120/ 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(3), p. 101–104.

4221. Proposed by Nguyen Viet Hung.

Let a, b, c, p, q be distinct positive real numbers satisfying

a2 b2 c2
+ + = p,
(b − c)2 (c − a)2 (a − b)2
1 1 1
+ + = q.
(b − c)2 (c − a)2 (a − b)2
Evaluate
a b c
+ +
(b − c)2 (c − a)2 (a − b)2
in terms of p and q.
We received 13 submissions, 12 of which were correct, and we present the same
solutions by Michel Bataille and Prithwijit De.
a b c p
Let r = 2
+ 2
+ 2
. We show that r = q(p − 2).
(b − c) (c − a) (a − b)
Let
1 1 1
s= + + ,
(c − a)(a − b) (a − b)(b − c) (b − c)(c − a)
a b c
t= + + ,
(c − a)(a − b) (a − b)(b − c) (b − c)(c − a)
bc ca ab
u= + + .
(c − a)(a − b) (a − b)(b − c) (b − c)(c − a)
Then
(b − c) + (c − a) + (a − b)
s= = 0,
(a − b)(b − c)(c − a)
a(b − c) + b(c − a) + c(a − b)
t= = 0,
(a − b)(b − c)(c − a)
bc(b − c) + ca(c − a) + ab(a − b)
u= = −1.
(a − b)(b − c)(c − a)
Hence,
Å ã2
1 1 1
+ + = q + 2s = q
b−c c−a a−b

Crux Mathematicorum, Vol. 44(3), March 2018

 SOLUTIONS /121

Å ã2
a b c
and + + = p + 2u = p − 2.
b−c c−a a−b
Thus,
ÅÅ ãÅ ãã2
a b c 1 1 1
q(p − 2) = + + + +
b−c c−a a−b b−c c−a a−b
Å ã2
b+c c+a a+b
= r+ + +
(c − a)(a − b) (a − b)(b − c) (b − c)(c − a)
= (r + (a + b + c)s − t)2
= r2 .
p
Since p > 2 and r > 0, r = q(p − 2) follows.

Editor’s comment. Steven Chow pointed out that the solution to this problem
includes the proof of problem 1 of the 2017 Canadian Mathematical Olympiad
which asks to show that p > 2 and is, in essence, a sub-problem of the current
problem.

4222. Proposed by Mihaela Berindeanu.

Let ABCD be a quadrilateral inscribed in a circle and X a mobile point on the
small arc CD. If E, F, G, H are the orthogonal projections of X on the lines
AD, BC, AC, BD show that the angle between EH and GF is always constant,
regardless of the position of X on the arc.
Soit ABCD un quadrilatère inscrit dans un cercle et soit X un point situé sur le
petit arc CD. Si E, F , G et H sont les projections orthogonales de X vers AD,
BC, AC et BD, démontrer que l’angle entre EH et GF est constant, quel que
soit le point X sur l’arc.

We received 12 submissions, all correct; most of the solutions were quite similar to
our featured solution by Jean-Claude Andrieux, which was singled out to remind
our readers that solutions can be submitted in either of our two official languages.

Rappelons le théorème de la droite de Simson: Soit ABC un triangle quelconque

et M un point du plan. On note P , Q et R les projetés orthogonaux de M respec-

Copyright
c Canadian Mathematical Society, 2018
122/ SOLUTIONS

tivement sur (AB), (BC) et (CA); alors, P , Q et R sont alignés si et seulement si

M appartient au cercle circonscrit à ABC.
Dans le problème posé, notons Y le projeté orthogonal de X sur (AB).
Considérons le triangle ABC: X appartient au cercle circonscrit au triangle ABC
donc les points F , G et Y projetés orthogonaux de X respectivement sur (BC),
(CA) et (AB) sont alignés.
Considérons le triangle ABD: X appartient au cercle circonscrit au triangle ABD
donc les points E, H et Y projetés orthogonaux de X respectivement sur (AD),
(BD) et (AB) sont alignés.

Les droites (F G) et (EH) se coupent donc en Y . Il faut montrer que l’angle EY

’ F
est indépendant de la position de X sur l’arc CD.
Les triangles XEA, XGA et XY A sont des triangles rectangles de même hy-
poténuse [XA]. Les points X, E, A, G et Y sont donc cocycliques.
On a alors:
EY
’ F = EY
’ G = EAG
’ = DAC.
’
L’angle EY
’ F est donc constant et on a

EY
’ F = DAC
’ = DBC.
’

Editor’s comments. Steven Chow observed that if directed angles are used (modulo
π), the point X need not be restricted to the small arc CD: the angle between EH
and GF remains constant (and the featured proof remains valid) for all positions
of X on the circle. Somasundaram Muralidharan observed, similarly, that the final
line of our argument shows that while C and D must be fixed points, A and B
are free to move about the circle without changing the angle between EH and
GF . Chow also suggested that the editors perhaps should not have included the
proposer’s diagram with the statement of his problem since it essentially provides
the solution. Maybe that explains the similarity of so many of the submissions.
Bataille’s solution, however, was based on the spiral similarity with fixed point X
that takes C to D (and therefore G to E and F to H). He added to his solution
the observation that the lines EG, F H, and CD are concurrent in a point common
to the circles on diameter XC and XD. For more information about arguments
that exploit intersecting circles and spiral similarities, see his article “Focus On...
No. 12” [40:5 (May 2014) 203-206].

4223. Proposed by Leonard Giugiuc and Dorin Marghidanu.

Let a, b and c be positive real numbers such that a + b + c ≤ 1. Prove that
»
3
(1 − a3 )(1 − b3 )(1 − c3 ) ≥ 26abc.

We received 15 correct solutions and we present a very succinct proof by Titu

Zvonaru.

Crux Mathematicorum, Vol. 44(3), March 2018

 SOLUTIONS /123

Since 1 ≥ (a + b + c)3 , we have by the AM-GM inequality that

1 − a3 ≥ (a + b + c)3 − a3
= b3 + c3 + 3(a2 b + ab2 + b2 c + bc2 + c2 a + ca2 ) + 6abc
√
26
≥ 26 a24 b27 c27 ,

with equality if and only if a + b + c = 1 and a = b = c or two of a, b, and c are 0.

Multiplying by the other two inequalities obtained by considering 1 − b3 and 1 − c3
we then obtain
√
26
(1 − a3 )(1 − b3 )(1 − c3 ) ≥ (26)3 a78 b78 c78 ,
p
from which 3 (1 − a3 )(1 − b3 )(1 − c3 ) ≥ 26abc follows.
The equality holds if and only if (a, b, c) = 13 , 31 , 13 , (1, 0, 0), (0, 1, 0), or (0, 0, 1).

Editor’s comments. Geupel remarked that the given inequality can be generalized
to the following result:
if n is a natural number and a1 , a2 , . . . , an are positive real numbers
such that a1 + a2 + · · · + an ≤ 1, then
»
n
(1 − an1 )(1 − an2 ) · · · (1 − ann ) ≥ (nn − 1)a1 a2 · · · an .

4224. Proposed by Michel Bataille.

Find the complex roots of the polynomial

16x6 − 24x5 + 12x4 + 8x3 − 12x2 + 6x − 1.

We received 16 solutions. We present 2 solutions.

Solution 1, by Prithwijit De.
Observe that

16x6 −24x5 +12x4 +8x3 −12x2 +6x−1 = (2x2 )3 +(2x2 )3 +(2x−1)3 −3(2x2 )(2x2 )(2x−1).

Using the identity

a3 + b3 + c3 − 3abc = (a + b + c)(a2 + b2 + c2 − ab − bc − ca),

we get

(2x2 )3 + (2x2 )3 + (2x − 1)3 − 3(2x2 )(2x2 )(2x − 1) = (4x2 + 2x − 1)(2x2 − 2x + 1)2 .
√
−1 ± 5 1 ± i
Thus the roots (ignoring multiplicity) are x = , .
4 2

Copyright
c Canadian Mathematical Society, 2018
124/ SOLUTIONS

Solution 2, by Somasundaram Muralidharan.

Let f (x) = 16x6 −24x5 +12x4 +8x3 −12x2 +6x−1. Let us first check whether f has
repeated roots. Such repeated roots, if any, will be roots of gcd(f (x), f 0 (x)), where
f 0 (x) is the derivative of f (x). In this case, it is easy to see that gcd(f (x), f 0 (x)) =
1±i
2x2 − 2x + 1 and hence the roots of this gcd, namely , are double roots of
2
f (x) = 0. Thus we have found four of the roots of f . We now find the remaining
two roots of f . We have

f (x) = (2x2 − 2x + 1)2 (4x2 + 2x − 1)

√
2 −1 ± 5
and hence the remaining roots are roots of 4x + 2x − 1 = 0. These are .
4
So, the complex roots of f are
√ √
−1 + 5 −1 − 5 1 + i 1 + i 1 − i 1 − i
, , , , , .
4 4 2 2 2 2

4225. Proposed by Leonard Giugiuc, Daniel Dan and Daniel Sitaru.

Prove that in any triangle ABC we have:

3(cos2 A + cos2 B + cos2 C) + cos A cos B + cos A cos C + cos B cos C ≥ 3.

We received 13 correct solutions. We present the solution by Arkady Alt.

Since cos2 A + cos2 B + cos2 C + 2 cos A cos B cos C = 1 in any triangle ABC, the
original inequality is successively equivalent to

3(1 − 2 cos A cos B cos C) + cos A cos B + cos A cos C + cos B cos C ≥ 3
⇐⇒ cos A cos B + cos A cos C + cos B cos C ≥ 6 cos A cos B cos C.
√
Setting t = 3 cos A cos B cos C and using the AM-GM inequality, we obtain

1 − 2t3 = 1 − 2 cos A cos B cos C

= cos2 A + cos2 B + cos2 C
√
3
≥ 3 · cos2 A · cos2 B · cos2 C
= 3t2 .

Therefore 2t3 +3t√ 2

−1 ≤ 0, implying successively that (2t−1)(t+1)2 ≤ 0 and then
1 3
t ≤ 2 . Hence, 2 · cos A cos B cos C ≤ 1, and again by the AM-GM inequality we
have
√
3
cos A cos B + cos A cos C + cos B cos C ≥ 3 · cos2 A · cos2 B · cos2 C
√
3
√3
≥ 3 · cos2 A · cos2 B · cos2 C · cos A cos B cos C
= 6 cos A cos B cos C.

Crux Mathematicorum, Vol. 44(3), March 2018

 SOLUTIONS /125

4226. Proposed by Daniel Sitaru.

Prove that if 0 < a < b then:
 Z b √1 + x 2  2 b
dx > (b − a)2 + ln2 .
a x a

We received nine submissions, eight of which are correct and the other is incorrect.
We present a composite of virtually the same solutions by Arkady Alt; Michel
Bataille; M. Bello, M. Benito, Ó. Ciaurri, E. Fernández, and L. Roncal (jointly);
and Digby Smith.
Note first that
ÇZ b
√ å2
1 + x2 b
dx > (b − a)2 + ln2
a x a
ÇZ b √ å2 ÇZ b å2
1 + x2 1
⇐⇒ dx − dx > (b − a)2
a x a x
Z b√ Z b√
1 + x2 + 1 1 + x2 − 1
⇐⇒ dx · dx > (b − a)2 . (1)
a x a x
√ √
1 + x2 + 1 1 1 + x2 − 1
Let f (x) = , x ∈ [a, b]. Then f (x) > 0 and = . By
x f (x) x
the integral form of the Cauchy-Schwarz Inequality, we have
2 å Z b Ç  1 å2
ÇZ b å ÇZ b å ÇZ b » !
1
f (x) dx dx = f (x) dx dx
a a f (x) a a f (x)
ÇZ b å2
≥ 1dx
a

= (b − a)2 . (2)

But equality cannot hold in (2) as f is not a constant on [a, b]. Hence, from (1)
and (2) the result follows.

4227. Proposed by Dan Marinescu and Leonard Giugiuc.

Let P be a point in the interior of an equilateral triangle ABC whose sides have
length 1, and let R0 and r0 be the circumradius and inradius of the triangle whose
sides are congruent to P A, P B and P C (which exists by Pompeiu’s theorem).
Prove that
3R0 ≥ 1 ≥ 6r0 .

Among the four submissions, three were complete and correct; in the fourth, Michel
Bataille simply provided a reference where the proof can be found: Proposition 7
in Jószef Sándor’s “On the Geometry of Equilateral Triangles”, Forum Geometri-
corum, vol. 5 (2005) 107-117. Here we present the solution by Roy Barbara.

Copyright
c Canadian Mathematical Society, 2018
126/ SOLUTIONS

Let square brackets denote area and let T = ∆A0 B 0 C 0 denote the given Pompeiu
triangle with sides a = P A, b = P B, c = P C (with a opposite A0 , etc.).
Lemma. √ Å
3 2 − (a2 + b2 + c2 )
ã
r0 = .
6 a+b+c
Proof of Lemma. Let X, Y , and Z be the reflections of P through BC, CA, and
AB, respectively. Triangle AZY satisfies AY = AZ = a and ∠Y AZ = 120◦ ;
hence,
√ √ √
3 2 3 2 3 2
[AZY ] = a and, similarly, [BXZ] = b and [CY X] = c . (1)
4 4 4
√ √ √
Triangle XY Z, having sides a 3, b 3, c 3 is similar to T . Clearly, the area of
the hexagon AZBXCY is twice [ABC] so,
√
3
[AZBXCY ] = . (2)
2
But also, [AZBXCY ] = [XY Z] + [AZY ] + [BXZ] + [CY X]. From this, (1), and
(2) we get
√
3
2 − (a2 + b2 + c2 ) .


[XY Z] = (3)
4
√
The inradius r00 of ∆XY Z is [XY Z] 00
s00 , where s =
3
2 (a + b + c) is the semiperimeter
of ∆XY Z. From this and (3) we get

2 − (a2 + b2 + c2 )
Å ã
1
r00 = .
2 a+b+c

Finally, since T is similar to ∆XY Z with ratio √1 ,

3
we obtain r0 = √1 r 00 ,
3
and the
lemma follows.
Proof that R0 ≥ 13 . Since ∆ABC is equilateral, its Fermat point (that minimizes
a + b + c) is its centroid. Hence,
√
a + b + c ≥ 3. (4)

By the Law of Sines we have

a b c a+b+c
2R0 = 0
= 0
= 0
= . (5)
sin A sin B sin C sin A + sin B 0 + sin C 0
0

√
Further, any triangle A0 B 0 C 0 satisfies sin A0 + sin B 0 + sin C 0 ≤ 3 2 3 . This together
with (5) and (4) yields
√
0 a+b+c 3 2
2R = ≥Ä √ ä= ;
sin A0 + sin B 0 + sin C 0 3 3 3
2

Crux Mathematicorum, Vol. 44(3), March 2018

 SOLUTIONS /127

that is, R0 ≥ 13 , as desired. Equality holds if and only if P is the centroid of

∆ABC (in which case ∆A0 B 0 C 0 is also equilateral).
√  2

+b2 +c2 )
Proof that r0 ≤ 16 . Set S = a + b + c. By the lemma, r0 = 63 2−(aa+b+c ;
S2
moreover, a2 + b2 + c2 ≥ 3 is a basic inequality. Consequently,
√ Å
3 6 − S2
ã
0
r ≤ . (6)
18 S
√ 6−S 2
By (4), S ≥ 3. Furthermore, S is decreasing with respect to S, whence
√
6 − S2 6 − ( 3)2 √
≤ √ = 3.
S 3

The desired result follows immediately from this together with (6); also here,
equality holds if and only if ∆A0 B 0 C 0 is equilateral.

4228. Proposed by Mihály Bencze.

Let zk ∈ C, k = 1, 2, . . . , n such that nk=1 zk = nk=1 zk2 = 0. Prove that
P P

n n
!2
X X
n |zk |2 ≤ (n − 2) |zk | .
k=1 k=1

We received three correct solutions and one incorrect solution. We present two
solutions here.

Solution 1, by the proposer.

For each k with 1 ≤ k ≤ n, we have that

2zk2 = zk2 + (z1 + z2 + · · · zˆk + · · · + zn )2

Xn X
= zk2 + 2 {zi zj : 1 ≤ i < j ≤ n; i, j 6= k}
k=1
X
=2 {zi zj : 1 ≤ i < j ≤ n; i, j 6= k}

from which X
2|zk |2 ≤ 2 {|zi zj | : 1 ≤ i < j ≤ n; i, j 6= k}.

Adding all these n inequalities leads to

n
X X
2 |zk |2 ≤ 2(n − 2) |zi zj |.
k=1 1≤i<j≤n

Copyright
c Canadian Mathematical Society, 2018
128/ SOLUTIONS

Pn
Adding (n − 2) k=1 |zk |2 to each side yields

Ñ é
n
X n
X X
n |zk |2 ≤ (n − 2) |zk |2 + 2 |zi zj |
k=1 k=1 1≤i<j≤n

n
!2
X
= (n − 2) |k | .
k=1

Crux Mathematicorum, Vol. 44(3), March 2018

 SOLUTIONS /129

Solution 2, by Michel Bataille.

When n = 1, 2, 3, we have equality on both sides. For n = 1, z1 = 0. For n = 2,
z1 + z2 = z1 z2 = 0 so that z1 = z2 = 0. For n = 3 and i 6= j,

2(zi3 − zj3 ) = (zi − zj )[zi2 + zj2 + (zi + zj )2 )] = 0,

so that z13 = z23 = z33 , whence |z1 | = |z2 | = |z3 |.

Let n ≥ 4. Then
n
X X
2
0 = |z1 + · · · + zn | = (z1 + · · · + zn )(z1 + · · · + zn ) = |zk |2 + 2 Re (zi zj ).
k=1 1≤i<j≤n

Therefore
n
X X X
|zk |2 = −2 Re (zi zk ) ≤ 2 |Re (zi zk )|
k=1 1≤i<j≤n 1≤i<j≤n
X X
≤2 |zi zj | = 2 |zi ||zj |
1≤i<j≤n 1≤i<j≤n
n
!2 n
X X
= |zk | − |zk |2 .
k=1 k=1

Rearranging like terms and multiplying by n/2 yields the inequality

n n
!2
X
2 n X
n |zk | ≤ |zk | ,
2
k=1 k=1

which is the required inequality when n = 4 and is stronger for larger n.

4229. Proposed by Leonard Giugiuc.

Let n be an integer with n ≥ 2 and let p be a prime number with p > n. Consider
an n × n matrix X over Zp with X p = In . Prove that (X − In )n = On .
There were 5 correct solutions. We present the solution obtained independently by
Roy Barbara and Trey Smith.

Since (X − In )p = X p − Inp = On , X − In is nilpotent. But this implies that

(X − In )n = On .

Editor’s comment. For the linear algebra result invoked, let D be nilpotent and m
the minimum exponent for which Dm = On . If m ≤ n, then Dn = On . If m > n,
suppose, if possible, (by Cayley’s theorem) that On = Dn +cn−1 Dn−1 +· · ·+ck Dk
with ck 6= 0 for some 0 ≤ k ≤ n − 1. Multiply the equation by Dm−k−1 to get a
contradiction.

Copyright
c Canadian Mathematical Society, 2018
130/ SOLUTIONS

4230. Proposed by Miguel Ochoa Sanchez and Leonard Giugiuc.

Let ABC be a triangle in which ∠B = 2∠C and let M be the midpoint of BC.
The internal bisector of ∠ACB intersects AM in D. Prove that ∠CDM ≤ 45◦
and find ∠C for which the equality holds.
Miguel Amengual Covas observed that this problem appeared as problem 1562
of Crux [1990:204], posed by Toshio Seimiya and that three solutions were given
in [1991 : 252-254]. He observes further that it appears in The Olympiad Corner
No. 161 [1995 : 9-10], with solution by Covas, distinct from those published in
1991 given in [1996 : 265-267].
References to further properties of triangles whose angles satisfy ∠B = 2∠C can
be found in J. Chris Fisher’s “Recurring Crux Configurations 7”, [2012 : 238-240].

Bouncing Balls

A pair of balls A and B of negligible radius (so you can treat them as points)
lie on a perfectly flat surface with ball B lying between ball A and a wall. Ball
A has mass 100n and ball B has mass 1. Ball A is pushed towards ball B and,
as the balls interact, we count the number of collisions.

When n = 0, ball A strikes ball B and stops. Ball B bounces off the wall and
returns to strike ball A. Ball B then stops and ball A rolls away into the distance.
A total of 3 collisions occurred.

When n = 1, there are a total of 31 collisions.

When n = 5, there are a total of 314159 collisions.

See a pattern? Can you prove it?

Crux Mathematicorum, Vol. 44(3), March 2018