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 2020. 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 2020 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
Shaun Fallat University of Regina
Chip Curtis Missouri Southern State University
Allen O’Hara University of Western Ontario

Guest Editors Cameron Morland University of Waterloo

Vasile Radu Birchmount Park Collegiate Institute
Aaron Slobodin University of Victoria
Ethan White University of British Columbia

Editor-at-Large Bill Sands University of Calgary

Managing Editor Denise Charron Canadian Mathematical Society
 IN THIS ISSUE / DANS CE NUMÉRO

3 Editorial Kseniya Garaschuk

4 MathemAttic: No. 11
4 Problems: MA51–MA55
8 Solutions: MA26–MA30
13 Teaching Problems: No. 8 Richard Hoshino
18 Olympiad Corner: No. 379
18 Problems: OC461–OC465
20 Solutions: OC436–OC440
26 Focus On . . . : No. 39 Michel Bataille
32 Problems: 4501–4510
36 Solutions: 4451–4460

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
 Editorial /3

EDITORIAL
New Year, New Volume! As my palindromic-loving friends would like to say,
Happy 2 × 505 × 2 !
My New Year’s resolution is to read more. I tend to flip through many books
(mostly math or math education related), but I find that I don’t fully read many.
This year, I’m hoping to change that and, in the absence of Book Reviews, I will
update you in my readings.
I am starting with something old: “On Being the Right Size And Other Essays”
by J. B. S. Haldane (1892–1964). An evolutionary biologist, Haldane was a pas-
sionate science popularizer and he writes with ease and charisma. The collection
was recommended to me by a colleague specifically for the first (and title) essay
after I lamented that my biology students use the surface area/volume “law” as
if it is something that holds independent of the shape you’re considering or pro-
portionality constants involved. “On Being the Right Size” addresses exactly that
question of proportion through a variety of examples: scaling of bones, danger of
falling or getting wet, how limits to gas diffusion limit insect size, why big animals
don’t have giant eyes and so on. “Comparative anatomy is largely the story of the
struggle to increase surface in proportion to volume”, hence large animals’ fractal
lungs and twisted guts. It is an interesting read, persuasive in its arguments (al-
beit not always infallible) and sprinkled with quantitative reasoning. I’m looking
forward to reading the rest of the book.
Kseniya Garaschuk

The book can be found, for example, here

https://archive.org/details/OnBeingTheRightSize-J.B.S.Haldane/

Copyright © Canadian Mathematical Society, 2020

4/ MathemAttic

MATHEMATTIC
No. 11
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, 2020.

MA51. Proposed by Nguyen Viet Hung.

Find all non-negative integers x, y, z satisfying the equation
2x + 3y = 4z .

MA52. The diagram shows part of a tessellation of the plane by a quadrilateral.

Khelen wants to colour each quadrilateral in the pattern.

1. What is the smallest number of colours he needs if no two quadrilaterals that

meet (even at a point) can have the same colour?
2. Suppose that quadrilaterals that meet along an edge must be coloured differ-
ently, but quadrilaterals that meet just at a point may have the same colour.
What is the smallest number of colours that Khelen would need in this case?
3. What is the smallest number of colours needed to colour the edges so that
edges that meet at a vertex are coloured differently?

MA53.
Find all positive integers m and n which satisfy the equation
23 − 1 33 − 1 m3 − 1 n3 − 1
3
· 3 ··· 3 = 3 .
2 +1 3 +1 m +1 n +2

Crux Mathematicorum, Vol. 46(1), January 2020

 MathemAttic /5

MA54. How many six-digit numbers are there, with leading 0s allowed, such
that the sum of the first three digits is equal to the sum of the last three digits,
and the sum of the digits in even positions is equal to the sum of the digits in odd
positions?

MA55. The diagram shows three touching semicircles with radius 1 inside
an equilateral triangle, which each semicircle also touches. The diameter of each
semicircle lies along a side of the triangle. What is the length of each side of the
equilateral triangle?

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

Copyright © Canadian Mathematical Society, 2020

6/ MathemAttic

Les problémes 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 2020.
La rédaction souhaite remercier Rolland Gaudet, professeur titulaire à la retraite à
l’Université de Saint-Boniface, d’avoir traduit les problèmes.

MA51. Proposé par Nguyen Viet Hung.

Déterminer tous les entiers non négatifs x, y, z satisfaisant à l’équation

2x + 3y = 4z .

MA52. Le diagramme montre une partie d’un pavage du plan par un quadri-
latère. Katherine désire en effectuer un colorage.

1. Déterminer le plus petit nombre de couleurs possible si Katherine exige que

deux quadrilatères se touchant, même en un seul point, aient besoin d’être
colorés différemment.
2. Supposons maintenant que deux quadrilatères partageant un côté doivent
être colorés différemment, mais pas nécessairement ceux se touchant en un
point seulement. Déterminer le plus petit nombre de couleurs requises pour
colorer les quadrilatères dans ce contexte.
3. Enfin, Katherine désire colorer les côtés seulement, mais de faon à ce que les
côtés se rencontrant en un point soient colorés différemment. Déterminer le
plus petit nombre de couleurs requises dans ce contexte.

Crux Mathematicorum, Vol. 46(1), January 2020

 MathemAttic /7

MA53.
Déterminer tous les entiers positifs m et n satisfaisant à l’équation

23 − 1 33 − 1 m3 − 1 n3 − 1
· · · · = .
23 + 1 33 + 1 m3 + 1 n3 + 2

MA54. Combien de nombres à six chiffres y a-t-il, tels que la somme des
trois premiers chiffres est égale à la somme des trois derniers chiffres et puis que
la somme des chiffres en positions paires égale la somme des chiffres en positions
impaires? La présence de 0s en première(s) position(s) est permise.

MA55. Le diagramme montre trois demi cercles de rayon 1 à l’intérieur d’un

triangle équilatéral, les diamètres étant situés sur les côtés du triangle. Chaque
demi cercle touche les deux autres et le triangle. Déterminer la longueur du côté
du triangle.

Copyright © Canadian Mathematical Society, 2020

8/ MathemAttic

MATHEMATTIC
SOLUTIONS
Statements of the problems in this section originally appear in 2019: 45(6), p. 303–305.

MA26. Nine (not necessarily distinct) 9-digit numbers are formed using each
digit 1 through 9 exactly once. What is the maximum possible number of zeros
that the sum of these nine numbers can end with?
Originally Problem M2430 of Kvant.
We received 3 submissions, all of which were correct and complete. We present
the solution by the Missouri State University Problem Solving Group.
The answer is eight. Since

8 × 987654321 + 198765432 = 8100000000,

the answer is at least 8. But the maximum value the sum can be is

9 × 987654321 = 8888888889,

so the only other possibility is to have nine zeros. Now each number whose digits
are a permutation of 1, . . . , 9 is a multiple of 9, since the sum of their digits is.
Therefore any sum of these numbers must also be a multiple of 9. But the only
10-digit number ending in nine zeros that is a multiple of 9 is 9000000000 and this
is larger than our upper bound.
We note that analogous methods extend this result to base b: if b − 1 numbers
consisting of permutations of 1, . . . , b−1 are added, the maximum possible number
of zeros that their sum can end in is b − 2.

MA27. You want to play Battleship on a 10 × 10 grid with 2 × 2 squares

removed from each of its corners:

Crux Mathematicorum, Vol. 46(1), January 2020

 MathemAttic /9

What is the maximum number of submarines (ships that occupy 3 consecutive

squares arranged either horizontally or vertically) that you can position on your
board if no two submarines are allowed to share any common side or corner?
Originally Problem 24 of 2018 Savin contest.
We received 1 submission, which was correct but incomplete. We present the so-
lution by Richard Hess and Taus Brock-Nannestad, and completed by the editor.
Consider the following diagram:

There is no way to place a submarine on the grid without its touching one of the
nine marked grid points. No two submarines can touch the same marked grid point
so nine submarines is the most that can be placed on the grid without touching.
It is possible to place nine submarines on the grid. There are many ways to do
this; here is one:

This is an example of a problem where a construction is a necessary part of the

proof. Without actually demonstrating that it is possible to place nine submarines,
we know only that we cannot place more than this many.

MA28. Prove that for all positive integers n, the number

1 4n+1
+ 44n+3 + 1


4
3
is not prime.
Originally Problem 27 of 2017 Savin contest.
We received 4 submissions which were correct and complete. We present the solu-
tion by the Missouri State University Problem Solving Group.

Copyright © Canadian Mathematical Society, 2020

10/ MathemAttic

The statement is false. If n = 6, we have

1 4n+1
(4 + 44n+3 + 1) = 6 380 099 472 108 203,
3
which is prime. (Mathematica claims that n = 861 and n = 5304 also yield prime
values).
However, it is true that if n 6≡ 0 mod 3, then (44n+1 + 44n+3 + 1)/3 is never prime.

If n ≡ 1 mod 3, then n = 3k + 1, k ∈ Z and

44n+1 + 44n+3 + 1 = 412k+5 + 412k+7 + 1

= 16 · 644k+1 + 4 · 644k+2 + 1
≡ 2 · 1 + 4 · 1 + 1 mod 7
≡ 0 mod 7

and
44n+1 + 44n+3 + 1 ≥ 4 + 43 + 1 = 69 > 7,

so 7 is a non-trivial factor of (44n+1 + 44n+3 + 1)/3.

If n ≡ 2 mod 3, then n = 3k + 2, k ∈ Z and

44n+1 + 44n+3 + 1 = 412k+9 + 412k+11 + 1

= 644k+3 + 16 · 644k+3 + 1
≡ 1 + 7 · 1 + 1 mod 9
≡ 0 mod 9

and
44n+1 + 44n+3 + 1 ≥ 69 > 9,

so 9 is a non-trivial factor of (44n+1 + 44n+3 + 1) and hence 3 is a non-trivial factor

of (44n+1 + 44n+3 + 1)/3.

MA29. Find all positive integers n satisfying the following condition: numbers
1, 2, 3, . . . , 2n can be split into pairs so that if numbers in each pair are added and
all the sums are multiplied together, the result is a perfect square.
Originally Problem 2 of Fall Junior A-level of XL Tournament of Towns 2017.
We received 3 submissions, all of which were correct and complete. We present
the solution by the Missouri State University Problem Solving Group, modified by
the editor.
We claim that n satisfies the condition if n > 1.

Crux Mathematicorum, Vol. 46(1), January 2020

 MathemAttic /11

We first observe that n = 1 fails the condition. For n = 1 the only pairing is
{1, 2}, the sum of which is the non-perfect square 3.
There are two cases:
1. n = 2k where k ≥ 1. By pairing i with 2n + 1 − i for i = 1, 2, . . . , n gives a

2
product of (2n + 1)k .
2. n = 2k + 1 where k ≥ 1. When k ≥ 1, we pair 1 and 5, 2 and 4, 3 and 6, and
6 + i with 2n + 1 − i for i = 1, 2, . . . , n − 3 = 2k − 2. The product is then

2
(1 + 5)(2 + 4)(3 + 6)(2n + 7)2k−2 = 18(2n + 7)k−1 .

MA30. Consider the two marked angles on a grid of equilateral triangles.

Prove that these angles are equal.

Originally Problem 18 of 2017 Savin contest.
We received 6 solutions, all of which were correct. We present the solution of
Missouri State University Problem Solving Group, modified by the editor.
Let the side lengths of the equilateral triangles be 1.
Method I. Consider the figure below.

Let α = m(∠ACB) and β = m(∠AED). Since AB = 2 and BC = 5, the Law of

Cosines gives p √
AC = 22 + 52 + 2 · 5 = 39.
Applying the Law of Cosines again
√ 2 …
39 + 52 − 22 12
cos α = √ = .
2 · 5 39 13

Copyright © Canadian Mathematical Society, 2020

12/ MathemAttic

√
Similarly, DE = 3 and applying the Law of Cosines to 4AF E we have
p √
AE = 12 + 32 + 1 · 3 = 13.

One more use of the Law of Cosines gives

√ 2 √ 2 …
13 + 3 − 22 12
cos β = √ √ = ,
2 3 39 13

so the angles in question are congruent.

Method II. Consider the figure below.

Triangle ABD in this figure is congruent to triangle DAE in the figure in Method I.
Thus, we wish to show that ∠ACB ∼ = ∠ADB. The point marked O is equidistant
from each of A, B, C, D (it lies on the intersection of the perpendicular bisectors
of AD, AB, and BC). Therefore, these points lie on a circle centered at O. Since
∠ACB and ∠ADB are subtended by the same arc, they must be congruent.

Crux Mathematicorum, Vol. 46(1), January 2020

 Richard Hoshino /13

TEACHING PROBLEMS
No. 8
Richard Hoshino
The Calendar Problem
In the Calendar Problem, your goal is to figure out the day of the week on which
you were born.
There are various YouTube videos of mathematicians (or “mathemagicians”) per-
forming this trick in their heads. For example, an audience member will call out
her birthday (e.g. May 25, 2004), and the mathematician will instantly reply,
“Tuesday”.
In this article, we will unpack this problem and determine an algorithm to solve
this problem.
First, let’s investigate the day of the week that our birthday falls on in the year
2020. To do this, all we need is the knowledge that January 1, 2020 is Wednesday.
Whenever I have presented this problem in a class, either with high school students
or undergraduates, one student always knows the number of days in each month:
January 31
February 29 (since 2020 is a leap year)
March 31
April 30
May 31
June 30
July 31
August 31
September 30
October 31
November 30
December 31

Notice that January 1 must be the same day of the week as January 8, January
15, January 22, and January 29. This is because each of these numbers in {1, 8,
15, 22, 29} gives the same remainder when divided by 7.
Thus, for some birthdays, the Calendar Problem can be easily solved. Let’s con-
sider someone born on January 17. For the year 2020, since January 1 is a Wednes-
day we know that January 15 is a Wednesday, which implies that January 16 is a
Thursday, from which it follows that January 17 is a Friday.
For birthdays in the month of January, notice that the answer can be found by
simply taking the date, dividing by 7, and calculating the remainder. Then we
can use this table to read off the answer:

Copyright © Canadian Mathematical Society, 2020

14/ Teaching Problems

0 Tuesday
1 Wednesday
2 Thursday
3 Friday
4 Saturday
5 Sunday
6 Monday

Here are two common approaches for solving the Calendar Problem.
Approach One: Count the number of days that have elapsed from the start of
the year (January 0) until the target date. For example, March 23 consists of
31 + 29 + 23 days, since we need to add up all the days in January and February
and then the twenty-three days in March. This adds up to 83. We divide by 7.
Since 83 = 7 × 11 + 6, the remainder is 6. From the above table, we see that a
remainder of 6 corresponds to Monday.
Approach Two: Determine the day of the week for the first date of each month,
showing that if January 1 falls on Wednesday, then February 1 must be a Saturday,
March 1 must be a Sunday, and so on. From this, students can solve their problem
for any given date by adding or subtracting increments of seven. For example,
March 23 has to be the same date as March 16, March 9, and March 2. Thus,
March 23 has to be a Monday, since March 1 is a Sunday.
A clever approach combines these two paradigms, using the first date of each
month to determine the appropriate “shift”. For example, March 1 is 31 + 29 =
60 = 8×7+4 days after January 1, and so March 1 is “shifted” by 4 days compared
to January 1. Thus, if we know that the shift number of March is 4, then we can
determine the day of week of March 23 by adding the date to the shift number
(4 + 23 = 27 = 3 × 7 + 6), dividing the number by 7 and taking the remainder
(which is 6), and then reading the above table to conclude that the answer is
Monday.
For leap years such as the year 2020, the shift dates of each month are as follows:
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
0 3 4 0 2 5 0 3 6 1 4 6

Notice this table forms four sets of three digits that can be remembered this way:

034 = 52 + 32 , 025 = 52 , 036 = 62 , 146 = 122 + 22 .

For example, the shift number for June is 5, since the number of days until the
start of June is 31 + 29 + 31 + 30 + 31 = 152 = 21 × 7 + 5, which has a remainder
of 5 upon division by 7. In other words, June 1 is exactly 21 weeks and 5 days
after January 1 which implies that the shift for June is 5.
Let D be the date and S be the shift number. For example, June 15 would have
D = 15 and S = 5. To perform this algorithm in our head, we just need to

Crux Mathematicorum, Vol. 46(1), January 2020

 Richard Hoshino /15

add D + S, divide by 7, and the remainder gives us our answer to the Calendar
Problem. Since this remainder is 6, we can conclude that June 15, 2020 will be a
Monday.
Now let’s extend this by replacing the year 2020 with our birth year. In solving
this harder problem, we realize that each 365-day year contributes one extra day
(52 weeks plus 1 day) and each 366-day leap year contributes two extra days (52
weeks plus 2 days). Thus, if January 1, 2020 is a Wednesday, then January 1,
2019 was a Tuesday, since we have shifted back one day. And similarly, January
1, 2021 will be a Friday since we will need to shift forward two days.
In one of my school visits (in 2019), one student made the powerful insight that
her birthday in 2002 must be the same day of week as her birthday in 2019, since
there are 17 “extra days” in addition to the four Feb 29 “leap days” that occurred
in 2004, 2008, 2012, and 2016. Since 17 + 4 = 21, the calendar shifted 21 days
between her birthday in 2002 and her birthday in 2019. And since 21 is a multiple
of 7, if her birthday fell on a Tuesday in 2019, then it must have fallen on a
Tuesday in 2002. This student provided a clear method for how to handle the
tricky concept of leap years.
A different student from the same class observed that the calendar repeats itself
every 28 years, since each year contributes one extra day (52 weeks plus 1 day),
and there are 7 occurrences of February 29 during any 28-year period. Thus, the
calendar shifts by 28 + 7 = 35 days, which is a multiple of 7. This observation
enabled the student to determine the day of the week on which her parents were
born.
Through this process of solving the Calendar Problem and determining an algo-
rithm that works for any birthday, students demonstrate the four principles of the
Computational Thinking process.
(i) Decomposition: break down the problem into smaller tasks
(ii) Pattern recognition: identify similarities, differences, and patterns within
the problem
(iii) Abstraction: identify general principles and filter out unnecessary informa-
tion
(iv) Algorithmic design: identify and organize the steps needed to solve the prob-
lem
As mathematicians we use these four principles in our research endeavours, and
the Calendar Problem offers a challenge for enabling our students to have similar
experiences.
During the 2018-2019 sabbatical year, I worked with the Callysto Project, a
federally-funded initiative to bring computational thinking and mathematical prob-
lem solving into Grade 5-12 Canadian classrooms (www.callysto.ca). Through
my work with Callysto, I visited over a dozen schools and worked with 700+ stu-
dents, sharing rich math problems that incorporated the Callysto technology (a

Copyright © Canadian Mathematical Society, 2020

16/ Teaching Problems

web-based platform known as a Jupyter Notebook, freely accessible to anyone with

an Internet connection). I created a Notebook for the Calendar Problem, to be
used by teachers and students. This free resource, which also includes a lesson
plan for teachers, can be found at www.bit.ly/CallystoCalendar.

We end with three questions for consideration.

Communications, including solutions, concerning these questions are welcomed via

email at [email protected].

Question #1

Here is an algorithm that determines the correct day of week for any date in the
20th century (Jan 1, 1901 to Dec 31, 2000).

Let Y be the last two digits of the year, D be the day, and S be the “shift” value
according to the following table that is correct for non-leap years:
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
0 3 3 6 1 4 6 2 5 0 3 5

For example, the author’s birthday (June 15, 1978) has Y = 78, D = 15, and
S = 4.

Now calculate the sum T = Y + bY /4c + D + S.

If the year corresponds to a leap year (i.e., Y is a multiple of 4) and the month is
January or February, subtract 1 from T . (Why do we need to do this?)

Divide T by 7 and determine its remainder. The remainder tells us our answer:
0 Sunday
1 Monday
2 Tuesday
3 Wednesday
4 Thursday
5 Friday
6 Saturday

For example, October 29, 1929 has T = 29 + 7 + 29 + 0 = 65, which gives a

remainder of 2 when divided by 7. Therefore, this date in history (known as Black
Tuesday) was indeed a Tuesday.

Here is the question: why does this algorithm work?

Question #2

What day of the week would it be on your 100th birthday?

Crux Mathematicorum, Vol. 46(1), January 2020

 Richard Hoshino /17

Question #3
Create your own algorithm for other famous dates before the 20th century, and
apply it to the famous dates such as the following:
(i) July 1, 1867 (Confederation Day in Canada)
(ii) July 4, 1776 (Independence Day in the USA)
(iii) April 23, 1616 (Death of William Shakespeare)
(iv) September 30, 1207 (Birthday of Rumi)
Note that you will need to be careful about ensuring the correct calculation of leap
years, due to the quirky rules that occur when the year is a multiple of 100 but
not a multiple of 400. Specifically, the years 1600 and 2000 are leap years, while
the years 1700, 1800, 1900 are not leap years.

Richard Hoshino teaches at Quest University Canada in Squamish, BC. He can be

reached via email at [email protected].
Note: Submissions for consideration in Teaching Problems are welcomed. Please
feel free to send along a contribution concerning a valuable teaching example from
your experience. It is also appreciated if you can include some related problems for
consideration as has been done here. Our readers welcome opportunities to solve
problems.

Copyright © Canadian Mathematical Society, 2020

18/ OLYMPIAD CORNER

OLYMPIAD CORNER
No. 379
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, 2020.

OC461. Let A and B be two finite sets. Determine the number of functions
f : A → A with the property that there exist two functions g : A → B and
h : B → A such that g(h(x)) = x ∀x ∈ B and h(g(x)) = f (x) ∀x ∈ A.

OC462. The integers a1 , a2 , . . . , an satisfy

1 < a1 < a2 < . . . < an < 2a1 .

If m is the number of distinct prime factors of a1 a2 · . . . · an , then prove that

(a1 a2 · . . . · an )m−1 ≥ (n!)m .

OC463. A 6 × 6 table is filled with the integers from 1 to 36.

(a) Give an example of such a fill of the table so that the sum of every two numbers
in the same row or column is greater than 11.
(b) Prove that in some row or column, no matter how you fill the table, you will
always find two numbers whose sum does not exceed 12.

OC464. Given an acute triangle ABC with orthocenter H. The angle bisector
of ∠BHC intersects side BC at D. Let E and F be the symmetric points of D
with respect to lines AB and AC, respectively. Prove that the circumcircle of
triangle AEF passes through the midpoint G of arc BAC.

OC465. The sequence (an ) is defined by

p
a1 = 1, an = b 2an−1 + an−2 + · · · + a1 c if n > 1.

Find a2017 .

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

Crux Mathematicorum, Vol. 46(1), January 2020

 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 2020.
La rédaction souhaite remercier Rolland Gaudet, professeur titulaire à la retraite à
l’Université de Saint-Boniface, d’avoir traduit les problèmes.

OC461. Soient A et B deux ensembles finis. Déterminer le nombre de

fonctions f : A → A telles qu’il existe deux fonctions g : A → B et h : B → A
pour lesquelles g(h(x)) = x ∀x ∈ B et h(g(x)) = f (x) ∀x ∈ A.

OC462. Les entiers a1 , a2 , . . . , an an satisfont à

1 < a1 < a2 < . . . < an < 2a1 .

Si m est le nombre de facteurs premiers distincts de a1 a2 · . . . · an , démontrer que

(a1 a2 · . . . · an )m−1 ≥ (n!)m .

OC463. Les cases d’une matrice de taille 6 × 6 sont remplies par les entiers
de 1 à 36.
(a) Déterminer une manière de remplir les cases de la matrice de façon à ce que la
somme de deux quelconques nombres d’une même rangée ou colonne donne
toujours supérieur à 11.
(b) Quelle que soit la façon de remplir les cases de la matrice, démontrer qu’au
moins une somme de deux éléments de même rangée ou colonne sera inférieure
ou égale à 12.

OC464. Soit ABC un triangle acutangle avec orthocentre H. La bissectrice

de ∠BCH intersecte le côté BC en D. Soient E et F less points symétriques à
D par rapport aux lignes AB et AC, respectivement. Démontrer que le cercle
circonscrit du triangle AEF passe par le mi point de l’arc BAC.

OC465. La suite (an ) est définie par

p
a1 = 1, an = b 2an−1 + an−2 + · · · + a1 c if n > 1.

Déterminer a2017 .

Copyright © Canadian Mathematical Society, 2020

20/ OLYMPIAD CORNER

OLYMPIAD CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2019: 45(6), p. 320–321.

OC436. In a non-isosceles triangle ABC, let O and I be its circumcenter

and incenter, respectively. Point B 0 , which is symmetric to point B with respect
to line OI, lies inside ∠ABI. Prove that the tangents to the circumcircle of the
triangle BB 0 I at points B 0 and I intersect on the line AC.
Originally Russia MO, 8th Problem, Grade 10, Final Round 2017 (Geometry).
We received 3 correct submissions. We present all three solutions.
Solution 1, by Lee Jang Yong

Let ω be the circumcircle of 4ABC. Let A0 , B 0 , and C 0 be the points that are
symmetric with respect to line OI to A, B, and C, respectively. Because the
symmetry line, OI, passes through the centre of ω we have that the symmetric
images A0 , B 0 , and C 0 belong to ω, as well.
Let D be the intersection of AC with the tangent at I to the circumcircle of
4BB 0 I. We show that DI = DB 0 .
Let D0 be the point symmetric to D with respect to line OI. The centre of the
circumcircle of triangle BB 0 I lies on OI, given that 4BB 0 I is isosceles. Therefore
the tangent ID is perpendicular to OI, and I, D, and D0 are co-linear with I
being the middle point of DD0 . Moreover, if we extend DD0 to intersect ω at M
and M 0 we find that I is the middle of the new segment, M M 0 .
We are in the setting of the butterfly theorem. Let X be the intersection of A0 I
with ω, and Z be the intersection of C 0 I with ω. D0 is a point on A0 C 0 , D, I, and
D0 are co-linear, and I is the middle point of DD0 . Due to the butterfly theorem
we conclude that D belongs to XZ.

Crux Mathematicorum, Vol. 46(1), January 2020

 OLYMPIAD CORNER /21

However, XZ is the perpendicular bisector of B 0 I, so DI = DB 0 . This is because I

is the centre of the incircle and the intersection of the bisectors of 4A0 B 0 C 0 . These
properties imply the equality of the following ∠B 0 IX = ∠IB 0 X = (∠C 0 A0 B 0 +
∠C 0 B 0 A0 )/2 and that 4B 0 IX is isosceles. Similarly, 4B 0 IZ is isosceles.
Since DI = DB 0 it follows that DB 0 is tangent to the circumcircle of 4BB 0 I
at B 0 , and that the tangents to the circumcircle of 4BB 0 I at points B 0 and I
intersect on the line AC at the point D.

Solution 2, by Ivko Dimitrić.

Consider, without loss of generality, triangle ABC in the plane of complex num-
bers, whose circumcircle is the unit circle centered at the origin O. For any point
included in our proof we associate a unique capital letter and a complex number.
The capital letter is used to refer to the complex number, as well. Let A = eiα ,
B = eiβ , and C = eiγ be the complex numbers that identify the triangle vertices,
with 0 < α < β < γ < 2π. Moreover, denote by a = eiα/2 , b = −eiβ/2 , c = eiγ/2
so that ā = 1/a, b̄ = 1/b, c̄ = 1/c, and A = a2 , B = b2 , C = c2 . Then,
a+b+c
I = −(ab + bc + ca) and I=− (1)
abc
(see p. 262 of M. Bataille’s article in Crux Mathematicorum, Vol 45:5 (May 2019)).
In general, the orthogonal projection S of a point X to a line P Q is given by
Ç å
1 PQ − PQ Q−P
S= +X + X (2)
2 Q−P Q−P
←→
and the point Y symmetric to X with respect to P Q is
PQ − PQ Q − P
Y = 2S − X = + X. (3)
Q−P Q−P
Thus, when P = O, Q = I and X = B we get B 0 = (I/I) B.
 
Let D = 12 (I + B 0 ) = I2 1 + BI be the midpoint of segment B 0 I and let K be the
center of the circumcircle of BB 0 I. The perpendicular bisector of B 0 I consists of
points Z for which
Z −D Z −D
Å ã
=− ,
I − B0 I − B0
yielding
ñ Ç åô Å ã ñ Å ãô Ç å
I B B I B B
Z− 1+ I 1− + Z− 1+ I 1− = 0,
2 I I 2 I I

After multiplying out, simplifying, and dividing by II the equation of the bisector,
KD, reduces to Ç å
1
Å ã
B Z B Z
1− + 1− =1− . (4)
I I I I II

Copyright © Canadian Mathematical Society, 2020

22/ OLYMPIAD CORNER

Since KB 0 = KI and KD is the angle bisector of ∠B 0 KI, the two tangents to the
circumcircle of BB 0 I at points B 0 and I intersect on the bisector KD. Therefore,
to prove that the two tangents intersect on the line AC it suffices to show that
the lines DK, AC and the perpendicular to IO at I intersect at one point.

The line perpendicular to IO at I is the locus of points Z such that Z − I is a real

multiple of i(I − O), i. e.

Z −I Z −I
=− ⇐⇒ IZ + IZ = 2II. (5)
I I

The line through arbitrary two points P and Q has an equation

(Q − P )Z − (Q − P )Z = P Q − P Q, (6)

so that the line AC through A = a2 and C = c2 is

Z + c2 a2 Z = a2 + c2 . (7)

Using (1) and combining (5) and (7), we find the intersection of the perpendicular
to IO at I and the line AC to be the point Z whose affix satisfies

2II − (a2 + c2 ) I (a + b + c)(a + 2b + c)

Z= = , (8)
I −c a I
2 2 ca (ca − b2 )

by factoring out a + c on the top and the bottom. Consequently,

(ab + bc + ca)(ab + bc + 2ca) I(ca − I)

Z= = . (9)
b − ca
2 ca − b2

It can be now shown that this point satisfies the equation (4) so the lines DK, AC
and IZ are concurrent at Z. Namely, using (1) we compute

1 (a + c)(ab + bc + ca + b2 ) a+c I −B
1− = =− · ,
II (a + b + c)(ab + bc + ca) abc I I

so that
Ç å
1 −I 1
Å ã Å ã Å ã
B abc I B
1− =− 1− and 1− = 1− .
I a+c II I b(a + c) II

Also, from (9) we get

Z ca − I Z b2 (ca I − 1)
= and = .
I ca − b2 I ca − b2

Crux Mathematicorum, Vol. 46(1), January 2020

 OLYMPIAD CORNER /23

Substituting these into the left-hand side of (4) we get

ñ ôÅ
abc I ca − I b2 (1 − ca I) 1
ã
I
− + 1−
a + c ca − b2 b(a + c) ca − b2 II
1 1
Å ã
2 2
= [−c a b I + abc II + bI − abc II ] 1 −
(a + c)(ca − b2 ) II
1 1
Å ã
= [ca(a + b + c) − b(ab + bc + ca) ] 1 −
(a + c)(ca − b2 ) II
1 1
Å ã
= [ca(a + c) − b2 (a + c) ] 1 −
(a + c)(ca − b2 ) II
1
=1− .
II
This verifies the equation (4) and proves the claim.

Solution 3, by Andrea Fanchini.

We use barycentric coordinates with reference to 4ABC. The line

IO : bc(cSC − bSB )x + ac(aSA − cSC )y + ab(bSB − aSA )z = 0
has infinite perpendicular point
IO∞⊥ (a(b − c) : b(c − a) : c(a − b)).
Therefore, the tangent to the circumcircle of 4BB 0 I at point I is given by
IIO∞⊥ : bc(b + c − 2a)x + ac(a + c − 2b)y + ab(a + b − 2c)z = 0.
Point B 0 , which is the symmetric image of point B with respect to line OI is
identified by
B 0 2a(s − b)(c − a)(c − b) : b2 (b − a)(b − c) : 2c(s − b)(b − a)(c − a) .

Copyright © Canadian Mathematical Society, 2020

24/ OLYMPIAD CORNER

Therefore, the tangent to the circumcircle of 4BB 0 I at B 0 is given by

B 0 X :b2 c(b − a)(b − c)(b + c − 2a)x − 2ac(s − b)(c − a)2 (a + c − 2b)y

+ ab2 (b − a)(b − c)(a + b − 2c)z = 0.

In conclusion, the two tangent lines intersect at the point identified by

X = IIO∞⊥ ∩ B 0 X = (a(2c − a − b) : 0 : c(b + c − 2a)) .

Clearly, this point lies on the line AC, since its second coordinate is 0.

OC437. The magician and his helper have a deck of cards. The cards all
have the same back, but their faces are coloured in one of 2017 colours (there are
1000000 cards of each colour). The magician and the helper are going to show the
following trick. The magician leaves the room; volunteers from the audience place
n > 1 cards in a row on a table, all face up. The helper looks at these cards, then
he turns all but one card face down (without changing their order). The magician
returns, looks at the cards, points to one of the face-down cards and states its
colour. What is the minimum number n such that the magician and his helper
can have a strategy to do the magic trick successfully?
Originally Russia MO, 4th Problem, Grade 11, Final Round 2017 (Game Theory).
No solutions were received.

OC438. A teacher gives the students a task of the following kind. He

informs them that he thought of a monic polynomial P (x) of degree 2017 with
integer coefficients. Then he tells them k integers n1 , n2 , . . . , nk and the value of
the expression P (n1 )P (n2 ) · . . . · P (nk ). According to these data, the students
should then find teacher’s polynomial. Find the smallest k for which the teacher
can compose such a problem so that the polynomial found by the students must
necessarily coincide with the one he thought of.
Originally Russia MO, 3rd Problem, Grade 11, Regional Round 2017 (Algebra).
No solutions were received.

OC439. Let (G, ·) be a group and let m and n be two nonzero natural numbers
that are relatively prime. Prove that if the functions f : G → G, f (x) = xm+1
and g : G → G, g(x) = xn+1 are surjective endomorphisms, then the group G is
abelian.
Originally Romania MO, 2nd Problem, Grade 12, District Round 2017 (Abstract
Algebra).
We received 2 correct submissions. We present the solution by Oliver Geupel.
Independently, Corneliu Manescu-Avram submitted a similar solution.

Crux Mathematicorum, Vol. 46(1), January 2020

 OLYMPIAD CORNER /25

Let a be an arbitrary element of G. Since f is a surjective endomorphism, we

deduce that, for every b ∈ G, there is a c ∈ G such that b = f (c), and it holds
am b = a−1 f (a)f (c) = a−1 f (ac) = a−1 (ac)m+1 = a−1 (ac) . . . (ac) = (ca) . . . (ca)a−1
= (ca)m+1 a−1 = f (ca)a−1 = f (c)f (a)a−1 = bam .
Hence, am commutes with every element of the group. Similarly, an commutes
with every element of the group.
It is well-known and easy to verify that the set of elements of a group, G, that
commute with every element of G is a subgroup of G, called the centre Z(G) of
the group. Thus, for integers q and r, we have amq+nr ∈ Z(G). Since m and n are
co-prime, integers q and r can be chosen such that mq + nr = 1. Consequently,
Z(G) = G, that is, G is abelian.

OC440. Let f : [a, b] → [a, b] be a differentiable function with continuous

and positive first derivative. Prove that there exists c ∈ (a, b) such that
f (f (b)) − f (f (a)) = (f 0 (c))2 (b − a).

Originally Romania MO, 4th Problem, Grade 11, Final Round 2017.
We received 4 correct submissions. We present the solution by Ivko Dimitrić. Simi-
lar solutions were submitted independently by Brian Bradie and Corneliu Manescu-
Avram.
Since f ([a, b]) ⊂ [a, b] and f is increasing and differentiable, the Mean Value Theo-
rem for f applied to the interval [f (a), f (b)] guarantees the existence of a number
q, a ≤ f (a) < q < f (b) ≤ b, such that
f (f (b)) − f (f (a)) = f 0 (q) (f (b) − f (a)).
Another application of the same theorem on the interval [a, b] tells us that
f (b) − f (a) = f 0 (p) (b − a)
for some number p, a < p < b. Combining the two formulas we get
f (f (b)) − f (f (a)) = f 0 (p)f 0 (q) (b − a), (1)
where p, q ∈ (a, b).
Next, we can assume that f 0 (p) ≤ f 0 (q). Since f 0 is positive we have
»
f 0 (p) ≤ f 0 (p)f 0 (q) ≤ f 0 (q).
Then, the value f 0 (p)f 0 (q) is between f 0 (p) and f 0 (q). Since f 0 is continuous,
p

by the Intermediate Value Theorem for f 0 on the interval [p, q], there exists c ∈
0
p
[p, q] ⊂ (a, b) such that f (c) = f (p)f 0 (q). Combining with (1)
0

f (f (b)) − f (f (a)) = (f 0 (c))2 (b − a),

and the statement follows.

Copyright © Canadian Mathematical Society, 2020

26/ Focus On... Introducing SA , SB , SC in Barycentric Coordinates

FOCUS ON...
No. 39
Michel Bataille
Introducing SA , SB , SC in Barycentric Coordinates
Introduction
The use of barycentric coordinates relative to a triangle ABC is quite appropriate
when solving problems involving affine properties such as collinearity of points,
concurrency of lines or even ratio of areas, but does not seem adapted to euclidean
properties such as lengths or perpendicularity. However, if a = BC, b = CA, c =
AB, a few results linked to the numbers

b2 + c2 − a2 c2 + a2 − b2 a2 + b2 − c2
SA = , SB = , SC =
2 2 2
(Conway’s triangle notation) sometimes lead to a simple coordinate solution to
euclidean problems. Besides, this is less surprising once one has remarked that
−−→ −→ −−→ −−→ −→ −−→
SA , SB , SC are nothing but the dot products AB · AC, BC · BA, CA · CB,
respectively!
After a paragraph offering useful relations concerning SA , SB , SC , we will present
some examples of situations that can prompt a resort to these numbers.

Becoming more familiar with SA , SB , SC

The obvious equalities

SB + SC = a2 , SB − SC = c2 − b2 ,
SC + SA = b2 , SC − SA = a2 − c2 ,
SA + SB = c2 , SA − SB = b2 − a2

are of constant use and should be kept in mind from now on!
Other interesting, readily checked relations are

c2 SC − b2 SB = (b2 − c2 )SA ,
a2 SA − c2 SC = (c2 − a2 )SB ,
b2 SB − a2 SA = (a2 − b2 )SC

and, denoting by s the semiperimeter of ∆ABC,

cSC − bSB = 2s(s − a)(b − c),

aSA − cSC = 2s(s − b)(c − a),
bSB − aSA = 2s(s − c)(a − b).

Crux Mathematicorum, Vol. 46(1), January 2020

 Michel Bataille /27

A connection to the area F of the triangle ABC is obtained with

SB SC + a2 SA = SC SA + b2 SB = SA SB + c2 SC = 4F 2

and
2(SB SC + SC SA + SA SB ) = a2 SA + b2 SB + c2 SC = 8F 2 .

These formulas are easily proved with the help of the known

16F 2 = 2(a2 b2 + b2 c2 + c2 a2 ) − (a4 + b4 + c4 ).

See also exercise 1 for more relations.

The coordinates of O and H

The numbers SA , SB , SC prove very useful when the barycentric coordinates of
the circumcentre O and the orthocenter H of ∆ABC are needed. In terms of
sidelengths and angles of ∆ABC, the coordinates are known to be

O = (a cos A : b cos B : c cos C)

and
H = (a cos B cos C : b cos C cos A : c cos A cos B).

Observing that for example

b2 + c2 − a2 SA bSB cSC SB SC
a cos A = a · = a2 and a cos B cos C = a · · =
2bc abc abc abc abc
we obtain that

O = (a2 SA : b2 SB : c2 SC ), H = (SB SC : SC SA : SA SB ).

These coordinates readily yield those of the centre N of the nine-point circle: since
N is the midpoint of OH, we have

(8F 2 )2N = (8F 2 )O + (8F 2 )H

= (a2 SA + 2SB SC )A + (b2 SB + 2SC SA )B + (c2 SC + 2SA SB )C

hence

N = (2SB SC + a2 SA : 2SC SA + b2 SB : 2SA SB + c2 SC )

= (SB SC + 4F 2 : SC SA + 4F 2 : SA SB + 4F 2 ),

a result to be used in our first example, problem OC 311 [2017 : 12 ; 2018 : 102]:
Let ∆ABC be an acute-scalene triangle, and let N be the center of
the circle which passes through the feet of the altitudes. Let D be the
intersection of the tangents to the circumcircle of ∆ABC at B and C.
Prove that A, D and N are collinear if and only if ∠BAC = 45◦ .

Copyright © Canadian Mathematical Society, 2020

28/ Focus On... Introducing SA , SB , SC in Barycentric Coordinates

Clearly, N is the point above and the point D lies on the symmedian through A
of ∆ABC (a well-known result), hence it is sufficient to prove that N is on this
symmedian if and only if ∠BAC = 45◦ .
With the previous notations, the symmedian point K is (a2 : b2 : c2 ) and so the
equation of the symmedian AK is c2 y − b2 z = 0.
Therefore N is on the symmedian AD if and only if

c2 (SC SA + 4F 2 ) = b2 (SA SB + 4F 2 ),

which successively rewrites as

SA (c2 SC − b2 SB ) = 4F 2 (b2 − c2 )
2
SA = 4F 2 (since ABC is scalene)
(b + c − a ) = 2a2 b2 + 2b2 c2 + 2c2 a2 − a4 − b4 − c4
2 2 2 2

Å 2 ã2
b + c2 − a2 1
=
2bc 2
1
cos2 (∠BAC) =
2
and since ABC is acute-angled, the latter means that ∠BAC = 45◦ .
As a second example, we next present an alternative solution to problem 4258
[2017 : 265 ; 2018 : 270]:
Let ABC be an acute triangle with circumcentre O, orthocentre H,
D ∈ BC, AD ⊥ BC, E ∈ AC, BE ⊥ AC. Define points F and G
to be the fourth vertices of parallelograms CADF and CBEG. If X
is the midpoint of F G, and Y is the point where XC intersects the
circumcircle again, prove that AHBY is a parallelogram.

Crux Mathematicorum, Vol. 46(1), January 2020

 Michel Bataille /29

Keeping the notations of this paragraph, we have

a2 D = (SC )B + (SB )C, b2 E = (SC )A + (SA )C

and, since G = E + C − B and F = D + C − A, we then deduce

b2 G = (SC )A − b2 B + (b2 + SA )C

and
a2 F = −a2 A + (SC )B + (a2 + SB )C.

A simple calculation then yields the midpoint X of F G:

2a2 b2 X = b2 (a2 F ) + a2 (b2 G)

= (−a2 SA )A + (−b2 SB )B + (2a2 b2 + b2 SB + a2 SA )C.

Now, noticing that the equation of the line CO is (b2 SB )x − (a2 SA )y = 0, we see
that X is on CO. It follows that the line CY is a diameter of the circumcircle and
therefore CB ⊥ BY and CA ⊥ AY . Thus, BY k AH and AY k BH and AHBY
is a parallelogram.

About perpendiculars
We shall illustrate the following result: If (f : g : h) is the infinite point of the line
`, then the infinite point (f 0 : g 0 : h0 ) of the perpendiculars to ` is given by

f 0 = gSB − hSC , g 0 = hSC − f SA , h0 = f SA − gSB .

We quickly repeat the known proof for completeness. Expressing that the vectors
−−→ −→ −−→ −→
g AB + hAC and g 0 AB + h0 AC are orthogonal yields
−−→ −→ −−→ −→
0 = (g AB + hAC) · (g 0 AB + h0 AC) = g 0 (gc2 + hSA ) + h0 (gSA + hb2 ).

Since f + g + h = f 0 + g 0 + h0 = 0, we obtain
g0 −h0 f0
= = ,
gSA + hb2 hSA + gc2 −gSA − hb2 + hSA + gc2
that is,
f0 g0 h0
= = .
gSB − hSC hSC − f SA f SA − gSB

To see this at work through a simple example, consider the line BC whose point
at infinity is (0 : 1 : −1). The point at infinity of the perpendiculars to BC then
is (SB + SC : −SC : −SB ) = (−a2 : SC : SB ). It follows that the equation of the
perpendicular bisector δA of BC is

x −a2 0
y SC 1 = 0,
z SB 1

Copyright © Canadian Mathematical Society, 2020

30/ Focus On... Introducing SA , SB , SC in Barycentric Coordinates

that is, x(c2 − b2 ) − a2 y + a2 z = 0.

With the help of this result, we can offer a variant of solution to the following
problem extracted from 3910 [2014 : 30 ; 2015 : 42]:

Two triangles ABC and A0 B 0 C 0 are homothetic. Show that if B 0 and

C 0 are on the perpendicular bisectors of CA and AB respectively, then
A0 is on the perpendicular bisector of BC.

From the above equation of δA , we get cyclically the equations of δB and δC , the
perpendicular bisectors of CA and AB:

b2 x + (a2 − c2 )y − b2 z = 0 and c2 x − c2 y + (a2 − b2 )z = 0.

Now, let Ω = (α : β : γ), with α + β + γ = 1, be the centre of the homothety h

transforming A, B, C into A0 , B 0 , C 0 , respectively. If λ denotes the factor of h, then
A0 −Ω = λ(A−Ω), hence A0 = (λ+(1−λ)α : (1−λ)β : (1−λ)γ). Similarly, we have
B 0 = ((1−λ)α : λ+(1−λ)β : (1−λ)γ) and C 0 = ((1−λ)α : (1−λ)β : λ+(1−λ)γ).

Expressing that B 0 and C 0 are on the lines δB and δC , respectively, we obtain

λ(a2 − c2 ) + (1 − λ)(αb2 + β(a2 − c2 ) − γb2 ) = 0 (1)

and
λ(a2 − b2 ) + (1 − λ)(αc2 − βc2 + γ(a2 − b2 )) = 0 (2).

The difference (2) − (1) gives λ(c2 − b2 ) + (1 − λ)(α(c2 − b2 ) − βa2 + γa2 ) = 0,

which implies that A0 is on δA , as desired.

Our second example gives a solution to problem 4313 [2018 : 71 ; 2019 : 93]:

Let I be the incenter of triangle ABC, and denote by Ha , Hb and Hc

the orthocenters of triangles IBC, ICA and IAB, respectively. Prove
that triangles ABC and Ha Hb Hc have the same area.

Let I = (a : b : c) be the incentre of ∆ABC. The equation of the line IB then is

cx − az = 0, its point at infinity is (a : −(a + c) : c) and so the perpendicular to
IB through C is
x −ca2 − aSB 0
y cSC − aSA 0 = 0,
z ac2 + cSB 1

that is,
(cSC − aSA )x + (ca2 + aSB )y = 0.

Similarly, the perpendicular to IC through B is (bSB − aSA )x + (ba2 + aSC )z = 0.

This provides their point of intersection Ha : Ha = (a : c − a : b − a). Cyclically,
we obtain Hb = (c − b : b : a − b) and Hc = (b − c : a − c : c). It follows that

Crux Mathematicorum, Vol. 46(1), January 2020

 Michel Bataille /31

Area(Ha Hb Hc ) = |δ|Area(ABC) where

a c−b b−c
1
δ= · c−a b a−c
(b + c − a)(c + a − b)(a + b − c)
b−a a−b c
1
= · δ0 .
8(s − a)(s − b)(s − c)

It is not difficult to check that δ 0 = 8(s−a)(s−b)(s−c) and the conclusion follows.

For another illustration of this paragraph and the previous one, we refer the reader
to my solution to problem 3878 [2013 : 371 ; 2014 : 359].
As usual, we end the number with a series of exercises.

Exercises
1. (Adapted from problem 11958 of The American Mathematical Monthly) Prove
the relations

a4 SA + b4 SB + c4 SC − 3SA SB SC = 2(a2 + b2 + c2 )F 2 = SA SB SC + a2 b2 c2

and deduce a condition on a, b, c for the nine-point centre N to lie on the circum-
circle of ∆ABC.
2. Use SA , SB , SC to show that O, H and the incenter I are collinear if and only
if the triangle ABC is isosceles.
3. Find the point at infinity of the perpendiculars to OI, where O and I are the
circumcentre and the incentre of a scalene triangle ABC.
4. If M1 = (x1 : y1 : z1 ), M2 = (x2 : y2 : z2 ) with x1 + y1 + z1 = x2 + y2 + z2 = 1,
show that

M1 M22 = SA (x2 − x1 )2 + SB (y2 − y1 )2 + SC (z2 − z1 )2 .

Copyright © Canadian Mathematical Society, 2020

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, 2020.

4501. Proposed by Vaclav Konecny, modified by the Board.

Given the rectangle whose vertices have Cartesian coordinates A(0, b), B(0, 0),
C(a, 0), D(a, b), find the equation of the locus of points P (x, y) in the third quad-
rant (with x, y < 0) for which ∠BP A = ∠CP D.
Comment from the proposer: this problem was inspired by problem #4301 in Crux
44(1) proposed by Bill Sands.

4502. Proposed by George Apostolopoulos.

Let a, b, c be the side lengths of triangle ABC with inradius r and circumradius
R. Prove that
3 r X a 3 R
· ≤ ≤ · .
2 R 2a + b + c 8 r
cyclic

4503. Proposed by Michel Bataille.

Let ABC be a triangle with ∠BAC = 90◦ and let Γ be the circle with centre B
and radius BC. A circle γ passing through B and A intersects Γ at X, Y with
X 6= Y . Let E and F be the orthogonal projections of X and Y onto CY and
CX, respectively. Prove that the line CA bisects EF .

4504. Proposed by Warut Suksompong.

Find all positive integers (a, b, c, x, y, z), a ≤ b ≤ c and x ≤ y ≤ z, for which the
following two equations hold:

a + b + c = xy + yz + zx,
x + y + z = abc.

4505. Proposed by Miguel Ochoa Sanchez and Leonard Giugiuc.

Let ABCD be a convex quadrilateral such that AC ⊥ BD and AB = BC. Let I
denote the point of intersection of AC and BD. A straight line l1 passes through
I and intersects BC and AD in R and S, respectively. Similarly, straight line l2
passes through I and intersects AB and CD in M and N , respectively. The lines
M S and RN intersect AC at P and Q, respectively. Prove that IP = IQ.

Crux Mathematicorum, Vol. 46(1), January 2020

 Problems /33

4506. Proposed by D. M. Bǎtineţu-Giurgiu and Neculai Stanciu.

an+1
Let (an ) be a sequence of positive real numbers such that lim = a, where
n→∞ nan
∗
a ∈ R+ . Compute
n
1X k
lim √ .
n→∞ n k a
k
k=2

4507. Proposed by Eduardo Silva.

Suppose that a0 < · · · < ar are integers. If {bi } are distinct integers with ai ≤ bi ,
for each i, and σ is a permutation so that bσ(0) < · · · < bσ(r) , prove that ai ≤ bσ(i)
for each i. Further, if aj = bσ(j) for some j, then σ(j) = j, so that aj = bj .

4508. Proposed by Hung Nguyen Viet.

Let x, y, z be nonzero real numbers such that x + y + z = 0. Find the minimum
possible value of
1 1 1
Å ã
2 2 2
(x + y + z ) + 2+ 2 .
x2 y z

4509. Proposed by Leonard Giugiuc and Dan Stefan Marinescu.

Let B and C be two distinct fixed points that lie in the plane α and let M
be the midpoint of BC. Find the locus of points A ∈ α, A ∈ / BC, for which
4R · AM = AB 2 + AC 2 , where R is the circumradius of ABC.

4510. Proposed by Leonard Giugiuc and Daniel Sitaru.

Let ABC be a non-obtuse triangle. Prove that
√
cos A cos B + cos A cos C + cos B cos C > 2 cos A cos B cos C.

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

Copyright © Canadian Mathematical Society, 2020

34/ Problems

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

offrir des solutions, commentaires ou généralisations aux problèmes
proposś dans cette section.

Pour faciliter l’examen des solutions, nous demandons aux lecteurs de les faire parvenir
au plus tard le 15 mars 2019.
La rédaction souhaite remercier Rolland Gaudet, professeur titulaire à la retraite à
l’Université de Saint-Boniface, d’avoir traduit les problèmes.

4501. Proposé par Vaclav Konecny, modifié par le conseul.

Pour un rectangle dont les sommets ont les coordonnées cartésiennes A(0, b),
B(0, 0), C(a, 0) et D(a, b), déterminer l’équation vérifiée par les points dans le
troisième quadrant (où x, y < 0) pour lesquels ∠BP A = ∠CP D.
Note du proposeur: ce problème tire son inspiration du problème # 4301 dans
Crux 44(1), proposé par Bill Sands.

4502. Proposé par George Apostolopoulos.

Soient a, b et c les longueurs des côtés du triangle ABC, dont les rayons des cercles
inscrit et circonscrit sont r et R respectivement. Démontrer que
3 r X a 3 R
· ≤ ≤ · .
2 R 2a + b + c 8 r
cyclic

4503. Proposé par Michel Bataille.

Soit ABC un triangle tel que ∠BAC = 90◦ et soit Γ le cercle de centre B et rayon
BC. Un cercle γ, passant par B et A, intersecte Γ en X et Y où X 6= Y . Soient
E et F les projections orthogonales de X et Y vers CY et CX, respectivement.
Démontrer que la ligne CA bissecte EF .

4504. Proposé par Warut Suksompong.

Déterminer tous les entiers positifs (a, b, c, x, y, z), a ≤ b ≤ c et x ≤ y ≤ z, pour
lesquels les deux équations suivantes tiennent:
a + b + c = xy + yz + zx,
x + y + z = abc.

4505. Proposé par Miguel Ochoa Sanchez et Leonard Giugiuc.

Soit ABCD un quadrilatère convexe tel que AC ⊥ BD et AB = BC. Dénotons
par I le point d’intersection de AC et BD. Une ligne l1 passe par I et intersecte

Crux Mathematicorum, Vol. 46(1), January 2020

 Problems /35

BC et AD en R et S respectivement. De façon similaire, la ligne l2 passe par

I et intersecte AB et CD en M et N , respectivement. Les lignes M S et RN
intersectent AC en P et Q respectivement. Démontrer que IP = IQ.

4506. Proposé par D. M. Bǎtineţu-Giurgiu et Neculai Stanciu.

an+1
Soit (an ) une suite de nombres réels positifs telle que lim = a, où a ∈ R∗+ .
n→∞ nan
Calculer
n
1X k
lim √ .
n→∞ n k a
k
k=2

4507. Proposé par Eduardo Silva.

Supposons que a0 < · · · < ar sont des entiers. Si {bi } sont des entiers distincts
tels que ai ≤ bi pour tout i, et σ est une permutation telle que bσ(0) < · · · < bσ(r) ,
démontrer que ai ≤ bσ(i) pour tout i. De plus, si aj = bσ(j) pour un certain j,
alors σ(j) = j, d’où aj = bj .

4508. Proposé par Hung Nguyen Viet.

Soient x, y, z des nombres réels non nuls tels que x + y + z = 0. Déterminer la
valeur minimale de
1 1 1
Å ã
2 2 2
(x + y + z ) + 2+ 2 .
x2 y z

4509. Proposé par Leonard Giugiuc et Dan Stefan Marinescu.

Soient B et C des points distincts dans le plan α et soit M le mi point de BC.
Déterminer le lieu des points A ∈ α, A ∈/ BC, pour lesquels 4R·AM = AB 2 +AC 2 ,
où R est le rayon du cercle circonscrit de ABC.

4510. Proposé par Leonard Giugiuc et Daniel Sitaru.

Soit ABC un triangle non obtus. Démontrer que
√
cos A cos B + cos A cos C + cos B cos C > 2 cos A cos B cos C.

Copyright © Canadian Mathematical Society, 2020

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 2019: 45(6), p. 346–349.

4451. Proposed by Michel Bataille.

For n ∈ N with n ≥ 2 and 0 < a < b < 1, let
Z b
(x + 1)((2n − 3)xn+1 − (2n − 1)xn + 3x − 1)
I(a, b) = dx.
a x2 (x − 1)2

Find
1
Å ã
lim+ + lim− I(a, b) .
a→0 a b→1

There were 15 correct solutions. We present the standard approach taken by most
of the solvers.
Note that

(2n−3)xn+1 −(2n−1)xn +3x−1 = (x−1)2 [(2n−3)xn−1 +(2n−5)xn−2 +· · ·+3x2 +x−1],

and

(x + 1)[(2n − 3)xn−1 + (2n − 5)xn−2 + · · · + 3x2 + x − 1]

= (2n − 3)xn + 4(n − 2)xn−1 + · · · + 4x2 − 1.

Therefore, the integrand is equal to

n−1
X 1
(2n − 3)xn−2 + 4 (k − 1)xk−2 − .
x2
k=2

and its antiderivative is

n−1
2n − 3 1
Å ã X
f (x) ≡ xn−1 + 4 xk−1 + .
n−1 x
k=2

Now
2n − 3 4n2 − 9n + 4
lim− f (b) = + 4(n − 2) + 1 = .
b→1 n−1 n−1
Therefore
ï 2 ò "Å n−1
#
1 4n − 9n + 4 2n − 3 n−1
ã X
n−1
+ lim I(a, b) = − a +4 a ,
a b→1− n−1 n−1
k=2

Crux Mathematicorum, Vol. 46(1), January 2020

 Solutions /37

whence
1 4n2 − 9n + 4
Å ã
lim+ + lim− I(a, b) = .
a→0 a b→1 n−1

Comment from the editor. This is essentially the solution supplied by all the
solvers. However, there were interesting aspects to the various manipulations.
While most used long division to find the cofactor of (x − 1)2 in the factorization
of
Pn (x) ≡ (2n − 3)xn+1 − (2n − 1)xn + 3x − 1,

Michel Bataille and Ivko Dimitrić relied on the respective recursions

Pn (x) = (2n − 3)xn−1 (x − 1)2 + Pn−1

and
Pn+1 (x) = xPn (x) + (x − 1)[2x(xn − 1) − (x − 1)]

to find the cofactor by an induction argument. However, Paul Bracken did not
need to bother with this, since, by some alchemy, he produced the antiderivative

1 2n − 3 2n − 1
ïÅ ã Å ã ò
n+1 n
x + x − 3x − 1 .
x(x − 1) n−1 n−1

4452. Proposed by Mihaela Berindeanu.

Let ABC be a triangle with orthocenter H. If A0 , B 0 , C 0 are the circumcenters of

−−→ −−→ −−→ →
−
M HBC, M HAC and M HAB, respectively, and AA0 + BB 0 + CC 0 = 0 , show
that ABC is an equilateral triangle.

We received 11 submissions, of which 9 were complete and correct. We present a

solution which combines parts of those submitted by Sorin Rubinescu and Alexan-
dru Pı̂rvuceanu and by Ivko Dimitrić.

Assume 4ABC is not a right triangle, as the problem is ill-posed in such a case.
Let O be the circumcenter of 4ABC and R its circumradius (see figure below).

First we show that A0 is the reflection of O over the side BC. Let M be the
midpoint of BC and A00 the reflection of O with respect to M . The relationship
between central and inscribed angles subtending the same arc gives us ∠BOC =
2∠A. Since the diagonals in quadrilateral BOCA00 are perpendicular and bisect
each other, BOCA00 is a rhombus. Thus,

∠BA00 C = ∠BOC = 2∠A.

Copyright © Canadian Mathematical Society, 2020

38/ Solutions

It is easy to see (and well-known) that ∠BHC = 180◦ − ∠A. Extend the segment
AH until it intersects the circumcircle of 4BHC, and denote the intersection by
T . Since the points B, H, C, T are concyclic, we have
∠BT C = 180◦ − ∠BHC = ∠A.
Thus we have shown ∠BA00 C = 2∠BT C, so A00 must be the centre of the circum-
circle of 4BHC; that is, A00 = A0 . In particular, BOCA0 is a rhombus.
We thus have −−→0 −→ −−→0 −→ −−→ −−→
AA = AO + OA = −OA + OB + OC.
Similar arguments show that OAC 0 B and OCBA0 are rhombi, and we calculate
−−→0 −−→0 −−→0 −→ −−→ −−→ −→ −−→0
AA + BB + CC = OA + OB + OC = OA + OA . (1)

Next, we note that OA0 = 2OM = 2R cos(∠A). But from known triangle formulas
we have AH = 2R cos(∠A) as well, so OA0 = AH. Moreover, AH and OA0 are
both perpendicular to BC, so AH k OA0 , which means that the quadrilateral
−−→ −−→
AHA0 O is a parallelogram. This gives us OA0 = AH, which we substitute into
(1) to get
−−→0 −−→0 −−→0 −→ −−→ −−→
AA + BB + CC = OA + AH = OH.

Crux Mathematicorum, Vol. 46(1), January 2020

 Solutions /39

−−→ −−→ −−→ →

−
Therefore, AA0 + BB 0 + CC 0 = 0 is equivalent to the condition that O and
H coincide, that is, the altitude of 4ABC from any vertex is the perpendicular
bisector of the opposite side i.e. the triangle ABC is equilateral.

4453. Proposed by Leonard Giugiuc and Miguel Ochoa Sanchez.

Let ABC be a triangle with no angle larger than 2π 3 and let T be its Fermat-
Torricelli point, that is the point such that the total distance from the three
vertices of ABC to T is minimum possible. Suppose BT intersects AC at D and
CT intersects AB at E. Prove that if AB + AC = 4DE, then ABC is equilateral.
The only submitted solution came from the proposers; it is presented below. In
addition, we received a comment from Walther Janous.
We use repeatedly the known theorem that the Fermat-Torricelli point T (of a
triangle that has all angles less than 2π3 ) is the point inside ∆ABC for which
∠AT B = ∠BT C = ∠CT A = 2π 3 . Set x = AT, y = BT , and z = CT . Then the
cosine law applied to triangles AT B and CT A yields
p p
AB = x2 + xy + y 2 and AC = z 2 + zx + x2 .
Furthermore, we have ∠AT E = π3 (because ∠AT E + CT A = ∠AT E + 2π 3 = π),
and (similarly) ∠ET B = π3 ; it follows that T E is the internal bisector of the angle
at T in ∆T AB, so that
xy xz
TE = ; similarly, TD = .
x+y x+z
Therefore, in ∆T DE we have
xy 2 xz 2
Å ã Å ãÅ ã Å ã
xy xz
DE 2 = + + .
x+y x+y x+z x+z
√ √
xy xy xz xz
Because 2 ≥ x+y and 2 ≥ x+z (by the GM-HM inequality), we have
√
xy + x yz + xz xy 2 xz 2
Å ã Å ãÅ ã Å ã
xy xz
≥ + + ,
4 x+y x+y x+z x+z
which implies that » √
2 xy + x yz + xz ≥ 4DE.
√ √
xy
xy
Moreover, because 2 = x+y if and only if x = y, and 2xz = x+z
xz
if and only if
p √
x = z, we have 2 xy + x yz + xz = 4DE if and only if x = y = z.
By the AM-GM inequality,
p p »
AB + AC = x2 + xy + y 2 + z 2 + zx + x2 ≥ 2 4 (x2 + xy + y 2 )(z 2 + zx + x2 ),
√ √
while by Cauchy’s inequality applied to the vectors (x, xy, y) and (z, zx, x) we
have
√
(x2 + xy + y 2 )(z 2 + zx + x2 ) ≥ (xy + x yz + xz)2 ,

Copyright © Canadian Mathematical Society, 2020

40/ Solutions

or » » √
2 4 (x2 + xy + y 2 )(z 2 + zx + x2 ) ≥ 2 xy + x yz + xz.
Putting the pieces together, we get
» √
AB + AC ≥ 2 xy + x yz + xz ≥ 4DE,

and conclude that if AB + AC = 4AD, then x = y = z. This immediately implies

that ∆ABC is equilateral, as desired.

Editor’s comments. Walther Janous observed that our problem has the following
immediate consequence:
If T is the Fermat-Torricelli point of a triangle with no angle larger
than 2π
3 , while D, E, F are the feet of the cevians through T , then the
perimeter of ∆DEF equals at most the semiperimeter of ∆ABC, with
equality if and only if both triangles are equilateral.

4454. Proposed by Nguyen Viet Hung.

Prove the identity
4n
Ç å Ç å Ç å
(−4)n + (−1)n


4n 4n n 4n 2n
− + · · · + (−1) = .
0 2 2n 2

We received 25 submissions, all of which were correct and complete. We present

the solution by Michel Bataille. Almost all solutions were based on the same idea.
From the binomial theorem, we have
4n Ç å
4n
X 4n k
(1 + i) = i .
k
k=0

Since for integers j we have i2j = (−1)j and i2j+1 = (−1)j i, the real part of
(1 + i)4n in the binomial expansion is
2n Ç å Ç å n−1 Ç å
X
k 4n n 4n
X
k 4n
R= (−1) = (−1) +2 (−1)) .
2k 2n 2k
k=0 k=0

The latter equality holds because

Ç å Ç å
k 4n 2n−k 4n
(−1) = (−1)
2k 2(2n − k))

for k = 0, 1, . . . , n − 1. On the other hand, since

√
(1 + i)4n = ( 2eiπ/4 )4n = 22n (−1))n = (−4)n ,

Crux Mathematicorum, Vol. 46(1), January 2020

 Solutions /41

we have R = (−4)n . Comparison with the above expression for R yields

n Ç å Ç å
k 4n n 4n
X
2 (−1)) − (−1) = R = (−4)n .
2k 2n
k=0

Solving for the summation in the above gives the required identity.

4455. Proposed by Marian Maciocha.

Find all integer solutions (if any) for the equation

(A + 3B)(5B + 7C)(9C + 11A) = 1357911.

We received 23 submissions, out of which 21 were correct and complete. We present

the solution by Corneliu Manescu-Avram.
Suppose A, B, C is an integer solution to the problem. Then the numbers A + 3B,
5B + 7C, and 9C + 11A are all odd, since their product is odd. But their sum
4(3A + 2B + 4C) is even, which is impossible. Thus the given equation has no
solutions in integers.

4456. Proposed by Leonard Giugiuc.

Let a, b, c be positive real numbers such that abc = 1. Show that

(a + b + c)(ab + bc + ac) + 3 ≥ 4(a + b + c).

We received 28 submissions, all correct. Most of these are similar to each other
and we present the solution by Boris C̆olaković.
3
The given inequality is equivalent to ab + bc + ca + a+b+c ≥ 4. By the AM-GM
inequality, we have

3 3(ab + bc + ca) 3 (ab + bc + ca)3

ab + bc + ca + = + ≥44 .
a+b+c 3 a+b+c 9(a + b + c)

Hence it suffices to prove that

(ab + bc + ca)3 ≥ 9(a + b + c). (1)

It is well known [Ed. and easy to show by simple algebra] that

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

Also, »
ab + bc + ca ≥ 3 3 (abc)2 = 3. (3)

Multiplying (2) and (3), inequality (1) follows, completing the proof.

Copyright © Canadian Mathematical Society, 2020

42/ Solutions

4457. Proposed by Hung Nguyen Viet.

Prove that for all − π2 < x, y < π
2, x 6= −y, we have that
tan x + tan2 y + cot2 (x + y) ≥ 1.
2

There were 22 correct solutions; we present four variants here.

Solution 1, due, independently, to Michel Bataille, Ganbat Batmunkh, Martin
Lukarevski, Marie-Nicole Gras, C.R. Pranesachar, Ioannis D. Sfikas, Kevin Soto
Palacios, and the Missouri State University Problem Solving Group.
Let a = tan x and b = tan y. Then cot(x + y) = (1 − ab)/(a + b) and
(1 − ab)2
tan2 x + tan2 y + cot2 (x + y) − 1 = a2 + b2 + − 1.
(a + b)2
Multiplying this quantity by (a + b)2 yields
(a4 + b4 + 2a3 b + 2ab3 + 3a2 b2 + 1) − (a2 + b2 + 4ab)
= (a2 + ab + b2 )2 + 1 − (a2 + b2 + 4ab)
= (a2 + ab + b2 − 1)2 + (a − b)2 .
Since this quantity is nonnegative, the result follows.
Equality occurs if and only if a = b and a2 +ab+b2 = 1, if and only if x = y = ±π/6.

Comment from the editor. For the difference between the two sides, Devis Alvarado
and Walther Janous obtained a fraction with the numerator
hÄ ä2 i hÄ ä2 i
a + b − √23 + (a − √13 )2 + (b − √13 )2 a + b + √23 + (a + √13 )2 + (b + √13 )2 .

Solution 2, by Digby Smith.

If |x| + |y| > π/2, then at least one of |x| and |y| exceeds π/4 and the left side
exceeds 1. Since 0 ≤ |x + y| ≤ |x| + |y|, then
cot2 (x + y) = cot2 (|x + y|) ≥ cot2 (|x| + |y|).
Thus, we may suppose that x and y are both nonnegative. If x + y = π/2, then the
left side exceeds 2. Suppose x + y 6= π/2. Then, using the arithmetic-geometric
means inequality, we have that
1
tan2 x + tan2 y +
tan2 (x + y)
1 1 1
ï Å ã Å ãò
= (tan2 x + tan2 y) + tan2 x + + tan2
y +
2 tan2 (x + y) tan2 (x + y)
tan x tan y tan x + tan y
≥ tan x tan y + + = tan x tan y +
tan(x + y) tan(x + y) tan(x + y)
= tan x tan y + 1 − tan x tan y = 1.

Crux Mathematicorum, Vol. 46(1), January 2020

 Solutions /43

Solution 3, built on ideas from Sefkat Arslanagić and Roy Barbara, independently.
If |x| + |y| > π/2, then at least one of |x| and |y| exceeds π/4 and the left side
exceeds 1. Since 0 ≤ |x + y| ≤ |x| + |y|, then

cot2 (x + y) = cot2 (|x + y|) ≥ cot2 (|x| + |y|).

Thus, it suffices to establish the result when 0 ≤ x ≤ y ≤ x + y ≤ π/2.

Since tan2 x is convex on (0, π/2),
x + y
tan2 x + tan2 y ≥ 2 tan2 .
2
Therefore, we need only establish that

2 tan2 θ + cot2 2θ ≥ 1

for 0 < θ ≤ π/2. Let t = tan2 θ. Then

(1 − t)2 9t2 − 6t + 1 (3t − 1)2

2 tan2 θ + cot2 2θ − 1 = 2t + −1= = ≥ 0.
4t 4t 4t
Equality occurs if and only if x = y = ±π/6.

Solution 4, by Vivek Mehra.

Since tan2 t = sec2 t − 1 and cot2 t = csc2 t − 1, the inequality is equivalent to
1 1 1
2
+ 2
+ 2 ≥ 4.
cos x cos y sin (x + y)

Let t = cos2 x + cos2 y. Applying the AM-HM inequality leads to

1 1 4 4
+ ≥ = .
cos2 x cos2 y cos2 x + cos2 y t
Also, from the AM-GM inequality, we have that

sin2 (x + y) = sin2 x cos2 y + cos2 x sin2 y + 2 sin x cos x sin y cos y

≤ sin2 x cos2 y + cos2 x sin2 y + 2| sin x cos x sin y cos y|
≤ sin2 x cos2 y + cos2 x sin2 y + sin2 x cos2 x + sin2 y cos2 y
= (cos2 x + cos2 y)(sin2 x + sin2 y) = t(2 − t).

Since
4 1 4t2 − 12t + 9 (2t − 3)2
+ −4= = ≥ 0,
t t(2 − t) t(2 − t) t(2 − t)
this, along with the foregoing inequalities, yields the result. Equality occurs if and
only if x = y and cos2 x = cos2 y = 3/4, if and only if x = y = ±π/6.

Copyright © Canadian Mathematical Society, 2020

44/ Solutions

4458. Proposed by Marian Cucoaneş and Marius Drăgan.

Let a, b, c, d be the sides of a cyclic quadrilateral with circumradius R and lengths
of diagonals d1 and d2 . Prove that
X a 4R
≥√ .
b+c+d−a d1 d2
cyclic

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

solution by Marie-Nicole Gras, lightly edited.
Denote by s the semiperimeter of the quadrilateral. In a cyclic quadrilateral we
can make use of the well-known formulas
1 (ab + cd)(ac + bd)(ad + bc)
R= and d1 d2 = ac + bd.
4 (s − a)(s − b)(s − c)(s − d)
The inequality we need to prove is thus equivalent to

1 (ab + cd)(ad + bc)

Å ã
a b c d
· + + + ≥ . (1)
2 s−a s−b s−c s−d (s − a)(s − b)(s − c)(s − d)
In order to simplify our calculations, we let x = s − a, y = s − b, z = s − c and
t = s − d; then x, y, z, t > 0. Note that x + y + z + t = 2s. On the right hand side
of (1), we have
ab + cd = (s − x)(s − y) + (s − z)(s − t)
= 2s2 − sx − sy − sz − st + xy + zt
= xy + zt,
and similarly ad + bc = xt + yz. Thus, the expression under the square root
becomes
(xy + zt)(xt + yz) x2 yt + xy 2 z + yz 2 t + xzt2 x y z t
= = + + + .
xyzt xyzt z t x y
√
Let F = xz + yt + xz + yt , so the right hand side of (1) is simply F .
On the left hand side of (1), we have
s−x −x + y + z + t 1 1
Å ã
a y z t
= = =− + · + + ,
2(s − a) 2x 4x 4 4 x x x
and similarly for the remaining terms, so that the left hand side of (1) becomes
1
Å ã
y z t x z t x y t x y z
−1 + · + + + + + + + + + + + .
4 x x x y y y z z z t t t
Moreover, we have that xy + xy ≥ 2, t
x + x
t ≥ 2, z
y + y
z ≥ 2, t
z + z
t ≥ 2, so in order
to show (1) it suffices to prove that
1 √
−1 + (8 + F ) ≥ F .
4

Crux Mathematicorum, Vol. 46(1), January 2020

 Solutions /45

√ √ √
Finally, observe that F + 8 = ( F − 2)2 + 4 + 4 F ≥ 4 + 4 F to conclude the
proof.

4459. Proposed by Leonard Giugiuc and Miguel Ochoa Sanchez.

Let ABC be an isosceles triangle with AB = AC. For a point P on side AB let Q
be a point of the extension of AC beyond C for which the midpoint N of P Q lies
on the segment BC; similarly, for a point R on side AC let S be a point of the
extension of AB beyond B for which the midpoint M of RS lies on the segment
BC. Prove that
PQ cos ∠RM N
= .
RS cos ∠P N M

We received 11 submissions, all of which were correct; we will sample two of the
variety of solutions.
Solution 1 is a composite of almost identical solutions submitted (independently)
by Marie-Nicole Gras, C.R. Pranesachar, and Titu Zvonaru.
Define P 0 and Q0 to be the feet of the perpendiculars from P and Q, respectively,
to the line BC. Since P N = N Q, it follows that the right triangles P N P 0 and
QN Q0 are congruent and, hence, P P 0 = QQ0 and P 0 N = Q0 N . The former implies
that the right triangles P P 0 B and QQ0 C are also congruent, because we have, in
addition, ∠P BP 0 = ∠ABC = ∠ACB = ∠QCQ0 . Thus BP 0 = CQ0 ; consequently,
the translation that takes B to P 0 will take C to Q0 , whence P 0 Q0 = BC. It
follows that
P 0N N Q0 P 0 Q0 BC
cos ∠P N M = = = = ,
PN NQ PQ PQ
and, therefore,
P Q cos ∠P N M = BC.
Similarly, we obtain RS cos ∠RM N = BC, and the desired conclusion follows.

Solution 2, by Walther Janous.

We place the origin of a vector space at the midpoint of BC and denote the vector
from the origin to a generic point X by X.~ Without loss of generality we set

~ = (0, t),
A ~ = (−1, 0),
B ~ = (1, 0),
C

where t > 0. We are given a point P on side AB, which means that

~ + (1 − λ)B
P = λA ~ = (λ − 1, tλ) and Q=C ~ + (C~ − A)s
~ = (s + 1, −st),
Ä ä
~ = s+λ , t(λ−s) ,
~ = 1 (P~ +Q)
for 0 < λ < 1 and s to be determined. Specifically, N 2 2 2
so that N is on BC if and only if s = λ. Consequently, we have

N = (λ, 0) and P~ − N
~ = (−1, tλ).

Copyright © Canadian Mathematical Society, 2020

46/ Solutions

~ = µA
Similarly, when R ~ + (1 − µ)C
~ = (1 − µ, tµ), with 0 < µ < 1, we have
~ = (−µ, 0),
M
so that
~ −M
R ~ = (1, tµ) and ~ −N
M ~ = (−λ − µ, 0).
Finally, we must verify that
(P~ − N~ ) · (M
~ −N
~) ~ ~ ~ ~
~ (R − M ) · (N − M ) ,
|P~ − Q|
~ ~ − S|
= |R
|P~ − N
~ | · |M
~ −N
~| ~ −M
|R ~ | · |N
~ −M~|
which is easy because
|P~ − Q|
~ = 2|P~ − N
~ |, ~ − S|
|R ~ = 2|R
~ −M
~ |, ~ −N
|M ~ | = |N
~ −M
~|=
6 0,
and
(P~ − N
~ ) · (M
~ −N
~ ) = (R
~ −M
~ ) · (N
~ −M
~ ) = λ + µ.

Editor’s comments. Note that the restriction of the points P and R to the sides
AB and AC can be omitted — Janous’s argument shows that the result continues
to hold starting with any point P on the line AB and any point R 6= Q on the line
AC (because λ and µ are free to be assigned any real values as long as λ + µ 6= 0).

4460. Proposed by Gantumur Choijilsuren and Leonard Giugiuc.

Let (xn )n≥1 be a sequence of real numbers such that (3xn+1 − 2xn )n≥1 is conver-
gent. Show that (xn )n≥1 is convergent.
We received 16 submissions of which 15 were correct. We present the solution by
Ángel Plaza.
Let yn = 3xn+1 − 2xn . Then
1 2
xn = yn−1 + xn−1
3 3
1 2 4
= yn−1 + yn−2 + xn−2
3 9 9
1 2 4 8
= yn−1 + yn−2 + yn−3 + xn−3
3 9 27 27
···
n−2 Å ãk Å ãn−1
X 1 2 2
= yn−1−k + x1 .
3 3 3
k=0

Since
n−2 Å ãk 1 1 2 n−1


X 1 2 3 − 3 3
= →1
k=0
3 3 1 − 32
and (yn )n≥1 converges, then (xn )n≥1 converges as well.

Crux Mathematicorum, Vol. 46(1), January 2020