Crux Mathematicorum

VOLUME 42, NO. 1 January / Janvier 2016

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

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

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

Editor-at-Large Bill Sands University of Calgary

Managing Editor Denise Charron Canadian Mathematical Society

Copyright
c Canadian Mathematical Society, 2016
 IN THIS ISSUE / DANS CE NUMÉRO

3 Editorial Kseniya Garaschuk

4 The Contest Corner: No. 41 John McLoughlin
4 Problems: CC201–CC205
6 Solutions: CC151–CC155
10 Wobbling Bicycle Luis Goddyn
11 The Olympiad Corner: No. 339 Carmen Bruni
11 Problems: OC261–OC265
13 Solutions: OC201–OC205
18 Book Reviews Robert Bilinski
20 Focus On . . . : No. 20 Michel Bataille
24 Generating inequalities using Schweitzer’s theorem
Daniel Sitaru and Claudia Nănuţi
28 Problems: 4101–4110
31 Solutions: 4001–4010
45 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
 3

EDITORIAL
In academia, we write a lot. As students, we are normally told how long our
writing pieces should be. But academic adulthood often comes with more nebulous
guidelines in terms of length of various writing pieces. So my question is, does the
size matter? In spirit, it should not (or should it?); in reality, we have some, often
untold, expectations about the “reasonable” minimum and maximum.
An abstract for a talk is expected to be no more than 200 words, an abstract
for a paper normally spans no more than a third of a page, a conventional Ph.D.
thesis spreads 100 or more pages. But what about articles? Do norms change
depending on the area or the journal? A thesis is supposed to be somewhat
comprehensive and include the necessary background for the reader not to be
bothered looking up all the sources. An article, on the other hand, assumes a
more specialized audience and hence does not require an in-depth background or
literature review. But who really decides on the necessary length as a measure of
an article’s worth? And is there such a thing as not long enough? Maybe not. John
H. Conway and Alexander Soifer win the prize for the shortest math article ever
published: their “Can n2 + 1 unit equilateral triangles cover an equilateral triangle
of side > n, say n + ?” published in the January 2005 edition of The American
Mathematical Monthly consists of just two words. (For the full story behind this
article and its publication, read the piece by Alexander Soifer in Mathematics
Competitions, 23 (1), available here: http://www.openculture.com/2015/04/
shortest-known-paper-in-a-serious-math-journal.html).
At Crux we have our own standards: an average article spans 5-6 pages while each
issue consists of approximately 45–50 pages. But problems and solutions come in
all shapes and sizes: short and long, pure and applied, “from the book” and brute
force. Flip through the following pages of Crux to see it all.
Kseniya Garaschuk

Copyright
c Canadian Mathematical Society, 2016
4/ THE CONTEST CORNER

THE CONTEST CORNER

No. 41
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 December

1, 2016, although late solutions will also be considered until a solution is published.

CC201. An expedition to the planet Bizarro finds the following equation

scrawled in the dust.

3x2 − 25x + 66 = 0 =⇒ x = 4 or x = 9.

What base is used for the number system on Bizarro?

CC202. The positive integers from 1 to n inclusive are written on a black-

board. After one number is erased, the average (arithmetic mean) of the remaining
n − 1 numbers is 46 20
23 . Determine n and the number that was erased.

CC203. Two circles, one of radius 1, the other of radius 2, intersect so that
the larger circle passes through the centre of the smaller circle. Find the distance
between the two points at which the circles intersect.

CC204. A 10 metre ladder rests against a vertical wall. The midpoint of the
ladder is twice as far from the ground as it is from the wall. At what height on
the wall does the ladder reach?

CC205. In the parallelogram ABCD, point X lies on AB such that XB is

twice the length of AX. Let Y be the point of intersection of XC and BD. What
fraction is the area of the triangle DCY of the area of the parallelogram ABCD?

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

CC201. Une expédition à la planète Bizarro découvre l’énoncé suivant inscrit

dans le sable.
3x2 − 25x + 66 = 0 =⇒ x = 4 ou x = 9.
Quelle est la base du système de numération de la planète Bizarro?

Crux Mathematicorum, Vol. 42(1), January 2016

 THE CONTEST CORNER /5

CC202. On écrit au tableau les entiers positifs de 1 à n. Un des nombres est

20
éffacé. La moyenne des n − 1 nombres qui restent est 46 23 . Déterminer la valeur
de n ainsi que le nombre éffacé.

CC203. Trouver la distance entre les deux points d’intersection de deux

cercles, de rayon 1 et 2 respectivement, qui se coupent de sorte le plus grand passe
par le centre du plus petit.

CC204. Une échelle longue de dix mètres est placée contre un mur vertical. Si
le milieu de l’échelle est deux fois plus distant du sol que du mur, à quelle hauteur
l’échelle s’appuie-t-elle contre le mur?

CC205. Dans le parallélogramme ABCD, soit X le point du segment AB

tel que XB est deux fois plus long que AX. Soit Y le point d’intersection de XC
et BD. Trouver le rapport de l’aire du triangle DCY à celle du parallélogramme
ABCD.

Math Quotes

Unfortunately what is little recognized is that the most worthwhile scientific

books are those in which the author clearly indicates what he does not know; for
an author most hurts his readers by concealing difficulties.

Evariste Galois, in N. Rose (ed.) “Mathematical Maxims and Minims”,

Raleigh NC: Rome Press Inc., 1988.

Copyright
c Canadian Mathematical Society, 2016
6/ THE CONTEST CORNER

CONTEST CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2015: 41(1), p. 4–5.

CC151. Consider a non-zero integer n such that n(n + 2013) is a perfect

square.
a) Show that n cannot be prime.
b) Find a value of n such that n(n + 2013) is a perfect square.
Originally question 1 from 2013 Pan African Mathematics Olympiad.
We received eleven correct solutions. We present two of the solutions.
Solution 1, by Andrea Fanchini.
a) We denote m = n(n + 2013). If m is a perfect square and n is prime then n
must divide n + 2013. By the divisibility properties n must then also be a factor
of 2013 = 3 · 11 · 61. Thus there are three possibilities for m:

m = 3(3 + 2013), m = 11(11 + 2013), m = 61(61 + 2013).

None of these numbers are square, so n cannot be prime.

b) We know that the sum of odd numbers gives a perfect square. If we set

n = 1 + 3 + · · · + 2011 = 10062 ,

then
n + 2013 = 1 + 3 + · · · + 2011 + 2013 = 10072
and m = 10062 · 10072 is a perfect square.

Solution 2, by Albert Stadler.

Put t = gcd(n, 2013). Note that gcd(n, n + 2013) = gcd(n, 2013) = t. Then

n n + 2013
n(n + 2013) = t2 · ·
t t
and gcd( nt , n+2013
t ) = 1. So n(n + 2013) is a perfect square if and only if both n
t
and n+2013
t are perfect squares. Set nt = a2 and n+2013
t = b2 . Then

2013
= b2 − a2 = (b − a)(b + a).
t
Since 2013 = 3 · 11 · 61, we have t ∈ {1, 3, 11, 33, 61, 183, 671, 2013}.

Crux Mathematicorum, Vol. 42(1), January 2016

 THE CONTEST CORNER /7

t = 2013 : (b − a)(b + a) = 1 has no solutions in positive integers a, b.

t = 671 : (b − a)(b + a) = 3 implies (a, b) = (1, 2).
t = 183 : (b − a)(b + a) = 11 implies (a, b) = (5, 6).
t = 61 : (b − a)(b + a) = 33 implies (a, b) = (16, 17) or (a, b) = (4, 7).
t = 33 : (b − a)(b + a) = 61 implies (a, b) = (30, 31).
t = 11 : (b − a)(b + a) = 183 implies (a, b) = (91, 92) or (a, b) = (29, 32).
t = 3 : (b − a)(b + a) = 671 implies (a, b) = (335, 336) or (a, b) = (25, 36).
t = 1 : (b − a)(b + a) = 2013 implies (a, b) = (1006, 1007) or
(a, b) = (334, 337) or (a, b) = (86, 97) or (a, b) = (14, 47).

With n = a2 t we obtain that n(n + 2013) is a perfect square if and only if n is one
of 196, 671, 976, 1875, 4575, 7396, 9251, 15616, 29700, 91091, 111556, 336675, or
1012036, none of which is prime.

CC152. A square of an n × n chessboard with n ≥ 5 is coloured in black

and white in such a way that three adjacent squares in either a line, a column
or a diagonal are not all the same colour. Show that for any 3 × 3 square inside
the chessboard, two of the squares in the corners are coloured white and the two
others are coloured black.
Originally question 5 from 2013 Pan African Mathematics Olympiad.
We received only one incorrect submission.

CC153. A sequence a0 , a1 , . . . , an , . . . of positive integers is constructed as

follows:
• if the last digit of an is less than or equal to 5, then this digit is deleted and
an+1 is the number consisting of the remaining digits; if an+1 contains no
digits, the process stops;
• otherwise, an+1 = 9an .
Can one choose a0 so that we can obtain an infinite sequence?
Originally question 5 from 2010 Pan African Mathematics Olympiad.
We received two correct solutions and one incomplete submission. We present the
solution by Titu Zvonaru.
It is not possible to obtain an infinite sequence. If the last digit of an is less than
or equal to 5, then it is obvious that an+1 < an . If the last digit of an is greater
than 5, then the last digit of an+1 is less than 5. It results that

an+2 = [an+1 /10] = [9an /10] < an .

Copyright
c Canadian Mathematical Society, 2016
8/ THE CONTEST CORNER

So if we had an infinite sequence (an ) of positive integers we would find an infinite

strictly decreasing subsequence, a contradiction.

CC154. The numbers 11 , 12 , . . . , 2012

1
are written on the blackboard. Alice
chooses any two numbers from the blackboard, say x and y, erases them and
instead writes the number x + y + xy. She continues to do so until there is only
one number left on the board. What are the possible values of the final number?
Originally question 4 from 2012 Pan African Mathematics Olympiad.
There was one correct solution for this problem and two incomplete submissions.
We present the solution by Konstantine Zelator.
Note that x = (x + 1) − 1 and xy + x + y = (x + 1)(y + 1) − 1.
We use the following lemma:
If X = (a1 + 1) · · · (ak + 1) − 1 and Y = (ak+1 + 1) · · · (ak+m + 1) − 1,
then XY + X + Y = (a1 + 1) · · · (ak+m + 1) − 1.
By writing XY + X + Y = (X + 1)(Y + 1) − 1, the lemma follows immediately.
From this lemma, it follows that if the board starts with numbers n1 , n2 , . . . , nt
and the given operation is applied to the numbers in any order until a single
number remains, that number will be (n1 + 1)(n2 + 1) · · · (nt + 1) − 1.
For the set of numbers 1, 21 , . . . , 2012
1
the final answer will thus be

2 3 4 2013
· · ··· − 1 = 2012.
1 2 3 2012

CC155. Find all real solutions x to the equation [x2 − 2x] + 2[x] = [x]2 . Here
[a] denotes the largest integer less than or equal to a.
Originally question 3 from 2012 Pan African Mathematics Olympiad.
There were three correct solutions for this problem and one incorrect submission.
We present the solution by the Missouri State University Problem Solving Group.
The equation is true for any integer x, so we need only fnid the non-integer so-
lutions. Suppose x is a non-integer solution and bxc = n. Then x = n +  for
some  with 0 <  < 1. We will make use of the fact that for any integer k,
ba + kc = bac + k. We have:

b(x − 1)2 − 1c + 2bxc = bxc2

b(x − 1)2 c = bxc2 − 2bxc + 1
b(n − 1 + )2 c = (n − 1)2
b(n − 1) + 2(n − 1) + 2 c
2
= (n − 1)2
b2(n − 1) + 2 c =0

Crux Mathematicorum, Vol. 42(1), January 2016

 THE CONTEST CORNER /9

Equivalently,
0 ≤ 2(n − 1) + 2 < 1. (1)

Since we assumed x is not an integer,  6= 0. Taking both sides of inequality (1)

we get:

 1 + 2 − 2
1− ≤n≤ .
2 2

Since 0 <  < 1, we have 0 < 1 − /2 < 1, implying n > 0.

Taking the right side of inequality (1) and completing the square gives:

( + (n − 1))2 < (n 2
p − 1) + 1
 + n − 1 < p(n − 1)2 + 1
x = n +  < (n − 1)2 + 1 + 1

For any positive integer n, this gives the following interval for the solution x:
»
(n, (n − 1)2 + 1 + 1).

Thus, the set of all solutions is given by

∞ 
[ » [
n, (n − 1)2 + 1 + 1 Z.
n=1

Copyright
c Canadian Mathematical Society, 2016
10/ WOBBLING BICYCLE

WOBBLING BICYCLE
Proposed by Luis Goddyn, Simon Fraser University, Burnaby, BC.

A wobbling bicycle passes through a mud patch. One of its wheels traces a part
of the curve y = sin x. The other wheel makes a curve with a vertical inflection
point.

How long is the bicycle?

In order to eliminate effects due to bicycle geometry, tilting and wheel size, assume
the bicycle has vanishingly thin tires with its front axle always positioned directly
below a vertical headset. Assume also that both wheels were in the mud patch
when the inflection point is traversed. Determine the distance ` between its axles.

Editor’s Comment. This problem is related to a famous puzzle of determining

in which way a bicycle was going from its tracks. You can find it in the book
Which Way Did the Bicycle Go?: And Other Intriguing Mathematical Mysteries
Side
by Joseph D. E. Konhauser, Dan View ofand
Velleman Bicycle
Stan Wagon.
The solution to this puzzle will appear in Crux 42(5).

Crux Mathematicorum, Vol. 42(1), January 2016

 THE OLYMPIAD CORNER /11

THE OLYMPIAD CORNER

No. 339
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 December
1, 2016, although late solutions will also be considered until a solution is published.
The editor thanks The editor thanks Rolland Gaudet, retired professor of the University
College of Saint Boniface, for translations of the problems.

OC261. Show that there are no 2-tuples (x, y) of positive integers satisfying
the equation (x + 1)(x + 2) · · · (x + 2014) = (y + 1)(y + 2) · · · (y + 4028).

OC262. In obtuse triangle ABC, with the obtuse angle at A, let D, E, F be

the feet of the altitudes through A, B, C respectively. DE is parallel to CF , and
DF is parallel to the angle bisector of ∠BAC. Find the angles of the triangle.

OC263. An integer n ≥ 3 is called special if it does not divide

Å ã
1 1
(n − 1)! 1 + + · · · + .
2 n−1

Find all special numbers n such that 10 ≤ n ≤ 100.

OC264. 2
A positive integer is called beautiful if it can be represented in the
+y 2
form xx+y for two distinct positive integers x, y. A positive integer that is not
beautiful is ugly.
1. Prove that 2014 is a product of a beautiful number and an ugly number.
2. Prove that the product of two ugly numbers is also ugly.

OC265. Five airway companies operate in a country consisting of 36 cities.

Between any pair of cities exactly one company operates two way flights. If some
air company operates between cities A, B and B, C we say that the ordered triple
A, B, C is properly-connected. Determine the largest possible value of k such that
no matter how these flights are arranged there are at least k properly-connected
triples.

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

Copyright
c Canadian Mathematical Society, 2016
12/ THE OLYMPIAD CORNER

OC261. Démontrer qu’il n’existe aucun couple d’entiers positifs (x, y) satis-
faisant à l’équation (x + 1)(x + 2) · · · (x + 2014) = (y + 1)(y + 2) · · · (y + 4028).

OC262. Soit un triangle obtus ABC, où l’angle obtus se situe à A, et soient
D, E, F les pieds des altitudes provenant de A, B, C respectivement. DE est
parallèle à CF et DF est parallèle à la bissectrice de ∠BAC. Déterminer les
angles du triangle.

OC263. Un entier n ≥ 3 est dit spécial s’il ne divise pas

Å ã
1 1
(n − 1)! 1 + + · · · + .
2 n−1

Déterminer tous les nombres spéciaux n tels que 10 ≤ n ≤ 100.

2 2
OC264. Un entier est dit adorable s’il peut être représenté sous la forme xx+y
+y

pour deux entiers positifs distincts x, y. Un entier positif qui n’est pas adorable
est dit moche.
1. Démontrer que 2014 est le produit d’un nombre adorable et un nombre
moche.
2. Démontrer que le produit de deux nombres moches est moche.

OC265. Cinq compagnies aériennes opèrent dans un pays comprenant 36 villes.

Entre toute paire de villes, exactement une compagnie aérienne opère un vol aller-
retour. Si une compagnie aérienne opère entre les villes A, B puis B, C, on dit
que le triplet A, B, C est proprement connecté. Déterminer la plus grande valeur
possible de k telle que, quelle que soit l’organisation des vols, il y aura toujours
au moins k triplets proprement connectés.

Crux Mathematicorum, Vol. 42(1), January 2016

 THE OLYMPIAD CORNER /13

OLYMPIAD SOLUTIONS
Statements of the problems in this section originally appear in 2014: 40(9), p. 374–375.

OC201. Find all functions f : R → R such that f (0) ∈ Q and

2
f (x + f (y)2 ) = f (x + y) .

Originally problem 2 from the third round algebra of the 2013 Iran National Math-
ematical Olympiad.
We received one correct solution. We present the solution by Oliver Geupel.
The two constant functions f (x) = 0 and f (x) = 1 have the required property,
and we show that there are no other solutions.
Let us refer to the functional equation as F (x, y). Let f be a solution and let
f (0) = q = a/b where a, b are integers and b ≥ 1. From F (x + y − q 2 , 0) we obtain
f (x + y) = f (x + y − q 2 + f (0)2 ) = f (x + y − q 2 )2 . Specialising x = 0, we also
have f (y) = f (y − q 2 )2 . Hence, using F (x, y − q 2 ), we see that

f (x + f (y)) = f (x + f (y − q 2 )2 ) = f (x + y − q 2 )2 = f (x + y). (1)

Setting y = 0 in (1), we obtain f (x + q) = f (x) and therefore

f (x + nq) = f (x) (2)

for every integer n. A further consequence of (1) is

f (f (x)) = f (x).

By F (x, 0), we have f (x + q 2 ) = f (x)2 . Mathematical induction yields

n
f (x + nq 2 ) = f (x)2 (n ∈ N). (3)

From F (x − q 2 , 0), we see that f (x) = f (x − q 2 + f (0)2 ) = f (x − q 2 )2 ≥ 0. By (2)

b
and (3), f (x) = f (x + aq) = f (x + bq 2 ) = f (x)2 , so that f (x) ∈ {0, 1}; whence
also q ∈ {0, 1}. We consider the cases q = 1 and q = 0 in succession.
Case q = 1. We show that f is the constant function f (x) = 1. The proof is by
contradiction. Suppose for some real number t it holds f (t) = 0. Then,

1 = f (0) = f (f (t)) = f (t) = 0,

a contradiction which shows that f is the constant function f (x) = 1.

Case q = 0. We prove that f is the constant function f (x) = 0. The proof is again
by contradiction. Assume f (t) = 1 for some real number t. By (1),
Å Å ãã Å ã
1 1 1 1
f +f =f + = f (1) = f (f (t)) = f (t) = 1.
2 2 2 2

Copyright
c Canadian Mathematical Society, 2016
14/ THE OLYMPIAD CORNER

If f 12 = 0,
then f 12 + f
12 = 0, which

is impossible. Thus f 21 = 1. Hence,

by (1), f 32

= f 12 + 1 = f
21 + f 12 f 21 − 12 =


= 1. Moreover, 0 =


f 12 + f −
12 . Note that 1 1 1

f − 2 is either 0 or 1. So f 2 + f − 2 is equal to
1 3
either f 2 = 1 or f 2 = 1. This is impossible and the proof is complete.

OC202. Let a, b be real numbers such that the equation x3 − ax2 + bx − a = 0

3
has three positive real roots. Find the minimum of 2a −3ab+3a
b+1 .
Originally problem 1 from day 1 of the 2013 South East Mathematical Olympiad.
We received six correct solutions. We present the solution by Michel Bataille.
First, let S be the set of all pairs of real numbers (a, b) such that the equation
3
x3 − ax2 + bx − a = 0 has three positive real roots and let R(a, b) = 2a −3ab+3a .
√ b+1
We show that min R(a, b) = 9 3.
(a,b)∈S
√ √
Now, if a = 3 3, b = 9, the equation becomes (x − 3)3 = 0 whose √ roots √
are
clearly positive real numbers and it is readily checked
√ that R(3 3, 9) = 9 3.
Thus, there just remains to prove that R(a, b) ≥ 9 3 whenever (a, b) ∈ S.
Let (a, b) ∈ S and let x1 , x2 , x3 be positive roots of x3 − ax2 + bx − a = 0. Then,

x1 + x2 + x3 = a, x1 x2 + x2 x3 + x3 x1 = b, x1 x2 x3 = a.

Note that the above shows that a, b > 0 since the roots are positive real numbers.
Observing that

(x1 + x2 + x3 )3 = x31 + x32 + x33 + 3(x1 + x2 + x3 )(x1 x2 + x2 x3 + x3 x1 ) − 3x1 x2 x3

we obtain 2a3 = 2(x31 + x32 + x33 ) + 6ab − 6a, hence

2a3 −3ab+3a = 2(x31 +x32 +x33 )+3ab−3a ≥ 2·3x1 x2 x3 +3ab−3a = 3ab+3a = 3a(b+1)

(using AM-GM for the inequality). It follows that

3a(b + 1)
R(a, b) ≥ = 3a. (1)
b+1

3 a3
√
Now, by AM-GM again, x1 +x32 +x3 ≥ x1 x2 x3 , hence ≥ a and so a ≥ 3 3.
√ 27
With (1), we deduce that R(a, b) ≥ 9 3, as desired.

OC203. Find all positive integers m and n satisfying 2n + n = m!.

Originally problem 1 from day 2 of the 2013 Turkey Mathematical Olympiad.
We received one correct solution. We present the solution by Oliver Geupel.
A solution is
(m, n) = (3, 2)
and we show that it is unique.

Crux Mathematicorum, Vol. 42(1), January 2016

 THE OLYMPIAD CORNER /15

Suppose that (m, n) is any solution. Then there exists a nonnegative integer a
and a positive odd integer b such that n = 2a b. The exact power of 2 that
divides m! = 2n + n = 2a (2n−a + b) is 2a . Thus m ≤ 2a + 1. If a ≤ 4 then
m ≤ 9. A straightforward inspection shows that when m ≤ 9, the only solution is
(m, n) = (3, 2). We now consider the case a ≥ 5.
We prove that for every a ≥ 5 it holds
a
22 > (2a + 1)!.
The proof is by mathematical induction on a. The base case a = 5 is satisfied
since
232 > 227 = 28 · 27 · 25 · 23 · 24 > 28 · 34 · 52 · 7 · 11 = 11!
a−1
Suppose that for some a ≥ 6 we have 22 > (2a − 1)!. Then
a
22 > (2a − 1)!2 > (2a − 1)! · 2(2a − 2) · 3(2a − 1) > (2a − 1)! · 2a · (2a + 1) = (2a + 1)!,
which completes the induction.
a
We conclude 2n + n > 22 > (2a + 1)! ≥ m!, a contradiction.

OC204. Let ABC be a triangle. Find all points P on segment BC satisfying

the following property: If X and Y are the intersections of line P A with the
common external tangent lines of the circumcircles of triangles P AB and P AC,
then Å ã
PA 2 PB · PC
+ = 1.
XY AB · AC

Originally problem 6 from day 2 of the 2013 USA Mathematical Olympiad.

We received one correct solution. We present the solution by Titu Zvonaru.
There are only two such points, namely the intersection of the internal bisector of
∠BAC with BC or its reflection with respect to the midpoint of BC.
Let a = BC, b = CA, c = AB and let Γ1 (O1 , R1 ) and Γ2 (O2 , R2 ) be the circum-
circles of triangles P AB and P AC respectively. Let M be the midpoint of AP
(and XY ) and let T1 and T2 be the points of tangency of the common external
tangent through X with the circles Γ1 and Γ2 respectively.
By the power of a point with respect to a circle, we have XT12 = XA · XP = XT22 .
Hence X is the midpoint of T1 T2 . Since the point M lies on O1 O2 and ∠XM O2 =
∠O2 T2 X = π/2, we obtain the following equivalences:
(XM )2 + (M O2 )2 = (O2 T2 )2 + (T2 X)2 ,
(T1 T2 )2
(XM )2 = −(R2 cos(C))2 + R22 + ,
4
4(XM )2 = 4R22 sin2 (C) + (O1 O22 − (R1 − R2 )2 ),
(XY )2 = 4R22 sin2 (C) + (R1 cos(B) + R2 cos(C))2 − R12 − R22 + 2R1 R2 ,
(XY )2 = 4R22 sin2 (C) − R12 sin2 (B) − R22 sin2 (C) + 2R1 R2 cos(B) cos(C) + 2R1 R2 .

Copyright
c Canadian Mathematical Society, 2016
16/ THE OLYMPIAD CORNER

Applying the law of sines, it follows that

(AP )2 (AP )2 (AP )2 cos(B) cos(C) AP 2

(XY )2 = (AP )2 − − + + ,
4 4 2 sin(B) sin(C) 2 sin(B) sin(C)
(AP )2 (1 + cos(B) cos(C) + sin(B) sin(C))
(XY )2 = ,
2 sin(B) sin(C)
(AP )2 2 sin(B) sin(C)
= .
(XY )2 1 + cos(B) cos(C) + sin(B) sin(C)

Letting x = BP/P C, we get BP = ax/(x + 1) and P C = a/(x + 1). This yields

Å ã2
PA PB · PC
+ =1
XY AB · AC
2 sin(B) sin(C) a2 x
⇔ + =1
1 + cos(B) cos(C) + sin(B) sin(C) bc(x + 1)2
x sin2 (A) 1 + cos(B) cos(C) − sin(B) sin(C)
⇔ =
(x + 1)2 sin(B) sin(C) 1 + cos(B) cos(C) + sin(B) sin(C)
x(1 − cos2 (A)) 1 + cos(B + C)
⇔ =
(x + 1)2 sin(B) sin(C) 1 + cos(B) cos(C) + sin(B) sin(C)
⇔ x2 sin(B) sin(C) + x(−1 + cos(B − C) cos(B + C)) + sin(B) sin(C) = 0.

Since
cos(2B) + cos(2C)
−1 + cos(B − C) cos(B + C) = −1 +
2
1 − 2 sin2 (B) + 1 − 2 sin2 (C)
= −1 + ,
2
we obtain the equation

x2 sin(B) sin(C) − x(sin2 (B) + sin2 (C)+ sin(B) sin(C) = 0,

which is a quadratic equation with roots x = sin(C)/ sin(B) and x = sin(B)/ sin(C).
Thus, the points P are the intersection of the internal bisector of ∠BAC with BC
or its reflection with respect to the midpoint of BC.

OC205. For each positive integer n determine the maximum number of points
in space creating the set A which has the following properties:
1. the coordinates of every point from the set A are integers from the range
[0, n];
2. for each pair of different points (x1 , x2 , x3 ), (y1 , y2 , y3 ) belonging to the set A
at least one of the following inequalities x1 < y1 , x2 < y2 , x3 < y3 is satisfied
and at least one of the following inequalities x1 > y1 , x2 > y2 , x3 > y3 is
satisfied.

Crux Mathematicorum, Vol. 42(1), January 2016

 THE OLYMPIAD CORNER /17

Originally problem 6 from day 2 of the 2013 Polish Mathematical Olympiad.

We received one correct solution. We present the solution by Oliver Geupel.
The answer is õ û
3(n + 1)2 + 1
an = .
4
We show that an an -element set with the desired properties is A = {(x, y, z) :
x + y + z = b3n/2c}. In fact, if n is an even number, n = 2m, then members of A
are points (x, y, 3m − x − y) where 0 ≤ x ≤ m and m − x ≤ y ≤ 2m, as well as
points (x, y, 3m − x − y) where m + 1 ≤ x ≤ 2m and 0 ≤ y ≤ 3m − x, the total
number of elements being
m
X 2m
X
(m + 1 + x) + (3m + 1 − x) = 3m2 + 3m + 1 = an .
x=0 x=m+1

If n is odd, n = 2m + 1, then members of A are points (x, y, 3m + 1 − x − y) where

0 ≤ x ≤ m and m − x ≤ y ≤ 2m + 1, and points (x, y, 3m + 1 − x − y) where
m + 1 ≤ x ≤ 2m + 1 and 0 ≤ y ≤ 3m + 1 − x, with the total number of elements
m
X 2m+1
X
(m + 2 + x) + (3m + 2 − x) = 3(m + 1)2 = an .
x=0 x=m+1

It remains to show that every set A with the required properties has not more
than an elements. Let us define subsets B0 , . . . , Bn of the lattice cube [0, n]3 . The
members of Bk are the points (x, n − k, z) where 0 ≤ x ≤ k − 1 and 0 ≤ z ≤ n, as
well as the points (k, y, z) where n − k ≤ y ≤ n and 0 ≤ z ≤ n. So Bk consists of
2k + 1 classes of n + 1 elements each, where the members of a single class vary only
in the third coordinate. Let P = (x, y, z) ∈ [0, n]3 . It follows that, if x + y < n
then P ∈ Bn−y , whereas if x + y ≥ n then P ∈ Bx . Hence the sets B0 , . . . , Bn
constitute a disjoint partition of the lattice cube.
Let A be a set with the required properties. Then A ∩ Bk has not more than
2k + 1 elements because it cannot contain any two members from the same class
by the given property 2. Also by property 2., the elements in A ∩ Bk have distinct
z-coordinates. Thus A ∩ Bk has not more than n + 1 elements. We obtain |A| =
n n
|A ∩ Bk | ≤
P P
min(2k + 1, n + 1). If n is an even number, n = 2m, then
k=0 k=0
n
X m
X 2m
X
min(2k + 1, n + 1) = (2k + 1) + (2m + 1) = 3m2 + 3m + 1 = an .
k=0 k=0 k=m+1

If n is odd, say n = 2m + 1, then

n
X m
X 2m+1
X
min(2k + 1, n + 1) = (2k + 1) + (2m + 2) = 3(m + 1)2 = an .
k=0 k=0 k=m+1

Consequently, A has at least an elements, which completes the proof.

Copyright
c Canadian Mathematical Society, 2016
18/ BOOK REVIEWS

BOOK REVIEWS
Robert Bilinski
Statistics Done Wrong: The Woefully Complete Guide by Alex Reinhart.
ISBN 9789-1-59327-620-1, 158 pages
Published by No starch press, 2015
Reviewed by Robert Bilinski, Collège Montmorency.
In his graduate studies in physics, M. Reinhart discovered that unintentional sta-
tistical errors are more common than all-out fraud. Ultimately, this physicist got
hooked on statistics and reoriented his career path: now his main research interest
is finding models that predict where crimes occur. This book, the author’s first,
is the result of his newfound passion for uncovering statistical fallacies.
Statistics Done Wrong is split into 12 chapters with titles such as “Statistical Power
and Underpowered Statistics”, “Pseudoreplication: Choose Your Data Wisely”,
“The p Value and the Base Rate Fallacy”, “Double Dipping in the Data”. If you
are unfamiliar with these terms, you should probably read this book. It will give
you an idea of how to apply critical thinking to published research and recognize
many possible sources of errors such as poorly planned experiments, bad data
collection and errors in compilation. At the beginning, the book assumes that
all the correct steps have been taken. The author then tries to answer “What is
statistics as a field?”, “What are the various statistics and how can they be used
correctly?”, “What do they mean and what do they not mean?”. Later, more
fundamental questions are broached: “How do you measure what you want to
study?”, “What answers does your data give you and does it answer the questions
you asked?”.
And if the titles of the chapters are familiar to you,
it probably means you have done some statistical
analysis, but I still recommend this book. The con-
tent will offer some interesting tidbits. The review
of the possible errors is quite extensive and the vari-
ety of the examples of misuse is not only wide rang-
ing, but also stemming from many fields. The last
few chapters cover the unethical side of the research
industry as well as structural flaws that encourage
errors in publishing. The very last chapter offers a
few guidelines on how to be more conscious and try
to avoid the pitfalls in statistical research.
Books that talk about what not to do are rare
whereas I feel that they should occupy a bigger place
in a scientist’s reading diet. It keeps the mind sharp;
it is all too easy to be stuck in a rut without knowing it. Reading a book like this
recharges the good “doubting of oneself and one’s approach” that makes science
advance in objectivity towards truth. Naturally, this book will be of much more

Crux Mathematicorum, Vol. 42(1), January 2016

 BOOK REVIEWS /19

apparent use to an applied mathematician dealing with authentic data and mod-
elling. However, the current overspecialization and the decrease in mathematical
and statistical content in other scientific fields make it all the more important
that all mathematicians and statisticians get more knowledgeable about resources
like this book. One becomes wise when reading the very subtle shortcomings of
others; as Socrates was quoted by Plato to have said : “The beginning of wisdom
is to know that one knows nothing.” Actually, the crux of the book comes down
to “even if a scientist is well meaning, he can make subtle methodological errors
that make his results unusable”. As an example, let us read (p.37) the analysis of
the error on “menstrual synchronization” apparently established in a 1971 Nature
article on (M. K. McClintock (1971), “Menstrual Synchrony and Suppression”,
Nature 229, p. 244-245):
Unfortunately, the statistical test they used assumed that if there was
no synchronization, the deviation would randomly increase or decrease
from one period to another. But imagine two women in the study who
start with aligned cycles. One has an average gap of 28 days between
periods and the other a gap of roughly 30 days. Their cycles will
diverge consistently over the course of the study, starting 2 days apart,
then four days, and so on, with only a bit of random variation because
periods are not perfectly timed. Similarly, two women can start the
study not aligned but gradually align.
This example shows us that statistical problems contain mathematics, that their
solution requires problem-solving skills much like Crux problems, that measuring
things is not particularly easy, and, moreover, as a society we have erred in lowering
math standards.
Statistics Done Wrong is not all about errors; it is also about avoiding them. Each
chapter ends with a brief list of do’s and don’ts that should limit the mistakes
exposed in the chapter. Naturally, these lists are not exhaustive and failsafe, but
they offer stepping stones to a better statistical practise.
The above example is one of many in the book. The reader will not learn statistics
while reading it though. There are no formulas or graphs or even data for that
matter. Statistics Done Wrong is a general interest book that should be read by
researchers in all fields. The style and writing is fluid and enjoyable, but, as is true
for all books of this level, one needs to be available mentally to fully benefit from
it. It is not light reading material, but oh so necessary. If this book review entices
you to read the book under scrutiny and if you find it interesting, I also recommend
Common Errors in Statistics and How to Avoid Them published at Wiley. This
second book is technical, with formulas, graphs and data. The next evolution
would be statistical case study books, supposing you already have knowledge of
advanced statistical techniques. If this kind of book is well written, you can have
a statistical apprentice’s journey which will further your skills. Statistics Done
Wrong has been done right. Good reading!

Copyright
c Canadian Mathematical Society, 2016
20/ FOCUS ON... INEQUALITIES VIA COMPLEX NUMBERS

FOCUS ON...
No. 20
Michel Bataille
Inequalities via Complex Numbers
Introduction
Consider the following famous inequality: If A, B, C, D are four points in the
plane, then AB.CD + BC.AD ≥ AC.BD (Ptolemy’s inequality). A very short
proof uses complex numbers: introducing the affixes a, b, c, d of A, B, C, D, the
equality (b − a)(d − c) + (c − b)(d − a) = (c − a)(d − b) is readily checked. The
familiar properties of the modulus of a complex number (in particular the triangle
inequality) then give
|c − a|.|d − b| = |(b − a)(d − c) + (c − b)(d − a)| ≤ |b − a|.|d − c| + |c − b|.|d − a|
and Ptolemy’s inequality follows at once! This gem of a proof, now well-known,
seems to date back to 1914 ([1]). In this number, we present some results in the
same vein, related to more or less recent problems.

Hayashi’s Inequality
Hayashi’s inequality, although less known, appears in problems from time to time.
To name a couple of recent examples, it is the main argument of the solutions to
the American Mathematical Monthly problem 11536 proposed in November 2010
and to problem OC41 [2011 : 424 ; 2012 : 361]. The inequality can be stated as
follows:
If P is a point in the plane of a triangle ABC, then
PA · PB PB · PC PC · PA
+ + ≥ 1.
CA · CB AB · AC BC · BA
Various identities for complex numbers can be taken as the starting point of the
proof. My favourite one follows from a decomposition in partial fractions which
leads to
1
=
(p − a)(p − b)(p − c)
1 1 1 1 1 1
· + · + · .
(b − a)(c − a) p − a (c − b)(a − b) p − b (a − c)(b − c) p − c
The proof then proceeds by multiplying by (p − a)(p − b)(p − c) and taking moduli
as in the proof of Ptolemy’s inequality above.
Let us connect Hayashi’s inequality to a close one which involves, besides triangle
ABC, two points M, N :
CM · CN AM · AN BM · BN
+ + ≥ 1. (1)
CA · CB AB · AC BC · BA

Crux Mathematicorum, Vol. 42(1), January 2016

 MICHEL BATAILLE /21

The method is similar, the inequality being deduced as above from the identity
(m − a)(n − a) (m − b)(n − b) (m − c)(n − c)
+ + = 1. (2)
(b − a)(c − a) (c − b)(a − b) (a − c)(b − c)
However, the proof of this identity is a good opportunity to apply some results
obtained in a prior Focus On (No 7). Indeed, introducing D(z) = (z − a)(z −
b)(z − c), the left-hand side L of (2) is
(m − a)(n − a) (m − b)(n − b) (m − c)(n − c)
+ +
D0 (a) D0 (b) D0 (c)
that is,
Å ã Å ã
1 1 1 a b c
mn + + − (m + n) + +
D0 (a) D0 (b) D0 (c) D0 (a) D0 (b) D0 (c)
Å 2 ã
a b2 c2
+ + +
D0 (a) D0 (b) D0 (c)
and finally L = mn × 0 − (m + n) × 0 + 1 = 1.
Incidentally, another interesting application of identity (2) is a variant of solution
to problem 2595 ([2000 : 498 ; 2001 : 557]), which offers a case of equality in (1).
Given that M and N are points inside the triangle ABC such that
∠M AB = ∠N AC and ∠M BA = ∠N BC, prove that
AM · AN BM · BN CM · CN
+ + = 1.
AB · AC BC · BA CA · CB

Keeping the notations a, b, c, m, n for the affixes of A, B, C, M, N , the additional

hypothesis on M and N leads to
Å ã Å ã
(m − a)(n − a) (m − b)(n − b)
arg = arg =0
(b − a)(c − a) (c − b)(a − b)
−−→ −−→ −−→ −−→ −−→ −−→ −→ −−→
(since ∠(BM , BA) = ∠(BC, BN ) and ∠(AM , AB) = ∠(AC, AN )).

Copyright
c Canadian Mathematical Society, 2016
22/ FOCUS ON... INEQUALITIES VIA COMPLEX NUMBERS

(m − a)(n − a) (m − b)(n − b)
As a result, both and are positive real numbers.
(b − a)(c − a) (c − b)(a − b)
(m − c)(n − c)
From (2), is a real number and because M and N are interior to
(a − c)(b − c)
−→ −−→ −−→ −−→ (m − c)(n − c)
the triangle, we must have ∠(CA, CN ) = ∠CM , CB), so that is a
(a − c)(b − c)
positive real number as well. Taking moduli in (2) then yields the desired equality.
Note in passing that M, N are isogonal conjugates with respect to ABC.

More examples
Prompted by expressions evoking the modulus of a complex number, one can
sometimes introduce complex numbers advantageously. Here are two examples.
We start with problem 3092, part (a) [2005 : 544,546 ; 2006 : 526]:
Let a, b, and c be positive
√ real numbers
√ such √
that a + b + c = abc. Find
the minimum value of 1 + a2 + 1 + b2 + 1 + c2 .
The statement then referred to the previous problem 2814 of which one of the
featured solutions (by Guersenzvaig) used complex numbers. We can mimic the
method as follows.
√ √ √
Since 1 + a2 + 1 + b2 + 1 + c2 = |1 + ia| + |1 + ib| + |1 + ic| ≥ |3 + i(a + b + c)|
(by the triangle inequality), we have
p p p »
1 + a2 + 1 + b2 + 1 + c2 ≥ 9 + (a + b + c)2 .

1 1 1
Now, since ab + bc + ca = 1 (from a + b + c = abc), the harmonic-arithmetic mean
inequality gives
3
ab + bc + ca ≥ 1 1 1 1

= 9.
3 ab + bc + ca

Thus, (a + b + c)2 = a2 + b2 + c2 + 2(ab + bc + ca) ≥ 3(ab + bc + ca) ≥ 27 and we

finally obtain
p p p
1 + a2 + 1 + b2 + 1 + c2 ≥ 6.
√
Observing that for a = b = c = 3, we have a + b + c = abc and the equality case
in the above inequality, we conclude that the required minimum value is 6.
Another example is provided by problem 3686 that I proposed in 2011 [2011 :
456, 458 ; 2012 : 391].
Let a, b, and c be real numbers such that abc = 1. Show that
Å ã Å ãÅ ãÅ ã
1 1 1 2 1 1 1
a− +b− +c− ≤2 a+ b+ c+ .
a b c a b c

Crux Mathematicorum, Vol. 42(1), January 2016

 MICHEL BATAILLE /23

The problem attracted various methods and three solutions were featured. My
proposed solution rested upon complex numbers: Since abc = 1, we have
Å ãÅ ãÅ ã
1 1 1
2 a+ b+ c+ = 2(a2 + 1)(b2 + 1)(c2 + 1)
a b c
= 2|(a + i)(b + i)(c + i)|2
= 2|(1 − a − b − c) + i(ab + bc + ca − 1)|2

so that
Å ãÅ ãÅ ã
1 1 1
= 2 (a + b + c − 1)2 + (1 − (ab + bc + ca))2 .
 
2 a+ b+ c+
a b c

But, 2(X 2 + Y 2 ) ≥ (X + Y )2 for all real numbers X, Y , hence

Å ãÅ ãÅ ã
1 1 1
2 a+ b+ c+ ≥ (a + b + c − ab − bc − ca)2
a b c
Å ã
1 1 1 2
= a− +b− +c− .
a b c

As usual, we conclude this number with a couple of exercises.

Exercises
1. Prove the identity

vw(v − w) + wu(w − u) + uv(u − v) + (v − w)(w − u)(u − v) = 0

where u, v, w are complex numbers and deduce another proof of Hayashi’s inequal-
ity.
2. Using complex numbers, prove the identity

(b2 + c2 )(c2 + a2 )(a2 + b2 ) = (a2 b + b2 c + c2 a − abc)2 + (ab2 + bc2 + ca2 − abc)2

for real numbers a, b, c. Deduce that if a, b, c are the side lengths of a triangle,
then
2(b2 + c2 )(c2 + a2 )(a2 + b2 ) > (a3 + b3 + c3 )2 .

References
[1] T. Hayashi, Two Theorems on Complex Numbers, Tôhoku Math. Journal,
p. 75–77, (1913/1914).

Copyright
c Canadian Mathematical Society, 2016
24/ GENERATING INEQUALITIES USING SCHWEITZER’S THEOREM

Generating inequalities using

Schweitzer’s theorem
Daniel Sitaru and Claudia Nănuţi
In 1914 P. Schweitzer published a theorem (see [1]) that later featured in the 1978
Russian Olympiad. Romanian mathematician Daniel Culea has since proposed
several applications of Schweitzer’s theorem [2]. In this article, we will present
other applications of this theorem.
Theorem. (Kantorovic) If pk ∈ (0, ∞); k ∈ 1 . . . n; xk ∈ R; 0 < m ≤ xk ≤ M then
n n n
X  X pk  (m + M )2  X 2 (m − M )2 X X 2
pk xk ≤ pk − ·min pi − pj ,
xk 4mM 4mM A
k=1 k=1 k=1 i∈A j∈Bpj

where A ∪ B = {1, 2, · · · , n}; A ∩ B = ∅

Proof. From (xk − m)(xk − M ) ≤ 0 we obtain successively:

x2k − (m + M )xk + mM ≤ 0,
mM
xk + ≤ m + M,
xk
mM
≤ m + M − xk ,
xk
1 m + M − xk
≤ ,
xk mM
pk (m + M )pk − pk xk
≤ , k ∈ 1 . . . n,
xk mM
n n
X pk 1 X 
≤ (m + M )pk − pk xk ,
xk mM
k=1 k=1

n n n n n
X  X pk   X  1  X X 
p k xk ≤ pk xk (m + M ) pk − pk xk . (1)
xk mM
k=1 k=1 k=1 k=1 k=1

We compute the maximum value of the right-hand side from (1). Let xi = m, i ∈
A, xj = M, j ∈ B, A ∩ B = ∅, A ∩ B = {1, 2, ..., n} and α = i∈A pi ; β = j∈B pj .
P P
Then
1  X h X X  X X i
m pi (m + M ) pi + pj − m pi − M pj
mM
i∈B i∈A j∈B i∈A j∈B
P
j∈B pj
X  P
i∈A pi
 X 
= m pi + M pj +
m M
i∈A j∈B
α β  (2mβ + 2M α)(2mα + 2M β)
= (mα + M β) + = =
m M 4mM

Crux Mathematicorum, Vol. 42(1), January 2016

 DANIEL SITARU AND CLAUDIA NĂNUŢI /25

(mα+mβ+M α+M β−mα+mβ+M α−M β)(mα+mβ+M α+M β+mα−mβ−M α+M β)

= 4mM

[(m + M )(α + β) − (m − M )(α − β)][(m + m)(α + β) + (m − M )(α − β)]

=
4mM
(m + M )2 (α + β)2 − (m − M )2 (α − β)2
=
4mM
(m + M )2 (α + β)2 (m − M )2 (α − β)2
= − .
4mM 4mM
The maximum value of the right-hand side from (1) is obtained when (α − β)2 is
minimum, namely when
X X 
min pi − pj , A ∪ B = {1, 2, ..., n}, A ∩ B = ∅.
A
i∈A j∈B

2
Theorem. (Schweitzer) If xk ∈ R; k ∈ 1 . . . n and 0 < m ≤ xk ≤ M then
n n
X  X 1  (m + M )2 n2 (m − M )2 [1 + (−1)n+1 ]
xk ≤ − .
xk 4mM 8mM
k=1 k=1

Proof. In the Kantorovic theorem, let the weights be pk = 1, k ∈ 1 . . . n, x =

|A|, n − x = |B|. It follows that:
n n
X  X 1  (m + M )2 n2 (m − M )2
xk ≤ − · min[4(x2 − nx) + n2 ]
xk 4mM 4mM
k=1 k=1

and
16n2
min(4x2 − 4nx) = − = −n2 .
16
n
The minimum value is reached for x = 2 if n is even when min(4x2 −4nx+n2 ) = 0.
n−1 n
If n is odd, the minimum is reached when x = 2 < 2.
h n − 1 2 n − 1i
min[4(x2 − nx) + n2 ] = 4 −n + n2
2 2
 n2 − 2n + 1 − 2n2 + 2n 
=4 + n2
4
= −n2 + 1 + n2 = 1.

For n even:
n n
X  X 1  (m + M )2 n2
xk ≤ .
xk 4mM
k=1 k=1

For n odd:
n n
X  X 1  (m + M )2 n2 (m − M )2
xk ≤ − .
xk 4mM 4mM
k=1 k=1

Copyright
c Canadian Mathematical Society, 2016
26/ GENERATING INEQUALITIES USING SCHWEITZER’S THEOREM

For n = 2, 3, 4 the inequality becomes, respectively,

1 1  (m + M )2
(x1 + x2 ) + ≤ ,
x1 x2 mM
1 1 1  (m + M )2 · 9 − (m − M )2 m M
(x1 + x2 + x3 ) + + ≤ =5+2 + ,
x1 x2 x3 4mM M m
1 1 1 1  (m + M )2 · 16 m M
(x1 + x2 + x3 + x4 ) + + + ≤ =8+4 + .
x1 x2 x3 x4 4mM M m
Let 0 < a ≤ b, m = a, M = b and x1 , x2 , x3 , x4 ∈ [a, b]. The inequality is
1 1  (a + b)2
(x1 + x2 ) + ≤ , (2)
x1 x2 ab
1 1 1 a b
(x1 + x2 + x3 ) + + ≤5+2 + , (3)
x1 x2 x3 b a
1 1 1 1 a b
(x1 + x2 + x3 + x4 ) + + + ≤8+4 + . (4)
x1 x2 x3 x4 b a
The following inequality is well known:
  …
2a2 b2 2ab √ a+b a2 + b2
0<a≤ 2 2
≤ ≤ ab ≤ ≤ ≤b (5)
a +b a+b 2 2
2
Problem. Prove that if x, y, z, t ∈ [a, b], 0 < a ≤ b then:

x+y+z+t a+b
√ √ √ √ ≤ √ .
xy + yz + zt + tx 2 ab

Proof. From (2) for m = a, M = b, x1 = x, x2 = y it follows that

1  (a + b)2
1
(x + y) ≤
+ ,
x y ab
(x + y)2 (a + b)2
≤
xy ab
ab(x + y) ≤ xy(a + b)2 ,
2
√ √
(x + y) ab ≤ xy(a + b).
√ √ √ √ √
Analogously, (y + z) ab ≤ yz(a + b), (z + t) ab ≤ zt(a + b) and (t + x) ab ≤
√
tx(a + b) and by adding
√ √ √ √ √
2(x + y + z + t) ab ≤ (a + b)( xy + yz + zt + tx),

and we obtain the result. 2

Crux Mathematicorum, Vol. 42(1), January 2016

 DANIEL SITARU AND CLAUDIA NĂNUŢI /27

Problem. In triangle ABC, let M, N, P ∈ [BC]. Prove that

√3
 1 1 1  5 2  AB AC 
AM · AN · AP + + ≤ + + .
AM AN AP 3 3 AC AB
Proof. WLOG we assume that AB < AC. In (3) we take m = AB, M = AC and
then AM, AN, AP ∈ [m, M ]. Let be x1 = AM, x2 = AN, x3 = AP . Then
 1 1 1   AB AC 
(AM + AN + AP ) + + ≤5+2 +
AM AN AP AC AB
From AM-GM inequality, we obtain:
√3
AM + AN + AP ≥ 3 AM · AN · AP .
It follows that
√3
 1 1 1   AB AC 
3 AM · AN · AP + + ≤5+2 + ,
AM AN AP AC AB
√3
 1 1 1  5 2  AB AC 
AM · AN · AP + + ≤ + + .
AM AN AP 3 3 AC AB
2
Problem. Prove that if 0 < a ≤ b then
√ a+b 1 1 2 1 a b
(a + ab + + b) +√ + + ≤8+4 + , (6)
2 a ab a + b b b a
… …
 2ab √ a+b a2 + b2  a + b 1 2 2 
+ ab + + +√ + +
a+b 2 2 2ab ab a + b a + b2
2
a b 
≤8+4 + . (7)
b a
√
Proof. For (6), in (4) take m = a, M = b, x1 = a, x2 = ab, x3 = a+b 2 , x4 = b.
√ » 2 2
2ab
For (7), in (4) take m = a, M = b, x1 = a+b , x2 = ab, x3 = a+b
2 , x4 =
a +b
2 .

References.
1. Matematikai és Fizikai Lapok, Vol. 23, pp. 257-251.
2. Daniel Culea, Commented Problems, Romanian Mathematical Gazzette, A
Series, Nr. 2, 1991, pp. 62-70.
3. Daniel Sitaru, Math Phenomenon, Paralela 45 Publishing House, Piteşti, 2016.
4. Daniel Sitaru, Radu Gologan, Leonard Giugiuc, 300 Romanian Mathematical
Challenges, Paralela 45 Publishing House, Piteşti, 2016.
5. Daniel Sitaru, Claudia Nănuţi, Diana Trăilescu, Leonard Giugiuc, Inequalities,
Ecko-Print Publishing House, Dr. Tr. Severin, 2015.
6. Romanian Mathematical Gazette, A and B series.

Copyright
c Canadian Mathematical Society, 2016
28/ PROBLEMS

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 December
1, 2016, although late solutions will also be considered until a solution is published.
The editor thanks Rolland Gaudet, retired professor of the University College of Saint
Boniface for translations of the problems.

4101. Proposed by Max Alekseyev.

Let n be an integer such that 3n ≡ 7 (mod n). Show that 127 cannot divide n.

4102. Proposed by Kimberly D. Apple and Eugen J. Ionascu.

Suppose the faces of a regular icosahedron are coloured with blue or yellow in such
a way that every blue face shares an edge with at most one other blue face. What
is the maximum possible number of blue faces?

4103. Proposed by Dan Stefan Marinescu, Leonard Giugiuc and Daniel Sitaru.
Let x, y and z be positive numbers such that x + y + z = 1. Show that
X » √
[(1 − x) 3yz(1 − y)(1 − z)] ≥ 4 xyz.
cyc

4104. Proposed by Daniel Sitaru.

Prove that for 0 < a ≤ b ≤ c ≤ d < 2, we have

5(ab4 + bc4 + cd4 + 16d) < 5(b5 + c5 + d5 + 16a) + 128.

4105. Proposed by Mihaela Berindeanu.

Let ABC be a triangle with centroid G. Let A0 , B 0 and C 0 be the feet of altitudes
on the sides of the triangle from the vertices A, B and C, respectively. Let G0 be
the centroid of A0 B 0 C 0 . If GG0 k BC, find all possible values of angle A.

4106. Proposed by D.M. Bătineţu-Giurgiu and Neculai Stanciu.

Let ABC be a triangle with BC = a, AC = b, AB = c and circumradius R. Show
that
b+c c+a a+b 2
+ 5 + 5 ≥ .
a5 b c 3R4

Crux Mathematicorum, Vol. 42(1), January 2016

 PROBLEMS /29

4107. Proposed by Lorian Saceanu.

Let ABC be an acute triangle with inradius r, circumradius R and semiperimeter
s. Prove that
 Å ã
9 r 2 r A B C s
+ + ≤ cos + cos + cos ≤ R .
4 2R R 2 2 2 2 +r

4108. Proposed by Alessandro Ventullo.

a) Write 2010 as a sum of consecutive squares.
b) Is it possible to write 2014 as the sum of several consecutive squares?

4109. Proposed by Mehtaab Sawhney.

Let k and n be positive integers. Compute the following sum in closed form:
k X k Ç åÇ åÇ å
X
k−r k nr k − r k−`
(−1) n .
r=1
r k+` k−`
`=r

4110. Proposed by Michel Bataille.

Find all functions f : R → R such that
Å ã
x
xf (x + y) = (x + y)f f (y)
y
for all real numbers x, y with y 6= 0.

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

4101. Proposé par Max Alekseyev.

Soit n un entier tel que 3n ≡ 7 (mod n). Démontrer que 127 ne peut pas diviser
n.

4102. Proposé par Kimberly D. Apple and Eugen J. Ionascu.

Supposons que les faces d’un icosaèdre sont colorées bleu ou jaune de façon à ce
que toute face bleue partage une arête avec au plus une autre face bleue. Quel est
le nombre maximum possible de faces bleues?

4103. Proposé par Dan Stefan Marinescu, Leonard Giugiuc and Daniel Sitaru.
Soient x, y et z des nombres positifs tels que x + y + z = 1. Démontrer que
X » √
[(1 − x) 3yz(1 − y)(1 − z)] ≥ 4 xyz.
cyc

Copyright
c Canadian Mathematical Society, 2016
30/ PROBLEMS

4104. Proposé par Daniel Sitaru.

Démontrer que si 0 < a ≤ b ≤ c ≤ d < 2, alors la suivante tient
5(ab4 + bc4 + cd4 + 16d) < 5(b5 + c5 + d5 + 16a) + 128.

4105. Proposé par Mihaela Berindeanu.

Soit ABC un triangle avec centroı̈de G. Soient A0 , B 0 et C 0 les pieds des altitudes
émanant de A, B et C respectivement. Soit G0 le centroı̈de de A0 B 0 C 0 . Si GG0 k
BC, déterminer toute valeur possible pour l’angle A.

4106. Proposé par D.M. Bătineţu-Giurgiu and Neculai Stanciu.

Soit ABC un triangle où BC = a, AC = b, AB = c et où R dénote le rayon du
cercle circonscrit. Démontrer que
b+c c+a a+b 2
5
+ 5 + 5 ≥ .
a b c 3R4

4107. Propoé par Lorian Saceanu.

Soit ABC un triangle aigu, où r est le rayon du cercle inscrit, R est le rayon du
cercle circonscrit et s est le demi périmètre. Démontrer que
 Å ã
9 r 2 r A B C s
+ + ≤ cos + cos + cos ≤ R .
4 2R R 2 2 2 2 +r

4108. Proposé par Alessandro Ventullo.

a) Représenter 2010 comme somme de carrés consécutifs.
b) Est-ce possible de représenter 2014 comme somme de plusieurs carrés consécutifs.

4109. Proposé par Mehtaab Sawhney.

Soient k et n des entiers positifs. Déterminer
k X k Ç åÇ åÇ å
X
k−r k nr k − r k−`
(−1) n
r=1
r k+` k−`
`=r

en forme close.

4110. Proposé par Michel Bataille.

Déterminer toutes les fonctions f : R → R telles que
Å ã
x
xf (x + y) = (x + y)f f (y)
y
pour tous nombres réels x, y tels que y 6= 0.

Crux Mathematicorum, Vol. 42(1), January 2016

 SOLUTIONS /31

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 2015: 41(1), p. 27–30.

4001. Proposed by Cristinel Mortici and Leonard Giugiuc.

Let a, b, c, d ∈ R with d > 2 such that
a b c
(2d + 1) · + + = 0.
6 2 d+1
Prove that there exists t ∈ (0, d) such that at2 + bt + c = 0.
We received four correct solutions and one solution that was almost complete. The
first solution is due to Digby Smith and the second consists of ingredients of others.
Solution 1, by Digby Smith.
The result actually holds when d > 1. Let

2v = d + 1 and 3w = 2d + 1.

Then 0 < v < w < d with 6(w − v) = d − 1. The given condition can be rewritten
as
0 = avw + bv + c = v(aw + b) + c.
Let f (t) = at2 + bt + c. Then

wc c(w − v) c(d − 1)
f (w) = aw2 + bw + c = w(aw + b) + c = − +c=− =− .
v v 6v
6 0, then it follows that f (0) = c and f (w) have opposite signs, so that f (t)
If c =
has a real root in the interval (0, w) ⊆ (0, d).
If c = 0, then f (w) = w(aw + b) = 0 since v(aw + b) = 0.

Solution 2.
Use the notation of Solution 1. Again 0 < v < w < d. Furthermore, when
a = 0, f (v) = 0. Otherwise, we may assume that a > 0, in which case f (v) <
avw + bv + c = 0.
When c = 0, then f (w) = 0. When c > 0, then f (0) > 0 and f (t) has a root in
(0, v). Finally, when c < 0, then avw + bv = −c > 0 and

f (w) = aw2 + bw − avw − bv = (w − v)(aw + b) > 0

and f (t) has a root in (v, w).

Copyright
c Canadian Mathematical Society, 2016
32/ SOLUTIONS

4002. Proposed by Henry Aniobi.

Let f be a convex function on an interval I. Let x1 ≤ x2 ≤ . . . ≤ xn and
y1 ≤ y2 ≤ . . . ≤ yn be numbers such that xi + yj is always in I for all 1 ≤ i, j ≤ n.
Let z1 , z2 , . . . , zn be an arbitrary permutation of y1 , y2 , . . . , yn . Show that

f (x1 + y1 ) + . . . + f (xn + yn ) ≥ f (x1 + z1 ) + . . . + f (xn + zn )

≥ f (x1 + yn ) + f (x2 + yn−1 ) + . . . + f (xn + y1 );

We received five submissions of which four were correct and complete. We present
the solution by Joseph DiMuro.
We can prove the above statement by proving the following simpler statement:
Claim. Let x1 < x2 and y1 < y2 be numbers such that xi + yj is always in I. Then

f (x1 + y1 ) + f (x2 + y2 ) ≥ f (x1 + y2 ) + f (x2 + y1 ).

The reason why this suffices: if we choose a permutation z1 , z2 , . . . , zn such that

zi > zj for some i < j, then we will have

f (xi + zi ) + f (xj + zj ) ≤ f (xi + zj ) + f (xj + zi ).

We would then be able to interchange zi and zj without decreasing the overall

sum. Thus, a permutation z1 , z2 , . . . , zn that gives us the largest overall sum is
one where zi ≤ zj whenever i < j; that is, zi = yi for all i. Similarly, a permutation
z1 , z2 , . . . , zn that gives us the smallest overall sum is one where zi ≥ zj whenever
i < j; that is, zi = yn−i+1 for all i.
Proof of claim. By the definition of convexity, for all a, b ∈ I and for all t ∈ [0, 1],
we have
f (ta + (1 − t)b) ≤ tf (a) + (1 − t)f (b).
Let a = x1 + y1 and b = x2 + y2 . Then a < x1 + y2 < b, so for some t ∈ [0, 1], we
have x1 + y2 = ta + (1 − t)b. From that, we have:

x2 + y1 = (x1 + y1 + x2 + y2 ) − (x1 + y2 )
= (a + b) − (ta + (1 − t)b)
= (1 − t)a + tb.

Therefore,

f (x1 + y2 ) + f (x2 + y1 ) = f (ta + (1 − t)b) + f ((1 − t)a + tb)

≤ (tf (a) + (1 − t)f (b)) + ((1 − t)f (a) + tf (b))
= f (a) + f (b)
= f (x1 + y1 ) + f (x2 + y2 ),

completing the proof of the claim.

Crux Mathematicorum, Vol. 42(1), January 2016

 SOLUTIONS /33

4003. Proposed by Martin Lukarevski.

Show that for any triangle ABC, the following inequality holds
Å ã
1 1 1
sin A sin B sin C + +
sin A + sin B sin B + sin C sin C + sin A
3
≤ (cos A + cos B + cos C).
4

We received 13 correct solutions. We present the solution by John G. Heuver,

modified slightly by the editor.
Let r, R and s denote the inradius, the circumradius and the semiperimeter of
4ABC, respectively. The following identities and inequalities are well known:
X s2 − 4Rr − r2
sin2 A = , (1)
2R2
X s2 + 4Rr + r2
sin A sin B = , (2)
4R2
X R+r
cos A = , (3)
R
R ≥ 2r Euler’s inequality (4)
2 2 2
s ≤ 4R + 4Rr + 3r Gerretsen’s inequality, (5)

(where all the summations are taken over all angles of 4ABC).
Let L denote the left-hand√side of the given inequality. By the AM-GM Inequality,
we have sin A + sin B ≥ 2 sin A sin B. Hence, by Cauchy-Schwarz Inequality we
have
1X √ 1 X 2
q qX
L≤ (sin A) sin B sin C ≤ sin A sin A sin B. (6)
2 2
By (1) and (5), we have
X 4R2 + 2r2
sin2 A ≤ . (7)
2R2
By (2) and (5), we have
X 4R2 + 8Rr + 4r2
sin A sin B ≤ . (8)
4R2
Using (6), (7) and (8) followed by (3) and (4), we then have
… … …
1 4R2 + 2r2 4R2 + 8Rr + 4r2 1  r 2 R + r
L≤ 2
· 2
= 2+ ·
2 2R 4R 2 R R
  Å ã2 X
1 1
≤ 2+ · cos A
2 2
3X
= cos A,
4

Copyright
c Canadian Mathematical Society, 2016
34/ SOLUTIONS

which completes the proof.

Editor’s comment. Digby Smith remembered that the following problem proposed
by Jack Garfunkel and George Tsintsifas appeared in the August–September 1982
issue (Vol. 8, no. 7, p. 210) of Crux and a solution given by Vedula N. Murty
appeared in the November 1983 issue (Vol. 9, no. 9, p. 282):

4X Y B−C 2X
sin B sin C ≤ cos ≤ cos A.
9 2 3
Smith gave a proof by first showing that 2L ≤
P
sin B sin C, which together with
the above inequality yields the result.

4004. Proposed by George Apostolopoulos.

Let x, y, z be positive real numbers such that x + y + z = 2. Prove that

x5 y5 z5
+ + ≥ 1.
yz(x2 + y 2 ) zx(y 2 + z 2 ) xy(z 2 + x2 )

We received 16 correct submissions. We present 3 solutions.

Solution 1, by Arkady Alt.
Since by Cauchy’s Inequality

2
X x5 X x6 x3 + y 3 + z 3
= ≥ P ,
cyc
yz (x2 + y 2 ) cyc
xyz (x2 + y 2 ) xyz (x2 + y 2 )
cyc

it suffices to prove the inequality

2
x3 + y 3 + z 3
≥ 1.
xyz (x2 + y 2 )
P
cyc

We have the following equivalences:

2
x3 + y 3 + z 3
2
x3 + y 3 + z 3 ≥ 2xyz x2 + y 2 + z 2


≥1 ⇐⇒
xyz (x2 + y 2 )
P
cyc

2
x3 + y 3 + z 3 ≥ xyz (x + y + z) x2 + y 2 + z 2 ,


⇐⇒

where the latter inequality holds because by AM-GM Inequality

x3 + y 3 + z 3 ≥ 3xyz

and by Chebyshev’s Inequality

(x + y + z) x2 + y 2 + z 2


3 3 3
x +y +z ≥ .
3

Crux Mathematicorum, Vol. 42(1), January 2016

 SOLUTIONS /35

Solution 2, by Michel Bataille.

y
Let a = x2 , b = 2 and c = z2 . With these notations, we are required to prove

a6 b6 c6 abc
2 2
+ 2 2
+ 2 2
≥ (1)
a +b b +c c +a 2

under the conditions a, b, c > 0 and a + b + c = 1.

The Cauchy-Schwarz inequality gives
Å ã
a6 b6 c6
2 2
+ 2 2
+ 2 ((a2 + b2 ) + (b2 + c2 ) + (c2 + a2 )) ≥ (a3 + b3 + c3 )2 .
a +b b +c c + a2

Hence, (1) will follow if we prove

(a3 + b3 + c3 )2
≥ abc.
a2 + b2 + c2

a3 +b3 +c3
Since abc ≤ 3 , it is sufficient to show that

3(a3 + b3 + c3 ) ≥ a2 + b2 + c2 .

Now, the latter follows from

3(a3 + b3 + c3 ) ≥ 2(a3 + b3 + c3 ) + 3abc

= a3 + b3 + c3 + (a3 + b3 + c3 + 3abc)
≥ a3 + b3 + c3 + ab2 + a2 b + bc2 + b2 c + ca2 + c2 a (Schur’s ineq.)
2 2 2 2 2 2
= (a + b + c )(a + b + c) = a + b + c (since a + b + c = 1)

so we are done.

Solution 3, by Oliver Geupel.

By hypothesis x + y + z = 2 and by the Cauchy-Schwarz inequality we have
Å ã
x5 y5 z5
+ + (x + y + z)xyz(x2 + y 2 + z 2 )
yz(x2 + y 2 ) zx(y 2 + z 2 ) xy(z 2 + x2 )
! !
X x5 X
= 2 + y2 )
xyz(x + y ) ≥ (x3 + y 3 + z 3 )2 .
2 2

cyc
yz(x cyc

By the power mean inequality, it holds

Å ã1/3 Å ã1/2
x3 + y 3 + z 3 x2 + y 2 + z 2 x+y+z
≥ ≥ ≥ (xyz)1/3 .
3 3 3

Copyright
c Canadian Mathematical Society, 2016
36/ SOLUTIONS

Putting together we obtain

x5 y5 z5
+ +
yz(x2 + y 2 ) zx(y 2 + z 2 ) xy(z 2 + x2 )
(x3 + y 3 + z 3 )2
≥
(x + y + z)xyz(x2 + y 2 + z 2 )
(x3 + y 3 + z 3 )1/3 x3 + y 3 + z 3 (x3 + y 3 + z 3 )2/3
= · ·
x+y+z xyz x2 + y 2 + z 2
≥ 3−2/3 · 3 · 3−1/3 = 1.
Hence the result. By the equality condition of the power mean inequality, the
equality holds if and only if x = y = z = 2/3.

4005. Proposed by Michel Bataille.

Let a, b, c be the sides of a triangle with area F . Suppose that some positive real
numbers x, y, z satisfy the equations
x + y + z = 4 and
Å ã
4 − yz 4 4 − zx 4 4 − xy 4
2xb2 c2 + 2yc2 a2 + 2za2 b2 − a + b + c = 16F 2 .
x y z
Show that the triangle is acute and find x, y, z.
We present the proposer’s solution — no others were submitted.
The second equation gives
(xyz)(16F 2 )
= xyz(2xb2 c2 + 2yc2 a2 + 2za2 b2 ) − yz(4 − yz)a4 − zx(4 − zx)b4 − xy(4 − xy)c4
= (a2 yz + b2 zx + c2 xy)2 − (4a4 yz + 4b4 zx + 4c4 xy)
x y z 2
= (b2 z + c2 y) + (c2 x + a2 z) + (b2 x + a2 y) − (4a4 yz + 4b4 zx + 4c4 xy).
2 2 2
Since t 7→ t2 is a convex function and x + y + z = 4, Jensen’s inequality yields
x 2 y z
(b z + c2 y)2 + (c2 x + a2 z)2 + (b2 x + a2 y)2
4 4 4
x y z 2
≥ (b z + c y) + (c2 x + a2 z) + (b2 x + a2 y)
2 2
(1)
4 4 4
and it follows that
(xyz)(16F 2 )
≤ x(b2 z + c2 y)2 + y(c2 x + a2 z)2 + z(b2 x + a2 y)2 − (4a4 yz + 4b4 zx + 4c4 xy)
= a4 yz(y + z − 4) + b4 zx(z + x − 4) + c4 xy(x + y − 4) + xyz(2b2 c2 + 2c2 a2 + 2a2 b2 )
= xyz(2b2 c2 + 2c2 a2 + 2a2 b2 − a4 − b4 − c4 )
= (xyz)(16F 2 ).

Crux Mathematicorum, Vol. 42(1), January 2016

 SOLUTIONS /37

Thus, equality must hold in (1) and because t 7→ t2 is a strictly convex function,
this calls for
b2 z + c2 y = c2 x + a2 z = b2 x + a2 y.
Setting these three expressions equal to λ and solving for x, y, z yields

b2 + c2 − a2 λ cos A
x=λ 2 2
= ,
2b c bc
c2 + a2 − b2 λ cos B
y=λ = ,
2c2 a2 ca
a2 + b2 − c2 λ cos C
z=λ = .
2a2 b2 ab
(As usual, A, B, C denote the angles of the triangle opposite sides a, b, c, respec-
tively.) Since at most one of A, B, C is not acute and x, y, z are positive, we
conclude that cos A, cos B, cos C, and λ are positive. Thus, the triangle is acute.
In addition, we have

λ cos A λ cos B λ cos C

4= + + .
bc ca ab
2F
Since a cos A+b cos B +c cos C = and 4RF = abc (where R is the circumradius
R
a2 b2 c2
of the triangle), we readily find λ = and obtain
2F 2

a2 (b2 + c2 − a2 ) b2 (c2 + a2 − b2 ) c2 (a2 + b2 − c2 )

x= , y= , z= .
4F 2 4F 2 4F 2
Note that conversely, if given an acute triangle, then these numbers x, y, z are
positive and satisfy the two equations: x + y + z = 4 is readily checked; also we
have b2 z + c2 y = c2 x + a2 z = b2 x + a2 y = λ, hence the calculations made at the
beginning (with equality in (1)) show that the second equation holds as well.

4006. Proposed by Dragolijub Milošević.

Let x, y, z be positive real numbers such that xyz = 1. Prove that

2 1 1
− ≤ .
xy + yz + zx x + y + z 3

We received 15 correct solutions from 14 submitters. Ten of these solutions were

along the lines of the solution presented below, with variations in how they justified
the ancillary inequalities and how straightforwardly they handled the algebra. In
addition, there was a MAPLE-based solution, which seemed heavy-handed for this
problem. There were four other solutions that were defective in some way. We
present the solution by Henry Ricardo.

Copyright
c Canadian Mathematical Society, 2016
38/ SOLUTIONS

Let p = x + y + z, q = xy + yz + zx and r = xyz = 1. Observe that q 2 ≥ 3rp = 3p,

since, by the arithmetic-geometric means inequality,
1 2 2 1 1
q2 = (x y + y 2 z 2 ) + (y 2 z 2 + z 2 x2 ) + (z 2 x2 + x2 y 2 ) + 2xyz(x + y + z)
2 2 2
≥ xy 2 z + yz 2 x + zx2 y + 2xyz(x + y + z)
= 3xyz(x + y + z) = 3rp = 3p.

The difference between the two sides of the inequality is one-third of

Å ã
6 3 6 9 3 2
1− + ≥1− + 2 = 1− ≥ 0,
q p q q q
and the result follows with equality if and only if x = y = z = 1.
Editor’s comment. Oliver Geupel notes that this problem is equivalent to a prob-
lem proposed by Vasile Cı̂rtoaje and Mircea Lascu for the Junior TST 2003 Ro-
mania. It is also Problem 72 in Chapter 20 of Inequalities, Theorems, Techniques
and Selected Problems by Zdravko Cvetkovski (Springer, 2012).

4007. Proposed by Mihaela Berindeanu.

Show that for any numbers a, b, c > 0 such that a2 + b2 + c2 = 12, we have

(a3 + 4a + 8)(b3 + 4b + 8)(c3 + 4c + 8) ≤ 243 .

We received nine submission of which eight were correct and complete. We present
two solutions.
Solution 1, by Ángel Plaza.
By taking logarithms, the proposed inequality may be written as
ln a3 + 4a + 8 + ln b3 + 4b + 8 + ln c3 + 4c + 8




≤ ln 24.
3
Changing variables a2 = x, b2 = y, c2 = z the problem becomes:
For any x, y, z > 0 such that x + y + z = 12, prove that

ln x3/2 + 4x1/2 + 8 + ln y 3/2 + 4y 1/2 + 8 + ln z 3/2 + 4z 1/2 + 8

≤ ln 24.
3
Let us consider function f (x) = ln x3/2 + 4x1/2 + 8 for x > 0. Then

√
00 −8x3/2 − 3x5/2 + 12x − 16 x − 16
f (x) = √
2
2x3/2 x3/2 + 4 x + 8

and since f 00 (x) < 0 for x > 0, the function f is concave. By Jensen’s inequality
f (x) + f (y) + f (z) x + y + z 
≤f = f (12/3) = f (4) = ln 24.
3 3

Crux Mathematicorum, Vol. 42(1), January 2016

 SOLUTIONS /39

Solution 2, by the proposer.

Observe that (a − 2)4 ≥ 0 implies that a4 − 8a3 + 24a2 − 32a + 16 ≥ 0, that is

a4 + 24a2 + 80 ≥ 8a3 + 32a + 64,

which gives
(a2 + 4)(a2 + 20) ≥ 8(a3 + 4a + 8)
and hence
(a2 + 4)(a2 + 20)
a3 + 4a + 8 ≤ .
8
So,

(a3 + 4a + 8)(b3 + 4b + 8)(c3 + 4c + 8)

(a2 + 4)(a2 + 20) (b2 + 4)(b2 + 20) (c2 + 4)(c2 + 20)
≤ · · ,
8 8 8
√ x+y+z
but we know that 3 xyz ≤ , therefore
3
Å 2 ã3
a + b2 + c2 + 12
(a2 + 4)(b2 + 4)(c2 + 4) ≤ = 83
3

and Å ã3
2 2 2 a2 + b2 + c2 + 60
(a + 20)(b + 20)(c + 20) ≤ = 243 .
3
Finally,

83 · 243
a3 + 4a + 8 b3 + 4b + 8 c3 + 4c + 8 ≤ = 243 .



83

Editor’s Comments. Ángel Plaza sent two solutions: the second solution consists
in taking the logarithms of the given inequality, setting a2 = x, b2 = y, c2 = z,
considering the concave function (on x ∈ (0, 12)) f (x) = ln(x3/2 + 4x1/2 + 8)
and using Jensen’s Inequality. A very similar approach was also used by Šefket
Arslanagić.

4008. Proposed by Mehmet Şahin.

Let ABC be a triangle with ∠ACB = 2α, ∠ABC = 3α, AD is an altitude and
AE is a median such that ∠DAE = α. If |BC| = a, |CA| = b, |AB| = c, prove
that …  
a c 2
=1+ 2 − 1.
b b

We received 15 correct solutions and one incorrect submission. We present the

solution given by Titu Zvonaru, modified slightly by the editor.

Copyright
c Canadian Mathematical Society, 2016
40/ SOLUTIONS

a
We have AD = c sin 3α, BD = c cos 3α, so DE = − cos 3α. By the law of sines,
2
a c c sin 5α
we have = , so a = . Then
sin (180◦ − 5α) sin 2α sin 2α
a
sin α DE 2 − cos 3α sin 5α − 2 sin 2α cos 3α
= tan α = = =
cos α AD c sin 3α 2 sin 2α sin 3α
sin 5α − (sin 5α + sin (−α))
=
2 sin 2α sin 3α
sin α
= ,
2 sin 2α sin 3α

so that
cos α = 2 sin 2α sin 3α = cos α − cos 5α,

which implies that

cos 5α = 0, so 5α = 90◦ , or α = 18◦ .

◦ ◦ ◦ ◦
Hence, ∠BAC = 180 √ − 5α = 90 , ∠ABC = 3α = 54 and ∠ACB = 2α = 36 .
1+ 5
Since cos 36◦ = , we have
4
Ç √ å
1+ 5
b = a cos 2α = a,
4

so
  √   √
p 3 + 5 5− 5
c= a2 − b2 = 2
a − 2
a =a .
8 8
a 4 √
Now, = √ = 5 − 1 and
b 1+ 5
Ç √ åÅ ã2 √ √
 c 2 5− 5 4 14 − 6 5 7−3 5
2 −1=2 √ −1= √ = √ .
b 8 1+ 5 6+2 5 3+ 5

Crux Mathematicorum, Vol. 42(1), January 2016

 SOLUTIONS /41

Therefore, we have the following equivalences:

…  
a c 2
=1+ 2 −1
b b
s √
√ 7−3 5
⇐⇒ 5−1=1+ √
3+ 5
√
√ 7−3 5
⇐⇒ ( 5 − 2)2 = √
3+ 5
√ √ √
⇐⇒ (9 − 4 5)(3 + 5) = 7 − 3 5
√ √
⇐⇒ 7 − 3 5 = 7 − 3 5,

which is true and our proof is complete.

4009. Proposed by George Apostolopoulos.

Let ma , mb , mc be the lengths of the medians of a triangle ABC. Prove that

1 1 1 R
+ + ≤ 2,
ma mb mc 2r

where r and R are inradius and circumradius of ABC, respectively.

We received eleven solutions, of which ten were correct. We present two solutions.

Solution 1, by Arkady Alt.

Let F, s and ha , hb , hc be the area, semiperimeter, and altitudes of the triangle.

Since mx ≥ hx , x ∈ {a, b, c} and

1 1 1 a b c s 1
+ + = + + = =
ha hb hc 2F 2F 2F 2F r

then
1 1 1 1 1 1 1 R
+ + ≤ + + = ≤ 2
ma mb mc ha hb hc r 2r
because
1 R
≤ 2 ⇐⇒ 2r ≤ R,
r 2r
by Euler’s Inequality.

Solution 2, by Edmund Swylan.

We take it as known that the triangle with side lengths 2ma , 2mb , 2mc has medians
of lengths 23 a, 32 b, 32 c. (See the drawing below.)

Copyright
c Canadian Mathematical Society, 2016
42/ SOLUTIONS

Let the area of 4ABC be F . The area of the big triangle is then 3F . Let the
altitudes of the big triangle be Ha , Hb , Hc .
6F 3
We have that = Hx and Hx ≤ x, for each x ∈ {a, b, c}. Therefore,
2mx 2

1 1 1 3
3F ( + + ) ≤ (a + b + c);
ma mb mc 2

equality occurs if and only if the big triangle, and consequently 4ABC too, is
equilateral. Finally,

3 1 1R R
(a + b + c) = 3F ≤ 3F = 3F 2 ;
2 r r 2r 2r

equality occurs if and only if 4ABC is equilateral.

4010. Proposed by Ovidiu Furdui.

Let f : [0, π2 ] → R be a continuous function. Calculate

π Å ã2n
cos x − sin x
Z 2
lim n f (x)dx.
n→∞ 0 cos x + sin x

There were eight submitted solutions for this problem, all of which were correct.
We present two solutions.

Solution 1, by the group of M. Bello, M. Benito, Ó. Ciaurri, E. Fernández, and

L. Roncal, expanded slightly by the editor.
1  π 
The value of the required limit is f (0) + f . Indeed, if we denote by L
4 2
the limit, then from the identity

cos(x) − sin(x) π 
= tan −x ,
cos(x) + sin(x) 4

Crux Mathematicorum, Vol. 42(1), January 2016

 SOLUTIONS /43

we have
Z π/2  π 2n
L = lim n tan −x f (x) dx
n→∞ 0 4
Z π/4 π 
= lim n (tan(s))2n f − s ds
n→∞ −π/4 4
Z π/4  π  π 
= lim n (tan(s))2n f −s +f +s ds
n→∞ 0 4 4
Z 1 1+1/n  
r π 1/n
 π
1/n

= lim f − arctan(r ) + f + arctan(r ) dr,
n→∞ 0 1 + r 2/n 4 4
where we have used symmetry, and in the last step we have used the change of
variable r = (tan(s))n .
Since f is a continuous function, ∃M such that |f (x)| ≤ M , for x ∈ [0, π2 ], and
 
 r1+1/n   π  π 
1/n 1/n
f − arctan(r ) + f + arctan(r )  ≤ M
 
 1 + r2/n 4 4



for all r ∈ [0, 1], using the bound for f and that the fraction in r is bounded above
by r/(1 + r2 ) (which is bounded by 1/2, by looking at (r − 1)2 ≥ 0). In this way,
we can apply the dominated convergence theorem to obtain
Z 1
r1+1/n   π 1/n
 π
1/n

L= lim 2/n
f − arctan(r ) + f + arctan(r ) dr
0 n→∞ 1 + r 4 4
1  π  1
Z
1  π 
= f (0) + f r dr = f (0) + f .
2 2 0 4 2

Solution 2, by Michel Bataille.

We show that the required limit is f (0)+f4 (π/2) . Let
Z π2 Å ã Z π2 
cos x − sin x 2n π 2n
In = f (x) dx = tan −x f (x) dx.
0 cos x + sin x 0 4
The change of variables x = π4 − tan−1 (y) yields
Z 1
y 2n π
−1

In = 2
f − tan (y) dy.
−1 1 + y 4
But we have
Z 0 Z 1
y 2n π
−1
 u2n π 
2
f − tan (y) dy = 2
f + tan−1 (u) du
−1 1+y 4 0 1+u 4
so that Z 1
In = y 2n g(y) dy,
0

Copyright
c Canadian Mathematical Society, 2016
44/ SOLUTIONS

π
− tan−1 (y) + f π4 + tan−1 (y) · y21+1 . It is known that if g




where g(y) = f 4
R1
is continuous on [0, 1], then lim n 0 xn g(x) dx = g(1) [for completeness, a quick
n→∞
proof is given at the end]. From this result, it follows that

f (0) + f (π/2)
lim (2n) · In = g(1) =
n→∞ 2
and so
f (0) + f (π/2)
lim n · In = ,
n→∞ 4
as claimed.
For the proof of the property used above, let  > 0. Using the continuity of g, we
choose δ ∈ (0, 1) such that |g(x) − g(1)| ≤  whenever x ∈ [δ, 1]. Then we have
 Z 1   Z 1 
   
n n
(n + 1) x · g(x) dx − g(1) = (n + 1) x · (g(x) − g(1)) dx
   
 0   0 
Z δ Z 1
≤ (n + 1) xn |g(x) − g(1)| dx + (n + 1) xn |g(x) − g(1)| dx
0 δ
≤ M · δ n+1 + 

where M denotes the maximum of the continuous function x 7→ |g(x) − g(1)| on

R1
[0, 1]. Since 0 < δ < 1, we deduce lim sup |(n + 1) 0 xn · g(x) dx − g(1)| ≤ . Since
n→∞
the latter holds for any positive , we must have
Ç Z 1 å
n
lim (n + 1) x · g(x) dx − g(1) = 0,
n→∞ 0

and the result follows.

Editor’s Comments. This type of problem has its roots in Fourier analysis, where
we are interested in limits such as the one in the problem statement. This particu-
lar limit picks out half the arithmetic mean of the function’s value at the endpoints
of the interval [0, π2 ]; more classical Fourier analysis will focus on limits which pick
out the function’s value at a specific point, like the ‘Dirac delta’ distribution. All
solutions aside from Solution 1 essentially followed Bataille’s Solution 2, including
the proposer’s; À. Plaza’s solution used a limit result from the proposer’s own
book (O. Furdui, Limits, Series and Fractional Part Integrals, Springer, Second
ed., 2013) to skip a substitution. The techniques in both solutions (i.e. utiliz-
ing dominated convergence in a clever way, and separating an integral up into two
parts which are handled using the two different functions involved in the integrand)
are common techniques in classical analysis. A. Stadler’s approach (namely, using
‘Big O’ notation instead of more precise estimates) is equally successful and is
common in analytic number theory.

Crux Mathematicorum, Vol. 42(1), January 2016

 AUTHORS’ INDEX /45

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

Proposers
Max Alekseyev, George Washington University, Washington, DC, USA : 4101
Kimberly D. Apple and Eugen J. Ionascu, Columbus State University, Columbus,
GA, USA : 4102
Michel Bataille, Rouen, France: 4110
Mihaela Berindeanu, Bucharest, Romania : 4105
Dan Stefan Marinescu, Leonard Giugiuc and Daniel Sitaru, Romania : 4103
Daniel Sitaru, Colegiul National Economic Theodor Costescu, Drobeta Turnu - Severin,
Romania : 4104
D.M. Bătineţu-Giurgiu and Neculai Stanciu, Romania : 4106
Lorian Saceanu, Norway : 4107
Alessandro Ventullo, Milan, Italy : 4108
Mehtaab Sawhney, Commack High School, Commack, NY, USA : 4109

Solvers - individuals
Arkady Alt, San Jose, CA, USA : 4001, 4003, 4004, 4006, 4008, 4009
Henry Aniobi, University of Waterloo, Canada : 4002
George Apostolopoulos, Messolonghi, Greece : 4003, 4004, 4006, 4008, 4009
Šefket Arslanagić, University of Sarajevo, Sarajevo, Bosnia and Herzegovina : CC151,
CC155, OC202, 4003, 4004, 4006 (2 solutions), 4007, 4008, 4009
Fernando Ballesta Yagũe, student, I.E.S. Infante don Juan Manuel, Murcia,
Spain : CC151
Ricardo Barroso Campos, University of Seville, Seville, Spain : 4008
Michel Bataille, Rouen, France : OC202, 4003, 4004, 4005, 4006, 4007, 4008, 4010
Mihaela Berindeanu, Bucharest, Romania : 4007
John Carter, Missouri State University, MO, USA : CC151
Matei Coiculescu, East Lyme High School, East Lyme, CT, USA : CC151
Prithwijit De, Homi Bhabha Centre for Science Education, Mumbai, India : 4008
Joseph DiMuro, Biola University, La Mirada, CA, USA : 4002
Marian Dincǎ, Bucharest, Romania : 4002, 4006
Andrea Fanchini, Cantu, Italy : CC151, OC202, 4003, 4004, 4009
Ovidiu Furdui, Technical University of Cluj-Napoca, Romania : 4010
Oliver Geupel, Brühl, NRW, Germany : OC201, OC203, OC205, 4003, 4004,
4006, 4010
Leonard Giugiuc, Drobeta Turnu Severin, Romania : 4010
John G. Heuver, Grande Prairie, AB : 4003
Dag Jonsson, Uppsala, Sweden : 4008
Kee-Wai Lau, Hong Kong, China : 4003, 4006
Martin Lukarevski, University “Goce Delcev”-Stip, Macedonia : 4003, 4009
Salem Malikić, student, Simon Fraser University, Burnaby, BC : 4004, 4006
David E. Manes, SUNY Oneonta, Oneonta, NY : CC151
Dragoljub Milošević, Gornji Milanovac, Serbia : 4006
Madhav R. Modak, formerly of Sir Parashurambhau College, Pune, India : 4001

Copyright
c Canadian Mathematical Society, 2016
46/ AUTHORS’ INDEX

Richard Peiró, i Estruch IES “Abastos” Valencia, Spain : CC151, 4008

Paolo Perfetti, Dipartimento di Matematica, Università degli studi di Tor Vergata Roma,
Rome, Italy : 4004, 4007, 4009, 4010
Ángel Plaza, University of Las Palmas de Gran Canaria, Spain : 4007 (2 solutions), 4010
C.R. Pranesachar, Indian Institute of Science, Bangalore, India : 4003, 4004
Cao Minh Quang, Nguyen Binh Khiem High School, Vinh Long, Vietnam : 4004, 4006
Henry Ricardo, Tappan, NY, USA : 4004, 4006
Wolfgang Remmel, Vienna : 4004, 4006
Mehmet Şahin, Ankara, Turkey : 4008
Cristóbal Sánchez-Rubio, I.B. Penyagolosa, Castellón, Spain : 4008
Joel Schlosberg, Bayside, NY, USA : 4002, 4007, 4008
Digby Smith, Mount Royal University, Calgary, AB : CC151, CC153, OC202, 4001,
4003, 4004, 4007, 4008, 4009
Albert Stadler, Herrliberg, Switzerland : CC151, 4004, 4006, 4009, 4010
Edmund Swylan, Riga, Latvia : 4003, 4007, 4008, 4009
Konstantine Zelator, University of Pittsburgh, Pittsburgh, PA, USA : CC151, CC154
Titu Zvonaru, Cománeşti, Romania : CC151, CC153, CC155, OC202, OC204, 4001,
4003, 4004, 4006, 4008, 4009

Solvers - collaborations
Dionne Bailey, Elsie Campbell, and Charles R. Diminnie, Angelo State University,
San Angelo, USA : 4004
M. Bello, M. Benito, Óscar Ciaurri, E. Fernández, L. Roncal, Logroño, Spain : 4010
Cristinel Mortici and Leonard Giugiuc, Romania : 4001
Missouri State University Problem Solving Group, Missouri State University, Springfield,
MO, USA : CC155, OC202, 4009
Skidmore College Problem Group, Skidmore College, Saratoga Springs, New York,
USA : 4008

Crux Mathematicorum, Vol. 42(1), January 2016