Crux Mathematicorum

VOLUME 42, NO. 3 March / Mars 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

Amanda Malloch 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

Copyright c Canadian Mathematical Society, 2016

 IN THIS ISSUE / DANS CE NUMÉRO

95 Editorial Kseniya Garaschuk

96 The Contest Corner: No. 43 John McLoughlin
96 Problems: CC211–CC215
99 Solutions: CC161–CC165
102 The Olympiad Corner: No. 341 Carmen Bruni
102 Problems: OC271–OC275
104 Solutions: OC211–OC215
109 Focus On . . . : No. 21 Michel Bataille
114 The use of coordinate systems before Descartes Florin Diacu
121 Problems: 4121–4130
126 Solutions: 4021–4030
139 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
 95

EDITORIAL
“All our knowledge has its origins in our perceptions.”
Leonardo da Vinci
If you asked me what was my least favourite class as an undergrad, I would tell you
Multivariate Calculus, hands down. All I remember doing in that course is com-
puting double and triple integrals, switching the order of integration, computing
numerous partial derivatives and memorizing everything about quadric surfaces.
The class was technical, there seemed no apparent reason for anything to be solved
the way we were asked to solve it and the methods used were nothing short of ar-
bitrary. (Of course, it didn’t help that the class was held at 6-8pm twice a week.)
That was my view of the course an as undergrad. Although I still did well in the
course and actually enjoyed the next course in Vector Calculus, I knew this branch
of mathematics was not for me. So imagine my mixed feelings when I was asked
to teach Multivariate Calculus last year.
I said yes. It was time to face my fears, so to speak. I expected to like this course
better than I did as an undergrad (low bar for comparison), but I didn’t expect
to like it quite as much. So I wondered, why was my perception of the material so
different this time around? Was it because I was not taking this course for credit
and could actually spend time enjoying the math as opposed to being stressed out
about it? Was it because I now had friends whose research involved multivariate
calculus tools and so it seemed more personally relevant? Was it because I was
more mathematically mature and could see more connections within the material
itself?
I have experienced similarly diverse feelings with various other things that I have
encountered as a young student and later as a more developed mathematician; this
includes contest and olympiad problems. While I did participate in these events
and did well in them, I can’t actually say I ever truly enjoyed them. And I most
definitely enjoy them now!
Whatever the reason for my changed appreciation for Multivariate Calculus or
math competition problems, it has taught me to give a second chance to things I
have made up my mind about. So I urge you to keep an open mind about various
types of problems or even whole areas of mathematics that you at some point
dismissed as not interesting to you. If you are a fan of inequalities, give geometry
problems a chance. If you normally stick to the Problems section of Crux, take a
look at our Contest Corner. You might be surprised by the math you discover.
(For a piece that inspired this Editorial, see “Changing the way we think about
mathematical ability” by Caroline Junkins, CMS Notes, September 2016.)
Kseniya Garaschuk

Copyright c Canadian Mathematical Society, 2016

96/ THE CONTEST CORNER

THE CONTEST CORNER

No. 43
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 January
1, 2017, 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.

CC211. A rectangular sheet of paper whose dimensions are 12 × 18 is folded

along a diagonal, which creates the M -shaped region drawn at the right. Find the
area of the shaded region.

CC212. A cube that is one inch wide has had its eight corners shaved
off. The cube’s vertices have been replaced by eight congruent equilateral tri-
angles, and the square faces have been replaced by six congruent octagons. If the
combined area of the eight triangles equals the area of one of the octagons, what
is that area? (Each octagonal face has two different edge lengths that occur in
alternating order.)

CC213. A pyramid is built from solid unit cubes that are stacked in square
layers. The top layer has 1 × 1 = 1 cube, the second 3 × 3 = 9 cubes and the layer
below that has 5 × 5 = 25 cubes, and so on, with each layer having two more cubes

Crux Mathematicorum, Vol. 42(3), March 2016

 THE CONTEST CORNER /97

on a side than the layer above it. The pyramid has a total of 12 layers. Find the
exposed surface area of this solid pyramid, including the bottom.

CC214. The points (2, 5) and (6, 5) are two of the vertices of a regular
hexagon of side length two on a coordinate plane. There is a line L that goes
through the point (0, 0) and cuts the hexagon into two pieces of equal area. What
is the slope of line L?

CC215. Each circle in this tree diagram is to be assigned a value, chosen from
a set S, in such a way that along every pathway down the tree the assigned values
never increase. That is, A ≥ B, A ≥ C, C ≥ D, C ≥ E and A, B, C, D, E ∈ S. (It
is permissible for a value in S to appear more than once.) How many ways can
the tree be so numbered using only values chosen from the set S = {1, . . . , 6}?

(Optional extension: Generalize to a case with S = {1, 2, 3, , . . . , n} by finding an

explicit algebraic expression for the number of ways the tree can be numbered.)

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

CC211. Une feuille de papier rectangulaire de taille 12 × 18 est pliée le

long de la diagonale, formant ainsi une région en forme de M , telle qu’illustrée.
Déterminer la surface de la région ombragée.

CC212. On a retranché les huit coins d’un cube dont les côtés mesurent cha-
cun un pouce. Les sommets ont ainsi été remplacés par huit triangles équilatéraux
congrus, et les faces carrées ont été remplacées par six octogones congrus. Si la

Copyright c Canadian Mathematical Society, 2016

98/ THE CONTEST CORNER

surface totale des huit triangles égale la surface d’un des octogones, quelle est cette
surface ? (Chaque face octogonale comporte deux longueurs différentes de côté,
en alternance.)

CC213. Une pyramide est construite à partir de cubes de taille unitaire,

empilés en tranches carrées. La tranche supérieure comporte 1 × 1 = 1 cube, la
seconde en a 3 × 3 = 9, celle en bas de ça en a 5 × 5 = 25, et ainsi de suite,
chaque tranche en ayant deux de plus sur chaque côté par rapport à la tranche
supérieure. La pyramide a 12 tranches au total. Déterminer la surface externe de
cette pyramide, incluant le fond.

CC214. Dans le plan, les points (2, 5) et (6, 5) sont deux sommets d’un
hexagone régulier de côté deux. Une certaine ligne L, passant par le point (0, 0),
coupe l’hexagone en deux parties de même surface. Quelle est la pente de la ligne
L?

CC215. À chaque cercle dans l’arbre indiqué ci-bas on assigne une valeur,
choisie dans un ensemble S, de façon à ce que dans chaque chemin vers le bas dans
l’arbre les valeurs assignées n’augmentent jamais. C’est-à-dire A ≥ B, A ≥ C, C ≥
D, C ≥ E où A, B, C, D, E ∈ S (Il est permis qu’une valeur dans S apparaisse
plus qu’une fois.) De combien de manières peut-on assigner des valeurs à l’arbre
si S = {1, . . . , 6}?

(Au choix: Généraliser au cas où S = {1, 2, 3, , . . . , n} à l’aide d’une expression

algébrique explicite pour le nombre d’assignations.)

Crux Mathematicorum, Vol. 42(3), March 2016

 THE CONTEST CORNER /99

CONTEST CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2015: 41(3), p. 96–97.

CC161. A number n written in base b reads 211, but it becomes 110 when
written in base b + 2. Find n and b in base 10.
A reformulation of #4 of the Santa Clara University High School Mathematics
2001 Contest.
We received ten solutions, of which eight were complete and correct. All eight
solutions were nearly identical so we present a composite solution here.
We have that n = 211 in base b. This requires b > 2 and means

n = (211)b = 2 · b2 + 1 · b + 1,

while n = 110 in base b + 2 gives us

n = (110)b+2 = 1 · (b + 2)2 + 1 · (b + 2) + 0.

Equating these two expressions gives

2b2 + b + 1 = b2 + 4 + 4b + b + 2
b2 − 4b − 5 = 0
b = −1, 5

We discard the negative solution both because of the restriction on b and the fact
that a base cannot be negative. Using b = 5 we can calculate

n = (211)5 = (110)7 = 56.

Therefore n = 56 and b = 5 in base 10.

CC162. What is the probability that 99 divides a randomly chosen 4-digit

palindrome?
A reformulation of #3 from the team section of the 2010 Raytheon MATHCOUNTS
State Competition.
We received eight submissions of which four were correct and complete. We present
the solution by Titu Zvonaru.
A 4-digit palindrome is a number of the form abba, with a = 1, 2, . . . , 9 and b =
0, 1, . . . , 9, hence there are 90 numbers which are 4-digit palindrome. Since abba =
1001a + 100b = 11(91a + 10b), we deduce that all 4-digit palindromes are divisible

Copyright c Canadian Mathematical Society, 2016

100/ THE CONTEST CORNER

by 11. The number abba is divisible by 9 if and only if a + b is divisible by 9. If a +

b = 9 we have the possibilities 1881, 2772, 3663, 4554, 5445, 6336, 7227, 8118, 9009;
if a + b = 18, then there is only the number 9999. The searched probability is
10/90 = 1/9.
Editor’s Comments. Some solvers counted also the case when a = b = 0 in the
solution, but there is a flaw. Indeed, they have counted a total of 90 palindrome
numbers (9 possibilities for the nonzero digit a and 10 possibilities for the digit
b), but then they counted the case when a = b = 0, a contradiction. The solution
could have been consistent if they also counted the degenerate case when a = 0
in the total number of palindromes, giving 10 · 10 = 100 palindrome numbers.
In this case we also have a = 0, b = 0 and a = 0, b = 9, giving the probability
12/100 = 3/25. The only consistent (but not correct in the strict sense) solutions
are the ones given by Kathleen E. Lewis and Hannes Geupel.

CC163. If x is randomly chosen in [−100, 100], what is the probability that

g[f (x)] is negative given that f (x) = x2 + 3x − 7 and g(x) = x2 − 2x − 99?
A reformulation of #8 of the 2014 University of North Colorado Math Contest.
We received eight submissions of which seven were correct and complete. We
present the solution by Titu Zvonaru.
Since g(x) = (x + 9)(x − 11), we have g(x) < 0 ⇐⇒ x ∈ (−9, 11). It follows that

g(f (x)) < 0 ⇐⇒ −9 < f (x) < 11

⇐⇒ −9 < x2 + 3x − 7 < 11
⇐⇒ (x + 6)(x − 3) < 0 and (x + 1)(x + 2) > 0.

Thus, g(f (x)) < 0 ⇐⇒ x ∈ (−6, −2)∪(−1, 3), and the probability is given by the
total length of the combined intervals divided by the total length of the domain
of x, which is
(−2 + 6) + (3 + 1) 1
= .
100 + 100 25

CC164. Build two equilateral triangles on the diameter of a circle with radius
5. What is the total area of the circle outside the equilateral triangles? (See the
diagram below.)
Proposed by the editor.
We received eleven submissions of which ten were correct. We present the solution
by Fernando Ballesta Yagüe, slightly modified by the editor.
Denote the center of the circle by O, the two equilateral triangles by ABC and
DBC (with BC being the diameter of the circle), and the intersections of AB and
AC with the circle by E and F respectively. Use r for the radius of the circle, and
recall that r = 5.

Crux Mathematicorum, Vol. 42(3), March 2016

 THE CONTEST CORNER /101

As 4ABC is equilateral, ∠BCA = 60◦ . Further, OC = OF = r, so it follows that

4OCF is also equilateral, and ∠F OC = 60◦ .
Hence the area of the circular segment between the chord CF and the circle is
equal to the area of a circular sector with central angle 60◦ minus the area of the
equilateral 4F OC, that is
√
π · r2 1 2 ◦ 25π 25 3
− · r · sin(60 ) = − .
6 2 6 4
We can reason the same way with 4OBE, and also with the matching construction
on 4DBC. It follows that the area contained inside the circle but outside the
triangles consists of four congruent circular segments, and the total area is
Ç √ å
25π 25 3 50π √
4· − = − 25 3.
6 4 3

CC165. Georges pays $50 on each of four gas refills but the prices per litre
were $1.32, $1.25, $1.11 and $1.18 as the price was fluctuating a lot in that time
period. What is the average price per litre?
Proposed by the editor.
We received five correct solutions and one incorrect solution. We present the so-
lution of Henry Ricardo.
$50 $50 $50 $50
The quantities of gas purchased were $1.32/L , $1.25/L , $1.11/L , and $1.18/L .

(Total cost of gas)

(Average price per litre) =
(Total quantity of gas purchased)

$200
= $50 $50 $50 $50
$1.32/L + $1.25/L+ $1.11/L + $1.18/L

≈ $1.21/L

Copyright c Canadian Mathematical Society, 2016

102/ THE OLYMPIAD CORNER

THE OLYMPIAD CORNER

No. 341
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 January
1, 2017, although late solutions will also be considered until a solution is published.
The editor thanks André Ladouceur, Ottawa, ON, for translations of the problems.

OC271. A scalene triangle ABC is inscribed within circle ω. The tangent to

the circle at point C intersects line AB at point D. Let I be the center of the
circle inscribed within 4ABC. Lines AI and BI intersect the bisector of ∠CDB
in points Q and P , respectively. Let M be the midpoint of QP . Prove that M I
passes through the middle of arc ACB of circle ω.

OC272. Find all real triples (a, b, c), for which

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

b(c2 + a) = a(a + bc),

c(a2 + b) = b(b + ca).

OC273. Find all functions f : R → R such that f (x2015 + (f (y))2015 ) =

2015
(f (x)) + y 2015 holds for all reals x, y.

OC274. Find all triplets (x, y, p) of positive integers such that p is a prime
xy 3
number and x+y = p.

OC275. Steve is piling m ≥ 1 indistinguishable stones on the squares of

an n × n grid. Each square can have an arbitrarily high pile of stones. After he
finishes piling his stones in some manner, he can then perform stone moves, defined
as follows. Consider any four grid squares, which are corners of a rectangle, i.e.
in positions (i, k), (i, l), (j, k), (j, l) for some 1 ≤ i, j, k, l ≤ n, such that i < j and
k < l. A stone move consists of either removing one stone from each of (i, k) and
(j, l) and moving them to (i, l) and (j, k) respectively, or removing one stone from
each of (i, l) and (j, k) and moving them to (i, k) and (j, l) respectively.
Two ways of piling the stones are equivalent if they can be obtained from one
another by a sequence of stone moves.
How many different non-equivalent ways can Steve pile the stones on the grid?

Crux Mathematicorum, Vol. 42(3), March 2016

 THE OLYMPIAD CORNER /103

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

OC271. Un triangle scalène ABC est inscrit dans un cercle ω. La tangente au

cercle au point C coupe la droite AB au point D. Soit I le centre du cercle inscrit
dans le triangle ABC. Les droites AI et BI coupent la bissectrice de l’angle CDB
aux points respectifs Q et P . Soit M le milieu du segment QP . Démontrer que
M I passe au milieu de l’arc ACB du cercle ω.

OC272. Déterminer tous les triplets (a, b, c) de réels tels que

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

b(c2 + a) = a(a + bc),

c(a2 + b) = b(b + ca).

OC273. Déterminer toutes les fonctions f : R → R qui vérifient

f (x2015 + (f (y))2015 ) = (f (x))2015 + y 2015

pour tous réels x, y.

OC274. Déterminer tous les triplets (x, y, p) d’entiers strictement positifs pour
xy 3
lesquels p est un nombre premier et x+y = p.

OC275. Steve empile m (m ≥ 1) pierres indifférenciables sur un carrelage

n × n. Chaque case du carrelage peut recevoir un nombre arbitraire de pierres.
Après avoir terminé d’empiler ses pierres d’une façon quelconque, il peut ensuite
accomplir des déplacements de pierres comme suit. On considère quatre cases qui
forment les coins d’un rectangle, c.-à-d. les positions (i, k), (i, l), (j, k), (j, l), k, j, k
et l étant des entiers tels que 1 ≤ i, j, k, l ≤ n, i < j et k < l. Un déplacement
de pierres consiste à enlever une pierre de chacune des cases (i, k) et (j, l) et les
ajouter aux cases respectives (i, l) et (j, k) ou à enlever une pierre de chacune des
cases (i, l) et (j, k) et les ajouter aux cases respectives (i, k) et (j, l).

On dit que deux façons d’empiler les pierres sont équivalentes si une façon peut
être obtenue à partir de l’autre par une série de déplacements de pierres.

Combien y a-t-il de façons non équivalentes d’empiler les pierres sur le carrelage?

Copyright c Canadian Mathematical Society, 2016

104/ THE OLYMPIAD CORNER

OLYMPIAD SOLUTIONS
Statements of the problems in this section originally appear in 2015: 41(1), p. 9–11.

OC211. Find maximum value of

|a2 − bc + 1| + |b2 − ac + 1| + |c2 − ba + 1|

where a, b, c are real numbers in the interval [−2, 2].

Originally problem 1 of day 2 of 2013 Kazakhstan National Olympiad Grade 11.
Editor’s Note. This problem is a duplicate from OC179. The editor used the
question from two different Kazakhstan Olympiad contests but did not realize that
the question could be repeated across grades. Since the editor received different
solutions to this problem as opposed to the first version, the editor will include
one here. My apologies.
We received 3 correct submissions. We present the solution by Šefket Arslanagić.
Let f (a, b, c) = |a2 − bc + 1| + |b2 − ac + 1| + |c2 − ba + 1|. By symmetry, we may
also suppose that a ≥ b ≥ c. Notice immediately that f (a, b, c) = f (−a, −b, −c).
By using these two properties, we can assume without loss of generality that at
least two values are positive , so a ≥ b ≥ 0.
Further, we can show that f (a, b, |c|) ≤ f (a, b, −|c|). This follows since

|a2 − b|c| + 1| ≤ a2 + b|c| + 1 and |b2 − a|c| + 1| ≤ b2 + a|c| + 1.

Hence, since we are looking for the maximum value, we may assume that a ≥ b ≥
0 ≥ c. Next, we show that f (a, b, c) ≤ f (a, b, −2). This inequality is equivalent to
showing that

a2 − bc + 1 + b2 − ac + 1 + |c2 − ab + 1| ≤ a2 + 2b + 1 + b2 + 2a + 1 + |4 − ab + 1|

which is equivalent to

|c2 − ab + 1| ≤ 5 − ab + (2 + c)(a + b).

If c2 − ab + 1 ≥ 0, then the equality is exact by adding the inequalities c2 ≤ 4 and

0 ≤ (2 + c)(a + b). If c2 − ab + 1 ≤ 0, then the above becomes via a sequence of if
and only if statements

2ab ≤ 6 + c2 + (2 + c)(a + b)
2ab − 2a − 2b + 2 ≤ 8 + c2 + c(a + b)
2(a − 1)(b − 1) ≤ 8 + c2 + c(a + b).

This last inequality is true since

2(a − 1)(b − 1) ≤ 2 ≤ 4 + 4 + c2 + c(a + b).

Crux Mathematicorum, Vol. 42(3), March 2016

 THE OLYMPIAD CORNER /105

The last inequality holds since

4 + c2 + c(a + b) ≥ 4|c| + c(a + b) ≥ (a + b)|c| + c(a + b) ≥ 0

using the fact that (2 − c)2 ≥ 0 in the first inequality. Therefore, it suffices to find
the maximum of the function f (a, b, −2) where a ≥ b ≥ 0. This reduces to finding
the maximum of

f (a, b, −2) = a2 + b2 − ab + 2a + 2b + 7.

with a ∈ [0, 2] and b ∈ [0, 2]. This maximum must occur when either a = b, b = 0
or a = 2 (by say Calculus). Checking each of these cases reveals that the maximum
value is
f (2, 2, −2) = |4 + 4 + 1| + |4 + 4 + 1| + |4 − 4 + 1| = 19.

OC212. Let ABCDE be a pentagon inscribe in a circle (O). Let BE ∩ AD =

T . Suppose the parallel line with CD which passes through T cuts AB, CE at
X, Y . If ω is the circumcircle of triangle AXY then prove that ω is tangent to
(O).
Originally problem 3 from level X of the 2013 Romanian National Olympiad.
We received 3 correct submissions. We present the solution by Andrea Fanchini.
We have that ∠AEC = ∠ADC because both are inscribed in the same arc of circle.
Then since the lines XY and CD are parallel, we have also ∠ADC = ∠DT Y .
Similarly ∠ECD = ∠T Y C = ∠EAD and therefore the quadrilateral AT Y E is
cyclic.
Now if we draw a tangent AW to the circle (O), we have

∠W AB = ∠AEB = ∠AET,

but the points A, T , Y and E are concyclic, so we have also

∠AET = ∠AY T = ∠AY X.

Therefore AW is also tangent to circle ω completing the proof.

OC213. Suppose p > 3 is a prime number and

X
S= ijk.
2≤i<j<k≤p−1

Prove that S + 1 is divisible by p.

Originally problem 4 of the 2013 Indonesian Mathematical Olympiad.
We received 2 correct submissions. We present the solution by Michel Bataille.

Copyright c Canadian Mathematical Society, 2016

106/ THE OLYMPIAD CORNER

From Fermat’s Little Theorem, each element of Zp = {0, 1, 2, . . . , p − 1} is a root

of the polynomial xp − x of Zp [x]. Thus, xp − x = x(x − 1)(x − 2) · · · (x − (p − 1)) in
Zp [x]. Since xp − x = x(xp−1 − 1), it follows that xp−1 − 1 = (x − 1)(x − 2) · · · (x −
(p − 1)) in Zp [x].
This said, let p(x) be the polynomial (x − 2)(x − 3) · · · (x − (p − 1)) of Zp [x]. Since
(x − 1)p(x) = xp−1 − 1, we see that

p(x) = xp−2 + xp−3 + · · · + x + 1. (1)

However, we also have

p(x) = xp−2 − e1 xp−3 + e2 xp−4 − e3 xp−5 + · · · − ep−2 (2)

with, modulo p,
p−1
X X X
e1 ≡ i, e2 ≡ i.j, e3 ≡ ijk, . . .
i=2 2≤i<j≤p−1 2≤i<j<k≤p−1

Comparing (1) and (2) yields 1 ≡ −e3 (mod p). Since S = e3 , we conclude that
S + 1 ≡ 0 (mod p), that is, p divides S + 1.

OC214. Let ABC be an acute-angled triangle with AC 6= BC, and let O be

the circumcentre and F the foot of the altitude through C. Furthermore, let X
and Y be the feet of the perpendiculars dropped from A and B respectively to
(the extension of) CO. The line F O intersects the circumcircle of F XY a second
time at P . Prove that OP < OF .
Originally problem 6 of the 2013 South Africa National Olympiad.
We received 2 correct submissions. We present the solution by Oliver Geupel.

Crux Mathematicorum, Vol. 42(3), March 2016

 THE OLYMPIAD CORNER /107

Let us denote α = ∠BAC and β = ∠CBA.

In the isosceles triangle AOC we have ∠ACO = 90◦ − ∠COA/2 = 90◦ − β. As
a consequence, in the right triangle AXC we have CX = AC sin β. Similarly
CY = BC sin α. The altitude from point C in the triangle ABC has length
CF = AC sin α = BC sin β. We obtain
CF 2 = CX · CY,
where the right-hand side is the power of point C with respect to the circumcircle
Γ of triangle F XY . Hence the line CF is a tangent to the circle Γ.
Therefore the centre of the circle Γ lies on the line AB. Also the centre of the
circle Γ lies on the perpendicular bisector p of the segment XY . Since AC 6= BC,
the line p is not parallel to the line AB so that the lines p and AB are concurrent.
The midpoint D of side AB lies on the line p because it is equidistant from the
lines AX and BY . Thus, the point D is the centre of the circle Γ.
In the right triangle DF O we have OD < OF . The power of point O with respect
to the circle Γ is
OF · OP = OD2 − DF 2 .
Consequently,
OD2 − DF 2 OF 2
OP = < = OF.
OF OF

OC215. Let n > 1 be an integer. The first n primes are p1 = 2, p2 = 3, . . . , pn .

A A
Set A = pp11 pp22 ...ppnn . Find all positive integers x, such that is even, and has
x x
exactly x divisors.
Originally problem 6 from day 2 of the 2013 South East Mathematical Olympiad.
We present the solution by Konstantine Zelator. There were no other submissions.
First, we claim that the only solution is x = p1 p2 ...pn . Note that this is a solution.
It suffices to show it is the only one. Since x is a divisor of A, we have that
n
Y
x= piei where 0 ≤ ei ≤ pi
i=1

Now, e1 cannot be 2 since then A/x is odd so e1 = 0 or e1 = 1. Assume towards

a contradiction that e1 = 0. Then
n
A
ppi i −ei
Y
=4·
x i=2

and the number of divisors this number has (which by the problem statement is
equal to x) is
Yn n
Y
pei i = x = 3 (pi − ei + 1)
i=2 i=2

Copyright c Canadian Mathematical Society, 2016

108/ THE OLYMPIAD CORNER

Now, if e2 = 0, then the left hand side above is not divisible by 3 but the right hand
side is, a contradiction. If e2 = 1, then the left hand side above is divisible by 1 and
the right hand side is divisible by at least 32 coming from 3(4 − e1 ) = 32 , another
contradiction. If e2 = 2, then the right hand side is even from the (4 − e1 ) = 2
term but the left hand side is odd, again a contradiction. Lastly, if e3 = 3, then
note that n ≥ 3 must be true by inspection. The remaining terms on the right
must all be odd numbers since the left hand side is odd and thus, since each pi is
an odd prime for i ≥ 3, we have that (pi − ei + 1) is odd and hence each ei is odd.
Thus ei ≥ 1 for all i ≥ 3. Hence, since pei i ≥ pi ≥ (pi − ei + 1), we see that
n
Y n
Y n
Y
x= pei i = 33 pei i > 3 · (4 − 3) · (pi − ei + 1) = x
i=2 i=3 i=3

which is a contradiction. Thus e1 6= 0 and hence e1 = 1. Then as before,

n
A
ppi i −ei
Y
=2·
x i=2

and the number of divisors this number has (which by the problem statement is
equal to x) is
Yn n
Y
2· pei i = x = 2 (pi − ei + 1).
i=2 i=2

Simplifying gives
n
Y n
Y
piei = (pi − ei + 1).
i=2 i=2

As before, each of the e1 terms must be odd otherwise the right hand side is even.
Hence ei ≥ 1 for all i ≥ 2. As before, pei i ≥ pi ≥ (pi − ei + 1) with equality holding
if and only if pei i + ei = pi + 1 and thus, since ei ≥ 1, equality holds if and only if
ei = 1 for all i ≥ 2. Thus, x = p1 p2 ...pn completing the proof.

Crux Mathematicorum, Vol. 42(3), March 2016

 MICHEL BATAILLE /109

FOCUS ON...
No. 21
Michel Bataille
The Product of Two Reflections in the Plane
Introduction
In this number, we exploit the two following theorems as a tool for geometrical
properties or problems. Rn denotes the reflection in the line n.
(a) If the lines ` and m are parallel, then Rm ◦ R` = T2→ −u , the translation by
vector 2→−
u where → −
u is the vector orthogonal to ` and m such that T→ u (`) = m.
−

(b) If the lines ` and m intersect in O, then Rm ◦ R` = ρO,2θ , the rotation with
centre O and angle 2θ where θ is the angle such that ρO,θ (`) = m.

Illustrating (a) and (b)

Our first problem offers an opportunity of using both (a) and (b):
Let ABC be a triangle and ρ be any rotation with centre A. Let
B 0 = ρ(B), C 0 = ρ(C), and B1 = RCA (B). The circle with centre B 0
and radius B 0 C 0 intersects CC 0 again at M . Show that M CB1 B 0 is a
parallelogram.
Let E be the midpoint of CC 0 . From (b), RAE ◦ RCA is a rotation with centre
A. In addition, it transforms C into RAE (C) = C 0 (since AC 0 = AC, AE is the
perpendicular bisector of CC 0 ). Thus, RAE ◦ RCA = ρ and so B 0 = RAE (B1 ).

In a similar way, if F is the midpoint of M C 0 , we have M = RF B 0 (C 0 ). Now,

RAE ◦ RF B 0 = T−−0−→ (from (a)) and so T−−0−→ (M ) = RAE (C 0 ) = RAE ◦ ρ(C) =
B B1 B B1
−−→ −−−→
RCA (C) = C. Thus M C = B 0 B1 and the result follows.

Copyright c Canadian Mathematical Society, 2016

110/ FOCUS ON... THE PRODUCT OF TWO REFLECTIONS IN THE PLANE

When ` and m are perpendicular

This is the simplest case of (b): the product Rm ◦ R` is the half-turn about the
point O of intersection of ` and m. The same is true of R` ◦ Rm . By way of
illustration, consider the gist of problem OC24 [2011 : 275 ; 2012 : 180]:

Let P be the midpoint of the line segment KL and O be any point not
on the line KL. Then, if M is the symmetric of O about P and N the
reflection of O in KL, the points K, L, M, N are concyclic.

Let ρP be the half-turn about P . Then N = RKL ◦ ρP (M ) that is, N = R` (M )

where ` is the perpendicular to KL at P (since ρP = RKL ◦ R` ). Now, let C be
the circle through K, L, M . As the perpendicular bisector of KL, the line ` is a
diameter of C, hence an axis of symmetry of C. Since M is on C, N = R` (M ) is
on C as well, and so K, L, M, N are concyclic.

Interestingly, this result readily leads to an unusual proof of the following well-
known property of the orthocentre H of a triangle ABC: the reflections of H in
the sides of the triangle lie on its circumcircle Γ.

For example, let us show that the reflection H 0 of H in BC is on Γ.

Let A0 be the midpoint of BC. From what we have just proved, it is sufficient to
show that the symmetric D of H about A0 is on Γ.

Since BC and HD have the same midpoint, BHCD is a parallelogram. Thus DC

is parallel to BH, hence is perpendicular to CA. Similarly, DB is perpendicular
to BA and so the circle with diameter AD, which passes through A, B, and C,
coincides with Γ. Thus D is on Γ.

Crux Mathematicorum, Vol. 42(3), March 2016

 MICHEL BATAILLE /111

Another unusual proof of a well-known property

Let M be a point of the circumcircle Γ of a triangle ABC and let
A0 = RBC (M ), B 0 = RCA (M ), C 0 = RAB (M ). Then A0 , B 0 , C 0
are collinear on a line through the orthocentre H (the Steiner line
associated with M ).
We discard the obvious case when M is a vertex of the triangle and assume that
M 6= A, B, C.

Let Γb and Γc be the circumcircles of ∆AB 0 C and ∆AC 0 B, respectively. Then

Γb = RCA (Γ) (since Γ is also the circumcircle of ∆AM C) and, similarly, Γc =
RAB (Γ). Let D be the point of intersection other than A of Γb and Γc . The
product RAB ◦ RCA is a rotation with centre A transforming Γb into Γc . A
rotation being a spiral similarity, it follows from the main result of Focus On...
No 12 that RAB (C) = RAB ◦ RCA (C) is the point of intersection other than D
of the line CD and Γc . Thus, CD is perpendicular to AB. In the same way, BD
is perpendicular to AC and consequently D coincides with the orthocentre H of
∆ABC. To conclude, it suffices to add that since RAB ◦ RCA (B 0 ) = RAB (M ) =
C 0 , the points B 0 , C 0 , H are collinear. The same being true of A0 , B 0 , H and of
C 0 , A0 , H (similarly), we conclude that A0 , B 0 , C 0 , H are collinear.

Reflections and billiards

A billiards table is a good place for observing reflections. Successive reflections in
the cushions are sometimes necessary to score a point! The following problem (a
part of American Mathematical Monthly problem 10749 posed in 1999) is likely to
appeal to the reader who is also a billiards player.
Let ABC be a triangle with a right angle at B and an angle of π/6 at

Copyright c Canadian Mathematical Society, 2016

112/ FOCUS ON... THE PRODUCT OF TWO REFLECTIONS IN THE PLANE

A. Consider a billiard path in the triangle that begins at A, reflects

successively off side BC at P , off side AC at Q, off side AB at R, off
side AC at S, and then ends at B. Show that AP, QR, and SB are
concurrent at a point X and that the angles formed at X measure π/3.

Without loss of generality, we suppose that the triangle ACB has positive orien-
tation and introduce the following points: B 0 = RCA (B), the point D of intersec-
tion of the lines BC and AB 0 , and the vertices A1 , B1 , D1 of equilateral triangles
A1 BD, B1 DA, D1 AB constructed outward ∆ABD. From the billiard path as de-
scribed, we have RBC (AP ) = P Q, RCA (P Q) = QR, RAB (QR) = RS, and
RCA (RS) = SB so that

SB = (RCA ◦ RAB ◦ RCA ) ◦ RBC (AP ) = RAB 0 ◦ RBC (AP ) (1)

(using the easily checked fact that Rm ◦R` ◦Rm is the reflection in the line Rm (`)).
Since RAB 0 ◦ RBC is the rotation ρD,π/3 , the point B1 = ρD,π/3 (A) is on the line
SB. Likewise, from AP = ρD,−π/3 (SB), we obtain that A1 is on AP . Further-
more, (1) yields RAB ◦ RCA (SB) = RCA ◦ RBC (AP ), that is, QR = ρC,2π/3 (AP )
and therefore the points ρC,2π/3 (A) = D and RAB ◦ RCA (B) = ρA,π/3 (B) = D1
lie on the line QR. In conclusion, the lines AP, QR, and SB coincide with the
lines AA1 , DD1 , and BB1 , respectively. This answers the questions since these
lines are concurrent at the Fermat point (say X) of ∆ABD, from which each side
subtends an angle of 120◦ (a well-known result).

Crux Mathematicorum, Vol. 42(3), March 2016

 MICHEL BATAILLE /113

Exercises
Our first exercise is problem 2439 [1999 : 238 ; 2000 : 241]. Three solutions
were featured and the reader is asked to find a fourth one! The second exercise
is problem 2485 [1999 : 431 ; 2000 : 508], slightly modified. Of course, solutions
should use reflections.
1. Suppose that ABCD is a square with side a. Let P and Q be points on
sides BC and CD, respectively, such that ∠P AQ = 45◦ . Let E and F be√the
intersections of P Q with AB and AD, respectively. Prove that AE + AF ≥ 2 2a.
[Hint: first show that RAP (B) = RAQ (D).]
2. Let ABCD be a convex quadrilateral with AB = BC = CD and such that
AD and BC are not parallel. Let P be the intersection of the diagonals AC and
BD. If AP : BD = DP : AC, prove that AB ⊥ CD. [Hint: if `, m, n are the
perpendicular bisectors of BC, CA, BD, respectively, and O is the circumcentre of
∆BP C, consider ROC ◦ R` and Rn ◦ Rm .]

Copyright c Canadian Mathematical Society, 2016

114/ THE USE OF COORDINATE SYSTEMS BEFORE DESCARTES

The use of coordinate systems before

Descartes
Florin Diacu

1 Introduction

It is basic knowledge among mathematicians that René Descartes introduced rect-

angular coordinate systems in the 17th century, thus creating analytic geometry.
This accomplishment was an early example of how a branch of mathematics, alge-
bra, can explore another area, geometry. The division of mathematics into various
fields was in its early stages at that time, so this way of looking at Descartes’s
achievement is a rather contemporary point of view, which reflects our current
(though perpetually changing) classification of mathematical subjects.
Descartes published his ideas on this topic in 1637 in the appendix La géométrie
(see Figure 1) to his book Discours de la méthode pour bien conduire sa raison et
chercher la vérité dans les sciences (Discourse on the Method of Reasoning Well
and Seeking Truth in the Sciences), a philosophical and autobiographical treatise
that had a strong influence on the further development of human thought. The
concept of Cartesian coordinate system was coined to honour Descartes, whose
Latanized name was Cartesius. Students of mathematics who have struggled to
solve geometry problems using the ancient synthetic methods can get a good feeling
about the power of this approach, the more so if they can surmount the compu-
tational hurdles that occur sometimes. Apart from this merit, Descartes’s contri-
bution opened the way towards laying bridges between branches of mathematics
that had developed separately, a feat that is highly regarded among researchers
today because it emphasizes the unity of our field of knowledge, in spite of the
often overlapping areas in which we divide it.
Some critics, however, contested Descartes’s priority for designing this method,
since Pierre Fermat had the same idea a few years before Descartes. A quote from
E.T. Bell, an American mathematician and colourful historian of mathematics,
sheds more light on this issue, [1]: “There is no doubt that he [Fermat] preceded
Descartes. But as his work of about 1629 was not communicated to others until
1636, and was published posthumously only in 1679, it could not possibly have
influenced Descartes in his own invention, and Fermat never hinted that it had.”
Since similar simultaneous independent contributions to science and mathematics
are common, a phenomenon that has been thoroughly researched, we will not
pursue this topic.
What we actually want to emphasize here is a little known fact, namely that hidden
rectangular coordinate systems have been used in geometry since antiquity. We
will further present two examples of how this idea was applied long before Descartes
made it universally known. The first is due to the Greek geometer and astronomer

Crux Mathematicorum, Vol. 42(3), March 2016

 FLORIN DIACU /115

Figure 1: An extract from La géométrie, the appendix of Discourse de la méthode

in which Cartesian coordinates were introduced.

Apollonius of Perga (c. 262–c. 190 BC), whereas the second is drawn from the work
of the Persian mathematician Umar ibn Ibrahim al-Khayyami (AD 1048–1131),
known among literati as the poet Omar Khayyam.

2 The conics of Apollonius

Apollonius was born in Perga, Pamphylia, today Murtina in Antalya, Turkey,

around 262 BC. He studied in Alexandria under disciples of Euclid, where he
learned geometry and astronomy, subjects he later taught there. We know very
little about his life. The information we can trust appears in the various prefaces of
his fundamental treatise, Kωνικά (Conics), whose original was lost (as it happened
with most ancient manuscripts), but which reached us mostly through Arabic
translations.
It is intuitively easy to understand why the ellipse, parabola, and hyperbola are
called conics: if we intersect a cone of two sheets with a plane at suitable an-
gles relative to the symmetry axis, we obtain these curves (along with circles,
points, and two intersecting straight lines). Apollonius was the first to coin their
names, though other Greek mathematicians studied them earlier. In his treatise,
he approached them without using algebra, as we define this branch of mathe-
matics today. So how did he do it? Unlike polygons or circles, conics are difficult
to understand without analytic geometry, which describes them through simple
quadratic equations. A close look at the work of Apollonius, however, reveals that
he used camouflaged coordinate systems, and in the absence of algebraic tools he
employed a geometric language. So, with his smart and systematic approach in
the study of conics, he anticipated the work of Descartes by almost two millennia.
As an example, we consider now the case of the parabola (see Figure 2), but let us

Copyright c Canadian Mathematical Society, 2016

116/ THE USE OF COORDINATE SYSTEMS BEFORE DESCARTES

remark that Apollonius used this method to understand hyperbolas and ellipses
too (see, e.g., [2]). Let us take a circular cone (Figure 2, left), which we intersect
with a plane parallel with the generatrix AC. The result of this intersection is what
we call a parabola, the curve F N H. Apollonius wanted to characterize this curve
in order to distinguish it from ellipses and hyperbolas. For this purpose he derived
the “symptom,” which in our modern language is nothing but the relationship
between the abscissa and the ordinate of an arbitrary point on the curve. So let
M be such a point. A plane through M orthogonal to the symmetry axis intersects
the cone along the circle of diameter DE. Denote by O the point at the base of
the perpendicular from M to DE. Then ON and OM are also perpendicular to
each other. Apollonius wanted to find the relationship between ON =: x and
OM =: y. Indeed, had he placed a coordinate system with the origin at N in the
plane of the parabola (Figure 2, right), M would have had coordinates (x, y), the
standard notation we use today to write the equation of the parabola.
Next, Apollonius started his geometric reasoning. He noticed that in the circle of
diameter DE he could write that

OM 2 = OD · OE. (1)

Then he constructed the segment N K perpendicular to ON in the plane of the

parabola (Figure 2, right), taking its length such that

NK BC BC
= · , (2)
NA AC AB
a brilliant choice he probably reached after long reflections on this problem. By
similarity of triangles, he noticed on one hand that

BC OD
= , (3)
AC ON
and on the other hand, after manipulating some proportions, that

BC OE
= . (4)
AB NA
Substituting the results from (3) and (4) into (2), he obtained that

NK OE · OD
= .
NA N A · ON
Multiplying N K/N A by ON/ON , i.e. by 1, he got

NK N K · ON
= .
NA N A · ON
Comparing the numerators of the above two relationships, he concluded that

OE · OD = N K · ON,

Crux Mathematicorum, Vol. 42(3), March 2016

 FLORIN DIACU /117

N M
M
D E N
O O
H K
B C
G
F

Figure 2: Left: the intersection of the cone with a plane parallel with the generatrix
AC yields a parabola. Right: the parabola F N H viewed in its own plane.

which in light of (1) can be written as

OM 2 = N K · ON.
If we denote the length of N K by p, which is the same as the length of the latus
rectum (defined below), then the above relationship becomes
y 2 = px,
easy to recognize as the equation of a parabola, albeit written in nonstandard
form.
So in spite of having no algebra tools at his disposal, Apollonius actually obtained
the equation of the parabola. He even had a value for the length of the latus
rectum, defined as the chord through the focus of the parabola parallel with the
directrix. All in all, he had a very good understanding of conics, which he achieved
through the same kind of analysis Descartes’s did, but using a very basic language.

3 The cubic equation of al-Khayyami

The mathematician, philosopher, and poet al-Khayyami was born in AD 1048 in
Nishapur, Persia, today a city of the same name in northeastern Iran. During
his long life he contributed to mathematics, mechanics, astronomy, geography,
mineralogy, and philosophy, and left an important body of poetry, which is still
widely read. Five centuries before Cardano’s formula for the cubic equation was
discovered, al-Khayyami approached the problem with innovative methods. Like
Apollonius, he used hidden Cartesian coordinates to express his solutions.
At that time all coefficients of an equation were considered positive. (Although
negative numbers had been used in China no later than the AD 100, see [2],

Copyright c Canadian Mathematical Society, 2016

118/ THE USE OF COORDINATE SYSTEMS BEFORE DESCARTES

Figure 3: A bust of al-Khayyami in Nishapur, Iran.

p. 198, they became universally accepted only after the 16th century.) So al-
Khayyami first listed all possible cubic equations: one binomial, x3 = d; six
trinomial, x3 + bx2 = d, x3 + d = bx2 , x3 = bx2 + d, x3 + cx = d, x3 + d = cx,
x3 = cx + d; and seven tetranomial, x3 + bx2 + cx = d, x3 + bx2 + d = cx,
x3 + cx + d = bx2 , x3 = bx2 + cx + d, x3 + bx2 = cx + d, x3 + cx = bx2 + d,
x3 + d = bx2 + cx. Then he provided for each of them a solution obtained with the
help of a conic section, as we will show in the following example. Let us consider
the case of
x3 + cx = d. (5)
√
He first took a segment AB = c and drew EB = d/c perpendicular to AB (see
Figure 4). Then he considered the parabola
1
y = √ x2
c
through D and B, and constructed the semicircle EDB given by
Å ã
d 2 d2
x− + y2 = 2 . (6)
2c 4c
The semicircle and the parabola intersect at D. Taking DF = y0 perpendicular
to EB, he obtained the segment F B = x0 and claimed that its length, x0 , is a
solution of equation (5). Notice the hidden upside-down frame centred at B that
occurs here, which allowed us to write the above equations in modern language.
Following [2], let us now prove that x0 is indeed a solution of equation (5). Since
D is on the semicircle, we have
Å ã
d
x0 − x0 = y02 ,
c
an equation equivalent to (6) that can be also expressed as
x0 y0
= d
. (7)
y0 c − x0
But D is also on the parabola, so we have
√
x20 = cy0 ,

Crux Mathematicorum, Vol. 42(3), March 2016

 FLORIN DIACU /119

E F x0 B

y0

C
D

Figure 4: A solution al-Khayyami gave for the equation he described as “a cube

and sides are equal to a number,” i.e. x3 + ax = b.

which is equivalent to
√
c x0
= .
x0 y0
Squaring this equality and using (7), we obtain

c x20 y02 y0 x0 x0
= =
= · = .
x20 y02 d
− x0 2 2 d
c − x0 y0 d
c − x0
c

But comparing the first and last expressions in the above sequence, we see that
x0 verifies equation (5), a remark that completes the argument.
Of course, al-Khayyami did not write his proof as presented here. He used only
words and figures, as it can be seen in one of his manuscripts (Figure 5). Even
equation (5) appeared without symbols. He described it as: “a cube and sides
are equal to a number.” Nevertheless, he did algebra, a branch of mathematics
that had recently formed. As in the case of Apollonius, coordinate systems were
essential for obtaining these results.

Figure 5: A page of al-Khayyami’s manuscript “Cubic equation and intersection

of conic sections,” kept at the University of Tehran, Iran.

Copyright c Canadian Mathematical Society, 2016

120/ THE USE OF COORDINATE SYSTEMS BEFORE DESCARTES

4 Conclusions
The modern language of mathematics is fairly recent. Even Isaac Newton’s Prin-
cipia, originally published in 1687, used geometry to express derivatives and inte-
grals. The advantage of introducing specialized notation is tremendous for progress
in research, but it also makes understanding difficult for the uninitiated. Since
plain language and figures was all that mathematicians used in antiquity and the
Middle Ages, the development of the field was very slow. It started booming only
after the 18th century, when modern symbolism allowed a better understanding
of mathematical objects and of the relationships that govern them.
Nevertheless, the ancients already had some of the ideas that later crystallized
and formed new branches of mathematics. Apart from the coordinate system
mentioned above, a well-known example is that of the integral, which lies at the
foundation of calculus. Some two millennia before the birth of mathematical anal-
ysis, Archimedes approximated the length of a circle with regular polygons of many
sides, a procedure that leads to an integral in the limit. In a way, these examples
resemble the wheel, which was invented in prehistoric times. We added ball bear-
ings and tires, improved and extended their use, but wheels are still based on the
same idea. So we should not be too surprised that the ancient Greeks employed
the rectangular coordinate system about 2,200 years ago, long before Fermat and
Descartes understood its value.

References
[1] E.T. Bell, Development of Mathematics, 2nd. ed., McGraw-Hill, New York,
1945.
[2] V.J. Katz, A History of Mathematics, 3rd. ed., Addison-Wesley, New York,
2009.

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

Florin Diacu
Pacific Institute for the Mathematical Sciences and
Department of Mathematics and Statistics
University of Victoria
P.O. Box 3060 STN CSC
Victoria, BC, Canada, V8W 3R4
[email protected]

Crux Mathematicorum, Vol. 42(3), March 2016

 PROBLEMS /121

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 January
1, 2017, 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.

4121. Proposed by Leonard Giugiuc and Daniel Sitaru.

Let s be a fixed real number such that s ≥ 1. Let a, b, c and d be non-negative
numbers that satisfy a + b + c + d = 4s and ab + bc + cd + da + ac + bd = 6. Express
the minimum value of the product abcd in terms of s.

4122. Proposed by Daniel Sitaru.

Prove that for n ∈ N, the following holds
 en − 1 2n+1 (e − 1)(e2 − 1)(e3 − 1) · ... · (e2n − 1)
≤ .
n (2n)!

4123. Proposed by Michel Bataille.

In 3-dimensional Euclidean space, a line ` is perpendicular to the plane of the
acute triangle A0 B 0 C 0 at its orthocentre K. Let A, B, C be the midpoints of
B 0 C 0 , C√
0 0
A and A0 B 0 , respectively. Show that BC > KA and if D on ` satisfies
KD = BC 2 − KA2 , that the tetrahedron ABCD is isosceles. (A tetrahedron is
called isosceles if its opposite edges are congruent.)

4124. Proposed by George Apostolopoulos.

Let A1 , B1 and C1 be points on the sides BC, CA and AB of a triangle ABC such
that
A1 B B1 C C1 A
= = = k.
A1 C B1 A C1 B
Prove that
Å ã2 Å ã2 Å ã2 Å ã2 Å ã4
AA1 BB1 CC1 3k 2r
+ + ≥ ,
BC CA AB k2 + 1 R

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

Copyright c Canadian Mathematical Society, 2016

122/ PROBLEMS

4125. Proposed by Stephen Su and Cheng-Shyong Lee.

Start with a triangle A1 A2 A3 in the Euclidean plane and three nonzero real num-
bers `1 , `2 , `3 . Define Mk and Ck to be points on the line Ak+1 Ak+2 such that

Ak+1 Mk
= `k and Ck Mk+1 ||Ak Ak+1 , k = 1, 2, 3
Mk Ak+2

(with subscripts reduced modulo 3 and distances taken to be signed, so that Mk

is between Ak+1 and Ak+2 precisely when `k is positive). Denote by Rk the point
where Ck Mk+1 intersects Ck+1 Mk+2 , k = 1, 2, 3. Show that
Å ã2
[R1 R2 R3 ] 2 + `1 + `2 + `3 − `1 `2 `3
= ,
[A1 A2 A3 ] (1 + `1 )(1 + `2 )(1 + `3 )

where square brackets denote area.

4126. Proposed by Mihaela Berindeanu.

Let ABC be an acute-angled triangle. Prove that
…
X tan A2 tan B2 3
» ≥ .
cyc 1 − tan A
tan B 2
2 2

4127. Proposed by D. M. Bătineţu-Giurgiu and Neculai Stanciu.

Calculate √
n+1
Z (n+1)! x
lim √ f dx,
n→∞ n
n! n
where f : R∗+ → R∗+ is a continuous function.

4128. Proposed by Valcho Milchev and Tsvetelina Karamfilova.

Let An be the number of domino tilings of a rectangular 3 × 2n grid. Let L(2n, 2k)
be the number of domino tilings of the grid composed of two rectangular grids of
dimensions 3 × 2n and 3 × 2k with n ≥ 2 and k ≥ 1 (depicted below):

Prove that L(2n, 2n) = A2n .

Crux Mathematicorum, Vol. 42(3), March 2016

 PROBLEMS /123

4129. Proposed by Lorean Saceanu.

√
Let ABC be an acute-angle triangle and let γ = 3(2 − 3). Prove that

sec A + sec B + sec C ≥ γ + tan A + tan B + tan C.

4130. Proposed by Leonard Giugiuc.

Let a, b and c be nonnegative real numbers such that a + b + c = ab + bc + ac > 0.
Prove that √ √ √ √
n n
n
a+ b+ nc≥2 2
for any integer n ≥ 3 and determine the case for equality to hold.

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

4121. Proposé par Leonard Giugiuc et Daniel Sitaru.

Soit s une nombre réel tel que s ≥ 1. Soient a, b, c et d des nombres non négatifs
tels que a + b + c + d = 4s et ab + bc + cd + da + ac + bd = 6. Déterminer la valeur
minimale du produit abcd en terme de s.

4122. Proposé par Daniel Sitaru.

Démontrer que pour n ∈ N, l’inégalité suivante tient
 en − 1 2n+1 (e − 1)(e2 − 1)(e3 − 1) · ... · (e2n − 1)
≤ .
n (2n)!

4123. Proposé par Michel Bataille.

Dans l’espace euclidien 3-dimensionnel, une ligne ` est perpendiculaire au plan du
triangle aigu A0 B 0 C 0 , à son orthocentre K. Soient A, B et C les mi points de
B 0 C 0 , C 0 A0 et A√
0 0
B , respectivement. Démontrer que BC > KA et que si D sur `
satisfait KD = BC 2 − KA2 , alors le tétraèdre ABCD est isocèle. (Un tétraèdre
est dit isocèle si ses côtés opposés sont congrus.)

4124. Proposé par George Apostolopoulos.

Soient A1 , B1 et C1 des points sur les côtés BC, CA et AB du triangle ABC, tels
que
A1 B B1 C C1 A
= = = k.
A1 C B1 A C1 B
Démontrer que
Å ã Å ã Å ã Å ã2 Å ã4
AA1 2 BB1 2 CC1 2 3k 2r
+ + ≥ ,
BC CA AB k2 + 1 R
où R et r sont, respectivement, les rayons des cercles circonscrit et inscrit du
triangle ABC.

Copyright c Canadian Mathematical Society, 2016

124/ PROBLEMS

4125. Proposé par Stephen Su et Cheng-Shyong Lee.

Soient un triangle A1 A2 A3 dans le plan euclidien et trois nombres réels non nuls
`1 , `2 , `3 . Définissons Mk et Ck comme étant les points sur la ligne Ak+1 Ak+2 tels
que
Ak+1 Mk
= `k and Ck Mk+1 ||Ak Ak+1 , k = 1, 2, 3
Mk Ak+2
(les indices étant réduits modulo 3 et les distances comportant un signe, de façon
à ce que Mk se trouve entre Ak+1 et Ak+2 précisément lorsque `k est positif).
Dénotons par Rk le point d’intersection de Ck Mk+1 et Ck+1 Mk+2 , k = 1, 2, 3.
Démontrer que
Å ã
[R1 R2 R3 ] 2 + `1 + `2 + `3 − `1 `2 `3 2
= ,
[A1 A2 A3 ] (1 + `1 )(1 + `2 )(1 + `3 )
où les crochets dénotent une surface.

4126. Proposé par Mihaela Berindeanu.

Soit ABC un triangle aigu. Démontrer que
…
X tan A2 tan B2 3
» ≥ .
cyc 1 − tan A
tan B 2
2 2

4127. Proposé par D. M. Bătineţu-Giurgiu et Neculai Stanciu.

Calculer √
n+1
Z (n+1)! x
lim √ f dx,
n→∞ n
n! n
où f : R∗+ → R∗+ est une fonction continue.

4128. Proposé par Valcho Milchev et Tsvetelina Karamfilova.

Soit An le nombre de carrelages par dominos d’une grille 3 × 2n. Soit L(2n, 2k) le
nombre de carrelages par dominos d’une grille formée de deux grilles rectangulaires
de tailles 3 × 2n et 3 × 2k, pour n ≥ 2 et k ≥ 1, comme ci-bas:

Démontrer que L(2n, 2n) = A2n .

Crux Mathematicorum, Vol. 42(3), March 2016

 PROBLEMS /125

4129. Proposé par Lorean Saceanu.

√
Soit ABC un triangle aigu et soit γ = 3(2 − 3). Démontrer que

sec A + sec B + sec C ≥ γ + tan A + tan B + tan C.

4130. Proposé par Leonard Giugiuc.

Soient a, b et c des nombres réels non négatifs tels que a + b + c = ab + bc + ac > 0.
Démontrer que √ √ √ √
n n
n
a+ b+ nc≥2 2
pour tout entier n ≥ 3 et déterminer tout cas où l’égalité tient.

Math Quotes

To the pure geometer the radius of curvature is an incidental characteristic - like

the grin of the Cheshire cat. To the physicist it is an indispensable characteristic.
It would be going too far to say that to the physicist the cat is merely incidental to
the grin. Physics is concerned with interrelatedness such as the interrelatedness
of cats and grins. In this case the ”cat without a grin” and the ”grin without a
cat” are equally set aside as purely mathematical phantasies.

Sir Arthur Eddington, “The Expanding Universe.” Cambridge University

Press, 1988.

Copyright c Canadian Mathematical Society, 2016

126/ SOLUTIONS

SOLUTIONS
No problem is ever permanently closed. The editor is always pleased to consider for
publication new solutions or new insights on past problems.
Statements of the problems in this section originally appear in 2015: 41(3), p. 119–123.

4021. Proposed by Arkady Alt.

Let (an )n≥0 be a sequence of Fibonacci vectors defined recursively by a0 = a, a1 =
b and an+1 = an + an−1 for all integers n ≥ 1. Prove that, for all integers n ≥ 1,
the sum of vectors a0 + a1 + · · · + a4n+1 equals kai for some i and constant k.
We received nine correct solutions. We present the solution by David Stone and
John Hawkins (joint).
We shall prove that ā0 + ā1 + · · · + ā4n+1 = L2n+1 ā2n+2 , where Lk is the kth
Lucas number. We use some easily proven results. Here, Fk is the kth Fibonacci
number.
1. F0 + F1 + · · · + Fm = Fm+2 − 1.
2. F4n+2 = L2n+1 F2n+1
3. F4n+3 = L2n+1 F2n+2 + 1
4. āk = Fk−1 ā0 + Fk ā1 for k ≥ 1.
Therefore,
m
X m
X
āk = ā0 + (Fk−1 ā0 + Fk ā1 )
k=0 k=1
m
! m
!
X X
= ā0 + Fk−1 ā0 + ā1
k=1 k=1
= ā0 + (Fm+1 − 1)ā0 + (Fm+2 − 1)ā1
= Fm+1 ā0 + Fm+2 ā1 − ā1
= ām+2 − ā1

Hence, with m = 4n + 1,
4n+1
X
āk = ā4n+3 − ā1 = F4n+2 ā0 + F4n+3 ā1 − ā1
k=0
= (L2n+1 F2n+1 )ā0 + (L2n+1 F2n+2 )ā1
= L2n+1 (F2n+1 ā0 + F2n+2 ā1 )
= L2n+1 ā2n+2 .

Crux Mathematicorum, Vol. 42(3), March 2016

 SOLUTIONS /127

Editor’s Comments. Various solvers expressed the coefficient of ā2n+2 as L2n+1 ,

F2n + F2n+2 , and F 4n+2
F2n+1 and variations of these resulting from different indexing
of the Fibonacci sequence. Swylan pointed out that if the word ‘constant’ is
interpreted to mean ‘independent of n’, then the claim of the problem is false.
Perhaps ‘scalar’ would have been a better word.

4022. Proposed by Leonard Giugiuc.

In a triangle ABC, let internal angle bisectors from angles A, B and C intersect
the sides BC, CA and AB in points D, E and F and let the incircle of ∆ABC
touch the sides in M, N , and P , respectively. Show that
PA MB NC F A DB EC
+ + ≥ + + .
PB MC NA FB DC EA

We received eleven submissions, of which seven were correct, two were incorrect,
and two were incomplete. We present the solution by Titu Zvonaru.
Define x = N A = P A, y = P B = M B, and z = M C = N C; then

BC = y + z, CA = z + x, and AB = x + y.

By the angle bisector theorem we have

FA CA z+x DB AB x+y EC BC y+z
= = , = = , and = = .
FB BC y+z DC CA z+x EA AB x+y

We therefore have to prove that for positive real numbers x, y, z,

x y z z+x x+y y+z
+ + ≥ + + . (1)
y z x y+z z+x x+y

After clearing denominators what we must prove reduces to

x2 y 4 + y 2 z 4 + z 2 x4 + x3 y 3 + y 3 z 3 + z 3 x3 ≥ x3 yz 2 + x2 y 3 z + xy 2 z 3 + 3x2 y 2 z 2 . (2)

By the AM-GM inequality we have

x2 y 4 + y 2 z 4 + z 2 x4 ≥ 3x2 y 2 z 2 ,
x3 y 3 + z 3 x3 + z 3 x3 ≥ 3x3 yz 2 ,
3 3 3 3 3 3
y z +x y +x y ≥ 3x2 y 3 z, and
z 3 x3 + y 3 z 3 + y 3 z 3 ≥ 3xy 2 z 3 ,

which together imply that (2) holds. Equality holds if and only if x = y = z,
which immediately implies that the triangle is equilateral.
Editor’s Comments. Most submissions reduced our problem to equation (1), but
then algebra caused difficulties with two of the faulty arguments. The solution
from Salem Malikić neatly avoided calculations by remarking that (1) is known;

Copyright c Canadian Mathematical Society, 2016

128/ SOLUTIONS

see, for example, the Belarussian IMO Team preparation tests of 1997, where cal-
culations are much simplified by exploiting the cyclic symmetry of the inequalities.
Beware, however, that one must not assume noncyclic symmetry (as in one of the
incomplete submissions).

4023. Proposed by Ali Behrouz.

Find all functions f : R+ 7→ R+ such that for all x, y ∈ R with x > y, we have
Å ã
x
f + f (xf (y)) = f (xf (x)).
x−y

We received two correct submissions. We present the solution by Joseph Ling.

1
It is easy to see that f (x) = ∀x > 0 satisfies
x
Å ã
x
f + f (xf (y)) = f (xf (x)) (1)
x−y

whenever 0 < y < x. We show that there are no other solutions f : (0, ∞) → (0, ∞)
to (1).
First, we note that f isÄone-to-one.
ä For if 0 < y < x are such that f (y) = f (x) ,
x
then (1) implies that f x−y = 0, which is impossible.

Second, we note that f (x) ≤ x1 for all x > 0. For if there exists x > 0 such that
f (x) > x1 , then y = x − f (x)
1 x
satisfies 0 < y < x. But then x−y = xf (x) and (1)
will imply that f (xf (y)) = 0, which is impossible.
1
Now, suppose that 0 < y1 < y2 . Consider x = y2 + f (y1 ) . Then 0 < y1 < y2 < x
and (1) implies that
Å ã Å ã
x x
f + f (xf (y2 )) = f (xf (x)) = f + f (xf (y1 )) . (2)
x − y2 x − y1
x
By the definition of x, x−y = xf (y1 ) . So, (2) is reduced to f (xf (y2 )) =
Ä ä 2
x x
f x−y1 . Since f is one-to-one, we have xf (y2 ) = x−y 1
, and so, x = y1 + f (y1 2 ) .
Using this and the definition of x, we see that f (y1 1 ) − y1 = f (y1 2 ) − y2 . Since y1 and
1
y2 are arbitrary, f (y) − y is independent of y, and so, it must be some constant,
say, c. That is,
1
f (y) =
y+c
for all y > 0.
1
It remains to show that c = 0. Since f (x) ≤ x for all x, c ≥ 0. Furthermore, if

Crux Mathematicorum, Vol. 42(3), March 2016

 SOLUTIONS /129

c > 0, then for any 0 < y < x, we have

Å ã
x x−y y+c
f + f (xf (y)) = +
x−y x + c (x − y) x + c (y + c)
x−y y+c x+c
> + =
x + c (x + c) x + c (x + c) x + c (x + c)
= f (xf (x)) ,

a contradiction to (1). So, c = 0 and our proof is complete.

4024. Proposed by Leonard Giugiuc.

Let a, b, c and d be real numbers such that a2 + b2 + c2 + d2 = 4. Prove that

abc + abd + acd + bcd + 4 ≥ a + b + c + d

and determine when equality holds.

We received three correct solutions. We present the solution by Titu Zvononu,
modified by the editor.
We consider several cases separately.
Case 1. If a, b, c, d ≥ 0, then by the Cauchy-Schwarz Inequality, we have

(a + b + c + d)2 ≤ (12 + 12 + 12 + 12 )(a2 + b2 + c2 + d2 ) = 16.

Thus, a + b + c + d ≤ 4 from which the result follows immediately.

Case 2. If a, b, c, d ≤ 0, we set x = −a, y = −b, z = −c, and t = −d. Then,
x, y, z, t ≥ 0 and we would now like to show that

−(xyz + xyt + xzt + yzt) + 4 ≥ −(x + y + z + t) or

2(x + y + z + t) − (xyz + xyt + xzt + yzt) + 4 ≥ x + y + z + t. (1)

Since we know x2 + y 2 + z 2 + t2 = 4, we have by the first case (with a, b, c, d
replaced with x, y, z, t, respectively and symbolically), that

x + y + z + t ≤ 4. (2)

Then,

4(x + y + z + t) − 2(xyz + xyt + xzt + yzt)

= (x2 + y 2 + z 2 + t2 )(x + y + z + t) − 2(xyz + xyt + xzt + yzt)
= x(y − z)2 + y(x − t)2 + z(x − t)2 + t(y − z)2 + x(x2 + t2 )
+ y(y 2 + z 2 ) + z(z 2 + y 2 ) + t(t2 + x2 )
≥ 0,

which together with (2) implies (1).

Copyright c Canadian Mathematical Society, 2016

130/ SOLUTIONS

Case 3. If one of a, b, c, d is nonnegative and the other three are nonpositive.

Due to the symmetry in the given equation and the one we wish to prove, we
may assume that a ≥ 0 and b, c, d ≤ 0. Here we let y = −b, z = −c, and
t = −d. Then, a, y, z, t ≥ 0 with a2 + y 2 + z 2 + t2 = 4 and we wish to prove
ayz + ayt + azt − yzt + 4 ≥ a − y − z − t or

2(y + z + t) + ayz + ayt + azt − yzt + 4 ≥ a + y + z + t. (3)

Now,

4(y + z + t) + 2(ayz + ayt + azt − yzt)

= (a2 + y 2 + z 2 + t2 )(y + z + t) + 2(ayz + ayt + azt − yzt)
≥ y(z 2 + t2 ) − 2yzt
= y(z − t)2
≥ 0,

so, 2(y + z + t) + ayz + ayt + azt − yzt + 4 ≥ 4, thus, establishing (3), as desired,
since a + y + z + t ≤ 4 (see Case 1).
Case 4. If a, b ≥ 0 and c, d ≤ 0, we set z = −c and t = −d. Then a, b, z, t ≥ 0 such
that a2 + b2 + z 2 + t2 = 4 and we would like to show that

−abz − abt + azt + bzt + 4 ≥ a + b − z − t or

2(z + t) − abz − abt + azt + bzt + 4 ≥ a + b + z + t. (4)

Now,

4(z + t) − 2(abz + abt − azt − bzt)

= (a2 + b2 + z 2 + t2 )(z + t) − 2(abz + abt − azt − bzt)
= z(a − b)2 + t(a − b)2 + (z + t)(z 2 + t2 ) + 2azt + 2bzt
≥ 0.

So, 2(z + t) − abz − abt + azt + bzt ≥ 4, thus establishing (4) since a + b + z + t ≤ 4.
Case 5. If a, b, c ≥ 0 and d ≤ 0, we set t = −d. Then a, b, c, t ≥ 0 with a2 + b2 +
c2 + t2 = 4 and we would like to show that

abc − abt − act − bct + 4 ≥ a + b + c − t or

2t + abc − abt − act − bct + 4 ≥ a + b + c + t. (5)

Since a + b + c + t ≤ 4, to establish (5), it suffices to show that

(a2 + b2 + c2 + t2 )t + 2(abc − abt − act − bct) ≥ 0. (6)

Crux Mathematicorum, Vol. 42(3), March 2016

 SOLUTIONS /131

We let L denote the left-hand side of (6) and assume, without loss of generality,
that a ≥ b ≥ c. Note that

L = t(b − c)2 + t(a − t)2 + 2a(t − b)(t − c) (7)

2 2
and L = t(a − b) + t(c − t) + 2c(t − a)(t − b). (8)

If t ≤ b, then from (7) we can see that L ≥ 0 and if t ≥ b, then L ≥ 0 from (8).
Hence, we can conclude that (6) is true, as desired.
Examining the five cases, it is readily seen that equality can only hold in Case 5
when a = b = c = t; that is, if and only if (a, b, c, d) = (1, 1, 1, −1) and all its
permutations.

4025. Proposed by Dragoljub Miloşević.

Prove that for positive numbers a, b and c, we have
Å ã2 Å ã2 Å ã2 √
3 a 3 b 3 c 3
+ + ≥ 3.
2b + c 2c + a 2a + b

We received eleven correct solutions. We present the solution by Salem Madikić.

Let f (a, b, c) denote the left hand side of the given inequality. By the AM-GM
Inequality, we have
Å ã2
a a a
= √ »
3
= p
2b + c 3
a(2b + c)2 3 2b+c 2b+c
9 a· 3 · 3
3

3a
≥ √
9 a + 3 + 2b+c
2b+c
3


3
√
3 3 3a
= .
3a + 4b + 2c
Using similar inequalities involving the other two summands, we then have
√
3
X a
f (a, b, c) ≥ 3 3 . (1)
cyc
3a + 4b + 2c

Now, by the Cauchy-Schwarz Inequality, we have

Ç ã2 å ÅX »
X Å… a 2 ã X
a(3a + 4b + 2c) ≥ a2 .
3a + 4b + 2c

So, P 2 P 2
X a a a 1
≥P = P 2 P = . (2)
3a + 4b + 2c a(3a + 4b + 2c) 3 a + 6 ab 3
√
Substituting (2) into (1), f (a, b, c) ≥ 3 3 follows immediately.

Copyright c Canadian Mathematical Society, 2016

132/ SOLUTIONS

To achieve equality, we must have 3a = 2b+c, 3b = 2c+a, and 3c = 2a+b. Without

loss of generality, we may assume that max{a, b, c} = a. Then, 3a = 2b + c implies
2(a − b) + (a − c) = 0, so a = b = c. Conversely, it is readily checked that if
a = b = c, then equality holds.
Editor’s Comments. Using convexity and Jensen’s Inequality, Stadler proved that
P Ä a äk
in general, 2b+c ≥ 31−k for all k ≥ 0.

4026. Proposed by Roy Barbara.

Prove or disprove the following property: if r is any non-zero rational number,
then the real number x = (1 + r)1/3 + (1 − r)1/3 is irrational.
We received two correct solutions. We present the solution by Joseph DiMuro.
Assume both r and x are rational numbers with r 6= 0. Setting y1 = (1 + r)1/3
and y2 = (1 − r)1/3 we can show that
x3 − 2
= y1 y2 .
3x
3
−2
That means that y1 and y2 are the two solutions for y of y 2 − xy + x 3x = 0. But
using the quadratic formula, we also obtain
» 3 …
x ± x2 − 4x3x−8 x 8 − x3
y= = ± .
2 2 12x
This shows that if x is rational then y1 and y2 are contained in quadratic extensions
of Q. On the other hand, if r is rational then y1 = (1 + r)1/3 and y2 = (1 − r)1/3
are contained in cubic extensions of Q as well. Both of these can only be true if
y1 and y2 are rational numbers themselves.
Let r = ab , where a, b are relatively prime non-zero integers. Then y1 = ( b+a
b )
1/3

and y2 = ( b−ab )
1/3
. The fractions b+a b−a
b and b are in lowest terms, so for them
to be perfect cubes, their numerators and denominators must be perfect cubes.
Then we have an arithmetic progression b − a, b, b + a of cubes, which is known to
be impossible (e.g. see P. Dénes, Über die Diophantische Gleichung xl + y l = cz l ,
Acta. Math. 88 (1952) 241-251).
Editor’s Comments. The statement that there is no arithmetic progression of three
cubes can be proven with elementary number theory and is an interesting exercise.

4027. Proposed by George Apostolopoulos.

Let a, b and c be positive real numbers such that a + b + c = 3. Prove that
ab bc ac
+ + ≤ 1.
a + ab + b b + bc + c a + ac + c

We received 24 submissions of which 22 were correct and complete. We present 5

solutions, each of them insightful in a different way.

Crux Mathematicorum, Vol. 42(3), March 2016

 SOLUTIONS /133

Solution 1, by Ali Adnan.

Observe that
X ab X 9
≤1 ⇐⇒ a+b
≤ 9. (1)
cyc
a + ab + b cyc ab
+1

Now, from Cauchy-Schwarz Inequality,

9 9
a+b
= 1 1 ≤ a + b + 1,
ab +1 a + b +1

and adding up analogous such inequalities cyclically, (1) follows.

Solution 2, by Ali Adnan.

We note that the inequality is equivalent to
a+b 1
≥2 ⇐⇒ ≥ 1,
P P
cyc cyc 2ab
a + b + ab 2+ a+b

which follows easily from the AM-HM and Cauchy-Schwarz Inequalities:

1 1 (1 + 1 + 1)2
≥ ≥
P P
cyc 2ab cyc a+b
= 1,
2+ a+b 2+ 2
6+a+b+c
thus completing the proof.

Solution 3, by Henry Ricardo.

We have
X ab X 1 X 1 3
= 1 1 1 =
cyclic
a + ab + b
cyclic b
3 +1+
cyclic b
+ 1 + a1
a
Å ã
1 X a+b+1 1 2(a + b + c) + 3
≤ = = 1,
3 3 3 3
cyclic

where we have used the harmonic mean-arithmetic mean inequality.

Equality holds if and only if a = b = c = 1.

Solution 4, by Salem Malikić.

Using the inequality between arithmetic and geometric mean for positive reals x
and y we have p
x + xy + y ≥ 3 3 x2 y 2
with equality if and only if x = xy = y, that gives x = y = 1 and implying that
√3 xy
xy
≤ .
x + xy + y 3

Copyright c Canadian Mathematical Society, 2016

134/ SOLUTIONS

Using this inequality we have

√
3
√
3 √
ab bc ca ab + bc + 3
ca
+ + ≤ .
a + ab + b b + bc + c c + ca + a 3
Now, using Power-mean inequality, we have
√ √ …
s
3 3 √ (a+b+c)2
ab + bc + 3
ca 3 ab + bc + ca 3
3
≤ ≤ = 1.
3 3 3
where in the last step we used the well known inequality

3(ab + bc + ca) ≤ (a + b + c)2 .

This completes our proof.

In order to achieve equality we must have a = b = c = 1. It is easy to verify that
this is indeed an equality case.

Solution 5, by Leonard Giugiuc.

The inequality is equivalent to
1 1 1
1 1 + 1 1 + 1 1 ≤ 1.
a + b +1 b + c +1 c + a +1

By AM-HM Inequality,
1 1 4 1 1 4 1 1 4
+ ≥ , + ≥ , + ≥ .
a b a+b b c b+c c a c+a
From here,
1 1 1 1 1 1
1 1 + 1 1 + 1 1 ≤ 4 + 4 + 4 .
a + b +1 b + c +1 c + a +1 a+b +1 b+c +1 c+a +1

But
1 1 1 a+b b+c c+a
4 + 4 + 4 = + + .
a+b +1 b+c +1 c+a +1 a+b+4 b+c+4 c+a+4

x
Since the function f (x) = is concave if x > 0, then by Jensen’s Inequality
x+4
we get
Å ã
2(a + b + c)
f (a + b) + f (b + c) + f (c + a) ≤ 3f = 3f (2) = 1.
3
So,
a+b b+c c+a
+ + ≤ 1.
a+b+4 b+c+4 c+a+4

Crux Mathematicorum, Vol. 42(3), March 2016

 SOLUTIONS /135

4028. Proposed by Michel Bataille.

In 3-dimensional Euclidean space, a line ` meets orthogonally two distinct parallel
planes P and P 0 at H and H 0 . Let r and r0 be positive real numbers with r ≤ r0 ;
let C be the circle in P with center H, radius r, and let C 0 in P 0 be similarly
defined. For a fixed point M 0 on C 0 , find the maximum distance between the lines
` and M M 0 as M moves about the circle C (where the distance between two lines
is the minimum distance from a point of one line to a point of the other).
We received four correct solutions and will feature two of them that are quite similar
except that the first makes use of coordinates.
Solution 1, by Oliver Geupel.
We prove that the required maximum distance is r. We use Cartesian coordinates
such that H 0 = (0, 0, 0), M 0 = (r0 , 0, 0), and H = (0, 0, h) where h ∈ R. For
every point M on C, the distance between ` and M M 0 is not greater than |M H| =
r (because that distance is, by definition, the length of the shortest among all
segments joining a point of ` to a point of M M 0 , which is therefore at most
|M H|). Moreover, the distance between two non-intersecting lines is measured
along a line that is perpendicular to both. Put
Ç … å
r r2
M= · r, 1 − 02 · r, h ,
r0 r

−−→ −−→
which is on C. We have HM · HH 0 = 0 and
Ç … å Ç … å
−−− → −−→ r2 r2 r2 r2
M 0 M · HM = − r 0
, 1 − · r, h · , 1 − · r, 0 = 0,
r0 r02 r0 r02

so that HM is perpendicular to both ` and M M 0 . Therefore the distance between

the lines ` and M M 0 is |HM | = r, which completes the proof.

Solution 2, by Edmund Swylan.

For every point M on C, let Q be the plane orthogonal to ` that contains a point
P of M M 0 nearest to `. Let O, D, D0 , N, N 0 be the orthogonal projections of
`, C, C 0 , M, M 0 , respectively, onto Q. The distance between the lines ` and M M 0
projects to |P O|. Our problem is thereby reduced to a 2-dimensional problem:
Given circles D and D0 in the same plane with common centre O and
radii r and r0 , a fixed point N 0 on D0 , and a point N moving about D,
what is the maximum distance from O to N N 0 ?
The answer is r.
For r < r0 the maximum is achieved if and only if N N 0 is tangent to D. For r = r0
it is achieved if and only if N = N 0 and the line N N 0 degenerates into a point
which occurs when M M 0 is parallel to `.

Copyright c Canadian Mathematical Society, 2016

136/ SOLUTIONS

4029. Proposed by Paul Bracken.

Suppose a > 0. Find the solutions of the following equation in the interval (0, ∞):
∞
1 X n!
+ = x − a.
x + 1 n=1 (x + 1)(x + 2) · · · (x + n + 1)

We received four correct solutions and will feature two different ones.
Solution 1. We present a composite of the very similar solutions by Arkady Alt
and the proposer, Paul Bracken. Another similar solution was received from Oliver
Geupel.
It is clear that
1 1 1 1 1 1 2
− = , − − = ,
x x+1 x(x + 1) x x + 1 (x + 1)(x + 2) x(x + 1)(x + 2)
and
n! n!
−
x(x + 1)(x + 2) · · · (x + n) (x + 1)(x + 2) · · · (x + n + 1)
(n + 1)!
= .
x(x + 1)(x + 2) · · · (x + n + 1)
It therefore follows by induction that
n−1
1 1 X k! n!
− − = .
x x+1 (x + 1)(x + 2) . . . (x + k + 1) x(x + 1)(x + 2) . . . (x + n)
k=1

However, for x > 0,

n!
lim = 0,
n→∞ x(x + 1) . . . (x + n)

since
n! 1
= x

x


x(x + 1)(x + 2) · · · (x + n) x (x + 1) 2 + 1 ··· n +1
and x x Å ã
  1 1
(x + 1) + 1 ··· + 1 > 1 + x 1 + + ··· + .
2 n 2 n
Hence the left-hand side of the original equation is given by
∞
1 X n! 1
+ = .
x + 1 n=1 (x + 1)(x + 2) · · · (x + n + 1) x

Therefore the original equation is equivalent to x2 − ax − 1 = 0. This quadratic

equation has the following unique solution in (0, ∞):
1 p
xr = (a + a2 + 4).
2

Crux Mathematicorum, Vol. 42(3), March 2016

 SOLUTIONS /137

Solution 2, by Albert Stadler.

We note that
∞
X n!
n=1
(x + 1)(x + 2) · · · (x + n + 1)
∞
X Γ(x + 1)Γ(n + 1)
=
n=1
Γ(x + 1)(x + 2) · · · (x + n + 1)
∞
X Γ(x + 1)Γ(n + 1)
=
n=1
Γ(x + n + 1)
∞
X
= β(x + 1, n + 1)
n=1
X∞ Z 1
= tx (1 − t)n dt
n=1 0
Z 1
= tx−1 (1 − t) dt
0
1 1
= − , x > 0.
x x+1

The original equation is therefore equivalent to x1 = x − a or x2 − ax − 1 = 0. This

√
quadratic equation has exactly one positive root, which is xr = 12 (a + a2 + 4).

4030. Proposed by Paolo Perfetti.

a) Prove that 4cos t + 4sin t ≥ 5 for t ∈ [0, π4 ].
b) Prove that 6cos t + 6sin t ≥ 7 for t ∈ [0, π4 ].
There were six submitted solutions for this problem, four of which were correct.
We present the solution by Michel Bataille.
Lemma. Let u(t) = sin t cos t(cos√t + sin t). Then, u is an increasing function on
[0, π4 ] with u(0) = 0 and u( π4 ) = 22 .
√
Proof. u(0) = 0, u( π4 ) = 22 are immediate. A simple calculation gives the deriva-
tive of u:
u0 (t) = (cos t − sin t)((cos t + sin t)2 + sin t cos t).
For t ∈ (0, π4 ), cos t > sin t, hence u0 (t) > 0 and so u is increasing on [0, π4 ]. 2
cos t sin t π
a) Let f (t) = 4 + 4 . We show that f is increasing on [0, 4]
(the required
0
result then follows since f (0) = 5). To this end, we prove that f (t) > 0 for all
t ∈ (0, π4 ). We calculate
Å ã
sin t
f 0 (t) = (ln 4)4cos t cos t 4sin t−cos t −
cos t

Copyright c Canadian Mathematical Society, 2016

138/ SOLUTIONS

so that it is sufficient to prove that φ(t) > 0 for t ∈ (0, π4 ) where

φ(t) = (sin t − cos t)(ln 4) − ln(sin t) + ln(cos t).

(ln 4)u(t)−1
Now, we easily obtain φ0 (t) = sin t cos t with, from the lemma,
√
2 ln 4
(ln 4)u(t) − 1 < − 1 < 0.
2
Therefore, φ0 (t) < 0 for t ∈ (0, π4 ) and φ(t) > φ π4 = 0, as desired.

b) Similarly, we introduce g(t) = 6cos t + 6sin t whose derivative has the same sign
as ψ(t) = (sin t − cos t)(ln 6) − ln(sin t) + ln(cos t). Here,
Å ã
ln 6 1
ψ 0 (t) = · u(t) − ,
sin t cos t ln 6
√
and since 0 < ln16 < 22 , u(t) − ln16 (and so ψ 0 (t)) vanishes at a unique t0 in (0, π4 ).
From the lemma, we deduce that ψ 0 (t) < 0 if 0 < t < t0 and ψ 0 (t) > 0 if t0 < t < π4 .
Thus, ψ is decreasing on (0, t0 ] and increasing on [t0 , π4 ). Since ψ( π4 ) = 0, we must
have ψ(t0 ) < 0, and since lim+ ψ(t) = ∞, we deduce that for some α ∈ (0, t0 ), we
t→0
have ψ(t) > 0 if t ∈ (0, α), ψ(α) = 0 and ψ(t) < 0 if t ∈ (α, π4 ). Thus, g 0 (t) > 0 if
t ∈ (0, α) and g 0 (t) < 0 if t ∈ (α, π4 ) and so g(t) ≥ (min(g(0), g(π/4)) = 7 for all
t ∈ [0, π4 ].
Editor’s Comments. It turns out that AM-GM is too weak to prove this inequal-
ity when used right at the beginning; the resulting right-hand-side is too small.
However, one may use AM-GM in a step of the proof, as A. Stadler did, and have
things work out well; the Stadler solution is an impressive use of Taylor series and
clever bounds. As well, the ‘general’ inequality, acos(t) + asin(t) ≥ a + 1, is not
true over the required interval for every a > 1; plotting it for a = 10, for example,
shows this.

Crux Mathematicorum, Vol. 42(3), March 2016

 AUTHORS’ INDEX /139

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

Proposers
George Apostolopoulos, Messolonghi, Greece: 4124
Michel Bataille, Rouen, France: 4123
D. M. Bătineţu-Giurgiu and Neculai Stanciu, Romania : 4127
Mihaela Berindeanu, Bucharest, Romania : 4126
Valcho Milchev and Tsvetelina Karamfilova, Kardzhali, Bulgaria : 4128
Leonard Giugiuc, Drobeta Turnu Severin, Romania : 4130
Leonard Giugiuc and Daniel Sitaru, Romania : 4121
Lorean Saceanu, Harstad, Norway : 4129
Daniel Sitaru, Drobeta Turnu Severin, Romania : 4122
Stephen Su and Cheng-Shyong Lee : 4125

Solvers - individuals
Ali Adnan, Mumbai, India : 4027
Arkady Alt, San Jose, CA, USA : 4021, 4025, 4027, 4029
George Apostolopoulos, Messolonghi, Greece : 4022, 4027
Šefket Arslanagić, University of Sarajevo, Sarajevo, Bosnia and Herzegovina : CC163,
CC164, OC211, 4022, 4024, 4025, 4027
Fernando Ballesta Yagüe, I.E.S. Infante don Juan Manuel, Murcia, Spain: CC161,
CC163, CC164, CC165
Roy Barbara, Lebanese University, Fanar, Lebanon : 4026
Michel Bataille, Rouen, France : OC213, 4021, 4022, 4025, 4027, 4028, 4030
Brian D. Beasley, Presbyterian College, Clinton, USA : 4021
Ali Behrouz, Sharif University of Technology, Tehran, Iran : 4023
Paul Bracken, University of Texas, Edinburg, TX, USA : 4029
Matei Coiculescu, East Lyme High School, East Lyme, CT, USA : CC161, 4027
Joseph DiMuro, Biola University, La Mirada, CA, USA : 4021, 4026
Andrea Fanchini, Cantù, Italy : 4022, 4027, CC161, CC162, CC164, OC212, OC214
Oliver Geupel, Brühl, NRW, Germany : CC164, OC211, OC212, OC213, OC214, 4021,
4025, 4028, 4029
Leonard Giugiuc, Drobeta Turnu Severin, Romania : 4022, 4024, 4027
John G. Heuver, Grande Prairie, AB : CC161, CC163, OC212
Dag Jonsson, Uppsala, Sweden : 4027
Kee-Wai Lau, Hong Kong, China : 4027, 4030
Kathleen E. Lewis, University of the Gambia, Brikama, Republic of the Gambia: CC161,
CC163, CC164, CC165
Joseph M. Ling, University of Calgary, Calgary, AB : 4023
Salem Malikić, Simon Fraser University, Burnaby, BC : 4022, 4025, 4027
Phil McCartney, Northern Kentucky University, Highland Heights, KY, USA : 4025, 4027
Dragoljub Milošević, Gornji Milanovac, Serbia : 4025
Madhav R. Modak, formerly of Sir Parashurambhau College, Pune, India : 4027
Ricard Peiró i Estruch. IES “Abastos” València, Spain : 4027

Copyright c Canadian Mathematical Society, 2016

140/ AUTHORS’ INDEX

Paolo Perfetti, Dipartimento di Matematica, Università degli studi di Tor Vergata Roma,
Rome, Italy : 4030
Ángel Plaza, University of Las Palmas de Gran Canaria, Spain : CC164, CC165
C.R. Pranesachar, Indian Institute of Science, Bangalore, India : 4021
Cao Minh Quang, Nguyen Binh Khiem High School, Vinh Long, Vietnam : 4025, 4027
Henry Ricardo, Tappan, NY, USA : CC161, CC162, CC163, CC165. 4027
Michael John Rod, Montgomery, USA : CC164
Digby Smith, Mount Royal University, Calgary, AB : CC161, CC162, CC163, CC164,
OC211, 4025, 4027
Albert Stadler, Herrliberg, Switzerland: 4021, 4025, 4027, 4029, 4030
Edmund Swylan, Riga, Latvia : 4021, 4027, 4028
Konstantine Zelator, University of Pittsburgh, Pittsburgh, PA, USA : CC161, CC164,
OC215
Titu Zvonaru, Cománeşti, Romania : CC162, CC163, CC164, CC165, 4022, 4024,
4025, 4027

Solvers - collaborations
Dionne Bailey, Elsie Campbell, and Charles R. Diminnie, Angelo State University,
San Angelo, USA : 4027
John Hawkins and David R. Stone, Georgia Southern University, Statesboro, USA : 4021
Missouri State University Problem Solving Group : 4028
Skidmore College Problem Group: 4027

Crux Mathematicorum, Vol. 42(3), March 2016