Mathproblems

ISSN: 2217-446X, url: http://www.mathproblems-ks.com

Volume 1, Issue 3 (2011), Pages 17–31

Editors: Valmir Krasniqi, José Luis Dı́az-Barrero, Mihály Bencze, Ovidiu

Furdui, Enkel Hysnelaj, Paolo Perfetti, József Sándor, Armend Sh. Shabani, David
R. Stone, Roberto Tauraso.
————————————————————————————————
PROBLEMS AND SOLUTIONS
————————————————————————————————
Proposals and solutions must be legible and should appear on separate sheets, each
indicating the name of the sender. Drawings must be suitable for reproduction.
Proposals should be accompanied by solutions. An asterisk (*) indicates that nei-
ther the proposer nor the editors have supplied a solution. The editors encourage
undergraduate and pre-college students to submit solutions. Teachers can help by
assisting their students in submitting solutions. Student solutions should include
the class and school name. Solutions will be evaluated for publication by a com-
mittee of professors according to a combination of criteria. Questions concerning
proposals and/or solutions can be sent by e-mail to: [email protected]

Solutions to the problems stated in this issue should arrive before

2 August 2011

Problems
15. Proposed by Valmir Bucaj, Texas Lutheran University, Seguin, TX.
Consider the set given by
1
La,r = {arn |n ∈ Z+ }, where r = , q ∈ Z+ , and gcd(a, q) = 1.
q
Show that
1
Bq = {La,r |r = , q ∈ Z+ , q ≥ 2 and gcd(a, q) = 1},
q
forms a basis for a topology on Q+ .
16. Proposed by José Luis Dı́az-Barrero, Technical University of Catalonia (BAR-
CELONA TECH), Barcelona, Spain. Find all triplets (x, y, z) of real numbers
which are solutions of the following system of equations

x2 (y + z)2 = (3x2 − x + 1)y 2 z 2 
y 2 (z + x)2 = (4y 2 − y + 1)z 2 x2
z 2 (x + y)2 = (5z 2 − z + 1)x2 y 2


2010
c Mathproblems, Universiteti i Prishtinës, Prishtinë, Kosovë.
17
18

17. Proposed by José Luis Dı́az-Barrero, Technical P

University of Catalonia (BAR-
n
CELONA TECH), Barcelona, Spain. Let A(z) = k=0 ak z k (ak 6= 0) be a non-
constant polynomial with complex coefficients. Prove that all its zeros lie in the
ring shaped region C = {z ∈ C : r1 ≤ |z| ≤ r2 }, where
(
 )1/k (  )1/k
4k nk  a0  5n − 1  an−k 
r1 = min and r2 = max
4k nk  an 


5n − 1  ak 
  
1≤k≤n 1≤k≤n

18. Proposed by Neculai√

Stanciu, George Emil Palade Secondary School, Buzău,
Romania. If, f (x) = 1− 21−2x , and fn−1 = f −1 ◦ f −1 ◦ ... ◦ f −1 , then evaluate:
| {z }
n
Z 1
2
lim fn−1 (x)dx.
n→∞ 0

19. Proposed by Paolo Perfetti, Department of Mathematics, Tor Vergata Univer-

sity, Rome,
PItaly. Let {ak }k≥1 be a monotonic sequence of real positive numbers
∞
such that n=1 an < ∞. Moreover, {ak }k≥1 fulfills the condition
ak − ak+1 ≥ 2−n a2n+1 , for all k with 2n ≤ k ≤ 2n+1 − 1.
Let α be a quadratic irrational. Prove that the following series converges for any
δ > 0.
∞ 2n
X an X 1
n(ln n)δ | sin(kπα)| kak ln k
n=1 k=n+1

20. Proposed by Ovidiu Furdui, Cluj, Romania. Let p > 1/2 be a real number.
Calculate !np
n
X 1
lim n p


n→∞
k=1 k

21. Proposed by Mihály Bencze, Braşov, Romania. Prove that

n
X k2 + k − 1  n2
tan arctan =
(k 2 + k + 1)(k 2 + k + 2) 2n2 + 5n + 5
k=1
 19

Solutions
No problem is ever permanently closed. We will be very pleased considering for
publication new solutions or comments on the past problems.

8. Proposed by Valmir Krasniqi and Armend Sh. Shabani, Department of Mathe-

matics, University of Prishtina, Republic of Kosova. If f is a nonnegative function
on [0, 1] and f 0 (x) ≥ 1, then
Z 1 Z 1 2
f 3 (x)dx ≥ f (x)dx
0 0

Solution by Ovidiu Furdui, Cluj, Romania. We prove that, under the hypo-
thesis of the problem, one has that for x ∈ [0, 1], the following stronger inequality
holds
Z x Z x 2 Z x
f 3 (t)dt − f (t)dt ≥ f 2 (0) f (t)dt
0 0 0
When x = 1 this implies that
Z 1 Z 1 2 Z 1
3
f (x)dx − f (x)dx ≥ f 2 (0) f (x)dx
0 0 0

Let F : [0, 1] → R be the function defined by

Z x Z x 2
F (x) = f 3 (t)dt − f (t)dt
0 0
0 2
Rx

Then F (x) = f (x) f (x) − 2 0 f (t)dt . Let g be the function defined on [0, 1] by
Rx
g(x) = f 2 (x) − 2 0 f (t)dt. A calculation shows that g 0 (x) = 2f (x)(f 0 (x) − 1) ≥ 0.
2 0
Hence g increases and it follows that g(x) ≥ g(0)R x = f (0).0 It follows that F (x) ≥
2 2
f (0)f (x), and this implies that (F (x) − f (0) 0 f (t)dt) ≥ 0. Thus, the function
Rx Rx
x → F (x) − f 2 (0) 0 f (t)dt increases. It follows that F (x) − f 2 (0) 0 f (t)dt ≥
F (0) = 0, and the problem is solved.
Comment by Henry Ricardo, USA: This problem is a special case of results
found in a paper by Mohamed Akkouchi ( Some integral inequalities, Divulgaciones
Matemticas 11 (2003), 121-125), which generalizes some results of Feng Qi (Several
integral inequalities, Journal of Inequalities in Pure and Applied Mathematics vol.
1, issue 2, Article 19, 2000).
Also solved by Henry Ricardo, USA; Arnau Massegué Buisan, Tech-
nical University of Barcelona, Spain, Paolo Perfetti, Department of
Mathematics, Tor Vergata University, Rome, Italy, Valmir Bucaj, Texas
Lutheran University, Seguin, TX and the proposers.

9. Proposed by Roberto Tauraso, Department of Mathematics, Tor Vergata Uni-

versity, Rome, Italy. Show that for any prime p and for any non-negative integer
20

n, p | Lpn − Ln , where Ln is the n-th Lucas number defined by L0 = 2, L1 = 1 and

for n ≥ 2, Ln = Ln−1 + Ln−2 .

Solution by the proposer. If p = 2 then L2
= 3 ≡ 1 = L1 (mod 2). Now we

assume that p is an odd prime. Since p divides kp for k = 1, . . . , p − 1, by Fermat’s
Little Theorem
√ !p √ !p
1+ 5 1− 5 1 X p
Lp = + = p−1 5k/2
2 2 2 k
0≤k≤p
k≡0 (mod 2)
 
1 1 X p k/2
= + 5 ≡ 1 = L1 (mod p)
2p−1 2p−1 k
1≤k≤p−1
k≡0 (mod 2)

Moreover, by the recurrence Ln+m = Lm Ln − (−1)m Ln−m , we have that for any
prime p
L(n+1)p ≡ Lp Lnp + L(n−1)p ≡ Lnp + L(n−1)p (mod p)
Hence by letting an = Lnp − Ln , it follows that
a0 ≡ 0, a1 ≡ 0, and an+1 = an + an−1 (mod p) for n ≥ 1,
which means that an ≡ 0 (mod p) for all n ≥ 0.

10. Proposed by Roberto Tauraso, Department of Mathematics, Tor Vergata Uni-

versity, Rome, Italy. Let n = 2010100 . Compute the cardinality of the set
Sn = d : d ∈ [1, n] ∩ N, d|n2 , d - n


Solution by Valmir Bucaj, Texas Lutheran University, Seguin, TX. We

will solve the more general problem instead: Let n be an integer given in the
standard prime factorization form: n = pe1 pe22 · · · pevv . Compute the cardinality of
Sm = {d : d ∈ [1, m] ∩ N, d|m2 , d - m},
where m = nk for some positive integer k. Let φ(m) be the number of divisors of
m excluding m. Then, we have φ(m) = (ke1 + 1)(ke2 + 1) · · · (kev + 1) − 1 and
φ(m2 ) = (2ke1 + 1)(2ke2 + 1) · · · (2kev + 1) − 1, √ as it is well-known. Also,√since
given an integer n, half of its divisors are less than n and half greater than n, in
our specific case follows that half of the divisors of m2 are less than m. Therefore,
φ(m2 )
|Sm | =
− φ(m)
2
For the original problem we have: n = 2010, k = 100, and m = 2010100 . So, since
2010 = 2 · 3 · 5 · 67, we have
φ(m) = 1014 − 1 and φ(m2 ) = 2014 − 1.
Therefore,
φ(m2 ) 2014 − 1
|Sm | = − φ(m) = − 1014 + 1 = 712060000
2 2
Also solved by the proposer.
 21

11. Proposed by Roberto Tauraso, Department of Mathematics, Tor Vergata Uni-

versity, Rome, Italy. Find a closed formula for
X X X
x
A⊂{1,...,n} B⊂{1,...,n} x∈A∪B
A6=∅ B6=∅

Solution 1 by the proposer. We first note that for any x ∈ {1, . . . , n} the
number of subsets of {1, . . . , n} which contain the element x is 2n−1 . Therefore
X X X X
g(n) = x= x [x ∈ A]
A⊂{1,...,n} x∈A x∈{1,...,n} A⊂{1,...,n}
A6=∅ A6=∅
 
X n+1
= 2n−1 x = 2n−1
2
x∈{1,...,n}

where [P ] is 1 if P is true and 0 otherwise. Similarly, for any x ∈ {1, . . . , n} the

number of couples of subsets of {1, . . . , n} whose union set contains the element x
is 3 · 2n−1 · 2n−1 ( 3 because x could belong to A \ B, A ∩ B or B \ A). Hence
X X X X X X
f (n) = x= x [x ∈ A ∪ B]
A⊂{1,...,n} B⊂{1,...,n} x∈A∪B x∈{1,...,n} A⊂{1,...,n} B⊂{1,...,n}
 
X n+1
= 3 · 4n−1 x = 3 · 4n−1
2
x∈{1,...,n}

Therefore
X X X X X X X X X
x = f (n) − x− x
A⊂{1,...,n} B⊂{1,...,n} x∈A∪B A⊂{1,...,n} B=∅ x∈A∪B A=∅ B⊂{1,...,n} x∈A∪B
A6=∅ B6=∅ A6=∅ B6=∅
 

n+1
= f (n) − 2g(n) = 3 · 4n−1 − 2n
2

Solution 2 by the Joaquı́n Rivero Rodrı́guez, I.E.S. Antonio de Nebrija,

Zalamea de la Serena, Spain. We have
!
X X X X X X X X
x = x+ x− x
A⊂{1,...,n} B⊂{1,...,n} x∈A∪B A⊂{1,...,n} B⊂{1,...,n} x∈A x∈B x∈A∩B
A6=Ø B6=Ø A6=Ø B6=Ø
X X X X X X
= 2· x− x
A⊂{1,...,n} B⊂{1,...,n} x∈A A⊂{1,...,n} B⊂{1,...,n} x∈A∩B
A6=Ø B6=Ø A6=Ø B6=Ø
n
X n
X
= 2 · 2n−1 · (2n − 1) · k − 2n−1 · 2n−1 · k
k=1 k=1
 n  n(n + 1)
= 2 · (2n − 1) − 22n−2

2
= 22n−1 − 22n−3 − 2n−1 n(n + 1)
 
n−1 n

n+1
= 3·4 −2
2
and we are done.
22

12. Proposed by Paolo Perfetti, Department of Mathematics, Tor Vergata Univer-

sity, Rome, Italy. Let a, b, c be positive numbers. Prove that
r s
X 5a2 + 5c2 + 8b2 8(a + b)2 (b + c)2 (c + a)2
≥3· 9
cyc
4ac (abc)2

Solution by the proposer. Since 5a2 + 5c2 + 8b2 = 4a2 + (a2 + 4b2 ) + (4b2 +
c2 ) + 5c2 ≥ 4(a2 + ab + bc + c2 ), as can be easily proven, then
r r s
X 5a2 + 5c2 + 8b2 X a+b b+c 8(a + b)2 (b + c)2 (c + a)2
≥ + ≥3· 9
cyc
4ac cyc
c a (abc)2
Setting
b+c a+c b+a 1 1 1
x= , y= , z= =⇒ a = , b= ,c=
a b c 1+x 1+y 1+z
from which follows x + y + z + 2 = xyz. The inequality claimed becomes
√ √ √ √
x + y + y + z + z + x ≥ 3 · 2(xyz)2/9
3
if x + y + z + 2 = xyz
p √
Squaring and using (a + b)(a + c) ≥ a + bc (Cauchy–Schwarz), we obtain
√ √ √
4(x + y + z) + 2 ( xy + yz + zx) ≥ 9 · 22/3 (xyz)4/9
By the AGM we show that
4(x + y + z) + 6(xyz)1/3 ≥ 9 · 22/3 (xyz)4/9
Now put r = (xyz)1/9 and use x + y + z + 2 = xyz to get
4r9 − 9 · 22/3 r4 + 6r3 − 8
1 1 2 1 2 1 2
= (r−2 )(4r8 +4·2 r7 +4·2 3 r6 +8r5 +8·2 3 r4 −2 3 r3 +4r2 +4·2 3 r+4·2 3 ) ≥ 0 (1)
3 3

The constraint x + y + z + 2 = xyz yields xyz ≥ 2. Indeed

xyz = 2 + x + y + z ≥ 2 + 3(xyz)1/3
1
thus by defining p = (xyz) 3 ≥ 2, we get p3 − 3p − 2 = (p − 2)(p + 1)2 ≥ 0 that is
1
p ≥ 2 implying r ≥ 2 3 . Since
1 2
8 · 2 3 r4 − 2 3 r3 ≥ 0
then (1) is proved and we are done.
Editorial comment: We claim that equality is never achieved.

13. Proposed by Mihály Bencze, Braşov, Romania. Let ak , 1 ≤ k ≤ n, be any

positive numbers. Prove that
 2
  X
n − 1 + 2 ai aj 
 n
X 
1   1≤i<j≤n
(n − 1) 
 ak + n ≥ X n
Y X
a2k
k=1
ak ai aj (ai + aj ) + (n − 1)

k=1 1≤i<j≤n k=1

Solution by the proposer. Using the Cauchy-Schwarz inequality we get:

 23

n n 2
X 1  X Y  X
ak + Q
n a1 · a2p + ak ≥ a1 ap + 1
k=1 ak k=1
k=1
where p ∈ {2, 3, . . . , n} or
P 2
n a1 ap + 1
X 1
ak + Q
n ≥P n
Q
k=1 ak a1 · a2p + ak
k=1 k=1
therefore
n n X
n
X 1  X 1 
(n − 1) ak + Q
n = ak + Q
n ≥
k=1 ak p=2 k=1 ak
k=1 k=1
P 2 n P
P 2
n a1 ap + 1 a1 ap + 1
X p=2
≥ P n
Q ≥ P
n P n
Q 
p=2 a1 · a2p + ak a1 · a2p + ak
k=1 p=2 k=1
 P 2
n−1+2 ai aj
1≤i<j≤n
= P n
Q
ai aj (ai + aj ) + (n − 1) ak
1≤i<j≤n k=1

14 (Correction). Proposed by Mihály Bencze, Braşov, Romania. Solve the equa-

tion
3
64x − 27 = 343x−1 +
· 28x
7
Solution by the proposer. Setting a = 3, b = −4x , c = 7x−1 the given equation
becomes a3 + b3 + c3 − 3abc = 0 or equivalently,
X X
(a + b + c)( a2 − ab) = 0
cyc cyc

Now we consider the following two possibilities:

X X
(1) a2 − ab = 0 ⇔ a = b = c which is impossible.
cyc cyc

(2) a + b + c = 0 ⇒ 3 − 4x + 7x−1 = 0 or 7x−1 − 4x−1 = 3 · (4x − 1). Applying

Lagrange Theorem to the function f (t) = tx−1 we have that there exist α ∈ (1, 4)
and β ∈ (4, 7) such that
f (7) − f (4) = 3(x − 1)β x−2 and f (4) − f (1) = 3(x − 1)αx−2
From the preceding immediately follows that 3(x − 1)β x−2 = 3(x − 1)αx−2 ⇒ x = 1
is the unique solution.
24

————————————————————————————————
MATHCONTEST SECTION
————————————————————————————————
This section of the Journal offers readers an opportunity to solve interesting and ele-
gant mathematical problems mainly appeared in Math Contest around the world
and most appropriate for training Math Olympiads. Proposals are always wel-
comed. The source of the proposals will appear when the solutions be published.

Proposals
11. Let n be a positive integer. Compute the following sum
n  
X k+4 n
(k + 1)(k + 2)(k + 3) k
k=0

12. Let α > 0 be a real number and let f : [−α, α] → R be a continuous function
two times derivable in (−α, α) such that f (0) = 0 and f 00 is bounded in (−α, α).
Show that the sequence {xn }n≥1 defined by
 n  
 Xf k , n > 1;

xn = n2 α
 k=1
n ≤ α1

0,
is convergent and determine its limit.

13. Let n be a positive integer. Compute

n
X X 2k
(i1 + 1)(i2 + 1) . . . (ik + 1)
k=1 1≤i1 <...<ik ≤n

14. Let Fn be the nth Fibonacci number defined by F0 = 0, F1 = 1 and for all
n ≥ 2, Fn = Fn−1 + Fn−2 . Prove that
n  2 2
1 X Tk Tn+1
≥ ,
n2 Fk 9Fn Fn+1
k=1

where Tk is the k th triangular number defined by Tk = k+1

2 for all k ≥ 1.

15. Prove that

Z 1p Z 1p
1 3 3
+ x + ln(1 + x) dx (x + ln(1 + x))2 dx < 2 ln 2
2 0 0
 25

Solutions
6. Let a, b, c be the lengths of the sides of a triangle ABC with circumradius r and
area A. Compute
cos A − cos B cos B − cos C cos C − cos A
+ +
A − rc A − ra A − rb

(Spanish First Stage 2007)

Solution by José Gibergans-Báguena and José Luis Dı́az-Barrero, Tech-

nical University of Catalonia (BARCELONA TECH), Barcelona, Spain.
Since A = pr, (p represents the semi-perimeter of 4ABC), then A−ra = rp−ra =
r(p − a) (cyclic) and the given expression can be written as
cos A − cos B cos B − cos C cos C − cos A
+ +
r(p − c) r(p − a) r(p − b)
 
1 cos A − cos B cos B − cos C cos C − cos A
= + +
r p−c p−a p−b
 
2 cos A − cos B cos B − cos C cos C − cos A
= + +
r a+b−c b+c−a a−b+c
Applying the Law of Cosine, yields
 2
c + b2 − a2 c2 + a2 − b2

cos A − cos B 1
= −
a+b−c a+b−c 2bc 2ac
b(c2 + b2 − a2 ) − a(c2 + a2 − b2
=
2abc(a + b − c)
(a − b)[c2 − (a + b)2 ] 2p(b − a)
= =
2abc(a + b − c) abc

cos B − cos C 2p(c − b) cos C − cos A 2p(a − c)

Likewise, = and = . Hence,
b+c−a abc a−b+c abc
cos A − cos B cos B − cos C cos C − cos A
+ +
A − rc A − ra A − rb
 
2 cos A − cos B cos B − cos C cos C − cos A
= + +
r a+b−c b+c−a a−b+c
 
2 2p(b − a) 2p(c − b) 2p(a − c)
= + + =0
r abc abc abc
and we are done.

7. Let ln a, ln b and ln c be the lengths of the sides of a triangle ABC. Prove that
3 ln a ln b ln c
≤ + + <1
5 ln(ab2 c2 ) ln(a2 bc2 ) ln(a2 b2 c)

(Shortlist XIX Ibero 2004)

26

Solution 1 by Valmir Bucaj, Texas Lutheran University, Seguin, TX.

Since the sum of any two sides of a triangle is greater than the third, then we have
X ln a X ln a X ln a
2 2
= <
ln(ab c ) ln a + 2(ln b + ln c) ln a + 2 ln a
cyclic cyclic cyclic
X ln a X 1 1 1 1
= = = + + =1
3 ln a 3 3 3 3
cyclic cyclic

To prove the LHS inequality, we suppose that ln a ≥ ln b ≥ ln c. Since the conditions

for Chebyshev’s Inequality are satisfied, then applying Chebyshev’s Inequality and
AM − HM successively, we get
 
X ln a 1 X 1
≥ ln(abc)  
ln(ab2 c2 ) 3 ln(ab2 c2 )
cyclic cyclic
 −1
ln(abc)  X
≥ 9 ln(ab2 c2 )
3
cyclic

3 ln(abc) 3
= = .
5 ln(abc) 5
This completes the proof.

Solution 2 by José Luis Dı́az-Barrero, Technical University of Catalonia

(BARCELONA TECH), Barcelona, Spain. First, we write the statement in
the most convenient form
3 ln a ln b ln c
≤ + + <1
5 ln a + 2(ln b + ln c) ln b + 2(ln c + ln a) ln c + 2(ln a + ln b)
RHS inequality trivially holds from
ln b + ln c ln c + ln a ln a + ln b
> 1, > 1 and >1
ln a ln b ln c
To prove LHS inequality, we put
ln a ln b ln c
x= , y= , z= ,
ln a + ln b + ln c ln a + ln b + ln c ln a + ln b + ln c
and we have
ln a ln b ln c
+ +
ln a + 2(ln b + ln c) ln b + 2(ln c + ln a) ln c + 2(ln a + ln b)
x y z
= + +
2−x 2−y 2−z
 h
1 1 1 i
Since + + (2 − x) + (2 − y) + (2 − z) ≥ 9, or equivalently,
2−x 2−y 2−z
 h
1 1 1 i
+ + 6 − (x + y + z) ≥ 9,
2−x 2−y 2−z
then
1 1 1 9
+ + ≥
2−x 2−y 2−z 5
 27

on account of the fact that x + y + z = 1. Therefore, from the preceding we have

x y z x+2−2 y+2−2 z+2−2
+ + = + +
2−x 2−y 2−z 2−x 2−y 2−z
 
1 1 1 3
=2 + + −3≥
2−x 2−y 2−z 5
Notice that equality holds when ln a = ln b = ln c. That is, when 4ABC is equilate-
ral.

Also solved by José Gibergans-Báguena, Technical University of Catalo-

nia (BARCELONA TECH), Barcelona, Spain, Paolo Perfetti, Depart-
ment of Mathematics, Tor Vergata University, Rome, Italy.

8. Suppose that the three roots of the equation t3 − at2 + t − b = 0 are positive real
numbers. Show that 9b2 (1 + 6ab) ≤ 1.
(Longlist OME 2006)

Solution by José Luis Dı́az-Barrero, Technical University of Catalonia

(BARCELONA TECH), Barcelona, Spain. First, we observe that the coe-
fficients b and c are nonzero. Dividing by b3 both sides of 9b2 (1 + 6ab) ≤ 1 and
taking the cubic root of both sides of the resulting inequality, yields
r √
3
2 3 1 3
9b (1 + 6ab) ≤ 1 ⇔ 3 + 6a ≤
b b
Let x, y, z be the roots of t3 − at2 + t − b = 0. On account of Cardan’s formulae,
we have
x + y + z = a,
xy + yz + zx = 1,
xyz = b,
and the last inequality becomes
r √3
1 3
3 3
+ 6(x + y + z) ≤ ,
xyz xyz
or equivalently,
r
3 1 1
√
3
3
+ 6(x + y + z) ≤
3 xyz xyz
We have
1 1 + 6x2 yz + 6xy 2 z + 6xyz 2
+ 6(x + y + z) =
xyz xyz
1 + 3xy(xz + yz) + 3yz(yx + zx) + 3zx(xy + yz)
=
xyz
Taking into account that xy + yz + zx = 1, we get
s
3 3 4 − 3[(xy)2 + (yz)2 + (zx)2 ] 1
√3
≤
3 xyz xyz
28

Now, from xy + yz + zx = 1 and

3[(xy)2 + (yz)2 + (zx)2 ] ≥ (xy + yz + zx)2 = 1,
to prove the last inequality it suffices to prove
r
3 3 3 1
√
3
≤
3 xyz xyz
which follows immediately from AM-GM inequality. Indeed,
 3
2 2 2 xy + yz + zx 1
x y z = (xy)(yz)(zx) ≤ = 3
3 3
Multiplying both sides by xyz and reordering terms, yields
33
r
1 3 3 3 1
≤ 3
⇔ √ 3
≤
xyz (xyz) 3 xyz xyz
√
1 √ 3
Equality holds when x = y = z = √ . That is, when a = 3 and b = .
3 9
Also solved by Bruno Salgueiro Fanego, Viveiro, Spain; and José Gibergans-
Báguena, Technical University of Catalonia (BARCELONA TECH),
Barcelona, Spain, Paolo Perfetti, Department of Mathematics, Tor Ver-
gata University, Rome, Italy.

9. Let {an }n≥0 be the sequence defined by a0 = 1, a1 = 2, a2 = 1 and for all n ≥ 3,

a3n = an−1 an−2 an−3 . Find lim an .
n→∞

(Longlist IMC 2006)

Solution by José Gibergans-Báguena and José Luis Dı́az-Barrero, Tech-

nical University of Catalonia (BARCELONA TECH), Barcelona, Spain.
Setting bn = log2 an , we get the sequence {bn }n≥0 defined by b0 = 0, b1 = 1, b2 = 0
and for all n ≥ 3, 3bn = bn−1 + bn−2 + bn−3 . The √ characteristic√ equation of bn is
3t3 − t3 − t − 1 = 0 which roots are 1, − 31 (1 − i 2), − 31 (1 + i 2). Therefore,
n n
√ √
 
1 1
bn = a + b − (1 − i 2) + c − (1 + i 2)
3 3
Taking into account of the initial conditions, we get the system of equations
a+b+c = 0
√ √
   
1 1
a + b − (1 − i 2) + c − (1 + i 2) = 1
3 3
2 2
√ √
 
1 1
a + b − (1 − i 2) + c − (1 + i 2) = 0
3 3
1 1 √ 1 √
with solutions a = , b = − (1 − 5i 2/2) and c = − (1 + 5i 2/2). Thus, if
3 6 6
limn→∞ an = L, then
1 √
3
log2 L = lim log2 an = lim bn = ⇒ L = 2
n→∞ n→∞ 3
and we are done.
 29

Also solved by Paolo Perfetti, Department of Mathematics, Tor Vergata

University, Rome, Italy.

10. Let x, y, z be three distinct positive real numbers. Prove that

1 X ln x2x 1
< <
max{x, y, z} (x − y)(x − z) min{x, y, z}
cyclic

(Longlist IMC 2009)

Solution by José Luis Dı́az-Barrero, Technical University of Catalonia

(BARCELONA TECH), Barcelona, Spain. The inequality claimed is equiv-
alent to
1 x ln x y ln y z ln z 1
< + + <
2 max{x, y, z} (x − y)(x − z) (y − x)(y − z) (z − x)(z − y) 2 min{x, y, z}
Taking into account that
x y z
+ + = 0,
(x − y)(x − z) (y − x)(y − z) (z − x)(z − y)
as can be easily proven, we have that the last inequality is equivalent to
1 x(ln x − 1) y(ln y − 1) z(ln z − 1) 1
< + + <
2 max{x, y, z} (x − y)(x − z) (y − x)(y − z) (z − x)(z − y) 2 min{x, y, z}
Applying the well-known result [1] from the theory of divided differences
n n
X Y 1
f [z0 , z1 , · · · , zn ] = f (zj )
j=0
zj − zk
k=0
k6=j

to the function f (t) = t(ln t − 1), we get

f (x) f (y) f (z)
f [x, y, z] = + +
(x − y)(x − z) (y − x)(y − z) (z − x)(z − y)
x(ln x − 1) y(ln y − 1) z(ln z − 1)
= + +
(x − y)(x − z) (y − x)(y − z) (z − x)(z − y)
Now we need the following result.
Lema 1. Let f : [a, b] → R be a real valued function with hsecond derivative f 00 (x) i
continuous in [a, b] and x, y, z ∈ [a, b]. Then there exists c ∈ min{x, y, z}, max{x, y, z}
f 00 (c)
such that f [x, y, z] = .
2
Proof. Since f 00 (x) is continuous in [a, b], then it has a maximum and a minimum
in [a, b]. Let m = min f 00 (x) and M = max f 00 (x). Then from the integral repre-
a≤x≤b a≤x≤b
sentation of f [x, y, z], we have
Z 1 Z t1 Z 1 Z t1
m dt1 dt2 ≤ f [x, y, z] ≤ M dt1 dt2
0 0 0 0
and
m M
≤ f [x, y, z] ≤ or m ≤ 2f [x, y, z] ≤ M
2 2
30

Since f 00 (x) is continuous, by applying

h the intermediate value itheorem to it, we
00
have 2f [x, y, z] = f (c) for some c ∈ min{x, y, z}, max{x, y, z} and the proof is
complete. 

Applying Lemma 1 to the function f (t) = t(lnh t−1) to which f 00 (t) is thei decreasing
function f 00 (t) = 1/t then there exists c ∈ min{x, y, z}, max{x, y, z} such that
1 1 1 1 1
f [x, y, z] = f 00 (c) = . So, ≤ ≤ . Since x, y, z
2 2c 2 max{x, y, z} 2c 2 min{x, y, z}
are distinct then the statement follows and we are done.

References
[1] E. Isaacson and H. B. Keller. Analysis of Numerical Methods. Dover, New York, 1994.

Solution 2 by Paolo Perfetti, Department of Mathematics, Tor Vergata

University, Rome, Italy. Both the LHS and the RHS are symmetric and ho-
mogeneous. The symmetry is evident while for the homogeneity we multiply each
variable by α obtaining
1 X 2αx(ln α + ln x) 1
< <
max{αx, αy, αz} α2 (x − y)(x − z) min{αx, αy, αz}
cyclic

Now we observe that X x

=0
(x − y)(x − z)
cyclic
thus we remain with
1 X 2 ln x 1
< <
α max{x, y, z} α(x − y)(x − z) α min{x, y, z}
cyclic

that is the homogeneity. By symmetry we can set x ≤ y ≤ z and by homogeneity

x = 1 thus we have
1 2y ln y 2z ln z 1
< + < =1
z (y − 1)(y − z) (z − 1)(z − y) x
The RHS becomes
2z ln z 2y ln y
−z ≤ −y (2)
z−1 y−1
2z ln z
Let f (z) = − z, z ≥ 1. Then,
z−1
z 2 − 4z + 3 + 2 ln z
f 0 (z) = − ≤ 0 ⇐⇒ z 2 − 4z + 3 + 2 ln z ≥ 0
(z − 1)2
2
(z 2 − 4z + 3 + 2 ln z)|z=1 = 0, (z 2 − 4z + 3 + 2 ln z)0 = 2z − 4 +
>0∀z≥1
z
It follows that f (z) does not increase yielding (2) and concluding the proof of the
RHS.

As for the LHS of (2) we need to prove that

z−y 2z ln z 2y ln y
< − (3)
z z−1 y−1
 31

2ξ ln ξ
Let f (ξ) = . By the Lagrange theorem f (z) − f (y) = f 0 (c)(z − y) where
ξ−1
ξ − 1 − ln ξ
y < c < z and therefore f 0 (ξ) = 2 . We conclude the computation
(ξ − 1)2
writing a Lemma
ξ − 1 − ln ξ 1
Lema 2. For all ξ > 1 holds 2 > .
(ξ − 1)2 ξ
Proof. It is equivalent to show
ξ 2 − ξ − ln ξ
F (ξ) = 2 − 1 > 0, F (1) = lim f (ξ) = 0
(ξ − 1)2 ξ→1

2(ln(ξ)ξ + ln(ξ) − 2ξ + 2)
F 0 (ξ) = ≥ 0 ⇐⇒ h(ξ) = (ln(ξ)ξ + ln(ξ) − 2ξ + 2) ≥ 0
(x − 1)3
2(−ξ + ln(ξ)ξ + 1)
h(1) = 0, h0 (ξ) = ≥ 0 ⇐⇒ k(ξ) = −ξ + ln(ξ)ξ + 1 ≥ 0
ξ
k(1) = 0, k 0 (ξ) = ln(ξ) ≥ 0, ξ ≥ 1
thus the assertion of the Lemma. q.e.d. 

Now the c in f (z) − f (y) = f 0 (c)(z − y) satisfies y < c < z and then by the Lemma
f 0 (c) ≥ 1c > z1 proving (3) and concluding the proof.
Also solved by José Gibergans-Báguena, Technical University of Catalo-
nia (BARCELONA TECH), Barcelona, Spain.