Volume/tome 50, issue/numéro 5

May/mai 2024
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 2024. 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 2024. 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 Grant McLoughlin University of New Brunswick

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

218 Editorial Kseniya Garaschuk

219 MathemAttic: No. 55
219 Problems: MA271–MA275
221 Solutions: MA246–MA250
227 Problem Solving Vignettes: No. 32 Shawn Godin
234 Olympiad Corner: No. 423
234 Problems: OC681–OC685
236 Solutions: OC656–OC660
244 Gregarious and Reclusive Triples: Part B Ed Barbeau
248 Reading a Math Book: No. 3 Yagub Aliyev
255 Problems: 4941–4950
260 Solutions: 4891–4900

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
218/ Editorial

EDITORIAL
Every year, I write a report on Crux for the Canadian Mathematical Society. I
highlight changes made to the publication (Editorial Board members, new submis-
sion procedures or timelines, formatting), I speak of the additions (new sections
or columns), I mention any special editions and in general I try to give a sense
of impact of the journal. The latter is hard to measure – Crux is an open-access
publication, so I don’t have a number of subscribers to cite or even a number of
downloads to provide. And while I know that the reach of the journal goes well
beyond the number of submissions we get, those are the only hard statistics I can
cite. So just for fun, here are some numbers describing two of the most recent is-
sues, namely Volume 50, issues 1 and 2. The two issues spanned 101 pages, which
included our regular problem sections as well as articles by 5 different authors. For
these two issues (a fifth of the yearly total), we received 402 solutions from solvers
in 25 countries: Austria, Azerbaijan, Brazil, Bulgaria, Canada, China, Colombia,
France, Germany, Greece, Honduras, Hong Kong, India, Indonesia, Italy, Lebanon,
Nicaragua, Peru, Romania, Saudi Arabia, South Korea, Spain, Syria, USA, Viet-
nam. So far from the beginning of 2024 to date (I am writing this on May 15th),
we received 127 problem proposals to be considered for publication and we publish
just over 100 authored problems a year.
Of course the number of readers is much higher than the number of those who end
up submitting solutions or proposals. Occasionally, I enjoy Googling the name of
the journal to see where it pops up. On my latest search, in the first two dozens
of links, I came across two websites of problem solving groups that track their
submissions to Crux, five personal websites/blogs mentioning the journal, two
pdf files with selected problems, not to mention the links of various libraries and
online repositories linking to Crux materials. Sometimes, I am gifted a pleasant
surprise of seeing Crux reference in the wild, for example as a reference in a paper
by Terence Tao.
Join our community of active participants: submit your solutions to the problems
of the issue – note that the section Reading a Math Book also has problems from
the chosen resource for you to submit solutions to. You can also submit problem
proposals, all at https://publications.cms.math.ca/cruxbox/. If Crux has
had an impact on you or you see a fun reference to it somewhere, drop me a line
at [email protected].
Kseniya Garaschuk

Crux Mathematicorum, Vol. 50(5), May 2024

 MathemAttic /219

MATHEMATTIC
No. 55
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 July 31, 2024.

MA271. Proposed by Daniel Sitaru.

Let a, b, c > 0 be real numbers. Show that
1 2  1 1 1  9
(a + b2 + c2 ) + 2 + + ≥ .
2 a+b b+c c+a 2

MA272. Consider the 8! distinct permutations of the digits 1, 3, 4, 5, 6, 7,

8, 9. How many of the eight-digit numbers will be multiples of 275?

MA273. Let
P (x) = x4 + ax3 + bx2 + cx + d,
where a, b, c and d are constants. If P (1) = 10, P (2) = 20 and P (3) = 30, compute
1
10 (P (12) + P (−8)).

MA274. If

sin θ + cos θ + tan θ + cot θ + sec θ + csc θ = 7,

√
then sin(2θ) = a − b 7. Determine the values of a and b.

MA275. Scalene triangle ABC is reflected through its own centroid G, the
image being triangle A0 B 0 C 0 . If AB = 2BC and the area of triangle A0 B 0 C 0
is 72, compute the area of the hexagonal region common to both triangle ABC
and triangle A0 B 0 C 0 . (Note: the centroid of a triangle is the intersection of its
medians.)

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

Copyright © Canadian Mathematical Society, 2024

220/ MathemAttic

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

Cliquez ici afin de soumettre vos solutions, commentaires ou

généralisations aux problèmes proposés dans cette section.

Pour faciliter l’examen des solutions, nous demandons aux lecteurs de les faire parvenir
au plus tard le 31 juillet 2024.

MA271. Soumis par Daniel Sitaru.

Soient a, b, c > 0 des nombres réels. Montrez que
1 2  1 1 1  9
(a + b2 + c2 ) + 2 + + ≥ .
2 a+b b+c c+a 2

MA272. Considérons les 8! permutations distinctes des chiffres 1, 3, 4, 5, 6,

7, 8, 9. Combien des nombres à huit chiffres ainsi formés seront des multiples de
275?

MA273. Soit
P (x) = x4 + ax3 + bx2 + cx + d,
où a, b, c et d sont des constantes. Si P (1) = 10, P (2) = 20 et P (3) = 30, calculez
1
10 (P (12) + P (−8)).

MA274. Si

sin θ + cos θ + tan θ + cot θ + sec θ + csc θ = 7,

√
alors sin(2θ) = a − b 7. Déterminez les valeurs de a et b.

MA275. L’image du triangle scalène ABC suivant une symétrie centrale

autour de son propre centre de masse (ou centroı̈de) G est le triangle A0 B 0 C 0 .
Si AB = 2BC et si l’aire du triangle A0 B 0 C 0 est 72, calculez l’aire de la région
hexagonale commune au triangle ABC et au triangle A0 B 0 C 0 . (Remarque : le
centre de masse (ou centroı̈de) d’un triangle est l’intersection de ses médianes.)

Crux Mathematicorum, Vol. 50(5), May 2024

 MathemAttic /221

MATHEMATTIC
SOLUTIONS
Statements of the problems in this section originally appear in 2023: 49(10), p. 516–518.

MA246. The length of a rectangular piece of paper is three times its width.
The paper is folded so that one vertex lies on top of the opposite vertex, thus
forming a pentagonal shape. What is the area of the pentagon as a fraction of the
area of the original rectangle?
Originally from the 2023 Senior Mathematical Challenge (organized by the United
Kingdom Mathematics Trust), Question 23.
We received 5 submissions, all of which were correct. We present a solution by the
Missouri State University Problem Solving Group.
More generally, consider a rectangle ABCD with AB = 1 and BC = a ≥ 1. Let
XY be the crease obtained by folding the paper and P its midpoint as shown in
the figure.

Now p 1p 2
BD = a2 + 1 and hence BP = a + 1.
2
Since 4BP X is similar to 4BCD,
√
a2 + 1
PX = .
2a
The area of 4BP X is therefore
a2 + 1
.
8a
The area of the pentagon is twice the area of quadrilateral ABP Y , which is the
area of rectangle ABCD minus twice the area of 4BP X. In other words, the area
of the pentagon is
a2 + 1 3a2 − 1
a− = .
4a 4a
Finally, the ratio of the pentagon’s area to that of the original rectangle is
3a2 − 1
.
4a2
For the original problem, this ratio is 13/18.

Copyright © Canadian Mathematical Society, 2024

222/ MathemAttic

MA247. A regular m-gon, a regular n-gon and a regular p-gon share a vertex
and pairwise share edges, as shown in the diagram. What is the largest possible
value of p?

Originally from the 2020 Senior Mathematical Challenge (organized by the United
Kingdom Mathematics Trust), Question 25.

There were 4 correct, 1 incorrect and 1 incomplete solutions submitted. We present

the solution by the Missouri State University Problem Solving Group.

Suppose that 3 ≤ m ≤ n ≤ p. The internal angles of the three polygons are

360 360 360
180 − , 180 − and 180 − degrees. Since their sum is 360◦ , we obtain
m n p
the equation
1 1 1 1
= + + . (1)
2 m n p

Therefore
1 3
≤ ,
2 m

whence m ≤ 6.

When m = 3, (1) yields (n − 6)(p − 6) = 36. Since p is largest when n is smallest,

we are led to the candidate (m, n, p) = (3, 7, 42).

When m = 4, (1) yields (n−4)(p−4) = 16 and the candidate (m, n, p) = (4, 5, 20).

When m = 5, (1) yields (3n − 10)(3p − 10) = 100 and (m, n, p) = (5, 5, 10).

Finally, when m = 6, (1) yields (n − 3)(p − 3) = 9 and (m, n, p) = (6, 6, 6).

We conclude that the largest possible value of p is 42, when a regular 42-gon meets
an equilateral triangle and a regular heptagon.

Comment from the editor. All of the solutions submitted arrived at the equation
1/m + 1/n + 1/p = 1/2. From then on, it was mainly a matter of trial and error
for most of them.

Crux Mathematicorum, Vol. 50(5), May 2024

 MathemAttic /223

MA248. A drinks carton is formed by arranging four congruent triangles as

shown. We have QP = RS = 4 and P R = P S = QR = QS = 10. What is the
volume of the carton?

Originally from the 2022 Senior Mathematical Challenge (organized by the United
Kingdom Mathematics Trust), Question 25.
There were 5 submissions, 4 of which were complete and correct. We present the
solution by Missouri State University Problem Solving Group.
We will solve the problem more generally when

QP = RS = b

and
P R = P S = QR = QS = a.
Let X denote the midpoint of segment P Q and Y the midpoint of segment RS.
Since 4P RX is a right triangle,
…
p b2
RX = P R − P X = a2 − .
2 2
4
Since 4RXY is also a right triangle,
…
p b2
XY = RX 2 − RY =
2 a2 − .
2
The volume of tetrahedron P RSX is Bh/3, where B is the area of 4RSX, i.e.,
…
1 b b2
B = RS · XY = a2 −
2 2 2
and h = b/2. The volume of the carton is twice the volume of tetrahedron P RSX,
so …
b2 b2
a2 − .
6 2
In the original problem, a = 10 and b = 4 giving a volume of
√
16 23
.
3

Copyright © Canadian Mathematical Society, 2024

224/ MathemAttic

MA249. Peter has 25 cards, each printed with a different integer from 1
to 25. He wishes to place N cards in a single row so that the numbers on every
adjacent pair of cards have a prime factor in common. What is the largest value
of N for which this is possible?
Originally from the 2015 Senior Mathematical Challenge (organized by the United
Kingdom Mathematics Trust), Question 24.
The problem has been correctly solved in all 6 submissions. We present the solution
by Mahadevan Aravind.
An integer can occur on a card in the row only if it shares a prime factor with at
least one other integer on a card in the row. This rules out 1, which has no prime
factors. It also rules out the primes 13, 17, 19 and 23 which are not factors of any
other integers in the range from 1 to 25. With 1, 13, 17, 19 and 23 excluded, this
leaves at most 20 cards that can be in the row.
It is indeed possible to arrange all these remaining 20 cards in a row so as to meet
the condition that integers on adjacent cards share a prime factor. For example,
one of many possible configurations is

7, 14, 21, 3, 18, 15, 5, 25, 10, 4, 20, 24, 9, 6, 8, 12, 16, 2, 22, 11.

MA250. Five line segments of length 2, 2, 2, 1 and 3 connect two corners of

a square as shown in the diagram. What is the shaded area?

Originally from the 2021 Senior Mathematical Challenge (organized by the United
Kingdom Mathematics Trust), Question 25.
We received 9 solutions, 8 of which were complete and correct. While there were
multiple approaches we will present two of them.
Solution 1, by Yeganeh Heydarizadeh.
We name the points of the diagram as follows for our convenience of solving.

Crux Mathematicorum, Vol. 50(5), May 2024

 MathemAttic /225

Let GF = x. Then F E = 1 − x. We note that AG = AK + KG = 3 + 2 = 5. It

is not hard to see that 4AGF and 4F EC are similar. Hence, we have x5 = 1−x
2
.
5 5 2
From this, x = 7 . This implies that GF = 7 and F E = 7 . With the help of
Pythagoras’ theorem, we have

Å ã2 √ Å ã2 √
5 25 2 2 10 2
AF = 52 + = and F C = 22 + = .
7 7 7 7

Thus,
√ √
25 2 10 2 √
AC = + = 5 2.
7 7

By Pythagoras’ theorem,

(AB)2 + (BC)2 = (AC)2 = 50

and so we have AB = 5. It can be seen that the area of the square GHJK is 2,
25
the area of the triangle AGF is 14 , the area of the triangle CEF is 72 , and area of
25
the triangle ADC is 2 . Hence, the shaded area is equal to

25 25 2
Å ã
− 2+ + = 9.
2 14 7

Solution 2, by Missouri State University Problem Solving Group

We have
» √
PQ = (2 + 2 + 3)2 + (2 − 1)2 = 5 2,

so the side length of the square is 5. Since the rays emanating from P and Q are
parallel, the shaded angles at each vertex are complementary.

Copyright © Canadian Mathematical Society, 2024

226/ MathemAttic

Rotating the shaded region 180◦ around the center of the square (and darkening
the shading) gives the image above. The unshaded region in the center consists of
seven unit squares. The area of the original shaded region is therefore
1 2

5 − 7 = 9.
2

Crux Mathematicorum, Vol. 50(5), May 2024

 Shawn Godin /227

PROBLEM SOLVING
VIGNETTES
No. 32
Shawn Godin
Two Problems from Honsberger

While doing my undergraduate studies, I was fortunate enough to take a problem

solving course from Ross Honsberger. I have brought up this course, and its
problems, many times over the years. In this course we were given an overview of
problem solving with some references to Pólya during the first lesson. At the end
of class we were presented with one hundred problems. In each subsequent class,
the “lesson” consisted of the solutions to five of the one hundred course problems.
However, the lessons were more than just the solutions to five problems. In some
cases, the solution would be presented in a manner similar to that of a person
attacking the problem for the first time. That is, we may have to recall (or look
up) some facts we haven’t used in a while (several themes from previous columns
are referenced throughout this article). A problem may also suggest a method of
attack to us which doesn’t yield the desired results forcing us to revisit and replan,
à la Pólya. In some cases a new technique is needed and the problem is the vehicle
for us to see the need for this new tool, be introduced to it and help understand its
facility. In this column we will revisit two problems from Professor Honsberger’s
course that were “left for homework” in previous columns.

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

The first problem is problem 9 [2018: 44(10), 419-422].

#9. Four points A, B, C, D are given on a straight line. Show how to

construct a pair of parallel lines through A and B, and another pair
of parallel lines through C and D, so that these pairs of parallel lines
intersect in the vertices of a square.

First we note that in order for the points of intersection of the parallel lines to
form a square, the two pairs of parallel lines must be perpendicular to each other.
In particular, the line through A and the line through D are perpendicular. We
will label the point of intersection of these two lines P . Then, as ∠AP D = 90◦ ,
P lies on the circle with diameter AD (see [2019: 45(8), 452-456] if you need a
refresher on some circle properties). Hence if P is any point on the circle with
diameter AD, then the quadrilateral, P QRS, formed by our two pairs of parallel
lines will necessarily be a rectangle.

Copyright © Canadian Mathematical Society, 2024

228/ Problem Solving Vignettes

P
S
Y
Q
X R
A D
B C

In order for P QRS to be a square, we require the distances between the pairs of
parallel lines to be the same, that is, BX = CY , where X and Y are shown in the
diagram above. Note that, by our method of construction, ∆ABX and ∆CDY
are similar right angled triangles. If BX = CY , we could translate and rotate
∆CDY to ∆C 0 D0 Y 0 such that C 0 and Y 0 coincide with B and X, respectively.
This process would create a larger triangle, ∆AD0 B, which is similar to the other
two triangles, as shown below.

D0
P

Y0
X C0
A D
B C

Then where AD0 (possibly extended) meets the circle with diameter AD is the
vertex P of our square. The line through B, parallel to AP ; the line through D
and P ; and the line through C parallel to DP complete the construction of square
P QRS.

The desired figure could be accomplished using the following constructions:

• construct the circle with diameter AD,

• construct segment BE such that BE = CD and BE ⊥ AD,

• draw the line through A and E,

• label intersection of line through A and E and the circle as P ,

• draw the line through D and P ,

• construct the line through B parallel to AE and the line through C parallel
to DP ,

Crux Mathematicorum, Vol. 50(5), May 2024

 Shawn Godin /229

• the other three points of intersection of the pairs of parallel lines are the
other vertices of square P QRS.
Interested readers might try the construction themselves using their favourite dy-
namic geometry software or, possibly, by hand with compass and straightedge
(see [2021: 47(5), 232-237] and [2021: 47(6), 282-286] for a review of geometric
constructions).

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

Next we will attack problem 25 [2020: 46(2), 57-60].

#25. Prove that

1 3 5 7 99 1
· · · ··· <
2 4 6 8 100 10

We could actually attack this problem with technology. A patient person could
punch this into a calculator and verify – after a few minutes – that the condition
is satisfied. You could also write a program, or use a spreadsheet to deliver the
result much more quickly.
Let’s explore the problem with a spreadsheet. Start by putting 1 in cell A1, 2
in cell B1, and 0.5 in cell C1. Then, in cells A2, B2 and C2, enter the formulas:
= A1 + 2; = B1 + 2; and = C1 ∗ A2/B2, respectively. Upon applying these
formulas, looking at the nth row we would find:
• the numerator of the nth fraction in cell An,
• the denominator of the nth fraction in cell Bn, and
• the product of the first n fractions in cell Cn.
Thus, we could see that the product at hand yields
1 3 5 7 99 1
· · · ··· ≈ 0.0795 · · · <
2 4 6 8 100 10
The nice thing with the spreadsheet is that, we can back up and see that
1 3 5 7 61 1
· · · ··· ≈ 0.1009 · · · >
2 4 6 8 62 10
while
1 3 5 7 61 63 1
· · · ··· · ≈ 0.0993 · · · < .
2 4 6 8 62 64 10
So the conditions for the inequality were satisfied quite some time before the end
of our product. This leads us to believe that the inequality can be strengthened.
For example, using the numbers from the spreadsheet we see that
1 1 3 5 7 99 1
< · · · ··· < .
13 2 4 6 8 100 12

Copyright © Canadian Mathematical Society, 2024

230/ Problem Solving Vignettes

Digging deeper with technology, we can use WolframAlpha – if we introduce the

idea of the double factorial. The double factorial is a generalization of the factorial
(see [2020: 46(7), 291 - 293] for more on factorials) where instead of multiplying
every natural number up to a given number, we multiply every second natural
number. As such, the definition is slightly different for odd and even numbers. If
n is odd, the double factorial of n is

n!! = n × (n − 2) × (n − 4) × · · · × 3 × 1

whereas if n is even, we have

n!! = n × (n − 2) × (n − 4) × · · · × 4 × 2.

This idea can be extended to things like:

13!!! = 13 × 10 × 7 × 4 × 1,
14!!!! = 14 × 10 × 6 × 2,
15!!!!! = 15 × 10 × 5.

The interested reader can explore these extended factorials further at their leisure.
The product of fractions in our problem can be rewritten as

99!!
100!!
which can be calculated with WolframAlpha as
2725392139750729502980713245400918633290796330545803413734328823443106201171875
34243224702511976248246432895208185975118675053719198827915654463488000000000000
2.72 × 1078
≈
3.42 × 1079
≈ 0.0795.

WolframAlpha even reduces the fraction to lowest terms as

12 611 418 068 195 524 166 851 562 157

158 456 325 028 528 675 187 087 900 672

However, even though we can verify the result with technology, it isn’t very sat-
isfying (unless it is to give us further insight – more on that later). As such, we
will attempt to verify the inequality without resorting to calculating the results
by electronic means.
√
As 100 = 10, we might see if we can generalize this situation, and conjecture
that

1 3 5 7 2n − 1 1
Pn = · · · ··· <√
2 4 6 8 2n 2n

Crux Mathematicorum, Vol. 50(5), May 2024

 Shawn Godin /231

This seems like a perfect situation to use mathematical induction (for an intro-
duction to induction, check out [2019 : 45(5), 236-240] ). Clearly, as 12 < √12 , the
statement is true for n = 1. Assume it is true for n = k, so
1 3 5 7 2k − 1 1
Pk = · · · ··· <√
2 4 6 8 2k 2k
Hence
2k + 1 1 2k + 1 1 (2k + 1)2
Pk+1 = Pk · <√ · =√ ·
2k + 2 2k 2k + 2 2k (2k + 2)2
and thus
1 4k 2 + 4k + 1 1
Pk+1 < √ · 2
>√
2k + 2 4k + 4k 2k + 2
which isn’t helpful since as it doesn’t tell us the larger of Pk+1 and √ 1 .
2k+2

Let’s step back and try another approach. It would be nice if we could get some
of the terms in our product to cancel out, simplifying our expression. With that
in mind, if we define
2 4 6 8 2n
Pn0 = · · · · · ·
3 5 7 9 2n + 1
then as 2
3 > 12 , 4
5 > 34 , 6
7 > 65 , . . . , we must have Pn0 > Pn for all n. Therefore,

Pn2 < Pn Pn0

1 3 5 7 2n − 1 2 4 6 8 2n
Å ãÅ ã
= · · · ··· · · · ···
2 4 6 8 2n 3 5 7 9 2n + 1
1 2 3 4 2n − 1 2n
= · · · · ··· ·
2 3 4 5 2n 2n + 1

1
which can be simplified, after cancelling terms out, to Pn2 < or
2n + 1
1
Pn < √
2n + 1
which is a stronger inequality than the one we originally conjectured. That is, for
the problem at hand we must have
1 3 5 7 99 1
· · · ··· <√
2 4 6 8 100 101
1 1 1
but as √101 < √100 = 10 , the original inequality is verified. The interested reader
is encouraged to retry the induction proof attempted above with the stronger
inequality.
As I mentioned earlier, the technologically aided “solutions” were not very satis-
fying. However, as I was writing this column, I added in the technological part

Copyright © Canadian Mathematical Society, 2024

232/ Problem Solving Vignettes

after I had done the more satisfying proofs. As noted in several places, it seems
that the generalized inequality could be strengthened. Looking and playing with
the numbers in the spreadsheet, I was able to conjecture and prove that
1
Pn < √
3n + 1

a significant improvement. However the spreadsheet suggested this could still be

strengthened. With a bit more exploration I came up with the very satisfying –
and to me, amazing – conjecture
1
Pn < √ .
πn

However, the induction proof fell through. When I tried, like before, the stronger
conjecture
1
Pn < √ .
πn + 1
I could prove the induction hypothesis but, unfortunately, didn’t have anywhere
to start as for every case I checked, Pn was larger, so I was stuck.
However, the appearance of π was intriguing, so I kept digging. If we want to
write Pn in terms of regular factorials, this can be accomplished by
1 3 2n − 1
Pn = · ···
2 4 2n
1 2 3 4 2n − 1 2n
= · · · ··· ·
2 2 4 4 2n 2n
(2n)!
=
(2 · 4 · · · 2n)2
(2n)!
=
(2 · 1 · 2 · · · n)2
n

(2n)!
=
4n (n!)2
Ç å
1 2n
=
4n n

This
looked familiar to me. I had seen that the central numbers in Pascal’s triangle,
2n
n , are related to the Catalan numbers, Cn . The Catalan numbers have many
applications. One application is if you take a convex polygon with n + 2 sides, it
can be cut into n triangles by connecting vertices with non-crossing line segments.
The number of different ways this can be accomplished is Cn .
For our problem, if we use the fact that
Ç å
1 2n
Cn =
n+1 n

Crux Mathematicorum, Vol. 50(5), May 2024

 Shawn Godin /233

then we can write

n+1
Pn = Cn
4n
However, the Catalan numbers have the asymptotic formula
4n
Cn ∼ 3√
n 2 π

Hence, we can deduce an asymptotic formula for our product Pn as

n+1 1
Pn ∼ ·√
n πn

Which shows that my conjecture is in the right ballpark.

The evidence from my spreadsheet suggests that
1 1
√ < Pn < √ .
πn + 1 πn

Interested readers are directed to the Wikipedia page for Stirling’s approximation
to the factorial. There are some bounds on the factorial that will be useful in
proving the upper bound of the inequality above. This just goes to show how far
down the rabbit hole you can go with some problems if you allow yourself the time
to do some playing.

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

I hope you enjoyed the two problems we looked at in this column. I leave you with
five more from Professor Honsberger’s course.

#36. Non-intersecting chords AB and CD of a circle are given. A point J on

CD is also given. Determine a point X on the circle such that XA and
XB meet CD at M and N , where J is the midpoint of M N .
#37. Two circles A and B and a line m are given. Construct a square which
has two opposite vertices on m and one vertex each on A and B.
#38. A point A and two lines m and n are given. Construct a square ABCD
with B and D on m and n, respectively.
#39. Construct an equilateral triangle which has one vertex at a given point A
and one on each of two given lines m and n.
#40. Two circles A and B and a point P are given. Construct a segment ST
with S on A and T on B which has P as midpoint.

Copyright © Canadian Mathematical Society, 2024

234/ OLYMPIAD CORNER

OLYMPIAD CORNER
No. 423
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 July 31, 2024.

OC681. Let p1 , p2 , . . . , p100 be one hundred distinct prime numbers. The nat-
ural numbers a1 , . . . , ak are greater than 1 and are such that each of the numbers
p1 p32 , p2 p33 , . . . , p99 p3100 , p100 p31 is equal to the product of some two of the numbers
a1 , . . . , ak . Prove that k ≥ 150.

OC682. Let p be a prime number and let k be a positive integer. Suppose

that the numbers ai = ik + i for i = 0, 1, . . . , p − 1 form a complete residue system
modulo p. What is the set of possible remainders of a2 upon division by p?

OC683. Given a natural number k > 1. A natural number n > 1 and coprime
to k is called correct if for any natural divisor d (d < n) of n the number d + k is
not coprime to n. Prove that there are a finite number of correct numbers.

OC684. Given a convex quadrilateral ABCD, in which AB = BC = CD = 4.

On sides AB and CD, points K and L are chosen, respectively, in such a way
that AK = DL = 1. On side AD, outside the quadrilateral, triangle AM D is
constructed, in which AM = M D = 2. It turns out that KL = 2. Prove that
BM = CM .

OC685. The cells of a 10×10 table contain the natural numbers 1, 2, . . . , 99, 100.
Let’s call the figure that is obtained by removing one cell from a 2 × 2 square a
corner. Let’s call a corner good if the number in its cell, bordering on the sides
with two others, is greater than the numbers in these two other cells. What is
the largest possible number of good corners? (Each corner is counted regardless
of how it is positioned in relation to the others. Different corners may overlap.)

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

Crux Mathematicorum, Vol. 50(5), May 2024

 OLYMPIAD CORNER /235

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 31 juillet 2024.

OC681. Soient p1 , p2 , . . . , p100 cent nombres premiers distincts. Les nom-

bres naturels a1 , . . . , ak sont supérieurs à 1 et sont tels que chacun des nombres
p1 p32 , p2 p33 , . . . , p99 p3100 , p100 p31 est égal au produit de deux des nombres a1 , . . . , ak .
Montrez que k ≥ 150.

OC682. Soit p un nombre premier et k un entier positif. Supposons que les

nombres ai = ik + i pour i = 0, 1, . . . , p − 1 forment un système complet de résidus
modulo p. Quel est l’ensemble des restes possibles de a2 après division par p ?

OC683. Soit k > 1 un nombre naturel. Un nombre naturel n > 1 copremier

avec k est dit correct si pour tout diviseur naturel d (d < n) de n le nombre
d + k n’est pas copremier avec n. Prouvez qu’il existe un nombre fini de nombres
corrects.

OC684. Soit ABCD un quadrilatère convexe dans lequel AB = BC =

CD = 4. Sur les côtés AB et CD, on choisit respectivement des points K et L de
telle sorte que AK = DL = 1. Sur le côté AD, à l’extérieur du quadrilatère, on
construit le triangle AM D, dans lequel AM = M D = 2. Il s’avère que KL = 2.
Montrez que BM = CM .

OC685. Les cellules d’un tableau 10 × 10 contiennent les nombres naturels

1, 2, . . . , 99, 100. Appelons coin la figure obtenue en enlevant une case d’un carré
2 × 2. Qualifions un coin de bon si le nombre dans sa cellule – limitrophe, sur
ses côtés, de deux autres cellules – est plus grand que les nombres dans ces deux
autres cellules. Quel est le plus grand nombre possible de bons coins ? (Chaque
coin est compté indépendamment de sa position par rapport aux autres. Des coins
différents peuvent se chevaucher).

Copyright © Canadian Mathematical Society, 2024

236/ OLYMPIAD CORNER

OLYMPIAD CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2023: 49(9), p. 532–533.

OC656. Let P be the set of all prime numbers. Find all functions f : P → P
such that:
f (p)f (q) + q p = f (q)f (p) + pq
holds for all p, q ∈ P.
Originally from the 2019 Balkan Mathematics Olympiad, Problem 1, Proposed by
Dorlir Ahmeti, Albania.
We received 2 solutions. We present the solution by UCLan Cyprus Problem Solv-
ing Group.
We know that this is from the Balkan Mathematical Olympiad of 2019.
Note that if f (p) = f (q), then q p = pq which gives that p = q. Therefore f is
injective.
If f (2) is odd, then taking p = 2 and q an odd prime we get that f (q)f (p) is
even which gives that f (q) is even and therefore f (q) = 2. This contradicts the
injectivity of f . Therefore f (2) = 2.
Now let p = 2 and q an odd prime. Let r = f (q). We get 2r + q 2 = r2 + 2q . We
will show that r = q.
We claim that the function g : N → N defined by g(n) = 2n − n2 is increasing for
n ≥ 3. Indeed

g(n + 1) − g(n) = (2n+1 − 2n ) − ((n + 1)2 − n2 ) = 2n − (2n + 1)

which can easily be shown by induction to be positive for n ≥ 3. Since g(2) = g(4),
then g(q) = g(r) =⇒ q = r or {q, r} = {2, 4}. In particular, since q, r are primes,
r = q as claimed.
Thus the only function that satisfies the given equation is f (p) = p for all p ∈ P.

OC657. Determine all real numbers x such that

√ √
x− x log1/√2 x ≥ x − x log1/√2 x.

Originally from the 2015, Science Atlantic Math Competition, Problem 1.

We received 9 solutions. We present the solution by Theo Koupelis.

Crux Mathematicorum, Vol. 50(5), May 2024

 OLYMPIAD CORNER /237

For real solutions we must have x > 0. Let

√ √
f (x) : = x − x log1/√2 x − x + x log1/√2 x
√
x log x √ x log x
=x+ √ − x− √
log 2 log 2
1 √ √
= √ (x − x)(log 2 − log x).
log 2
Clearly, for x > 0 we get
√
x≥ x ⇐⇒ x(x − 1) ≥ 0 ⇐⇒ x ≥ 1.
Also, √ √
2 ≥ log x ⇐⇒ x ≤ 2.
log
√
√ all real numbers 1 ≤ x ≤ 2 we get f (x) ≥ 0, with equality when x = 1
Thus, for
or x = 2.
Editor’s Comment. G.C. Greubel generalized the problem and considered for
n > 1 the inequality
√ √
x − x log1/√n (x) ≥ x − x log1/√n (x).

OC658. A quadrilateral ABCD is inscribed in a circle k where AB > CD

and AB is not parallel to CD. Point M is the intersection of diagonals AC and
BD, and the perpendicular from M to AB intersects the segment AB at a point
E. If EM bisects the angle CED, prove that AB is a diameter of k.
Originally from the 2018 Balkan Mathematics Olympiad, Problem 1, Proposed by
Emil Stoyanov, Bulgaria.
We received 7 solutions and we present 2 of them.
Solution 1, by UCLan Cyprus Problem Solving Group.
Let X be the intersection of AB and CD and let Y be the intersection of EM
with CD.

Copyright © Canadian Mathematical Society, 2024

238/ OLYMPIAD CORNER

Since Y E bisects ∠CED and Y E ⊥ XE, then (C, D; Y, X) is harmonic. Project-

ing from M we get that (A, B; E, X) is also harmonic. Therefore E is on the polar
`X of X.
Let Z be the intersection of AD with BC. By Brocard’s theorem, M Z is the polar
`X of X and therefore passes through E. So `X is perpendicular to AB. Since it
is also perpendicular to XO and X belongs to AB, then O also belongs to AB,
i.e. AB is a diameter of k.

Solution 2, by Theo Koupelis.

Let

∠DEM = ∠M EC = θ, ∠CAB = α, ∠ABD = β, ∠ECM = γ, ∠EDM = δ.

Using the law of sines in triangles DM E and CM E, we get

CM sin δ
= .
DM sin γ

In triangles ADB and ACB we have ∠ADB = ∠ACB, and thus, using the law of
sines, we get
BC sin α
= .
AD sin β
Triangles DAM and CBM are similar because ∠DM A = ∠CM B and

∠ADM = ∠ADB = ∠BCA = ∠BCM,

and thus CM/DM = BC/AD. From the above we get sin δ/ sin γ = sin α/ sin β.
But ∠AEC = ∠DEB = 90◦ + θ and α + γ = δ + β = 90◦ − θ. Thus,

sin δ sin α
= ⇐⇒ sin(θ + 2δ) = sin(θ + 2α).
sin(90◦ − θ − α) sin(90◦ − θ − δ)

Therefore, we get α = δ or

θ + δ = 90◦ − α ⇐⇒ ∠EM B = 90◦ − α ⇐⇒ α = β.

The latter contradicts the condition that AB is not parallel to CD, and thus α = δ.
Therefore, ADM E is cyclic; thus

∠ADB = ∠ADM = ∠AEM = 90◦

and AB is a diameter of k. Also,

∠M DC = ∠BDC = ∠BAC = α = δ = ∠EDM

and thus M is the incenter of triangle CED.

Crux Mathematicorum, Vol. 50(5), May 2024

 OLYMPIAD CORNER /239

OC659. Find all sequences of integers a0 , a1 , a2 , . . . such that for any integers
k, l ≥ 0, we have
ak − al | k 2 − l2 .
That is for any integers k, l ≥ 0 there exists an integer z such that

(ak − al )z = k 2 − l2 .

Originally from the 2023 37th Nordic Mathematical Contest, Problem 3.

We received a single solution by UCLan Cyprus Problem Solving Group, presented
below.
We will show that the sequence must be in one of the following forms:
• an = n2 + m for all n ∈ N and some fixed m ∈ Z.
• an = −n2 + m for all n ∈ N and some fixed m ∈ Z.
• an = n + m or an = −n + m for all n ∈ N and some fixed m ∈ Z.
It is easy to check that these sequences satisfy the condition, so it is enough to
check that there are no more.
Note that if a sequence (an ) satisfies the condition, so does the sequence (an + m)
where m is a fixed integer. So we may assume a0 = 0.
Let p be a prime. For k = p, ` = 0 we get ap |p2 and therefore ap = ±1, ±p, ±p2 .
The sequence is injective since ak = a` =⇒ 0|k 2 − `2 =⇒ k = `. Therefore there
is an infinite set S of prime numbers such that one of the following holds:
(a) ap = p for all p ∈ S
(b) ap = −p for all p ∈ S
(c) ap = p2 for all p ∈ S
(d) ap = −p2 for all p ∈ S
Assume (c) holds. Then for every n ∈ N and every p ∈ S we have an − p2 |n2 − p2 .
This implies that an − p2 |n2 − an . Since n2 − an is divisible by infinitely many
distinct numbers, then an = n2 for every n ∈ N.
Similarly if (d) holds we get an = −n2 for every n ∈ N.
Now assume (a) holds. Then for every n ∈ N and every p ∈ S we have an −p|n2 −p2 .
Since also an |n2 , then n2 = bn an for some integer bn . Since an − p|bn an − bn p we
deduce that an − p|p(bn − p). For n > 0 we have an 6= 0 and taking p ∈ S with
p > |an | we have p - an − p. Therefore an − p|bn − p which gives that an − p|an − bn .
Since this holds for infinitely many primes p, then an = bn and since an bn = n2
we get that an = ±n for every n ∈ N.
Similarly if (d) holds we get an = ±n for every n ∈ N.
So in all cases we get that the sequence has one of the claimed forms.

Copyright © Canadian Mathematical Society, 2024

240/ OLYMPIAD CORNER

OC660. Let ABC be a triangle, and M the midpoint of the side BC. Let E
and F be points on the sides AC and AB, respectively, so that M E = M F . Let D
be the second intersection of the circumcircle of M EF and the side BC. Consider
the lines `D , `E and `F through D, E and F , respectively, such that `D ⊥ BC,
`E ⊥ CA, and `F ⊥ AB. Show that `D , `E and `F are concurrent.
Originally from the 2023 37th Nordic Mathematical Contest, Problem 4.
We received 4 solutions and we present 2 of them.
Solution 1, by Michel Bataille.
Let Γ be the circumcircle of ∆M EF and let K be the point of Γ diametrically
opposite to M . Since D, E, F is on Γ, we have

KD ⊥ BC, KE ⊥ EM, KF ⊥ F M

and since M E = M F , we have KE = KF . Hence cos(∠KEA) = sin(∠M EC) and

cos(∠KF A) = sin(∠M F B). Applying the Law of Sines in ∆M EC and ∆M F B
yields
sin(∠M EC) MC MB sin(∠M F B)
= = = ,
sin C ME MF sin B
so that c sin(∠M F B) = b sin(∠M EC), or c cos(∠KF A) = b cos(∠KEA) (where
b = CA, c = AB, as usual).
Since
−→ −−→
CA · EK = bEK cos(∠KEA)

and
−−→ −−→
AB · F K = cF K cos(180◦ − ∠KF A),

we finally obtain
−→ −−→ −−→ −−→
CA · EK + AB · F K = 0. (1)

We will make use of the following lemma: if I is the midpoint of the segment U V ,
then the point H of U V is the orthogonal projection of X on the line U V if and
−→ −−→
only if IH · U V = 12 (XU 2 − XV 2 ).
Proof. The result immediately follows from:

−→ −−→ −→ −−→ −−→ −−→ −−→ 1 −−→ −−→ −−→ −−→

IH · U V = (IX + XH) · U V = XH · U V − (XU + XV ) · (XV − XU )
2
−−→ −−→ 1
= XH · U V + (XU 2 − XV 2 ).
2
Now, suppose that `E and `F intersect at Q. From the lemma, we have

−−→ −→ 1 −−→ −−→ 1

N E · CA = (QC 2 − QA2 ) and P F · AB = (QA2 − QB 2 ),
2 2

Crux Mathematicorum, Vol. 50(5), May 2024

 OLYMPIAD CORNER /241

where N and P are the respective midpoints of CA and AB. Hence

−−→ −−→ 1 −−→ −−→ −−→ −→ −−→ −−→

M D · BC − (QB 2 − QC 2 ) = M D · BC + N E · CA + P F · AB
2
−−→ −−→ −−→ −→ −−→ −−→
= M K · BC + N K · CA + P K · AB (using (1) and KD ⊥ BC)
−→ −−→ −−→ −−→ −−→ −−→
= CA · (N K − M K) + AB · (P K − M K)
−→ −−→ −−→ −−→
= CA · N M + AB · P M
−→ 1 −−→ −−→ 1 −→
= CA · AB + AB · AC = 0
2 2
From the lemma again, D is the orthogonal projection of Q onto BC, hence
`D , `R , `F concur at Q.

Solution 2, by UCLan Cyprus Problem Solving Group.

It is enough to show the following:
Let D, E, F, M be concyclic points with M E = M F , let ` be the line
perpendicular to DM and let B, C be points on the line M D such that
M B = M C. Then the other point of intersection of the circumcircles
of triangles DEB and DF C belongs on `.
To prove this we consider the points D, E, F, M as fixed and the points B, C as
moving points. Let f be the transformation from the line M D to the line ` which
maps B to the point of intersection of ` with the circumcircle of the triangle DEB.
Let g be the transformation from the line M D to the line ` which maps B first
to the unique point C 6= B on M D with M B = M C and then to the point of

Copyright © Canadian Mathematical Society, 2024

242/ OLYMPIAD CORNER

intersection of ` with the circumcircle of the triangle DF C. It is enough to show

that the maps f and g are identical.

The transformation f is projective. We can see this by writing f as the composition

of the following three projective transformations: Consider the circle ω centred at
D with radius DE. We first invert B on ω to obtain a point B 0 on M D, then
project B 0 through E on ` to obtain a point B 00 and finally invert B 00 on ω to get
B 000 . Since the image of the circumcircle of triangle DEB under inversion is the
line B 0 E, then B 000 = f (B).

Similarly g is also projective as it is a composition of a homothety together with

a transformation defined analogously to f . So to show that f and g are identical,
it is enough to show that they agree on three points.

The transformations f and g trivially agree when B = M as then f (B) and g(B)
are the other point of intersection of the circumcircle of DEF M with `.

Now let B be the point of intersection of the bisector of ∠DF E with the line M D.
Since
∠BDE = ∠EF M = ∠F EM = ∠F DM,

then B is the F -excenter of triangle DEF .

Crux Mathematicorum, Vol. 50(5), May 2024

 OLYMPIAD CORNER /243

Letting I be the incenter of the triangle DEF we have

∠DBI = 180◦ − ∠IF D − ∠F DB

Å ã
∠DEF ∠EDF
= 180◦ − − 90◦ +
2 2
∠DEF
=
2
= ∠DEI,

so f (B) = I. To show that g(B) = I, it is enough to show that EI passes through

C and then use an analogous argument. To show this let C 0 be the E-excenter
of triangle DEF and let A be the point of intersection of BE and C 0 F . Observe
that (in this particular case) DEF is the orthic triangle of triangle ABC 0 . The
nine-point circle of triangle ABC 0 passes through D, E, F as well as the midpoint
M 0 of BC 0 . So M 0 = M and therefore C 0 = C.
Now let B be the E-excenter of triangle DEF and let A be the D-excenter. Then
B, D, E, A are concyclic in a circle of diameter AB. Thus f (B) = A and by similar
arguments as in this and the previous paragraph we get that g(B) = A as well.
So f, g are identical as claimed.

Copyright © Canadian Mathematical Society, 2024

244/ Gregarious and Reclusive Triples: Part B

Gregarious and Reclusive Triples

Ed Barbeau
PART B: A PLETHORA OF TRIPLES AND QUADRUPLES

This article is the second of three on this topic. For Part A, please refer to Crux,
50(4), p. 190-193.

How to construct lots of k−triples

Exercise 7. Suppose that x, y and c are arbitrary integers. Let z = x + y + 2c
and k = c2 − xy. Prove that xz + k = (x + c)2 and yz + k = (y + c)2 , so that
(x, y, z) is a k−triple.
A k−triple for which z and k are related in this way is said to be superbly gregarious
or simply superb.
Exercise 8. Prove that the right and left associates of a superb k−triple are also
superb k−triples.
Comment. Note that for the triple (y, z, 2(y + z) − x), the role of c is now played
by (y + c). This result allows a simple way of establishing that the sequences
{u(k, n)} and {v(k, n)} are k−gregarious since it is necessary only to find any
superbly gregarious consecutive triple in each sequence.
If we permute the terms of (x, y, z) to (x, z, y), we find that y = x+z −2(c+x) and
xz = [−(c + x)]2 and we can embed this triple in another sequence of k−triples.
We can look at this construction in three ways. Suppose we are given a triple
(x, y, z) and want to know if it is a k−triple for some k. If x + y + z is even, then
x + y and z have the same parity, and we can take c = 21 (z − x − y).
Suppose we are given an integer pair (x, y) and we want to embed it into a k−triple
(x, y, z) for some k such that xy + k is equal to a given square c2 . Then simply
define z = x + y + 2c.
Finally, suppose that we are interested in k−triples for a specific value of k. Pick
any square c2 and choose x, y such that their product is c2 − k. In this way, for
example, we can find any number of 1−triples. With c = 5, we find (1, 24, 35),
(2, 12, 24), (3, 8, 21), (4, 6, 20). More generically, we have the infinite families
(1, c2 − 1, (c + 1)2 − 1), (c − 1, c + 1, 4c), (2, 2c(c + 1), 2(c + 1)(c + 2)).
Thus we see that k−triples are prolific and many interesting infinite families of
such triples can be found. For example:
k (x, y, z) (a, b, c)
r2 + s2 + t2 − 2(rs + st + rt) (2r, 2s, 2t) (−r + s + t, r − s + t, r + s − t)
r2 + s2 + t2 − 2(rs + st + rt) − 2r (2r, 2s + 1, 2t + 1) (s + t − r + 1, r − s + t, r + s − t)

Crux Mathematicorum, Vol. 50(5), May 2024

 Ed Barbeau /245

How to construct lots of k−quadruples

It is natural to ask whether, for any value of k, there are k−quadruples of numbers
for which the product of any pair plus k is a square. The construction described
in the previous section makes it quite straightforward to answer this in the affir-
mative. If we extend the triple (x, y, x + y + 2c) to the left, we get the quadruple

(x + y − 2c, x, y, x + y + 2c).

Since (x + y − 2c, x, y) and (x, y, x + y + 2c) are k−triples, it is necessary only to

arrange that

(x + y − 2c)(x + y + 2c) + k = (x + y)2 − 4c2 + (c2 − xy) = (x2 + xy + y 2 ) − 3c2

is equal to d2 for some integer d. In other words, we need to find numbers express-
ible in each of the forms φ(x, y) = x2 + xy + y 2 and ψ(c, d) = 3c2 + d2 .

Exercise 9. Prove that φ(x, y) = φ(x + y, −y) = φ(−x, x + y).

Exercise 10. Prove that φ(x, y) = x2 + xy + y 2 and ψ(c, d) = 3c2 + d2 take the
same set of integer values, where x, y, c and d are integers. (Hint: given (c, d), let
(x, y) = (c + d, c − d). How can you go from (x, y) to a corresponding (c, d)?)

In order to get k−quadruples whose entries are distinct, we can exploit the fact
the some numbers can be represented by both of the forms φ(x, y) or ψ(c, d) in
several ways, so that we can get numerous examples of k−quadruples by using
each c with each of the pairs (x, y) involved.

Exercise 11. There are several ways of representing each of the numbers 49, 91
and 133 by φ(x, y) and ψ(c, d). For each, use all of the possible triples (x, y, c) to
construct k−triples.

Exercise 12. There are parametric families of k−quadruples. Determine k−quadruples

when (x, y) = (2r, s), (2r, 2s), (2r, 2s + 1), where r and s are arbitrary integers.

Exercise 13. Verify that each of the following are 1−quadruples:

(r − 1, r + 1, 4r, 4r(4r2 − 1));

(1, r2 − 1, r(r + 2), 4r(r3 + 2r2 − 1));
(r, s2 − 1 + (r − 1)(s − 1)2 , s(rs + 2);
4r3 s4 + 8r2 (2 − r)s3 + 4r(r − 1)(r − 5)s2 + 4(2r − 1)(r − 2)s + 4(r − 1));
(r, 4(r − 1), r − 2, 4(2r − 3)(2r − 1)(r − 1);
(r, s, r + s + 2c, 2c(r + c)(s + c)).

Copyright © Canadian Mathematical Society, 2024

246/ Gregarious and Reclusive Triples: Part B

Reclusive k−triples and their families

Exercise 14. Not every k−triple generates a succession of k−triples when embed-
ded in a sequence satisfying the congenial recurrence. For example, when x = y,
there are triples for which (x, x, z) is a k−triple, but its right associate (x, z, x+2z)
is not. With xy + k = c2 , zx + k = b2 , yz + k = a2 , we have the examples:

k (x, y, z) (a, b, c)
4r4 + 8r3 − 4r + 1 (2r + 1, 2r + 1, 2(2r + 1)) (2r2 + 2r + 1, 2r2 + 2r + 1, 2r2 + 2r)
r4 − 6r2 s2 + s4 (2rs, 2rs, 4rs) (r2 + s2 , r2 + s2 , r2 − s2 )

When z = x + y, we can make use of Pythagorean triples to construct k−triples.

Suppose that we have values k and c for which xy + k = c2 . Then we want to find
a and b for which a2 = yz + k = x2 + xy + k = x2 + c2 and b2 = xz + k = y 2 + c2 .
Thus (x, c, a) and (y, c, b) are both Pythagorean triples sharing the value of a leg.
(Such pairs of triples are easy to find; there are as many ways to express c2 as a
difference of squares as c2 can be factored as a product of two integers of the same
parity.) These triples allow us to isolate the values of a, b, c, x, y.
Exercise 15. The three Pythagorean triples (5, 12, 13), (9, 12, 15) and (35, 12, 37)
share the term c = 12. Using the three pairs of them, arrive at the reclusive
k−triples (5, 9, 14), (5, 35, 40) and (9, 35, 44) with values of k respectively equal to
99, −31 and −171.
The right associate of (5, 35, 40) is (35, 40, 145) and we note that

35 × 145 − 31 = 712 + 3,

a near miss. This is not the only occurrence of this.

The k−triple
(2r + 1, r2 (r + 1)2 − 1, r2 (r + 1)2 + 2r)
with k = −(2r5 + r4 − 4r3 − 3r2 − 2r − 1) has right associate

(r4 + 2r3 + r2 − 1, r2 + 2r3 + r2 + 2r, 4r4 + 8r3 + 4r2 + 2r − 3).

We find that

(r4 + 2r3 + r2 − 1)(4r4 + 8r3 + 4r3 + 2r − 3) − (2r5 + r4 − 4r3 − 3r2 − 2r − 1)

= (4r8 + 16r7 + 24r6 + 18r5 + r4 − 12r3 − 7r2 − 2r + 3)
+ (−2r5 − r4 + 4r3 + 3r2 + 2r + 1)
= 4r8 + 16r7 + 24r6 + 16r5 − 8r3 − 4r2 + 1 + 3
= (2r4 + 4r3 + 2r2 − 1)2 + 3
= [2r2 (r + 1)2 − 1]2 + 3.

Crux Mathematicorum, Vol. 50(5), May 2024

 Ed Barbeau /247

Finding pairs of Pythagorean triples with a common leg arise in the determination
of Heronian triangles, whose sides and area are all integers. Such triangles can be
constructed by pasting together two right triangles that share a common leg, as
shown in the diagram.

a c b

x y

Exercise 16.
(a) We can take as a particular example x = r −2, y = r +2, a = 2r −1, b = 2r +1,
where r is an integer that exceeds 2. Show that, if there is a solution in integers,
then c = 3s where r2 − 3s2 = 1.
(b) r2 − 3s2 = 1 is an example of a Pell’s equation which has infinitely many
solutions. Determine the solution (r1 , s1 ) with the smallest positive integers, and
show that, for each n, (rn , sn ) is a solution where
√ √
rn + sn 3 = (r1 + s1 3)n .

Both the sequences {rn } and {sn } satisfy the same second order recursion; find it.
Also show how each of rn+1 and sn+1 can be written as linear combinations of rn
and sn .

Copyright © Canadian Mathematical Society, 2024

248/ Reading a Math Book

Reading a Math Book

No. 3
Yagub Aliyev
Titu Andreescu (Ed.), Mathematical Reflections, Two
Special Years (2014–2015)

In each appearance of this column, we feature one math book by collecting, com-
menting, and solving problems from it. In this issue we focus on a recent problem
book. There will also be some problems at the end for you to solve and send us.

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

The book [5] is a collection of problems, their so-

lutions, and articles published in the online journal
Mathematical Reflections during the years 2014–2015.
The books in this series appear every two years col-
lecting and saving the material of the journal, some
of which is not available online anymore. The most
recent one covers the two lockdown years 2020–2021.
The editor, Titu Andreescu, who is a veteran prob-
lem composer himself, did excellent work by collecting
problems of different levels (junior, senior, olympiad,
and undergraduate) and articles about a variety of
topics in one voluminous book (583 pages). To the
delight of the readers the editor did not try to make
the solutions and proofs shorter, which would make it more challenging to read, as
is usually done in such books for economy of space. As in Crux Mathematicorum,
the solutions in Mathematical Reflections are selected from many others submitted
by enthusiastic readers. In most cases the choice of featured solution is done based
on aesthetic qualities and unexpectedness of the methods. One can see this effect
of beauty and surprise when reading the solutions in the book. The book will be
a precious resource for students and their instructors to prepare for mathematical
contests and organize math training. Learning trending themes and new ideas in
the proposals of active problem composers can be beneficial for the participants
of mathematical competitions. The first half of the book is composed of carefully
written solutions with beautiful diagrams and detailed explanations. In the sec-
ond half of the book there are 19 articles, one of which [1] was contributed by the
author of this column. The following text can be considered as a continuation of
that chapter and the problems are selected from or inspired by the material there
[2–4, 6, 7]. Enjoy!

Crux Mathematicorum, Vol. 50(5), May 2024

 Yagub Aliyev /249

Solved Problems
Problem 1. A circle w passing through points A and B is externally tangent to
a circle w1 . Line AB intersects circle w1 at points C and D. Let r1 and R be the
radii of circles w1 and w, respectively. Prove that

|AB| R |AB|
p p < <p p .
|AC| · |BD| + |BC| · |AD| r1 |AC| · |BD| − |BC| · |AD|

Solution. Let M be the point of tangency between circles w and w
1 . Extend line
AM to intersect w1 at A0 . Since M A0 = rR1 · |M A|, AA0 = 1 + rR1 · |M A|. Using
|AD| · |AC| = |AM | · |AA0 | one can show that
p
|AC| · |AD|
|M A| =
1 + rR1
p

and similarly p
|BC| · |BD|
|M B| = .
1 + rR1
p

It remains to substitute these formulas into the triangle inequality

|M B| − |M A| < AB < |M A| + |M B|

and simplify.
Problem 2. Under the conditions of Problem 1, prove also that if w2 is another
circle passing through points C and D, and is externally tangent to circle w, then
p p
|AC| · |BD| − |BC| · |AD|
…
r1
p p < ,
|AC| · |BD| + |BC| · |AD| r2

where r2 is the radius of circle w2 .

Solution. By the results obtained in the previous problem, we get
p p
|AC| · |BD| − |BC| · |AD| |AB|
…
r1 R
< , p < .
|AB|
p
R |AC| · |BD| + |BC| · |AD| r2

By multiplying these two inequalities we obtain the required inequality.

Copyright © Canadian Mathematical Society, 2024

250/ Reading a Math Book

Problem 3. Let circles w and w1 of radii R and r be externally tangent. Let the
extension of chord AB of circle w be tangent to circle w1 at point C. Let CD be
tangent to circle w. Prove that
…
|AB| R
< .
2|CD| r

Solution. Like the previous problem, this one is a direct corollary of Problem 1
(case |CD| = 0). But we will give an independent solution here.

Let us denote |CA| = m and |CB| = n. Denote the midpoint of AB by P , and

the centers of w and w1 by O and K, respectively. Then
n−m
|AP | = |BP | =
2
and
n+m
|CP | = .
2
By the Pythagorean theorem,
 n − m 2
|OP | = R2 − .
2

Crux Mathematicorum, Vol. 50(5), May 2024

 Yagub Aliyev /251

From the right-angled trapezoid CP OK we obtain

|KO|2 = |CP |2 + (|KC| ± |OP |)2 ,

where the ± is because of the position of O with respect to line AB. So,
!2
 n + m 2  n − m 2
2
(R + r) = + r± R2 − .
2 2

Simplifying, we obtain
s »
R R 2 n−m 2



n−m r ± r − 2r
√ = .
2 mn 2
Consequently, …
|AB| n−m R
= √ < .
2|CD| 2 mn r

Problem 4. Let circles w and w1 of radii R and r be externally tangent. A

line through the center of circle w1 is tangent to circle w at point A. Let AB be
tangent to circle w1 at point B. Similarly, a line through the center of circle w is
tangent to circle w1 at
√ point C. Let CD be tangent to circle w at point D. Prove
that |AB| = |CD| > Rr.

Solution. The inequality follows from the previous problem (this is the special
case |AB| = 2R in Problem 4), but we will give an independent proof here. Let
O1 and O be the centers of w1 and w. Then

|O1 O| = R + r, |OA| = R and ∠O1 AO = 90◦ .

Using the Pythagorean theorem for 4O1 AO one can show that
p
|O1 A| = r2 + 2Rr.
√
√ same method for 4O1 AB, one can √
Using the show that |AB| = 2Rr. Similarly,
|CD| = 2Rr and therefore, |AB| = |CD| > Rr is true.

Copyright © Canadian Mathematical Society, 2024

252/ Reading a Math Book

√
Problem 5. Do equality |AB| = |CD| and inequality |AB| > Rr remain true
if circles w and w1 are (1) non-intersecting (2) properly intersecting?

Solution. Using the method of the previous problem one can show that
p
|AB| = |CD| = d2 − R2 − r2 ,

where d is the distance √

between the centers of the circles, is always true. Therefore,
the inequality |AB| > Rr is true only when

d2 > R2 + Rr + r2 .

So, if the
√ circles are non-intersecting, then d > R + r and therefore inequality
|AB| > Rr is always true. But if the circles are properly intersecting, then the
inequality holds true when ∠O1 XO > 120◦ , where O1 and O are the centers of
circles w1 and w, and X is one of the intersection points of these circles.

Problems to solve

Do you want to try some problems yourselves?

Submit your solutions at https://publications.cms.math.ca/cruxbox/ before

the 31st of July 2024. Feel free to also send us your opinion about the featured
book or your recommendations for future issues of this column.

S11. Two circles w1 and w2 intersect at distinct points points B and C, and
a third circle w is internally tangent to w1 and w2 at M and N , respectively.
Line BC intersects w at A and D, with A nearer B than D to B. Prove that
the tangents of w at M and N and the line BC are concurrent. Prove also that
∠AM B = ∠CM D and
|BC| sin ∠CM B
> .
|AD| sin ∠AM D

Crux Mathematicorum, Vol. 50(5), May 2024

 Yagub Aliyev /253

S12. Under the conditions of S11, prove also that the ratio of |AB| · |CD| to
|AC| · |BD| is independent of the locations of A and D (for B and C fixed).

S13. Under the conditions of S11, prove also that if r, r1 , and r2 are the radii
of w, w1 and w2 , respectively, with r1 ≤ r2 , then for i = 1, 2, we have

|AD| |AD|
…
r
p p < <p p .
|AC| · |BD| + |AB| · |CD| ri |AC| · |BD| − |AB| · |CD|

S14. Under the conditions of S11 and S13, prove also that
p p
|AC| · |BD| − |AB| · |CD|
…
r1
p p < .
|AC| · |BD| + |AB| · |CD| r2

S15. Let A, B, C, and D be points on line k in that order, and M be a point

not on k such that ∠AM B = ∠CM D. Prove that

|AD| · |BC| ∠CM B sin ∠CM B

ä2 < < .
∠AM D sin ∠AM D
Äp p
|AC| · |BD| + |BC| · |AD|

Copyright © Canadian Mathematical Society, 2024

254/ Reading a Math Book

References
[1] Y.N. Aliyev, Inequalities on Ratios of Radii of Tangent Circles, Mathematical
Reflections, 5 (2015) 1–6.
[2] Y.N. Aliyev, Use of dynamic geometry software in teaching and research, 5th
International Conference on AICT, 12–14 October 2011, Baku, 212–214. https:
//doi.org/10.1109/ICAICT.2011.6110920
[3] Y.N. Aliyev, Problem 11689, Amer. Math. Month., 120(1), January 2013, 77.
[4] Y.N. Aliyev, Aufgabe 1318, Elem. Math., 68(2013) 128.
[5] T. Andreescu, (Ed.), Mathematical Reflections Two Special Years (2014–2015),
XYZ Press, 2016.
[6] G. Bercea, Solution: Aufgabe 1318, Elem. Math., 69(2014), 164–165.
[7] R. Stong, Solution to Problem 11689, A Tale of Three Circles, Amer. Math.
Month., 122(3) (2015) 285–286.

Crux Mathematicorum, Vol. 50(5), May 2024

 Problems /255

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 July 31, 2024.

4941. Proposed by Michel Bataille.

Let a non-equilateral triangle ABC be inscribed in a circle Γ with center O. Its
incircle γ (with center I) touches BC, CA, AB at D, E, F , respectively. The circle
ΓA passes through A and is tangent to ID at I and ΓB , ΓC are cyclically defined.
Prove that ΓA , ΓB , ΓC and the line OI have a common point other than I.

4942. Proposed by Ivan Hadinata.

Let n ∈ Z+ . It is known that (ak )∞ ∞
k=1 and (bk )k=1 are positive real sequences
satisfying the following equation.
Ç åk ã Ç å
n1 n2 nk n1 n2 nk
Å ãÅ
akk + + ··· + = b1 + b2 + · · · bk + , ∀k ∈ Z+ .
1! 2! k! 1! 2! k!

Prove that (ak )∞ ∞

k=1 is upper-bounded if and only if (bk )k=1 is upper-bounded.

4943. Proposed by Mihaela Berindeanu.

Let ABC be a triangle with the circumcircle Γ and orthocenter H. The projections
of the point H on the tangents to Γ at B and C are B1 and C1 and B2 , C2 are
the midpoints of sides AC and AB, respectively. Show that if B1 B2 = C1 C2 , then
ABC is an isosceles triangle.

4944. Proposed by Vasile Córtoaje.

Let a, b, c, d, e be nonnegative real numbers satisfying ab + bc + cd + de + ea = 5.
Prove that:
(a) (a + 2)2 + (b + 2)2 + (c + 2)2 + (d + 2)2 + (e + 2)2 ≥ 45.
(b) a3/2 + b3/2 + c3/2 + d3/2 + e3/2 ≥ 5.

4945. Proposed by Daniel Sitaru.

Let x, y, z ∈ R. Prove that
X»
(− cos x + cos y + cos z)2 + (− sin x + sin y + sin z)2 ≥ 3.
cyc

Copyright © Canadian Mathematical Society, 2024

256/ Problems

4946. Proposed by Chikara Tsugawa.

Let ABC and DEF be two triangles. Moreover, P = BC ∩ F D, Q = BC ∩ DE,

R = CA ∩ DE, S = CA ∩ EF , T = AB ∩ EF , and U = AB ∩ F D (as shown).
Suppose that
BP CR AT
· · = 1.
CQ AS BU

Let X, Y , Z be the midpoints of P Q, RS, T U , respectively. Prove that DX, EY ,

F Z are concurrent.

4947. Proposed by George Apostolopoulos.

Let ABC be a given triangle and let M, N be the interior points on the side BC
such that BM = CN . Prove that

AB 2 AC 2
+ > AB + AC.
AN AM

4948. Proposed by Mihaela Berindeanu.

2 2
Show that (log7 133) + (log19 133) > 9.

4949. Proposed by Goran Conar, modified by the Editorial Board.

Let a, b, c > 0 be positive real numbers such that a + b + c = 3. Prove that

ab bc ca 3
+ + ≥ ,
c(a + b)2 a(b + c)2 b(c + a)2 4

with equality if and only if a = b = c = 1.

Crux Mathematicorum, Vol. 50(5), May 2024

 Problems /257

4950. Proposed by Michel Bataille.

Let Fn denote the n-th Fibonacci number defined by Fn+1 = Fn + Fn−1 with
F0 = 0, F1 = 1 for any positive integer n. Prove that
∞ ∞
X Fn+3 2
X F2n+6
2 · 240 · = 379 and 2 · (240) · 2F 2
= 34801.
F F
n=1 n n+6
F
n=1 n n+6

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

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

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

Pour faciliter l’examen des solutions, nous demandons aux lecteurs de les faire parvenir
au plus tard le 31 juillet 2024.

4941. Soumis par Michel Bataille.

Soit un triangle ABC, non équilatéral, inscrit dans un cercle Γ de centre O. Son
cercle inscrit γ, de centre I, rencontre BC, CA, AB en D, E, F , respectivement.
Le cercle ΓA passe par A et est tangent à ID en I; ΓB et ΓC sont définis de façon
similaire. Démontrer que ΓA , ΓB , ΓC et la ligne OI ont un point en commun autre
que I.

4942. Soumis par Ivan Hadinata.

Soit n ∈ Z+ . Or, on sait que les suites (ak )∞ ∞
k=1 et (bk )k=1 sont des suites de
nombres réels positifs telles que
Ç åk Å ã Ç å
n1 n2 nk n1 n2 nk
ãÅ
k
ak + + ··· + = b1 + b2 + · · · bk + , ∀k ∈ Z+ .
1! 2! k! 1! 2! k!

Démontrer que (ak )∞ ∞

k=1 est supérieurement bornée si et seulement si (bk )k=1 est
supérieurement bornée.

4943. Soumis par Mihaela Berindeanu.

Soit ABC un triangle de cercle circonscrit Γ et orthocentre H. Les projections
du point H vers les tangentes au cercle Γ passant par B et C sont dénotées B1 et
C1 respectivement; les points milieux des côtés AC et AB sont dénotés B2 et C2
respectivement. Démontrer que si B1 B2 = C1 C2 , alors ABC est isocèle.

Copyright © Canadian Mathematical Society, 2024

258/ Problems

4944. Soumis par Vasile Córtoaje.

Soient a, b, c, d, e des nombres réels non négatifs tels que ab + bc + cd + de + ea = 5.
Démontrer que:
(a) (a + 2)2 + (b + 2)2 + (c + 2)2 + (d + 2)2 + (e + 2)2 ≥ 45.
(b) a3/2 + b3/2 + c3/2 + d3/2 + e3/2 ≥ 5.

4945. Soumis par Daniel Sitaru.

Si x, y, z ∈ R, démontrer que
X»
(− cos x + cos y + cos z)2 + (− sin x + sin y + sin z)2 ≥ 3.
cyc

4946. Soumis par Chikara Tsugawa.

Soient ABC et DEF deux triangles. Soient aussi P = BC ∩ F D, Q = BC ∩ DE,
R = CA ∩ DE, S = CA ∩ EF , T = AB ∩ EF et U = AB ∩ F D (tel qu’illustré
ci-bas). Supposer que
BP CR AT
· · = 1.
CQ AS BU
Enfin, soient X, Y , Z les points milieux de P Q, RS, T U , respectivement. Démontrer
que DX, EY , F Z sont concourantes.

4947. Soumis par George Apostolopoulos.

Soit ABC un certain triangle et soient M , N des points à l’intérieur du côté BC
tels que BM = CN . Démontrer que

AB 2 AC 2
+ > AB + AC.
AN AM

Crux Mathematicorum, Vol. 50(5), May 2024

 Problems /259

4948. Soumis par Mihaela Berindeanu.

2 2
Démontrer que (log7 133) + (log19 133) > 9.

4949. Soumis par Goran Conar, modifié par le Comité de Rédaction.

Soient a, b, c des nombres réels positifs tels que a + b + c = 3. Démontrer que
ab bc ca 3
+ + ≥ ,
c(a + b)2 a(b + c)2 b(c + a)2 4

avec égalité si et seulement si a = b = c = 1.

4950. Soumis par Michel Bataille.

Soit Fn le n-ième nombre de Fibonacci selon la définition F0 = 0, F1 = 1 et
Fn+1 = Fn + Fn−1 pour tout entier positif n. Démontrer que
∞ ∞
X Fn+3 X F2n+6
2 · 240 · = 379 and 2 · (240)2 · = 34801.
F F
n=1 n n+6
F 2F 2
n=1 n n+6

Copyright © Canadian Mathematical Society, 2024

260/ 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 2023: 49(10), p. 539–543.

4891. Proposed by Mihaela Berindeanu.

Let x, y, z be nonegative real numbers such that x + y + z > 0. Show that
√
p p p xy + xz + yz
4x + y + z + x + 4y + z + x + y + 4z ≥ 18 √ √ √ .
x+y+ x+z+ y+z
√ √ √
(Hint. You may consider the points A = ( x, 0, 0), B = (0, y, 0), C = (0, 0, z)
in an Euclidean system of coordinates.)
We received 10 submissions, 8 of which were correct and complete. We present
two solutions.
Solution 1, by Theo Koupelis.
Using the AM-GM inequality, we get

(x + y)(x + y + 4z) = (x + y)2 + 4z(x + y) ≥ 4(xy + yz + zx).

Similarly, we obtain

(y + z)(y + z + 4x) ≥ 4(xy + yz + zx) and (z + x)(z + x + 4y) ≥ 4(xy + yz + zx).

Thus, Cauchy-Schwarz gives

√ √ √
Äp p p ä
x+y+ y+z+ z+x x + y + 4z + y + z + 4x + z + x + 4y
 » 2
≥ 3 · 4 4(xy + yz + zx)
√
= 18 xy + yz + zx.

Equality occurs when x = y = z > 0.

Solution 2, by Marie-Nicole Gras, slightly altered by the editor.

Let
hp p p i h√ √ √ i
F = 4x + y + z + x + 4y + z + x + y + 4z · y+z+ x+z+ x+y
√
and we will prove that F ≥ 18 xy + yz + zx.
Using the AM-GM inequality, we obtain F ≥ G, where
» »
G = 3 6 (4x + y + z)(x + 4y + z)(x + y + 4z) · 3 6 (y + z)(x + z)(x + y).

Crux Mathematicorum, Vol. 50(5), May 2024

 Solutions /261

p
It now suffices to prove G ≥ 18 6 (xy + yz + zx)3 , which is equivalent to showing

3
that H ≥ 4(xy + yz + zx) , where

H = (4x + y + z)(x + 4y + z)(x + y + 4z)(y + z)(x + z)(x + y)

   
= (4x + y + z)(y + z) (x + 4y + z)(x + z) (x + y + 4z)(x + y)
= (4x(y + z) + (y + z)2 (4y(x + z) + (x + z)2 (4z(x + y) + (x + y)2
   

= (4(xy + xz + yz) + (y − z)2 (4(yx + yz + xz) + (x − z)2 (4(zx + zy + xy) + (x − y)2 .

   

3
Expressed in this form it is clear that H ≥ 4(xy + yz + zx) .

4892. Proposed by Dong Luu.

Consider an acute triangle ABC, (AB < AC). Let (I), (O) be the inscribed and
circumscribed circles, respectively. Let D, E, F be the points of contact of (I) and
BC, CA, AB and let H be the midpoint of arc BAC. Suppose HD cuts (O) a
second time at K and suppose AK cuts EF at J. Finally, let M, N be midpoints
of BC and AJ, respectively. Prove that the four points M, N, O, K lie on the same
circle.
We received 4 submissions, all correct, and feature the solution by Theo Koupelis.
Referring to the figure below, let T be the second point of intersection of (O) with
the line AO, and let S be the second point of intersection of the line T I with (O).
We have 90◦ = ∠IF A = ∠ISA = ∠IEA, and thus the points S, F, I, E, A are on
the circle whose diameter is AI; denote the circle by (L), where L is its center.

Let P be the midpoint of the minor arc BC. Since OM is the perpendicular
bisector of BC, the points H, O, M, P are collinear; so are points A, I, P. Points
B, I, C are on the circle (P, P I) (because ∠P BI = ∠P IB = (∠A + ∠B)/2, and
∠P IC = ∠P CI = (∠A + ∠C)/2).

Copyright © Canadian Mathematical Society, 2024

262/ Solutions

Let V be the point of intersection of the lines AS and BC. Then points V, S, I, D
are concyclic because ∠V SI = ∠V DI = 90◦ . Considering the power of point V
with respect to the circles (O), (L), and (P ) we get V S · V A = V B · V C. But the
circles (L) and (P ) are tangent at I, and thus the line V I is the common tangent
at I of the circles (L) and (P ). Therefore,

∠DSI = ∠DV I = ∠IP H = ∠AP H = ∠P HT = ∠P AT = ∠P ST.

Thus the points P, D, S are collinear.

The quadrilateral KDM P is cyclic because ∠DKP = ∠DM P = 90◦ . From the
isosceles triangle KOS we get

∠KSO = (180◦ −∠SOK)/2 = 90◦ −∠SHK = 90◦ −∠SP K = ∠KDP = ∠KM P.

Therefore, the quadrilateral SOM K is cyclic. It remains to prove that N lies on

this circle.
Let U be the point of intersection of the line SH with the line BC. Then we
have ∠U SP = ∠U M P = 90◦ , and thus U, S, M, P are concyclic. But also since
∠DSH = ∠DM H = 90◦ , we conclude that S, H, M, D are concyclic. Therefore,

∠P U M = ∠P SM = ∠DSM = ∠DHM = ∠KHP,

so that P U ⊥ HK. But P K ⊥ HK at K (because P H is a diameter of the circle

(P KH)), whence K is on U P.

Let R be the point of intersection of the line ID with the line SH. In the right
triangle U DR we have DS ⊥ U R, and thus

U D2 = U S · U R.

Crux Mathematicorum, Vol. 50(5), May 2024

 Solutions /263

Also,
∠SOB + ∠P OC ∠SOB + ∠BOP ∠SOP
∠SRI = ∠U RD = ∠SDU = = =
2 2 2
= ∠SAP = ∠SAI. (1)

Therefore, point R is on the circle (L), and point U has the same power with
respect to circles (L) and (I), which have F E as a common chord. Therefore, U
is on the line F E.
Let J 0 be the point of intersection of ST with the line F E, and let X be the
midpoint of F E. We prove that J 0 coincides with the given point J = F E ∩ AK:
because AI is the perpendicular bisector of F E, it follows that SAXJ 0 is cyclic.
Therefore, with the help of equation (1), ∠SJ 0 U = ∠SAI = ∠SDU. Thus, J 0 is
on the circle (U SDK), whence
∠P OH − ∠SOK ∠HOS + ∠KOP ∠AOS + ∠KOT
∠KJ 0 T = ∠KU S = = = .
2 2 2
Therefore, the points K, J 0 , A are collinear, and thus J 0 ≡ J, as desired.
Points O, L, N are the midpoints of the segments AT, AI, AJ, respectively. There-
fore, these points are on the line OL; it is parallel to ST and is the perpendicular
bisector of segment AS, the common chord of the circles (O) and (L). If W is the
midpoint of segment SA, then

∠KN O = ∠AN W = 90◦ − ∠SAK = 90◦ − ∠SHK = ∠KSO

(where for the last equality, ∠KSO is the complement of the angle between the
chord SK and the tangent at S to the circumcircle; that angle equals ∠SHK).
Therefore, the quadrilateral SN OK is cyclic. This completes the proof that the
points S, N, O, M, K are concyclic.
(Note: Some of the incidences suggested by the diagrams are indeed further prop-
erties of the configuration; for example, the midpoint Q of the segment DH is also
on the circle (SN OM K), because QO k DP and thus

∠KQO = ∠KDP = ∠KM P = ∠KSO.

Similarly, the midpoint Z of the segment DP is also on circle (SN OM K), because
QZ k HP and thus ∠SZQ = ∠SP H = ∠SKH.)
Editor’s comments. Two of the solutions used cartesian coordinates (and consid-
erable computer power) to verify the result. Such arguments are unreadable by
humans, but they are useful in showing that there is no need to assume that the
given triangle be acute (except to ensure that the diagram can be used to guide the
proof). On the other hand, the assumption that AB < AC is significant — should
AB = AC, the featured proof collapses and the points M, N, O, K are collinear
(not concyclic). The proposer’s solution was based on the interesting observation
that the points S and K on the circumcircle are harmonic conjugates with respect
to the vertices B and C.

Copyright © Canadian Mathematical Society, 2024

264/ Solutions

4893. Proposed by Albert Natian.

Find all continuous real functions f on [−1, 1] that satisfy the integral equation
Z 1
x
x2 + f x2 t dt = 1.


1

We received 16 submissions, 14 of which were correct and complete. We present

the solution submitted by Michel Bataille, which is representative of the majority
of solutions.
It is readily checked that the function x 7→ 2x is continuous on [−1, 1] and satisfies
the integral equation. We show that there is no other solution.
u
Let f be a solution. Let x ∈ [−1, 1], x 6= 0. The change of variables t = 2 in the
x
integral of the equation yields
Z x
f (u) du = x2 − x4 ,
x2

so that, F (x) − F (x2 ) = x2 − x4 where F is a primitive of f on [−1, 1] (F exists

since f is continuous). By differentiation, we obtain

f (x) − 2xf (x2 ) = 2x − 4x3 . (1)

We first deduce that

f (−x) + 2xf (x2 ) = −2x + 4x3

and by addition with (1) that f (x) + f (−x) = 0. From the continuity of f at 0,
we see that 2f (0) = 0, hence f (−x) = −f (x) for all x ∈ [−1, 1]: f is odd. Note
that (1) also provides f (1) = 2.
Now, suppose that 0 < x < 1. For any positive integer n, we have
n n n
Ä ä
f (x) − 2x = 2n x2 −1 f (x2 ) − 2x2 (H(n))

The proof is by induction. (1) shows that H(1) holds. Assume that H(n) holds
n
for some n ≥ 1. From (1) applied to x2 instead of x, we obtain
n n n n+1 n
f (x2 ) − 2x2 = 2x2 f (x2 ) − 4x3·2 ,

which, combined with H(n), gives

n n n+1 n+1 n+1 n+1 n+1
Ä ä Ä ä
−1
f (x)−2x = 2n x2 ·2x2 f (x2 ) − 2x2 = 2n+1 x2 −1 f (x2 ) − 2x2 ,

hence H(n + 1) holds, completing the proof.

Crux Mathematicorum, Vol. 50(5), May 2024

 Solutions /265

Now, since 0 < x < 1, we have

n
lim x2 = 0
n→∞

and, f being continuous at 0,

n
lim f (x2 ) = f (0) = 0,
n→∞

hence H(n) gives f (x) = 2x. Since f (0) = 0, f (1) = 2 and f is odd, we see that
f (x) = 2x for all x ∈ [−1, 1] and we are done.

Editor’s Comments. In their solution, Raymond Mortini and Rudolf Rupp explic-
itly noted the need to interpret this functional equation at x = 0 as
Ç Z 1/x å
2 2
lim x + f (x t)dt = 1.
x→0 1

4894. Proposed by Ovidiu Furdui and Alina Sı̂ntămărian.

Calculate
∞
X Hn−1 Hn+1
,
n=1
n(n + 1)
1 1
where Hn = 1 + 2 + ··· + n denotes the nth harmonic number and H0 = 0.
We received 17 submissions, 16 of which were correct. We present 2 solutions,
both slightly abridged.
Solution 1, by Kee-Wai Lau. We have
M
X Hn−1 Hn+1
S(M ) :=
n=1
n(n + 1)
M 1
X 1 
= Hn−1 Hn+1 −
n=1
n n+1
M M +1
X Hn−1 Hn+1 X Hn−2 Hn
= −
n=2
n n=2
n
M
X Hn−1 Hn+1 − Hn−2 Hn HM −1 HM +1
= − .
n=2
n M +1

Now
 1  1   1 1 
Hn−1 Hn+1 − Hn−2 Hn = Hn − Hn + − Hn − − Hn
n n+1 n n−1
 1 1  1
= + Hn − ,
n−1 n+1 n(n + 1)

Copyright © Canadian Mathematical Society, 2024

266/ Solutions

so
HM −1 HM +1
S(M ) = S1 (M ) + S2 (M ) − S3 (M ) − ,
M +1
where
M M M
X Hn X Hn X 1
S1 (M ) = , S2 (M ) = , S3 (M ) = .
n=2
(n − 1)n n=2
n(n + 1) n=2
2
n (n + 1)

It is easy to prove by induction that

1 HM
S1 (M ) = 2 − − ,
M M
M
X 1 1 HM
S2 (M ) = 2
+ − , and
n=2
n 2 M +1
M
X 1 1 1
S3 (M ) = 2
− + .
n=2
n 2 M +1

It follows that
 1 1  HM −1 HM +1
S(M ) = 3 − + (1 + HM ) − .
M M +1 M +1
Since HM = O(ln M ) as M tends to infinity, we conclude lim S(M ) = 3.
M →∞

Solution 2, by Michel Bataille. From

 1  1  Hn 1
Hn−1 Hn+1 = Hn − Hn + = Hn2 − −
n n+1 n(n + 1) n(n + 1)

and the convergence (since Hn ∼ ln(n) as n → ∞) of the three series

∞ ∞ ∞
X Hn2 X Hn X 1
A= , B= , and C = ,
n=1
n(n + 1) n=1
n (n + 1)2
2
n=1
n2 (n + 1)2

we deduce that the required sum S is A − B − C.

By Problem 3417 of Crux [2009 : 109, 111 ; 2010 : 114], we have A = 3ζ(3) and
B = 3ζ(3) − 2ζ(2), while from

1 1 1 1 1 
= + − 2 −
n2 (n + 1)2 n2 (n + 1)2 n n+1


we obtain C = ζ(2) + ζ(2) − 1 − 2 = 2ζ(2) − 3. Thus S = 3.

Editor’s Comments. Every submission lay somewhere between these two. Of the
many closed forms for the partial sums obtained by solvers, Solution 1’s stood out

Crux Mathematicorum, Vol. 50(5), May 2024

 Solutions /267

as the cleanest with the simplest derivation. R. Mortini and R. Rupp report that
WolframAlpha has trouble evaluating this series directly.

4895. Proposed by Shawn Godin and J. Chris Fisher.

For any triangle ABC with centroid G and circumcenter O, denote by G0 the
centroid of the triangle A0 B 0 C 0 formed by the lines through the midpoints D, E,
and F of the sides that are perpendicular to the corresponding medians (namely,
AD, BE, CF ). Prove that G is the midpoint of the segment G0 O.
We received 10 submissions of which 9 were correct and complete. We present a
solution by Michel Bataille, modified by the editor.
Let A1 be the image of A0 , B1 the image of B 0 and C1 the image of C 0 under the
homothety with center G and factor 21 (see Figure 1).

A'

A
A1

F G' E
G1
G

B1
O
C1
B' C
D
B
C'

Figure 1: Construction of 4A1 B1 C1 with centroid G1 .

−−→ −−→
Denote by G1 the centroid of 4A1 B1 C1 ; then G1 satisfies GG1 = 21 GG0 . Therefore
−−→ −−→
it suffices to show that GG1 = − 12 GO; referring to the properties of the Euler line,
this equality will hold if G1 is the center of the nine-point circle for 4ABC, that
is, if G1 is the circumcenter of 4F DE.
Denote by M the midpoint of B1 C1 ; we will prove that the median A1 M is the

Copyright © Canadian Mathematical Society, 2024

268/ Solutions

perpendicular bisector of F E. Denote by N the midpoint of F E, and note that N

is the intersection of the median AD in 4ABC with the midline F E; in particular,
D, G and N are collinear (see Figure 2).

A'

A1

U N V E
F

B1
M
C1
B'
C
D
B
C'

Figure 2: Additional constructions used in the proof that A1 M ⊥ EF .

Note: Does not show 4A1 B1 C1 .

From the construction described in the problem, ∠A0 F G = ∠A0 EG = 90◦ , so A1

is the circumcenter of the cyclic quadrilateral A0 F GE. In particular, A1 F = A1 E
and so A1 N ⊥ F E. That is, A1 is on the perpendicular bisector of F E and we
just have to show that M N ⊥ F E.

Let ΓF (center B1 , radius B1 G) and ΓE (center C1 , radius C1 G) be the circumcir-

cles of the cyclic quadrilaterals B 0 F GD and C 0 EGD, respectively. Suppose F E
intersects ΓF at F and U and ΓE at V and E. Denote by H the midpoint of F U
and by K the midpoint of V E (see Figure 3).

Crux Mathematicorum, Vol. 50(5), May 2024

 Solutions /269

K E
H U N V
F

B1
M
C1

Figure 3: Zooming in on circles ΓE and ΓF

Since 4B1 F U and 4C1 V E are isosceles, B1 H and C1 K are both perpendicular
to F E. Since N is on the radical axis DG of ΓE and ΓF , the power of a point
theorem applied to N and each of the circles gives us N U · N F = N V · N E,
from which N U = N V . Hence F U = EV and, since H and K are midpoints,
HU = KV . Thus HU + U N = N V + V K, so N is also the midpoint of HK, and
the line N M is the midline in the trapezoid HB1 C1 K. Hence N M is parallel to
HB1 , so N M ⊥ F E, as desired.
We have thus shown that A1 M is the perpendicular bisector of F E. A similar
property will hold for the other two medians of 4A1 B1 C1 , allowing us to conclude
that the intersection G1 of the medians of 4A1 B1 C1 is the circumcenter of 4F DE.
Thus G1 is the center of the nine-point circle for 4ABC, concluding the proof.
Further comments. Observe that the four points H, G0 , G, O are equally spaced
along the Euler line of ∆ABC (where H is the orthocenter of ∆ABC). A fifth
point on the Euler line, namely the nine-point center N , is the midpoint of both
G0 G and HO. In Clark Kimberling’s Encyclopedia of Triangle Centers, G0 is
assigned the number X(381). Many properties are listed there, but none are as
interesting as its role as the centroid of ∆A0 B 0 C 0 . One further property: the
bisector of ∠B 0 A0 C 0 also bisects ∠G0 A0 G, and similarly for the angles at B 0 and
C 0 ; in other words, G is the symmedian point of ∆A0 B 0 C 0 .

4896. Proposed by Ivan Hadinata.

Find all functions f : R −→ R such that for all x, y, z ∈ R the following equation
holds:
f (f (x) + yf (z) − 1) + f (z + 1) = zf (y) + f (x + z).

We received 17 solutions from 15 solvers. Of these, 16 were correct and 1 was

incomplete. The first two solutions presented below are quite efficient; the third

Copyright © Canadian Mathematical Society, 2024

270/ Solutions

uses elements from various submissions.

Solution 1, by Raymond Mortini and Rudolf Rupp.
The function f (x) = 0 for all x is a solution. Suppose there is a solution for which
f (k) 6= 0 for some real k. Denote the equation by E(x, y, z). From E(1, k, z), we
find that
f (f (1) + kf (z) − 1) = f (k)z.
Since the right side can assume any real value, f is surjective. The equation
E(x, 0, 0) is equivalent to

f (f (x) − 1) = f (x) − f (1),

whence f (u − 1) = u − f (1) for every real u. Setting u = 2 yields f (1) = 1, so

that f (t) = t for real t = u − 1. Therefore, there are two solutions: f (x) = 0 and
f (x) = x.

Solution 2, by José Luis Arregui.

Denote the equation by E(x, y, z). From E(1, 0, z), we find that f (f (1)−1) = zf (0)
for all z, from which f (f (1) − 1) = f (0) = 0. Comparing the equations E(x, 1, z)
and E(z, 1, x), we arrive at f (z + 1) − zf (1) = f (x + 1) − xf (1) for all x, z. Let
z = −1. Then f (x + 1) = (x + 1)f (1), or equivalently f (x) = xf (1) for all x.
Either f (1) = 0, whereupon f (x) is identically 0. Or f (1) 6= 0, whereupon f (x) = 0
if and only if x = 0. But then f (1) − 1 = 0, so that f (x) = x for all x.

Solution 3, a composite of several solutions.

Suppose, if possible, that f (0) 6= 0. Then E(1, 1−f (1))
f (0) , 0) implies that f (0) = 0,
yielding a contradiction. Thus, f (0) = 0. The zero function is a solution. Suppose
f is a solution with f (k) 6= 0 for some k. If f (z1 ) = f (z2 ), then comparing
E(1, k, z1 ) and E(1, k, z2 ) leads to f (k)z1 = f (k)z2 . Hence, z1 = z2 and f is
injective. From E(1, 0, 1), f (f (1) − 1) = f (0), so that f (1) = 1.
E(1, 1, z) leads to f (f (z)) = z, so f is surjective. Finally, E(x, y, 0) implies that
f (f (x) − 1) = f (x) − 1, so that f (t) = t for each real t.

4897. Proposed by Michel Bataille.

Let the sequence (xn )n≥0 be defined by x0 = 2, x1 = 7 and xn+1 = 4xn − xn−1 for
all n ≥ 1. Prove that for all n ≥ 0,
p p
b 8x2n+1 + 6c − b 2x2n+1 + 1c = 2xn .

We received 10 interesting solutions for this problem, all correct. The following is
the solution by M. Bello, M. Benito, Ó. Ciaurri and E. Fernández and presents
the main ideas of all the solutions in more details.

Crux Mathematicorum, Vol. 50(5), May 2024

 Solutions /271

The recurrence law xn+1 = 4xn − xn−1 is a homogeneous discrete linear equation
of constant coefficients and characteristic equation

r2 − 4r + 1 = 0.
√
Characteristic roots: r = 2 ± 3. General solution:
√ √
xn = A(2 + 3)n + B(2 − 3)n .

The initial terms x0 = 2, x1 = 7 provide the equations

®
A + B = 2,
√ √
(2 + 3)A + (2 − 3)B = 7,

from which it follows

1 √ 1 √
A= (2 + 3), B= (2 − 3),
2 2
so that
1Ä √ √ ä
xn = (2 + 3)n+1 + (2 − 3)n+1 . (1)
2
√ √
From (1) and taking into account that (2 + 3)(2 − 3) = 1 we have, on the one
hand,
√ √ √ √
2x2n+1 + 1 = (2 + 3)2n+2 + (2 − 3)2n+2 + 2(2 + 3)n+1 (2 − 3)n+1 − 1
Ä √ √ ä2
= (2 + 3)n+1 + (2 − 3)n+1 − 1
= (2xn )2 − 1,

from which we can easily deduce

p »
2xn − 1 ≤ 2x2n+1 + 1 = (2xn )2 − 1 < 2xn ,
√
given that for any a ≥ 1 the inequalities a − 1 ≤ a2 − 1 < a both hold. And so,
taking integer parts (we can see by induction that xn is a positive integer for all
n), p
2xn − 1 ≤ b 2x2n+1 + 1c < 2xn . (2)
On the other hand, we also have
√ √ √ √
8x2n+1 + 6 = 4(2 + 3)2n+2 + 4(2 − 3)2n+2 + 6(2 + 3)n+1 (2 − 3)n+1
Ä √ √ ä2
= 2(2 + 3)n+1 + 2(2 − 3)n+1 − 2
= (4xn )2 − 2,

from which we can deduce

p »
4xn − 1 ≤ 8x2n+1 + 6 = (4xn )2 − 2 < 4xn ,

Copyright © Canadian Mathematical Society, 2024

272/ Solutions

√
given that for any a ≥ 3/2 the inequalities a − 1 ≤ a2 − 2 < a both hold. So,
again taking integer parts
p
4xn − 1 ≤ b 8x2n+1 + 6c < 4xn . (3)

Finally, from (2) and (3) it follows

p p
2xn −1 = (4xn −1)−2xn < b 8x2n+1 + 6c−b 2x2n+1 + 1c < 4xn −(2xn −1) = 2xn +1
√ √
and, since b 8x2n+1 + 6c − b 2x2n+1 + 1c is an integer,
p p
b 8x2n+1 + 6c − b 2x2n+1 + 1c = 2xn .

4898. Proposed by George Apostolopoulos.

Let ABC be a triangle with ∠B = 2∠C. Consider an interior point D on the side
BC such that ∠C = 2∠CAD. If b and c are the lengths of sides AC and AB
respectively, prove that
b p
AD = bc + 2c2 .
b+c

We received 22 solutions for this problem. We present 2 solutions.

Solution 1, by Rob Downes.
Without loss of generality, let BC = 1. Then CD = n, and DB = 1 − n. Also, let
d = AD. The triangle with the given information is shown below:

By summing areas, we have [ACD] + [ADB] = [ACB]. This immediately gives:

1 1 1
bn sin 2θ + (1 − n)c sin 4θ = b sin 2θ.
2 2 2
Using the fact that sin 4θ = 2 sin 2θ cos 2θ, rearranging terms, and simplifying
yields:
b = 2c cos 2θ.

Crux Mathematicorum, Vol. 50(5), May 2024

 Solutions /273

Since cos 2θ = 2 cos2 θ − 1, we have:

b + 2c
= cos2 θ. (1)
4c
Applying the Law of Sines to 4ACD gives:

sin 2θ sin (π − 3θ)

= . (2)
d b
Using elementary trig identities such as the angle sum formulas, it is easily shown
that sin (π − 3θ) = sin θ(4 cos2 θ − 1). Substituting this result into (2) and simpli-
fying, we get:
2b cos θ = d(4 cos2 θ − 1). (3)
Lastly we use (1) to further simplify (3) giving:
…
b + 2c b + 2c
Å ã
2b =d 4 −1 .
4c 4c
Simplifying and solving for d, we arrive at the desired result:
b p
AD = bc + 2c2 .
b+c

Solution 2, by Miguel Amengual Covas.

Let a be the length of side BC. We put DC = m. Then BD = a − m. By
Stewart’s theorem, applied to 4ABC and cevian AD,

b2 (a − m) + c2 m − a (a − m) m
AD2 = . (4)
a
Now we use the fact that ∠B = 2∠C is equivalent to the condition

b2 = c (c + a) , (5)

which has appeared before in Crux [1996:74], [1978:278], [1984:287] and [1996:265-
267].
In turn, the fact that ∠C = 2∠CAD is equivalent to the condition

AD2 = DC · DC + CA = m (m + b) .



(6)

In (4) we substitute c (c + a) for b2 from (5), obtaining

AD2 = c2 + (a − m) (c − m) .

We subtract this equation from (6) and find that

c (c + a) c (c + a) b2 bc
m= = = (using (2)) = 2 = .
a+b+c b + (c + a) b + bc b+c

Copyright © Canadian Mathematical Society, 2024

274/ Solutions

When this is substituted into (6), we get

b2 bc + 2c2


b2 + 2bc
Å ã
2 bc bc bc
AD = +b = · = 2 ,
b+c b+c b+c b+c (b + c)
so
b p
AD = bc + 2c2 ,
b+c
as desired.

4899. Proposed by Aravind Mahadevan.

Let ABC be a right-angle triangle with ∠B = 90◦ . Suppose AE and CD are
√ angle
bisectors of angles BAC and ACB, respectively. If AE = 9 and CD = 8 2, find
the length of AC.
We received 16 complete and correct submissions. Presented is the one by Antoine
Mhanna, slightly expanded by the editor.

π
Let θ = ∠DCB. Then ∠BAE = − θ. The given side lengths imply
4
√
BC = 8 2 cos θ
and using the sum of angles formula for cosine,
π 
AB = 9 cos −θ
 4 π  π 
= 9 cos cos θ + sin sin θ
√ 4 4
9 2

= cos θ + sin θ .
2

Therefore
9 cos θ + sin θ 9
Å ã
AB
tan (2θ) = = = (1 + tan θ) .
BC 16 cos θ 16

Crux Mathematicorum, Vol. 50(5), May 2024

 Solutions /275

On the other hand, using the double angle formula for tangent, we get
2 tan θ
tan (2θ) = .
1 − tan2 θ

Setting x = tan θ we obtain from the last two equations

2x 9
= (1 + x),
1−x 2 16
1
which has x = as its only real solution. From 4DCB we get
3
√ 10
(8 2)2 = BC 2 + (tan θ · BC)2 = BC 2
9
24 3
or BC = √
5
and from 4ABC and tan (2θ) = 4 we get
25
AC 2 = BC 2 + (tan (2θ) · BC)2 = BC 2
16
√
or AC = 6 5.

4900. Proposed by Daniel Sitaru.

For a positive integer m, let Hm denote the m-th harmonic number, that is,
Hm = 1 + 21 + · · · + m
1
. For m, n, p, q positive integers, prove that
Hm + Hn + Hp + Hq ≤ 3 + Hmnpq .

We received 21 submissions and they were all correct. We present the following
solution by the majority of solvers.
Observe that if m > 1, then
mn
X 1 X Å (k+1)n
m−1 X 1 ã m−1X 1
Hmn − Hn = = ≥ = Hm − 1.
j=n+1
j j k+1
k=1 j=kn+1 k=1

It follows that Hm + Hn ≤ 1 + Hmn for all positive integers m, n. Thus, for any
positive integers m, n, p, q, we have
Hm + Hn + Hp + Hq ≤ Hmn + 1 + Hpq + 1 ≤ 3 + Hmnpq .

Editor’s Comment. Another common solution by the solvers uses the well-known
fact that Hn can be well approximated by log n + γ, where γ ≈ 0.577 is the
Euler-Mascheroni constant. In fact, several readers pointed out the inequality
γ + Hmn ≤ Hm + Hn ≤ 1 + Hmn has appeared in Problem E819, Amer. Math.
Monthly 56 (1949), no.2.

Copyright © Canadian Mathematical Society, 2024