Crux Mathematicorum

VOLUME 41, NO. 9 November / Novembre 2015

Editorial Board
Editor-in-Chief Kseniya Garaschuk University of British Columbia

Contest Corner Editor John McLoughlin University of New Brunswick

Olympiad Corner Editor Carmen Bruni University of Waterloo
Book Reviews Editor Robert Bilinski Collège Montmorency
Articles Editor Robert Dawson Saint Mary’s University

Problems Editors Edward Barbeau University of Toronto

Chris Fisher University of Regina
Edward Wang Wilfrid Laurier University
Dennis D. A. Epple Berlin, Germany
Magdalena Georgescu University of Toronto

Assistant Editors Chip Curtis Missouri Southern State University

Lino Demasi Ottawa, ON
Allen O’Hara University of Western Ontario

Guest Editors Joseph Horan University of Victoria

Mallory Flynn University of British Columbia
Kelly Paton University of British Columbia
Alessandro Ventullo University of Milan
Kyle MacDonald McMaster University

Editor-at-Large Bill Sands University of Calgary

Managing Editor Denise Charron Canadian Mathematical Society

c Canadian Mathematical Society, 2016

 IN THIS ISSUE / DANS CE NUMÉRO

371 How To Write A Crux Article Robert Dawson

373 The Contest Corner: No. 39 John McLoughlin
373 Problems: CC191–CC195
374 Solutions: CC141–CC145
377 The Olympiad Corner: No. 337 Carmen Bruni
377 Problems: OC251–OC255
378 Solutions: OC191–OC195
383 Book Reviews Robert Bilinski
386 Focus On . . . : No. 19 Michel Bataille
392 A Mathematical Performance
Hee-Joo Nam, Giavanna Valacco and Ling-Feng Zhu
397 Problems: 4081–4090
401 Solutions: 3981–3990
413 Solvers and proposers index

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
 ROBERT DAWSON /371

How To Write A Crux Article

Robert Dawson
Longtime Crux readers will notice that the dates on your magazines are getting
closer to the dates on your calendar. This is good news, and is largely thanks to
the hard work of our dedicated editor-in-chief (stand up and take a bow, Kseniya!)
Once we catch up, we do not plan to keep charging into the future at the same
rate – production will slow to one year per year. This means that our current
demand for articles may slow somewhat, though we’ll still need plenty.

This seems like a good time to remind people what a Crux article is, and isn’t.

(1) Crux is accessible. Crux is read by university professors and graduate stu-
dents. It’s also read by undergraduates, school teachers, school students, and
amateurs whose day jobs have nothing to do with mathematics. We ask prospec-
tive writers to write for a very clever high school student. Assume high intelligence
but not a lot of specialized knowledge. That said:

(2) Crux is primarily for problem solvers, and by “problems” I mean the sort
of thing that might appear on a regional or national math contest. We may run
puzzles and alphametics, but they aren’t our main topic, and we probably wouldn’t
run articles about them. The natural home for articles on those topics used to
be the late and much lamented Journal of Recreational Mathematics. Right now
perhaps the best place, if the article has some mathematical content, is probably
Mathematics Gazette, the Mathematical Intelligencer, or the College Math Journal.

You should assume that most Crux readers know the standard tricks of the trade.
Don’t stop and explain mathematical induction or double counting unless you’re
explaining something new or unusual about those topics.

A good Crux article is not one where readers watch you solve problems. It’s
one that tells a significant proportion of them something they didn’t know before
about how to solve problems.

(3) Crux is short. Our articles don’t usually run more than five or six pages, and
we’re more likely to run it if it’s three or four. That’s the right length for the
sort of thing we publish. We do like well-done illustrations, nice examples, and
interesting asides – but please keep it all brief.

(4) Crux is not a research journal. If you’ve written a research paper, please send
it somewhere else, where it has a chance of getting published. We publish a dozen
or so articles per year – the College Math Journal, at a similar level, publishes
almost that many in a month. None of our articles are research articles in the
conventional sense.

Crux is definitely not a “research journal of last resort.” We don’t even publish
good research papers, and we have no need at all for “research” that other journals
won’t print.

Copyright c Canadian Mathematical Society, 2016

372/ HOW TO WRITE A CRUX ARTICLE

*You cannot trisect the angle, square the circle, or duplicate the cube using clas-
sical tools. If you understand Wantzel’s classic 1837 proof, you won’t try. If you
don’t understand it, you have not done your preparatory work and you have no
business trying.
*You probably cannot prove the Riemann Conjecture, Fermat’s Last Theorem,
the ABC conjecture, or the Four Color Theorem in five pages. If you could, you
would not be sending your proof to Crux : you would be sending it to one of the
very top math journals. Finally, please don’t use a theorem-proof-lemma form.
It’s not our style.
Okay? Now you know what we’re looking for. Write it – get somebody to check it
over to make sure the math and English (or French) is correct – and send it to us.

Mathematics Books for Sale

Over 1500 used books from all areas of pure and applied mathematics, including
some computer science, statistics, and mathematical physics, are for sale at
http://www.mathstat.dal.ca/~dilcher/oldbooks.html
Of particular interest to Crux readers will be the sections on Problem Solving,
Puzzles and Games, Biographies, History, and General and Popular Mathematics.
This is a fundraiser, and prices are moderate. All proceeds go, in equal parts, to the
Canadian Mathematical Society and to the Dalhousie Department of Mathematics
and Statistics.
(Illustrations by Vladimı́r Jiránek.)

Crux Mathematicorum, Vol. 41(9), November 2015

 THE CONTEST CORNER /373

THE CONTEST CORNER

No. 39
John McLoughlin
The problems featured in this section have appeared in, or have been inspired by, a math-
ematics contest question at either the high school or the undergraduate level. Readers
are invited to submit solutions, comments and generalizations to any problem. Please see
submission guidelines inside the back cover or online.
To facilitate their consideration, solutions should be received by the editor by November
1, 2016, although late solutions will also be considered until a solution is published.
The editor thanks André Ladouceur, Ottawa, ON, for translations of the problems.

CC191. There are 32 competitors in a tournament. No two of them are equal

in playing strength, and in a one against one match the better one always wins.
Show that the gold, silver, and bronze medal winners can be found in 39 matches.

CC192. Let M be a 3 × 3 matrix with all entries drawn randomly (and with
equal probability) from {0, 1}. What is the probability that det M will be odd?

CC193. Consider the set of numbers {1, 2, . . . , 10}. Let {a1 , a2 , . . . , a10 } be
some permutation of these numbers and compute

|a1 − a2 | + |a3 − a4 | + · · · + |a9 − a10 |.

What is the maximum possible value of the above sum over all possible permuta-
tions and how many permutations give you this maximum value?

CC194. At a strange party, each person knew exactly 22 others. For any
pair of people X and Y who knew one another, there was no other person at the
party that they both knew. For any pair of people X and Y who did not know
one another, there were exactly 6 other people that they both knew. How many
people were at the party?

CC195. A bisecting curve is one that divides a given region into two sub-
regions of equal area. The shortest bisecting curve of a circle is clearly a diameter.
What is the shortest bisecting curve of an equilateral triangle?

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

CC191. Il y a 32 concurrents dans un tournoi et il n’y a pas deux concurrents

de force égale. Dans n’importe quel match entre deux concurrents, le plus fort
l’emporte toujours. Démontrer qu’il est possible de déterminer les récipiendaires
des médailles d’or, d’argent et de bronze après 39 parties.

Copyright c Canadian Mathematical Society, 2016

374/ THE CONTEST CORNER

CC192. Soit M une matrice 3 × 3 dont les éléments sont tous choisis de
façon aléatoire dans l’ensemble {0, 1}. Quelle est la probabilité pour que det M
soit impair?

CC193. On considère l’ensemble {1, . . . , 10} et une permutation {a1 , . . . , a10 }

de cet ensemble. On calcule

|a1 − a2 | + |a3 − a4 | + · · · + |a9 − a10 |.

Parmi toutes les permutations de l’ensemble, quelle est la valeur maximale de cette
somme et combien de permutations donnent cette valeur maximale?

CC194. Lors d’une drôle de fête, chaque personne connaı̂t 22 autres person-
nes. Pour chaque paire de personnes X et Y qui se connaissent l’une l’autre, il
n’y a aucune autre personne à la fête que X et Y connaissent tous les deux. Pour
chaque paire de personnes X et Y qui ne se connaissent pas l’une l’autre, il y a
6 personnes à la fête que X et Y connaissent tous les deux. Combien y a-t-il de
personnes à la fête?

CC195. Une courbe bissectrice d’une surface est une courbe qui coupe
la surface en deux régions de même aire. La courbe bissectrice d’un cercle est
évidemment un diamètre. Quelle est la courbe bissectrice la plus courte d’un
triangle équilatéral?

CONTEST CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2014: 40(9), p. 368–369.
All problems are from 42nd Ural tournament of young mathematicians, as printed in
Kvant 2014(1).

CC141. Alice writes down 100 consecutive natural numbers. Bob multiplies
50 of them: 25 smallest ones and 25 largest ones. He then multiplies the remaining
50 numbers. Can the sum of the two products be equal to 100! = 1 · 2 · . . . · 100?
Problem 2 of grade 7 level of 42nd Ural tournament of young mathematicians.
We received three solutions, two of which were correct. We present a composite of
the two correct solutions from Konstantine Zelator and Fernando Ballesta Yagüe.

Crux Mathematicorum, Vol. 41(9), November 2015

 THE CONTEST CORNER /375

Write the 100 consecutive natural numbers as n + 1, n + 2, . . . , n + 100, where n

is a natural number. Let P1 = (n + 1)(n + 2) · · · (n + 25) be the product of the
25 smallest numbers, P2 = (n + 26)(n + 27) · · · (n + 75) be the product of the 50
middle numbers, and P3 = (n + 76)(n + 77) · · · (n + 100) be the product of the
largest 25 numbers. We want to know if it’s possible to have P1 · P3 + P2 = 100!.
The factors in the products P1 , P2 and P3 contain 100 consecutive natural numbers.
Thus, if P1 · P3 + P2 = 100!, the greatest prime number smaller than 100, which
is 97, must be a divisor of each of P1 · P3 and P2 .
If 97 is a divisor of P2 , that means it is a divisor of one of its factors n + k,
k ∈ 26, . . . , 75. The neighbouring multiples of 97 are n + k − 97 and n + k + 97.
Both of these fall outside the range of the natural numbers in P1 (which, for all
options of n + k in P2 , must fall within n + k − 74 to n + k − 1) and P3 (which
must fall within n + k + 1 to n + k + 74). Thus if 97 is a divisor of P2 , then it
cannot be a divisor of either P1 or P3 . We cannot have P1 · P3 + P2 = 100!.

CC142. Roboto writes down a number. Every minute, he increases the existing
number by the double of the number of its natural divisors (including 1 and itself).
For example, if he started with 5, the sequence would be 5, 9, 15, 23, . . .. What is
the maximum number of perfect squares that appears on the board within 24
hours?
Problem 3 of grade 7 level of 42nd Ural tournament of young mathematicians.
We received one correct solution. We feature the solution by Konstantine Zelator.
We show that such a sequence contains at most one perfect square. We make two
observations. First, if n is a perfect square, then n ≡ 0 or 1 (mod 4) . Second,
the number of positive divisors of an integer is odd if and only if that integer is a
perfect square.
Suppose such a sequence contains a perfect square k. Then k has 2m + 1 divisors
for some integer m, and twice the number of divisors of k is 4m + 2. Since k is a
perfect square k ≡ 0 or 1 (mod 4) and k+4m+2 ≡ 2 or 3 (mod 4) . We notice that
numbers of these forms cannot be perfect squares, and so have an even number of
divisors. Thus, twice the number of their divisors is a multiple of 4. So after this
point, the terms in the sequence will be the same modulo 4 and hence can never
be perfect squares. Thus, there can be at most one perfect square in the sequence.

CC143. Summer Camp has attracted 300 students this year. On the first day,
the students discovered (as mathematicians would) that the number of triples of
students who mutually know each other is greater than the number of pairs of
students who know each other. Prove that there is a student who knows at least
5 other students.
Problem 9 of grade 7 level of 42nd Ural tournament of young mathematicians.
We received no solutions to this problem.

Copyright c Canadian Mathematical Society, 2016

376/ THE CONTEST CORNER

CC144. Year 2013 is the first one since Middle Ages that uses 4 consecutive
digits in its base 10 representation. How many other years like this will there be
before year 10,000?
Problem 1 of grade 8 level of 42nd Ural tournament of young mathematicians.
We received three solutions, of which two were correct and complete. We present
the most succinct, by Andrea Fanchini.
We have seven possible groups of four consecutive digits:

{0123, 1234, 2345, 3456, 4567, 5678, 6789}.

For each group we have 4! = 24 possible permutations. In total, this makes

7 · 24 = 168 possibilities. But from these we have to cancel those that start with
0, those that start with 1, and the year 2013.
The possibilities that start with 0 are 3! = 6 permutations (0 is present only in
the first group). The possibilities that start with 1 are 3! + 3! = 12 permutations
(1 is present in the first and second groups).
Finally, we remain with 168 − 6 − 12 − 1 = 149 years with the property requested.

CC145. Can a natural number be divisible by all numbers between 1 and 500
except for some two consecutive ones? If so, find these two numbers (show all
possible cases).
Problem 3 of grade 8 level of 42nd Ural tournament of young mathematicians.
We received two correct solutions. We present the solution of Titu Zvonaru below.
Let n be a natural number less than 500 satisfying the statement of the problem.
Let p be a number such that n is not divisible by p. If p = a · b with a, b > 1
relatively prime, we deduce that n is not divisible by a or is not divisible by b.
Since a (or b) and p are not consecutive, we obtain a contradiction. This yields
that p is a power of a prime. If p ≤ 250 then n is not divisible by 2p. It follows
that p > 250.
By the same reasoning, p + 1 must be a power of a prime, and p + 1 > 250. One
of them is even, meaning it must be a power of 2 (the only even prime). It is
not hard to see that 256 = 28 is the only number that fits our criteria. So either
p = 256 or p + 1 = 256. If p = 256 then p + 1 = 257 which is easily seen to be
prime. If p + 1 = 256 then p = 255 which factors as 255 = 3 · 5 · 17, not a power
of a prime. As a conclusion, the number

n = lcm(1, 2, . . . , 254, 255, 258, 259, . . . , 500)

is divisible by all numbers between 1 and 500, except for consecutive numbers 256
and 257.

Crux Mathematicorum, Vol. 41(9), November 2015

 THE OLYMPIAD CORNER /377

THE OLYMPIAD CORNER

No. 337
Carmen Bruni
The problems featured in this section have appeared in a regional or national mathematical
Olympiad. Readers are invited to submit solutions, comments and generalizations to any
problem. Please see submission guidelines inside the back cover or online.
To facilitate their consideration, solutions should be received by the editor by November
1, 2016, although late solutions will also be considered until a solution is published.
The editor thanks Rolland Gaudet, retired professor of Université de Saint-Boniface in
Winnipeg, for translations of the problems.

OC251. Let a, b, c and d be real numbers such that b − d ≥ 5 and all zeros
x1 , x2 , x3 , and x4 of the polynomial P (x) = x4 + ax3 + bx2 + cx + d are real. Find
the smallest value of the product
(x21 + 1)(x22 + 1)(x23 + 1)(x24 + 1).

OC252. In an obtuse triangle ABC (AB > AC), O is the circumcentre and
D, E and F are the midpoints of BC, CA and AB respectively. Median AD
intersects OF and OE at M and N respectively and BM meets CN at point P .
Prove that OP ⊥ AP .

OC253. Prove that there exist infinitely many positive integers n such that
3n + 2 and 5n + 2 are all composite numbers.

OC254. Find all non-negative integers k, n which satisfy

22k+1 + 9 · 2k + 5 = n2 .

OC255. Let n be a positive integer and x1 , x2 , . . . , xn be positive reals. Show

that there are numbers a1 , a2 , . . . , an ∈ {−1, 1} such that the following holds:
a1 x21 + a2 x22 + · · · + an x2n ≥ (a1 x1 + a2 x2 + · · · + an xn )2 .

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

OC251. Soit a, b, c et d des nombres réels tels que b − d ≥ 5 et tels que les
racines x1 , x2 , x3 et x4 du polynôme P (x) = x4 + ax3 + bx2 + cx + d sont toutes
réelles. Déterminer la plus petite valeur possible pour le produit
(x21 + 1)(x22 + 1)(x23 + 1)(x24 + 1).

Copyright c Canadian Mathematical Society, 2016

378/ THE OLYMPIAD CORNER

OC252. Pour un triangle obtus ABC (AB > AC), soit O le centre du cercle
circonscrit et soit D, E et F les mi points de BC, CA et AB respectivement. La
médiane AD intersecte OF et OE à M et N respectivement; BM rencontre CN
au point P . Démontrer que OP ⊥ AP .

OC253. Démontrer qu’il existe un nombre infini d’entiers positifs n tels que
3n + 2 et 5n + 2 sont composés.

OC254. Déterminer tous les entiers non négatifs k et n satisfaisant

22k+1 + 9 · 2k + 5 = n2 .

OC255. Soit n un entier positif et soit x1 , x2 , . . . , xn des nombres réels positifs.

Démontrer qu’il existe des nombres a1 , a2 , . . . , an ∈ {−1, 1} tels que l’inégalité
suivante tient:
a1 x21 + a2 x22 + · · · + an x2n ≥ (a1 x1 + a2 x2 + · · · + an xn )2 .

OLYMPIAD SOLUTIONS
Statements of the problems in this section originally appear in 2014: 40(7), p. 282–283.

OC191. Let P be a point in the interior of triangle ABC. Extend AP , BP ,

and CP to meet BC, AC, and AB at D, E, and F , respectively. If triangle AP F ,
triangle BP D and triangle CP E have equal areas, prove that P is the centroid of
triangle ABC.
Originally problem 2 from the 2013 Philippines Mathematical Olympiad.
We received four correct submissions. We present the solution by Michel Bataille.

Crux Mathematicorum, Vol. 41(9), November 2015

 THE OLYMPIAD CORNER /379

Let P = αA + βB + γC where α, β, γ > 0 and α + β + γ = 1. Since the point

P − αA = βB + γC is on AP and on BC, we have

(1 − α)D = P − αA = βB + γC
AP 1−α
and so PD = α . Similarly,

(1 − γ)F = P − γC = αA + βB
AF β
and so AB = 1−γ .

Since the altitudes F H and BK in triangles AF P and BP D, respectively, satisfy

FH AF
BK = AB (note that ∆AF H and ∆ABK are similar), we obtain

[AP F ] F H · AP AF.AP β(1 − α)

1= = = =
[BP D] BK · P D AB.P D α(1 − γ)
where [XY Z] denotes the area of ∆XY Z. It follows that α(α + β) = β(β + γ); in
the same way, we have β(β + γ) = γ(γ + α).
Now, let k = α2 + αβ = β 2 + βγ = γ 2 + γα. Then, recalling that α + β + γ = 1,
we successively obtain

k = kβ + kγ + kα = α2 β + αβ 2 + β 2 γ + βγ 2 + γ 2 α + γα2

and
k = kγ + kα + kβ = α2 γ + αβγ + β 2 α + αβγ + γ 2 β + αβγ.
As a result, we have
p
α2 β + β 2 γ + γ 2 α = 3αβγ = 3 3 α2 β · β 2 γ · γ 2 α.

From the case of equality in AM-GM, this means that α2 β = β 2 γ = γ 2 α. This

easily yields α3 = αβγ = β 3 = γ 3 and so α = β = γ. Thus, P is the centroid of
∆ABC.

OC192. Find all possible values of a positive integer n for which the expression
Sn = xn + y n + z n is constant for all real x, y, z with xyz = 1 and x + y + z = 0.
Originally problem 2 from the 2013 Spain Mathematical Olympiad.
We received two correct submissions. We present the solution by the Missouri
State University Problem Solving Group.
We claim that the only solutions are n = 1 and n = 3 (and n = 0 if non-negative
integers are allowed). Let σ1 = x + y + z, σ2 = xy + xz + yz, and σ3 = xyz (the
elementary symmetric functions in x, y, z). It is straightforward to verify that for
any x, y, z,
Sn = σ1 Sn−1 − σ2 Sn−2 + σ3 Sn−3 .
In our case, this gives

Sn = −σ2 Sn−2 + Sn−3 . (1)

Copyright c Canadian Mathematical Society, 2016

380/ THE OLYMPIAD CORNER

We have S0 = 3, S1 = σ1 = 0, and S2 = σ12 − 2σ2 = −2σ2 . By recurrence (1), we

have S3 = −σ2 S1 + S0 = 0 + 3 = 3.
We claim that for n > 3, Sn is √
never constant. We
√ first note that σ2 is not constant.
The values x = −1, y = (1 + 5)/2, z = (1 − 5)/2 satisfy the conditions √ of the
problem√ and here σ 2 = −2. On the other hand, x = 2, y = (−2 + 2)/2, z =
(−2 − 2)/2 also satisfy the conditions, but σ2 = −7/2. Since σ2 is not constant,
the following lemma will suffice to prove our claim.
Lemma. For k ≥ 1, S2k is a polynomial of degree k in σ2 with leading coefficient
(−1)k · 2 and S2k+1 is a polynomial of degree k − 1 in σ2 with leading coefficient
(−1)k−1 (2k + 1).
Proof. We have S2 = −2σ2 and S3 = 3, so the result follows when k = 1.
Assuming the result holds for all k < N , we have

S2N = −σ2 S2N −2 + S2N −3 .

Since by the induction hypothesis, S2N −2 is polynomial in σ2 of degree N − 1 with

leading coefficient (−1)N −1 ·2 and S2N −3 is a polynomial in σ2 of degree N −2, S2N
is a polynomial of degree N with leading coefficient (−1)(−1)N −1 · 2 = (−1)N · 2.
Similarly,
S2N +1 = −σ2 S2N −1 + S2N −2
with −σ2 S2N −1 being a polynomial of degree N − 1 with leading coefficient

−(−1)N −2 (2N − 1) = (−1)N −1 (2N − 1)

and S2N −2 is a polynomial of degree N − 1 with leading coefficient (−1)N −1 · 2,

so their sum is a polynomial of degree N − 1 with leading coefficient

(−1)N −1 (2N − 1) + (−1)N −1 · 2 = (−1)N −1 (2N + 1).

OC193. Let {an } be a positive integer sequence such that ai+2 = ai+1 + ai
for all i ≥ 1. For positive integer n, define {bn } as
4n−2
1 X
bn = ai .
a2n+1 i=1

Prove that bn is a positive integer, and find the general form of bn .

Originally problem 4 from day 1 of the 2013 Korea National Olympiad.
We present the solution by Ángel Plaza. There were no other submissions.
From the definition of {an } it is deduced that an = Fn−2 a1 + Fn−1 a2 for n > 2
where Fn is the n-th Fibonacci number beginning with F1 = F2 = 1. This fact
can be proved easily by induction.

Crux Mathematicorum, Vol. 41(9), November 2015

 THE OLYMPIAD CORNER /381

Then, using well known facts about Fibonacci numbers,

4n−2 4n−4
! 4n−3
!
X X X
ai = 1+ Fi a1 + Fi a2 = F4n−2 a1 + (F4n−1 − 1)a2 .
i=1 i=1 i=1

On the other hand, a2n+1 = F2n−1 a1 + F2n a2 . Finally, since L2n−1 F2n−1 = F4n−2
and L2n−1 F2n = F4n−1 − 1, if follows that bn = L2n−1 , where Ln is the n-th Lucas
number beginning with L1 = 2 and L2 = 1 and the problem is done.

OC194. Let Q+ be the set of all positive rational numbers. Let f : Q+ → R

be a function satisfying the following three conditions:

1. for all x, y ∈ Q+ , f (x)f (y) ≥ f (xy);

2. for all x, y ∈ Q+ , f (x + y) ≥ f (x) + f (y) ;

3. there exists a rational number a > 1 such that f (a) = a.

Prove that f (x) = x for all x ∈ Q+ .

Originally problem 5 from day 2 of the 2013 International Mathematical Olympiad.

There were no submitted solutions.

OC195. Let O denote the circumcentre of an acute-angled triangle ABC. Let

point P on side AB be such that ∠BOP = ∠ABC, and let point Q on side AC
be such that ∠COQ = ∠ACB. Prove that the reflection of BC in the line P Q is
tangent to the circumcircle of triangle AP Q.

Originally problem 5 from the 2013 Canadian Mathematical Olympiad.

We received two correct submissions. We present the solution by Michel Bataille.

Let Γ and Γ0 be the circumcircles of ∆ABC and ∆AP Q, respectively. Let O0 be

the centre of Γ0 and let α = ∠BAC, β = ∠CBA and γ = ∠BCA. Since ∆ABC is
acute-angled, ∠BOC = 2α and so

∠P OQ = 360◦ − 2α − β − γ = 180◦ − α.

It follows that O lies on Γ0 .

Let A0 be the second point of intersection of Γ and Γ0 . From Focus On... No 12,
the spiral similarity with centre A0 transforming Γ into Γ0 transforms B into P and
C into Q and so ∠(BC, P Q) = ∠(A0 O, A0 O0 ). (Here and in what follows, ∠(`, `0 )
denotes the directed angle of lines from the line ` to the line `0 and we suppose that
the orientation is such that ∠(AB, AC) = α, ∠(BC, BA) = β, ∠(CA, CB) = γ.)

Copyright c Canadian Mathematical Society, 2016

382/ THE OLYMPIAD CORNER

Assume that A0 O ⊥ P Q has been proved. Then, since the tangent t to Γ0 at

A0 is perpendicular to A0 O0 , we have ∠(A0 O, A0 O0 ) = ∠(P Q, t). It follows that
∠(BC, P Q) = ∠(P Q, t), hence the reflection of BC in P Q is t. Therefore, it is
sufficient to prove the assumption A0 O ⊥ P Q.
We observe that OB = OA and ∠(BC, BO) = 90◦ − α, hence

∠(BO, BA) = ∠(AB, AO) = ∠(AP, AO) = β − (90◦ − α) = 90◦ − γ.

Similarly,
∠(CA, CO) = ∠(CQ, CO) = ∠(AO, AQ) = 90◦ − β.
Since O, A, A0 and Q are concyclic, we obtain

∠(A0 O, A0 Q) = ∠(AO, AQ) = ∠(CQ, CO).

Thus, the circle Γ0 is the locus of all points M such that ∠(M O, M Q) = 90◦ −β and
since ∠(CO, CQ) = −∠(AO, AQ), C is on the reflection of Γ0 in the line OQ. Since
Γ is its own reflection in its diameter OQ, the reflection of C in OQ is on Γ and Γ0 ,
hence is A0 (not A, unless CQ ⊥ OQ but then γ = ∠COQ = 90◦ − ∠OCQ = β,
hence AB = AC, in which case it is easily seen that A = A0 ). As a result,
∠(OQ, OA0 ) = ∠(OC, OQ) = γ and so

∠(P Q, A0 O) = ∠(QP, QO) + ∠(OQ, OA0 )

= ∠(AP, AO) + ∠(OC, OQ)
= 90◦ − γ + γ = 90◦ .

and we are done.

Crux Mathematicorum, Vol. 41(9), November 2015

 BOOK REVIEWS /383

BOOK REVIEWS
Robert Bilinski
As a first for my present tenure as book reviewer for Crux, I present multiple
books in this column. The two books are radically different, one appealing to the
brain and the other the heart. What links them is that I see both as possible gifts.
Trigonometry: A clever study guide by James Tanton
ISBN 978-088385-836-3, 212 pages
Published by MAA Press: Problem book series, 2015

James Tanton is an education consultant, who used

to teach in high school and university. He worked on
several outreach programs for the MAA and serves
as an ambassador for the MAA in Washington D.C.
This is his third book for the MAA.
Trigonometry problems in the book are split in 20
categories, each occupying its own chapter, each fol-
lowing the same pattern : a list of Common Core
State Standards covered in the chapter, a brief syn-
opsis of the chapter with, where possible, some his-
torical notes, a simple calculation example, a worked
out example with comments and, in most cases, a se-
ries of problems to solve. All the proposed problems
come from either the American Mathematics Com-
petitions (AMC) or American High School Math-
ematics Examination (AHSME) contests ranging from 1952 to 2013, but most
coming from the 1980–2010 period. The 20 chapters take up about three quarters
of the book and cover mostly classical sections of a standard trigonometry course
from Chapter 1 “The Pythagorean Theorem” to Chapter 17 “Heron’s formula”.
The last three chapters stand out for several reasons : their originality and the
fact they do not come with practise problems. This seems to be at odds with
the aim of the book, which is to train the reader for math competitions, but they
do round out the knowledge a true mathlete should have about the subject. The
first of these is “Fitting trigonometric functions to periodic data”. In my opinion,
the name of the chapter is misleading as you might think that this is a chap-
ter on statistics, namely trigonometric regression, and you would be wrong. The
following non-trivial and fun problem illustrates the chapter well:
For each integer n > 1, let F (n) be the number of solutions of the
equation sin x = sin nx on the interval [0, π]. What is 2007
P
n=2 F (n)?
(#24, AMC 12A, 2007)
Not an easy cookie to crack, but far from the only interesting math morsel in the
book! Sadly, there are no practise problems of the same kind to hone one’s skills.
The last two chapters, “Polar Coordinates” and “Polar Graphs”, are even sparser,

Copyright c Canadian Mathematical Society, 2016

384/ BOOK REVIEWS

but they are clearly labelled as extra chapters. After all, they do not correspond
to any Common Core State Standards, but maybe, that is the mistake.
The book is a small paperback and can easily be carried everywhere, even to that
coffeeshop around the corner where you can dabble in math problems while sipping
a latte. The book is abundantly illustrated and the problems are definitely worthy
of the attention of Crux readers. Here are two more:
All three vertices of an equilateral triangle lie on a parabola y = x2 ,
and one of its sides has a slope of 2. The x-coordinates of the three
vertices have a sum of m/n, where m and n are relatively prime positive
integers. What is the value of m + n? (#24, AMC 12B, 2005)
Inscribed in a circle is a quadrilateral having sides of lengths 25, 39, 52
and 60 taken consecutively. What is the diameter of this circle? (#25,
AHSME, 1972)
The book is clearly not a standalone resource on the subject and has no real
introductory level exercises. The reader must have a basic comprehension of basic
trigonometry and can use this book as a further study guide. An interesting feature
is the presence of links to webpages, although I must admit I did not explore them.
The last 50 pages of the book are filled with complete solutions to the problems,
which are labelled from 1 to 100 in the order they appear and independent of
the chapter they are in. This provides for an easy way to find the solutions. A
little warning : the solutions proposed in the book are bareboned. But if you are
working on a solution, it should be enough to see if you are right or help you find
your mistake if you took the same approach. At the end, the author also presents
a 10-step strategy to conquer math contest problems, though I do not really see
its use since the book itself is aimed at problem solvers who should already have
developed and practiced these strategies.
This book has a well defined readership, namely all prospective mathletes and
various honours students who need extra stimulation to keep them interested.
Crux readers and problem solvers of all ages and strengths will also find this book
interesting. I hear a lot of people reminisce about the good old days when they
had time to play around with math, especially trigonometry. Maybe this book
might make a good gift for that nostalgic engineer friend of yours? In any case,
good reading!

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

Patterns of the universe : a coloring adventure in math and beauty by Alex Bellos
and Edmund Harriss
ISBN 978-1-61519-323-3, 144 pages
Published by The experiment publishing, 2015
Alex Bellos wears many hats : he is math and puzzles blogger for The Guardian,
he is a science presenter on the BBC and he has written a few mathematics general
interest books. Edmund Harriss is a mathematics professor at the University of

Crux Mathematicorum, Vol. 41(9), November 2015

 BOOK REVIEWS /385

Arkansas who actively participated in many math outreach programs. Together,

they created the most mathematical colouring I’ve ever seen.
This book is made to stimulate artistic beauty and
creativity while subtly enticing mathematical cu-
riosity. Each drawing is accompanied by a small
description of the the mathematical object to be
drawn. The book is separated into 2 sections, which
are further divided into subcategories : coloring
(Voronoi diagrams, transformations, fractals, peri-
odic tiling, non-periodic tiling, knots, mechanical
curves, polyhedra, proportions, space-filling curves,
. . .) and creating (randomness, latin squares, cellu-
lar automata, . . .).
Naturally, any artistically inclined mathematician would be interested in such a
book. It would make a good gift for artists who say they were “never good at
math”. Maybe, they would be surprised how much of mathematics has been
discovered by artists or how much good artistic representations helped advance
mathematics. Good reading!

Colour the above Voronoi diagram (image is from Patterns of the universe : a
coloring adventure in math and beauty courtesy of Amazon).

Copyright c Canadian Mathematical Society, 2016

386/ FOCUS ON... SOLUTIONS TO EXERCISES FROM FOCUS ON No. 12–16

FOCUS ON...
No. 19
Michel Bataille
Solutions to Exercises from Focus On... No. 12 – 16

From Focus On... No. 12

Let ABCD be a quadrilateral such that AD and BC intersect at E. Suppose that

−→ −→ −→ −→ −−→ −−→
ID = IC, JA = JC, KB = KD and ∠(ID, IC) = ∠(JA, JC) = ∠(KD, KB) =
−→ −−→
∠(EA, EB). Show that E, I, J, K are collinear.

If E = I = J, we are done, so we may suppose that E is distinct from I or J,

say E 6= I. From the hypothesis about the angles of the figure, E lies on the
circles (DIC), (AJC) and (BKD). Let σ be the spiral similarity with centre
D transforming the circle (DKB) into the circle (DIC). Since these two circles
intersect at D and E and B, C, E are collinear, we have σ(B) = C.
−→ −→
Now, ∆BKD and ∆CID are isosceles at K and I, respectively, and ∠(ID, IC) =
−−→ −−→ −−→ −−→ −−→ −→ −−→ −→ −−→ −−→
∠(KD, KB), hence ∠(DB, DK) = ∠(DC, DI) and so ∠(DK, DI) = ∠(DB, DC).
Since K is on (DKB) and I is on (DIC), it follows that σ(K) = I.

Thus, E, K, I are collinear. In a similar way, we obtain that E, J, I are collinear

and conclude that J, K are on the line EI.

Crux Mathematicorum, Vol. 41(9), November 2015

 MICHEL BATAILLE /387

From Focus On... No. 13

(a) Let the angle bisectors of triangle ABC meet its sides at D, E, F . Show that
∆DEF is right-angled if and only if one of the angles of ∆ABC equals 120◦ .
We shall use the familiar notations for the sides and the angles of ∆ABC and
we suppose that D, E, F are on sides BC, CA, AB, respectively. We show that
∆DEF is right-angled at D if and only if A = 120◦ .
We know that (b + c)D = bB + cC, (c + a)E = aA + cC, (a + b)F = aA + bB,
and deduce the following vectorial relations:
−−→ −−→ −→
(a + b)(b + c)F D = b(a − c)AB + c(a + b)AC
−−→ −−→ −→
(a + c)(b + c)ED = b(a + c)AB + c(a − b)AC.
−−→ −−→
Now, ∆DEF is right-angled at D if and only if F D · ED = 0, which is successively
equivalent to
−−→ −→
b2 c2 (a2 − c2 ) + b2 c2 (a2 − b2 ) + bc(AB · AC)((a − c)(a − b) + (a + c)(a + b)) = 0,
2a2 − b2 − c2 + 2(a2 + bc) cos A = 0,
1 + 2 cos A = 0 (since b2 + c2 = a2 + 2bc cos A),
A = 120◦ .

(b) We have n distinct points A1 , . . . , An in the plane. To each point Ai a real

number λi 6= 0 is assigned, in such a way that Ai A2j = λi + λj for all i, j with
i 6= j. Show that n ≤ 4 and that λ11 + λ12 + λ13 + λ14 = 0 if n = 4.
Assume that n ≥ 5. For distinct i, j, k we have
−−−→ −−−→ −−−→ −−−→
λj + λk = Aj A2k = (Aj Ai + Ai Ak )2 = Aj A2i + Ai A2k + 2Aj Ai · Ai Ak
−−−→ −−−→
= λj + λi + λi + λk + 2Aj Ai · Ai Ak
−−−→ −−−→
and so λi = Ai Aj · Ai Ak .
−−−→ −−−→ −−−→ −−−→ −−−→ −−−→
From λ1 = A1 A2 · A1 A3 = A1 A2 · A1 A4 = A1 A2 · A1 A5 , we deduce (by difference)
−−−→ −−−→ −−−→ −−−→
A1 A2 · A4 A3 = A1 A2 · A5 A3 = 0.

Therefore the points A3 , A4 , A5 lie on a perpendicular to the line A1 A2 . In the

same way, A1 , A2 , A3 must lie on a perpendicular to A4 A5 and consequently A3
must be the point of intersection of the lines A1 A2 and A4 A5 . However, this
implies A1 A24 = A1 A23 + A3 A24 , which leads to λ3 = 0, in contradiction with the
hypothesis. As a result, we must have n ≤ 4.
From now on, we suppose that n = 4. As in the first part, we see that A2 A3 ⊥
A1 A4 , A3 A4 ⊥ A1 A2 , A1 A3 ⊥ A2 A4 and deduce that A1 A2 A3 A4 is formed by
the vertices of a triangle together with its orthocenter.

Copyright c Canadian Mathematical Society, 2016

388/ FOCUS ON... SOLUTIONS TO EXERCISES FROM FOCUS ON No. 12–16

−−−→ −−→
Let H be the point of intersection of A3 A4 and A1 A2 . Then, λ1 = A1 A2 · A1 H and
−−−→ −−→ −−→ −−−→
λ2 = A1 A2 · HA2 (since HA3 is orthogonal to A1 A2 ). Since A1 , H, A2 are collinear,
−−→ −−→ −−→ −−→
this yields λ1 λ2 = A1 A22 (HA2 · A1 H). Similarly, we find λ3 λ4 = A3 A24 (HA4 · A3 H)
and so
−−→ −−→ −−→ −−→
λ1 λ2 (λ3 + λ4 ) + λ3 λ4 (λ1 + λ2 ) = A1 A22 .A3 A24 (HA2 · A1 H + HA4 · A3 H). (1)

We may consider A4 as the orthocenter of ∆A1 A2 A3 and we know that its reflection
A04 in A1 A2 is on the circumcircle Γ of ∆A1 A2 A3 . It follows that the power of H
−−→ −−→ −−→ −−→ −−→ −−→
with respect to Γ is HA2 · HA1 as well as HA3 · HA04 = −HA3 · HA4 . From (1),
λ1 λ2 (λ3 + λ4 ) + λ3 λ4 (λ1 + λ2 ) = 0, and the desired equality is derived at once.

From Focus On... No. 15

1. Show that for each integer n ≥ 2
Ç å
X n (−1)k−1 Å
n−1 ã n−1
k n X 1
1− = .
k k n k+1
k=1 k=1

With the help of the binomial theorem, we rewrite the left-hand side Ln as
n Ç å n−1 Ç å
X (−1)j n X (−1)k−1
j n
Ln = aj , where aj = ·k .
j=0
nj j k k
k=1
Pn 1
Let Hn = k=1 k denote the nth harmonic number. Recalling the well-known
Pn (−1)k−1 n

equality Hn = k=1 k k (easily proved by induction), we see that
n
(−1)
a 0 = Hn + .
n
Then, we calculate a1 as follows:
n−1 Ç å n Ç å !
k−1 n k n
X X
n
a1 = (−1) =− (−1) − 1 − (−1)
k k
k=1 k=0
= −(0 − 1 − (−1)n ) = 1 + (−1)n .

Crux Mathematicorum, Vol. 41(9), November 2015

 MICHEL BATAILLE /389

Lastly, if 2 ≤ j ≤ n, we obtain
n−1 Ç å Ç Ç å å
n−k n n j−1
X
n+1 j−1 n+1
aj = (−1) (−1) k = (−1) S(n, j − 1) − n
k n
k=1
= (−1)n nj−1 .

(Note that S(n, j − 1) = 0 since 1 ≤ j − 1 < n.)

Back to Ln , all this leads to
n Ç å
1 X (−1)j n
Ln = a0 − · n · a1 + (−1)n nj−1
n j=2
nj j
(−1)n (−1)n
= Hn + − 1 − (−1)n + (−1 + n).
n n
Thus
n−1
X 1
Ln = Hn − 1 = .
k+1
k=1

2. For nonnegative integer n, evaluate in closed form

n
X (−1)k (n + k + 2)!
· .
(k!)2 (n − k)!
k=0

(−1)n
Let Sn be the given sum. We show that Sn = 2 · (n + 1)2 (n + 2)2 .
(−1)n Sn
Consider Tn = (n+1)(n+2) . We calculate:

(−1)n (Sn )n!

Tn =
(n + 2)!
n Ç åÇ å
n−k n n+k+2
X
= (−1)
k n+2
k=0
n Ç å
1 X
n−k n
= (−1) (k + n + 2)(k + n + 1) · · · (k + 2)(k + 1)
(n + 2)! k
k=0
n Ç å
1 X
n−k n
= (−1) (k n+2 + ak n+1 + bk n + P (k))
(n + 2)! k
k=0

(n+2)(n+3)
where P (x) is a polynomial whose degree is less than n, a = n+2
P
j=1 j = 2
and
Ñ !2 n+2 é
n+2
X 1 X X (n + 1)(n + 2)(n + 3)(3n + 8)
b= i·j = j − j2 =
2 j=1 j=1
24
1≤i<j≤n+2

Copyright c Canadian Mathematical Society, 2016

390/ FOCUS ON... SOLUTIONS TO EXERCISES FROM FOCUS ON No. 12–16

Pn+2 (n+2)(n+3)(2n+5)
using j=1 j2 = 6 .
It follows that
1
Tn = (S(n, n + 2) + aS(n, n + 1) + bS(n, n)) .
(n + 2)!
n(n+1)!
Now, S(n, n + 1) = 2 (as seen in the number) and in a similar way, using

x x2 n xn+2 n(n − 1) xn+2

xn (1 + + + · · · )n = xn + xn+1 + n · + · + ··· ,
2 6 2 6 2 4
n(3n+1)(n+2)!
we readily find S(n, n + 2) = 24 .
Finally, Tn is equal to
Å
1 n(3n + 1)(n + 2)! (n + 2)(n + 3) n(n + 1)!
+ ·
(n + 2)! 24 2 2
ã
(n + 1)(n + 2)(n + 3)(3n + 8)
+ · (n!)
24
that is, to
(n + 1)(n + 2)
2
and the announced value of Sn follows.

From Focus On... No. 16

1. Let P be an arbitrary point in the plane of a triangle ABC with sidelengths
a, b, c. Prove that
a2 + b2 + c2
P A2 + P B 2 + P C 2 ≥ .
3

Let G be the centroid of ∆ABC. Leibniz’s relation gives

P A2 + P B 2 + P C 2 = 3P G2 + GA2 + GB 2 + GC 2 . (2)

Now, if ma , mb , mc denote the medians from A, B, C, respectively, we have

4
GA2 + GB 2 + GC 2 = · (m2a + m2b + m2c )
9
1
= · (2b2 + 2c2 − a2 + 2c2 + 2a2 − b2 + 2a2 + 2b2 − c2 )
9
a2 + b2 + c2
= .
3
a2 +b2 +c2
From (2), we then deduce P A2 + P B 2 + P C 2 ≥ 3 (with equality if and
only if P = G).

Crux Mathematicorum, Vol. 41(9), November 2015

 MICHEL BATAILLE /391

2. Let A, B, C, D be four points on a line ` in this order and let M not on ` be

such that ∠AM B = ∠CM D. Prove that
M A2 AB M B2
> > .
M C2 CD M D2

Let us apply Stewart’s relation to points A, B, D and M :

AB · M D2 + BD · M A2 + DA · M B 2 + AB · BD · DA = 0.

It follows that AB · M D2 + BD · M A2 − AD · M B 2 = AB · BD · AD > 0, or, since

AD = AC + CD,

(AB · M D2 − CD · M B 2 ) + (BD · M A2 − AC · M B 2 ) > 0. (3)

2
MB AB 2 2
The inequality M D 2 < CD clearly holds if AB · M D − CD · M B > 0. But
otherwise, from (3) we must have

BD · M A2 − AC · M B 2 > 0 (4)

However, since ∠AM B = ∠CM D, we have

Å ã
MA · MB AB area(M AB) MA · MC AC
= = and =
MC · MD CD area(M CD) MB · MD BD
AB·AC
hence M A2 = M D2 · CD·BD . Inequality (4) now yields

AB · AC
M D2 · − AC · M B 2 > 0
CD
M B2 AB
and again M D2 < CD .

Further, since
MA · MB MB · MD AB BD
· = · ,
MC · MD MA · MC CD AC
we have
M B2 AB · BD
=
M C2 CD · AC
CD
and the inequality M D2 > M B 2 · AB already obtained gives

CD · BD CD AB · BD
M A2 · > · · M C 2.
AB · AC AB CD · AC
AB
The desired inequality M A2 > M C 2 · CD follows.
Alternately, the latter inequality can be obtained from Stewart’s relation applied
to A, C, D and M in the same way as above. The details are left to the reader.

Copyright c Canadian Mathematical Society, 2016

392/ A MATHEMATICAL PERFORMANCE

A Mathematical Performance
Hee-Joo Nam, Giavanna Valacco and Ling-Feng Zhu
Two of us girls were members of the S.M.A.R.T. Circle in Edmonton. The acronym
stands for Saturday Mathematical Activities, Recreations & Tutorials. In 2010, the
Circle sent two teams of four to the Junior High Division of the International Math-
ematics Competition (IMC). Five Circle members, reinforced by three students
elsewhere from Alberta, went to the host city Iuncheon for the event. Hee-Joo,
the third girl, came from Calgary. She was particularly excited since she was orig-
inally from South Korea. We had the highest percentage of female representation
in the history of the competition, except that Iran, not fully understanding the
rules in their first year of participation, sent nine all-girl teams!

The individual and team contest papers are given in the second part of this article,
which will appear in Volume 41, issue 10 of Crux. For further details, see their
website http://www.imc-official.org/en\_US/; for solutions, see the book [1].
Our lack of international experience showed in that Giavanna was the only one
who managed a Bronze Medal in the individual contest.

Apart from academic matters, the I.M.C. emphasized international friendship and
understanding. The most wonderful feature was the Cultural Evening, when teams
from various countries gave short performances. These are usually in the form of
songs and dances, but there had been innovative presentations as well.

The activity we chose was Platonic Metamorphosis. We used six coloured strings
to construct the skeleton of each of the five Platonic solids, in six steps in a
continuous transformation. The clever idea came from Karl Schaffer. See his
paper [2]. However, we made our own adaptation. Ten students were required,
preferably in a six-to-four gender mix. Fortunately, Giavanna’s younger brother
and sister came along on a family holiday, and we had the perfect combination.

The official languages for the I.M.C. were English and Chinese, along with Korean
for 2010. During our performace, narration was done by Giavanna in English, Ling-
Feng in Chinese and Hee-Joo in Korean. This was extremely well-received by the
audience.

Step 1. Construction of the Tetrahedron

Start off with the four girls identified as N(orth), S(outh), E(ast) and W(est). Each
designates one hand as the U(pper) hand and the other hand as the L(ower) hand.
N and S hold out their U hands while E and W hold out their L hands. String 1
is held between UN and LW, string 2 between UN and LE, string 3 between LW
and LE, string 4 between LW and US, string 5 between LE and US, and string 6
between UN and US. The completed tetrahedron is shown in Figure 1, with string
6 drawn in such a way to facilitate the description of the next step.

Crux Mathematicorum, Vol. 41(9), November 2015

 HEE-JOO NAM, GIAVANNA VALACCO AND LING-FENG ZHU /393

Figure 1

Step 2. Transformation into the Cube

Figure 2
Each of the four girls holds out the other hand and places it at the center of one
of the four faces of the tetrahedron, as shown on the left side of Figure 2. Each
of these hands will grab the three sides of the triangular face. The end result is a
cube, as shown on the right side of Figure 2. Each string forms a face of the cube.

Step 3. Transformation into the Dodecahedron

First, the cube is redrawn as shown in Figure 3.

Figure 3

Copyright c Canadian Mathematical Society, 2016

394/ A MATHEMATICAL PERFORMANCE

Now the six boys enter the picture. They are identified as T(op face), B(ottom
face), H (northwest face), I (southwest face), J (southeast face) and K (northeast
face). Each of them holds out both hands and places them symmetrically about
the center of the assigned face of the cube. The line segment joining the two hands
of each student is parallel to a side of the cube, and the segments on adjacent faces
are perpendicular to each other.
Each pair of these hands will grab the two sides of the square face parallel to the
segment they form. Each hand will also grab the nearer one of the remaining two
sides of the square face. It should be emphasized that while each face of
the cube is formed of one string, no part of this string is to be grabbed
by the hands assigned to this face. Instead, the other four strings joining
adjacent pairs of vertices of the face are grabbed, as illustrated in Figure 4.

Figure 4

The end result is a dodecahedron, as shown in Figure 5. Failure to observe the

caution in the preceding paragraph will still produce a dodecahedron, but the
whole structure will then fall apart in Step 4.

Figure 5

Crux Mathematicorum, Vol. 41(9), November 2015

 HEE-JOO NAM, GIAVANNA VALACCO AND LING-FENG ZHU /395

Step 4. Transformation into the Icosahedron

The four girls let go of their strings. The end result is an icosahedron, as shown
in Figure 6.

Figure 6

Step 5. Transformation into the Octahedron

Each of the six boys slides both hands together. The end result is an octahedron, as
shown in Figure 7. Two strings which are opposite sides of the original tetrahedron
now form the same square cross-section of the octahedron.

Figure 7

Copyright c Canadian Mathematical Society, 2016

396/ A MATHEMATICAL PERFORMANCE

Step 6. Return to the Tetrahedron

The four girls N, S, E and W re-enter the picture. N puts the U hand in triangle
HKT (north and top), S puts the U hand in triangle IJT (south and top), E puts
the L hand in triangle JKB (bottom and east) and W puts the L hand in triangle
HIB (bottom and west). This is shown in Figure 8.

Figure 8

Each hands grabs the three strings it originally holds, and then the six boys let
go of theirs. The end result is once again a tetrahedron, as shown in Figure 9.

Figure 9
Bibliography:
[1] Wen-Hsien Sun, Huan Zheng and Hua-Wei Zhu, An Indepth Study of the
International Mathematics Competition, Chiu Chang Math. Publ., Taipei,
(2014) 204–211, 325–333.
[2] Karl Schaffer, A Platonic sextet in strings, College Mathematics Journal, 43
(2012) 64–69.

Crux Mathematicorum, Vol. 41(9), November 2015

 PROBLEMS /397

PROBLEMS
Readers are invited to submit solutions, comments and generalizations to any problem in
this section. Moreover, readers are encouraged to submit problem proposals. Please see
submission guidelines inside the back cover or online.
To facilitate their consideration, solutions should be received by the editor by November
1, 2016, although late solutions will also be considered until a solution is published.
The editor thanks André Ladouceur, Ottawa, ON, for translations of the problems.
An asterisk (?) after a number indicates that a problem was proposed without a solution.

4081. Proposed by Daniel Sitaru.

Determine all A, B ∈ M2 (R) such that:
 Ç å
2 2 22 44
A + B = 14 28 ,



Ç å
 10 20
AB + BA = .


2 4

4082. Proposed by D. M. Bătineţu-Giurgiu and Neculai Stanciu.

Let ABC be a right-angle triangle with ∠A = 90◦ and BC = a, AC = b and AB =
c. Consider the Fibonacci sequence Fn with F0 = F1 = 1 and Fn+2 = Fn+1 + Fn
for all non-negative integers n. Prove that
2
Fm Fn2 Fp2 3
+ + ≥ 2
(bFn + cFp )2 (bFp + cFm )2 (bFm + cFn )2 2a
or all non-negative integers m, n, p.

4083. Proposed by Ovidiu Furdui.

Calculate Z n
1 x
lim √ dx.
n→∞ n n 0 1 + n cos2 x

4084. Proposed by Michel Bataille.

In the plane, let Γ be a circle and A, B be two points not on Γ. Determine when
MA
M B is not independent of M on Γ and, in these cases, construct with ruler and
compass I and S on Γ such that
ß ™ ß ™
IA MA SA MA
= inf :M ∈Γ and = sup :M ∈Γ .
IB MB SB MB

Copyright c Canadian Mathematical Society, 2016

398/ PROBLEMS

4085. Proposed by José Luis Dı́az-Barrero. Correction.

Let ABC be an acute triangle. Prove that
» » » √
4 4 4 3 2
sin(cos A) · cos B + sin(cos B) · cos C + sin(cos C) · cos A < .
2

4086. Proposed by Daniel Sitaru.

Let be f : [0, 1] → R; f twice differentiable on [0, 1] and f 00 (x) < 0 for all x ∈ [0, 1].
Prove that Z 1 Z 1
25 f (x)dx ≥ 16 f (x)dx + 4f (1).
1
5 0

4087. Proposed by Lorian Saceanu.

If S is the area of triangle ABC, prove that

ma (b + c) + 2m2a ≥ 4S sin A,

where b and c are the lengths of sides that meet in vertex A, and ma is the length
of the median from that vertex; furthermore, equality holds if and only if b = c
and ∠A = 120◦ .

4088. Proposed by Ardak Mirzakhmedov.

Let a, b and c be positive real numbers such that a2 b + b2 c + c2 a + a2 b2 c2 = 4.
Prove that
a2 + b2 + c2 + abc(a + b + c) ≥ 2(ab + bc + ca).

4089. Proposed by Daniel Sitaru and Leonard Giugiuc.

Let a, b, c and d be real numbers with 0 < a < b < c < d. Prove that

b c d d
+ + > 3 + ln .
a b c a

4090. Proposed by Nermin Hodz̆ić and Salem Malikić.

Let a, b and c be non-negative real numbers such that a2 + b2 + c2 = 3. Prove that

a b c 3
+ + ≥ .
3b2 + 6c − bc 3c2 + 6a − ca 3a2 + 6b − ab 8

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

Crux Mathematicorum, Vol. 41(9), November 2015

 PROBLEMS /399

4081. Proposé par Daniel Sitaru.

Déterminer toutes les matrices A, B ∈ M2 (R) telles que:
 Ç å
2 2 22 44
A + B = ,


 14 28
Ç å
 10 20
AB + BA = .


2 4

4082. Proposé par D. M. Bătineţu-Giurgiu et Neculai Stanciu.

Soit ABC un triangle rectangle tel que ∠A = 90◦ , BC = a, AC = b et AB = c.
Soit Fn la suite de Fibonacci définie par F0 = F1 = 1 et Fn+2 = Fn+1 + Fn pour
tout entier non négatif n. Démontrer que
2
Fm Fn2 Fp2 3
2
+ 2
+ ≥ 2
(bFn + cFp ) (bFp + cFm ) (bFm + cFn )2 2a
pour tout entier non négatif m, n et p.

4083. Proposé par Ovidiu Furdui.

Évaluer Z n
1 x
lim √ dx.
n→∞ n n 0 1 + n cos2 x

4084. Proposé par Michel Bataille.

Dans le plan, soit un cercle Γ et deux points, A et B, qui ne sont pas sur Γ.
MA
Déterminer les conditions qui font que M B n’est pas indépendant d’un point M
sur Γ. De plus, construire avec compas et règle deux points I et S sur Γ tels que
ß ™ ß ™
IA MA SA MA
= inf :M ∈Γ et = sup :M ∈Γ .
IB MB SB MB

4085. Proposé par José Luis Dı́az-Barrero. Correction.

Soit ABC un triangle acutangle. Démontrer que
» » » √
4 4 4 3 2
sin(cos A) · cos B + sin(cos B) · cos C + sin(cos C) · cos A < .
2

4086. Proposé par Daniel Sitaru.

Soit une fonction f : [0, 1] → R qui est deux fois dérivable sur [0, 1] et telle que
f 00 (x) < 0 pour tout x ∈ [0, 1]. Démontrer que
Z 1 Z 1
25 f (x)dx ≥ 16 f (x)dx + 4f (1).
1
5 0

Copyright c Canadian Mathematical Society, 2016

400/ PROBLEMS

4087. Proposé par Lorian Saceanu.

Soit S l’aire d’un triangle ABC, b = AC, c = AB et ma la longueur de la médiane
issue de A. Démontrer que

ma (b + c) + 2m2a ≥ 4S sin A

et démontrer qu’il y a égalité si et seulement si b = c et ∠A = 120◦ .

4088. Proposé par Ardak Mirzakhmedov.

Soit a, b et c des réels strictement positifs tels que a2 b + b2 c + c2 a + a2 b2 c2 = 4.
Démontrer que

a2 + b2 + c2 + abc(a + b + c) ≥ 2(ab + bc + ca).

4089. Proposé par Daniel Sitaru et Leonard Giugiuc.

Soit a, b, c et d des réels tels que 0 < a < b < c < d. Démontrer que
b c d d
+ + > 3 + ln .
a b c a

4090. Proposé par Nermin Hodz̆ić et Salem Malikić.

Soit a, b et c des réels non négatifs tels que a2 + b2 + c2 = 3. Démontrer que
a b c 3
+ 2 + 2 ≥ .
3b2 + 6c − bc 3c + 6a − ca 3a + 6b − ab 8

Crux Mathematicorum, Vol. 41(9), November 2015

 SOLUTIONS /401

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 2014: 40(9), p. 391–394.

3981. Proposed by José Luis Dı́az-Barrero.

Let a, b, c be three positive numbers such that ab + bc + ca = 6abc. For all positive
integers n ≥ 2, show that
bc ca ab
+ n + n ≥ 3 · 2n−2 .
an (b+ c) b (c + a) c (a + b)

We received 16 submissions, of which 15 were correct and complete. We present

the solution by Wolfgang Remmel, slightly modified by the editor.
We make the substitutions x = a1 , y = 1b and z = 1c . Dividing both sides of
ab + bc + ac = 6abc by abc we get the condition x + y + z = 6. For the first
summand of the inequality we have
1
bc an xn
= b+c
= ;
an (b + c) bc
z+y

applying a similar treatment to the other two summands, the inequality we have
to prove becomes
xn yn zn
+ + ≥ 3 · 2n−2 .
z+y z+x x+y
To start with, by the Cauchy-Schwarz Inequality,
Ä n ä
z yn xn
2(x + y + z) · x+y + x+z + y+z
Å ã
√ √ √
 z n2 2  z n2 2  z n2 2
= ( x + y)2 + ( x + z)2 + ( y + z)2 √
y+x
+ √
y+x
+ √
y+x
n n n
2
≥ z +y +x
2 2 2 .

Using the fact that x + y + z = 6, we can divide both sides by 12 to get

zn yn xn 1 n n n
2
+ + ≥ · z2 + y2 + x2 . (1)
x+y x+z y+z 12
If n = 2 we are done. Else, by Hölder’s inequality,
n n n
n2 n−2
z2 + y2 + x2 · (1 + 1 + 1) n ≥ z + y + x = 6.

Raising both sides to the power n and then dividing by 3n−2 gives us
n n n
2
z 2 + y 2 + x 2 ≥ 2n · 32 . (2)

Copyright c Canadian Mathematical Society, 2016

402/ SOLUTIONS

Combining (1) and (2) we get

zn yn xn 1
+ + ≥ · (2n · 32 ) = 3 · 2n−2 ,
x+y x+z y+z 12
which is what we needed to prove. Note that equality holds if and only if

x = y = z = 2,

which in terms of the original variables is

1
a=b=c= .
2

3982. Proposed by Michel Bataille.

Let n ∈ N, u > 0 and for k = 0, 1, . . . , n − 1, let ak be such that 0 < ak ≤ sinh(u).
Prove that if x ≥ eu , then
xn
an−1 xn−1 − an−2 xn−2 + · · · + (−1)n−2 a1 x + (−1)n−1 a0 < .
2

There were 3 submitted solutions for this problem, all of which were correct. We
present the solution by Joel Schlosberg.
Since the natural exponential, natural logarithm, and hyperbolic sine functions
are strictly increasing, x ≥ eu > e0 = 1 and

eln x − e− ln x x − x−1
0 < ak ≤ sinh(u) ≤ sinh[ln(x)] = = .
2 2
By the formula for the sum of an infinite geometric series with common ratio
x−2 ∈ (0, 1),

an−1 xn−1 −an−2 xn−2 + · · · + (−1)n−2 a1 x + (−1)n−1 a0

x − x−1 n−1 x − x−1 n−3 x − x−1 n−5
< ·x + ·x + ·x + ···
2 2 2
x − x−1 xn−1 xn
= · −2
= .
2 1−x 2

Editor’s Comments. The other two solutions involve a little more effort; induction,
or a bit of case work between odd and even m. The weakness of the inequality
is shown in the featured solution. In fact, the solution shows that if we have a
sequence of positive ai , all of which are less than or equal to sinh(u), and x ≥ eu ,
then we can bound an infinite series:
∞
X xn
ai xn+1−2i ≤ .
i=1
2

Crux Mathematicorum, Vol. 41(9), November 2015

 SOLUTIONS /403

3983. Proposed by Marcel Chiriţă.

Find all differentiable functions f : R → R such that
xf (x) − yf (y) = (x2 − y 2 ) max (f 0 (x), f 0 (y))
for all real numbers x, y.
We received three solutions, all of which were correct and complete. We present
two solutions.
Solution 1, by Michel Bataille.
It is readily checked that the functions x 7→ kx (k ∈ R) are solutions. We show
that there are no others. To this aim, let f satisfy the conditions and let a 6= 0.
We show that f 0 (a) ≤ f 0 (0). Assume that f 0 (0) < f 0 (a). Let α ∈ (f 0 (0), f 0 (a)).
Since a derivative satisfies the Darboux property, we have α = f 0 (u) for some
u ∈ (0, a) or (a, 0).
From the condition xf (x) − yf (y) = (x2 − y 2 ) max(f 0 (x), f 0 (y)), we obtain
af (a) = a2 max(f 0 (a), f 0 (0)) = a2 f 0 (a)
and
uf (u) = u2 max(f 0 (u), f 0 (0)) = u2 f 0 (u);

that is, f (a) = af 0 (a) and f (u) = uf 0 (u) = αu. But we also get
a2 f 0 (a) − u2 f 0 (u) = af (a) − uf (u) = (a2 − u2 )f 0 (a)
and so f 0 (a) = f 0 (u) = α, a contradiction with α ∈ (f 0 (0), f 0 (a)). It follows that
f 0 (x) ≤ f 0 (0) for any real number x.
Now, the condition gives xf (x) = x2 f 0 (0) so that f (x) = kx if x 6= 0 where
k = f 0 (0). Since f is continuous (even differentiable),
f (0) = lim f (x) = lim kx = 0 = k · 0.
x→0 x→0

Finally, f (x) = kx for any real number x.

Solution 2, by Marcel Chiriţǎ.

Consider the differentiable function g : R → R such that g(x) = f (x) − f (0) −
xf 0 (0). Clearly g(0) = 0; we differentiate to find that g 0 (x) = f 0 (x) − f 0 (0), and
so g 0 (0) = 0. We can rewrite the functional equation in terms of g and simplify:
x[g(x) + f (0) + xf 0 (0)] − y[g(y) + f (0) + yf 0 (0)]
= (x2 − y 2 ) max(g 0 (x) + f 0 (0), g 0 (y) + f 0 (0))
xg(x) − yg(y) + f (0)(x − y) + (x2 − y 2 )f 0 (0)
= (x2 − y 2 ) max(g 0 (x) + f 0 (0), g 0 (y) + f 0 (0))
xg(x) − yg(y) + f (0)(x − y) = (x2 − y 2 ) max(g 0 (x), g 0 (y)).

Copyright c Canadian Mathematical Society, 2016

404/ SOLUTIONS

If y = 0, the above relation reduces to x(g) + xf (0) = x2 max(g 0 (x), 0), which
reduces further to g(x) + f (0) = x max(g 0 (x), 0) when x 6= 0. We take the limit of
both sides as x approaches zero:

lim (g(x) + f (0)) = lim x max(g 0 (x), 0),

x→0 x→0

yielding g(0) + f (0) = 0, hence f (0) = 0. We then have g(x) = f (x) − xf 0 (0) and
in particular g(x) = x max(g 0 (x), 0) for x 6= 0 .
If x ≥ 0, then g(x) ≥ 0 and g(x) ≥ xg 0 (x), or equivalently g(x) − xg 0 (x) ≥ 0.
On the other hand, if x ≤ 0, then g(x) ≤ 0 and g(x) ≤ xg 0 (x), or equivalently
g(x) − xg 0 (x) ≤ 0.
g(x)
Define the function h(x) = x , which is differentiable and non-negative for x 6= 0.
0 xg 0 (x)−g(x)
In fact, h (x) = x2 ,
which we just showed is non-negative for positive x
and non-positive for negative x. Moreover,

g(x)
lim h(x) = lim = lim g 0 (x) = 0.
x→0 x→0 x x→0

Therefore h(x) is non-negative and non-increasing for negative x; zero for x = 0;

and non-negative and non-decreasing for positive x. Clearly h(x) = 0 for all x ∈ R.
But then g(x) = f (x) − xf 0 (0) = 0 for all x ∈ R, from which the result follows
that f (x) = kx, k = f 0 (0).

3984. Proposed by Dragoljub Milošević.

Let ABC be any right-angled triangle with ∠C = 90◦ . Let wa be the length of
the internal bisector of ∠A from A to the side BC; define wb similarly. If [ABC]
is the area of ABC, prove that
√
wa wb ≥ 4[ABC](2 − 2).

We received 22 correct solutions. We present a composite of nearly identical proofs

given by Brian Beasley, Miguel Amengual Covas, Cristóbal Sánchez-Rubio, and
C. R. Pranesachar (done independently).
Clearly cos(A/2) = b/wa , cos(B/2) = a/wb , and [ABC] = ab/2. Hence,

ab 2ab
wa wb = =
cos A2 cos B2 cos( A+B
2 ) + cos( A−B
2 )
4[ABC] 4[ABC] 8[ABC]
= π A−B
≥ √ = √
cos 4 + cos 2 2 2+ 2
2 +1
√
= 4[ABC](2 − 2)

Equality occurs if and only if A = B, that is, when ∆ABC is an isosceles right-
angled triangle.

Crux Mathematicorum, Vol. 41(9), November 2015

 SOLUTIONS /405

3985. Proposed by Mihaela Berindeanu.

Prove that if a, b, c are positive numbers with sum of 18, then
a b c 1
+ + ≥ .
b2 + 36 c2 + 36 a2 + 36 4

We received 22 solutions of which 21 were correct and complete. We present two

solutions.
Solution 1, by AN-anduud Problem Solving Group.
Using AM-GM inequality and using inequality
1
ab + bc + ca ≤ (a + b + c)2 ,
3
we have
X a X a(b2 + 36) − ab2 X a 1 X ab2
= = −
cyc
b2 + 36 cyc
2
36(b + 36) cyc
36 36 cyc b2 + 36
1 1 X ab2
≥ − √
2 36 cyc 2 b2 · 36
1 1 1 X
= − · ab
2 36 12 cyc
1 1 1 1
≥ − · · (a + b + c)2
2 36 12 3
1 1 1
= − = .
2 4 4
Equality holds only when a = b = c = 6.

Solution 2, by Michel Bataille.

Setting a = 6x, b = 6y, c = 6z, we are led to prove that
x y z 3
+ + ≥ (1)
y 2 + 1 z 2 + 1 x2 + 1 2
whenever x, y, z ≥ 0 and x + y + z = 3.
Now, since y 2 + 1 ≥ 2y,

x xy 2 xy 2 xy
=x− 2 ≥x− =x− .
y2 +1 y +1 2y 2
y yz z zx
Similarly, z 2 +1 ≥y− 2 and x2 +1 ≥z− 2 and so

x y z xy + yz + zx xy + yz + zx
+ 2 + 2 ≥x+y+z− =3− . (2)
y2 +1 z +1 x +1 2 2

Copyright c Canadian Mathematical Society, 2016

406/ SOLUTIONS

But, 9 = (x + y + z)2 = x2 + y 2 + z 2 + 2(xy + yz + zx) ≥ 3(xy + yz + zx), hence

xy + yz + zx ≤ 3. Back to (2), we obtain

x y z 3 3
+ 2 + 2 ≥3− =
y2 +1 z +1 x +1 2 2

that is, (1).

Editor’s Comments. From Solution 2, it is clear that the given inequality is equiv-
alent to
If x, y, z are positive real numbers such that x + y + z = 3, prove that

x y z 3
+ + ≥ .
y 2 + 1 z 2 + 1 x2 + 1 2

This was observed by several solvers, and Arslanagić and Malikić also noticed
that this inequality is problem 2994 part (d) of Crux 31(8) proposed by Faruk
Zejnulahi and Šefket Arslanagić.
Dragojlub Milošević gave the following generalization:
If a, b, c are positive real numbers such that a + b + c = 3k with k > 0,
then
a b c 3
+ 2 + 2 ≥ .
b2 + k 2 c + k2 a + k2 2k

It is easy to see that if k = 1 we obtain problem 2994 and if k = 6 we obtain the

problem 3985.

3986. Proposed by George Apostolopoulos.

Let a, b, c be the lengths of the sides of a triangle ABC with circumradius R. Prove
that
1 1 1 1
+ + ≥ .
(a + b)(b + c) (b + c)(c + a) (c + a)(a + b) 4R2

There were 19 correct solutions submitted. We present the one below which is both
elegant and short. Solution by Joel Schlosberg.
It is well known (O. Bottema et al., Geometric Inequalities, Wolters-Noordhoff,
Groningen, 1969, p. 49) that
√
a + b + c ≤ 3 3R.

By the AM-GM Inequality we have

(a + b) + (b + c) + (c + a) »
≥ 3 (a + b)(b + c)(c + a).
3

Crux Mathematicorum, Vol. 41(9), November 2015

 SOLUTIONS /407

Therefore,
1 1 1
+ +
(a + b)(b + c) (b + c)(c + a) (c + a)(a + b)
(a + b) + (b + c) + (c + a) 27
= ≥
(a + b)(b + c)(c + a) [(a + b) + (b + c) + (c + a)]2
Ç √ å2
3 3 1
= ≥ .
2(a + b + c) 4R2

3987. Proposed by Michel Bataille.

Let ABC be a triangle with circumcircle Γ and let A0 be the point of Γ diamet-
rically opposite to A. The lines AB and AC intersect the tangent to Γ at A0 in
B 0 and C 0 , respectively. Prove that the tangents to Γ at B and C intersect at the
centroid of AB 0 C 0 if and only if 2 cos A = 3 sin B sin C.
We received four correct submissions and present the solution by Oliver Geupel.
Consider the problem in the plane of complex numbers where a, b, c, . . . are the
complex numbers representing the points A, B, C, . . . . Assume that Γ with centre
O is the unit circle, a = 1, and a0 = −1.
As is easily verified, if P , Q, R, S are points on the unit circle centred at O, the
points of intersection of the relevant chords and tangents are represented as in the
following table:
The intersection of is represented by
pq(r+s)−rs(p+q)
chords P Q and RS pq−rs
2pqr−r 2 (p+q)
chord P Q and the tangent at R pq−r 2
2pr
the tangents at P and at R p+r

Consequently, the intersection T of tangents to Γ at B and C is represented by

2bc
t= ,
b+c
and the centroid G of triangle AB 0 C 0 is represented by
Å ã
1 1 3b + 1 3c + 1
g = (1 + b0 + c0 ) = 1+ + .
3 3 1−b 1−c
(Note that G is always well defined.) Hence the condition T = G is equivalent to
t − g = 0, which factors as (2bc − b − c)(3bc + b + c + 3) = 0. If 2bc − b − c = 0
then t = 1 = a (that is, T = A), which is impossible. Therefore
T =G ⇐⇒ 3bc + b + c + 3 = 0. (1)
Next, we show the following:
T =G =⇒ cos A > 0. (2)

Copyright c Canadian Mathematical Society, 2016

408/ SOLUTIONS

If, to the contrary, cos A = 0 then the tangents at B and C would be parallel so
that they could not intersect at G. Similarly, if cos A < 0, in which case A would
be obtuse, then A would be an interior point of triangle BCT . As a consequence,
T would lie on one side of the tangent to Γ at A while B 0 and C 0 (and therefore
G) would lie on the other, and we would have T 6= G. This proves (2).
Note that cos A is the distance of point O from the line BC, that is
2 2
AC 2 (1−b)2
cos2 A = (b+c) 2
= − (1−c) 2


4bc ; furthermore, sin B = 2 4c , and sin C = − 4b .
Hence

(2 cos A)2 = (3 sin B sin C)2 ⇐⇒ (3bc − 7b − 7c + 3)(3bc + b + c + 3) = 0. (3)

Next, we show the following:

2 cos A = 3 sin B sin C =⇒ 3bc + b + c + 3 = 0. (4)

The proof is by contradiction. Suppose 3bc−7b−7c+3 = 0. Then the intersection

of chords AA0 and BC is represented by 1+bc
b+c
= 73 , which implies that angle A is
obtuse, so that cos A < 0. This contradicts the premise of (4).
Finally we are ready to prove that the two conditions

T =G and 2 cos A = 3 sin B sin C

are equivalent. First assume T = G. Then by (1) and (3), we have (2 cos A)2 =
(3 sin B sin C)2 . From (2) we obtain cos A > 0. Therefore 2 cos A = 3 sin B sin C.
The converse is immediate from (4) and (1).
Editor’s Comments. Let D be the foot of the altitude from A to BC, and E be
the point where the altitude again meets the circumcircle. Then Pranesachar, who
used Cartesian coordinates for his solution, observed that T = G if and only if
E is the midpoint of AD; furthermore, because D is known to be the midpoint
of EH (where H is the orthocentre of ∆ABC), he concluded that E and D are
points of trisection of the line segment AH if and only if T = G, if and only if
2 cos A = 3 sin B sin C.
Bataille noted that because the line AT is known to be the symmedian from A of
∆ABC, this problem provides an alternative construction of this symmedian as
the median from A of ∆AB 0 C 0 .

3988? . Proposed by George Apostolopoulos.

Let a, b, c be positive real numbers. Find the maximum and minimum values of
the expression
a b c
√ +√ +√ .
a2 + 3b2 b2 + 3c2 c2 + 3a2

We received four correct solutions and three flawed submissions. We present the
solution by S. Arslanagić, N. Hodz̆ić and S. Malikić.

Crux Mathematicorum, Vol. 41(9), November 2015

 SOLUTIONS /409

Setting (a, b, c) equal to (x2 , x, y) and (x, y 2 , y) respectively and letting x tend to
infinity and y tend to 0, we obtain the respective limits 2 and 1 for the given
expression. We show that these are its supremum and infimum, although neither
value is actually attained.
For the upper bound, we exploit an inequality of V. Cı̂rtoaje and G. Dospinescu:
Suppose that n ≥ 3, that xi (1 ≤ i ≤ n) are positive reals with x1 x2 · · · xn = 1,
and that 0 < p ≤ (2n − 1)/(n − 1)2 . Then
1 1 1 n
√ +√ + ··· + √ ≤√ .
1 + px1 1 + px2 1 + pxn 1+p
The proof of this can be found online in the book, Algebraic inequalities: old and
new methods, by V. Cı̂rtoaje (GIL Publishing House, 2006). See Problem 8 in
Sections 4.2 and 4.3, pages 199, 205-206.
Applying this to n = 3, p = 5/4, x1 = b2 /a2 , x2 = c2 /b2 and x3 = a2 /c2 , we
deduce that
3 3
X a2 X 1 X 1
= < ≤ 2.
cyclic
a2 + 3b2 i=1
1 + 3xi i=1
1 + 45 xi

For the lower bound, recall a generalization of the Hölder inequality, to wit: Let
n be a positive integer, p, q, r > 1 with p−1 + q −1 + r−1 = 1, and xi , yi and zi
(1 ≤ i ≤ n) be positive reals. Then
n
!1/p n !1/q n
!1/r n
X p
X q X X
r
xi yi zi ≥ xi yi zi .
i=1 i=1 i=1 i=1

Apply this to
n = p = q = r = 3, x1 = [a4 (a2 + 3b2 )]1/3 , y1 = z1 = a1/3 (a2 + 3b2 )−1/6 ,
with analogous substitutions for x2 , x3 , y2 , y3 , z2 , z3 , to obtain
#2
îX ó X
"
4 2 2 a2
a (a + 3b )
a2 + 3b2
ïX  » 3 ò X
 !3   !3 
a2 X a2
3 6 6
= a4 (a2 + 3b2 )   
a2 + 3b2 a2 + 3b2

X» 6 a2 6 a2
3
≥ a4 (a2 + 3b2 ) · ·
a2 + 3b2 a2 + 3b2
≥ (a2 + b2 + c2 )3 .
where each sum is cyclic with three terms. Therefore,
Å ã2
a b c (a2 + b2 + c2 )3
√ +√ +√ ≥ 6
a2 + 3b2 b2 + 3c2 c2 + 3a2 a + b6 + c6 + 3(a4 b2 + b4 c2 + c4 a2 )
> 1.

Copyright c Canadian Mathematical Society, 2016

410/ SOLUTIONS

We conclude from this and the continuity of the expression that its values fill up
the open interval (0, 1).
Editor’s Comments. A. Alt and K.-W. Lau each avoided an appeal to advanced in-
equalities by intricate algebraic arguments involving several cases. One submitter
used multivariate calculus to identify the critical points of

(1 + 3x)−1/2 + (1 + 3y)−1/2 + (1 + 3/(xy))−1/2

in the plane, but failed to provide a comprehensive analysis as x, y → ∞.

3989. Proposed by Dragoljub Milošević.

Let ha , hb and hc be the altitudes, ra , rb and rc be the exradii, r the inradius and
R the circumradius of a triangle. Prove that

ra2 r2 r2
+ b + c ≥ 3(2R − r).
ha hb hc

We received 12 solutions. We present the solution by Miguel Amengual Covas.

We will use the following relations:
(i) ra + rb + rc = 4R + r
(ii) ra2 + rb2 + rc2 = (ra + rb + rc )2 − 2(ra rb + rb rc + rc ra ) = (4R + r)2 − 2s2
1 1 2
(iii) r − ra = ha

(iv) s2 ≤ 4R2 + 4Rr + 3r2

Equalities (i) and (ii) are well-known. The proof of (iii) is nearly trivial as
1 1 s s−a a 2
− = − = = ,
r ra K K K ha

and (iv) is obtained from IH 2 = 4R2 + 4Rr + 3r2 − s2 . Here, s, K, I, and H

denote, respectively, the semiperimeter, area, incenter, and orthocenter.
With sums cyclic over a, b, c, we obtain
X r2 Å ã ÅP 2 X ã
a 1X 2 1 1 1 ra
= ra − = − ra
ha 2 r ra 2 r
Å ã
1 (4R + r)2 − 2s2 8R2 + 2Rr − s2
= − (4R + r) =
2 r r
8R2 + 2Rr − (4R2 + 4Rr + 3r2 ) 4R2 − 2Rr − 3r2
≥ =
r r
4(R − 2r) + 3r(2R − r) 3r(2R − r)
= ≥ = 3(2R − r).
r r

Equality holds if and only if R − 2r = 0, if and only if the triangle is equilateral.

Crux Mathematicorum, Vol. 41(9), November 2015

 SOLUTIONS /411

3990. Proposed by Ángel Plaza.

Let a1 , a2 , . . . , an be positive real numbers such that a1 > a2 > . . . > an . Prove
that Å ã
1 1 1
(a1 − an ) + + ··· + ≥ (n − 1)2 .
a1 − a2 a2 − a3 an−1 − an
When does equality hold?
There were 26 submitted solutions for this problem, all of which proved the inequal-
ity, 3 of which forgot or missed the equality case, 20 of which stated an equality
case that was not as precise as possible, and 3 of which stated the most precise
equality case possible.
Solution 1, by Daniel Vacaru (expanded slightly by the editor). By Bergström’s
Inequality, one has

1 1 1
+ + ··· +
a1 − a2 a2 − a3 an−1 − an
12 12 12
= + + ··· +
a1 − a2 a2 − a3 an−1 − an
(1 + 1 + · · · + 1)2 (n − 1)2
≥ = .
(a1 − a2 ) + (a2 − a3 ) + · · · + (an−1 − an ) a1 − an

A trivial rearrangement yields the result. For the required equality, one must have

a1 − a2 = a2 − a3 = · · · = an−1 − an ,

so
a1 − an = (a1 − a2 ) + (a2 − a3 ) + · · · + (an−1 − an ) = λ(n − 1),
a1 − an
and hence the common difference ak − ak+1 is λ = . It follows that:
n−1
Å ã
a1 − an
ai = an + (n − i) , i = 1, . . . , n − 1.
n−1

Solution 2, by Joseph DiMuro. For any such choice of a1 > a2 > . . . an , let
bi = ai − ai+1 for 1 ≤ i < n (then bi > 0 for all i). We then have a1 − an =
b1 + b2 + · · · + bn−1 , so we can rewrite the desired inequality as follows:
n−1
! n−1 !
X X 1
bi ≥ (n − 1)2 .
i=1 j=1
b j

This can be rewritten as

X bi
≥ (n − 1)2 .
bj
1≤i,j≤n−1

Copyright c Canadian Mathematical Society, 2016

412/ SOLUTIONS

And if we break the summation into cases where i = j and cases where i 6= j, we
obtain
X bi
(n − 1) + ≥ (n − 1)2 .
bj
i6=j

bi
Now, for i > j, let cij = . We can then rewrite the desired inequality as:
bj
XÅ 1
ã
(n − 1) + cij + ≥ (n − 1)2 .
i>j
c ij

1
However, for any positive number c, we have c + ≥ 2, with equality if and only
c
if c = 1. Thus, we have
XÅ 1
ã
(n − 1)(n − 2)
(n − 1) + cij + ≥ (n − 1) + 2 · = (n − 1)2 ,
i>j
cij 2

as desired. And the only way to have equality is to have cij = 1 for all i > j.
That means bi = bj for all i and j, which means that the ai ’s form an arithmetic
sequence.
Editor’s Comments. There are a multitude of solution methods for this one. Only
Vacaru’s solution used Bergström’s inequality; three other solutions used the direct
computation in DiMuro’s solution. 5 solutions used the AM-GM inequality, 6 used
the AM-HM inequality, and 6 used the Cauchy-Schwarz inequality. S. Malikić
listed six different ways to obtain the inequality, five of which are completely
general (the three just listed, the direct method, and Jensen’s inequality), and
Chebyshev’s sum inequality, which requires the difference of the terms to either
increase or decrease. DiMuro and J. Schlosberg both remarked that we don’t
actually need the original terms to be positive.

Crux Mathematicorum, Vol. 41(9), November 2015

 AUTHORS’ INDEX /413

AUTHORS’ INDEX
Solvers and proposers appearing in this issue
(Bold font indicates featured solution.)

Proposers
Michel Bataille, Rouen, France: 4084
D. M. Bătineţu-Giurgiu, Bucharest, and Neculai Stanciu, Buzău, Romania : 4082
José Luis Dı́az-Barrero, Universitat Politècnica de Catalunya, Barcelona, Spain : 4085
Norman Hodžić, Bosnia and Herzegovina, and Salem Malikić, Burnaby, BC : 4090
Ovidiu Furdui, Campia Turzii, Cluj, Romania: 4083
Ardak Mirzakhmedov, Kazakhstan, Almaty : 4088
Lorian Saceanu, Harstad, Norway : 4087
Daniel Sitaru, Romania : 4081, 4086
Daniel Sitaru and Leonard Giugiuc, Romania : 4089

Solvers - individuals
Arkady Alt, San Jose, CA, USA : 3985, 3986, 3988, 3989, 3990
Miguel Amengual Covas, Cala Figuera, Mallorca, Spain : 3984, 3986, 3989
George Apostolopoulos, Messolonghi, Greece : 3981, 3984, 3985, 3986, 3989, 3990
Šefket Arslanagić, University of Sarajevo, Sarajevo, Bosnia and Herzegovina : 3981, 3984,
3985, 3986, 3988, 3989, 3990
Fernando Ballesta, IES Infante Don Juan Manuel, Murcia, Spain : CC141
Roy Barbara, Lebanese University, Fanar, Lebanon : 3984, 3990
Michel Bataille, Rouen, France : OC191, OC195, 3981, 3982, 3983, 3984, 3985, 3986,
3987, 3989, 3990
Brian D. Beasley, Presbyterian College, Clinton, USA : 3984
Mihaela Berindeanu, Bucharest, Romania : 3985
Marcel Chiriţǎ, Bucharest, Romania : 3983
José Luis Dı́az-Barrero, Universitat Politècnica de Catalunya, Barcelona, Spain : 3981
Joseph DiMuro, Biola University, La Mirada, CA, USA : 3990
Andrea Fanchini, Cantu, Italy : CC144, OC195, 3984, 3985, 3986, 3987, 3989
Oliver Geupel, Brühl, NRW, Germany : 3981, 3984, 3987
John G. Heuver, Grande Prairie, AB : 3985, 3986, 3989, 3990
Dag Jonsson, Uppsala, Sweden : 3984, 3985, 3990
Kee-Wai Lau, Hong Kong, China : 3984, 3985, 3986, 3988, 3989
Salem Malikić, Simon Fraser University, Burnaby, BC : 3981, 3984, 3985, 3986,
3990 (6 solutions)
Phil McCartney, Northern Kentucky University, Highland Heights, KY, USA : 3985
Dragoljub Milošević, Gornji Milanovac, Serbia : 3984, 3985, 3986, 3989, 3990
Theoklitos K. Parayiou, Agioi Theodoroi Samos, Greece: 3981, 3984, 3985, 3986
Ricard Peiro i Estruch, IES “Abastos” Valencia, Spain : OC191, 3984, 3985, 3986
Paolo Perfetti, Dipartimento di Matematica, Università degli studi di Tor Vergata Roma,
Rome, Italy : 3986, 3989, 3990
Ángel Plaza, University of Las Palmas de Gran Canaria, Spain : OC193, 3990
C.R. Pranesachar, Indian Institute of Science, Bangalore, India : 3984, 3986, 3987, 3989

Copyright c Canadian Mathematical Society, 2016

414/ AUTHORS’ INDEX

Cao Minh Quang, Nguyen Binh Khiem High School, Vinh Long, Vietnam : 3981, 3985,
3986, 3990
Wolfgang Remmel, Vienna, Austria : 3981, 3985, 3990
Henry Ricardo, Tappan, NY, USA : 3984
Barbara Roy, Lebanese University, Fanar, Lebanon : 3990
Cristóbal Sánchez-Rubio, I.B. Penyagolosa, Castellón, Spain : 3984
Joel Schlosberg, Bayside, NY, USA: 3981, 3982, 3986, 3990 (2 solutions)
Digby Smith, Mount Royal University, Calgary, AB : OC192, 3981, 3982, 3983, 3984,
3985, 3986
Daniel Văcaru, Pitesti, Romania : 3990
Konstantine Zelator, University of Pittsburgh, Pittsburgh, PA, USA : CC141, CC142,
CC144, CC145
Titu Zvonaru, Cománeşti, Romania : CC145, OC191, 3981, 3984, 3985, 3986, 3989, 3990

Solvers - collaborations
AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia : 3981, 3984, 3985, 3986,
3990
Dionne Bailey, Elsie Campbell, and Charles R. Diminnie, Angelo State University,
San Angelo, USA : 3981, 3984, 3985, 3990
John Hawkins and David R. Stone, Georgia Southern University, Statesboro, USA : 3984
Norman Hodžić, Bosnia and Herzegovina, and Salem Malikić, Burnaby, BC : 3988
Missouri State University Problem Solving Group : OC191, OC192
Haohao Wang and Jerzy Wojdylo, Southeast MO State University: 3981, 3985

Crux Mathematicorum, Vol. 41(9), November 2015