CRUXv 47 N 2

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 2021. 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 2021 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

Ana Duff Ontario Tech University
Andrew McEachern York University
Vasile Radu Birchmount Park Collegiate Institute
Aaron Slobodin University of Victoria
Chi Hoi Yip 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
71 Editorial Kseniya Garaschuk

72 MathemAttic: No. 22
72 Problems: MA106–MA110
74 Solutions: MA81–MA85
77 Teaching Problems: No. 13 Erick Lee
82 Olympiad Corner: No. 390
82 Problems: OC516–OC520
84 Solutions: OC491–OC495
90 Multifaceted Solutions to a Remarkable Geometry Puzzle
H. S. Hoffman and S. I.Warshaw
97 Problems: 4611–4620
101 Solutions: 4561–4570
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 /71
EDITORIAL
As you likely noticed when you opened this issue, Crux is boasting a new cover.
In the process of moving online, the journal lost its iconic purple cover and I’m
very excited to have a new beautiful design to represent Crux . The cover was
designed by Rebekah Brackett and you can find more of her work on her website
https://www.rebekahbrackettart.com/.
Rebekah is one of the people that pro-
vided inspiration and guidance in my
own journey to understand and em-
brace First People’s principles of know-
ing and learning. As we were organiz-
ing Fraser Valley Math Education Sq’ep
(Sq’ep meaning a meeting, gathering in
Halq’eméylem), we explored the connec-
tions between math, language, art, land.
With the help of Tasheena Boulier and
her family, consisting of the few last flu-
ent speakers of Halq’eméylem, we cre-
ated a counting booklet featuring num-
ber words in Halq’eméylem and images
of the lands of the Sto:lo people. To
me, number systems are fascinating as
they offer a unique insight into the cul-
ture. For example, Sto:lo have different
counting words depending on what is be-
ing counted, highlighting the fundamen-
tal differences between how they treat
objects, animals and people. Take a look
at the booklet, explore the numbers and
enjoy the views of the beautiful Fraser Valley:
https://www.ufv.ca/media/assets/mathematics/halq-booklet-j.pdf
Pandemic has offered us an opportunity to see the importance of human connec-
tions. So where do we start in math? Veselin Jungic and I write more about our
journeys in exploring Indigenous ways of knowing in mathematics in the March
edition of CMS Notes: https://cms.math.ca/publications/cms-notes/
Let us learn together.
Kseniya Garaschuk
Copyright © Canadian Mathematical Society, 2021

72/ MathemAttic
MATHEMATTIC
No. 22
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 April 30, 2021.
MA106. Suppose
N = 1 + 11 + 101 + 1001 + 10001 + · · · + 1000 · · · 01,
where there are 50 zeros in the last term. When N is written as a single integer
in decimal form, find the sum of its digits.
MA107. A wooden cube is painted red on five of its six sides and then cut
into identical small cubes, of which 52 have exactly two red sides. How many small
cubes have no red sides?
MA108. Suppose that a, b, c and d are positive integers that satisfy the
equations
ab + cd = 38, ac + bd = 34, ad + bc = 43.
What is the value of a + b + c + d?
MA109. Ten equal spheres are stacked to form a regular tetrahedron. How
many points of contact are there between the spheres?
MA110. In the figure, ABCDEF is a regular

hexagon and P is the midpoint of AB.
Find the ratio
Area(DEQR)
.
Area(F P Q)
.................................................................
Crux Mathematicorum, Vol. 47(2), February 2021

MathemAttic /73
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 30 avril 2021.
La rédaction souhaite remercier Rolland Gaudet, professeur titulaire à la retraite à
l’Université de Saint-Boniface, d’avoir traduit les problèmes.
MA106. Supposer que
N = 1 + 11 + 101 + 1001 + 10001 + · · · + 1000 · · · 01,
où on trouve 50 zéros dans le dernier terme. Si N est écrit en forme décimale,
déterminer la somme de ses chiffres.
MA107. Un cube en bois est peint rouge sur cinq de ses six côtés et puis taillé
en petits cubes identiques, dont 52 ont exactement deux faces rouges. Déterminer
le nombre de petits cubes ayant aucune face rouge.
MA108. Supposer que a, b, c et d sont des entiers positifs tels que
ab + cd = 38, ac + bd = 34, ad + bc = 43.
Déterminer la valeur de a + b + c + d.
MA109. Dix sphères de même rayon sont empilées pour former un tétraèdre.
Déterminer le nombre de points de contact entre les sphères.
MA110. Dans la figure, ABCDEF est un hexagone

régulier et P est le point milieu de AB.
Déterminer le ratio
Area(DEQR)
.
Area(F P Q)

74/ MathemAttic
MATHEMATTIC
SOLUTIONS
Statements of the problems in this section originally appear in 2020: 46(7), p. 285–286.
MA81. Find the sum of all positive integers smaller than 1260 which are not
divisible by 2 and not divisible by 3.
Originally modified problem 8 from the 2018 Alberta High School Mathematics
Competition.
We received 8 submissions, all of which were correct and complete. We present
the solution by Alin Popescu and Daniel Văcaru, modified by the editor.
(1+1260)·1260
The sum of the first 1260 natural numbers is 2 = 1261 · 630.
We want to remove the sum of the natural numbers smaller than 1260 that are
divisible by 2. This sum is
630 · 631
2 · 1 + 2 · 2 + . . . + 2 · 630 = 2 (1 + 2 + . . . + 630) = 2 = 630 · 631.
2
We also want to remove the sum of the odd numbers under 1260 that are divisible
by 3, i.e. numbers of the form 3 (2i + 1) for i ≤ k ∈ N where the maximum such
number is 3 (2k + 1) ≤ 1260. We find 2k + 1 ≤ 420 or 2k ≤ 419, and since k is an
integer, k = 209. Thus the sum of the numbers 3 (2i + 1) under 1260 is
X209 X209 X209
3 (2i + 1) = 6i + 3
i=0 i=0 i=0
= 6 (1 + 2 + . . . + 209) + 3 · 210
209 · 210
=6 + 3 · 210
2
= 630 (209 + 1)
= 630 · 210.
Thus, the desired sum is
1261 · 630 − 630 · 631 − 630 · 210 = 630 · (1261 − 631 − 210) = 630 · 420 = 264600.
MA82. Let an = n2 + 2n + 50, n = 1, 2, . . .. Let dn be the largest positive

integer that is a divisor of both an and an+1 . Find the maximum value of dn ,
n = 1, 2, . . .
Originally problem 12 from the 2018 Alberta High School Mathematics Competi-
tion.

MathemAttic /75
We received 5 submissions, four of which were correct and complete. We present

the solution by Corneliu Mănescu-Avram, lightly edited.
Since dn divides both an and an+1 , it also divides
an+1 − an = 2n + 3
and
an − (2n + 3) = n2 + 47.
Therefore dn divides
4(n2 + 47) − (2n − 3)(2n + 3) = (4n2 + 188) − (4n2 − 9) = 197.
Since 197 is prime, dn ∈ {1, 197} for all n. For n = 97 we have
a97 = 49 · 197, a98 = 50 · 197.
Therefore the maximum value of dn is 197.
MA83. Prove that the numbers 26n and 26n + 2n have the same number of
digits, for any non-negative integer n.
Originally problem 3 from Part II of the 2018 Alberta High School Mathematics
Competition.
We received 2 solutions. We present the one by Corneliu Mănescu-Avram, modified
by the editor.
The statement is easily checked for n = 1, 2. Let n ≥ 3 and suppose that there
exists a positive integer m such that
26n < 10m ≤ 26n + 2n .
Since n ≥ 3 and 263 > 104 , we must have m ≥ n + 2. Dividing by 2n we obtain
13n < 2m−n 5m ≤ 13n + 1,
where 2m−n 5m is an integer divisible by 4. Since 13n + 1 ≡ 2 (mod 4), we arrive

at a contradiction.
MA84. The area of the trapezoid ABCD with AB k CD, AD ⊥ AB and

AB = 3CD is equal to 4. A circle inside the trapezoid is tangent to all of its sides.
Find the radius of the circle.
Originally problem 15 from the 2016 Alberta High School Mathematics Competi-
tion.
We received 11 submissions, 10 of which are correct. We present the solution by
Alin Popescu and Daniel Văcaru, modified by the editor.

76/ MathemAttic
We take CD = b. It follows that AB = 3CD = 3b. The area of the trapezoid is
(CD + AB) · h 4b · h
4= = ,
2 2
which gives bh = 2 or
2
h=. (1)
b
However, AD ⊥ AB ⇒ h = AD. According to the Pitot theorem, AD + CB =
AB + CD which gives h + CB = 4b or CB = 4b − h. Let CP ⊥ AB, where
{P } = AB ∩ CP . It follows that AD k CP, AB k CD implies AD = CP and
AP = CD, so P B = 2b.
In the triangle CP B the angle ∠CP B = 90◦ . We use the Pythagorean theorem
2
to write CB 2 = P B 2 + P C 2 or (4b − h) = 4b2 + h2 , which gives 16b2 − 8bh = 4b2
2
or 4b (4b − 2h) = 4b or 4b − 2h = b, which finally results in
2h = 3b. (2)
√
Together, (1) and (2) show that 2h = 4b and 3b = 4b , so b = √2
3
and h = 3. Then
√ √
the diameter of the circle is 3 and the radius is 23 .
MA85. A collection of items weighing 3, 4 or 5 kg has a total weight of 120

kg. Prove that there is a subcollection of the items weighing exactly 60 kg.
Originally problem 4 from Part II of the 2018 Alberta High School Mathematics
Competition.
We received 2 submissions, neither of which was fully correct and complete. You
can fine the official solution at
https://drive.google.com/file/d/0B5b6n_Nz71-rRVVxa1g5ZlBTOEdvMXNCLUFVbTczSTRWSWJr/
view

Erick Lee /77
TEACHING PROBLEMS
No. 13
Erick Lee
Four Triangles: An Example of Interleaved Practice
Triangles, it seems, are everywhere. You see them daily through art, architecture
and nature. Triangles are common elements of school mathematics from the initial
naming and categorizing of two-dimensional shapes through to deeper excursions
into the realms of geometry and trigonometry. Students spend much time deter-
mining areas, side lengths and angles of triangles They should have a toolbox of
strategies and techniques to solve a wide range of problems involving this basic
figure; however, students often find problems with triangles to be challenging.
As an example of this unexpected complexity, I’ve gathered four problems involv-
ing triangles in the grid below. These questions share some commonalities on the
surface. They all involve triangles and feature numbers that are alike. In order to
solve these problems, students must first understand how the given information
will inform the selection of a strategy. Once they have identified a strategy, they
must then carry it out to determine a solution.

78/ Teaching Problems
Solving the Problems
Question 1. When students see a question relating the areas of squares to the sides
of a right triangle, they should think of the Pythagorean Theorem. This problem
asks students to find the area of the central triangle given the areas of squares on
two of its sides. The most straightforward way to solve this problem is to find the
measures of the base and height of the triangle and then apply the triangle area
formula. The Pythagorean Theorem states that the area of the square whose side
is the hypotenuse is equal to the sum of the areas of the squares on the other two
sides. The unmarked area of the square whose side is the base of the triangle is
1 since 8 + 1 = 9. The length of the base and height are the square √ roots of the
areas of the squares of those
√ sides. The height of the triangle is 2 2 and the base
is 1. The area is therefore 2.
Question 2. This question asks students to determine the area of a triangle given
the coordinates of each of its vertices. One method students might use to determine
this area is to find an altitude and base of the triangle using coordinate geometry
and the distance formula. If students choose the line segment between (7,8) and
(8,7) as the base, the calculations will be simplified compared to other choices for
the base.
A less complicated and perhaps more mathematically elegant way to find the area
of this triangle is to find the area of the rectangle that encloses it and then subtract
the areas of the shaded triangles shown on the right.

Erick Lee /79
The area of the encompassing square is 4 square units. From this area we subtract
the areas of the three shaded triangles.
1
Areablue4 = · 1 · 2 = 1,
2
1
Areagreen4 = · 2 · 1 = 1,
2
1 1
Areapurple4 = · 1 · 1 = ,
2 2
3
Area4 = Area − Areablue4 − Areagreen4 − Areapurple4 = .
2
An alternative way to find the area of any simple polygon with vertices identified
by coordinate points is to use Gauss’ Area Formula, more commonly known as the
Shoelace Formula. It is rarely taught in Canadian secondary schools except as an
enrichment activity. If you haven’t seen a description of this formula in the past,
I recommend you check out James Tanton’s description in his Cool Math Essay
from June 2014.
Question 3. When students see right triangles and are asked about angles, they
should be reminded of inverse trigonometric ratios. In this question, they can
calculate the measure of ∠XY Z by calculating and then adding ∠XY W and
∠W Y Z. As they have a hypotenuse and an adjacent leg of ∠XY W , they will
need inverse cosine to find this angle. With ∠W Y Z, students have an adjacent
leg and an opposite leg and hence will need inverse tangent.

−1 7
∠XY W = cos ≈ 38.9◦ ,
9

−1 8
∠W Y Z = tan ≈ 88.5◦ ,
7
∠XY Z = ∠XY W + ∠W Y Z ≈ 127.4◦ .
Question 4. Finding the sides and angles of a non-right triangle should lead
students to consider the Sine and Cosine Laws. In this case, students can use
the Cosine Law to find one of the angles. They can then use two sides and the
contained angle to find the area of the triangle:
y 2 + z 2 − x2 92 + 72 − 82 66
cos X = = = ,
2yz 2·9·7 126
so then X = cos−1 66
≈ 58.4◦ .

126
Now that this angle is known, students can use this angle to find the area of the
triangle:
1 1
Area = yz sin X = · 9 · 7 · sin(58.4◦ ) ≈ 26.8.
2 2
An alternate solution method is using Heron’s Formula. This formula is rarely
taught in Canadian secondary schools. The formula calculates the area of a triangle

80/ Teaching Problems
given the lengths of its three sides. With this formula, there is no need to calculate
any other lengths or angles.
p
Area = s(s − a)(s − b)(s − b),
where s is semi-perimeter of the triangle and a, b, c are side lengths. The semi-
perimeter is half of the sum of the lengths of the three sides. For the given triangle
this is (7 + 8 + 9)/2 = 12, so we have
p √
Area = 12(12 − 7)(12 − 8)(12 − 9) = 12 5 ≈ 26.8.
The Benefits of Interleaved Practice

I created the set of problems shown at the beginning of this article to give stu-
dents “interleaved” practice instead of “blocked” practice. Interleaved and blocked
practice are two different methods of practicing newly acquired skills.
In “blocked” practice, a block of practice questions focused on a single skill are
assigned. Blocked practice is the type of practice that is often found in textbooks.
The practice questions are often subtle variations of examples that were previously
demonstrated. While blocked practice gives students lots of practice on a targeted
skill, it doesn’t help them become better problem solvers. Because the strategy
required to solve the problem is known up front, students are never challenged to
analyze a problem to determine what solution strategy is needed. Students never
learn to recognize the characteristics of a problem that might suggest a certain
strategy.
During “interleaved” practice, students are given practice problems which require
numerous different skills and strategies to solve. Since students don’t know up
front what strategy might be needed, they must consider various problem solving
strategies and select the one that would be most useful for a given situation.
This helps students become more flexible problem solvers through consideration
of a wider range of strategies instead of developing an over-reliance on a specific
formula or strategy. For example, when students first learn the formula for a
definite integral, they sometimes jump to applying the formula to every situation.
They might even rush to apply an integral to a rectangular region when its area
could be calculated much more simply with length times width.
While both interleaved and blocked methods of practice are useful, many of the
resources that are provided to classrooms contain predominantly blocked practice.
In a recent survey of widely used US seventh grade mathematics textbooks, Dr.
Doug Rohrer, et. al. found that there were more than eight blocked problems for
every interleaved problem in these textbooks. Even the review assignments in each
textbook were moderately blocked (see Rohrer, D., Dedrick, R. F., & Hartwig, M.
K. (2020). The scarcity of interleaved practice in mathematics textbooks. Edu-
cational Psychology Review, 32, 873-883). While blocked practice appears to be
predominant in textbooks, interleaved practice has shown significant benefits. In-
terested readers can see a recent published study by Rohrer, D. et al, A randomized
controlled trial of interleaved mathematics practice.

Erick Lee /81
Teachers that are interested in including more interleaved practice with their stu-
dents don’t necessarily need a new textbook. Teachers can quickly create an
interleaved assignment by combining a selection of problems from several different
sections throughout the textbook.
For more information and tips about interleaved practice, teachers can visit the
site https://www.retrievalpractice.org/interleaving.
Same Surface, Different Depth

The four triangle problems at the beginning of this article are presented in a math
routine useful for interleaved practice called ÒSame Surface, Different DepthÓ or
SSDD. Each set of four problems is related in shape, appearance or context. While
they may look similar on the surface, underneath, they require different problem
solving strategies to solve.
UK mathematician Craig Barton has created a website where educators from
around the world can create and share problems in this format: check it out at
https://ssddproblems.com/. Hundreds of these problem sets have been shared
which makes it easy to find one that is suitable for any secondary school mathe-
matics outcome.
If you create an “SSDD” problem, consider sharing it with the wider mathematics
education community by contributing it to the website. Criteria for submissions
are included at https://ssddproblems.com/submission-guidelines/
.................................................................
Erick Lee is a Mathematics Support Consultant for the Halifax Regional Centre
for Education in Dartmouth, NS. Erick blogs at https://pbbmath.weebly.com/
and can be reached via email at elee@hrce.ca and on Twitter at @TheErickLee.

82/ OLYMPIAD CORNER
OLYMPIAD CORNER
No. 390
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
OC516. Pasha placed numbers from 1 to 100 in the cells of the square 10×10,
each number exactly once. After that, Dima considered all sorts of squares, with
the sides going along the grid lines, consisting of more than one cell, and painted
in green the largest number in each such square (one number could be coloured
many times). Is it possible that all two-digit numbers are painted green?
OC517. Denote by N the set of positive integers 1, 2, 3, . . . Find all functions

f : N → N such that n! + f (m)! divides f (n)! + f (m!) for all m, n ∈ N.
OC518. In a triangle ABC with AB 6= AC let M be the midpoint of AB,

let K be the midpoint of the arc BAC in the circumcircle of ABC, and let the
perpendicular bisector of AC meet the bisector of the angle BAC at P . Prove
that A, M , K, P are concyclic.
OC519. Show that the number x is rational if and only if three distinct terms
that form a geometric progression can be chosen from the sequence:
x, x + 1, x + 2, x + 3, . . .
OC520. Larry and Rob are two robots travelling in one car from Argovia
to Zillis. Both robots have control over the steering and steer according to the
following algorithm: Larry makes a 90◦ left turn after every l kilometer driving
from start; Rob makes a 90◦ right turn after every r kilometer driving from start,
where l and r are relatively prime positive integers. In the event of both turns
occurring simultaneously, the car will keep going without changing direction. As-
sume that the ground is flat and the car can move in any direction. Let the car
start from Argovia facing towards Zillis. For which choices of the pair (l, r) is the
car guaranteed to reach Zillis, regardless of how far it is from Argovia?
.................................................................

OLYMPIAD CORNER /83
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.
OC516. Pasha inscrit les nombres de 1 à 100 dans les cellules d’un grillage
de taille 10 par 10, chacun exactement une fois. Par la suite, Dima considère tous
les carrés contenant plus d’une cellule et dont les côtés suivent l’alignement du
grillage, et puis elle colore en vert le plus grand nombre dans chaque tel carré,
il étant entendu qu’un nombre pourrait bien être coloré plus d’une fois. Est-ce
possible que tous les nombres à deux chiffres soient ainsi colorés vert ?
OC517. Soit N l’ensemble des entiers positifs 1, 2, 3, . . . Déterminer toutes

les fonctions f : N → N telles que n! + f (m)! divise f (n)! + f (m!) pour tout m,
n ∈ N.
OC518. Soit M le mi point de AB dans un triangle ABC tel que AB 6= AC.

Soit aussi K le mi point de l’arc BAC du cercle circonscrit de ABC. Enfin,
supposons que la bissectrice perpendiculaire de AC rencontre la bissectrice de
l’angle BAC en P . Démontrer que A, M, K, P sont cocycliques.
OC519. Démontrer que le nombre x est rationnel si et seulement si trois

termes distincts en progression géométrique peuvent être choisis dans la suite
suivante:
x, x + 1, x + 2, x + 3, . . .
OC520. Laurent et Rolland sont deux robots se déplaçant en une même

voiture, allant de Argovia à Zillis. Ces deux robots ont contrôle du volant et
pilotent la voiture, selon les règles suivantes. À chaque l kilomètres depuis le
départ, Laurent tourne à gauche par 90◦ , tandis que Rolland tourne à droite par
90◦ chaque r kilomètres depuis le départ, où l et r sont des entiers relativement
premiers ; advenant que les deux voudraient tourner au même moment, la voiture
ne change pas de direction. Supposons que le terrain est plat et que la voiture
peut se déplacer en toute direction. La voiture débute sa randonnée à Argovia,
pointant directement vers Zillis. Pour quelles valeurs de la paire (l, r) est-on assuré
que les robots vont se rendre à Zillis, quelle que soit sa distance de Argovia ?

84/ OLYMPIAD CORNER
OLYMPIAD CORNER
SOLUTIONS
OC491. Let ABC be a triangle such that AB 6= AC. Prove that there
exists a point D 6= A on its circumcircle satisfying the following property: For
any points M, N outside the circumcircle on the rays AB and AC, respectively,
satisfying BM = CN , the circumcircle of AM N passes through D.
Originally problem 2, Grade 11-12, Day 1, Final Round of 2017 Germany Math
Olympiad.
We received 12 submissions, all correct. We present 2 solutions.
Solution 1, by UCLan Cyprus Problem Solving Group.
Let D be the point of intersection of the perpendicular bisector of BC with the

arc of BC containing A, of the circumcircle of the triangle 4ABC. Note that the
definition of D is independent of the points M and N .
We claim that the triangles 4DBM and 4DCN are equal. Indeed it is given that
BM = CN . We also have DB = DC since D is on the perpendicular bisector of
BC. Finally we have
∠DBM = 180◦ − ∠DBA = 180◦ − ∠DCA = ∠DCN ,

OLYMPIAD CORNER /85
as D is on the circumcircle of the triangle 4ABC. (D and A are on the same arc
of BC, while M, N are outside of the circumcircle on the rays AB and AC.)
By the equality of the above triangles, we get that ∠DM A = ∠DN A, showing
that D is on the circumcircle of the triangle 4AM N .
Solution 2, by Miguel Amengual Covas.

Let AB = c, CA = b and ∠CAB = ϕ.
We consider a Cartesian coordinate system with the unit of measurement the same
along both coordinate axes, the x axis along the side AB of 4ABC and the y axis
along the side CA.
The coordinates of A then are (0, 0), the coordinates of B are (c, 0) and those of
C are (0, b). Therefore M will have coordinates (c + λ, 0) and N is at (0, b + λ),
where λ is a positive real number.
N (0, b + λ)
C (0, b)
A (0, 0) ϕ
B (c, 0) M (c + λ, 0)
Since the general form of the equation of a circle passing through the origin is
x2 + y 2 + 2xy cos ϕ − P x − Qy = 0,
where P and Q are real numbers, the equations of the circumcircles of 4ABC and
AM N are
x2 + y 2 + 2xy cos ϕ − cx − by = 0 (1)
and
x2 + y 2 + 2xy cos ϕ − (c + λ) x − (b + λ) y = 0, (2)
respectively.
Solving simultaneously (1) and (2), we find x = 0, y = 0 and
−b + c b−c
x= , y= .
2 (1 − cos ϕ) 2 (1 − cos ϕ)
Hence
−b + c b−c

D , ,
2 (1 − cos ϕ) 2 (1 − cos ϕ)
which does not depends of λ, is the required point.

86/ OLYMPIAD CORNER
OC492. Let ABC be a triangle with AB = AC and let I be its incenter. Let
Γ be the circumcircle of ABC. Lines BI and CI intersect Γ in two new points M
and N respectively. Let D be another point on Γ lying on arc BC not containing
A, and let E, F be the intersections of AD with BI and CI, respectively. Let P, Q
be the intersections of DM with CI and of DN with BI respectively.
(i) Prove that D, I, P, Q lie on the same circle Ω.
(ii) Prove that lines CE and BF intersect on Ω.
Originally problem 6, Final Round of 2018 Italy Math Olympiad.
We received 4 correct submissions. We present the solution by UCLan Cyprus
Problem Solving Group.
(i) We have
B̂ + Ĉ
∠QIP = ∠BIC = 180◦ −
2
and
B̂ + Ĉ
∠QDP = ∠N DM = ∠N DA + ∠ADM = ∠N CA + ∠ABM = .
2
So (i) follows. Note that this holds even if the triangle 4ABC is not isosceles.
(ii) Let Γ1 and Γ2 be the circumcircles of triangles 4BED and 4CF D respec-
tively. Let X 6= D be the other point of intersection of Γ1 and Γ2 .

OLYMPIAD CORNER /87
We have
∠BXC = ∠BXD + ∠DXC

= ∠BED + ∠DF C
= (∠EDM + ∠EM D) + (∠F AC + ∠F CA)
= ∠ADM + ∠BM D + ∠DAC + ∠N CA
B̂ Ĉ
= + (∠BAD + ∠DAC) +
2 2
B̂ + Ĉ B̂ + Ĉ
= Â + = 180◦ −
2 2
= ∠BIC
It follows that X belongs on the circumcircle of 4BIC.

We now have
B̂ + Ĉ
∠BXF = ∠BXC + ∠CXF = ∠BIC + ∠CDF = 180◦ − + B̂ = 180◦ ,
2
where here, we have for the first time used that the triangle 4ABC is isosceles.
Thus X belongs on BF . Furthermore, we also have
∠BXE = 180◦ − ∠BDE = 180◦ − Ĉ = ∠BXC ,
as B, X, I, C are concyclic. So X belongs on EC as well.

So to complete the proof it remains to show that X ∈ Ω. To this end, it is enough
to show that ∠DXI + ∠IP D = 180◦ .
We have
B̂
∠DXI = ∠DXC + ∠CXI = ∠DXE + ∠CBI = ∠DBE + .
2
So
B̂
∠DXI + ∠IP D = ∠DBE + ∠IP D +
2
B̂
= 360◦ − ∠BIP − ∠BDP +
2
B̂
= 360◦ − (180◦ − B̂) − (∠BDA + ∠ADM ) + = 180◦ ,
2
as ∠BDA = ∠BCA = B̂ and ∠ADM = ∠ABM = B̂/2.
The result follows.
Editor’s Comment. Sergey Sadov also proved that the first part holds true without
the assumption that AB = AC.

88/ OLYMPIAD CORNER
OC493. Let a, b be real numbers such that a < b and let f : (a, b) → R
be a function such that the functions g : (a, b) → R, g(x) = (x − a)f (x) and
h : (a, b) → R, h(x) = (x − b)f (x) are increasing. Prove that the function f is
continuous on (a, b).
Originally problem 4, Grade 11, District Round of 2018 Romania Math Olympiad.
We received 10 submissions. We present the solution by Oliver Geupel.
It is enough to show that, for every x0 ∈ (a, b), it holds
lim f (x) = f (x0 ) and lim f (x) = f (x0 ). (1)

x%x0 x&x0
Let a < x < x0 < b. By the monotonicity of g and h, we have
b − x0 x0 − a
· f (x0 ) ≤ f (x) ≤ · f (x0 ). (2)
b−x x−a
Both the lower and the upper bound in (2) tend to f (x0 ) as x % x0 . Hence f (x)
tends to f (x0 ), which proves the first limit (1). The second limit (1) is analogous,
using the similar relation
x0 − a b − x0
· f (x0 ) ≤ f (x) ≤ · f (x0 ).
x−a b−x
which holds for a < x0 < x < b.
OC494. Let n and q be two natural numbers, n ≥ 2, q ≥ 2 and q 6≡ 1 (mod 4)

and let K be a finite field having exactly q elements. Prove that for every a ∈ K
n n
there exist x, y ∈ K such that a = x2 + y 2 .
Originally problem 4, Grade 12, District Round of 2018 Romania Math Olympiad.
We received 5 submissions. We present the solution by Corneliu Avram Manescu.
Let p be the characteristic of K. Then, p is a prime number and q = pα , where
α is a positive integer. From q 6≡ 1 (mod 4), we deduce p 6≡ 1 (mod 4), therefore
p = 2 or p ≡ 3 (mod 4) and in this last case α is odd.
n n n
If p = 2 and x, y ∈ K such that x2 = y 2 , then x = y = 0 or (xy −1 )2 = 1 for
α
y 6= 0. Then, (xy −1 )2 −1 = (xy −1 )q−1 = 1. Since (2n , 2α − 1) = 1, we get that
n
xy −1 = 1, whence x = y. Consequently, the function f : K → K, f (x) = x2 is
injective, hence surjective. If a ∈ K, then there exists x ∈ K such that a = f (x),
n n
therefore a = x2 + 02 .
If p ≡ 3 (mod 4) and α is odd, then q ≡ 3 (mod 4), i.e. q = 4k + 3, where k is
n
a natural number. Define g : K ∗ → K ∗ , g(x) = x2 and take x, y ∈ K for which
g(x) = g(y). Then
n
(xy −1 )2 = 1, (xy −1 )4k+2 = (xy −1 )q−1 = 1

OLYMPIAD CORNER /89
and since (2n , 4k + 2) = 2, we get (xy −1 )2 = 1. Hence, y = ±x. Since 1 6= −1, we

q−1
deduce that the image of the function g has exactly elements. Define
2
n
Kn = {x2 | x ∈ K} = {0} ∪ Im g.
q−1 q+1
Then |Kn | = 1 + = .
2 2
q+1
If a ∈ K, then we also have |a − Kn | = |Kn | = and so, by the Pigeonhole
2
Principle Kn and a − Kn have an element in common. So, there exists u, v ∈ Kn
n n
such that u = a − v. Since u = x2 for some x ∈ K and v = y 2 for some y ∈ K,
n n
we conclude that a = x2 + y 2 , where x, y ∈ K.
OC495. A box contains 2017 balls. On each ball is written exactly one
integer. We randomly select two balls with replacement from the box and add the
numbers written on them. Prove that the probability of getting an even sum is
greater than 1/2.
Originally problem 2, First Round of 2017 Poland Math Olympiad.
We received 16 submissions. We present the solution by UCLan Cyprus Problem
Solving Group.
Assume that m balls have an even number written on them and n balls have an
odd number written on them. Then m + n = 2017 and therefore m 6= n. To get
an even sum we must either pick two balls with an even number written on them,
or pick two balls with an odd number written on them. So the probability that
we get an even sum is
m2 + n2 (m + n)2 + (m − n)2 20172 + 1 1

= > > .
20172 2 · 20172 2 · 20172 2
Editor’s Comment. Roy Barbara, Noah Garson, Kathleen E. Lewis, De Prithwijit

and Jason L. Smith generalized the problem and proved the statement in the case
in which the number of balls is any odd number. The proof is basically the same
as the one presented.

90/ Multifaceted Solutions to a Remarkable Geometry Puzzle
Multifaceted Solutions to a
Remarkable Geometry Puzzle
H. S. Hoffman and S. I.Warshaw
Introduction
In their book “Mathematical Curiosities” [1], Alfred Posamentier and Ingmar
Lehmann present and solve a mathematical puzzle involving the geometrical con-
figuration shown in Figure 1 [see 1, pp. 184–185 and pp. 237–238, respectively.]
Triangle ABP is inscribed in a semicircle having fixed radius R and center point
M , with vertices A and B on the semi-circle arc and vertex P on diameter DC,
such that sides AP and BP (of length S and T respectively) make fixed angles of
60◦ with the diameter and with each other. The puzzle asks us to show that length
U of chord AB is invariant for any position of P on the diameter between endpoints
C and D, with lengths S and T correspondingly changed to accommodate different
positions of P .
What is remarkable about this assertion is that we know from standard circle
theorems that chord AB is invariant when apex point P with fixed subtending
angle lies on the circle’s circumference. Seeing this invariance when P lies on a
diameter of the same circle is unexpected.
In this monograph we report different solutions to this puzzle and key properties
imbedded in the configuration geometry.
Section I
We draw radii AM and BM (each of length R) from center point M to points A
and B on the semi-circle of Figure 1 as shown in Figure 2, indicating the length
P M by X.
Since cos 60◦ = − cos 120◦ = 1/2, the cosine law equations for three triangles in
the figure simplify considerably. The lengths U , S and T are determined from this

H. S. Hoffman and S. I.Warshaw /91
cosine law triad:

4BP M : R2 = X 2 + T 2 − XT,
4AP B : U 2 = S 2 + T 2 − ST,
4AP M : R2 = X 2 + S 2 + XS.
Combining the first and third equations, we obtain T 2 − XT = S 2 + XS, which

becomes T 2 − S 2 = XS + XT , and thus (T − S)(T + S) = X(S + T ). This is
satisfied when T +S = 0 and when T −S = X. The first condition is unacceptable,
so the second one applies. On substituting T − S for X in the equation for BP M
(say), we obtain R2 = X 2 + T 2 − XT = S 2 + T 2 − ST . This in turn is seen equal
to U 2 in the equation for AP B. Thus U 2 = R2 , and U = R. 2
Triangle ABM is seen to be equilateral with fixed side length R for any position
of P on CD.
Section II
Using the proven result U = R, we circumscribe equilateral triangle AM B in
Figure 2, with circle center at Q and radii QA, QB and QM , as in Figure 3.

We find – unexpectedly – this circle now appears to include point P .
Four different approaches can be invoked to show that A, B, M and P indeed lie
on the same circle, as follows.
(1) A standard circle theorem establishes that the angles subtended by a given
chord of a circle from two different points on the same circle that lie on one side
of the chord are equal, and its converse is also true. We have proven that ABM
is equilateral, and that the angle at M that subtends chord AB is also 60◦ . This
equality of vertex angles AP B and AM B satisfies the converse of the theorem and
suffices to put P on the dotted circle.
(2) The internal angle property of a cyclic quadrilateral asserts that its opposing
vertex angles are supplementary, i.e., add up to 180◦ [See 1, pp. 155 - 158]. In
Figure 3, we see by inspection of ABP M that the vertex angles at P and B are
respectively 120◦ and 60◦ ; in triangle AP M , the vertex angles at A and M add
up to 60◦ , while for equilateral triangle ABM its vertex angles at A and M are
each 60◦ . The vertex angle pairs at A and M add up to 180◦ .
(3) Ptolemy’s theorem [see 1, pp. 157 - 158 and ref. 2, pp. 42 - 49] relates the
diagonals and opposing sides of a cyclic quadrilateral, and the form the Ptolemy
relationship takes for ABM P here is RT = RS + XR, which, immediately, is
T = X + S. This expression appeared in the algebraic development in Section I.
It is also known as van Schooten’s theorem. [See 2 (pp. 184 - 186) and 3].
(4) Directly establishing that point P actually lies on the circumcircle of triangle
AM B is equivalent to showing that the distance from P to the circumcenter Q of
AM B is equal to the circumscribed circle radius, or, alternatively, to the distance
between Q and any vertex at A, M or B. We provide this calculation in Section
III.
Our realization that a cyclic quadrilateral might be imbedded in the puzzle first
arose when we noticed the perpendicular bisector of chord P M (in Figures 1 – 3)
appeared to pass through the centroid Q of triangle AM B, which itself is equilat-
eral and the intersection of its side bisectors. Creating perpendicular bisectors of
the sides of all triangles in Figure 2 with a drawing app and seeing them intersect
in a single point at Q further strengthened this consideration.
Section III
We now provide solutions of the challenge proposed in item (4) of the previous
section, while using the convenience of coordinate geometry. Define a coordinate
system (x, y) with M as origin and x-axis on CD of Figure 3, as shown next in
Figure 4.

Because of the 60◦ angles at P , the x and y coordinates of the equilateral triangle
vertices and point P are as follows:
Point x y
M 0 √0
A −X − S/2 S √3/2
B −X + T /2 T 3/2
P −X 0
We also know from Section I that X = T − S. The coordinates of Q (the center of

equilateral triangle ABM ) are also the mean values of the coordinates of vertices
A, B and M , which are one-third of the sum of their respective x and y values.
Thus
xQ = (−2X + (T − S)/2 + 0)/3 = −X/2,

√ √
yQ = ((S + T ) 3/2)/3 = (S + T )/(2 3).
The value of xQ shows that if a vertical line x = −X/2 passes through point Q in
Figure 4, it indeed bisects the line segment P M on the semicircle diameter CD.
Thus triangle P QM is isosceles and length P Q equals circle radius QM . This
immediately proves that P is also a point of the circle that circumscribes triangle
ABM . 2
We close this section by calculating the lengths of each segment AQ, BQ, P Q and
M Q from their endpoint coordinates given above, and find that these lengths all
have the same squared value (X/2)2 + (S + T )2 /12. If one now uses X = T − S,
this becomes (S 2 + T 2 − T S)/3, which, from the second equation in Section√I (with
U = R), is exactly R2 /3. Each of these four segments thus has length R/ 3.
Section IV
A mysterious aspect of the puzzle configuration as presented in Figure 1 is the
triplet of adjacent 60◦ angles set up at point P . This angle arrangement is not as

ad hoc as it appears, because it is a consequence of having four points A, B, M and

P lie on a circle. We make this more explicit in Figure 5: an equilateral triangle
ABM is inscribed in a circle, with a fourth point P located elsewhere on the circle
rim to which lines from the triangle vertices A, B and M are drawn.
This arrangement of geometric elements is a copy of those shown in the preceding

Figures 2, 3 and 4 without the overarching semicircle. What is now evident is that
the two angles α and β in Figure 5 are each always 60◦ for any location of point
P on the circumscribing circle. Those are seen to be the same corresponding pair
of 60◦ angles at point P in all the previous figures.
We can now point out that Figure 5 (and each of the prior figures) shows that van
Schooten’s Theorem [see 3 and pp. 184 - 186 of 2] applies, which is an interesting

special case of Ptolemy’s theorem [see pp. 157 - 158 of 1 and pp. 42 - 49 of 2]
when three of the quadrilateral vertices form an equilateral triangle. We first draw
a horizontal line through point B in Figure 5 and extend line P A until it meets
this horizontal line at E. Let BE have length Z and AE have length Y as shown
in Figure 6. Then S + Y = Z = T. Also AEB and M P B are congruent triangles.
Thus X = Y and S + X = T , which is van Schooten’s theorem. [See 3 and pp.
184 - 186 of 2.]
Section V
The quantified geometry of the cyclic quadrilaterals provided so far leads one to
additional and immediate geometric insights. To that end we display in Figure
7 next the actual numerical values of the vertex angles actually used to draw all
Figures 1 through 4, which have been kept the same from figure to figure.
The length of P M (denoted by X in all earlier figures) that makes ∠P QM = 40◦

is calculated from X 2 = 2R2 (1 − cos 40◦ )/3.
Some well-known “circle theorems” are quickly seen to be embedded in Figure
7, and the “internal angles” discussions in Section II apply. For instance, of the
three angles at A, B and Q that subtend chord P M , the angle at Q is twice the
angles at A and B, since Q is at the center of the circle and A and B are on its
circumference. Thus vertex angles at A and B that subtend the same arc and
chord P M are also equal. Numerical length values and relationships are explicitly
evident: starting with equilateral triangle
√ ABM , we have AM = M B = BA = R,
and so QB = QA = QM = QP = R/ 3.

Acknowledgments
We wish to thank Alfred Posamentier and Ingmar Lehmann for graciously re-
sponding to our e- mailed inquiries about their puzzle. In particular, AP and IL
encouraged SW in the observation that the problem has a circle-circumscribed
quadrilateral imbedded in it, and HH was informed that copyright issues of ref-
erencing their Problem 39 can be obviated with appropriate source acknowledge-
ment, which we have provided in the references section herewith.
We also thank Gerald Minerbo, scholar, colleague and friend, for discussions of
the solutions presented, their ramifications, and pertinent editorial critiques.
References
[1] Posamentier, Alfred and Lehmann, Ingmar, Mathematical Curiosities, Prometheus
Books, 2014. (paperback)
[2] Pritchard, Chris (Ed.), The Changing Shape of Geometry, Mathematical Asso-
ciation of America, Cambridge University Press, 2003 (paperback)
[3] Viglione, Raymond, Proof Without Words: van Schooten’s Theorem, Mathe-
matics Magazine 89(2):132, April 2016. Also https://www.researchgate.net/
publication/303865413

Problems /97
PROBLEMS
Click here to submit problems proposals as well as solutions, comments
and generalizations to any problem in this section.
4611. Proposed by Nguyen Viet Hung.

Evaluate
1 1 1
4 π + 4 3π + 4 5π .
sin 14 sin 14 sin 14
4612. Proposed by Mihaela Berindeanu.

In the convex quadrilateral ABCD, we have
] (BAC) = ] (CAD) and ] (CDA) = ] (BCA) .
Denote O ∈ AC, X ∈ BC, Y ∈ CD such that OA = OC, AX ⊥ BC and
AY ⊥ CD. The perpendicular line from A to XY cuts BD at Z. Show that
−→ −→ −−→ −−→
OZ = OA + OX + OY .
4613. Proposed by Daniel Sitaru.

Let A and B be n × n real matrices with n ∈ N, n ≥ 2 such that AB = BA. Show
that
det(4(A2 + B 2 ) + AB + 3(A + B) + In ) ≥ 0.
4614. Proposed by Florin Stanescu.

Let k be a given natural number and let (an )n≥1 be a sequence such that
1 a1 a2 an
lim k + + ··· + = 1.
n→∞ n 1 2 n
a1 + a2 + · · · + an

Prove that the sequence is convergent by finding its
nk+1 n≥1
limit.
4615. Proposed by Anthony Garcia.

R1 f (1)
Let f be a twice differentiable function on [0, 1] such that 0
f (x)dx = 2 . Prove
that Z 1
(f 00 (x))2 dx ≥ 30(f (0))2 .
0

98/ Problems
4616. Proposed by Marius Drăgan, modified by the Editorial Board.

For each suitable point N on side AC of ∆ABC define P to be the point where
the line parallel to AB meets the side BC, and M to be the point on side AB for
which ∠M N A = ∠B. If the area of ∆ABC equals 1, determine the maximum
area of triangle M P N .
4617. Proposed by Nermin Hodzic, Adnan Ali and Salem Malikic.

Let a, b, c be positive real numbers such that
a b c
+ + = 2.
b+c c+a a+b
√
3
Show that max(a, b, c) ≥ 9abc.
4618. Proposed by Cherng-tiao Perng.

Let C be a nondegenerate conic and L be a line. Let O, P be two distinct points
such that O, P ∈/ L and P ∈ C. Denote the alternative intersection of OP and C
by Q0 . Furthermore let P 0 be a point on OP such that P 0 ∈
/ L. For any Q on C
other than Q0 , let
QP ∩ L = {D} and DP 0 ∩ QO = {Q0 }.
Prove that when Q varies on C, Q0 moves on a fixed conic through P 0 .
4619. Proposed by D. M. Bătineţu-Giurgiu and Neculai Stanciu.

Pn 1 Pn 1
Consider the sequences an and bn such that an = k=1 2 and bn = k=1 .
4 k (2k − 1)2
π
Compute lim − an bn n.
n→∞ 48
4620. Proposed by Alpaslan Ceran.

Consider three semicircles in the configuration below:
1 1 1
Prove that = + .
x a b
.................................................................

Problems /99
Cliquez ici afin de proposer de nouveaux problèmes, de même que pour

offrir des solutions, commentaires ou généralisations aux problèmes
proposś dans cette section.
4611. Proposée par Nguyen Viet Hung.

Évaluer
1 1 1
4 π + 4 3π + 4 5π .
sin 14 sin 14 sin 14
4612. Proposée par Mihaela Berindeanu.

Dans le quadrilatère convexe ABCD, ] (BAC) = ] (CAD) et ] (CDA) = ] (BCA) .
Dénoter O ∈ AC, X ∈ BC, Y ∈ CD tels que OA = OC, AX ⊥ BC et AY ⊥ CD.
La ligne perpendiculaire de A vers XY intersecte BD en Z. Démontrer que
−→ −→ −−→ −−→
OZ = OA + OX + OY .
4613. Proposée par Daniel Sitaru.

Soient A et B des matrices n × n ré elles tels que AB = BA, où n ∈ N, n ≥ 2.
Démontrer que
det(4(A2 + B 2 ) + AB + 3(A + B) + In ) ≥ 0.
4614. Proposée par Florin Stanescu.

Soit k un nombre naturel et soit (an )n≥1 une suite telle que
1 a1 a2 an
lim + + · · · + = 1.
n→∞ nk 1 2 n
a1 + a2 + · · · + an

Démontrer que la suite est convergente et calculer sa
nk+1 n≥1
limite.
4615. Proposée par Anthony Garcia.

R1 f (1)
Soit fonction f qui est deux fois dérivable sur [0, 1] et telle que 0
f (x)dx = 2 .
Démontrer que Z 1
(f 00 (x))2 dx ≥ 30(f (0))2 .
0

100/ Problems
4616. Proposée par Marius Drăgan, modifié par le Comité de rd́action.

Soit le point N sur le côté AC de ∆ABC, soit P le point où la ligne parallèle à
AB intersecte BC, et soit M le point sur le côté AB tel que ∠M N A = ∠B. Si la
surface de ∆ABC est égale à 1, déterminer la plus grande valeur possible pour la
surface de ∆M P N .
4617. Proposée par Nermin Hodzic, Adnan Ali et Salem Malikic.

Soient a, b, c des nombres réels positifs tels que
a b c
+ + = 2.
b+c c+a a+b
√
Démontrer que max(a, b, c) ≥ 3 9abc.
4618. Proposée par Cherng-tiao Perng.

Soit C une conique non dégénérée et soit L une ligne. Soient O, P des points
distincts tels que O, P ∈
/ L et P ∈ C. Soit alors Q0 le deuxième point d’intersection
de OP et C. De plus, soit P 0 un point sur OP tel que P 0 ∈ / L. Pour tout Q sur C
autre que Q0 , soit
QP ∩ L = {D} et DP 0 ∩ QO = {Q0 }.
Démontrer que lorsque Q varie le long de C, Q0 se déplace sur une certaine conique
passant par P 0 .
4619. Proposée par D. M. Bătineţu-Giurgiu et Neculai Stanciu.

Pn 1 Pn 1
Soient des suites an et bn telles que an = k=1 2 et bn = k=1 .
4 k (2k − 1)2
π
Calculer lim − an bn n.
n→∞ 48
4620. Proposée par Alpaslan Ceran.

Soient trois demi cercles, tels qu’indiqués:
1 1 1
Démontrer que = + .
x a b

Solutions /101
SOLUTIONS
No problem is ever permanently closed. The editor is always pleased to consider for
publication new solutions or new insights on past problems.
4561. Proposed by Michel Bataille.

Let n be an integer with n ≥ 2 and let w1 , w2 , . . . , wn be distinct complex numbers
n
Q
such that w1 + w2 + · · · + wn = 1. For k = 1, 2, , . . . , n, let Pk (x) = (x − wj ).
j=1,j6=k
If z is a complex number, evaluate
n
X wk Pk (zwk )
.
Pk (wk )
k=1
There were 7 correct solutions, three of which used calculus of residues. We present
4 solutions.
The sum is equal to z n−1 .
Solution 1, by UCLan Cyprus Problem Solving Group.
Let " #
n
X wk Pk (zwk )
g(z) = (z − 1) .
Pk (wk )
k=1
Observe that the quantity in square brackets equals 1 when z = 1. We will show
that g(0) = g 0 (0) = · · · = g (n−2) (0) = 0 so that g(z) = (z − 1)z n−1 .
Qn
Let P (z) = k=1 (z − wk ). When z 6= 1, for 1 ≤ k ≤ n,
(z − 1)wk P (zwk )

wk Pk (zwk ) P (zwk )
(z − 1) = = 0 .
Pk (wk ) (zwk − wk )Pk (wk ) P (wk )
This equation also holds for z = 1, so that, for 0 ≤ m ≤ n − 1,

n n
X P (zwk ) X wm P (m) (zwk )
g(z) = and g (m) (z) = k
.
P 0 (wk ) P 0 (wk )
k=1 k=1
Hence
n
X wkm
g (m) (0) = P (m) (0) .
P 0 (wk )
k=1
m
Let hm (z) = z /P (z) for 0 ≤ m ≤ n − 2. The function hm (z) has a simple pole
at wk with residue
wkm /Pk (wk ) = wkm /P 0 (wk ).

102/ Solutions
Suppose that CR is a circle centred at the origin whose interior contains all the
values wk . Then
P (m) (0)
I
(m)
g (0) = hm (z) dz
2πi CR
and
|P (m) (0)|
|g (m) (0)| ≤ (2πR)[max |hm (z)|].
2π CR
Since the degree of P (z) exceeds m by at least 2, limR→∞ maxCR |hm (z)| = 0, and
the result follows. Hence
n
X wk Pk (zwk )
= z n−1 .
Pk (wk )
k=1
Solution 2, by Madhav Modak.

Let Y
D ≡ D(w1 , . . . , wk , . . . , wn ) = (wi − wj ),
1≤i<j≤n
and
m n
X wk Pk (zwk ) X wk (−1)i−1 D(w1 , . . . , zwk , . . . , wn )
f (z) = = .
Pk (wk ) (−1)i−1 D(w1 , . . . , wk , . . . wn )
k=1 k=1
The Vandermonde matrix V whose (i, j)th element is wji−1 has determinant
n
X
n D = wji−1 Tij ,
i=1
n
where n = (−1)( ) and Tij is the cofactor of wji−1 and the determinant is expanded
2
according to the jth column.

By replacing wk by zwk , we find that
n
X
n D(w1 , . . . , zwk , . . . , wn ) = z i−1 wki−1 Tik ,
i=1
whence !
n n
n X X i
f (z) = wk Tik z i−1 .
D i=1
k=1
n−1
Pn
The coefficient of z is equal to n /D times k=1 wkn Tnk , which is the expansion
of the matrix Vn obtained from V by replacing the last row by (w1n , w2n , . . . , wnn ).
The determinant of Vn is equal to n D multiplied by w1 + w2 + · · · + wn = 1. Thus
the coefficient of z n−1 is 1.
Pn
When i ≤ n − 1, k=1 wki Tik is the expansion of the matrix Vi obtained from V by
replacing the ith row by (w1i , w2i , . . . , wni ), making it identical to the following row.
Hence the coefficient of z i−1 is 0 when 1 ≤ i ≤ n − 1. It follows that f (z) = z n−1 .

Solutions /103
Solution 3, by C.R. Pranesachar.

Let Qn
s (zs − wk )
g(s) = Qnk=1 .
k=1 (s − wk )
Since the degree of the numerator as a polynomial in s exceeds that of the denom-
inator by 1, we can write
n
X Ak
g(s) = sz n + C + ,
s − wk
k=1
where C and Ak are polynomials in z.

Hence
n
Y n
Y n
X
s (zs − wk ) = (sz n + C) (s − wk ) + Ak Pk (s).
k=1 k=1 k=1
Setting s = wk , we find that
wk2 (z − 1)Pk (zwk )
Ak = .
Pk (wk )
Equating the coefficients of sn leads to
−z n−1 = z n−1 (−w1 − w2 − · · · − wn ) = C + z n (−w1 − w2 − · · · − wn ) = C − z n .
Hence
n
X wk2 Pk (zw)
f (s) = sz n + z n−1 (z − 1) + (z − 1) .
(s − wk )Pk (wk )
k=1
n−1
Setting s = 0 gives the required function as z .
Solution 4, by the proposer.

Qn Pn
Let P (s) = k=1 (s − wk ) = sn − sn−1 + j=2 aj sn−j . Fix k and define Um (s) =
sm + wk sm−1 + · · · + wkm−1 + wkm . Using the fact that
sm sm − wkm wkm wkm
= + = Um−1 (s) +
s − wk s − wk s − wk s − wk
for 1 ≤ m ≤ n + 1, we find that
n
sP (s) sn+1 sn X aj sn+1−j
sPk (s) = = − +
s − wk s − wk s − wk j=2 s − wk
n
X wk P (wk )
= Un (s) − Un−1 (s) + aj Un−j (s) +
j=2
s − wk
n
X
= Un (s) − Un−1 (s) + aj Un−j (s)
j=2

104/ Solutions
Since Um (zwk ) = wkm (1 + z + · · · + z m ),
zwk Pk (zwk ) = (1 + z + · · · + z n )wkn − (1 + z + · · · + z n−1 )wkn−1 + Q(wk ),

Pn
where Q(s) = j=2 bj sn−j is a polynomial whose coefficients bj depend on z but
are independent of k. Hence
n
X wk Pk (zwk )
Pk (wk )
k=1
 
n n n n−1 n n n−j
1 X w X w X X w
= (1 + z + · · · + z n ) k
− (1 + z + · · · + z n−1 ) k
+ bj k 
z Pk (wk ) Pk (wk ) j=2 Pk (wk )
k=1 k=1 k=1
1
= [(1 + z + · · · + z n ) − (1 + z + · · · + z n−1 ) + 0] = z n−1 ,
z
using the result established in the following appendix.
Appendix. The following result, used in Solution 4, Solution 1 and by one other
solver, is of independent interest and may not be readily accessible to the reader.
This proof is supplied by the proposer.
n n n
X wkn X wkn−1 X wkm
= w1 + w2 + · · · + wn , = 1, = 0,
Pk (wk ) Pk (wk ) Pk (wk )
k=1 k=1 k=1
for 0 ≤ m ≤ n − 2.
Proof. Let w = w1 + w2 + · · · + wn and define
n
Y
P (x) = (x − wk ) = xn − wxn−1 + U (x),
k=1
where U (x) is a polynomial of degree less than n − 1. The Lagrange polynomial

Q(x) of degree less than n taking values Q(wk ) when x = wk is given by
n n
X Q(wk )Pk (x) X Q(wk )P (x)
Q(x) = = .
P 0 (wk ) Pk (wk )(x − wk )
k=1 k=1
Set Q(x) ≡ 1 to obtain

n
1 X 1
= .
P (x) Pk (x)(x − wk )
k=1
Observe that xn+1 = (x + w)(xn − wxn−1 + U (x)) + V (x), with the degree of V (x)
less than n, so that
xn+1 V (x)
=x+w+ .
P (x) P (x)

Solutions /105
Also
n n
X xn+1 − wn+1 n
xn+1 X xn+1 k
X wkn+1
= = +
P (x) Pk (wk )(x − wk ) Pk (wk )(x − wk ) Pk (wk )(x − wk )
k=1 k=1 k=1
n n
X xn + wk xn−1 + · · · + wkn X wkn+1
= +
Pk (wk ) Pk (wk )(x − wk )
k=1 k=1
n n
!
X X wm R(x)
= k
xn−m + ,
m=0
P k (w k ) P k (wk )P (x)
k=1
where the degree of R(x) is less than n. Hence, by equating polynomials parts,
n n
!
X X wkm
x+w = xn−m ,
m=0
Pk (wk )
k=1
and the result follows from a comparision of coefficients.
4562. Proposed by Pericles Papadopoulos.

Let P be the intersection point of the diagonals AC and BD of a convex quadri-
lateral ABCD. The angle bisector of the opposite angles ∠AP D and ∠BP C
intersects AD and BC at points K and M respectively, and the angle bisector of
the opposite angles ∠AP B and ∠CP D intersects AB and DC at points L and N
respectively. Show that:
(a) (DK)(AL)(BM )(CN ) = (KA)(LB)(M C)(N D).
(b) Cevians AM , BP , CL concur at point Q, cevians BN , CP , DM concur at
point R, cevians AN , DP , CK concur at point S, and cevians DL, BK,
P A concur at point T .

106/ Solutions
We received 13 solutions, all essentially the same. Here, then, is the common
solution.
(a) P K, P L, P M, and P N are (interior) bisectors of the angles at P in triangles
DP A, AP B, BP C, and CP D respectively. Therefore,
DK PD AL PA BM PB CN PC
= , = , = , and = .
KA PA LB PB MC PC ND PD
Multiplying these expressions, we get
DK AL BM CN
· · · = 1,
KA LB M C N D
which proves part (a).
Comment by Sergey Sadov. For part (a) it is not necessary that P be the in-
tersection point of the diagonals — the same argument (word for word) proves
the identity for any point P inside a convex quadrilateral that is joined by line
segments to the four vertices. Indeed, the analogous identity holds similarly for
an arbitrary point P inside a convex n-gon for any n ≥ 3.
(b) Using recriprocals of two of the equal ratios from part (a), we have
AP CM BL AP P C P B
· · = · · = 1.
P C M B LA P C P B AP
Because we assume here that P lies between A and C on a diagonal of the given
convex quadrilateral, Ceva’s theorem applied to ∆ABC implies that AM, BP, CL
are concurrent at a point inside the triangle. The rest of part (b) follows by a
cyclic relabeling of points.
Editor’s comments. Sadov describes several further properties related to the given
configuration. For example, he finds that the lines AB, KM , and SR are concur-
rent or parallel, and those six points lie on a conic.

Solutions /107
Other incidences arise by replacing points K, L, M, N in part (b) by K 0 , L0 , M 0 , N 0 ,

where now the bisectors of angles AP B and CP D meet the lines AD and BC at
K 0 and M 0 , while the other bisector meets AB and CD at L0 and M 0 . He also
discusses a number of other conics associated with the figure; for example, if S 0 is a
point on the line AN and we define Q0 = P S 0 ∩ CL, then the points Q, Q0 , S, S 0 , R,
and T lie on a conic. Consequently,
• The points Q, R, S, T, L, N lie on a conic.
• The ellipse through the points Q, R, S, T that is tangent to the line AN at
S is tangent to CL at Q.
4563. Proposed by George Stoica.

Find all perfect squares in the sequence x0 = 1, x1 = 2, xn+1 = 4xn − xn−1 , n ≥ 1.
We received 11 submissions, of which 7 were correct and complete. We present
the solution by Theo Koupelis.
2
√ the given recurrence relation is r − 4r + 1 = 0,
The characteristic equation of
whose solutions are r√= 2 ± 3. Therefore,
√ the general term of the sequence is
given by xn = A(2 + 3)n + B(2 − 3)n , for all n ≥ 0, where A, B are constants.
Taking into account that x0 = 1 and x1 = 2, we find that A = B = 1/2, and
therefore, for n ≥ 0,
1h √ √ i
xn = (2 + 3)n + (2 − 3)n . (1)
2
Clearly x0 = 1 is a perfect square. We will show that no other term xn , where

n ≥ 1, is a perfect square.
We start by examining the terms of the sequence modulo 3 and modulo 5. Using
the first two terms and the recursive definition, the residues of the sequence modulo
3 are 1, 2, 1, 2, 1, 2, 1, · · · ; in particular, 3 - xn for any n. Similarly, modulo 5
the residues are 1, 2, 2, 1, 2, 2, 1, · · · ; that is, x3k ≡ 1 mod 3 for k a non-negative
integer and xn ≡ 2 mod 5 for n not a multiple of 3. Since 2 is not a quadratic
residue mod 5, if xn is to be a perfect square, then we must have n = 3k for k a
non-negative integer.
√ √
From (1) and the fact that (2 + 3)(2 − 3) = 1 we have
√ √ i3 √ √ i

1 h h
x3k = (2 + 3)k + (2 − 3)k − 3 (2 + 3)k + (2 − 3)k ,
2
or
x3k = xk · (4x2k − 3).
We know that x0 = 1 is a perfect square while x3 = 26 is not. Now let n = 3k

be the smallest positive integer for which x3k is a perfect square. Then if d =
gcd(xk , 4x2k − 3), then d|3 and therefore d = 3 or d = 1. But d 6= 3 because

108/ Solutions
3 - xn . Therefore d = 1. But if x3k is a perfect square, and it is a product of two

terms that have no common divisor, each term must be a perfect square. Thus
xk must be a perfect square, which contradicts the assumption that n = 3k was
the smallest positive integer for which xn is a perfect square. Therefore the only
perfect square in the given sequence is x0 = 1.
4564. Proposed by Alijadallah Belabess.

Let a, b, c and d be non-negative real numbers with ab + bc + cd + da = 4. Prove
that:
a3 + b3 + c3 + d3 + 4abcd ≥ 8
We received 10 submissions, all correct. We present the solution by Marie-Nicole

Gras, slightly modified by the editor.
By hypothesis, we have ab + bc + cd + da = (a + c)(b + d) = 4. Without loss of
generality, we may assume that a + c ≥ 2.
The given inequality is equivalent to
F := 4(a + c)3 a3 + b3 + c3 + d3 + 4abcd − 8 ≥ 0,

or F = 4(a + c)3 (a + c)3 + (b + d)3 − 3ac(a + c) − 3bd(b + d) + 4abcd − 8 . Since
(a + c)(b + d) = 4, and 4ac = (a + c)2 − (a − c)2 , we can write F = G + H where
G = 4(a + c)6 + 256 − 3(a + c)4 (a + c)2 − (a − c)2 − 32(a + c)3

= (a + c)6 − 32(a + c)3 + 256 + 3(a + c)4 (a − c)2

2
= (a + c)3 − 16 + 3(a + c)4 (a − c)2 , and

H = −48bd(a + c)2 + 16abcd(a + c)3

= −48 b(a + c) d(a + c) + 16ac(a + c) b(a + c) d(a + c) .
Set x = b(a+c). Then by (a+c)(b+d) = 4, we have d(a+c) = 4−b(a+c) = 4−x,

x(4 − x) ≥ 0, and

H = −48x(4 − x) + 16ac(a + c)x(4 − x) = 16x(4 − x) ac(a + c) − 3 . (1)

Then F = (a + c)3 − 16 + 3(a + c)4 (a − c)2 + 16x(4 − x)[ac(a + c) − 3]. Hence
F ≥ 0 if ac(a + c) ≥ 3.
If ac(a + c) < 3, then by (1) we can write

F = H + G = 16x(4 − x) ac(a + c) − 3 + G
= 16(x − 2)2 3 − ac(a + c) + 64 ac(a + c) − 3 + G.

(2)

Solutions /109

Let T = 64 ac(a + c) − 3 + G. Then by (2) we get
T = 16(a + c) (a + c)2 − (a − c)2 − 192 + (a + c)6 − 32(a + c)3 + 256 + 3(a + c)4 (a − c)2

= (a + c)6 − 16(a + c)3 + 64 + 3(a + c)4 (a − c)2 − 16(a + c)(a − c)2

2
= (a + c)3 − 8 + (a + c)(a − c)2 3(a + c)3 − 16 .

(3)
Since a + c ≥ 2, we see from (3) that T ≥ 0, so finally we have from (2) that
F = 16(x − 2)2 [3 − ac(a + c)] + T ≥ 0, completing the proof.
Editor’s comment. Out of the ten solvers, four of them also showed that equality
holds if and only if (a, b, c, d) = (1, 1, 1, 1) or (21/3 , 22/3 , 21/3 , 0) together with all
its cyclic permutations.
4565. Proposed by Daniel Sitaru.

Let ma , mb and mc be the lengths of the medians of a triangle ABC. Prove that
4(amb mc + bmc ma + cma mb ) ≥ 9abc.
We received 11 solutions, one of which was incorrect. We present the solution by

Sergey Sadov.
Consider the triangle in the complex plane. Let the origin (complex zero) be at
the center of mass of the triangle and u, v, w be the complex coordinates of the
midponts of the sides a, b, and c, respectively. Then
ma = 3|u| mb = 3|v|, mc = 3|w|,
and
a = 2|v − w|, b = 2|w − u|, c = 2|u − v|.
Put
4 ma mb u v
ξ= · · = · ,
9 a b v−w w−u
4 mb mc v w
η= · · = · ,
9 b c w−u u−v
4 mc ma w u
ζ= · · = · .
9 c a u−v v−w
The required identity takes the form |ξ| + |η| + |ζ| ≥ 1, and it follows, by the
triangle inequality, from the identity ξ + η + ζ = −1, which we are about to prove.
Equivalently, we want to prove that
(u − v)(v − w)(w − u) + uv(u − v) + vw(v − w) + wu(w − u) = 0.

110/ Solutions
Consider the coefficients at powers of u:

u2 : (w − v) + v − w = 0,
u1 : (v − w)(v + w) − v 2 + w2 = 0,
u0 : vw(w − v) + vw(v − w) = 0.
The proof is finished.
A generalization. In the above proof we did not use the relation u + v + w = 0.

Therefore we have in fact proved a more general fact:
Let D be any point in the plane of triangle ABC. Then
AD · BD · c + BD · CD · a + CD · AD · b ≥ abc.
The given problem is equivalent to the particular case of this proposition with D
being the center of mass.
Case of equality. A natural question to ask is: when, in the described generaliza-
tion, does the inequality turn to equality. I will show that this happens if and only
if D is the orthocenter. As a corollary, in the original problem the equality takes
place only for the equilateral triangle.
For the equality
| − 1| = |ξ + η + ζ| = |ξ| + η| + ζ|
to hold, it is necessary and sufficient that ξ, η, ζ be real and nonpositive. At least

one of them is nonzero. Suppose ξ 6= 0 and consider the condition ξ < 0. It means
that
w−v w−u
· > 0.
v u
Hence the arguments of the complex numbers (w − v)/v and (w − u)/u have equal
magnitudes and opposite signs. Geometrically it means that the signed magni-
tudes of the angles DBA and ACD (considering the counterclockwise direction as
positive) are equal.
Denote the unsigned magnitude of the angles as ∠DBA = ∠DCA = α0 , ∠DAB =
∠DCB = β 0 and ∠DAC = ∠DBC = γ 0 . Then
β 0 + γ 0 = α ( = ∠A), α0 + β 0 = γ, γ 0 + α0 = β, 2(α0 + β 0 + γ 0 ) = π.
It follows that α0 = π/2 − α etc. This condition defines the orthocenter.
Editor’s note. Several other solvers (Gayen, Giugiuc, Janous, Văcaru, and the
proposer) also asserted the generalization of the inequality to an arbitrary point
in the plane of the triangle, with Gayen and Giugiuc citing this generalization
as Hyashi’s Inequality and Janous indicating it as a generalization of Murray
Klamkin’s “Polar Moment of Inertia” Inequality.

Solutions /111
4566. Proposed by J. Chris Fisher.

Given three circles, α, β, and γ with centers A, B, C and radii a, b, c, respectively,
where γ is tangent to α at A0 and to β at B 0 , either both circles externally or both
internally. One exterior common tangent line is tangent to α at S and to β at T .
(a) Prove that the lines A0 S and B 0 T intersect at a point of γ.

(b) Show that
c2 · ST 2
(A0 B 0 )2 = ,
(c ± a)(c ± b)
where the plus signs are used when α and β are externally tangent to γ, and the
negative signs when internally tangent to γ.
Comment. Part (b) is problem 1.2.8 on page 5 of H. Fukagawa and D. Pedoe,
Japanese Temple Geometry Problems, San Gaku (The Charles Babbage Research
Centre, 1989). Instead of a proof, the authors provide (on page 82) a reference to a
19th century Japanese geometry text together with the comment, “Called ‘Three
Circles and Tangent Problem’, or ‘Sanen Bousha’, and applied in the solution to
many problems.”
All 6 submissions were complete and correct. We have selected a different corre-
spondent for each part.
(a) Solution by Sergey Sadov.
Let P1 be the second point of intersection of the line A0 S with circle γ. Similarly,
let P2 be the second point of intersection of the line B 0 T with circle γ. We will
prove that P1 = P2 .
The circles α and γ are homothetic with homothety centre A0 . Under this ho-
mothety, the triangle A0 AS corresponds to the triangle A0 CP1 . It follows that
CP1 ||AS. Similarly, CP2 ||BT .
Since AS ⊥ ST and BT ⊥ ST , we conclude that AS||BT , hence CP1 ||CP2 ; hence,
P1 = P2 .

112/ Solutions
(b) Solution by Marie-Nicole Gras.

Let L be the point of AS such that AL = BT . The quadrilateral ABT L is a
parallelogram so that T L = AB; in the right-angled 4LST , we have
AB 2 = LT 2 = ST 2 + (a − b)2 .
The cosine law applied to the isosceles 4A0 CB 0 gives us A0 B 02 = 2c2 (1−cos(∠C)),
and in 4ACB, we have
CA2 + CB 2 − AB 2
cos(∠C) = ,
2 CA·CB
2 CA·CB − CA2 − CB 2 + AB 2 AB 2 − (CA − CB)2
1 − cos(∠C) = = ·
2 CA·CB 2 CA·CB
When α and β are internally tangent to γ,

CA = c − a, CB = c − b, and (CA − CB)2 = (a − b)2 .
When α and β are externally tangent to γ,
CA = c + a, CB = c + b, and (CA − CB)2 = (a − b)2 .
We deduce that
ST 2 + (a − b)2 − (a − b)2 c2 ·ST 2
A0 B 02 = c2 = ·
CA·CB (c ± a)(c ± b)
4567. Proposed by Paul Bracken.

Prove that for any n ∈ {0, 1, 2, 3, . . .}, the following holds
n
k 2n + 1
X
(−1) (2k + 1)2n+1 = (−1)n 22n (2n + 1)!
n−k
k=0
We received 10 submissions and they were all correct. We present 3 solutions.

The first step is a reduction. We can transform the left-hand side in order to
obtain a sum over all values of k from 0 to 2n + 1, which was observed by all
solvers. Let Sn be the left-hand side of the equation. Note that
n
n−k 2n + 1
X
Sn = (−1) (2n + 1 − 2k)2n+1
k
k=0
n
X
n−k 2n + 1
= (−1) (2n + 1 − 2(2n + 1 − k))2n+1 (−1)2n+1
2n + 1 − k
k=0
2n+1
n−m 2n + 1
X
= (−1) (2n + 1 − 2m)2n+1 ,
m=n+1
m

Solutions /113
where we change to the variable m = 2n + 1 − k in the last step. So it suffices to

show that
2n+1
X 2n + 1
2Sn = (−1)n−k (2n + 1 − 2k)2n+1 = (−1)n 22n+1 (2n + 1)!. (1)
k
k=0
The second step is to prove the binomial identity (1).
Solution 1, by Seán M. Stewart, simplified by the editor.

We have the following well-known identity: for any n ∈ N ∪ {0},
2n+1
(−1)n X

k 2n + 1
sin2n+1 x = (−1) ei(2n−2k+1)x . (2)
22n+1 i k
k=0
Indeed, by Euler’s formula and the binomial theorem, we have

2n+1 2n+1
eix − e−ix (−1)n X

2n+1 k 2n + 1
sin x= = 2n+1 (−1) ei(2n−2k+1)x .
2i 2 i k
k=0
Now we differentiate both sides of equation (2) with respect to x by 2n + 1 times

and then evaluate the result at x = 0. We start by differentiating the left-hand
side using the general Leibniz rule:
2n+1
2n+1 (2n+1)
X (2n + 1)! Y
(sin x) = (sin x)(kj ) . (3)
k1 !k2 ! · · · k2n+1 ! j=1
k1 +k2 +···+k2n+1 =2n+1
Note that sin 0 = 0, so when we are evaluating at x = 0, on the right-hand side

of the equation (3), the only term that survives is the term with k1 = k2 = · · · =
k2n+1 = 1. So when x = 0, equation (3) can be simplified to
(2n+1)
sin2n+1 x (0) = (2n + 1)!(sin0 (0))2n+1 = (2n + 1)!.
For the right-hand side of equation (2), the (2n + 1)-st order derivative is
2n+1
(−1)n X

k 2n + 1 2k+1 2n+1 i(2n−2k+1)x

2n+1
(−1) i (2n − 2k + 1) e
2 i k
x=0
k=0
2n+1
(−1)n X

k 2n + 1
= 2n+1 (−1) (2n − 2k + 1)2n+1 .
2 k
k=0
Combining the two ways of the computation, we get (1), as required.

114/ Solutions
Solution 2, by Sergey Sadov.

For any polynomial P (x) the expression
r
X r
∇r P (x) = (−1)m P (x − m)
m=0
m
is known as the (backward) finite difference of order r of P (·) at the point x.

We will employ the known fact that if P (x) = axr + (terms of degree < r), then
∇r xr = ar! (the constant function). Applying this formula to P (x) = (2x+1)2n+1 ,
we get
2Sn = (−1)n ∇2n+1 P (x)|x=2n+1 = (−1)n · 22n+1 (2n + 1)! = (−1)n 22n (2n + 1)!.
Solution 3, by the majority of the solvers, slightly modified by the editor.

We will show that for any positive integer m and an integer t, we have
m
(
if 0 ≤ t ≤ m − 1,

X m 0,
(−1)j jt = m
(4)
j=0
j (−1) m!, if t = m.
By the binomial theorem, it is clear that equation (4) implies that

m
X
jm
(−1) (a + bj)m = (−b)m · m!. (5)
j=0
j
Taking m = a = 2n + 1 and b = −2 in equation (5), we obtain (1).

d
To prove (4), we apply t times the differential operator x on the binomial
dx
identity
m
m
X m j
(1 + x) = x .
j=0
j
Note that the right-hand side becomes

m
X m m j−1
j x .
j=1
j
If t ≤ m − 1, then (1 + x) is a factor of all terms of the left-hand side, and if t = m,

the left-hand side is
(1 + x)Q(x) + m! xm
for some polynomial Q(x). Then we substitute x = −1 and obtain identity (4).
Editor’s Comment. As pointed out by Marie-Nicole Gras, this problem is similar
to Problem 4463 in Crux Vol. 46 (2).

Solutions /115
4568. Proposed by Song Qing, Leonard Giugiuc and Michael Rozenberg.

Let k be a fixed positive real number. Consider positive real numbers x, y and z
such that
xy + yz + zx = 1 and (1 + y 2 )(1 + z 2 ) = k 2 (1 + x2 ).
Express the maximum value of the product xyz as a function of k.

We received 9 submissions of which 6 were correct and complete. We present the
solution by Arkady Alt, slightly modified.
Since xy + yz + zx = 1, the equation (1 + y 2 )(1 + z 2 ) = k 2 (1 + x2 ) is equivalent
to each of:
xy + yz + zx + y 2 xy + yz + zx + z 2 = k 2 xy + yz + zx + x2 ,

(y + z)(x + y)(y + z)(x + z) = k 2 (x + z)(x + y),

(y + z)2 = k 2 ,
y + z = k.
Let t = xyz, then since y + z = k, we have

t
1 = (xy + zx) + yz = kx +
x
and so t = x(1 − kx). Thus, yz = 1 − kx and y + z = k, so, by the AM-GM
inequality,
(y + z)2 k2
1 − kx = yz 6 = .
4 4
1 k 1
Hence, x > k − 4 and we are to maximize h(x) = x(1 − kx) when x > k − k4 .
1 1 k
Since h0 (x) = 1 − 2kx, h(x) is decreasing when < − . That is, when
√ 2k k 4
0 < k < 2. For such k,

k 4 − k2

1 k 1 k 1 k
max t = h − = − 1−k − = .
k 4 k 4 k 4 16
√ 1 k 1 1
Likewise, if k > 2, then − 6 so 2k is in the domain of h(x) and
k 4 2k

1 1 1 1
max t = h = 1−k· = .
2k 2k 2k 4k
√

k 4 − k2
if k ∈ 0, 2


Thus, max (xyz) = 16 √
1
if k ≥ 2


4k

116/ Solutions
4569. Proposed by Nguyen Viet Hung.

Solve the following equation in the set of real numbers
1 x+1 2x+1
8x + 27 x + 2x+1 · 3 x + 2x · 3 x = 125.
We received 20 submissions, of which 18 were correct and complete. We present

the solution by the UCLan Cyprus Problem Solving Group.
Let a = 2x and b = 31/x . Then a3 + b3 + 6ab + 9ab = 125 and therefore
1
0 = a3 + b3 + (−5)3 − 3ab(−5) = (a + b − 5)((a − b)2 + (a + 5)2 + (b + 5)2 ).
2
It follows that a + b = 5 or a = b = −5. However, a = b = −5 gives us no solutions
for x.
Consider the equation a + b = 5, where a = 2x and b = 31/x . Observe that
log a log 3
x= = .
log 2 log b
Thus a + b = 5 and log a log b = log 2 log 3. Two obvious solutions are a = 2, b = 3
and a = 3, b = 2 which give the solutions x = 1 and x = log2 3 respectively.
We will show that there are no more solutions. Note that 31/x is undefined at
x = 0. Moreover, we must have x > 0 since otherwise 2x + 31/x < 1 + 1 = 2.
We consider the function f : R>0 → R defined by f (x) = 2x + 31/x . We have
31/x log 3
f 0 (x) = 2x log 2 − and
x2
31/x (log 3)2 2 · 31/x log 3

f 00 (x) = 2x (log 2)2 + + .
x4 x3
It is clear that f 00 (x) > 0 for x > 0. So f is strictly convex and therefore the
equation f (x) = 5 can have at most two solutions as claimed.
4570. Proposed by Lorian Saceanu.

If ABC is an acute angled triangle, then
√
A B C 9 3 3
cos + cos + cos ≤ q ≤ .
2 2 2 11 + tan2 A
+ tan2 B
+ tan2 C 2
2 2 2
We received 6 solutions, one of which was incorrect. We present the solution by

Walther Janous, condensed by the editor.

Solutions /117
Left-hand inequality
Letting
A B C
x = cos , y = cos , z = cos
2 2 2
the inequality becomes (upon squaring and clearing fractions)

1 1 1
(x + y + z)2 · ∗ + 2 + 2 + 2 ≤ 80
x y z
In all triangles, however,

√
3 3
x+y+z ≤ ,
2
√
and since we are here concerned only with acute triangles, we have x, y, z > 2/2.
Thus, there exist positive real numbers ξ, η, and ψ such that
√ √ √
2 2 2
x= + ξ, y = + η, z = + ψ.
2 2 2
Setting
Σ = ξ + η + ψ,
we then have √ √
3 3− 2
Σ≤ . (1)
2
Our goal is to prove the inequality
√ !2 " #
3 2 2 2 2
Σ+ · √ 2 + √ 2 + √ 2 + 8 ≤ 81.
2 2ξ + 1 2η + 1 2ψ + 1
Keeping Σ fixed, we find the maximum of the second left-hand factor. Letting
2 2 2
Φ= √ 2 + √ 2 + √ 2 + 8 − λ · (ξ + η + ψ),
2ξ + 1 2η + 1 2ψ + 1
we get
√
d 4 2
Φ=0⇔ √ 3 − λ = 0,
dξ 2ξ + 1
and two similar expressions for η and ψ. This gives ξ = η = ψ = Σ/3 as the only
stationary point of Φ in the interior of B = {ξ + η + ψ = Σ}. But the required
inequality is then
√ !2 " #
3 2 6
Σ+ · √ 2 + 8 ≤ 81,
2 2 · Σ3 + 1

118/ Solutions
which is succesively equivalent to

√ !2
3 2 27
Σ+ ≤
2 4
√
4 · Σ2 + 12 2 · Σ − 9 = 0
√ √ √ √
3 3+ 2 3 3− 2
− ≤Σ≤ ,
2 2
the last of which holds by (1).
We now consider the boundary of B. If say, ψ = 0, we have C = π/2. The
inequality then becomes
" √ #2
A π A 2 2 A 2 π A
cos + cos − + · 12 + tan + tan − ≤ 81.
2 4 2 2 2 4 2
A

Setting w = tan 2 , the inequality is
" √ √ #2 " 2 #
1−w

1 2 w+1 2 2
√ + ·√ + · 12 + w + ≤ 81,
w2 + 1 2 w2 + 1 2 1+w
which is equivalent to
h √ √ p i2
2(w + 1) + 2 · w2 + 1 · 12 + w2 (1 + w)2 + (1 − w)2

2+
≤ 324 w2 + 1 (1 + w)2 .

It is tedious but straightforward to show that this inequality holds for w ∈ [1, 2].
Right-hand inequality
This inequality is equivalent to

2 A 2 B 2 C
tan + tan + tan ≤ 1,
2 2 2
which holds by the convexity of the function f (x) = tan2 x2 .


CRUXv 47 N 2

Uploaded by

Copyright:

Available Formats

CRUXv 47 N 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CRUXv 47 N 2

Uploaded by

Copyright:

Available Formats

Crux Mathematicorum is a problem-solving journal at the secondary and university undergraduate levels,

c SOCIÉTÉ MATHÉMATIQUE DU CANADA 2021 TOUS DROITS RÉSERVÉS.

Supported by / Soutenu par :

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

MathemAttic Editors John McLoughlin University of New Brunswick

Olympiad Corner Editors Alessandro Ventullo University of Milan

Articles Editor Robert Dawson Saint Mary’s University

Associate Editors Edward Barbeau University of Toronto

Guest Editors Yagub Aliyev ADA University, Baku, Azerbaijan

Editor-at-Large Bill Sands University of Calgary

71 Editorial Kseniya Garaschuk

Copyright © Canadian Mathematical Society, 2021

Click here to submit solutions, comments and generalizations to any

To facilitate their consideration, solutions should be received by April 30, 2021.

N = 1 + 11 + 101 + 1001 + 10001 + · · · + 1000 · · · 01,

MA110. In the figure, ABCDEF is a regular

Crux Mathematicorum, Vol. 47(2), February 2021

Cliquez ici afin de soumettre vos solutions, commentaires ou

MA106. Supposer que

N = 1 + 11 + 101 + 1001 + 10001 + · · · + 1000 · · · 01,

MA108. Supposer que a, b, c et d sont des entiers positifs tels que

ab + cd = 38, ac + bd = 34, ad + bc = 43.

MA110. Dans la figure, ABCDEF est un hexagone

Copyright © Canadian Mathematical Society, 2021

Thus, the desired sum is

MA82. Let an = n2 + 2n + 50, n = 1, 2, . . .. Let dn be the largest positive

Crux Mathematicorum, Vol. 47(2), February 2021

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

4(n2 + 47) − (2n − 3)(2n + 3) = (4n2 + 188) − (4n2 − 9) = 197.

Since 197 is prime, dn ∈ {1, 197} for all n. For n = 97 we have

a97 = 49 · 197, a98 = 50 · 197.

Therefore the maximum value of dn is 197.

26n < 10m ≤ 26n + 2n .

Since n ≥ 3 and 263 > 104 , we must have m ≥ n + 2. Dividing by 2n we obtain

13n < 2m−n 5m ≤ 13n + 1,

where 2m−n 5m is an integer divisible by 4. Since 13n + 1 ≡ 2 (mod 4), we arrive

MA84. The area of the trapezoid ABCD with AB k CD, AD ⊥ AB and

Copyright © Canadian Mathematical Society, 2021

We take CD = b. It follows that AB = 3CD = 3b. The area of the trapezoid is

MA85. A collection of items weighing 3, 4 or 5 kg has a total weight of 120

Crux Mathematicorum, Vol. 47(2), February 2021

Copyright © Canadian Mathematical Society, 2021

Solving the Problems

Crux Mathematicorum, Vol. 47(2), February 2021

Copyright © Canadian Mathematical Society, 2021

The Benefits of Interleaved Practice

Crux Mathematicorum, Vol. 47(2), February 2021

Same Surface, Different Depth

Copyright © Canadian Mathematical Society, 2021

Click here to submit solutions, comments and generalizations to any

To facilitate their consideration, solutions should be received by April 30, 2021.

OC517. Denote by N the set of positive integers 1, 2, 3, . . . Find all functions

OC518. In a triangle ABC with AB 6= AC let M be the midpoint of AB,

Crux Mathematicorum, Vol. 47(2), February 2021

Cliquez ici afin de soumettre vos solutions, commentaires ou

OC517. Soit N l’ensemble des entiers positifs 1, 2, 3, . . . Déterminer toutes

OC518. Soit M le mi point de AB dans un triangle ABC tel que AB 6= AC.

OC519. Démontrer que le nombre x est rationnel si et seulement si trois

OC520. Laurent et Rolland sont deux robots se déplaçant en une même