First OJIMC 2021

Online International Mathematical Cup

 Day I Problems

Problem 1. Find all functions f : R → R such that any real numbers x, y satisfy

xf (y) + f (x + y) ≥ (y + 1)f (x) + f (y).

Problem 2. Call a polynomial P ∈ Z[x] i-good if exactly one of i and P (i) can be written as the
sum of two squares. Prove that for any positive integer n there exists some P which is i-good for
any i ∈ {1, 2, ..., n}. Does the exist a polynomial which is i-good for any i ∈ N?

Problem 3. Alice has n (possibly empty) bags of coins labelled 1 to n. She performs moves as such:
At the k th move, she lists all the bags which have (strictly) less coins than the k th bag (indices
reduced modulo n), but differ by (strictly) less than n coins from it.
If there is at least one listed bag, Alice takes one coin from each such bag and puts them all in the
k th bag. Otherwise, she takes n − 1 coins from the k th bag and places one in each of the other bags.
If at some point Alice cannot perform a move, the process ends. Find the greatest total number
of coins Alice can have, such that the process will end regardless of the initial distribution of the
coins in the bags.

Day II Problems

Problem 4. Let ABC be a triangle such that ∠A = 75◦ , ∠B = 60◦ , and ∠C = 45◦ . Let N9 be the
9-point center of △ABC, and let M be the midpoint of AC. What is the measure of the acute
angle formed by the lines AN9 and BM ?

Problem 5. Prove that no matter how we tile a fixed grid polygon with (possibly rotated) 1 × 2021
tiles, the number of tiles whose bottom left vertex has the sum of coordinates divisible by 2021 is
a constant (which may depend on the polygon).

Problem 6. Let a, b, c be positive real numbers. For any positive integer n, prove that
n n n
4a2 + 2ab 4b2 + 2bc 4c2 + 2ac
  
+ + ≥ an + bn + cn .
3a + 3b 3b + 3c 3c + 3a

1
Problem 1. Find all functions f : R → R such that any real numbers x, y satisfy

xf (y) + f (x + y) ≥ (y + 1)f (x) + f (y).

(Dorlir Ahmeti, Kosovo)

Solution 1. Let P (x, y) denote the given condition. Taking P (x, 0) we get that (x − 1)f (0) ≥ 0 for
all real numbers x, which clearly implies f (0) = 0.

Taking P (x, −x) we have, (x − 1)f (−x) ≥ −(x − 1)f (x). Hence, we get that f (−x) ≥ −f (x) for
all x > 1. Analogously, P (−x, x) gives us that f (−x) ≤ −f (x) for all x > −1.

Therefore, by combining the two latter inequalities, we get that f (−x) = −f (x) for all x > 1.

By adding P (x, y) and P (y, x), and reducing the common terms, we get f (x + y) ≥ f (x) + f (y).
Furthermore, P (−x, −y) and P (−y, −x) analogously give us f (−x − y) ≥ f (−x) + f (−y).

Now, let x, y > 1 in f (−x − y) ≥ f (−x) + f (−y). Since f is odd on (1, ∞), we have −f (x +
y) ≥ −f (x) − f (y) or, in other words, f (x + y) ≤ f (x) + f (y). However, we also know that
f (x + y) ≥ f (x) + f (y) for all x and y.

In conclusion f (x + y) = f (x) + f (y) for all real numbers x, y > 1. Now by taking P (x, y) for
x, y > 1 and using latter relationship, we have

f (x) f (y)
xf (y) ≥ yf (x) ⇐⇒ ≤
x y.

Similarly, P (y, x) implies f (x)/x ≥ f (y)/y. Hence, for all x, y > 1, f (x)/x = f (y)/y so for all
x > 1, f (x) = cx for some constant c.

Now, fix x and take y such that y > max{1, 1 − x}. Then, P (x, y) yields

cxy + c(x + y) ≥ (y + 1)f (x) + cy ⇒ (y + 1)cx ≥ (y + 1)f (x) ⇒ cx ≥ f (x).

On other hand, in the same conditions, P (y, x) gives us

yf (x) + c(x + y) ≥ c(y + 1)x + cxy ⇒ yf (x) ≥ cxy ⇒ f (x) ≥ cx.

By combining the last two inequalities, since x can have any real value, we get that f (x) = cx for
all real numbers x. This function can be easily checked to satisfy the statement.

2
Problem 2. Call a polynomial P ∈ Z[x] i-good if exactly one of i and P (i) can be written as the
sum of two squares. Prove that for any positive integer n there exists some P which is i-good for
any i ∈ {1, 2, ..., n}. Does there exist a polynomial which is i-good for any i ∈ N?

(Archit Manas, India)

Solution. In what follows, Q2 will denote the set of positive integers expressible as the sum of two
squares. We will first prove that, indeed, for any positive integer n, there exists P which is i-good
for each i ∈ {1, 2, ..., n}. Let us consider a polynomial P of the form
n
Y n
Y n
Y
P (x) = 1 + a1 (x − i) + a2 (x − i) + ... + an (x − i).
i̸=1 i̸=2 i̸=n

One can easily observe that for each i ∈ {1, 2, ..., n}, all the terms which coefficients different from
ai will vanish, so P (i) will be equal to ai (−1)n−i (n − i)!(i − 1)! + 1.

2
/ Q2 simply let ai = (−1)n−i (n − i)!(i − 1)! so P (i) = (−1)n−i (n − i)!(i − 1)! + 1 ∈ Q2 .
If i ∈

If i ∈ Q2 , then P (i) = ai (−1)n−i (n − i)!(i − 1)! + 1 must not be in Q2 . Hence, we choose a prime
number p ≡ 3 (mod 4) bigger than n and consider ai to be a solution to the congruence

−1
k ≡ (p − 1) · (−1)n−i (n − i)!(i − 1)! (mod p2 ).

Note that p > n and thus, (−1)n−i (n − i)!(i − 1)! has an inverse modulo p2 . Clearly, since ai satisfies
the above congruence, then

νp P (i) = νp ai (−1)n−i (n − i)!(i − 1)! + 1 = 1

and therefore, P (i) ∈

/ Q2 . So by combining the two cases we discussed, we can choose a1 , a2 , ..., an
such that, for each i ∈ {1, 2, ..., n}, P is i-good, which completes the proof.

We will proceed to show that there is no polynomial P which is i-good for all i ∈ N. Assume,
for the sake of contradiction, that such a polynomial P exists. Begin by observing that constant
polynomials do not work, so deg P > 0.

Furthermore, if P has a negative leading coefficient, then for big enough n ≡ 3 (mod 4), P (n) will
be negative. Since n ∈
/ Q2 , then P (n) must be in Q2 , but there are no negative numbers in Q2 .

Now, choose some big enough n such that P (n2 )
> 1. Since n2 ∈ Q2 , then P (n2 ) ∈
/ Q2 so there
exists
some prime p ≡ 3 (mod 4) such that νp P (n2 ) is odd. For the sake of brevity, let νp P (n2 ) = t.
Then, for any positive integer k, pt | P (n2 + kpt ).

Choose k = p · m such that, for some prime number q ≡ 3 (mod 4), greater than p, m satisfies

−1
m ≡ (q − n2 ) · pt+1 (mod q 2 ).

Note that q > p so pt+1 has an inverse modulo q. Since m satisfies the latter congruence, then
νq (n2 + kpt ) = νq (n2 + mpt+1 ) = 1, so n2 + mpt+1 ∈
/ Q2 . This implies the fact that P (n2 + kpt ) =
P (n2 + mpt+1 ) ∈ Q2 .

However, pt divides P (n2 + mpt+1 ) and t is odd, and thus, pt+1 must divide 2 t+1

P (n + mp ), which
yields p t+1 2 2
| P (n ). This is a contradiction, since we chose t to be νp P (n ) . Hence, no polynomial
P which is i-good for each i ∈ N exists.

3
Problem 3. Alice has n (possibly empty) bags of coins labelled 1 to n. She performs moves as such:
At the k th move, she lists all the bags which have (strictly) less coins than the k th bag (indices
reduced modulo n), but differ by (strictly) less than n coins from it.
If there is at least one listed bag, Alice takes one coin from each such bag and puts them all in the
k th bag. Otherwise, she takes n − 1 coins from the k th bag and places one in each of the other bags.
If at some point Alice cannot perform a move, the process ends. Find the greatest total number
of coins Alice can have, such that the process will end regardless of the initial distribution of the
coins in the bags.

(Rafael Yanto, Indonesia)

Solution. The answer is n2 − n − 1. By putting 0 coins in the nth bag and n coins in each of the
other bags, it can easily be shown inductively that the process never ends:

. . . → (n, n, n, . . . , 0) → (1, n + 1, n + 1, . . . , 1) → (2, 2, n + 2, . . . , 2) → . . . → (n, n, n . . . , 0) → . . .

It follows that the desired maximum is strictly smaller than n2 − n. To show that it is indeed equal
to n2 − n − 1, it suffices to prove that any process starting in a configuration with n2 − n − 1 coins
must eventually end.

Let min(k) denote the minimum number of coins among all bags after the k th move. Assume that
for some configuration of n2 − n − 1 coins, the process never ends.
Claim. If B has the least number of coins before a move, and we perform the move on a bag B ′
with more coins than B, then B will still have the least amount of coins.
Proof. We split the proof in two cases:

Case 1. If B ′ has t ≥ min(k) + n coins then we have two possibilities:

If there is no other bag with at least t − (n − 1) coins, then n − 1 coins are taken out of B ′ and
distributed among the other bags. In this scenario, B remains minimal, since all bags gain 1 coin,
B ′ which will have t − (n − 1) ≥ min(k) + 1 coins.

If there exists a bag with at least t − (n − 1) coins, then B coins remains unchanged, while some
other bags receive coins from B ′ . Of course, B remains minimal in this case too.

Case 2. If B ′ has at most min(k) + n − 1 coins, then bag B, as well as all other bags which differ
by at most n coins from B ′ lose one coin each, so B will still be minimal.

Assume that we never make a move on a bag with the least amount of coins. It follows from the
claim that throughout the whole process, the bag with the least amount of coins stays the same.

However, since we operate on bags cyclically, it follows that at some point, after enough moves
(which we can guarantee we will make, because we assumed the process never ends), we will make
a move on the fixed minimal bag, contradicting our assumption!

It follows that at some point, we will indeed make a move on the bag B with the least number of
coins. Because no bag has strictly less coins than B, then we must remove n − 1 coins from B and
under the assumption that the process never ends, we infer that B has at least n − 1 coins.

Since B has the least amount of coins among all bags however, it follows that there is a total of at
least n(n − 1) > n2 − n − 1 coins in all the bags, which is a contradiction.

4
Problem 4. Let ABC be a triangle such that ∠A = 75◦ , ∠B = 60◦ , and ∠C = 45◦ . Let N9 be the
9-point center of △ABC, and let M be the midpoint of AC. What is the measure of the acute
angle formed by the lines AN9 and BM ?

(Orestis Lignos, Greece)

Solution. Let H and O be the orthocenter and circumcenter of △ABC respectively.

Note that ∡HAO = ∡HAC − ∡OAC = 45◦ − 30◦ = 15◦ and that ∡AHO = 360◦ − ∡AHB −
∡BHO = 360◦ − 135◦ − 75◦ = 150◦ where ∡BHO = 75◦ follows since BH = 2OM and BO =
AO = 2OM because ∡OAC = 30◦ , hence △BHO is isosceles with ∡HBO = 30◦ .

Thus, ∡HOA = 180◦ − ∡AHO − ∡HAO = 15◦ = ∡HAO. Therefore, the triangle HAO is isosceles,
implyng that HA = HO. Since N is the midpoint of HO, we deduce that HA = 2HN .

In addition, BO = 2OM , as noted before, and ∡BOM = ∡AOB + ∡AOM = 150◦ = ∡AHN .
Thus, ∡AHN = ∡BOM and AH/HN = BO/OM , implying that △AHN ∼ △BOM.

The latter similarity yields that ∡M BO = ∡HAN . In addition, ∡ABO = 45◦ = ∡HAC, hence

∡ABM = ∡ABO − ∡M BO = ∡HAC − ∡HAN = ∡N AC

or, in other words, ∡ABM = ∡N AC. Finally, we get that desired angle is equal to 75◦ as such:

∡(AN, BM ) = ∡N AB + ∡ABM = ∡N AB + ∡N AC = 75◦ .

H
M
N9

O
B

5
Problem 5. Prove that no matter how we tile a fixed grid polygon with (possibly rotated) 1 × 2021
tiles, the number of tiles whose bottom left vertex has the sum of coordinates divisible by 2021 is
a constant (which may depend on the polygon).

(Pranjal Srivastava, India)

Solution. For each (i, j) ∈ Z × Z, to the 1 × 1 cell whose bottom left vertex has coordinates (i, j)
we will associate the integer f (i, j) defined by

i + j, if i + j ≡ 0 mod 2021

f (i, j) := −(i + j), if i + j ≡ −1 mod 2021

0, if i + j ̸≡ 0, −1 mod 2021


Assume that a 1 × 2021 tile has a bottom left vertex with the sum of coordinates divisible by
2021. Let that vertex be (i, j). Without loss of generality, assume that the tile is vertical, hence the
bottom left vertices of the remaining 2020 cells have coordinates (i, j + 1), . . . , (i, j + 2020).

Therefore, the sum of the integers f (i, j) assigned to the cells of the 1 × 2021 tile is equal to
2020
X 2019
X
f (i, j + k) = f (i, j) + f (i, j + k) + f (i, j + 2020)
k=0 k=1
= f (i, j) + f (i, j + 2020) = −2020

Now, assume that a 1 × 2021 tile has a bottom left vertex with the sum of coordinates congruent to
c ̸= 0 modulo 2021. By letting the tile be horizontal again without loss of generality, we can infer
analogously that the sum of the integers assigned to the cells of this tile is equal to
2020
X
f (i, j + k) = f (i, j + 2020 − c) + f (i, j + 2021 − c) = 1
k=0

Now, let C1 be the total number of cells contained withing the polygon and let C2 be the sum of
the integers f (i, j) assigned to those cells. Clearly, C1 and C2 are constants which only depend on
our polygon.

Consider an arbitrary tiling T and let NT be the number of of tiles whose bottom left vertex has
the sum of coordinates divisible by 2021. We therefore have C1 /2021 − NT tiles whose bottom left
vertex has the sum of coordinates congruent to some c ̸= 0 modulo 2021.

However, by the above observations, by summing up all integers f (i, j) assigned to the cells of our
polygon, grouped by tiles, we get that C2 = (C1 /2021 − NT ) · 1 − NT · 2020. We can conclude that
NT = (C1 /2021 − C2 )/2021 which is a constant depending solely on our polygon.

6
Problem 6. Let a, b, c be positive real numbers. For any positive integer n, prove that
4a + 2ab n 4b + 2bc n 4c + 2ac n
 2   2   2 
+ + ≥ an + bn + cn .
3a + 3b 3b + 3c 3c + 3a
(Orestis Lignos, Greece)

Solution. Observe that we can rewrite the general term of the sum as follows
n   n  
4a + 2ab n 2n
 2  2 n
2n X n a2k an+k

a n−k
X n
= n +a = n · a =
3a + 3b 3 a+b 3 k (a + b)k k (a + b)k
k=0 k=0

Doing the same for the other two fractions and adding cyclically, we get
n  
4a + 2ab n 4b + 2bc n 4c + 2ca n 2n X n
 2  2  2
an+k bn+k cn+k
   
+ + = n + + .
3a + 3b 3b + 3c 3c + 3a 3 k (a + b)k (b + c)k (c + a)k
k=0

We now proceed to prove the following claim:

Claim. For all a, b, c > 0 and for all k ≤ n we have
an+k bn+k cn+k an + bn + cn
+ + ≥ .
(a + b)k (b + c)k (c + a)k 2k
Proof. We begin by rewriting the left hand side as
an+k bn+k cn+k a2n b2n c2n
+ + = + + .
(a + b)k (b + c)k (c + a)k (a + b)k an−k (b + c)k bn−k (c + a)k cn−k
Furthermore, by applying the Cauchy Schwarz inequality, we obtain
a2n b2n c2n (an + bn + cn )2
+ + ≥ .
(a + b)k an−k (b + c)k bn−k (c + a)k cn−k (a + b)k an−k + (b + c)k bn−k + (c + a)k cn−k
Therefore, in order to prove the claim, it suffices to show that
(a + b)k an−k + (b + c)k bn−k + (c + a)k cn−k ≤ 2k (an + bn + cn ).
The above inequality may be further rewritten as
a + b k n−k b + c k n−k c + a k n−k
     
a + b + c ≤ an + bn + cn
2 2 2
However, by the AM-GM inequality and the Power Mean Inequality, we have
n n n
a + b k n−k (n − k)an + k(a+b) (n − k)an + k(a 2+b )
 
2n
a ≤ ≤
2 n n
By adding cyclically the above terms, we obtain the desired inequality.

Back to the problem, using our claim we obtain

n   n  
2n X n an+k bn+k cn+k 2n X n an + bn + cn

+ + ≥ n .
3n k (a + b)k (b + c)k (c + a)k 3 k 2k
k=0 k=0

Therefore, it remains to be proven that

n   n  !
2n X n 1 X
n−k n
≥ 1 ⇐⇒ 2 ≥ 3n .
3n k 2k k
k=0 k=0

However, the latter equation is obviously true, using the binomial expansion of (1 + 2)n . Thus, the
inequality is true. Equality holds when a = b = c, when n ≥ 1, and for all a, b, c > 0, when n = 0.