Volume/tome 51, issue/numéro 1

January/janvier 2025
Crux Mathematicorum is a problem-solving journal at the secondary and university undergraduate levels,
published online by the Canadian Mathematical Society. Its aim is primarily educational; it is not a research
journal. Online submission:
https://publications.cms.math.ca/cruxbox/

Crux Mathematicorum est une publication de résolution de problèmes de niveau secondaire et de premier
cycle universitaire publiée par la Société mathématique du Canada. Principalement de nature éducative,
le Crux n’est pas une revue scientifique. Soumission en ligne:
https://publications.cms.math.ca/cruxbox/

The Canadian Mathematical Society grants permission to individual readers of this publication to copy articles for
their own personal use.
c CANADIAN MATHEMATICAL SOCIETY 2025. ALL RIGHTS RESERVED.
ISSN 1496-4309 (Online)
La Société mathématique du Canada permet aux lecteurs de reproduire des articles de la présente publication à des
fins personnelles uniquement.

c SOCIÉTÉ MATHÉMATIQUE DU CANADA 2025. TOUS DROITS RÉSERVÉS.

ISSN 1496-4309 (électronique)

Editorial Board

Editor-in-Chief Kseniya Garaschuk University of the Fraser Valley

MathemAttic Editors John Grant McLoughlin University of New Brunswick

Shawn Godin Cairine Wilson Secondary School
Olympiad Corner Editors Alessandro Ventullo University of Milan
Anamaria Savu University of Alberta
Articles Editor Robert Dawson Saint Mary’s University
Associate Editors Edward Barbeau University of Toronto
Chris Fisher University of Regina
Dennis D. A. Epple Toronto, Canada
Magdalena Georgescu Toronto, Canada
Chip Curtis Missouri Southern State University
Philip McCartney Northern Kentucky University
Guest Editors Yagub Aliyev ADA University, Baku, Azerbaijan
Almaz Butaev University of the Fraser Valley
Mateusz Buczek Warsaw, Poland
Ana Duff Ontario Tech University
Andrew McEachern York University
Matt Olechnowicz Concordia University
Vasile Radu Birchmount Park Collegiate Institute
Chi Hoi Yip Georgia Institute of Technology
Translators Rolland Gaudet Université de Saint-Boniface
Frédéric Morneau-Guérin Université TÉLUQ
Editor-at-Large Bill Sands University of Calgary
 IN THIS ISSUE / DANS CE NUMÉRO

4 Happy New Year

5 MathemAttic: No. 61
5 Problems: MA301–MA305
8 Solutions: MA276–MA280
13 Problem Solving Vignettes: No. 35 Shawn Godin
18 Olympiad Corner: No. 429
18 Problems: OC711–OC715
20 Solutions: OC686–OC690
24 Integer Values of a Rational Function Ted Barbeau
32 Problems: 5001–5010
36 Solutions: 4951–4960

Crux Mathematicorum
Founding Editors / Rédacteurs-fondateurs: Léopold Sauvé & Frederick G.B. Maskell
Former Editors / Anciens Rédacteurs: G.W. Sands, R.E. Woodrow, Bruce L.R. Shawyer,
Shawn Godin

Crux Mathematicorum
with Mathematical Mayhem
Former Editors / Anciens Rédacteurs: Bruce L.R. Shawyer, James E. Totten, Václav Linek,
Shawn Godin
4/ Editorial

Crux Mathematicorum, Vol. 51(1), January 2025

 MathemAttic /5

MATHEMATTIC
No. 61
The problems featured in this section are intended for students at the secondary school
level.

Click here to submit solutions, comments and generalizations to any

problem in this section.

To facilitate their consideration, solutions should be received by March 15, 2025.

MA301. Suppose n and m are positive integers. Mike and Mary each have a
bag of N = nm marbles labelled 1 to N . Mike randomly takes out a marble from
his bag. Suppose a is the sum of the numbers on the remaining marbles in Mike’s
bag. Similarly, Mary randomly takes out a marble from her bag. Suppose b is the
sum of the numbers on the remaining marbles in Mary’s bag. Find the probability
that a − b is a multiple of n. Express your answer as a reduced rational function
in terms of n and m.

MA302. A 5 × 5 × 5 wooden cube is painted on all 6 faces and then cut up

into unit cubes. One unit is randomly selected and rolled. What is the probability
that exactly one of the five visible faces (on the rolled die) is painted?

MA303. The rectangle below is 12×5. The diagonal forms two right triangles
and a circle is inscribed in each triangle. Find the distance between the centers of
the two circles.

MA304. At a picnic, there are c children, m mothers, and f fathers, with

2 ≤ f < m < c. Every person shakes hand with every other person. The sum
of the number of handshakes amongst the children, amongst the mothers, and
amongst the fathers is 80. How many persons attended the picnic?

Copyright © Canadian Mathematical Society, 2025

6/ MathemAttic

MA305. Choose points D, E, and F on the sides of the equilateral triangle

ABC so that |AF | = |EC| = |DB| = 1 and |F B| = |DC| = |AE| = 3. The line
segments EB, AD, and CF enclose a triangle that is shaded in the diagram. Find
the ratio of the area of the shaded region and the area of triangle ABC.

.................................................................

Les problèmes proposés dans cette section sont appropriés aux étudiants de l’école sec-
ondaire.

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 mars 2025.

MA301. Supposons que n et m soient des entiers positifs. Mike et Mary ont
chacun un sac de N = nm billes étiquetées de 1 à N . Mike prend au hasard une
bille dans son sac. Supposons que a soit la somme des nombres inscrits sur les billes
restantes dans le sac de Mike. De même, Marie prend au hasard une bille dans
son sac. Supposons que b soit la somme des nombres sur les billes restantes dans
le sac de Marie. Trouvez la probabilité que a − b soit un multiple de n. Exprimez
votre réponse sous la forme d’une fonction rationnelle réduite en fonction de n et
m.

MA302. Un cube de bois de taille 5 × 5 × 5 est peint sur ses 6 faces puis
découpé en cubes unitaires. Une unité est choisie au hasard et lancée. Quelle est
la probabilité que l’une des cinq faces visibles (sur le dé lancé) soit peinte ?

Crux Mathematicorum, Vol. 51(1), January 2025

 MathemAttic /7

MA303. Le rectangle ci-dessous est de taille 12 × 5. La diagonale forme deux

triangles droits et un cercle est inscrit dans chaque triangle. Trouvez la distance
entre les centres des deux cercles.

MA304. Lors d’un pique-nique, il y a e enfants, m mères, et p pères, où

2 ≤ p < m < e. Chaque personne serre la main de chaque autre personne. La
somme des poignées de main des enfants, des mères et des pères est de 80. Combien
de personnes ont participé au pique-nique ?

MA305. Choisissons des points D, E et F sur les côtés du triangle équilatéral

ABC de sorte que |AF | = |EC| = |DB| = 1 et |F B| = |DC| = |AE| = 3. Les
segments de droite EB, AD et CF entourent un triangle qui est ombré dans le
diagramme. Trouvez le rapport entre l’aire de la région ombrée et l’aire du triangle
ABC.

Copyright © Canadian Mathematical Society, 2025

8/ MathemAttic

MATHEMATTIC
SOLUTIONS
Statements of the problems in this section originally appear in 2024: 50(6), p. 282-284.

MA276. Table-tennis balls come in boxes of 8 and in boxes of 15. For

example, an order for 38 balls may be filled without breaking open any box by
sending 2 boxes of 15 and 1 box of 8 while an order for 37 balls cannot be filled
without breaking open a box. What is the maximum number of balls that cannot
be bought without breaking open a box?
Originally Peter’s problem – Irish Mathematics Teachers’ Association, 2019 Prob-
lem No. 18.
We received 23 submissions, of which 22 were correct and complete. We present
the solution by Meryem Bourget.
The problem consists of finding the largest integer that cannot be expressed as the
sum of a multiple of 15 and a multiple of 8. We start by listing a few terms for
the arithmetic progressions 15m + 8n when m takes the values 0, 1, 2, 3, 4, 5, 6, 7:

m 15m + 8n
0 0, 8, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104
1 15, 23, 31, 39, 47, 55, 63, 71, 79, 87, 95, 103
2 30, 38, 46, 54, 62, 70, 78, 86, 94, 102
3 45, 53, 61, 69, 77, 85, 93, 101
4 60, 68, 76, 84, 92, 100
5 75, 83, 91, 99
6 90, 98
7 105

We make two key observations from the table above:

(i) 97 is the largest number not listed, i.e. 97 cannot be written as the sum of
a multiple of 15 and a multiple of 8.
(ii) 98, 99, 100, 101, 102, 103, 104, 105 are all listed: these consecutive numbers
can all be written as the sum of a multiple of 15 and a multiple of 8.
We claim that every integer greater or equal to 98 can be written as the sum of a
multiple of 15 and a multiple of 8. To see this, let n ≥ 98, and write
n = 98 + k
for some k ≥ 0. By the division algorithm, there exist a non-negative integer q
and an integer r between 0 and 7 such that k = 8q + r. Therefore, we have
n = 98 + 8q + r = (98 + r) + 8q.

Crux Mathematicorum, Vol. 51(1), January 2025

 MathemAttic /9

It follows that n is the sum of a multiple of 15 and a multiple of 8 if and only if

98 + r is too. But this is obvious by (ii) since 98 + r is an integer between 98 and
105.

Consequently, (i) implies that 97 is the largest number that cannot be written
as a multiple of 15 and a multiple of 8. In other words, the largest amount of
table-tennis balls that cannot be bought in boxes of 15 and in boxes of 8 without
breaking them is 97.

MA277. Two congruent rectangles have a common vertex and overlap, as

shown in the diagram.

Find the area of the drawing that is not common to both rectangles.

Originally Peter’s problem – Irish Mathematics Teachers’ Association, 2010 Prob-

lem No. 21.

We received 21 submissions of which 17 were correct and complete. We present

the solution by the Missouri State University Problem Solving Group.

Let AC = x. Then BD = x − 2, AD = 8 and BC = 4. By the Pythagorean

theorem,

82 + (x − 2)2 = x2 + 42 ,

which yields x = 13. The area common to the two rectangles is

1
(13 · 4 + 11 · 8) = 70,
2

so the area that is not common to both is

2 · 8 · 13 − 2 · 70 = 68.

Copyright © Canadian Mathematical Society, 2025

10/ MathemAttic

MA278. The points A, B, C, D, E, F are collinear and the points P , Q, C,

R, S are collinear, as shown.

How many different triangles may be formed by using three of these given points
as vertices?
Originally Peter’s problem – Irish Mathematics Teachers’ Association, 2022 Prob-
lem No. 15.
We received 23 submissions, of which 17 were correct and complete. Majority of the
solvers used combinatorial arguments. Here we present two common approaches.
Solution 1. There are 10


3 ways of choosing three points to form a triangle. How-
ever, a
triangle cannot be formed if all three chosen points are collinear. There
are 53 ways to choose three points on horizontal line that won’t yield a triangle
and there are 63 ways to choose three points on a vertical line that will fail for


form a triangle. This gives
Ç å Ç å Ç å
10 5 6
− − = 90
3 3 3

total triangles.

Solution 2. There are two cases to consider.

Case 1: Two of the vertices of the triangle are among A, B, C, D, E, F and the
third among P , Q, R, S. This can be done in 62 × 4 = 60 ways.

Case 2: Two of the vertices of the triangle are
among P , Q, R, S and the third
among A, B, D, E, F . This can be done in 42 × 5 = 30 ways.
Adding the two cases we get total of 90 triangles.

MA279. A 4-digit number is a perfect square. When 1 is added to each of

the digits, another 4-digit perfect square is formed. What number is the smaller
of these perfect squares?
Originally Peter’s problem – Irish Mathematics Teachers’ Association, 2022 Prob-
lem No. 18.
We received 23 submissions of which 22 were correct and complete. We present
the solution by Adam Mawani.

Crux Mathematicorum, Vol. 51(1), January 2025

 MathemAttic /11

Assume n2 represents the smaller perfect square, while m2 represents the larger
perfect square. Since each digit of m2 is one greater than the corresponding digit
of n2 , we find that:

m2 − n2 = 1111 ⇒ (m − n)(m + n) = 1111

The number 1111 has factors 1, 11, 101, 1111, so the possible values for (m − n)
and (m + n), considering that (m − n) < (m + n), are the pairs (1, 1111) and
(11, 101). Using systems of equations, we can see that (1, 1111)
√ yields 2n = 1110
and so n = 555. This is impossible because 32 ≤ n ≤ 99 (since 1000 ≈ 31.62 and
√
9999 ≈ 99.99), meaning n2 would not be a four-digit number. However, when
we evaluate the pair (11, 101), we find that:

2m = 112 ⇒ m = 56 ⇒ m2 = 3136

2n = 90 ⇒ n = 45 ⇒ n2 = 2025

This matches the condition 2025 + 1111 = 3136, so the smaller of the perfect
squares is 2025.

MA280. Prove that every odd integer n lies exactly halfway between a mul-
tiple of 3 and a multiple of 4 where n ≥ 5.

Originally Peter’s problem – Irish Mathematics Teachers’ Association, 2019 Prob-

lem No. 20.

There were 19 submissions, 16 of them were complete and correct. We present 2

different approaches below.

Solution 1, by Ioan Viorel Codreanu.

We have n = 2k + 1, where k ∈ N, k ≥ 2. We observe that

4k − 4 + 6
2k + 1 = ,
2

where 4 | (4k − 4) and 3 | 6, and the conclusion follows.

Solution 2, by Missouri State University Problem Solving Group.

In fact, every integer n is exactly halfway between a multiple of 3 and a multiple

of 4, namely between −6n and 8n.

If the intention was to restrict to positive integers then an odd integer n = 2k + 1

is midway between 6 and 4k − 4 and the latter is positive if k ≥ 2. We note that
an even integer n = 2k is midway between 12 and 4k − 12, the latter being positive
if k ≥ 4.

Copyright © Canadian Mathematical Society, 2025

12/ MathemAttic

Two additional observations provided by Meryem Bourget.

1. It can be proven that the result still holds for every even integer n ≥ 8.
One can obtain positive multiples of 3 and 4 by choosing r = 2, so that
3p = 6r = 12 and 4q = 2n − 12.
2. A possible generalization could be: Given two relatively prime integers
m, n ∈ N, show there exists an integer N such that every integer k ≥ N
is the middle point of positive multiples of m and n.

Crux Mathematicorum, Vol. 51(1), January 2025

 Shawn Godin /13

PROBLEM SOLVING
VIGNETTES
No. 35
Shawn Godin
A Problem from the CSMC

For years, I have marked contests (Hypatia or Euclid) in April for The Centre for
Education in Mathematics and Computing at the University of Waterloo. Since
my retirement, I have added the Canadian Senior Mathematics Contest to the list
in December. A few weeks ago I was in Waterloo marking problems A6 and B3
from this year’s test. We will take a look at B3 in this issue.

B3. In the diagram, ABCD is a rectangle with AB > BC.

A B

D P C

Point P is on CD so that P D = P B.

(a) Let P D = 53 and BC = 28. Determine the length of AB.

(b) Suppose that AB = 101. If the length of BC is an integer, prove

that the length of P D cannot be an integer.

(c) Suppose that BC = m for some positive integer m. Suppose further

that, for this value of m, there are exactly 7 positive integers n so that
when AB = n, the length of P D is an integer. Determine all possible
values of m with 1 ≤ m ≤ 100.

I will leave part (a) to the interested reader.

In parts (b) and (c), since we are interested in all the lengths in the diagram being
integral, the important point is that since P D = P B, then the lengths of the sides
of triangle BCP are a Pythagorean triple (for more on Pythagorean triples, check
out number 21 of this column [2022: 48(3), 128-132]), and BP + P C = AB.

Copyright © Canadian Mathematical Society, 2025

14/ Problem Solving Vignettes

(b) Solution: Setting P D = P B = c and BC = a ∈ Z+ (where Z+ is the set of

positive integers), then P C = 101 − c.

A 101 B

c a

D c P 101 − c C

Thus

a2 = c2 − (101 − c)2
a2 = c2 − (10 201 − 202c + c2 )
a2 = 202c − 10 201 (1)

Upon reaching (1), or something equivalent, many students resorted to hand-

waving, making statements similar to “for 0 < c < 101, 202c − 10 201 produces
−9999, −9797, . . . , −101, 101, . . . , 9797, 9999 none of which is a perfect square”
without any justification (although I did have one student compute all the val-
ues and show that none was a perfect square).
Justification can be delivered by factoring the right hand side of (1) to give

a2 = 101(2c − 101)

Thus, since 101 is prime and assuming c is an integer, 101 | a2 , and hence 101 | a,
which is impossible, since a < 101. Therefore c cannot be an integer.
A student who knew about Pythagorean triples could have gotten there a bit
quicker by noting that since 101 is prime, and BP + P C = 101, then if P D is an
integer, and hence also BP and P C, then BP and P C must be relatively prime
(or else their common factor would divide 101, which is impossible). Also, since
101 is odd, then BP and P C must be of different parities, which forces them to
be of the form

P B = m2 + n2
P C = 2mn

from the formulas for primitive Pythagorean triples. However, this means that

P B + P C = m2 + n2 + 2mn = (m + n)2 = 101 = AB

which is impossible as 101 is prime, and hence not a perfect square.

Crux Mathematicorum, Vol. 51(1), January 2025

 Shawn Godin /15

(c) Solution: Once more setting BP = c ∈ Z+ , we get

A n B

c m

D c P n−c C

The Pythagorean theorem yields

c2 = m2 + (n − c)2
c2 = m2 + n2 − 2nc + c2
n(2c − n) = m2 (2)

from which, since all values are integers, we must have n | m2 . Also, since c is the
hypotenuse of a right triangle, then

c > n − c ⇒ 2c − n > 0

and clearly
n > c ⇒ 2n > 2c ⇒ n > 2c − n.
Since m < n, by the conditions of the problem, we must have

2c − n < m < n < m2 .

However, since our two factors of m2 , n and 2c − n, necessarily have the same
parity, we have to be careful. We will examine two cases.
Case 1: n and 2c − n are odd (and thus, so is m).
In this case, since n is the larger of the two factors of m2 and n 6= m, then m2
must have 15 factors (m, plus 7 factors larger than m and 7 factors smaller than
m). Recalling that if the prime factorization of a number k is

k = pα α2 αt
1 × p2 × · · · × pt
1

then the number of factors of k, d(k) is given by

d(k) = (α1 + 1)(α2 + 1) · · · (αt + 1).

Since d(m2 ) = 15 = 3 × 5, we must have either m2 = p14 or m2 = p2 q 4 for some

primes p and q. However m2 = p14 is impossible, as that would give m = p7 > 100
for all primes. Therefore we must have m2 = p2 q 4 or m = pq 2 , for distinct primes
p and q. Since m is odd, p and q are both odd. Checking odd primes, recalling
that m ≤ 100 we get:

Copyright © Canadian Mathematical Society, 2025

16/ Problem Solving Vignettes

p q m = pq 2
3 5 m = 3 × 52 = 75
5 3 m = 5 × 32 = 45
7 3 m = 7 × 32 = 63
11 3 m = 11 × 32 = 99

Case 2: n and 2c − n are even (and thus, so is m).

In this case n has to be an even divisor of m2 . To simplify things, let n = 2N and

m = 2M , thus (2) becomes

2N (2c − 2N ) = (2M )2
4N (c − N ) = 4M 2
N (c − N ) = M 2

where similarly we have

c − N < M < N < M2

and again M 2 must have 15 factors, so M = pq 2 for primes p and q, as the M = p7

case can again be eliminated. Thus m = 2pq 2 , but p and q can be any primes.
Checking, we get

p q m = 2pq 2
2 3 m = 2 × 2 × 32 = 36
2 5 m = 2 × 2 × 52 = 100
3 2 m = 2 × 3 × 22 = 24
5 2 m = 2 × 5 × 22 = 40
5 3 m = 2 × 5 × 32 = 90
7 2 m = 2 × 7 × 22 = 56
11 2 m = 2 × 11 × 22 = 88

Therefore, we get a total of 11 solutions, m ∈ {24, 36, 40, 45, 56, 63, 75, 88, 90, 99, 100}.

When I originally saw the problem, I thought the student who knew the parametric
expressions for the general Pythagorean triples would have an advantage. As I
started playing with it, it became apparent that it would not readily produce a
nice solution. To give a sense of what problems one would run into, we will look at
a case that works, m = 99. This value would be one of the legs of a Pythagorean
triple and would have to be

m = k(α2 − β 2 ) = k(α − β)(α + β),

where α and β are relatively prime, one is even and one is odd. This yields
a = 2αβ, c = k(α2 + β 2 ) and whence n = a + c = k(α + β)2 . However, when we
look at factoring 99 in acceptable ways we run into a problem:

Crux Mathematicorum, Vol. 51(1), January 2025

 Shawn Godin /17

m = k(α − β)(α + β) k α−β α+β n = k(α + β)2

99 1 1 99 9801
99 1 3 33 1089
99 1 9 11 121
99 3 1 33 3267
99 3 3 11 363
99 9 1 11 1089
99 11 1 9 891
99 33 1 3 297

Two of the eight acceptable ways of factoring m, to yield a side of a Pythagorean

triple, yield the same value of n. Using this method it will be challenging to find
the form that m has to take. In the case where m is even, it becomes even more
interesting as m could be of the form 2kαβ or k(α − β)(α + β). The interested
reader may try to see if there is a way to salvage this method.

Copyright © Canadian Mathematical Society, 2025

18/ OLYMPIAD CORNER

OLYMPIAD CORNER
No. 429
The problems featured in this section have 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 March 15, 2025.

OC711. Place different positive integers, not greater than 25, in the cells of
a 3 × 3 square so that in any pair of adjacent cells one number is divisible by the
other.

OC712. The teacher dictated to Vovochka the slopes and y-intercepts of three
different linear functions whose graphs are all parallel. Inattentive Vovochka, when
writing down each of the functions, swapped the slope and the y-intercept of each
function and plotted the graphs of the resulting lines. How many points could
there be through which at least two graphs pass?

OC713. Let a, b, c, d be real numbers such that

a b c d
+ = + .
b a d c
Prove that the product of any two numbers from a, b, c, d is equal to the product
of the other two.

OC714. Petya and Vasya play on the segment [0, 1] in which the points 0
and 1 are marked. The players take turns, Petya starts. Each move, the player
marks a previously unmarked point on the segment. If, after the next player’s
move, three consecutive segments are found between adjacent marked points from
which a triangle can be formed, then the player who made such a move is declared
the winner, and the game ends. Will Petya be guaranteed to win?

OC715. A mathematician has 19 different weights, the masses of which in

kilograms are equal to ln 2, ln 3, ln 4, . . . , ln 20, and an absolutely precise two-pan
scale. He puts several weights on the scale so that equilibrium is established. What
is the greatest number of weights that could be on the scale?

.................................................................

Crux Mathematicorum, Vol. 51(1), January 2025

 OLYMPIAD CORNER /19

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 mars 2025.

OC711. Placez différents nombres entiers positifs, ne dépassant pas 25, dans
les cases d’un carré 3 × 3 de façon à ce que dans toute paire de cases adjacentes,
l’un des nombres soit divisible par l’autre.

OC712. Le professeur a dicté à Vovochka les coefficients de proportionnalité

et les ordonnées à l’origine de trois fonctions linéaires différentes dont les graphes
sont parallèles. Inattentive, Vovochka a interverti – en écrivant chacune des fonc-
tions – le coefficient de proportionnalité et l’ordonnée à l’origine et a tracé les
graphes des fonctions résultantes. Combien y a-t-il de points par lesquels passent
au moins deux graphes?

OC713. Soient a, b, c et d des nombres réels tels que

a b c d
+ = + .
b a d c
Montrez que le produit de deux nombres quelconques parmi a, b, c et d est égal
au produit des deux autres.

OC714. Petya et Vasya jouent sur le segment [0, 1] dans lequel les points 0
et 1 sont marqués. Les joueurs jouent à tour de rôle, Petya commence. À chaque
coup, le joueur marque un point précédemment non marqué sur le segment. Si,
après le déplacement du joueur suivant, trois segments consécutifs sont trouvés
entre des points marqués adjacents à partir desquels un triangle peut être formé,
le joueur qui a effectué ce déplacement est déclaré vainqueur et la partie se termine.
La victoire de Petya est-elle garantie ?

OC715. Un mathématicien dispose de 19 poids différents dont les masses en

kilogrammes sont égales à ln 2, ln 3, ln 4, . . . , ln 20, et d’une balance à deux plateaux
d’une précision absolue. Il place plusieurs poids sur la balance de façon à ce que
l’équilibre s’établisse. Quel est le plus grand nombre de poids pouvant se trouver
sur la balance ?

Copyright © Canadian Mathematical Society, 2025

20/ OLYMPIAD CORNER

OLYMPIAD CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2024: 50(6), p. 299–301.

OC686. A sequence of Xs and Os is given, such that no three consecutive

characters in the sequence are all the same, and let N be the number of characters
in this sequence. Maia may swap two consecutive characters in the sequence. After
each swap, any consecutive block of three or more of the same character will be
erased (if there are multiple consecutive blocks of three or more characters after a
swap, then they will be erased at the same time), until there are no more consec-
utive blocks of three or more of the same character. For example, if the original
sequence were XXOOXOXO and Maia swaps the fifth and sixth character, the
end result will be XXOOOXXO → XXXXO → O. Find the maximum value
N for which Maia can’t necessarily erase all the characters after a series of swaps.
Originally from the January 2024, CMO Qualifying Repechage.
We received no correct submissions to this problem. The official solutions can be
found at this link.

OC687. For certain real constants p, q, r, we are given a system of equations

 2
a + b + c = p

a + b2 + c = q

a + b + c2 = r


What is the maximum number of solutions of real triplets (a, b, c) across all possible
p, q, r? Give an example of the p, q, r that achieves this maximum.
Originally from the January 2024, CMO Qualifying Repechage.
We received 3 submissions, all of which were correct and complete. We present
the solution by Sicheng Du.
Number the given equations as follows,

2
a + b + c = p, (1)

b2 + c + a = q, (2)

 2
c + a + b = r. (3)

From (1) we get c = p − a2 − b. Plugging this into (2) and (3) respectively gives

b2 − b − a2 + a + p − q = 0, (4)

2
b2 + 2a2 − 2p + 1 b + a + a2 − p − r = 0.


(5)

Crux Mathematicorum, Vol. 51(1), January 2025

 OLYMPIAD CORNER /21

Using the resultant and Sylvester’s Matrix to eliminate b from (4) and (5), we get

1 −1 −a2 + a + p + q 0
0 1 −1 −a2 + a + p + q
2 2

2 = 0.
1 2a − 2p + 1 a+ a −p −r 0
2
2a2 − 2p + 1 a + a2 − p − r


0 1

When expanding this determinant, the result is a polynomial f (a) of the 8th
degree. Hence, there are at most 8 possibilities of a.
For each value of a, (1) gives b + c = p − a2 . Subtracting (3) from (2) gives

(b − c)(b + c − 1) = q − r,

and so
q−r
b−c= .
p − a2 − 1
Hence, the values of b and c can be uniquely determined; thereby there are at most
8 triplets (a, b, c).
If we take p = q = r = 4, then
Ä√ √ √ ä Ä √ √ √ ä
(a, b, c) = 3, 3, 1 − 3 , − 3, − 3, 1 + 3

and their cyclic permutations as well as

√ √
a = b = c = −1 − 5, a = b = c = −1 + 5

are all valid solutions, making a total of 8 triplets.

OC688. Fix integers a and b greater than 1. For any positive integer n, let
rn be the (non-negative) remainder that bn leaves upon division by an . Assume
there exists a positive integer N such that rn < 2n /n for all integers n ≥ N . Prove
that a divides b.
Originally from the February 2024, The 15th Romanian Master of Mathematics
Competition.
We received 4 submissions, all of which were correct and complete. We present
the solution by Theo Koupelis.
Let a, b be integers with a, b ≥ 2. If a = b the problem is solved. If a > b then
rn = bn and rn ≥ 2n ≥ 2n /n, contradicting the original assumption. Therefore
a < b. It is given that

bn = kan + rn , and bn+1 = λan+1 + rn+1 , (∗)

where k, λ are non-negative integers, 0 ≤ rn < an , and 0 ≤ rn+1 < an+1 . Then
bn+1 = kb · an + brn , and thus

an (kb − λa) = rn+1 − brn . (∗∗)

Copyright © Canadian Mathematical Society, 2025

22/ OLYMPIAD CORNER

Choosing No = max{b, N } we get

2n an
rn < ≤
n b
because Å ã1/n
b
a≥2≥2 ,
n
for n ≥ No . Thus, brn ≤ an for n ≥ No . Also,

2n+1
rn+1 < ≤ an ,
n+1
because ã1/n
2
Å
a≥2≥2 .
n+1
Therefore, the right-hand-side of (∗∗) has magnitude less than an , and thus we
must have kb = λa and rn+1 = brn . From the latter we get
rn rn+1 rNo bn 2n
n
= n+1 = constant = N =⇒ rn = rNo · No ≥ rNo · No .
b b b o b b
2n
But we also have rn < n , and thus

2n 2n bNo
rNo · < =⇒ rN
< for all n ≥ No .
bNo n o
n
Thus, rNo = 0, and therefore rn = 0. Then from (∗) we get a | b.

OC689. The product of five consecutive positive integers is 120 times greater
than ABABAB, where A and B are non-zero digits. What is the largest of those
five numbers?
Originally from the July 2023, Bulgaria International Mathematics Competition,
Elementary Mathematics International Contest, Individual Contest.
We received 11 submissions, all of which were correct and complete. We present
the solution by UCLan Cyprus Problem Solving Group.
Note that

ABABAB = 10101(10A + B) = 3 · 7 · 13 · 37(10A + B).

If one of the five consecutive numbers is 74 or larger, then their product is at least

70 · 71 · 72 · 73 · 74 > 705 = 16807 · 105 > 120 · 107 > 120 · ABABAB

which is impossible. Hence all five consecutive numbers need to be 73 or lower.

Since the product of the five numbers is a multiple of 37, and no number greater
or equal to 74 can be used, then one of the five numbers must be 37. We must

Crux Mathematicorum, Vol. 51(1), January 2025

 OLYMPIAD CORNER /23

also have a multiple of 13 and so 39 is also one of the five numbers. We must also
have a multiple of 7 and so 35 must also be one of the five numbers. (It cannot
be 42 as then we would have at least 6 numbers.) So the five numbers must be
35, 36, 37, 38 and 39. In fact, we verify that the product of these five numbers is

35 · 36 · 37 · 38 · 39 = 120 · 575757,

as required.
Therefore the largest of these five numbers is 39.

OC690. Let ABC be an equilateral triangle whose area is 36 cm2 , and EF D

be an isosceles triangle with EF = F D. The point F is the centre of triangle
ABC, and the points B and C are the midpoints of EF and F D, respectively, as
shown in the diagram below. If BG ⊥ EF and CH ⊥ DF , what is the area, in
cm2 , of the shaded region?

Originally from the July 2023, Bulgaria International Mathematics Competition,

Invitational World Youth Mathematics Intercity Competition, Individual Contest.
We received 19 submissions, all of which were correct and complete. We present
a typical solution.
Let I be the midpoint of BC. Then the 30◦ -60◦ -90◦ triangle BF I is similar to
triangle EGB with scale factor √23 . The same is true for triangles CF I and DHC.
Since area scales as the square of the scale factor, the sum of the areas of triangles
Ä ä2
EGB and DHC is √23 = 34 of the sum of the areas of triangles BF I and CF I,
that is, 43 of the area of triangle BCF . So the shaded area, the sum of the area of
quadrilateral ABF C and the area of triangle BCF , is
2
4 1
36 cm2 + · 36 cm2 = 24 cm2 + 16 cm2 = 40 cm2 .


3 3 3

Copyright © Canadian Mathematical Society, 2025

24/ Integer Values of a Rational Function

Integer Values of a Rational

Function
Ted Barbeau
Sometimes, the editors of Crux receive a problem proposal that offers a more
extensive opportunity for investigation than a one-shot appearance in the Problems
section allows. Such a problem was received from Ibrahim Aghazada, a student at
ADA University in Baku, Azerbaijan. The student wants to know the solutions in
positive integers to x2 + y 2 + z 2 = 2025n (xy + xz + yz). As the proposer’s solution
indicated, the coefficient of the right side was a stand-in for a general positive
integer k. Differing values of k exhibit some quite distinct possibilities.
Let
x2 + y 2 + z 2
f (x, y, z) = ,
xy + yz + zx
where x, y, z are integers, at least two of which are nonzero. We will investigate
which integer values k are assumed by this rational function. This is equivalent to
determining values of k for which the diophantine equation

x2 + y 2 + z 2 = k(xy + yz + zx) (1)

has a solution. Because of its homogeneity, if it does have a solution, then it must
have one for which the greatest common divisor of x, y and z is 1; call such a
solution basic.

Exercises
Exercise 1. Determine solutions, if any, for f (x, y, z) = k when k is equal to −2,
−1, 0 and 1.
Exercise 2. Show that f (x, y, z) can never be a multiple of 4.
Exercise 3. Show that f (x, y, z) can never be congruent to 3 modulo 4.
Exercise 4.
(a) Show that (1) is equivalent to each of

(x + y + z)2 = (k + 2)(xy + yz + zx) (2)

and
k(x + y + z)2 = (k + 2)(x2 + y 2 + z 2 ). (3)

(b) Suppose that k + 2 = m2 n where n is divisible by no square except 1. Prove

that, if f (x, y, z) = k, then x2 + y 2 + z 2 is divisible by the least common
multiple of k and n.

Crux Mathematicorum, Vol. 51(1), January 2025

 Ted Barbeau /25

Exercise 5. Suppose that f (x, y, z) = f (u, v, w) = k. Show that in the ratio

(xy + yz + zx) : (uv + vw + wv) reduced to lowest terms, the antecedent and
consequent are both squares. In particular, if there is a solution for which we have
x2 + y 2 + z 2 = ±k, then for each solution (u, v, w), |uv + vw + wu| is a square.

Exercise 6. Prove that f (x, y, z) can never assume a value k that is either a
positive multiple of 3 or negative integer k for which |k| ≡ 1 (mod 3).

Exercise 7. Prove that f (x, y, z) cannot assume a value k = 4r m2 − 2 where

r ≥ 2 and m is odd.

Exercise 8. Does the equation f (x, y, z) = −3 have a solution?

Exercise 9. List all the values of k for which −10 ≤ k ≤ 10 that have not been
ruled out as values of f (x, y, z) by the foregoing exercises.

Exercise 10.

(a) Write equation (1) as a quadratic in x and determine its discriminant in

terms of k, y, z. For which values of k does f (x, y, z) = k have infinitely
many basic solutions?

(b) If f (x, y, z) = k has a solution and k + 2 = m2 n, where n is square free, show

that −3 must be a square (mod p) for every odd prime divisor p > 3 of n.
Use the law of quadratic reciprocity to show that p ≡ 1 (mod 3).

Exercise 11. Determine infinitely many basic triples (x, y, z) for which f (x, y, z) =
2.

Exercise 12. Determine all integer values assumed by f (x, y, z) when x, y, z are
in arithmetic progression.

Exercise 13. One way to find integer values of f (x, y, z) is to guarantee that
xy + yz + zx is equal to 1 or -1, or some other small value. Provide some examples.

Exercise 14.

(a) Suppose that we have a solution for f (x, y, z) = k for which the quadratic
equation for x in terms of y, z, k has a double root. Then we would have
2x = k(y + z). Investigate the possibility that x = k.

(b) In the situation of (a), with (x, y, z) = (k, y, z) and k a double root, suppose
instead that f (x, y, z) = 2k.

Exercise 15. Investigate which values of k < 100 can be assumed by f (x, y, z),
for which the only prime divisors of k + 2 are 2 and 3.

Copyright © Canadian Mathematical Society, 2025

26/ Integer Values of a Rational Function

Hints and comments for the exercises

Exercise 1. When k = −2, the equation is equivalent to (x + y + z)2 = 0. When
k = ±1, it is equivalent to

(x ∓ y)2 + (y ∓ z)2 + (z ∓ x)2 = 0.

Exercise 2. Let f (x, y, z) be a multiple of 4. Since k + 2 is even but not a

multiple of 4, x2 + y 2 + z 2 must be even. But then x + y + z is even as well. so
that x2 + y 2 + z 2 ≡ 0 (mod 4). However, for any basic solution, at least one of
x, y, z is odd and x2 + y 2 + z 2 ≡ 1, 2, 3 (mod 4).
Exercise 3. Let (x, y, z) be a basic solution of f (x, y, z) = k, where k ≡ 3 (mod
4). Then if exactly one of (x, y, z) is odd, then the left side of (1) is odd and the
right is even. If exactly two of (x, y, z) are odd, then the left side is even and the
right odd. Finally, if all three of (x, y, z) are odd, then the left side is congruent
to 3 (mod 4) while the right side is congruent to 3 × 3 ≡ 1 (mod 4).
The possibility that k ≡ 1 (mod 4) remains. When x, y, z are all odd, x2 +y 2 +z 2 6≡
7 (mod 8), so that, when k ≡ 1, then xy + yz + zx ≡ 3, and when k ≡ 5, then
xy + yz + zx ≡ 7, modulo 8.
Exercise 6. If k < 0 and |k| ≡ 1 (mod 3), then k ≤ −3 and

|k + 2| = −k − 2 = |k| − 2 ≡ 1 − 2 ≡ 2

modulo 3. In any case |k| + 2 is congruent to 2 modulo 3, and therefore has a

prime divisor p congruent to 2 that divides it to an odd power. Since p divides
(x + y + z)2 , it must do so to an even power. Hence p must be a divisor of

4(xy + yz + zx) = 4[xy + z(y + x)] ≡ 4[xy − (y + x)2 ] = (2y + x)2 + 3x2 .

We may assume that x is not a multiple of 3. Then −3 must be a square, modulo

p. However (using the Legendre symbol and the quadratic reciprocity law),

−3
Å ã p
= (−1)(p−1)/2 (−1)(p−1)/2 = −1,
p 3

from which −3 is not a square modulo p. This yields a contradiction. This was
the argument in the solution of Ibrahim Aghazade.
Exercise 7. (x, y, z) must have the form (2r u, 2r v + n, 2r w − n) for some odd
integer n. Substituting this into the equation (x + y + z)2 = 44 m2 (xy + yz + zx)
leads to (u + v + w)2 ≡ −m2 n2 ≡ −1 (mod 4), which is impossible.
Exercise 8. One can easily guess the solution (x, y, z) = (−1, 1, 1), from which
others can be found. The equation is equivalent to

(x2 + y 2 + z 2 ) + 3(xy + yz + zx) = 0.

Crux Mathematicorum, Vol. 51(1), January 2025

 Ted Barbeau /27

For a basic solution, at least one of x, y and z is odd. If any are even, then the
left side is odd. Hence we may write (x, y, z) = (2u + 1, 2v + 1, 2w + 1) for some
integers u, v, w. Then we obtain

(u2 + v 2 + w2 ) + 4(u + v + w) + 3(uv + vw + wu) + 3 = 0.

If we set w = 0, then we get the following quadratic in u in terms of v:

u2 + (4 + 3v)u + (v 2 + 4v + 3) = 0.

Its discriminant is 5v 2 + 8v + 4, which assumes a square value when v = −1, 0, 5.

This leads to

f (−1, 1, 0) = f (−5, 1, 1) = f (−5, 11, 1) = f (−31, 11, 1) = −3.

We can fix two of the variables in equation (1) and regard it as a quadratic in the
third, and in this way, generate an infinite set of solutions.
Exercise 9. −6, −3, −2, 2, 5, 10.
Exercise 10. The quadratic equation Ek in x

x2 − k(y + z)x + (y 2 − kyz + z 2 ) = 0

has discriminant Dk equal to

k 2 (y 2 + 2yz + z 2 ) − 4(y 2 − kyz + z 2 ) = (k 2 − 4)(y 2 + z 2 ) + 2k(k + 2)yz

= (k + 2)[(k − 2)(y 2 + z 2 + yz)] + (k + 2)2 yz.

Suppose that p ≥ 5 is a prime divisor to an odd power of k + 2. Since Dk must

be square, it is a multiple of p2 . Since p does not divide k − 2, p must divide
y 2 + z 2 + yz, and therefore it divides

4(y 2 + z 2 + yz) = (2y + z)2 + 3z 2 .

Since we can arrange that z 6= 0, −3 must be a square (mod p). This can happen
only if p ≡ 1 (mod 6). (When p ≡ 5 (mod 6), then y 3 ≡ z 3 if and only if y ≡ z.)
For example, when k = 5 and (y, z) = (3, 5), y 2 + z 2 + yz = 49 = 72 . This puts us
on track for a solution. Indeed, D5 = 21 × 34 + 70 × 15 = 1764 = 422 and we find
that f (−1, 3, 5) = 5.
If (1) has a solution with k 6= 1, then it has a solution with y 6= z. Since gcd
(x, y, z) = 1, x, y, z are pairwise coprime. If D = 0, and k 6= −2, we can start with
a solution and arrange y and z so that it has two distinct solutions x, and we can
similarly continue spawning new solutions.
Exercise 11. When k = 2, the discriminant D2 is equal to 16yz, which is square
only if yz is square. Since y and z are coprime, we may take y = u2 and z = v 2 ,
and the equation E2 becomes

0 = x2 − 2(u2 + v 2 ) + (u4 − 2u2 v 2 + v 2 ) = [x − (u + v)2 ][x − (u − v)2 ].

Copyright © Canadian Mathematical Society, 2025

28/ Integer Values of a Rational Function

We get infinitely many solutions

(x, y, z) = ((u ± v)2 , u2 , v 2 ),

where u and v are arbitrary integers.

We can form any sequence of integers satisfying the recursion mn+1 = mn + mn−1
and find that (x, y, z) = (m2n−1 , m2n , m2n+1 ) satisfies f (x, y, z) = 2. In particu-
2
lar (x, y, z) = (fn−1 , fn2 , fn+1
2
) satisfies the equation where {fn } is the bilateral
Fibonacci sequence indexed over the complete set of integers. In this case, deter-
mine the values of x2 + y 2 + z 2 and xy + yz + zx.
Another sequence for example is

{. . . , 72 , 52 , 22 , 32 , 12 , 42 , 52 , 92 , 142 , 232 , . . . .}

It is interesting that the basic solutions in this case consist of squares.

Exercise 12. Suppose that (x, y, z) = (v − u, v, v + u) for some integers u and v.
Then
(k + 2)u2 = 3(k − 1)v 2 .
Since each side is divisible by 3, we assume that k = 3l − 2, whereupon lu2 =
3(l − 1)v 2 . Taking l = m2 , we get m2 u2 = (3m2 − 3)v 2 . Thus, 3m2 − 3 must be a
square, (3n)2 and so
m2 − 3n2 = 1.
This Pell’s equation has infinitely many solutions (mi , ni ) where (m0 , n0 ) = (1, 0),
(m1 , n1 ) = (2, 1) and, for each integer i,

mi+1 = 4mi − mi−1 and ni+1 = 4ni − ni−1 .

(Solutions are (1, 0), (2, 1), (7, 4), (26, 15), (97, 56), . . . .)
When k = 3m2 − 2 and m2 − 3n2 = 1, we get the solution

(x, y, z) = (m − 3n, m, m + 3n).

Thus

x2 + y 2 + z 2 = 3m2 + 18n2 = 3(m2 + 6n2 ) = 3(m2 + 2m2 − 2) = 3(3m2 − 2) = 3k,

and
xy + yz + zx = 3m2 − 9n2 = 3(m2 − 3n2 ) = 3.
When m = 2, k = 10, and we get the solutions:

(x, y, z) = (−1, 2, 5), (2, 5, 71), (−1, 5, 38), . . .

from which we can generate an infinite supply.

When m = 7, k = 145, and we get the solution (x, y, z) = (−5, 7, 19).

Crux Mathematicorum, Vol. 51(1), January 2025

 Ted Barbeau /29

Exercise 13. There are several ways one can look at this. One approach is to
look at situations for which xy + yz + zx = ±1. When xy + yz + zx = 1, then
x2 + y 2 + z 2 = k and
yz − 1
Å ã
x=− .
y+z
When xy + yz + zx = −1, then x2 + y 2 + z 2 = −k and
yz + 1
Å ã
x=− .
y+z
To make x an integer in both cases, we can make z = 1 − y or z = −(y + 1).
In particular, we can take (y, z) = (2r − 1, 2r + 1) so that yz + 1 = 4r2 and x = −2.
Alternatively, we can note that

(y + x)(z + x) = x2 ± 1,

and choosing values of y and z that will make this product valid.
Interestingly, the Fibonacci numbers make another appearance. Let {fn } be the
Fibonacci sequence with f0 = 0, f1 = 1. When (x, y, z) = (−fn−1 , fn , fn+1 ), then

x2 + y 2 + z 2 = 4fn2 + 2(−1)n

and
xy + yz + zx = fn2 − fn−1 fn+1 = (−1)n+1 .
This solution corresponds to k = (−1)n+1 [4fn2 + 2(−1)n ].
Exercise 14.
(a) We must have y + z = 2, so we may assume that y = −(r − 1) and z = r + 1.
If (x, y, z) = (k, −(r − 1), r + 1) is a solution, then

0 = k 2 − (r2 − 1)k + 2(r2 + 1) = (k − (r2 + 1))(k − 2).

(b) In this situation, we take y + z = 1, so we can assume that y = −(r − 1) and

z = r, and then solve for k.
Exercise 15. We have to investigate k = 22, 25, 34, 46, 70 and 94. (k = 14, 62
have already been dealt with.) Since 2 divides 24, 72, 96 to an odd power, then
x + y + z is divisible by 4 and xy + yz + zx is even. Therefore, up to order,
(x, y, z) ≡ (0, 1, 3) (mod 4), which would make xy + yz + zx odd. Hence k = 22,
70 and 94 are not possible.
If k = 25, then
(x + y + z)2 = 27(xy + yz + zx).
Therefore x + y + z is divisible by 9 and xy + yz + zx is divisible by 3. Therefore,
modulo 9, up to order,

(x, y, z) = (9u ± 1, 9v ± 1, 9w ± 7), (9u ± 1, 9v ± 4, 9w ± 4), (9u ± 2, 9v ± 2, 9w ± 5)

Copyright © Canadian Mathematical Society, 2025

30/ Integer Values of a Rational Function

for some integers u, v, w. Substituting any of these into the equation and dividing
by 81 yields
(u + v + w ± 1)2 ≡ 2 mod 3.
Since this is not possible, there is no solution.
If k = 34, then (x + y + z)2 = 36(xy + yz + zx), so that x + y + z = 0 (mod 6).
Therefore (x, y, z) ≡ (0, 1, 5), (1, 2, 3) (mod 6).
In the first instance, if (x, y, z) = (6u, 6v + 1, 6w − 1), then (u + v + w)2 =
36(uv + vw + wu) + 6(w − v) − 1. But this is not possible, since −1 is not a
quadratic residue, modulo 6.
If (x, y, z) = (6u + 1, 6v + 2, 6w + 3), then (u + v + w)2 ≡ 11 (mod 6), which again
is not possible. Therefore k = 34 is not a possible value.
In the case of k = 46,

(x + y + z)2 = 16 × 3(xy + yz + zx),

so that x + y + z is divisible by 4. Therefore (x, y, z) ≡ (0, 1, 3) (mod 4). If

(x, y, z) = (4u, 4v + 1, 4w − 1), then

(u + v + w)2 = 3[(uv + vw + wu − v + w) − 1].

Since u + v + w is divisible by 3, (u, v, w) ≡ (1, 1, 1), (0, 1, −1) (mod 3) up to order.

Therefore, in both cases,

uv + vw + wu − v + w − 1 ≡ −1

(mod 3). But then 32 divides the left side, but not the right side. Therefore there
is no solution when k = 46.
Tables of possibly acceptable values
We have the following families:

k Solution of f (x, y, z) = k
r(r + 1)(r2 + r + 4) + 2 = (r2 + r + 2)2 − 2 (−r, r + 1, r2 + r + 1)
−[r2 (r + 1)2 + 2] (−r, r + 1, r2 + r − 1)
−(r2 + 2) (−1, 1, r)
−(9r2 + 2) (−r, 2r − 1, 2r + 1)
r2 + 1 (−(r − 1), r + 1, r2 + 1)
4r(r − 1) + 2 (−(r − 1), r, 2r(r − 1) + 1)

Tabulation of some numerical solutions

Positive values of k are not possible values of f (x, y, z) when k is a multiple of 3,
k ≡ 0, 3 (mod 4), k = 4k m2 − 2, or k + 2 is divisible to an odd power by a prime
congruent to 2 (mod 3). Negative values of k are not possible when |k| − 2 ≡ 2
(mod 3), k ≡ 0, 3 (mod 4) or |k| − 2 is divisible to an odd power by a prime

Crux Mathematicorum, Vol. 51(1), January 2025

 Ted Barbeau /31

congruent to 2 (mod 3). The values of k that are candidates are listed below,
along with solutions when available.

k Some solution of f (x, y, z) = k

−102 (−1, 1, 10)
−83 (−1, 1, 9), (−3, 5, 7), (−993, 5, 7)
−66 (−1, 1, 8)
−51 (−1, 1, 7)
−38 (−1, 1, 6), (−2, 3, 5), (−3, 2, 43), (−302, 3, 5)
−27 (−1, 1, 5)
−18 (−1, 1, 4)
−11 (−1, 1, 3), (−3, 1, 23), (−43, 1, 3)
−6 (−1, 1, 2), (−2, 1, 7), (−17, 1, 2)
−3 (−1, 1, 1), (−5, 1, 1), (−5, 1, 11), (−11, 5, 19), (−31, 1, 11)
−2 (−1, 0, 1)
2 (0, 1, 1), (1, 1, 4), (1, 4, 9), (1, 9, 16)
5 (−1, 3, 5), (−1, 5, 17), (−1, 17, 75), (3, 5, 41)
10 (−1, 2, 5), (−1, 5, 38), (2, 5, 71), (5, 71, 758)
14 (−1, 2, 3), (−1, 2, 11), (−1, 3, 26), (2, 3, 71)
17 (−3, 5, 17)
26 (−2, 3, 13)
29
37 (−5, 7, 37)
41
46
50 (−3, 4, 25)
61
62 (−2, 3, 7), (−2, 3, 55), (−2, 7, 307), (3, 7, 622)
65 (−7, 9, 65)
70
73
74
77
82 (−4, 5, 41)
85
89
94
98 (−3, 5, 8), (−3, 5, 188), (5, 8, 1277)
101 (−9, 11, 101)

Copyright © Canadian Mathematical Society, 2025

32/ Problems

PROBLEMS
Click here to submit problems proposals as well as solutions, comments
and generalizations to any problem in this section.

To facilitate their consideration, solutions should be received by March 15, 2025.

5001. Proposed by Mihaela Berindeanu.

Let ABC be a triangle with the incircle Γ , the incenter I and A0 , B 0 , C 0 the
points of tangency of circle Γ with the sides BC, AC, AB. If A0 I ∩ B 0 C 0 =
A1 , B 0 I ∩ A0 C 0 = B1 , C 0 I ∩ A0 B 0 = C1 , AA1 ∩ BC = A2 , BB1 ∩ AC = B2 and
−−→ −−→ −−→ → −
CC1 ∩ AB = C2 , show that AA2 + BB2 + CC2 = 0 .

5002. Proposed by Max A. Alekseyev.

Let m and n be positive integers such that n does not divide m. Prove that
n−1
πm 1X 2πmj
cot = (2n − 1 − 2j) sin .
n n j=0 n

5003. Proposed by Michael Friday, modified by the Editorial Board.

Let H be the orthocenter of a triangle ABC for which B = A + 90◦ . Denote by
D, E, F the feet of the altitudes from A, B, C respectively, and by T the foot of
the bisector of ∠C. Prove that the circle with center F and radius F H contains
the points C, D, E, and T .

5004. Proposed by Rajesh Sharma and Vijay Kumar.

Let z1 , z2 , z3 , z4 denote four distinct complex numbers. Prove that following con-
ditions are equivalent:
(i) The points z1 , z2 , z3 , z4 form the vertices of a rectangle in the plane.
r
(ii) m3 = 0 and Re m2 m4 − m22 = 0, where mr = 41 4j=1 (zj − ze) and
p P

ze = 14 4j=1 zj .
P

5005. Proposed by Daniel Sitaru.

Let x, y, z > 0 such that xyz = 1. Show that
 x y z 3 x3 y3 z3
+ + ≤ + +
1 + x + xy 1 + y + yz 1 + z + zx 1 + x + xy 1 + y + yz 1 + z + zx

Crux Mathematicorum, Vol. 51(1), January 2025

 Problems /33

5006. Proposed by Pericles Papadopoulos.

In the plane of a triangle ABC, consider a point P not lying on a side of the
triangle. The P -symmedians of the triangles P BC, P AC and P AB meet the
sides BC, AC and AB at A1 , B1 and C1 respectively. Prove that AA1 , BB1 and
CC1 concur at a point L satisfying

AL BL CL (P A2 + P B 2 )(P A2 + P C 2 )(P B 2 + P C 2 )
+ + = − 2.
LA1 LB1 LC1 P A2 P B 2 P C 2

5007. Proposed by Michel Bataille.

Let m and n be non-negative integers. Evaluate in closed form:
n Ç å m Ç å
X m + k n−k X n + k m−k
2 + 2 .
k k
k=0 k=0

5008. Proposed by George Apostolopoulos.

Let ABC be a triangle with angles denoted by A, B, C. Prove that
ã √
X Å k 3 k 3
sec − tan ≤ .
2 2 2 2
k=A,B,C

5009. Proposed by Vasile Cortoaje.

Let a, b, c, d be nonnegative real numbers such that ab + bc + cd = 7. Prove that
1 1 1 1 3
+ + + ≥ .
a+1 b+1 c+1 d+1 2

5010. Proposed by Nguyen Viet Hung, modified by the Editorial Board.

Let M be on the arc BC opposite A of the circumcircle of triangle ABC. Prove
that
b+c [ABM C]
≤ ,
a [M CB]
where a = BC, b = CA, c = AB, and square brackets represent areas. When does
equality hold?

.................................................................

Copyright © Canadian Mathematical Society, 2025

34/ Problems

Cliquez ici afin de proposer de nouveaux problèmes, de même que pour

offrir des 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 mars 2025.

5001. Soumis par Mihaela Berindeanu.

Soit ABC un triangle dont le cercle inscrit est Γ , le centre du cercle inscrit est I,
et pour lequel A0 , B 0 et C 0 désignent les points de tangence du cercle Γ avec les
côtés BC, AC, et AB respectivement. Si A0 I ∩ B 0 C 0 = A1 , B 0 I ∩ A0 C 0 = B1 ,
C 0 I ∩ A0 B 0 = C1 , AA1 ∩ BC = A2 , BB1 ∩ AC = B2 et CC1 ∩ AB = C2 , montrez
−−→ −−→ −−→ → −
que AA2 + BB2 + CC2 = 0 .

5002. Soumis par Max A. Alekseyev.

Soient m et n des entiers positifs tels que n ne divise pas m. Montrez que
n−1
πm 1X 2πmj
cot = (2n − 1 − 2j) sin .
n n j=0 n

5003. Soumis par Michael Friday, modifié par le comité de rédaction.

Soit H l’orthocentre d’un triangle ABC pour lequel B = A + 90◦ . On note D, E
et F les pieds des hauteurs de A, B et C respectivement. De plus, on note T le
pied de la bissectrice de ∠C. Montrez que le cercle de centre F et de rayon F H
contient les points C, D, E et T .

5004. Soumis par Rajesh Sharma et Vijay Kumar.

Soient z1 , z2 , z3 et z4 quatre nombres complexes distincts. Montrez que les condi-
tions suivantes sont équivalentes:
(i) Les points z1 , z2 , z3 , z4 forment les sommets d’un rectangle dans le plan.
r
(ii) m3 = 0 et Re m2 m4 − m22 = 0, où mr = 14 4j=1 (zj − ze) et
p P
1 P4
ze = 4 j=1 zj .

5005. Soumis par Daniel Sitaru.

Soient x, y, z > 0 tels que xyz = 1. Montrez que
 x y z 3 x3 y3 z3
+ + ≤ + + .
1 + x + xy 1 + y + yz 1 + z + zx 1 + x + xy 1 + y + yz 1 + z + zx

Crux Mathematicorum, Vol. 51(1), January 2025

 Problems /35

5006. Soumis par Pericles Papadopoulos.

Dans le plan d’un triangle ABC, on considère un point P qui n’est pas situé sur un
côté du triangle. Les P -symédianes des triangles P BC, P AC et P AB rencontrent
les côtés BC, AC et AB en A1 , B1 et C1 respectivement. Montrez que AA1 , BB1
et CC1 sont concourantes en un point L satisfaisant

AL BL CL (P A2 + P B 2 )(P A2 + P C 2 )(P B 2 + P C 2 )
+ + = − 2.
LA1 LB1 LC1 P A2 P B 2 P C 2

5007. Soumis par Michel Bataille.

Soient m et n des entiers non négatifs. Evaluez sous forme close l’expression
suivante :
n Ç å m Ç å
X m + k n−k X n + k m−k
2 + 2 .
k k
k=0 k=0

5008. Soumis par George Apostolopoulos.

Soit ABC un triangle dont les angles sont identifiés par A, B et C. Montrez que
ã √
X Å k 3 k 3
sec − tan ≤ .
2 2 2 2
k=A,B,C

5009. Soumis par Vasile Cortoaje.

Soient a, b, c et d des nombres réels non négatifs tels que ab + bc + cd = 7. Montrez
que
1 1 1 1 3
+ + + ≥ .
a+1 b+1 c+1 d+1 2

5010. Soumis par Nguyen Viet Hung, modifié par le comité de rédaction.
Soit M un point sur l’arc BC opposé A du cercle circonscrit au triangle ABC.
Montrez que
b+c [ABM C]
≤ ,
a [M CB]
où a = BC, b = CA et c = AB et où les crochets représentent les aires. Quand
a-t-on égalité ?

Copyright © Canadian Mathematical Society, 2025

36/ Solutions

SOLUTIONS
No problem is ever permanently closed. The editor is always pleased to consider for
publication new solutions or new insights on past problems.
Statements of the problems in this section originally appear in 2024: 50(6), p. 313–316.

4951. Proposed by Michel Bataille.

Let n be a positive integer. Prove that the sums
n Ç å−1 n Ç å
X (−1)k−1 n X (−1)k−1 n
and
k k k k−1
k=1 k=1

are equal and find their common value.

We received 25 submissions, all of which were correct and complete. We present
an amalgamation of solutions based upon the independent submissions of Chikara
Tsugawa, Giuseppe Fera, Changyi Yu, and many others.
Let
n Ç å−1 n n
X (−1)k−1 n X (−1)k−1 k!(n − k)! 1 X
an := = · = (−1)k−1 (k − 1)!(n − k)!
k k k n! n!
k=1 k=1 k=1

and let g(k) := (−1)k−1 (k − 1)!(n − k + 1)!. Then,

g(k + 1) − g(k) = (−1)k k!(n − k)! − (−1)k−1 (k − 1)!(n − k + 1)!

= (−1)k−1 (k − 1)!(n − k)!(−k − n + k − 1)
= −(n + 1)(−1)k−1 (k − 1)!(n − k)!

Hence,
n
1 X 1 1 + (−1)n+1
an = − (g(k + 1) − g(k)) = − (g(n + 1) − g(1)) = .
(n + 1)! (n + 1)! n+1
k=1

Next, we define bn and reindex the sum to obtain:

n Ç å n−1 Ç å
X (−1)k−1 n X (−1)k n
bn := = .
k k−1 k+1 k
k=1 k=0

By applying the binomial formula,

n Ç å
n
X
n−k n k
(x − 1) = (−1) x .
k
k=0

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /37

Hence,
Z 1 Z n
1X
Ç å
n n−k n k
(x − 1) dx = (−1) x dx
0 0 k=0 k
and ô1
ò1 n ñ Ç å
(x − 1)n+1 (−1)n−k n k+1
ï X
= x .
n+1 0 k+1 k 0
k=0

Therefore,
n Ç å
X (−1)n−k n (−1)n+1
=−
k+1 k n+1
k=0

and after multiplying both sides by (−1)n we have,

n Ç å
X (−1)k n 1
= .
k+1 k n+1
k=0

That is,
n−1 Ç å
X (−1)k n (−1)n 1
+ =
k+1 k n+1 n+1
k=0

and so,
1 + (−1)n+1
bn = = an .
n+1

Editor’s Comments. Giuseppe Fera wrote bn as a telescoping sum by noting that

if we define Ç å
(−1)k n
h(n, k) = ,
n+1 k−1
then
Ç å Ç å ÇÇ å Ç åå
(−1)k+1 n (−1)k n (−1)k−1 n n
h(n, k + 1) − h(n, k) = − = +
n+1 k n+1 k−1 n+1 k k−1
Ç åÅ Ç å
(−1)k−1 n−k+1 (−1)k−1
ã
n n
= +1 =
n+1 k−1 k k k−1

and so,
n Ç å n
X (−1)k−1 n X
bn = = (h(n, k + 1) − h(n, k))
k k−1
k=1 k=1
(−1)n+1 + 1
= h(n, n + 1) − h(n, 1) = = an .
n+1
Several submissions included references to closely related results.

Copyright © Canadian Mathematical Society, 2025

38/ Solutions

4952. Proposed by Mihaela Berindeanu.

Let ABC be a non-isosceles triangle with the circumcircle Γ , the centroid G, AG∩
Γ = {X} , BG ∩ Γ = {Y } and CG ∩ Γ = {Z}. Prove that the triangle XY Z is
isosceles with XY = XZ if and only if BXCG is a parallelogram.
Of the 13 submissions received, all but one was complete and correct. We feature
two of them.
Solution 1 is a composite of the similar solutions by Michal Adamaszek and Sarthak
Dattatray Dhobale.
The result is an easy consequence of the theorem that two chords of the same
circle have the same length if and only if they subtend congruent angles. Thus,
under the assumption that BXCG is a parallelogram, because the opposite angles
of a parallelogram are equal (in particular, the angles at B and at C), they are
subtended by equal chords, namely XY = XZ, as desired. But even more is true:
These equal chords have the same length as the side BC of the given
triangle.
This fact is seen as follows: the angles at Y and C (marked p in the diagram)
are subtended by the chord ZB and are therefore equal; similarly for the angles
at Y and B subtended by XC and marked q. But in the parallelogram BXCG,
the transversal BC forms alternate interior angles p and q with the parallel lines
ZC and BX, whence ∠p = ∠q so that the chord XZ that subtends ∠XY Z must
equal the chord BC that subtends ∠CY B = ∠CAB.

For the converse we assume that XY = XZ and will prove that BXCG is a
parallelogram. For this we need a simple lemma.
Lemma. Let P QRS be a convex quadrilateral such that ∠P = ∠R, the diagonal
QS meets the diagonal P R in its midpoint, and the diagonals are not perpendic-
ular. Then P QRS is a parallelogram.

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /39

Proof. Let D = QS ∩ P R and let T be the reflection of R in the line QS. Since
P R is not perpendicular to QS, we have T 6= P . By symmetry we note that
∠QT S = ∠QRS = ∠QP S so that the points Q, S, T, P are concyclic. Since
DP = DR = DT , the triangle P RT is a right triangle with circumcenter D
and ∠RT P = 90◦ . Since RT ⊥ P T and RT ⊥ QS, we get P T ||QS. A cyclic
quadrilateral with a pair of parallel sides is an isosceles trapezoid. From this and
symmetry we get
P Q = T S = SR,

P S = QT = QR,

whence P QRS is a parallelogram and the lemma is proved.

For the proof that BXCG is a parallelogram, we recall that XY = XZ is equiv-

alent to equal angles of the parallelogram at B and C, whence BXCG is a par-
allelogram by the lemma because it is convex (as the centroid G is inside ∆ABC
and X is outside) while AD is the median of 4ABC and (as the triangle is not
isosceles) it is not perpendicular to BC.

Alternative proof of the converse by Miguel Amengual Covas.

We assume that XY = XZ and will prove that the midpoint D of the side BC of
the given triangle ABC is also the midpoint of GX; it immediately will follow that
BXCG is a parallelogram, as claimed. Let 4ABC have sides a, b, c in the usual
order and let ma , mb , mc denote the lengths of the medians to a, b, c, respectively.
XY GX
The pairs of similar triangles GXY and GBA yield = , while the similar
BA GB
XZ GX
pair GXZ and GCA, yield = . Thus,
CA GC

GX · BA GX · CA
= XY = XZ = .
GB GC

With c = BA, b = CA, BG = 32 mb , and CG = 32 mc the previous equation becomes

b · mb = c · mc . (1)

Square both sides, substitute 14 2c2 + 2a2 − b2 for m2b , 1

2a2 + 2b2 − c2 for m2c ,



4
and obtain, after multiplying both sides by 4,

b2 2c2 + 2a2 − b2 = c2 2a2 + 2b2 − c2 .

(2)

Equation (2) is easily brought into the form

2 b2 − c2 b2 + c2 − 2a2 = 0.

Copyright © Canadian Mathematical Society, 2025

40/ Solutions

Now b2 − c2 cannot be zero since otherwise, triangle ABC would be isosceles,

which is a contradiction. Therefore,

b2 + c2 − 2a2 = 0. (3)

Eliminating b2 + c2 between this and 4m2a = 2 b2 + c2 − a2 , we get

4m2a = 3a2 . (4)

ma
Let {D} = BC ∩ AX. Then D is the midpoint of side BC and GD = 3 .

Since A, B, X, C are concyclic,

AD · DX = BD · DC,

whence
a a
BD · DC · a2 ma
DX = = 2 2 = = (substituting from (4)) = = GD,
AD ma 4ma 3
implying that D is the midpoint of GX (since G, D, X are collinear). Thus, BC
and GX bisect each other, making BXCG a parallelogram.

Editor’s comments. Equation (3) describes a property of triangles that has often
appeared on the pages of Crux Mathematicorum: » 2 because a is the root mean square
b +c2
of the other two sides b and c (that is, a = 2 ), it is natural to call a triangle
whose sides satisfy 2a2 = b2 + c2 a root-mean-square triangle. In a series of
Mathesis articles [1], 69 properties of these triangles were listed, generally with an
abundance of earlier references instead of proofs. Many of those properties were
discussed in [2]. With emphasis on other properties, these triangles have also been
called automedian and self-median, as well as quasi-isosceles. Just two months ago
(Problem 4936 in the November, 2024 issue, pages 477-481), two further properties
were added to the list.
Miguel Amengual Covas kindly sent along a list of 10 automedian triangle prop-
erties [3] compiled by Francisco Bellot Rosado:

1. The Euler’s line is perpendicular to the median AX.

2. Triangles BGX and CXG are inversely similar to 4ABC.
3. The line GK is parallel to BC, where K denotes the Lemoine (or symmedian)
point of 4ABC.
4. The lengths of segments BT , AT , CT , where T denotes the Fermat’s point
of 4ABC, are in arithmetic progression in this order.
5. 2m2a = m2b + m2c .
6. 2 cot A = cot B + cot C.

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /41

7. OK ⊥ AK, where O denotes the circumcenter of 4ABC.

8. 2a2 = b2 +c2 if and only if the bisectors of the three pairs of angles ABC and
AGC, BCA and BGA, CAB and CGB, meet on CA, AB, BC, respectively.
9. ΩΩ0 ⊥ AK, where Ω, Ω0 denote the first and second Brocard’s points.
√ √
10. The triangle with sides 3, 2, 1 is the only right-angled automedian tri-
angle.

Bibliography
[1] Mathesis: (1902) 205-208; (1903) 196-200, 226-230, 245-248; (1926) 68-69.
[2] J. Chris Fisher, Recurring Crux Configurations 1: Triangles for which 2b2 =
c2 + a2 . Crux Mathematicorum, 37(5) (September, 2011) 304-307.
[3] Francisco Bellot Rosado, Revista Escolar de la Olimpiada Iberoamericana,
Num.19, (May - June 2005) ISSN-1698-277X.

4953. Proposed by Michel Bataille.

Let a ∈ R − {0, 1}. Find all a and all functions f : R → R that are not identically
zero such that
(x + y)f (x + y) − xf (x) − yf (y) = a(yf (x) + xf (y))
for all x, y.
There were 14 solutions, 12 of which were correct and 2 were incomplete. Our
solution essentially follows those of Devis Alvarado, Theo Koupelis, UCLan Cyprus
Problem Solving Group and the proposer.
The substitutions (x, y) = (1, 0) and (x, y) = (t, −t) lead respectively to f (0) = 0
and f (t) = f (−t) for t ∈ R. The substitutions (x, y) = (1, 1) and (x, y) = (2, −1)
lead respectively to
f (2) = (1 + a)f (1)
and
(a − 2)f (2) = 2(a − 1)f (1).
Eliminating f (2), we find that
a(a − 3)f (1) = [(a + 1)(a − 2) − 2(a − 1)]f (1) = 0,
whereupon either a = 3 or f (1) = 0.
From (x, y) = (t, 1), we obtain
(t + 1)f (t + 1) − tf (t) − f (1) = af (t) + atf (1).
From (x, y) = (t + 1, −1), we obtain
tf (t) − (t + 1)f (t + 1) + f (1) = −af (t + 1) + a(t + 1)f (1).

Copyright © Canadian Mathematical Society, 2025

42/ Solutions

Adding these equations, we find that

f (t + 1) = f (t) + (2t + 1)f (1).

Substituting this into one of the previous equations yields

(a − 1)f (t) = t[2t + (3 − a)]f (1).

If f (1) = 0, then f (t) = 0 for all t. If a = 3, then f (t) = f (1)t2 for all t. These
are easily checked.

Comments by the editor. Didier Pichon and C.R. Pranesachar, independently,

showed that f (2x) = (a + 1)f (x) and 3f (3x) = (a2 + 5a + 3)f (x).

4954. Proposed by Sicheng Du.

In triangle ABC, let I be the incenter and D, E, F be the points of tangency of the
incircle on BC, CA, AB, respectively. If the circumcircles of triangles ABD and
AIC intersect at A and G, show that the lines AG, DE, and F I are concurrent.
All 10 submissions were correct; we feature the solution by the UCLan Cyprus
Problem Solving Group.
Let X be the point of intersection of AG and F I; we are to show that X ∈ ED.
Because of the right angles at F and D, we know that B, F, I, D are concyclic,
and
∠XGD = ∠AGD = 180◦ − ∠ABD = ∠F ID = 180◦ − ∠XID.
Therefore X, G, D, I are concyclic. Since C, D, I, E are also concyclic, then

∠ACB 180◦ − ∠DIE

∠XDI = ∠XGI = ∠AGI = ∠ACI = = = ∠EDI.
2 2
Thus X also lies on ED as required.

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /43

4955. Proposed by Vasile Córtoaje.

For given n ≥ 6, find the largest positive constant kn such that
1 1 1 n
+ + ··· + ≥
a1 + kn a2 + kn an + kn 1 + kn
whenever a1 ≥ 1 ≥ a2 ≥ · · · ≥ an ≥ 0 satisfying a1 a2 + a2 a3 + · · · + an a1 = n.
We received 3 submissions, only 1 of which was correct and complete. We present
the solution by the proposer, slightly altered by the editor.
By setting a1 = 3, a2 = · · · = an−1 = 1 and an = 0, the desired inequality leads to
the necessary condition kn ≤ 3. To show that 3 is the largest value of kn whenever
n ≥ 6, we need to prove the inequality
1 1 1 n
+ + ··· + ≥ .
a1 + 3 a2 + 3 an + 3 4
Let
a1 + an a2 + a3 + · · · + an−1
S= , x= , where a1 ≥ 1 ≥ x ≥ an .
2 n−2
By Titu’s inequality we have

1 1 (n − 2)2 n−2
+ ··· + ≥ = .
a2 + 3 an−1 + 3 (a2 + 3) + · · · + (an−1 + 3) x+3
So, it suffices to show that
1 1 n−2 n
+ + ≥ . (∗)
a1 + 3 an + 3 x + 3 4
By Lemma below, we have

(n − 3)x2 + x(a1 + an ) + a1 an ≤ n.

Since the left side of (∗) decreases when a1 increases, we may replace this inequality
constraint with the equality constraint

(n − 3)x2 + x(a1 + an ) + a1 an = n, i.e. a1 an = n − 2xS − (n − 3)x2 .

From (a1 − x)(an − x) ≤ 0, we get a1 an ≤ 2xS − x2 . Since

n − 2xS − (n − 3)x2 = a1 an ≤ 2xS − x2 ,

we get
n − (n − 4)x2
S ≥ S1 = .
4x
Since
1 1 2S + 6 2S + 6
+ = = ,
a1 + 3 an + 3 a1 an + 9 + 6S n + 9 + 2(3 − x)S − (n − 3)x2

Copyright © Canadian Mathematical Society, 2025

44/ Solutions

we need to show that

2S + 6 n − 8 − nx
+ ≥ 0,
n + 9 + 2(3 − x)S − (n − 3)x2 4(x + 3)

which can be written as 2A(x)S + B(x) ≥ 0, where

A(x) = nx2 − 4(n − 3)x + 3(n − 4) = (1 − x)(3n − 12 − nx),

B(x) = n(n − 3)x3 − (n − 3)(n − 8)x2 − (n2 + 9n − 24)x + n(n + 4).

Since x ≤ 1 and

3n − 12 − nx ≥ 3n − 12 − n = 2(n − 6) ≥ 0,

we have A(x) ≥ 0. So, it suffices to show that 2A(x)S1 + B(x) ≥ 0, which is

equivalent to the obvious inequality (x − 1)2 h(x) ≥ 0, where

h(x) = (n − 2)x2 + 2(2n − 5)x + 3(n − 4) > 0.

Thus, the proof is completed.

For kn = 3, the equality occurs when a2 = · · · = an−1 = 1 and a1 + an + a1 an = 3
such that a1 ≥ 1 ≥ an ≥ 0.
Lemma. Let a1 ≥ a2 ≥ . . . ≥ an ≥ 0 such that a1 a2 + a2 a3 + · · · + an a1 = n. If
n ≥ 4 and
a2 + · · · + an−1
x= , then (n − 3)x2 + x(a1 + an ) + a1 an ≤ n,
n−2
with equality for a2 = · · · = an−1 .
Proof. Write the desired inequality as follows:

(n − 3)x2 + x(a1 + an ) + a1 an ≤ a1 a2 + a2 a3 + · · · + an a1 ,

(n − 3)x2 + a1 (x − a2 ) + an (x − an−1 ) ≤ a2 a3 + · · · + an−2 an−1 .

Since x − a2 ≤ 0 and x − an−1 ≥ 0, it suffices to show that

(n − 3)x2 + a2 (x − a2 ) + an−1 (x − an−1 ) ≤ a2 a3 + · · · + an−2 an−1 ,

which can be rewritten as

a2 a3 + · · · an−2 an−1 ≥ (n − 3)x2 + (a2 + an−1 )x − a22 − a2n−1 .

Since the sequences a2 , a3 , . . . , an−2 and a3 , a4 , . . . , an−1 are decreasing, by Cheby-

shev’s inequality we have

(n − 3)(a2 a3 + · · · + an−2 an−1 ) ≥ (a2 + · · · + an−2 )(a3 + · · · + an−1 )

= ((n − 2)x − an−1 ) ((n − 2)x − a2 ) .

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /45

Thus, it suffices to show that

((n − 2)x − an−1 ) ((n − 2)x − a2 )
≥ (n − 3)x2 + (a2 + an−1 )x − a22 − a2n−1 ,
n−3
which is equivalent to
(2n − 5)x2 − (2n − 5)(a2 + an−1 )x + (n − 3)(a22 + a2n−1 ) + a2 an−1 ≥ 0,
(2n − 5)(2x − a2 − an−1 )2 + (2n − 7)(a2 − an−1 )2 ≥ 0.
Clearly, the last inequality is true.

Editor’s Comments. For n = 4 and n = 5, the largest positive constant kn is

k4 ≈ 2.9260 and k5 ≈ 2.9938, respectively.

4956. Proposed by Toyesh Sharma.

If a, b, c > 1 and x, y, z > 0, show that
1 ≤ logabc (ax by cz ) · logabc (a1/x b1/y c1/z ).

We received 22 submissions, only 14 of which were correct and complete. We

present the solution by Giuseppe Fera.
Let u = logabc a, v = logabc b, w = logabc c. [Note that u, v, w > 0 because
a, b, c > 1.] Thus
logabc ax by cz · logabc a1/x b1/y c1/z
1 1 1 
= (xu + yv + zw) u + v + w
x y z
x y  y z z x
2 2 2
=u +v +w + + uv + + vw + + wu
y x z y x z
≥ u2 + v 2 + w2 + 2uv + 2vw + 2wu = (u + v + w)2
[because t + 1/t ≥ 2 for all t > 0]. But u + v + w = logabc abc = 1 and the proof is
complete.

Editor’s Comments. All (correct) solutions boiled down to the inequality

(ux + vy + wz)(u/x + v/y + w/z) ≥ (u + v + w)2
and differed only in their proofs of this. The elementary argument featured above
was the most common. One solver (the proposer) applied Jensen’s inequality to the
convex function t 7→ 1/t; another (M. Adamaszek) recognized this as the inequality
of the weighted harmonic and arithmetic
√ means. The rest
p applied
p the pCauchy–
√ √
Schwarz inequality to the vectors ( ux, vy, wz) and ( u/x, v/y, w/z), an
approach which generalizes the result to any number of variables (W. Janous):
n
Y  n
Y 
x 1/x
loga1 ...an aj j · loga1 ...an aj j ≥ 1
j=1 j=1

Copyright © Canadian Mathematical Society, 2025

46/ Solutions

for all positive integers n and real numbers a1 , . . . , an > 1 and x1 , . . . , xn > 0. A
few solvers (M. Bataille, T. Koupelis, P. Perfetti, C. R. Pranesachar) considered
a, b, c > 0 and proved the result whenever a, b, c all lie on the same side of 1 (as
long as abc 6= 1), giving counterexamples otherwise.

4957. Proposed by George Apostolopoulos.

Let ABC be a triangle with I as its incenter. Let R and r be the circumradius
and inradius of ABC, respectively. Prove that
√ AB BC CA 3R p
3 3≤ + + ≤ 2 R2 − r 2 .
IC IA IB 2r

We received 12 solutions for this problem. The following solution is by Tapas Das.
Let
r r r
AB = c, BC = a, CA = b, IC = C
, IA = A
, IB = .
sin 2 sin 2 sin B2

So,
1
Å ã
AB BC CA C A B
+ + = c sin + a sin + b sin
IC IA IB r 2 2 2
ÃÇ åÇ å
CBS 1 X X C
≤ c2 sin2
r 2
… …
Leibniz 1
2
 r  3R r2
≤ (9R ) 1 − = 1−
r 2R r 2Rr
s …
Euler 3R r2 3R r2 3R p 2
≤ 1− R
= 1− 2 = R − r2
r 2R · 2 r R Rr
Euler 3R p
≤ R2 − r 2
2r · r
3R p
= 2 R2 − r 2 .
2r
Similarly,
1
Å ã
AB BC CA C A B
+ + = c sin + a sin + b sin
IC IA IB r 2 2 2
… …
AM - GM 3 3 A B C 3 r
≥ abc sin sin sin = 3 4Rrs
r 2 2 2 r 4R
√
(s≥3 3r) 3 Ä √ ä 13 3 √
≥ r2 3 3r = · 3r
√ r r
=3 3.

Equality holds for an equilateral triangle.

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /47

√
Editor’s Comments. Inequality s ≥ 3 3r used in the proof can be found on p.
52 of O. Bottema et al., Geometric inequalities, Wolters-Noordhoff Publishing,
Gröningen, 1969. Walther Janous generalized the result in the following way:
Ç √ å
2s 1 r s 3 AB BC CA 1 (s2 − r2 − 4Rr)(2R − r)
− + ≤ + + ≤ .
3r 2 4R 4R IC IA IB R R

4958. Proposed by Nguyen Viet Hung.

Let a, b, c be non-negative real numbers, no two of which are zero, satisfying a2 +
b2 + c2 = 2. Prove that

a(2 + bc)2 b(2 + ca)2 c(2 + ab)2 (3abc)2

+ + ≥8+ .
b+c c+a a+b 2(ab + bc + ca)

We received 9 submissions, of which 6 were correct and complete. We present two

solutions.
Solution 1, by the proposer, slightly altered by the editor.
Using the given condition and the AM-GM inequality we have:

a(2 + bc)2 a(4 + 4bc + b2 c2 ) 2a(2 + 2bc) ab2 c2

= = +
b+c b+c b+c b+c
2a(a2 + b2 + c2 + 2bc) ab2 c2
= +
b+c b+c
2a3 + 2a(b + c)2 ab2 c2
= +
b+c b+c
2a3 ab2 c2
= + 2a(b + c) +
b+c b+c
√ ab2 2
c
≥ 2 2a3 · 2a +
b+c
2 2
ab c
= 4a2 + .
b+c
After adding corresponding sides of similar inequalities, Titu’s inequality yields:
X a(2 + bc)2 ab2 c2 bc2 a2 ca2 b2
≥ 4(a2 + b2 + c2 ) + + +
cyc
b+c b+c c+a a+b
1 1 1
Å ã
= 8 + a2 b2 c2 + +
ab + ac bc + ba ca + cb
(3abc)2
≥8+ .
2(ab + bc + ca)

Copyright © Canadian Mathematical Society, 2025

48/ Solutions

Solution 2, by Theo Koupelis, slightly altered by the editor.

Without loss of generality, assume a ≥ b ≥ c ≥ 0. By the Cauchy-Schwarz inequal-

ity:
ñ ô ñ ô Ç å2
X X a(2 + bc)2 X
a(b + c) · ≥ 2 a + 3abc .
c c
b+c c

Thus, it suffices to show that

[2(a + b + c) + 3abc]2 (3abc)2

≥8+ .
2(ab + bc + ca) 2(ab + bc + ca)

We multiply by 2(ab + bc + ca) and simplify to reduce the problem to a quadratic

inequality in c:

p(c) := c2 (1 + 3ab) + c(a + b)(3ab − 2) + (a − b)2 ≥ 0,

which clearly holds if ab ≥ 2/3, with equality when c = 0 and a = b.

p
If ab < 2/3, from the condition a2 + b2 + c2 = 2 we have c ≤ 2/3 and so

a2 + b2 = 2 − c2 ≥ 4/3.

The discriminant of p(c) is given by

∆ = ab (a2 + b2 )(9ab − 24) + 18a2 b2 + 16

 

4
ï ò
≤ ab 18a2 b2 + 16 − (24 − 9ab) ·
3
2 2


= ab 18a b + 12ab − 16
< ab(8 + 8 − 16)
= 0.

Therefore, the original inequality holds with equality when (a, b, c) = (1, 1, 0) and
cyclical permutations.

4959. Proposed by Jeromin Rocklage.

For α > 0, evaluate the limit

2n Å ãα
X
k k
lim (−1) .
n→∞ 2n
k=1

We received 19 solutions. Some solutions used the Stolz-Cesàro theorem and others
focused more on integral or sum estimates. We present 2 solutions.

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /49

Solution 1, by Mingcan Fan.

By the Stolz-Cesàro theorem, and the equivalent infinitesimals (1 + x)α − 1 ∼ αx
when x → 0, we have
2n ãα P2n+2
(−1)k k α − 2n k α
k=1 (−1) k
Å P
X k
lim (−1)k = lim k=1
n→∞ 2n n→∞ (2n + 2)α − (2n)α
k=1
(2n + 2)α − (2n + 1)α
= lim
n→∞ (2n + 2)α − (2n)α
[(1 + 1/n)α − 1] − [(1 + 1/(2n))α − 1]
= lim
n→∞ (1 + 1/n)α − 1
α/n − α/(2n)
= lim
n→∞ α/n
1
= .
2

Solution 2, by Didier Pinchon.

Let fα be the function defined for x ∈ (0, 1) by fα (x) = xα , and Sn (α) the sum
2n Å ã
X k
Sn (α) = (−1)k fα .
2n
k=1

Sn (α) may be rewritten

n ï
2k 2k − 1
X Å ã Å ãò
Sn (α) = fα − fα .
2n 2n
k=1

Because fα is a differentiable function on interval 2k−1 2k

 
2n , 2n for 1 ≤ k ≤ n,

2k 2k − 1 2k 2k − 1 1 0
Å ã Å ã Å ã
α
fα − fα = − fα0 (zk ) = f (zk ) = gα (zk ),
2n 2n 2n 2n 2n α 2n
with gα (x) = xα−1 , x ∈ (0, 1), and zk ∈ 2k−1 2k


2n , 2n by the mean value theorem.

So
n
1X 2k − 1 2k k−1 k
Å ã Å ã
α
Sn (α) = × gα (zk ), zk ∈ , ⊂ , , 1 ≤ k ≤ n.
2 n 2n 2n n n
k=1

A Riemann sum is identified for the Riemann integrable function gα , and

n Z 1
1X 1 1
lim gα (zk ) = gα (t) dt = (fα (1) − fα (0)) = ·
n→∞ n 0 α α
k=1

It follows that
1
lim Sn (α) = ·
n→∞ 2

Copyright © Canadian Mathematical Society, 2025

50/ Solutions

4960. Proposed by Yagub N. Aliyev.

a) Suppose that a, b and c are real numbers. Prove that if b2 = a2 + ac and
c2 = b2 + ab, then abc = (a + b + c)(−a2 − b2 + c2 ).
b) Show that there are no positive integers a, b and c such that b2 = a2 + ac
and c2 = b2 + ab.
c) Find positive integers a, b and c such that abc = (a + b + c)(−a2 − b2 + c2 ).
We received 13 solutions, some answering only parts a) and b). Part c) was the
hardest part of the problem. The majority of the solvers to this part used the idea
of point addition on elliptic curves, taking several steps to reach positive solutions
from trivial ones. We present two solutions.
Solution 1, by Sicheng Du.
a) From the conditions, c2 − b2 = ab and b2 − a2 − ac = 0, so
RHS − LHS = (a + b + c) c2 − b2 − a2 − abc



= a [(a + b + c)(b − a) − bc] = a b2 − a2 − ac = 0.

ä2
b2 −a2
Ä
b) Assume that there are. From the conditions, b2 + ab = c2 = a ; dividing

2 î
2 ó2
both sides by a2 gives ab + ab = ab − 1 .

Let u = ab , then u4 − 3u2 − u + 1 = 0. By a rational roots test, there is no rational

root, so ab is irrational, which contradicts with a, b, c being positive integers.
c) Here is a valid solution and how it can be obtained.
a = 49105016933436320224063,
b = 301361533449900458837600,
c = 316629033253501281102807.
We shall find positive rational numbers a, b satisfying
ab = (a + b + 1)(1 − a2 − b2 ). (1)
Let
−12x − 24y − 29 −12x + 24y − 29
a= , b=
48x + 44 48x + 44
so that
22a + 22b + 29 3(a − b)
x=− , y= ,
12(2a + 2b + 1) 2(2a + 2b + 1)
then (1) is transformed into
864x3 − 2178x + 864y 2 − 845 = 0. (2)

Since (1) has trivial solutions
(a, b) =
(0, ±1), (±1, 0), then (2) has corresponding
17
solutions (x, y) = − 12 7
, ± 12 , 12 , ± 23 .

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /51

Let Γ be the graph of (2). If points P1 , P2 with rational coordinates both lie on
Γ, then the line P1 P2 intersects Γ at another point P3 (if P1 P2 is tangent to Γ,
then P3 coincides with P1 or P2 ) because a cubic equation is produced. Moreover,
Vieta’s theorem implies that P3 has rational coordinates as well.

0
0 7 3

1. Let A − 17 1
− 17 1 7 3


12 , − 2 , A 12 , 2 , E 12 , − 2 , E 12 , 2 , which are points
on Γ with rational coefficients.
A0 E, AE 0 intersect Γ again at H, H 0 respectively, then H − 16 , − 34 ,


2. Let the lines
H 0 − 16 , 34 .

3. Let the line AH intersect Γ again at L, then L 463 273

300 , − 250 .

4. Let the line A0 L intersect Γ again at N , then N − 16428 6833 3873

, − 101306 .
5. Let the line LE intersect Γ again at R, then R − 443 1397


192 , − 512 .
230662356007 14098278361684893


6. Let the line RN intersect Γ again at S, then S 329529580428 , 9101222560244194 .
7. The line H 0 S intersect Γ again at − 26913592970611051 252256516516464138613537


19454486844752964 , − 678374756013449892817422 .

So (2) has a solution

26913592970611051 252256516516464138613537
x=− , y=−
19454486844752964 678374756013449892817422
and
49105016933436320224063 301361533449900458837600
a= , b= .
316629033253501281102807 316629033253501281102807

Copyright © Canadian Mathematical Society, 2025

52/ Solutions

Hence, the original equation has a solution with c = 1 and a, b the same as above.
Due to homogeneity, multiplying by the denominator gives the final answer.

Solution 2 for part c), by Giuseppe Fera.

Substituting a = xc, b = yc into abc − (a + b + c) −a2 − b2 + c2 = 0 and


then dividing by c3 , we need to find a rational point, with positive coordinates, on
the plane curve

C : x3 + x2 y + xy 2 + y 3 + x2 + xy + y 2 − x − y − 1 = 0, (x, y) ∈ R2

This is a non-singular cubic of genus 1, namely an elliptic curve with two real
components. The projective transformation
12X + 24Y − 29 12X − 24Y − 29
ß ™
T : x= , y=
44 − 48X 44 − 48X
puts the curve C in Weierstrauss normal form
121 845
CW : Y 2 = X 3 − X +
48 864
It is well-known that if PW (x1 , y1 ), QW (x2 , y2 ) are two points with rational coor-
dinates on CW , and RW is the intersection of the line PW QW with CW , then RW
is a point with rational coordinates. Obviously, the same property holds for the
points P = T −1 (Pw ) , Q = T −1 (Qw ), R = T −1 (Rw ) on C.
For CW in the form
Y 2 = X 3 + aX 2 + bX + c
and x2 6= x1 , it is immediate that

RW m2 − a − x1 − x2 , m m2 − a − x1 − x2 + y2 − mx2 ,

y2 −y1
where m = x2 −x1 . In our case, a = 0.
A numerical search provides the points P (0, −1), Q − 45 , − 14 on C. Using the


inverse transformation
22x + 22y + 29 3(y − x)
ß ™
T −1 : X = , Y =
12(2x + 2y + 1) 2(2x + 2y + 1)
7 3
, 2
, QW 61 , − 34 on CW . The corresponding RW is 113 57




we have PW − 12 12 , − 2
namely R 25 32
17 , − 17 on C.

The symmetry of the curve implies that if 61 , − 34 ∈ CW , also 61 , 34 ∈ CW .

the composition of PW 61 , 34 and QW
113 57

Next up, we consider 12 , − 2 , that is
6833 3873 17536 16245


RW 16428 , − 101306 . This corresponds to R − 16909 , − 16909 on C.
3 1
it with Q − 17536 16245



We use the point P − 5 , − 5
on C, and then combine 16909 , − 16909
.
In such case PW − 19 6833 3873 2566229 25978488


12 , −1 , QW 16428 , − 101306 and RW 1834572 , 59776471 ,

Crux Mathematicorum, Vol. 51(1), January 2025

 Solutions /53

1481089 18800081


which finally corresponds to R 19214131 , 19214131 on C. Hence, we have a ra-
tional point, with positive coordinates, on C and it gives a requested solution
a = 1481089, b = 18800081, c = 19214131.
The procedure is shown in the figure, where the triplets of collinear points on C
are on the lines red, green, blue, in that order:

Editor’s Comments. Various equivalent forms of the equation

abc = (a + b + c)(c2 − b2 − a2 ),

which can also be written as

a b c
+ = ,
b+c a+c a+b
and its connection with “Sharygin triangles” appeared in several sources, for ex-
ample see [1, 2, 3, 4, 5, 6]. A Sharygin triangle has the property that the base
points of its angle bisectors form an isosceles triangle. Nikolai Osipov found the
same trio of integers as in the above Solution 1 but using only 4 steps on a dif-
ferent elliptical curve. The trio (a, b, c) = (1481089, 18800081, 19214131) found in
Solution 2 above is the smallest one [3]. Michal Adamaszek found a trio of much
larger numbers using mathematical software. Lyubomir Lyubenov found that a
triangle with sides a, b, c such that c2 − b2 = ab and b2 − a2 = ac has angles π/7,
2π/7, and 4π/7. This triangle which is sometimes called a heptagonal triangle,
because it is formed by the 1st, 2nd and 4th vertices of a regular heptagon, is also
a Sharygin triangle [7].
References:
[1] Sharygin I. F., Problems in Geometry (Planimetry), Nauka [in Russian], 1982.

Copyright © Canadian Mathematical Society, 2025

54/ Solutions

[2] Sharygin I. F., About bisectors. Kvant [in Russian], 1983, no. 8, p. 32 – 36.
[3] Netay, I. V., Savvateev, A. V. (2017). Sharygin triangles and elliptic curves.
Bulletin of the Korean Mathematical Society, 54(5), p. 1597 – 1617. https:
//doi.org/10.4134/BKMS.B160681
[4] Belopolsky A., Sharygin equation, https://hackmd.io/@aab/H1HYhyqiv#Sharygin-triangle
[5] Nash O., Sharygin’s group of triangles, http://olivernash.org/2016/12/18/
sharygins-group-of-triangles/index.html
[6] Aliyev Y. N., Comments to solution of Problem 4895, Mathematics in School
[in Russian], 2006 (10), p. 65.
[7] Chelyabov I.M., Bakmaev Sh.A., Variations on Sharygin triangle theme, Math-
ematics in School [in Russian], 2006 (10), p. 66 – 68.

Crux Mathematicorum, Vol. 51(1), January 2025