Crux Mathematicorum

VOLUME 44, NO. 7 September / Septembre 2018

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

Contest Corner Editor John McLoughlin University of New Brunswick

Olympiad Corner Editor Alessandro Ventullo University of Milan
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 BGU, Be’er Sheva, Israel
Shaun Fallat University of Regina

Assistant Editors Chip Curtis Missouri Southern State University

Allen O’Hara University of Western Ontario

Guest Editors Kelly Paton University of British Columbia

Anamaria Savu University of Alberta
Andrew McEachern University of Victoria
Vasile Radu Birchmount Park Collegiate Institute
Aaron Slobodin Quest University Canada

Editor-at-Large Bill Sands University of Calgary

Managing Editor Denise Charron Canadian Mathematical Society

Copyright
c Canadian Mathematical Society, 2018
278

IN THIS ISSUE / DANS CE NUMÉRO

279 The Contest Corner: No. 67 John McLoughlin

279 Problems: CC331–CC335
280 Solutions: CC281–CC285
284 The Olympiad Corner: No. 365 Anamaria Savu
284 Problems: OC391–OC395
286 Solutions: OC331–OC335
291 Focus On . . . : No. 32 Michel Bataille
297 Linear Recurrence Sequences and Polynomial Division in
Number theory Valcho Milchev
302 Problems: 4361–4370
306 Solutions: 4261–4270

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

Crux Mathematicorum, Vol. 44(7), September 2018

 THE CONTEST CORNER /279

THE CONTEST CORNER

No. 67
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 February 1, 2019.
The editor thanks Valérie Lapointe, Carignan, QC, for translations of the problems.

CC331. Consider triangle ABC with ∠B = ∠C = 70◦ . On the sides AB

and AC, we take the points F and E, respectively, so that ∠ABE = 15◦ and
∠ACF = 30◦ . Find ∠AEF .

CC332. Find the largest integer k such that 135k divides 2016!. Note that
n! = 1 · 2 · 3 · · · n.

1
CC333. Let θ = arctan 2 + arctan 3. Find and simplify fully.
sin2 θ

CC334. Find the sum of all positive integers x for which x + 56 and x + 113
are perfect squares.

CC335. In the triangle ABC, BD is the median to the side AC, DG is

parallel to the base BC (G is the point of intersection of the parallel with AB).
In the triangle ABD, AE is the median to the side BD and F is the intersection
BC
point of DG and AE. Find .
FG

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

CC331. Considérez le triangle ABC tel que ∠B = ∠C = 70◦ . Sur les côtés
AB et AC, on prend les points F and E tels que ∠ABE = 15◦ and ∠ACF = 30◦ .
Trouvez ∠AEF .

CC332. Trouvez le plus grand entier k tel que 135k divise 2016!, où n! =
1 · 2 · 3 · · · n.

1
CC333. Soit θ = arctan 2 + arctan 3. Trouvez et simplifiez le plus
sin2 θ
possible.

Copyright
c Canadian Mathematical Society, 2018
280/ THE CONTEST CORNER

CC334. Trouvez la somme de tous les entiers positifs x tels que x + 56 et

x + 113 sont des carrés parfaits.

CC335. Dans le triangle ABC, BD est la médiane du côté AC et DG est

parallèle à la base BC (G est le point d’intersection de la droite parallèle avec le
côté AB). Dans le triangle ABD, AE est la médiane du côté BD et F est le point
BC
d’intersection des segments DG et AE. Trouvez le rapport .
FG

CONTEST CORNER
SOLUTIONS
Statements of the problems in this section originally appear in 2017: 43(7), p. 281–281.

CC281. In the Original Six era of the NHL, one particular season, each team
played 20 games (each team played the other 5 teams 4 times each). Each game
ended as a win, a loss or a tie (there were no ‘overtime losses’). At the end of this
certain season, the standings were as below. What were all the possible outcomes
for Montreal’s number of wins X, losses Y and ties Z?
Team Wins Losses Ties
Toronto 2 12 6
Boston 6 10 4
Detroit 7 12 1
New York 7 9 4
Chicago 11 7 2
Montreal x y z
Originally Question 6 from the 2015 W.J. Blundon Mathematics Contest.
We received 8 solutions, all of which were correct and complete. We present the
solution by Fernando Ballesta Yagüe.
The known results are:
• 2 + 6 + 7 + 7 + 11 = 33 wins.
• 12 + 10 + 12 + 9 + 7 = 50 losses.
• 6 + 4 + 1 + 4 + 2 = 17 ties.

Crux Mathematicorum, Vol. 44(7), September 2018

 THE CONTEST CORNER /281

As we eventually need to have the same total number of wins and losses (if a team
loses a match, another team has had to win that match), we have that x + 33,
which is the total number of wins, has to be equal to y + 50, which is the total
number of losses. Since the total number of matches played by Montreal’s team is
20, we have:

x + y + z = 20,
x + 33 = y + 50.

From that we can deduce that

x = y + 17 → (y + 17) + y + z = 20 → z = 3 − 2y.

Since the total number of ties, which is z + 17, has to be even (if a match has
ended in a tie, it has ended in a tie for both of the teams), we have that z > 0 and
z is odd. The only two possibilities for z > 0 are y = 0 and y = 1. If y = 0, then
z = 3 and x = 17; if y = 10, then z = 1 and x = 18. So the possible outcomes for
Montreal’s number of wins, losses and ties are:

x = 17, y = 0, z = 3
x = 18, y = 1, z = 1

CC282. Calculate the value of

Ä ä3 Ä ä3 Ä ä3 Ä ä3
34/3 − 31/3 + 35/3 − 32/3 + 36/3 − 33/3 + · · · + 32006/3 − 32003/3 .

Originally Question 7 from the 2015 W.J. Blundon Mathematics Contest.

We received 20 submissions of which 17 were correct and complete. We present
two solutions.
Solution 1, by Ivko Dimitrić.
Note that the difference of the two exponents of 3 of the pair of terms within each
set of parentheses is 1, which means that the first term in each pair is 3 times the
second one, i.e. those expressions are of the form
√
3
3(k+3)/3 − 3k/3 = 3 · 3k/3 − 3k/3 = 2 · 3k/3 = 2 · 3k ,

for k = 1, 2, . . . , 2003. Then the sum equals

√ √
3
√
3
√
3
(2 · 3)3 + (2 · 32 )3 + (2 · 33 )3 + · · · + (2 · 32003 )3
3

= 23 · 3 + 23 · 32 + 23 · 33 + · · · + 23 · 32003
= 8 · (3 + 32 + 33 + · · · + 32003 )
3 (32003 − 1)
=8·
2
= 12 (32003 − 1).

Copyright
c Canadian Mathematical Society, 2018
282/ THE CONTEST CORNER

Solution 2, by Miguel Amengual Covas.

The answer is 12 32003 − 1 .

We give a generalization which, in the case n = 2003, yields the solution to the
proposed problem.

1 3 2 3 3 3 n 3
Ä 4 ä Ä 5 ä Ä 6 ä Ä n+3 ä
33 − 33 + 33 − 33 + 33 − 33 + · · · + 3 3 − 3 3 (1)
î 1 ó3 î 2 ó3 î 3 ó3  n 3
= 3 3 (3 − 1) + 3 3 (3 − 1) + 3 3 (3 − 1) + · · · + 3 3 (3 − 1)
= 3 · 23 + 3 2 · 23 + 3 3 · 23 · · · + 3 n · 23
= 23 3 + 32 + 33 + · · · + 3n .

The sum 3 + 32 + 33 + · · · + 3n is a geometric progression with value

3n · 3 − 3 3 (3n − 1)
= .
3−1 2
Substituting this value in (1), we get

12 32003 − 1 .

CC283. Two bags, Bag A and Bag B, each contain 9 balls. The 9 balls in
each bag are numbered from 1 to 9. Suppose one ball is removed randomly from
Bag A and another ball from Bag B. If X is the sum of the numbers on the balls
left in Bag A and Y is the sum of the numbers of the balls remaining in Bag B,
what is the probability that X and Y differ by a multiple of 4?
Originally Question 10 from the 2015 W.J. Blundon Mathematics Contest.
We received eight submissions to this problem, all of which were correct. We
present the solution by Steven Chow, Miguel Amengual Covas, and Ballesta Yagüe
Fernando (all done independently), modified by the editor.
Let x, y be the number of the ball removed from the Bag A and B, respectively.
We have that
|X − Y | = | (45 − x) − (45 − y) | = | − x + y|.
X and Y differ by a multiple of 4 if and only if x and y are congruent modulo
4. From 1 to 9, the number of integers congruent to 0, 1, 2, 3 modulo 4 are 2, 3,
2, 2, respectively, so the probability that the numbers on the balls removed are
congruent modulo 4 is
Å ã2 Å ã2 Å ã2 Å ã2
2 3 2 2 7
+ + + = .
9 9 9 9 27
7
Therefore the probability that X and Y differ by a multiple of 4 is 27 .

Crux Mathematicorum, Vol. 44(7), September 2018

 THE CONTEST CORNER /283

CC284. Define the function f (x) to be the largest integer less than or equal
to x for any real x. For example, f (1) = 1, f (3/2) = 1, f (7/2) = 3, f (7/3) = 2.
Let
g(x) = f (x) + f (x/2) + f (x/3) + · · · + f (x/(x − 1)) + f (x/x).
a) Calculate g(4) − g(3) and g(7) − g(6).
b) What is g(116) − g(115)?
Originally Question 10 from the 2016 W.J. Blundon Mathematics Contest.
We received 8 correct solutions. We present the solution by Titu Zvonaru.
For positive integersn, we denote by d(n) the number of divisors of n. Let k =
1, 2, ..., n. Dividing n by k, we have n = pk + r, with 0 ≤ r < k.
If 0 < r, then n − 1 = pk + (r − 1). Hence, f nk = f n−1



k .
n n−1



If r = 0, then n − 1 = (p − 1)k + (k − 1). Hence, f k − f k = p − (p − 1) = 1.
We deduce that
g(n) − g(n − 1) = d(n),
Since 4 = 2 , 7 = 7, and 116 = 22 · 29, we have
2

g(4) − g(3) = d(4) = 3

g(7) − g(6) = d(7) = 2
g(116) − g(115) = d(116) = 6.

CC285. Find all values of k so that x2 + y 2 = k 2 will intersect the circle with
equation (x − 5)2 + (y + 12)2 = 49 at exactly one point.
Originally Question 6 from the 2016 W.J. Blundon Mathematics Contest.
We received 15 solutions, of which 10 were correct and complete and 5 were incom-
plete. One of the correct solutions was in Spanish. We present here the solution
by Dan Daniel.
The two circles intersect in one point (internally or externally) if either O1 O2 =
r1 + r2 or O1 O2 = pkr1 − r2 k, where we define O1 as (0,0) and O2 as (5,12). We
then have O1 O2 = (5 − 0)2 + (−12 − 0)2 = 13. So r1 = kkk, and r2 = 7.
The first case gives
kkk + 7 = 13 =⇒ kkk = 6 =⇒ k = ±6,
while the second gives
kkkk − 7k = 13 =⇒ kkk − 7 = ±13 =⇒ kkk = 20 =⇒ k = ±20.
Therefore k ∈ {−20, −6, 6, 20}.

Copyright
c Canadian Mathematical Society, 2018
284/ THE OLYMPIAD CORNER

THE OLYMPIAD CORNER

No. 365
Anamaria Savu
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 February 1, 2019.

The editor thanks Valérie Lapointe, Carignan, QC, for translations of the problems.

OC391. Let x1 , x2 , x3 , . . . be a sequence of positive integers such that for every

pair of positive integers (m, n) we have xmn 6= xm(n+1) . Prove that there exists a
positive integer i such that xi ≥ 2017.

OC392. In a convex hexagon ABCDEF all sides are equal and also AD =
BE = CF . Prove that a circle can be inscribed into this hexagon.

OC393. The point O is the center of the circumcircle Ω of the acute triangle
ABC. The circumcircle ω of the triangle AOC intersects the sides AB and BC
again at the points E and F . Moreover, the line EF divides the area of the triangle
ABC in half. Find ∠B.

OC394. In Chicago, there are 36 criminal gangs, some of which are at war with
each other. Each gangster belongs to several gangs and every pair of gangsters
belongs to a different set of gangs. It is known that no gangster is a member of
two gangs that are at war with each other. Furthermore, each gang that some
gangster does not belong to is at war with some gang he does belong to. What is
the largest possible number of gangsters in Chicago?

OC395. Let A1 , A2 , . . . , Ak ∈ Mn (R) be symmetric matrices. Prove that the

following statements are equivalent:
(a) det(A21 + A22 + · · · + A2k ) = 0;
(b) for all matrices B1 , B2 , . . . , Bk ∈ Mn (R) it holds

det(A1 B1 + A2 B2 + · · · + Ak Bk ) = 0.

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

Crux Mathematicorum, Vol. 44(7), September 2018

 THE OLYMPIAD CORNER /285

OC391. Soit x1 , x2 , x3 , . . . une suite d’entiers positifs telle que pour chaque
paire d’entiers positifs (m, n), on a xmn 6= xm(n+1) . Prouvez qu’il existe un entier
positif i tel que xi ≥ 2017.

OC392. Soit un hexagone convexe ABCDEF dont tous les côtés sont égaux et
dont AD = BE = CF . Prouvez qu’un cercle peut être inscrit dans cet hexagone.

OC393. Le point O est le centre du cercle circonscrit Ω du triangle acutangle

ABC. Le cercle circonscrit ω du triangle AOC intercepte les côtés AB et BC aux
points E et F . De plus, le segment EF divise l’aire du triangle ABC en deux.
Trouvez ∠B.

OC394. À Chicago, il y a 36 bandes criminelles, dont certaines sont en guerre

une contre l’autre. Chaque bandit appartient à diverses bandes et chaque paire
de bandits appartient à des groupes de bandes différents. Un bandit ne peut pas
appartenir à deux bandes qui sont en guerre. De plus, chaque bande à laquelle un
bandit n’appartient pas est en guerre avec certaines bandes auxquelles ce bandit
appartient. Quel est le plus grand nombre possible de bandits à Chicago?

OC395. Soit A1 , A2 , . . . , Ak ∈ Mn (R)des matrices symétriques. Prouvez que

les énoncés suivants sont équivalents :
(a) det(A21 + A22 + · · · + A2k ) = 0;
(b) pour toutes matrices B1 , B2 , . . . , Bk ∈ Mn (R) on a

det(A1 B1 + A2 B2 + · · · + Ak Bk ) = 0.

Copyright
c Canadian Mathematical Society, 2018
286/ THE OLYMPIAD CORNER

OLYMPIAD SOLUTIONS
Statements of the problems in this section originally appear in 2017: 43(5), p. 194–195.

OC331. Find all triples of nonnegative integers (x, y, z) and x ≤ y such that

x2 + y 2 = 3 · 2016z + 77.

Originally 2016 Greece National Olympiad Problem 1.

We received 4 solutions of which 2 solutions were correct and complete and 2 were
incorrect. We present the solution by Steven Chow.
We distinguish two cases.
First, assume z = 0. Then the equation reduces to x2 + y 2 = 80. Since 0 ≤ x ≤ y,
the equation is satisfied by x = 4 and y = 8, only.
Second, assume that z ≥ 1. For any integer a, a2 is congruent to 0, 1, 2, or 4
modulo 7, and a2 is congruent to 0 modulo 7 if and only if a is congruent to 0
modulo 7. Since for z ≥ 1, 3 · 2016z + 77 ≡ 0 (mod 7), it follows that x2 + y 2 ≡ 0
(mod 7). However, the only two remainders among 0, 1, 2, 4 that sum to 0
(mod 7) are 0, 0; therefore x ≡ 0 (mod 7) and y ≡ 0 (mod 7). Let x1 and y1 be
non-negative integers such that x
= 7x1 and y = 7y1 . In terms of x1 , y1 , and z,
the equation becomes 72 x21 + y12 = 3 · 2016z + 77. The left side is divisible by
72 . However, 77 is not divisible by 72 , and the right side is divisible by 72 if only
if z = 1. It follows that z = 1. Therefore x21 + y12 = (3 · 2016 + 77) /72 = 125.
Since 0 ≤ x1 ≤ y1 , (x1 , y1 ) are either (2, 11) or (5, 10). This implies that (x, y) are
either (14, 77) or (35, 70).
All non-negative integer solutions (x, y, z) of the given equation are the triplets
(4, 8, 0), (14, 77, 1), and (35, 70, 1).

OC332. Let ABCD be a convex quadrilateral. Show that there exists a

square A0 B 0 C 0 D0 (where vertices may be ordered clockwise or counter-clockwise)
such that A 6= A0 , B 6= B 0 , C 6= C 0 , D 6= D0 and AA0 , BB 0 , CC 0 , DD0 are all
concurrent.
Originally Problem 5 of Day 2 of the 2016 China National Olympiad.
We received only one solution that was incomplete.

OC333. Find all functions f : R → R so that for all real numbers x and y,

(f (x) + xy) · f (x − 3y) + (f (y) + xy) · f (3x − y) = (f (x + y))2 .

Originally 2016 USAMO Day 2 Problem 4.

Crux Mathematicorum, Vol. 44(7), September 2018

 THE OLYMPIAD CORNER /287

We received 2 correct solutions. We present the solution by Mohammed Aassila.

We will establish that the only solutions are the functions f (x) = 0 and f (x) = x2 .
We evaluate the statement equation at specific points to obtain several properties
of the function.
At x = 0 and y = 0: 2(f (0))2 = (f (0))2 implies f (0) = 0.
At x = 0 and y arbitrary: f (0)f (−3y)+f (y)f (−y) = (f (y))2 implies f (y)f (−y) =
(f (y))2 . Also f (y)f (−y) = (f (−y))2 , from which we conclude f (y) = f (−y),
equivalently f is even.
At x arbitrary and y = −x: (f (x) − x2 )f (4x) + (f (−x) − x2 )f (4x) = (f (0))2
implies that for arbitrary x

f (4x) = 0, or f (x) = x2 . (1)

Let t 6= 0 such that f (t) 6= 0, then f (t/4) = t2 /16. At x = t/4 and y = 3t/4:
(f (t/4)+3t2 /16)f (−2t)+(f (3t/4)+3t2 /16)f (0) = (f (t))2 implies t2 f (2t)/4 = f (t).
Consequently,

f (t) 6= 0 implies f (2t) 6= 0, f (4t) 6= 0, f (8t) 6= 0, . . . (2)

In addition, because f (t) 6= 0 implies f (4t) 6= 0 and because of (1), it follows that
f (t) = t2 . Hence for arbitrary x

f (x) = 0, or f (x) = x2 . (3)

Next, we shall prove that either f (x) = 0 for all x’s or f (x) = x2 for all x’s.
Assume the contrary: there exist a 6= 0 and b 6= 0 such that f (a) = 0 and
f (b) = b2 . Evaluate the statement equation at x = (a + b)/4 and y = (−a + 3b)/4:
Å Å ã ã
a+b (a + b)(−a + 3b)
f + f (a − 2b)
4 16
−a + 3b
Å Å ã ã
(a + b)(−a + 3b) 2
+ f + f (a) = (f (b))
4 16

or, equivalently,
Å Å ã ã
a+b (a + b)(−a + 3b) 2
f + f (a − 2b) = (f (b)) . (4)
4 16

Relation (4) implies that if f (b) = b2 , then f (a − 2b) 6= 0, f (a − 2b) = (a − 2b)2 ,

and then b and a are related via two polynomial equations:
2
(a + b)2 + (a + b)(−a + 3b) (a − 2b) = 16b2 , or

2
(a + b)(−a + 3b) (a − 2b) = 16b2 . (5)

Copyright
c Canadian Mathematical Society, 2018
288/ THE OLYMPIAD CORNER

Since a polynomial has finitely many roots, there exist finitely many b such that
f (b) 6= 0. However, this contradicts (2) that states the existence of infinitely many
b’s such that f (b) 6= 0. Therefore our assumption that there exist a 6= 0 and b 6= 0
such that f (a) = 0 and f (b) = b2 is incorrect.
In conclusion, there are only two solutions for the statement equation: f (x) = 0
for all x, or f (x) = x2 for all x.

OC334. Let p be an odd prime number. For positive integers k satisfying

1 ≤ k ≤ p − 1, the number of divisors of kp + 1 between k and p exclusive is ak .
Find the value of a1 + a2 + · · · + ap−1 .
Originally 2016 Japan Mathematical Olympiad Finals Problem 1.
We received 2 solutions of which 1 was correct and 1 was incomplete. We present
the solution by Steven Chow slightly modified by the editor.
For each 1 ≤ k ≤ p − 1, let Ak be the set consisting of the divisors of kp + 1
between k and p exclusive. Therefore for each k, ak = |Ak | and

a1 + · · · + ap−1 = |A1 | + · · · + |Ap−1 | .

We establish two facts about the sets A1 , . . . , Ap−1 .

First, A1 , . . . , Ap−1 have no common elements, equivalently are mutually exclusive.
Assume the opposite that there exists integers k1 , k2 , and j such that

1 ≤ k1 ≤ p − 1, 1 ≤ k2 ≤ p − 1, k1 < k2 < j < p

and j belongs to the intersection of Ak1 and Ak2 . It follows that j is a common
divisor of k1 p + 1 and k2 p + 1, and consequently j is a divisor of the difference

(k2 p + 1) − (k1 p + 1) = (k2 − k1 ) p.

Since p is prime and j < p we can conclude that j is a divisor of k2 − k1 . However

this implies j ≤ k2 − k1 < k2 < j, which is a contradiction. Therefore the sets
A1 , . . . , Ap−1 are mutually exclusive.
Second, the union of the sets A1 , . . . , Ap−1 is {2, 3, . . . , p − 1}. Let j be an integer
such that 2 ≤ j ≤ p − 1. Since j and p are relatively prime integers, there exist two
integers k and m such that kp = mj−1 and 1 ≤ k ≤ j−1. A short argument for this
fact is provided in the editor’s comments at the end of the solution. Rearranging
the last equation into kp + 1 = mj, we find that j is a divisor of kp + 1 , and hence
j ∈ Ak . Since j was arbitrary, the union A1 ∪ · · · ∪ Ap−1 is {2, 3, . . . , p − 1}, a set
with p − 2 elements.
In conclusion p − 2 = |A1 ∪ · · · ∪ Ap−1 | = |A1 | + · · · + |Ap−1 | = a1 + · · · + ap−1 .
Editor’s comments. Regardless of the value of p, we have ap−1 = 0 as there are no
integers between p − 1 and p. Therefore the question could have been asked with
1 ≤ k ≤ p − 2 and led to the conclusion that p − 2 = a1 + · · · + ap−2 .

Crux Mathematicorum, Vol. 44(7), September 2018

 THE OLYMPIAD CORNER /289

The proof used the fact that if j and p are relatively prime integers then there
exist two integers k and m such that kp = mj − 1 and 1 ≤ k ≤ j − 1. This can
be obtained by looking at the remainders of p, 2p, . . . , (j − 1)p when divided by
j. There are j − 1 such remainders, all different and taking only j − 1 values:
1, . . . , j − 1. Therefore there exists k such that 1 ≤ k ≤ j − 1 and the remainder
of kp when divided by j is j − 1. Equivalently, there exists k and m such that
1 ≤ k ≤ j − 1 and kp = mj + j − 1, or kp = (m + 1)j − 1.

OC335. Medians AMA , BMB and CMC of a triangle ABC intersect at M .

Let ΩA be the circumcircle of the triangle that passes through the midpoint of
AM and is tangent to BC at MA . Define ΩB and ΩC analogously. Prove that
ΩA , ΩB and ΩC intersect at one point.
Originally 2016 All Russian Olympiad Grade 11 Day 2 Problem 8.
We received 2 solutions of which 1 was correct and 1 was incomplete. We present
the solution by Steven Chow slightly modified by the editor.
We use barycentric coordinates. Let (1, 0, 0) = A, (0, 1, 0) = B, and (0, 0, 1) = C.
Let a = BC, b = CA, and c = AB. Let x, y, and z represent the first, second
and third coordinate of an arbitrary point, respectively. Let u, v, and w be the
numbers such that the equation of ΩA is

0 = −a2 yz − b2 zx − c2 xy + (ux + vy + wz) (x + y + z) .

Since MA = (0, 1/2, 1/2) is on ΩA , v + w = a2 /2.

Since ΩA is tangent to the line BC described by x = 0, the equation

0 = −a2 yz + (vy + wz) (y + z)

has unique solution y and z. This is equivalent to the quadratic equation

0 = v(y/z)2 − (a2 /2)(y/z) + w

in y/z having unique solution, or the equation having zero discriminant. The zero

2
discriminant implies that vw = a2 /4 . Therefore v = w = a2 /4.
Since the midpoint of segment AM has barycentric coordinates (4/6, 1/6, 1/6) and
lies on ΩA ,
0 = −a2 − 4b2 − 4c2 + 4u + a2 /2 (6) ,

which leads to u = −a2 /12 + b2 /6 + c2 /6.

Therefore the equation of ΩA is
ÅÅ 2
b2 c2 a2 a2
ã ã
a
0 = −a2 yz − b2 zx − c2 xy + − + + x + y + z (x + y + z) . (1)
12 6 6 4 4

Similarly, the equations of ΩB and ΩC are the cyclic forms of (1).

Copyright
c Canadian Mathematical Society, 2018
290/ THE OLYMPIAD CORNER

Next, we claim that the point Ω with barycentric coordinates in proportion

Ä
2
2
2 ä
−2a2 + b2 + c2 : a2 − 2b2 + c2 : a2 + b2 − 2c2
satisfies the equation of ΩA , and hence lies on ΩA . Similarly, by cyclicity, this
point is also on ΩB and ΩC , and so is the intersection point of the three circles.
The intersection point was identified by computing the radical centre of the three
circles, which is the point of intersection of two of the three radical axes. The
equation of a radical axis is computed by subtracting the equations of two circles.
To see that the point Ω satisfies the equation (1) of ΩA , we introduce some nota-
tions m = −2a2 + b2 + c2 , n = a2 − 2b2 + c2 , and p = a2 + b2 − 2c2 . Note that
m + n + p = 0. Consequently, after taking the square of
m2 + n2 + p2 = −2(mn + np + pm),
we have:

2
m2 + n2 + p2 = 4(mn + np + pm)2
= 4 m2 n2 + n2 p2 + p2 m2 + 2mnp (m + n + p)

= 4 m 2 n2 + n2 p 2 + p 2 m 2 .


(2)
Also, since m = −(n + p),
m2 + n2 + p2 = (n + p)2 + n2 + p2 = 2(n2 + p2 + np). (3)
The coordinates of Ω are in proportion m2 : n2 : p2 . We evaluate the second


part of the equation (1) of ΩA at xΩ = m2 , yΩ = n2 , and zΩ = p2 . We use (2),
(3), and m + n + p = 0 in the following computations:
ÅÅ 2
b2 c2 a2 a2
ã ã
a
− + + xΩ + yΩ + zΩ (xΩ + yΩ + zΩ )
12 6 6 4 4
Å 2 2 2ã
a b c a2
= − + + xΩ (xΩ + yΩ + zΩ ) + (xΩ + yΩ + zΩ )2
3 6 6 4
2
m
a
2
= m2 m2 + n2 + p2 + m2 + n2 + p2
6 4
m
= m2 n2 + p2 + np + a2 (m2 n2 + n2 p2 + p2 m2 )


3  m 2 2  2 m 2 2 mnp
= a2 n2 p2 + a2 + m p + a + n m + m2
3 3 3
2 2 2

2 n 2 2  2 p 2 2 2 mnp
=a n p + b + m p + c + n m +m
3 3 3
2
m np
= a2 n2 p2 + b2 p2 m2 + c2 m2 n2 + (m + n + p)
3
= a2 n2 p2 + b2 p2 m2 + c2 m2 n2 = a2 yΩ zΩ + b2 zΩ xΩ + c2 xΩ yΩ .

Therefore ΩA , ΩB , and ΩC intersect in one point.

Crux Mathematicorum, Vol. 44(7), September 2018

 MICHEL BATAILLE /291

FOCUS ON...
No. 32
Michel Bataille
Harmonic Ranges and Pencils
Introduction
Elementary properties of harmonic conjugacy can lead to simple and elegant so-
lutions to some geometry problems. Before considering examples, let us recall the
basic definitions. Let A, B, C, D be four distinct points on a line. We say that C, D
are harmonic conjugates w.r.t A, B when C, D divide AB in the same ratio, that
CA DA
is, if CB = − DB (here and in what follows, the bar indicates signed distance and
w.r.t. means ”with respect to”). Clearly, the latter is equivalent to AD
AC
= − BD
BC
,
meaning that A, B are harmonic conjugates w.r.t. C, D. If either condition is
satisfied, we say that A, B, C, D is a harmonic division or a harmonic range. Let
−−→ −→ −−→ −−→
I be the midpoint of AB. Starting with AD = −k AC and BD = k BC for some
−→ −
→ −→ −→
real number k, easy manipulations give ID = k IA and IA = k IC and conversely.
2
Thus, the condition IA = IC.ID can also be used to prove the harmonicity of
the range of collinear points A, B, C, D.

Harmonic pencil
Let `1 , `2 , `3 , `4 be four distinct lines which are either parallel or concurrent, and
let transversals m, m0 meet them in A, B, C, D and in A0 , B 0 , C 0 , D0 , respectively.
C 0 A0 D 0 A0
If `1 , `2 , `3 , `3 are parallel, then C 0 B0
CA
= CB and D 0 B0
DA
= DB so that A0 , B 0 , C 0 , D0
is a harmonic range as soon as A, B, C, D is one (Figure 1).

Figure 1
This conservation of harmonicity remains true when `1 , `2 , `3 , `4 are concurrent
lines. To prove this, we shall use the following lemma (for a proof, we refer the
reader to [1] p. 169).
Let A, B, C, D be four distinct points on a line and S a point not on
this line. Let the parallel to SC through A intersect SD at M and
SB at E. Then A, B, C, D is a harmonic range if and only if M is the
midpoint of AE (Figure 2).

Copyright
c Canadian Mathematical Society, 2018
292/ FOCUS ON... HARMONIC RANGES AND PENCILS

Figure 2
Now, let `1 , `2 , `3 , `4 be concurrent at S and let m intersect them along a harmonic
range A, B, C, D. If m0 intersects them along A0 , B 0 , C 0 , D0 , we draw the parallels
to SC through A and through A0 , which intersect SD and SB, respectively at M
and E and at M 0 and E 0 (Figure 2). Since A, B, C, D is a harmonic range, M is
the midpoint of AE; since AE is parallel to A0 E 0 , M 0 is the midpoint of A0 E 0 and
so A0 , B 0 , C 0 , D0 is a harmonic range as well.
This justifies the following definition: `1 , `2 , `3 , `4 is called a harmonic pencil when
a transversal m intersects `1 , `2 , `3 , `4 along a harmonic range. From the lemma
above, an example is given by the lines `, AM, AB, AC if M is the midpoint of the
side BC of ∆ABC and ` is the parallel to BC through A.
We are now ready to examine a few situations involving harmonic ranges or pencils
and illustrate them with problems.

An angle and its bisectors

If ABC is a triangle (with AB 6= AC) and the internal and external bisectors of
∠BAC meet BC at D and D0 , respectively, we know that D, D0 divide BC in
0 0
the ratio AB
AC . Thus, B, C, D, D is a harmonic range and AB, AC, AD, AD is a
0
harmonic pencil. Note that AD, AD are perpendicular. Interestingly, a kind of
converse holds (easily proved or see [1] p. 170):
Let `1 , `2 , `3 , `4 be a harmonic pencil of concurrent lines at S. If `3 , `4
are perpendicular, then they are the axes of symmetry of `1 , `2 .
To illustrate these results, we consider problem 3036 [2005 : 175 ; 2006 : 244],
slightly modified:
Let A, B, C be three distinct collinear fixed points. Let M be an arbi-
trary point not on the line ABC. The internal angle bisector of ∠M AB
intersects the line M B at a point X. The perpendicular at A to the
line AX intersects the line M C at a point Y .
(a) Prove that the line XY passes through a fixed point D.
(b) Let Z be the projection of the point A onto the line XY . Prove
that ZA is a bisector of ∠BZC.

Crux Mathematicorum, Vol. 44(7), September 2018

 MICHEL BATAILLE /293

Figure 3

(a) Let XY intersect M A at U and the line m through A, B, C at D (Figure 3).

Since AY and AX are the bisectors of ∠M AB, the pencil AY, AX, AM, AB is
harmonic. Therefore, the range Y, X, U, D is harmonic and in consequence the
pencil M Y ,M B,M U ,M D is harmonic. Finally, considering the latter and the
transversal m, we see that C, B, A, D is harmonic and conclude that XY always
passes through the harmonic conjugate of A w.r.t. B, C.

(b) Since ZA and ZD are perpendicular and C, B, A, D is a harmonic range,

ZA, ZD are the bisectors of ∠BZC.

About polars with respect to a circle

Consider a circle Γ with centre O and radius r and let M be a point distinct from
−−→ −−→
O. The locus ΠM of points P such that the dot product OM · OP equals r2 is
the polar of M w.r.t. Γ. If M lies on Γ, M itself is a point of ΠM ; otherwise,
denoting by A and B the points of intersection of Γ and the line OM , we see that
the harmonic conjugate M 0 of M w.r.t. A, B is a point of ΠM (since O is the
midpoint of AB and OM .OM 0 = OA2 ). Moreover, P is on ΠM if and only if
−−→ −−→ −−→ −−−→0 −−→ −−−→
OM · OP = OM · OM , which is equivalent to OM · M 0 P = 0, and therefore ΠM
is the perpendicular to OM through M 0 (Figure 4). In the same way, we obtain
that if M is on Γ, then ΠM is the tangent to Γ at M .

Figure 4

Copyright
c Canadian Mathematical Society, 2018
294/ FOCUS ON... HARMONIC RANGES AND PENCILS

In passing, note that M 0 is the inverse of M in Γ and that M is on ΠN as soon as

N is on ΠM (polar reciprocity).

The result that M, M 0 , A, B is a harmonic range can be generalized as follows:

If a line through M intersects ΠM at N and the circle Γ at U and V ,

then M, N, U, V is a harmonic range.

Let the perpendicular bisector of U V intersect U V at I, ΠM at K and the per-

pendicular to OM through U at L (Figure 4). Clearly, U is the orthocenter of
∆OM L, hence OU is perpendicular to M L. But M is on ΠK (since K is on ΠM )
and so U V is the polar of K. It follows that K is on the polar of U , that is, the
tangent to Γ at U . Being both perpendicular to OU , M L and KU are parallel, and
−→ −→ −→ −→ −−→ −→
so are KN and LU . As a result, if IU = k IN , then IL = k IK and so IM = k IU .
The result follows.

As an application, we propose here a problem of the 50th Olympiad of Moldova

[2009 : 377]:

The quadrilateral ABCD is inscribed in a circle. The tangents to the

circle at A and C intersect at a point P not on BD and such that
P A2 = P B.P D. Prove that BD passes through the midpoint of AC.

Let Γ be the circumcircle of ABCD and let O be its centre. The line P D intersects
Γ again at B 0 with B 0 6= B (since P is not on BD). Since P B 0 · P D is the power
of P w.r.t. Γ, we have P B 0 · P D = P A2 = P B · P D, so that P B 0 = P B and the
line OP is the perpendicular bisector of BB 0 (Figure 5).

Figure 5

It follows that OP is a bisector of the angle ∠BP D and so is the line m perpen-
dicular to OP at P . As a result, P O, m, P D, P B is a harmonic pencil and the line
BD intersects P O and m at Q and S such that Q, S, B, D is a harmonic range.
From the property above, we then deduce that the polar of Q w.r.t. Γ passes
through S, hence is m (since m is perpendicular to OQ). By polar reciprocity, Q
is on the polar of P , which is AC, and the conclusion immediately follows.

Crux Mathematicorum, Vol. 44(7), September 2018

 MICHEL BATAILLE /295

Constructions with the straightedge alone

Harmonic ranges or pencils can be constructed with the straightedge alone. This
interesting feature rests upon the following property:
Let `1 , `2 be two lines intersecting at S and A a point not on these lines.
Through A we draw two transversals intersecting `1 , `2 at M1 , M2 and
N1 , N2 . If the lines M1 N2 and M2 N1 intersect at U , then SA, SU, `1 , `2
is a harmonic pencil.
The proof is easy: If B is the harmonic conjugate of A w.r.t. M1 , M2 , the line
through A, N1 , N2 is a transversal of the harmonic pencil U A, U B, U M1 , U M2 ,
hence intersects U B at C such that A, C, N2 , N1 is a harmonic range. The line
SB, which also passes through C, must coincide with SU .
Of course, if `1 , `2 are parallel, a similar conclusion holds provided that SA and
SU are replaced by the parallels to `1 , `2 through A and U , respectively.
To see this at work, a good example is Problem 2965 [2004 : 367, 370; 2005 :
405]:
Let ABCD be a parallelogram. Using only an unmarked straightedge,
find a point M on AB such that AM = 51 AB.
Here are the steps of the construction. First, we obtain the reflection B1 of B in A
by drawing ` such that DC, DA, DB, ` is a harmonic pencil (Figure 6a). The line
` intersects the line AB at B1 . Second, we construct B10 such that A, B, B1 , B10
is a harmonic range (Figure 6b). Finally, we repeat the first two steps with B10
instead of B. This yields the desired point M as the harmonic conjugate w.r.t.
A, B of the reflection B2 of B10 in A.

Figure 6a Figure 6b

B10 A B1 A
Indeed, we have = −B = − 12 and so
B10 B 1B

AM B2 A B10 A −B10 A/B10 B 1/2 1

= =− 0
= 0 0 0
= =
MB B2 B AB1 + AB −B1 A/B1 B + AB/B1 B 1/2 + 3/2 4

The relation AM = 15 AB follows.

Copyright
c Canadian Mathematical Society, 2018
296/ FOCUS ON... HARMONIC RANGES AND PENCILS

Exercises
1. Through a point P exterior to a given circle pass a secant and a tangent to
the circle. The secant intersects the circle at A and B, and the tangent touches
the circle at C on the same side of the diameter through P as A and B. The
projection of C onto the diameter is Q. Prove that QC bisects ∠AQB. (Set at
the competition Baltic Way in 2004.)
2. The standard construction for bisecting a line segment involves the use of two
arcs and one straight line. Show that it can, in fact, be done with straight lines and
just one arc. (Problem 88.I of the Mathematical Gazette, proposed in November
2004.)

Reference
[1] N. Altshiller-Court, College Geometry, Dover, 2007.

Crux Mathematicorum, Vol. 44(7), September 2018

 VALCHO MILCHEV /297

Linear Recurrence Sequences and

Polynomial Division in Number Theory
Valcho Milchev

1 Introduction

The paper discusses some characteristics of linear recurrence sequences. It is in-

spired by the article “Polynomial Division in Number Theory” by James Rickards
(Crux, 43(10), December 2017.).
Let us look at the following Diophantine equation which is symmetric in the un-
knowns x and y:
x2 + pxy + y 2 = q, (1)
where p and q are integers, and p > 2 (if p ≤ 2, then there can only be a finite
number of integer solutions). The integer solutions for p > 2 satisfy a linear
recurrence equation
an+2 = pan+1 − an , (2)
whose terms are integer coordinates on the hyperbola with equation (1).
Theorem 1 Let p > 2 in the Diophantine equation. Then for any solution (x, y):
(i) if q < 0, x and y always have the same sign.
(ii) if q > 0, x and y, if they exist, can have the same or opposite signs.
Theorem 2 If the Diophantine equation (1) given q 6= 0 and p > 2 has solutions,
then there is an infinite number of such solutions, which are pairs of neighbour
terms of an integer sequence with linear recurrence equation (2).
Proof. Let’s assume that the equation (1) has a solution (x1 , y1 ) in positive inte-
gers. By symmetry, we can assume without loss of generality that x1 ≤ y1 . Thus,
considered as a quadratic in x, the equation x2 − pxy1 + y12 − q = 0 has a root x1 .
As p > 2, the discriminant D = (p − 4) y12 + 4q > 0, so there exists a second root
x2 . By Vieta’s formulas, we have x2 = py1 − x1 ; as p > 2 and x1 ≤ y1 , we have
x2 > y1 .
Considering the equation x22 − px2 y + y 2 − q = 0 as a quadratic in y, we find
that a positive integer y2 exists so that y2 > x2 and x22 − p2 x2 + y22 − q = 0.
Continuing this process, we obtain an increasing sequence of natural numbers.
Every three neighbour terms respectively fulfil Vieta’s formula. The process is
known as “Vieta jumping”, and yields an endless linear recurrence sequence of
integers with a recurrence equation (2). Analogously, we can develop a jumping
“downwards”. Every two neighbour terms of the resulting sequence {an } satisfy
the Diophantine equation (1). 2

Copyright
c Canadian Mathematical Society, 2018
298/ LINEAR RECURENCES AND POLYNOMIAL DIVISION IN NUMBER THEORY

Figure 1 shows the two branches of the hyperbola x2 − 3xy + y 2 = 1 (an example
with q > 0). The integer points on its graph have coordinates that are pairs
of consecutive terms of the sequence {. . . , −8, −3, −1, 0, 1, 3, 8, . . . } satisfying
an+2 = 3an+1 −an . Figure 2 shows an example with q < 0, the hyperbola x2 −3xy+
y 2 = −1. The integer points have coordinates that are pairs of consecutive terms
of the sequence {. . . , 5, 2, 1, 1, 2, 5, . . . }, which also satisfies an+2 = 3an+1 − an .

Figure 1 Figure 2

Theorem 3 If the Diophantine equation (1) with p > 2, q > 0 has integer so-
lutions, then q is a perfect square if and only if there is an integer point on the
hyperbola, one of whose coordinates is equal to zero.
Criterion 1 If the Diophantine equation (1) with p > 2, q > 0, and q ≤ p has a
solution, then q is the perfect square of an integer.
Below, we will apply these results to some contest problems.
Problem 1 (1988 International Mathematical Olympiad) Let a and b be
a2 + b2
such positive integers that ab + 1 divides a2 + b2 . Prove that the number
ab + 1
is a perfect square.
(The problem was considered the hardest at the 29th IMO. It is said that five
number theorists from Australia were not able to solve the problem in five hours!
With modern techniques, this problem does not look that tough.)
Solution. We must solve the Diophantine equation a2 − pab + b2 = q for p =
q = 1. According to the formulated criterion, if this equation has a solution in
integers, then p is a perfect square of an integer. All solutions are given by a linear
recurrence sequence which includes the number zero.
Let p = m2 . Since we have already concluded that zero is part of the se-
quence, then by “Vieta Jumping” the pair (0, m) yields a sequence of polynomi-
als 0, m, m2 , m3 , m5 − m, m7 − 2m3 , . . . , obtained through the recurrence relation
an+1 = m2 an − an−1 . This sequence of polynomials provides all solutions of the
problem. The hyperbola is of the type illustrated in Figure 1, and both branches
cross the coordinate axis in integer points. 2
Problem 2 (1998 Canadian Mathematical Olympiad, National Round)

Crux Mathematicorum, Vol. 44(7), September 2018

 VALCHO MILCHEV /299

Let the sequence {an } be defined by a0 = 0, a1 = m, an+1 = m2 an − an−1 . Prove

a 2 + b2
that all integer solutions (a, b) of the equation = m given a < b coincide
ab + 1
with the pairs (an , an+1 ).
Problem 3 (Kvant magazine, M1225) Prove that if for the natural numbers
a2 + b2 a 2 + b2
a and b the number is also natural, then = 5.
ab − 1 ab − 1
2 2
Solution. Denote aab−1
+b
= p. Then a2 − pab − b2 = −p. As a 6= b and b > 0, we
reach a version of (1) with q = −p < 0. In this case, the hyperbola does not cross
the coordinate axis, and the respective sequences of solutions, if such, include only
positive or negative terms. Let’s look at the sequence with positive numbers; let
y0 be its least element. The equation x2 − pxy0 + y 2 = −p has integer solutions
x1 and x2 , according to the assumption. Let y0 < x1 ≤ x2 , which means
1 » 
py0 − (p2 − 4) y02 − 4p > y0 .
2
Thus p > 2, (p − 2)y02 < p. As p and y0 are natural numbers, y0 = 1. The
discriminant becomes p2 − 4 − 4p; as it is a perfect square, and differs by 8 from
the perfect square p2 + 4p − 4, we must have p = 5. This way we find x1 = 2,
x2 = 3 and deduce that the Diophantine equation x2 − 5xy + y 2 + 5 = 0 has
an infinite number of solutions in natural numbers. These are adjacent pairs of
terms from the sequence {. . . , 14, 3, 1, 2, 9, . . . } with linear recurrence equation
ai+2 = 5ai+1 − ai . They are also the pairs of coordinates of the integer points on
the hyperbola x2 − 5xy + y 2 + 5 = 0. 2
Problem 4 (2007 Spain Mathematical Olympiad, National Round) Find
m2 + mn + n2
all the possible positive integers which the expression can take where
mn − 1
m and n are natural numbers and mn 6= 1.
Problem 5 (2013 British Mathematical Olympiad) Find all pairs of natu-
ral numbers x and y for which x divides y 2 + 1 and y divides x2 + 1.
Problem 6 (2012 Vietnam Mathematical Olympiad) Let a and b be two
odd natural numbers, where a is a divisor of b2 + 2, and b is a divisor of a2 + 2.
Prove that a and b are terms of the sequence of natural numbers {vn } for which
v1 = v2 = 1, vn = 4vn−1 − vn−2 if n ≥ 3.
Problem 7 (1999 Bulgaria Team Selection Test) Prove that the number n4 +
1 has a divisor of the nm−1 type (m and n are natural numbers) if and only if n is a
term of the sequence {ai }ni=1 for which a1 = 1, a2 = 2 or a2 = 3, ai+2 = 5ai+1 −ai .
Comment. This problem is based on the fact that the defined sequence {ai }ni=1
has the property a4k + 1 = (ak−1 ak − 1) (ak ak+1 − 1). This property is peculiar to
the specific sequence of this problem; it is not shared by all sequences with the
linear recurrence equation ai+2 = 5ai+1 − ai .
We finish with a pair of harder problems (with solutions).

Copyright
c Canadian Mathematical Society, 2018
300/ LINEAR RECURENCES AND POLYNOMIAL DIVISION IN NUMBER THEORY

Problem 8 (2007 International Mathematical Olympiad) Let a and b be

2
natural numbers for which 4ab − 1 divides 4a2 − 1 . Prove that a = b.
Solution. In this problem the relation (a − b)2 = k(4ab − 1) is valid where k ≥ 0
is an integer. This follows from the equation

2
2
4a2 − 1 = 4a2 − 4ab + 4ab − 1 = 16a2 (a − b)2 + (4ab − 1)(8a2 − 4ab − 1),
when considered that 4ab − 1 and 16a2 are mutually prime. When k = 0 we have
a = b. Let’s assume the possibility that k ≥ 1, i.e. a 6= b. If integer solutions (a, b)
exist, then they are pairs of consecutive terms of the linear recurrence sequence
obtained through “Vieta Jumping”. Let b0 be the least positive integer in this
sequence. From Vieta’s formulas we can deduce that positive integers a1 and a2
exist so that the pairs (b0 , a1 ) and (b0 , a2 ) are solutions to the equation in this
case. Moreover, a1 > b0 and a2 > b0 , which means that (a1 − 1)(a2 − 1) ≥ b20 , and
so we obtain
b20 + k − 2(2k + 1)b0 = a1 a2 − (a1 + a2 ) = (a1 − 1)(a2 − 1) − 1 ≥ b20 − 1,
k+1
whence b0 ≤ ≤ 1, a contradiction. 2
4k + 2
Problem 9 (La Gaceta de la RSME, Vol. 17 (2014), Problema 241) Find
a4 − a2 + 1
all positive integers a and b for which the expression is a positive in-
ab − 1
teger.
Solution. If a = 1, then b = 2. Let a > 1, and let a and b fulfil the condition of the
problem. We can write a4 − a2 + 1 = (ab − 1) B; then a4 − a2 = abB − (B + 1).
Consequently, a divides B + 1, so a4 − a2 + 1 = (ab − 1) (ka − 1). We rewrite the
numerator:
a4 − a2 + 1 = b4 − b2 + 1 a4 − a2 b2 − 1 a2 b2 − a2 + 1

= b4 − b2 + 1 a4 − (ab − 1)A,

where A is a positive integer. As a2 and ab − 1 are mutually prime, b4 − b2 + 1 is

divisible by ab − 1. In the same way, we find that k 4 − k 2 + 1 is divisible by ka − 1.
Obviously a 6= b and a 6= k. If a < b, then k < a, because otherwise we would find

2
that a4 − a2 + 1 = (ab − 1)(ka − 1) ≥ a2 + a − 1 , which is impossible. Therefore
a2 + k 2 − 1
ab(ka − 1) = a4 − a2 + 1 + (ka − 1) or b = a − k = pa − k.
ka − 1
Then p is also a positive integer. We obtain the equation
a2 − pka + k 2 + p − 1 − 0,
which is symmetric in relation to a and k (and has no solutions for p ≤ 2). Let
(a0 , k0 ) be the integer solution minimizing the value of k0 (note that k0 ≤ a0 ), and
let a1 be the other root of
a2 − pk0 a + k02 + p − 1 − 0.

Crux Mathematicorum, Vol. 44(7), September 2018

 VALCHO MILCHEV /301

We can consider a1 ≥ a0 > k0 . The root a1 also is a positive integer due to the
relation a1 = pk0 − a0 . Then
p
pk0 − p2 k02 − 4k02 − 4p + 4
k0 < a0 = .
2
The only possibility is k0 = 1, from where it follows that p = 4. This way we find
a0 = 2 and a1 = 2. Because a, b, and k are symmetrical, we conclude that the
solutions a, b of the problem are consecutive terms of the sequence {1, 2, 7, 26, . . . }
with linear recurrence equation xn+2 = 4xn+1 − xn . 2

Quadrature of the figure

Take a grid paper and cut out the figure shown below on the left. Can you cut
it into 5 pieces and arrange them to form an 8 × 8 square?

Puzzle by Nikolai Avilov.

Copyright
c Canadian Mathematical Society, 2018
302/ 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 February 1, 2019.
The editor thanks Valérie Lapointe, Carignan, QC, for translations of the problems.
An asterisk (?) after a number indicates that a problem was proposed without a solution.

4361. Proposed by Andrew Wu.

Let ABC be a scalene triangle with circumcircle Γ, circumcenter O and incenter I.
Suppose that L is the midpoint of the arc BAC of Γ. The perpendicular bisector of
AI meets at X the arc AC that contains B, and at Y the arc AB that contains C.
Let XL and AC meet at P ; let Y L and AB meet at Q. Show that the orthocenter
of triangle OP Q lies on XY .

4362. Proposed by Oai Thanh Dao and Leonard Giugiuc.

Let ABCD be a convex quadrilateral and let F be the midpoint of CD. Consider
a point E inside ABCD such that AE · CE = BE · DE. The lines EF and AB
intersect at G. If ∠AED + ∠CEB = 180◦ , prove that ∠AED = ∠AGE.

4363. Proposed by Michel Bataille.

Let (an )n≥0 be the sequence defined by a0 > 0 and the recursion
an
an+1 = .
1 + (n + 1)a2n
∞ ∞
Å ã
a2n is convergent and find lim n· a2k .
P P
Prove that the series
n=0 n→∞ k=n

4364. Proposed by George Stoica.

Let f : N × N → N be a binary operation, and define g : N × N → N by
ß
a if b = 1
g(a, b) =
f (g(a, b − 1), a) if b ≥ 2.

If f is associative and g is commutative, prove that f (a, b) = a+b and g(a, b) = ab.

4365. Proposed by Marius Drăgan and Neculai Stanciu.

Let a and b be real numbers such that a + b, a4 and b4 are rational numbers and
a + b 6= 0. Prove that a ad b are rational numbers.

Crux Mathematicorum, Vol. 44(7), September 2018

 PROBLEMS /303

4366. Proposed by Daniel Sitaru.

Let xn be the base angle of a right triangle with base n and altitude 1. Find
∞
X
xk2 +k+1 .
k=1

4367. Proposed by Kadir Altintas and Leonard Giugiuc.

Let a, b and c be distinct complex numbers such that |a| = |b| = |c| = 1 and
|a + b + c| ≤ 1. Prove that
     
 (a + b)(b + c)   (b + c)(c + a)   (c + a)(a + b) 
 (a − b)(b − c)   (b − c)(c − a)   (c − a)(a − b)  = 1.
 + + 

4368. Proposed by Ovidiu Furdui and Alina Sı̂ntămărian.

Calculate
∞
X
[2n (ζ(n) − 1) − 1] ,
n=2
P∞ 1
where ζ denotes the Riemann zeta function defined as ζ(z) = z
.
n=1 n

4369. Proposed by Mihaela Berindeanu.

On the sides of M ABC take points A1 , A2 ∈ (BC) , B1 , B2 ∈ (AC) , C1 , C2 ∈
(AB), so that BA1 = A2 C, CB1 = B2 A, AC1 = C2 B. On B2 C1 , A1 C2 , A2 B1
C 1 A3 A1 B 3 B1 C 3
take A3 , B3 , C3 so that = = = k. Find all values of k for
A3 B 2 B3 C 2 C 3 A2
which AA3 , BB3 , CC3 are concurrent lines.

4370. Proposed by Leonard Giugiuc and Sladjan Stankovik.

Solve the following system of equations:

a + b + c + d = 4,

a2 + b2 + c2 + d2 = 7,
15

abc + abd + acd + bcd − abcd = 16 .


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

4361. Proposé par Andrew Wu.

Soit ABC un triangle scalène et le cercle circonscrit Γ de centre O et soit le point I,
le centre du cercle inscrit. Soit L le point milieu de l’arc BAC de Γ. La médiatrice

Copyright
c Canadian Mathematical Society, 2018
304/ PROBLEMS

du segment AI rencontre au point X l’arc AC qui contient B et rencontre au point

Y l’arc AB qui contient C. Soit P , le point d’intersection de XL et AC et soit Q,
le point d’intersection de Y L et AB. Montrez que l’orthocentre du triangle OP Q
est sur le segment XY .

4362. Proposé par Oai Thanh Dao et Leonard Giugiuc.

Soit ABCD un quadrilatère convexe et soit F le point milieu du segment CD.
Considérez un point E dans ABCD tel que AE · CE = BE · DE. Les segments
EF et AB s’interceptent au point G. Si ∠AED + ∠CEB = 180◦ , prouvez que
∠AED = ∠AGE.

4363. Proposé par Michel Bataille.

Soit (an )n≥0 une suite définie par a0 > 0 et la récurrence
an
an+1 = .
1 + (n + 1)a2n
∞ ∞
Å ã
a2n est convergente et trouvez lim n· a2k .
P P
Prouvez que la série
n=0 n→∞ k=n

4364. Proposé par George Stoica.

Soit f : N × N → N une opération binaire et soit g : N × N → N définie telle que
ß
a si b = 1
g(a, b) =
f (g(a, b − 1), a) si b ≥ 2.
Si f est associative et g est commutative, prouvez que f (a, b) = a+b et g(a, b) = ab.

4365. Proposé par Marius Drăgan et Neculai Stanciu.

Soit a et b des nombres réels tels que a + b, a4 et b4 sont des nombres rationnels
et a + b 6= 0. Prouvez que a et b sont des nombres rationnels.

4366. Proposé par Daniel Sitaru.

Soit xn l’angle à la base d’un triangle rectangle de base n et de hauteur 1. Trouvez
∞
X
xk2 +k+1 .
k=1

4367. Proposé par Kadir Altintas et Leonard Giugiuc.

Soit a, b et c des nombres complexes distincts tels que |a| = |b| = |c| = 1 et
|a + b + c| ≤ 1. Prouvez que
     
 (a + b)(b + c)   (b + c)(c + a)   (c + a)(a + b) 
 (a − b)(b − c)   (b − c)(c − a)   (c − a)(a − b)  = 1.
 + + 

Crux Mathematicorum, Vol. 44(7), September 2018

 PROBLEMS /305

4368. Proposé par Ovidiu Furdui et Alina Sı̂ntămărian.

Calculez
∞
X
[2n (ζ(n) − 1) − 1] ,
n=2

P∞ 1
où ζ indique la fonction zeta de Riemann définie par ζ(z) = z
.
n=1 n

4369. Proposé par Mihaela Berindeanu.

Sur le triangle ABC, on prend les points A1 , A2 ∈ (BC) , B1 , B2 ∈ (AC) , C1 , C2 ∈
(AB), tel que BA1 = A2 C, CB1 = B2 A, AC1 = C2 B. Sur les segments
B2 C1 , A1 C2 , A2 B1 on prend les points A3 , B3 , C3 tels que
C 1 A3 A1 B 3 B1 C3
= = = k.
A3 B 2 B3 C2 C 3 A2
Trouvez la valeur de k pour laquelle AA3 , BB3 , CC3 sont des droites concourantes.

4370. Proposé par Leonard Giugiuc et Sladjan Stankovik.

Résolvez le système d’équations suivant :

a + b + c + d = 4,

a2 + b2 + c2 + d2 = 7,
15

abc + abd + acd + bcd − abcd = 16 .


Copyright
c Canadian Mathematical Society, 2018
306/ 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 2017: 43(7), p. 302–306.

4261. Proposed by Margarita Maksakova.

Consider the chess board. A baron can move only on the black squares and in
one move he can go from one black square to any of the diagonally adjacent black
squares. What is the smallest number of moves he needs to go to every black
square?
We received no correct and complete solution, so the problem remains open. One
submission showed that it is possible to go through every square in 34 moves. Can
you show whether this is the smallest number of moves needed?

4262. Proposed by Prithwijit De.

n
P
Let a1 , a2 , . . . , an be positive integers and suppose ak = S. Find the smallest
k=1
positive value of c such that the equation
n
X ak xk
=c
1 + x2k
k=1

has a unique real solution.

We received 6 correct solutions and one incomplete submission. We present the
solution of Ivko Dimitrić, slightly modified by the editor.
Let L(x) denote the left hand side of the equation. Then L(0) = 0. Suppose x 6= 0.
Then
n n
X ak xk X ak
L(x) = = .
1+x
k=1
2k x + x1k
k
k=1

Note that L(x) = L(1/x). We have

n n n
X ak 1X 2 1X 2
L(x) = k + 1
≤ ak k 1 = ak k 1
k=1
x xk
2
k=1
|x + xk
| 2
k=1
|x| + |x|k
n n
 
1X 1 1X S
≤ ak |x|k · k = ak = ,
2 |x| 2 2
k=1 k=1

where the last inequality is the AM-GM inequality. Note that equality L(x) = S2
is satisfied if and only if x > 0 and xk = 1/xk for all k, implying x = 1 is a

Crux Mathematicorum, Vol. 44(7), September 2018

 SOLUTIONS /307

unique solution for L(x) = S/2. Now suppose 0 < c < S/2. By the intermediate
value theorem and L(0) = 0, there exists 0 < r < 1 with L(r) = c. But since
L(1/r) = L(r), the equation L(x) = c does not have a unique solution. Thus
c = S/2.

4263. Proposed by Michel Bataille.

Let ABC be a triangle. Let Γ, with centre O and radius R, be the circumcircle
of ABC and γ, with centre I 6= O and radius r, be the incircle of ABC. Let
D, E, F be the orthogonal projections of the inverse of I in Γ onto BC, CA, AB,
respectively. Express the circumradius of ∆DEF as a function of R and r.
We received 3 submissions, all of which were correct; we feature the solution by
Andrew David Ionascu, slightly modified by the editor.
We denote the inverse of I with respect to Γ by J (that is, OJ × OI = R2 ), and
denote by D0 , E 0 , and F 0 the points of tangency where the incircle touches the
OJ
respective sides BC, CA, and AB. Note that because OA = OAOI , the triangles
JA
JAO and AIO are similar by side-angle-side. Therefore, AI = AO R
IO = IO , which
2 2
by Euler’s formula (namely OI = R − 2Rr) becomes

JA R
=√ .
AI 2
R − 2Rr

Because of the right angles at E and F , the circle whose diameter is JA contains
E and F and is therefore the circumcircle of ∆EAF . The Law of Sines applied to
this triangle gives us
EF
= JA. (1)
sin ∠EAF
Similarly, because of right angles at E 0 and F 0 , the circle whose diameter is IA
contains E 0 and F 0 and is therefore the circumcircle of ∆E 0 AF 0 , whence

E0F 0
= IA. (2)
sin ∠E 0 AF 0
Because E 0 ∈ AE and F 0 ∈ AF , the angles ∠EAF and E 0 AF 0 are equal or
supplementary, and division using equations (1) and (2) yields

EF JA R
0 0
= =√ .
EF AI 2
R − 2Rr

In the same way, we can show that

FD DE R
= 0 0 =√ .
F 0 D0 DE 2
R − 2Rr

Thus triangles DEF and D0 E 0 F 0 are similar (by side-side-side), so that if x is the
circumradius of ∆DEF (which we seek) while the inradius r is the circumradius

Copyright
c Canadian Mathematical Society, 2018
308/ SOLUTIONS

of ∆D0 E 0 F 0 , we must also have x

r = √ R
R2 −2Rr
. We conclude that the circumradius
of ∆DEF satisfies
Rr
x= √ .
R2 − 2Rr

Editor’s comments. The proposer, whose key step was the same as that of our
featured solution, observed that the result can be found as a theorem in Nathan
Altshiller-Court, College Geometry, Dover, 1980, paragraph 362, p. 173: The pedal
triangles of two points for a given triangle are similar if and only if the two points
are inverse with respect to the circumcircle of the given triangle.

4264. Proposed by Dorin Marghidanu and Leonard Giugiuc.

Let (an ) and (bn ) be two sequences such that a0 , b0 > 0 and
1 1
an+1 = an + and bn+1 = bn +
2bn 2an
for all n ≥ 0. Prove that
max(a2017 , b2017 ) > 44.

We received 11 correct solutions. We present here the solution by Paolo Perfetti.

The quantity
an
f (an , bn ) =
bn
is invariant under the recurrence. Indeed,
1
an + 2bn an
f (an+1 , bn+1 ) = 1 = = f (an , bn )
bn + 2an
bn

This means that

an a0
= f (an , bn ) = f (a0 , b0 ) =
bn b0
and then
1 a0 1 b0
an+1 = an + and bn+1 = bn + .
2an b0 2bn a0
Thus,
1 a20 1 b20
Å ã Å ã
a0 b0
a2n+1 + b2n+1 = a2n + b2n + + + + 2 2 ≥ a2n + b2n + 2,
b0 a0 4a2n b20 4b a
| {z } | {z n 0 }
≥2 (AGM ) ≥0

and
a2n+1 + b2n+1 > 2(n + 1) + a20 + b20 > 2(n + 1).
Moreover,
√
p
a22017 + b22017
max(a2017 , b2017 ) ≥ √ > 2017 > 44.9.
2

Crux Mathematicorum, Vol. 44(7), September 2018

 SOLUTIONS /309

4265. Proposed by Daniel Sitaru.

Consider real numbers a, b, c ∈ (0, 1) such that a + b + c = 1. Show that
4 1
(arctan a + arctan b + arctan c) > .
π 2 − (ab + bc + ca)

There were 9 correct solutions and 2 incorrect submissions submitted. We follow

the independent solutions of Daniel Dan; and the team D. Bailey, E. Campbell,
and C. Diminnie.
Since (4/π) arctan x is concave for x ≥ 0 and is equal to x for x = 0 and x = 1,
4
arctan x ≥ x
π
for 0 ≤ x ≤ 1. Therefore the left side of the inequality is not less than a+b+c = 1.
Since

2(ab + bc + ca) = (a + b + c)2 − (a2 + b2 + c2 )

= 1 − (a2 + b2 + c2 )
≤ 1 − (ab + bc + ca),

then ab + bc + ca ≤ 1/3 and

1 3
≤ < 1.
2 − (ab + bc + ca) 5

The result follows.

Editor’s comments. Some solvers used arctan x ≤ x−(x3 /3) and standard inequal-
ities to get the lower bound 8/3π for the left side. One solver used Karamata’s
inequality for the concave function arctan x and the triples (a, b, c), (1, 0, 0) to
show that the left side was not less than 4/π.

4266. Proposed by Marius Stănean.

Let ABC be a triangle with orthocenter H. Let HM be the median and HS be
the symmedian in triangle BHC. Denote by P the orthogonal projection of A onto
HS. Prove that the circumcircle of triangle M P S is tangent to the circumcircle
of triangle ABC.
We received 3 solutions. We present the solution by Michel Bataille modified by
the editor.
To ensure the existence of 4BHC, we assume that ∆ABC is not right-angled
at B or C. Consider the case when 4ABC is right-angled at A; then A, H
and P coincide. The circumcentre of 4ABC is M . Moreover, one can show
that S coincides with the foot of the altitude from A, hence the circumcentre of

Copyright
c Canadian Mathematical Society, 2018
310/ SOLUTIONS

4P SM = 4ASM is the midpoint of AM . The required result follows in this

case.
From now on, we suppose that ∆ABC is not right-angled. Below is included a
diagram for the case when 4ABC is acute (the argument will also work when
4ABC is obtuse, but points might be in a different order on lines and circles).

Let A0 , B 0 , C 0 be the feet of the altitudes of 4ABC (with A0 on BC, and so on),
and J be the midpoint of the line segment AH. Denote by I the inversion with
centre H such that I(B) = B 0 . Consider the effect of this inversion on the two
circles in which we are interested.
Look at ABC, the circumcircle of 4ABC. Since ∠BB 0 C = ∠CC 0 B = 90◦ ,
the points B, B 0 , C and C 0 are concyclic. The power of the point H gives us
HB 0 · HB = HC 0 · HC, whence I(C) = C 0 . Similarly, I(A) = A0 . Hence the
image of the circumcircle of 4ABC under I is the Euler circle A0 B 0 C 0 (a.k.a.
the 9-point circle, which is known to also go through M and J, and which is shown
dotted in the diagram).
Now consider the circumcircle of 4P SM . Similar to the above, we have ∠AP S =
∠AA0 S = 90◦ , so the points A, A0 , P and S are concyclic and the power of the
point H gives us HA0 · HA = HP · HS. Recall that I(A) = A0 , so it follows that
I(P ) = S. Since the two points P and S on P SM get mapped to each other,

Crux Mathematicorum, Vol. 44(7), September 2018

 SOLUTIONS /311

we conclude that the image of P SM is itself.

It follows that, in order to prove that ABC and P SM are tangent, it suffices
to show that A0 B 0 C 0 and P SM are tangent. To this end, we will show that
the centre of the Euler circle, the centre of P SM , and the point M are collinear.
Denote I(M ) by M 0 ; note that M 0 is on both ABC and P SM .

From earlier, I(S) = P and I(A0 ) = A; it follows that the line through SA0 gets
mapped to the circumcircle of 4P AH (recall that an inversion with center H will
map lines to circles that go through H). Thus M 0 , B 0 and C 0 , which are images
of points on the line segment BC, are all on P AH. From the given setup of
the problem, ∠BHS = ∠CHM , so ∠M 0 HC 0 = ∠P HB 0 ; thus the arcs M 0 C 0
and P B 0 on P AH are congruent, which implies M 0 P k C 0 B 0 . Denote by l the
perpendicular bisector of M 0 P , which is necessarily also the perpendicular bisector
of C 0 B 0 .

M 0 P is a chord in P SM , so the centre of P SM is on l. C 0 B 0 is a chord in

A0 B 0 C 0 , so the centre of the Euler circle is on l. Finally, note that 4M B 0 C 0 is
isosceles, since M is the midpoint of the hypotenuse of 4BB 0 C, implying M B 0 =
M C, and also the midpoint of the hypotenuse of 4CC 0 B, implying M C 0 = M C.
Hence M is also on the perpendicular bisector l of B 0 C 0 . This concludes the
argument that P SM and the Euler circle are tangent at M , and applying the
inversion I gives us that P SM and ABC are tangent at M 0 .

4267. Proposed by Leonard Giugiuc.

Let a, b, c and d be real numbers such that 0 < a, b, c ≤ 1 and abcd = 1. Prove
that
4
5(a + b + c + d) + ≥ 21.
abc + abd + acd + bcd

There were 7 correct solutions and two incorrect submissions, as well as one that
made use of Maple. Some of the solutions were quite complicated. We present the
solution by Kee-Wai Lau and Angel Plaza, done independently.

We first note that d = 1/abc and

(a + b + c + d) − (abc + abd + acd + bcd)

= d(a2 bc + ab2 c + abc2 + 1 − a2 b2 c2 − ab − ac − bc)
= d(1 − ab)(1 − ac)(1 − bc) ≥ 0,

so that
a+b+c+d
≥ 1.
abc + abd + acd + bcd

Copyright
c Canadian Mathematical Society, 2018
312/ SOLUTIONS

Therefore, applying the arithmetic-geometric means inequality twice, we find that

4
5(a + b + c + d) +
abc + abd + acd + bcd
Å ã ï ò
a+b+c+d a+b+c+d 4
= 19 + +
4 4 abc + abd + acd + bcd
ï ò1/2
a+b+c+d
≥ 19(abcd)1/4 + 2 ≥ 19 + 2 = 21.
abc + abd + acd + bcd

4268. Proposed by Mihaela Berindeanu.

Let I be the incenter of the acute triangle ABC, and let the triangle’s internal
angle bisectors intersect the circles IBC, ICA, and IAB again at A1 , B1 , and C1 ,
−−→ −−→ −−→ → −
respectively. Show that IA1 + IB1 + IC1 = 0 if and only if M ABC is equilateral.
We received 14 submissions, all of which were correct; we feature a composite of
similar solutions by Ivko Dimitrić and Titu Zvonaru.

The exterior angle at I of 4AIC satisfies ∠A1 IC = A+C

2 . Since I, B, A1 , C are
concyclic (in that order), we have ∠CA1 I = ∠CBI = B/2. Hence,

A+C B
∠ICA1 = 180◦ − − = 90◦
2 2
and in the same manner, ∠B1 CI = 90◦ , so that ∠B1 CA1 = 180◦ , which means
that the points A1 , C, and B1 are collinear with C1 C ⊥ A1 B1 . Analogously,
B1 , A, C1 are collinear with A1 A ⊥ B1 C1 , and C1 , B, A1 are collinear with B1 B ⊥
C1 A1 . Hence, A1 A, B1 B and C1 C are the altitudes of the triangle A1 B1 C1 so that
the incenter I of the given triangle ABC is the orthocenter of 4A1 B1 C1 .

Crux Mathematicorum, Vol. 44(7), September 2018

 SOLUTIONS /313

Since for any point I in the plane of triangle A1 B1 C1 one has

−−→ −−→ −−→ −−→
IA1 + IB1 + IC1 = 3IG1 ,
where G1 is the centroid of 4A1 B1 C1 , this sum will be zero if and only if I is the
centroid of the said triangle. But, the centroid and incenter of a triangle coincide
if and only if the triangle is equilateral. The problem has therefore been reduced
to proving that ∆ABC is equilateral if and only if ∆A1 B1 C1 is equilateral. We
have ∆ABC is equilateral if and only if ∠AIB = ∠BIC = ∠CIA = 120◦ , if and
only if ∠BC1 A = ∠CA1 B = ∠AB1 C = 60◦ , if and only if ∆A1 B1 C1 is equilateral.

Editor’s comments.
(1) Only Leonard Giugiuc observed explicitly that there is no need to require that
∆ABC be acute (as our featured solution shows).
(2) It is a standard result that the vertices of ∆A1 B1 C1 are the excenters of
∆ABC (see, for example, Chapter X of Roger A. Johnson, Advanced Euclidean
Geometry), and many of the submissions made use of well-known properties of
this pair of triangles to shorten their arguments.
(3) Anna Valkova Tomova used an argument much like our featured solution to
extend the result to
∆ABC is isosceles with apex at A if and only if there exists a nonzero
−−→ −−→ −−→
real number λ for which λIA1 + IB1 + IC1 = 0.

4269. Proposed by Hung Nguyen Viet.

Let x1 , x2 , . . . , xn be real numbers such that
n
sin x1 cos x2 + sin x2 cos x3 + · · · + sin xn cos x1 = .
2
Prove that cos 2x1 + cos 2x2 + · · · + cos 2xn = 0.

There were 15 correct solutions submitted, 9 of which had essentially the argument
given below. The remainder relied on an inequality forced to equality by the same
upper and lower bounds.
With xn+1 = x1 , we have that
n
X n
X n
X
(sin xk − cos xk+1 )2 = (sin2 xk + cos2 xk+1 ) − 2 sin xk cos xk+1
k=1 k=1 k=1
Xn
= (sin2 xk + cos2 xk ) − 2(n/2) = n − n = 0,
k=1

so that sin xk = cos xk+1 for each k. Therefore,

n
X n
X n
X
cos 2xk = (cos2 xk − sin2 xk ) = (cos2 xk+1 − sin2 xk ) = 0.
k=1 k=1 k=1

Copyright
c Canadian Mathematical Society, 2018
314/ SOLUTIONS

4270. Proposed by Leonard Giugiuc.

Let k and t be real numbers with k ∈ (0, 1) and t ∈ [ π4 , π2 ]. Prove that
Z t Z t
cos x sin x
k
dx ≥ dx.
0 x 0 xk

We received 10 solutions and will feature just one of them here, by Michel Bataille.
Let t ∈ [ π4 , π2 ]. For x ∈ (0, t], we have
cos x 1
0≤
k
≤ k
x x
Rt 1 R t cos x Rt x
and 0 xk dx exists, hence the integral 0 xk dx exists. The integral 0 sin xk
dx
also exists since
sin x sin x
lim = lim+ x1−k · = 0 · 1 = 0.
x→0+ xk x→0 x
Now, let
Z π/4 Z t
cos x − sin x cos x − sin x
I= k
dx, F (t) = dx, G(t) = I + F (t).
0 x π/4 xk
We are required to prove that G(t) ≥ 0.
Since
cos x − sin x
x 7→
xk
is continuous on [ π4 , π2 ], the function F is differentiable on this interval and so is
the function G with G0 (t) = F 0 (t) = cos t−sin
tk
t
. For t ∈ ( π4 , π2 ], we have cos t < sin t,
0
hence G (t) < 0 and therefore G is decreasing on [ π4 , π2 ]. As a result, it is sufficient
to show that G(π/2) ≥ 0. To this aim, we consider
√ Z π/4
sin( π4 − x)
Z π/2
2 sin( π4 − x)
· G(π/2) = dx + dx.
2 0 xk π/4 xk
The substitutions x = π4 − u in the first integral and x = π4 + u in the second one
lead to √ Z π/4 Ç å
2 1 1
· G(π/2) = (sin u) − π du.
2 0 ( π4 − u)k ( 4 + u)k
But for u ∈ (0, π4 ), we have
π π
( + u)k ≥ ( − u)k > 0
4 4
and sin u > 0, hence
Ç å
1 1
(sin u) π k
− π ≥0
( 4 − u) ( 4 + u)k
√
2
and so 2 · G(π/2) ≥ 0 and we are done.

Crux Mathematicorum, Vol. 44(7), September 2018