Volume/tome 48, issue/numéro 2

February/février 2022
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 2022. 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 2022. TOUS DROITS RÉSERVÉS.

ISSN 1496-4309 (électronique)

Supported by / Soutenu par :

• Intact Financial Corporation
• University of the Fraser Valley

Editorial Board

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

MathemAttic Editors John McLoughlin University of New Brunswick

Shawn Godin Cairine Wilson Secondary School
Kelly Paton Quest University Canada
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
Edward Wang Wilfrid Laurier University
Dennis D. A. Epple Berlin, Germany
Magdalena Georgescu BGU, Be’er Sheva, Israel
Chip Curtis Missouri Southern State University
Philip McCartney Northern Kentucky University

Guest Editors Yagub Aliyev ADA University, Baku, Azerbaijan

Andrew McEachern York University
Vasile Radu Birchmount Park Collegiate Institute
Aaron Slobodin University of Victoria
Chi Hoi Yip University of British Columbia
Samer Seraj Existsforall Academy
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

60 Upcoming special issue in memory of Bruce Shawyer

61 MathemAttic: No. 32
61 Problems: MA155–MA160
63 Solutions: MA131–MA135
67 From the bookshelf of . . . Andy Liu
70 Teaching Problems: No. 15 Margo Kondratieva
80 Olympiad Corner: No. 400
80 Problems: OC566–OC570
82 Solutions: OC541–OC545
89 The Last Problem: Demystified Sam Hopkins
94 Problems: 4711–4720
99 Solutions: 4661–4670

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
60/ Upcoming special issue in memory of Bruce Shawyer

Upcoming special issue in memory

of Bruce Shawyer
We were saddened by the news of Bruce Shawyer’s recent passing. Bruce con-
tributed in many ways to the Canadian mathematics community, including as a
coach for several Canadian teams at the International Mathematical Olympiad
(IMO), organizing the IMO held in Canada in 1995, and serving as the Editor-
in-Chief of Crux Mathematicorum with Mathematical Mayhem from 1996 to 2001.
To honour his memory, we will have a special issue in fall 2022.
This issue will be dedicated to Bruce Shawyer in November 2022. This is a call
for submissions. We encourage problem proposals, articles, expositions of Bruce’s
contributions to mathematics and math education, tributes to Bruce, and remi-
niscences.
Please circulate this call to others in the math community, particularly those who
may have known Bruce Shawyer. If you would like to contribute to the issue,
please send any materials to [email protected] by August 15th.

Crux Mathematicorum, Vol. 48(2), February 2022

 MathemAttic /61

MATHEMATTIC
No. 32
The problems 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 May 1, 2022.

MA156. In a kindergarten, 17 children made an even number of postcards.

Any group of 5 children made no more than 25 postcards while any group of 3
children made no less than 14 postcards. Determine the total number of postcards
made.

MA157. The four sides and one diagonal of a quadrilateral have lengths 1,
2, 2.8, 5 and 7.5, not necessarily in that order. Determine which number was the
length of the diagonal.

MA158. Proposed by Aravind Mahadevan.

A semi-circle is inscribed in ∆ABC such that it is tangent to AB and AC and
its diameter lies along the side BC. If AB = 13, AC = 14 and BC = 15,
find the radius of the semi-circle. (Solvers may find
p Heron’s formula for the area
of a triangle with sides a, b, and c useful: A = s(s − a)(s − b)(s − c), where
s = a+b+c
2 .)

MA159. A 5-by-5 square consists of 25 1-by-1 small squares. If one corner

square is removed, prove that it is not possible to cover the rest of the squares by
eight 3-by-1 rectangles as shown in the figure.

MA160. In a right triangle, the smallest height is one-quarter the length of

the hypotenuse. Determine the measure, in degrees, of the smallest angle of this
triangle.

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

Copyright © Canadian Mathematical Society, 2022

62/ MathemAttic

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 1 mai 2022.

MA156. Dans une certaine garderie, les 17 enfants ont produit un nombre
pair de cartes postales. Tout ensemble de 5 enfants a produit au plus 25 cartes
postales, tandis que tout ensemble de 3 enfants a produit au moins 14 cartes
postales. Déterminer le nombre total de cartes produites en cette garderie.

MA157. Les quatre côtés et une des diagonales d’un certain quadrilatère ont
les longueurs 1, 2, 2.8, 5 et 7.5, mais pas nécessairement dans cet ordre. Déterminer
lequel de ces nombres représente la diagonale de ce quadrilatère.

MA158. Proposé par Aravind Mahadevan.

Un demi cercle est inscrit dans ∆ABC de façon à ce qu’il soit tangent à AB et
AC, puis que son diamètre se situe sur le côté BC. Si AB = 13, AC = 14 et
BC = 15, déterminer le rayon de ce demi cercle. (Il sera possiblement utile de
faire intervenir
p la formule de Héron pour la surface d’un triangle de côtés a, b et
c : A = s(s − a)(s − b)(s − c), où s = a+b+c
2 .)

MA159. Un carré de taille 5 par 5 consiste de 25 petits carrés de taille 1 par

1. On lui enlève un carré de coin, tel qu’illustré. Démontrer qu’il est impossible
de paver le grand carré modifié, à l’aide de huit tuiles rectangulaires de taille 3
par 1.

MA160. Dans un certain triangle rectangle, la plus courte altitude est de

longueur égale au quart de la longueur de l’hypoténuse. Déterminer, en degrés, le
plus petit angle de ce triangle.

Crux Mathematicorum, Vol. 48(2), February 2022

 MathemAttic /63

MATHEMATTIC
SOLUTIONS
Statements of the problems in this section originally appear in 2021: 47(7), p. 326–328.

MA131. Prove that there are infinitely many positive integers k such that
k k can be expressed as the sum of the cubes of two positive integers.
Originally from 2009 Alberta High School Mathematics Competition, Part II, prob-
lem 5.
We received 11 submissions of which 10 were correct and complete. We present
the solution by Richard Hess, slightly edited.
Suppose that k = a3 + b3 for positive integers a and b and that k ≡ 1 (mod 3). In
this case, write k = 3m + 1 for m a positive integer, and note that

k k = k 3m · k = k 3m · (a3 + b3 ) = (ak m )3 + (bk m )3 ;

that is, k k is a sum of two cubes.

Any choice of a and b that satisfies either a ≡ 1 (mod 3) and b ≡ 0 (mod 3) or
a ≡ b ≡ 2 (mod 3) results in k ≡ 1 (mod 3), so there are infinitely many choices
of a and b that will satisfy the desired conditions.
For the case a ≡ 1 (mod 3) and b ≡ 0 (mod 3), we get the solutions k = 28, 91,
217, 280 and so on.
For the case a ≡ b ≡ 2 (mod 3), we get the solutions k = 16, 133, 250, 520 and so
on.

MA132. Proposed by Ed Barbeau.

Determine all sets consisting of an odd number 2m + 1 of consecutive positive
integers, for some integer m ≥ 1 such that the sum of the smallest m + 1 integers
is equal to the sum of the largest m integers.
We received 8 solutions. We present the solution by Digby Smith, lightly edited.
Given the positive integer m, let k be a positive integer such that

k + (k + 1) + · · · + (k + m) = (k + m + 1) + (k + m + 2) + · · · + (k + 2m).

Rearranging gives

(m + 1)k + (1 + 2 + · · · + m) = mk + (1 + 2 + · · · + m) + m2 ,

making k = m2 . Hence for each positive integer m there exists exactly one set
satisfying the property, namely {m2 , m2 + 1, . . . , m2 + 2m}.

Copyright © Canadian Mathematical Society, 2022

64/ MathemAttic

MA133. Proposed by Nguyen Viet Hung.

Find all pairs (x, y) of positive integers satisfying the equation

x2 − 2x + 29 = 7x y.

We received 9 solutions, of which we present the one by Mohammad Bakkar,

slightly expanded by the editor.
Notice that the right hand side is divisible by 7. Since the left hand side can be
rewritten as
x2 − 2x + 29 = (x − 1)2 + 28,
we see that x ≡ 1 (mod 7). Set x = 7k + 1 for some non-negative integer k. Then

(7k)2 + 28 = 7x · y, or equivalently 7k 2 + 4 = 7x−1 · y.

If x > 1, then the right hand side is divisible by 7, so we have to have x = 1. We

obtain x = 1, y = 4 as the only solution.

MA134. If the perimeter of an isosceles right-angled triangle is 8, what is its

area?
Originally from 2003 Manitoba Mathematical Contest, problem 3b.
We received 11 submissions, all of which were correct. We present the solution by
Bala Venkataraman, modified by the editor.
Let a be the √ side length of one of the√triangle’s legs. The hypotenuse therefore
has length 2a. By assumption, 2a + 2a = 8 ⇒ a = 2+8√2 . As the area A of an
2 √
isosceles right triangle is given by A = a2 , we have that A = 16(3 − 2 2).

MA135. 80 students responded to a survey about sports they played.

30 played basketball.
26 played rugby.
28 played hockey.
12 played basketball and rugby.
8 played hockey and rugby.
x played basketball and hockey only.
4 played all 3 sports.
Twice as many played none of the 3 sports as played basketball and hockey
only.
If a student is picked at random from the whole group, what is the probability
that the student plays only 1 of the 3 sports?

Crux Mathematicorum, Vol. 48(2), February 2022

 MathemAttic /65

Originally question 2 of The π Quiz 2017, Round 8 by the Irish Maths Teachers’
Association.

We received 5 submissions of which 3 were correct and complete. We present the

solution by Miguel Amengual Covas.

Let B, H, R represent all students of the group that played basketball, hockey,
rugby, respectively. According to the following Venn diagram

18 − x
8

x 10 R
4

20 − x

16 + x

(18 − x) + (20 − x) + 10 24 − x
the required probability is i.e., . We are told also
80 40
that 16 + x = 2x, which gives x = 16.
24 − 16
Thus, the required probability is = 0.2
40

MA136. Sent in by Ed Barbeau, from correspondence with Harold Reiter.

Solve the alphametic

SET A − AT ES = EAST

where S > E > T > A are digits in the 4-digit numbers.

We received 5 solutions. The following is by Huang Aaron.

From the equation

SET A − AT ES = (S − A)(E − T − 1)(T − E + 9)(A − S + 10) = EAST

Copyright © Canadian Mathematical Society, 2022

66/ MathemAttic

we obtain

S−A=E
E−T −1=A
T −E+9=S
A − S + 10 = T.

Adding the first and fourth equation gives

E + T = 10.

Adding the second and third equation gives

A + S = 8.

Observe that E ≥ 6 and S ≤ 7, we must have

S=7

and
E = 6.
Hence T = 4 and A = 1 and the problem is solved.

Crux Mathematicorum, Vol. 48(2), February 2022

 Andy Liu /67

From the bookshelf of . . .

Andy Liu
This new feature of MathemAttic brings attention to books of potential interest to
the readers. Some of these will be reviews whereas others will be hearty recommen-
dations from the contributors. If you have a book related to mathematics that would
be of interest to secondary school students and/or teachers, feel welcome to send
along a submission to [email protected]. Publishers are also welcome to
send along books for possible review.

A Mixed Bag
by Raymond Smullyan
ISBN 978-0-9861445-7-8, softcover, 144+ pages
Published by Sagging Meniscus in 2016.

I call Raymond Smullyan, who passed away in 2017 at age 98, a 3M logician (ex-
plained later). He himself regarded Kurt Gödel as the greatest of all logicians.
Indeed, Gödel’s Incompleteness Theorem in 1931 was a major result in metamath-
ematics that shook the foundation of the world of mathematics. Raymond had
devoted his life to disseminating Gödel’s idea to the general public in an ingenious,
illuminating and invigorating manner.
This is primarily done through a sequence of eleven books of logic puzzles. In
order of publication, they are:
[1] What is the Name of this Book?
[2] The Lady or the Tiger?
[3] Alice in Puzzleland.
[4] To Mock a Mocking Bird.
[5] Forever Undecided.
[6] Satan, Cantor, and Infinity.
[7] The Riddles of Scheherazade.
[8] Logical Labyrinth.
[9[ King Arthur in Search of his Dog.
[10] The Gödelian Puzzle Book.
[11] The Magic Garden of George B.
A Mixed Bag consists of largely a collection of personal reminiscences: how he got
married to the musician Blanche, how he got hired by the prestigious Dartmouth
College before he even had a bachelor’s degree, how he got that degree from the
University of Chicago on the strength of courses he never took but had taught,
how he met Kurt Gödel at the Institute of Advanced Studies, and so on. Many of

Copyright © Canadian Mathematical Society, 2022

68/ From the bookshelf of . . .

the accounts reflected on his additional lives as a mathematician, a magician and

a musician.
Humor was an essential component of Raymond’s character. This book is spiced
with many fantastic jokes. Here are some samples. They are chosen because they
are short. There are much better ones.
1. Teacher: If your father has ten dollars and you ask for six, how many will
he still have?
Kid: Ten.
Teacher: You don’t know your math.
Kid: You don’t know my father.
2. Teacher: There are three kinds of people — those who can count and those
who can’t.
Kid: Five out of four people can’t count.
3. Patient: My trouble is that I believe I am a dog.
Doctor: Since when?
Patient: Since I was a puppy.
4. Patient: My trouble is that I am losing my memory.
Doctor: Since when?
Patient: Since when what?
What is a book by Raymond without puzzles? There are plenty of them. They
may loosely be classified into two kinds. The first is what Raymond called monkey
tricks. Here are two of them.
Puzzle 1. A certain man had great grandchildren, yet none of his grandchildren
had any children! How is this possible?
Puzzle 2. A man was driving along a highway. His headlights were broken, there
were no street lights on and there was no moon out. There was a pedestrian
crossing the street about a hundred and fifty yards in from of him. The driver
knew that the pedestrian was there and stopped his car in time to avoid hitting
him. How did he know that the pedestrian was there?
The second is the more traditional logical-reasoning problems, even though Ray-
mond had a knack of making traditions unconventional. Here are two of them.
Puzzle 3. You have a line of people. The first one in line is married and the last
one is not. Prove that at least one married one is directly in front of an unmarried
one.
Puzzle 4. In a certain flower garden, each flower was either red, yellow or blue. All
three colors were actually represented. One statistician observed that whichever
three flowers were picked, at least one was bound to be yellow. Another observed
that whatever three flowers were picked, at least one was bound to be red. From
these two observations, does it logically follow that given any three of the flowers,
at least one is bound to be blue?

Crux Mathematicorum, Vol. 48(2), February 2022

 Andy Liu /69

The answers to these puzzles are given in the book, which also has many stories,
often with a philosophical bent. Here is one of them.
A monk came up the mountain to interview the Master, who asked him whether
he came from the North or the South. “The South,” was the reply. “In that case,”
said the Master, “have a cup of tea.” The next morning, another monk came
up the mountain for an interview, and the Master likewise asked him whether he
came from the North or the South. This time, the monk said he had come from
the North. “In that case,” said the Master, “have a cup of tea.” Later on, the
Master’s assistant said to him: “I don’t understand, Master; you told one that
since he was from the South, he should have a cup of tea, and the other, that since
he was from the North, he should have a cup of tea. How come?” The Master
replied: “Have a cup of tea.”
The book ended with another such story.
A certain great Sage in the East was reputed to be the wisest man in the world.
A philosopher heard about him and was anxious to meet him. It took him fifteen
years to find him, but when he finally did, he asked him: “What is the best
question that can be asked, and what is the best answer that can be given?” The
great Sage replied: “The best question that can be asked is the question you have
asked, and the best answer that can be given is the answer I am now giving.”

The review was provided by Andy Liu. Andy ran a Mathematical Circle for Ed-
monton upper elementary and junior high students from 1981 to 2012. He has
given lectures to students in six continents. He has been the vice-president of the
International Mathematics Tournament of the Towns since 1992. He had been in-
volved in various capacities in the International Mathematical Olympiad from 1981
to 2016. He regularly attended the International Puzzle Party and the Gathering
for Gardner from 1991 to 2018. Andy was involved with Crux Editorial Board as
a Book Review editor from 1990 to 1998. He has authored eighteen mathematics
books so far, and edited several others.

Copyright © Canadian Mathematical Society, 2022

70/ Teaching Problems

TEACHING PROBLEMS
No.15
Margo Kondratieva
Basic geometric configurations: Which one do you see?
.................................................................

Teaching Problems usually features problems that have been integrated into the
teaching experiences of the contributors. This issue is a little different in that
the experience with a problem in a math contest suggests its potential merit as
one to be incorporated into teaching. The example draws forth a range of meth-
ods of solution as well as identifying links to different known geometrical results.
Others involved in math contests may wish to share such examples from their ex-
periences in future issues. Contributions to Teaching Problems are welcomed via
[email protected].
.................................................................

Why do people suggest different solutions to a given problem? George Pólya in

his famous book, How to solve it?, proposed that in problem solving one starts
with understanding and analysing what is given and what is to be found. Once
a problem is understood the solver needs to design an action plan. For that, one
looks at relevant facts that could connect the known and unknown data. Speaking
about Euclidean geometry, some geometrical facts can be stored in our memory in
the form of figures that illustrate a required property, possibly indicating a reason
for it. I call such figures basic geometric configurations (BGCs). The choice of
your solution depends on the BGCs that come to mind in association with the
problem. Here is an illustration of how this works.
Consider a geometry problem from the 2021 Canadian Open Mathematics Chal-
lenge (part B, number 3). Note, we will use the notation [P QR] to denote the
area of triangle P QR.
Two right triangles ∆AXY and ∆BXY have a common hypotenuse
XY and side lengths (in units) |AX| = 5, |AY | = 10, and |BY | = 2.
Sides AY and BX intersect at P . Determine the area (in square units)
of ∆P XY .
A
|AY | = 10

5 B
P
2
[P XY ] =?
X Y

Figure 1: The problem.

Crux Mathematicorum, Vol. 48(2), February 2022

 Margo Kondratieva /71

The BGC “Pythagorean theorem” allows us to find one side √ in a right triangle if
two other sides are known. Thus, we find the length |XY | = 125 from the right
triangle XAY and then the length of |XB| = 11 from the right triangle XBY . In
order to use the theorem one needs to focus on the relevant part of the figure and
ignore the rest, as illustrated in Figure 2.

√ B
10 125 − 22 = 11
5
2
X √ √ Y X √ Y
52 + 102 = 125 125

Figure 2: Applications of the BGC “Pythagorean theorem” to the cases of right

triangles XAY and XBY .

Another well known BGC depicts the fact that the area of a triangle is equal to
“a half of base times height”. Drop the height from P on XY . Call the foot D.
Then we have

1 1 1
[P XY ] = |XY | · |P D| = |XP | · |BY | = |Y P | · |AX|
2 2 2

5 B
P
2

X D Y

Figure 3: The heights and bases in the triangle P XY .

√ √
Since we know |XY | = 125 = 5 5, |AX| = 5 and |BY | = 2, the problem then
reduces to finding the length of either |P D| or |XP | or |Y P |.

Below we consider six different solutions of the problem, each of which is related
to a more sophisticated BGC. We start with presenting three methods of finding
|P D|.

Method 1. Figure 4 depicts the following properties. In an acute triangle KLM ,

let the segments KK 0 , LL0 and M M 0 be the heights and P be the orthocentre.
Then we have:

Copyright © Canadian Mathematical Society, 2022

72/ Teaching Problems

L0
M0
P

M L
K0

Figure 4a: BGCs depicting properties of the heights in acute and right triangles.

BGC 1: The ratio of two heights in an acute triangle is equal to the ratio of the
lengths from the feet of these heights to the third vertex of the triangle:

|KK 0 | |M K 0 |
0
= . (1)
|LL | |M L0 |

Proof: As seen in Figure 4a,

|KK 0 | |LL0 |
tan(∠L0 M K 0 ) = =
|M K 0 | |M L0 |

which is equivalent to (1).

BGC 2: The product of a height and the distance from the orthocentre to the
foot of this height is equal to the product of the distances from that foot to the
two other vertices of the triangle:

|KK 0 | · |P K 0 | = |M K 0 | · |LK 0 |. (2)

0 0
|KK | |LK |
Proof: Relation (2) is equivalent to |M K 0 | = |P K 0 | . The latter is true because
∠P KL = ∠P LK , as indicated in Figure 4a, and so cot(∠P KL0 ) = cot(∠P LK 0 ).
0 0

M L
K0

Figure 4b: BGCs depicting properties of the heights in right triangles.

In the special case when ∠M KL = 90◦ , points P , L0 , and M 0 coincide with K.

BGC1 then becomes
|KK 0 | |M K 0 |
= , (10 )
|LK| |M K|
which is true due to the fact that ∠LKK 0 = ∠KM K 0 , as indicated in Figure 4b.
Then BGC 2 reduces to the following fact.

Crux Mathematicorum, Vol. 48(2), February 2022

 Margo Kondratieva /73

BGC 20 : If ∠M KL = 90◦ , we have, from Figure 4b:

|KK 0 |2 = |M K 0 | · |LK 0 |. (20 )

Remark: This formula could be alternatively derived by noticing that M KK 0

K0| |KK 0 |
and KLK 0 are similar right triangles, and so |M
|KK 0 | = |LK 0 | .

The facts (1) and (2) become useful for solving our problem if we identify the
points M, L, L0 , M 0 with the points X, Y, A, B respectively. By extending XA
and Y B we obtain the point of intersection K. Point K 0 = D is the foot of the
perpendicular dropped from P on XY .
K

|AY | = 10

|BX| = 11
A
B
P
5
2
X Y
D

Figure 5: Setting up our problem for the application of BGC 1 and BGC 2.

These BGCs lead to the following solution of our problem.

Solution 1: Applying BGC 1 to Figure 5, we get:

|KD| |XD| |KD| |XD|

= ⇒ =
|Y A| |XA| 10 5

and hence
1
|XD| = |KD|.
2
Similarly,
|KD| |Y D| 2
= ⇒ |Y D| = |KD|.
|XB| |Y B| 11
Then,
1 2 15
Å ã
|XD| + |Y D| = + |KD| = |KD|.
2 11 22

On the other hand, √

|XD| + |Y D| = |XY | = 5 5,
15
√ √ √ √
so, 22 |KD| = 5 5. Therefore, |KD| = 22
3 5, |XD| = 11
3 5 and |Y D| = 43 5.

Copyright © Canadian Mathematical Society, 2022

74/ Teaching Problems

√
|XD|·|Y D| 2 5
Now from BGC 2, |P D| = |KD| = 3 . Therefore

1 25
[XP Y ] = |XY | · |P D| = .
2 3

Method 2. The following fact relates the vertical segments depicted in Figure 6.

S
P

X D Y

Figure 6: BGC 3 “two right triangles with common leg”

BGC 3: Let RXY and SXY be two right triangles for which leg XY is shared
and the hypotenuses intersect at P . Let P D be perpendicular to XY . Then

1 1 1
= + . (3)
|P D| |RX| |SY |

|Y D| |XY |
Proof: Since P DY and RXY are similar right triangles, we have |P D| = |RX| .
|XD| |XY |
Since P DX and SY X are similar right triangles, we have |P D| = |SY | . Therefore,

|Y D| |XD| |XY | |XY |

+ = + .
|P D| |P D| |RX| |SY |

On the other hand,

|Y D| |XD| |Y D| + |DX| |XY |

+ = = .
|P D| |P D| |P D| |P D|

Thus, we obtain
|XY | |XY | |XY |
+ = .
|RX| |SY | |P D|
Dividing through by |XY | gives the required relation.
This BGC leads to the following solution of our problem.
Solution 2: Extend Y A to intersect at R with the line perpendicular to XY
through X, as shown. Likewise let S be the intersection of XB (extended beyond
B) with the line perpendicular to XY through Y in figure 7.

Crux Mathematicorum, Vol. 48(2), February 2022

 Margo Kondratieva /75

R
|AY | = 10
A
|BX| = 11
5 B S
P
2
X D Y

Figure 7: Application of BGC 3 in Solution 2.

In the right triangle RXY , the segment XA is the height dropped to the hy-
potenuse RY . From BGC 20 we have the relation |XA|2 = |RA| · |AY |. Thus,
|RA| = 52 /10 = 5/2. Applying the Pythagorean theorem to right triangle XAR
yields … √
25 5 5 25
|RX| = 25 + = = √ .
4 2 2 5
BGC 20 applied to the right triangle SXY gives |Y 2
»B| = |SB| · √
|BX|, so |SB| =
16
4/11. Then from the right triangle Y BS, |SY | = 4 + 121 = 11 = 1150
10 5 √ .
5

Then, from BGC 3, √ √ √

1 2 5 11 5 3 5
= + =
|P D| 25 50 10
√ √
10 5 2 5
so |P D| = 15 = 3 .

Method 3. This approach is based on a BGC depicting two right triangles that
share an acute angle in the following way.
K

M L
D

∆M P D ∼ ∆M LK

Figure 8: Similar right triangles BGC 4.

BGC 4: Drop a perpendicular P D from any point P on a leg of a right triangle

KLM to its hypotenuse LM . Then the two right triangles KLM and DP M are
similar.
BGC 4 leads to the following solution.

Copyright © Canadian Mathematical Society, 2022

76/ Teaching Problems

Solution 3: Let |P D| = h.

A
|AY | = 10
|BX| = 11
5 B
P
h 2

X 11
2 h
D 2h Y
√
|XY | = 5 5

Figure 9: Application of BGC 4 in Solution 3.

Triangles XP D and XY B are similar by BGC 4. Thus,

|XD| |P D| |XD| h
= ⇒ =
|XB| |Y B| 11 2

11
and hence |XD| = 2 h. Similarly, triangles Y P D and Y XA are similar by BGC
4 which yields,
|Y D| |P D| |Y D| h
= ⇒ =
|Y A| |XA| 10 5

and therefore |Y D| = 2h.

Then
11 15
|XD| + |Y D| = h + 2h = h.
2 2
On the other hand,
√
|XD| + |Y D| = |XY | = 5 5,

therefore,
√
2 √ 2 5
h= 5 5= = |P D|.
15 3

One may identify a geometric configuration involving another pair of similar right
triangles.

BGC 5: Given two lines intersecting at P , drop a perpendicular XA from any

point X of one line on another one and a perpendicular Y B from any point Y
of the second line on the first one. Then the right triangles XAP and Y BP are
similar as shown in Figure 10. This is because ∠AP X = ∠BP Y as vertical angles
and ∠AXP = ∠BY P as complementary to them.

Crux Mathematicorum, Vol. 48(2), February 2022

 Margo Kondratieva /77

B
P

X Y

∆XAP ∼ ∆Y BP

Figure 10: Similar right triangles BGC 5.

Recognition of this BGC may yet lead to different solutions that use distinct
properties and relations in similar triangles. We present three of them below.

Solution 4: Triangles XAP and Y BP are similar by BGC 5. Then, since the ratio
5 2
of the sides is |AX| 5 [XAP ]

|BY | = 2 we can deduce that the ratio of the areas is [Y BP ] = 2 .
Thus, [XAP ] = 25 4 [Y BP ].

Labelling by x, y the area of the triangles P XY and Y BP respectively, as shown

in Figure 11, we get the following expressions:

25
[XY A] = [P XY ] + [XAP ] = x + y
4
[XY B] = [P XY ] + [Y BP ] = x + y.

On the other hand, [XY A] = 21 |XA| · |Y A| = 25 and [XY B] = 21 |XB| · |Y B| = 11.

5 25 B
4 y P
y 2
x
X Y

Figure 11: Application of BGC 5 in Solution 4

Thus, we obtain the system of equations

25
x+ y = 25
4
x + y = 11

25
Solving the system we get x = 3 and y = 83 . Thus, the desired area is 25
3 square
units.

Copyright © Canadian Mathematical Society, 2022

78/ Teaching Problems

Solution 5:
A |AY | = 10
|BX| = 11
5u
5 2u B
P
5v 2v 2

X Y

Figure 12: Application of BGC 5 in Solution 5

|AP | |XP | 5
Triangles XAP and Y BP are similar by BGC 5. We have |P B| = |P Y | = 2 , so
|AP | = 5u, |P B| = 2u, |XP | = 5v, |P Y | = 2v, for some numbers u and v. Thus,
|BX| = 2u + 5v = 11
|AY | = 5u + 2v = 10
4
Solving this system we obtain u = 3 and v = 35 . Thus, |XP | = 25
3 .

Therefore, the desired area is 21 |XP | · |BY | = 25

3 square units.

Solution 6: Let |AP | = x. Then |Y P | = 10 − x.

A

x
2x
5 5 B
P
2
10 − x
X Y

Figure 13: Application of BGC 5 in Solution 6

Triangles XAP and Y BP are similar by BGC 5. We have

|P B| 2 2 4x2
= ⇒ |P B| = x ⇒ |P B|2 = .
|AP | 5 5 25
On the other hand, from the right triangle P BY ,
|P B|2 = |P Y |2 − |Y B|2 = (10 − x)2 − 4 = 96 − 20x + x2 .
4x2
Thus, 25 = 96 − 20x + x2 and so

21x2 − 500x + 2400 = 0.

Solving this equation we obtain x1 = 20 120
3 and x2 = 7 . We take the first root
x = 3 because of the restriction |AP | = x < 10 = |AY |. Then |Y P | = 10− 20
20 10
3 = 3
1 25
and the area [P XY ] = 2 |Y P | · |AX| = 3 square units.

Crux Mathematicorum, Vol. 48(2), February 2022

 Margo Kondratieva /79

Closing comments
Considering the range of solution methods is there one that spoke to you as being
most appealing? Was it an aspect of the presentation or the geometrical ideas in
play or some other element that resonated with your problem-solving style?
Note that solutions 1, 2, and 3 require solving a linear equation, while solutions 4
and 5 require solving a system of linear equations and solution 6 requires solving
a quadratic equation. Solution 3 is probably the most “economical” which gives
an advantage in a competition setting, especially if calculators are not allowed.
However, each solution highlights an interesting point of view and is valuable for
making mathematical connections. Do you have yet another way of solving this
problem?

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

Margo Kondratieva is an Associate Professor of Mathemat-

ics and Mathematics Education at Memorial University,
where she studies various aspects of mathematical teach-
ing and learning. She is involved in organization of mathe-
matical contests at the provincial and national levels. She
enjoys traveling the world, walking nature trails and she
relishes occasional blooms from her garden in St. John’s.

Copyright © Canadian Mathematical Society, 2022

80/ OLYMPIAD CORNER

OLYMPIAD CORNER
No. 400
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 May 1, 2022.

OC566. Prove that if a and b are real numbers such that a + b > 2, then

(a − 1)x + b < x2 < ax + (b − 1)

for infinitely many real numbers x.

OC567. In a group of people, there are some mutually friendly pairs. For
a positive integer k ≥ 3, we say that the group is k-good if every k people in the
group can be seated around a round table so that every two neighbors are mutually
friends. Prove that if the group is 6-good, then it is also 7-good.

OC568. Point K is marked inside a parallelogram ABCD. Point M is

the midpoint of BC, point P is the midpoint of KM . Prove that if ∠AP B =
∠CP D = 90◦ , then AK = DK.

OC569. Let ABC be a triangle with ∠A = 80◦ and ∠C = 30◦ . Let M be

an internal point to triangle ABC such that ∠M AC = 60◦ and ∠M CA = 20◦ .
If N is the intersection point of lines BM and AC, prove that M N is the angle
bisector of ∠AM C.

OC570. Let n be a positive integer. Assumejthat in the set {1, 2, . . . , n}

nk
there are exactly M squarefree integers k such that is odd. Prove that M is
k
odd.

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

Crux Mathematicorum, Vol. 48(2), February 2022

 OLYMPIAD CORNER /81

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 1 mai 2022.

OC566. Démontrer que si a et b sont des nombres réels tels que a + b > 2,
alors
(a − 1)x + b < x2 < ax + (b − 1)
pour un nombre infini de réels x.

OC567. Dans tout ensemble de personnes, il existe des amitiés. Pour un

entier k ≥ 3, on dit qu’un ensemble de personnes est k-amical si tout sous ensemble
de k personnes peut siéger autour d’une table ronde de façon à ce chacun soit ami
avec ses deux voisins. Démontrer que tout ensemble 6-amical est aussi 7-amical.

OC568. Le point K se trouve à l’intérieur d’un certain parallélogramme

ABCD; M est le point milieu de BC, tandis que P est le point milieu de KM .
Démontrer que si ∠AP B = ∠CP D = 90◦ , alors AK = DK.

OC569. Soit ABC un triangle tel que ∠A = 80◦ et ∠C = 30◦ ; soit aussi M
un point à l’intérieur du triangle, tel que ∠M AC = 60◦ et ∠M CA = 20◦ . Si N est
le point d’intersection des lignes BM et AC, démontrer que M N est bissectrice
de ∠AM C.

OC570. Soit n un entier positif et supposons que l’ensemble

jnk {1, 2, . . . , n} con-
tient exactement M entiers k sans facteur carré tels que est impair. Démontrer
k
que M est impair.

Copyright © Canadian Mathematical Society, 2022

82/ OLYMPIAD CORNER

OLYMPIAD CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2021: 47(7), p. 338–339.

OC541. In a convex quadrilateral ABCD, suppose ∠ABC = ∠ACD and

∠ACB = ∠ADC. Assume that the center O of the circle circumscribed to the
triangle BCD is different from point A. Prove that triangle OAC is a right
triangle.
Originally from 2018 Czech-Slovakia Math Olympiad, 5th Problem, Category A,
Final Round.
We received 9 correct solutions. We present two solutions.
Solution 1, by UCLan Cyprus Problem Solving Group.
Let ∠ABC = ∠ACD = ϑ and ∠ACB = ∠ADC = ϕ. Then

∠DAC = ∠CAB = 180◦ − ϑ − ϕ.

So
∠DAB = 360◦ − 2(ϑ + ϕ) = 2(180◦ − ∠BCD) = ∠DOB .
So A, O, B, D belong on the same circle, say ω. Let A0 be the other point of
intersection of AC with ω. Since AC is the angle bisector of ∠DAB, then A0
belongs on the perpendicular bisector of BD. But so does O, as OB = OD. Thus
OA0 is a diameter of ω and so OA ⊥ AC as required.

Crux Mathematicorum, Vol. 48(2), February 2022

 OLYMPIAD CORNER /83

Solution 2, by Michel Bataille.

We embed the problem in the complex plane with origin at A. Let b, c, d be the
affixes of B, C, D respectively, and let U , with affix u, be the point of intersection
of the perpendicular bisector λ of BC and the line µ perpendicular to AC through
A. We answer the problem by showing that U is also on the perpendicular bisector
of CD (so that O = U and OA ⊥ AC).
The equation of µ is zc + zc = 0, that is, zc + zc = 0. The equation of λ is
|z − b|2 = |z − c|2 , that is, (zb + zb) − (zc + zc) = |b|2 − |c|2 . Solving the system
of these two equations, an easy calculation gives

−c(|b|2 − |c|2 )
u= .
bc − bc

(We show below that bc − bc 6= 0).

From the hypotheses about the quadrilateral ABCD, a spiral similarity with centre
A transforms B into C and C into D, hence, c = αb and d = αc for some nonzero
complex number α. First, it follows that bc − bc = |b|2 (α − α) 6= 0 since b 6= 0
(because B 6= A) and α 6= α (otherwise α is a real number and A, B, C would be
collinear).
c(|α|2 −1)
Second, we can write u as u = α−α .

Finally, we have to check that |u − c|2 = |u − αc|2 , that is

−(uc + uc) + |c|2 = −(αuc + αuc) + |α|2 |c|2 .

Now, recalling that uc + uc = 0, we obtain

c(|α|2 − 1) c(|α|2 − 1)
αuc + αuc = αc + αc = |c|2 (|α|2 − 1)
α−α α−α
hence |u − c|2 = |u − αc|2 holds and we are done.

OC542. Let x1 , x2 , . . . , xn be positive integers. Assume that in their decimal

representations no xi “is an extension” of another xj . For instance, 123 is an
extension of 12, 459 is an extension of 4, but 134 is not an extension of 123. Prove
that
1 1 1
+ + ··· + < 3.
x1 x2 xn

Originally from 2018 Italy Math Olympiad, 3rd Problem, Final Round.
We received 6 correct solutions. We present two solutions.
Solution 1, by Oliver Geupel.
Let us say that the positive integer x reduces to the positive integer y, if y =
bx/10c. We write x → y if x reduces to y. Observe that the positive integer x ”is

Copyright © Canadian Mathematical Society, 2022

84/ OLYMPIAD CORNER

an extension” of the positive integer y if and only if there is a chain x → · · · → y of

reductions from x to y. We say that the finite set M of positive integers reduces to
the set N of positive integers (and write M → N ) if N is obtained by the following
construction: Put the greatest element a1 of M and reduce it to, say, the positive
integer b. Include b, and remove all elements a1 , a2 , . . . , am that reduce to b (i.e.
a1 → b, a2 → b, . . . , am → b), to obtain

N = (M ∪ {b}) r {a1 , a2 , . . . , am }. (1)

We say that a finite set M of positive integers is good if no element of M is an

extension of another element of M . Note that if M is good and M → N , then
N is also a good set. Call a good set M of positive integers irreducible if it does
not reduce to any other set. It is clear that the irreducible sets are Pthe subsets of
{1, 2, . . . , 9}. For a finite set M of positive integers, let f (M ) = x∈M 1/x. For
every irreducible set M , it holds
9
1 1 1 1 1 1 1 1 1
X Å ã Å ã Å ã
f (M ) ≤ =1+ + + + + + + +
k 2 3 6 4 5 7 8 9
k=1
1 3
<1+1+ + < 3.
2 7
If M is good and M → N , then we have with the notation (1):
1 1 1
f (M ) = f (N ) + + ··· + −
a1 am b
1 1 1 1
< f (N ) + + + ··· + − < f (N ).
10b 10b + 1 10b + 9 b

The set M1 = {x1 , x2 , . . . , xn } reduces in a finite chain M1 → M2 → · · · → Mk to

an irreducible set Mk , where f (M1 ) < f (M2 ) < · · · < f (Mk ) < 3.

Solution 2, by UCLan Cyprus Problem Solving Group.

Fix a digit k and suppose that for each m, there are dm of the xi ’s which begin
with k and have m digits. We may assume that the largest xi which begins with
k (if there is one) has N digits. The condition of the problem guarantees that
for each N -digit number M beginning with k there is at most one xi which is a
prefix of M . Since an m-digit number is a prefix of exactly 10N −m numbers with
M -digits, we get that

d1 · 10N −1 + d2 · 10N −2 + · · · + dN ≤ 10N −1

or equivalently
d2 dN
d1 + + · · · + N −1 ≤ 1 .
10 10
This is actually an immediate application of Kraft’s inequality but we chose to
spell out the proof as it is not that known in Olympiad circles.

Crux Mathematicorum, Vol. 48(2), February 2022

 OLYMPIAD CORNER /85

Now we observe that each m-digit number beginning with k is at least 10m k and
so contributes at most 101m k in the required sum. So the total contribution of the
xi ’s beginning with k is at most

d1 d2 dN 1
+ + · · · + N −1 ≤
k 10k 10 k k
Therefore
1 1 1 1 1 1 1 1 1 1 2 3
+···+ < 1+ + +···+ = 2+ + + + + < 2+ + < 3.
x1 xn 2 3 9 4 5 7 8 9 4 7

OC543. There are 50 cards in a box with the first 100 positive integers
written on them. That is, the first card has number 1 on one side and number 2
on the other side, the second card has number 3 on one side and number 4 on the
other, and so on up to the 50-th card which has number 99 on one side and 100
on the other side. Eliza takes four cards out of the box and calculates the sum of
the eight numbers written on them. How many distinct sums can Eliza get?
Originally from 2018 Romania Math Olympiad, 4th Problem, Grade 5, District
Round.
We received 5 submissions, of which 4 were correct and complete. We present the
solution by Oliver Geupel.
We show that the answer is 185.
For 1 ≤ n ≤ 50, the sum of the two numbers written on the card with number n
is (2n − 1) + 2n = 4n − 1. Hence, the sum of the eight numbers written on four
distinct cards with numbers i, j, k, and ` is 4(i + j + k + ` − 1). We have

10 = 1 + 2 + 3 + 4 ≤ i + j + k + ` ≤ 47 + 48 + 49 + 50 = 194.

Thus, Eliza cannot get more than 194 − 9 = 185 distinct sums.
It remains to show that the sum s = i + j + k + ` can attain every integer value
in the range from 10 to 194.
If s = 4n where 3 ≤ n ≤ 48, then we put {i, j, k, `} = {n − 2, n − 1, n + 1, n + 2}.
If s = 4n + 1 where 3 ≤ n ≤ 48, we take {i, j, k, `} = {n − 2, n, n + 1, n + 2}.
If s = 4n + 2 where 2 ≤ n ≤ 48}, put {i, j, k, `} = {n − 1, n, n + 1, n + 2}.
Finally, if s = 4n + 3 where 2 ≤ n ≤ 47, then {i, j, k, `} = {n − 1, n, n + 1, n + 3}
does the job.

OC544. Prove that if n ≥ 2 is an integer, then there exist invertible matrices

A1 , A2 , . . . , An ∈ M2 (R) with nonzero entries such that

A−1 −1 −1
1 + A2 + · · · + An = (A1 + A2 + · · · + An )
−1
.

Copyright © Canadian Mathematical Society, 2022

86/ OLYMPIAD CORNER

Originally 2018 Romania Math Olympiad, 1st Problem, Grade 11, District Round.
We received 6 correct solutions. We present 2 solutions.
Solution 1, by Michel Bataille.
Å ã
ba
Let S be the set of all matrices such that a, b, c, d, a + b, c + d, ad − bc are
d c
Å ã
0 −1
nonzero real numbers and let C = .
1 −1
Å ã Å ã
−1 1 1 0
Note that C −1 = = −C − I2 where I2 = is the unit matrix.
−1 0 0 1
Pick any A1 in S. It is readily checked that A2 = A1 C and

A1 + A2 = A1 (I2 + C) = −A1 C −1

are in S. Moreover, we have

(A1 + A2 )−1 = −CA−1

1

and
A−1 −1 −1
1 + A2 = A1 + C
−1 −1
A1 = (I2 + C −1 )A−1 −1
1 = −CA1 .

Thus, A1 , A2 answer the problem for n = 2.

We continue the proof by induction: assume that for some integer n ≥ 2 we have
A1 , A2 , . . . , An in S such that A1 + A2 + · · · + An ∈ S and

A−1 −1 −1
1 + A2 + · · · + An = (A1 + A2 + · · · + An )
−1
.

Consider An+1 = (A1 + A2 + · · · + An )C. From the case n = 2 above, An+1 ∈ S

and

A1 + A2 + · · · + An + An+1 = (A1 + A2 + · · · + An ) + (A1 + A2 + · · · + An )C ∈ S.

In addition, again from the case n = 2, we have

(A1 + A2 + · · · + An + An+1 )−1 = (A1 + A2 + · · · + An )−1 + A−1

n+1
= A−1 −1 −1 −1
1 + A2 + · · · + An + An+1 .

This completes the induction step and the answer to the problem.

Solution 2, by Corneliu-Avram Manescu.

The given equality is equivalent to

(A1 + A2 + . . . + An )(A−1 −1 −1
1 + A2 + . . . + An ) = I 2 .

Let A1 = A, A2 = . . . = An = B. The condition becomes

(A + (n − 1)B)(A−1 + (n − 1)B −1 ) = I2 .

Crux Mathematicorum, Vol. 48(2), February 2022

 OLYMPIAD CORNER /87

Denote X = BA−1 and multiply the two parentheses. It follows that

I2 + (n − 1)(X + X −1 ) + (n − 1)2 I2 = I2 ,

that is,
X + X −1 + (n − 1)I2 = O2 ,
i.e.
X 2 + (n − 1)X + I2 = O2 .
The matrix Å ã
1 n+1
X=
−1 −n
satisfies this equation because tr(X) = −(n − 1) and det(X) = 1 by the Cayley-
Hamilton theorem. Now, choose
Å ã
2 1
A=
1 1

and then
Å ãÅ ã Å ã
1 n+1 2 1 n+3 n+2
B = XA = = .
−1 −n 1 1 −n − 2 −n − 1

OC545. Solve in real numbers the system of equations

 2
 x y+2 = x + 2yz
y2 z + 2 = y + 2zx
 2
z x+2 = z + 2xy

Originally from 2018 Poland Math Olympiad, 3rd Problem, First Round.
We received 11 submissions of which 10 were correct and complete. We present
the solution by Oliver Geupel.
A straightforward check shows that (−1, −1, −1), (1, 1, 1), and (2, 2, 2) are solu-
tions for (x, y, z). We show that there are no other solutions. Let us refer to the
given equations as to (1), (2), and (3).
Rearrange the terms to get

x(xy − 1) = 2(yz − 1), y(yz − 1) = 2(zx − 1), z(zx − 1) = 2(xy − 1).

Multiplying the terms on either sides, we arrive at

xyz(xy − 1)(yz − 1)(zx − 1) = 8(xy − 1)(yz − 1)(zx − 1).

This means that either xy = 1, or yz = 1, or zx = 1, or xyz = 8. We consider the

cases in succession.

Copyright © Canadian Mathematical Society, 2022

88/ OLYMPIAD CORNER

First, suppose that xy = 1. By (1), we obtain x + 2 = x + 2yz and hence yz = 1.

Similarly, zx = 1. Thus, z 2 = yz · zx/(xy) = 1, i.e. z = −1 or z = 1. If z = −1,
then we obtain x = zx/z = −1 and y = xy/x = −1. On the other hand, if z = 1,
then we arrive at x = zx/z = 1 and y = xy/x = 1. The cases yz = 1 and zx = 1
are each similar by the cyclic structure of the given equations.
It remains to consider the case when xyz = 8. If two of the numbers x, y, and z are
negative, say, x < 0 and y < 0, then we get y 2 z + 2 > 0 > y + 2zx, a contradiction
to (2). Therefore, x, y, and z are positive. Make the substitution (a, b, c) =
(2/x, 2/y, 2/z). Then, a, b, and c are positive real numbers such that abc = 1. By
the AM-GM inequality, we have 3a2 + a ≥ 4a7/4 with similar inequalities in b and
c, respectively. We rewrite the given system of equations as

 4c + a = 1 + 4a2
4a + b = 1 + 4b2
4b + c = 1 + 4c2 .


It follows that

4c = 1 + 4a2 − a = (1 − a)2 + 3a2 + a ≥ (1 − a)2 + 4a7/4 .

Analogously, 4a ≥ (1 − b)2 + 4b7/4 and 4b ≥ (1 − c)2 + 4c7/4 . We prove by

contradiction that a = b = c = 1. Assuming the contrary, we obtain

64 = 4a · 4b · 4c > 4a7/4 · 4b7/4 · 4c7/4 = 64,

which is absurd. This shows that a = b = c = 1, i.e. x = y = z = 2.

Crux Mathematicorum, Vol. 48(2), February 2022

 Sam Hopkins /89

The Last Problem: Demystified

Sam Hopkins
The June 2021 issue of Crux Mathematicorum included, on page 287, an intriguing
problem entitled simply “The Last Problem.” Here it is:

A m × n rectangular array is made up of the positive integers 1, 2,

3, . . . , mn arranged in such a way that each row and each column is
monotonically decreasing. In particular, mn must appear in the upper
left corner and 1 in the lower right corner. An operator on the array is
as follows. The number in the lower right corner is circled. Once any
number is circled, the smaller of two of its neighbours, one immediately
to the left in the same row and the other immediately above in the
same column, is also circled. If there is only one such number, it is
circled. In this way, a track of m + n − 1 circled numbers from the
lower right to the upper left is obtained. Now the number in the lower
right is transferred to the upper left position and the rest of the circled
numbers are displaced one position along the track. The uncircled
numbers are not moved. The same operation is then repeated, with
the understanding that, once any number k is transferred from the
lower right position to the upper left position, it is treated as though
its magnitude were mn + k.
Prove or disprove:
(a) After mn operations, each number in the array is restored to its
initial position;
(b) If i moves down on the jth move, then j moves down on the ith
move;
(c) If i moves right on the jth move, then j moves right on the ith
move.

Here we aim to demystify this problem, and, if not exactly explain its solution, at
least situate it in its proper mathematical context.
A partition of a positive integer n is a sequence λ = (λ1 , λ2 , . . . , λk ) of integers
satisfying λ1 ≥ λ2 ≥ · · · ≥ λk > 0 and λ1 + λ2 + · · · + λk = n. For example,
λ = (4, 2, 2, 1) is a partition of 9. Associated to any partition is its Young diagram:
the left- and top-justified array of boxes which has λi boxes in the ith row. The
Young diagram of (4, 2, 2, 1) is

Copyright © Canadian Mathematical Society, 2022

90/ The Last Problem: Demystified

A standard Young tableau (SYT) of shape λ is a filling of the Young diagram of λ

with the numbers 1, 2, . . . , n so that these numbers are strictly increasing along
rows and down columns. An SYT of shape (4, 2, 2, 1) is

1 3 5 8
2 4
6 9
7

Young diagrams and tableaux are named after Alfred Young (1873–1940), a British
mathematician who pioneered the study of the group of permutations of a finite
set. Much is known about SYTs because of their connection to algebra. For
instance, there is a beautiful formula for the number of SYTs of given shape. The
hook of a box u in a Young diagram consists of all boxes directly below, or directly
to the right of, that box, including the box itself. The hook length of a box is the
number of boxes in its hook. For example, the box with entry 2 in the above SYT
has a hook length of 4. The celebrated hook length formula says that the number
of SYTs of shape λ, a partition of n, is
n!
Y ,
h(u)
u

where the product runs over all boxes u of the Young diagram of λ, and h(u) is the
hook length of the box u. So for example there are 9!/(7 · 5 · 4 · 3 · 2 · 2 · 1 · 1 · 1) = 216
SYTs of shape (4, 2, 2, 1). To learn more about tableaux in general, see Yong’s
short note [3] or Sagan’s survey [1].
Our present interest in SYTs lies not in their enumeration but rather in a certain
operation on them, which we now explain using a somewhat fanciful analogy.
We can view a Young diagram as a building whose rooms are the boxes of the
diagram. (This presents some engineering challenges because the rows get longer
towards the top, but never mind that.) If this building belongs to a hierarchical
organization, like a company, then we can view a filling of the Young diagram with
numbers 1, 2 . . ., as an assignment of rooms to the person of rank 1, the person of
rank 2, and so on. Suppose that to the left of our Young diagram office building
is a beautiful ocean. Naturally, everyone in the building wants to have the best
view of this ocean, and hence would always prefer to have a room as much to the
left (to be closer to the ocean) and above (to have a higher viewpoint) as possible.
So, in order to respect the pecking order, we might require room assignments to
be such that every person has a lesser rank than the people in the rooms to their
left and above them. Room assignments like this are precisely SYTs.
But now suppose that the CEO (the person of rank 1) leaves the company for
a better opportunity elsewhere. Their departure creates an opening in a very
desirable room. Those of lesser rank will fill this opening. Since the company does

Crux Mathematicorum, Vol. 48(2), February 2022

 Sam Hopkins /91

not want people moving their stuff a long way across the building, only the people
whose rooms are adjacent, either to the right or below, can compete to fill that
open room. Of course, among these two, the room is awarded to the person of
greater rank. They move from their current room to the more desirable one, and in
doing so they create a new room opening, which is filled in the same manner: with
the people currently adjacent to the right and below competing. In this way the
departure of the CEO causes a series of room re-assignments, which eventually
terminates with an undesirable room at the bottom-right of the building being
emptied. Then, two final things happen to complete the corporate restructuring.
First, everyone in the building gets a “promotion,” meaning that the person of
rank 2 becomes rank 1, the person of rank 3 becomes rank 2, and so on. And
second, a new intern, of the least rank n, gets hired to fill the empty room.
Here is an example of this procedure:

1 3 5 8 • 3 5 8 2 3 5 8 2 3 5 8
2 4 1 leaves 2 4 2 upgrades • 4 4 upgrades 4 •
−−−−−→ −−−−−−−→ −−−−−−−→ ···
6 9 6 9 6 9 6 9
7 7 7 7

2 3 5 8 1 2 4 7 1 2 4 7
9 upgrades 4 9 promotions 3 8 n is hired 3 8
· · · −−−−−−−→ −−−−−−−→ −−−−−−→
6 • 5 • 5 9
7 6 6

This entire operation, which takes one SYT of shape λ to another one, is in fact
called promotion. The promotion operation on tableaux was introduced, together
with another closely related operation called evacuation, by M.P. Schützenberger1 .
The sliding process which goes into the definition of both promotion and evacuation
was termed jeu de taquin by Schützenberger. “Jeu de taquin” literally translates
to “teasing game,” but is the name in French for what is usually called the “15
Puzzle” in English. For an excellent introduction to promotion and evacuation,
see Stanley’s survey [2].
Promotion is an invertible operation. To see this, we can imagine doing all of
the steps backwards: firing the intern, demoting everyone, and forcing them into
1 The French mathematician Marcel-Paul Schützenberger (1920–1996) had a wide range of

scientific interests: e.g., he obtained a doctorate in medicine in 1948; and in the 1960s he worked
with the famous linguist Noam Chomsky on the analysis of formal languages. In algebraic
combinatorics he is especially remembered for seminal contributions to the theory of tableaux,
symmetric functions, and Schubert calculus.

Copyright © Canadian Mathematical Society, 2022

92/ The Last Problem: Demystified

worse rooms until there is a spot at the top for a new CEO. Hence, for any given
tableau T there must be some number of times we can apply promotion to T that
will get us back to T . But for most shapes, promotion behaves quite chaotically
and it takes a long time for us to get back to where we started. For instance, we
would need to apply promotion 60 times to our running example SYT of shape
(4, 4, 2, 1) in order to return to it. (In contrast, evacuation is always an involution,
meaning if we apply it twice we get back to where we started.)
There are a very small number of partition shapes for which promotion behaves
in an orderly fashion: see [2, §4]. These nice shapes include the rectangle 2
a times
z }| {
a × b := (b, b, . . . , b).

For any SYT of rectangular shape a × b, if we apply promotion ab times we get

back to where we started; this is Theorem 4.1(a) in [2]. Note that we might get
back to where we started even before ab applications of promotion. For example,
for the following SYT of shape 2 × 3, we can apply promotion 3 times to return
to it:

1 2 3 Promotion 1 2 5 Promotion 1 3 4 Promotion 1 2 3

−−−−−−→ −−−−−−→ −−−−−−→
4 5 6 3 4 6 2 5 6 4 5 6

By now the reader may recognize that “The Last Problem” from the June 2021
issue of Crux precisely concerns the promotion operation applied to SYTs of rect-
angular shape3 . Part (a) of the problem asks the reader to show that ab applica-
tions of promotion applied to an SYT T of shape a × b returns the initial tableau
T . There is no really simple proof of this fact: it does follow from the fundamental
properties of jeu de taquin as developed by Schützenberger, but it takes quite a
while to develop this theory. We would be very impressed if any reader submitted
a correct solution to this problem.
Parts (b) and (c) also follow from known properties of promotion: see Theorem
2.3 in [2], which relates the “principal chain” and “trajectory” of a tableau, and
Theorem 4.1(a) of [2], which explains that evacuation for rectangular SYTs is 180◦
rotation plus swapping each number i for n + 1 − i.
Although we were not able to explain the full solution to “‘The Last Problem” in
this short space, we hope that we have inspired the reader to learn more about
tableaux and their fascinating properties, as well as the dynamical operations
defined on them.

2 Shapes that behave well under promotion also include the staircase δ = (n, n − 1, . . . , 1),
n
though understanding promotion for the staircase is even more involved than for the rectangle.
3 But note that the arrays there are 180◦ rotations of SYTs as we defined them here.

Crux Mathematicorum, Vol. 48(2), February 2022

 Sam Hopkins /93

References
1. B. E. Sagan, The ubiquitous Young tableau, Invariant theory and tableaux
(Minneapolis, MN, 1988), Volume 19 of IMA Vol. Math. Appl., pp. 262–298,
Springer, New York, 1990. Available online here.
2. R. P. Stanley, Promotion and evacuation, Electron. J. Combin., Volume 16
(2): paper 2.9, 24, 2009. Available online here.
3. A. Yong, What is . . . a Young tableau?, Notices Amer. Math. Soc., 54 (2), pp.
240–241, 2–7. Available online here.

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

Sam Hopkins
[email protected]
Department of Mathematics, Howard University, Washington, DC

Copyright © Canadian Mathematical Society, 2022

94/ 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 May 1, 2022.

4711. Proposed by Sergey Sadov.

Let ABC be a triangle with incenter I and excenters IA , IB , IC . Prove that the
centroid of the four-point system {I, IA , IB , IC } is the circumcenter of 4ABC.

4712. Proposed by Michel Bataille.

Let n be a positive integer and for z ∈ C − {1, 2, . . . , n} let
n n Ç å
X 1 X
k−1 n k
Un (z) = and Vn (z) = (−1) .
k−z k (k − z)2
k=1 k=1

Un (z)
Evaluate the ratio Vn (z) in closed form.

4713. Proposed by András Szilárd.

Let x, y be distinct positive real numbers. Prove that if there exists n1 ∈ N such
that [nx] divides [ny] for all natural numbers n ≥ n1 , then x and y are integers
([a] denotes the integer part of the real number a).

4714. Proposed by András Szilárd.

Let 0 < a < b be two numbers and let f be a function on [a, b] that is twice
differentiable, increasing and concave with nonnegative values. Prove that if a ≤
x1 ≤ x2 ≤ . . . ≤ xn ≤ b, then
n
X f (xi+1 ) − f (xi )
≥ 0,
i=1
xi+1 + xi

where xn+1 = x1 and n ≥ 3.

4715. Proposed by George Stoica.

a) Let A be a 3 by 3 matrix all of whose entries are complex numbers on the
unit circle, and so that det(A) = 0. Must A have two proportional rows or
columns?
b) What if A is 4 by 4?

Crux Mathematicorum, Vol. 48(2), February 2022

 Problems /95

4716. Proposed by Michael Friday.

The three roots of the cubic x3 + 4x2 + 4x + 1 = 0 are the slopes of the sides of a
triangle. Find the slope of its Euler line.

4717. Proposed by Toyesh Sharma.

Find the value of the following integral:
1 1 1
x4 y 3 z 2
Z Z Z
dxdydz.
0 0 0 (x + y + z)(x2 + y 2 + z 2 ) − (x3 + y 3 + z 3 )

4718. Proposed by Pericles Papadopoulos.

Let X, Y and Z be arbitrary points on the sides BC, AC and AB of 4ABC,
respectively. The parallel through X to AB meets AC at Y 0 ; the parallel through
Y to BC meets AB at Z 0 ; the parallel through Z to AC meets BC at X 0 . Assuming
A, B, C, X, Y, Z are distinct, prove the following:
a) points D = X 0 Y ∩ AB, E = XZ 0 ∩ AC and F = ZY 0 ∩ BC are collinear;
b) points P = ZX ∩ X 0 Z 0 , Q = Z 0 Y ∩ XY 0 and R = ZY ∩ X 0 Y 0 are collinear.

4719. Proposed by Neculai Stanciu, modified by the Editorial Board.

We are given a triangle ABC with circumcenter O. For any point P1 on the line
CA define the following:
P2 is the point where the line through P1 perpendicular to OA intersects AB,
P3 is the point where the line through P2 perpendicular to OB intersects BC,
P4 is the point where the line through P3 perpendicular to OC intersects CA,
P5 is the point where the line through P4 perpendicular to OA intersects AB, and
P6 is the point where the line through P5 perpendicular to OB intersects BC.
Prove that P6 P1 is perpendicular to OC.

Copyright © Canadian Mathematical Society, 2022

96/ Problems

4720. Proposed by George Apostolopoulos.

Let a, b and c be positive real numbers with a2 + b2 + c2 = 12. Prove that
a4 b4 c4
√ +√ +√ ≥ 16.
a3 + 1 b3 + 1 c3 + 1

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

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 1 mai 2022.

4711. Proposeé par Sergey Sadov.

Soit ABC un triangle et I le centre de son cercle inscrit ; soient aussi IA , IB et
IC les centres des cercles exinscrits. Démontrer que le centroı̈de de l’ensemble à 4
points {I, IA , IB , IC } est le centre du cercle circonscrit de 4ABC.

4712. Proposeé par Michel Bataille.

Soit n un entier positif; pour z ∈ C − {1, 2, . . . , n}, soient
n n Ç å
X 1 X n k
Un (z) = et Vn (z) = (−1)k−1 .
k−z k (k − z)2
k=1 k=1

Un (z)
Déterminer le ratio Vn (z) en forme close.

4713. Proposeé par András Szilárd.

Soient x, y des nombres réels positifs distincts. Démontrer que s’il existe n1 ∈ N
tel que [nx] divise [ny] pour tout nombre naturel n ≥ n1 , alors x et y sont des
entiers ([a] dénote la partie entière du nombre réel a).

4714. Proposeé par András Szilárd.

Soient 0 < a < b deux nombres réels et f une fonction sur [a, b], deux fois
différentiable, croissante, concave, puis à valeurs non négatives. Démontrer que si
a ≤ x1 ≤ x2 ≤ . . . ≤ xn ≤ b, alors
n
X f (xi+1 ) − f (xi )
≥ 0,
i=1
xi+1 + xi

où xn+1 = x1 et n ≥ 3.

Crux Mathematicorum, Vol. 48(2), February 2022

 Problems /97

4715. Proposeé par George Stoica.

a) Soit A une matrice de taille 3 par 3 formée de nombres complexes se trouvant
sur le cercle unitaire, puis telle que dét(A) = 0. A doit-elle alors avoir deux
rangées ou colonnes proportionnelles?
b) Qu’en est-il si A est plutôt de taille 4 par 4?

4716. Proposeé par Michael Friday.

Les trois racines de la cubique x3 + 4x2 + 4x + 1 = 0 sont les pentes d’un certain
triangle. Déterminer la pente de la droite d’Euler de ce triangle.

4717. Proposeé par Toyesh Sharma.

Déterminer la valeur de l’intégrale suivante:
Z 1Z 1Z 1
x4 y 3 z 2
dxdydz.
0 (x + y + z)(x + y + z ) − (x + y + z )
2 2 2 3 3 3
0 0

4718. Proposeé par Pericles Papadopoulos.

Soient X, Y et Z des points quelconques sur les côtés BC, AC et AB de 4ABC,
respectivement. La ligne parallèle à AB et passant par X rencontre AC en Y 0 ;
la ligne parallèle à BC et passant par Y rencontre AB en Z 0 ; la ligne parallèle à
AC et passant par Z rencontre BC en X 0 . Supposer que A, B, C, X, Y et Z sont
distincts et démontrer les suivantes:
a) les points D = X 0 Y ∩ AB, E = XZ 0 ∩ AC et F = ZY 0 ∩ BC sont alignés;
b) les points P = ZX ∩ X 0 Z 0 , Q = Z 0 Y ∩ XY 0 et R = ZY ∩ X 0 Y 0 sont alignés.

Copyright © Canadian Mathematical Society, 2022

98/ Problems

4719. Proposeé par Neculai Stanciu, avec modification venant de l’éditeur.

Soit le triangle ABC, où O dénote le centre du cercle circonscrit. Pour P1 un
point quelconque sur la ligne CA, on définit les points suivants:
P2 est le point où la ligne passant par P1 et perpendiculaire à OA intersecte AB,
P3 est le point où la ligne passant par P2 et perpendiculaire à OB intersecte BC,
P4 est le point où la ligne passant par P3 et perpendiculaire à OC intersecte CA,
P5 est le point où la ligne passant par P4 et perpendiculaire à OA intersecte AB,
et
P6 est le point où la ligne passant par P5 et perpendiculaire à OB intersecte BC.
Démontrer que P6 P1 est perpendiculaire à OC.

4720. Proposeé par George Apostolopoulos.

Soient a, b et c des nombres réels positifs tels que a2 + b2 + c2 = 12. Démontrer
que
a4 b4 c4
√ +√ +√ ≥ 16.
a3 + 1 b3 + 1 c3 + 1

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /99

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 2021: 47(7), p. 352–356.

4661. Proposed by Mihaela Berindeanu.

Let ABC be a triangle with the point M ∈ BC such that

AC 2 − AB 2
MC − MB = .
2BC
The centroids of triangles AM B and AM C are G1 and G2 , respectively. Prove
that A, G1 , M, C are concyclic points if and only if A, B, M, G2 are also concyclic
points.
We received 14 solutions. We present 4 of them here.
Solution 1, by Michel Bataille.
Let BC = a, CA = b, AB = c, as usual, and let M B = u, M C = v (so that
u + v = a). In barycentric coordinates relative to (A, B, C), we have M = (0, v, u)
and G1 = (a : a + v : u) (since 3G1 = A + B + M ) and G2 = (a : v : a + u). We
know that the circle Γ1 through A, M, C has an equation of the form

a2 yz + b2 zx + c2 xy = (x + y + z)(αx + βy + γz).

Expressing that A(1 : 0 : 0), M (0 : v : u), C(0 : 0 : 1) are on this circle, we obtain
that the equation of Γ1 is

a2 yz + b2 zx + c2 xy = auy(x + y + z).

Similarly, the equation of the circle Γ2 through A, M, B is

a2 yz + b2 zx + c2 xy = avz(x + y + z).

Thus, for i = 1, 2, Gi is on the circle Γi if and only if δi = 0 where

δ1 = au(a+v)+b2 u+c2 (a+v)−3au(a+v), δ2 = av(a+u)+b2 (a+u)+c2 v−3av(a+u).

From the hypothesis, we have 2a(v − u) = b2 − c2 , hence

δ2 − δ1 = a(b2 − c2 − 2a(v − u)) = 0

and therefore δ1 = 0 if and only if δ2 = 0, that is, G1 ∈ Γ1 if and only if G2 ∈ Γ2 ,

as desired.
Solution 2, by Prithwijit De.

Copyright © Canadian Mathematical Society, 2022

100/ Solutions

Denote by a, b, c the lengths of BC, CA, and AB respectively. Let X1 = AG1 ∩BC
and X2 = AG2 ∩ BC. If A, G1 , M , C are concyclic points then

X1 A2
X1 M.X1 C = X1 G1 .X1 A = ⇒ 4X1 A2 = 3.M B.(2a − M B).
3
We also have

2c2 + 2AM 2 = 4X1 A2 + M B 2 ⇒ c2 + AM 2 = M B.(3a − M B).

If A, B, M , G2 are concyclic points then analogously we obtain

b2 + AM 2 = M C.(3a − M C).

But
b2 + AM 2 = M C.(3a − M C) ⇔ c2 + AM 2 = M B.(3a − M B)
because
(b2 + AM 2 ) − (c2 + AM 2 ) = b2 − c2
and

M C.(3a − M C) − M B.(3a − M B) = (M C − M B).(2a) = b2 − c2 ,

which shows that

b2 + AM 2 − M C.(3a − M C) = c2 + AM 2 − M B.(3a − M B).

The equivalence of concyclicity of {A, G1 , M, C} and {A, B, M, G2 } occurs when

b2 + AM 2 − M C.(3a − M C) = c2 + AM 2 − M B.(3a − M B) = 0.

Solution 3, by Theo Koupelis.

Let a, b, c be the lengths of the sides of 4ABC. Without loss of generality let b ≥ c
so that M C ≥ M B. Let D be the foot of the perpendicular from A to BC, and let
F, E, N be the midpoints of BM, CM, BC, respectively. From the given condition
we get
AC 2 = AB 2 + BC 2 + BC · (2M C − 2M B − BC),
and from the law of cosines we have

AC 2 = AB 2 + BC 2 − 2BC · BD.

Therefore,
BC = 2BN = 2(M C − M B + BD)
and thus
M B − BD = M C − BN = M C − CN.
That is, M is the midpoint of DN.

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /101

We will now show that F M · F C = F G1 · F A = 13 AF 2 (that is, A, G1 , M, C are

concyclic) if and only if EM · EB = EG2 · EA = 31 AE 2 (that is, A, B, M, G2 are
concyclic). We have:

1 DN 2
Å ã Å ã
BM BM BM
FM · FC = AF 2 ⇐⇒ 3 · · + CM = AD2 + −
3 2 2 2 2
⇐⇒ 2BM · (BM + DN ) = 4AD2 + DN 2 − 6CM · BM,

and similarly

1 DN 2
Å ã Å ã
2 CM CM 2 CM
EM · EB = AE ⇐⇒ 3 · · + BM = AD + +
3 2 2 2 2
2 2
⇐⇒ 2CM · (CM − DN ) = 4AD + DN − 6CM · BM.

But CM = 21 (a+DN ) and BM = 21 (a−DN ), and thus CM ·(CM −DN ) = BM ·

(BM + DN ). Therefore A, G1 , M, C are concyclic points if and only if A, B, M, G2
are also concyclic points.
Expressing the segments CM, BM, DN and AD in terms of a, b, c, we find that
the above condition is equivalent to 12a4 − 8(b2 + c2 )a2 + (b2 − c2 )2 = 0.

Solution 4, by Miguel Amengual Covas.

Let a = BC, b = CA, c = AB. We put BM = x. Then M C = a − x and the
2
−c2
given equality becomes a − 2x = b 2a which we rewrite as

4ax − c2 = 2a2 − b2

Subtracting x (a + x) from each side gives

3ax − c2 − x2 = 2a2 − b2 − x (a + x) . (1)

Let AG1 and AG2 (extended) meet BC at P and Q, respectively.

G1 G2

B C
P M Q

Copyright © Canadian Mathematical Society, 2022

102/ Solutions

Points A, G1 , M , C are cyclic if and only if

P G1 · P A = P M · P C,

that is,
1
Åã
x x
PA · PA = a− , (2)
3 2 2
where, in 4ABM , P A2 = 14 2 AB 2 + AM 2 − BM 2 = 1
2 c2 + AM 2 − x2 .





4

Substituting for P A2 into (2) and solving for AM 2 yields

AM 2 = 3ax − c2 − x2 . (3)

Similarly, points A, B, M , G2 are cyclic if and only if

QG2 · QA = QM · QB,

that is, Å
1
ã a − x a + x
QA · QA = , (4)
3 2 2
where, in 4AM C,

1
1Ä 2
ä
QA2 = 2 AM 2 + AC 2 − M C 2 = 2 AM 2 + b2 − (a − x) .



4 4

Substituting for QA2 into (4) and solving for AM 2 yields

AM 2 = 2a2 − b2 − x (a + x) . (5)

Since, by hypothesis, (1) holds, then (3) and (5) hold simultaneously, and the
conclusion follows.

4662. Proposed by Michel Bataille.

Let A and B be complex p × p matrices such that AB = BA and A3 B = A and
let m, n be integers with m ≥ n ≥ 1 and m 6= 2n. Show that Am B n is equal to a
power of A or a power of AB.
We received 14 submissions and they were all correct. We present the solution by
the majority of solvers.
First, it is easy to verify by induction that A2k+1 B k = A for any non-negative
integer k. Next we proceed according to the sign of m − 2n:
• If m > 2n, then m = 2n + 1 + ` for some ` ≥ 0. We have

Am B n = A` A2n+1 B n = A`+1 = Am−2n .

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /103

• If m < 2n, then m = n + ` for some 0 ≤ ` ≤ n − 1. We have

Am B n = Am−2`−1 (A2`+1 B ` )B n−` = Am−2` B n−` = (AB)n−` = (AB)2n−m .

Editor’s Comment. As Eagle Problem Solvers pointed out, the assumption that
A and B are complex p × p matrices is not crucial. More generally, the proof
given above holds for any two elements A and B of a semigroup with the given
properties.

4663. Proposed by Vijay Dasari.

Let M be any point in the plane of an acute triangle ABC with sides a, b, c. Prove
that
AM 2 BM 2 CM 2
+ 2 + 2 ≥ 1,
b +c −a
2 2 2 c +a −b
2 2 a + b2 − c 2
with equality when M is the orthocenter.
We received 16 solutions, all of which were correct. We present the solution by
Mohamed Amine Ben Ajiba.
Since ABC is an acute triangle, we have

b2 + c2 − a2 ≥ 0, c2 + a2 − b2 ≥ 0, and a2 + b2 − c2 ≥ 0.

By Bergström’s inequality, we have

AM 2 BM 2 CM 2
+ 2 + 2
b2 +c −a
2 2 c +a −b2 2 a + b2 − c2
(aAM )2 bBM 2 cCM 2
= + +
(ab) + (ca) − a
2 2 4 (bc) + (ab) − b
2 2 4 (ca) + (bc)2 − c4
2

(aAM + bBM + cCM )2

≥ .
2 [(ab) + (bc)2 + (ca)2 ] − (a4 + b4 + c4 )
2

From the formula

2 (ab)2 + (bc)2 + (ca)2 − a4 + b4 + c4 = 16F 2 ,

 

where F is the area of triangle ABC, we obtain

ã2
AM 2 BM 2 CM 2 aAB + bBM + cCM
Å
+ + ≥ .
b2 + c2 − a2 c 2 + a 2 − b2 a2 + b2 − c2 4F

However we know that (see below)

aAB + bBM + cCM ≥ 4F, (1)

with equality exactly when M is the orthocenter. Therefore

AM 2 BM 2 CM 2
+ 2 + 2 ≥ 1,
b2 +c −a
2 2 c +a −b
2 2 a + b2 − c 2

Copyright © Canadian Mathematical Society, 2022

104/ Solutions

with equality exactly when M is the orthocenter.

For the proof of (1), let A1 , B1 , C1 be the feet of the altitudes of ABC, and let A0 ,
B 0 , C 0 be the projections of M onto its sides. We then have AM + A0 M ≥ AA1
or
aAM ≥ aAA1 − aA0 M = 2F − 2[BM C],
where [X] is the area of X. Similarly, we have

bBM ≥ 2F − 2[CM A] and cCM ≥ 2F − 2[AM B].

Adding these inequalities and using the relation

[AM B] + [BM C] + [CM A] = F,

we obtain
aAB + bBM + cCM ≥ 4F.
Equality holds when M is the orthocenter of triangle ABC.

4664. Proposed by Marian Cucoanes and Lorian Saceanu.

Let ABCDEF be a convex cyclic hexagon that respects the following rules:
a) The lines AD, BE, CF are concurrent;
b) (1/3)(AF + BC + DE) = AB = CD = EF .
Prove that ABCDEF is a regular hexagon.
Almost all of the 10 submissions we received used an approach similar to that of
our featured solution by Anay Aggarwal.
Denote by O the point where the chords AD, BE, CF concur. In triangles ABO
and EDO we have

∠ABO = ∠ABE = ∠ADE = ∠ODE,

while the (vertical angles) at O satisfy ∠AOB = ∠EOD. Consequently, ∆ABO ∼

EDO and
AB AO
= .
ED EO
Analogously,
CD CO EF EO
= and = .
AF AO CB CO
Hence
AB · CD · EF
= 1.
CB · AF · DE
AF +BC+DE
Consequently, our assumption AB = CD = EF = 3 implies that

AF + BC + DE √
3
√
3
= AB · CD · EF = CB · AF · DE.
3

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /105

However, the AM-GM inequality tells us that

AF + BC + DE √3
≥ AF · BC · DE,
3
with equality if and only if AF = BC = DE. Thus, all six sides of the hexagon
are equal and, because it is inscribed in a circle, it is regular.

4665. Proposed by Daniel Sitaru.

Find π
!
cos x
Z 2
lim dx .
n→∞ π
6
sin x(1 + sinn x)

We received 25 submissions, all of which are correct. We present the similar solu-
tions by Brian Bradie, UCLan Cyprus Solving Group, and Eagle Problem Solvers.
Since
cos x cos x (sinn−1 x)(cos x)
n = − ,
(sin x)(1 + sin x) sin x 1 + sinn x
we have
π/2 ã π/2
cos x 1
Z Å
n
dx = ln sin x − ln(1 + sin x)
π/6 sin x(1 + sinn x) n
π/6
1 1 1 1
Å ã
= − ln 2 − ln + ln 1 + n .
n 2 n 2
Hence,
π/2
ÇZ å
cos x
lim dx = − ln 1/2 = ln 2.
n→∞ π/6 sin x(1 + sinn x)

4666. Proposed by Dong Luu.

Let ABC be a triangle and let the circle I be tangent to BC, CA and AB at
points D, E and F , respectively. Let M , N be the points on the line EF such
that BM is parallel to AC and CN is parallel to AB. Let P and Q be points
on DM and DN , respectively such that BP is parallel to CQ. Denote by S the
intersection point of P F and QE. Prove that S lies on the circle I.

Copyright © Canadian Mathematical Society, 2022

106/ Solutions

We received 12 solutions. We present the solution by Jason Fang, edited.

In the following, ]A, ]B and ]C denote the angles of 4ABC.
In 4AF E, AF = AE and hence
]B + ]C
]AF E = ]AEF = .
2
From CN k AF we get 4CN E ∼ 4AF E, and so
]B + ]C
]N CE = ]A and ]CN E = ]CEN = ;
2
also CN = CE. From CE = CD it follows that CN = CD; since

]N CD = ]N CE + ]ECD = ]A + ]C

we get
]B
]CN D = and hence (1)
2
]C
]DN E = ]CN E − ]CN D = . (2)
2
We repeat the first part of the argument with B, M and E instead of C, N and
F (starting from BM k AE) to get
]C
]DM B = ; (3)
2
we also note that 4BM F ∼ 4CN E.
Combining (2) and (3) we get ]QN E = ]P M B. As angles whose sides are parallel
lines, we also have ]P BM = ]QCE. It follows that the similarity transformation
which maps 4BM F to 4CN E maps P to the isogonal conjugate of Q. Hence
]P F M = ]QEC. From 4F SE, using this observation and the fact that opposite
angles are equal, we calculate

]F SE = 180◦ − (]SEF + ]SF E) = 180◦ − (]QEN + ]P F M )

= 180◦ − (]QEN + ]QEC)
= 180◦ − ]N EC
= 180◦ − ]AEF.

Finally, note that ]AEF = ]F DE (since the intercepted arc on circle I is the
same for both angles). Hence ]F SE = 180◦ − ]F DE, so F SED is a cyclic
quadrilateral. Therefore, S lies on the circle I.
Editor’s comment. It was brought to the attention of the editors that this problem
also appeared as M2672 in Issue 10 (2021) of the Russian magazine Kvant. We take
this opportunity to remind our readers that problems submitted to any journal
should be original and not actively under consideration at another publication.

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /107

4667. Proposed by Conar Goran.

Pn
Let x1 , . . . , xn > 0 be real numbers and s = i=1 xi . Prove
n n
!s
Y 1X 2
xxi i ≤ x .
i=1
s i=1 i

When does equality occur?

We received 8 submissions, all correct. Most solutions are similar to one another
and use either the weighted AM-GM inequality or the Jensen’s inequality. We
present two such solutions.
Solution 1, by Benjamin Braiman.
n
X xi
Since each xi is positive and = 1, we have by the weighted AM-GM in-
i=1
s
equality that
n n  n
Y xi X xi  1X 2
xi s ≤ xi = x ,
i=1 i=1
s s i=1 i
from which it follows that
n n
!s
Y 1X 2
xxi i ≤ x .
i=1
s i=1 i

Equality holds if and only if all xi s are equal.

Solution 2, by Brian Bradie.

By Jensen’s inequality we have
n n
!
X xi X x2 i
ln xi ≤ ln ,
i=1
s i=1
s
so ! !
n n
Y x /s
X x2 i
ln xi i ≤ ln .
i=1 i=1
s
Exponentiating both sides then yields
n n n n
!s
Y x /s 1X 2 Y 1X 2
xi i ≤ x , or xxi i ≤ x ,
i=1
s i=1 i i=1
s i=1 i

with equality when x1 = x2 = · · · = xn .

4668. Proposed by Jiahao Chen.

Let Γ be the inscribed circle of triangle ABC, and I is the center of Γ. Suppose
Γ touches BC, CA and AB at D, E and F , respectively. Let X be an arbitrary

Copyright © Canadian Mathematical Society, 2022

108/ Solutions

point on the smaller arc DF , and the line perpendicular to XE passing through I
intersects line BX in point Y . Show that IY is the external angle bisector of the
angle AY C.
We received 5 submissions: two were complete and essentially correct, and another
two correctly proved that the line IY was an angle bisector but did not address the
matter of whether it was internal or external; the fifth was computer aided so it is
hard to determine its status. We feature the solution by Marie-Nicole Gras, with
the final step modified by the editor.
A

F
•

Y
X

E
θ
I B

•
D

We shall prove that, more generally, IY is the internal bisector of ∠AY C when X
is on the arc DF that contains E, and the external bisector on the other (smaller)
arc DF . Let r be the inradius; we denote by ∠A, ∠B and ∠C the value of ∠BAC,
∠CBA and ∠ACB, respectively. Since I is the incenter, we have the equalities:

π ∠B
β :=∠BIF = − ;
2 2
π ∠B π ∠B
α :=∠BIE = − + π − ∠A = + + ∠C.
2 2 2 2
−
→ −→
We introduce cartesian coordinates with the origin at I and axis Ix on IB. We put
−→ −→
θ = ∠(IB, IX); since X is on the smaller arc DF , we have − π2 < β < θ < β < π2 ;
we find
 r 
B∼ ,0 ,
cos β
E ∼ (r cos α, r sin α), F ∼ (r cos β, r sin β), D ∼ (r cos β, −r sin β),
X ∼ (r cos θ, r sin θ).

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /109

I) Coordinates of Y .
Equation of line BX is
 r  sin θ
y = r2


r sin θ x − r cos θ − ·
cos β cos β

Line `, passing through I and perpendicular to the chord XE, has parametric
equation:
θ+α θ+α
x = t cos , y = t sin ,
2 2
where t is given by

θ+α  r  θ+α sin θ

= r2


r sin θ t cos − r cos θ − t sin
2 cos β 2 cos β
 θ+α θ+α 1 θ + α sin θ
⇐⇒ t sin θ cos − cos θ sin + sin =r
2 2 cos β 2 cos β
 θ−α 1 θ + α sin θ
⇐⇒ t sin + sin =r
2 cos β 2 cos β
 θ−α θ+α 
⇐⇒ t sin cos β + sin = r sin θ.
2 2
We put
θ−α θ+α
w = sin cos β + sin ;
2 2
then, since lines BX and ` are concurrent, w 6= 0, and t = r sin θ
w · The coordinates
0
(z, z ) of Y are
r sin θ θ + α r sin θ θ+α
Å ã
Y ∼ cos , sin ·
w 2 w 2

II) Equation of line AY .

Vertex A is the intersection of lines tangent to the incircle at points F and E;
then, its coordinates (a, a0 ) are given by the system

a cos β + a0 sin β = r
®

a cos α + a0 sin α = r;

the determinant is sin(α − β) = sin(π − ∠A) 6= 0. Then, the coordinates of A are

sin α − sin β 2 sin α−β

2 cos 2
α+β
cos α+β
2
a=r =r = r
sin(α − β) 2 sin α−β
2 cos α−β
2 cos α−β
2
cos β − cos α 2 sin α−β
2 sin 2
α+β
sin α+β
a0 = r =r = r 2
·
sin(α − β) 2 sin α−β
2 cos α−β
2 cos α−β
2

Copyright © Canadian Mathematical Society, 2022

110/ Solutions

It follows that the equation of line AY is

Ç å Ç å
r sin θ θ+α r α+β r sin θ θ+α r α+β
sin − α−β
sin x− cos − α−β
cos y
w 2 cos 2 2 w 2 cos 2 2
r2 sin θ θ−β
= α−β
sin
w cos 2 2
or
θ−β
ux − vy = r sin θ sin , with
2
α−β θ+α α+β
u = sin θ cos sin − w sin ,
2 2 2
α−β θ+α α+β
v = sin θ cos cos − w cos ·
2 2 2

III) Distance d from I to line AY .

Since the coordinates of I are (0,0), we have

sin2 θ sin2 θ−β sin2 θ sin2 θ−β

2 sin
2 θ+β
d2 = 2
r 2
= 2
r2 .
u2 + v 2 (u2 + v 2 ) sin2 θ+β
2

We compute the denominator:

α−β α−β θ−β

u2 + v 2 = sin2 θ cos2 + w2 − 2w sin θ cos cos ,
2 2 2
θ+β  θ+β α − β 2 θ+β 2
4(u2 + v 2 ) sin2 = sin2 θ 2 sin cos + 4 sin2 w
2 2 2 2
 θ+β α − β  θ+β θ −β
− 2w sin θ 2 sin cos 2 sin cos .
2 2 2 2

We notice that

θ+β α−β θ+α θ − α 

2 sin cos = sin + sin +β
2 2 2 2
θ+α θ−α θ−α
= sin + sin cos β + cos sin β
2 2 2
θ−α
= w + cos sin β;
2
θ+β
2 sin2 = 1 − cos(θ + β) = 1 − cos θ cos β + sin θ sin β;
2
θ+β θ−β
2 sin cos = sin θ + sin β.
2 2

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /111

We deduce that
θ+β  θ−α 2
4(u2 + v 2 ) sin2 = sin2 θ w + cos sin β + 2 1 − cos θ cos β + sin θ sin β w2


2 2
 θ−α 

− 2w sin θ w + cos sin β sin θ + sin β
2
 θ−α 
= sin2 θ w2 + cos2 sin2 β + 2 1 − cos θ cos β w2


2
 θ−α 
− 2w sin θ w sin θ + cos sin2 β
2
− θ−α
ï ò
θ α 
+ 2w sin2 θ cos + 2w2 sin θ − 2w sin θ w + cos sin θ sin β,
2 2
and the coefficient of sin β in the brackets is equal to zero.
IV) Line IY is the bisector of ∠AY C.
All calculations relating to vertex C can be deduced from those obtained with A
by replacing β by −β. Note that the expression for d is invariant if we replace β
by −β. It follows that I is equidistant from lines AY and CY ; consequently, IY
is a bisector of one pair of the angles defined by those two lines. It remains to
prove that it is the external bisector of ∠AY C when X is on the smaller arc DF ,
and the internal bisector when on the larger. The lines AY and CY divide the
plane into four quadrants; as usual, we call the first quadrant that which is defined
by ∠AY C — that is, the first quadrant is the region bounded by the rays Y A
and Y C. The second quadrant is bounded by the rays Y C and −Y A, and so on.
The point I is in the first or third quadrants (making Y I the internal bisector of
∠AY C) if the rotation that takes the ray IA to the ray IY is in the same direction
as the rotation that takes the ray IY to the ray IC. The point I is in the second
or fourth quadrants if those two rotations are in opposite directions. The sense
a a0 z z0
of the rotations is given by the sign of the determinants 0 and ,
z z c c0
namely,
r sin θ θ+α α+β α+β θ+α
Å ã
0 0 r
az − a z = sin cos − sin cos
cos α−β
2
w 2 2 2 2
r2 sin θ θ−β
= sin ,
w cos α−β
2
2

and (by replacing β by −β),

r2 sin θ θ+β
zc0 − z 0 c = −(cz 0 − c0 z) = − α+β
sin .
w cos 2 2

Consider their product

r4 sin2 θ β−θ β+θ

(az 0 − a0 z)(zc0 − z 0 c) = sin sin ·
w2 cos α−β
2 cos α+β
2
2 2

Copyright © Canadian Mathematical Society, 2022

112/ Solutions

π ∠B π ∠B
Since β = 2 − 2 and α = 2 + 2 + ∠C, we have

α−β B+C π A
= = −
2 2 2 2
α+β π+C π C
= = + ;
2 2 2 2

α−β α+β
We deduce that cos cos < 0, which clearly does not depend on θ.
2 2
When −β < θ < β (where X is on the small arc DF as in the original statement
of the problem), we have

β−θ β+θ 1 
sin sin = cos θ − cos β > 0.
2 2 2

It follows that (az 0 − a0 z)(zc0 − z 0 c) < 0, and I is in the second or fourth quadrant,
whence IY is the external bisector of ∠AY C. Otherwise, when θ > β or θ < −β,
(az 0 − a0 z)(zc0 − z 0 c) > 0, I is in the first or third quadrant, and IY is the internal
angle bisector.

Editor’s comments. Note that when X is at D or F the angle AY C is undefined

(because Y is at A or C); when X = E, “external” has no meaning because
∠AY C = 180◦ ; and when X is at either point where BI meets the incircle, then
Y = I and the line Y I is undefined. When X is permitted to be an arbitrary
point of the incircle different from those five points, the problem of proving that
the line IY bisects ∠AY C is relatively easy; the hard work comes in showing that
it is external if and only if X is on the smaller arc DF ; this is apparently not so
easy.

4669. Proposed by Warut Suksompong.

For a given positive integer n, a 4n×4n table is partitioned into 16n2 unit squares,
each of which is coloured in one of 4 given colours. A set of four cells is called
colourful if the centers of the cells form a rectangle with sides parallel to the sides
of the table, and the cells are coloured in all four different colours. Determine the
maximum number of colourful sets.

We received 2 correct solutions. There was an additional incorrect solution and a

solution that conjectured the correct answer on the basis of some examples. We
present a solution based on that of UC Lan Cyprus Problem Solving Group.

The maximum number is 24n4 . Denote the colours by 1, 2, 3, 4.

In a given row, suppose that there are bk cells with colour k, so that

b1 + b2 + b3 + b4 = 4n.

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /113

Then the number of pairs of cells in that row with different colours is

b1 b2 + b1 b3 + b1 b4 + b2 b3 + b2 b4 + b3 b4
(b1 + b2 + b3 + b4 )2 − (b21 + b22 + b23 + b24 )
=
2
(4n)2 − (b21 + b22 + b23 + b24 ) 16n2 − (1/4)(b1 + b2 + b3 + b4 )2
= ≤
2 2
16n2 − 4n2 2
= = 6n .
2
Hence the number of pairs of cells lying in the same row of the table with different
colours is not greater than (4n)(6n2 ) = 24n3 .
Fix a pair (i, j) of indices and let there be mij pairs of cells in the same row, one
in each of columns i and j, with different colours. Let a1 , a2 , a3 , a4 , a5 , a6 be the
number of pairs for which the colours are, respectively, (1, 2), (3, 4), (1, 3), (2, 4),
(1, 4), (2, 3). Then the number of colourful sets in these two columns is
 a + a 2  a + a 2  a + a 2
1 2 3 4 5 6
a1 a2 + a3 a4 + a5 a6 ≤ + +
2 2 2
 a + a + a + a + a + a 2
1 2 3 4 5 6 m2ij
≤ = .
2 4
The total number of colorful sets does not exceed
1X 2
{mij : 1 ≤ i, j ≤ 4n}.
4
The
P quantities mij are subject to the constraints mij ≤ 4n for each (i, j) and
mij ≤ 24n3 .
In obtaining an upper bound for mij , we can systematically use the following
P
procedure for replacing pairs {u, v} of summands, where u ≥ v, by pairs {u +
e, u − e} where e > 0, thus increasing the sum since

(u + e)2 + (v − e)2 = (u2 + v 2 ) + 2e(u − v) + 2e2 > u2 + v 2 .

This can be done to lead us to a sum where the entries add up to 24n3 , and 6n2
P 2
entries are equal to 4n with the rest equal to 0. Thus mij ≤ (6n2 )(4n)2 =
4(24n ) and the number of colourful sets does not exceed 24n4 . It remains to find
4

a configuration that realizes this bound.

We can construct a maximal table in this way. Let Ak be a n × n table with each
cell coloured k. These are put together to form a (4n) × (4n) table:

A1 A2 A3 A4
A2 A1 A4 A3
A3 A4 A1 A2
A4 A3 A2 A1

Copyright © Canadian Mathematical Society, 2022

114/ Solutions

There are 42 n2 ways of picking the rows containing the cells of a colourful set.


For each choice, there are 4n2 ways of picking the columns. Hence there are 24n4
colourful sets in all.

Editor’s comment. The P

proposer had a similar solution, with an alternative argu-
ment for the bound on bi :

b1 b2 + b1 b3 + b1 b4 + b2 b3 + b2 b4 + b3 b4
Ç å 4 Ç å Ç å 4 4
4n X bi 4n 1X 1X 2
= − = + bi − b
2 i=1
2 2 2 i=1 2 i=1 i
≤ 2n(4n − 1) + 21 (4n) − 18 (b1 + b2 + b3 + b4 )2 = 8n2 − 81 (16n2 ) = 6n2 .

Alternatively, we have

b1 b2 + b1 b3 + b1 b4 + b2 b3 + b2 b4 + b3 b4
= 21 [b1 (b2 + b3 + b4 ) + b2 (b3 + b4 + b1 ) + b3 (b4 + b1 + b2 ) + b4 (b1 + b2 + b3 )]
4 4 4
1X X 1X 2
= bi (4n − bi ) = 2n bi − bi ≤ 8n2 − 2n2 = 6n2 .
2 i=1 i=1
2 i=1

For the maximal configuration, the proposer considered the 4n×4n array consisting
of n2 4 × 4 tables identically coloured as follows:

1 2 3 4
2 1 4 3
3 4 1 2
4 3 2 1

where successive quartets of rows or columns are the same. There are
Ç å Ç å
4n n
−4 = 6n2
2 2

choices of pairs of distinct rows from which colourful sets can be selected. Now
consider a distinct pair of rows each consisting of n quartets of colours. There are
4 ways of selecting a colourful set if all its elements are in the same corresponding
quartet in the two rows. There are 8 ways of selecting a colourful set if its left two
elements lie in one quartet and the right two in another quartet in the two rows.
Hence the total number of colourful sets involving these two rows is
Ç å
n
4n + 8 = 4n2 .
2

Therefore the total number of colourful sets in this array is 24n4 .

Crux Mathematicorum, Vol. 48(2), February 2022

 Solutions /115

4670. Proposed by Nguyen Viet Hung.

Let a, b, c be real numbers such that (a + b)(b + c)(c + a) 6= 0. Prove that
ã2 ã2 ã2
4abc
Å Å Å
a b c
+ + + ≥ 1.
a+b b+c c+a (a + b)(b + c)(c + a)

We received 20 submissions of which 14 were correct and complete. We present

the solution by Michel Bataille utilizing an approach shared in most submissions.
b c a
With x = b+c , y= c+a , z= a+b , the inequality to be proved becomes

x2 + y 2 + z 2 + 4xyz ≥ 1.
c a b
Since 1 − x = b+c , 1−y = c+a , 1−z = a+b , we have
Å ã
abc
xyz = (1 − x)(1 − y)(1 − z) = .
(a + b)(b + c)(c + a)

Hence,
4xyz = 2 − 2(x + y + z) + 2(xy + yz + zx),
and by adding x2 + y 2 + z 2 on both sides we obtain
2
x2 + y 2 + z 2 + 4xyz = 1 + ((x + y + z) − 1) ≥ 1.

Editor’s comment. Oliver Geupel’s proof included an example in which equality

is achieved.

Copyright © Canadian Mathematical Society, 2022