 −1
√ √
we can find more than 2020 divisors of the form pa q b in the interval ( n, 1.01 n), which completes the proof.
Second Solution:
√
1. Let’s select n = (4000!)2 . Then n = 4000!.
√ √ √
2. Define ci = 1 + 1i . It is obvious, that for all i = 1001, 1002, . . . 3020 we have n < ci n < 1, 01 · n.
√
3. Let’s prove that all√numbers ci · n for i = 1001, . . . 3020 are integer and are divisors of n. Any such number
is integer because nisdivisiblebydenominatorof ci .
√
n √n n 4000!·i
4. And √nc i
is integer, because nci
= ci = i+1 and 4000! is divisible by i + 1. Thus, all such numbers are
divisors of n.

Problem 5. Positive a, b, c satisfy the equality a2 + b2 + c2 + abc = 4. Prove that a + b + c ⩽ 3.

Solution 1: Without loss of generality, assume that a ≤ b ≤ c. We will consider two cases: either a ≤ 1 and
b ≤ 1, or b > 1 and c > 1.
Case 1: a ≤ 1 and b ≤ 1.
Since a ≤ 1 and b ≤ 1, we have:

(1 − a)(1 − b) ≥ 0 which implies 1 + ab ≥ a + b.

Thus, we can write:

a + b + c ≤ 1 + ab + c.
Now, we only need to prove that ab+c ≤ 2. Suppose, for the sake of contradiction, that ab+c > 2. Then, substituting
this into the given equation a2 + b2 + c2 + abc = 4, we get:

a2 + b2 + c2 + abc ≥ 2ab + c(c + ab) > 2ab + 2c = 2(ab + c) > 4.

This would contradict the given condition a2 + b2 + c2 + abc = 4. Therefore, it must be the case that ab + c ≤ 2,
and consequently a + b + c ≤ 3.
Case 2: b > 1 and c > 1.
In this case, use that
(b − 1)(c − 1) > 0
and then continue as the previous case.
Solution 2.

1. Suppose the opposite and there is such triple of numbers a0 , b0 , c0 with a0 +b0 +c0 > 3 and a20 +b20 +c20 +a0 b0 c0 =
4. Consider numbers a = a0 · a0 +b30 +c0 , b = b0 · a0 +b30 +c0 , c = c0 · a0 +b30 +c0 . Since a < a0 , b < b0 and c < c0 , it
is clear that a2 + b2 + c2 + abc < a20 + b20 + c20 + a0 b0 c0 = 4.
2. So to get a contradiction we need to prove inequality that for any positive real numbers a, b, c which satisfy
a + b + c = 3 we have a2 + b2 + c2 + abc ⩾ 4.
 3. First of all, we’ll make the inequality homogeneous

a+b+c (a + b + c)3
(a2 + b2 + c2 ) + abc ⩾ 4
3 27
9(a2 + b2 + c2 )(a + b + c) + 27abc ⩾ 4(a + b + c)3

4. Now we’ll continue our proof. Denote T3 = a3 + b3 + c3 , T21 = a2 b + a2 c + b2 c + b2 a + c2 a + c2 b,

9(a2 + b2 + c2 )(a + b + c) + 27abc ⩾ 4(a + b + c)3

9T3 + 9T21 + 27abc ⩾ 4T3 + 12T21 + 24abc
5T3 + 3abc ⩾ 3T21

5. This inequality is the sum of two classic inequalities: Schur’s inequality T3 + 3abc ⩾ T21 and Muirhead’s
inequality 2T3 ⩾ T21 .

Solution 3. We will prove inequality 9(a2 + b2 + c2 )(a + b + c) + 27abc ⩾ 4(a + b + c)3 using pqr-method.

1. Define p = a + b + c, q = ab + bc + ca, r = abc. In this terms we can rewrite the inequality we want to prove.

9(p2 − 2q)p + 27r ⩾ 4p3

5p3 − 18pq + 27r ⩾ 0.

2. We will use following.

Definition. The polynomial (a − b)2 (a − c)2 (b − c)2 = −4p3 r + p2 q 2 + 18pqr − 4q 3 − 27r2 = T (p, q, r) is called
the discriminant.
Criterion for reality. Let (p, q, r) be a triple of real numbers. Three roots of cubic polynomial x3 −px2 +qx−r
are real if and only if T (p, q, r) = −4p3 r + p2 q 2 + 18pqr − 4q 3 − 27r2 ⩾ 0.
Non-negativity lemma. Let (p, q, r) be a triple of real numbers. Three roots of cubic polynomial x3 − px2 +
qx − r are non-negative if and only if T (p, q, r) = −4p3 r + p2 q 2 + 18pqr − 4q 3 − 27r2 ⩾ 0 and p, q, r ⩾ 0.
3. We want to prove 5p3 − 18pq + 27r ⩾ 0. If we fix p and q then T (p, q, r) is a quadratic polynomial with respect
to variable r with negative leading coefficient. Hence, T (p, q, r) ⩾ 0 for values of r belonging to some segment
[rmin , rmax ]. So we have two conditions on r: r ⩾ 0 and rmin ⩽ r ⩽ rmax .
4. Consider the case rmin < 0. We know that for triple (p, q, 0) there are three non-negative roots of polynomial
x3 − px2 + qx. Denote them a1 , b1 , c1 = 0. If we prove the initial inequality for a1 , b1 , c1 then we’ll check
5p3 − 18pq0 and 5p3 − 18pq + 27r ⩾ 0 will follow automatically. If c1 = 0, then we have a + b = 3 and we need
to prove that a2 + b2 ⩾ 4. This is true because a2 + b2 ⩾ 21 (a + b)2 = 29 .

5. Consider the case rmin > 0. We know that for triple (p, q, rmin ) there are three non-negative roots of
polynomial x3 − px2 + qx − rmin . Denote them a1 , b1 , c1 . Two of them are equal because T (p, q, rmin ) = 0. If
we prove the initial inequality for a1 , b1 , c1 then we’ll check 5p3 − 18pq + 27rmin 0 and 5p3 − 18pq + 27r ⩾ 0
will follow automatically because r ⩾ rmin .
6. So we need to check the inequality in the case when two numbers are equal. WLOG we can assume that a = b.
So we need to check that for 0 ⩽ a ⩽ 32 inequality a2 + a2 + (3 − 2a)2 + a2 (3 − 2a) ⩾ 4 holds.

−2a3 + 9a2 − 12a + 9 ⩾ 4

2a3 − 9a2 + 12a − 5 ⩽ 0
(a − 1)2 (2a − 5) ⩽ 0

The last is true because 2a − 5 < 0 for a ⩽ 23 .

 Solution 4. Derivatives.
1. We do the same manipulations as in solution 2, so we need to prove inequality that for any non-negative real
numbers a, b, c which satisfy a + b + c = 3 we have a2 + b2 + c2 + abc ⩾ 4.
2. Consider f (a, b) = a2 + b2 + (3 − a − b)2 + ab(3 − a − b) = −a2 b − ab2 + 2a2 + 2b2 + 5ab − 6a − 6b + 9. If we
want to find minimum value of f (a, b) on triangle with borders a = 0, b = 0 and a + b = 3, then we need to
check borders or the points where all partial derivatives are equal to zero.
3. Checking the border. All three sides of the triangle lead somehow to the case when one of the numbers
a, b, c is zero. This case was considered in solution 3.
4. Finding partial derivatives.
∂f (a, b)
= −2ab + b2 + 4a + 5b − 6 = −(b − 2)(2a + b − 3)
∂a
∂f (a, b)
= −2ab + a2 + 4b + 5a − 6 = −(a − 2)(2b + a − 3)
∂a
5. So we need to solve the system (b − 2)(2a + b − 3) = 0, (a − 2)(2b + a − 3) = 0. Both equation holds in cases
where one of the number is 2 or 2a + b = 3 and 2b + a = 3.
6. If one of the numbers is 2, then its square equals 4, so a2 + b2 + c2 + abc ⩾ 4.
7. If 2a + b = 3 and 2b + a = 3 then a = b = 1. And we can see that f (1, 1) = 4. So f (a, b) ⩾ 4 for all (a, b)
inside the triangle.
Solution 5.
a2 + b2 + c2 + abc = 4
(a + b + c)2 − 2ab − 2ac − 2bc + abc = 4
(a + b + c)2 + (a − 2)(b − 2)(c − 2) − 4(a + b + c) + 8 = 4
(a + b + c − 2)2 + (a − 2)(b − 2)(c − 2) = 0 (1)

(a + b + c − 2)2 = (2 − a)(2 − b)(2 − c)

1
(a + b + c − 2)2 ⩽ (2 − a + 2 − b + 2 − c)3
27
If a + b + c > 3 then LHS is bigger than 1, and RHS is smaller than 1. So we get a contradiction.

Problem 6. The edges of a complete graph on n vertices are colored in

(a) 2 colors;
(b) 3 colors.
What is the smallest n for which it can be guaranteed that there exists a monochromatic triangle?
(a) Two-Color Case
We aim to show that in any two-coloring of the edges of a complete graph on 6 vertices, there must exist a
monochromatic triangle.
Consider a complete graph K6 with vertices labeled v1 , v2 , v3 , v4 , v5 , v6 . Fix one vertex, say v1 . This vertex is
connected to the other 5 vertices by edges that can be colored either red or blue.
By the pigeonhole principle, at least 3 of these edges must be of the same color. Without loss of generality,
suppose that the edges v1 v2 , v1 v3 , v1 v4 are red. Now we need to examine the edges between the vertices v2 , v3 , and
v4 .
- If any edge among v2 v3 , v2 v4 , or v3 v4 is red, we have a red triangle v1 , v2 , v3 . - If none of these edges is red,
then all of them must be blue, which creates a blue triangle among v2 , v3 , and v4 .
Thus, in any two-coloring of K6 , there must exist a monochromatic triangle, either red or blue. Therefore, it is
guaranteed that n ≤ 6, and we can find a K5 where there is no monochromatic triangle , hence the answer is 6.
 Red Edges: v1 v2 , v1 v3 , v2 v4 , v3 v5 , v4 v5
Blue Edges: v1 v4 , v1 v5 , v2 v3 , v2 v5 , v3 v4
(b) Three-Color Case
Now, we prove that in any three-coloring of the edges of a complete graph on 17 vertices, there must exist a
monochromatic triangle.
Consider the complete graph K17 . Fix one vertex, v1 . This vertex is connected to the other 16 vertices with
edges colored using three colors, say red, blue, and green. By the pigeonhole principle, at least 6 of these edges
must be of the same color. Without loss of generality, let’s assume these edges are red, connecting v1 to vertices
v2 , v3 , v4 , v5 , v6 , v7 .
Now, we will look at the subgraph formed by the vertices v2 , v3 , v4 , v5 , v6 , v7 . The edges among these vertices
are colored either blue or green.
Since there are 6 vertices, we apply the reasoning from the two-color case. In the subgraph with these 6 vertices:
- If any edge among v2 v3 , v2 v4 , v2 v5 , v3 v4 , v3 v5 , v4 v5 , etc. is red, we form a red triangle with v1 and the two
vertices connected by this edge. - If none of the edges among these vertices is red, then we examine the edges
connecting v2 , v3 , v4 , v5 , v6 , v7 , which must all be colored either blue or green.
Therefore, it is impossible to avoid forming a monochromatic triangle either through v1 or entirely within the
set v2 , v3 , v4 , v5 , v6 , v7 .
Thus, in any three-coloring of K17 , there must exist a monochromatic triangle. Therefore, n ≤ 17, and we can
also find a K16 colored with 3 colors where there is no monochromatic triangle.
Now we will provide an example of a graph with 16 vertices and no monochromatic triangles.
We assign each of 16 vertices a sequence of four zeros and ones (binary string).
We will use the bitwise sum operation.
Definition. The bitwise sum of two binary strings is an operation in which each symbol of one string is added
to the corresponding symbol of the other string, bit by bit. The bitwise sum follows the rules of binary addition
(without carry).
We will color an edge between two strings depending on the value of their bitwise sum. In this way, we need to
divide all 15 non-zero strings into three groups of 5. If there is a monochromatic triangle with vertices a, b, and c,
this means that a ⊕ b, a ⊕ c, and b ⊕ c are assigned the same color, i.e., within a group of the same color, there are
three strings whose bitwise sum is 0.
In a group of five strings, there will be no three strings with a bitwise sum of 0 if the sum of all five strings in
this group is 0.
The remaining task is to divide all 15 non-zero strings into three groups of 5 with a total sum of 0. This can be
done, for example, as follows:
1) 1000, 0100, 0010, 0001, 1111
2) 1100, 0110, 0011, 1110, 0111
3) 1010, 0101, 1101, 1001, 1011.