MR3 Solutions

Junior problems
J193. Let ABCD be a square of center O. The parallel through O to AD intersects AB and CD at M and
N and a parallel to AB intersects diagonal AC at P. Prove that
OP
4
+
MN
2
4
= MP
2
NP
2
.
Proposed by Titu Andreescu, University of Texas at Dallas, USA
First solution by Mihai Stoenescu, Bischwiller, France
The parallel to AB intersects MN at T. It is clear that AMO and OTP are right isosceles triangles.
So, we can denote AM = MO = x and OT = TP = t. Let MP = y, NP = z, and OP = u for
convenience. So we have TN = x t. So we must prove that u
4
+x
4
= y
2
z
2
. By Pythagoras, we have
y
2
z
2
= [(x +t)
2
+t
2
] [(x t)
2
+t
2
] = (x
2
+ 2t
2
+ 2xt) (x
2
+ 2t
2
2xt)
= (x
2
+ 2t
2
)
2
4x
2
t
2
= x
4
+ 4t
4
.
From triangle OTP we have 2t
2
= u
2
then 4t
4
= u
4
and we are done.
Second solution by Christopher Wiriawan, Jakarta, Indonesia
Denote the intersection between line MN and the parallel to AB as Q and let the perpendicular from
P to AB as R.
It is easy to see that QN = RB, thus by Pythagorean Theorem we have,
OP
4
+
MN
2
4
= (OQ
2
+QP
2
)
2
+
MN
2
2
=
MN
2
QN
2
+
MN
2
RB
2
2
+
MN
2
4
which is also equivalent to
MN
2
QN
2
+
MN
2
4
= 4
MN
2
QN
4
+
MN
2
4
Thus, we want to prove that this last expression equals the RHS. Now, by using Pythagorean Theorem
we have,
NP
2
= QN
2
+QP
2
= QN
2
+
MN
2
QN
2
= 2QN
2
MN QN +
MN
2
2
MP
2
= QP
2
+MQ
2
=
MN
2
QN
2
+ (MN QN)
2
= 5
MN
2
2
3MN QN + 2QN
2
Thus,
NP
2
MP
2
=
2QN
2
MN QN +
MN
2
MN
2
2
3MN QN + 2QN
2
which is equivalent to
4QN
4
16QN
3
MN
2
+ 24QN
2
MN
2
2
16QN
MN
2
3
+ 4
MN
2
4
+
MN
2
4
which is 4
MN
2
QN
4
+
MN
2
4
, thus proving the desired expression.
Also solved by Arber Selimi, Bedri Pejani - Peje, Kosovo; Daniel Lasaosa, Universidad P ublica de
Navarra, Spain; Ercole Suppa, Teramo, Italy; Gabriel Alexander Chicas Reyes, El Salvador; Miguel
Amengual Covas, Cala Figuera, Mallorca, Spain; Roberto Bosch Cabrera, Havana, Cuba; Kannappan
Sampath, Tiruchirappalli, India.
Mathematical Reflections 3 (2011) 1
J194. Let a, b, c be the side-lenghts of a triangle with the largest side c. Prove that
ab (2c +a +b)
(a +c) (b +c)

a +b +c
3
.
Proposed by Arkady Alt, San Jose, California, USA
First solution by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia
Without loss of generality assume that c a b. Consider the function
f(x) =
a +b +x
3
ab
1
a +x
+
1
b +x
.
We have
f(c) f(a) = (c a)
1
3
+
ab
2a(a +c)
+
1
(b +a)(b +c)
0 (1)
and
f(a) =
2a +b
3
ab(
1
2a
+
1
a +b
)
2a +b
3

b
2

ab
4
1
a
+
1
b
=
5
12
(a b) 0.
Note that, the given inequality holds for any positive a, b, c with the largest c.
Second solution by Hoang Quoc Viet, University of Auckland, New Zealand
Assume that c = max{a, b, c}, hence
a(c b) +b(c a) 0. (1)
The original inequality can be written as
3abc
a +b +c
c
2
+a(c b) +b(c a).
We have
3abc
a +b +c
c
c
2
+c(a +b)
a +b +c
= c
2
. (2)
Combining (1) and (2) the inequality is proved.
Navarra, Spain; Ercole Suppa, Teramo, Italy; Henry Ricardo New York, USA; Mihai Stoenescu, Bis-
chwiller, France; Perfetti Paolo, Dipartimento di Matematica, Università degli studi di Tor Vergata
Roma, Italy; Roberto Bosch Cabrera, Havana, Cuba; Christopher Wiriawan, Jakarta, Indonesia.
J195. Find all primes p and q such that both pq 555p and pq + 555q are perfect squares.
First solution by Ercole Suppa, Teramo, Italy
Since pq 555p = p(q 555) is a perfect square, p divides q 555 and q > 555. Therefore there exists
an integer a 1 such that
q 555 = ap (1)
Likewise q divides p + 555, so there exists an integer b 1 such that
p + 555 = bq (2)
From (1) and (2) it follows that
p + 555 = b(555 +ap)
(1 ab)p = 555(b 1) 0
1 ab 0 a = 1, b = 1
Therefore q p = 555, so p = 2 (otherwise q p would be an even number) and q = 557.
Second solution by Roberto Bosch Cabrera, Florida, USA
We have pq 555p = p(q 555) and pq + 555q = q(p + 555), hence p | q 555 and q | p + 555, so
pq | (q 555)(p + 555) = pq + 555q 555p 555
2
. Thus pq | 3 5 37(p + 555 q). But note that
p + 555 = qk hence p | 3 5 37 (k 1). If p = 3 3 | q 555 3 | q q = 3 contradiction since
q > 555. Analogously for p = 5 and p = 37. Thus we arrive to p | k 1, hence k = ph + 1. But
q 555 = pr and hence
p + 555 = q(ph + 1) p = pqh +pr 1 = qh +r h = 0, r = 1 k = 1
Hence q = p + 555, if p 3 we obtain that p + 555 is even, contradiction, so p = 2 and q = 557 is the
solution to our problem.
Navarra, Spain; Prasanna Ramakrishnan, Trinidad and Tobago; Christopher Wiriawan, Jakarta, In-
donesia.
J196. Let I be the incenter of triangle ABC and let A
, B
, C
be the feet of altitudes from vertices A, B, C.

If IA
= IB
= IC
, then prove that triangle ABC is equilateral.

Proposed by Dorin Andrica and Liana Topan, Babes-Bolyai University, Romania
Solution by Anthony Erb Lugo
We start by noting that IA
= IB
= IC
implies that I is the circumcenter of triangle A
whose
circumcircle is of course the nine-point circle. Thus, I is the center of the nine-point circle. Furthermore,
Feuerbachs Theorem tells us that the nine-point circle is internally tangent to the incircle. Since the
nine-point circle and the incircle have the same center and are internally tangent we can conclude that
these circles have the same radius. Moreover, if we let r be inradius and R be the circumradius of
triangle ABC then the condition that the radius of the nine-point circle,
R
2
, is the same as the radius
of the incircle tells us that
R
2
= r = R = 2r
which is the equality case of Eulers Triangle Inequality. But Eulers Triangle Inequality, R 2r, has
equality if and only if triangle ABC is equilateral and so were done.
Navarra, Spain; Ercole Suppa, Teramo, Italy; Mihai Stoenescu, Bischwiller, France.
J197. Let x, y, z be positive real numbers. Prove that
2 (x
2
y
2
+y
2
z
2
+z
2
x
2
)
1
x
3
+
1
y
3
+
1
z
3
1
y
+
1
z
+y
1
z
+
1
x
+z
1
x
+
1
y
.
Proposed by Vazgen Mikayelyan, Yerevan, Armenia
Solution by Arkady Alt, San Jose, California, USA
Note that by the Cauchy-Schwarz Inequality we have that
cyc
x
1
y
+
1
z

cyc
x
2
cyc
1
y
+
1
z
cyc
x
2
cyc
1
x
,
so it would suce to prove that
x
2
y
2
+y
2
z
2
+z
2
x
2
1
x
3
+
1
y
3
+
1
z
3
x
2
+y
2
+z
2
1
x
+
1
y
+
1
z
.
Now, let a :=
1
x
, b :=
1
y
, c :=
1
z
; then the inequality to be proven can be rewritten as
1
a
2
b
2
+
1
b
2
c
2
+
1
c
2
a
2
a
3
+b
3
+c
3
1
a
2
+
1
b
2
+
1
c
2
(a +b +c) ,
which is equivalent with
a
2
+b
2
+c
2

a
3
+b
3
+c
3
a
2
b
2
+b
2
c
2
+c
2
a
2
(a +b +c) ,
i.e.
cyc
a
5
+
cyc
a
3
b
2
+c
2
cyc
a
3
b
2
+c
2
cyc
ab
2
c
2
,
which turns out to be just the immediate
cyc
a
5
cyc
ab
2
c
2
= abc (ab +bc +ca) ,
which can be seen for example as a consequence of the AM-GM Inequality.
Also solved by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Anthony Erb Lugo; Arber
Selimi, Bedri Pejani - Peje, Kosovo; Daniel Lasaosa, Universidad P ublica de Navarra, Spain; Henry
Ricardo New York, USA; Perfetti Paolo, Dipartimento di Matematica, Università degli studi di Tor
Vergata Roma, Italy.
J198. Find all pairs (x, y) for which x! +y! + 3 is a perfect cube.
Solution by Ivan Borsenco, MIT, USA and Roberto Bosch Cabrera, Florida, USA
Let x! +y! + 3 = z
3
with x, y 7. We have that z
3
3 mod 7, so this gives a contradiction since the
cubic residues modulo 7 are only 0, 1, 6. Thus, we can suppose without loss of generality that x 6.
Now let us do the case work:
A) x = 1
y! + 4 = z
3
and y 7 z
3
4 mod 7; impossible, so y 6, but for any value of y we obtain a cube.
B) x = 2
y! + 5 = z
3
and y 7 z
3
5 mod 7; impossible, so y 6, and we obtain the solution y = 5.
C) x = 3
y! + 9 = z
3
and y 7 z
3
D) x = 4
y! + 27 = z
3
; we postpone this for now.
E) x = 5
y! + 123 = z
3
and y 7 z
3
F) x = 6
y! + 723 = z
3
and y 7 z
3
Now, lets turn back to D). First, note that the case y = 1, 2, 3, ..., 8 can be easily checked manually.
Fory > 8 we have 81 | y!; so z = 3z
and y!/27 = z
3
1. Now, 3 | z
3
1, so z
1 mod 3. But if
z
= 3k + 1; then
y!
27
= (3k + 1)
3
1 = 27k
3
+ 27k
2
+ 9k = 9k(3k
2
+ 3k + 1).
Thus y! = 243k(3k
2
+3k +1). Furthermore, let f(n) to be equal to the power of 3 dividing n and note
that f(n) log
3
n, so f(y!) = [243k(3k
2
+3k+1)/3]+[243k(3k
2
+3k+1)/9]+[243k(3k
2
+3k+1)/27]+. . .,
where [] denotes as usual the oor function.
But f(y!) = f(243k(3k
2
+ 3k + 1)) = 5 + f(k) 5 + log
3
k, so we can see that for k 1, [243k(3k
2
+
3k +1)/3] = 81k(3k
2
+k +1) > 5 +log
3
k and hence the equation has no solution. Finally, we get that
the solutions to original equation are (x, y) = (2, 5); (x, y) = (5, 2); (x, y) = (3, 6); (x, y) = (6, 3).
Also solved by Daniel Lasaosa, Universidad P ublica de Navarra, Spain; Ercole Suppa, Teramo, Italy;
Kannappan Sampath, Tiruchirappalli, India.
Senior problems
S193. Find all pairs (x, y) of positive integers such that x
2
+y
2
= p
6
+q
6
+ 1, for some primes p and q.
Solution by Carlo Pagano, Università di Roma Tor Vergata, Roma, Italy
Suppose that 3 = q and 3 = p, then 3 is coprime with both p and q (because they are prime). So 9
is coprime with both p and q. Then, being (9) = 6 we have that p
6
+ q
6
+ 1 3 (mod 9) by Euler
theorem. But then we should have x
2
+ y
2
0 (mod 3) which implies that x 0 (mod 3) and y 0
(mod 3). Hence x
2
+ y
2
0 (mod 9) and 3 0 (mod 9), contradiction. Then we should have that
without loss of generality p = 3. Now 3
6
+ q
6
+ 1 2 + q
6
(mod 4). So if q is odd, then x
2
+ y
2
3
(mod 4), contradiction. Then q = 2. So we have to solve x
2
+ y
2
= 794 = 2 (mod 397). But we have
that 397 1 (mod 4) and it is prime. So we get this factorization in primes of Z[i]:
794 = (1 +i)(1 i)(6 + 19i)(6 19i).
Therefore the unique solutions are (25, 13) and (13, 25).
Navarra, Spain; Prasanna Ramakrishnan, Trinidad and Tobago; Roberto Bosch Cabrera, Havana,
Cuba.
S194. Let p be a prime of the form 4k +3 and let n be a positive integer. Prove that for each integer m there
are integers a and b such that a
2
n
+b
2
n
m (mod p).
Proposed by Tigran Hakobyan, Yerevan, Armenia
Solution by Omran Kouba, Damascus, Syria
Let F
p
= Z/pZ, and let Q
p
= {x
2
: x F
p
} be the set quadratic residues modulo p. It is well-known
that |Q
p
| = 2k + 2 = (p + 1)/2. Also, note that the function : Q
p
Q
p
dened by (x) = x
2
is a
bijection. Indeed, suppose that (x) = (x
).
If x
0 mod p then p|x

2
, thus p|x and x x
( mod p).
If x
0 mod p, then (x x
)(x +x
) 0 mod p.
But x + x
0 mod p since, otherwise, we would have a quadratic residue x equal to a non-quadratic

residue x
, (here we used the fact that 1 is not a quadratic residue modulo p as p 3 mod 4,) and
since F
p
is an integral domain, we conclude that xx
0 mod p. So x x
mod p and thus it follows

that is injective. Furthermore, we get that : Q
p
Q
p
is also surjective since the set Q
p
is nite.
Now, in what is about to follow we dene =
1
, that is, the square root function dened on Q
p
.
Furthermore, consider an integer m, and the sets A = Q
p
and B = {(mx) mod p : x Q
p
}, which as
a matter of fact are two subsets of F
p
of cardinality |A| = |B| = 2k + 2 > p/2. Therefore, A and B can
not be disjoint and so there must be a common element AB. Hence, there exist (, ) Q
p
Q
p
such that m + mod p. But now we can set a =
n
and b =
n
, and obtain precisely that
m a
2
n
+b
2
n
mod p, which is the desired conclusion.
Also solved by Daniel Lasaosa, Universidad P ublica de Navarra, Spain; Carlo Pagano, Università di
Roma Tor Vergata, Roma, Italy.
S195. Let ABC be a triangle with incenter I and circumcenter O and let M be the midpoint of BC. The
bisector of angle A intersects lines BC and OM at L and Q, respectively. Prove that
AI LQ = IL IQ.
Proposed by Ivan Borsenco, Massachusetts Institute of Technology, USA
First solution by Gabriel Alexander Chicas Reyes, El Salvador
It is a well-known that the bisector of CAB and the perpendicular bisector OM of BC intersect on
the circumcircle of ABC. In other words, Q is the midpoint of the arc BC not containing A.
Observe that BQ = IQ, since by angle chasing QBI = QIB = /2 BCA/2 (in fact Q is the
circumcenter of BIC). On the other hand, since the triangles BLQ and ALC are similar we can write
BQ
LQ
=
AC
LC
. Now repeated use of the angle bisector theorem in the triangles ABC and ABL yields
IQ
LQ
=
BQ
LQ
=
AC
LC
=
AB
BL
=
AI
IL
.
Therefore AI LQ = IQ IL, as desired.
Second solution by Roberto Bosch Cabrera, Florida, USA
Let D the foot of perpendicular from I to BC. We need to prove that
AI
IL
=
IQ
LQ
=
IL
LQ
+ 1. We
have that
AI
IL
=
AB
BL
=
sin(A/2+C)
sin A/2
because I is incenter and by Sines Law. Besides
IL
LQ
=
DL
LM
=
BLBD
BMBL
=
BL(sb)
a/2BL
because triangles IDL and LQM are similar. Where a = BC, b = AC, c = AB
and s is the semiperimeter. We have that BL =
c sin A/2
sin(A/2+C)
=
2Rsin C sin A/2
sin(A/2+C)
and s b =
a+cb
2
=
R(sin A+ sin C sin B), where R is the circumradius. So we need to prove that
sin(A/2 +C)
sin A/2
=
2 sin C sin A/2
sin(A/2+C)
+ sin B sin Asin C
sin A
2 sin C sin A/2
sin(A/2+C)
+ 1
which after several transformations is equivalent to prove
sin B + sin C = 2 cos A/2 sin(A/2 +C)
but
sin B +sin C = 2 sin
B +C
2
cos
B C
2
= 2 cos A/2 cos(90
A/2 C) = 2 cos A/2 sin(A/2 +C).

Navarra, Spain; Ercole Suppa, Teramo, Italy; Leandro Remolina Ardila, Santander, Colombia; Omran
Kouba, Damascus, Syria.
S196. Find the least prime that can be written as
a
3
+b
3
2011
for some positive integers a and b.
Solution by Daniel Lasaosa, Universidad P ublica de Navarra, Spain
Since 2011 is prime (it is not divisible by any prime up to 43, and it is less than 47
2
= 2209), then
a
3
+ b
3
= (a + b)(a
2
ab + b
2
) is the product of two primes. Since a, b 1, clearly a + b > 1. If
a
2
ab +b
2
= 1, then ab (a b)
2
+ab = 1, or a = b = 1, and a
3
+b
3
= 2 is clearly not a multiple of
2011, hence either a + b is the lowest prime and a
2
ab + b
2
= 2011, or a + b = 2011 and a
2
ab + b
2
is the lowest prime. Note however that
a
2
ab +b
2
=
3(a b)
2
+ (a +b)
2
4

(a +b)
2
4
,
or if a + b = 2011, then a
2
ab + b
2
> 2011 500, while if a
2
ab + b
2
= 2011, then a + b
4 2011
and a +b < 90 because 2011 < 45
2
= 2025.
Let us therefore look for values a, b such that a + b = p is prime and a
2
ab + b
2
= 2011. Clearly,
3ab = (a + b)
2
(a
2
ab + b
2
) = p
2
2011, or p 47. Note
p
2
2011
3
is always an integer since
p
2
2011 1 (mod 3) for primes p = 3. Now, if p 3 (mod 50), then p
2
9 (mod 100), and
p
2
2011 2 (mod 100), or ab =
p
2
2011
3
66 (mod 100). Then (a b)
2
= p
2
4ab 9 64 45
(mod 100) cannot be a perfect square, or p = 47, 53, and p 59.
Note however that for (a, b) a permutation of (10, 49), we have
49
3
+ 10
3
2011
=
118649
2011
= 59,
and this is therefore the lowest prime of this form that can be found.
Also solved by Arber Selimi, Bedri Pejani - Peje, Kosovo; Prasanna Ramakrishnan, Trinidad and
Tobago; Roberto Bosch Cabrera, Havana, Cuba.
S197. Let (F
n
)
n0
be the Fibonacci sequence. Prove that for any prime p 3, p divides F
2p
F
p
.
Proposed by Dorin Andrica, Babes-Bolyai University, Cluj-Napoca, Romania
First solution by Daniel Lasaosa, Universidad P ublica de Navarra, Spain
The Fibonacci sequence for n 0 is 0, 1, 1, 2, 3, 5, . . . . Note that for p = 2, F
2p
F
p
= F
4
F
2
= 31 = 2
is a multiple of 2, and the result is clearly true in this case. Let now p be an odd prime. The general
form of the terms in the sequence is well known to be
F
n
=

+
n
5
, where
+
=
1 +
5
2
,
=
1
5
2
,
are the roots of the characteristic equation x
2
x 1 = 0. Dene now (G
n
)
n0
as G
0
= 2, G
1
= 1, and
for all n 0, G
n+2
= G
n+1
+ G
n
. It is easy to show (it is done in the same way as for the Fibonacci
sequence), that the general form of the terms in this sequence is G
n
=
+
n
+
n
, or F
2n
= F
n
G
n
.
It therefore suces to show that G
p
1 (mod p). Now, by Fermats little theorem, 2
p
2 2
p1
2
(mod p) because p is an odd prime, or
G
p
=
+
p
+
p1
2
k=0
p
2k
5
k
p
0
5
0
1 (mod p),
since for any k = 1, 2, . . . ,
p1
2
, both 2k, p 2k are positive integers less than p, and p does not appear
as a factor in (2k)! or (p 2k)!, but it does appear as a factor in p!. The conclusion follows.
Second solution by Roberto Bosch Cabrera, Havana, Cuba
We have the well-known identity F
m+n
= F
m1
F
n
+ F
m
F
n+1
, so F
2p
= F
p1
F
p
+ F
p
F
p+1
and hence
F
2p
F
p
= F
p
(F
p1
+F
p+1
1) = F
p
(L
p
1) where L
p
is the p th Lucas number. We will prove that
p | L
p
1.
L
p
=
1 +
5
2
p
+
5
2
p
=
p
k=0
p
k
5)
k
+
p
k=0
p
k
(1)
k
(
5)
k
2
p
=
p
k=0
p
k
5)
k
[1 + (1)
k
]
2
p
Hence
2
p
(L
p
1) =
p
2
5 2 +
p
4
5
2
2 + +
p
p 1
5
p1
2
2 (2
p
2)
But
p
2
, ,
p
p 1
are divisible by p and 2

p
2 is divisible by p by Fermats theorem, thus
p | 2
p
(L
p
1) and nally p | L
p
1.
Remark: There exists n composite such that n | L
n
1, this numbers are called Lucas pseudoprimes,
the rst few of these are 705, 2465, 2737, 3745, 4181, 5777, 6721, ...(Sloanes A005845).
S198. Let x, y, z be positive real numbers such that (x 2)(y 2)(z 2) xyz 2. Prove that
x
x
5
+y
3
+z
+
y
y
5
+z
3
+x
+
z
z
5
+x
3
+y
x +y +z
.
Solution by Hoang Quoc Viet, University of Auckland, New Zealand
I shall demonstrate the new method that I temporarily call Partition. To see how it works, we compare
the following evaluations (Holders inequality is applied for 3 variables in this case)
(x
5
+y
3
+z)
1
x
+ 1 +z
1
x
+ 1 +z
(x +y +z)
3
(3)
and
3(x
5
+y
3
+z)
1
x
2
+ 1 +z
2
(x +y +z)
3
(4)
Clearly, by comparing the above estimations, we can recognize that destination is unchanged (x+y+z)
3
except for the multiplying factors that really mattter. Which one is better? (3) or (4)
By comparing the multiplying factors, we have
3
1
x
2
+ 1 +z
2
1
x
+ 1 +z
2
which is a direct result of Cauchy- Schwarz inequality for 3 variables.
Hence, (3) is better.
Now, we may use (3) to prove the original inequality. (3) is equivalent to
cyc
x
x
5
+y
3
+z
cyc
1 +x +xz
(x +y +z)
x +y +z
x +y +z
3 +xy +yz +zx +x +y +z

x +y +z
From the given condition, it implies that

2(x +y +z) xy +yz +zx + 3
Thus, the above inequality is proven completely.
Also solved by Arber Selimi, Bedri Pejani - Peje, Kosovo.
Undergraduate problems
U193. Let n be a positive integer. Find the largest constant c
n
> 0 such that, for all positive real numbers
x
1
, . . . , x
n
,
1
x
2
1
+ +
1
x
2
n
+
1
(x
1
+ +x
n
)
2
c
n
1
x
1
+ +
1
x
n
+
1
x
1
+... +x
n
2
.
Proposed by Titu Andreescu, University of Texas at Dallas, USA and
Dorin Andrica, Babes-Bolyai University, Romania
First solution by N.J. Buitrago A., Universidad Nacional, Colombia
Let n be a positive integer. Find the largest constant c
n
> 0 such that, for all positive real numbers
x
1
, . . . , x
n
,
1
x
2
1
+ +
1
x
2
n
+
1
(x
1
+ +x
n
)
2
c
n
1
x
1
+ +
1
x
n
+
1
x
1
+ +x
n
2
.
Solution by N.J. Buitrago A., Universidad Nacional, Colombia.
Using the Arithmetic-Quadratic Mean Inequality to the n + 1 numbers
1
x
1
, . . . ,
1
xn
,
1
x
1
++xn
, we nd
1
x
2
1
+ +
1
x
2
n
+
1
(x
1
++xn)
2
n + 1

1
x
1
+ +
1
xn
+
1
x
1
++xn
n + 1
.
Therefore
1
x
2
1
+ +
1
x
2
n
+
1
(x
1
+ +x
n
)
2

1
n + 1
1
x
1
+ +
1
x
n
+
1
x
1
+ +x
n
2
.
Then is enough to take c
n
=
1
n+1
.
Second solution by Perfetti Paolo, Dipartimento di Matematica, Università degli studi di Tor Vergata
Roma, Italy
By the homogeneity of the inequality we set x
1
+ . . . + x
n
= 1. If x
k
= 1/n for any k we get c
n
=
(n
3
+ 1)/(n
2
+ 1)
2
thus the actual value of c
n
cannot be greater than c
n
.
Moreover powermeans-inequality yields if a
k
0 (a
2
1
+. . .+a
2
n
) (a
1
+. . .+a
n
)
2
/n. thus the inequality
is implied by (S = 1/x
1
+. . . + 1/x
n
)
S
2
(nc
n
1) + 2nc
n
S +nc
n
n 0, S n
2
The lower bound on S follows by
1 = x
1
+. . . +x
n
n
k=1
x
k
1/n
, S n
k=1
x
k
1/n
whence S n
2
. We observe that nc
n
< 1 since
c
n
c
n
=
n
3
+ 1
(n
2
+ 1)
2
<
1
n
and this follows by 2n
2
+ 1 > n. Thus the parabola P(S)
.
= S
2
(nc
n
1) + 2nc
n
S +nc
n
n is concave
and we have two cases.
First case. = nc
n
+ n
2
c
n
n 0. If c
n
1/(n + 1) the maximum has ordinate

4(ncn1)
0
evidently
Second case. = nc
n
+ n
2
c
n
n > 0 but the zeroes of P(S) are less than n
2
. In such a case
P(S) P(n
2
) 0 if and only if c
n
c
n
. Indeed the zeroes of P(S) are
S
1,2
nc
n
nc
n
+n
2
c
n
n
nc
n
1
and we must show that
nc
n
+
nc
n
+n
2
c
n
n
nc
n
+ 1
n
2
If c
n
1/(n + 1) but c
n
c
n
we get clearing the denominators and squaring
n
4
2n
5
c
n
2n
3
c
n
+n
6
c
2
n
+ 2n
4
c
2
n
+c
2
n
n
2
c
n
n n
2
c
n
+n =
n(c
n
(n
2
+ 1)
2
n
3
1)(c
n
n 1) 0
which clearly holds in the range of c
n
since c
n
c
n
< 1/n.
Also solved by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Arkady Alt, San Jose, USA;
Daniel Lasaosa, Universidad P ublica de Navarra, Spain; Roberto Bosch Cabrera, Florida, USA.
U194. Prove that the set of positive integers n for which n divides 2
n
2
+1
+ 3
n
has density 0.
Proposed by Gabriel Dospinescu, Ecole Polytechnique, France
Solution by Gabriel Dospinescu, Ecole Polytechnique, France
Let A
x
be the set of those positive integers n x such that n divides 2
n
2
+1
+ 3
n
. Let n A
x
and
observe that n is odd. Moreover, if p is a prime factor of n, then 2
n
2
+1
is a perfect square mod p. We
deduce that 3
n
is a square mod p and so 3 is a square mod p. Using the quadratic reciprocity law,
it follows that p = 1 (mod 3). Thus A
x
is included in the set of those n x all of whose prime factors
are 1 (mod 3). An inclusion-exclusion argument combined with the fact that
p=1 (mod 3)
1
p
=
p=2 (mod 3)
1
p
=
(as it follows from basic estimates used in the proof of Dirichlets theorem) easily shows that the last
set has density 0. The result follows.
U195. Given a positive integer n, let f(n) be the square of the number of its digits. For example f(2) = 1
and f(123) = 9. Show that
n=1
1
nf(n)
is convergent.
Proposed by Roberto Bosch Cabrera, Florida, USA
Solution by Jose Hernandez Santiago, Oaxaca, Mexico
Let k be a xed but arbitrary natural number. We know that, if m
k
is the number of digits in the
decimal expansion of k, then 10
m
k
1
k < 10
m
k
. It follows that
k
n=1
1
nf(n)
=
101
n=1
1
nf(n)
+
10
2
1
n=10
1
nf(n)
+. . . +
k
n=10
m
k
1
1
nf(n)
=
101
n=1
1
n 1
2
+
10
2
1
n=10
1
n 2
2
+. . . +
k
n=10
m
k
1
1
n m
2
k
101
n=1
1
1 1
2
+
10
2
1
n=10
1
10 2
2
+. . . +
k
n=10
m
k
1
1
10
m
k
1
m
2
k
<
10
1 1
2
+
10
2
10 2
2
+. . . +
10
m
k
10
m
k
1
m
2
k
= 10
1
1
2
+
1
2
2
+. . . +
1
m
2
k
10 (2)
and we are done.
Also solved by Daniel Lasaosa, Universidad P ublica de Navarra, Spain; Ercole Suppa, Teramo, Italy;
Henry Ricardo New York, USA; N.J. Buitrago A., Universidad Nacional, Colombia; Luigino Capone,
University of Rome Tor Vergata, Italy; Moubinool Omarjee, Paris , France; Perfetti Paolo, Diparti-
mento di Matematica, Università degli studi di Tor Vergata Roma, Italy; Emanuele Natale, Università
di Roma Tor Vergata, Roma, Italy.
U196. Let A, B M
2
(Z) be commuting matrices such that for any positive integer n there exists C M
2
(Z)
such that A
n
+B
n
= C
n
. Prove that A
2
= 0 or B
2
= 0 or AB = 0.
Solution by Gabriel Dospinescu, Ecole Polytechnique, France
We begin with the following preliminary result:
Lemma. Let z
1
, z
2
, ..., z
k
be complex numbers such that z
n
1
+z
n
2
+... +z
n
k
is the nth power of an integer
for any n. Then at most one of the numbers z
1
, z
2
, ..., z
k
is nonzero.
Proof. We may assume that all z
i
s are nonzero and we will prove that k = 1. If all z
i
are real numbers,
this can be proved easily with analytic tools, but the case of complex numbers seems to be more delicate.
First, since z
n
1
+ ... + z
n
k
is an integer for all n, the Newtons formulae show that all symmetric sums
in the z
i
s are rational numbers and so z
i
are all algebraic numbers. Pick a number eld K containing
all z
i
s and pick a prime ideal I of K lying above a prime p suciently large. We may ensure that I is
relatively prime to all z
i
s. Let n be the norm of I. Then x
n
= 1 (mod I) for any x O
K
prime to I.
We deduce that z
n
1
+ ... + z
n
k
= k (mod I). On the other hand, we know that z
n
1
+ ... + z
n
k
= a
n
= 1
(mod I) for some integer a, which is necessarily prime to I (otherwise p would divide k, but we may
choose p > k from the very beginning). We deduce that k = 1 (mod I) and then k = 1 (mod p). As p
was arbitrarily large, we obtain k = 1 and we are done.
Coming back to the solution of the problem, we recall that a standard argument shows that if A is
nonscalar and if B commutes with A, then B = aA+bI
2
for some numbers a, b. If both A, B are scalar,
the result is immediate from the lemma (we are even in a very easy case of that lemma). So, assume
that A is not scalar, let z
1
, z
2
be its eigenvalues. Pick a, b such that B = aA+bI
2
. Thus the eigenvalues
of B are az
1
+b and az
2
+b. By assumption det(A
n
+B
n
) is the nth power of an integer for all n. But
det(A
n
+B
n
) = (det A)
n
+ (det B)
n
+ (a det A+bz
1
)
n
+ (a det A+bz
2
)
n
.
Using the lemma, we deduce that at most one of the numbers det A, det B, a det A+bz
1
and a det A+bz
2
is nonzero. A small case analysis easily yields the result.
U197. Let n 2 be an integer. Find all continuous functions f : R R such that for all x
1
, x
2
, ..., x
n
R,
n
i=1
f(x
i
)
1i<jn
f(x
i
+x
j
) + + (1)
n1
f(x
1
+... +x
n
) = 0.
Solution by Daniel Lasaosa, Universidad P ublica de Navarra, Spain;
Since

n
1
n
2
+ + (1)
n1
n
n
n
0
(1 + (1))
n
= 1,
taking x
1
= = x
n
= 0 yields f(0) = 0. For n = 2, the problem becomes the Cauchy functional
equation f(x + y) = f(x) + f(y), whose continuous solutions are f(x) = kx for any real constant k.
We shall now show by induction that, for all n 2, f(x) is necessarily an n 1-degree polynomial in
x without a constant term; this is clearly true for the base n = 2.
For the step, dene rst g(x, y) = f(x+y)f(x)f(y), and x y. Taking x
n
= y, note that substitution
in the proposed equation yields
n1
i=1
g(x
i
, y)
1i<jn1
g(x
i
+x
j
, y) + + (1)
n2
g(x
1
+ +x
n1
, y) = 0,
ie, for a xed y, g(x, y) is a solution of the problem for n 1, hence by hypothesis of induction, an
n2-degree polynomial in x without a constant term, whose coecients will depend on the value of y.
Since g(x, y) must be invariant upon exchange of x, y, g(x, y) is also an n 2-degree polynomial in y
without a constant term, whose coecients will depend on the value of x. We conclude that g(x, y) is a
polynomial in x, y, where the degree of each variable is at most n2, and where all terms are multiples
of xy.
Dene h(x) = g(x, x) = f(2x) 2f(x), which is clearly a polynomial in x, whose least-degree term is
x
2
. Therefore,
lim
x0
h(x)
x
2
= lim
x0
f(2x) 2f(x) +f(0)
x
2
exists and is nite, ie f(x) is twice dierentiable at x = 0. Therefore, for all x and y 0, we nd
lim
y=0
f(x +y) f(x)
y
= lim
y0
f(y)
y
+ lim
y0
g(x, y)
y
,
which clearly exists and is nite for all real x, since the rst limit in the RHS is clearly f
(0), which
exists and is nite since f is twice dierentiable at 0, and the second limit is the coecient of y (which
is a function of x) in g(x, y). It follows that, since the coecients of each term in g(x, y) (considered as
a function of y with x as a parameter) are clearly n 2-degree polynomials in x, then f
(x) exists and

is an n 2-degree polynomial in x, hence f(x) is an n 1-degree polynomial in x, without a constant
term since f(0) = 0.
We now show that any n 1-degree polynomial in x without a constant term is indeed a solution of
the proposed equation. By linearity in the functional equation, note that it suces to show that x
k
is
a solution for all 1 k n 1. Now, if f(x) = x
k
, then
n
i=1
f(x
i
)
1i<jn
f(x
i
+x
j
) + + (1)
n1
f(x
1
+ +x
n
)
is the sum of terms of the form x
1
i
1
x
2
i
2
. . . x
p
ip
, where p n, and
1
,
2
, . . . ,
p
are positive integers with
sum k. For a specic set of indices i
1
, i
2
, . . . , i
p
, and specic values of
1
,
2
, . . . ,
p
, let us calculate
the coecient of this term. Clearly, this term will appear exactly in all terms of the form (1)
u1
f(s),
where s is the sum of u of the x
i
, such that x
i
1
, x
i
2
, . . . , x
ip
are part of these u x
i
. Clearly p u n,
and exactly
np
up
such terms will exist for each u (as many as possible forms to choose u p other x
i
out of the remaining n p). The coecient of x
1
i
1
x
2
i
2
. . . x
p
ip
in each term will clearly be
K =
k!
(
1
)!(
2
)! . . . (
p
)!
,
clearly independent on u. The total coecient will then be
n
u=p
(1)
u1
K
n p
u p
= (1)
p1
K
np
v=0
(1)
v
n p
v
= (1)
p1
K(1 + (1))
np
= 0,
where we have made the variable exchange v = u p. It follows that all terms have 0 coecient, and
indeed f(x) = x
k
satises the proposed equation for all 1 k n 1. We conclude that, for each n,
f(x) is a solution i it is an n 1-degree polynomial with no constant term.
Also solved by Omran Kouba, Damascus, Syria.
U198. Dene a sequence (x
n
)
n
by x
0
= 1 and x
n+1
= 1 +x
n
+
1
xn
for n 0. Prove that there is a real number
a such that
lim
n
n
log n
(a +n + log n x
n
) = 1.
Solution by Perfetti Paolo, Dipartimento di Matematica, Università degli studi di Tor Vergata Roma,
Italy
We will prove the assertion
x
n
= n + ln n + 1
ln n
n
+O(
ln(n + 1)
n
2
)
which yields the desired limit. To prove the assertion we proceed by induction For n = 1 clearly holds.
Lets suppose it holds for 1 n r. For n = r + 1 we need to show that (q
n+1
.
= O(
ln(n+2)
(n+1)
2
))
x
r+1
= r + 1 + ln(r + 1) + 1
ln(r + 1)
r + 1
+q
r+1
=
1 +
r + ln r
ln r
r
+ 1 +q
r
+
1
r + ln r + 1
ln r
r
+q
r
= 1 +x
r
+
1
x
r
whence
ln(r + 1) ln r =
ln(r + 1)
r + 1

ln r
r
+q
r
q
r+1
+
1
r + ln r + 1
ln r
r
+q
r
(1)
By employing ln(r + 1) = ln r + ln(1 +
1
r
) = ln r +
1
r

1
2r
2
+o(
1
r
2
) we get
ln(r + 1)
r + 1

ln r
r
=
1
r
2

ln r
r
2
+o(
1
r
2
)
and
1
r + ln r + 1
ln r
r
+q
r
=
1
r

ln r
r
2

1
r
2
+o(
1
r
2
)
Inserting in (1) we have
1
r

1
2r
2
+o(
1
r
2
) =
1
r
2

ln r
r
2
+o(
1
r
2
)
+q
r
q
r+1
+
1
r

ln r
r
2

1
r
2
+o(
1
r
2
)
that is
q
r
= q
r+1
+
2 ln r
r
2

1
2r
2
+o(
1
r
2
)
and the right hand side is O(
ln(r+1)
r
2
).
Also solved by Daniel Lasaosa, Universidad P ublica de Navarra, Spain; Moubinool Omarjee, Paris ,
France; Omran Kouba, Damascus, Syria; Adrian Neacsu, Pitesti, Romania.
Olympiad problems
O193. Let a, b, c be positive real numbers. Prove that
1
a +b +
1
abc
+ 1
+
1
b +c +
1
abc
+ 1
+
1
c +a +
1
abc
+ 1
a +b +c
a +b +c + 1
.
Denote d =
1
abc
, a +b +c +d + 1 = s, or the inequality is equivalent to
1
s a
+
1
s b
+
1
s c
+
1
s d
1,
where abcd = 1, a +b +c +d = s 1. After some algebra, the proposed inequality is equivalent to
s(s 2) 3s
2
+ (abc +bcd +cda +dab)(s 1) 1,
where we have dened = ab+bc+cd+da+ac+bd, q = a
2
+b
2
+c
2
+d
2
, and clearly (s1)
2
= q +2.
Now,
2
(a
2
b
2
+b
2
c
2
+c
2
d
2
+d
2
a
2
+a
2
c
2
+b
2
d
2
)
2
= (s 1)(abc +bcd +cda +dab) 1,
or the problem is equivalent to showing that
2(s 1)
2
+ (a
2
b
2
+b
2
c
2
+c
2
d
2
+d
2
a
2
+a
2
c
2
+b
2
d
2
) 6(s 1)
2
+
2
+ 2 + 12(s 1) + 6.
Now, by the inequality between arithmetic and quadratic means,
a
2
b
2
+b
2
c
2
+c
2
d
2
+d
2
a
2
+a
2
c
2
+b
2
d
2

2
6
, 3(s 1)
2
8,
whereas by the AM-GM inequality, s1 4 and 6, with equality in all cases i a = b = c = d = 1.
Therefore,
2(s 1)
2
6(s 1)
2
(s 1)
2
5
16

8
3
+

8
4
2
+
s 1
2
6 4 +
1
16
6 4
2

2
+ 2 + 12(s 1) + 6 (a
2
b
2
+b
2
c
2
+c
2
d
2
+d
2
a
2
+a
2
c
2
+b
2
d
2
).
The conclusion follows, equality holds i a = b = c = 1.
O194. Let A be a set of nonnegative integers, containing 0 and let a
n
be the number of solutions of the
equation x
1
+x
2
+ +x
n
= n, with x
1
, . . . , x
n
A, a
0
= 1. Find A, if for all n 0,
n
k=0
a
k
a
nk
=
3
n+1
+ (1)
n
4
.
Solution by G.R.A.20 Problem Solving Group, Roma, Italy
The number of solutions of the equation x
1
+x
2
+ +x
n
= n with x
1
, . . . , x
n
0, 1, 2 is given by
a
n
=
n
k=0
n
k
n k
k
because there are
n
k
ways to assign k times the value 0 and
nk
k
ways to assign k times the value 2

to the remaining n k components. Moreover
a
n
=
n
k=0
n
k
n k
k
=
n
k=0
2k
k
n
2k
= [x
n
]
1
1 x

1
1
4x
2
1x
2
= [x
n
]
1
1 2x 3x
2
and
n
k=0
a
k
a
nk
= [x
n
]
1
(
1 2x 3x
2
)
2
= [x
n
]
1
1 2x 3x
2
=
3
n+1
+ (1)
n
4
.
Also solved by Daniel Lasaosa, Universidad P ublica de Navarra, Spain; Emanuele Tron, Marie Curie
High School, Pinerolo, Italy; Omran Kouba, Damascus, Syria.
O195. Let O, I, H be the circumcenter, incenter, and orthocenter of a triangle ABC, and let D be an interior
point to triangle ABC such that BC DA = CA DB = AB DC. Prove that A, B, D, O, I, H are
concyclic if and only if C = 60
.
Proposed by Titu Andreescu, Dorin Andrica, and Catalin Barbu
Claim: Denote by U, V, W the projections of D onto BC, CA, AB. Then, UV W is equilateral, and
ADB = C + 60
.
Proof: Since AWD = AV D = 90
, AV DW is cyclic with diameter DA, or V W = ADsin A =

BCAD
2R
, and by cyclic permutation of A, B, C, this quantity equals UV, WU. Moreover, ADW =
AV W = 180
AAWV , and
ADB = ADW +BDW = 360
AB AWV BWU = C + 60
.
Incidentally, D is called the rst isodynamic point, and it is inside ABC i no angle of ABC exceeds
120
.
It is well known that AIB = 90
+
1
2
C and AHB = 180
C; if ABC is obtuse at C, then O, C

are on opposite sides of AB and AOB = 360
2C, while if ABC is not obtuse at C, then O, C

are on the same side of AB and AOB = 2C. Note therefore that A, B, O, I can only be concyclic i
ABC is acute at C, since otherwise we would need AOB+AIB = 180
, or equivalently 270
=
3
2
C,
ie C = 180
, absurd since ABC would be degenerate and O, I could not be dened. We may thus
assume that ABC is acute at C.
But if ABC is acute at C, then A, B, and any two of D, O, I, H are concyclic i the corresponding pair
from the following four angles are equal:
AIB = 90
+
1
2
C, AHB = 180
C,
AOB = 2C, ADB = C + 60
,
ie i C = 60
. The conclusion follows.

Second solution by Henry Ricardo New York, USA
First of all, observe that D lies on the A-Apollonian circle of ABC since
DB
DC
=
AB
AC
. Likewise D lies
on the B-Apollonian circle. Hence D is the rst point isodynamic, being interior to ABC.
Let LMN be the pedal triangle of D and let R be the circumradius of ABC (see gure).
A B
C
D
N
L
M
From cyclic quadrilaterals ANDM, BLDN we have
MN = DA sin A, NL = DB sin B
MN
NL
=
DA
DB

sin A
sin B
=
CA
CB

sin A
sin B
=
2Rsin B
2Rsin A

sin A
sin B
= 1
Therefore MN = NL and similarly we obtain NL = LM, so LMN is an equilateral triangle. Thus
we have
ADB = 360
MDALDM BDL =
= 360
MNA(180
C) BNL =
= 180
+C (180
MNL) =
= 180
+C 120
= 60
+C
Now it is easy to prove the result.
If A, B, D, O, I, H are concyclic then
AIB = ADB 90
+
C
2
= 60
+C C = 60
Conversely if C = 60
we have
ADB = 60
+C = 120
AOB = 2C = 120
AIB = 90
+
C
2
= 120
AHB = A+B = 180
C = 120
so A, B, D, O, I, H are concyclic, and we are done.

O196. Let ABC be a triangle such that ABC > ACB and let P be an exterior point in its plane such that
PB
PC
=
AB
AC
Prove that
ACB +APB +APC = ABC.
Proposed by Mircea Becheanu, Bucharest, Romania
Note that the relation
PB
PC
=
AB
AC
clearly denes an Apollonius circle with center on line BC, passes
through A and through the point D where the internal bisector of angle A intersects BC, leaving B
inside and C outside because ABC > ACB. Note that the powers of B, C with respect to
are respectively p
B
, p
C
, such that
p
B
p
C
=
BD BD
CD CD
=
BA
2
CA
2
=
c
2
b
2
,
where D
is the point diametrally opposite D in . Let now T, U be the second points where PB, PC
meet (the rst one being clearly P in both cases). Note therefore that
CT
CU
=
b BT
c CU
=
b p
B
c CU PB
=
b
2
p
B
c
2
CU PC
=
b
2
p
B
c
2
p
C
= 1,
or CT = CU, and similarly BT = BU, ie BC is the perpendicular bisector of TU, which are therefore
symmetric with respect to BC. Therefore, if P is on the same half plane as A,
APB = APT = 180
ADT = 180
ADB BDT =
= 180
ADB BDU = 180
2ADB ADU = 180
2ADB APU =
= 180
2ADB APC,
and similarly we nd the same result if P is on the opposite half plane. In either case, we nd
APB +APC = 180
2ADB = 180
180
B
A
2
= 2B +A180
= B C.
The conclusion follows.
O197. Let x, y, z be integers such that 3xyz is a perfect cube. Prove that (x +y +z)
3
is a sum of four cubes
of nonzero integers.
Proposed by Dorin Andrica, Babes-Bolyai University, Cluj-Napoca, Romania
If xyz, x +y z, y +z x, z +x y = 0, note that
(x +y +z)
3
= (x +y z)
3
+ (y +z x)
3
+ (z +x y)
3
+ 24xyz,
where the rst three terms in the RHS are clearly nonzero cubes, and 24xyz = 2
3
(3xyz) is also a
nonzero cube. This is however not a solution when at least one of xyz, x +y z, y +z x, z +x y is
zero. In that case, we can use that 7
3
= 343 = 216+125+1+1 = 6
3
+5
3
+1
3
+1
3
, or if x+y+z = s = 0,
then
(x +y +z)
3
= s
3
= (7s)
3
+ (6s)
3
+ (5s)
3
+ (s)
3
is the sum of four nonzero cubes. Finally, if x +y +z = 0. Then for any a, b = 0,
(x +y +z)
3
= 0 = a
3
+b
3
+ (a)
3
+ (b)
3
is the sum of four nonzero cubes. Note therefore that for any three integers x, y, z, (x +y +z)
3
may be
written as the sum of four nonzero cubes, regardless of whether 3xyz is a perfect cube or not.
Second solution by Roberto Bosch Cabrera, Florida, USA.
Let 3xyz = w
3
. We have the identity
(x +y +z)
3
= (x +y z)
3
+ (y +z x)
3
+ (z +x y)
3
+ (2w)
3
which can be proved expanding the cubes in both sides. Now if x+yz = 0 then (x+y+z)
3
= 8z
3
but
1 = (1)
3
+ (7)
3
+ (5)
3
+ (6)
3
and hence 8z
3
= (2z)
3
+ (14z)
3
+ (10z)
3
+ (12z)
3
, analogously
if y +z x = 0 or z +x y = 0.
O198. Let a, b, c be positive real numbers such that
(a
2
+ 1)(b
2
+ 1)(c
2
+ 1)
1
a
2
b
2
c
2
+ 1
= 2011.
Find the greatest possible value of max(a(b +c), b(c +a), c(a +b)).
Proposed by Titu Andreescu, University of Texas at Dallas, USA and
Gabriel Dospinescu, Ecole Polytechnique, France
Assume that when a(b + c) reaches its maximum possible value (which by symmetry in the variables
will also be the maximum that we are looking for), the values of p = abc and s = b + c are known. It
remains thus to nd the maximum value of a under these two constraints; it would seem at rst sight
that bc would need to be minimized, since p = abc is known, and a needs to be maximized; let us see
that it is otherwise. Note rst that
(b
2
+ 1)(c
2
+ 1) = s
2
+b
2
c
2
2bc + 1 = s
2
+ (bc 1)
2
,
a
2
+ 1 =
2011
s
2
+ (bc 1)
2

p
2
p
2
+ 1
.
Thus the maximum of the RHS, hence of a
2
and of a, is obtained when bc = 1, since p, s are known
and xed. We may thus perform substitution bc = 1, and the relation given in the problem statement
becomes
(a
2
+ 1)
2
s
2
= 2011a
2
.
Under this constraint, we need to maximize as, or equivalently, a
2
s
2
= x. Now, the previous relation
between a and s may be rewritten as
x
2
(2011 2s
2
)x +s
4
= 0,
wherefrom the largest of both (clearly positive real) roots is
x =
2011 2s
2
+
(2011 2s
2
)
2
4s
4
2
=
2011 2s
2
+
2011
2011 4s
2
2
.
It nally follows that, since bc = 1, and s needs to be minimized so that x = a
2
s
2
is maximized, and
s
2
4bc = 4 by the AM-GM inequality, then the maximum is achieved when b = c = 1, and the square
of this maximum is
x =
2003 +
2011
1995
2
,
and nally,
max {a(b +c), b(c +a), c(a +b)} =
2003 +
2011
1995
2
=
2011 +
1995
2
with equality i two of a, b, c are 1, and the other one is
2011+
1995
4
. These values can be easily shown
to satisfy the relation given in the problem statement.

MR3 Solutions

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

MR3 Solutions

Загружено:

Авторское право:

Доступные форматы

Junior problems

be the feet of altitudes from vertices A, B, C.

, then prove that triangle ABC is equilateral.

implies that I is the circumcenter of triangle A

0 mod p then p|x

0 mod p since, otherwise, we would have a quadratic residue x equal to a non-quadratic

mod p and thus it follows

= 2 cos A/2 cos(90

A/2 C) = 2 cos A/2 sin(A/2 +C).

are divisible by p and 2

3 +xy +yz +zx +x +y +z

From the given condition, it implies that

(x) exists and

because there are

ways to assign k times the value 0 and

ways to assign k times the value 2

, AV DW is cyclic with diameter DA, or V W = ADsin A =

C; if ABC is obtuse at C, then O, C

2C, while if ABC is not obtuse at C, then O, C

. The conclusion follows.

AHB = A+B = 180

so A, B, D, O, I, H are concyclic, and we are done.

ADB BDU = 180

2ADB ADU = 180

Вам также может понравиться