Junior problems

J343. Prove that the number 102400 . . . 002401, having a total of 2014 zeros, is composite.
Proposed by Titu Andreescu, University of Texas at Dallas, USA

First solution by Polyahedra, Polk State College, USA

By 4a4 + b4 = (2a2 2ab + b2 )(2a2 + 2ab + b2 ), the given number equals
1024 102016 + 2401 = 4 4 10504
=

4

+ 74

32 101008 56 10504 + 49

32 101008 + 56 10504 + 49 .

Second solution by Polyahedra, Polk State College, USA

Because 104 1 (mod 137), the given number equals
1024 104

504

+ 2401 1024(1)504 + 2401 3425 0

(mod 137).

Also solved by Adnan Ali, A.E.C.S-4, Mumbai, India; Alok Kumar, Delhi, India; Arkady Alt, San Jose,
CA, USA; Corneliu Mnescu-Avram, Transportation High School, Ploieti, Romania; David Stoner, Harvard University, Cambridge, MA, USA; Devesh Rajpal, Raipur, Chhattisgarh, India; Ercole Suppa, Teramo,
Italy; Haimoshri Das, South Point High School, India; Jos Hernndez Santiago, Mxico; Paul Revenant,
Lyce Champollion, Grenoble, France; Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick
Lpez, IMPA, Brazil; Albert Stadler, Herrliberg, Switzerland; Hyun Jin Kim, Stuyvesant High School, New
York, NY, USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Tho Lenoir, Institut Saint-L,
Agneaux, France; Michael Tang, Edina High School, MN, USA; Ioan Viorel Codreanu, Satulung, Maramures,
Romania.

Mathematical Reflections 4 (2015)

J344. Find the maximum possible value of k for which

a2 + b2 + c2

a+b+c
3

2

k max (a b)2 , (b c)2 , (c a)2

for all a, b, c R.
Proposed by Dominik Teiml, University of Oxford, United Kingdom

Solution by Daniel Lasaosa, Pamplona, Spain

Note that the problem is invariant under exchange of any two of a, b, c, or we may assume wlog a b c,
and denote u = a c, v = a b with u v 0, and
a2 + b2 + c2

3
=

a+b+c
3

2
=

(a b)2 + (b c)2 + (c a)2

v 2 + (u v)2 + u2
=
=
9
9

3u2 + (u 2v)2
u2
1

= max (a b)2 , (b c)2 , (c a)2 ,

18
6
6

or the maximum value is k = 61 , with equality iff u = 2v, ie iff a, b, c are in arithmetic progression.
Also solved by Polyahedra, Polk State College, USA; Hyun Jin Kim, Stuyvesant High School, New York,
NY, USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Tho Lenoir, Institut Saint-L,
Agneaux, France; Alok Kumar, Delhi, India; Arkady Alt, San Jose, CA, USA; Corneliu Mnescu-Avram,
Transportation High School, Ploieti, Romania; Ercole Suppa, Teramo, Italy; Joel Schlosberg, Bayside, NY,
USA; Nick Iliopoulos, Music Junior HS, Trikala, Greece; Nicusor Zlota, Traian Vuia Technical College, Focsani, Romania; Paolo Perfetti, Universit degli studi di Tor Vergata Roma, Roma, Italy; Paul Revenant,
Lyce Champollion, Grenoble, France; Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick
Lpez, IMPA, Brazil; Albert Stadler, Herrliberg, Switzerland.

Mathematical Reflections 4 (2015)

J345. Let a and b be positive real numbers such that

a6 3b5 + 5b3 3a = b6 + 3a5 5a3 + 3b.
Find a b.
Proposed by Titu Andreescu, University of Texas at Dallas, USA

Solution by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil
We have
a6 3a5 + 5a3 3a = b6 + 3b5 5b3 + 3b,
(a2 a 1)3 + 1 = (b2 + b 1)3 + 1,
(a2 a 1)3 = (b2 + b 1)3 ,
but x3 = y 3 for real x, y if and only if x = y. Obtaining a2 a1 = b2 +b1, it follows that (ab1)(a+b) = 0
and finally a b = 1 since a + b > 0.
Also solved by Daniel Lasaosa, Pamplona, Spain; Polyahedra, Polk State College, USA; Hyun Jin Kim,
Stuyvesant High School, New York, NY, USA; Michael Tang, Edina High School, MN, USA; Adnan Ali,
A.E.C.S-4, Mumbai, India; Arkady Alt, San Jose, CA, USA; Corneliu Mnescu-Avram, Transportation High
School, Ploieti, Romania; Ercole Suppa, Teramo, Italy; Joseph Lee, Loomis Chaffee School, Windsor, CT,
USA; Nicusor Zlota, Traian Vuia Technical College, Focsani, Romania; Paul Revenant, Lyce Champollion,
Grenoble, France; Rade Krenkov, SOUUD, Dimitar Vlahov, Strumica, Macedonia.

Mathematical Reflections 4 (2015)

J346. We are given a sequence of 12n numbers that are consecutive terms of an arithmetic progression. We
randomly choose four numbers. What is the probability that among the chosen numbers there will be
three in arithmetic progression?
Proposed by Ivan Borsenco, Massachusetts Institute of Technology, USA

Solution by Polyahedra, Polk State College, USA

Without loss of generality,
assume the given sequence

is S = {1,
 2, . . . , 12n}.

For d 1, let Ad = (a, a + d, a + 2d, a + 3d) S 4 and Bd = (a, a + d, a + 2d, b) S 4 : b a 6= 0, d, 2d, d2 , 3d
2 .
Then |Ad | = 12n 3d, |Bd | = (12n 2d)(12n 3) for odd d, and |Bd | = (12n 2d)(12n
P4n d. By
P 5) for even
|B
|

PIE (the principle of inclusion-exclusion), the total number of favorable outcomes is 6n

d
d=1 |Ad |,
d=1
which equals
(12n 3)

3n
X

(12n 4i + 2) + (12n 5)

i=1

3n
4n
X
X
(12n 4i)
(12n 3i)
i=1

i=1

= (12n 3)(3n)(6n) + (12n 5)(3n)(6n 2) 2n(12n 3) = 12n(36n2 20n + 3).

Hence, the probability we are seeking is
12n(36n2 20n + 3)


=
12n
4

4(36n2 20n + 3)
.
(12n 1)(6n 1)(4n 1)

Also solved by Daniel Lasaosa, Pamplona, Spain; Robert Bosch, Archimedean Academy, Florida, USA
and Jorge Erick Lpez, IMPA, Brazil.

Mathematical Reflections 4 (2015)

J347. Let x, y, z be positive real numbers. Prove that

x
y
z
6xyz
+
+

.
2x + y + z x + 2y + z x + y + 2z
(x + y)(y + z)(z + x)
Proposed by Titu Zvonaru and Neculai Stanciu, Romania

Solution by Adnan Ali, A.E.C.S-4, Mumbai, India

Observe that the inequality is equivalent to
P

cyc

1
6

.
yz(2x + y + z) (x + y)(y + z)(z + x)

But by the Cauchy-Schwartz Inequality,

P

cyc

9
9
1
X
=
.
yz(2x + y + z)
yz(2x + y + z) (x + y)(y + z)(z + x) + 4xyz
cyc

Hence, it suffices to prove that

9
6

(x + y)(y + z)(z + x) 8xyz,

(x + y)(y + z)(z + x) + 4xyz (x + y)(y + z)(z + x)
which follows by the AM-GM Inequality. Equality holds iff x = y = z.
Also solved by Rade Krenkov, SOUUD, Dimitar Vlahov, Strumica, Macedonia; Daniel Lasaosa, Pamplona, Spain; Polyahedra, Polk State College, USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Ioan Viorel Codreanu, Satulung, Maramures, Romania; George Gavrilopoulos, Nea Makri High School,
Athens, Greece; Alok Kumar, Delhi, India; Andrea Fanchini, Cant, Italy; Arkady Alt, San Jose, CA, USA;
David Stoner, Harvard University, Cambridge, MA, USA; Ercole Suppa, Teramo, Italy; Haimoshri Das,
South Point High School, India; Nicusor Zlota, Traian Vuia Technical College, Focsani, Romania; Paolo
Perfetti, Universit degli studi di Tor Vergata Roma, Roma, Italy; Paul Revenant, Lyce Champollion, Grenoble, France; Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil;
Brian Bradie, Christopher Newport University, Newport News, VA, USA.

Mathematical Reflections 4 (2015)

J348. Let ABCDEF G be a regular heptagon. Prove that

AD3
AC + 2AD
3 = AB + AD .
AC
Proposed by Dragoljub Milosevic, Gornji Milanovac, Serbia

Solution by Ercole Suppa, Teramo, Italy

Let AB = a, AC = b, AD = c as shown in the following figure.
E
F

c
C

G
b

Ptolemys theorem applied to quadrilaterals ABCD, ABCE,ABDF yields

a2 + ac = b2

(1)

ab + ac = bc

(2)

b + ab = c

(3)

Now, by using (1),(2),(3) we get

AC + 2AD
b + 2c
a(b + 2c) (1) a(b + 2c)
=
= 2
=
=
AB + AD
a+c
a + ac
b2
ab(b + 2c) (2) (bc ac)(b + 2c)
=
=
=
b3
b
3


c b2 ab + 2(bc ac) (2)
c b2 + 2bc ab 2ac
=
=
=
b3

b3


c b2 ab + 2ab
c b2 + ab (3)
=
=
=
b3
b3
c3
AD3
= 3 =
b
AC 3
as desired. The proof is complete.
Also solved by Rade Krenkov, SOUUD, Dimitar Vlahov, Strumica, Macedonia; Daniel Lasaosa, Pamplona, Spain; Polyahedra, Polk State College, USA; Michael Tang, Edina High School, MN, USA; Adnan
Ali, A.E.C.S-4, Mumbai, India; Corneliu Mnescu-Avram, Transportation High School, Ploieti, Romania;
Nicusor Zlota, Traian Vuia Technical College, Focsani, Romania; Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil; SooYoung Choi, Mount Michael Benedictine, Elkhorn, NE,
USA.

Mathematical Reflections 4 (2015)

Senior problems

S343. Let a, b, c, d be positive real numbers. Prove that

a+b+c+d
16abcd

+
5.
4
(a + b)(b + c)(c + d)(d + a)
abcd
Proposed by Titu Andreescu, USA and Alok Kumar, India

Solution by the authors

b+c
c+d
d+a
16
a+b

+
+
+
+
5
4
4
4
4
2 abcd 2 abcd 2 abcd 2 abcd (a + b)(b + c)(c + d)(d + a)
by AM-GM.
Also solved by Brian Bradie, Christopher Newport University, Newport News, VA, USA; Daniel Lasaosa,
Pamplona, Spain; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Michael Tang, Edina High
School, MN, USA; Ioan Viorel Codreanu, Satulung, Maramures, Romania; SooYoung Choi, Mount Michael
Benedictine, Elkhorn, NE, USA; Adnan Ali, A.E.C.S-4, Mumbai, India; David Stoner, Harvard University,
Cambridge, MA, USA; Ercole Suppa, Teramo, Italy; Francesco De Sclavis, Universit degli Studi di Roma
Tor Vergata, Rome, Italy; George Gavrilopoulos, Nea Makri High School, Athens, Greece; Nicusor Zlota,
Traian Vuia Technical College, Focsani, Romania; Paolo Perfetti, Universit degli studi di Tor Vergata
Roma, Roma, Italy; Paul Revenant, Lyce Champollion, Grenoble, France; Rade Krenkov, SOUUD, Dimitar
Vlahov, Strumica, Macedonia; Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez,
IMPA, Brazil; Anant Mudgal, Delhi Public School, Faridabad, Haryana, India; Li Zhou, Polk State College,
Winter Haven, FL, USA.

Mathematical Reflections 4 (2015)

S344. Find all non-zero polynomials P Z[X] such that a2 + b2 c2 |P (a) + P (b) P (c).
Proposed by Vlad Matei, University of Wisconsin, USA

Solution by Daniel Lasaosa, Pamplona, Spain

Taking b = c, we find that for all a Z, we have a2 |P (a). If the constant or linear term coefficients in P (x)
are non-zero, we may choose a = p, where p is a prime that does not divide the constant or the linear term
coefficients, reaching a contradiction, or P (x) = x2 Q(x) for some polynomial Q(x) with integer coefficients.
Let now u, v be any two distinct positive integers with u > v, and let a = u2 v 2 , b = 2uv and c = u2 +v 2 ,
or a2 + b2 = c2 , or by the condition given in the problem statement we need P (c) = P (a) + P (b). Now, if
the degree of Q(x) is at least 1, then the absolute value of Q(x) is increasing for all sufficiently large x, and
taking a, b, c sufficiently large, we would have Q(c) > Q(a), Q(b), for
P (c) = c2 Q(c) = a2 Q(c) + b2 Q(c) > a2 Q(a) + b2 Q(b) = P (a) + P (b),
reaching a contradiction, or Q(x) must be constant, ie an integer. It follows
kx2 for

that necessarily P2 (x) =
2
2
2
2
some nonzero integer k, and clearly P (a) + P (b) P (c) = k a + b c is a multiple of a + b c2 , or
P (x) is a solution iff it is a polynomial of this form.
Also solved by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazi;
Anant Mudgal, Delhi Public School, Faridabad, Haryana, India.

Mathematical Reflections 4 (2015)

S345. Solve in positive integers the system of equations

(x 3)(yz + 3) = 6x + 5y + 6z,
(y 3)(zx + 3) = 2x + 6y,

(z 3)(xy + 3) = 4x + y + 6z.
Proposed by Titu Andreescu, University of Texas at Dallas, USA

First solution by the author

Summing the three equations we obtain
3xyz 3(xy + yz + zx) = 9(x + y + z) + 27,
and so
4xyz = (x + 3)(y + 3)(z + 3),
This is equivalent to


3
1+
x




3
3
1+
1+
= 4.
y
z

From the given equations it follows that x, y, z 4 and that x, y, z are not all equal. If min(x, y, z) > 4,
then



 



3
3
3
3
3
3
96
1+
1+
1+
< 1+
1+
1+
=
< 4,
x
y
z
5
5
6
25
a contradiction. Hence min(x, y, z) = 4 and let {a, b} = {x, y, z} {4} . Then
 


7
3
3
1+
1+
= 4,
4
a
b
which reduces to (3a 7)(3b 7) = 7 16.
It follows that 7 divides a or b, and since a and b are greater than 3, the only possibility is {a, b} = {5, 7} .
So {x, y, z} = {4, 5, 7} and, looking back at the original equations, we see that the only solution is
(x, y, z) = (7, 4, 5).
Second solution by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA,
Brazil
The only solution is (x, y, z) = (7, 4, 5). The right side on each equation is positive, so x, y, z 4. Now by
the second equation we obtain
4 x 2x + 24 8x 2x + 24 2zx y =

2x + 3zx + 9
5.
zx 3

Hence, y = 4 or y = 5. When y = 5 we have equality in the previous inequality, therefore x = z = 4,

but (4, 5, 4) is not solution of the system. For y = 4 in the second equation we get zx + 3 = 2x + 24, or
equivalently x(z 2) = 21, and finally x = 7 and z = 5, that satisfies the other two equations.
Also solved by Daniel Lasaosa, Pamplona, Spain; Antoine Faisant; David E. Manes, Oneonta, NY,
USA; Ercole Suppa, Teramo, Italy; Nicusor Zlota, Traian Vuia Technical College, Focsani, Romania; Paul
Revenant, Lyce Champollion, Grenoble, France; Rade Krenkov, SOUUD, Dimitar Vlahov, Strumica, Macedonia.

Mathematical Reflections 4 (2015)

S346. Let a, b, c be positive real numbers. Prove that

1
1
1
15
6
+
+
+

.
2
2
2
2
(a + b)
(b + c)
(c + a)
(a + b + c)
ab + bc + ca
Proposed by Marius Stnean, Zalu, Romania

Solution by Adnan Ali, Student in A.E.C.S-4, Mumbai, India

W.L.O.G assume that a b c. Let
f (a, b, c) =

1
1
1
15
6
+
+
+

.
2
2
2
2
(a + b)
(b + c)
(c + a)
(a + b + c)
ab + bc + ca

Then
f (a, b, c) f (a, b + c, 0) =

1
1
1
6
6
1

+
2
+
2
2
2
(a + b)
(a + b + c)
(c + a)
a
ab + bc + ca a(b + c)
6bc
c(2a + c)

.
a(b + c)(ab + bc + ca) a2 (a + c)2

Now notice that since a b c, we have

6ab (2a + c)(b + c),

(a + c)2 (ab + bc + ca)

proving that f (a, b, c) f (a, b + c, 0). So, now it remains to show that f (a, b + c, 0) 0. Let b + c = d, then
by the AM-GM Inequality,
!

1
1
16
(a + d)2
16ad
ad
+
+
=
2
+
+
2 + 2 16 = 6,
2
2
2
2
a
d
(a + d)
ad
(a + d)
and the conclusion follows. Equality holds when a = b, c = 0 and its respective permutations.
Also solved by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Arkady Alt, San Jose, CA,
USA; David Stoner, Harvard University, Cambridge, MA, USA; Ercole Suppa, Teramo, Italy; George Gavrilopoulos, Nea Makri High School, Athens, Greece; Nick Iliopoulos, Music Junior HS, Trikala, Greece;
Nicusor Zlota, Traian Vuia Technical College, Focsani, Romania; Paolo Perfetti, Universit degli studi di
Tor Vergata Roma, Roma, Italy; Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez,
IMPA, Brazil; Subhadeep Dey, Shyamnagar, West Bengal, India.

Mathematical Reflections 4 (2015)

10

S347. Prove that a convex quadrilateral ABCD is cyclic if and only if the common tangent to the incircles
of triangles ABD and ACD, different from AD, is parallel to BC.
Proposed by Nairi Sedrakyan, Armenia
Solution by the author
Denote the centers of the circles inscribed in triangles ABD and ACD as O1 and O2 respectively, and the
common tabgent, different from AD, as l.

Notince that l and AD are symmetric to each other with respect to O1 O2 , thus l and BC are parallel if
and only if CBD ADB = 2(O2 O1 D O1 DA), i.e.
CBD = 2O2 O1 D

(1)

If quadrilateral ABCD can be inscribed into a circle, then ABD = ACD and AO1 D = 90 +
1
1
ABD = 90 + ACD = AO2 D points A, O1 , O2 , D are laying on the same circle. Hence, CBD =
2
2
CAD = 2O2 AD = 2O2 O1 D, i.e. lP BC .
Now let us prove that quadrilateral ABCD is circumscribable. Let CAD = , CAB = 1 , ABD =
, DBC = 1 , ACB = , ACD = 1 , CDB = , BDA = 1 .
According to the Laq of Sines
sin
sin sin sin
AB BC CD DA

BC CD DA AB
sin 1 sin 1 sin 1 sin 1

1=
thus

sin sin beta sin sin = sin 1 sin beta1 sin 1 sin 1

(2)

Therefore, according to (1)

1

and AO2 O1 =
2
2

1
On the other side O2 AD = , O2 AO1 =
, AO1 D = 90 + , AO2 D = 90 + , O1 DA =
2
2
2
2
1

, O2 DO1 = .
2
2
Writing down expresison (2) for AO1 O2 D, get
O2 O1 D =

sin

1
1
1
1
cos sin sin = sin
sin
cos
sin
2
2
2
2
2
2
2
2

(3)

1
1
1
1
sin cos cos = cos
cos
sin
cos
2
2
2
2
2
2
2
2

(4)

Divide (2) by (3) and get

cos
Mathematical Reflections 4 (2015)

11

Let us prove that = 1 . Indeed, assume 6= 1 , then WLOG we can assume < 1 . In this case point
C will be located inside the circle circumscribed around triangle ABD. Therefore, > 1 and > 1
from (3) we get
cos

1
1
sin < sin
cos
2
2
2
2

and from (4) we get

1
1

sin
< cos sin
2
2
2
2

1 1
Adding up last two expressions we get sin
< sin
, which is impossible since = 1 1 .
2
2
Hence = 1 and the conclusion follows.
cos

Mathematical Reflections 4 (2015)

12

S348. Find all functions f : R R R such that for all a, b, c R

f a2 , f (b, c) + 1 = a2 (bc + 1).
Proposed by Mehtaab Sawhney, USA

Solution by Daniel Lasaosa, Pamplona, Spain

For any real x, take a = b = 1, c = x 1, or f (a2 , f (b, c) + 1) = x, and f is surjective. Assume now that
f (b, c) = f (u, v) for some (b, c) 6= (u, v). Then, for any nonzero a, we have




a2 (bc uv) = f a2 , f (b, c) + 1 f a2 , f (u, v) + 1 = 0,
or if f (b, c) = f (u, v), we have bc = uv. Reciprocally, assume that bc = uv, or for any nonzero a we have
f (a2 , f (b, c) + 1) = a2 (bc + 1) = a2 (uv + 1) = f (a2 , f (u, v) + 1),
hence by the previous result we have
a2 (f (b, c) + 1) = a2 (f (u, v) + 1) ,

a2 (f (b, c) f (u, v)) = 0,

and we conclude that f (b, c) = f (u, v) iff bc = uv. It then suffices to find f (1, x) = f (x, 1) for all real x,
since for any (y, z) R, f (y, z) = f (1, yz) = f (yz, 1). Now, since f is surjective, clearly b, c exist such that
f (b, c) = 0, and choosing any such pair we have f (a2 , 1) = a2 for any real a, hence f (1, x) = f (x, 1) = x for
all non-negative real x (this is consistent for x = 0 with taking a = 0. Now, taking a = b = 1, we obtain
f (1, f (1, c) + 1) = c + 1.
Note that this is zero when c = 1, and since f (1, 0) = f (0, 1) = 0, by previous results we obtain that
f (b, c) = 0 iff bc = 0. This produces f (1, 1) + 1 = 0, or f (1, 1) = 1. Finally, choose b, c such that
f (b, c) = 2, which is always possible since f is surjective, or for any real a, we have


f a2 , 1 = a2 (bc + 1).
In particular, taking a = 1 we have 1 = f (1, 1) = bc + 1, or bc = 2, ie f (a2 , 1) = a2 for all real a,
and f (1, x) = x for all non-negative real x, hence f (x) = x for all negative x.
Or f (b, c) = f (1, bc) = bc for all (b, c) RR is the only function that may satisfy the proposed equation,
and direct substitution easily yields that it is indeed a solution. The conclusion follows.
Also solved by Tho Lenoir, Institut Saint-L, Agneaux, France; Joel Schlosberg, Bayside, NY, USA;
Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil.

Mathematical Reflections 4 (2015)

13

Undergraduate problems

U343. Evaluate

lim

n
X
k=1

1
n + (k) log n

where (k) denotes the number of prime not exceeding k.

Proposed by Albert Stadler, Herrliberg, Switzerland

Solution by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil
log k
log k
and um := sup (k)
, then lm 1 and um 1 by the Prime Number
Let lm := inf (k)
km
k
k
km
Theorem. For n k m > 1 we have
log n
n + lm k,
log k
log(n/k)
e1 n
n + (k) log(n) n + um k + um k
n + um k + um
,
log k
log m

n + (k) log(n) n + lm k

using

k
n

log nk e1 . Then
lim sup
n

n
X
k=1

1
n + (k) log(n)

= lim sup
n

lim sup
n

= lim sup
n
1

n
X
k=m
n
X

1
,
n + (k) log(n)
1
,
n + lm k

k=m
n
X

1
n

k=1

dx
.
1 + lm x

1
,
1 + lm k/n

In a similar way we arrive to

lim inf
n

Z
Taking m , the limit is
0

n
X
k=1

n + (k) log(n)

dx
1+

um e1
log m

+ um x

dx
= log(1 + x)|10 = log 2.
1+x

Also solved by Paolo Perfetti, Universit degli studi di Tor Vergata Roma, Roma, Italy

Mathematical Reflections 4 (2015)

14

U344. Evaluate the following sum

n

X
3n 23 1 + 1
.
43n + 23n + 1

n=0

Proposed by Titu Andreescu, University of Texas at Dallas, USA

Solution by Joel Schlosberg, Bayside, NY, USA

For n a nonnegative integer,
n

3 + 2)(23 1) 43 23 1
3n
3n (23 + 2)
n (2

=
3

n
n
n
43 + 23 + 1 23 1
83n 1

3n+1
,
2
1
so for N a nonnegative integer, by a telescoping sum,
=

n
N
X
3n (23 1 + 1)

n=0

43n + 23n + 1
1
=
2

3n+1


N
X
1
n=0

3n
3n+1

23n 1 23n+1 1

3N +1
1 3N +1
2
1


.

Since x = o(2x 1),

n

X
3n (23 1 + 1)

n=0

43n

23n

1
= lim
N 2
+1


1

3N +1
23N +1 1

1
= .
2

Also solved by Daniel Lasaosa, Pamplona, Spain; Hyun Jin Kim, Stuyvesant High School, New York,
NY, USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Brian Bradie, Christopher Newport
University, Newport News, VA, USA; Li Zhou, Polk State College, Winter Haven, FL, USA; Adnan Ali,
A.E.C.S-4, Mumbai, India; Arkady Alt, San Jose, CA, USA; G. C. Greubel, Newport News, VA, USA;
Moubinool Omarjee, Lyce Henri IV, Paris, France; Robert Bosch, Archimedean Academy, Florida, USA
and Jorge Erick Lpez, IMPA, Brazil; Albert Stadler, Herrliberg, Switzerland.

Mathematical Reflections 4 (2015)

15

U345. Let R be a ring. We say that the pair (a, b) R R satisfies property (P ) if the unique solution of
the equation axa = bxb is x = 0. Prove that if (a, b) has property (P ) and a b is invertible, then the
equation axa bxb = a + b has a unique solution in R.
Proposed by Dorin Andrica, "Babes-Bolyai" University, Cluj Napoca, Romania

Solution by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia

Let (a, b) has property (P ) and (a b) is invertible.
c R :c(a b) = 1 ca cb = 1
(a b)c = 1 ac bc = 1
Since prove that c is solution of axa bxb = a + b.
aca bcb = (1 + bc)a bcb = a + bca bcb = a + bc(a b) = a + b.
Now prove that this solution is unique. Assume by contradiction there exists 2 solution.
ax1 a bx1 b = ax2 a bx2 b a(x1 x2 )a = b(x1 x2 )b
(a, b) has property (P ), hence x1 x2 = 0. Hence we have contradiction. Thus
axa bxb = a + b
equation has unique solution.
Also solved by Daniel Lasaosa, Pamplona, Spain; Shohruh Ibragimov, National University of Uzbekistan,
Uzbekistan; Corneliu Mnescu-Avram, Transportation High School, Ploieti, Romania; David Stoner, Harvard University, Cambridge, MA, USA; Joel Schlosberg, Bayside, NY, USA; Robert Bosch, Archimedean
Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil.

Mathematical Reflections 4 (2015)

16

U346. Let f : [0, 1] R be a twice differentiable function for which [f 0 (x)]2 + f (x)f 00 (x) 1. Prove that
 
Z 1
1
2
2 1
f (x)dx f
+ .
2
12
0
Proposed by Marcel Chirita, Bucharest, Romania

Solution by Li Zhou, Polk State College, USA

Let g(x) = f 2 (x). Then g 00 (x) = 2 [f 0 (x)]2 + f (x)f 00 (x) 2 for all x (0, 1). By Taylors theorem, for
each x [0, 1], there exists c(x) between x and 21 such that
 


 

1
1 00
1
1 2
0 1
+g
x
+ g (c(x)) x
g(x) = g
.
2
2
2
2
2
Hence,
Z
0

 Z 1


 

Z
1
1 1 00
1
1 2
0 1
+g
x
dx +
g (c(x)) x
dx
f (x)dx =
g(x)dx = g
2
2
2
2 0
2
0
0

 
 
Z 1
1 2
1
1
2 1
+0+
x
dx = f
+ .
g
2
2
2
12
0
2

Also solved by Daniel Lasaosa, Pamplona, Spain; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Brian Bradie, Christopher Newport University, Newport News, VA, USA; George Gavrilopoulos, Nea
Makri High School, Athens, Greece; Joel Schlosberg, Bayside, NY, USA; Paolo Perfetti, Universit degli studi di Tor Vergata Roma, Roma, Italy; Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick
Lpez, IMPA, Brazil; Shohruh Ibragimov, National University of Uzbekistan, Uzbekistan; Albert Stadler,
Herrliberg, Switzerland; Stanescu Florin, Serban Cioculescu School, Gaesti, Dambovita, Romania.

Mathematical Reflections 4 (2015)

17

U347. Find all differentiable functions f : [0, ) R such that f (0) = 0, f 0 is increasing and for all x 0
x2 f 0 (x) = f 2 (f (x))
Proposed by Stanescu Florin, Gaesti, Romania

Solution by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil
f (x) = 0 and f (x) = x are the only solutions. By the given equation f 0 (x) 0, but f (0) = 0, then f (x) is
a positive and increasing function, thus the same remains valid for the function f f (x). If (f f )(x) = 0,
by the given equation f 0 (x) = 0 with f (0) = 0, then f (x) = 0. If f f 6 0, from f (f (0)) = 0 and (f f )(x)
increasing, a = sup{x, f (f (x)) = 0} is well-defined and finite with f (f (a)) = 0 and f (f (x)) > 0 for all x > a.
For x > a


1
1 0
1
f 0 (x)

= 2
= 0,
f (f (x)) x
x
(f (f (x)))2
x
therefore f (f (x)) =
(for some constant c) for all x a, hence c 0 and a = 0, since f (f (a)) = 0.
1 + cx
1
Substituting in the original equation we have f 0 (x) =
for all x 0, but f 0 (x) is an increasing
(1 + cx)2
function, so c = 0 and f 0 (x) = 1 with f (0) = 0, and finally f (x) = x for all x 0.

Mathematical Reflections 4 (2015)

18

U348. Evaluate the linear integral

I
c

(1 + x2 y 2 ) dx + 2xy dy
(1 + x2 y 2 )2 + 4x2 y 2

where c is the square [sic] with vertices (2, 0), (2, 2), (2, 2), and (2, 0) traversed counterclockwise.
Proposed by ngel Plaza, Universidad de Las Palmas de Gran Canaria, Spain

Solution by Jos Hernndez Santiago, Mxico

The integral under consideration is equal to the real part of the following complex integral
I
1
dz.
2
c z +1

(1)

Indeed, if we write z in the form x + iy, where x, y R, then


I

I
1
1
dz
= Re
(dx + idy)
Re
2
2
2
c (1 + x y ) + 2ixy
c z +1
I

(1 + x2 y 2 ) 2ixy
= Re
(dx
+
idy)
2
2 2
2 2
c (1 + x y ) + 4x y
I
(1 + x2 y 2 ) dx + 2xy dy
.
=
2
2 2
2 2
c (1 + x y ) + 4x y
Thus, in order to solve the problem it suffices to evaluate the integral in (1). This can be done by means of
the Cauchy Integral Formula. Let U be any region of the complex plane containing c. If i C \ U , then
1
the function f : U C determined by the assignation z z+i
is holomorphic on U ; the Cauchy Integral
Formula gives us in this case that
I
c

1
dz =
2
z +1

I
c

1
z+i

zi

dz = 2if (i) = .

Hence,
I
c

(1 + x2 y 2 ) dx + 2xy dy
= Re
(1 + x2 y 2 )2 + 4x2 y 2

I
c

1
dz
2
z +1


=

and we are done.

Also solved by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Brian Bradie, Christopher
Newport University, Newport News, VA, USA; Li Zhou, Polk State College, Winter Haven, FL, USA; Paolo
Perfetti, Universit degli studi di Tor Vergata Roma, Roma, Italy; Robert Bosch, Archimedean Academy,
Florida, USA and Jorge Erick Lpez, IMPA, Brazil.

Mathematical Reflections 4 (2015)

19

Olympiad problems

O343. Let a1 , . . . , an be positive real numbers such that

a1 + a2 + + an = a1 + a2 + + an .
Prove that

p
p
p

a1 2 + 1 + a2 2 + 1 + + an 2 + 1 n 2.
Proposed by Titu Andreescu, University of Texas at Dallas, USA

Solution by Li Zhou, Polk State College, USA

Let A = a1 + + an = a1 + + an . By the Cauchy-Schwarz inequality,

q
p
a21 + 1 + + a2n + 1
q
q

=
(a1 + 2a1 + 1)(a1 2a1 + 1) + + (an + 2an + 1)(an 2an + 1)
q
p

(A + 2A + n)(A 2A + n) = 2n2 (n A)2 n 2.

Also solved by Rade Krenkov, SOUUD, Dimitar Vlahov, Strumica, Macedonia; Daniel Lasaosa, Pamplona, Spain; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Adnan Ali, A.E.C.S-4, Mumbai,
India; David Stoner, Harvard University, Cambridge, MA, USA; Ercole Suppa, Teramo, Italy; George Gavrilopoulos, Nea Makri High School, Athens, Greece; Joel Schlosberg, Bayside, NY, USA; Nicusor Zlota, Traian
Vuia Technical College, Focsani, Romania; Robert Bosch, Archimedean Academy, Florida, USA and Jorge
Erick Lpez, IMPA, Brazil.

Mathematical Reflections 4 (2015)

20

h
3

O344. Consider the sequence an =

65 4

n i

, where n N . Prove that an 2, 3 (mod 15).

Proposed by Vlad Matei, University of Wisconsin, USA

Solution by Daniel Lasaosa, Pamplona, Spain

1
2
= 3 65 + 4 65 + 16 = r2 + 4r + 16. Note that
Denote r = 3 65 and u = 3 65 4
u2 = 48r2 + 193r + 776,

u3 = 2316r2 + 9312r + 37441,

or u is one of the roots of equation

x3 48x2 12x 1 = 0,
and denoting by v, w the other two, we have v + w = 48 u = 32 4r r2 = 12(r 4) (r 4)2 and
vw = u1 = r 4. Note first that
3 42
r =4 +1=4 +
<
48
3

or 0 < r 4 <

1
48 ,



1 3
4+
,
48

hence 0 > v + w > 1 and 1 > vw > 0, or v, w < 0, and for every positive integer n,
0 < |v n + wn | = |v n | + |wn | |v| + |w| < 1,

or 1 > v n + wn > 0 iff n is even and 0 > v n + wn > 1 iff n is odd. Note next that
v 2 + w2 = (v + w)2 2vw = 1552 193r 48r2 ,
v 3 + w3 = (v + w)3 3vw(v + w) = 74882 9312r 2316r2 ,
for
u + v + w = 48 3

(mod 15),

u2 + v 2 + w2 = 2328 3

u3 + v 3 + w3 = 112323 3

(mod 15),

(mod 15).

Now, sequence (bn )n1 defined by bn+3 = 48bn+2 + 12bn+1 + bn , with initial conditions b1 = 48, b2 = 2328,
b3 = 112323, has characteristic equation with roots u, v, w, or bn = un + v n + wn for n N . Clearly all
bn are integers, and bn 3 (mod 15) for all n N , since b1 b2 b3 3 (mod 15), and by bn+3
3 (bn+2 bn+1 ) + bn (mod 15). We therefore conclude that, for all even n N , we have
an = [un ] = un + v n + wn 1 2

(mod 15),

while for all odd n N , we have

an = [un ] = un + v n + wn 3

(mod 15).

Therefore, we conclude that an 2, 3 (mod 15).

Also solved by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil;
Li Zhou, Polk State College, Winter Haven, FL, USA.

Mathematical Reflections 4 (2015)

21

O345. Let A1 , B1 , C1 be points on the sides BC, CA, AB of a triangle ABC. Let rA , rB , rC , r1 be the inradii
of triangles AB1 C1 , BC1 A1 , CA1 B1 and A1 B1 C1 respectively. Prove that
R1 r1 2 min(rA rB , rB rC , rC rA ),
where R1 is the circumcircle of triangle A1 B1 C1 .
Proposed by Nairi Sedrakyan, Armenia

Solution by the author

a

First we prove that for any triangle ha cot .

2
2

(a)

(b)

Indeed,
ha
sin sin
cos( ) cos( + )
1 + cos
1

=
=

= cot
a
sin
sin
sin
2
2
ha
1

cot
a
2
2
Denote the centers of circumscribed triangles AB1 C1 , BC1 A1 , CA1 B1 as O1 , O2 , O3 respectively.
1
1
1

u = B1 O1 C1 , v = C1 O2 A1 , w = A1 O3 B1 . Notice that u = +
,v = +
,w = +
2
2
2
4
4
4
4
4
therefore u + v + w = and

cot u cot v + cot v cot w + cot u cot w = 1

(1)
Let
C
,
4
(2)

Apply (1) and (2) for triangles B1 O1 C1 , C1 O2 A1 , A1 O3 B1 and get

2rA
2rB
2rB
2rC
2rC
2rA

1.
B1 C1 A1 C1 A1 C1 A1 B1 A1 B1 B1 C1
Therefore,
2 min(rA rB , rB rC , rC rA )

A1 B1 B1 C1 A1 C1
4R1 SA1 B1 C1
= R1 r1
=
2SA1 B1 C1
2(A1 B1 + B1 C1 + A1 C1 )
2
r1

and the conclusion folllows.

Also solved by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil.
Mathematical Reflections 4 (2015)

22

O346. Define the sequence (an )n0 by a0 = 0, a1 = 1, a2 = 2, a3 = 6 and

an+4 = 2an+3 + an+2 2an+1 an ,

n 0.

Prove that n2 divides an for infinitely many positive integers.

Proposed by Dorin Andrica, Babes-Bolyai University, Cluj Napoca, Romania

Solution by Li Zhou, Polk State College, USA

Let (Fn ) be the Fibonacci sequence, that is, F0 = 0, F1 = 1, and Fn+2 = Fn+1 + Fn for all n 0. Notice
that an = nFn for n = 0, 1, 2, 3. As an induction hypothesis, assume that k 3 and an = nFn for all
n = 0, 1, . . . , k. Then
ak+1 = 2kFk + (k 1)Fk1 2(k 2)Fk2 (k 3)Fk3
= 2kFk + 2Fk1 (k 1)Fk2 = (k + 1)Fk + (k + 1)Fk1 = (k + 1)Fk+1 ,
completing the induction step. Hence an = nFn for all n 0. Thus, it suffices to show that n|Fn for infinitely
many n.
Indeed, in the solutions to U316, Math. Reflections, 5 (2014), it is shown that 2m+2 |F32m for all
m 1. Also, for m 2, 144 = F(12,32m ) = (F12 , F32m ), so 3|F32m . Therefore n|Fn for all n = 3 2m with
m 2.
Also solved by Corneliu Mnescu-Avram, Transportation High School, Ploieti, Romania; AN-anduud
Problem Solving Group, Ulaanbaatar, Mongolia; ngel Plaza, Universidad de Las Palmas de Gran Canaria, Spain; Antoine Faisant; Arkady Alt, San Jose, CA, USA; Ercole Suppa, Teramo, Italy; G. C. Greubel,
Newport News, VA, USA; Adnan Ali, Student in A.E.C.S-4, Mumbai, India; Paul Revenant, Lyce Champollion, Grenoble, France; Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA,
Brazil.

Mathematical Reflections 4 (2015)

23

O347. Let a, b, c, d 0 be real numbers such that a + b + c + d = 1. Prove that

r

2(b c)2 2(c d)2 2(d b)2

+
+
+ b+ c+ d2
a+
9
9
9
Proposed by Marius Stanean, Zalau, Romania
Solution by the author
First we prove the following Lemma.
Lemma: Let x, y, z be non-negative real numbers. Prove that
X
X
2
(y 2 z 2 )2 3(x2 + y 2 + z 2 )
(y z)2 .
cyc

cyc

Proof: Since the inequality is symmetric and homogeneous, WLOG we assume that x + y + z = 3 and
for an easy computing, denote q = xy + yz + zx. Rewrite the inequality to the following forms
2(x4 + y 4 + z 4 x2 y 2 y 2 z 2 z 2 x2 ) 3(x2 + y 2 + z 2 )(x2 + y 2 + z 2 xy yz zx)
2[(x2 + y 2 + z 2 )2 3(x2 y 2 + y 2 z 2 + z 2 x2 )] 3(9 2q)(9 3q)
2[(9 2q)2 3(q 2 6xyz)] 3(9 2q)(9 3q)
36xyz 81 63q + 16q 2 .
Now, since 3q (x + y + z)2 = 9 q 3, there are a real number t [0, 1] such that q = 3(1 t2 ).
Therefore, the inequality becomes
4xyz 16t4 11t2 + 4.
(1)
Now we have xyz (1 t2 )(1 + 2t) = 1 3t2 + 2t3 Returning back to (1) we only need to prove that
4 12t2 + 8t3 16t4 11t2 + 4
t2 (16t2 8t + 1) 0 t2 (4t 1)2 0
1
which means
which is obviously true. Equality holds for t = 0 which means x = y = z or for t =
4
2x = 2y = z.
Let us now return to out inequality. According to Lemma we have
X
X

2
(b c)2 3(b + c + d)
( b c)2
cyc

cyc

but b + c + d 1, so

2
2X
1X
( b + c + d)2
2
(b c)
( b c) = b + c + d
.
9 cyc
3 cyc
3

(2)

Applying the Cauchy-Schwarz inequality we have

!2

2(b c)2 2(c d)2 2(d b)2

a+
+
+
+ b+ c+ d

9
9
9

(2)
2(b c)2 2(c d)2 2(d b)2 ( b + c + d)2
a+
+
+
+
(1 + 3)
9
9
9
3
r

1
4(a = b = c = d = )
4
Equality holds when a = b = c = d =

1
4

Also solved by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil
Mathematical Reflections 4 (2015)

24

O348. Let ABCDE be a convex pentagon with area S, and let R1 , R2 , R3 , R4 , R5 be the circumradii of
triangles ABC, BCD, CDE, DEA, EAB, respectively. Prove that
R14 + R24 + R34 + R44 + R54

4
2

5 sin 108

S2.

Proposed by Nairi Sedrakyan, Armenia

Solution by Robert Bosch, Archimedean Academy, Florida, USA and Jorge Erick Lpez, IMPA, Brazil
First, let us only consider the triangle ABC. Let O1 be its circumcenter, and M1 , M2 the midpoints of
AB and BC respectively. Denote the area of the quadrilateral O1 M1 BM2 by S1 . We have the following
inequality
BC R
S1
.
4
The proof is as follows,
AC R
,
4
AC
,
sin(2A) + sin(2C)
R
sin(2A) + sin(2C) 2 sin(B),

1 2
R (sin(2A) + sin(2C))
4

2 sin(A + C) cos(A C) 2 sin(B),

cos(A C) 1.
Clearly, S S1 + S2 + S3 + S4 + S5 , suppose without loss of generality that min{R1 , R2 , R3 , R4 , R5 } = R1
and move clockwise. Thus
2S R52 sin A + R12 sin B + R22 sin C + R32 sin D + R42 sin E.
Now, by Cauchy-Schwarz inequality it follows
!
5
X


4S 2
Ri4 sin2 A + sin2 B + sin2 C + sin2 D + sin2 E .
i=1

To finish let us prove that

sin2 A + sin2 B + sin2 C + sin2 D + sin2 E 5 sin2 (108 ).
We know that A + B + C + D + E = 540 . Assuming 0 < A B C D E < 180 we obtain if A = 108
then B = C = D = E = 108 . A < 108 since if A > 108 then A+B +C +D +E > 540 a contradiction. In
a similar way E > 108 . Notice that A + E < 270 since if A + E 270 then B + C + D 270 , so B 90
and hence A 90 , therefore E 180 , a contradiction. By the identity sin2 x sin2 y = sin(x y) sin(x + y)
we obtain
sin2 A + sin2 E < sin2 (108 ) + sin2 (A + E 108 ).
Let us see the proof. The difference sin2 (108 ) + sin2 (A + E 108 ) sin2 A sin2 E is positive due to
sin2 (108 ) sin2 A + sin2 (A + E 108 ) sin2 E,
= sin(108 A) sin(108 + A) + sin(A 108 ) sin(A + 2E 108 ),
= sin(A 108 ) (sin(A + 2E 108 ) sin(108 + A)) ,
= 2 cos(A + E) sin(A 108 ) sin(E 108 ) > 0,
since 90 < A + E < 270 . This means that by the replacement of A with 108 and of E with A + E 108
we increase the sum of the squared sinuses. Repeating this operation, we will make all the angles equal to
108 , and the inequality is proved.
Mathematical Reflections 4 (2015)

25