MR Year 2014 Problem Column PDF

Junior problems
J289. Let a be a real number such that 0 ≤ a < 1. Prove that

1 1
a 1+ +1= .
1−a 1−a
Proposed by Arkady Alt, San Jose, California, USA
J290. Let a, b, c be nonnegative real numbers such that a + b + c = 1. Prove that

p3
p
3
p
3
13a3 + 14b3 + 13b3 + 14c3 + 13c3 + 14a3 ≥ 3.
Proposed by Titu Andreescu, University of Texas at Dallas, USA
J291. Let ABC be a triangle such that ∠BCA = 2∠ABC and let P be a point in its interior
such that P A = AC and P B = P C. Evaluate the ratio of areas of triangles P AB and
P AC.
Proposed by Panagiote Ligouras, Noci, Italy
J292. Find the least real number k such that for every positive real numbers x, y, z, the following
inequality holds: Y
(2xy + yz + zx) ≤ k(x + y + z)6 .
cyc
Proposed by Dorin Andrica, Babes-Bolyai University, Romania
J293. Find all positive integers x, y, z such that
(x + y 2 + z 2 )2 − 8xyz = 1.
Proposed by Aaron Doman, University of California, Berkeley , USA
1 ≤ (a2 − a + 1)(b2 − b + 1)(c2 − c + 1) ≤ 7.
Proposed by An Zhen-ping, Xianyang Normal University, China
Mathematical Reflections 1 (2014) 1

Senior problems
S289. Let x, y, z be positive real numbers such that x ≤ 4, y ≤ 9 and x + y + z = 49. Prove
that
1 1 1
√ + √ + √ ≥ 1.
x y z
Proposed by Marius Stanean, Zalau, Romania
S290. Prove that there is no integer n for which

2
1 1 1 4
+ + ··· + 2 = .
22 32 n 5
Proposed by Ivan Borsenco, Massachusetts Institute of Technology, USA
S291. Let a, b, c be nonnegative real numbers such that ab + bc + ca = 3. Prove that

5
(2a2 − 3ab + 2b2 )(2b2 − 3bc + 2c2 )(2c2 − 3ca + 2a2 ) ≥ (a2 + b2 + c2 ) − 4.
3
Proposed by Titu Andreescu, USA and Marius Stanean, Romania
S292. Given triangle ABC, prove that there exists X on the side BC such that the inradii of
triangles AXB and AXC are equal and find a ruler and compass construction.
Proposed by Cosmin Pohoata, Princeton University, USA
S293. Let a, b, c be distinct real numbers and let n be a positive integer. Find all nonzero
complex numbers z such that
c a b
az n + bz + = bz n + cz + = cz n + az + .
z z z
S294. Let s(n) be the sum of digits of n2 + 1. Define the sequence (an )n≥0 by an+1 = s(an ),
with a0 an arbitrary positive integer. Prove that there is n0 such that an+3 = an for all
n ≥ n0 .
Proposed by Roberto Bosch Cabrera, Havana, Cuba

Undergraduate problems
1
U289. Let a ≥ 1 be such that (ban c) n ∈ Z for all sufficiently large integers n. Prove that a ∈ Z.
Proposed by Mihai Piticari, Campulung Moldovenesc, Romania
U290. Prove that there are infinitely many consecutive triples of primes (pn−1 , pn , pn+1 ) such
that 12 (pn+1 + pn−1 ) ≤ pn .
U291. Let f : R → R be a bounded function and let S be the set of all increasing maps
ϕ : R → R. Prove that there is a unique function g in S satisfying the conditions:
a) f (x) ≤ g(x), for all x ∈ R.
b) If h ∈ S and f (x) ≤ h(x) for all x ∈ R, then g(x) ≤ h(x) for all x ∈ R.
Proposed by Marius Cavachi, Constanta, Romania
U292. Let r be a positive real number. Evaluate

Z π/2
1
dx.
0 1 + cotr x
Proposed by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain.
U293. Let f : (0, ∞) → R be a bounded continuous function P and let α ∈ [0, 1). Suppose there
exist real numbers a0 , . . . , ak , with k ≥ 2, so that kp=0 ap = 0 and

k
X
α

lim x ap f (x + p) = α.
x→∞ p=0
Prove that α = 0.
Proposed by Marcel Chirita, Bucharest, Romania
U294. Let p1 , p2 , . . . , pn be pairwise distinct prime numbers. Prove that

√ √ √ √ √ √
Q( p1 , p2 , . . . , pn ) = Q( p1 + p2 + · · · + pn ).

Olympiad problems
O289. Let a, b, x, y be positive real numbers such that x2 − x + 1 = a2 , y 2 + y + 1 = b2 , and

(2x − 1)(2y + 1) = 2ab + 3. Prove that x + y = ab.
O290. Let Ω1 and Ω2 be the two circles in the plane of triangle ABC. Let α1 , α2 be the circles
through A that are tangent to both Ω1 and Ω2 . Similarly, define β1 , β2 for B and γ1 , γ2
for C. Let A1 be the second intersection of circles α1 and α2 . Similarly, define B1 and
C1 . Prove that the lines AA1 , BB1 , CC1 are concurrent.
O291. Let a, b, c be positive real numbers. Prove that
a2 b2 c2 a+b+c
√ +√ +√ ≥ .
2
4a + ab + 4b2 2
4b + bc + 4c2 2
4c + ca + 4a2 3
O292. For each positive integer n let

n
X 1
Tn = .
k · 2k
k=1
Find all prime numbers p for which

p−2
X Tk
≡0 (mod p).
k+1
k=1
Proposed by Gabriel Dospinescu, Ecole Normale Superieure, Lyon
xyz
O293. Let x, y, z be positive real numbers and let t2 = max(x,y,z) . Prove that
4(x3 + y 3 + z 3 + xyz)2 ≥ (x2 + y 2 + z 2 + t2 )3 .
Proposed by Nairi Sedrakyan, Yerevan, Armenia
O294. Let ABC be a triangle with orthocenter H and let D, E, F be the feet of the altitudes
from A, B and C. Let X, Y , Z be the reflections of D, E, F across EF , F D, and
DE, respecitvely. Prove that the circumcircles of triangles HAX, HBY , HCZ share a
common point, other than H.

Junior problems
J295. Let a, b, c be positive integers such that (a − b)2 + (b − c)2 + (c − a)2 = 6abc. Prove that
a3 + b3 + c3 + 1 is not divisible by a + b + c + 1.
Proposed by Mihaly Bencze, Brasov, Romania
J296. Several positive integers are written on a board. At each step, we can pick any two
numbers u and v, where u ≥ v, and replace them with u + v and u − v. Prove that after
a finite number of steps we can never obtain the initial set of numbers.
J297. Let a, b, c be digits in base x ≥ 4. Prove that
ab bc ca
+ + ≥ 3,
ba cb ac
where all numbers are written in base x.
Proposed by Titu Zvonaru, Comanesti and Neculai Stanciu, Buzau, Romania
J298. Consider a right angle ∠BAC and circles ω1 , ω2 , ω3 , ω4 passing through A. The centers
of circles ω1 and ω2 lie on ray AB and the centers of circles ω3 and ω4 lie on ray AC.
Prove that the four points of intersection, other than A, of the four circles are concyclic.
J299. Prove that no matter how we choose n numbers from the set {1, 2, . . . , 2n}, one of them
will be a square-free integer.
J300. Let a, b, c be positive real numbers. Prove that

b+c c+a a+b
√ +√ +√ ≥ 1.
2a2 + 16ab + 7b2 + c 2b2 + 16bc + 7c2 + b 2c2 + 16ca + 7a2 + b
Mathematical Reflections 2 (2014)

Senior problems
S295. Let a, b, c be positive real numbers such that a + b + c = 3. Prove that

√
X (a + b)2
√ ≤ 12.
cyc a2 − ab + b2
S296. A ball in Vienna is attended by n ladies (some of which are wearing red dresses) and m
gentlemen. Some ladies and some gentlemen are acquainted. Dancing floor is occupied
by acquainted mixed pairs. At some point during the night, all the present gentlemen
were seen on the dancing floor. At some other time, all the ladies wearing red dresses
were on the dancing floor. Show that at some point there could be all gentlemen and all
red-dressed ladies on the dancing floor.
Proposed by Michal Rolinek, Institute of Science and Technology, Vienna
S297. Let ABC be a triangle and let A1 , A2 , B1 , B2 , C1 , C2 be points that trisect segments BC,
CA, AB, respectively. Cevians AA1 , AA2 , BB1 , BB2 , CC1 , CC2 intersect each other at
the vertices of a convex hexagon that does not have any intersection points inside it.
Prove that if the hexagon is cyclic then our triangle is equilateral.
S298. Prove the following identity

a
X
a−k a b+k b
(−1) = .
k c c−a
k=0
Proposed by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain
S299. Let ABCD be a trapezoid with AB k CD and let P an arbitrary point in its plane. If
{E} = P D ∩ AC, {F } = P C ∩ BD, {M } = P D ∩ AF and {N } = P C ∩ BE, prove that
M N k AB.
Proposed by Mihai Miculita and Marius Stanean, Romania
S300. Let x, y, z be positive numbers and a, b > 0 such that a + b = 1. Prove that
(x + y)3 (y + z)3 ≥ 64abxy 2 z (ax + y + bz)2 .

U295. Let a be a real number such that (bnac)n≥1 is an arithmetic sequence. Prove that a is
an integer.
U296. Let a and b be real nonzero numbers and let z0 ∈ C\R be a root √to the equation
z n+1 + az + nb = 0, where n is a positive integer. Prove that |z0 | ≥ n+1 b.

q
an−1
U297. Let a0 = 0, a1 = 2, and an+1 = 2− an for n ≥ 0. Find limn→∞ 2n an .
U298. Determine all pairs (m, n) of positive integers such that the polynomial
f = (X + Y )2 (mXY + n) + 1
is irreducible in Z[X, Y ].
U299. Let ABC be a triangle with incircle ω and let A0 , B0 , C0 be points outside ω. Tangents
from A0 to ω intersect BC at A1 and A2 . Points B1 , B2 and C1 , C2 are defined similarly.
Prove that A1 , A2 , B1 , B2 , C1 , C2 lie on a conic if and only if triangle ABC and A0 B0 C0
are perspective.
Proposed by Luis Gonzalez, Maracaibo, Venezuela
U300. Let f : [a, b] → [a, b] be a function having lateral limits in every point. If
lim f (t) ≤ lim f (t)

t→x− t→x+
for all x ∈ [a, b], prove that there is an x0 ∈ [a, b] such that limt→x0 f (t) = x0 .
Proposed by Dan Marinescu and Mihai Piticari, Romania

Olympiad problems
O295. Let a, b, c, x, y, z be positive real numbers such that x + y + z = 1 and
2ab + 2bc + 2ca > a2 + b2 + c2 .
Prove that
2
a (x + 3yz) + b(y + 3xz) + c (z + 3xy) ≤ (a + b + c).
3
O296. Let m be a positive integer. Prove that φ(n) divides mn, only for finitely many square-
free integers n, where φ is Euler’s totient function.
O297. Cells of an 11 × 11 square are colored in n colors. It is known that the number of cells
of each color is greater than 6 and less than 14. Prove that one can find a row and a
column whose cells are colored in at least four different colors.
O298. Let n be a square-free positive integer. Find the number of functions f : {1, 2, . . . , n} →
{1, 2, . . . , n} such that f (1)f (2) · · · f (n) divides n.
O299. Let a, b, c be positive real integers such that a2 + b2 + c2 + abc = 4.

√
1 − abc(3 − a − b − c) ≥ |(a − 1)(b − 1)(c − 1)|.
O300. Let ABC be a triangle with circumcircle Γ and incircle ω. Let D, E, F be the tangency
points of ω with BC, CA, AB, respectively, let Q be the second intersection of AD
with Γ, and let the T be the intersection of the tangents at B and C with respect to Γ.
Furthermore, let QT intersect Γ for the second time at R. Prove that AR, EF , BC are
concurrent.
Proposed by Faraz Masroor, Gulliver Preparatory, Florida, USA

Junior problems
1 1
J301. Let a and b be nonzero real numbers such that ab ≥ a + b + 3. Prove that
3
1 1
ab ≥ √
3
+√
3
.
a b
J302. Given that the real numbers x, y, z satisfy x + y + z = 0 and
x4 y4 z4
+ + = 1,
2x2 + yz 2y 2 + zx 2z 2 + xy
determine, with proof, all possible values of x4 + y 4 + z 4 .
Proposed by Razvan Gelca, Texas Tech University, USA
J303. Let ABC be an equilateral triangle. Consider a diameter XY of the circle centered at C
which passes through A and B such that lines AB and XY as well as lines AX and BY
meet outside this circle. Let Z be the point of intersection of AX and BY . Prove that
AX · XZ + BY · Y Z + 2CZ 2 = XZ · Y Z + 6AB 2 .
J304. Let √
a, b, c be real numbers such that a + b + c = 1. LetpM1 be the maximum value of
√ √
a + b + 3 c and let M2 be the maximum value of a + b + 3 c. Prove that M1 = M2
and find this value.
Proposed by Aaron Doman, University of California, Berkeley, USA
J305. Consider a triangle ABC with ∠ABC = 30◦ . Suppose the length of the angle bisector
from vertex B is twice the length of the angle bisector from vertex A. Find the measure
of ∠BAC.
Proposed by Mircea Lascu and Marius Stanean, Zalau, Romania
J306. Let S be a nonempty set of positive real numbers such that for any a, b, c in S, the
number ab + bc + ca is rational. Prove that for any a and b in S, ab is a rational number.
Proposed by Bogdan Enescu, Buzau, Romania

Senior problems
S301. Let a, b, c be positive real numbers. Prove that
(a + b + c)(ab + bc + ca)(a3 + b3 + c3 ) ≤ (a2 + b2 + c2 )3 .

√ √ √
S302. If triangle ABC has sidelengths a, b, c and triangle A0 B 0 C 0 has sidelengths a, b, c,
prove that
A B C
sin sin sin = cos A0 cos B 0 cos C 0 .
2 2 2

1 n
S303. Let a1 = 1 and an+1 = 2 an + an , for n ≥ 1. Find ba2014 c.
Proposed by Marius Cavachi, Romania
S304. Let M be a point inside triangle ABC. Line AM intersects the circumcircle of triangle
M BC for the second time at D. Similarly, line BM intersects the circumcircle of triangle
M CA for the second time at E and line CM intersects the circumcircle of triangle M AB
for the second time at F . Prove that
AD BE CF 9
+ + ≥ .
MD ME MF 2
S305. Solve in integers the following equation:
x2 + y 2 + z 2 = 2(xy + yz + zx) + 1.
S306. Points M, N, K lie on sides BC, CA, AB of a triangle ABC, respectively and are different
from its vertices. Triangle M N K is called beautiful if ∠BAC = ∠KM N and ∠ABC =
∠KN M . If in triangle ABC there are two beautiful triangles with a common vertex,
prove that triangle ABC is right.

U301. Let x, y, z, t > 0 such that x ≤ 2, x + y ≤ 6, x + y + z ≤ 12, and x + y + z + t ≤ 24.
Prove that
1 1 1 1
+ + + ≥ 1.
x y z t
U302. Let a be a real number. Evaluate
r q p
a− a2 − a4 − a8 − . . . .
Proposed by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain

U303. Let p1 , p2 , . . ., pk be distinct primes and let n = p1 p2 . . . pk . For each function
f : {1, 2, . . . , n} → {1, 2, . . . , n}, denote Pf (n) = f (1)f (2) . . . f (n).
(a) For how many functions f are n and Pf (n) are relatively prime?
(b) For how many functions f is gcd(n, Pf (n)) a prime?
Proposed by Vladimir Cerbu and Mihai Piticari, Romania
U304. In a finite graph G, we call a subset S of the set of vertices a dominating set if the
following conditions are satisfied
(i) the subgraph induced by S is connected,
(ii) every vertex of G is either in S or is adjacent to a vertex of S.
Given a positive integer k, find the maximum number of edges of G such that there are
no dominating sets of size k or less.
U305. Let (an )n≥1 be a sequence of positive real numbers such that a1 + a2 + . . . + an < n2 for
all n ≥ 1. Prove that
1 1 1
lim + + ... + = ∞.
n→∞ a1 a2 an
Proposed by Mihai Piticari Campulung, Moldovenesc, Romania
U306. Let n be a natural number. Prove the identity
n Z ∞
X 2k+1 4n+1 1
π= + dx
2k 2n 1 (1 + x2 )n+1
k=1 k
k n
and derive the estimate
n
X 2k+1
2 4
n
√ < π − < n√ .
2 n 2k 2 n
k=1 k
k
Proposed by Albert Stadler, Herrliberg, Switzerland

Olympiad problems
O301. Let a, b, c, d be nonnegative real numbers such that a2 + b2 + c2 + d2 = 4. Prove that

a b c d
+ + + ≤ 1.
b+3 c+3 d+3 a+3
O302. Let ABC be an isosceles triangle with AB = AC and let M ∈ (BC) and N ∈ (AC)
such that ∠BAM = ∠M N C. Suppose that lines M N and AB intersect at P . Prove
that the bisectors of angles BAM and BP M intersect at a point lying on line BC.
O303. Let a, b, c be real numbers greater than 2 such that

1 1 1 1
+ 2 + 2 = .
a2 −4 b −4 c −4 7
Prove that
1 1 1 3
+ + ≤ .
a+2 b+2 c+2 7
O304. Let C1 and C2 be non-intersecting circles centered at O1 and O2 . One common external
tangent of these circles touches Ci at Pi (i = 1, 2). The other common external tangent
touches Ci at Qi (i = 1, 2). Denote by M the midpoint of Q1 Q2 . Let Pi M intersect Ci at
Ri and R1 R2 intersect Ci again at Si (i = 1, 2). P1 S1 intersects P2 S2 at A. The tangent
to C1 at R1 and the tangent to C2 at R2 intersect at B. Prove that AB ⊥ O1 O2 .
Proposed by Alex Anderson, UC Berkeley, USA
O305. Prove that for any positive integers m and a, there is a positive integer n such that
an + n is divisible by m.
Proposed by Gregory Galperin, Eastern Illinois University, USA
O306. Let ABC be a triangle with incircle γ and circumcircle Γ. Let Ω be the circle tangent
to rays AB, AC, and to Γ externally, and let A0 be the tangency point of Ω with Γ. Let
the tangents from A0 to γ intersect Γ again at B 0 and C 0 . Finally, let X be the tangency
point of the chord B 0 C 0 with γ. Prove that the circumcircle of triangle BXC is tangent
to γ.

Junior problems
J307. Prove that for each positive integer n there is a perfect square whose sum of digits is
equal to 4n .
J308. Are there triples (p, q, r) of primes for which (p2 − 7)(q 2 − 7)(r2 − 7) is a perfect square?
J309. Let n be an integer greater than 3 and let S be a set of n points in the plane that are not
the vertices of a convex polygon and such that no three are collinear. Prove that there
is a triangle with the vertices among these points having exactly one other point from S
in its interior.
J310. Alice puts checkers in some cells of an 8 × 8 board such that:
a) there is at least one checker in any 2 × 1 or 1 × 2 rectangle.

b) there are at least two adjacent checkers in any 7 × 1 or 1 × 7 rectangle.
Find the least amount of checkers that Alice needs to satisfy both conditions.
J311. Let a, b, c be real numbers greater than or equal to 1. Prove that
a(b2 + 3) b(c2 + 3) c(a2 + 3)

+ + ≥ 3.
3c2 + 1 3a2 + 1 3b2 + 1
J312. Let ABC be a triangle with circumcircle Γ and let P be a point in its interior. Let M
be the midpoint of side BC and let lines AP , BP , CP intersect BC, CA, AB at X, Y ,
Z, respectively. Furthermore, let line Y Z intersect Γ at points U and V . Prove that M ,
X, U , V are concyclic.

Senior problems
S307. Let ABC be a triangle such that ∠ABC − ∠ACB = 60◦ . Suppose that the length of the
altitude from A is 41 BC. Find ∠ABC.
Proposed by Omer Cerrahoglu and Mircea Lascu, Romania
S308. Let n be a positive integer and let Gn be an n × n grid with the number 1 written in
each of its unit squares. An operation consists of multiplying all entries of a column or
all entries of a row by −1. Determine the number of distinct grids that can be obtained
after applying a finite number of operations on Gn .
S309. Let ABCD be a circumscriptible quadrilateral, which lies stricly inside a circle ω. Let
ωA be the circle outside of ABCD that is tangent to AB, AD, and to ω at A0 . Similarly,
define B 0 , C 0 , D0 . Prove that lines AA0 , BB 0 , CC 0 , DD0 are concurrent.
Proposed by Khakimboy Egamberganov, Tashkent, Uzbekistan
S310. Let a, b, c be nonzero complex numbers such that |a| = |b| = |c| = k. Prove that
p p p √
| − a + b + c| + |a − b + c| + |a + b − c| ≤ 3 k.
S311. Let n be a positive integer. Prove that

bn
2
c bn
2
c
n n
(x + 2j + 1)( )− (x + 2j)(2j )
Y Y
2j+1
j=0 j=0
is a polynomial of degree 2n−1 − n, whose highest term’s coefficient is (n − 1)!.
S312. Let a, b, c, d be positive real numbers such that a3 + b3 + c3 + d3 = 1. Prove that

1 1 1 1 16
+ + + ≤ .
1 − bcd 1 − cda 1 − dab 1 − abc 3

U307. Prove that any polynomial f ∈ R[X] can be written as a difference of increasing poly-
nomials.
Proposed by Jishnu Bose, Calcutta, India
U308. Let a1 , b1 , c1 , a2 , b2 , c2 be positive real numbers. Consider the functions X(x, y) and
Y (x, y) which satisfy the system of functional equations
x
= 1 + a1 x + b1 y + c1 Y,
X
y
= 1 + a2 x + b2 y + c2 X.
Y
Prove that if 0 < x1 ≤ x2 and 0 < y2 ≤ y1 , then X(x1 , y1 ) ≤ X(x2 , y2 ) and Y (x1 , y1 ) ≥
Y (x2 , y2 ).
U309. Let a1 , . . . , an be positive real numbers such that a1 + . . . + an = 1, n ≥ 2. Prove that

for every positive integer m,
n
X am+1
k 1
m ≥ m .
1 − ak n −1
k=1
Proposed by Titu Zvonaru, Comanesti and Neculai Stanciu, Romania
U310. Let E be an ellipse with foci F and G, and let P be a point in its exterior. Let A and
B be the points where the tangents from P to E intersect E, such that A is closer to F .
Furthermore, let X be the intersection of AG with BF . Prove that XP bisects ∠AXB.
U311. Let f : [0, 1] → [0, 1] be a nondecreasing concave function such that f (0) = 0 and
f (1) = 1. Prove that
Z 1
1
(f (x)f −1 (x))2 dx ≥ .
0 12
U312. Let p be a prime and let R be a commutative ring with characteristic p. Prove that the
sets Sk = {x ∈ R | xp = k}, where k ∈ {1, . . . , p}, have the same number of elements.
Proposed by Corneliu Manescu-Avram, Ploiesti, Romania

Olympiad problems
O307. Let a, b, c, d be positive real numbers such that a + b + c + d = 4. Prove that

1 1 1 1 1
+ + + ≤ .
a+3 b+3 c+3 d+3 abcd
O308. Let ABC be a triangle and let X, Y be points in its plane such that
AX : BX : CX = AY : BY : CY.
Prove that the circumcenter of triangle ABC lies on the line XY .
Proposed by Cosmin Pohoata, USA and Josef Tkadlec, Czech Republic
O309. Determine the least real number µ such that

√ √ √ p
µ( ab + bc + ca) + a2 + b2 + c2 ≥ a + b + c
for all nonnegative real numbers a, b, c with ab + bc + ca > 0. Find when equality holds.
O310. Let ABC be a triangle and let P be a point in its interior. Let X, Y , Z be the
intersections of AP , BP , CP with sides BC, CA, AB, respectively. Prove that
XB Y C ZA R
· · ≤ .
XY Y Z ZX 2r
O311. Let ABC be a triangle with circumcircle Γ centered at O. Let the tangents to Γ at
vertices B and C intersect each other at X. Consider the circle X centered at X with
radius XB, and let M be the point of intersection of the internal angle bisector of angle
A with X such that M lies in the interior of triangle ABC. Denote by P the intersection
of OM with the side BC and by E and F be the orthogonal projections of M on CA
and AB, respectively. Prove that P E and F P are perpendicular.
O312. Find all increasing bijections f : (0, ∞) → (0, ∞) satisfying
f (f (x)) − 3f (x) + 2x = 0
and for which there exists x0 > 0 such that f (x0 ) = 2x0 .

Junior problems
J313. Solve in real numbers the system of equations
x(y + z − x3 ) = y(z + x − y 3 ) = z(x + y − z 3 ) = 1.
J314. Alice was dreaming. In her dream, she thought that primes of the form 3k + 1 are weird.
Then she thought it would be interesting to find a sequence of consecutive integers all of
which are greater than 1 and which are not divisible by weird primes. She quickly found
five consecutive numbers with this property:
8 = 23 , 9 = 32 , 10 = 2 · 5, 11 = 11, 12 = 22 · 3.
What is the length of the longest sequence she can find?
J315. Let a, b, c be non-negative real numbers such that a + b + c = 1. Prove that

√ √ √ √
4a + 1 + 4b + 1 + 4c + 1 ≥ 5 + 2.
Proposed by Cosmin Pohoata, Columbia University, USA
J316. Solve in prime numbers the equation
x3 + y 3 + z 3 + u3 + v 3 + w3 = 53353.
J317. In triangle ABC, the angle-bisector of angle A intersects line BC at D and the circum-
ference of triangle ABC at E. The external angle-bisector of angle A intersects line BC
at F and the circumference of triangle ABC at G. Prove that DG ⊥ EF .
J318. Determine the functions f : R → R satisfying f (x − y) − xf (y) ≤ 1 − x for all real

numbers x and y.

Senior problems
√ √ √
S313. Let a, b, c be nonnegative real numbers such that a + b + c = 3. Prove that
p p p
(a + b + 1)(c + 2) + (b + c + 1)(a + 2) + (c + a + 1)(b + 2) ≥ 9.
S314. Let p, q, x, y, z be real numbers satisfying
x2 y + y 2 z + z 2 x = p and xy 2 + yz 2 + zx2 = q.
Evaluate (x3 − y 3 )(y 3 − z 3 )(z 3 − x3 ) in terms of p and q.
S315. Consider triangle ABC with inradius r. Let M and M 0 be two points inside the triangle
such that ∠M AB = ∠M 0 AC and ∠M BA = ∠M 0 BC. Denote by da , db , dc and d0a , d0b , d0c
the distances from M and M 0 to the sides BC, CA, AB, respectively. Prove that
da db dc d0a d0b d0c ≤ r6 .
S316. Circles C1 (O1 , R1 ) and C2 (O2 , R2 ) intersect in points U and V . Points A1 , A2 , A3 lie on
C1 and points B1 , B2 , B3 lie on C2 such that A1 B1 , A2 B2 , A3 B3 are passing through U .
Denote by M1 , M2 , M3 the midpoints of A1 B1 , A2 B2 , A3 B3 . Prove that M1 M2 M3 V is a
cyclic quadrilateral.
S317. Let ABC be an acute triangle inscribed in a circle of radius 1. Prove that

tan A tan B tan C 1 1 1
+ + ≥ 4 + + − 3.
tan3 B tan3 C tan3 A a2 b2 c2
S318. Points A1 , B1 , C1 , D1 , E1 , F1 are lying on the sides AB, BC, CD, DE, EF, F A of a convex
hexagon ABCDEF such that
AA1 AF1 CC1 CB1 ED1 EE1
= = = = = = λ.
AB AF CD BC ED EF
2
[ACE] λ
Prove that A1 D1 , B1 E1 , C1 F1 are concurrent if and only if [BDF ] = 1−λ .

U313. Let X and Y be nonnegative definite Hermitian matrices such that X − Y is also non-
negative definite. Prove that tr(X 2 ) ≥ tr(Y 2 )
Proposed by Radouan Boukharfane, Sidislimane, Morocco
U314. Prove that for any positive integer k,

√
n
√
n
!n
1+ 2 + ··· + k k
lim > ,
n→∞ k e
where e is Euler constant.
U315. Let X and Y be complex matrices of the same order with XY 2 − Y 2 X = Y . Prove that
Y is nilpotent.
U316. The sequence {Fn } is defined by F1 = F2 = 1, Fn+2 = Fn+1 + Fn for n ≥ 1. For any
nonnegative integer m, let v2 (m) be the highest power of 2 dividing m. Prove that there
is exactly one positive real number µ such that the equation
v2 (bµnc!) = v2 (F1 . . . Fn )
is satisfied by infinitely many positive integers n. Find µ.
U317. For any positive integers s, t, p, prove that there is a number M (s, t, p) such that every
graph G with a matching of size at least M (s, t, p) contains either a complete graph
Ks , an induced copy of the complete bipartite graph Kt,t , or a matching of size p as an
induced subgraph. Does the result remain true if we replace the word “matching” by
“path”?
(−1)q(k)
U318. Determine all possible values of ∞
P
k=1 k2
, where q(x) is a quadratic polynomial that
assumes only integer values at integer places.

Olympiad problems
O313. Find all positive integers n for which there are positive integers a0 , a1 , . . . , an such that
a0 + a1 + · · · + an = 5(n − 1) and
1 1 1
+ + ··· + = 2.
a0 a1 an
O314. Prove that every polynomial p(x) with integer coefficients can be represented as a sum
of cubes of several polynomials that return integer values for any integer x.
(a3 + 3b2 + 5)(b3 + 3c2 + 5)(c3 + 3a2 + 5) ≥ 27(a + b + c)3 .
O316. Prove that for all integers k ≥ 2 there exists a power of 2 such that at least half of the
last k digits are nines. For example, for k = 2 and k = 3 we have 212 = . . . 96 and
253 = . . . 992.
O317. Twelve scientists met at a math conference. It is known that every two scientists have a
common friend among the rest of the people. Prove that there is a scientist who knows
at least five people from the attendees of the conference.
O318. Find all polynomials f ∈ Z[X] with the property that for any distinct primes p and q,
f (p) and f (q) are relatively prime.

Junior problems
J319. Let 0 = a0 < a1 < . . . < an < an+1 = 1 such that a1 + a2 + · · · + an = 1. Prove that
a1 a2 an 1
+ + ··· + ≥ .
a2 − a0 a3 − a1 an+1 − an−1 an
J320. Find all positive integers n for which 2014n + 11n is a perfect square.
J321. Let x, y, z be positive real numbers such that xyz(x + y + z) = 3. Prove that
1 1 1 54
2
+ 2+ 2+ ≥ 9.
x y z (x + y + z)2
Proposed by Marius Stânean, Zalau, Romania
J322. Let ABC be a triangle with centroid G. The parallel lines through a point P situated in
0 0 0
the plane of the triangle to the medians AA , BB , CC intersect lines BC, CA, AB at
A1 , B1 , C1 , respectively. Prove that
0 0 0 3
A A1 + B B1 + C C1 ≥ P G.
2
Proposed by Dorin Andrica, Babes, -Bolyai University, Cluj-Napoca, Romania
J323. In triangle ABC, √

5−1
sin A + sin B + sin C = .
2
Prove that max(A, B, C) > 162◦ .
J324. Let ABC be a triangle and let X, Y , Z be the reflections of A, B, C in the opposite
sides. Let Xb , Xc be the orthogonal projections of X on AC, AB, Yc , Ya the orthogonal
projections of Y on BA, BC, and Za , Zb the orthogonal projections of Z on CB, CA,
respectively. Prove that Xb , Xc , Yc , Ya , Za , Zb are concyclic.
Proposed by Cosmin Pohoat, ă, Columbia University, USA

Senior problems
S319. Let a, b, c be positive real numbers such that a + b + c = 1. Prove that for any positive
real number t,
(at2 + bt + c)(bt2 + ct + a)(ct2 + at + b) ≥ t3 .
S320. Let ABC be a triangle with circumcenter O and incenter I. Let D, E, F be the tangency
points of the incircle with BC, CA, AB, respectively. Prove that line OI is perpendicular
to angle bisector of ∠EDF if and only if ∠BAC = 60◦ .
S321. Let x be a real number such that xm (x + 1) and xn (x + 1) are rational for some relatively
prime positive integers m and n. Prove that x is rational.
S322. Let ABCD be a cyclic quadrilateral. Points E and F lie on the sides AB and BC,
respectively, such that ∠BF E = 2∠BDE. Prove that
EF FC CD
= + .
AE AE AD
S323. Solve in positive integers the equation
x + y + (x − y)2 = xy.
Proposed by Neculai Stanciu and Titu Zvonaru, Romania
S324. Find all functions f : S → S satisfying
f (x)f (y) + f (x) + f (y) = f (xy) + f (x + y)
for all x, y ∈ S when (i) S = Z; (ii) S = R.
Proposed by Prasanna Ramakrishnan, Port of Spain, Trinidad and Tobago

U319. Let A, B, C be the measures (in radians) of the angles of a triangle with circumradius
R and inradius r. Prove that
A B C 2R
+ + ≤ − 1.
B C A r
Proposed by Nermin Hodžić, Bosnia and Herzegovina and Salem Malikić, Canada
U320. Evaluate X 2n
.
22n + 1
n≥0
Proposed by Titu Andreescu, University of Texas at Dallas
U321. Consider the sequence of polynomials (Ps )s≥1 defined by

0
Pk+1 (x) = (xa − 1)Pk (x) − (k + 1)Pk (x), k = 1, 2, . . . ,
where P1 (x) = xa−1 and a is an integer greater than 1.
1. Find the degree of Pk .

2. Determine Pk (0)
U322. Evaluate
∞
X 16n2 − 12n + 1
.
n(4n − 2)!
n=1
Proposed by Titu Andreescu, USA and Oleg Mushkarov, Bulgaria
U323. Let X and Y be independent random variables following a uniform distribution

(
1 0 < x < 1,
pX (x) =
0 otherwise.
What is the probability that inequality X 2 + Y 2 ≥ 3XY is true?
U324. Let f : [0, 1] → R be a differentiable function such that f (1) = 0. Prove that there is
c ∈ (0, 1) such that |f (c)| ≤ |f 0 (c)|.

Olympiad problems
O319. Let f (x) and g(x) be arbitrary functions defined for all x ∈ R. Prove that there is a
function h(x) such that (f (x) + h(x))2014 + (g(x) + h(x))2014 is an even function for all
x ∈ R.
O320. Let n be a positive integer and let 0 < yi ≤ xi < 1 for 1 ≤ i ≤ n. Prove that
1 − x1 · · · xn 1 − x1 1 − xn
≤ + ··· + .
1 − y1 · · · yn 1 − y1 1 − yn
Proposed by Angel Plaza, Universidad de Las Palmas de Gran Canaria, Spain
O321. Each of the diagonals AD, BE, CF of the convex hexagon ABCDEF divide its area in
half. Prove that
AB 2 + CD2 + EF 2 = BC 2 + DE 2 + F A2 .
O322. Let ABC be a triangle with circumcircle Γ and let M be the midpoint of arc BC not
containing A. Lines `b and `c passing through B and C, respectively, are parallel to
AM and meet Γ at P 6= B and Q 6= C. Line P Q intersects AB and AC at X and Y ,
respectively, and the circumcircle of AXY intersects AM again at N .
Prove that the perpendicular bisectors of BC, XY , and M N are concurrent.

1 2 n
O323. Prove that the sequence 22 +1, 22 +1, . . . 22 +1, . . . and an arbitrary infinite increasing
arithmetic sequence have either infinitely many terms in common or at most one term in
common.
O324. Let a, b, c, d be nonnegative real numbers such that a3 + b3 + c3 + d3 + abcd = 5. Prove

that
abc + bcd + cda + dab − abcd ≤ 3.

Junior problems
J289. Let a be a real number such that 0 ≤ a < 1. Prove that

1 1
a 1+ +1= .
1−a 1−a
Solution by Ángel Plaza, University of Las Palmas de Gran Canaria, Spain

Since 0 ≤ a < 1, then 0 < 1 − a ≤ 1.
1 1 1 k−1 k 1
If k ∈ N such that < 1 − a ≤ , then k ≤ < k + 1, and ≤a< , so = k.
1+k k 1−a k k+1 1−a
On the other hand, for the left-hand side of the proposed identity we have

1
a 1+ + 1 = ba (1 + k)c + 1
1−a
= k − 1 + 1 = k.
Also solved by Archisman Gupta, RKMV, Agartala, Tripura, India; Joshua Benabou, Manhasset High
School, NY, USA; Daniel Lasaosa, Pamplona, Navarra, Spain; Arber Igrishta, Eqrem Qabej, Vushtrri, Ko-
sovo; Mathematical Group Galaktika shqiptare, Albania; Paolo Perfetti, Università degli studi di Tor Vergata
Roma, Italy; Alessandro Ventullo, Milan, Italy; Polyahedra, Polk State College, FL, USA; AN-anduud Pro-
blem Solving Group, Ulaanbaatar, Mongolia; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Viet
Quoc Hoang, University of Auckland, New Zealand.

p
3
p
3
p
3
13a3 + 14b3 + 13b3 + 14c3 + 13c3 + 14a3 ≥ 3.
Solution by Shohruh Ibragimov, Lyceum Nr.2 under the SamIES, Samarkand, Uzbekistan
From Hölder’s inequality we easily obtain (13a3 + 14b3 )(13 + 14)(13 + 14) ≥ (13a + 14b)3 =>
p
3 13a + 14b
13a3 + 14b3 ≥ (1)
9
Similarly, we have
p
3 13b + 14c p
3 13c + 14a
13b3 + 14c3 ≥ , 13c3 + 14a3 ≥ (2)
9 9
From (1) and (2) we get
p
3
p
3
p
3 27(a + b + c)
13a3 + 14b3 + 13b3 + 14c3 + 13c3 + 14a3 ≥ =3
9
Also solved by Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Daniel Lasaosa, Pamplona,
Navarra, Spain; Nicuşor Zlota, Traian Vuia Technical College, Focşani, Romania; Arkady Alt, San Jose,
California, USA; An Zhen-ping, Xianyang Normal University, China; Viet Quoc Hoang, University of Auc-
kland, New Zealand; Ioan Viorel Codreanu, Satulung, Maramures, Romania; AN-anduud Problem Solving
Group, Ulaanbaatar, Mongolia; Polyahedra, Polk State College, FL, USA; Paolo Perfetti, Università degli
studi di Tor Vergata Roma, Italy; Ángel Plaza, University of Las Palmas de Gran Canaria, Spain; Sayak
Mukherjee, Kolkata, India; Sayan Das, Indian Statistical Institute, Kolkata, India; Neculai Stanciu, Buzău,
Romania and Titu Zvonaru, Comănes, ti, Romania; Alessandro Ventullo, Milan, Italy; Georgios Batzolis,
Mandoulides High School, Thessaloniki, Greece.

J291. Let ABC be a triangle such that ∠BCA = 2∠ABC and let P be a point in its interior such that
P A = AC and P B = P C. Evaluate the ratio of areas of triangles P AB and P AC.
Proposed by Panagiote Ligouras, Noci, Italy
Solution by Polyahedra, Polk State College, USA
D ∑
B A
Let Σ be the circle with center A and radius AC. Suppose that the bisector of ∠ACB intersects AB at D and
Σ at E. Then P D is the perpendicular bisector of BC. Since ∠AEC = ∠ACE = ∠BCE, EA k BC. Thus
∠BAE = ∠ABC, so P D is the perpendicular bisector of AE as well. Hence 4AP E is equilateral. Therefore,
∠P CE = 21 ∠P AE = 30◦ , ∠AP C = 30◦ + B, and ∠AP B = 60◦ + ∠EP B = 60◦ + ∠AP C = 90◦ + B. Finally,
let [·] denote area, then
[P AB] sin ∠AP B cos B 2

= = ◦
= √ .
[P AC] sin ∠AP C sin (30 + B) 1 + 3 tan B
Also solved by Andrea Fanchini, Cantú, Italy; Daniel Lasaosa, Pamplona, Navarra, Spain; Arkady Alt,
San Jose, California, USA.

J292. Find the least real number k such that for every positive real numbers x, y, z, the following inequality
holds: Y
(2xy + yz + zx) ≤ k(x + y + z)6 .
cyc
Solution by Albert Stadler, Herrliberg, Switzerland

64
We claim that k = .
729
Let x = y = z = 1 ⇒ Q
(2xy + yz + zx)
cyc 64
= ≤k
(x + y + z)6 729
By the Cauchy-Schwarz inequality , xy + yz + zx ≤ x2 + y 2 + z 2 .
So 3(xy + yz + zx) ≤ 2xy + 2yz + 2zx + x2 + y 2 + z 2 = (x + y + z)2 .
By AM-GM:
(2xy + yz + zx) 3
P  P 3
4 xy

3
4(x + y + z)2

Y cyc cyc 64
(2xy + yz + zx) ≤   =   ≤ = (x + y + z)6 ,
cyc
3 3 9 729
64
Which proves that k ≤ and the conclusion follows.
729
Also solved by Daniel Lasaosa, Pamplona, Navarra, Spain; Alessandro Ventullo, Milan, Italy; Titu Zvona-
ru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania; Shohruh Ibragimov, Lyceum Nr.2 under the
SamIES, Samarkand, Uzbekistan; Sayan Das, Indian Statistical Institute, Kolkata, India; Sayak Mukherjee,
Kolkata, India; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Sun Mengyue Lansheng,
Fudan Middle School, Shanghai, China; Arkady Alt, San Jose, California, USA; Georgios Batzolis, Mandou-
lides High School, Thessaloniki, Greece; Ioan Viorel Codreanu, Satulung, Maramures, Romania; AN-anduud
Problem Solving Group, Ulaanbaatar, Mongolia; Polyahedra, Polk State College, FL, USA; Jan Jurka, Brno,
Czech Republic.

J293. Find all positive integers x, y, z such that
2
x + y2 + z2 − 8xyz = 1.
Proposed by Aaron Doman, University of California, Berkeley , USA
Solution by Alessandro Ventullo, Milan, Italy

We rewrite the equation as
x2 + 2x(y 2 + z 2 − 4yz) + (y 2 + z 2 )2 − 1 = 0.
Since x must be a positive integer, the discriminant of this quadratic equation in x must be non-negative,
i.e.
(y 2 + z 2 − 4yz)2 − (y 2 + z 2 )2 + 1 ≥ 0,
which is equivalent to
−8yz(y − z)2 + 1 ≥ 0,
which gives yz(y − z)2 ≤ 1/8. Since y and z are positive integers, it follows that
yz(y − z)2 = 0,
so y − z = 0, i.e. y = z. The given equation becomes (x − 2y 2 )2 = 1, which yields x = 2y 2 ± 1. Therefore,

all the positive integer solutions to the given equation are
(2n2 − 1, n, n), (2n2 + 1, n, n), n ∈ Z+ .
Also solved by Albert Stadler, Herrliberg, Switzerland; Daniel Lasaosa, Pamplona, Navarra, Spain; Sima
Sharifi, College at Brockport, SUNY, USA; Sayan Das, Indian Statistical Institute, Kolkata, India; Georgios
Batzolis, Mandoulides High School, Thessaloniki, Greece; Viet Quoc Hoang, University of Auckland, New
Zealand; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Polyahedra, Polk State College, FL, USA.

1 ≤ (a2 − a + 1)(b2 − b + 1)(c2 − c + 1) ≤ 7.
Solution by Daniel Lasaosa, Pamplona, Navarra, Spain

Denote p = abc and s = ab + bc + ca, where clearly 0 ≤ p ≤ 1, with p = 0 iff at least one of a, b, c is zero,
and p = 1 iff a = b = c = 1 by the AM-GM inequality, while 0 ≤ s ≤ 3, with s = 0 iff two out of a, b, c are
zero, and s = 3 iff a = b = c = 1 by the scalar product inequality. Note that
(a2 − a + 1)(b2 − b + 1)(c2 − c + 1) = p2 + s2 − ps − 4s − p + 7.
The lower bound then rewrites as
(2 + p − s)2 + 2(1 − p) ≥ p(3 − s).
Now, 9p = 3abc(a + b + c) ≤ (ab + bc + ca)2 = s2 , or 9(1 − p) ≥ (3 + s)(3 − s), or it suffices to show that
6 + 2s − 9p
(2 + p − s)2 + (3 − s) ≥ 0.
9
p √
Assume that 9p > 6, or p > 23 , hence by the AM-GM inequality, s ≥ 3 3 p2 ≥ 3 12 > 2, or 6 + 2s > 10 > 9p.
It follows that both terms in the LHS are non-negative, being zero iff s = 3 and simultaneously s = p + 2,
for p = 1.
On the other hand, the upper bound rewrites as
p(1 − p) + s(3 − s) + s + ps ≥ 0,
clearly true because s, ps, p(1 − p), s(3 − s) are all non-negative. Note that equality requires s = 0, and since
a, b, c are non-negative, this requires at least two out of a, b, c to be zero, resulting in p = 0.
The conclusion follows, equality holds in the lower bound iff a = b = c = 1, and in the upper bound iff
(a, b, c) is a permutation of (3, 0, 0).
Also solved by Polyahedra, Polk State College, FL, USA; AN-anduud Problem Solving Group, Ulaanbaa-
tar, Mongolia; Viet Quoc Hoang, University of Auckland, New Zealand; Arkady Alt, San Jose, California,
USA; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Paolo Perfetti, Università de-
gli studi di Tor Vergata Roma, Italy; Shivang jindal, Jaipur, Rajasthan, India; Albert Stadler, Herrliberg,
Switzerland.

Senior problems
S289. Let x, y, z be positive real numbers such that x ≤ 4, y ≤ 9 and x + y + z = 49. Prove that
1 1 1
√ + √ + √ ≥ 1.
x y z
Solution by Li Zhou, Polk State College, FL, USA √

Applying Jensen’s inequality to the convex function f (t) = 1/ t for t > 0, we get

1 1 1 1 x 1 y 1 z 1 x 1 y 1 z
√ +√ +√ = f + f + f ≥f · + · + ·
x y z 2 4 3 9 6 36 2 4 3 9 6 36
√ √ √
216 216 216
= √ =√ ≥p = 1.
27x + 8y + z 26x + 7y + 49 26(4) + 7(9) + 49
Also solved by Daniel Lasaosa, Pamplona, Navarra, Spain; Sima Sharifi, College at Brockport, SUNY,
USA; Arkady Alt, San Jose, California, USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia;
Ioan Viorel Codreanu, Satulung, Maramures, Romania; Shohruh Ibragimov, Lyceum Nr.2 under the SamIES,
Samarkand, Uzbekistan; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Sun Mengyue ,
Lansheng Fudan Middle School, Shanghai, China.

S290. Prove that there is no integer n for which
2
1 1 1 4
2
+ 2 + ··· + 2 =
2 3 n 5

Let p be the greatest prime number such that p ≤ n < 2p. Then the given equality can be written as
52 k = (4 · n!)2 ,
n
X (n!)2 (n!)2
where k = . Observe that k ≡ 6≡ 0 (mod p). Since p|52 k and p does not divide k, it follows
i2 p2
i=2
that p|52 , i.e. p = 5. So, n ∈ {5, 6, 7, 8, 9}. An easy check shows that none of these values satisfies the
equality.
Also solved by Daniel Lasaosa, Pamplona, Navarra, Spain; Yassine Hamdi, Lycée du Parc, Lyon, France;
Sima Sharifi, College at Brockport, SUNY, USA; Li Zhou, Polk State College, FL, USA; Arghya Datta,
Hooghly Collegiate School, Kolkata, India; Albert Stadler, Herrliberg, Switzerland.

S291. Let a, b, c be nonnegative real numbers such that ab + bc + ca = 3. Prove that
5 2
2a2 − 3ab + 2b2 2b2 − 3bc + 2c2 2c2 − 3ca + 2a2 ≥ a + b2 + c2 − 4.

3
Proposed by Titu Andreescu, USA and Marius Stanean, Romania

Note that
27 2a2 − 3ab + 2b2 2b2 − 3bc + 2c2 2c2 − 3ca + 2a2 + 27 · 4 − 45 a2 + b2 + c2 =

= 27 2a2 − 3ab + 2b2 2b2 − 3bc + 2c2 2c2 − 3ca + 2a2 + 4 (ab + bc + ca)3 −

−5 a2 + b2 + c2 (ab + bc + ca)2 =

= 211(a4 b2 + b4 c2 + c4 a2 + a2 b4 + b2 c4 + c2 a4 ) − 320(a3 b3 + b3 c3 + c3 a3 )−
−334abc(a3 + b3 + c3 ) + 164abc(a2 b + b2 c + c2 a + ab2 + bc2 + ca2 ) − 288a2 b2 c2 ,
or it suffices to show that this last expression is non-negative. Note that it can be rearranged as
X
3c4 + 160a2 b2 + 48c2 (a + b)2 − 164abc(a + b) (a − b)2 =

cyc
X
3c4 + 5c2 (a + b)2 + 39(bc + ca − 2ab)2 + 4(bc + ca − ab)2 (a − b)2 ,

=
cyc
clearly non-negative, and being zero iff a = b = c. The conclusion follows.
Also solved by Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy; Nicuşor Zlota, “Traian
Vuia” Technical College, Focşani, Romania; Arkady Alt, San Jose, California, USA.

S292. Given triangle ABC, prove that there exists X on the side BC such that the inradii of triangles AXB
and AXC are equal and find a ruler and compass construction.
Solution by Li Zhou, Polk State College, USA
I
J K
B E X F D C
As usual, let a, b, c, s, and r be the sides BC, CA, AB, semiperimeter, and inradius of 4ABC. Let I, J, K
be the incenters of triangles ABC, ABX, AXC, and D, E, F be the feet of perpendiculars from A, J, K
onto BC. Let h = AD, r1 = JE, and r2 = KF . As X moves from B to C, r1 increases from 0 to r while r2
BE
decreases from r to 0. Hence, there exists X such that r1 = r2 = t. Now rt = s−b , so BE = t(s−b)
r . Likewise,
t(s−c)
FC = r . Let [·] denote area. Then
1
ah = [ABC] = [ABX] + [AXC] = t(c + EX) + t(b + XF )
2
t t
= t(c + b + a − BE − F C) = (2rs − ta) = (ah − at).
r r

Hence, t2 − ht + 12 hr = 0, which yields t = 12 h − h(h − 2r) . This suggests an easy construction:
p
p
Construct the length 2rpon AD. Then construct the length AY = h(h − 2r). Taking away the length
AY from AD gives us h − h(h − 2r). Halving this yields t, as in the figure.
Also solved by Titu Zvonaru, Comănes, ti, Romania, and Neculai Stanciu, ”George Emil Palade” School,
Buzău, Romania; Arkady Alt, San Jose, California, USA.

S293. Let a, b, c be distinct real numbers and let n be a positive integer. Find all nonzero complex numbers
z such that
c a b
az n + bz̄ + = bz n + cz̄ + = cz n + az̄ + .
z z z

Since z 6= 0, we may multiply by z all terms in the proposed equation, or
az n+1 + b|z|2 + c = bz n+1 + c|z|2 + a = cz n+1 + a|z|2 + b,
yielding
(a − b)z n+1 + (b − c)|z|2 + (c − a) = (b − c)z n+1 + (c − a)|z|2 + (a − b) = 0,
and eliminating the terms with z n+1 , we obtain
(a − b)2 − (b − c)(c − a) a2 + b2 + c2 − ab − bc − ca
|z|2 = = = 1,
(b − c)2 − (a − b)(c − a) a2 + b2 + c2 − ab − bc − ca
since a2 + b2 + c2 > ab + bc + ca because of the Cauchy-Schwarz Inequality, a, b, c being distinct. Inserting

this result in the original equation and rearranging terms, we obtain
a z n+1 − 1 = b z n+1 − 1 = c z n+1 − 1 ,

1
or z must be one of the n + 1-th roots of unity. For any one of those n + 1 roots, we have z̄ = z = z n , and
all n + 1-th roots of unity are clearly solutions of the proposed equation.
Also solved by Arkady Alt, San Jose, California, USA; Moubinool Omarjee Lycée Henri IV, Paris, France;
Ángel Plaza, University of Las Palmas de Gran Canaria, Spain; Albert Stadler, Herrliberg, Switzerland; Li
Zhou, Polk State College, FL, USA.

S294. Let s(n) be the sum of digits of n2 + 1. Define the sequence (an )n≥0 by an+1 = s(an ), with a0 an
arbitrary positive integer. Prove that there is n0 such that an+3 = an for all n ≥ n0 .

We have to prove that the given sequence is 3-periodic. Since f (5) = 8, f (8) = 11 and f (11) = 5, it suffices
to prove that for every positive integer a0 there exists some n ∈ N such that an ∈ {5, 8, 11}. Let m be the
number of digits of a0 . We prove the statement by induction on m. For m ≤ 2 we proceed by a direct check.
If a0 ∈ {5, 8, 11} there is nothing to prove. If a0 is a two-digit number, then a20 ≤ 10000, so a1 ≤ 37 and we
reduce to analyze the cases for a0 ≤ 37.
(i) If a0 ∈ {2, 7, 20}, then a1 = 5. If a0 ∈ {1, 10, 26, 28}, then a1 ∈ {2, 20}, so a2 = 5. Finally, if
a0 ∈ {3, 6, 9, 12, 15, 18, 27, 30, 33}, then a3 = 5.
(ii) If a0 ∈ {4, 13, 23, 32}, then a1 = 8.
(iii) If a0 ∈ {17, 19, 21, 35, 37}, then a1 = 11. If a0 ∈ {14, 22, 24, 31, 36}, then a1 ∈ {17, 19}, so a2 = 11.
Finally, if a0 ∈ {16, 25, 29, 34}, then a3 = 11.
Thus, we have proved that if a0 is a one or two-digit number, then an ∈ {5, 8, 11} for some n ∈ N, i.e. the
sequence is 3-periodic. Let m ≥ 2 and suppose that the statement is true for all k ≤ m. Let a0 be an
(m + 1)-digit number. Then, 10m ≤ a0 < 10m+1 which implies 102m ≤ a20 < 102(m+1) . Hence,
a1 = f (a0 ) ≤ 9 · 2(m + 1) + 1 < 10m ,
and by the induction hypothesis, the sequence (an )n≥1 is 3-periodic, which implies that the sequence (an )n≥0
is 3-periodic, as we wanted to prove.
Also solved by Daniel Lasaosa, Pamplona, Navarra, Spain; Arghya Datta, Hooghly Collegiate School,
Kolkata, India; Li Zhou, Polk State College, FL, USA; AN-anduud Problem Solving Group, Ulaanbaatar,
Mongolia.

1
U289. Let a ≥ 1 be such that (ban c) n ∈ Z for all sufficiently large integers n. Prove that a ∈ Z.
Solution by Li Zhou, Polk State College, FL, USA

For the purpose of contradiction, suppose that a = m+h for some positive integer m and some real h ∈ (0, 1).
Then for all n ≥ h1 ,
mn + 1 ≤ mn + nh < an < (m + 1)n ,
1
and thus mn < ban c < (m + 1)n . Hence, m < (ban c) n < m + 1 for all sufficiently large n, a desired
contradiction.
Also solved by Arkady Alt, San Jose, California, USA; Daniel Lasaosa, Pamplona, Navarra, Spain;
Alessandro Ventullo, Milan, Italy; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia.

U290. Prove that there are infinitely many triples of primes (pn−1 , pn , pn+1 ) such that 12 (pn+1 + pn−1 ) ≤ pn .
Solution by G.R.A.20 Problem Solving Group, Roma, Italy

Assume for the sake of contradiction that there is an integer n0 such that for all n > n0 ,
pn−1 + pn+1 > 2pn
that is, dn > dn−1 ≥ 1 where dn = pn+1 − pn . Hence, for k ≥ 1,
dn0 +k−1 ≥ dn0 +k−2 + 1 ≥ · · · ≥ dn0 + k − 1 ≥ k
and
k(k + 1) k2
pn0 +k = dn0 +k−1 + dn0 +k−2 + · · · + dn0 + pn0 > k + (k − 1) + · · · + 1 = > .
2 2
On the other hand, by the Prime Number Theorem, π(n) ∼ n/ ln(n) and since x/ ln(x) is strictly increasing
for x ≥ e, it follows that
π(pn0 +k ) ln(pn0 +k ) (n0 + k) ln(pn0 +k ) (n0 + k) ln(k 2 /2)

1← = < →0
pn0 +k pn0 +k k 2 /2
which yields a contradiction.
Also solved by Julien Portier, Francois 1er, France; Daniel Lasaosa, Pamplona, Navarra, Spain; Arpan
Sadhukhan, Indian Statistical Institute, Kolkata; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia;
Li Zhou, Polk State College, FL, USA.

U291. Let f : R → R be a bounded function and let S be the set of all increasing maps ϕ : R → R. Prove that
there is a unique function g in S satisfying the conditions
a) f (x) ≤ g(x) for all x ∈ R.
b) If h ∈ S and f (x) ≤ h(x) for all x ∈ R then g(x) ≤ h(x) for all x ∈ R.
Solution by Arkady Alt, San Jose, California, USA

a) Since f is bounded then for any x ∈ R set G (x) := {f (t) | t ∈ R and t ≤ x} is bounded. Therefore for
any x ∈ R we can define g (x) := sup G (x) and, obviously, that function g (x) defined by such way satisfy to
condition (a).
b) Let now h ∈ S and f (x) ≤ h(x) for all x ∈ R.

Since f (t) ≤ h(t) for any t ≤ x then g (x) = sup f (t) ≤ sup h(t) = h (x) (since h is increasing in t ∈ (−∞, x] ).
t≤x t≤x
Also solved by Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy.

U292. Let r be a positive real number. Evaluate
Z π
2 1
dx.
0 1 + cotr x
Proposed by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain
Solution by Robinson Higuita, Universidad de Antioquia , Colombia

We denote by
Z π Z π
2 1 2 sinr x
I(r) := r dx = r r
dx.
0 1 + cot (x) 0 sin x + cos x
π
If we make the substitution x = 2 − y, we have
π
0 sinr ( π2 − y) cosr y
Z Z
2
I(r) = (−dy) = .
π sinr ( π2 − y) + cosr ( π2 − y) 0 cosr y + sinr y
2
Therefore
π π
sinr x cosr y
Z Z
2 2
2I(r) = dx +
0 sinr x + cosr x 0 cosr y + sinr y
π π
sinr y + cosr y
Z Z
2 2 π
= dx = dx = .
0 cosr y + sinr y 0 2
Thus
π
I(r) = .
4
We note that the hypothesis on r is not important.
Also solved by Daniel Lasaosa, Pamplona, Navarra, Spain; Albert Stadler, Herrliberg, Switzerland; Sayak
Mukherjee, Kolkata, India; Samin Riasat, Dhaka, Bangladesh; Paolo Perfetti, Università degli studi di Tor
Vergata Roma, Italy; Antoine Barré and Dmitry Chernyak, Lycée Stanislas, Paris, France; Arkady Alt, San
Jose, California, USA; Moubinool Omarjee Lycée Henri IV, Paris, France; Li Zhou, Polk State College, FL,
USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia.

U293. Let f : (0, ∞) → R be a bounded continuous Pk function and let α ∈ [0, 1). Suppose there exist real
numbers a0 , ..., ak , whith k ≥ 2, so that p=0 ap = 0 and

X k
α

lim x ap f (x + p) = a.
x→∞ p=0
Prove that a = 0.
Solution by the author

Let us denote M = sup|f (x)| and g(x) = kp=0 ap f (x + p).
P
g : (0, ∞) → R is a continuous bounded function.
Assume that a > 0.
Then there is an N > 0, such that xα |g(x)| > a, for ∀x > N (1)
From (1) we conclude that function g has the same sign ∀x > N , since it is continuous, and therefore has
the intermediate value property.
First, suppose that function g is positive on the interval (N, ∞).
a
From (1) we have |g(x)| > (2)
xα
Denoting n1 = [N ] + 1 and x = n1 , n1 + 1, n1 + 2, ... and summing up yields to
X X 1
|g(k)| > a .
nα
n≥n1 n≥n1
P 1 P 1
Since the series α
is divergent for α ∈ (0, 1) it follows that α
is divergent as well
n≥1 n n≥n1 n
X
⇒ |g(k)| = ∞.
n≥n1
Now, let T = max{|a0 |, |a1 |, |a2 |, ..., |ak |}. Taking into consideration positiveness of g on (N, ∞) and the
same values of n we get
X X X
Sn = |g(k)| = |ap f (n + p)| = | ap f (n + p)| =
n≥n1 n≥n1 n≥n1
|ak f (n + k) + (ak + ak−1 )f (n + k − 1) + ... + (ak + ak−1 + ... + a1 )f (n + 1) + (ak−1 + ak−2 + ... + a0 )f (n1 +
k − 1) + (ak−2 + ak−3 + ... + a0 )f (n1 + k − 2) + ... + (a1 + a0 )f (n1 + 1) + a0 f (n1 )| ≤
|ak |M + |ak + ak−1 |M + ... + |ak + ak−1 + ... + a1 |M +
|ak−1 + ak−2 + ... + a0 |M + |ak−2 + ak−3 + ... + a0 |M + ... + |a1 + a0 |M + |a0 |M ≤ M T + 2M T + ... + kM T +
kM T + ... + 2M T + M T = k(k + 1)M T and the series converges.
Therefore, we have reached a contradiction ⇒ a = 0. Similarly to prove for g(x) ≤ 0.

U294. Let p1 , p2 , . . . , pn be pairwise distinct prime numbers. Prove that
√ √ √ √ √ √
Q ( p1 , p2 , . . . , pn ) = Q ( p1 + p2 + · · · + pn ) .

√ √ √ √ √ √ √ √ √
Clearly, Q( p1 + p2 +. . .+ pn ) is a subfield of Q( p1 , p2 , . . . , pn ). Observe that Q( p1 , p2 , . . . , pn )
√ √
is Galois over Q, since it is the splitting field of the polynomial (x2 − p1 ) · · · (x2 − pn ). Every automorphism
√ √ √ √
σ is completely determined by its action on p1 , . . . , pn , which must be mapped to ± p1 , . . . , ± pn ,
√ √ √
respectively. Therefore, Gal(Q( p1 , p2 , . . . , pn )/Q) is the group generated by {σ1 , σ2 , . . . , σn }, where σi
is the automorphism defined by ( √
√ − pj if i = j
σi ( pj ) = √
pj if i 6= j.
√ √ √
Clearly, the only automorphism that fixes Q( p1 , p2 , . . . , pn ) is the identity. Moreover, it’s easy to see
√ √
that the only automorphism that fixes the element p1 + . . . + pn is the identity, which means that the
√ √
only automorphism that fixes Q( p1 + . . . + pn ) is the identity. Hence, by the Fundamental Theorem of
√ √ √ √ √ √
Galois Theory, Q( p1 + p2 + . . . + pn ) = Q( p1 , p2 , . . . , pn ).
Also solved by Daniel Lasaosa, Pamplona, Navarra, Spain.

Olympiad problems
O289. Let a, b, x, y be positive real numbers such that x2 − x + 1 = a2 , y 2 + y + 1 = b2 , and

(2x − 1)(2y + 1) = 2ab + 3. Prove that x + y = ab.
Solution by Khakimboy Egamberganov, S.H.Sirojiddinov lyceum, Tashkent, Uzbekistan

From the conditions (2x − 1)2 = 4a2 − 3 and (2y + 1)2 = 4b2 − 3 and
(2ab + 3)2 = (2x − 1)2 (2y + 1)2 = (4a2 − 3)(4b2 − 3),
a2 b2 − ab − a2 − b2 = 0 (1).
We have (2x − 1)2 + (2y + 1)2 = 4(a2 + b2 ) − 6, and since (2x − 1)(2y + 1) = 2ab + 3, we get that
(2(x + y))2 = 4(a2 + b2 ) − 6 + (4ab + 6),
(x + y)2 = a2 + b2 + ab (2).
Since (1), (2), we get that (x + y)2 = a2 b2 and
x + y = ab.
and we are done.
Also solved by Daniel Lasaosa, Pamplona, Navarra, Spain; Moubinool Omarjee Lycée Henri IV, Paris,
France; Shohruh Ibragimov, Lyceum Nr.2 under the SamIES, Samarkand, Uzbekistan; Daniel Văcaru, Piteşti,
Romania; Alessandro Ventullo, Milan, Italy; Li Zhou, Polk State College, FL, USA; AN-anduud Problem
Solving Group, Ulaanbaatar, Mongolia; Arkady Alt, San Jose, California, USA; Ayoub Hafid, Elaraki School,
Morroco.

O290. Let Ω1 and Ω2 be the two circles in the plane of triangle ABC. Let α1 , α2 be the circles through A that
are tangent to both Ω1 and Ω2 . Similarly, define β1 , β2 for B and γ1 , γ2 for C. Let A1 be the second
intersection of circles α1 and α2 . Similarly, define B1 and C1 . Prove that the lines AA1 , BB1 , CC1 are
concurrent.

Note that we must assume that Ω1 , Ω2 have different radii. Otherwise, the figure clearly has symmetry
around the perpendicular bisector of the segment joining their centers, and lines AA1 , BB1 , CC1 would be
parallel to the line joining their centers (which could be considered equivalent to AA1 , BB1 , CC1 meeting at
infinity). We will assume in the rest of the problem that Ω1 , Ω2 have distinct radii, hence a homothety with
center O and scaling factor ρ with |ρ| = 6 1 can be found, which transforms Ω1 into Ω2 .
Claim 1: Let Ω1 , Ω2 be two circles with different radii, and let ω be a circle simultaneously tangent to
both, respectively at points T1 , T2 . Then, T1 , T2 and the center O of the homothety that transforms Ω1 into
Ω2 are collinear.
Proof 1: Let O1 , O2 be the respective centers, and R1 , R2 the radii, of Ω1 , Ω2 . Let O0 , r be the center
and radius of ω. Consider triangle O1 O2 O0 , where T1 , T2 are clearly inside segments O1 O0 and O2 O0 . Let P
be the second point where T1 T2 intersects O1 O2 , where by Menelaus’ theorem, we have
O1 P T2 O0 T1 O1 r R1 R1
= · 0 = · = ,
P O2 O2 T2 O T1 R2 r R2
or indeed P is a center of a homothety that transforms O1 into O2 with scaling factor with absolute value
R1
R2 , hence P = O. The Claim 1 follows.
Claim 2: The power of O with respect to ω is invariant for all the possible circles ω which are simulta-
neously tangent to Ω1 , Ω2 .
Proof 2: Since T1 , T2 are points on ω which are collinear with O, the power of O with respect to ω is
OT1 · OT2 . Let P1 , P2 be the respective powers of O with respect to Ω1 , Ω2 , which clearly satisfy P2 = ρ2 P1 .
Consider now points T10 , resulting from applying the homothety to T1 . Clearly, T10 is collinear with O, T1 , hence
on line OT2 ; at the same time, T1 ∈ Ω1 , or T10 ∈ Ω2 by construction, hence T10 , T2 are the two points where a
line through O intersects T2 , or OT10 · OT2 = P2 . But since OT10 = |ρ|OT1 , we have OT1 · OT2 = P|ρ|2 = |ρ|P1 ,
independently on the choice of ω. The Claim 2 follows.
By the Claim 2, the power of O with respect to α1 , α2 is the same, or since AA1 is their radical axis, then
A, A1 , O are collinear, or line AA1 passes through O. Similarly, BB1 , CC1 pass through O. The conclusion
follows.
Also solved by Saturnino Campo Ruiz, Salamanca, Spain; Shohruh Ibragimov, Lyceum Nr.2 under the
SamIES, Samarkand, Uzbekistan; Khakimboy Egamberganov, S.H.Sirojiddinov lyceum, Tashkent, Uzbekistan;
Alessandro Pacanowski, PECI, Rio de Janeiro, Brazil.

a2 b2 a2 a+b+c
√ +√ +√ ≥ .
2
4a + ab + 4b2 2
4b + bc + 4c2 2
4c + ca + 4a2 3
First solution by Marius Stanean

From Hölder’s Inequality, we have
!2 !
X a2 X
√ a2 (4a2 + ab + 4b2 ) ≥ (a2 + b2 + c2 )3 . (1)
cyc 4a2 + ab + 4b2 cyc
But
X
a2 (4a2 + ab + 4b2 ) = 4(a4 + b4 + c4 ) + a3 b + b3 c + c3 a + 4(a2 b2 + b2 c2 + c2 a2 )
cyc
= 4(a2 + b2 + c2 )2 + a3 b + b3 c + c3 a − 4(a2 b2 + b2 c2 + c2 a2 )
13(a2 + b2 + c2 )2 − 12(a2 b2 + b2 c2 + c2 a2 )
≤ ,
4
where the last line follows from Cartoaje’s Inequality,
(a2 + b2 + c2 )2 ≥ 3(a3 b + b3 c + c3 a).
Hence, considering (1), it follows that it is sufficient to prove that
27(a2 + b2 + c2 )3 ≥ [13(a2 + b2 + c2 )2 − 12(a2 b2 + b2 c2 + c2 a2 )](a + b + c)2

9(a2 + b2 + c2 )2 [3(a2 + b2 + c2 ) − (a + b + c)2 ] − 4(a + b + c)2 [(a2 + b2 + c2 )2 − 3(a2 b2 + b2 c2 + c2 a2 )] ≥ 0
9(a2 + b2 + c2 )2 [(a − b)2 + (a − c)(b − c)] − 2(a + b + c)2 [(a + b)2 (a − b)2 + (a + c)(b + c)(a − c)(b − c)] ≥ 0
[9(a2 + b2 + c2 )2 − 2(a + b + c)2 (a + b)2 ](a − b)2 + [9(a2 + b2 + c2 )2 − 2(a + b + c)2 (a + c)(b + c)](a − c)(b − c) ≥ 0.
Without loss of generality, we may assume a ≤ b ≤ c. Then it suffices to show that
9(a2 + b2 + c2 )2 ≥ 2(a + b + c)2 (a + c)(b + c),
but this is true because

3(a2 + b2 + c2 ) ≥ (a + b + c)2
by Cauchy-Schwarz and
3(a2 + b2 + c2 ) − 2(a + c)(b + c) ≥ 0

⇐⇒ c2 − 2(a + b)c + 3(a2 + b2 ) − 2ab ≥ 0
⇐⇒ (c − a − b)2 + 2(a − b)2 ≥ 0,
so we are done.

Second solution by the author
We will begin by finding x, y such that
a2
√ ≥ xa + yb
4a2 + ab + 4b2
for all a, b > 0. Letting a = b = 1, the ”best” such x, y will satisfy x + y = 13 . Note that the inequality is
homogeneous, so letting t = ab , we have
t2 1
√ ≥ xt + y = x(t − 1) + . (1)
2
4t + t + 4 3
The inequality clearly holds if the RHS is negative. Otherwise, squaring and multiplying both sides by 9
yields
9t4
≥ [3x(t − 1) + 1]2
4t2 + t + 4
9t4 − (4t2 + t + 4)[3x(t − 1) + 1]2 ≥ 0.
Considering the LHS as a function of t, say f (t), we want it to have a double root at t = 1. This means
f 0 (1) = 27 − 54x = 0,
1
so x = 2 and y = − 16 . This implies
1
f (t) = (15t − 4)(t − 1)2 .
4
4 4
Clearly, f is positive for t > 15 . For t ≤ 15 , (1) becomes
3t2 3t − 1
√ ≥ ,
2
4t + 4t + 4 2
which must be true since the RHS is negative. Thus,
a2 3a − b
√ ≥
2
4a + ab + 4b2 6
for all a, b > 0. Adding up the similar inequalities with a and c and b and c yields
X a2 X 3a − b
√ ≥
cyc 4a2 + ab + 4b2 cyc
6
a+b+c
= ,
3
as desired.
Also solved by Arkady Alt, San Jose, California, USA; Paolo Perfetti, Università degli studi di Tor
Vergata Roma, Italy; Batzolis, Mandoulides High School, Thessaloniki, Greece; Nicuşor Zlota, “Traian Vuia”
Technical College, Focşani, Romania.

O292. For each positive integer k let
k
X 1
Tk = .
j2j
j=1
Find all prime numbers p for which

p−2
X Tk
≡0 (mod p).
k+1
k=1
Proposed by Gabriel Dospinescu, Ecole Normale Superieure, Lyon

It is known that the following identity holds
n n
n (−1)k (1 − x)j xk X 1
X X
= − .
k jk k2 k2
1≤j≤k≤n k=1 k=1
Let p > 2 be
a prime (for p = 2 the congruence trivially holds).
Then, p−1
k ≡ (−1)k (mod p) and the above identity imply
p−2 p−2 p−2 p−1

X Tk X 1 X 1 X (1/2)j X (1/2)j X (1/2)k
= = = −
k+1 j2j k+1 jk jk k2
k=1 j=1 k=j 1≤j<k≤p−1 1≤j≤k≤p−1 k=1
p−1
p − 1 (−1)k (1 − 1/2)j X (1/2)k
X
≡ − (mod p)
k jk k2
1≤j≤k≤p−1 k=1
p−1 p−1 p−1 p−1
(1/2)k (1/2)k

X X 1 X X 1 1 if p = 3,
= − − =− ≡ (mod p).
k2 k2 k 2 k2 0 if p > 3.
k=1 k=1 k=1 k=1

xyz
O293. Let x, y, z be positive real numbers and let t2 = max(x,y,z) . Prove that
2 3
4 x3 + y 3 + z 3 + xyz ≥ x2 + y 2 + z 2 + t 2 .
Solution by Paolo Perfetti, Università degli studi di Tor Vergata Roma, Italy
The inequality is implied by
(a+b+c)3 3
!
4(x3 + y 3 + z 3 + xyz)2 ≥ x2 + y 2 + z 2 + 27
a+b+c
3
Now define
a + b + c = 3u, ab + bc + ca = 3v 2 , abc = w3
We have
a2 + b2 + c2 = 9u2 − 6v 2 , a3 + b3 + c3 = 27u3 − 27uv 2 + 3w3

The inequality reads as
3
4(27u3 − 27uv 2 + 4w3 )2 ≥ (9u2 − 6v 2 ) + u2
that is
64w6 + 864w3 (−uv 2 + u3 ) + 1836u2 v 4 + 1916u6 − 4032u4 v 2 + 216v 6 ≥ 0

This a convex parabola in w3 whose minimum has abscissa negative. Since w3 ≥ 0, it
follows that the parabola is nonnegative if and only it is nonnegative its value for w3 = 0.
The theory states that, fixed the values of (u, v), the minimum of w3 occurs
for a = 0 or a = b.
If a = 0 we have
479(b6 + c6 ) + 1156b3 c3 ≥ 780(b4 c2 + b2 c4 ) + 150(c5 b + cb5 )

which implied by
479(b6 + c6 ) + 1156b3 c3 ≥ 930(c5 b + cb5 )

This follows by AM-GM since
r
479 6 479 6 2
3 3 3 (479) 578 5
b + c + 578b c ≥ 3 b c > 963b5 c
2 2 4
and the same with (c, b) in place of (b, c).
If a = b we have
(254b4 + 1972b3 c + 525b2 c2 + 658c3 b + 479c4 )(b − c)2 ≥ 0

and the proof is complete.
Also solved by Ioan Viorel Codreanu, Satulung, Maramures, Romania; Daniel Lasaosa, Pamplona, Na-
varra, Spain; Shohruh Ibragimov, Lyceum Nr.2 under the SamIES, Samarkand, Uzbekistan; Ángel Plaza,
University of Las Palmas de Gran Canaria, Spain; Khakimboy Egamberganov, S.H.Sirojiddinov lyceum,
Tashkent, Uzbekistan; Arkady Alt, San Jose, California, USA.

O294. Let ABC be a triangle with orthocenter H and let D, E, F be the feet of the altitudes from A, B and
C. Let X, Y, Z be the reflections of D, E, F across EF , F D, and DE, respectively. Prove that the
circumcircles of triangles HAX, HBY, HCZ share a common point, other than H.

Claim: Let ABC be any triangle, I its incenter, Ia its excenter opposite vertex A, and A0 the symmetric of
A with respect to side BC. Define similarly Ib , Ic and B 0 , C 0 . Then, the circles through I, Ia , A0 , through
I, Ib , B 0 and through I, Ic , C 0 , pass through a second common point other than I.
Proof 1: In exact trilinear coordinates, I ≡ (r, r, r), Ia ≡ (−ra , ra , ra ) and A0 ≡ (−ha , 2ha cos C, 2ha cos B),
or using non-exact trilinear coordinates, we have
I ≡ (1, 1, 1), Ia ≡ (−1, 1, 1), A0 ≡ (−1, 2 cos C, 2 cos B).
The equation of a circle in trilinear coordinates is given by
(`α + mβ + nγ)(aα + bβ + cγ) + k(aβγ + bγα + cαβ) = 0,
where substitution of the coordinates of the three given points yields k = −(` + m + n) when applied to the
incenter, and consequently m + n = 0 and k = −` when applied to the excenter, yielding finally
1 − 2 cos A
m = −n = `
2(cos C − cos B)
when applied to A0 . Indeed, the circle equation

(1 − 2 cos A)(β − γ) aβγ + bγα + cαβ
α+ =
2(cos C − cos B) aα + bβ + cγ
is easily checked to be satisfied by I, Ia , A0 . Analogous equations may be found for the circles through I, Ib , B 0
and I, Ic , C 0 . The intersection of the circles through I, Ib , B 0 and I, Ic , C 0 must clearly satisfy
(1 − 2 cos B)(γ − α) (1 − 2 cos C)(α − β)

β+ =γ+ ,
2(cos A − cos C) 2(cos B − cos A)
or
1 + 2 cos A − 2 cos B − 2 cos C 1 + 2 cos B − 2 cos C − 2 cos A
α+ )β+
(cos A − cos B)(cos A − cos C) (cos B − cos A)(cos B − cos C)
1 + 2 cos C − 2 cos B − 2 cos A
+ γ = 0.
(cos C − cos A)(cos C − cos B)
This line equation represents the radical axis of these two circles, and is clearly satified by I, and since it
is invariant under cyclic permutation of A, B, C and α, β, γ, it is also therefore the equation of the radical
axes of the other two pairs of circles, which intersect the three circles at I, and at another point. The Claim
follows.
Consider triangle DEF . Clearly, X, Y, Z are the reflections of its vertices with respect to its sides, H is
its incenter, and A, B, C its excenters. The conclusion follows by direct application of the Claim.
Also solved by Khakimboy Egamberganov, S.H.Sirojiddinov lyceum, Tashkent, Uzbekistan; Sebastiano

Mosca, Pescara, Italy

Junior problems
J295. Let a, b, c be positive integers such that (a − b)2 + (b − c)2 + (c − a)2 = 6abc. Prove that a3 + b3 + c3 + 1
is not divisible by a + b + c + 1.
Solution by Raffaella Rodoquino and Alessandro Ventullo, Milan, Italy

Clearly,
(a − b)2 + (b − c)2 + (c − a)2 = 2(a2 + b2 + c2 − ab − bc − ac)
which implies
a2 + b2 + c2 − ab − bc − ac = 3abc.
Now, using the identity
a3 + b3 + c3 − 3abc = (a + b + c)(a2 + b2 + c2 − ab − bc − ac)
we obtain that
a3 + b3 + c3 − 3abc = 3abc(a + b + c),
or
a3 + b3 + c3 = 3abc(a + b + c + 1)
which means that a + b + c + 1 divides a3 + b3 + c3 . If a + b + c + 1 divides a3 + b3 + c3 + 1, then a + b + c + 1
divides (a3 + b3 + c3 + 1) − (a3 + b3 + c3 ) = 1, which is impossible, since a + b + c + 1 > 1.
Thus, a3 + b3 + c3 + 1 is not divisible by a + b + c + 1.
Also solved by Daniel Lasaosa, Universidad Pública de Navarra, Spain; Jin Hwan An, Seoul Internatio-
nal School, Seoul, South Korea; Seung Hwan An, Taft School, Watertown, CT, USA; Nikolaos Evgenidis,
student, High-School of Agia,Thessalia, Greece; Sun Mengyue, Lansheng Fudan Middle School, Shanghai,
China; Amedeo Sgueglia, Universitá degli studi di Padova, Italy; Arber Avdullahu, Mehmet Akif College,
Kosovo; Arghya Datta, Hooghly Collegiate School, Kolkata, India; Corneliu Mănescu-Avram, Transporta-
tion High School, Ploieşti, Romania; Debojyoti Biswas, Uttarpara Govt. High School, West Bengal, In-
dia; Dimitris Oikonomou, 2nd Hight School, Nauplio, Greece; Khakimboy Egamberganov, Academic Lyceum
S.H.Sirojiddinov, Tashkent, Uzbekistan; Ilyes Hamdi, Lycée Voltaire, Doha, Qatar; Joshua Benabou, Ma-
nhasset High School, NY, USA; Titu Zvonaru, Comanesti and Neculai Stanciu, Buzau, Romania; Nicuşor
Zlota, “Traian Vuia” Technical College, Focşani, Romania; Prithwijit De, HBCSE, Mumbai, India; Yeongwoo
Hong,Seoul International School, South Korea; Polyahedra, Polk State College, FL, USA; Jishnu Bose, Ut-
tarpara Govt. High School, Kolkata, India; Jeong Ho Ha, Ross School, East Hampton, NY, USA; Ioan Viorel
Codreanu, Satulung, Maramures, Romania; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia.

J296. Several positive integers are written on a board. At each step, we can pick any two numbers u and v,
where u ≥ v, and replace them with u + v and u − v. Prove that after a finite number of steps we can
never obtain the initial set of numbers.
Solution by Khakimboy Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan

Let S0 be the sum of all numbers on board and Si be the sum of all numbers on board after the i-step.
u ≥ v yields to (u + v) + (u − v) ≥ u + v and Sn is a non-decreasing sequence, Si ≤ Si+1 .
If u > v are the selected numbers at the (i + 1)-step, then Si+1 > Si and we can’t obtain the initial set
of numbers anymore. (1)
Thus, at the 1-step we need to consider equal numbers u = v. Hence, after the step, the board has one 0,
all other numbers are positive integers. Due to our result in (1) we can’t pick u together with 0. Therefore,
always 0 has to be put on the board, but the numbers of the initial set are positive integers ⇒ contradiction.
Also solved by Daniel Lasaosa, Universidad Pública de Navarra, Spain; Titu Zvonaru, Comanesti, Roma-
nia and Neculai Stanciu, Buzau, Romania; Yeongwoo Hong, Seoul International School, South Korea; Seong
Kweon Hong, The Hotchkiss School, Lakeville, CT, USA; Debojyoti Biswas, Uttarpara Govt. High School,
West Bengal, India; Arber Avdullahu, Mehmet Akif College, Kosovo; Alessandro Ventullo, Milan, Italy; Dmi-
try Chernyak, Lycée Stanislas, Paris, France; Jeong Ho Ha, Ross School, East Hampton, NY, USA; Jishnu
Bose, Uttarpara Govt. High School, Kolkata, India; Polyahedra, Polk State College, FL, USA; Seung Hwan
An, Taft School, Watertown, CT, USA; Jin Hwan An, Seoul International School, South Korea.

J297. Let a, b, c be digits in base x ≥ 4. Prove that
ab bc ca
+ + ≥ 3,
ba cb ac
where all numbers are written in base x.
Proposed by Titu Zvonaru, Comanesti and Neculai Stanciu, Buzau, Romania
ab bc ca ax + b bx + c cx + a
Note that + + ≥ 3 can be written as + + ≥ 3.
ba cb ac bx + a cx + b ax + c
Since by Cauchy Inequality
!2
P
(ax + b)
X ax + b X (ax + b)2 cyc
= ≥ P =
cyc
bx + a cyc
(ax + b) (bx + a) (ax + b) (bx + a)
cyc
(a + b + c) (x + 1)2
2
(a + b + c)2 (x + 1)2
=
(abx2 + (a2 + b2 ) x + ab) (ab + bc + ca) x2 + 2 (a2 + b2 + c2 ) x + ab + bc + ca
P
cyc
it suffice to prove inequality
(a + b + c)2 (x + 1)2 ≥ 3 (ab + bc + ca) x2 + 2 a2 + b2 + c2 x + ab + bc + ca .

We have
(a + b + c)2 (x + 1)2 − 3 (ab + bc + ca) x2 + 2 a2 + b2 + c2 x + ab + bc + ca =

a2 + b2 + c2 − ab − bc − ca x2 − 4x + 1 ≥ 0 because a2 + b2 + c2 ≥ ab + bc + ca and

x2 − 4x + 1 = x (x − 4) + 1 ≥ 1.
Also solved by Daniel Lasaosa, Universidad Pública de Navarra, Spain; Nicuşor Zlota, “Traian Vuia”
Technical College, Focşani, Romania; Yeongwoo Hong,Seoul International School, South Korea; Khakimboy
Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan; Seong Kweon Hong, The Hot-
chkiss School, Lakeville, CT, USA; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Jeong Ho Ha,
Ross School, East Hampton, NY, USA; Polyahedra, Polk State College, FL, USA; Jin Hwan An, Seoul
International School, Seoul, South Korea; Seung Hwan An, Taft School, Watertown, CT, USA.

J298. Consider a right angle ∠BAC and circles ω1 , ω2 , ω3 , ω4 passing through A. The centers of circles ω1
and ω2 lie on ray AB and the centers of circles ω3 and ω4 lie on ray AC. Prove that the four points of
intersection, other than A, of the four circles are concyclic.
A B
Let σ be the inversion in a circle Σ with center A. Then σ(ω1 ) and σ(ω2 ) are lines perpendicular to AB
and σ(ω3 ) and σ(ω4 ) are lines perpendicular to AC. So the intersection points of these four lines form a
rectangle and thus lie on a circle Ω0 . Hence the other four intersection points of ω1 , ω2 , ω3 , ω4 lie on the circle
Ω = σ(Ω0 ).
Also solved by Nicusor Zlota, “Traian Vuia” Technical College, Focsani, Romania; Titu Zvonaru, Co-
manesti, Romania and Neculai Stanciu, Buzau, Romania; Daniel Lasaosa, Universidad Pública de Navarra,
Spain; Khakimboy Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan; Dimitris Oi-
konomou, 2nd Hight School, Nauplio, Greece; Georgios Batzolis, Mandoulides High School, Thessaloniki,
Greece; by Corneliu Mănescu-Avram, Transportation High School, Ploieşti, Romania; Dmitry Chernyak, Ly-
cée Stanislas, Paris, France; Seong Kweon Hong, The Hotchkiss School, Lakeville, CT, USA; Polyahedra,
Polk State College, USA; Jin Hwan An, Seoul International School, Seoul, South Korea; Seung Hwan An,
Taft School, Watertown, CT, USA.

J299. Prove that no matter how we choose n numbers from the set {1, 2, . . . , 2n}, one of them will be a
square-free integer.
Solution by José Hernández Santiago, México

It suffices to show that, for every n ∈ N, the number of natural numbers in the interval [1, 2n] which are
not square-free is less than n. Since the thesis of the problem is trivially true for n = 1, in what follows we
suppose that n > 1.
Let us denote by Q2n the set of natural numbers in the interval [1, 2n] which are not square-free. Besides,
if p ∈ [1, 2n] is a prime a number, let us denote by M(p2√ ) the set of multiples of p2 which belong to the
interval [1, 2n]. So, if q is the greatest prime number in [1, 2n] then
Q2n = M(22 ) ∪ M(32 ) ∪ M(52 ) ∪ . . . ∪ M(q 2 ). (1)
2n
Since any natural number N has N multiples in the interval [1, 2n], we obtain from (1) that

2n 2n 2n 2n
|Q2n | ≤ 2
+ 2 + 2 + ... + 2
2 3 5 q
2n 2n 2n 2n
≤ + 2 + 2 + ... + 2
22 3 5 q

1 1 1 1
= 2n + + + ... + 2
22 32 52 q
X 1
< 2n (2)
p
p2
where p p12 is the series of the squared reciprocals of the prime numbers. This series is convergent and the
P
π2
well-known fact that ∞ 1
P
n=1 n2 = 6 implies that
∞
π2 X 1
=
6 n2
n=1
X 1 ∞
X 1
> 1+ 2
+ 2
p
p (2n)
n=2
1 π2
X 1
= 1+ + − 1
p
p2 4 6
3 π2 X 1
= + + . (3)
4 24 p
p2
From (2) and (3) we conclude that

π2 3 π2 − 6

|Q2n | < 2n − =n <n
8 4 4
which was what we desired to establish.
Also solved by Daniel Lasaosa, Universidad Pública de Navarra, Spain; Titu Zvonaru, Comanesti and
Neculai Stanciu, Buzau, Romania; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania;
Yeongwoo Hong,Seoul International School, South Korea; Khakimboy Egamberganov, Academic Lyceum
S.H.Sirojiddinov, Tashkent, Uzbekistan; by Corneliu Mănescu-Avram, Transportation High School, Ploieşti,
Romania; Arkady Alt, San Jose, California, USA; Arghya Datta, Hooghly Collegiate School, Kolkata, India;
Arber Avdullahu, Mehmet Akif College, Kosovo; Alessandro Ventullo, Milan, Italy; Seong Kweon Hong, The
Hotchkiss School, Lakeville, CT, USA; Jeong Ho Ha, Ross School, East Hampton, NY, USA; Polyahedra,
Polk State College, FL, USA; Jin Hwan An, Seoul International School, Seoul, South Korea; Seung Hwan
An, Taft School, Watertown, CT, USA.

J300. Let a, b, c be positive real numbers. Prove that
b+c c+a a+b
√ +√ +√ ≥ 1.
2a2 2
+ 16ab + 7b + c 2 2
2b + 16bc + 7c + a 2c + 16ca + 7a2 + b
2
Solution by Sayan Das, Indian Statistical Institute, Kolkata

Note that for positive real x, y we have
p
2x2 + 16xy + 7y 2 ≤ 2x + 3y
since on squaring it is equivalent to 2(x − y)2 ≥ 0. Therefore we have

X b+c X b+c
√ ≥
cyc
2 2
2a + 16ab + 7b + c cyc
2a + 3b + c
Again by Cauchy-Schwarz inequality we have

X b+c X (b + c)2 (2(a + b + c))2
= ≥ =1
2ab + 2ac + 4bc + 3b2 + c2 4 a2 + 8 ab
P P
cyc
2a + 3b + c cyc
cyc cyc
Also solved by Daniel Lasaosa, Universidad Pública de Navarra, Spain; Titu Zvonaru, Comanesti, Ro-
mania and Neculai Stanciu, Buzau, Romania; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani,
Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy; Yeongwoo Hong,Seoul
International School, South Korea; Jishnu Bose, Uttarpara Govt. High School, Kolkata, India; Khakimboy
Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan; Arkady Alt, San Jose, California,
USA; Ángel Plaza, University of Las Palmas de Gran Canaria, Spain; An Zhen-ping, Xianyang Normal Uni-
versity, Xianyang, Shaanxi, China; Sun Mengyue, Lansheng Fudan Middle School, Shanghai, China; Seong
Kweon Hong, The Hotchkiss School, Lakeville, CT, USA; AN-anduud Problem Solving Group, Ulaanbaatar,
Mongolia; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Jeong Ho Ha, Ross School, East Hampton,
NY, USA; Polyahedra, Polk State College, FL, USA; Jin Hwan An, Seoul International School, Seoul, South
Korea; Seung Hwan An, Taft School, Watertown, CT, USA.

Senior problems
S295. Let a, b, c be positive real numbers such that a + b + c = 3. Prove that

√
X (a + b)2
√ ≤ 12.
a 2 − ab + b2
cyc
Solution by Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy
We know that 4(a2 + b2 − ab) ≥ (a + b)2 since (a − b)2 ≥ 0. Consequently,
√
X (a + b)2
2 ≤ 12
cyc
a+b
or
√
X a2 + 2a b + b
2 ≤ 12
cyc
a+b
Moreover,
a2 ab b a
=a− , =1−
a+b a+b a+b a+b
and then we come to
√ √
X a b X a b X ab + a
2(a + b + c) + 6 + 4 = 12 + 4 ≤ 12 + 2
cyc
a+b cyc
a+b cyc
a+b
which is actually
√
X a b X ab + a
2 ≤
cyc
a+b cyc
a+b
The proof concludes by observing that

√
ab + a ≥ 2a b
Also solved by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Daniel Lasaosa, Universidad
Pública de Navarra, Spain; Arber Avdullahu, Mehmet Akif College, Kosovo; Arkady Alt, San Jose, Califor-
nia, USA; Mai Quoc Thang, Ho Chi Minh City, Vietnam; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Peter Tirtowijoyo Young, SMAK St. Louis 1 Surabaya, Indonesia; Sayak Mukherjee, Kol-
kata, India; Titu Zvonaru, Comanesti, Romania and Neculai Stanciu, Buzau, Romania; Philip Radoslavov
Grozdanov, Yambol, Bulgaria.

S296. A ball in Vienna is attended by n ladies (some of which are wearing red dresses) and m gentlemen.
Some ladies and some gentlemen are acquainted. Dancing floor is occupied by acquainted mixed pairs.
At some point during the night, all the present gentlemen were seen on the dancing floor. At some
other time, all the ladies wearing red dresses were on the dancing floor. Show that at some point there
could be all gentlemen and all red-dressed ladies on the dancing floor.
Proposed by Michal Rolinek, Institute of Science and Technology, Vienna
Solution by Daniel Lasaosa, Universidad Pública de Navarra, Spain

Let r be the number of ladies in red dresses. Clearly r ≤ m, since otherwise not all ladies could ever be seen
on the dancing floor simultaneously. If r = m, then all gentlemen must have been on the dancing floor when
all the ladies in red dresses were on it. Otherwise, assume that r < m, and we will find an algorithm to
simultaneously pair all r ladies in red dresses, and all m gentlemen, so that the resulting m pairs containing
the m + r people can be simultaneously on the dancing floor. Pair first all gentlemen with the ladies they
were dancing with, when all m of them where on the dancing floor. If all ladies with red dresses are paired
we are finished, otherwise pick a random lady with a red dress which is not paired, and pair her with the
gentleman she was dancing with when all r ladies with red dresses were on the dancing floor. A new lady is
left unpaired; if she does not wear a red dress, the number of unpaired ladies with red dresses has decreased;
otherwise proceed with this new lady with a red dress as with the previous one. The process of pairing and
unpairing ladies with red dresses cannot continue indefinitely, because 1) there is a finite number of them,
and 2) the process cannot be cyclic since once a lady in a red dress is paired with a gentleman after being
unpaired, he is the gentleman who was dancing with her when all r ladies with red dresses were on the
dancing floor, and no other lady but her will be paired with him ever again. Note that at no point in the
process any gentleman is left unpaired, and the number of unpaired ladies with red dresses decreases. The
conclusion follows.

S297. Let ABC be a triangle and let A1 , A2 , B1 , B2 , C1 , C2 be points that trisect segments BC, CA, AB,
respectively. Cevians AA1 , AA2 , BB1 , BB2 , CC1 , CC2 intersect each other at the vertices of a convex
hexagon that does not have any intersection points inside it. Prove that if the hexagon is cyclic then
our triangle is equilateral.

−−→ −→
Let AB = ~u and AC = ~v . Wlog, we have
−−→ ~u −−→ 2~u −−→ 2~u + ~v
AC1 = , AC2 = , AA1 = ,
3 3 3
−−→ ~u + 2~v −−→ 2~v −−→ ~v
AA2 = , AB1 = , AB2 = .
3 3 3
Note that the vertices of the hexagon are AA1 ∩ BB2 , AA1 ∩ CC2 , BB1 ∩ CC2 , BB1 ∩ AA2 , CC1 ∩ AA2
−−→ −−→
and CC1 ∩ BB2 . For instance, segment AA1 is the set of points X such that AX = ρAA1 for ρ ∈ [0, 1],
−→ −−→ −−→
while segment BB2 is the set of points Y such that AY = κAB + (1 − κ)AB2 for κ ∈ [0, 1]. It follows that
−−−−→
PA,B = AA1 ∩ BB2 occurs for ρ = 53 , κ = 52 , or APA,B = 2~u5+~v . Similarly, we can define PA,C = AA1 ∩ CC2 ,
PB,C = BB1 ∩ CC2 , PB,A = BB1 ∩ AA2 , PC,A = CC1 ∩ AA2 and PC,B = CC1 ∩ BB2 , obtaining after some
algebra
−−−−→ 2~u + ~v −−−−→ 2~u + 2~v −−−−→ ~u + 2~v
APA,C = , APB,C = , APB,A = ,
4 5 4
−−−−→ ~u + 2~v −−−−→ ~u + ~v
APC,A = , APC,B = .
5 4
−−−−−−→
Now, convex quadrilateral PC,B PB,C PB,A PC,A is cyclic iff the sine of the angle between PC,B PB,C and
−−−−−−→ −−−−−−→ −−−−−−→
PC,B PC,A equals the sine between PB,A PC,A and PB,A PB,C , or using the vector product and dropping
irrelevant multiplicative constant factors, we must have
(~u + ~v ) × (3~v − ~u) (−~u − 2~v ) × (3~u − 2~v )
=
|~u + ~v | · |3~v − ~u| |−~u − 2~v | · |3~u − 2~v |
Note now that both numerators equal 4~u × ~v and 8~u × ~v respectively, or this result is equivalent to
(c2 + 4b2 + 4k)(9c2 + 4b2 − 12k) = |−~u − 2~v |2 · |3~u − 2~v |2 = 4 |~u + ~v |2 · |3~v − ~u|2 =
= 4(c2 + b2 + 2k)(9b2 + c2 − 6k),
where we have used that ~u · ~u = c2 , ~v · ~v = b2 and we have denoted k = ~u · ~v . It follows that
4b4 − c4
k = ~u · ~v = .
8(c2 − 2b2 )
Inverting the role of ~u, ~v and b, c, we conclude that quadrilateral PC,B PB,C PA,C PA,B is cyclic iff
4c4 − b4
k = ~u · ~v = ,
8(b2 − 2c2 )
or for both quadrilaterals to be simultaneously cyclic, it is necessary that
0 = 2b6 − 7b4 c2 + 7b2 c4 − 2c6 = b2 − c2 2b4 − 5b2 c2 + 2c4 .

k
In either case, it is not hard to see that cos A = bc must be negative, or ABC must be obtuse at A. By
cyclic symmetry, it follows that ABC must be obtuse at each angle, absurd. Therefore, the hexagon can
never be cyclic. Indeed, drawing an equilateral triangle and forming the described hexagon, it is not hard to
check that it is NOT cyclic, even when ABC is equilateral.
Also solved by Peter Tirtowijoyo Young, SMAK St. Louis 1 Surabaya, Indonesia.

S298. Prove the following identity
a
X a b+k b
(−1)a−k = .
k c c−a
k=0
First solution by G.R.A.20 Problem Solving Group, Roma, Italy

Given a ≥ 0, let
a
a−k a b+k
X
F (b, c) = (−1) .
k c
k=0
n

For c ≥ 0, for b = c − a the identity follows from the fact that for k > n, k = 0:
a
X
a−k a c−a+k
F (c − a, c) = (−1)
k c
k=0

a−a a c−a+a c−a
= (−1) =1= .
a c c−a
Now we prove the full identity by induction on b ≥ c − a.

a
X
a−k a b+1+k
F (b + 1, c) = (−1)
k c
k=0
a
a−k a b+k b+k
X
= (−1) +
k c−1 c
k=0

b b b+1
= F (b, c − 1) + F (b, c) = + = .
c−1−a c−a c−a

Second solution by Albert Stadler, Herrliberg, Switzerland
We express the binomial coefficients in terms of complex integrals, using Cauchy’s theorem:
(1 + u)a (1 + v)b+k
I I
a 1 b+k 1
= = k+1
du, = dv
k 2πi |u|=3 u c 2πi |v|=1 v c+1
So,
a X ∞
X
a−k a b+k a−k a b+k
(−1) = (−1) =
k c k c
k=0 k=0
∞
(1 + u)a (1 + v)b+k
I I
a−k 1 1
X
(−1) du dv =
2πi |u|=3 uk+1 2πi |v|=1 v c+1
k=0
(1 + u)a 1 (1 + v)b
I I
a 1 1
(−1) · dvdu =
2πi |u|=3 u 2πi |v|=1 v c+1 1 +v
1+
u
1
I
(1 + v) b 1
I
(1 + u)a
a
(−1) · dudv =
2πi |v|=1 v c+1 2πi |u|=3 u + v + 1
(1 + v)b (−v)a (1 + v)b

I I
a1 1 b
(−1) c+1
dv = dv = .
2πi |v|=1 v 2πi |v|=1 v c−a+1 c−a
Remark: We have used the residue theorem to deduce that:
(1 + u)a
I
1
du = (−v)a .
2πi |u|=3 u + v + 1
Also solved by G.R.A.20 Problem Solving Group, Roma, Italy; Daniel Lasaosa, Universidad Pública de
Navarra, Spain; Khakimboy Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan; G.
C. Greubel, Newport News, VA, USA; Dimitris Oikonomou, 2nd Hight School, Nauplio, Greece; Arkady Alt,
San Jose, California, USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia.

S299. Let ABCD be a trapezoid with AB k CD and let P an arbitrary point in its plane. If {E} = P D ∩AC,
{F } = P C ∩ BD, {M } = P D ∩ AF and {N } = P C ∩ BE, prove that M N k AB.
Proposed by Mihai Miculita and Marius Stanean, Romania

We prove the more general result that if ABCD is a quadrilateral then AB, CD and M N are concurrent.
Q
A
B
M
D
N
C
R
P F
7 X 0 be a projective transformation such that P 0 M 0 Q0 N 0

Let Q = AF ∩ BE and R = AB ∩ M N . Let f : X →
0 0 0 0 0 0 0 0
is a square. Then A C k M P and B D k N P . Hence, by Menelaus’ theorem,
M 0 R0 B 0 Q0 A0 M 0 D0 M 0 C 0 P 0
− = · = · ,
R0 N 0 N 0 B 0 Q 0 A0 P 0 D0 N 0 C 0
which implies that R = CD ∩ M N as well.
Also solved by Titu Zvonaru, Comanesti, Romania and Neculai Stanciu, Buzau, Romania; Daniel Lasao-
sa, Universidad Pública de Navarra, Spain.

S300. Let x, y, z be positive real numbers and a, b > 0 such that a + b = 1. Prove that
(x + y)3 (y + z)3 ≥ 64abxy 2 z(ax + y + bz)2

Let A, B ∈ (0, π/2) such that tan2 A = x/y and tan2 B = z/y. Then the claimed inequality becomes
2
sec6 A sec6 B ≥ 64ab tan2 A tan2 B a sec2 A + b sec2 B ,
which is equivalent to √
1 ≥ 8 ab sin A sin B a cos2 B + b cos2 A .

Now by the AM-GM inequality,

√ √ 2
8 ab sin A sin B a cos2 B + b cos2 A ≤ 2 ab sin A sin B + a cos2 B + b cos2 A

√ √ 2 2
= 1− a sin B − b sin A ≤ 1,
completing the proof.

U295. Let a be a real number such that (bnac)n≥1 is an arithmetic sequence. Prove that a is an integer.
Solution by Corneliu Manescu-Avram, Transportation High School, Ploiesti, Romania

We have a = bac + {a} and bnac = nbac + bn{a}c, so we can suppose without loss of generality, that
0 ≤ a < 1, since the first terms form an arithmetic sequence.
It remains to prove that f (n) = b(n + 1)ac − bnac is a constant only for a = 0. Indeed, for a 6= 0 there is a
1 1
positive integer m such that m+1 ≤a< m , which implies b(m − 1)ac = bmac = 0, b(m + 1)ac = 1 , so that
f (m − 1) = 0 and f (m) = 1.
For a 6= 0 that function f is nonconstant, which ends the proof.
Also solved by Philip Radoslavov Grozdanov, Yambol, Bulgaria; Daniel Lasaosa, Universidad Pública
de Navarra, Spain; Khakimboy Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan;
Alessandro Ventullo, Milan, Italy; Arghya Datta, Hooghly Collegiate School, Kolkata, India; Arkady Alt,
San Jose, California, USA; Jishnu Bose, Uttarpara Govt. High School, Kolkata, India; Sayak Mukherjee,
Kolkata, India.

U296. Let a and b be real nonzero numbers and let z0 ∈ √ C \ R be a root to the equation z n+1 + az + nb = 0,
n+1
where n is a positive integer. Prove that |z0 | ≥ b.
Solution by Khakimboy Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan

Let z0 = |z0 |(cos α + i sin α) = |z0 |eiα , where sin α 6= 0.
Since the given equation we have
|z0 |n+1 cos(n + 1)α + a|z0 | cos α + nb = 0
and
|z0 |n+1 sin(n + 1)α + a|z0 | sin α = 0.
So
|z0 |n+1 sin nα = nb sin α

and since sin α 6= 0 we get that sin nα 6= 0 and
nb sin α
|z0 |n+1 = (1).
sin nα
We can use by induction and will prove that | sin nα| ≤ n| sin α| and since sin α 6= 0, | sin nα| < n| sin α|.
Hence see (1) and we have
n| sin α|
|z0 |n+1 = |b| ≥ |b|.
| sin nα|
p √
This gives us, |z0 | ≥ n+1 |b| ≥ n+1 b and we are done.
Also solved by Corneliu-Mănescu-Avram, Transportation High School, Ploiesti, Romania.

q
an−1
U297. Let a0 = 0, a1 = 2, and an+1 = 2− an for n ≥ 1. Find limn→∞ 2n an .
Solution by Brian Bradie, Christopher Newport University, Newport News, VA, USA
First, we will show, by induction on n, that
π
an = 2 sin .
2n
Note
π π π
2 sin = 2 sin 0 = 0 = a0 and 2 sin = 2 sin = 2 · 1 = 2 = a1 .
20 21 2
Then
π
v
u
r
an−1 u
u 2 sin n−1
an+1 = 2− = t2 − 2
an π
2 sin n
2
r r
π π π
= 2 − 2 cos n = 4 sin2 n+1 = 2 sin n+1 ,
2 2 2
as required. Finally,
π 

π sin n
lim 2n an = lim 2n+1 sin n = lim 2π · π2  = 2π.
n→∞ n→∞ 2 n→∞
2n
Also solved by Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Daniel Lasaosa, Uni-
versidad Pública de Navarra, Spain; Moubinool Omarjee, Lycée Henri IV, Paris, France; Moubinool Omarjee,
Lycée Henri IV, Paris, France; G. C. Greubel, Newport News, VA; Corneliu Mănescu-Avram, Transportation
High School, Ploieşti, Romania; Arkady Alt, San Jose, California, USA; Alessandro Ventullo, Milan, Italy;
Dmitry Chernyak, Lycée Stanislas, Paris, France; Li Zhou, Polk State College, Winter Haven, FL, USA;
AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; N.J. Buitrago A., Universidade de São Paulo,
São Paulo, SP, Brazil.

U298. Determine all pairs (m, n) of positive integers such that the polynomial
f = (X + Y )2 (mXY + n) + 1
is irreducible in Z[X, Y ].

When m = 0, we have f (X, Y ) = nX 2 + 2nXY + nY 2 + 1. Assume that his polynomial is reducible, or since
its degree is 2, it must be the product of first degree polynomials, ie integers a, b, c and u, v, w exist such that
nX 2 + 2nXY + nY 2 + 1 = (aX + bY + c)(uX + vY + w) =

auX 2 + (av + bu)XY + bvY 2 + (aw + cu)X + (bw + cv)Y + cw.
Since cw = 1 for c, w integers, and we may simultaneously change the signs of a, b, c, u, v, w without altering
the problem, we may assume wlog that c = w = 1, yielding u = −a, v = −b, for a2 = b2 = ab = −n, or the
polynomial is reducible when m = 0 iff n = −k 2 for some nonzero integer k, in which case
f (X, Y ) = −k 2 X 2 − 2k 2 XY − k 2 Y 2 + 1 = (1 + kX + kY ) (1 − kX − kY ) .
For f to be reducible and not of this form, m must be nonzero, which we will assume to hold for during the
rest of this solution.
Assume now that m is nonzero, or the degree of f is 4. If f is reducible, then every factoring polynomial
that is not symmetric on X, Y (because f is symmetric on X, Y ), must be paired with another one, such that
one becomes the other after exchanging X, Y . Since the highest degree terms are X 3 Y, X 2 Y 2 , XY 3 but not
X 4 or Y 4 , then X 2 , Y 2 cannot appear in the same polynomial unless it is symmetric on X, Y , and X 3 , Y 3
cannot appear in any polynomial. We conclude that all possible ways to factor f are (or can be reduced to)
the following:
(aX 2 + bXY + cX + dY + e)(aY 2 + bXY + dX + cY + e) =

ab(X 3 Y + XY 3 ) + (a2 + b2 )X 2 Y 2 + ad(X 3 + Y 3 ) + (ac + bc + bd)(X 2 Y + XY 2 )+
(ae + cd)(X 2 + Y 2 ) + (c2 + d2 + 2be)XY + (c + d)e(X + Y ) + e2 ,
or
(aX 2 + aY 2 + bXY + cX + cY + d)(uXY + vX + vY + w) =

au(X 3 Y + XY 3 ) + buX 2 Y 2 + av(X 3 + Y 3 ) + (av + bv + cu)(X 2 Y + XY 2 )+
(aw + cv)(X 2 + Y 2 ) + (bw + 2cv + du)XY + (cw + dv)(X + Y ) + dw,
or
(aX 2 Y + aXY 2 + bX 2 + bY 2 + cXY + dX + dY + e)(uX + uY + v) =

au(X 3 Y + XY 3 ) + 2auX 2 Y 2 + bu(X 3 + Y 3 ) + (av + bu + cu)(X 2 Y + XY 2 )+
(bv + du)(X 2 + Y 2 ) + (cv + 2du)XY + (dv + eu)(X + Y ) + ve,
and there can be no other, since the product of two polynomials in X, Y , one of which is the result of
exchanging X, Y in the other, becomes a symmetric polynomial in X, Y .

We analyze case by case, identifying the coefficients of corresponding terms and noting that
f (X, Y ) = m(X 3 Y + XY 3 ) + 2mX 2 Y 2 + n(X 2 + Y 2 ) + 2nXY + 1
Case 1: We must have e2 = 1, or since we may exchange signs in all coefficients without altering the
problem, we may assume wlog that e = 1, in which case d = −c, a = n + c2 and b = 2n − 2c2 . It follows that
0 = ac + bc + bd = ac = cn + c3 , yielding either c = 0 or n = −c2 . In the first case, we have a = n, b = 2n,
which result in m = ab = 2n2 and 2m = a2 + b2 = 5n2 , in contradiction with m 6= 0. Therefore, f cannot
be expressed as the product of two polynomials in the first form.
Case 2: We must have dw = 1, or again wlog d = w = 1, yielding v = −c. If c 6= 0, then a = 0 for
m = 0, in contradiction with m 6= 0, or c = v = 0, for a = n and b + u = 2n. Moreover, bu = 2m = 2au,
or since m 6= 0, we have b = 2a = 2n, for u = 0, contradiction. Therefore, f cannot be expressed as the
product of two polynomials in the second form.
Case 3: Note that if u = 0, then m = 0, contradiction, or u 6= 0 yielding b = 0, for du = n and cv = 0.
At the same time, ve = 1, or again v = e = 1, yielding c = 0, and now a = 0, or m = 0, contradiction.
Therefore, f cannot be expressed either as the product of two polynomials in the third and last form.
We conclude that f is irreducible for any nonzero m, and for m = 0 as long as n is not of the form −k 2
for some nonzero integer k.
Also solved by Jishnu Bose, Uttarpara Govt. High School, Kolkata, India.

U299. Let ABC be a triangle with incircle ω and let A0 , B0 , C0 be points outside ω. Tangents
from A0 to ω intersect BC at A1 and A2 . Points B1 , B2 and C1 , C2 are defined similarly.
Prove that A1 , A2 , B1 , B2 , C1 , C2 lie on a conic if and only if triangle ABC and A0 B0 C0
are perspective.
Proposed by Luis Gonzalez, Maracaibo, Venezuela
No solutions have been received yet.

U300. Let f : [a, b] → [a, b] be a function having lateral limits in every point. If
lim f (t) ≤ lim f (t)

t→x− t→x+
for all x ∈ [a, b], prove that there is an x0 ∈ [a, b] such that limt→x0 f (t) = x0 .
Proposed by Dan Marinescu and Mihai Piticari, Romania

Denote
g(x) = lim f (t) − x, h(x) = lim f (t) − x.
t→x+ t→x−
Note that g is defined over [a, b), and h over (a, b]. The proposed result is equivalent to the following: x0
exists such that either x0 = a and g(a) = 0, x0 = b and h(b) = 0, or x0 ∈ (a, b) and g(x0 ) = h(x0 ) = 0. Note
also that g(x) ≥ h(x) at every point in (a, b), where g(x) = h(x) iff f is continuous at x, and that there are
finitely many discontinuity points in (a, b); if there would be infinitely many, there would be a cumulation
point x ∈ [a, b] in whose vecinity there would be infinitely many discontinuities, and at least one of the lateral
limits at such x would not exist. Clearly, g, h are continuous wherever f is continuous. Note finally that,
since f takes values in [a, b], we have g(a) ≥ 0 and h(b) ≤ 0.
Assume that the proposed result is false, or g(a) > 0 and h(b) < 0. Let a < x1 < x2 < · · · < xn < b
be the finitely many discontinuity points in (a, b). Since f is continuous in (a, x1 ), and g(x) = h(x) 6= 0 in
(a, x1 ), we have h(x1 ) > 0. But g(x1 ) > h(x1 ) because x1 is a discontinuity point, or g(x1 ) > 0. For all
i = 1, 2, . . . , n − 1, if g(xi ) > 0, and since f is continuous in (xi , xi+1 ), we have again g(xi+1 ) > h(xi+1 ) > 0,
or after trivial induction g(xn ) > 0, and h(b) > 0. But h(b) < 0, or we have reached a contradiction, hence
the proposed result follows.
Also solved by Sayak Mukherjee, Kolkata, India; Paolo Perfetti, Università degli studi di Tor Vergata
Roma, Roma, Italy.

Olympiad problems
O295. Let a, b, c, x, y, z be positive real numbers such that x + y + z = 1 and
2ab + 2bc + 2ca > a2 + b2 + c2 .
Prove that
2
a(x + 3yz) + b(y + 3xz) + c(z + 3xy) ≤ (a + b + c).
3

Using x + y + z = 1 we get
B = 2(a + b + c) − 3 [a(x + 3yz) + b(y + 3zx) + c(z + 3xy)]

X
= 2(a + b + c)(x + y + z)2 − 3 a [x(x + y + z) + 3yz]
cyc
= 2au + avw + 2bv + bwu + 2cw2 + cuv,

2 2
where u = y − z, v = z − x, and w = x − y. Replacing w by −(u + v), we obtain further
B = (2c + 2a − b)u2 + (5c − a − b)uv + (2b + 2c − a)v 2 .
Now the discriminant
(5c − a − b)2 − 4(2c + 2a − b)(2b + 2c − a) = 9(a2 + b2 + c2 − 2ab − 2bc − 2ca) < 0.
Also, (2c + 2a − b) + (2b + 2c − a) = a + b + 4c > 0. So both 2c + 2a − b > 0 and 2b + 2c − a > 0. Therefore,

B ≥ 0, completing the proof.
Also solved by Daniel Lasaosa, Universidad Pública de Navarra, Spain; Khakimboy Egamberganov, Acade-
mic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani,
Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy.

O296. Let m be a positive integer. Prove that φ(n) divides mn, only for finitely many square-free integers n,
where φ is Euler’s totient function.

Let a ≥ 0 be the largest power of 2 dividing m. Suppose that n = p1 p2 · · · pk is such a square-free integer,
with primes pk > · · · > p1 ≥ 2. Then φ(n) = (p1 − 1) · · · (pk − 1). Since p2 − 1, . . . , pk − 1 are even,
i−1
k − 1 ≤ a + 1, i.e., k ≤ a + 2. Next, we show inductively the claim that pi ≤ (m + 1)2 for 1 ≤ i ≤ k.
Indeed, any prime factor of p1 − 1 is smaller than p1 , so p1 − 1 must divide m, thus p1 ≤ m + 1. Now assume
that the claim is true for all i up to some j < k. Again, all prime factors of pj+1 − 1 are smaller than pj+1 ,
so pj+1 − 1 must divide mp1 · · · pj . Hence,
j−1 j
pj+1 ≤ mp1 · pj + 1 ≤ m(m + 1)(m + 1)2 · · · (m + 1)2 + 1 ≤ (m + 1)2 ,
completing the induction. Therefore, the total number of such n is no greater than
a+1 a+2 −1
(m + 1)(m + 1)2 (m + 1)4 · · · (m + 1)2 = (m + 1)2 .
Also solved by Dimitris Oikonomou, 2nd Hight School, Nauplio, Greece; Arpan Sadhukhan, Indian Stati-
stical Institute, Kolkata, India; Jishnu Bose, Uttarpara Govt. High School, Kolkata, India.

O297. Cells of an 11 × 11 square are colored in n colors. It is known that the number of cells of each color is
greater than 6 and less than 14. Prove that one can find a row and a column whose cells are colored
in at least four different colors.
Solution by Dimitris Oikonomou, 2nd Hight School, Nauplio, Greece

First of all, name each color with one of the the numbers 1, 2, . . . , n. Next, name the cells as follows:
1 2 ... 11
12 13 ... 21
... ... ... ...
111 112 ... 121
Now we will the bound n. Let di be the number of cells in the matrix colored with the i-th color.
Obviously from the condition 7 ≤ di ≤ 13.
Now let A = (ai,j ) be a n × 121 (0, 1)-matrix. The rows represent the colors and the columns represent
the cells. Let ai,j = 1 if the j-th cell is colored by the i-th color, otherwise ai,j = 0. Now since the sum of
the elements of the matrix is the same if we count it either by columns or rows and each column has exactly
one 1 we have:
Xn
di = 121 ⇒ 7n ≤ 121 ≤ 13n ⇒ 10 ≤ n ≤ 17.
i=1
Now we will use the probabilistic method in order to solve the problem. Construct a 21- regular
hyper-graph H with 121 vertices, representing the cells of the matrix and the hyper-edges representing
each combination of a row and a line. Color each vertex of H with one of the n colors as it was in the
matrix(for example 1-st row and the 1-st column is represented by the line which pass from the vertices
(1, 2, . . . , 11, 12, 23, 34, . . . , 111) and if the i-th cell is colored with the j-th color then the i-th vertex will be
colored with the j-th color as well.)
At this point we will use Lovasz Local Lemma(LLL) which states:
Let A1 , A2 , ..., Ak be a series of events such that each event occurs with probability at most p and such that
each event is independent of all the other events except for at most d of them. If ep(d + 1) ≤ 1 then there is
a nonzero probability that none of the events occurs.
Let Af be the event that the hyper-edge f is painted with at most 3 colors. Obviously
n 17
321 321

3 3
p = P(Af ) ≤ ≤ .
n21 1021
0
Moreover, each event Af is clearly mutually independent of all the other events A( f ) for all hyper-edges f 0
that do not intersect f . Now since any edge f of such an H contains 21 vertices, each of which is incident
with 21 edges (including f ), it follows that f intersects at most d = 21(21 − 1) other edges.

421 · e · 17
21
3 ·3 8983769744864520
Now we have ep(d + 1) ≤ 21
< < 1, after some tedious calculations
10 1021
using the obvious inequality e < 3. So according to the Lovasz Local Lemma there is a nonzero probability
that none of the Af occurs which means that there exists a hyper-edge with at least 4 different colors or one
can find a row and a column whose cells are colored in at least four different colors.
Remark: Using exactly the same argument we can prove that there is a row and a column with at least
17 21
4
5 different colors. Indeed, we have p0 ≤ 4 2 , d + 1 = 421,and, using a calculator
10 1
421 · e · 17
21
0 4 ·4 11978801069077172091240802754973
ep (d + 1) ≤ 21
≤ < 1.
10 1021
Note that for 6 different colors this method fails.

O298. Let n be a square-free positive integer. Find the number of functions f : {1, 2, . . . , n} → {1, 2, . . . , n}
such that f (1)f (2) · · · f (n) divides n.

We solve the problem for any positive integer n with prime factorization
pα1 1 pα2 2 · · · pαr r .
Any function f with the required property is uniquely determined by

x(1,j) x(2,j)
f (j) = p1 p2 · · · px(r,j)
r
where x(i, j) are non-negative integers for i = 1, . . . , r and j = 1, . . . , n, such that
x(i, 1) + · · · + x(i, n) ≤ αi , for i = 1, . . . , r.
The number of solutions of the above inequality is n+α i

αi , hence the number of such functions is
r
Y n + αi
.
αi
i=1
If n is square-free then this number is (n + 1)r .
Also solved by Khakimboy Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan; Da-
niel Lasaosa, Universidad Pública de Navarra, Spain; Sayak Mukherjee, Kolkata, India; AN-anduud Problem
Solving Group, Ulaanbaatar, Mongolia; Jishnu Bose, Uttarpara Govt. High School, Kolkata, India.

O299. Let a, b, c be positive real integers such that a2 + b2 + c2 + abc = 4. Prove that
√
1 − abc(3 − a − b − c) ≥ |(a − 1)(b − 1)(c − 1)| .

Note first that if a, b, c ≥ 1, then a2 + b2 + c2 + abc ≥ 4, with equality iff a = b = c = 1, and similarly if
a, b, c ≤ 1. Note also that a, b, c < 2, since if wlog a ≥ 2, then a2 ≥ 4, and b2 + c2 + abc > 0, in contradiction
with a2 + b2 + c2 + abc = 4. Note further that if one of a, b, c equals 1 (wlog c = 1 by symmetry in the
variables), then the proposed inequality rewrites as
√
1 − ab(2 − a − b) ≥ 0,
with condition 3 = a2 + b2 + ab.

√ By the √ AM-GM, it follows that 3 = a2 + b2 + ab ≥ 3ab, or ab = 1 − δ for
some 0 < δ ≤ 1, and a + b = 3 + ab = 4 − δ ≤ 2, with equality iff δ = 0 and equivalently a = b = 1. It
follows that whenever one of a, b, c equals 1, the proposed result holds, with equality iff a = b = c = 1. Any
case where none of a, b, c equals 1 falls necessarily under one of the following two cases:
Case 1: Wlog by symmetry in the variables, a > 1 > b ≥ c. Let us denote a = 1 + δ where 0 < δ < 1,
and d = b − c where 0 ≤ d < 1. Then, the condition rewrites as (1 − bc √ − δ)(3 + δ) = d2 ≥ 0, or bc ≤ 1 − δ
with equality iff b = c, for 1 − abc ≥ 1 − (1 + δ)(1 − δ) = δ 2 , and 1 − abc ≥ δ = a − 1, or it suffices
to show that 3 − a − b − c ≥ (1 − b)(1 − c), or equivalently, bc ≤ 2 − a = 1 − δ, which we √ have already
= c = 2 − a. Direct
proved. The proposed inequality holds in this case, with equality iff bc = 1 − δ, ie iff b √
substitution shows that indeed a2 + b2 + c2 + abc = a2 + 4 − 2a + a(2 − a) = 4, while 1 − abc = a − 1 and
√ 2
3 − a − b − c = 1 − 2 − a = (1 − b)(1 − c).
Case 2: Wlog by symmetry in the variables, a < 1 < b ≤ c. Denoting a = 1 − δ where 0 < δ < 1, and
d = c − b where 0 ≤ √ d < 1, the condition rewrites as (3 − δ)bc = (1 + δ)(3 − δ) − d2 , or bc ≤ 1 + δ with
equality iff b = c, for 1 − abc ≥ δ = 1 − a, and again it suffices to show that 3 − a − b − c ≥ (b − 1)(c − 1),
or bc ≤ √
2 − a = 1 − δ, already proved. Analogously as in case 1, it is readily shown that equality holds iff
b = c = 2 − a by direct substitution in the condition and in the inequality.
√
The conclusion follows, equality holds iff (a, b, c) is a permutation of (k, k, 2−k 2 ) for any real 0 < k < 2.
In the case k = 1, we revert to the trivial equality case a = b = c = 1.

O300. Let ABC be a triangle with circumcircle Γ and incircle ω. Let D, E, F be the tangency points of ω with
BC, CA, AB, respectively, let Q be the second intersection of AD with Γ, and let T be the intersection
of the tangents at B and C with respect to Γ. Furthermore, let QT intersect Γ for the second time at
R. Prove that AR, EF, BC are concurrent.
Proposed by Faraz Masroor, Gulliver Preparatory, Florida, USA
Solution by Cosmin Pohoata, Princeton University, USA

Let EF meet BC at X. Since lines AD, BE, CF are concurrent, the quadruplet (X, B, D, C) is a harmonic
division; hence pencil (AX, AB, AD, AC) is harmonic too. By intersecting it with Γ, it follows that AX ∩ Γ,
B, Q, C are the vertices of a harmonic quadrilateral. But the tangents at B and C with respect to Γ meet
at T , which lies on the line QR, therefore RBQC is a harmonic quadrilateral. It thus follows that R is the
second intersection of AX with Γ. This proves that the lines AR, EF , BC are concurrent, as claimed.
Also solved by Daniel Lasaosa, Universidad Pública de Navarra, Spain; Peter Tirtowijoyo Young, SMAK
St. Louis 1 Surabaya, Indonesia; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Kha-
kimboy Egamberganov, Academic Lyceum S.H.Sirojiddinov, Tashkent, Uzbekistan; Georgios Batzolis, Man-
doulides High School, Thessaloniki, Greece; Dimitris Oikonomou, 2nd Hight School, Nauplio, Greece; Arkady
Alt, San Jose, California, USA.

Junior problems
1 1
J301. Let a and b be nonzero real numbers such that ab ≥ a + b + 3. Prove that
3
1 1
ab ≥ √
3
+√
3
.
a b

√ 1 1 1 1
Let x := 3 ab, y := − √ 3
, z := − √ 3
; then, xyz = 1 and ab ≥ + + 3 becomes equivalent with
a b a b
x3 + y 3 + z 3 ≥ 3. However,
(x + y + z) x2 + y 2 + z 2 − xy − yz − zx = x3 + y 3 + z 3 − 3xyz = x3 + y 3 + z 3 − 3 ≥ 0

and
x2 + y 2 + z 2 ≥ xy + yz + zx ;
therefore,
x + y + z ≥ 0,
which implies
3
3 3 1 1
x ≥ − (y + z) ⇐⇒ x ≥ (−y − z) ⇐⇒ ab ≥ √
3
+√
3
.
a b
Also solved by Polyahedra, Polk State College, USA, Himansu Mookherjee, Kolkata, India; Radouan Bo-
ukharfane, Morocco; Daniel Lasaosa, Pamplona, Spain; David Yang, Bergen County Academies, NJ, USA;
Debojyoti Biswas, Kolkata, India; Eliott S. Kim, The Lawrenceville School, NJ, USA; Jaesung Son, Rid-
gewood, NJ, USA; Jishnu Bose,Indian Statistical Institute, Kolkata, India; Kevin Ren; Corneliu Mănescu-
Avram, Transportation High School, Ploieşti, Romania; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy; Peter C.Shim,
Pingry School, Basking Ridge, NJ, USA; Woosung Jung, Korea International School, South Korea; Hyun-
seo Yang, Daecheong Middle School, Seoul, South Korea; Joshua An, Washington University in St. Louis,
MO, USA; Yeonjune Kang, Peddie School, Hightstown, NJ, USA; Seung Hwan An, Taft School, Watertown,
CT, USA; Joe Hong, Seoul International School, South Korea; Seong Kweon Hong, The Hotchkiss School,
Lakeville, CT, USA; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

J302. Given that the real numbers x, y, z satisfy x + y + z = 0 and
x4 y4 z4
+ + = 1,
2x2 + yz 2y 2 + zx 2z 2 + xy
determine, with proof, all possible values of x4 + y 4 + z 4 .
Solution by Chakib Belgani, Youssoufia, Morocco and Mahmoud Ezzaki, Oujda, Morocco
Notice that 2x2 + yz = x2 + yz − x(y + z) = (x − y)(x − z) and x4 = 12 x2 (2x2 + yz) − 21 x2 yz, so

1 2 xyz x y z
(x + y 2 + z 2 ) − + + = 1.
2 2 (x − y)(x − z) (y − x)(y − z) (z − x)(z − y)
It is clear that
x y z
+ + = 0,
(x − y)(x − z) (y − x)(y − z) (z − x)(z − y)
so
x2 + y 2 + z 2 = 2
and
2
(x + y + z)2 − (x2 + y 2 + z 2 ) = 4(xy + yz + zx)2 = 4(x2 y 2 + y 2 z 2 + z 2 x2 + 2xyz(x + y + z)),
which it is equal to
4(x2 y 2 + y 2 z 2 + z 2 x2 ) = 2 (x2 + y 2 + z 2 )2 − (x4 + y 4 + z 4 )

Thus
4 = 2(4 − (x4 + y 4 + z 4 ))
and the result follows.
Also solved by Polyahedra, Polk State College, USA, Himansu Mookherjee, Kolkata, India; Daniel La-
saosa, Pamplona, Spain; Radouan Boukharfane, Morocco; Hyunseo Yang, Daecheong Middle School, Seoul,
South Korea; Joshua An, Washington University in St. Louis, MO, USA; Seung Hwan An, Taft School,
Watertown, CT, USA; Joe Hong, Seoul International School, South Korea; Seong Kweon Hong, The Hot-
chkiss School, Lakeville, CT, USA; Adnan Ali, Mumbai, India; Arkady Alt, San Jose, California, USA;
David Yang, Bergen County Academies, NJ, USA; Debojyoti Biswas, Kolkata, India; Eliott S. Kim, The
Lawrenceville School, NJ, USA; Jaesung Son, Ridgewood, NJ, USA; Jishnu Bose,Indian Statistical Institute,
Kolkata, India; Corneliu Mănescu-Avram, Transportation High School, Ploieşti, Romania; Peter C.Shim,
Pingry School, Basking Ridge, NJ, USA; Neculai Stanciu, Buzău, Romania and Titu Zvonaru, Comănes, ti,
Romania; Woosung Jung, Korea International School, South Korea; Jin Hyup Hong, Great Neck South High
School, New Hyde Park, NY, USA.

J303. Let ABC be an equilateral triangle. Consider a diameter XY of the circle centered at C which passes
through A and B such that lines AB and XY as well as lines AX and BY meet outside this circle.
Let Z be the point of intersection of AX and BY . Prove that
AX · XZ + BY · Y Z + 2CZ 2 = XZ · Y Z + 6AB 2 .
Solution by Nikolaos Evgenidis, Gymnasium of Agia, Thessalia, Greece

Triangles ABZ and XY Z are similar, so it will be Y Z = 2AZ , XZ = 2BZ and XY = 2AB. Now, since
∠Y XZ + ∠XY Z + ∠BAC = ∠Y XZ + ∠XY Z + ∠Y ZX, it will be ∠Y ZX = 60◦ . It is also
XY 2
2CZ 2 = ZX 2 + ZY 2 − (1)
2
and
XY 2 = XZ 2 + Y Z 2 − 2XZ · Y Z · cos 60◦ ⇔ XZ · ZY = XZ 2 + Y Z 2 − XY 2 (2)
By (1) and (2) the given is written
AX · XZ + BY · Y Z = XY 2 .
But it is
AX · XZ = XZ(XZ − AZ) = XZ 2 − AZ · XZ = 4BZ 2 − AZ · XZ
and
BY · Y Z = Y Z(Y Z − BZ) = Y Z 2 − Y Z · BZ = 4AZ 2 − Y Z · BZ.
By (2), it is
XY 2 = 4BZ 2 + 4AZ 2 − 4BZ · AZ.
So, it now suffices to prove
4BZ · AZ = AZ · XZ + BZ · Y Z(3).
It is known that AZ · XZ = BZ · Y Z, which gives that (3) holds if
4BZ · AZ = 2BZ · Y Z
that is true since Y Z = 2AZ.
Also solved by Polyahedra, Polk State College, USA, Neculai Stanciu, Buzău, Romania and Titu Zvonaru,
Comănes, ti, Romania; Daniel Lasaosa, Pamplona, Spain; Hyunseo Yang, Daecheong Middle School, Seoul,
South Korea; Joshua An, Washington University in St. Louis, MO, USA; Seung Hwan An, Taft School, Wa-
tertown, CT, USA; Joe Hong, Seoul International School, South Korea; Seong Kweon Hong, The Hotchkiss
School, Lakeville, CT, USA; Chakib Belgani, Youssoufia, Morocco and Mahmoud Ezzaki, Oujda, Morocco;
Adnan Ali, Mumbai, India; Arkady Alt, San Jose, California, USA; David Yang, Bergen County Academies,
NJ, USA; Himansu Mookherjee, Kolkata, India; Eliott S. Kim, The Lawrenceville School, NJ, USA; Kevin
Ren; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Woosung Jung, Korea International School,
South Korea; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

√ √
+ b + c = 1. Let M1 be the maximum value of a + b + 3 c and
J304. Let a, b, c be real numbers such that ap
√
let M2 be the maximum value of a + b + 3 c. Prove that M1 = M2 and find this value.
Proposed by Aaron Doman, University of California, Berkeley, USA
Solution by Daniel Lasaosa, Pamplona, Spain √ √

By Lagrange’s multiplier method, the maximum of f (a, b, c) = a + b + 3 c when a + b + c = 1 occurs for a
1 1 1 1
real constant λ such that 1 = λ, 2√ = λ, √ 2 = λ, ie for b = 4 , c =
√ , and consequently a = 34 − 3√ 1
,
b 33c 3 3 3
for
3 1 1 1 5 2
M1 = − √ + + √ = + √
4 3 3 2 3 4 3 3
p √
Similarly, the maximum of g(a, b, c) = a + b + 3 c occurs when 1 = λ, √ 1 √ = λ, and √ 1√ √
3 = λ.
2 b+ 3 c 6 b+ 3 c c2
√
From the second relation we find b + 3 c = 14 , which inserted in the third yields c = 3√ 1
3
, for b = 14 − √13 ,
3 2
and consequently a = 4 + √
3 3
, yielding
3 2 1 5 2
M2 = + √ + = + √ = M1 .
4 3 3 2 4 3 3
The conclusion follows.
Also solved by Polyahedra, Polk State College, USA, Radouan Boukharfane, Morocco; Hyunseo Yang,
Daecheong Middle School, Seoul, South Korea; Joshua An, Washington University in St. Louis, MO, USA;
Seung Hwan An, Taft School, Watertown, CT, USA; Joe Hong, Seoul International School, South Korea;
Seong Kweon Hong, The Hotchkiss School, Lakeville, CT, USA; Arkady Alt, San Jose, California, USA;
David Yang, Bergen County Academies, NJ, USA; Eliott S. Kim, The Lawrenceville School, NJ, USA;
Jishnu Bose,Indian Statistical Institute, Kolkata, India; Kevin Ren; Nicuşor Zlota, “Traian Vuia” Technical
College, Focşani, Romania; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Woosung Jung, Korea
International School, South Korea; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY,
USA.

J305. Consider a triangle ABC with ∠ABC = 30◦ . Suppose the length of the angle bisector from vertex B
is twice the length of the angle bisector from vertex A. Find the measure of ∠BAC.
Solution by Daniel Lasaosa, Pamplona, Spain

Let à , `b be the respective lengths of the angle bisectors from A, B. It is relatively well known (or can easily
be found using Stewart’s theorem and the bisector theorem) that
2bc A 2ca B
à = cos , `b = cos .
b+c 2 c+a 2
It follows, after using the Sine Law and setting C = 180◦ − A − B and B = 30◦ , that `b = 2à is equivalent
to
A B
sin (b + c) = 2 sin (c + a),
2 2

A B−C C −A
sin A cos 60◦ − = sin A cos = 2 sin B cos = cos (75◦ − A) .
2 2 2
In turn this can be further expressed as

◦ ◦ A ◦
cos (75 ) cos A = sin A cos 60 − − cos (15 ) .
2
Now, if A > 90◦ , cos A < 0, hence cos 60◦ − A2 < cos (15◦ ), yielding either 60◦ − A2 > 15◦ , for A < 90◦ ,

contradiction, or 60◦ − A2 < −15◦ , for A > 150◦ = 180 − B, absurd. On the other hand, if A < 90◦ , then
cos A > 0, and 60◦ − A2 < 15◦ , for A > 90◦ , with contradiction again. It follows that A = 90◦ , in which case
clearly both sides are equal to cos (15◦ ), and consequently `b = 2à .
Also solved by Polyahedra, Polk State College USA, Hyunseo Yang, Daecheong Middle School, Seoul,
South Korea; Joshua An, Washington University in St. Louis, MO, USA; Seung Hwan An, Taft School,
Watertown, CT, USA; Joe Hong, Seoul International School, South Korea; David Yang, Bergen County
Academies, NJ, USA; Himansu Mookherjee, Kolkata, India; Eliott S. Kim, The Lawrenceville School, NJ,
USA; Jaesung Son, Ridgewood, NJ, USA; Neculai Stanciu, Buzău, Romania and Titu Zvonaru, Comănes, ti,
Romania; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Woosung Jung, Korea International
School, South Korea; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

J306. Let S be a nonempty set of positive real numbers such that for any a, b, c in S, the number ab + bc + ca
is rational. Prove that for any a and b in S, ab is a rational number.
Solution by Henry Ricardo, New York Math Circle

Suppose that a, b, c ∈ S. Replacing b by a—that is, taking the triplet (a, a, c)—we get a2 +2ac = a(b+2c) ∈ Q.
Now consider the triplet (b, b, c) to get b2 + 2bc = b(b + 2c) ∈ Q. Thus ab = a(b+2c)
b(b+2c ∈ Q.
Also solved by Polyahedra, Polk State College, USA, Daniel Lasaosa, Pamplona, Spain; Hyunseo Yang,
Daecheong Middle School, Seoul, South Korea; Joshua An, Washington University in St. Louis, MO, USA;
Seung Hwan An, Taft School, Watertown, CT, USA; Joe Hong, Seoul International School, South Korea;
Seong Kweon Hong, The Hotchkiss School, Lakeville, CT, USA; Chakib Belgani, Youssoufia, Morocco and
Mahmoud Ezzaki, Oujda, Morocco; Adnan Ali, Mumbai, India; Amedeo Sgueglia, Universitá degli studi di
Padova, Italy; David Yang, Bergen County Academies, NJ, USA; Eliott S. Kim, The Lawrenceville School,
NJ, USA; Jishnu Bose,Indian Statistical Institute, Kolkata, India; Kevin Ren; Corneliu Mănescu-Avram,
Transportation High School, Ploieşti, Romania; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA;
Neculai Stanciu, Buzău, Romania and Titu Zvonaru, Comănes, ti, Romania; Woosung Jung, Korea Interna-
tional School, South Korea; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

Senior problems
S301. Let a, b, c be positive real numbers. Prove that
(a + b + c)(ab + bc + ca)(a3 + b3 + c3 ) ≤ (a2 + b2 + c2 )3 .
Solution by Mai Quốc Thắng, Ho Chi Minh City, Vietnam

Since the inequality is homogeneous, without loss of generality, we can assume that p = a + b + c = 1.
Let
q = ab + bc + ca > 0 ; r = abc > 0
We have
a2 + b2 + c2 = 1 − 2q ; a3 + b3 + c3 = 1 − 3q + 3r ;
and
1 = (a + b + c)2 ≥ 3 (ab + bc + ca) = 3q
The inequality can be rewritten as
q (1 − 3q + 3r) ≤ (1 − 2q)3
Using q 2 = (ab + bc + ca)2 ≥ 3abc (a + b + c) = 3r, we have
q (1 − 3q + 3r) = q − 3q 2 + 3qr ≤ q − 3q 2 + q 3
It suffices to prove
(1 − 2q)3 ≥ q − 3q 2 + q 3
Have
(1 − 2q)3 − q − 3q 2 + q 3 = (1 − q) (3q − 1)2 ≥ 0

The inequality is proved. Equality occurs if and only if a = b = c.
Also solved by Evgenidis Nikolaos, Gimnasium of Agia, Thessalia,Greece; Daniel Lasaosa, Pamplona,
Spain; Yassine Hamdi, Lycée du Parc, Lyon, France; Li Zhou, Polk State College, Winter Haven, FL, USA;
Joshua An, Washington University in St. Louis, MO, USA; Hyunseo Yang, Daecheong Middle School, Seoul,
South Korea; Woosung Jung, Korea International School, South Korea; Neculai Stanciu, Buzău, Romania
and Titu Zvonaru, Comănes, ti, Romania; Sayak Mukherjee, Kolkata, India; Peter C.Shim, Pingry School,
Basking Ridge, NJ, USA; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy; Nicuşor
Zlota, “Traian Vuia” Technical College, Focşani, Romania; Corneliu Mănescu-Avram, Transportation High
School, Ploieşti, Romania; Jishnu Bose,Indian Statistical Institute, Kolkata, India; Henry Ricardo, New
York Math Circle; Himansu Mookherjee, Kolkata, India; Arkady Alt, San Jose, California, USA; Philip
Radoslavov Grozdanov, Yambol, Bulgaria; Radouan Boukharfane, Morocco; Seong Kweon Hong, The Hotch-
kiss School, Lakeville, CT, USA; Joe Hong, Seoul International School, South Korea; Seung Hwan An, Taft
School, Watertown, CT, USA; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

√ √ √
S302. If triangle ABC has sidelengths a, b, c and triangle A0 B 0 C 0 has sidelengths a, b, c, prove that
A B C
sin sin sin = cos A0 cos B 0 cos C 0 .
2 2 2
Solution by Brian Bradie, Christopher Newport University, Newport News, VA

By the Law of Cosines
b2 + c2 − a2 a2 − (b − c)2
cos A = , so that 1 − cos A = ,
2bc 2bc
and r p
A 1 − cos A a2 − (b − c)2
sin = = √ .
2 2 2 bc
Similarly, p p
B b2 − (a − c)2 C c2 − (a − b)2
sin = √ and sin = √ .
2 2 ac 2 2 ab
Thus,
p p p
A B C a2 − (b − c)2 · b2 − (a − c)2 · c2 − (a − b)2
sin sin sin = √ √ √
2 2 2 2 bc · 2 ac · 2 ab
p p p
(b + c − a)2 (c + a − b)2 (a + b − c)2
= √ · √ · √
2 bc 2 ac 2 ab
b+c−a c+a−b a+b−c
= √ · √ · √
2 bc 2 ac 2 ab
= cos A0 cos B 0 cos C 0 .
Also solved by Daniel Lasaosa, Pamplona, Spain; Yassine Hamdi, Lycée du Parc, Lyon, France; Jae-
sung Son, Ridgewood, NJ, USA; Li Zhou, Polk State College, Winter Haven, FL, USA; Woosung Jung,
Korea International School, South Korea; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Nicuşor
Zlota, “Traian Vuia” Technical College, Focşani, Romania; Corneliu Mănescu-Avram, Transportation High
School, Ploieşti, Romania; Jishnu Bose,Indian Statistical Institute, Kolkata, India; Himansu Mookherjee,
Kolkata, India; Debojyoti Biswas, Kolkata, India; Arkady Alt, San Jose, California, USA; Andrea Fan-
chini,Cantu’,Italy; An Zhen-ping, Xianyang Normal University, Xianyang, Shaanxi, China; Adnan Ali,
Mumbai, India; Philip Radoslavov Grozdanov, Yambol, Bulgaria; Joshua An, Washington University in St.
Louis, MO, USA; Hyunseo Yang, Daecheong Middle School, Seoul, South Korea; Chakib Belgani, Youssou-
fia, Morocco and Mahmoud Ezzaki, Oujda, Morocco; Seong Kweon Hong, The Hotchkiss School, Lakeville,
CT, USA; Joe Hong, Seoul International School, South Korea; Seung Hwan An, Taft School, Watertown,
CT, USA; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

1 n
S303. Let a1 = 1 and an+1 = 2 an + an , for n ≥ 1. Find ba2014 c.
Proposed by Marius Cavachi, Romania
Solution by Chakib Belgani, Youssoufia, Morocco and Mahmoud Ezzaki, Oujda, Morocco
By AM-GM we have for every n ≥ 1
1 n √
an+1 = (an + ) ≥ n
2 an
Hence,
1 n
an+1 ≤ (an + √ )
2 n−1
By induction it is straightforward from this that ∀n ≥ 2
2n an+1 ≤ a1 + 2n−1 bn + ... + 2b3 + b2
where
n
bn = √
n−1
and
bn+1 > bn ⇐⇒ n2 > n + 1
so (bn ) is an increasing sequence and
1 + (2n−1 + ... + 1)bn 1 + (2n − 1)bn 1

an+1 ≤ n
= n
< n + bn
2 2 2
so
√ 1 n
n ≤ an+1 ≤ n
+√
2 n−1
we apply this result for n = 2013 to obtain that
ba2014 c = 44
Also solved by Daniel Lasaosa, Pamplona, Spain; Woosung Jung, Korea International School, South
Korea; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Jishnu Bose,Indian Statistical Institute,
Kolkata, India; Henry Ricardo, New York Math Circle; Haroun Meghaichi, University of Science and Tech-
nology Houari Boumediene, Algiers, Algeria; Arkady Alt, San Jose, California, USA; Li Zhou, Polk State
College, Winter Haven, FL, USA; Joshua An, Washington University in St. Louis, MO, USA; Hyunseo
Yang, Daecheong Middle School, Seoul, South Korea; Radouan Boukharfane, Morocco; Seong Kweon Hong,
The Hotchkiss School, Lakeville, CT, USA; Joe Hong, Seoul International School, South Korea; Seung Hwan
An, Taft School, Watertown, CT, USA; Jin Hyup Hong, Great Neck South High School, New Hyde Park,
NY, USA.

S304. Let M be a point inside triangle ABC. Line AM intersects the circumcircle of triangle M BC for the
second time at D. Similarly, line BM intersects the circumcircle of triangle M CA for the second time
at E and line CM intersects the circumcircle of triangle M AB for the second time at F . Prove that
AD BE CF 9
+ + ≥ .
MD ME MF 2
F E
M
C
D
B
By Ptolemy’s theorem, M D · BC = CM · BD + BM · CD. By the law of sines,

BD sin ∠AM B CD sin ∠CM A
= , = .
BC sin ∠BM C BC sin ∠BM C
AM x AD
Hence M D = y+z , where x = AM sin ∠BM C, y = BM sin ∠CM A, and z = CM sin ∠AM B. Thus MD =
AM x+y+z
M D + 1 = y+z . Therefore, by symmetry,

AD BE CF 1 1 1
+ + = (x + y + z) + +
MD ME MF y+z z+x x+y
9 9
≥ (x + y + z) = ,
(y + z) + (z + x) + (x + y) 2
where the inequality is by virtue of the AM-HM inequality.
Also solved by Seung Hwan An, Taft School, Watertown, CT, USA; Joe Hong, Seoul International School,
South Korea; Seong Kweon Hong, The Hotchkiss School, Lakeville, CT, USA; Chakib Belgani, Youssoufia,
Morocco and Mahmoud Ezzaki, Oujda, Morocco; Hyunseo Yang, Daecheong Middle School, Seoul, South
Korea; Joshua An, Washington University in St. Louis, MO, USA; Jishnu Bose,Indian Statistical Institute,
Kolkata, India; Philip Radoslavov Grozdanov, Yambol, Bulgaria; Peter C.Shim, Pingry School, Basking
Ridge, NJ, USA; Woosung Jung, Korea International School, South Korea; Jin Hyup Hong, Great Neck
South High School, New Hyde Park, NY, USA.

S305. Solve in integers the following equation:
x2 + y 2 + z 2 = 2(xy + yz + zx) + 1.

If two of the integers are equal (wlog y = z because of the symmetry in the proposed equation), we need to
solve x(x − 4y) = 1, for x = ±1 since it must divide 1, yielding in either case y = z = 0. Any other solution
must have x, y, z all distinct.
If one of the integers is zero (wlog z = 0), the equation becomes (x − y)2 = 1, with solution iff x, y are
consecutive integers. Any other solution must have x, y, z all nonzero.
Since neither changing signs simultaneously to x, y, z, nor exchanging any two of them, changes the
proposed equation, the previous arguments allow us to conclude that any solution must either be of the
form (x, y, z) = (k, k − 1, 0) or one of its permutations, for any integer k, or must satisfy wlog x > y > z > 0,
or must satisfy wlog x > y > 0 > z.
Case 1: If x > y > 0 > z, note that the equation rewrites
4xy + 1 = (x + y − z)2 ≥ (x + y + 1)2 > (x + y)2 + 1 > 4xy + 1,
the last inequality holding strictly because x, y are distinct. We have reached a contradiction, hence no
solutions exist in this case.
Case 2: If x > y > z > 0, note that the equation rewrites as
(x + y − z − 1)(x + y − z + 1) = 4xy.
Now, x + y − z − 1 < 2x, or x + y − z + 1 > 2y, for x > y + z − 1, hence x = y + z + δ, where δ ≥ 0 is a
non-negative integer. Now,
δ 2 = (x − y − z)2 = 4yz + 1.
n2 −1
Assume now that y, z are any two positive integers such that yz = 4 for some odd positive integer n,
and take x = y + z + n. It follows that
x2 + y 2 + z 2 − 2xy − 2zx = n2 − 2yz = 2yz + 1,
and the equation is clearly satisfied. It follows that all solutions in this case are of the form
p
y + z + 4yz + 1, y, z ,
where y, z are positive integers such that 4yz + 1 is a perfect square.
Restoring generality, all solutions are either a permutation of
(k, k − 1, 0),
where k takes any integral value, or a permutation of
(2k + n)2 − 1 n2 − 1

, ,k ,
4k 4k
n2 −1
where n is any positive odd integral value, and k is any integral divisor (positive or negative) of 4 .
Also solved by Jaesung Son, Ridgewood, NJ, USA; Woosung Jung, Korea International School, South
Kolkata, India; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Himansu Mookherjee, Kolkata, In-
dia; Debojyoti Biswas, Kolkata, India; Philip Radoslavov Grozdanov, Yambol, Bulgaria; Li Zhou, Polk State
College, Winter Haven, FL, USA; Joshua An, Washington University in St. Louis, MO, USA; Radouan
Boukharfane, Morocco; Chakib Belgani, Youssoufia, Morocco and Mahmoud Ezzaki, Oujda, Morocco; Seong
Kweon Hong, The Hotchkiss School, Lakeville, CT, USA; Joe Hong, Seoul International School, South Ko-
rea; Seung Hwan An, Taft School, Watertown, CT, USA; Jin Hyup Hong, Great Neck South High School,
New Hyde Park, NY, USA.

S306. Points M, N, K lie on sides BC, CA, AB of a triangle ABC, respectively and are different from its
vertices. Triangle M N K is called beautiful if ∠BAC = ∠KM N and ∠ABC = ∠KN M . If in triangle
ABC there are two beautiful triangles with a common vertex, prove that triangle ABC is right.
Solution by Marius Stanean, Zalau, Romania

Suppose that there are two beautiful triangles (M N K and M N1 K1 ) with a common vertex on the side BC.
Is obviously that the two triangles are similar to the triangle ABC.
There is a spiral similarity (centered at M ) swapping triangles 4M N K ∼ 4M N1 K1 , i.e. there is another
spiral similarity centered at M swapping triangles 4M N N1 ∼ 4M KK1 . Therefore ∠M N N1 = ∠M KK1
and ∠M N1 N = ∠M K1 K hence M N AK and M N1 AK1 are cyclic quadrilaterals, so ∠BAC + ∠N M K =
180◦ ⇐⇒ ∠BAC = 90◦ .
Now, let N0 , K0 the feet of the perpendiculars to AC, AB trough M . We have 4M N N0 ∼ 4M KK0
and from this it follows that
M K0 MK AC M K0 M N0
= = ⇐⇒ =
M N0 MN AB AC AB
M K0 MB M N0 MC
But = and = , this means that M must be the middle of BC. In this situation, any
AC BC AB BC
right triangle in M , is a beautiful triangle.
Also solved by Daniel Lasaosa, Pamplona, Spain; Seung Hwan An, Taft School, Watertown, CT, USA;
Joshua An, Washington University in St. Louis, MO, USA; Li Zhou, Polk State College, Winter Haven,
FL, USA; Philip Radoslavov Grozdanov, Yambol, Bulgaria; Jaesung Son, Ridgewood, NJ, USA; Jin Hyup
Hong; Shatlyk Mamedov, Dashoguz, Turkmenistan; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA;
Woosung Jung, Korea International School, South Korea.

U301. Let x, y, z, t > 0 such that x ≤ 2, x + y ≤ 6, x + y + z ≤ 12, and x + y + z + t ≤ 24. Prove that
1 1 1 1
+ + + ≥ 1.
x y z t
First solution by Brian Bradie, Christopher Newport University, Newport News, VA

Applying Jensen’s inequality to the function f (w) = 1/w for w > 0 yields

1 1 1 1 1 x 1 y 1 z 1 t
+ + + = f + f + f + f
x y z t 2 2 4 4 6 6 12 12

x y z t
≥ f + + +
4 16 36 144
144
=
36x + 9y + 4z + t
144
=
27x + 5(x + y) + 3(x + y + z) + (x + y + z + t)
144
≥ = 1.
27(2) + 5(6) + 3(12) + 24
Second solution by Khakimboy Egamberganov, Tashkent, Uzbekistan

We have that x1 ≥ 12 , x + y2 + y2 ≤ 6 and by Cauchy-Schwartz, x1 + y2 + y2 ≥ 3
2 and 1
x + 4
y ≥ 32 . Similarly, we
can x1 + y4 + z9 ≥ 3 and x1 + y4 + z9 + 36
t ≥ 6. Hence
1 1 1 4 3 1 4 9 1 4 9 36
≥ , + ≥ , + + ≥ 3, + + + ≥ 6.
x 2 x y 2 x y z x y z t
So
1 1 1 1 1 1 4 1 4 9 1 4 9 36
36 + + + = 27 +5 + +3 + + + + + + ≥ 36
x y z z x x y x y z x y z t
and
1 1 1 1
+ + + ≥ 1.
x y z t
Korea; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Paolo Perfetti, Università degli studi di Tor
Vergata Roma, Roma, Italy; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Kevin Ren;
Jishnu Bose,Indian Statistical Institute, Kolkata, India; Jaesung Son, Ridgewood, NJ, USA; Hyunseo Yang,
Daecheong Middle School, Seoul, South Korea; Khakimboy Egamberganov, Tashkent, Uzbekistan; Mai Quốc
Thắng, Ho Chi Minh City, Vietnam; Himansu Mookherjee, Kolkata, India; Arkady Alt, San Jose, California,
USA; An Zhen-ping, Xianyang Normal University, Xianyang, Shaanxi, China; Adnan Ali, Mumbai, India;
Corneliu Mănescu-Avram, Transportation High School, Ploieşti, Romania; Radouan Boukharfane, Morocco;
Li Zhou, Polk State College, Winter Haven, FL, USA; Chakib Belgani, Youssoufia, Morocco and Mahmoud
Ezzaki, Oujda, Morocco; Joe Hong, Seoul International School, South Korea; Seung Hwan An, Taft School,
Watertown, CT, USA; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

U302. Let a be a real number. Evaluate
r q p
a− a2 − a4 − a8 − . . ..

The problem’s conclusion is flawed, because the expression is divergent for all a 6= 0. Indeed, if a > 0,
then the sequence is
√ q √ p
a, a − a = 0, a − a2 − a4 = a − a2 − a2 = a, . . . ,
2
that is, it alternates between a and 0. Likewise, if a < 0, then the sequence is
√ q √ p
a, a − a2 = 2a, a − a2 − a4 = a − a2 − a2 = a, ...,
that is, it alternates between a and 2a.
Also solved by Juan José Granier Torres, Universidad de Chile, Santiago, Chile; Seung Hwan An, Taft
School, Watertown, CT, USA; Joe Hong, Seoul International School, South Korea; Chakib Belgani, Yous-
soufia, Morocco and Mahmoud Ezzaki, Oujda, Morocco; Li Zhou, Polk State College, Winter Haven, FL,
USA; Radouan Boukharfane, Morocco; Haroun Meghaichi, University of Science and Technology Houari
Boumediene, Algiers, Algeria; Jaesung Son, Ridgewood, NJ, USA; Moubinool Omarjee, Lycée Henri IV,
Paris France; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Woosung Jung, Korea International
School, South Korea; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

U303. Let p1 , p2 , . . . , pk be distinct primes and let n = p1 p2 · · · pk . For each function f : {1, 2, . . . , n} →
{1, 2, . . . , n}, denote Pf (n) = f (1)f (2) · · · f (n).
(a) For how many functions f are n and Pf (n) relatively prime?
(b) For how many functions f is gcd(n, Pf (n)) a prime?
Proposed by Vladimir Cerbu and Mihai Piticari, Romania

(a) Let [n] = {1, 2, . . . , n}, A = {a ∈ [n] : (a, n) = 1}, and F be the set of all such functions. Then
|A| = φ(n) = (p1 − 1) · · · (pk − 1), where φ is Euler’s totient function. Now for each f ∈ F , f (i) ∈ A for each
i ∈ [n]. So |F | = (φ(n))n .
(b) Let A1 = {a ∈ [n] : (a, n) = p1 } and F1 = {f ∈ F : (n, Pf (n)) = p1 }. Then |A1 | = φ(n/p1 ) =
(p2 − 1) · · · (pk − 1). Now if f ∈ F1 , then f (i) ∈ A ∪ A1 for each i ∈ [n], but not f (i) ∈ A for all i ∈ [n].
Therefore,
pn1

n n n n n
|F1 | = |A ∪ A1 | − |A| = (|A| + |A1 |) − |A| = (φ(n)) −1 .
(p1 − 1)n
Hence the desired answer is
k
!
n
X pni
(φ(n)) −k .
(pi − 1)n
i=1
Kolkata, India; Jaesung Son, Ridgewood, NJ, USA; Chakib Belgani, Youssoufia, Morocco and Mahmoud
Ezzaki, Oujda, Morocco; Joe Hong, Seoul International School, South Korea; Seung Hwan An, Taft School,
Watertown, CT, USA; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

U304. No solutions have yet been received.

U305. Let (an )n≥1 be a sequence of positive real numbers such that a1 + a2 + . . . + an < n2 for all n ≥ 1.
Prove that
1 1 1
lim + + ... + = ∞.
n→∞ a1 a2 an
Proposed by Mihai Piticari Campulung, Moldovenesc, Romania
Solution by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain

By rearranging the sequence if necessary, we may suppose that (an )n≥1 is an increasing sequence. We argue
by contradiction.

1 1 1
Let us suppose that lim + + ... + < ∞. This implies by comparison with the harmonic
n→∞ a1 a2 an
n
series that lim = 0. Now, since the last limit exists, the following limit exists as well
n→∞ an
n2 2n − 1
lim = lim =0
n→∞ a1 + a2 + . . . + an n→∞ an
by the Cezaro-Stolz criteria.
n2
But this contradicts with the fact that 1 ≤ .
a1 + a2 + . . . + an
South Korea; Chakib Belgani, Youssoufia, Morocco and Mahmoud Ezzaki, Oujda, Morocco; Li Zhou, Polk
State College, Winter Haven, FL, USA; Brian Bradie, Christopher Newport University, Newport News, VA;
Salem Malikic, Simon Fraser University, Burnaby, BC, Canada; Radouan Boukharfane, Morocco; Khakim-
boy Egamberganov, Tashkent, Uzbekistan; Henry Ricardo, New York Math Circle; Jaesung Son, Ridgewood,
NJ, USA; Jishnu Bose,Indian Statistical Institute, Kolkata, India; Juan José Granier Torres, Universidad
de Chile, Santiago, Chile; Moubinool Omarjee, Lycée Henri IV, Paris France; Nicuşor Zlota, Traian Vuia
Technical College, Focşani, Romania; Paolo Perfetti, Universitá degli studi di Tor Vergata Roma, Roma,
Italy; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Stanescu Florin, School Serban Cioculescu,
Gaesti, Dambovita, Romania; Woosung Jung, Korea International School, South Korea; Haroun Meghaichi,
University of Science and Technology Houari Boumediene, Algiers, Algeria; Jin Hyup Hong, Great Neck
South High School, New Hyde Park, NY, USA.

U306. Let n be a natural number. Prove the identity
n
2k+1 4n+1 ∞
Z
X 1
π= 2k
+ 2n dx
k n 1 (1 + x2 )n+1
k=1 k
and derive the estimate

X 2k+1 n
2 4
n
√ <π− 2k
< n√ .
2 n k k 2 n
k=1

Let
n ∞
2k+1 4n+1
Z
X 1
Pn = , In = dx, Rn = 2n In .
k 2k (1 + x2 )n+1

k=1 k 1 n
Using the substitution x = tan θ and integration by parts, we get

Z π/2 π/2
Z π/2
2n
θdθ = cos2n−1 θ sin θ π/4 + (2n − 1) cos2n−2 θ sin2 θdθ

In = cos
π/4 π/4
1
= − + (2n − 1) (In−1 − In ) .
2n
Hence 1 + 2n+1 nIn = 2n (2n − 1)In−1 for all n ≥ 1, with I0 = π4 . Since I1 = − 14 + π8 , it is easy to see that
P1 + R1 = π. As an induction hypothesis, assume that π = Pn−1 + Rn−1 for some n ≥ 2. Then
4n 2n+1 n

1
π = Pn−1 + Rn−1 = Pn−1 + 2n−2 + In = Pn + Rn ,
n−1
2n (2n − 1) 2n (2n − 1)
completing the induction.

To prove the estimate, we shall use the well-known Wallis’ inequality

1 1 1 2n 1 1
p <q < n <q <√ .
π(n + 1) π n + 12
4 n π n + 14
3n + 1
First,
Z π/2 Z π/2
1 1
In = (1 + cos 2θ)n dθ = (1 − cos z)n dz.
2n π/4 2n+1 0
Using the well-known refinement of Kober’s inequality that 1 − cos z > (2z/π)2 for 0 < z < π/2, we get
2n Z π/2
2 π 3
2n+1 In > z 2n dz = > .
π 0 2(2n + 1) 2(2n + 1)
Thus √
12 3n + 1 2
π − Pn = Rn > n+2 ≥ n√ ,
2 (2n + 1) 2 n
p 2
where the last inequality follows from 3 (3n + 1)n − (4n + 2)2 = (11n + 4)(n − 1) ≥ 0.
Next, it is easy to check directly the upper bound of Rn for n = 1, 2, 3. Consider n ≥ 4. Note that
3
1 − cos z < 2π z for 0 < z < π3 and 1 − cos z < π3 z − 12 for π3 < z < π2 . Therefore
n Z π/3 n Z π/2
n+1 3 n 3 π n π 1 11π
2 In < z dz + z− dz = 1 + n+1 ≤ .
2π 0 π π/3 6 3(n + 1) 2 32(n + 1)

Hence p
44π π(n + 1) 44 4
Rn < n+1
< p < n√ ,
32(n + 1)2 2 n+1 32(n + 1) 2 n
√
where the second inequality follows from π 3 < 32 and the last inequality follows from 11 < 2 32.
South Korea; Radouan Boukharfane, Morocco; Jishnu Bose,Indian Statistical Institute, Kolkata, India; Peter
C.Shim, Pingry School, Basking Ridge, NJ, USA; Paolo Perfetti, Università degli studi di Tor Vergata Roma,
Roma, Italy; Li Zhou, Polk State College, Winter Haven, FL, USA; Jin Hyup Hong, Great Neck South High
School, New Hyde Park, NY, USA.

Olympiad problems
O301. Let a, b, c, d be nonnegative real numbers such that a2 + b2 + c2 + d2 = 4. Prove that

a b c d
+ + + ≤ 1.
b+3 c+3 d+3 a+3
Solution by Daniel Lasaosa, Pamplona, Spain and

We use in this solution the following
Lemma: For non-negative reals a, b, c, d, we have
p
a2 + b2 + c2 + d2 a2 + b2 + c2 + d2 ≥ 2(abc + bcd + cda + dab),

with equality iff a = b = c = d.

Proof: The proposed inequality is equivalent to
3
a2 + b2 + c2 + d2 ≥ 4(abc + bcd + cda + dab)2 ,
or to
a6 + b6 + c6 + d6 + 2 a2 b2 c2 + b2 c2 d2 + c2 d2 a2 + d2 a2 b2 +

+3 a4 b2 + a2 b4 + b4 c2 + b2 c4 + c4 d2 + c2 d4 + d4 a2 + d2 a4 + a4 c2 + a2 c4 + b4 d2 + b2 d4 ≥

≥ 8 a2 b2 cd + b2 c2 da + c2 d2 ab + d2 a2 bc + a2 c2 bd + b2 d2 ca .

Note now that, by the AM-GM inequality, we have a4 c2 + b4 d2 ≥ 2a2 b2 cd, 2a6 + 2b6 + c6 + d6 ≥ 6a2 b2 cd,
and a2 b2 c2 + d2 a2 b2 ≥ 2a2 b2 cd, with equality in all of them simultaneously, iff a = b = c = d. Using these
inequalities and the result of permuting in them a, b, c, d, the Lemma follows. Note that in this particular
problem, the Lemma results in abc + bcd + cda + dab ≤ 4, with equality iff a = b = c = d = 1.
After some algebra, the inequality proposed in the problem statement is equivalent to
3(ab + bc + cd + da + ac + bd) + 2(abc + bcd + cda + dab) + abcd ≤ 27.
Now, ab+bc+cd+da ≤ a2 +b2 +c2 +d2 because of the scalar product inequality applied to vectors (a, b, c, d)
2 2 2 2
and (b, c, d, a), while ac ≤ a +c
2 and bd ≤ b +d
2 because of the AM-GM inequality, or
a2 + b2 + c2 + d2
ab + bc + cd + da + ac + bd ≤ 3 = 6,
2
with equality iff a = b = c = d = 1. Moreover, by the QM-GM inequality, we find
r
a2 + b2 + c2 + d2 √
4
1= ≥ abcd,
4
or abcd ≤ 1, with equality iff a = b = c = d = 1. Finally, by the Lemma we have abc + bcd + cda + dab ≤ 4.
The conclusion follows, equality holds iff a = b = c = d = 1.
Also solved by Seung Hwan An, Taft School, Watertown, CT, USA; Chakib Belgani, Youssoufia, Morocco
and Mahmoud Ezzaki, Oujda, Morocco; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Jishnu
Bose,Indian Statistical Institute, Kolkata, India; Woosung Jung, Korea International School, South Korea.

O302. Let ABC be an isosceles triangle with AB = AC and let M ∈ (BC) and N ∈ (AC) such that
∠BAM = ∠M N C. Suppose that lines M N and AB intersect at P . Prove that the bisectors of angles
BAM and BP M intersect at a point lying on line BC.
Solution by Marius Stanean, Zalau, Romania

Let A0 reflection of A across BC. Because ∠BAM = ∠M N C and ∠ABM = ∠N CM we have 4ABM ∼
4N CM and from this it follows that ∠N M C = ∠AM B. Therefore ∠A0 M B = ∠AM B = ∠N M C so
A0 , M, N, P are collinear points.
Now in triangle A0 BP , because BC is angle bisector of ∠A0 BP we deduce that the other two angle
bisectors intersect on BC. Let X this point (i.e. the incenter of 4A0 BP ). But the fact that A0 is the
reflection of A with respect to BC, it follows that AX is angle bisector of ∠BAM .
Also solved by Mriganka Basu Roy Chowdhury, South Point High School, Kolkata, West Bengal, India;
Daniel Lasaosa, Pamplona, Spain; Li Zhou, Polk State College, Winter Haven, FL, USA; Neculai Stanciu,
Buzău, Romania and Titu Zvonaru, Comănes, ti, Romania; Saturnino Campo Ruiz, Salamanca, Spain; Jishnu
Bose,Indian Statistical Institute, Kolkata, India; Jaesung Son, Ridgewood, NJ, USA; Khakimboy Egamber-
ganov, Tashkent, Uzbekistan; Himansu Mookherjee, Kolkata, India; Arkady Alt, San Jose, California, USA;
Andrea Fanchini,Cantu’,Italy; Adnan Ali, Mumbai, India; Peter C.Shim, Pingry School, Basking Ridge, NJ,
USA; Philip Radoslavov Grozdanov, Yambol, Bulgaria; Joe Hong, Seoul International School, South Korea;
Seung Hwan An, Taft School, Watertown, CT, USA; Jin Hyup Hong, Great Neck South High School, New
Hyde Park, NY, USA.

O303. Let a, b, c be real numbers greater than 2 such that
1 1 1 1
+ 2 + 2 = .
a2 −4 b −4 c −4 7
Prove that
1 1 1 3
+ + ≤ .
a+2 b+2 c+2 7
Solution by Mai Quốc Thắng, Ho Chi Minh City, Vietnam

With t > 2, we have
1 9 1
≤ +
t+2 10 (t2 − 4) 10
Because
9 1 1 (t − 5)2
+ − = ≥0
10 (t2 − 4) 10 t + 2 10 (t + 2) (t − 2)
With a, b, c > 2, we have

1 1 1 9 1 1 1 3 3
+ + ≤ · 2
+ 2 + 2 + =
a+2 b+2 c+2 10 a −4 b −4 c −4 10 7
Equality holds clearly when a = b = c = 5.
Also solved by Seung Hwan An, Taft School, Watertown, CT, USA; Daniel Lasaosa, Pamplona, Spain;
Joe Hong, Seoul International School, South Korea; Chakib Belgani, Youssoufia, Morocco and Mahmoud Ez-
zaki, Oujda, Morocco; Li Zhou, Polk State College, Winter Haven, FL, USA; Philip Radoslavov Grozdanov,
Yambol, Bulgaria; Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Adnan Ali, Mumbai, India; An
Zhen-ping, Xianyang Normal University, Xianyang, Shaanxi, China; Arkady Alt, San Jose, California, USA;
Khakimboy Egamberganov, Tashkent, Uzbekistan; Jaesung Son, Ridgewood, NJ, USA; Jishnu Bose,Indian
Statistical Institute, Kolkata, India; Shatlyk Mamedov, Dashoguz, Turkmenistan; Corneliu Mănescu-Avram,
Transportation High School, Ploieşti, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma,
Roma, Italy; Salem Malikic, Simon Fraser University, Burnaby, BC, Canada; Woosung Jung, Korea In-
ternational School, South Korea; Zarif Ibragimov, Samarkand, Uzbekistan; Yassine Hamdi, Lycée du Parc,
Lyon, France; Jin Hyup Hong, Great Neck South High School, New Hyde Park, NY, USA.

O304. Let C1 and C2 be non-intersecting circles centered at O1 and O2 . One common external tangent of
these circles touches Ci at Pi (i = 1, 2). The other common external tangent touches Ci at Qi (i = 1, 2).
Denote by M the midpoint of Q1 Q2 . Let Pi M intersect Ci at Ri and R1 R2 intersect Ci again at Si
(i = 1, 2). P1 S1 intersects P2 S2 at A. The tangent to C1 at R1 and the tangent to C2 at R2 intersect
at B. Prove that AB ⊥ O1 O2 .
Solution by Li Zhou, Polk State College, Winter Haven, FL, USA

A
B
P1 P2
O1
S1
R1 O2
Q1 R2
M
S2
Q2
By definition, M P1 · M R1 = M Q21 = M Q22 = M P2 · M R2 . So P1 , P2 , R2 , R1 are concyclic. Therefore

∠AP1 P2 = ∠P1 R1 S1 = ∠P1 P2 R2 = ∠P2 S2 R2 . Likewise ∠AP2 P1 = ∠P1 S1 R1 . Hence 4AP1 P2 ∼ 4AS2 S1 ,
and thus AP1 · AS1 = AP2 · AS2 . Therefore AM is the radical axis of C1 and C2 , which implies that
AM ⊥ O1 O2 . Moreover
∠BR1 R2 = ∠S1 P1 R1 = ∠P1 AP2 = ∠S2 P2 R2 = ∠BR2 R1 ,
so BR1 = BR2 , that is, B is on the radical axis AM as well. This completes the proof.
South Korea; Chakib Belgani, Youssoufia, Morocco and Mahmoud Ezzaki, Oujda, Morocco; Daniel Lasaosa,
Pamplona, Spain; Mriganka Basu Roy Chowdhury, South Point High School, Kolkata, West Bengal, India;
Peter C.Shim, Pingry School, Basking Ridge, NJ, USA; Marius Stanean, Zalau, Romania; Jin Hyup Hong,
Great Neck South High School, New Hyde Park, NY, USA.

O305. Prove that for any positive integers m and a, there is a positive integer n such that an + n is divisible
by m.

Let a = uv, where all prime factors of u divide m, and all prime factors of v are coprime with m. If v = 1,
then an is a multiple of m for all sufficiently large n, and it suffices to take n equal to a sufficiently large
multiple of m. We will assume henceforth that there are prime factors of a which are coprime with m.
Lemma: Given a positive integer a, for all positive integer m, there exist arbitrarily large positive integers
n0 , n1 , n2 , . . . , nm−1 such that
ani + ni ≡ i (mod m).
Proof: If m = 1, the result is trivially true. If m = 2, it suffices to take n0 arbitrarily large with the
same parity as a, and n1 arbitrarily large with opposite parity to a. Assume that the result is true for
1, 2, . . . , m − 1, and let k be the periodicity of the remainders of an modulus m for sufficiently large n, ie
k is the least positive integer such that an ≡ an+k (mod m) for all sufficiently large n. This clearly exists
because writing a = uv as above, for all sufficiently large n the remainder of un modulus m0 is constant,
where m0 is the product of all factors of m which are not coprime with a, and the remainders of un modulus
m n
m0 , and of v modulus m, are periodic because of the Euler-Fermat theorem, which applies because u is
m
coprime with m 0 , and v is coprime with m. Let d be the greatest common divider of m and k, or since
k ≤ ϕ(m) < m, where ϕ denotes Euler’s totient function, then d < m, and by hypothesis of induction there
exist arbitrarily large positive integers n0 , n1 , . . . , nd−1 such that, for each i = 0, 1, . . . , d − 1, we have
ani + ni ≡ i (mod d).

m
Define µ = d, clearly an integer, and consider the integers
ani + ni , ani +k + ni + k, ani +2k + ni + 2k, ... ani +(µ−1)k + ni + (µ − 1)k,
which are clearly congruent modulus m with the integers
ani + ni , ani + ni + k, ani + ni + 2k, ... ani + ni + (µ − 1)k.
If we repeat this process for each i = 0, 1, . . . , d − 1, we generate dµ = m such integers. Assume that any
two of them are congruent modulus m. If this happens for the same value of i, then δk is a multiple of m
for some 0 < δ ≤ µ − 1. Since d is the greatest common divider of k, m, then kδ is a multiple of m iff dδ is a
multiple of m, absurd since 0 < dδ < dµ = m. Otherwise, there exist indices i, j and not necessarily distinct
factors ci , cj ∈ {0, 1, . . . , µ − 1} such that
(ani + ni − anj − nj ) + (ci − cj )k ≡ 0 (mod m).
Now, ani + ni and anj + nj correspond to distinct remainders modulus d, or the first term in the LHS is
not a multiple of d, but the second one is, since it contains a factor of k, yielding a contradiction. Since we
have thus generated m integers with distinct remainders modulus m, each remainder modulus m appears
exactly once in these integers. The Lemma follows.
The conclusion to the proposed problem clearly follows from taking i = 0 for the desired value of m in
the Lemma.
Also solved by Arber Igrishta, Eqrem Qabej, Vushtrri, Kosovo; Radouan Boukharfane, Morocco; Li Zhou,
Polk State College, Winter Haven, FL, USA; Khakimboy Egamberganov, Tashkent, Uzbekistan; Navid Sa-
fei,Student of Sharif,University of Technoogy in Policy Making of Science and Technology, Iran; Jishnu
Bose,Indian Statistical Institute, Kolkata, India.

O306. Let ABC be a triangle with incircle γ and circumcircle Γ. Let Ω be the circle tangent to rays AB,
AC, and to Γ externally, and let A0 be the tangency point of Ω with Γ. Let the tangents from A0 to γ
intersect Γ again at B 0 and C 0 . Finally, let X be the tangency point of the chord B 0 C 0 with γ. Prove
that the circumcircle of triangle BXC is tangent to γ.
Let I be the center of γ, D the tangency point of γ with √ BC, and E the reflection of D across AI. Let
I be the inversion in the circle with center A and radius AB · AC. Then I(Γ) is the line tangent to γ at
E and I(Ω) = γ. So I(A0 ) = E. Now we need to prove a lemma.
Lemma. Suppose AI and A0 I intersect Γ again at M and M 0 , respectively. Then M M 0 passes through
the midpoint S of ID.
Proof. We use homogeneous barycentric coordinates and set A = (1 : 0 : 0), B = (0 : 1 : 0), and
0 0
C
= (0 : 0 : 1).
2 2
Then with the usual notations, I = (a : b : c), D = (0 : s − c : s − b), and D = AA ∩ BC =
b
0 : s−c c
: s−b . Hence the equation of AD0 is c2 (s − c)y − b2 (s − b)z = 0. Solving this with the equation of

b2 c2
Γ : a2 yz + b2 zx + c2 xy = 0, we get A0 = −a : s−c : s−b . Also, AI has equation cy − bz = 0. Solving this

a2
with the equation of Γ we get M = − b+c : b : c . Next, A0 I has equation
bc(b − c) ac(s − b + c) ab(s − c + b)

− x+ y− z = 0.
(s − b)(s − c) (s − a)(s − b) (s − a)(s − c)

a2
Solving this with the equation of Γ we get M 0 = (b−c)(s−a) b
: s−b+c c
: − s−c+b . Finally, S has coordinates

1 a b c s−c s−b
= a2 : 2s(s − c) + ab : 2s(s − b) + ac ,

, , + 0, ,
2 2s 2s 2s a a
so M S has equation
b2 − c2 (s − a)x − a2 (s − b + c)y + a2 (s − c + b)z = 0.

Then it is easy to see this equation is satisfied by the coordinates of M 0 . This proves the lemma.
Now let O be the center of Γ and T the internal center of similitude of Γ and γ (X55 in Kimberling’s
Encyclopedia of Triangle Centers [ETC]). Then it is well known that the Gergonne point of 4ABC (X7 in
ETC) is the isogonal conjugate (with respect to 4ABC) of T . Hence T = IO ∩ AA0 . Since the Gergonne
point of 4A0 B 0 C 0 is also the isogonal conjugate (with respect to 4A0 B 0 C 0 ) of T , A0 X is the reflection of
A0 A across A0 I. Thus I is the incenter of 4AQA0 , where Q = AD ∩ A0 X. Hence BC, B 0 C 0 , and IQ concur
at a point P . Next, it is easy to see that M and M 0 are the centers of the circumcircles Σ and Σ0 of 4BCI
and 4B 0 C 0 I, respectively. Note that P is the radical center of Γ, Σ, and Σ0 . Finally, let R be the second
intersection point of Σ and Σ0 . Then M M 0 is the perpendicular bisector of IR, so the lemma implies that R
is on DX as well. Therefore, P is the radical center of Γ, Σ, Σ0 , and γ, from which the required conclusion
follows immediately.

Junior problems
J307. Prove that for each positive integer n there is a perfect square whose sum of digits is equal to 4n .
Solution by Albert Stadler, Herrliberg, Switzerland

Let aj = 10j and consider
 2
Xk k
X X
aj 
m=  10 = 102aj + 2 10ai +aj
j=1 j=1 1≤i≤j≤k
m is a perfect square whose only digits in the base 10 representation are only zeros, ones and twos, since
ai + aj = ar + as , 1 ≤ i ≤ j ≤ k , 1 ≤ r ≤ s ≤ k, implies i = r, j = s. The sum of the digits of m equals
2k(k − 1)
k+ = k2 .
2
This result goes beyond what is asserted, since we have produced for a given natural number k a perfect
square m whose digit sum equals k 2 . In particular, if k = 2n , then k 2 = 4n .
Also solved by Daniel Lasaosa, Pamplona, Spain; Bedri Hajrizi, Gjimnazi "Frang Bardhi", Mitrovicë,
Kosovë; Corneliu Mănescu- Avram, Transportation High School, Ploieşti, Romania; Henry Ricardo, New
York Math Circle; Alessandro Ventullo, Milan, Italy; Zachary Chase, University School of NSU, FL, USA;
Jaesung Son, Ridgewood, NJ, USA; Jongyeob Lee, Stuyvesant High School, NY, USA; Yeonjune Kang,
Peddie School, Hightstown, NJ, USA; William Kang, Bergen County Academies, Hackensack, NJ, USA;
Chaeyeon Oh, Episcopal High School, Alexandra, VA, USA; Kyoung A Lee, The Hotchkiss School, Lakevi-
lle, CT, USA; Seung Hwan An, Taft School, Watertown, CT, USA; Polyahedra, Polk State College, USA;
Seonmin Chung, Stuyvesant High School, New York, NY, USA; Alyssa Hwang, Kent Place School Summit,
NJ, USA; Woosung Jung, Korea International School, South Korea; Daniel Jhiseung Hahn, Phillips Exeter
Academy, Exeter, NH, USA; Seong Kweon Hong,The Hotchkiss School, Lakeville, CT; Radouan Boukhar-
fane, Sidislimane, Morocco; Farrukh Mukhammadiev, Academic lyceum under the SamIES Nr.1, Samarkand,
Uzbekistan.

J308. Are there triples (p, q, r) of primes for which (p2 − 7)(q 2 − 7)(r2 − 7) is a perfect square?
Solution by Michael Tang, Edina High School, MN, USA

We claim that there is no triple (p, q, r) of primes so that S = (p2 − 7)(q 2 − 7)(r2 − 7) is a perfect square.
Suppose for the sake of contradiction that there is such a triple. If, say, p = 2, then p2 − 7 < 0, so we
must have (q 2 − 7)(r2 − 7) < 0, since S must be nonnegative. However, if q, r ≥ 3, then q 2 − 7, r2 − 7 > 0,
a contradiction, so one of q, r must equal 2. Without loss of generality, take q = 2. Then the expression
becomes
S = (22 − 7)(22 − 7)(r2 − 7) = 9(r2 − 7),
so r2 −7 must be a perfect square. Let r2 −7 = k 2 for some nonnegative integer k. Rearranging and factoring,
we get (r − k)(r + k) = 7. Since r + k is positive and greater than r − k, we must have r + k = 7 and r − k = 1,
which gives (r, k) = (4, 3). But r is prime, a contradiction.
Therefore, none of p, q, r can equal 2, so they must all be odd. Then, by a well-known result, p2 , q 2 , r2 ≡ 1
(mod 8). (To see this, write p = 2n − 1, giving p2 = 4n2 − 4n + 1 = 4n(n − 1) + 1, where n(n − 1) is clearly
even, so 4n(n − 1) + 1 ≡ 0 + 1 = 1 (mod 8).) Thus, p2 − 7, q 2 − 7, r2 − 7 ≡ 2 (mod 8). This means that the
power of 2 in each of p2 − 7, q 2 − 7, r2 − 7 is exactly one (since they are of the form 8m + 2 = 2(4m + 1),
twice an odd integer), so the power of 2 in their product, S, is exactly 3. But the power of 2 must be even
for S to be a perfect square, a contradiction.
Therefore, we conclude that there is no triple (p, q, r) of primes so that (p2 − 7)(q 2 − 7)(r2 − 7) is a perfect
square.
Also solved by Adnan Ali, A.E.C.S-4, Mumbai, India; Daniel Lasaosa, Pamplona, Spain; Ali Baouan,
Rabat, Morocco; Jaesung Son, Ridgewood, NJ, USA; Ioan-Andrei Nicolae, Romania; Jongyeob Lee, Stuyve-
sant High School, NY, USA; Yeonjune Kang, Peddie School, Hightstown, NJ, USA; William Kang, Bergen
County Academies, Hackensack, NJ, USA; Chaeyeon Oh, Episcopal High School, Alexandra, VA, USA;
Kyoung A Lee, The Hotchkiss School, Lakeville, CT, USA; Seung Hwan An, Taft School, Watertown, CT,
USA; Polyahedra, Polk State College, USA; Seonmin Chung, Stuyvesant High School, New York, NY, USA;
Alyssa Hwang, Kent Place School Summit, NJ, USA; Woosung Jung, Korea International School, South
Korea; Arber Avdullahu, Mehmet Akif College, Kosovo; Arkady Alt, San Jose, CA, USA; Ilyes Hamdi, Ly-
cée Voltaire, Doha, Qatar; Evgenidis Nikolaos, M. N. Raptou High School, Larissa, Greece; Prithwijit De,
HBCSE, Mumbai, India; Alessandro Ventullo, Milan, Italy; Zachary Chase, University School of NSU, FL,
USA; Radouan Boukharfane, Sidislimane, Morocco.

J309. Let n be an integer greater than 3 and let S be a set of n points in the plane that are not the vertices of
a convex polygon and such that no three are collinear. Prove that there is a triangle with the vertices
among these points having exactly one other point from S in its interior.
First solution by Daniel Lasaosa, Pamplona, Spain

We solve the problem by induction on n. The base case is clearly n = 4, and if 4 points in the plane are not
the vertices of a convex polygon, and no 3 are collinear, then three of them form a triangle which has the
fourth in its interior, or the proposed result is clearly true. Assume that the result is true for 4, 5, . . . , n − 1,
and consider n points in the plane which are not the vertices of a convex polygon, and such that no three
are collinear. There is a convex polygon, with vertices among the n points, such that the rest of the n points
are in its interior. We consider three cases:
• If the convex polygon is a triangle ABC, consider any point P of the remaining n − 3, which is clearly
in its interior, and consider triangles P AB, P BC, P CA. Since n ≥ 5, at least one of these triangles,
wlog P AB, has at least one point in its interior. Apply now the hypothesis of induction to the set
formed by the vertices of P AB and all points in its interior, which contains at least 4 and at most
n − 1, and the proposed result is true in this case.
• If the convex polygon is not a triangle and three consecutive of its vertices A, B, C are such that none
of the n points are in the interior of triangle ABC, consider the set of n − 1 points once B is removed,
and apply the hypothesis of induction to this set, and the proposed result is also true in this case.
• If the convex polygon is not a triangle and three consecutive of its vertices A, B, C are such that at
least one of the n points is in the interior of triangle ABC, consider the set of points formed by A, B, C,
and all points inside triangle ABC, which clearly contains at least 4 points and at most n − 1, and
apply the hypothesis of induction to this set, and the proposed result is true again in this case.
Second solution by Alessandro Ventullo, Milan, Italy

We use the following theorem.
Carathéodory’s Theorem. If x ∈ Rd lies in the convex hull of a set S, there is a subset S 0 of S consisting of
d + 1 or fewer points such that x lies in the convex hull of S 0 .
Let x ∈ S such that x lies in the interior of the convex hull of S. Hence, by Caratheodory’s Theorem
(d = 2), there is a subset S 0 of S having at most 3 points such that x lies in the convex hull of S 0 . Since the
points are not collinear, then |S 0 | 6= 2 and since x lies in the interior of the convex hull of S, then |S 0 | 6= 1.
So, |S 0 | = 3 and there exists a triangle T whose vertices are in S which contains x. If x is the unique point
of S in T , we are done. If there is another point y 6= x of S in T , joining x with the three points in S 0 we
have that y lies in the interior of one among these triangles. Proceeding in this way and using the fact that
S has finitely many points, then there exist a triangle whose vertices are in S which has exactly one point
of S in its interior.
Also solved by Jaesung Son, Ridgewood, NJ, USA; Jongyeob Lee, Stuyvesant High School, NY, USA; Ye-
onjune Kang, Peddie School, Hightstown, NJ, USA; William Kang, Bergen County Academies, Hackensack,
NJ, USA; Chaeyeon Oh, Episcopal High School, Alexandra, VA, USA; Kyoung A Lee, The Hotchkiss School,
Lakeville, CT, USA; Seung Hwan An, Taft School, Watertown, CT, USA; Polyahedra, Polk State College,
USA; Adnan Ali, A.E.C.S-4, Mumbai, India; Arber Igrishta, Eqrem Qabej, Vushtrri, Kosovo; Ioan-Andrei
Nicolae, Romania; Zachary Chase, University School of NSU, FL, USA.

J310. Alice puts checkers in some cells of an 8 × 8 board such that:
a) there is at least one checker in any 2 × 1 or 1 × 2 rectangle.

b) there are at least two adjacent checkers in any 7 × 1 or 1 × 7 rectangle.
Find the least amount of checkers that Alice needs to satisfy both conditions.

First, the two placements of 37 checkers below satisfy both conditions.
Now each row requires at least 4 checkers. Also, if a row has only 4 checkers, then the two squares at the
ends of the row must be both empty. Thus two such 4-checker rows cannot be adjacent, to avoid two empty
1 × 2 rectangles in the first and last columns. So there cannot be 5 or more 4-checker rows. If there are 3 or
fewer 4-checker rows (such as in the left figure above), then Alice needs at least s × 2 + 5 × 5 = 37 checkers.
Next, suppose that there are 4 nonadjacent 4-checker rows. Then the two adjacent checkers in the first and
last columns must appear in the same rows (such as the fourth and fifth rows in the right figure above).
Now between the 4 checkers in the first and last columns, these two adjacent rows require at least 7 more
checkers in the middle columns: 6 for the 6 1 × 2 rectangles and 1 more to ensure condition b). Again, Alice
needs at least 4 × 4 + 2 × 5 + (4 + 7) = 37 checkers.
Also solved by Ioan-Andrei Nicolae, Romania; Jaesung Son, Ridgewood, NJ, USA; Jongyeob Lee, Stuy-
vesant High School, NY, USA; Yeonjune Kang, Peddie School, Hightstown, NJ, USA; William Kang, Bergen
County Academies, Hackensack, NJ, USA; Chaeyeon Oh, Episcopal High School, Alexandra, VA, USA; Kyo-
ung A Lee, The Hotchkiss School, Lakeville, CT, USA; Seung Hwan An, Taft School, Watertown, CT, USA;
Seonmin Chung, Stuyvesant High School, New York, NY, USA; Alyssa Hwang, Kent Place School Summit,
NJ, USA; Woosung Jung, Korea International School, South Korea; Ilyes Hamdi, Lycée Voltaire, Doha,
Qatar; Zachary Chase, University School of NSU, FL, USA; Adnan Ali, A.E.C.S-4, Mumbai, India.

J311. Let a, b, c be real numbers greater than or equal to 1. Prove that
a b2 + 3 b c2 + 3 c a2 + 3

+ + ≥ 3.
3c2 + 1 3a2 + 1 3b2 + 1
Solution by Arkady Alt, San Jose, CA, USA

By AM-GM Inequality
v v
X a(b2 + 3) uY
u a(b2 + 3) uY
u a(a2 + 3)
≥ 3 3
t = 3 3
t ≥3
cyc
3c2 + 1 cyc
3c2 + 1 cyc
3a2 + 1
since
a(a2 + 3)
≥ 1 ⇐⇒ (a − 1)3 ≥ 0.
3a2 + 1
Also solved by Bedri Hajrizi, Gjimnazi "Frang Bardhi", Mitrovicë, Kosovë; Daniel Lasaosa, Pamplona,
Spain; Jaesung Son, Ridgewood, NJ, USA; Jongyeob Lee, Stuyvesant High School, NY, USA; Yeonjune
Kang, Peddie School, Hightstown, NJ, USA; William Kang, Bergen County Academies, Hackensack, NJ,
USA; Chaeyeon Oh, Episcopal High School, Alexandra, VA, USA; Kyoung A Lee, The Hotchkiss School, La-
keville, CT, USA; Seung Hwan An, Taft School, Watertown, CT, USA; Polyahedra, Polk State College, USA;
Alyssa Hwang, Kent Place School Summit, NJ, USA; Woosung Jung, Korea International School, South Ko-
rea; Daniel Jhiseung Hahn, Phillips Exeter Academy, Exeter, NH, USA; Seong Kweon Hong,The Hotchkiss
School, Lakeville, CT; Michael Tang, Edina High School, MN, USA; Adnan Ali, A.E.C.S-4, Mumbai, India;
Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Ilyes Hamdi, Lycée Voltaire, Doha, Qatar; Pri-
thwijit De, HBCSE, Mumbai, India; Alessandro Ventullo, Milan, Italy; Zachary Chase, University School
of NSU, FL, USA; George Gavrilopoulos, Nea Makri High School, Athens, Greece; Ioan Viorel Codreanu,
Satulung, Maramures, Romania; Florin Stanescu, Cioculescu Serban High School, Gaesti, Romania; Rado-
uan Boukharfane, Sidislimane, Morocco; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma,
Italy; Moubinool Omarjee, Lycée Henri IV, Paris, France.

J312. Let ABC be a triangle with circumcircle Γ and let P be a point in its interior. Let M be the midpoint
of side BC and let lines AP, BP, CP intersect BC, CA, AB at X, Y, Z, respectively. Furthermore, let
line Y Z intersect Γ at points U and V . Prove that M, X, U, V are concyclic.

AZ
Denote ρ = ZB and κ = CYY A . Clearly, by Ceva’s theorem ρκ = 1 iff M = X. In this case, by Thales’ theorem
Y Z k BC, and parallel chords BC and U V in the circumcircle of BC have a common perpendicular bisector,
which is an axis of symmetry. A circle with center in this axis through U, V , and tangent to BC at M = X
clearly exists, which is the limiting case of the general case with ρκ 6= 1, which we study next.
Assume wlog (since we may invert the roles of B, C without altering the problem) that ρκ > 1, ie line Y Z
intersects line BC at a point T such that B is inside segment T C, and by Menelaus’ theorem, T C = ρκT B.
Since T C = T B + BC, it follows that
BC ρκ BC BC(ρκ + 1)
TB = , TC = , TM = + TB = ,
ρκ − 1 ρκ − 1 2 2(ρκ − 1)
BC
and since by Ceva’s theorem we have CX = ρκBX, or BX = ρκ+1 , it follows that
2ρκBC ρκBC 2
T X = BX + T B = , TM · TX = = T B · T C.
ρ2 κ2 − 1 (ρκ − 1)2
Now, since U V is a common chord of the circumcircle of ABC and the circumcircle of XU V , it is also their
radical axis, or the power of T with respect to the circumcircle of ABC (which clearly equals T B · T C) also
equals the power of T with respect to the circumcircle of XU V , and since this power equals T M · T X, the
circumcircle of XU V also passes through M . The conclusion follows.
Also solved by Jaesung Son, Ridgewood, NJ, USA; Jongyeob Lee, Stuyvesant High School, NY, USA;
Yeonjune Kang, Peddie School, Hightstown, NJ, USA; William Kang, Bergen County Academies, Hacken-
sack, NJ, USA; Chaeyeon Oh, Episcopal High School, Alexandra, VA, USA; Kyoung A Lee, The Hotchkiss
School, Lakeville, CT, USA; Seung Hwan An, Taft School, Watertown, CT, USA; Polyahedra, Polk State
College, USA; George Gavrilopoulos, Nea Makri High School, Athens, Greece; Nicuşor Zlota, "Traian Vuia"
Technical College, Focşani, Romania; Prithwijit De, HBCSE, Mumbai, India; Zachary Chase, University
School of NSU, FL, USA; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania.

Senior problems
S307. Let ABC be a triangle such that ∠ABC − ∠ACB = 60◦ . Suppose that the length of the altitude from
A is 14 BC. Find ∠ABC.
Proposed by Omer Cerrahoglu and Mircea Lascu, Romania
Solution by Miguel Amengual Covas, Cala Figuera, Mallorca, Spain

Let a and h be lengths of the side BC and the altitude from A respectively. Then the area of 4ABC may
1 2 sin B sin(B−60◦ )
be expressed as 21 ah = 81 a2 , and also as 12 a2 sin B sin C
sin(B+C) = 2 a sin(2B−60◦ ) .
Therefore
sin B sin (B − 60◦ ) 1
◦
=
sin (2B − 60 ) 4
and
sin (2B − 60◦ ) = 4 sin B sin (B − 60◦ )
= −2 [cos (2B − 60◦ ) − cos 60◦ ]
= −2 cos (2B − 60◦ ) + 1
i.e.
sin (2B − 60◦ ) + 2 cos (2B − 60◦ ) = 1 (1)
Let t = tan (B − 30◦ ). Then
2t 1 − t2
sin (2B − 60◦ ) = and cos (2B − 60◦
) = (2)
1 + t2 1 + t2
Substitute (2) into (1) and get
3t2 − 2t − 1 = 0
and the only valid solution is t = 1. We conclude that B − 30◦ = 45◦ , and therefore ∠ABC = 75◦
Also solved by George Tsapakidis, "Panagia Prousiotissa" High School, Agrinio, Greece; Daniel Lasa-
osa, Pamplona, Spain; Li Zhou, Polk State College, Winter Haven, FL, USA; Titu Zvonaru, Comănes, ti,
Romania and Neculai Stanciu, Buzău, Romania; Yassine Hamdi, Lycée du Parc, Lyon, France; Alessan-
dro Ventullo, Milan, Italy; Prithwijit De, HBCSE, Mumbai, India; Nicuşor Zlota, "Traian Vuia" Technical
College, Focşani, Romania; Jongyeob Lee, Stuyvesant High School, NY, USA; Corneliu Mănescu- Avram,
Transportation High School, Ploieşti, Romania; Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India;
Arkady Alt, San Jose, CA, USA; Alok Kumar, Delhi, India; Adnan Ali, A.E.C.S-4, Mumbai, India; Ioan
Viorel Codreanu, Satulung, Maramures, Romania; Ali Baouan, Rabat, Morocco; Alyssa Hwang, Kent Place
School Summit, NJ, USA; Woosung Jung, Korea International School, South Korea; Seung Hwan An, Taft
School, Watertown, CT, USA; Kyoung A Lee, The Hotchkiss School, Lakeville, CT, USA; Chaeyeon Oh,
Episcopal High School, Alexandra, VA, USA; William Kang, Bergen County Academies, Hackensack, NJ,
USA; Radouan Boukharfane, Sidislimane, Morocco.

S308. Let n be a positive integer and let Gn be an n × n grid with the number 1 written in each of its
unit squares. An operation consists of multiplying all entries of a column or all entries of a row by
−1. Determine the number of distinct grids that can be obtained after applying a finite number of
operations on Gn .

Operations as defined are clearly commutative and associative. Note further that applying twice (or any
even number of times) an operation over the same row, or over the same column, leaves the grid unchanged.
Therefore, all possible sequences of a finite number of operations on Gn are equivalent to choosing any
number of the n rows and n columns, and multiplying exactly the chosen rows and columns by −1, but not
the rest. This results in a priori 22n distinct grids. However, note that two different choices of rows and
columns may provide the same final grid. Indeed, we say that two sequences of operations are complementary
if the rows and columns which are multiplied by −1 in the first sequence, are exactly those that are not
multiplied by −1 in the second sequence. Note that applying two complementary sequences leaves the grid
unchanged, since each number is multiplied by −1 exactly twice, once when its row is chosen, and once when
its column is chosen. It follows that two complementary sequences yield the same grid.
Assume now that two sequences yield the same grid. For every position in the grid where a 1 appears,
either both its row and column, or niether its row and column, were chose, whereas for every position where
a −1 appears, either its row was chosen but not its column, or its column was chosen but not its row. It
follows that, if a row was chosen in both sequences, then every column was either chosen in both sequences,
or not chosen in both sequences, and similarly all rows, or both sequences are the same. On the other hand,
if a row was chosen in one sequence but not in the other, then all columns chosen in the first sequence
were not chosen in the second, and vice versa, and similarly with all rows, resulting in both sequences being
complementary.
We conclude that for every grid, there are exactly two sequences that produce it, which are complemen-
tary, and the total number of distinct grids is therefore 22n−1 .
Also solved by Erlang Wiratama Surya, Ipeka International, Indonesia; Li Zhou, Polk State College,
Winter Haven, FL, USA; Kyoung A Lee, The Hotchkiss School, Lakeville, CT, USA; Seung Hwan An, Taft
School, Watertown, CT, USA.

S309. Let ABCD be a circumscriptible quadrilateral, which lies strictly inside a circle ω. Let ωA be the circle
outside of ABCD that is tangent to AB, AD, and to ω at A0 . Similarly, define B 0 , C 0 , D0 . Prove that
lines AA0 , BB 0 , CC 0 , DD0 are concurrent.
Proposed by Khakimboy Egamberganov, Tashkent, Uzbekistan

Let the circle k is incircle of ABCD and point P is insimilicenter of the circles ω and k. Now, we will prove
that the lines AA0 , BB 0 , CC 0 , DD0 are pass through the point P .
For the line AA0 , we have :
0
A is exsimilicenter of the circles ω and ωA ;
A is insimilicenter of the circles k and ωA ;
Since P is the insimilicenter of the circles ω and k and by Monge -D’Alemberts’ circles theorem we can
that A0 , A and P are collinear. The line AA0 passes through P . Anologously, the lines BB 0 , CC 0 and DD0
are pass through P . Hence AA0 , BB 0 , CC 0 and DD0 are concurrent at the point P , which the insimilicenter
of the circles k and ω and we are done.
Also solved by Li Zhou, Polk State College, Winter Haven, FL, USA; Yassine Hamdi, Lycée du Parc,
Lyon, France.

S310. Let a, b, c be nonzero complex numbers such that |a| = |b| = |c| = k. Prove that
p p p √
| − a + b + c| + |a − b + c| + |a + b − c| ≤ 3 k.
Solution by Daniel Lasaosa, Pamplona,√ Spain

Since we may divide both sides by k, it follows that we may assume wlog k = 1. Moreover, since we
may rotate a, b, c by the same angle without altering the problem, we may assume wlog that c = 1, with
a = cos α + i sin α, b = cos β + i sin β for appropriately chosen angles 0 ≤ α ≤ β < 2π. Then,
| − a + b + c|2 = (1 + cos β − cos α)2 + (sin β − sin α)2 = 3 − 2 cos α + 2 cos β − 2 cos(α − β) =
= 3 − 4 sin µ sin δ − 2 cos2 (δ) + 2 sin2 δ = 1 − 4 sin µ sin δ + 4 sin2 δ,
α+β β−α
where we have defined µ = 2 and δ = 2 . Similarly,
|a − b + c| = 3 + 2 cos α − 2 cos β − 2 cos(α − β) = 1 + 4 sin µ sin δ + 4 sin2 δ,
2
|a + b − c|2 = 3 − 2 cos α − 2 cos β + 2 cos(α − β) = 1 − 4 cos µ cos δ + 4 cos2 δ.

By the power mean inequality, we have
p p p q
4
| − a + b + c| + |a − b + c| + |a + b − c| = 1 − 4 sin µ sin δ + 4 sin2 δ+
q p
4
+ 1 + 4 sin µ sin δ + 4 sin2 δ + 4 1 − 4 cos µ cos δ + 4 cos2 δ ≤
r
2
4 7 − 4 cos µ cos δ + 4 sin δ
3 ,
3
with equality iff
1 − 4 sin µ sin δ + 4 sin2 δ = 1 + 4 sin µ sin δ + 4 sin2 δ = 1 − 4 cos µ cos δ + 4 cos2 δ,
ie iff sin µ sin δ = 0 and cos µ cos δ = cos(2δ). If sin µ = 0, then either µ = α = β = 0 and δ = 0 yielding the
bound given in the problem statement, or µ = π and (2 cos δ − 1)(cos δ + 1) = 0. If cos δ = −1, then α = 0
and β = 2π, yielding the same result as in the previous case, while if cos δ = 21 , then α = 2π 4π
3 and β = 3 .
Moreover, note that the maximum of the expression 7 − 4 cos µ cos δ + 4 sin2 δ is reached for values of
µ, δ such that their cosines have opposite signs (or one of them is zero), otherwise we could add or subtract
π to one of them, making the second term positive instead of negative, thus producing a larger value than
the maximum. It follows that it suffices to find the maximum of 7 + 4 cos δ + 4 sin2 δ. This expression has
first and second derivatives with respect to δ which are respectively equal to
−4 2 + cos δ − 4 cos2 δ .

−4 sin δ(1 − 2 cos δ),
The first expression is zero for δ = 0, in which case the second expression is positive, or a minimum is
reached, and for δ = π3 , in which case the second expression is negative, and the maximum of the expression
√
3
is reached. Since sin π3 = 2 , we conclude that
1 3
7 − 4 cos µ cos δ + 4 sin2 δ ≤ 7 + 4 · + 4 · = 12,
2 4
with equality iff α = 2π 4π
3 and β = 3 , which is also clearly a case of equality in the previous inequality.
Restoring generality, we have
p p p √
4
√ √
| − a + b + c| + |a − b + c| + |a + b − c| ≤ 3 4 k = 3 2k,
with equality iff a, b, c are the vertices of an equilateral triangle centered at the origin of the complex plane.
Note that this maximum is larger than the one proposed in the problem statement.
Also solved by Florin Stanescu, Cioculescu Serban High School, Gaesti, Romania; George Gavrilopoulos,
Nea Makri High School, Athens, Greece; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Li Zhou
Polk State College, FL; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy.

S311. Let n be a positive integer. Prove that
bY
2c
n
bY
2c
n
n n
(x + 2j + 1) ( 2j+1) − (x + 2j)(2j )
j=0 j=0
is a polynomial of degree 2n−1 − n, whose highest term’s coefficient is (n − 1)!.
Solution by Khakimboy
Egamberganov,
Tashkent,
Uzbekistan
n n n
Since (x − 1)n = xn + xn−1 + ... + x0 we have that at x = 1
0 1 n
n n
b2c b2c
X n X n
= = 2n−1 .
2j + 1 2j
j=0 j=0

bn c n n bn c
2 2
2j +1 2j
Y Y
Let P (x) = (x + 2j + 1) and G(x) = (x + 2j) , Ki (A(x)) is i- coefficient of the
j=0 j=0
polynomial A(x), whose the coefficient of xi . Since
n n
b2c b2c
X n X n
=
2j + 1 2j
j=0 j=0
we have that
Pb n
2c
n n Pb n
2c
j=0 j=0
x =x 2j + 1
2j
K n n (P (x)) = K n n (G(x)).
Pb 2 c Pb 2 c
j=0 j=0
2j + 1 2j
Let f (x) = (x − 1)n is a function(polynomial) and x = 1 is the root(n times) of the polynomial. Then x = 1
00 00
be root of the polynomials f 0 (x), f 00 (x), ..., f .. (x) - (n − 1)-derivative of the f (x) and n-derivative of the
00 00
f (x) is equal to n!. So f (1) = f 0 (1) = ... = f .. (1) = 0. Now, we can

bn c
2
n bn c
2
n
X Y X Y
2j + 1 = 2j
α +α +...+α n =k<n,(α ≥0) j=0
αj α +α +...+α n =k<n,(α ≥0) j=0
αj
1 2 b2c i 1 2 b2c i
and
bn c
2
n n
bn c
2
X Y X Y
2j + 1 − 2j = (n − 1)!.
α1 +α2 +...+αb n c =n,(αi ≥0) j=0
αj α +...+α n =n,(α ≥0) j=0 α j
1 +α2 b2c i
2
Since n n
b2c b2c
X n X n
= = 2n−1 ,
2j + 1 2j
j=0 j=0
we get that
n
b2c bn c
2
X n X n
−n= − n = 2n−1 − n.
2j + 1 2j
j=0 j=0
and we are done.
Also solved by Li Zhou, Polk State College, Winter Haven, FL, USA.

S312. Let a, b, c, d be positive real numbers such that a3 + b3 + c3 + d3 = 1. Prove that
1 1 1 1 16
+ + + ≤ .
1 − bcd 1 − bcd 1 − bcd 1 − bcd 3
Solution by Adnan Ali, A.E.C.S-4, Mumbai, India

From the AM-GM Inequality, we obtain that
1 1 1 1
abcd ≤ √ ⇐⇒ abc ≤ √ ⇐⇒ ≤
3
4 4 3
4 4d 1 − abc 1
1− √
3
4 4d
So, we obtain √
X 1 X 4 3 4d
≤ √
cyc
1 − abc 3
cyc 4 4d − 1
Hence it suffices to show that

√
X 4 3 4d 16 X −1 16
√
3
≤ ⇐⇒ 4 − √
3
≤
cyc 4 4d − 1 3 cyc 4 4d − 1
3
X −1 −4 X 1 −4
⇐⇒ √
3
≥ ⇐⇒ √3
≥
cyc 4 4d − 1 3 cyc 1 − 4 4d
3
So, from Cauchy-Schwartz Inequality(Titu’s Lemma), we obtain
X 1 16
√3
≥ √
3
cyc 1 − 4 4d 4 − 4 4 (a + b + c + d)
Hence it suffices to show that
16 −4 1 −1
√
3
⇐⇒ √
3
≥ ≥
4 − 4 4 (a + b + c + d) 3 1 − 4 (a + b + c + d) 3
√3
√3
⇐⇒ 3 ≥ −1 + 4 (a + b + c + d) ⇐⇒ 4 ≥ 4 (a + b + c + d)
√
3
⇐⇒ 16 ≥ a + b + c + d.
The last inequality follows from Hölder Inequality as
(a3 + b3 + c3 + d3 )(1 + 1 + 1 + 1)(1 + 1 + 1 + 1) ≥ (a + b + c + d)3

√
3
⇐⇒ 16 ≥ a + b + c + d.
1
Hence the conclusion follows and equality holds for a = b = c = d = √
3
.
4
Also solved by Marius Stanean, Zalau, Romania; Navid Safei, Sharif University of Technology, Tehran,
Iran; Nicuşor Zlota, "Traian Vuia" Technical College, Focşani, Romania; Paolo Perfetti, Università degli
studi di Tor Vergata Roma, Roma, Italy.

U307. Prove that any polynomial f ∈ R[X] can be written as a difference of increasing polynomials.

Let h(x) = (2d + 1)x2d + f 0 (x) where d is the degree of f . If d > 0 then lim h(x) = +∞, while if d = 0
x→±∞
then h ≡ 1. In either case the polynomial h attains a minimum value a.
Let Q(x) = x2d+1 + (|a| + 1)x and P (x) = Q(x) + f (x) then P and Q are polynomials such that P − Q = f
and
Q0 (x) = (2d + 1)x2d + (|a| + 1) ≥ 1 > 0,

P 0 (x) = Q0 (x) + f 0 (x) = h(x) + (|a| + 1) ≥ 1 > 0,
so they are increasing.
Also solved by Daniel Lasaosa, Pamplona, Spain; Li Zhou, Polk State College, Winter Haven, FL, USA;
Khakimboy Egamberganov, Tashkent, Uzbekistan; Moubinool Omarjee, Lycée Henri IV, Paris, France; Ales-
sandro Ventullo, Milan, Italy; Radouan Boukharfane, Sidislimane, Morocco; Paolo Perfetti, Università degli
studi di Tor Vergata Roma, Roma, Italy.

U308. Let a1 , b1 , c1 , a2 , b2 , c2 be positive real numbers. Consider the functions X(x, y) and Y (x, y) which
satisfy the functional equations
x
= 1 + a1 x + b1 y + c1 Y
X
y
= 1 + a2 x + b2 y + c2 X.
Y
Prove that if 0 < x1 ≤ x2 and 0 < y2 ≤ y1 , then X(x1 , y1 ) ≤ X(x2 , y2 ) and Y (x1 , y1 ) ≥ Y (x2 , y2 ).

We will show first that if 0 < x1 ≤ x2 then X(x1 , y) ≤ X(x2 , y) and Y (x1 , y) ≥ Y (x2 , y). Set 1 + b1 y = d1 ,
1 + b2 y = d2 . Rewrite the system as
a1 xX + c1 XY + d1 X = x
a2 xY + c2 XY + d2 Y = y.
Differentiating with respect to x and rearranging the terms we obtain
(a1 x + c1 Y + d1 )X 0 + (c1 X)Y 0 = 1 − a1 X

(c2 Y )X 0 + (a2 x + c2 X + d2 )Y 0 = −a2 Y.
Solving for the unknowns X and Y we obtain

1 − a1 X c1 Y

−a2 Y a2 x + c2 X + d2
0
X =
a1 x + c1 Y + d1 c1 X

c2 Y a2 x + c2 X + d2
(1 − a1 X)(a2 x + c2 X + d2 ) + c1 b2 XY
2
.
a1 a2 x + a1 c2 xX + a1 xd2 + a2 c1 xY + c1 d2 Y + d1 a2 x + d1 c2 X + d1 d2
The denominator is positive. The (1 − a1 x) in the numerator equals x1 (c1 XY + d1 X), which is also positive.
This shows that X 0 > 0. Because
y
Y =
d2 + a2 x + c2 X
it follows that Y is decreasing.
Exchanging the roles of x, y respectively X, Y we deduce that if 0 < x1 = x2 = x and 0 < y2 ≤ y1 then
X(x, y1 ) ≤ X(x, y2 ) and Y (x, y1 ) ≥ Y (x, y2 ).
So for x1 , x2 , y1 , y2 as specified in the statement we have
X(x1 , y1 ) ≤ X(x2 , y1 ) ≤ X(x2 , y2 ) and Y (x1 , y1 ) ≥ Y (x2 , y1 ) ≥ Y (x2 , y2 ),
as desired.

U309. Let a1 , . . . , an be positive real numbers such that a1 + · · · + an = 1, n ≥ 2. Prove that for every positive
integer m,
n
X ak m+1 1
m
≥ m .
1 − ak n −1
k=1
Proposed by Titu Zvonaru, Comanesti and Neculai Stanciu, Romania

xm+1
Function f (x) = 1−x m is convex for x ∈ (0, 1) because
2m2 x( − 1 + 3m) 2m2 x−1+2m ) m(1 + m)x−1+2m m(1 + m)x−1+m

f 00 (x) = + + + >0
(1 − xm )3 (1 − xm )2 (1 − xm )2 1 − xm
for x ∈ (0, 1)
Therefore, by Jensen’s inequality
n P ak ! 1 m+1
am+1

X
k n 1 n 1
≥ nf = nf =n m = m .
1 − am
P1 1
k n
n 1− n n −1
k=1
Also solved by Daniel Lasaosa, Pamplona, Spain; Prithwijit De, HBCSE, Mumbai, India; Titouan Mor-
van, Lyée Millet, France; Nicuşor Zlota, "Traian Vuia" Technical College, Focşani, Romania; Khakimboy
Egamberganov, Tashkent, Uzbekistan; Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Arkady Alt,
San Jose, CA, USA; Adnan Ali, A.E.C.S-4, Mumbai, India; Li Zhou, Polk State College, Winter Haven,
FL, USA; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Ali Baouan, Rabat, Morocco; Moubinool
Omarjee, Lycée Henri IV, Paris, France; Radouan Boukharfane, Sidislimane, Morocco; Paolo Perfetti, Uni-
versità degli studi di Tor Vergata Roma, Roma, Italy; Prasanna Ramakrishnan, International School of Port
of Spain, Trinidad and Tobago.

U310. Let E be an ellipse with foci F and G, and let P be a point in its exterior. Let A and B be the points
where the tangents from P to E intersect E, such that A is closer to F . Furthermore, let X be the
intersection of AG with BF . Prove that XP bisects ∠AXB.
D
C
P
B
A
X
F G
Locate C on the ray F A and D on the ray GB such that AC = AG and BD = BF . Since P A
is tangent to E, ∠P AC = ∠P AG, so 4P AC ∼= 4P AG. Hence P C = P G. Likewise P D = P F . Also,
F C = F A + AC = F A + AG = GB + BF = GB + BD = GD, thus 4P CF ∼ = 4P GD. Therefore
∠CF P = ∠GDP = ∠BF P , which implies that P is the excenter, opposite F , of 4AF X. Hence P X bisects
∠AXB.
Also solved by Daniel Lasaosa, Pamplona, Spain; Radouan Boukharfane, Sidislimane, Morocco; Prasanna
Ramakrishnan, International School of Port of Spain, Trinidad and Tobago.

U311. Let f : [0, 1] → [0, 1] be a nondecreasing concave function such that f (0) = 0 and f (1) = 1. Prove that
Z 1 2 1
f (x)f −1 (x) dx ≥ .
0 12

1
Let f (x) = x k for some positive real k > 1, which clearly satisfies the conditions given in the problem
statement. Note that f −1 (x) = xk , or
Z 1 Z 1
1
−1
2 2k2 +2 k 2k2 +k+2 k
f (x)f (x) dx = x k dx = 2 x k =
2
,
0 0 2k + k + 2
0 2k + k + 2
which can be made as small as desired by letting k be as large as needed. It follows that the proposed result
is not necessarily true.
Also solved by Li Zhou, Polk State College, Winter Haven, FL, USA; Moubinool Omarjee, Lycée Henri
IV, Paris, France; Radouan Boukharfane, Sidislimane, Morocco; Paolo Perfetti, Università degli studi di
Tor Vergata Roma, Roma, Italy.

U312. Let p be a prime and let R be a commutative ring with characteristic p. Prove that the sets Sk = {x ∈
R|xp = k}, where k ∈ {1, . . . , p}, have the same number of elements.
Proposed by Corneliu Manescu-Avram, Ploiesti, Romania

If for any x ∈ R and k ∈ {1, . . . , p} it holds xp 6= k, then there is nothing to prove. Assume that there is
some x ∈ R such that xp = k for some k ∈ {1, . . . , p}. Hence, k ∈ R. Consider the set S = {x + k | x ∈ R}.
Observe that the mapping ϕ : R −→ S defined by ϕ(x) = x + k is bijective. Indeed, if ϕ(x) = ϕ(y), then
x + k = y + k, i.e. x = y and this proves that ϕ is injective. If y ∈ S, then y = x + k for some x ∈ R,
so it’s enough to take x = y − k ∈ R in order to have y = ϕ(x), and this proves that ϕ is surjective. Let
n ∈ {1, . . . , p} and consider the set Sn = {x ∈ R | xp = n}. Since (x + k)p = xp + k p = xp + k, then
(
Sn−k if n > k
ϕ(Sn ) =
Sp−k+n if k ≤ n.
It follows that |Sn | = |Sn−k | or |Sn | = |Sp−k+n |. By arbitrarily of n ∈ {1, . . . , p}, we conclude that |S1 | =
|S2 | = . . . = |Sp |.
Also solved by Daniel Lasaosa, Pamplona, Spain.

Olympiad problems
O307. Let a, b, c, d be positive real numbers such that a + b + c + d = 4. Prove that

1 1 1 1 1
+ + + ≤ .
a+3 b+3 c+3 d+3 abcd

Denote s1 = a + b + c + d = 4, s2 = ab + bc + cd + da + ac + bd, s3 = abc + bcd + cda + dab, s4 = abcd. After
multiplying both sides by s4 (a + 3)(b + 3)(c + 3)(d + 3), and rearranging terms, the inequality is equivalent
to
√ p
4
√ √ p
4 √
3 (s2 − 6 s4 ) + 2 s3 − 4 s4 3 + 18 s4 (1 − s4 ) + 8 s4 3 (1 − 4 s4 ) +
+ (81 + 27s1 + 6s2 + s3 ) (1 − s4 ) ≥ 0.

Now, by the AM-GM inequality, we have
a+b+c+d √
4 √
1= ≥ abcd = 4 s4 ,
4
s2 ab + bc + cd + da + ac + bd √6 √
= ≥ a3 b3 c3 d3 = s4 ,
6 6
s3 abc + bcd + cda + dab √ 4
p
4
= ≥ a3 b3 c3 d3 = s4 3 ,
4 4
or clearly all terms in the LHS are non-negative, being all simultaneously zero iff equality holds in all the
AM-GM inequalities, ie iff a = b = c = d = 1. The conclusion follows.
Also solved by Ioan Viorel Codreanu, Satulung, Maramures, Romania; Nicuşor Zlota, "Traian Vuia" Te-
chnical College, Focşani, Romania; Adnan Ali, A.E.C.S-4, Mumbai, India; Arkady Alt, San Jose, CA, USA;
Khakimboy Egamberganov, Tashkent, Uzbekistan; Marius Stanean, Zalau, Romania; Navid Safei, Sharif Uni-
versity of Technology, Tehran, Iran; Radouan Boukharfane, Sidislimane, Morocco; Paolo Perfetti, Università
degli studi di Tor Vergata Roma, Roma, Italy.

O308. Let ABC be a triangle and let X, Y be points in its plane such that
AX : BX : CX = AY : BY : CY.
Prove that the circumcenter of triangle ABC lies on the line XY .
Proposed by Cosmin Pohoata, USA and Josef Tkadlec, Czech Republic
X C
P
M T
Y O
A
Q
S
α B
γ
β
Let u = AX : BX = AY : BY and v = AX : CX = AY : CY . Then the locus of all points U such

that AU : BU = u is the circle α of diameter P Q, where P, Q are on AB with P A : P B = AQ : QB = u.
Likewise, the locus of all points V such that AV : CV = v is the circle β of diameter ST , where S, T are on
AC with SA : SC = AT : T C = v. Let M be the midpoint of P Q. Then
(P M + M A)(M B − P M ) = P A · QB = P B · AQ = (P M + M B)(P M − M A),
that is, P M 2 = M A · M B. Thus α is orthogonal to the circumcircle γ of 4ABC. Likewise, β is orthogonal

to γ. Therefore, the center O of γ is on the radical axis XY of α and β.
Also solved by Daniel Lasaosa, Pamplona, Spain; Khakimboy Egamberganov, Tashkent, Uzbekistan; Pra-
sanna Ramakrishnan, International School of Port of Spain, Trinidad and Tobago.

O309. Determine the least real number µ such that
√ √ √ p
µ ab + bc + ca + a2 + b2 + c2 ≥ a + b + c
for all nonnegative real numbers a, b, c with ab + bc + ca > 0. Find when equality holds.
Solution by Daniel Lasaosa, Pamplona, Spain √ √ √

Let R be the region of three-dimensional space defined by the curve g(a, b, c) = ab + bc + ca = k
for some positive real k, where moreover at most one of a, b, c is zero, and√a, b, c are all non-negative. The
problem is equivalent to findind the maximum of f (a, b, c) = a + b + c − a2 + b2 + c2 , and dividing this
maximum by k, thus yielding the minimum value of µ which satisfies the proposed inequality. Given the
nature of the region R, we need to analyze wlog (by symmetry in the variables) the following cases: (1)
a ≥ b > 0 and c = 0; (2) the limit of f (a, b, c) when b, c → 0; (3) the interior of region R, ie a, b, c all positive.
(1) We have √
a + b − a2 + b2
µ≥ √ .
ab
√
Taking a = b, note that µ ≥ 2 − 2, or it suffices to show that for all positive reals a, b, the following
inequality holds:
√ √ 2 √ √ 2
√ √ 2 a+ b a− b
a− b ≤ √ √ .
a2 + b2 + 2ab
In turn, and unless a = b (in which case the inequality clearly holds with equality), it suffices to prove that
p √ √
a2 + b2 + 2ab ≤ a + b + 2 ab,
√
2 > a2 + b2 . It follows that µ ≥ 2 − 2, and
clearly true and strict since a, b are
√ positive reals, hence (a + b)
the inequality holds for µ = 2 − 2.
(2) If b, c → 0 for fixed k, then a → ∞, while
p b2 + c2
+ O b4 + O c4 ,

a2 + b2 + c2 = a +
2a
or 2 2
a + b + c − a − b 2a
+c
b+c
lim µ = lim √ √ √ √ = lim √ √ √ = 0,
b,c→0 b,c→0
a b + c + bc b,c→0 a b+ c
or the proposed inequality holds for any positive µ under these conditions.
(3) For the interior of R, we may use Lagrange’s multiplier method, or a constant λ to be determined
exists, such that √ √
a b+ c
1− √ =λ √ ,
a2 + b2 + c2 2 a
and similarly for the cyclic permutations of a, b, c. If a 6= b when f (a, b,√
c) reaches its maximum for fixed k,
√
then subtracting their corresponding equalities and dividing by ( a − b)2 (clearly nonzero) yields
√ √ √ √ √
a+ b a+ b+ c
√ =λ √ .
a2 + b2 + c2 2 ab
Therefore, if a, b, c are all distinct and positive, it follows that
√ √ √ p √ √ √ √ √ √
2 ab a + b = a2 + b2 + c2 a + b + c = 2 bc b+ c ,

or equivalently
√ √ √ √ √
a− c a + b + c = 0,
contradiction. Or at least two of a, b, c are equal when f (a, b, c) reaches √ its maximum in the interior of R.
If a, b, c are all equal, then g(a, b, c) = k = 3a, and f (a, b, c) = a(3 − 3), yielding µ ≥ 1 − √13 . Note that
√ √ √ √ √
1 − √13 < 2 − 2 is equivalent to 24 < 2 + 12, clearly true because 24 < 5 and 12 > 3, hence if
the maximum inside R occurs when a = b = c, then µ is determined by case (1). Otherwise, if a maximum
occurs√in the interior of R which supersedes case (1), it happens when wlog a 6= b = c. Denote therefore
2 +8b2 4 +32b4
∆ = 2 ab, or k = b + ∆, a + b + c = ∆ 4b , and a2 + b2 + c2 = ∆ 16b2 , or
√
∆2 + 8b2 − ∆4 + 32b4
µ≥ .
4b(b + ∆)
√
Assume that the RHS is larger than 2 − 2 for some combination of b, ∆. Then,
√ √ p
∆2 − 4(2 − 2)b∆ + 4 2b2 > ∆4 + 32b4 ,
√ √ √ √
8 2( 2 − 1)b∆ ∆2 − (5 − 2)b∆ + 4 2b2 < 0.
The quadratic expression in the LHS needs therefore to be negative, or its discriminant must be positive, ie
√ √ √ √
16 2 < (5 − 2)2 = 29 − 10 2, 26 2 < 29,
√
clearly false. Therefore, no point inside R produces µ > 2 − 2
√
We conclude that the minimum allowed value of µ is 2 − 2, in which case equality is reached in the
proposed inequality iff one of a, b, c is zero, and the other two are equal, ie iff (a, b, c) is a permutation of
(k, k, 0) for some positive real k.
Also solved by Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Arkady Alt, San Jose, CA,
USA; Khakimboy Egamberganov, Tashkent, Uzbekistan; Radouan Boukharfane, Sidislimane, Morocco; Paolo
Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy.

O310. Let ABC be a triangle and let P be a point in its interior. Let X, Y, Z be the intersections of
AP, BP, CP with sides BC, CA, AB, respectively. Prove that
XB Y C ZA R
· · ≤ .
XY Y Z ZX 2r

Note first that, because of Ceva’s theorem, we have
XB · Y C · ZA = XC · Y A · ZB,
while as it is well known, r = 4R sin A2 sin B2 sin C2 , or squaring the proposed inequality and substituting
these results, the problem is equivalent to
Y A · ZA ZB · XB XC · Y C XB 2 Y C 2 ZA2 R2 1
· · = · · ≤ = ,
Y Z2 ZX 2 XY 2 XY 2 Y Z 2 ZX 2 4r2 64 sin2 A
2 sin2 B
2 sin2 C
2
and by cyclic symmetry, it suffices to show that

A
Y Z 2 ≥ 4Y A · ZA sin2 = 2Y A · ZA(1 − cos A).
2
Now, by the Cosine Law, Y Z 2 = Y A2 + ZA2 − 2Y A · ZA cos A, or the inequality is clearly true by the
AM-GM, with equality iff Y A = ZA. The conclusion follows, equality holds iff simultaneously Y A = ZA,
ZB = XB and XC = Y C, or iff X, Y, Z are the points where the incircle of ABC is tangent to its sides, ie
iff P is the Gergonne point of ABC.
Also solved by Ioan Viorel Codreanu, Satulung, Maramures, Romania; Arkady Alt, San Jose, CA, USA;
Khakimboy Egamberganov, Tashkent, Uzbekistan; Neculai Stanciu, Buzău, Romania; Nicuşor Zlota, "Traian
Vuia" Technical College, Focşani, Romania; Li Zhou, Polk State College, Winter Haven, FL, USA.

O311. Let ABC be a triangle with circumcircle Γ centered at O. Let the tangents to Γ at vertices B and
C intersect each other at X. Consider the circle χ centered at X with radius XB, and let M be the
point of intersection of the internal angle bisector of angle A with χ such that M lies in the interior
of triangle ABC. Denote by P the intersection of OM with the side BC and by E and F be the
orthogonal projections of M on CA and AB, respectively. Prove that P E and F P are perpendicular.
Solution by Ercole Suppa, Teramo, Italy

We begin with a preliminary result.
Lemma. The lines BE, CF and AP are concurrent.

Proof.
O E
F
M
B C
P
Let D be the second intersection point between AB and χ. We have

1 1
∠BDC = · ∠BXC = (180◦ − 2A) = 90◦ − A ⇒ CA ⊥ CD
2 2
Therefore CD k M E so, from the cyclic quadrilateral BM CD we get
∠M BF = ∠M CD = ∠CM E ⇒ 4M BF ∼ 4CM E ⇒
2
MB MF BF MB M F BF BF
= = ⇒ = · = (1)
MC CE ME MC CE M E CE

On the other hand, since OM is the M -symmedian of 4M BC we have
2
MB PB
= (2)
MC PC
From (1) and (2) it follows that
P B EC P B EC F A
· =1 ⇒ · · =1
PC FB P C EA F B
and this implies that AP , BE, CF are concurrent by Ceva’s theorem.
Returning to initial problem, let Q = P E ∩ AB, R = P E ∩ AM , S = P F ∩ AM , T = P F ∩ AC,

N = EF ∩ AM , as shown in figure.
O E
F N
M
R
B C
P
S
Q
X
By the previous lemma BE, CF , AP are concurrent so (A, B; F, Q) = −1 and the pencil R(A, B, F, Q)
is harmonic. Since F N = N E, using a well known property, we have BR k F E so
∠F BR = ∠AF E = ∠F M A
Therefore the quadrilateral BRM F is cyclic, so
∠M BF = ∠M RF = ∠N RF = ∠N RE (3)

Similarly we have S(A, C, E, T ) = −1, CS k F E, CEM S is cyclic and
∠M CE = ∠M SE = ∠M SF = ∠N SP (4)
From (3) and (4), since ∠N RE + ∠N ER = 90◦ and ∠M BF + ∠M CE = 90◦ (by Lemma), we have
∠N RE + ∠N ER = ∠M BF + ∠M CE ⇒
∠N EP = ∠N ER = ∠M CE = ∠M SE = ∠N SP (5)
hence the quadrilateral EN P S is cyclic.
Therefore ∠SP E = ∠SN E = 90◦ , i.e. P E ⊥ P F and we are done.
Also solved by Khakimboy Egamberganov, Tashkent, Uzbekistan; Sebastiano Mosca, Pescara, Italy; Ra-
douan Boukharfane, Sidislimane, Morocco; Prasanna Ramakrishnan, International School of Port of Spain,
Trinidad and Tobago.

O312. Find all increasing bijections f : (0, ∞) → (0, ∞) satisfying
f (f (x)) − 3f (x) + 2x = 0
and for which there exists x0 > 0 such that f (x0 ) = 2x0 .
Solution by Khakimboy Egamberganov, Tashkent, Uzbekistan

Suppose that there is ξ > 0 such that f (ξ) 6= 2ξ. We have f (f (ξ)) − 3f (ξ) + 2ξ = 0 and
f n+1 (ξ) − 3f n (ξ) + 2f n−1 (ξ) = 0,
where f n (x) = f (f (..f (f (x))..)), n times f .
Let f n (ξ) = an , and (an )n∈Z is positive sequence. By Characteristic equation for an , we get that t2 −
3t + 2 = 0 and t1,2 = {1, 2}. So
f n (ξ) = an = A + B · 2n (1)
for all integer n ∈ Z and some constant real numbers A and B. Why we said n ∈ Z? Because, us given that
f is bijective function. So f −1 (x) exists and satisfies the condition (1). Similarly, the f −n (x) also exists.
Clearly, B > 0 and we will prove that A = 0. If A < 0 then there is a n0 ∈ Z such that f n0 (ξ) =
A + B · 2n0 ≤ 0, for example n0 → −∞. Therefore, A ≥ 0.
Assume that A > 0. By the condition, there exists x0 > 0 such that f (x0 ) = 2x0 . We can see that there
exist infinitely many x > 0 such that f (x) = 2x, it’s not hard look at the (1). For x0 also there is one linear
equation similar to (1) and there will be A = 0,
f n (x0 ) = 2n x0
for all n ∈ Z.
So there exists x0 > 0 such that satisfies the condition f (x0 ) = 2x0 and x0 < 2A. If x0 ≤ A then there
exists n0 ∈ Z such that x0 < A + B · 2n0 < x0 + A. If A < x0 < 2A then
x0
(log2 x0 − log2 B) − (log2 (x0 − A) − log2 B) = log2 >1
x0 − A
and there exists n0 ∈ Z such that log2 (x0 −A)−log2 B < n0 < log2 x0 −log2 B and x0 < A+B ·2n0 < x0 +A.
Hence there exists n0 ∈ Z such that
x0 < A + B · 2n0 < x0 + A.
Let α = A + B · 2n0 . Since the f is increasing bijection, we get that 2x0 < f (x0 ) < f (α) = A + B · 2n0 +1 =
2α − A and x0 + A2 < α. So 2m x0 = f m (x0 ) < f m (α) = 2m α − (2m − 1) · A for all m ∈ Z and as m → +∞
2m − 1
α > lim (x0 + · A),
m→∞ 2m
α ≥ x0 + A
and a contradiction.
Hence A = 0 and f (ξ) = 2ξ again a contradiction. There isn’t ξ > 0 such that f (ξ) 6= 2ξ and for all
ξ > 0 we get that f (ξ) = 2ξ. The solution of the equation
f (f (x)) − 3f (x) + 2x = 0
is f (x) = 2x for all x > 0.
Also solved by Daniel Lasaosa, Pamplona, Spain; Henry Ricardo, New York Math Circle; Navid Safei,
Sharif University of Technology, Tehran, Iran; Radouan Boukharfane, Sidislimane, Morocco; Moubinool
Omarjee, Lycée Henri IV, Paris, France.

Junior problems
J313. Solve in real numbers the system of equations

x y + z − x3 = y z + x − y 3 = z x + y − z 3 = 1.

Solution by Ashley Colopy, College at Brockport, SUNY

We can write the equations as
 1
 1
3 + x3
y + z − x = x
 y + z
 = x
3 1 1
z + x − y = y ⇐⇒ z + x = y + y3
 
x + y − z 3 = z1 x+y = 1
+ z3
 
z
From the first equation we can see that y + z has the same sign as x. Therefore there is another variable
having the same sign as x. Repeating this for the other two equations we conclude that for each variable
there is another one having the same sign as it. Thus either all three are negative or all three are positive.
We will solve the system assuming that all three variables are positive.
Using the AM - GM inequality we get
 q
1 3 ≥ 2 x3 1 = 2x
 y + z = x + x
q x



z + x = y + y ≥ 2 y 3 y1 = 2y
1 3

 q
x + y = 1 + z 3 ≥ 2 z 3 1 = 2z

z z
Adding all three inequalities we get

1 1 1
2x + 2y + 2z = + x3 + + y 3 + + z 3 ≥ 2x + 2y + 2y.
x y z
Therefore all inequalities must be equalities and we get
1
+ x3 = 2x
x
and similar equations for y and z.
The equation is equivalent to
x4 − 2x2 + 1 = 0 ⇐⇒ (x2 − 1)2 = 0 ⇐⇒ x = 1.
Thus the solution of the system is x = y = z = 1.
The negative case produces x = y = z = −1.
Also solved by Daniel Lasaosa, Pamplona, Spain; Peter C. Shim, The Pingry School, Basking Ridge, NJ;
Seonmin Chung, Stuyvesant High School, NY; Adnan Ali, student at A.E.C.S-4, Mumbai, India; Albert Sta-
dler, Herrliberg, Switzerland; Alok Kumar, Delhi, India; Arbër Avdullahu,Mehmet Akif College,Kosovo; Ar-
kady Alt, San Jose, California, USA; Francesc Gispert Sánchez, CFIS, Universitat Politécnica de Catalunya,
Barcelona, Spain; Joshua Noah Benabou, Manhasset High School, NY; Farrukh Mukhammadiev, Academic
Lyceum Nr.1 under the SamIES, Samarkand, Uzbekistan; Evgenidis Nikolaos, M. N. Raptou High School,
Larissa, Greece; Prithwijit De, HBCSE, Mumbai, India; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Ta-
shkent, Uzbekistan; Shatlyk Mamedov, School Nr. 21, Dashoguz, Turkmenistan; Titu Zvonaru, Comănes, ti,
Romania and Neculai Stanciu, Buzău, Romania; Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter,
NH, USA; William Kang, Bergen County Academies, Hackensack, NJ, USA; Jongyeob Lee, Stuyvesant
High School, NY, USA; Michael Tang, Edina High School, MN, USA; Jaesung Son, Ridgewood, NJ, USA;
Polyahedra, Polk State College, FL, USA; Kwon Il Ko, Cushing Academy, MA, USA.

J314. Alice was dreaming. In her dream, she thought that primes of the form 3k + 1 are weird. Then she
thought it would be interesting to find a sequence of consecutive integers all of which are greater than
1 and which are not divisible by weird primes. She quickly found five consecutive numbers with this
property:
8 = 23 , 9 = 32 , 10 = 2 · 5, 11 = 11, 12 = 22 · 3.
What is the length of the longest sequence she can find?
Solution by Adnan Ali, Student at A.E.C.S-4, Mumbai, India

First of all we note that we can have a sequence of maximum of 6 integers, as, in the sequence of any 7
integers, there is obviously one which is divisible by the prime 7 = 3(2) + 1. Therefore, we show that the
required number is 6. It suffices to see that
953 = 1 · 953, 954 = 2 · 32 · 53, 955 = 5 · 191, 956 = 22 · 239, 957 = 3 · 11 · 29, 958 = 2 · 479.
Also solved by Daniel Lasaosa, Pamplona, Spain; Peter C. Shim, The Pingry School, Basking Ridge,
NJ; Seonmin Chung, Stuyvesant High School, NY; Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter,
NH, USA; William Kang, Bergen County Academies, Hackensack, NJ, USA; Rebecca Buranich, College at
Brockport, SUNY; Jongyeob Lee, Stuyvesant High School, NY, USA; Michael Tang, Edina High School, MN,
USA; Jaesung Son, Ridgewood, NJ, USA; Polyahedra, Polk State College, FL, USA; Kwon Il Ko, Cushing
Academy, MA, USA.

J315. Let a, b, c be non-negative real numbers such that a + b + c = 1. Prove that
√ √ √ √
4a + 1 + 4b + 1 + 4c + 1 ≥ 5 + 2.
Solution √
by Daniel Lasaosa,
√ Pamplona, Spain
√
Denote x = 4a + 1 − 1, y = 4b + 1 − 1, z = 4c + 1 − 1, and s = x + y + z. Note therefore that
s2 ≥ x2 + y 2 + z 2 = 4(a + b + c) + 3 − 2(x + 1 + y + 1 + z + 1) + 3 = 4 − 2s,
or √ √
0≤ s+1+ 5 s+1− 5 ,
√
for s ≥ 5 − 1, with equality iff xy + yz + zx = 0, ie iff two of x, y, z are zero, or iff two of a, b, c are zero.
It follows that √ √ √ √
4a + 1 + 4b + 1 + 4c + 1 = s + 3 ≥ 5 + 2,
with equality iff (a, b, c) is a permutation of (1, 0, 0).
Also solved by Ángel Plaza, Department of Mathematics, University of Las Palmas de Gran Canaria,
Spain; Peter C. Shim, The Pingry School, Basking Ridge, NJ; Seonmin Chung, Stuyvesant High School, NY;
Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter, NH, USA; William Kang, Bergen County Acade-
mies, Hackensack, NJ, USA; Albert Stadler, Herrliberg, Switzerland; Arkady Alt, San Jose, California, USA;
Debojyoti Biswas, Kolkata, India; Erdenebayar Bayarmagnai, School Nr.11, Ulaanbaatar, Mongolia; Henry
Ricardo, New York Math Circle; Marius Stanean, Zalau, Romania; Farrukh Mukhammadiev, Academic Ly-
ceum Nr.1 under the SamIES, Samarkand, Uzbekistan; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy; Salem Malikić,
Simon Fraser University, Burnaby, BC, Canada; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent,
Uzbekistan; Shatlyk Mamedov, School Nr. 21, Dashoguz, Turkmenistan; Titu Zvonaru, Comănes, ti, Romania
and Neculai Stanciu, Buzău, Romania; Utsab Sarkar, Chennai Mathematical Institute, India; Vincent Hu-
ang, Schimelpfenig Middle School, Plano, TX; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia;
George Gavrilopoulos, High School of Nea Makri, Athens, Greece; An Zhen-ping, Mathematics Department,
Xianyang Normal University, Xianyang, Shaanxi, China; Jongyeob Lee, Stuyvesant High School, NY, USA;
Michael Tang, Edina High School, MN, USA; Jaesung Son, Ridgewood, NJ, USA; Polyahedra, Polk State
College, FL, USA; Kwon Il Ko, Cushing Academy, MA, USA.

J316. Solve in prime numbers the equation
x3 + y 3 + z 3 + u3 + v 3 + w3 = 53353.
Solution by Evgenidis Nikolaos, M.N. Raptou High School, Larissa, Greece

Lemma 1: Every cube of an integer is conguent with 1, 0, −1 (mod 9).
Lemma 2: Every cube of an integer is congruent with 1, 0, −1 (mod 7).
Both of these lemmas can be easily proved by examining all forms of an integer (mod 9) and (mod 7),
correspondingly.
First, we observe that if all of x, y, z, u, w, v are odd primes, then their cubes’ summary would be an
even number, contradiction. So at least one of them will be equal to 2.
• if five of x, y, z, u, v, w are equal to 2 then the initial equation has no solutions.
• if three of x, y, z, u, v, w are equal to 2, let us say w = v = u = 2, then we have
x3 + y 3 + z 3 = 53329,
which by Lemma 1 does not hold.
Suppose that only w = 2. Then it will be
x3 + y 3 + z 3 + u3 + v 3 = 53345 (1)
By Lemma 1, we have that if any of the remaining x, y, z, u, v is not equal to 3, (1) does not hold, since
53345 ≡ 2 (mod 9), while LHS cannot be conguent with it.
Hence, let v = 3.
• if it is also u = 3, we obtain the equation
x3 + y 3 + z 3 = 53291.
WLOG, suppose that x ≥ y ≥ z. This way, we get to the conclusion that x = 29, x = 31 and x = 37,
for none of which we have a solution.
Back to the case only v = 3. This gives
x3 + y 3 + z 3 + u3 = 53318. (2)
By Lemma 2, we can similarly conclude that if any of x, y, z, u is not equivalent to 7, we get to a
contradiction since 53318 ≡ 6 (mod 7) and LHS of (2) cannot be congruent with it.
• if any of x, y, z equals 7 (apart from u that is taken), for instance z = 7 ⇔ z 3 = 343, then we have
y 3 + z 3 = 52632.
Now supposing x ≥ y, we can find that either x = 31 or x = 37. In each case there is no solution in
prime integers.
So, let u = 7 and we take that

x3 + y 3 + z 3 = 52975.

Now suppose that
x ≥ y ≥ z. (3)
Then, we have
52975
3x3 ≥ 52975 ⇔ x ≥ d e = 26
3
and
x ≤ 38
since if x ≥ 38 ⇔ x3 ≥ 54872, that is impossible.
We have taken that 26 ≤ x ≤ 38. Consequently, x = 29, x = 31 and x = 37. By (3), we can drop out
cases x = 29 and x = 31. For x = 37, we have
y 3 + z 3 = 2322.
Relation (3) implies that y 3 ≥ 1311 and because y is an odd prime and 1311 ≤ y 3 ≤ 2322, it must be
y = 13 or y = 11 . For y = 11 the given equation has no solution, while for y = 13 we get z = 5. Hence, the
given equation has a solution (x, y, z, u, v, w) = (37, 13, 5, 7, 3, 2) and all its permutations in prime integers.
Also solved by Daniel Lasaosa, Pamplona, Spain; Théo Lenoir, Institut Saint-Lô, Agneaux, France; Peter
C. Shim, The Pingry School, Basking Ridge, NJ; Seonmin Chung, Stuyvesant High School, NY; William
Kang, Bergen County Academies, Hackensack, NJ, USA; Albert Stadler, Herrliberg, Switzerland; Alessandro
Ventullo, Milan, Italy; Evgenidis Nikolaos, M. N. Raptou High School, Larissa, Greece; Samantha Paradis,
College at Brockport, SUNY; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania;
Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter, NH, USA; Michael Tang, Edina High School, MN,
USA; Jaesung Son, Ridgewood, NJ, USA; Polyahedra, Polk State College, FL, USA; Kwon Il Ko, Cushing
Academy, MA, USA.

J317. In triangle ABC, the angle-bisector of angle A intersects line BC at D and the circumference of triangle
ABC at E. The external angle-bisector of angle A intersects line BC at F and the circumference of
triangle ABC at G. Prove that DG ⊥ EF .
Solution by Adnan Ali, Student at A.E.C.S-4, Mumbai, India

Since ∠BAC + ∠CAY = 180◦ , ∠DAC + ∠CAF = 90◦ = ∠GAE. Hence, GE is the diameter of the
circumcircle of 4ABC. Consider 4F GE and let BC ∩ GE = Z. Then, since ∠BAE = ∠EAC, E is the
midpoint of BC. \ Hence GE ⊥ BC ⇒ ∠CZE = 90◦ . Thus F Z ⊥ GE,
d Similarly G is the midpoint of BAC.
or that F Z is an altitude of 4F GE through F . Similarly, we know that ∠EAC + ∠CAF = 90◦ . So,
EA ⊥ F G. And so EA is the altitude of 4F GE through E. One can see that AE ∩ F Z = D. Hence to
prove that DG ⊥ EF , it suffices to show that GX is an altitude of 4F GE which passes through D where
X = GD ∩ EF . Since the altitudes of a triangle concur at a point, we conclude that the altitude through G
also passes through D. And that forces DG ⊥ EF and thus the result.
Also solved by Daniel Lasaosa, Pamplona, Spain; Ricardo Barroso Campos, Sevilla, Spain; Peter C. Shim,
The Pingry School, Basking Ridge, NJ; Seonmin Chung, Stuyvesant High School, NY; William Kang, Bergen
County Academies, Hackensack, NJ, USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia;
George Gavrilopoulos, High School of Nea Makri, Athens, Greece; Andrea Fanchini, Cantú, Italy; Georgios
Batzolis, Mandoulides High School, Thessaloniki, Greece; Yassine Hamdi, Lycée du Parc, Lyon, France;
Farrukh Mukhammadiev, Academic Lyceum Nr.1 under the SamIES, Samarkand, Uzbekistan; Prithwijit
De, HBCSE, Mumbai, India; Jongyeob Lee, Stuyvesant High School, NY, USA; Jaesung Son, Ridgewood,
NJ, USA; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania; Evgenidis Nikolaos,
M. N. Raptou High School, Larissa, Greece; Polyahedra, Polk State College, FL, USA; Kwon Il Ko, Cushing
Academy, MA, USA.

J318. Determine the functions f : R → R satisfying f (x − y) − xf (y) ≤ 1 − x for all real numbers x and y.

Clearly, f (x) ≡ 1 is a solution. We show that it is the only one.
First, f (0 − (−x)) − 0f (−x) ≤ 1 − 0, so f (x) ≤ 1 for all x. Next, f (2x − x) − 2xf (x) ≤ 1 − 2x, that is,
(1 − 2x) [f (x) − 1] ≤ 0.
So f (x) ≥ 1 for all x > 21 . Thus f (x) = 1 for all x > 21 .

Finally, if x ≤ 12 then 2 − x > 1, so f (2 − x) = 1. Since f (2 − x) − 2f (x) ≤ 1 − 2, f (x) ≥ 1.
Hence f (x) = 1 for all x ≤ 21 as well.
Also solved by Solution by Alessandro Ventullo, Milan, Italy; Daniel Lasaosa, Pamplona, Spain; Peter
C. Shim, The Pingry School, Basking Ridge, NJ; Seonmin Chung, Stuyvesant High School, NY; William
Kang, Bergen County Academies, Hackensack, NJ, USA; AN-anduud Problem Solving Group, Ulaanbaatar,
Mongolia; Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter, NH, USA; Arbër Avdullahu, Mehmet
Akif College,Kosovo; Arbër Igrishta, Eqrem Qabej, Vushtrri, Kosovo; Arkady Alt, San Jose, California,
USA; Paul Revenant, Lycée Champollion, Grenoble, France; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov,
Tashkent, Uzbekistan; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania; Jongyeob
Lee, Stuyvesant High School, NY, USA; Michael Tang, Edina High School, MN, USA; Jaesung Son, Rid-
gewood, NJ, USA; Kwon Il Ko, Cushing Academy, MA, USA.

Senior problems
√ √ √
S313. Let a, b, c be nonnegative real numbers such that a + b + c = 3. Prove that
p p p
(a + b + 1)(c + 2) + (b + c + 1)(a + 2) + (c + a + 1)(b + 2) ≥ 9.
First solution by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia

√ √ √ √ √ √
a+ b+ c=3⇔3− c= a+ b
√ √
⇔ 9 − 6 c + c = a + b + 2 ab
√
⇒ 9 − 6 c + c ≤ a + b + (a + b)
9 √ c
⇔ −3 c+ ≤a+b
2 2
11 √ c
⇔ − 3 c + ≤ a + b + 1.
2 2
Hence
√

11 c
(a + b + 1)(c + 2) ≥ −3 c+ (c + 2)
2 2
√ 2 √
=9+ c − 1 ( c − 2)2 ≥ 9
p
⇒ (a + b + 1)(c + 2) ≥ 3
Similarly,
p
(b + c + 1)(a + 2) ≥ 3
p
(c + a + 1)(b + 2) ≥ 3.
Adding all above inequality the desired inequality is proved. The equality holds only when a = b = c = 1 .
Second solution by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia

Xp Xp
(a + b + 1)(c + 2) = (a + b + 1)(1 + 1 + c)
cyc cyc
Applying Cauchy-Schwartz’s inequality,

X√ √ √
≥ ( a · 1 + b · 1 + 1 · c) = 9.
cyc
Hence the inequality is proved.

The equality holds only when a = b = c = 1.
Also solved by Li Zhou, Polk State College, USA; Daniel Lasaosa, Pamplona, Spain; William Kang, Ber-
gen County Academies, Hackensack, NJ, USA; George Gavrilopoulos, High School of Nea Makri, Athens,
Greece; An Zhen-ping, Mathematics Department, Xianyang Normal University, Xianyang, Shaanxi, China;
Jongyeob Lee, Stuyvesant High School, NY, USA; Adnan Ali, Student in A.E.C.S-4, Mumbai, India; Arbër
Igrishta, Eqrem Qabej, Vushtrri, Kosovo; Arkady Alt, San Jose, California, USA; Bodhisattwa Bhowmik,
RKMV, Agartala, Tripura, India; Erdenebayar Bayarmagnai, School Nr.11, Ulaanbaatar, Mongolia; Far-
rukh Mukhammadiev, Academic Lyceum Nr.1 under the SamIES, Samarkand, Uzbekistan; Nicuşor Zlota,
“Traian Vuia” Technical College, Focşani, Romania; Evgenidis Nikolaos, M. N. Raptou High School, La-
rissa, Greece; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy; Sardor Bozorboyev,
Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan; Shatlyk Mamedov, School Nr. 21, Dashoguz, Turkmenis-
tan; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania; Utsab Sarkar, Chennai
Mathematical Institute, India; Ioan Viorel Codreanu, Satulung, Maramures, Romania.

S314. Let p, q, x, y, z be real numbers satisfying
x2 y + y 2 z + z 2 x = p and xy 2 + yz 2 + zx2 = q.
Evaluate x3 − y 3 y 3 − z 3 z 3 − x3 in terms of p and q.

Solution by Prithwijit De, HBCSE, Mumbai, India

Let ω be a non-real cube root of unity. We can write (x3 − y 3 )(y 3 − z 3 )(z 3 − x3 ) = U V W where
U = (x − y)(y − z)(z − x) = q − p (1)
V = (x − yω)(y − zω)(z − xω) = qω 2 − pω (2)

and
W = (x − yω 2 )(y − zω 2 )(z − xω 2 ) = qω − pω 2 . (3)
Thus U V W = (q − p)(qω 2 − pω)(qω − pω 2 ) = q 3 − p3 . Therefore
(x3 − y 3 )(y 3 − z 3 )(z 3 − x3 ) = q 3 − p3 .
Also solved by Li Zhou, Polk State College, USA; Daniel Lasaosa, Pamplona, Spain; William Kang,
Bergen County Academies, Hackensack, NJ, USA; Adnan Ali, student at A.E.C.S-4, Mumbai, India; Al-
bert Stadler, Herrliberg, Switzerland; Alok Kumar, Delhi, India; Arbër Avdullahu, Mehmet Akif College,
Kosovo; Arkady Alt, San Jose, California, USA; Bunyod Boltayev, Khorezm, Uzbekistan; Debojyoti Biswas,
Kolkata, India; Erdenebayar Bayarmagnai, School Nr.11, Ulaanbaatar, Mongolia; Farrukh Mukhammadiev,
Academic Lyceum Nr.1 under the SamIES, Samarkand, Uzbekistan; Navid Safei, University of Technoogy
in Policy Making of Science and Technology, Iran; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani,
Romania; Evgenidis Nikolaos, M. N. Raptou High School, Larissa, Greece; Sardor Bozorboyev, Lyceum
S.H.Sirojjidinov, Tashkent, Uzbekistan; Shatlyk Mamedov, School Nr. 21, Dashoguz, Turkmenistan; Titu
Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania.

S315. Consider triangle ABC with inradius r. Let M and M 0 be two points inside the triangle such that
∠M AB = ∠M 0 AC and ∠M BA = ∠M 0 BC. Denote by da , db , dc and d0a , d0b , d0c the distances from M
and M 0 to the sides BC, CA, AB, respectively. Prove that
da db dc d0a d0b d0c ≤ r6 .
Solution by Khakimboy Egamberganov, National University of Uzbekistan, Tashkent, Uzbekistan

We have that the points M and M 0 are isogonal points in the triangle ABC with da · d0a = db · d0b = dc · d0c .
Let
k = da · d0a = db · d0b = dc · d0c
and
Sa = [BM C], Sb = [AM C], Sc = [AM B],
Sa0 = [BM 0 C], Sb0 = [AM 0 C], Sc0 = [AM 0 B].
So we get that
2
4Sb Sb0 Sa Sa0 + 2 Sb Sb0 + 2 Sc Sc0
p p p
4Sa Sa0 4Sc Sc0 2
k= = = =
a2 b2 c2 a+b+c
and by Cauchy-Schwartz inequality,
(2Sa + 2Sb + 2Sc )(2Sa0 + 2Sb0 + 2Sc0 )

k≤ 2 = r2 .
(a + b + c)
Hence da db dc d0a d0b d0c = k 3 ≤ r6 and the equality holds if and only if M ≡ M 0 ≡ I, where I is incenter of the
triangle ABC.
Also solved by Li Zhou, Polk State College, USA; Daniel Lasaosa, Pamplona, Spain; Titu Zvonaru,
Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania; Bodhisattwa Bhowmik, RKMV, Agartala, Tri-
pura, India; Marius Stanean, Zalau, Romania; Farrukh Mukhammadiev, Academic Lyceum Nr.1 under the
SamIES, Samarkand, Uzbekistan; Prithwijit De, HBCSE, Mumbai, India.

S316. Circles C1 (O1 , R1 ) and C2 (O2 , R2 ) intersect in points U and V . Points A1 , A2 , A3 lie on C1 and points
B1 , B2 , B3 lie on C2 such that A1 B1 , A2 B2 , A3 B3 are passing through U . Denote by M1 , M2 , M3 the
midpoints of A1 B1 , A2 B2 , A3 B3 . Prove that M1 M2 M3 V is a cyclic quadrilateral.
First solution by Li Zhou, Polk State College, USA
A1 U
X
M1
Z Y
O1 O O2 B1
C1 C
V C2
Let O, X, Y, Z be the midpoints of O1 O2 , A1 U, U B1 , U M1 , respectively. Then

X +Y 1 A1 + U U + B1 1 A1 + B1 M1 + U
= + = +U = = Z.
2 2 2 2 2 2 2
Hence OZ k O1 X, and thus OZ ⊥ U M1 . Therefore, M1 is on the circle C centered at O and of radius

OU = OV . Likewise M2 and M3 are on C as well.

Second solution by Daniel Lasaosa, Pamplona, Spain
Lemma: Two secant circles (O1 , R1 ) and (O2 , R2 ) intersect at U, V . Let A, B be the respective points where a
line through either U or V meets again both circles. Then, the midpoint M of AB satisfies OM = OU = OV ,
where O is the midpoint of O1 O2 .
Proof: Consider a cartesian coordinate system with center at O, such that O1 ≡ (−d, 0), O2 ≡ (d, 0),
where 2d < R1 + R2 because the circles are secant. The equations of the circles are then
(x + d)2 + y 2 = R1 2 , (x − d)2 + y 2 = R2 2 .
Let U ≡ (∆, h) and V ≡ (∆, −h), where (∆ + d)2 + h2 = R1 2 and (∆ − d)2 + h2 = R2 2 . Any line through
U has equation y = h + m(x − ∆), or its intersections with (O1 , R1 ) satisfy
m2 + 1 x2 + 2dx − 2m2 ∆x + 2mhx + m2 − 1 ∆2 − 2d∆ − 2mh∆ = 0.

Since one of the roots, corresponding to U , is ∆, the other root equals the independent term divided by
(m2 + 1)∆, and substitution in the equation for the line through U yields
!
m2 − 1 ∆ − 2d − 2mh m2 − 1 h + 2m∆ + 2md

A≡ ,− .
m2 + 1 m2 + 1
Similarly, !
m2 − 1 ∆ + 2d − 2mh m2 − 1 h + 2m∆ − 2md

B≡ ,− ,
m2 + 1 m2 + 1
yielding !
m2 − 1 ∆ − 2mh m2 − 1 h + 2m∆

M≡ ,− .
m2 + 1 m2 + 1
It follows that 2
m2 − 1 + 4m2 ∆2 + h2

OM 2 = = ∆2 + h2 = OU 2 = OV 2 .
(m2 + 1)2
The Lemma follows, also when the line passes through V because of the symmetry with respect to line
O1 O2 .
From the Lemma, it follows that the circle through U, V with center in the midpoint O of O1 O2 passes
also through M1 , M2 , M3 . The conclusion follows.
Also solved by Adnan Ali, A.E.C.S-4, Mumbai, India; Marius Stanean, Zalau, Romania; Prithwijit De,
HBCSE, Mumbai, India; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan.

S317. Let ABC be an acute triangle inscribed in a circle of radius 1. Prove that

tan A tan B tan C 1 1 1
+ + ≥4 + + − 3.
tan3 B tan3 C tan3 A a2 b2 c2

Since a = 2 sin A, b = 2 sin B, c = 2 sin C and
(tan A − tan B)2 tan2 A + tan A tan B + tan2 B

1 1
(tan A − tan B) − =− ≤0
tan3 A tan3 B tan3 A tan3 B
then by Rearrangement Inequality
P 1 P 1 P 1 P 2
P 1
cyc tan A· ≥ cyc tan A · = cyc = cyc cot A = cyc −1 =
tan3 B tan3 A tan2 A sin2 A
X 1
1 1 1
− 3 = 4 + + − 3.
cyc
sin2 A a2 b2 c2
Also solved by Daniel Lasaosa, Pamplona, Spain; AN-anduud Problem Solving Group, Ulaanbaatar,
Mongolia; Adnan Ali, Student in A.E.C.S-4, Mumbai, India; Bodhisattwa Bhowmik, RKMV, Agartala,
Tripura, India; Marius Stanean, Zalau, Romania; Farrukh Mukhammadiev, Academic Lyceum Nr.1 under
the SamIES, Samarkand, Uzbekistan; Prithwijit De, HBCSE, Mumbai, India; Sardor Bozorboyev, Lyceum
S.H.Sirojjidinov, Tashkent, Uzbekistan; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău,
Romania; Li Zhou, Polk State College, USA.

S318. Points A1 , B1 , C1 , D1 , E1 , F1 are lying on the sides of AB, BC, CD, DE, EF, F A of a convex hexagon
ABCDEF such that
AA1 AF1 CC1 CB1 ED1 EE1
= = = = = = λ.
AB AF CD BC ED EF
2
[ACE] λ
Prove that A1 D1 , B1 E1 , C1 F1 are concurrent if and only if [BDF ] = 1−λ .
Solution by Khakimboy Egamberganov, National University of Uzbekistan, Tashkent, Uzbekistan

We have that
(1 − λ)2 [ACE] − λ2 [BDF ] = (1 − λ)2 ([OAE] + [OEC] + [OCA]) − λ2 ([OBD] + [ODF ] + [OF B]) (1).
Suppose that A1 D1 , B1 E1 , C1 F1 are concurrent at one point O. Then, we get that

0 = [OA1 D1 ] = (1−λ)[OA1 E]−λ[OA1 D] = (1−λ) (1−λ)[OAE]−λ[OBE] −λ λ[OBD]−(1−λ)[OAD] =

= (1 − λ)2 [OAE] − λ2 [OBD] + λ(1 − λ) [OAD] − [OBE] ,
so
(1 − λ)2 [OAE] − λ2 [OBD] = λ(1 − λ) [OBE] − [OAD] .
Analoguosly,
(1 − λ)2 [OEC] − λ2 [OF B] = λ(1 − λ) [OF C] − [OBE]

(1 − λ)2 [OCA] − λ2 [ODF ] = λ(1 − λ) [OAD] − [OF C] .
2
[ACE] λ
Thus, since (1) we have that [BDF ] = 1−λ .
2
[ACE] λ
Suppose that [BDF ] = 1−λ and we will prove that A1 D1 , B1 E1 , C1 F1 are concurrent. Since (1) we
can find easily that there exists a point O such that
0 = (1 − λ)2 [ACE] − λ2 [BDF ] = [OA1 D1 ] + [OB1 E1 ] + [OC1 F1 ].
So
[OA1 D1 ] = [OB1 E1 ] = [OC1 F1 ]
and O be the intersection point of the lines A1 D1 , B1 E1 , C1 F1 .
Also solved by Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan; Li Zhou, Polk State
College, USA.

U313. Let X and Y be nonnegative definite Hermitian matrices such that X − Y is also nonnegative definite.
Prove that tr X 2 ≥ tr Y 2 .

Solution by Henry Ricardo, New York Math Circle

First we show that tr(XY ) ≥ 0 : For nonnegative definite Hermitian matrices X and Y , there exist ma-
trices A and B such that X = AA∗ and Y = BB ∗ , giving us tr(XY ) = tr(AA∗ BB ∗ ) = tr(A∗ BB ∗ A) =
tr ((A∗ B)(A∗ B)∗ ) ≥ 0.
Now we write
tr(X 2 ) − tr(Y 2 ) = tr ((X − Y )X) + tr (Y (X − Y )) ≥ 0,
which follows from the result proved above since X − Y is nonnegative definite.
Also solved by Daniel Lasaosa, Pamplona, Spain; AN-anduud Problem Solving Group, Ulaanbaatar,
Mongolia; Radouan Boukharfane, Sidislimane, Morocco; Alessandro Ventullo, Milan, Italy; Li Zhou, Polk
State College, USA.

U314. Prove that for any positive integer k,
√
n
√
n
!n
1+ 2 + ··· + k k
lim > ,
n→∞ k e
where e is Euler constant.

√ x √
Let f (x) = n x. This function is clearly increasing (with first derivative f (x) = n
x) and concave (with
√ n
second derivative f 00 (x) = − nn−1
2 x2
n
x). It follows that, for x ∈ [k − 1, k], we have
√
n n−1√ n
f (x) ≤ k− 2 2
k(k − x),
n k
with equality iff x = k, or
√ Z k k Z j k p n Z j
nk n k X X
n
X n−1p n
= f (x)dx = f (x)dx < j− j (j − x)dx =
n+1 0 j−1 n2 j 2 j−1
j=1 j=1 j=1
k p k
X
n
X n−1 p n
= j− j,
2n2 j 2
j=1 j=1
or equivalently,
√
n
√
n
!n  √
n k √
n
1+ 2 + ··· + k n k n−1 X n
j
> + >
k n+1 2n2 k j2
j=1
n √
k √
n n √ k
n−1 nk X 1

n n−1 k X nj n
>k + >k + .
n+1 n + 1 2k j2 n+1 n + 1 2k j2
j=1 j=1
The limit of the second term when n → ∞ is clearly positive, since the factors outside the sum have limit
1
2k , and the sum is positive and does not depend on n. It follows that
√
n
√
n
!n n
1+ 2 + ··· + k n 1 k
lim > k lim = k lim = .
1 n
n→∞ k n→∞ n+1 n→∞ 1+ n e
Also solved by Daniel Lasaosa, Pamplona, Spain; Henry Ricardo, New York Math Circle; Arkady Alt ,
San Jose,California, USA; Ioan Viorel Codreanu, Satulung, Maramures, Romania; Alessandro Ventullo, Mi-
lan, Italy; Jaesung Son, Ridgewood, NJ, USA; Ángel Plaza, Department of Mathematics, University of Las
Palmas de Gran Canaria, Spain; Albert Stadler, Herrliberg, Switzerland; Utsab Sarkar, Chennai Mathemati-
cal Institute, India; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Corneliu Mănescu-Avram,
Transportation High School, Ploieşti, Romania; Khakimboy Egamberganov, National University of Uzbekis-
tan, Tashkent, Uzbekistan; William Kang, Bergen County Academies, Hackensack, NJ, USA; Nicuşor Zlota,
“Traian Vuia” Technical College, Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata
Roma, Roma, Italy; Li Zhou, Polk State College, USA.

U315. Let X and Y be complex matrices of the same order with XY 2 − Y 2 X = Y . Prove that Y is nilpotent.

Let J be the Jordan canonical form of Yn×n . Then there is P such that P Y P −1 = J. Let Z = P XP −1 , then
the given condition becomes ZJ 2 − J 2 Z = J. Let J1 , . . . , Jk be the Jordan blocks of J corresponding to the
eigenvalues λ1 , . . . , λk , not necessarily distinct. Then
   
J1 0 · · · 0 Z1 ∗ · · · ∗
 0 J2 · · · 0   ∗ Z2 · · · ∗ 
J = . , Z = ..  ,
   
. .
. . . .
.   .. .. . .
 . . . .   . . . . 
0 0 · · · Jk ∗ ∗ · · · Zk
where Zi and Ji have the same order for each i. By block multiplication, the given condition implies Zi Ji2 −
Ji2 Zi = Ji for each i. Now let ni be the order of Ji , then
ni λi = tr(Ji ) = tr(Zi Ji2 ) − tr(Ji2 Zi ) = tr(Ji2 Zi ) − tr(Ji2 Zi ) = 0,
so λi = 0 for 1 ≤ i ≤ k. Hence Jini = 0, thus J m = 0, where m = max{n1 , . . . , nk }. Therefore, Y m =

P −1 J m P = 0 as well.
Also solved by Khakimboy Egamberganov, National University of Uzbekistan, Tashkent, Uzbekistan; AN-
anduud Problem Solving Group, Ulaanbaatar, Mongolia; Alessandro Ventullo, Milan, Italy.

U316. The sequence {Fn } is defined by F1 = F2 = 1, Fn+2 = Fn+1 + Fn for n ≥ 1. For any natural number
m define v2 (m) as v2 (m) = n if 2n |m and 2n+1 |m. Prove that there is exactly one positive number µ
such that the equation
v2 ([µn]!) = v2 (F1 F2 · · · Fn )
is satisfied by the infinitely many positive integers n. determine that number µ.
First solution by by the proposer

We claim that the equation
v2 ([µn]!) = v2 (F1 F2 · · · Fn ) (1)
has infinitely many solutions in n if and only if µ = 56 . We start with the following:
Lemma 1:
Let p be a prime. Let n be a natural number whose base p representation equals n = kj=0 nj pj , where
P
0 ≤ nj ≤ p − 1, 0 ≤ j ≤ k. For any natural number m define vp (m) = n if pn |m and pn+1 6 |m. Then
P n n − n0 − n1 − n2 − · · · − nk
vp (n!) = j≥1 j = .
p p−1
Proof: h i
Among the integers 1, 2, . . . , n there are exactly np which are divisible by p, namely

n
p, 2p, ..., p. (2)
p
The integers between 1 and n which are divisible by p2 (a subset of (2)) are

n
2 2
p , 2p , . . . , 2 p2 ,
p
h i
which are pn2 in number, and so on. The number of integers between 1 and n, which are divisible by
h i h i
pj but not by pj+1 is exactly pnj − pj+1
n
. Hence p divides n! exactly
X n n X n
j − j+1 =
pj p pj
j≥1 j≥1
times. We have
X n X n X n
= −
pj pj pj
j≥1 j≥1 j≥1
X n0 + n1 p + n2 p2 + · · · + nk pk

n
= − =
p−1 pj
j≥1
k
n X n0 + n1 p + n2 p2 + · · · + nj−1 pj−1 X n0 + n1 p + n2 p2 + · · · + nk pk
= − − =
p−1 pj pj
j=1 j≥k+1
n 1 P 1 n 1
− kj=1 j ji=1 ni−1 pi−1 − k
Pk
− ki=1 ni−1 kj=i pi−j−1 − k
i
Pk
ni p i =
P P P
i=0 ni p =
p−1 p p (p − 1) p−1 p (p − 1) i=0
k k
n 1 X 1 X
− ni−1 (1 − pi−1−k ) − k ni p i =
p−1 p−1 p (p − 1)
i=1 i=0

n − n0 − n1 − n2 − · · · − nk
= .
p−1
This concludes the proof of Lemma 1.
Lemma 2: 


0, if n ≡ 1, 2 mod 3

1, if n ≡ 3 mod 6
v2 (Fn ) =


3, if n ≡ 6 mod 12

v (n) + 2,
2 if n ≡ 0 mod 12
Proof:
See Lemma 2 of Lengyel’s paper
"The order of the
P Fibonacci h nand
i Lucas
P numbers" n
Let [µn] = i ei 2i and = i fi 2i be the binary representations of [µn] and 12 respectively. Then,
12
by Lemma 1 and Lemma 2:
X [µn] X X
v2 ([µn]!) = k
= ei (2i − 1) = [µn] − ei ,
2
k≥1 i i
n
X X X X X
v2 (F1 F2 · · · Fn ) = v2 (Fk ) = 1+ 3+ 2+ v2 (12k) =
k=1 0<6k+3≤n 0<12k+6≤n 0<12k≤n 0<12k≤n

X X X X n+3 n+6 hni h n i
= 1+ 3+ 4+ v2 (k) = +3 +4 + v2 ! =
n 6 12 12 12
0<6k+3≤n 0<12k+6≤n 0<12k≤n 0<k≤ 12

n 1 n 1 hni X
+ +3 + +5 − fi .
6 2 12 2 12
i
P P
We have i ei = O(log n) and i fi = O(log n). So, as n tends to infinity,
X
v2 ([µn]!) = [µn] − ei = µn + O(log n)
i
and
n 1 n 1 hni X 5n
v2 (F1 F2 · · · Fn ) = + +3 + +5 − fi = + O(log n).
6 2 12 2 12 6
i
5
So a necessary condition that (1) has infinitely many solutions is µ = .
6
k 5n k−1 7 2 1 0 k−1 5n
Consider the numbers n = 81·2 , k ≥ 2 Then = 135·2 = (2 +2 +2 +2 )2 , v2 ! =
hni 6 h n i 6
135 · 2k−1 − 4, = 27 · 2k−2 = (24 + 23 + 21 + 20 )2k−2 , v2 ! = 27 · 2k−1 − 4, v2 (F1 F2 · · · Fn ) =
12 12
n 1 n 1 hni P
+ +3 + +5 − i fi = 27 · 2k−1 + 3 · 27 · 2k−2 + 5 · 27 · 2k−2 − 4 =
6 2 12 2 12

5n
135 · 2k−1 − 4 = v2 ! .
6
This concludes the proof.

Second solution by Li Zhou, Polk State College, USA
We show that µ = 56 and the equation is satisfied by n = 3 · 2m + 17 for all m ≥ 5.
From the recurrence we see that Fn is odd for all n not divisible by 3. Note that F3 = 2, F6 = 23 , and
F12 = 24 · 9. Assume q is an odd integer below. Then (F3q , F6 ) = F(3q,6) = F3 and (F6q , F12 ) = F(6q,12) = F6 ,
so v2 (F3q ) = 1 and v2 (F6q ) = 3. NowP
F2n = F n Ln , where L0 = 2, L1 = 1, and Ln+2 = Ln+1 + Ln for n ≥ 0.
By the well-known fact that L3n = ni=0 ni 2i Li , we get L6k ≡ L0 (mod 4), that is, v2 (L6k ) = 1 for all
k ≥ 0. It is then easy to see by induction on m that v2 (F6·2m q ) = m + 3 for all m ≥ 0. Therefore,
j k j n k 1 j n k ∞
1 n jnk jnk X bn/6c
v2 (F1 · · · Fn ) = +3 + v2 ! = +3 +
2 3 6 6 2 3 6 2i
i=1
n n n 5n
∼ + + = , n → ∞.
6 2 6 6
On the other hand,
∞
X bµnc
v2 (bµnc!) = ∼ µn, n → ∞.
2i
i=1
Hence, to satisfy the given equation for infinitely many n, µ must necessarily be 56 .
Finally, for all n = 3 · 2m + 17 with m ≥ 5, the two formulas above become
v2 (F1 · · · Fn ) = 2m−1 + 3 + 3 2m−1 + 2 + 2m−1 = 2m+1 + 2m−1 + 9,

∞ ∞ m+1
5 · 2m−1 + 14 + 2m−1 + 14
X
5n X 2
v2 ! = =
6 2i 2i
i=1 i=1
X m m−2
14 14 14 X
= + + + 2i + 2i
2 4 8
i=0 i=0
m+1 m−1
− 1 = 2m+1 + 2m−1 + 9.

= 11 + 2 −1 + 2

U317. For any natural numbers s, t and p, prove that there is a number M (s, t, p) such that every graph with
a matching of size at least M (s, t, p) contains either a clique Ks , an induced complete bipartite graph
Kt,t or an induced matching Mp . Does the result remain true if we replace the word "matching" by
"path"?
Solution by the proposer

The result is true for both "matching" and "path" cases. We’ll only include a proof of it for "matching";
for the "path" proof, we refer to A. Atminas, V.V. Lozin, I. Razgon, Linear time algorithm for computing
a small complete biclique in graphs without long induced paths, SWAT, 2012: 142-152.
The key part of the proof is the following analogue for bipartite graphs, which is easier to prove.
Lemma:.
For any natural numbers t and p, there is a natural number N (t, p) such that every bipartite graph with a
matching of size at least N (t, p) contains either a complete bipartite graph Kt,t or an induced matching Mp .
Proof:
For p = 1 and arbitrary t, we can define N (t, p) = 1. Now, for each fixed t, we prove the result by induction
on p. Without loss of generality, we prove it only for values of the form p = 2s . Suppose we have shown the
lemma for p = 2s for some s ≥ 0. Let us now show that it is sufficient to set N (t, 2p) = RB(t, RB(t, N (t, p))),
where RB is the bipartite Ramsey number.
Consider a graph G with a matching of size at least RB(t, RB(t, N (t, p))). Without loss of generality,
we may assume that G contains no vertices outside of this matching. We also assume that G does not
contain an induced Kt,t , since otherwise we are done. Then G must contain the bipartite complement of
KRB(t,N (t,p)),RB(t,N (t,p)) with vertex classes, say A and B. Now let C and D consist of the vertices matched
to vertices in A and B respectively in the original matching in G.
Note that A, B, C, D are pairwise disjoint. Graphs G[A ∪ C] and G[B ∪ D] (induced on sets A ∪ C and
B ∪ D) now each contain a matching of size RB(t, N (t, p)). There are no edges between A and B, yet there
might be edges between C and D. By our assumption, G[C ∪ D] is however Kt,t -free, therefore, by Ramsey’s
theorem, it must contain the bipartite complement of KN (t,p),N (t,p) , with vertex sets C 0 ⊂ C, D0 ⊂ D. Let
A0 ⊂ A and B 0 ⊂ B be the set of vertices matched to C 0 and D0 respectively in the original matching in G.
Now there are no edges in G[A0 ∪ B 0 ] and none in G[C 0 ∪ D0 ], but G[A0 ∪ C 0 ] and G[B 0 ∪ D0 ] both contain
a matching of size N (t, p). Since G is Kt,t -free, by the induction hypothesis, we conclude that they both
contain an induced matching Mp . Putting these together we find that G contains an induced M2p . This
proves the Lemma.
Returning to the problem, define M (s, t, p) := R(s, R(s, N (t, p))), where R is the classical Ramsey
number and N (t, p) is the N from the Lemma. Suppose that G is a (Ks , Kt,t )-free graph with a matching
of size R(s, R(s, N (t, p))). Since G is Ks -free, by Ramsey’s theorem, it must contain an independent set A
of size R(s, N (t, p)). Let B be the set of vertices matched to A. Since G[B] is Ks -free, it must contain an
independent set B 0 of size N (t, p). Let A0 be the set of vertices matched to B 0 ; the graph H = G[A0 ∪ B 0 ] is
a bipartite graph with a matching of size N (t, p). By the above, since H is Kt,t -free, it thus follows that H
contains an induced matching Mp . This completes the proof.

P∞ (−1)q(k)
U318. Determine all possible values of k=1 k2
, where q(x) is a quadratic polynomial that assumes only
integer values at integer places.

It is well known that
x x x 1
q(x) = a +b +c = ax(x − 1) + bx + c
2 1 0 2
for some a, b, c ∈ Z with a 6= 0. Note that for k ∈ N,
q(k + 4) = q(k) + 4ak + 6a + 4b ≡ q(k) (mod 2).
Now denote 12 ax(x − 1) + bx + c by qa,b,c (x), with a, b, c ≡ 0 or 1 (mod 2). Then

∞
X (−1)q0,0,0 (k) 1 1 1 1 π2
S0,0,0 = = + 2 + 2 + 2 + ··· = ,
k2 1 2 2 3 4 6
k=1
∞
X (−1)q0,1,0 (k) 1 1 1 1 π2
S0,1,0 = =− + − + − · · · = − ,
k2 12 22 32 42 12
k=1
∞
X (−1)q1,0,0 (k)
1 1 1 1 1 1 1 1
S1,0,0 = − 2 − 2 + 2 + 2 − 2 − 2 + 2 + ···
=
k2
1 2 2 3 4 5 6 7 8
k=1
π2

1 1 1 1 1 1 1 1 1
= − + − + · · · − − + − + · · · = G − ,
12 32 52 72 4 12 22 32 42 48
where G is Catalan’s constant, and

∞
X (−1)q1,1,0 (k)
1 1 1 1 1 1 1 1
S1,1,0 = − =− + + − − + + − ···
k2 12 22 32 42 52 62 72 82
k=1
π2

1 1 1 1 1 1 1 1 1
= − − + − + · · · − − + − + · · · = −G − .
12 32 52 72 4 12 22 32 42 48
Finally, S0,0,1 = −S0,0,0 , S0,1,1 = −S0,1,0 , S1,0,1 = −S1,0,0 , and S1,1,1 = −S1,1,0 .
Also solved by Roberto Mastropietro and Emiliano Torti, University of Rome Tor Vergata", Italy; AN-
anduud Problem Solving Group, Ulaanbaatar, Mongolia; Daniel Lasaosa, Pamplona, Spain.

Olympiad problems
O313. Find all positive integers n for which there are positive integers a0 , a1 , . . . , an such that a0 + a1 + · · · +
an = 5(n − 1) and
1 1 1
+ + ··· + = 2.
a0 a1 an
Solution by Adnan Ali, Student in A.E.C.S-4, Mumbai, India

From the AM-HM Inequality,
5(n − 1) n + 1
≥ .
n+1 2
√ √
So, n2 − 8n + 11 ≤ 0 ⇒ 1 < 4 − 5 ≤ n ≤ 4 + 5 < 7. Hence 2 ≤ n ≤ 6. But it is easy to see that
1 1 1
2 + 2 + 1 = 5, + + = 2, (for n = 2)
2 2 1
1 1 1 1
3 + 3 + 3 + 1 = 10, + + + = 2, (for n = 3)
3 3 3 1
1 1 1 1 1
2 + 2 + 2 + 3 + 6 = 15, + + + + = 2, (for n = 4)
2 2 2 3 6
1 1 1 1 1 1
3 + 3 + 3 + 2 + 3 + 6 = 20, + + + + + = 2, (for n = 5)
3 3 3 2 3 6
1 1 1 1 1 1 1
3 + 3 + 3 + 4 + 4 + 4 + 4 = 25, + + + + + + = 2, (for n = 6)
3 3 3 4 4 4 4
Hence n ∈ {2, 3, 4, 5, 6}.
Also solved by Daniel Lasaosa, Pamplona, Spain; William Kang, Bergen County Academies, Hackensack,
NJ, USA; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Michael Tang, Edina High School,
MN, USA; Ángel Plaza, Department of Mathematics, University of Las Palmas de Gran Canaria, Spain;
Albert Stadler, Herrliberg, Switzerland; Alessandro Ventullo, Milan, Italy; Arbër Avdullahu, Mehmet Akif
College, Kosovo; Arbër Igrishta, Eqrem Qabej, Vushtrri, Kosovo; Bodhisattwa Bhowmik, RKMV, Agartala,
Tripura, India; Erdenebayar Bayarmagnai, School Nr.11, Ulaanbaatar, Mongolia; Francesc Gispert Sán-
chez, CFIS, Universitat Politécnica de Catalunya, Barcelona, Spain; Henry Ricardo, New York Math Circle;
Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan; Shatlyk Mamedov, School Nr. 21, Da-
shoguz, Turkmenistan; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania Li Zhou,
Polk State College, USA.

O314. Prove that every polynomial p(x) with integer coefficients can be represented as a sum of cubes of
several polynomials that return integer values for any integer x.
Solution by Navid Safei, University of Technology in Policy Making of Science and Technology, Iran
At first thanks to the following identity one can find that the polynomial x, could be written as sum of
cubes of five integer valued polynomials as follows:
3 3 3 3
x3 − x x3 − x x3 − x x3 − x

3
x=x + + + −1 − + 1−
6 6 6 6
x3 − x
It is obvious that the polynomial is an integer-valued polynomial. By use of the following we can
6
2
find that the polynomial x , could be written as sum of cubes of integer-valued polynomials( just take place
in the above identity x through x2 ) . Now consider the polynomial P (x) with integer coefficients then we
can write the polynomial by the form P (x) = P1 (x3 ) + xP2 (x3 ) + x2 P3 (x3 ) where P1 , P2 , P3 are polynomials
with integer coefficients . it is obvious that these polynomials (i.e. P1 , P2 , P3 ) could be written as the sum
of cubes of polynomials with integer coefficients and thus integer-valued. Since polynomials x, x2 could be
written so , the statement was proved.
Also solved by Henri Godefroy, Stanislas Secondary School, Paris, France; Li Zhou, Polk State College,
USA.

a3 + 3b2 + 5 b3 + 3c2 + 5 c3 + 3a2 + 5 ≥ 27(a + b + c)3 .

The AGM yields a3 + 1 + 1 ≥ 3a hence
(a3 + 3b2 + 5)(b3 + 3c2 + 5)(c3 + 3a2 + 5) ≥

≥ (3a + 3b2 + 3)(3b + 3c2 + 3)(3c + 3a2 + 3) =
= 27(a + b2 + 1)(b + c2 + 1)(c + a2 + 1)
we come to
(a + b2 + 1)(b + c2 + 1)(c + a2 + 1) ≥ (a + b + c)3

Now Hölder comes in
!3
X 1 1 2
(a + b2 + 1)(1 + b + c2 )(a2 + 1 + c) ≥ a 1 a
3 3 3
cyc
whence the result.
Also solved by George Gavrilopoulos, High School of Nea Makri, Athens, Greece; An Zhen-ping, Ma-
thematics Department, Xianyang Normal University, Xianyang, Shaanxi, China; Albert Stadler, Herrliberg,
Switzerland; Erdenebayar Bayarmagnai, School Nr.11, Ulaanbaatar, Mongolia; Khakimboy Egamberganov,
National University of Uzbekistan, Tashkent, Uzbekistan; Daniel Lasaosa, Pamplona, Spain; Marius Sta-
nean, Zalau, Romania; Farrukh Mukhammadiev, Academic Lyceum Nr.1 under the SamIES, Samarkand,
Uzbekistan; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan; Titu Zvonaru, Comănes, ti,
Romania and Neculai Stanciu, Buzău, Romania; Li Zhou, Polk State College, USA.

O316. Prove that for all integers k ≥ 2 there exists a power of 2 such that at least half of the last k digits
are nines. For example, for k = 2 and k = 3 we have 212 = . . . 96 and 253 = . . . 992.

Claim: If x ≡ −1 (mod 5k ) for some integers x and k ≥ 1, then x5 ≡ −1 (mod 5k+1 ).
k−1
Corollary: For every positive integer k, we have 22·5 ≡ −1 (mod 5k ).
Proof: If x ≡ −1 (mod 5k ), an integer u exists such that x = u · 5k − 1, or

5 5 4k−1
x = u ·5 − u · 5 + 2u · 5 − 2u · 5 + u 5k+1 − 1.
4 3k 3 2k 2 k
The Claim follows. Since 22 = 5 · 1 − 1, the result holds for k = 1, serving as the base case for induction over
k, the step being guaranteed by the Claim. The Corollary follows.
By the Corollary, for every positive integer k, a positive integer u exists such that
k−1 k−1 +k
22·5 = u5k − 1, 0 < u10k − 22·5 = 2k .
Note that if k = 2` ≥ 2 is an even positive integer, then 2k < 10` , clearly true since it is equivalent to
5 `
> 1, whereas if k = 2` + 1 ≥ 3 is an odd integer larger than 1, then 2k < 10` , which is equivalent to

2
2`+1 < 5` , true for ` = 1 (equivalent to k = 3), and clearly true for any ` ≥ 1 since the RHS is multiplied by
5 every time ` grows in one unit, the LHS being multiplied only by 2. It follows that, whatever the last `
digits are, the previous k − ` digits (which are clearly at least half of the last k digits) are 9’s. The conclusion
follows.
Also solved by AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Albert Stadler, Herrliberg,
Switzerland; Li Zhou, Polk State College, USA.

O317. Twelve scientists met at a math conference. It is known that every two scientists have a common friend
among the rest of the people. Prove that there is a scientist who knows at least five people from the
attendees of the conference.
First solution by Li Zhou, Polk State College, USA

Number the scientists 1 through 12 and let S = {1, . . . , 12}. For each i ∈ S, denote by Fi the set of friends
of i. Assume that |Fi | ≤ 4 for all i.
(a) Claim 1: |Fi | = 4 for all i.
Indeed, if |Fi | ≤ 3 for some i, then

[ X

(F k \ {i})≤
|Fk \ {i}| ≤ 3|Fi | ≤ 9 < 11,
k∈Fi k∈Fi
so there is j ∈ S \ {i} such that i and j do not have a friend in common, a contradiction.
(b) Claim 2: If i and j are friends then |Fi ∩ Fj | = 1.
Indeed, suppose that Fi = {j, k, l, m} and k, l ∈ Fj as well. Note that i and m must have a friend
in common, so |Fm ∩ {j, k, l}| ≥ 1. Hence, |(Fj ∪ Fk ∪ Fl ∪ Fm ) \ {i, j, k, l, m}| ≤ 6 < 7, so there is n ∈
S \ {i, j, k, l, m} such that i and n do not have a friend in common, a contradiction.
(c) Without loss of generality, assume now that F1 = {2, 3, 4, 5} with F1 ∩ F2 = {3} and F1 ∩ F4 = {5}, as
in the left graph below. For 1 to have a friend in common with each scientist in {6, . . . , 12}, we must have
{6, . . . , 12} ⊆ (F2 ∪ F3 ∪ F4 ∪ F5 ). Now each of 2, 3, 4, 5 needs to have two friends in {6, . . . , 12}, so exactly
one scientist in {6, . . . , 12} has two friends in {2, 3, 4, 5} who are not friends. Without loss of generality, say
6 is a friend of 2 and 4, then we have the complete left graph below. However, for 2 to have a friend in
common with each of 11 and 12, 7 has to be a friend of both 11 and 12, which contradicts Claim 2 since
{5, 7} ⊆ F11 ∩ F12 . This contradiction completes the proof.
2 9 3 8 10
7
3 1 6
10 4
6 1
11 7
5 2 5 9
8 4 12 11
(d) Finally we notice that it does require twelve scientists to draw the conclusion, as shown by the right
graph above with eleven scientists.

Second solution by Michael Ma, Plano, TX
Let each of the scientists represent a vertex on a graph G. Now draw an edge between two vertices if the two
corresponding scientists are friends. Now we know that every two vertices share a common neighbor. We will
proceed now by contradiction. So assume that no vertex is of degree at least 5. Now say there exists a vertex
with degree 3 or less. Then in G there are 11 other vertices that share a common neighbor. This common
neighbor must be one of the 3 or less neighbors. Then by the pigeonhole principle there is a neighbor that
has at least 44 edges to the other 11 vertices. Since this neighbor also has an edge to the original vertex,
this vertex has degree at least 5. Contradiction.
So now every vertex has degree at least and at most and therefore exactly 4. Now take a vertex A. Say
it’s neighbors are B, C, D, E. Now the other 7 points have to share one of B, C, D, E with A. Say one of
B, C, D, E is adjacent to 3 or more of these 7 points. WLOG say B. Then there are only 3 since the degree
of B is 4. Then B must share a vertex with A. But none of these 3 points are neighbors with A and A is
not a neighbor of itself. Contradiction.
So one of B, C, D, E is adjacent to at least 1 of the seven points and the other 3 are neighbors of exactly
2 of the 7 points. WLOG say E is adjacent to at least 1 and E is adjacent to F, D is adjacent to G, H, C is
adjacent to I, L, and B is adjacent to J, K. Now E has to be adjacent with on of B, C, D. WLOG say D.
Therefore since B and C cannot both be adjacent to E, B and C must be adjacent. Therefore E is adjacent
to one of G, H, I, L, J, K. Assume E is connected to one of G, H. WLOG let it be H. Then J, K, I, L must
share a neighbor with D. Since they are not adjacent to A or E, they must be adjacent to H or G. But H
can only have one more neighbor and G only 2. Contradiction.
Hence E is connected to one of I, J, K, L. WLOG let it be L. Now since H shares a neighbor with D, G
and H are adjacent. Similarly I, L and K, J are adjacent.
Now since L and E must share a common neighbor and L cannot be adjacent to A or D, its must
be adjacent to F . Similarly, since K and J cannot be neighbors of L, since it’s degree is 4, K and J are
neighbors of F . Now since H and B share a common neighbor, H is connected to KorJ. WLOG assume
it’s K. Similarly we get that J is adjacent to G. Now since H and C share a common neighbor, H and I
are adjacent. Similarly, so are G and I. Now H and K share no common neighbors. Contradiction. So there
is a vertex of degree at least 5.
Also solved by Francesc Gispert Sánchez, CFIS, Universitat Politécnica de Catalunya, Barcelona, Spain;
Philip Radoslavov Grozdanov, Yambol, Bulgaria.

O318. Find all polynomials f ∈ Z[X] with the property that for any distinct primes p and q, f (p) and f (q)
are relatively prime.

For each degree n ≥ 0, f (x) = ±xn clearly have this property. We show that they are the only ones.
Suppose that f (x) = an xn + · · · + a1 x + a0 is such a polynomial, with n ≥ 0. Let p be any prime such
that p > |an | + · · · + |a0 |. If there is a prime p1 6= p such that p1 |f (p), then by Dirichlet’s theorem, there is
a prime q = p1 u + p for some positive integer u. Hence f (q) ≡ f (p) ≡ 0 (mod p1 ), that is, p1 | (f (p), f (q)), a
contradiction. Therefore |f (p)| must be a power of p. Observe that
|f (p)| ≤ |an |pn + · · · + |a1 |p + |a0 | ≤ (|an | + · · · + |a0 |) pn < pn+1 ,
|f (p)| ≥ |an |pn − (|an−1 | + · · · + |a0 |) pn−1 ≥ |an |pn − (p − 2)pn−1 .

If |an | ≥ 2 then |f (p)| > pn , so f (p) cannot be a power of p. Thus |an | = 1. Then |f (p)| > pn−1 , so
|f (p)| = pn , which implies f (p) − an pn ∈ {0, ±2pn }. But
|f (p) − an pn | = |an−1 pn−1 + · · · + a1 p + a0 | ≤ (|an−1 | + · · · + |a0 |) pn−1 < pn ,
so f (p) − an pn = 0. This means that f (x) − an xn has all primes greater than |an | + · · · + |a0 | as zeros, thus
it must be identically 0, that is, f (x) = an xn with |an | = 1.
Also solved by Daniel Lasaosa, Pamplona, Spain; Navid Safei, University of Technoogy in Policy Making
of Science and Technology, Iran; Khakimboy Egamberganov, National University of Uzbekistan, Tashkent,
Uzbekistan.

Junior problems
J319. Let 0 = a0 < a1 < · · · < an < an+1 = 1 such that a1 + a2 + · · · + an = 1. Prove that
a1 a2 an 1
+ + ··· + ≥ .
a2 − a0 a3 − a1 an+1 − an−1 an

The expression on the left-hand-side can be rewritten as
a21 a22 a2n
+ + ··· +
a1 a2 − a0 a1 a2 a3 − a1 a2 an an+1 − an−1 an
Now applying the Cauchy-Schwartz Inequality on the expression, we have
a21 a22 a2n
+ + ··· +
a1 a2 − a0 a1 a2 a3 − a1 a2 an an+1 − an−1 an
(a1 + a2 + · · · + an )2 1 1
≥ = = .
a1 a2 − a0 a1 + a2 a3 − a1 a2 + · · · + an an+1 − an−1 an an an+1 − a0 a1 an
Also solved by Daniel Lasaosa, Pamplona, Spain; Ioan Viorel Codreanu, Satulung, Maramures, Romania;
Alok Kumar, Delhi, India; Ángel Plaza, University of Las Palmas de Gran Canaria, Spain; Arber Igrishta,
Eqrem Qabej, Vushtrri, Kosovo; Arkady Alt, San Jose, California, USA; Bodhisattwa Bhowmik, RKMV,
Agartala, Tripura, India; Daniel Văcaru, Pites, ti, Romania; David E. Manes, Oneonta, NY, USA; Ilyes
Hamdi, Lycée Voltaire, Doha, Qatar; Farrukh Mukhammadiev, Academic Lyceum Nr1, Samarkand, Uzbe-
kistan; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Paolo Perfetti, Università degli
studi di Tor Vergata Roma, Roma, Italy; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekis-
tan; Shatlyk Mamedov, Dashoguz , Turkmenistan; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu,
Buzău, Romania; Seung Hwan An, Taft School, Watertown, CT, USA; Chaeyeon Oh, Episcopal High School,
Alexandra, VA, USA; Mehtaab Sawhney, USA; Yujin Kim, Stony Brook School, Stony Brook, NY, USA;
Misiakos Panagiotis, Athens College (HAEF), Nea Penteli; William Kang, Bergen County Academies, Hac-
kensack, NJ, USA; Ji Eun Kim, Tabor Academy, Marion, MA, USA; Timothy Chon, Horace Mann School,
Bronx, NY, USA; Cody Johnson, USA; Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter, NH, USA;
Michael Tang, Edina High School, MN, USA; Yong Xi Wang,East China Institute Of Technology, China;
Yooree Ha, Ponte Vedra High School, Ponte Vedra, FL, USA; AN-anduud Problem Solving Group, Ulaan-
baatar, Mongolia.

J320. Find all positive integers n for which 2014n + 11n is a perfect square.
Solution by Prithwijit De, HBCSE, Mumbai, India

Let p(n) = 2014n + 11n for n ≥ 1. Then p(1) = 2025 = 452 . Also p(n) is odd for all n. We will show that for
no other value of n is p(n) a perfect square. For n even note that the last digit of p(n) is 7. Thus it cannot
be a perfect square.
For n odd and n > 1 note that p(n) leaves 3 as remainder when divided by 8. But the square of any odd
integer leaves 1 as remainder when divided by 8. Thus p(n) cannot be a square for odd positive integers n
greater than 1.
Also solved by Daniel Lasaosa, Pamplona, Spain; Ioan Viorel Codreanu, Satulung, Maramures, Romania;
Seung Hwan An, Taft School, Watertown, CT, USA; Chaeyeon Oh, Episcopal High School, Alexandra,
VA, USA; Mehtaab Sawhney, USA; Yujin Kim, Stony Brook School, Stony Brook, NY, USA; Misiakos
Panagiotis ,Athens College (HAEF), Nea Penteli; William Kang, Bergen County Academies, Hackensack,
NJ, USA; Ji Eun Kim, Tabor Academy, Marion, MA, USA; Timothy Chon, Horace Mann School, Bronx,
NY, USA; Cody Johnson, USA; Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter, NH, USA; Michael
Tang, Edina High School, MN, USA; Adnan Ali, Student in A.E.C.S-4, Mumbai, India; Albert Stadler,
Herrliberg, Switzerland; Alok Kumar, Delhi, India; Arber Avdullahu, Mehmet Akif College, Kosovo; Arkady
Alt, San Jose, California, USA; Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Corneliu Mănescu-
Avram, Transportation High School, Ploieşti, Romania; David E. Manes, Oneonta, NY, USA; Ilyes Hamdi,
Lycée Voltaire, Doha, Qatar; Jean Heibig, Paris, France; Farrukh Mukhammadiev, Academic Lyceum Nr1,
Samarkand, Uzbekistan; Paul Revenant,Lycée Champollion,Grenoble,France; Sardor Bozorboyev, Lyceum
Romania; Yooree Ha, Ponte Vedra High School, Ponte Vedra, FL, USA.

J321. Let x, y, z be positive real numbers such that xyz(x + y + z) = 3. Prove that
1 1 1 54
+ + + ≥ 9.
x2 y 2 z 2 (x + y + z)2
Rewriting the inequality yields to
(xy)2 + (yz)2 + (zx)2 54

2 2 2
+ ≥9
x y z (x + y + z)2
Define now the new variables
x + y + z = 3u, xy + uyz + zx = 3v 2 , xyz = w3
and trivial AGM yields u ≥ v ≥ w. We also use the well known inequality
p
xy + yz + zx ≥ 3(xyz)(x + y + z) ⇐⇒ 3v 2 ≥ 3
that is v ≥ 1.
The inequality becomes
9v 4 − 6uw3 9
w3 u = 1 =⇒ 6
+ 2 ≥9
w u
that is
. 2
f (u) = (3v 4 − 2)u2 + 2 − 3 ≥ 0,
u
r
0 4 4 4 2
f (u) = 2u(3v − 2) − 3 = 0 ⇐⇒ u = u0 (v) =
u 3v 2 −2
√
If v ≤ 2/ 3 then u0 (v) ≤ 1, and this implies f (u) ≥ f (1).
2
f (1) = (3v 4 − 2)u2 + − 3 ≥ 3v 4 − 2 + 2 − 3 ≥ 3 · 1 − 2 + 2 − 3 = 0
u2
√
and this part of the proof is complete. Now let v > 2/ 3.
2 2
f (u) = (3v 4 − 2)u2 + − 3 > (4 − 2)u2 + 2 − 3 ≥ 4 − 3 = 1
u2 u
and also this part is complete.
Also solved by Seung Hwan An, Taft School, Watertown, CT, USA; Chaeyeon Oh, Episcopal High School,
Alexandra, VA, USA; Mehtaab Sawhney, USA; Yujin Kim, Stony Brook School, Stony Brook, NY, USA;
Misiakos Panagiotis ,Athens College (HAEF), Nea Penteli; Timothy Chon, Horace Mann School, Bronx,
NY, USA; Cody Johnson, USA; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania.

J322. Let ABC be a triangle with centroid G. The parallel lines through a point P situated in the plane
of the triangle to the medians AA0 , BB 0 , CC 0 intersect lines BC, CA, AB at A1 , B1 , C1 , respectively.
Prove that
3
A0 A1 + B 0 B1 + C 0 C1 ≥ P G.
2
Proposed by Dorin Andrica, Babeş-Bolyai University, Cluj-Napoca, Romania
Solution by Marius Stânean, Zalau, Romania

Let (x : y : z) be the barycentric coordinates of P with respect to triangle ABC so that the three (signed)
areas [P BC], [P CA], and [P AB] are in the ratio x : y : z. We have
P A1 [P AB]
= = x =⇒
AA0 [ABC]
−−0−→ −−→ −−→ −−→0 −−→ −−→ −−→
A A1 = P A1 − P G − GA = 3x · GA0 − P G − GA0 =

1 →− 1 1 →
− 1 1 →
− −−→
(3x − 1) − · A + − ·B+ − · C − PG =
3 2 3 2 3
3x − 1 →
− → − → − −−→
· −2 · A + B + C − P G
6
Analog obtain
−−0−→ 3y − 1 →
− →
− → − −−→
B B1 = · A − 2 · B + C − PG
6
−−0−→ 3z − 1 →
− → − →
− −−→
C C1 = · A + B − 2 · C − PG
6
Therefore
−−0−→ −−0−→ −−0−→ 1 − 3x → − 1 − 3y → − 1 − 3z →− −−→
A A1 + B B 1 + C C 1 = ·A+ ·B+ · C − 3 · PG =
2 2 2

3 1 →
− 1 →
− 1 →
− −−→
−x · A + −y · B + − z · C − 3 · PG =
2 3 3 3
3 −−→ −−→ 3 −−→
· P G − 3 · P G = − · P G.
2 2
Considering this we have
3 −−→ −−0−→ −−0−→ −−0−→

3
P G = − · P G = A A1 + B B1 + C C1 ≤

2 2
−−−→ −−−→ −−−→
0 0 0
A A1 + B B1 + C C1 = A0 A1 + B 0 B1 + C 0 C1 .
William Kang, Bergen County Academies, Hackensack, NJ, USA; Timothy Chon, Horace Mann School,
Bronx, NY, USA.

J323. In triangle ABC, √
A B C 5−1
cos + cos + cos = .
2 2 2 2
Prove that max(A, B, C) > 144◦ .
Solution by Henry Ricardo, New York Math Circle √

WLOG, assume that max{A, B, C} = A ≤ 162◦ . Then sin A ≥ sin 162◦ = sin 18◦ = 5−1
4 , 18◦ ≤ B + C <
√
180◦ , and sin(B + C) ≥ 5−1
4 .
Since sin B + sin C > sin(B + C) for 0 < B, C < 180◦ , we have
√
5−1
= sin A + sin B + sin C > sin A + sin(B + C)
2 √ √ √
5−1 5−1 5−1
> + = .
4 4 2
This contradiction establishes that max{A, B, C} > 162◦ .
Chaeyeon Oh, Episcopal High School, Alexandra, VA, USA; Mehtaab Sawhney, USA; Yujin Kim, Stony
Brook School, Stony Brook, NY, USA; William Kang, Bergen County Academies, Hackensack, NJ, USA; Ji
Eun Kim, Tabor Academy, Marion, MA, USA; Timothy Chon, Horace Mann School, Bronx, NY, USA; Cody
Johnson, USA; Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter, NH, USA; Arkady Alt, San Jose,
California, USA; Nicuşor Zlota, “Traian Vuia” Technical College, Focşani, Romania; Sardor Bozorboyev,
Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan; Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu,
Buzău, Romania; Yooree Ha, Ponte Vedra High School, Ponte Vedra, FL, USA.

J324. Let ABC be a triangle and let X, Y , Z be the reflections of A, B, C in the opposite sides. Let Xb ,
Xc be the orthogonal projections of X on AC, AB, Yc , Ya the orthogonal projections of Y on BA,
BC, and Za , Zb the orthogonal projections of Z on CB, CA, respectively. Prove that Xb , Xc , Yc , Ya ,
Za , Zb are concyclic.

We first prove the following claims:
Claim 1. The lines Xb Xc , Ya Yc , Za Zb are antiparallel to BC, AC, AB respectively.
E
F
B C
D
Xb
Xc
Proof of Claim 1. From the similar triangles 4AHE ∼ 4AXXb and 4AHF ∼ 4AXXc we get
AE : AXb = AH : AX, AF : AXc = AH : AX ⇒ AE : AXb = AF : AXc
Therefore 4AF E and 4AXc Xb are homotetic, so Xb Xc k EF , i.e. Xb Xc is antiparallel to BC. A similar
reasoning show that Ya Yc is antiparallel to AC and Za Zb is antiparallel to AB, as claimed.
Claim 2. The lines Xb Ya , Yc Zb , Xc Za are parallel to AB, BC, AC respectively.

Proof of Claim 2. Let α = ∠BAC, β = ∠ABC, γ = ∠ACB and let D, E, F be the orthogonal projections
of A, B, C on BC, CA, AB respectively, as shown in figure.

Yc
F Xb
H
B D Ya C
Xc
X
From the cyclic quadrilaterals BYa Y Yc and AXc XXb we have
Ya Yc = BY · sin β = 2 · AB · sin α sin β = AX sin α = Xb Xc

∠Xc Yc Ya = ∠BY Ya = γ = ∠AXXb = ∠Yc Xc Xb (1)
whence it follows that 4Xc Yc Xb and 4Xc Yc Ya are congruent (SAS). Therefore
∠Xc Xb Yc = ∠Xc Ya Yc
and this implies that Xc , Xb , Yc , Ya are concyclic, so
∠Ya Yc Xb = ∠Xb Xc Ya (2)
Adding (1) and (2) gives
∠Xc Yc Xb = ∠Xc Yc Ya + ∠Ya Yc Xb = ∠Yc Xc Xb + ∠Xb Xc Ya = ∠Yc Xc Ya
Therefore Xc Ya Xb Yc is an isosceles trapezoid. Thus Xc Yc k Ya Xb , i.e. AB k Xb Ya . We can argue similarly

to show that BC k Yc Zb and AC k Xc Za and the claim follows.
Returning to the original problem it suffices to notice that the above claims tell us that Xc Xb Ya Yc Zb Za
is a Tucker hexagon, so Xb , Xc , Yc , Ya , Za , Zb are concyclic, which is what we wanted to prove.
Also solved by Andrea Fanchini, Cantú, Italy; Mehtaab Sawhney, USA; Misiakos Panagiotis ,Athens
College (HAEF), Nea Penteli; Cody Johnson, USA; Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India;
Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania.

Senior problems
S319. Let a, b, c be positive real numbers such that a + b + c = 1. Prove that for any positive real number t,
at2 + bt + c bt2 + ct + a ct2 + at + b ≥ t3 .

First solution by the proposer

By Hölder’s inequality:
(at2 + bt + c)(bt2 + ct + a)(ct2 + at + b) = (at2 + bt + c)(a + bt2 + ct)(at + b + ct2 ) ≥ (at + bt + ct)3 = t3 .
Second solution by Daniel Lasaosa, Pamplona, Spain

After some algebra, the inequality rewrites as
abc t6 + 1 + a2 b + b2 c + c2 a t5 + 2t2 + ab2 + bc2 + ca2 2t4 + t +

+ a3 + b3 + c3 + 4abc t3 ≥ t3 .

Now, by the AM-GM inequality, we have t6 + 1 ≥ 2t3 , t5 + 2t2 ≥ 3t3 and 2t4 + t = 3t3 , with equality iff
t = 1. It therefore remains only to prove that
2abc + 3 a2 b + b2 c + c2 a + 3 ab2 + bc2 + ca2 + a3 + b3 + c3 + 4abc ≥ 1,

clearly true and with equality always because the LHS is nothing other than (a+b+c)3 = 13 . The conclusion
follows, equality holds iff t = 1.
Also solved by Ioan Viorel Codreanu, Satulung, Maramures, Romania; Seung Hwan An, Taft School, Wa-
tertown, CT, USA; Chaeyeon Oh, Episcopal High School, Alexandra, VA, USA; Mehtaab Sawhney, USA;
Yujin Kim, Stony Brook School, Stony Brook, NY, USA; Misiakos Panagiotis ,Athens College (HAEF), Nea
Penteli; William Kang, Bergen County Academies, Hackensack, NJ, USA; Timothy Chon, Horace Mann
School, Bronx, NY, USA; Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter, NH, USA; Yong Xi
Wang,East China Institute Of Technology, China; Adnan Ali, Student in A.E.C.S-4, Mumbai, India; Albert
Stadler, Herrliberg, Switzerland; Alok Kumar, Delhi, India; Arkady Alt, San Jose, California, USA; Bo-
dhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Marin Chirciu, Pites, ti, Romania; Farrukh Mukham-
madiev, Academic Lyceum Nr1, Samarkand, Uzbekistan; Nicuşor Zlota, “Traian Vuia” Technical College,
Focşani, Romania; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy; Samin Riasat,
University of Waterloo, ON, Canada; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan;
Titu Zvonaru, Comănes, ti, Romania and Neculai Stanciu, Buzău, Romania; Ángel Plaza, University of Las
Palmas de Gran Canaria, Spain; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia.

S320. Let ABC be a triangle with circumcenter O and incenter I. Let D, E, F be the tangency points of
the incircle with BC, CA, AB, respectively. Prove that line OI is perpendicular to angle bisector of
∠EDF if and only if ∠BAC = 60◦ .
Solution by Andrea Fanchini,Cantú, Italy
Using barycentric coordinates and the Conway’s notation, if ∠BAC = 60◦ , then
√
2 2 2 bc 2c2 − bc 2b2 − bc 3
a = b + c − bc, SA = S60◦ = , SB = , SC = , S= bc, (1)
2 2 2 2
Now line DE have equation

0
s−c s−b

s−c 0 s−a =0 ⇒ DE ≡ (s − a)x + (s − b)y − (s − c)z = 0

x y z
and line DF have equation

0
s−c s−b

s−b s−a 0 =0 ⇒ DF ≡ (s − a)x − (s − b)y + (s − c)z = 0

x y z
Now, the formula to calculate the angle between two lines is
SA (q1 − r1 )(q2 − r2 ) + SB (r1 − p1 )(r2 − p2 ) + SC (p1 − q1 )(p2 − q2 )

Sθ = S cot θ =

1 1 1

p1 q1 r1

p2 q2 r2
so applying this to the ∠EDF between the lines DE and DF we have
−a2 SA − b(a − c)SB + c(b − a)SC

SEDF =

1 1 1

s−a s−b c−s

s−a b−s s−c
and keeping in mind the (1), we have that
bc
SEDF = = S60◦
2
so ∠EDF = 60◦ .
Now we know that if θ is the oriented angle between the line px + qy + rz = 0 and a line d, the coordinates
of the infinite point of this line are
(pa2 + q(Sθ − SC ) − r(Sθ + SB ) : qb2 + r(Sθ − SA ) − p(Sθ + SC ) : rc2 + p(Sθ − SB ) − q(Sθ + SA ))
in our case, the angle between line DE ≡ (s − a)x + (s − b)y − (s − c)z = 0 and the angle bisector DL is
30◦ , so the coordinates of the infinite point of DL are
√ √
DL∞ (a2 (s − a) + (s −√ b)( 3S − SC ) + (s −√c)( 3S + SB ) :
b2 (s − b) − (s − c)( √3S − SA ) − (s − a)( √
3S + SC ) :
2
−c (s − c) + (s − a)( 3S − SB ) − (s − b)( 3S + SA ))
and using the (1) we obtain
DL∞ (a(a − b + 2c) : b(a − b − c) : c(2b − 2a − c)).
Line OI have equation

a b c
2
a SA b2 SB c2 SC =0 ⇒ OI ≡ bc(cSC − bSB )x − ac(cSC − aSA )y + ab(bSB − aSA )z = 0

x y z
so the infinite point of line OI using the (1) is
OI∞ (a(ab − 4bc + ac + b2 + c2 ) : b(−ab − 2bc + 2ac − b2 + 2c2 ) : c(2ab − 2bc − ac + 2b2 − c2 ))
Now we know that two lines with infinite points (f : g : h) and (f 0 : g 0 : h0 ) are perpendicular to each other
if and only if
SA f f 0 + SB gg 0 + SC hh0 = 0
and this is the case of our two infinite points DL∞ and OI∞ , so line OI is perpendicular to angle bisector
of ∠EDF and also the contrary is immediately proved.
Also solved by Daniel Lasaosa, Pamplona, Spain; Mehtaab Sawhney, USA; Yujin Kim, Stony Brook
School, Stony Brook, NY, USA; Misiakos Panagiotis ,Athens College (HAEF), Nea Penteli; Ercole Su-
ppa, Teramo, Italy; Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Sardor Bozorboyev, Lyceum
Romania.

S321. Let x be a real number such that xm (x + 1) and xn (x + 1) are rational for some relatively prime
positive integers m and n. Prove that x is rational.
Solution by Yujin Kim, Stony Brook School, Stony Brook, NY, USA
Suppose m > n and denote m − n = q > 0. One can assume x 6= 0, −1.
xm (x + 1)
∈ Q → x2 ∈ Q
xn (x + 1)
[xn (x + 1)]m xmn (x + 1)m

∈ Q → ∈ Q → (x + 1)q ∈ Q
[xm (x + 1)]n xmn (x + 1)n
We will show that if x2 , (x + 1)2 are both rational and x is real.Thus x is rational.
Consider two polynomials: R(z) = z 2 − x2 and S(z) = (z + 1)2 − (x + 1)2
Note that both R and S have rational coefficients and R(x) = S(x) = 0. Hence R and S have x as a
common root. Suppose that x ∈ / Q. Denote by m(z) the minimal polynomial of x that has the smallest degree
polynomial with rational coefficients which has x as a root. Such polynomial exists since R(x) = S(x) = 0.
Since x is irrational, it follows that degree m ≥ 2.
On the other hand, both R(z) and S(z) are multiples of m(z). This means that R and S have some other
common root besides x. Denote such a root by y. We have y 6= x. Note that y could be a complex unreal
number.
We have R(y) = S(y) = 0 → y Σ = xΣ and (1 + y)Σ = (1 + x)Σ .
Consequently, |y|Σ = |x|Σ and |1 + y|Σ = |1 + x|Σ →
(
|y| = |x| Recall that x ∈ R
|1 + y| = |1 + x| Denote y = a + ib
√ p
Then, |a + ib| = |x| → a2 + b2 = |x| → a2 + b2 = x2 Also |1 + a + ib| = |1 + x| → (1 + a)2 + b2 =
|1 + x| → (1 + a)2 + b2 = (1 + x)2 → 1 + 2a + a2 + b2 = 1 + 2x + x2 → a = x and since a2 + b2 = x2 this
implies b = 0.
But then y = a + ib = x + i · 0 = x, contradiction. The conclusion follows.
Also solved by Seung Hwan An, Taft School, Watertown, CT, USA; William Kang, Bergen County
Academies, Hackensack, NJ, USA; Timothy Chon, Horace Mann School, Bronx, NY, USA; Yooree Ha,
Ponte Vedra High School, Ponte Vedra, FL, USA.

S322. Let ABCD be a cyclic quadrilateral. Points E and F lie on the sides AB and BC, respectively, such
that ∠BF E = 2∠BDE. Prove that
EF FC CD
= + .
AE AE AD
Solution by Andrea Fanchini, Cantú, Italy
Let Z be the intersection point between BC and the circumcircle of 4DEB.

Then ∠BZE = ∠BDE = δ
Now if we consider 4EF Z, we have
∠F EZ + ∠EZF + ∠ZF E = π ⇒ ∠F EZ = π − δ − (π − 2δ) = δ
Therefore 4EF Z is isosceles and then EF = F Z.
Now being ABCD a cyclic quadrilateral
∠BAD + ∠BCD = π ⇒ ∠DCZ = ∠BAD
Furthermore being BEDZ a cyclic quadrilateral
∠BED + ∠BZD = π ⇒ ∠AED = ∠BZD
So 4DCZ and 4DAE are similar, that is

CZ AE
=
CD AD
but CZ = F Z − F C or also, being EF = F Z, we have that CZ = EF − F C.Therefore
EF − F C AE EF FC CD
= ⇒ = +
CD AD AE AE AD
so the equality is proved.
Also solved by Yujin Kim, Stony Brook School, Stony Brook, NY, USA; William Kang, Bergen County
Academies, Hackensack, NJ, USA; Jhiseung Daniel Hahn, Phillips Exeter Academy, Exeter, NH, USA;
Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Ta-
shkent, Uzbekistan.

S323. Solve in positive integers the equation
x + y + (x − y)2 = xy.
Proposed by Neculai Stanciu and Titu Zvonaru, Romania

Denote s = x + y, d = x − y, and note that the proposed equation rewrites as
s2 − d2 = 4xy = 4x + 4y + 4(x − y)2 = 4s + 4d2 , (s − 2)2 − 5d2 = 4.
This is clearly a Pell-like equation c2 − 5d2 = 4, where s − 2 = c and d are integers of the same parity iff
x, y are integers, and all of whose infinite solutions (cn , dn ) may be found through the recurrent relations
cn+2 = 3cn+1 − cn , dn+2 = 3dn+1 − dn , n ≥ 0,
with initial conditions (c0 , d0 ) = (2, 0) and (c1 , d1 ) = (3, 1). Therefore, all solutions are of the form
√ !2n √ !2n √ !2n √ !2n
5+1 5−1 1 5+1 1 5−1
sn = + + 2, dn = √ −√ ,
2 2 5 2 5 2
or equivalently,
√ !2n+1 √ !2n+1
sn + dn 1 5+1 1 5−1
xn = =√ +√ + 1,
2 5 2 5 2
√ !2n−1 √ !2n−1
sn − dn 1 5+1 1 5−1
yn = =√ +√ + 1,
2 5 2 5 2
where n can take any non-negative integer value. We can readily check that indeed the following relation
holds:
√ √
 !4n !4n 
1 5+1 5−1
xn yn − xn − yn =  + − 2 = dn 2 = (xn − yn )2 .
5 2 2
We conclude that all solutions of the proposed equation are those (xn , yn ) already found, or the result of
interchanging xn , yn by symmetry in the problem.
Also solved by Brian Bradie, Christopher Newport University, Newport News, VA, USA; Chaeyeon Oh,
Episcopal High School, Alexandra, VA, USA; Mehtaab Sawhney, USA; Cody Johnson, USA; Jhiseung Daniel
Hahn, Phillips Exeter Academy, Exeter, NH, USA; Adnan Ali, Student in A.E.C.S-4, Mumbai, India; Albert
Stadler, Herrliberg, Switzerland; Alok Kumar, Delhi, India; Arber Avdullahu, Mehmet Akif College, Kosovo;
Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; David E. Manes, Oneonta, NY, USA; Sardor
Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan.

S324. Find all functions f : S → S satisfying
f (x)f (y) + f (x) + f (y) = f (xy) + f (x + y)
for all x, y ∈ S when (i) S = Z; (ii) S = R.

For the first part where S = Z, putting x = y = 0 we obtain f (0) = 0. Then, putting (x, y) = (1, −1), we
have
f (1) + f (−1) + f (1)f (−1) = f (0) + f (−1) ⇒ f (1)[f (−1) + 1] = 0.
(a) f (1) = 0, then,
f (x − 1) + f (1) + f (1)f (x − 1) = f (x) + f (x − 1) ⇒ f (x) = 0 ∀x ∈ Z
(b) f (−1) = −1, then,
f (2) + f (−1) + f (−1)f (2) = f (1) + f (−2) ⇒ f (1) + f (−2) = f (−1)

f (−2) + f (1) + f (1)f (−2) = f (−1) + f (−2) ⇒ f (−1) = f (1) + f (1)f (−2)
Combining the above two results, f (−2)[f (1) − 1] = 0.

(i) f (1) = 1, then,
f (x) + f (1) + f (x)f (1) = f (x) + f (x + 1) ⇒ f (x) + 1 = f (x + 1) ∀x ∈ Z
and due to induction, f (x) = x ∀x ∈ Z.

(ii) f (−2) = 0, then, f (1) + f (−2) = f (−1) = −1 ⇒ f (1) = −1 and so,
f (x) + f (1) + f (x)f (1) = f (x) + f (x + 1) ⇒ f (x) + f (x + 1) + 1 = 0
and replacing x by x + 1 in the above relation,
f (x + 1) + f (x + 2) + 1 = 0
and so,
(1.0) · · · f (x) = f (x + 2),
thus applying induction, we have f (2n) = 0 and f (2n − 1) = −1 for all n ∈ Z or

0, when x is even
f (x) =
−1, when x is odd
Now we come to the second part where S = R. We have the same conditions as before except that the
domain is the real set. So,
(a) f (1) = 0, then,
f (x − 1) + f (1) + f (1)f (x − 1) = f (x) + f (x − 1) ⇒ f (x) = 0 ∀x ∈ R
(b) f (−1) = −1 and so, f (−2)[f (1) − 1] = 0.

(i) f (1) = 1, then,
f (x) + f (−1) + f (x)f (−1) = f (x − 1) + f (−x) ⇒ f (x − 1) = −f (−x) − 1.
And also, in

(1.1) · · · f (x + 1) = f (x) + 1 (as it was in the previous part),
replacing x + 1 by x, we have f (x) = f (x − 1) + 1 = −f (−x). Thus we conclude that −f (x) = f (−x) ∀x ∈ R.

And replacing x by x + 1 in (1.1), f (x + 2) = f (x + 1) + 1 = f (x) + 2.
So, f (x + 1) = f (x) + 1 ⇒ f (2) = f (1) + 1 = 2. Thus,
f (2) + f (x) + f (2)f (x) = f (2x) + f (x + 2) ⇒ 2f (x) = f (2x) ∀x ∈ R.
Now,
f (x + y) + f (x − y) + f (x + y)f (x − y) = f (x2 − y 2 ) + f (2x)

f (x + y) + f (y − x) + f (x + y)f (y − x) = f (y 2 − x2 ) + f (2y)
adding up the above two relations and using f (−a) = −f (a) ∀a ∈ R, we have f (x + y) = f (x) + f (y)
implying f (x)f (y) = f (xy), thus the function is both additive and multiplicative, implying that f (x) = x
∀x ∈ R and f (x) = 0 ∀x ∈ R. It is easy to see that both of them satisfy the functional equation.
(ii)
0, when x is even ∀x ∈ Z
f (x) =
−1, when x is odd ∀x ∈ Z
So, according to the previous part we have the result (1.0),
f (x + 2) = f (x)
1
which here, holds true for all real x. So, x = yields
2
! !
1 5
f =f · · · (1.2)
2 2
! ! !
1 1 5
Whereas, f + f (2) + f f (2) = f (1) + f implying that
2 2 2
! !
1 5
f = −1 + f
2 2
contradicting (1.2). Thus in the real domain, f (−1) = −1 and f (−2) = 0 cannot both be simultaneously
true. Thus in summary, we have the functions as:
(1) When S = Z,
f (x) = 0
f (x) = x

0, when x is even
f (x) =
−1, when x is odd
(2) When S = R.
f (x) = 0
f (x) = x
Also solved by Mehtaab Sawhney, USA; Misiakos Panagiotis ,Athens College (HAEF), Nea Penteli; Cody
Johnson, USA; Arber Avdullahu, Mehmet Akif College, Kosovo; Bodhisattwa Bhowmik, RKMV, Agartala,
Tripura, India; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan.

U319. Let A, B, C be the measured (in radians) of the angles of a triangle with circumradius R and inradius
r . Prove that
A B C 2R
+ + ≤ −1
B C A r
Proposed by Nermin Hodžić, Bosnia and Herzegovina and Salem Malikić, Canada
We employ two known inequalities. The first one is
1 27 R
≤ 3
ABC 2π r
and can be found in problem 3757 of Crux Mathematicorum, vol.39–6, 2013. The second one is a known
inequality
4
a2 c + b2 a + c2 b ≤ (a + b + c)3 − abc
27
given that a, b, c ≥ 0.
Therefore, we have
A B C A2 C + B 2 A + C 2 B
+ + = ≤
B C A ABC
1 4
(A + B + C)3 − ABC =
ABC 27
1 4 3 27 R 4 3 R
= π −1≤ 3 π −1=2 −1
ABC 27 2π r 27 r
Also solved by Dragoljub Milošević, Gornji Milanovac, Serbia; Cody Johnson, USA; AN-anduud Problem
Solving Group, Ulaanbaatar, Mongolia.

U320. Evaluate X 2n
.
22n + 1
n≥0
Proposed by Titu Andreescu, University of Texas at Dallas

Denote for all non-negative integer N ,
N
X 2n
SN = 2n
.
2 +1
n=0
We show that for all positive integer N , we have
2N +1
SN = 1 − 2N +1
.
2 −1
The result is clearly true for N = 0 and N = 1, since
1 2 21 2 4 22
S0 = = 1 − = 1 − 21 , S1 = S0 + =1− = 1 − 22 .
3 3 2 −1 5 15 2 −1
If the result is true for N − 1, then for N we have
N
N 22 + 1 − 22N − 1
2 2N +1
SN = SN −1 + 2N = 1 − 2N = 1 − ,
22N + 1 22N − 1 22·2N − 1

2 +1
or the result is true by inducton for all non-negative integer N . It follows that
X 2n 2N x
2n
= lim SN −1 = 1 − lim 2N = 1 − lim x = 1,
2 + 1 N →∞ N →∞ 2 −1 x→∞ 2 −1
n≥0
since as it is well known, the exponential function increases much more rapidly than the linear function, and
where we have defined x = 2N . The conclusion follows.
Also solved by Reiner Martin, Bad Soden-Neuenhain, Germany; AN-anduud Problem Solving Group,
Ulaanbaatar, Mongolia; David E. Manes, Oneonta, NY, USA; Arkady Alt, San Jose, California, USA; Alok
Kumar, Delhi, India; Albert Stadler, Herrliberg, Switzerland; Adnan Ali, Student in A.E.C.S-4, Mumbai,
India; Yong Xi Wang,East China Institute Of Technology, China; Cody Johnson, USA; Ji Eun Kim, Tabor
Academy, Marion, MA, USA; William Kang, Bergen County Academies, Hackensack, NJ, USA; Yujin Kim,
Stony Brook School, Stony Brook, NY, USA; Mehtaab Sawhney, USA.

U321. Consider the sequence of polynomials (Ps )s≥1 defined by
0
Pk+1 (x) = (xa − 1)Pk (x) − (k + 1)Pk (x), k = 1, 2, . . . ,
where P1 (x) = xa−1 and a is an integer greater than 1.

1. Find the degree of Pk
2. Determine Pk (0)
Solution by the proposer

1. We have deg P1 = a − 1, and from the recursive relation we obtain deg P2 = 2a − 2. Now, a simple
inductive argument shows that deg Pk = k(a − 1).
1
2. Let us consider the function f : (−1, 1) → R, defined by f (x) = xa −1 . Observe that
axa−1 P1 (x)
f 0 (x) = a 2
=a a .
(x − 1) (x − 1)2
One can prove immediately by induction that
Pk (x)
f (k) (x) = a ,
(xa− 1)k+1
where we use the recursive relation in the definition of the sequence. This implies
(−1)k+1 (k)
Pk (0) = f (0),
a
and we reduce the problem of finding the value of Pk (0) to determine f (k) (0). In this respect, using the
geometric series we have
∞
1 X
f (x) = a =− xas ,
x −1
s=0
and we obtain (
(k) −k! if a|k
f (0) =
0 otherwise
Finally,
(−1)k k!
(
(−1)k+1 (k) if a|k
a
Pk (0) = f (0) =
a 0 otherwise
Also solved by Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India.

U322. Evaluate
∞
X 16n2 − 12n + 1
.
n(4n − 2)!
n=1
Proposed by Titu Andreescu, USA and Oleg Mushkarov, Bulgaria
First solution by Brian Bradie, Christopher Newport University, Newport News, VA, USA
Note that
16n2 − 12n + 1 4n(4n − 3) 1
= +
n(4n − 2)! n(4n − 2)! n(4n − 2)!
4n − 3 4n − 1
= 4· +4·
(4n − 2)! (4n)!

1 1 1 1
= 4 − + − .
(4n − 3)! (4n − 2)! (4n − 1)! (4n)!
Therefore,
∞ ∞
16n2 − 12n + 1

X X 1 1 1 1
= 4 − + −
n(4n − 2)! (4n − 3)! (4n − 2)! (4n − 1)! (4n)!
n=1 n=1
∞ ∞
!
X 1 X 1
= 4 (−1)n−1 = 4 1 − (−1)n
n! n!
n=1 n=0

1
= 4 1− .
e
Second solution by Brian Bradie, Christopher Newport University, Newport News, VA, USA
Write
16n2 − 12n + 1 = (4n − 2)(4n − 3) + 2(4n − 2) − 1,
so that
∞ ∞ ∞ ∞
X 16n2 − 12n + 1 X 1 X 1 X 1
= +2 −
n(4n − 2)! n(4n − 4)! n(4n − 3)! n(4n − 2)!
n=1 n=1 n=1 n=1
= f 00 (1) + 2f 0 (1) − f (1),
where
∞
X x4n−2
f (x) = .
n(4n − 2)!
n=1
Next, consider the function
∞
X x4n−2
g(x) = ,
(4n − 2)!
n=1
and note that g (4) (x) − g(x) = 0 subject to the initial conditions g(0) = 0, g 0 (0) = 0, g 00 (0) = 1, and
g 000 (0) = 0. Thus,
1
g(x) = (cosh x − cos x).
2
To obtain f , multiply g by x, integrate term-by-term, determine the constant of integration using the initial
condition f (0) = 0, multiply by 4, and divide by x2 . This yields
2
f (x) = (2 + x sinh x − cosh x − x sin x − cos x).
x2

Now,
2 4
f 0 (x) = 2
(x cosh x − x cos x) − 3 (2 + x sinh x − cosh x − x sin x − cos x),
x x
2 8
f 00 (x) = 2
(x sinh x + cosh x + x sin x − cos x) − 3 (x cosh x − x cos x) +
x x
12
(2 + x sinh x − cosh x − x sin x − cos x),
x4
so that

1
f (1) = 2 2 − − sin 1 − cos 1 ,
e

1
f 0 (1) = 2(cosh 1 − cos 1) − 4 2 − − sin 1 − cos 1 ,
e

00 1
f (1) = 2(e + sin 1 − cos 1) − 8(cosh 1 − cos 1) + 12 2 − − sin 1 − cos 1 ,
e
and
∞
16n2 − 12n + 1

X
00 0 1
= f (1) + 2f (1) − f (1) = 4 1 − .
n(4n − 2)! e
n=1
Also solved by Daniel Lasaosa, Pamplona, Spain; Bodhisattwa Bhowmik, RKMV, Agartala, Tripura,
India; Arkady Alt, San Jose, California, USA; Albert Stadler, Herrliberg, Switzerland; Cemal Kadirov, Is-
tanbul University, Turkey; AN-anduud Problem Solving Group, Ulaanbaatar, Mongolia; Cody Johnson, USA;
Ji Eun Kim, Tabor Academy, Marion, MA, USA; William Kang, Bergen County Academies, Hackensack,
NJ, USA; Yujin Kim, Stony Brook School, Stony Brook, NY, USA; Mehtaab Sawhney, USA; Chaeyeon Oh,
Episcopal High School, Alexandra, VA, USA.

U323. Let X and Y be independent random variables following a uniform distribution

1 0 < x < 1,
pX (x) =
0 otherwise.
What is the probability that inequality X 2 + Y 2 ≥ 3XY is true?

Note first that the proposed inequality can be written as
√ ! √ !
3+ 5 3− 5
Y − X Y − ≥ 0,
2 2
√ √ √
3+ 5 3− 5 3− 5
ie since X, Y take only non-negative values, we must have either Y ≥ 2 X ≥ 2 X or Y ≤ 2 X ≤
√
3+ 5
2 X. This defines inside the unit square
with vertices (X, Y ) = (0, 0), (0, 1), (1, 1), (1, 0) two triangles,
√ √
3− 5
one with vertices (X, Y ) = (0, 0), (1, 0), 1, 2 , and one with vertices (X, Y ) = (0, 0), (0, 1), 3−2 5 , 1 ,
inside which or on whose boundaries the inequality is true, being false in the rest of the unit square. Since
both variables X, Y are independent and uniform,
√
the probability of (X, Y ) falling inside or on the boundary
3− 5
of one of these triangles equals the sum 2 of the areas of both triangles, divided by the area 1 of the
unit square. The probability is therefore √
3− 5
.
2
Also solved by Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy; Cody John-
son, USA; Ji Eun Kim, Tabor Academy, Marion, MA, USA; William Kang, Bergen County Academies,
Hackensack, NJ, USA; Yujin Kim, Stony Brook School, Stony Brook, NY, USA.

U324. Let f : [0, 1] → R be a differentiable function such that f (1) = 0. Prove that there is c ∈ (0, 1) such
that |f (c)| ≤ |f 0 (c)|

We will prove that for all M > 0 there is c ∈ (0, 1) such that M |f (c)| ≤ |f 0 (c)|.
If there is a c ∈ (0, 1) such that f (c) = 0 then the inequality is trivial. Otherwise by continuity f has constant
sign and without loss of generality we may assume that f (x) > 0 in (0, 1). By the Mean Value Theorem, for
all t ∈ (1/2, 1) there is ct ∈ (1/2, t) such that
ln(f (t)) − ln(f (1/2)) f 0 (ct )

= .
t − 1/2 f (ct )
The condition f (1) = 0 implies that the LHS tends to −∞ as t → 1− .

Hence for all M > 0 there is a ct such that
f 0 (ct )
≤ −M.
f (ct )
Therefore f 0 (ct ) < 0 and

|f 0 (ct )| −f 0 (ct )
= ≥ M.
|f (ct )| f (ct )
Also solved by Reiner Martin, Bad Soden-Neuenhain, Germany; Paolo Perfetti, Università degli studi di
Tor Vergata Roma, Roma, Italy; Corneliu Mănescu- Avram, Transportation High School, Ploieşti, Romania;
Arkady Alt, San Jose, California, USA; Cody Johnson, USA.

Olympiad problems
O319. Let f (x) and g(x) be arbitrary functions defined for all x ∈ R. Prove that there is a function h(x)
such that (f (x) + h(x))2014 + (g(x) + h(x))2014 is an even function for all x ∈ R.

It is well-known that any function f may be written as f (x) = fe (x) + fo (x) where fe and fo (x) are
f (x) + f (−x) f (x) − f (−x)
respectively the even and odd part of function f . Also fe (x) = while fo (x) = .
2 2
Let us take h(x) = −fo (x) − ge (x). Then, f (x) + h(x) = fe (x) − ge (x), and g(x) + h(x) = −fo (x) + go (x).
Therefore,
F (x) = (f (x) + h(x))2014 + (g(x) + h(x))2014

= (fe (x) − ge (x))2014 + (−fo (x) + go (x))2014 .
Then,
F (−x) = (fe (−x) − ge (−x))2014 + (−fo (−x) + go (−x))2014

= (fe (x) − ge (x))2014 + (fo (x) − ge (x))2014
= F (x).
Also solved by Daniel Lasaosa, Pamplona, Spain; Yassine Hamdi, Lycee du Parc, Lyon, France; George
- Petru Scărlătescu, Pites, ti, Romania; Misiakos Panagiotis, Athens College (HAEF), Nea Penteli; Mehtaab
Sawhney, USA; Michael Tang, Edina High School, MN, USA.

O320. Let n be a positive integer and let 0 < yi ≤ xi < 1 for 1 ≤ i ≤ n. Prove that
1 − x1 . . . xn 1 − x1 1 − xn
≤ + ··· + .
1 − y1 . . . yn 1 − y1 1 − yn
Proposed by Angel Plaza, Universidad de Las Palmas de Gran Canaria, Spain
Solution by Marius Stânean, Zalau, Romania

We can prove this inequality by mathematical induction. For n = 1 it’s obvious. For n = 2 we have
1 − x1 x2 1 − x1 1 − x2
≤ + .
1 − y1 y2 1 − y1 1 − y2
y1 y2
Let a = x1 ≤ 1 and b = x2 ≤ 1, so the inequality becomes
1 − x1 x2 1 − x1 1 − x2
≤ + ⇐⇒
1 − abx1 x2 1 − ax1 1 − bx2
(1 − x1 )(1 − x2 ) + ax1 x2 (1 − x1 ) + bx1 x2 (1 − x2 ) − 3abx1 x2 + abx21 x22 +
abx1 x2 (x1 + x2 ) + a2 bx21 x2 + ab2 x1 x22 − ab(a + b)x21 x22 ≥ 0
dividing this inequality with x1 x2 we obtain
(1 − x1 )(1 − x2 )
+ ab(1 − x1 )(1 − x2 ) + 2ab(x1 + x2 − 2)+
x1 x2
a2 x1 (1 − x2 ) + ab2 x2 (1 − x1 ) + a(1 − x1 ) + b(1 − x2 ) ≥ 0 ⇐⇒
(1 − x1 )(1 − x2 )
+ a(1 − x1 )(b2 x2 − 2b + 1) + b(1 − x2 )(a2 x1 − 2a + 1) + ab(1 − x1 )(1 − x2 ) ≥ 0 ⇐⇒
x1 x2
(1 − x1 )(1 − x2 )
+ ab(1 − x1 )(1 − x2 ) + a(1 − x1 )(b − 1)2 +
x1 x2
b(1 − x2 )(a − 1)2 − ab(a + b)(1 − x1 )(1 − x2 ) ≥ 0 ⇐⇒

1
(1 − x1 )(1 − x2 ) + ab − ab(a + b) + a(1 − x1 )(b − 1)2 + b(1 − x2 )(a − 1)2 ≥ 0
x1 x2
which is true because
1
+ ab − ab(a + b) ≥ 1 + ab − a − b = (1 − a)(1 − b) ≥ 0.
x1 x2
Suppose that the original inequality holds for n − 1 ∈ N and we want to prove it for n. Therefore we
have
1 − x1 · · · xn−1 1 − x1 1 − xn−1
≤ + ... +
1 − y1 · · · yn−1 1 − y1 1 − yn−1
but Y = y1 y2 · · · yn−1 ≤ x1 x2 · · · xn−1 = X < 1 and then applying the inequality for X, Y , respectively xn ,
yn we obtain
1 − x1 · · · xn 1 − X · xn 1−X 1 − xn
= ≤ + ≤
1 − y1 · · · yn 1 − Y · yn 1−Y 1 − yn
1 − x1 1 − xn−1 1 − xn
+ ... + + .
1 − y1 1 − yn−1 1 − yn
Also solved by Daniel Lasaosa, Pamplona, Spain; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent,
Uzbekistan; Paolo Perfetti, Università degli studi di Tor Vergata Roma, Roma, Italy; Yassine Hamdi, Lycee
du Parc, Lyon, France; Arkady Alt, San Jose, California, USA; Adnan Ali, Student in A.E.C.S-4, Mum-
bai, India; William Kang, Bergen County Academies, Hackensack, NJ, USA; Misiakos Panagiotis, Athens
College (HAEF), Nea Penteli; Yujin Kim, Stony Brook School, Stony Brook, NY, USA; Mehtaab Sawhney,
USA.

O321. Each of the diagonals AD, BE, CF of the convex hexagon ABCDEF divides its area in half. Prove
that
AB 2 + CD2 + EF 2 = BC 2 + DE 2 + F A2 .
B
Q
E
P
C D
We begin by proving the following preliminary results:
Claim 1. We have AC/DF = CE/F B = EA/BD.

Proof of Claim 1. Let P = AD ∩ CF . Observe that the areas of triangles 4AP F and 4DP C are equal
because
[AP F ] = [ADEF ] − [DEF P ] = [CDEF ] − [DEF P ] = [DP C]
Therefore AP · P F = CP · P D, or AP/P D = CP/P F .
Thus 4AP C ∼ 4DP F (SAS) whence ∠CAP = ∠F DP and AC k DF .
In a similar way, we can prove that BF k CE and AE k BD, so that 4ACE ∼ 4DF B and the Claim
1 follows.
Claim 2. The three diagonals AD, BE, CF are concurrent.

Proof of Claim 2. Let P = AD ∩ CF , Q = AD ∩ BE. As proved in Claim 1 we have 4AP C ∼ 4DP F ,
so
AP AC
= (1)
PD DF
In a similar way, we can prove that
AQ EA
= (2)
QD BD
From (1) and (2), taking into account of Claim 1, we have AP/P D = AQ/QD so that P = Q. Hence,
the three diagonals AD, BE, CF concur in P , as claimed.
Coming back to the proposed problem, note first that the pairs of triangles 4F AP and 4CDP , 4ABP
and 4DEP , 4BCP and 4EF P are equivalent.

A
F
a
f
B
b 

P


 e E
c
d
C D
Therefore denoting by a, b, c, d, e, f the lengths of P A, P B, P C, P D, P E, P F respectively, we have
a · f = c · d, a · b = d · e, b·c=e·f (3)
Finally, putting α = ∠F P A, β = ∠AP B, γ = ∠BP C, the cosinus law yields
AB 2 + CD2 + EF 2 = a2 + b2 − 2ab cos β + c2 + d2 − 2cd cos α + e2 + f 2 − 2ef cos γ

BC 2 + DE 2 + F A2 = b2 + c2 − 2bc cos γ + d2 + e2 − 2de cos β + f 2 + a2 − 2f a cos α
The above equalities and (3) gives
AB 2 + CD2 + EF 2 = BC 2 + DE 2 + F A2
which is precisely what we want to prove
Also solved by Daniel Lasaosa, Pamplona, Spain; Titu Zvonaru, Comănes, ti, Romania and Neculai Stan-
ciu, Buzău, Romania; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan; Farrukh Mu-
khammadiev, Academic Lyceum Nr1, Samarkand, Uzbekistan; William Kang, Bergen County Academies,
Hackensack, NJ, USA; Yujin Kim, Stony Brook School, Stony Brook, NY, USA.

O322. Let ABC be a triangle with circumcircle Γ and let M be the midpoint of arc BC not containing A.
Lines `b and `c passing through B and C, respectively, are parallel to AM and meet Γ at P 6= B
and Q 6= C. Line P Q intersects AB and AC at X and Y , respectively, and the circumcircle of AXY
intersects AM again at N .
Prove that the perpendicular bisectors of BC, XY , and M N are concurrent.
A
Q
Y
V
X O
P N
B C
W
The parallelisms BP k AM and CQ k AM imply

1
∠P BX = ∠BAC = ∠QCY (1)
2
The cyclic quadrilateral BCQP yields
∠P BC + ∠P QC = 180◦ (2)
From (1) and (2) we obtain
∠XBC + ∠XY C = 180◦ − ∠P QC − ∠P BX + ∠XY C =

= 180◦ − ∠P QC − ∠QCY + ∠XY C =
= ∠QY C + ∠XY C = 180◦
so BXY C is a cyclic quadrilateral.
Let V , W denote the centers of (AXY ) and (BXY C) respectively. Observe that V , N , W are colli-
near since N is the midpoint of the arc XY and V W is the perpendicular bisector of XY .
Since ∠P BX = ∠QCY we have AP = AQ, so A lies on the perpendicular bisector of P Q. Therefore

OA is the perpendicular bisector of P Q and this implies that OA ⊥ P Q.
Now, from V W ⊥ P Q and OA ⊥ P Q, it follows that OA k V W .

Therefore we have
∠W M N = ∠OAM = ∠AN V = ∠W N M ⇒ WN = WM
hence W belongs to the perpendicular bisector of M N .
Clearly W also belongs to the perpendicular bisectors of BC and XY so the proof is complete.
Also solved by Daniel Lasaosa, Pamplona, Spain; Andrea Fanchini, Cantú, Italy; Sardor Bozorboyev, Ly-
ceum S.H.Sirojjidinov, Tashkent, Uzbekistan; Farrukh Mukhammadiev, Academic Lyceum Nr1, Samarkand,
Uzbekistan; Bodhisattwa Bhowmik, RKMV, Agartala, Tripura, India; Misiakos Panagiotis, Athens College
(HAEF), Nea Penteli; Mehtaab Sawhney, USA.

1 2 n
O323. Prove that the sequence 22 + 1, 22 + 1, . . . , 22 + 1, . . . and an arbitrary infinite increasing arithmetic
sequence have either infinitely many terms in common or at most one term in common.

Assume to the contrary that there exists an arbitrary infinite increasing arithmetic progression (or sequence)
{an }n≥1 with common difference equal to d, such that it has a terms in common with the sequence {Fn }n≥1 =
1 2 n
22 + 1, 22 + 1, · · · , 22 + 1, · · · , and 2 ≤ a < ∞. Then assume that the last two terms which are common to
j m j m j
both the sequences, are ak = 22 +1 and a` = 22 +1 and k < ` ⇔ j < m. Then, d|a` −ak ⇔ d|22 (22 −2 −1).
m m+1 m
If d = 2b , b ≤ 2n , then we have d|22 (22 −2 − 1) ⇔ d|Fm+1 − Fm and thus there ∃ a u such that
a`+u = Fm+1 . But this is a contradiction to the assumption that the sequences {an } and {Fn } share only a
terms in common and 2 ≤ a < ∞. Thus if {an } and {Fn } share more than one term, then using the above
method we can generate infinitely many common terms.
m j m j m−j
Next, let d have an odd factor say d0 . Then, 22 −2 ≡ 1 (mod d0 ) ⇒ (22 −2 )2 ≡ 1 (mod d0 ). Thus
m 2m−j −2m
d0 |22 (22 − 1) ⇔ d|F2m−j − Fm . This shows the existence of a positive integer r such that a`+r =
F2m−j . This again contradicts our assumption, thus there exist infinitely many common terms in {an } and
{Fn } if they have more than one term in common, or they share at most one common term.
Note: An example of the arithmetic sequence which shares only one term with {Fn }n≥1 is the sequence
5, 15, 25, 35, · · · which shares only 5 as a common term and none else.
By the Claim, the two describe sequences have either zero, one, or infinitely many terms in common.
Note: We may easily construct infinite increasing arithmetic sequences with no terms in common by
taking sequences of non-integers, or of even integers. We may also easily construct infinite increasing ari-
thmetic sequences with exactly one term in common, by taking one of its terms in common and an irra-
tional difference, or by appropriately choosing the starting point and the difference. For example, defining
1
sm = 22 + 1 + (m − 1)25 for m = 1, 2, . . . , every term of this arithmetic sequence is congruent to 1 modulus
1 n
22 = 4, hence not of the form 22 + 1, except for n = 1 when m = 1, which is the only common term, and
there are no others.
Also solved by Reiner Martin, Bad Soden-Neuenhain, Germany; Mehtaab Sawhney, USA; Misiakos Pa-
nagiotis, Athens College (HAEF), Nea Penteli; Daniel Lasaosa, Pamplona, Spain; Samin Riasat, University
of Waterloo, ON, Canada; Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan.

O324. Let a, b, c, d be nonnegative real numbers such that a3 + b3 + c3 + d3 + abcd = 5. Prove that
abc + bcd + cda + dab − abcd ≤ 3
Solution by Semchankau Aliaksei, Moscow Institute of Physics and Technology

We will prove more general problem: let (a3 + b3 + c3 + d3 )t + abcd = 5t4 , t > 0, prove that
(abc + bcd + cda + dab)t − abcd ≤ 3t4

In required ineaquality t = 1.
Let r = abc + bcd + cda + dab, S3 = a3 + b3 + c3 + d3 , s = abcd. Then we have S3 t + s = 5t4 and we have
to prove that rt − s ≤ 3t4 . Obviously t4 = S3 t+s
5 , so, our inequality transforms to
3
rt − s ≤ (S3 t + s) ⇔ 5rt − 5s ≤ 3S3 t + 3s ⇔ (5r − S3 )t ≤ 8s
5
If 5r − 3S3 ≤ 0, then it is obvious, so we can conclude that 5r − 3S3 > 0. Let’s suppose, that inequality
isn’t hold (and after that we will get a contradiction). So, now we conclude that
8s
(5r − 3S3 )t > 8s ⇔ t > .
5r − 3S3
Using it, we get:
S3 t + s = 5t4 ⇒
8s
s = t(5t3 − S3 ) > (5t3 − S3 ) ⇒
5r − 3S3
⇒ 5r − 3S3 > 8(5t3 − S3 ) ⇒ 5r + 5S3 > 40t3 ⇔ r + S3 > 8t3
8s 8s
We know, that t > 5r−3S3 , so r + S3 > 8( 5r−3S 3
)3 , and it gives us that
(r + S3 )(5r − 3S3 )3 > 84 s3
(r + S3 )(5r − 3S3 ) = (3r − S3 )2 − (2r − 2S3 )2 ≤ (3r − S3 )2

3
So, we can get that 84 s3 < (3r − S3 )2 (5r − 3S3 )2 ⇒ 82 s 3 < (3r − S3 )(5r − 3S3 ).
(3r − S3 )(5r − 3S3 ) = (4r − 2S3 )2 − (r − S3 )2 ≤ (4r − 2S3 )2 ,

so we get that
3 3 3 3 3 3
8s 4 < 4r − 2S3 ⇔ 4s 4 < 2r − S3 ⇔ 4a 4 b 4 c 4 d 4 ≤ 2(abc + bcd + cda + dab) − (a3 + b3 + c3 + d3 )

This is a contradiction because of the following lemma:
Lemma.
3 3 3 3
a3 + b3 + c3 + d3 + 4a 4 b 4 c 4 d 4 ≥ 2(abc + bcd + cda + dab)
Proof
3 3 3 3
Let f (a, b, c, d) = a3 + b3 + c3 + d3 + 4a 4 b 4 c 4 d 4 − 2(abc + bcd + cda + dab). Our goal is to prove that
f (a, b, c, d) ≥ 0. At first we will prove it in case a = b, c = d:
√ √ √
f (x, x, y, y) ≥ 0 ⇔ 2(x3 − x2 y − xy 2 + y 3 ) − 2xy(x − 2 xy + y) ≥ 0 ⇔ ( x + y)2 (x + y) ≥ xy
which is obvious. √ √
Now we will find, when inequality f (a, b, c, d) ≥ f (a, b, cd, cd) holds:
√ √ √ √ √ √
f (a, b, c, d) ≥ f (a, b, cd, cd) ⇔ (c c − d d)2 ≥ 2ab( c − d)2 ⇐
√ √
⇐ (c + cd + d)2 ≥ 2ab ⇐ (3 cd)2 ≥ 2ab ⇐ 9cd ≥ 2ab ⇐ 4cd ≥ ab
√
So, if 4cd ≥ ab, then we can change c, d to x, x, where x = cd. We will try to do it with our numbers.
Lets rearrange a, b, c, d in such way that a ≥ b ≥ c ≥ d.
Obviously, 4ab ≥ cd, so we can change a, b to x, x. Now we need to prove, that f (x, x, c, d) ≥ 0. If
4cd ≥ x2 ,then we can replace c, d to y, y and we are done.
So, we will consider the case, when x2 ≥ 4cd. Let’s take a look at√pairs x, c and x, d. If 4xc ≥ xd and
√
4xd ≥ xc, then we can replace them to pairs l, l and r, r, l = xc, r = xd.
4xc ≥ xd obviously holds, so we have to check 4xd ≥ xc. If it is true, then we are done. Else, let’s suppose
that 4xd ≤ xc ⇔ 4d < c.
6 3 3
f (x, x, c, d) ≥ 0 ⇔ 2x3 + c3 + 4x 4 c 4 d 4 ≥ 2x2 (c + d) + 4xcd
6 3 3
4x 4 c 4 d 4 ≥ 4xcd - obvious. It remains to prove, that
2x3 + c3 + d3 ≥ 2x2 (c + d)
2x3 + c3 + d3 ≥ 2x3 + c3 = x3 + x3 + c3 ≥ 3x2 c. 3x2 c ≥ 2x2 (c + d) ⇔ 3c ≥ 2c + 2d ⇔ c ≥ 2d, but we
already have c ≥ 4d.
Also solved by Sardor Bozorboyev, Lyceum S.H.Sirojjidinov, Tashkent, Uzbekistan; Nicuşor Zlota, “Traian
Vuia” Technical College, Focşani, Romania; Arber Avdullahu, Mehmet Akif College, Kosovo; Adnan Ali,
Student in A.E.C.S-4, Mumbai, India.

MR Year 2014 Problem Column PDF

Încărcat de

Informații document

Descriere originală:

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

MR Year 2014 Problem Column PDF

Încărcat de

Drepturi de autor:

Formate disponibile

Junior problems

J289. Let a be a real number such that 0 ≤ a < 1. Prove that

Proposed by Arkady Alt, San Jose, California, USA

J290. Let a, b, c be nonnegative real numbers such that a + b + c = 1. Prove that

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

Proposed by Panagiote Ligouras, Noci, Italy

Proposed by Dorin Andrica, Babes-Bolyai University, Romania

J293. Find all positive integers x, y, z such that

Proposed by Aaron Doman, University of California, Berkeley , USA

J294. Let a, b, c be nonnegative real numbers such that a + b + c = 3. Prove that

1 ≤ (a2 − a + 1)(b2 − b + 1)(c2 − c + 1) ≤ 7.

Proposed by An Zhen-ping, Xianyang Normal University, China

Mathematical Reflections 1 (2014) 1

Proposed by Marius Stanean, Zalau, Romania

S290. Prove that there is no integer n for which

Proposed by Ivan Borsenco, Massachusetts Institute of Technology, USA

S291. Let a, b, c be nonnegative real numbers such that ab + bc + ca = 3. Prove that

Proposed by Titu Andreescu, USA and Marius Stanean, Romania

Proposed by Cosmin Pohoata, Princeton University, USA

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

Proposed by Roberto Bosch Cabrera, Havana, Cuba

Mathematical Reflections 1 (2014) 2

Proposed by Mihai Piticari, Campulung Moldovenesc, Romania

Proposed by Ivan Borsenco, Massachusetts Institute of Technology, USA

Proposed by Marius Cavachi, Constanta, Romania

U292. Let r be a positive real number. Evaluate

Proposed by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain.

Proposed by Marcel Chirita, Bucharest, Romania

U294. Let p1 , p2 , . . . , pn be pairwise distinct prime numbers. Prove that

Proposed by Marius Cavachi, Constanta, Romania

Mathematical Reflections 1 (2014) 3

O289. Let a, b, x, y be positive real numbers such that x2 − x + 1 = a2 , y 2 + y + 1 = b2 , and

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

Proposed by Cosmin Pohoata, Princeton University, USA

O291. Let a, b, c be positive real numbers. Prove that

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

O292. For each positive integer n let

Find all prime numbers p for which

Proposed by Gabriel Dospinescu, Ecole Normale Superieure, Lyon

4(x3 + y 3 + z 3 + xyz)2 ≥ (x2 + y 2 + z 2 + t2 )3 .

Proposed by Nairi Sedrakyan, Yerevan, Armenia

Proposed by Cosmin Pohoata, Princeton University, USA

Mathematical Reflections 1 (2014) 4

Proposed by Mihaly Bencze, Brasov, Romania

Proposed by Marius Cavachi, Constanta, Romania

J297. Let a, b, c be digits in base x ≥ 4. Prove that

Proposed by Titu Zvonaru, Comanesti and Neculai Stanciu, Buzau, Romania

Proposed by Nairi Sedrakyan, Yerevan, Armenia

Proposed by Ivan Borsenco, Massachusetts Institute of Technology, USA

J300. Let a, b, c be positive real numbers. Prove that

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

Mathematical Reflections 2 (2014)

S295. Let a, b, c be positive real numbers such that a + b + c = 3. Prove that

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

Proposed by Michal Rolinek, Institute of Science and Technology, Vienna

Proposed by Ivan Borsenco, Massachusetts Institute of Technology, USA

S298. Prove the following identity

Proposed by Ángel Plaza, Universidad de Las Palmas de Gran Canaria, Spain

Proposed by Mihai Miculita and Marius Stanean, Romania

(x + y)3 (y + z)3 ≥ 64abxy 2 z (ax + y + bz)2 .

Proposed by Marius Stanean, Zalau, Romania

Mathematical Reflections 2 (2014)