122/ Olympiad Corner

122/ OLYMPIAD CORNER
OLYMPIAD CORNER
No. 371
The problems in this section appeared in a regional or national mathematical Olympiad.
Click here to submit solutions, comments and generalizations to any

problem in this section
To facilitate their consideration, solutions should be received by June 15, 2019.
OC416. Mim has a deck of 52 cards, stacked in a pile with their backs facing
up. Mim separates the small pile consisting of the seven cards on the top of the
deck, turns it upside down, and places it at the bottom of the deck. All cards are
again in one pile, but not all of them face down; the seven cards at the bottom
do, in fact, face up. Mim repeats this move until all cards have their backs facing
up again. In total, how many moves did Mim make?
OC417. A 2017 × 2017 table is filled with nonzero digits. Among the 4034
numbers whose decimal expansion is formed with the rows and columns of this
table, read from left to right and from top to bottom, respectively, all but one are
divisible by a prime number p, and the remaining number is not divisible by p.
Find all possible values of p.
OC418. There are 100 gnomes with weight 1, 2, . . . , 100 kg gathered on the
left bank of the river. They cannot swim, but they have one boat with capacity
100 kg. Because of the current, it is hard to row back, so each gnome has enough
power only for one passage from right side to left as oarsman. Can all gnomes get
to the right bank?
OC419. Let n be a nonzero natural number, let a1 < a2 < . . . < an be real
numbers and let b1 , b2 , . . . , bn be real numbers. Prove that:
(a) if all the numbers bi are positive, then there exists a polynomial f with real
coefficients and having no real roots such that f (ai ) = bi for i = 1, 2 . . . , n;
(b) there exists a polynomial f of degree at least 1 having all real roots and such
that f (ai ) = bi for i = 1, 2 . . . , n.
OC420. Consider a triangle ABC with ∠A < ∠C. Point E is on the internal
angle bisector of ∠B such that ∠EAB = ∠ACB. Let D be a point on line BC
such that B ∈ CD and BD = AB. Prove that the midpoint M of the segment
AC is on the line DE.
.................................................................
Crux Mathematicorum, Vol. 45(3), March 2019

OLYMPIAD CORNER /123
Les problèmes présentés dans cette section ont déjà été présentés dans le cadre d’une
olympiade mathématique régionale ou nationale.
Cliquez ici afin de soumettre vos solutions, commentaires ou

généralisations aux problèmes proposés dans cette section.
Pour faciliter l’examen des solutions, nous demandons aux lecteurs de les faire parvenir
au plus tard le 15 juin 2019.
La rédaction souhaite remercier Rolland Gaudet, professeur titulaire à la retraite à
l’Université de Saint-Boniface, d’avoir traduit les problèmes.
OC416. Mim a devant elle un paquet de 52 cartes, dans une pile avec les
versos vers le haut. Elle retire la toute petite pile de sept cartes du haut, la renverse
en un mouvement et la place au fond du paquet. Les cartes sont ainsi de nouveau
en une seule pile, mais les versos ne sont plus tous vers le haut; les sept cartes
au fond ont leurs versos vers le bas. Mim répète la même série de déplacements,
jusqu’à ce que tous les versos se retrouvent de nouveau vers le haut. Déterminer
le nombre de déplacements effectués.
OC417. Un tableau 2017 × 2017 consiste de chiffres non nuls. Les rangées,
lues de gauche à droite, et les colonnes, lues du haut vers le bas, donnent les
représentations décimales de 4034 entiers. Pour un certain nombre premier p,
tous ces 4034 entiers sont divisibles par p, sauf un seul. Déterminer toute valeur
possible de p.
OC418. Se trouvent 100 gnomes de poids 1, 2, . . . , 100 kg sur la rive gauche

d’une rivière. Ils ne peuvent pas nager, mais ils disposent d’un bateau de capacité
totale de 100 kg. Aucun gnome peut faire le passage de retour de la rive droite
vers la rive gauche plus d’une fois. Est-ce possible pour tous les 100 gnomes de se
rendre de la rive gauche à la rive droite ?
OC419. Soit n un nombre naturel non nul et soient a1 < a2 < . . . < an puis
b1 , b2 , . . . , bn des nombres réels. Démontrer que:
(a) si tous les nombres bi sont positifs, alors il existe un polynôme f à coefficients
réels et n’ayant aucune racine réelle tel que f (ai ) = bi pour i = 1, 2 . . . , n;
(b) il existe un polynôme de degré au moins 1 ayant seulement des racines réelles,
tel que f (ai ) = bi pour i = 1, 2 . . . , n.
OC420. Soit un triangle ABC tel que ∠A < ∠C. Le point E se trouve sur la
bissectrice interne de ∠B de façon à ce que ∠EAB = ∠ACB. Enfin, soit D un
point sur la ligne BC tel que B ∈ CD et BD = AB. Démontrer que le mi point
M du segment AC se trouve sur la ligne DE.
Copyright © Canadian Mathematical Society, 2019

OLYMPIAD CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2018: 44(2), p. 52–53,
and 44(3), p. 100–101.
OC366. Prove that there exist infinitely many positive integer triples (a, b, c)
such that a, b, c are pairwise relatively prime, and ab + c, bc + a, ca + b are pairwise
relatively prime.
Originally 2016 China Western Mathematical Olympiad Day 2 Problem 5.
We received 6 submissions. We present 3 solutions.
Solution 1, by Shuborno Das.
Let a = 2, b = 3 and c be any prime number greater than 3 such that c ≡ 4
(mod 5). We claim that ab + c, bc + a, ca + b are pairwise co-prime for the above
choice. Note that gcd(2, 3) = gcd(2, c) = gcd(3, c) = 1. Now,
ab + c = c + 6, bc + a = 3c + 2, ca + b = 2c + 3.
Therefore,
gcd(c + 6, 3c + 2) = gcd(3c + 18, 3c + 2) = gcd(16, 3c + 2) = 1.
(Note gcd(3, 3c + 2) = 1).

We have
gcd(c + 6, 2c + 3) = gcd(2c + 12, 2c + 3) = gcd(9, 2c + 3) = 1.
(Note gcd(2, 2c + 3) = 1).

We also have
gcd(3c + 2, 2c + 3) = gcd(6c + 4, 2c + 3) = gcd(6c + 4, 6c + 9) = gcd(5, 6c + 4).
Now, 6c+4 ≡ c−1 (mod 5). We have assumed that c ≡ 4 (mod 5), that is c−1 6≡ 0
(mod 5), so gcd(5, 6c + 4) = 1. By Dirichlet’s theorem, there are infinitely many
primes of the form 5k + 4. This means there are infinitely many triples (a, b, c)
satisfying the condition.
Solution 2, by Oliver Geupel.

It is enough to show that, for every positive integer k, the triple
(a, b, c) = (10k − 1, 10k, 10k + 1)

satisfies the required conditions.

Evidently, a, b, and c are pairwise relatively prime. Let
n = 10k, u = ab + c, v = bc + a, w = ca + b.
Then,
(u, v, w) = (n2 + 1, n2 + 2n − 1, n2 + n − 1).
We have
(n + 3)u − (n + 1)v = 4,
so that every common divisor of u and v is a divisor of 4. Since u = 100k 2 + 1 is
odd, gcd(u, v) = 1. From the identity
(n + 1)v − (n + 2)w = 1,
we see that v and w are coprime. Finally, we have
(n + 3)u − (n + 2)w = 5,
which implies that every common divisor of u and w is a divisor of 5. But u is

not divisible by 5. Therefore, u and w are relatively prime, which completes the
proof.
Solution 3, by David Manes.

For each positive integer n, let an = 2n , bn = 3 and cn = 5. Then
(an , bn , cn ) = (2n , 3, 5)
define infinitely many triples such that gcd(2n , 3) = gcd(2n , 5) = gcd(3, 5) = 1.

Therefore, (2n , 3, 5) are pairwise relatively prime for each n. For a given positive
integer n, let
x = an bn + cn = 3 · 2n + 5, y = bn cn + an = 3 · 5 + 2n , z = cn an + bn = 5 · 2n + 3.
Then x, y and z are all odd positive integers. We consider 3 cases.

1) Let d = gcd(x, y). Then d divides all linear combinations of x and y. In
particular, d divides
3x − 1 · y = 9 · 2n − 2n = 2n+3 .
Hence, d = 1 since the only odd number that divides a power of 2 is 1. Therefore,
x and y are relatively prime.
2) Let d = gcd(x, z). Then d divides the linear combination
5x − 3z = 25 − 9 = 24 .
Hence, d = 1 so that x and z are relatively prime.

3) Let d = gcd(y, z). Then d divides
5z − y = 24 · 2n = 3 · 2n+3 .
The only odd positive divisors of 3 · 2n+3 are 1 and 3. If d = 3, then d divides y
and d divides 15 implies d divides y − 15 = 2n , a contradiction. Hence, d = 1 so
that y and z are relatively prime.
Therefore, x, y and z are pairwise relatively prime for each positive integer n.
Summarizing, if for each positive integer n, (2n , 3, 5) define infinitely many positive
integer triples that are pairwise relatively prime, then the triples
(3 · 2n + 5, 5 · 3 + 2n , 5 · 2n + 3)
are also pairwise relatively prime.

Editor’s Comments. These are not the only triples that satisfy the conditions
of the problem. Mohammed Aassila found the triples (a, b, c) = (2, 3, 30k + 5),
where k ∈ N, Richard Hess found the triples (a, b, c) = (2, 3, 30k + 7), where
k ∈ N and the Missouri State University Problem Solving Group found the triples
(a, b, c) = (10k + 5, 10k + 3, 2), where k ∈ N. We leave to the reader the pleasure
to check that these triples will do the trick.
OC367. A mathematical contest had 3 problems, each of which was given

a score between 0 and 7, inclusive. It is known that, for any two contestants,
there exists at most one problem in which they have obtained the same score (for
example, there are no two contestants whose ordered scores are 7, 1, 2 and 7, 1, 5,
but there might be two contestants whose ordered scores are 7, 1, 2 and 7, 2, 1).
Find the maximum number of contestants.
Originally 2016 Italian Mathematical Olympiad Problem 2.
We received 2 submissions. We present the solution by Mohammed Aassila.
Consider the lattice points (x, y, z) such that x, y, z ∈ [0, 7] and x, y, z ∈ Z. Colour
a lattice point (x, y, z) red if there exists a student who got x, y, and z as their
scores for the first, second, and third problems, respectively.
The problem condition is equivalent to the condition that if (x, y, z) is red, then
(x, y, k), (x, k, z), (k, y, z) are all not red for k ∈ [0, 7] (with the exception of when
the point is equal to (x, y, z)), which correspond to the up-down, left-right, and
front-back rows that contain (x, y, z).
Looking at all lattice points (x, y, k) such that x, y ∈ [0, 7] and k is a constant, by
Pigeonhole principle there exists at most 8 red lattice points, else there exists two
red lattice points in the same row, contradiction.
Thus, the maximum possible is 8×8 = 64 and it remains to show this is attainable.
Indeed, just take all points of the form (x, y, x − y) where values are calculated
mod 8 and x, y ∈ [0, 7].

OC368. Let n be a positive integer. Find the number of solutions of

x2 + 2016y 2 = 2017n
as a function of n.
Originally 2016 Korea National Olympiad Day 1 Problem 1.
Problem in not complete as stated. The statement should specify that the solutions
sought are positive integer solutions.
We received one correct submission. We present the solution by Shengda Hu.
For n = 1, the only integral solutions are (±1, ±1).
We see that the following generate integral solutions of the equation for integers
l, m ≥ 0:
√ √
am,l + 2bm,l −504 := 2017m (1 + 2 −504)l =⇒ a2m,l + 2016b2m,l = 2017l+2m
We prove that for each n the set of integral solutions is
{(±am,l , ±bm,l ), where l + 2m = n} .
For k > 1, assume that solutions for n < k are given above. Let (u, v) be a solution
of the equation for n = k > 1, we compute
u2 + 2016v 2 = 2017k =⇒ u2 ≡ v 2 (mod 2017) =⇒ u ≡ ±v (mod 2017) .
Suppose that u ≡ v mod 2017 (the other case reduces to this one by changing the
sign of v), then
√ √ √
u + 2v −504 (u + 2v −504)(1 − 2 −504) u + 2016v v − u√
√ = = +2 −504,
1 + 2 −504 2017 2017 2017
which implies that
u + 2016v v − u
Å ã
(a, b) := ,
2017 2017
is an integral solution to the equation for n = k − 1. By the inductive assumption,
we see that
(a, b) ∈ {2017m (±al , ±bl ) : l + 2m = k − 1}
and so √ √
a + 2b −504 ∈ {±2017m (1 ± 2 −504)l : l + 2m = k − 1}
and
√ √ √
u + 2v −504 ∈ {±2017m (1 ± 2 −504)l (1 + 2 −504) : l + 2m = k − 1}.
We compute that
√ √ √ √
±(1 + 2 −504)l (1 + 2 −504) = ±(1 + 2 −504)l+1 = ±(a0,l+1 + 2b0,l+1 −504),
√ √ √
±(1 − 2 −504)l (1 + 2 −504) = ±2017(1 − 2 −504)l−1
√
= ±2017(a0,l−1 − 2b0,l−1 −504).

Thus (u, v) is of the claimed form.

For each (m, l), among the four solutions (±am,l , ±bm,l ), there is only one positive
solution. It follows that the number of positive integral solutions to the equation in
the problem for n > 0 coincides with the number of non-negative integral solutions
to the equation below
l + 2m = n, where l, m ≥ 0
jnk
Thus the original equation has + 1 positive solutions.
2
We note that the explicit description above also shows that the number of integer
solutions to the equation in the problem is 2(n + 1).
OC369. Let I be the incenter of 4ABC. Let D be the point of intersection

of AI with BC and let S be the point of intersection of AI with the circumcircle
of ABC (S 6= A). Let K and L be incenters of 4DSB and 4DCS. Let P be a
reflection of I with respect to KL. Prove that BP ⊥ CP .
Originally 2016 Polish Mathematical Olympiad Finals Day 2 Problem 6.
We received 3 submissions. We present the solution by Shuborno Das.
We first prove that BKDI and CLDI are cyclic.

As K and I are incenters of 4DSB and 4ABC, respectively, we have
∠A + ∠B
Å ã
∠BKD + ∠BID = (180◦ − ∠DBK − ∠BDK) +
2
∠A + ∠B
Å ã Å ã
◦ ∠DBS ∠A ∠B
= 180 − − − +
2 4 2 2
∠A + ∠B
Å ã Å ã
∠A ∠B
= 180◦ − − +
2 2 2
◦
= 180 ,
where the penultimate equality follows from
∠DBS = ∠CBS = ∠A/2.
So, BKDI is cyclic and similarly CLDI is cyclic.

Next, we prove that the circumcenter of 4BP I is K and the circumcenter of
4CP I is L. We have
∠A ∠B
∠SBI = ∠SBC + ∠CBI = + = ∠BAI + ∠ABI = ∠SIB,
2 2
so SB = SI. As K is the incenter of 4DSB, then KS bisects ∠ISB, which gives
KI = KB.
As point P is the reflection of point I over the line KL, then IK = P K, so
KB = IK = P K. Therefore, the circumcenter of 4BP I is point K and similarly
the circumcenter of 4CP I is point L.
Now, we have
∠BKI + ∠CLI ∠BDI + ∠CDI ∠BDC

∠BP C = ∠BP I+∠CP I = = = = 90◦ ,
2 2 2
which gives the desired conclusion.
OC370. Integers n and k are given, with n ≥ k ≥ 2. You play the following
game against an evil wizard. The wizard has 2n cards; for each i = 1, . . . , n, there
are two cards labeled i. Initially, the wizard places all cards face down in a row,
in unknown order. You may repeatedly make moves of the following form: you
point to any k of the cards. The wizard then turns those cards face up. If any
two of the cards match, the game is over and you win. Otherwise, you must look
away, while the wizard arbitrarily permutes the k chosen cards and then turns
them back face-down. Then, it is your turn again.
We say this game is winnable if there exist some positive integer m and some
strategy that is guaranteed to win in at most m moves, no matter how the wizard
responds. For which values of n and k is the game winnable?
Originally 2016 USAMO Day 2 Problem 6.

We received 2 submissions. We present the solution by Oliver Geupel.

We assert that the game is winnable if and only if k < n.
First, suppose k < n. We give a way to win in not more than n + 3 moves.
Denote the positions of cards by the integers 1, . . . , 2n. In the first move, we point
to the cards at positions 1, . . . , k. If the game is not over, we point to the cards
at positions 2, . . . , k + 1 in the second move. By inspecting the card at position
k + 1 and the distinction of the labels shown in both moves, we can determine
the label of the card at position 1. We proceed, if necessary, pointing to cards
positions 3, . . . , k + 2 in the third move and so on. We point to cards at positions
n + 2, . . . , n + k + 1 in the (n + 2)nd move which gives us enough information to
determine the label of the card at position n + 1.
At that time, we have learned the labels of n+1 cards. By the Pigeonhole principle,
we know two matching cards, and we can point to them and k − 2 redundant cards
in the (n + 3)rd move, which wins the game.
It remains to show that there is no winning strategy when k = n.
Let P1 be the set of positions of cards we point to in the first move. Assume
that there is no match in the first move. Then, after the first move, we have the
situation that we know a collection P1 of n positions with no matching pair of
cards, but we do not know the specific position of any label in P1 . The latter is
because the wizard has arbitrarily permuted the n chosen cards.
In the second move, it will not help to point to the set P1 or to its complement P1 ,
which has also no matching pair. So we will point to a combination P2 of some
positions in P1 and some positions in P1 . It may happen that the positions chosen
from P1 give exactly the labels that are missing at the positions chosen from P1 .
Then, after the second move, we have the situation that we know a collection P2
of n positions with no matching pair of cards, but we do not know the specific
position of any label in P2 .
Proceeding this way, it may happen that, after every move, we know a set of n
positions with no matching pair of cards, but we do not know the specific position
of any label in this set. This prevents us from winning certainly in the next move.
So there is no fixed number m of moves, in which we can certainly win the game.
OC371. Let a, b and c ∈ R+ such that abc = 1. Prove that
a+b b+c c+a 2

2
+ 2
+ 2
≥ .
(a + b + 1) (b + c + 1) (c + a + 1) a+b+c
Originally 2016 Iranian Mathematical Olympiad 3rd Round Algebra Problem 2.

We received 5 correct submissions. We present the solution submitted indepen-
dently by Šefket Arslanagić and Shuborno Das.

The inequality in the statement will be obtained via AM-GM inequality and
Cauchy-Schwarz inequality.
First, we establish that for any positive real numbers x, y, and, z, we have
xyz xyz xyz
+ 3 + 3 ≤1 (1)
x3 + y 3 + xyz y + z 3 + xyz z + x3 + xyz
Indeed AM-GM inequality implies that
x3 + x3 + y 3 x3 + y 3 + y 3
≥ x2 y, ≥ xy 2
3 3
and so x3 + y 3 ≥ x2 y + xy 2 . Similarly
y 3 + z 3 ≥ y 2 z + yz 2 and z 3 + x3 ≥ z 2 x + zx2 .
These consequences of AM-GM inequality imply and prove (1)

xyz xyz xyz
+ 3 + 3
x3 + y 3 + xyz y + z 3 + xyz z + x3 + xyz
xyz xyz xyz
≤ + 2 + 2
x2 y + xy 2 + xyz y z + yz 2 + xyz z x + zx2 + xyz
xyz xyz xyz
= + +
xy(x + y + z) yz(y + z + x) zx(z + x + y)
z+x+y
=
x+y+z
= 1.
Next we take a = x3 , b = y 3 , and c = z 3 in (1). Taking into consideration that

abc = 1, we obtain
1 1 1
+ + ≤ 1,
a+b+1 b+c+1 c+a+1
1 1 1
2=3−1≤1− +1− +1− and,
a+b+1 b+c+1 c+a+1
!2
a+b b+c c+a
4≤ + + (2)
a+b+1 b+c+1 c+a+1
Lastly, Cauchy-Schwarz inequality provides an upper bound for the right side of
inequality (2), and allows us to write
" #
a+b b+c c+a
4 ≤ [(a + b) + (b + c) + (c + a)] + + (3)
(a + b + 1)2 (b + c + 1)2 (c + a + 1)2
After dividing both sides of (3) by 2(a + b + c), we obtain the statement inequality.
Equality holds if and only if a = b = c = 1.

OC372. In the circumcircle of a triangle ABC, let A1 be the point diamet-

rically opposite to the vertex A. Let A0 be the intersection point of A01 and BC.
The perpendicular to the line AA1 from A0 meets the sides AB and AC at M and
N , respectively. Prove that the points A, M, A1 and N lie on a circle whose center
lies on the altitude from A of the triangle ABC.
Originally 2016 Spain Mathematical Olympiad Day 1 Problem 3.
We received 4 correct submissions. We present the solution by Ivko Dimitrić.
We solve the question in several cases.
Case 1. If ∠B = ∠C, then 4ABC is isosceles, M = B, and N = C. In this case,
it is obvious that the statement holds.
Case 2. Assume ∠B < ∠C. Let D be the foot of the altitude from A.
Case 2a. Assume ∠B < ∠C < 90◦ . If vertices A, B, and C are arranged
in counterclockwise order, then rays AB, AA1 , AD, and AB are also in counter-
clockwise order. Since AA1 is a diameter of the circumcircle, we have
∠M BA1 = ∠ABA1 = 90◦
and
∠N CA1 = 180◦ − ∠ACA1 = 90◦ .
In addition, as AA0 ⊥ M N , ∠M A0 A1 = 90◦ and ∠N A0 A1 = 90◦ . Therefore the

quadrilaterals M BA1 A0 and N CA0 A1 are cyclic, inscribed in circles whose centers
are midpoints of M A1 and N A1 , respectively.
Furthermore, ∠AM N = ∠BA1 A = ∠C as an exterior angle of the cyclic quadri-
lateral M BA1 A0 , and ∠AA1 N = ∠C, as ∠C is an exterior angle of cyclic quadri-
lateral N CA0 A1 . As ∠AM N = ∠AA1 N it follows that AM A1 N is a cyclic quadri-
lateral.
Let P be the center of the circumcircle of AM A1 N . We have
180◦ − ∠AP N 180◦ − 2∠AM N

∠N AP = = = 90◦ − ∠C.
2 2
On the other hand, since AD ⊥ BC, ∠CAD = 90◦ −∠C. Hence ∠N AP = ∠CAD
and P belongs to AD . Its location is found as the intersection point of the
perpendicular bisector of the diameter AA1 with altitude AD.
Case 2b. Assume ∠B < ∠C = 90◦ . Then A0 = A1 = M = B and the quadri-
lateral AM A1 N degenerates into a right triangle, 4ABN , with hypotenuse, AN ,
along AC. However AC is the altitude from A of 4ABC. Hence the circumcenter
of quadrilateral AM A1 N , now 4ABN , belongs to that altitude from A of 4ABC.
Case 2c. Assume ∠B < 90◦ < ∠C. The proof for this case follows closely the
proof of Case 2a. If vertices A, B, and C are arranged in counterclockwise order,

then rays AA1 , AB, AC, and AD are also in counterclockwise order and P and
B are on different sides of AC. Moreover, ∠CAD = ∠C − 90◦ and
∠AP N = 2 (180◦ − ∠AM N ) = 2 ∠BM A0 = 2 ∠BA1 A = 2 (180◦ − ∠C),
leading to
180◦ − ∠AP N
∠N AP = = ∠C − 90◦ = ∠CAD,
2
so, again, P belongs to the altitude AD.
OC373. Let a and b be positive integers. Denote by f (a, b) the number of

sequences s1 , s2 , . . . , sa ∈ Z such that |s1 | + |s2 | + · · · + |sa | ≤ b. Show that
f (a, b) = f (b, a).
Originally 2016 Polish Mathematical Olympiad Finals Day 1 Problem 3.
We received 2 correct submissions. We present both solutions modified for editorial
presentation.
Solution 1, by Mohammed Aassila.
Let f (a, b, 0) = 1. Let k be an integer such that 1 ≤ k ≤ min(a, b). Let f (a, b, k) be
the number of a-tuples, (s1 , s2 , . . . , sa ), of integers such that |s1 |+|s2 |+· · ·+|sa | ≤ b
and exactly k of s1 , . . . , sa are nonzero. Such a tuple is uniquely defined by the
k positions of the nonzero si ’s, the signs, + or −, of the nonzero si ’s, and an
ordered k-tuple of positive integers with sum b or less. There are ka ways to
choose the unordered subset of k positions from the fixed set of positions 1, . . . , a.
There are two ways to choose the sign of a specific nonzero si , and there are
2k ways to choose the signs of the nonzero si ’s. There are kb ways to choose
the ordered k-tuple of positive integers that sum up to b or less. This counting
result follows from the next observation. Choosing the ordered k-tuple of nonzero
positive integers u1 , . . . , uk with u1 + · · · + uk ≤ b is equivalent to choosing k
numbers u1 , u1 + u2 , . . . , u1 + · · · + uk , disregarding the order, from b numbers
1, . . . , b.
By the multiplicative priciple of counting
Ç åÇ å
a b
f (a, b, k) = 2k ,
k k
for any a ≥ 1, b ≥ 1, and k ≥ 1. Moreover, the above formula holds for k = 0, as

f (a, b, 0) = 1.
Subsequently,
min(a,b) min(a,b) Ç åÇ å
X X
k a b
f (a, b) = f (a, b, k) = 2 .
k k
k=0 k=0
As the above is symmetric in a and b, it follows that f (b, a) = f (a, b) for any
a ≥ 1, b ≥ 1.

Solution 2, by Missouri State University Problem Solving Group.

First, f (a, b) = f (b, a) is true if either a = 1 or b = 1. This is because the number
of 1-tuples, (s1 ), with |s1 | ≤ b is 2b + 1, hence f (1, b) = 2b + 1. Also, the number
of a-tuples such that |s1 | + · · · + |sa | ≤ 1 is 2a + 1, hence f (a, 1) = 2a + 1.
For a ≥ 2 and b ≥ 2, f (a, b) = f (b, a) is established as a consequence of the more
general result
f (a, b) = f (a, b − 1) + f (a − 1, b) + f (a − 1, b − 1). (1)
Relation (1) follows from writing the set T of a-tuples (s1 , . . . , sa ) of integers such
that |s1 | + · · · + |sa | ≤ b as the union of three mutually exclusive sets. Namely,
T1 = {(s1 , . . . , sa ) | |s1 | + · · · + |sa | ≤ b − 1},
T2 = {(s1 , . . . , sa ) | |s1 | + · · · + |sa | = b, sa > 0},

T3 = {(s1 , . . . , sa ) | |s1 | + · · · + |sa | = b, sa ≤ 0}.
The cardinal of set T1 is f (a, b − 1). The set T2 is in one-to-one correspondence

with
{(s1 , . . . , sa−1 ) | |s1 | + · · · + |sa−1 | ≤ b}
and has cardinal f (a − 1, b). Lastly, T3 is in one-to-one correspondence with
{(s1 , . . . , sa−1 ) | |s1 | + · · · + |sa−1 | ≤ b − 1}
and has cardinal f (a − 1, b − 1). Relation (1) follows.
Editor’s Comments. Solution 2 was submitted as a proof by induction. However,

the editor felt that the induction step can be avoided.
OC374. Let p be an odd prime and let a1 , a2 , . . . , ap be integers. Prove that

the following two conditions are equivalent:
p−1
1) There exists a polynomial P (x) of degree less than or equal to 2 such that
P (i) ≡ ai (mod p) for all 1 ≤ i ≤ p.
p−1
2) For any natural number d ≤ 2 ,
p
X
(ai+d − ai )2 ≡ 0 (mod p) ,
i=1
where indices are taken modulo p.

Originally 2016 China National Olympiad Day 1 Problem 3.
We received one correct submission. We present the solution by Mohammed Aas-
sila with editorial changes.

First, we establish the following two results.

(A) Let p be a prime number, then for natural numbers k ≥ 0
p ß
X −1 (mod p) if p − 1 | k
S= ik ≡
0 (mod p) if p − 1 - k.
i=1
Proof of A. If p − 1 | k, the result follows from Fermat’s little theorem. For any
1 ≤ i ≤ p − 1, i is not divisible by p, hence we have ip−1 ≡ 1(mod p), and
ik ≡ 1(mod p). Therefore
p
X
ik ≡ p − 1 + pk (mod p) ≡ −1(mod p).
i=1
If p − 1 - k, consider g a primitive root of p. Any odd prime number has a

primitive root. A primitive root is an integer number, g, that is coprime with p,
such that the powers g, g 2 , . . . , g p−1 are congruent to 1, 2, . . . , p − 1 modulo p, not
necessarily in the specified order. In addition, by Fermat’s little theorem we know
that g p−1 ≡ 1(mod p). Therefore,
p
X p−1
X p
X
gk S = gk ik ≡ g k (g i )k (mod p) ≡ (g i )k (mod p) ≡ S(mod p).
i=1 i=1 i=2
Since p − 1 - k, we have g k 6≡ 1(mod p). This together with g k S ≡ S(mod p)

implies that S ≡ 0(mod p), which establishes the second part of (A).
(B) Let Q(x, d) = (P (x+d)−P (x))2 . Let k be the degree of P as a polynomial in x.
Then Q(x, d) = m Cm (d)xm can be viewed as a polynomial in x with coefficients
P
Cm (d) that are polynomials in d. The following can easily be established:
(B1) The degree of Q in x is less than or equal to 2(k − 1).
(B2) The degree of Q in d is less than or equal to 2k.
(B3) The degree of Cm (d) is less than or equal to 2k − m.
We now will prove the main results.
(1) ⇒ (2)
Fix d. Note that the sum in question is pi=1 Q(i), where Q(x) is the polynomial
P
defined at (B). The degree of Q in x is less than or equal to
2(deg(P ) − 1) = 2((p − 1)/2 − 1) = p − 3.
Therefore pi=1 Q(i) is a linear combination of pi=1 ik , where 0 ≤ k ≤ p − 3. By

P P
result (A),
Xp
Q(i) ≡ 0(mod p).
i=1

(2) ⇒ (1)
Lagrange polynomials and modular multiplicative inverse under p can be used to
construct a polynomial P of degree at most p − 1, such that P (i) ≡ ai (mod p) for
any 1 ≤ i ≤ p. The polynomial P can be thought of as interpolating the data set
of points {(i, ai ), 1 ≤ i ≤ p} mod p.
Assuming (2), we will prove by contradiction that the degree of P is at most
(p − 1)/2, mod p.
Assume deg(P ) ≥ (p − 1)/2 + 1, mod p. Let k be the degree of P mod p, and
let R be the coefficient of xk in P , R 6≡ 0(mod p) . From (B) we know that the
degree of Q in x is at most 2(k − 1) ≤ 2(p − 2). Note that since (2) is valid for any
1 ≤ d ≤ (p − 1)/2, in fact, it is valid for 0 ≤ d ≤ p − 1. Due to (A) and (2) we have
that the coefficient of xp−1 in Q(x), Cp−1 (d) ≡ 0(mod p) for any 0 ≤ d ≤ p − 1.
However, Cp−1 (d) is a polynomial in d of degree at most 2k − (p − 1) ≤ p − 1.
Having more roots modulo p than its degree, it follows that Cp−1 (d) is the zero
polynomial modulo p and that all its coefficients are 0 modulo p.
The coefficient of d2k−p+1 , the monomial with largest exponent in Cp−1 (d), origi-
nates from the expansion of R2 (x + d)2k − 2R2 (x + d)k xk and is a combination of
binomial coefficients
Ç å Ç å
2 2k k
R −2 .
p−1 2k − p + 1
2k k

However, notice that p divides p−1 and does not divide 2k−p+1 , since p+1 ≤ 2k
and k ≤ p − 1. It follows that R ≡ 0(mod p) which is a contradiction. It must be
that deg(P ) ≤ (p − 1)/2.
Editor’s Comments. The question and its solution can be better understood when
viewed in the finite field, Zp , of integers mod p.
OC375. Let ABCD be a non-cyclic convex quadrilateral with no parallel

sides. Suppose the lines AB and CD meet in E. Let M 6= E be the intersection
of circumcircles of ADE and BCE. Further, suppose that the internal angle
bisectors of ABCD form a convex cyclic quadrilateral with circumcenter I while
the external angle bisectors of ABCD form a convex cyclic quadrilateral with
circumcenter J. Show that I, J, M are collinear.
Originally 2016 Brazil National Olympiad Day 2 Problem 6.
We received no solutions to this problem.

122/ Olympiad Corner

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

122/ Olympiad Corner

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

122/ Olympiad Corner

Uploaded by

Copyright:

Available Formats

122/ OLYMPIAD CORNER

Click here to submit solutions, comments and generalizations to any

To facilitate their consideration, solutions should be received by June 15, 2019.

Crux Mathematicorum, Vol. 45(3), March 2019

Cliquez ici afin de soumettre vos solutions, commentaires ou

OC418. Se trouvent 100 gnomes de poids 1, 2, . . . , 100 kg sur la rive gauche

Copyright © Canadian Mathematical Society, 2019

gcd(c + 6, 3c + 2) = gcd(3c + 18, 3c + 2) = gcd(16, 3c + 2) = 1.

(Note gcd(3, 3c + 2) = 1).

gcd(c + 6, 2c + 3) = gcd(2c + 12, 2c + 3) = gcd(9, 2c + 3) = 1.

(Note gcd(2, 2c + 3) = 1).

gcd(3c + 2, 2c + 3) = gcd(6c + 4, 2c + 3) = gcd(6c + 4, 6c + 9) = gcd(5, 6c + 4).

Solution 2, by Oliver Geupel.

(a, b, c) = (10k − 1, 10k, 10k + 1)

Crux Mathematicorum, Vol. 45(3), March 2019

satisfies the required conditions.

we see that v and w are coprime. Finally, we have

which implies that every common divisor of u and w is a divisor of 5. But u is

Solution 3, by David Manes.

define infinitely many triples such that gcd(2n , 3) = gcd(2n , 5) = gcd(3, 5) = 1.

Then x, y and z are all odd positive integers. We consider 3 cases.

Hence, d = 1 so that x and z are relatively prime.

Copyright © Canadian Mathematical Society, 2019

3) Let d = gcd(y, z). Then d divides

are also pairwise relatively prime.

OC367. A mathematical contest had 3 problems, each of which was given

Crux Mathematicorum, Vol. 45(3), March 2019

OC368. Let n be a positive integer. Find the number of solutions of

Copyright © Canadian Mathematical Society, 2019

Thus (u, v) is of the claimed form.

OC369. Let I be the incenter of 4ABC. Let D be the point of intersection

We first prove that BKDI and CLDI are cyclic.

Crux Mathematicorum, Vol. 45(3), March 2019

As K and I are incenters of 4DSB and 4ABC, respectively, we have

where the penultimate equality follows from

∠DBS = ∠CBS = ∠A/2.

So, BKDI is cyclic and similarly CLDI is cyclic.

∠BKI + ∠CLI ∠BDI + ∠CDI ∠BDC

Copyright © Canadian Mathematical Society, 2019

We received 2 submissions. We present the solution by Oliver Geupel.

OC371. Let a, b and c ∈ R+ such that abc = 1. Prove that

a+b b+c c+a 2

Originally 2016 Iranian Mathematical Olympiad 3rd Round Algebra Problem 2.

Crux Mathematicorum, Vol. 45(3), March 2019

These consequences of AM-GM inequality imply and prove (1)

Next we take a = x3 , b = y 3 , and c = z 3 in (1). Taking into consideration that

Copyright © Canadian Mathematical Society, 2019

OC372. In the circumcircle of a triangle ABC, let A1 be the point diamet-

∠M BA1 = ∠ABA1 = 90◦

In addition, as AA0 ⊥ M N , ∠M A0 A1 = 90◦ and ∠N A0 A1 = 90◦ . Therefore the

180◦ − ∠AP N 180◦ − 2∠AM N

Crux Mathematicorum, Vol. 45(3), March 2019

∠AP N = 2 (180◦ − ∠AM N ) = 2 ∠BM A0 = 2 ∠BA1 A = 2 (180◦ − ∠C),

OC373. Let a and b be positive integers. Denote by f (a, b) the number of

for any a ≥ 1, b ≥ 1, and k ≥ 1. Moreover, the above formula holds for k = 0, as

Copyright © Canadian Mathematical Society, 2019

Solution 2, by Missouri State University Problem Solving Group.