Carte Olimpiada - Balcanica-2010

c o l e c i a
Aceast lucrare este rezultatul colaborrii dintre

Editura Paralela 45 i Societatea Romn de tiine Matematice.

Editor: Clin Vlasie

Corectur: autorii
Tehnoredactare: Carmen Rdulescu
Coperta coleciei: Andrei Mnescu
Machetare & prepress: ART CREATIV

Descrierea CIP a Bibliotecii Naionale a Romniei
ERBNESCU, DINU
Training Problems for the Junior Balkan Mathematical Olympiads
The Romanian Experience / Dinu erbnescu, Mircea Fianu,
Ioan erdean. - Piteti : Paralela 45, 2010
ISBN 978-973-47-0957-1

I. Fianu, Mircea
II. erdean, Ioan

51(075.33)(076)

Copyright Editura Paralela 45, 2010

D DD DI NU I NU I NU I NU ERBNESCU ERBNESCU ERBNESCU ERBNESCU, ,, , F FF FLORI CA LORI CA LORI CA LORI CA B BB BANU ANU ANU ANU, ,, , M MM MI RCEA I RCEA I RCEA I RCEA F FF FI ANU I ANU I ANU I ANU, ,, ,
TEFAN TEFAN TEFAN TEFAN S SS SM MM M R RR R ND ND ND NDOI U OI U OI U OI U, ,, , M MM MARI US ARI US ARI US ARI US P PP PERI A ERI A ERI A ERI AN NN NU UU U, ,, , I II I OAN OAN OAN OAN ERDEAN ERDEAN ERDEAN ERDEAN, ,, , G GG GABRI EL ABRI EL ABRI EL ABRI EL P PP POPA OPA OPA OPA

Tra Tra Tra Training Problems ining Problems ining Problems ining Problems
for the Junior Balkan for the Junior Balkan for the Junior Balkan for the Junior Balkan
Matehmatical Olympiads Matehmatical Olympiads Matehmatical Olympiads Matehmatical Olympiads

The Romanian Experience The Romanian Experience The Romanian Experience The Romanian Experience

FOREWORD
The Junior Balkan Mathematical Olympiad started fourteen years ago as an
alternative of the Balkan Olympiads for Junior High students.
Since, the competition has become very popular not only in South-Eastern
European countries but all over the world. Some teams from Asia and Europe
participate as invited countries to the competition.
But the most important achievement of this competition is in my opinion the
long list of nice problems produced for these kids; usually extremely elementary
but containing deep mathematical ideas. Maybe this is why, in the Balkan
countries, the junior teams usually graw to become in years the senior teams for
the BMO and IMO.
The booklet is devoted to the Romanian problems used for the qualification
tests of the Romanian team in the last decade. One can remark the great number of
authors, some of them being IMO students at the time the problems were given.
It is a great job the authors have done here, collecting all those problems and
editing nice solutions.

Radu Gologan

PROBL PROBL PROBL PROBLE EE EM MM MS SS S

9 99 9
Chapter I Chapter I Chapter I Chapter I

AL AL AL ALGEBRA GEBRA GEBRA GEBRA
Problem Problem Problem Problem 1 11 1
Prove that for any real numbers a, b, c such that 0 < a, b, c < 1, the following inequality holds
1 ) 1 )( 1 )( 1 ( < + c b a abc .
Dinu erbnescu, 2002
Let a, b, c be positive real numbers with abc = 1. Prove that
1 +
ca bc ab c b a + +
+ +
6 3
.
Mircea Lascu and Vasile Crtoaje, 2003
Find the positive real numbers a, b, c which satisfy the inequalities
4(ab + bc + ca) 1 a
2
+ b
2
+ c
2
3(a
3
+ b
3
+ c
3
).
Laureniu Panaitopol, 2004
Let a, b, c be positive real numbers such that a + b + c
c b a
1 1 1
+ + . Prove that
a + b + c
abc
3
.
Cezar Lupu, 2005
Let a, b, c be positive real numbers such that (a + b)(b + c)(c + a) = 1. Show that
ab + bc + ca
4
3
.
Cezar Lupu, 2005
Let a, b, c be positive real numbers with a + b + c = 3. Prove that
(3 2a)(3 2b)(3 2c) a
2
b
2
c
2
.
Robert Szasz, 2005

10 10 10 10
Pr Pr Pr Problem oblem oblem oblem 7 77 7
Prove that c b a
ba
c
ca
b
bc
a
+ + + +
3 3 3
, for all positive real numbers a, b and c.
* * * , 2005
Prove that
|
\
| +
+
+
+
+
|
\
|
+ +
b
a c
a
c b
c
b a
a
c
c
b
b
a
2
3
2

for all positive real numbers a, b and c.
Cezar Lupu, 2006
Suppose a, b, c are positive real numbers with the sum equal to 1. Prove that:
) ( 3
2 2 2
2 2 2
c b a
a
c
c
b
b
a
+ + + + .
Mircea Lascu, 2006
Problem Problem Problem Problem 10 10 10 10
Let x, y, z be positive real numbers such that
2
1
1
1
1
1
1
=
+
+
+
+
+ z y x
.
Prove that 8xyz 1.
Mircea Lascu, 2006
Let a and b be integer numbers. Show that there exists a unique pair of integers x, y so that
(x + 2y a)
2
+ (2x y b)
2
1.
Adrian Zahariuc, 2007
Suppose a, b, c are positive real numbers satisfying:
1
1
1
1
1
1
1
+ +
+
+ +
+
+ + a c c b b a
.
Show that
a + b + c ab + bc + ca.
Andrei Ciupan, 2007

11 11 11 11
Problem Problem Problem Problem 1 11 13 33 3
Prove that
| ) )( )( ( |
4
3
3
3 3 3
x z z y y x xyz
z y x
+
+ +
,
for any real numbers x, y, z 0.
Viorel Vjitu, 2007
Let a, b, c be positive real numbers with ab + bc + ca = 3. Prove that
abc b a c a c b c b a
1
) ( 1
1
) ( 1
1
) ( 1
1
2 2 2

+ +
+
+ +
+
+ +
.
Vlad Matei, 2008

Determine the maximum value of the real number k such that
(a + b + c) k k
c a c b b a
|
\
|

+
+
+
+
+
1 1 1
,
for all real numbers a, b, c 0 with a + b + c = ab + bc + ca.
Andrei Ciupan, 2008
Let a, b, c > 0 be real numbers with the sum equal to 3. Show that:
ab c
c
ca b
b
bc a
a
+
+
+
+
+
+
+
+
3
3
3
3
3
3
3.
Dinu erbnescu, 2009
Let a, b, c be positive real numbers such that a + b + c
c b a
1 1 1
+ + . Prove

that
ca bc ab a
c
c
b
b
a 1 1 1
+ + + + .
Cezar Lupu, 2009
Let A be a non-empty subset of R with the property that for every real numbers x, y, if x + y A,
then xy A. Prove that A = R.
Eugen Pltnea, 2001

12 12 12 12
For any positive integer n, let
1 2 1 2
1 4 4
) (
2
+ +
+
=
n n
n n
n f .
Compute the sum f (1) + f (2) + + f (40).
Titu Andreescu, 2002
Five real numbers of absolute values not greater than 1 and having the sum equal to 1 are written
on the circumference of a circle.
Prove that one can choose three consecutively disposed numbers a, b, c, such that all the sums
a + b, b + c and a + b + c are nonnegative.

Dinu erbnescu, 2003
A set of 2003 positive integers is given. Show that one can find two elements such that their sum is
not a divisor of the sum of the other elements.
Valentin Vornicu, 2003
Consider the numbers
1 3
2
+ = n n a
n

and
) ( 2
2 2
n n n n b
n
+ + = ,
for all n = 1, 2, , 49.
Prove that there are integers A, B so that
2 ...
49 49 2 2 1 1
B A b a b a b a + = + + + .
Titu Andreescu, 2004
Let a < b c < d be positive integers such that ad = bc and 1 a d . Prove that a is a square.
Dinu erbnescu, 2004
The real numbers a
1
, a
2
, , a
n
satisfy the relation
101 ) ... ( ...
2
100 2 1
2
100
2
2
2
1
= + + + + + + + a a a a a a .
Prove that ,a
k
, 10, for all k = 1, 2, , 100.
Dinu erbnescu, 2004

13 13 13 13
Consider the triangular array

0 1 1 2 3 5
0 1 1 2 3
2 3 5 8
4 7 11
defined by the conditions:
i) on the first two rows, each element starting with the third is the sum of the two preceding
elements;
ii) on the other rows each element is the sum of the two elements placed above on the same
column.

a) Prove that all the rows are defined according to condition i).
b) Consider 4 consecutive rows and let a, b, c, d be the first element in each of these rows,
respectively. Find d in terms of a, b and c.
Dinu erbnescu, 2004
Let A be a set of positive integers with the properties:
i) if a A, then all positive divisors of a are elements of A;
ii) if a, b A and 1 < a < b, then 1 + abc A.
Prove that if the set A has at least 3 elements, then A = N
*
.
Let n > 3 be a positive integer. Consider n sets, each having two elements, such that the intersection
of any two of them is a set with one element. Prove that the intersects of all sets is nonempty.
Sever Moldoveanu, 2005
Consider two distinct positive integers a and b having integer arithmetic, geometric and harmonic
means. Find the minimum value of ,a b,.
Mircea Fianu, 2005
Find all real numbers a and b satisfying
2(a
2
+ 1)(b
2
+ 1) = (a + 1)(b + 1)(ab + 1).
Show that the set of real numbers can be partitioned into subsets having two elements.
* * *, 2006

14 14 14 14
The set of positive integers is partitioned in subsets with infinite elements each. The question (in
each of the following cases) is if there exists a subset in the partition such that any positive integer has
a multiple in this subset.
a) Prove that if the number of subsets in the partition is finite the answer is yes.
b) Prove that if the number of subsets in the partition is infinite, then the answer can be no (for a
certain partition).
* * *, 2006
Find all non-empty subsets A of the set {2, 3, 4, 5, } so that for any n A, both n
2
+ 4 and
] [ n + 1 also belong to A.
Lucian urea, 2007
An irrational number x, 0 < x < 1 is called suitable if its first 4 decimals in the decimal
representation are equal. Find the smallest positive integer n such that any real number t, 0 < t < 1 may
be written as a sum of n distinct suitable numbers.
Lucian urea, 2007
Let n N
*
and let a
1
, a
2
, , a
n
be positive real numbers so that
2 2
2
2
1
2 1
1
...
1 1
...
n
n
a a a
a a a + + + = + + + .
Prove that for any m = 1, 2 , n, there exist m numbers among the given ones with the sum not
less than m.
Andrei Ciupan and Flavian Georgescu, 2008
Let A be a finite set of positive real numbers satisfying the property:

For any real numbers a > 0, the sets
{x A , X > a} and
)
`
<
a
x A x
1
|
have the cardinals of the same parity.

Show that the product of all elements in A is equal to 1.
Dinu erbnescu, 2009

15 15 15 15
Let a and b be positive integers. Consider the set of all non-negative integers n for which the
number
n n
b a |
\
|
+ + |
\
|
+
2
1
2
1

is an integer. Show that the set is finite.

* * *, 2009

16 16 16 16
Chapter II Chapter II Chapter II Chapter II

GEOMETRY GEOMETRY GEOMETRY GEOMETRY
Let ABC be an arbitrary triangle. A circle passes through B and C and cuts again the lines AB and
AC in D and E, respectively. The projections of the points B and E on CD are denoted by B and E,
respectively. The projections of the points D and C on BE are denoted by D and C, respectively.
Prove that the points B, D, E and C lie on the same circle.
Dan Brnzei, 2001
Let ABCD be a rectangle. We consider the points E CA, F AB, G BC such that DE CA,
EF AB and EG BC. Solve in the set of rational numbers the equation AC
x
= EF
x
+ EG
x
.
Dan Brnzei, 2001
Let ABCD be a quadrilateral inscribed in the circle O. For a point E O, its projections K, L, M,
N on the lines DA, AB, BC, CD, respectively, are considered. Prove that if N is the orthocenter of the
triangle KLM for some point E, different from A, B, C, D, then this holds for every point E of the
circle O.
Dan Brnzei, 2001
Let ABCDEF be a hexagon with AB ,, DE, BC ,, EF, CD ,, FA and in which the diagonals AD, BE
and CF are congruent. Prove that the hexagon can be inscribed in a circle.
Dan Brnzei, 2001
Find the minimal area of a rectangular box of volume strictly greater than 1000, given that the side
lengths are integers.
Dinu erbnescu, 2001
Pr Pr Pr Problem oblem oblem oblem 4 44 42 22 2
Let ABCD be a parallelogram of center O. Let M and N be the midpoints of BO and CD,
respectively. Prove that if the triangles ABC and AMN are similar, then ABCD is a square.
Dinu erbnescu, 2002

17 17 17 17
Let ABC be an isosceles triangle such that AB = AC and 'A = 20. Let M be the foot of the altitude
from C and let N be a point on the side AC such that CN = BC
2
1
. Find the measure of the angle 'AMN.
Dinu erbnescu, 2002
Let ABCD be a unit square. For any interior points M and N such that the line MN does not contain
any vertices of the square, denote by s(M, N) the least area of a triangle having vertices in the set {A,
B, C, D, M, N}. Find the least number k such that s(M, N) k, for all such points M, N.
Dinu erbnescu, 2002
The diagonals AC and BD of a convex quadrilateral meet at O. Let m be the measure of the acute
angle formed by these diagonals. For any angle xOy of measure m, the area inside the angle that is in
the interior of the quadrilateral is constant. Prove that ABCD is a square.
Mircea Fianu, 2002
Let C
1
(O
1
) and C
2
(O
2
) be two circles such that C
1
passes through O
2
. Point M lies on C
1
such that
M O
1
O
2
. The tangents from M at C
2
meet again C
1
at A and B. Prove that the tangents from A and B
at C
2
others than MA and MB meet at a point located on C
1
.
Dinu erbnescu, 2002
Five points are given in the plane such that each of 10 triangles defined by them has area greater
than 2. Prove that there exists a triangle of area greater than 3.
Consider n > 2 concentric circles and two lines D
1
, D
2
which meet at P, a point inside all the
circles. The rays determined by P on the line D
1
meet the circles in points A
1
, A
2
, , A
n
and
n
A A A ..., , ,
2 1
respectively and the rays on D
2
meet the circles at points B
1
, B
2
, , B
n
and
n
B B B ..., , ,
2 1

(points with the same indices lie on the same circle). Prove that if the arcs A
1
B
1
and A
2
B
2
are equal,
then the arcs A
i
B
i
and
i i
B A are equal, for all i = 1, 2, , n.
Dinu erbnescu, 2002
Let ABC be a triangle and a = BC, b = CA and c = AB be the lengths of its sides. Points D and E lie
in the same halfplane determined by BC as A. Suppose that DB = c, CE = b and that the area of DECB
is maximal. Let F be the midpoint of DE and let FB = x.
Prove that FC = x and
4x
3
= (a
2
+ b
2
+ c
2
)x + abc.
Dan Brnzei, 2002

18 18 18 18
Consider a rhombus ABCD with center O. A point P is given inside the rhombus, but not situated
on the diagonals. Let M, N, Q, R be the projection of P on the sides (AB), (BC), (CD), (DA),
respectively. The perpendicular bisectors of the segments MN and RQ meet at S and the perpendicular
bisectors of the segments NQ and MR meet at T. Prove that P, S, T and O are the vertices of a
rectangle.
Mircea Fianu, 2003
Two circles C
1
(O
1
) and C
2
(O
2
) with distinct radii meet at points A and B. The tangent from A to C
1

intersects the tangent from B to C
2
at point M. Show that both circles are seen from M under the same
angle.
Dinu erbnescu, 2003
Let E be the midpoint of the side CD of a square ABCD. Consider the point M inside the square
such that 'MAB = 'MBC = 'BME = x. Find the angle x.
Suppose ABCD and AEFG are rectangles such that the points B, E, D, G are collinear (in this
order). Let the lines BC and GF intersect at point T and let the lines DC and EF intersect at point H.
Prove that points A, H and T are collinear.
Mircea Fianu, 2003
Two unit squares with parallel sides overlap by a rectangle of area 1/8. Find the extreme values of
the distance between the centers of the squares.
Radu Gologan, 2003
Consider a circle of center O and V a point outside the circle. The tangents from V touch the circle
at points T
1
, T
2
. Let T be a point on the small arc T
1
T
2
of the circle. The tangent at T intersects the line
VT
1
in the point A and the lines TT
1
and VT
2
intersect in the point B. Let M be the intersection point of
the lines OM and AB. Prove that lines OM and AB are perpendicular.
Mircea Fianu, 2004
Let ABC be an acute triangle and let D be a point on the side BC. Points E and F are the projections
of D on the sides AB and AC, respectively. Lines BF and CE meet at point P. Prove that AD is the
bisector line of the angle BAC if and only if lines AP and BC are perpendicular.

19 19 19 19
Consider a triangle ABC with the side lenghts a, b, c so that a is the largest. Prove that the triangle
is right-angled if and only if
2 ) ( ) )( ( c b a c a c a b a b a + + = + + + + .
Virgil Nicula, 2004
Consider the triangle ABC with AB = AC and a variable point M on the line BC so that B is
between M and C. Prove that the sum between the inradius of AMB and the exradius of AMC
corresponding to the angle M is constant.
Virgil Nicula, 2004
Let ABC be a triangle inscribed in the circle K and consider a point M on the arc BC that does not
contain A. The tangents from M to the incircle of ABC intersect the circle K at the points N and P.
Prove that if 'BAC = 'NMP, then triangles ABC and MNP are congruent.
Let M, N, P be the midpoints of the sides BC, CA, AB of the triangle ABC, respectively, and let G
be the centroid of the triangle. Prove that if the quadrilateral BMGP is cyclic and 2BN = AB 3 , then
the triangle ABC is equilateral.
BMO shortlist 2004
Circles C
1
and C
2
intersect at points A and B. The tangent line from A to C
2
meets C
1
at point C and
the tangent line from A to C
1
meets C
2
at point D. A ray from A, interior to the angle 'CAD, intersects
C
1
at M, C
2
at N and the circumcircle of the triangle ACD at P. Prove that AM = NP.
Mircea Fianu, 2005
Points M and N are given on the sides AD and BC of a rhombus ABCD. Line MC meets the
segment BD at T and line MN meets the segment BD at U. Line CU intersects the side AB at Q and
finally! line QT intersects the side CD at P. Show that the triangles QCP and MCN have the same
area.
Mircea Fianu, 2005
A point M is given inside an equilateral triangle ABC. Denote by A, B, C the projections of the
point M on the sides BC, CA, AB, respectively. Prove that lines AA, BB, CC are concurrent if and
only if point M lies on an altitude of the triangle.
Laurentiu Panaitopol, 2005

20 20 20 20
Let ABC be a triangle with BC > CA > AB and let G be the centroid of the triangle. Prove that
'GCA + 'GBC < 'BAC < 'GAC + 'GBA.
Dinu erbnescu, 2005
Three circles C
1
(O
1
), C
2
(O
2
), C
3
(O
3
) share a common point Q and meet again pairwisely at points
A, B, C. Show that if points A, B, C are collinear then points Q, O
1
, O
2
, O
3
are cocyclic.
Simpson, 2005
Let AB and BC be two consecutive sides of a regular polygon with 9 vertices inscribed in a circle of
center O. Let M be the midpoint of AB and let N be the midpoint of the radius perpendicular to BC.
Find the measure of the angle 'OMN.
* * *, 2005
A piece of cardboard has the shape of a pentagon ABCDE in which BCDE is a square and ABE is
an isosceles triangle right-angled at A. Prove that the pentagon can be divided in 2 different ways in
three parts that can be rearranged in order to recompose a right isosceles triangle.
* * *, 2005
Let ABC be a triangle right at C and consider points D, E on the sides BC, CA, respectively such
that k
CD
AE
AC
BD
= = . Lines BE and

AD meet at point O. Show that 'BOD = 60 if and only if
k = 3 .
Marcel Chiri, 2006
In a plane 5 points are given such that all triangles having vertices at these points are of area not
greater than 1. Show that there exists a trapezoid which contains all points in the interior (or on the
sides) and having the area not exceeding 3.
Marcel Chiri, 2006
Consider a circle C of center O and let A, B be points on the circle with 'AOB = 90. Circles
C
1
(O
1
) and C
2
(O
2
) are internally tangent to C at points A, B, respectively, and moreover are tangent
to themselves. Circle C
3
(O
3
), located inside the angle 'AOB, is externally tangent to C
1
, C
2
and
internally tangent to C. Prove that points O, O
1
, O
2
, O
3
are vertices of a rectangle.
* * *, 2006

21 21 21 21
Suppose ABCD is a cyclic quadrilateral of area 8. Prove that if there exists a point O in the plane of
the triangle such that OA + OB + OC + OD = 8, then ABCD is either an isosceles trapezoid or a square.
Flavian Georgescu, 2006
Let ABC be a triangle and let A
1
, B
1
, C
1
be the midpoints of the sides BC, CA, AB, respectively.
Show that if M is a point in the plane of the triangle such that
2
1 1 1
= = =
MC
MC
MB
MB
MA
MA
,
then M is the centroid of the triangle.
Dinu erbnescu, 2006
Let ABC be a triangle and D a point inside the triangle, located on the median from A. Show that if
'BDC = 180 'BAC, then AB CD = AC BD.
Eduard Bzvan, 2006
Consider a trapezoid ABCD with the bases AB and CD so that with the diameters AD and BC are
secant; denote by M and N their common points. Prove that the midsection point of the diagonals AC
and BD belongs to the line MN.
Let ABCD be a convex quadrilateral. The incircle
1
of triangle ABD touches the sides AB, AD at
points M, N respectively, while the incircle
2
of triangle CBD touches the sides CD, CB at points P,
Q, respectively. Given that
1
and
2
are tangent, show that:
a) the quadrilateral ABCD is circumscriptible;
b) the quadrilateral MNPQ is cyclic;
c) the incircles of triangles ABC and ADC are tangent.
Vasile Pop, 2007
Let ABC be an acute-angled triangle with AB = AC. For any point P inside the triangle ABC
consider the circle centered at A with radius AP and let M and N be the intersection points of the sides
AB and AC with the circle. Determine the position of the point P such that MN + BP + CP is
minimum.
Francisc Bozgan, 2007
Let ABC be a triangle. Points M, N, P are given on the sides AB, BC, CA respectively, so that
CPMN is a parallelogram. Lines AN and MP meet at point R, lines BP and MN meet at point S, while
Q is the intersection point of the lines AN and BP. Show that S[MRQS] = S[NQP].
Mircea Lascu, 2007

22 22 22 22
Circles
1
and
2
meet at points A and B. A third circle
3
, which intersects
1
at points D and B,
is internally tangent to
2
at point C and tangent to the line AB at point F, and lines DE and AB meet
at point G. Let H be the mirror image of F across G. Find the measure of the angle 'HCF.
Lucian urea, 2007
Let be a semicircle of diameter AB. A parallel line to AB intersects semicircle in C and D so that
points B and C lie on opposite sides of the line AD. The parallel line from C to AD meets again at
point E. Lines BE and CD meet at point F and the parallel line from F to AD intersects AB at point P.
Prove that the line PC is tangent to the semicircle .
Cosmin Pohoa, 2007
Let ABC be a triangle right-angled at A and let D be a point on the side AC. Point E is the mirror
image of A across BD and point F is the intersection of the line CE with the perpendicular line from D
to CB. Show that the lines AF, DE and CB are concurrent.
Dinu erbnescu, 2007
Let ABC be an acute-angled triangle. Consider the equilateral triangle AUV, with A (BC), U
(AC), V (AB) such that UV ,, BC. Similarly, define points B (AC) and C (AB). Show that the
lines AA, BB and CC are concurrent.
Vasile Pop, 2008
Let ABC be a triangle and D the midpoint of BC. On the sides AB and AC there are points M, N,
respectively, other than the midpoints of these segments, so that AM
2
+ AN
2
= BM
2
+ CN
2
and
'MDN = 'BAC. Prove that A = 90.
Consider an acute-angled triangle ABC, the height AD and the point E where the diameter from A
of the circumcircle intersects the line BC. Let M, N be the mirror images of D across the lines AC and
AB. Show that 'EMC = 'BNE.
Dinu erbnescu, 2008
Let d be a line and let M, N be two points on d. Circles , , , centered at A, B, C, D are tangent
to d in such a manner that circles , are externally tangent at M, while circles , are externally
tangent at N. Moreover, points A and C lie on the same side of line d. Prove that if there exists a circle
tangent to all circles , , , , containing all of them in the interior, then lines AC, BD and d are
concurrent or parallel.

23 23 23 23
Let ABCD be a quadrilateral with no two opposite sides parallel. The parallel from A to BD meets
the line CD at point F and the parallel from D at AC meets the line AB at point E. Consider the
midpoints M, N, P, Q of the segments AC, BD, AF, DE respectively. Show that lines MN, PQ and AD
are concurrent.
Dinu erbnescu, 2008
Consider a rhombus ABCD. Point M and N are given on the line segments AC and BC respectively,
such that DM = MN. Lines AC and DN meet at point P and lines AB and DM meet at point R. Prove
that RP = PD.
Cristinel Mortici, 2009
Let ABCD be a quadrilateral. The diagonals AC and BD are perpendicular at point O. The
perpendiculars from O on the sides of the quadrilateral meet AB, BC, CD, DA at M, N, P, Q,
respectively, and meet again CD, DA, AB, BC at M, N, P, Q, respectively. Prove that points M, N,
P, Q, M, N, P, Q are concyclic.
Cosmin Pohoa, 2O09
Consider a regular polygon A
0
A
1
A
n1
, n 3, and m {1, 2, , n 1}, m n/2. For any number
i {1, 2, , n 1}, r(i) be the remainder of i + m at the division by n. Prove that no three segments
A
i
A
r(i)
are concurrent.
* * *, 2009
Consider K a polygon in plane, such that the distance between any two vertices is not greater than 1.
Let X and Y be two points inside K. Show that there exist a point Z, lying on the border of K, such that
XZ + YZ 1.
* * *, 2009
Show that in any triangle ABC with A = 90 the following inequality holds:
(AB AC)
2
(BC
2
+ 4AB AC)
2
2BC
6
.
Lucian Petrescu, 2009
Let ABC be a triangle and A the foot of the internal bisector of angle BAC. Consider the
perpendicular line from A on BC.
Define analogously the lines d
B
and d
C
. Prove that lines d
A
, d
B
and d
C
are concurrent if and only if
triangle ABC is isosceles.
* * *, 2009

24 24 24 24

Chapter III Chapter III Chapter III Chapter III

NUMBER THEORY NUMBER THEORY NUMBER THEORY NUMBER THEORY
Find n Z such that the number
5
2 4
+
n
n
is rational.
Dan Popescu, 2001
Three students write on the blackboard next to each other three two-digit squares. In the end, they
observe that the 6-digit number thus obtained is also a square. Find this number!
Mircea Becheanu, 2001
Determine all positive integers a < b < c < d with the property that each of them divides the sum of
the other three.
* * *, 2001
Let n be a non-negative integer. Find the non-negative integers a, b, c, d, such that
a
2
+ b
2
+ c
2
+ d
2
= 7 4
n
.
Let n 2 be a positive integer. Find the positive integers x such that
n x x x < + + + ... ,
for any number of radicals.
Ion Dobrot, 2001
Let k, n, p be positive integers such that p is a prime number, k < 1000 and p n k = .
a) Prove that if the equation p x n x k ) ( 100 + = + has a non-zero integer solution, then p is a
divisor of 10.
b) Find the number of all non-negative solutions of the above equation.
Mircea Fianu, 2002

25 25 25 25
Find all positive integers a, b, c, d such that
a + b + c + d 3 = ab + cd.
Dinu erbnescu, 2002
Let n be an even positive integer and let a, b be two coprime positive integers.
Find a and b such that a + b is a divisor of a
n
+ b
n
.
Dinu erbnescu, 2002
Problem Problem Problem Problem 10 10 10 100 00 0
Let a be an integer. Prove that for any real number x, x
3
< 3, both numbers
2
3 x and
3 3
x a
cannot be rational.
The last four digits of a perfect square are equal. Prove that all of them are zeros.
Let m, n > 1 be integer numbers. Solve in positive integers
x
n
+ y
n
=2
m
.
Let p, q be two distinct primes. Prove that there are positive integers a, b such that the arithmetic
mean of all positive divisors of the number n = p
a
q
b
is an integer.
Consider the prime numbers n
1
< n
2
< < n
31
. Prove that if 30 divides
4
31
4
2
4
1
... n n n + + + , then
among these numbers one can find three consecutive primes.
Vasile Berghea, 2003
Probl Probl Probl Problem em em em 1 11 105 05 05 05
Let n be a positive integer. Prove that there are no positive integers x and y such as
2 4 1 + < + < + + n y x n n .
Dinu erbnescu, 2003
Let a be a positive integer such that the number a
n
has an odd number of digits in the decimal
representation, for all n > 0. Prove that the number a is an even power of 10.
Vasile Zidaru, 2003

26 26 26 26
Find all positive integers n for which there exist distinct integers a
1
, a
2
, , a
n
such that
2
...
...
2 1
2 1
2 1
n
n
a a a
a
n
a a
+ + +
= + + + .
Dinu erbnescu, 2004
Let p, q, r be primes and let n be a positive integer such that
p
n
+ q
n
= r
2
.
Prove that n = 1.
A finite set of positive integers is called isolated if the sum of the elements in any proper subset is a
number relatively prime with the sum of the elements of the isolated set. Find all nonprime integers n
for which there exist positive integers a, b such that the set A = {(a + b)
2
, (a + 2b)
2
, , (a + nb)
2
} is
isolated.
Gabriel Dospinescu, 2004
Find the greatest integer n, n > 10 such that the remainder of n when divided by each square
between 2 and
2
n
is an odd integer.
Adrian Stoica, 2005
Let k, r N
*
and let x (0, 1) be a rational number given in decimal representation:
x = 0,a
1
a
2
a
3
a
4

Show that if the decimals a
k
, a
k+r
, a
k+2r
, a
k+3r
, are canceled, the new number thus obtained is still
rational.
Dan Schwarz, 2005
Find all positive integers n and p if p is prime and
n
8
p
5
= n
2
+ p
2
.
Adrian Stoica, 2005
Prob Prob Prob Problem lem lem lem 11 11 11 113 33 3
For any positive integer n let s(n) be the sum of its digits in decimal representation. Find all
numbers n for which s(n) is the largest proper divisor of n.

27 27 27 27
Consider the integers a
1
, a
2
, a
3
, a
4
, b
1
, b
2
, b
3
, b
4
with a
k
b
k
for all k =1, 2, 3, 4. If
{a
1
, b
1
} + {a
2
, b
2
} = {a
3
, b
3
} + {a
4
, b
4
},
show that the number |(a
1
b
1
)(a
2
b
2
)(a
3
b
3
)(a
4
b
4
), is a square.
Note. For any sets A and B, we denote A + B = {x + y , x A, y B}.
For a positive integer n denote r(n) the number having the digits of n in reverse order for
example, r(2006) = 6002. Prove that for any positive integers a and b the numbers 4a
2
+ r(b) and
4b
2
+ r(a) can not be simultaneously squares.
Marius Ghergu, 2006
Find all integers n, n 4 such that ] [ n + 1 divides n 1 and ] [ n 1 divides n + 1.
Marian Andronache, 2007
P PP Problem roblem roblem roblem 1 11 117 17 17 17
Solve in positive integers the equation
(x
2
+ 2)(y
2
+ 3)(z
2
+ 4) = 60xyz.
Determine all positive integers n which can be represented in the form
n = [a, b] + [b, c] + [c, a],
where a, b, c are positive integers.
Note: [p, q] is the lowest common multiple of the integers p and q.
Let p be a prime number, p 3, and let a, b be integer numbers such that p | a + b and p
2
| a
3
+ b
3
.
Show that p
2
| a + b or p
3
| a
3
+ b
3
.
* * *, 2008
Prove that for any positive integer n there exists a multiple of n with the sum of its digits equal to n.
Mihai Blun, 2008
Let n N, n 2. Consider the integers a
1
, a
2
, , a
n
with 0 < a
k
k, for all k = 1, 2, , n. Given
that a
1
+ a
2
+ + a
n
is an even number, prove that one can choose the signs + and such that
a
1
a
2
a
n
= 0.
* * *, 2008

28 28 28 28
A sequence of integers a
1
, a
2
, , a
n
is given so that a
k
is the number of all multiples of k in this
sequence, for any k = 1, 2, , n. Find all possible values for n.
Cristian Mangra, 2008
Let a, b be real numbers with the property that the integer part of an + b is an even number, for all
n N. Show that a is an even integer.
Dinu erbnescu, 2002
Find all primes p, q satisfying the equation 2p
q
q
p
= 7.
Find all pairs of integers (m, n), n, m > 1 so that mn 1 divides n
3
1.

For all positive integers n define a
n
=

times
3 ... 33 2
n
, where digit 3 occurs n times. Show that the number
a
2009
has infinitely many multiples in the set {a
n
, n N
*
}.
Cristian Lazr, 2009
Find all non-negative integers a, b, c, d such that:
7
a
= 4
b
+ 5
c
+ 6
d
.
Petre Stngescu, 2009
A positive integer is called saturated if any prime factor occurs at a power greater than or equal to
2 in its factorisation. For example, numbers 8 = 2
3
and 9 = 3
2
are saturated; moreover, they are
consecutive. Prove that there exist infinitely many saturated consecutive numbers.
* * *, 2009

29 29 29 29

Chapter IV Chapter IV Chapter IV Chapter IV

COMBINATORICS COMBINATORICS COMBINATORICS COMBINATORICS
Consider a 1 n rectangle and some tiles of size 1 1 of four different colours. The rectangle is
tiled in such a way that no two neighboring square tiles have the same colour.
a) Find the number of distinct symmetrical tilings.
b) Find the number of tilings such that any consecutive square tiles have distinct colours.
Dan Brnzei, 2002
A square of side 1 is decomposed into 9 equal squares of sides
3
1
and the central square is colored
black. The remaining eight squares are analogously divided into nine squares each, and central squares
are colored in black.
Prove that after 1000 steps the total area of the black region exceeds 0.999.
Cristinel Mortici and Costel Chite, 2002
An equilateral triangle of side 10 is divided into 100 unit equilateral triangles by lines parallel to
the sides of the triangle. Find the number of (not necessarily unit) equilateral triangles in the
configuration described above such that the sides of the triangles are parallel to the sides of the initial
one.
Dinu erbnescu, 2002
Show that one can color all the points of a plane using only two colors such that no line segment
has all points of the same color.
Consider a cube and let M, N be two of its vertices. Assign the number 1 to these vertices and 0 to
the other six vertices. We are allowed to select a vertex and to increase with a unit the numbers
assigned to the 3 adjacent vertices call this a movement.
Prove that there is a sequence of movements after which all the numbers assigned to the vertices of
the cube became equal if and only if MN is not a diagonal of a face of the cube.
Marius Ghergu and Dinu erbnescu, 2004

30 30 30 30
Problem Problem Problem Problem 1 11 13 33 34 44 4
An array 8 8 consists of 64 unit squares. Inside each square are written the numbers 1 or 1 so
that in any 2 2 subarray the sum of the four numbers equals 2 or 2. Prove that there exist two rows
of the array which are equal.
Marius Ghergu, 2004
In a chess tournament each of the players have played with all the others two games, one time with
the white pieces and then with the black pieces. In each game the winners sets one point and both
players receive 0.5 points if the game ends with draw. At the end of the tournament, all the players end
with the same number of points.
a) Prove that there are two players with the same number of draws.
b) Prove that there are two players with the same number of losses when playing the white.
Marius Ghergu, 2004
A regular polygon with 1000 sides has the vertices colored in red, yellow or blue. A move consists
in choosing to adjacent vertices colored differently and coloring them in the third color. Prove that
there is a sequence of moves after which all the vertices of the polygon will have the same color.
Marius Ghergu, 2004
A country has six cities with airports and two rival flight companies. Any two cities are connected by
flights so that on each route between two cities one may travel with exactly one of the two flight com-
panies. Prove that you can visit 4 cities in a cycle flying with the same air company (that is, there exist
four cities A, B, C, D and a company which operates on the routes A B, B C, C D and D A).
Dan Schwarz, 2005
A phone company starts a new type of customer service. A new client can choose k phone numbers
in this network which are call-free regardless if is called or if calling. A group of n students decide to
take advantage of this promotion.
Show that if n 2k + 2 then there will exist 2 students which will be charged when speaking.
Show that if n = 2k + 1 then there exists a way of arranging the free calls so that in this group
everybody speaks free to anyone else.
The positive integers from 1 to n
2
are placed arbitrarily on squares of an n n chessboard. Two
squares are called adjacent if they have a common side. Show that two opposite corner squares can be
joined by a path of 2n 1 adjacent squares so that the sum of the numbers places on them is at least
1
2
2
3
+ +
(
n n
n
.
Radu Gologan, 2005

31 31 31 31
An 7 7 array is divided in 49 unit squares. Find all integers n N
*
for which n checkers can be
placed on the unit squares so that each row and each line have an even number of checkers.
(0 is an even number, so there may exist empty rows or columns. A square may be occupied by at
most 1 checker.)
Dinu erbnescu, 2006
A rectangular cardboard is divided successively into smaller pieces by a straight cut; at each step,
only one single piece is divided in two. Find the smallest number of cuts required in order to obtain
among others 251 polygons with 11 sides.
Marian Andronache, 2007
Consider a n n array divided into unit squares which are randomly colored in black or white.
Three of the four comer squares are colored in white and the fourth is colored in black. Prove that
there exists a 2 2 square which contains an odd number of white squares.
Lidia Ilie, 2007
Consider the numbers from 1 to 16. A solitaire game is played in the following manner: the
numbers are paired and each pair is replaced by the greatest prime divisor of the sum of the numbers
in that pair for example, (1, 2); (3, 4); (5, 6); ; (15, 16) produces the sequence 3, 7, 11, 5, 19, 23, 3,
31. The game continues similarly until one single number is left. Find the greatest possible value of
the number which ends the game.
Adrian Stoica, 2007
Eight persons attend a party, and each participant has at most three others to whom he/she cannot
speak. Show that the persons can be grouped in 4 pairs so that each pair can converse.
Mihai Blun, 2007
Problem Problem Problem Problem 1 11 14 44 45 55 5
A set of points is called free if there is no equilateral triangle whose vertices are among the points
in the set. Show that any set of n points in the plane contains a free subset of at least n points.
Clin Popescu, 2007
A 8 8 square board is divided into 64 unit squares. A skew-diagonal of the board is a set of 8
unit squares with the property that each row or column of the board contains only one unit square of
the set. Checkers are placed in some of the unit squares so that each skew-diagonal has exactly 2
squares occupied by checkers. Prove that there exist two rows or two columns which contain all the
checkers.
Dinu erbnescu, 2007

32 32 32 32
To obtain a square P of side length 2 cm divided into 4 unit squares it is sufficient to draw 3
squares: P and another 2 unit squares with a common vertex, as shown below:

Find the minimum number of squares sufficient to obtain a square of side length n cm divided into
n
2
unit squares (n 3 is an integer).
* * *, 2009
The plane is divided into a net of equilateral triangles of side length 1, with disjoint interiors. A
checker is placed initially inside a triangle. The checker can be moved into another triangle sharing a
common vertex (with the triangle hosting the checker) and having the opposite sides (with respect to
this vertex) parallel. A path consists in a finite sequence of moves. Prove that there is no path between
two triangles sharing a common side.
Vasile Pop, 2009
Show that there exist (at least) a rearrangement a
0
, a
1
, a
2
, , a
63
of the numbers 0, 1, 2, , 63,
such that a
i
a
j
a
j
a
k
, for any i < j < k {0, 1, 2, , 63}.
* * *, 2009
Let A = {1, 2, , 2006}. Find the maximal number of subsets of A that can be selected such that
the intersection of any 2 distinct subsets has 2004 elements.
* * *, 2006
Let 1 m < n be positive integers, and consider the set
M = {(x, y) , x, y N
*
, 1 x, y n}.
Determine the least value v(m, n) with the property that for any subset P M with ,P, = v(m, n)
there exist m + 1 elements A
i
= (x
i
, y
i
) P, i = 1, 2, , m + 1, for which the values x
i
are all distinct,
and y
i
are also all distinct.
Vasile Pop, 2007
Ten numbers are chosen at random from the set 1, 2, 3, , 37. Show that one can select four
distinct numbers from the chosen ones so that the sum of two of them is equal to the sum of the other
two.
Vasile Pop, 2008

33 33 33 33
Let m, n N
*
and A = {1, 2, , n}, B = {1, 2, , m}. A subset S of the set product A B has the
property that for any pairs (a, b)(x, y) S, then (a x)(b y) 0. Show that S has at most m + n 1
elements.
Dinu erbnescu, 2007
In the interior of a circle centered in O a number of 1200 points A
1
, A
2
, , A
1200
are considered
such that for every i, j with 1 i j 1200, the points O, A
i
and A
j
are not collinear. Prove that there
exist the points M and N on the circle, with m('MON) = 30, such that in the interior of the angle
'MON lie exactly 100 points.
* * *, 2001
Consider a convex polygon with n 5 sides. Prove that there are at most
3
) 5 2 ( n n
triangles of
area 1 with the vertices among the vertices of the polygon.
Andrei Negu, 2004

34 34 34 34

35 35 35 35

FORMAL FORMAL FORMAL FORMAL
SOLUTIONS SOLUTIONS SOLUTIONS SOLUTIONS

36 36 36 36

37 37 37 37

Chapter I

ALGEBRA
Solution to Solution to Solution to Solution to Problem Problem Problem Problem 1 1 1 1
Observe that
3
x x < for x [0, 1]. Thus, we have
3
abc abc < and
3
) 1 )( 1 )( 1 ( ) 1 )( 1 )( 1 ( c b a c b a < .
By the AM-GM inequality, we get
3
3
c b a
abc abc
+ +
<
and
3
) 1 ( ) 1 ( ) 1 (
) 1 )( 1 )( 1 ( ) 1 )( 1 )( 1 (
3
c b a
c b a c b a
+ +
< .
Summing up, we obtain
1
3
1 1 1
) 1 )( 1 )( 1 ( =
+ + + + +
< +
c c b b a a
c b a abc ,
as desired.

Solution to Solution to Solution to Solution to Problem Problem Problem Problem 2 22 2
Setting x =
a
1
, y =
b
1
, z =
c
1
, we have xyz = 1. The inequality rewrites as
1 +
z y x zx yz xy + +
+ +
6 3
.
Since (x + y + z)
2
3(xy + yz + zx), it follows that
1 +
2
) (
9
1
3
z y x zx yz xy + +
+
+ +
.
It suffices to observe that
1 +
z y x z y x + +
+ +
6
) (
9
2
,
which reduces to
0
3
1
2
|
|
\
|
+ +
z y x
.

38 38 38 38
Using the Chebyshev inequality we derive
(a + b + c)(a
2
+ b
2
+ c
2
) 3(a
3
+ b
3
+ c
3
),
hence a + b + c 1. On the other hand,
4(ab + bc + ca) 1 a
2
+ b
2
+ c
2
ab + bc + ca,
therefore ab + bc + ca 1. As
3(ab + bc + ca) (a + b + c)
2
1,
we obtain a + b + c 1, thus a + b + c = 1. Consequently, a + b + c = 1 and
3(ab + bc + ca) = (a + b + c)
2
,
which imply
a = b = c =
3
1
.

The given condition rewrites
abc(a + b + c) ab + bc + ca.
From the Cauchy-Schwarz inequality we have
ab + bc + ca ) ( 3 c b a abc + + ,
so the conclusion follows.

Let s = a + b + c. The given condition rewrites as 1 = (s a)(s b)(s c) = s(ab + bc + ca) abc,
so ab + bc + ca =
c b a
abc
+ +
+ 1
. From the AM-GM inequality we obtain
2
3
) )( )( (
2
3
)) ( ) ( ) ((
2
1
= + + + + + + + + = + + a c c b b a a c c b b a c b a .
On the other side, 1 = (a + b)(b + c)(c + a) 8 ca bc ab , hence abc
8
1
. Consequently
ab + bc + ca
4
3
3
2
8
1
1 = |
\
|
+ ,
as needed.

Wlog, assume that a b c.
If a + b c, then c >
2
3
, while a <
2
3
and b <
2
3
. In this case (3 2a)(3 2b)(3 2c) < 0 a
2
b
2
c
2
and we are done.
If a + b > c, then a, b, c are the side lengths of a triangle with semiperimeter p =
2
3
. The inequality
rewrites
8(p a)(p b)(p c) a
2
b
2
c
2
.

39 39 39 39
Using the formulas S
2
= p(p a)(p b)(p c) and abc = 4RS we obtain 8S
2
16pR
2
S
2
or 1 3R
2
.
This inequality can be proved in many ways, for example from 2p 3 3 R, since p =
2
3
.

The inequality rewrites as a
4
+ b
4
+ c
4
abc(a + b + c). Using successively the inequality
x
2
+ y
2
+ z
2
xy + yz + zx
we get
a
4
+ b
4
+ c
4
a
2
b
2
+ b
2
c
2
+ c
2
a
2
= (ab)
2
+ (bc)
2
+ (ca)
2
abc(a + b + c)
as desired.

Let y
c
b
x
b
a
= = , and
a
c
= z. The inequality rewrites successively:
x
2
+ y
2
+ z
2
+ 2xy + 2yz + 2zx
|
|
\
|
+ + + + +
z y x
z y x
1 1 1
2
3

x
2
+ y
2
+ z
2
+ 2
|
|
\
|
+ +
z y x
1 1 1

|
|
\
|
+ + + + +
z y x
z y x
1 1 1
2
3

2(x
2
+ y
2
+ z
2
) +
z y x
1 1 1
+ + 3(x + y + z).
From AM-GM inequality we get
2x
2
+ x
x
x x
x
x x
x
3
1
3
1 1
3
2 2 2 2
= + + = .
Summing with the analogously relations we get the conclusion.

First solution. Using Cauchy-Schwarz inequality we get:
a c c b b a
c b a
ac
c
cb
b
ba
a
a
c
c
b
b
a
2 2 2
2 2 2 2
2
4
2
4
2
4 2 2 2
) (
+ +
+ +
+ + = + + .
The inequality is reduced to a
2
+ b
2
+ c
2
3(a
2
b + b
2
c + c
2
a) or
(a + b + c)(a
2
+ b
2
+ c
2
) 3(a
2
b + b
2
c + c
2
a).
The latter rewrites as 0 ) (
2

b a a , which is obvious.
Second solution. Since a + b + c = (a + b + c)
2
= 1, the inequality gives successively:
2 2 2 2
2 2 2
) ( ) ( 3 ) ( c b a c b a c b a
a
c
c
b
b
a
+ + + + + + + + ,
or

|
|
\
|
+
2
2
) ( 2 b a b a
b
a
,

40 40 40 40
hence

2
2
) (
) (
b a
b
b a
.

Since a, b, c 1, we are done.

Canceling the denominators we get 1 = xy + yz + zx + 2xyz. From AM-GM inequality we get
1 4
4 3 3 3
2 z y x ,
so 1 (8xyz)
3
. The conclusion follows.

Solving for a and b the system of equations
=
= +
, 2
2
t b y x
s a y x

one has
+
=
+ + +
=
.
5
) 2 ( ) 2 (
5
) 2 ( ) 2 (
t s b a
y
t s b a
x

Restating the claim, one has to prove that there exists a unique pair of integers s, t Z with
s
2
+ t
2
1 so that both numbers (a + 2b) + (s + 2t), (2a b) + (2s t) are divisible by 5.
Notice that (a + 2b) + (s + 2t) + 2[(2a b) + (2s t)] = 5(a + s), so
x Z y Z.
Since s
2
+ t
2
1 (s, t) {(0, 0), (1, 0), (0, 1), (1, 0), (0, 1)}, it follows that
(a + 2b) + (s + 2t) {a + 2b 2, a + 2b 1, a + 2b, a + 2b + 1, a + 2b + 2},
so there is exactly one pair (s, t) with 5 | (a + 2b) + (s + 2t),, as needed.

The Cauchy-Schwarz inequality gives
(a + b + 1)(a + b + c
2
) (a + b + c)
2
,
so
1
1
1
) (
2
2
+ +
+ +
+ +

cyc cyc
b a c b a
c b a
.
Then

+ = + + +
cyc cyc cyc cyc
ab a c b a a a 2 ) ( 2
2 2 2
,
and the claim follows immediately.

41 41 41 41
Let p = |(x y)(y z)(z x)|. Recall the identities:
x
3
+ y
3
+ z
3
3xyz = (x + y + z)(x
2
+ y
2
+ z
2
xy yz zx)
and
x
2
+ y
2
+ z
2
xy yz zx =
2
1
[(a y)
2
+ (y z)
2
+ (z x)
2
].
Using AM-GM inequality, we have
(1) x
2
+ y
2
+ z
2
xy yz zx
3 2
2
3
p .
On the other hand, since ,x y, x + y, ,y z, y + z and ,z x, z + x, it follows that
2(x + y + z) ,x y, + ,y z, + ,z x,.
Applying again the AM-GM inequality gives
(2) 2(x + y + z)
3
2
3
p ,
and the claim follows from the inequalities (1) and (2).

Using the AM-GM inequality we derive
3 2
) (
3
abc
ca bc ab
+ +
. As ab + bc + ca = 3, then
abc 1. Now
abc abc
ca bc ab
a abc a bc a ac ab a c b a
1
3 3
1
) 1 ( 3
1
) 3 ( 1
1
) ( 1
1
) ( 1
1
2
=
+ +
=
+
=
+
=
+ +
=
+ +

,
as claimed.

Observe that the numbers a = b = 2, c = 0 fulfill the condition ab + bc + ca = a + b + c. Plugging
into the given inequality, we derive that k k |
\
|
+ +
2
1
2
1
4
1
4 , hence k 1.
We claim that the inequality holds for k = 1, proving that the maximum value of k is 1. For this,
rewrite the inequality as
1 1
1 1 1
) ( |
\
|

+
+
+
+
+
+ +
c a c b b a
ca bc ab 1 + + +
+
+ +
ca bc ab
b a
ca bc ab

1 + + + |
\
|
+
+
ca bc ab c
b a
ab
1
+
b a
ab
.
Notice that
c b a
ab
b a
ab
+ +
+
, since a, b, c 0. Summing over a cyclic permutation of a, b, c we
get
b a
ab
1 =
+ +
+ +
=
+ +
c b a
ca bc ab
c b a
ab
,
as needed.

42 42 42 42
Alternative solution. The inequality is equivalent to the following:
k
c a c b b a c b a
c b a
S |
\
|
+
+
+
+
+ + + +
+ +
=
1 1 1
1
.

Using the given condition, we get
=
+ + +
+ + + + +
=
+
+
+
+
+ ) )( )( (
) ( 3 1 1 1
2 2 2
a c c b b a
ca bc ab c b a
c a c b b a

abc ca bc ab c b a
c b a ca bc ab c b a
+ + + +
+ + + + + + + +
=
) )( (
) ( 2
2 2 2
=
abc c b a
c b a c b a
+ +
+ + + + +
2
) (
) 1 )( (
,
hence
S =
abc c b a
c b a
+ +
+ +
2
2
) (
) (
.
Now it is clear that S 1, and the equality occurs for abc = 0. Therefore k = 1 is the maximum
value.

Observe that 3a + bc = a
2
+ ab + bc + ca = (b + a)(c + a), hence

= + +
+ + +
=
+
+
cyc cyc
) )( 3 (
) )( )( (
1
3
3
c b a
a c c b b a bc a
a

3
) )( )( (
) 27 (
) 9 (
) )( )( (
1
2 2 2
cyc
2
+ + +

=
+ + +
=

a c c b b a
c b a
a
a c c b b a

+ + + + + + = + ) ) ( 3 ) ( 3 27 ( 3 ) 3 ( 3 27
2 2
abc ca bc ab c b a a a a

+ abc ab a 3 9 27
2
18 + 3abc 7(ab + bc + ca).
Multiplying by 3 in the last inequality gives:
2(a + b + c)
3
+ 9abc 7(ab + bc + ca)(a + b + c)
2(a
3
+ b
3
+ c
3
+ 6abc + ) 3 ( 7 9 )
2 2
b a abc abc b a
sym sym

+ +
0 ) )( ( 2
2 2 3
+

b a b a b a a
sym
.

The inequality rewrites as a
2
c + b
2
a + c
2
b a + b + c.
Using Cauchy-Schwarz inequality, we have
2
2 2 2
) (
1 1 1
1 1 1
c b a
b
c
a
b
c
a
b a c
+ +
|
|
|
|
\
|
+ + |
\
|
+ + .

43 43 43 43
This implies a
2
c + b
2
a + c
2
b ) ( 1 ) (
1 1 1
) (
1 1 1
) (
2
c b a c b a
c b a
c b a
c b a
c b a
+ + + +
+ +
+ +
+ +
+ +
, as claimed.

Let a A. Then a + 0 A, hence 0 = 0 a A. For any real number b, b + (b) = 0 A, hence
b
2
A. Thus, A contains all negative numbers. Let c > 0; we have 0 < c c , so c c A.
It follows that c = ) )( ( c c A, hence A = R.

Let a = 1 2 + n and b = 1 2 n . Then a
2
+ b
2
= 4n, ab = 4n
2
1 and a
2
b
2
= 2. Hence
|
\
|
+ =
=
+
+ +
=
3 3
2 2
3 3 2 2
) 1 2 ( ) 1 2 (
2
1
) ( n n
b a
b a
b a
ab b a
n f
and thus
= + + + = + + + ) 79 81 ... 3 5 1 3 (
2
1
) 40 ( ... ) 2 ( ) 1 (
3 3 3 3 3 3
f f f
364 ) 1 729 (
2
1
) 1 9 (
2
1
) 1 81 (
2
1
3 3 3
= = = = .

First, we prove that at most two of the sums a + b, b + c, c + d, d + e and e + a can be negative.

Indeed, assume that two non-consecutive sums (say a + b and c + d) are less than 0. Then 1 e =
= (a + b) + (c + d) < 0 and so 1 < e, a contradiction. Thus, if three sums are negative, then two of them
are not consecutive, which is false. Moreover, if two sums are negative, then these must be
consecutive; in other words, at least three consecutive sums are nonnegative.
Let a + b, b + c, c + d be greater than or equal to zero. If one of the sums d + e or e + a is negative,
then a + b + c = 1 (d + e) or b + c + d = 1 (e + a) are at least 1, hence is positive and we are done.
Finally, consider the case when all sums a + b, b + c, c + d, d + e, e + a are positive. Suppose that
a + b + c < 0; then b > (a + b) + (b + c) > 0. Thus, if a + b + c, b + c + d, c + d + e are negative,
then b, c, d are positive and we are done.

a
b
c
d
e

44 44 44 44
Let a
1
< a
2
< < a
2003
be the elements of the set. We prove the claim by contradiction.
Numbers a
1
+ a
2003
, a
2
+ a
2003
, , a
2002
+ a
2003
divide the sum S = a
1
+ a
2
+ + a
2003
, since
a + b | S a b if and only if a + b , S. Hence S = k
i
(a
i
+ a
2003
) for all i = 2002 , 1 , where k
i
are
integers.
Since a
i
+ a
2003
< S < 2003a
2003
< 2003(a
i
+ a
2003
), it follows that k
i
{2, 3, , 2002} for all i =
= 2002 , 1 . By Pigeonhole principle, there is a pair of indices i j such that k
i
= k
j
, a contradiction.

The key idea is to observe that
a
n
b
n
=
2
) 1 2 1 (
2
1
+ + n n n .
As 1 1 2 + n n n , it follows that the sum is 5 2 4 .

Solution t Solution t Solution t Solution to o o o Problem Problem Problem Problem 23 23 23 23
Consider the integers 0 < m n < p so that b = a + m, c = a + n and d = a + p.
Then a(a + p) = (a + m)(a + n) and a + p a + 1 + a 2 . As p = m + n +
a
mn
is an integer, then
a mn and p m + n + 1. On the other hand, 1 + a 2 n + m + 1 mn
n m
a
+
2
, hence
a mn.
Consequently, a = mn and m = n, so a is a square.

Solution to Solution to Solution to Solution to Problem Problem Problem Problem 2 22 24 44 4
Assume by contradiction that |a
k
| > 10, for some k. Wlog, let k = 1. Then
2
1
a > 100 and
1 ...
2 2
100
2
3
2
2
< + + + + s a a a ,
where s = a
1
+ a
2
+ + a
100
.
On the other hand, the Cauchy-Schwarz inequality yields
2
1
a = (s a
2
a
3
a
100
)
2
100( 100 ) ...
2 2
100
2
3
2
2
< + + + + s a a a ,
a contradiction.

a) In the array below

a b b
d e f
g h i

where i is an element of the third row, observe that i = c + f = (a + b) + (d + e) = (a +d) + (b + c) =
= g + h. The same argument holds for all the other rows, by induction.
b) We prove that d = 2b + 2c a. Indeed, from the array

45 45 45 45
a x y z
b t u
c v
d
we derive d = u + v = b + t + (u + z) = 2(b + t) +(x + y)= 2b + 2(t + y) + x y =
= 2b + 2c + x (x + a) = 2b + 2c a.

It is obvious that 1 A, since 1 is a divisor of any integer. Consider a, b two elements of A with
1 < a < b. Since at least one of a, b or 1 + ab is even, then 2 is an element of A.
We induct on n 6 to prove that n A. Assume that k A for all k = 1, 2, , n 1. If n is odd,
then n = 2p + 1 with 1 < 2 < p A, hence n A. If n is even, then n = 2p. As above, 2p 1 and 2p + 1
are elements of A and consequently 1 + (2p 1)(2p + 1) = 4p
2
A. The first property implies n =
= 2p A, as needed.
To complete the proof, we need to show that 3, 4, 5 A. For this, consider a > 2 an element of A.
Then 1 + 2 a A, 1 + 2(1 + 2a) = 3 + 4a A and 1 + (1 + 2a)(3 + 4a) = 4 + 10a + 8a
2
A. If a is
even, then 4 | 4 + 10a + 8a
2
and so 4 A. If a is odd, then choose a to be 4 + 10a + 8a
2
and again
4 A. Next, as 1 < 2 < 4 A we have 1 + 2 4 = 9 A and so 3 A. Finally, 7 = 1 + 2 3 A, 15 =
= 1 + 2 7 A, hence 5 A and we are done.

Let A = {a, b} and B = {a, c} be two of the given sets. A third set C intersect both of them, so C is
{b, c} or {a, d}. In the first case, a fourth set cannot have in common with each of the first three sets
exactly one element, so C = {a, d}.
Any set from the other n 3 > 1 sets that do not contain a should contain simultaneously the
elements b, c and d, a contradiction. Hence the claim is proved.

Let
2
b a
x
+
= and
b a
ab
y
+
=
2
. Then ab = xy and a
2
a(a + b) + ab = a
2
2ax + xy = 0, so x(x y) =
= (a x)
2
is a square.
Let d be the greatest common divisor of x and y and write x = dx, y = dy with x, y Z and (x, y) =
= 1. Then d
2
x(x y) is a square and (x, x y) = 1, so both x and x y are squares; write x = m
2

and x y = n
2
.
Recall that the geometric mean xy ab = is an integer, so xy = d
2
xy = d
2
m
2
(m
2
n
2
) is a
square. Consequently m
2
n
2
= p
2
, where p is an integer.
To conclude, to this point we have x = dm
2
, y = dp
2
and since a = x ) ( y x x , b =
= x ) ( ( y x x , it follows that a = dm(m n) and b = dm(m n). Therefore |a b| = 2dmn with the
minimum of 30 obtained for d = 1, m = 5, n = 3 dont forget the Pythagorean triples!

46 46 46 46
First solution. Consider the given equation as quadratic in a:
a
2
(b
2
b + 2) a(b + l)
2
+ 2b
2
b + 1 = 0.
The discriminant is = (b 1)
2
(7b
2
2b + 7), hence we have solutions only for b = 1. It follows
that a = 1.
Second solution. Using Cauchy-Schwarz inequality, we get 2(a
2
+ 1) (a + 1)
2
, 2(b
2
+ 1) (b + 1)
2

and (a
2
+ 1)(b
2
+ 1) (ab + 1)
2
. Multiplying, we obtain 2(a
2
+ 1)(b
2
+ 1) (a + 1)(b + 1)(ab + 1),
hence the equality case occur in all the inequalities, so a = b = 1.

For example, consider R \ Z partitioned in {x, x} and Z in {2n, 2n + 1}.
Another example: split R into disjoint intervals [2n, 2n + 1), with n Z. Then take pairs (x, x + 1)
from each interval [2n, 2n + 1), with x [2n, 2n + 1).

a) Let A
k
be the partition classes, with k = 1, 2, , r. Assuming that the answer is no, there exist n
k
,
k = 1, 2, , r, such that no multiples of n
k
is in A
k
. But n
1
n
2
n
r
lies in one of the sets A
k
and is
multiple of any n
k
, false.
b) We exhibit a partition for which the answer is no.
Let A
k
be the set of all numbers written only with the first k primes at any positive power;
moreover, put 1 A
1
. Fixing k, the number p
1
p
2
p
k+1
have no multiples in A
k
.

We claim that A = {2, 3, 4, 5, }.
Let m be the smallest element of the set A. Since ] [ m + 1 A, we have m ] [ m + 1 m + 1,
which gives m = 2.
Notice that ] 4 [
2
+ n = n for all n 2. Indeed, n
2
n
2
+ 4 < (n + 1)
2
= n
2
+ 2n + 1, for all n 2.
Using both hypothesis, we have
n A n
2
+ 4 A ] 4 [
2
+ n + 1 A n + 1 A.
The conclusion follows by induction.

At first, we look for a lower margin of n. The number 0.1111 can be written as a sum of n distinct
suitable numbers, all starting with 0.0000, therefore lower than 0.0001. Hence if 0.1111 = a
1
+ a
2
+
+ + a
n
, with a
k
< 0.0001, then 0.1111 < n 0.0001, or n > 1111. Thus n is at least 1112.
We claim that 1112 is the requested number. Let t (0, 1) be a real number. If t > 0.1111, choose a
suitable number of the form y = 0.xxxx so that y < t y + 0.1111. The other suitable numbers will
have the form 0.0000, so they are different from y. We have 0 < t y 0.1111. Because
0 <
1112
1111 . 0
1112

y t
< 0.0001, the first four decimals of the number u =
1112
y t
are all equal to 0. Choose
an irrational number e, small enough not to change the first four decimals of the numbers u + e,

47 47 47 47
u + 2e, , u + 1111e, and such that all plus y + u + e are left irrationals. Then
t y = (u + e) + (u + 2e) + (u + 3e) + + (u + 1111e) + |
\
|
2
1112 1111 e
u .
As previously stated, number e was selected so that all summands are irrational, suitable numbers
with the first four decimals 0 and paiwisely distinct in other words, different from the last one.
Consequently,
t = (y + u + e) + (u + 2e) + (u + 3e) + + (u + 1111e) + |
\
|
2
1112 1111 e
u .

The number y + u + e starts with 0.xxxx, while the others with 0.0000, so they are suitable. Since t
is now represented as required, the proof is concluded.

It is clear that we need to prove that a
1
+ a
2
+ + a
n
n. Let us notice that this is enough: let m =
= 1, 2, , n and assume that any selection of m numbers from the given ones has the sum less than 1.
Then add the n inequalities
, ...
, ...
, ...
1 1
1 3 2
2 1
m a a a
m a a a
m a a a
m n
m
m
< + + +
< + + +
< + + +

to get n(a
1
+ a
2
+ + a
n
) < nm, which is a contradiction.
Back to the top, let g be the geometrical mean of the numbers a
1
, a
2
, , a
n
and suppose that
a
1
+ a
2
+ + a
n
< n.
By AM-GM inequality, we have g 1
...
2 1
<
+ + +
n
a a a
n
, while
1 >
2
2 2
2
2
1 2 1
1
1
...
1 1
...
g n
a a a
n
a a a
n n
+ + +
=
+ + +
,
which gives g > 1, a contradiction.

The main idea is to show that x A if and only if
x
1
A. Since A is finite, one may pair the
elements of A in |
\
|
x
x
1
, , with number 1 left alone if belongs to A. Then it is obvious that the product
of all elements is equal to 1.
Let b > 1 be an arbitrary real number. Applying the hypotesis for a = b and a =
b
1
we find that the
sets {x A , x > b} and
)
`
<
b
x A x
1
| have the cardinals of the same parity, and also the sets

48 48 48 48
)
`
>
b
x A x
1
| and
= < b
b
x A x
1
1
| have the cardinals of the same parity.
This implies that the sets L =
)
`
>
b
x A x
1
| \ {x A , x > b} =
)
`
\
|
b
b
x A x ,
1
| and R =
= {x A | x < b} \
)
`
<
b
x A x
1
| =
)
`
b
b
x A x ,
1
| have the cardinals of the same parity. But the
set L may have at most one element which is not in R and vice versa, which shows the sets L and R
have the same number of elements. Thus both numbers b and
b
1
are or arent elements of A, as
claimed.

The number
n n
b a |
\
|
+ + |
\
|
+
2
1
2
1
is an integer if and only if 2
n
divide (2a + 1)
n
+ (2b + 1)
n
.
For n = 2k, 2
n
= 4
k
divides (2a + 1)
n
+ (2b + 1)
n
= (4a
2
+ 4a + 1)
k
+ (4b
2
+ 4b + 1)
k
2 (mod 4),
implying k 1, hence the set of even integers satisfying the claim is finite.
For n = 2k + 1, one has
(2a + 1)
n
+ (2b + 1)
n
= (2a + 1 + 2b + 1)[(2a + 1)
2k
(2a + 1)
2k1
(2b + 1) +
+ (2a + 1)
2k2
(2b + 1)
2
+ (2b + 1)
2k
].
The second factor is a sum of odd summands, each summand being odd, thus being also an odd
integer. The number 2a + 2b + 2 has only a finite number of divisors of the form 2
k
, therefore the
claim is proved.

49 49 49 49
Chapter II

GEOMETRY
First solution. Let I be the intersection point of the lines BE and CD. The quadrilaterals BDBD
and CECE are cyclic, hence 'BDB = 'BDI and 'CEC= 'IEC. Since BDEC is also cyclic,
'BDB= 'CEC. It follows that 'BDI = 'IEC, so BDEC is a cyclic quadrilateral.

Second solution. Using the power of a point theorem, one has:
IB ID = ID IB
IC IE = IE IC
IE IB = ID IC
From these one easily obtains IB IE= ID IC, which proves that the quadrilateral BDEC is
cyclic, using the reciprocal of the power of a point theorem.

Denote AD = a, AB = b. We have AC
2
= a
2
+ b
2
, CE = b
2
/AC, AE = a
2
/AC and EF/a = AE/AC =
= a
2
/AC. It follows that EF =
2
3
AC
a
and, analogously, EG =
2
3
AC
b
. Thus, the equation is equivalent to
(a
2
+ b
2
) = (a
3x
+ b
3x
)
2
.
Notice that x =
3
2
is a solution. Moreover, for a = b the solution is unique.
Suppose a b; wlog a > b. Set k =
a
b
(0, 1) to obtain (1 + k
2
)
3x
= (1 + k
3
)
x
.
A
D
E
E
D
C
B
B
C

50 50 50 50
If x <
3
2
, then k
3x
> k
2
1 + k
3x
> 1 + k
2
> 1 (1 + k
3x
)
2
> (1 + k
2
)
2
> (1 + k
2
)
3x
.
If x <
3
2
, we use a similar argument.
Thus, the only solution is x =
3
2
.

Let F and G be the projections of E on the diagonals BD and AC. From Simsons theorem, it
follows that the triplets of points (K, L, F), (M, N, F), (K, G, N) and (M, L, G) are collinear. The point
N is the orthocenter of the triangle KLM if and only if KL MN and ML KN. Let F and G be the
points in which EF and EG intersect the second time the circle. We have KF ,, AF and MG ,, CF.
Thus KL MN is equivalent to AF CF and then to O AC. Similarly, ML KN is equivalent to
O BD. Thus, ABCD is a rectangle. It is easy to see that in this case, N is the orthocenter of the
triangle KLM for any position of the point E.

We first notice that ABDE is an isosceles trapezoid. The segments AB and DE have the same
perpendicular bisector. Let O and R the center and radius of the circumcircle of the triangle ABC. One
can see that the perpendicular bisectors of DE and CF also pass through O, hence O is the center of the
circle circumscribed around DCF, with radius R. Finally, since ACDF is an isosceles trapezoid, it
follows that R = R.

Let a b c the lengths of the parallelepipeds. We have abc 1001 and c 11. By analysing the
cases c {11, , 21} one finds that a = 8, b = 9 and c = 14 is the solution.

From the similarity of the triangles AMN and ABC, we obtain
(1)
AC
AN
AB
AM
=
and
(2) 'MAN = 'BAC or 'BAM = 'CAN.
The relations (1) and (2) imply the similarity of the triangles BAM and CAN. Hence, we obtain the
proportions:
(3)
CN
BM
AC
AB
AN
AN
= =
and 'ABM 'ACN. The last equality implyies that ABCD is a rectangle.
To conclude the proof, notice that BM =
4
1
BD =
4
1
AC and CN =
2
1
AB. Hence the last equality in

51 51 51 51
(3) becomes
AB
AC
AC
AB
2
= , that is 2AB
2
= AC
2
= AB
2
+ BC
2
, which proves that ABCD is a square.

Let P the midpoint of BC. Since MP is a median in the right-angled triangle MBC, it follows that
PB = MP = PC = CN.

The point R is considered such that PCNR is a parallelogram (in fact a rhombus). Notice that
'RPM = 'RPB 'MPB = 'ACB (180 2'MBC) = 60,
and BP = MP, therefore MPR is an equilateral triangle. Hence MR = RP = RN and 'MRN = 'MRP +
+ 'PRN = 60 + 80 = 140. Then 'RMN = 'RNM = 20, 'ANM = 20 + 80 = 100 and the required
angle 'AMN is equal to 60.

Let E, F be the midpoints of AD and BC repectively and let M, N be the midpoints of OE and OF.
It is easy to check that
8
1
) , ( = N M S , therefore
8
1
k . We claim that
8
1
= k .

Notice that area[BMC] + area[AMD] =
2
1
, for any interior point M. We may assume that N is an
interior point of the triangle AMD. Therefore area[AND] + area[ANM] + area[DNM] + area[BMC] =
=
2
1
. Consequently, one of the summands exceeds
8
1
. The conclusion follows.

A
M
B P
C
N
R
20
20
A
B
C D
E F
M
N
A
B
C D
M
N

52 52 52 52
Consider 'AOD = m 90. As the angles 'AOD and 'BOC are equal to m, we find area[AOD] =
= area[BOC]. It follows
m CO BO m DO AO sin
2
1
sin
2
1
= ,
that is
DO
BO
CO
AO
= . Since 'AOB = 'DOC, the triangles AOB and DOC are similar and consequently
AB is parallel to DC.

Draw line EF that contains O, such that 'AOE = 'COF = m and E (AB), F (DC). The triangles
AOE and COF are similar and have the same area, that is they are congruent. It follows that AO = OC
and in the same way BO = OD, implying AD || BC. Moreover area(COF) = area(BOC), and since
ABCD is a parallelogram, we find area[BOC] = area[DOC]. Hence D = F and m = 'COF = 'DOF =
= 'BOC = 90. We thus proved that ABCD is a rhombus.
To conclude, consider the bisector lines OM and ON of angles 'AOD and 'DOC respectively,
where M (AD), N (DC). It is easy to check that 'MON = m = 90, whence area[MON] =
= area[AOD]. Thus area[DON] = area[ACM], that is area[AOM] = area[DOM] =
2
1
area[AOD]. It
follows that OM is a median in the triangle AOD, that is AO = OD, which proves that the rhombus
ABCD is a square.

Since O
2
is at equal distances from the tangents MA and MB, it follows that MO
2
is a bisector line
of the angle 'AMB or of the exterior angle defined by MA and MB.

In each case one obtains O
2
A = O
2
B.
Reflecting the figure with respect to the line O
1
O
2
, the circles C
1
and C
2
remain fixed, A reflects in
B and M reflects in N. It is obvious that NA, the reflected of MB is tangent to C
2
and the same is valid
for NB. Observe that N

is on C
1
, proving thus the claim.
A
N
B
M
O
1 O
2
A E
B
C F
D
O
m
m
m

53 53 53 53
Solution to Solution to Solution to Solution to P PP Problem roblem roblem roblem 47 47 47 47
Denote by A, B, C, D, E the five given points. If the pentagon ABCDE is concave, we can suppose
that D is located inside the triangle ABC or inside the quadrilateral ABCD.
In first case area[ABC] = area[ABD] + area[DBC] + area[DAC] > 6 > 3.
In the second case, D is inside the one of triangles BCE, ACE, ABC or ABD. Suppose, without loss
of generality, that D is inside to the triangle BCE. Then
area[BCE] area[BDC] + area[CDE] > 4 > 3.
Consider now the case when ABCD is a convex pentagon. Let M and N be the intersection points of
BE with AC and AD respectively.

The following result will be useful.
Lemma. Let PQRS be a quadrilateral and T a point on the side PQ. Then
area[TRSR] min(area[PRS], area[QSR]).
The proof consists of simply observing that the distance from T to SR is bounded up and below by
the distances from P and Q to SR.
In our case, suppose that BM
3
1
BE, which yields BM
3
1
ME. Then
area[BDE] = area[BDM] + area[MDE]
2
1
area[MDE] + area[MDE] =
=
2
3
area[MDE]
2
3
min(area[CDE], area[ADE]) >
2
3
2 = 3.
The case when NE
3
1
BE is similar. It is left to consider the case when MN
3
1
BE. We then
have:
area[AMN]
3
1
area[ABE] >
3
2
,
area[MND]
3
1
area[BED] >
3
2
,
and
area[MCD] min[area[BCD], area[ECD] > 2.
Summing up, we conclude
area[ACD] > 2 +
3
2
3
2
+ > 3,
and the proof is complete.

A
B
C
D
E
N
M

54 54 54 54
Let O be the common center of the n circles and = A
1
B
1
= A
2
B
2
(the arcs are directly oriented).
Rotate the figure around the center O by angle such that A
1
, A
2
become B
1
, B
2
respectively. The
above rotation R maps lines into lines, that is R(D
1
) = D
2
, since D
1
= A
1
A
2
and D
2
= B
1
B
2
. Moreover,
any circle C
i
is invariant under the rotation. As R(A
i
) lies on D
2
and on C
i
, we get that R(A
i
) = B
i
, that is
A
i
B
i
= .
In the same way we get
i i
B A R = ) ( and
i i
B A = . This concludes the proof.

Let DECB be the quadrilateral of maximal area. It is easy to prove that 'DBE = 'DCE = 90.

It follows FB = FC =
2
DE
= x and that the quadrilateral DBCE is cyclic. By Ptolemeys theorem
we have DC BE = BC DE + DB CE.
Squaring, we get
(4x
2
b
2
)(4x
2
c
2
) = (2ax + bc)
2
,
that is
4x
2
(b
2
+ c
2
)x
2
+ b
2
c
2
= 4a
2
x
2
+ 4abcx + b
2
c
2
.
From this we obtain 4x
3
= (a
2
+ b
2
+ c
2
)x + abc, as desired.

Firstly, observe that triangles RSN and QSM are congruent (S.S.S.), hence 'PMS = 'PNS and
'PQS = 'PQS. It follows that P, S, M, N are concyclic and P, S, Q, R are concyclic.

On the other hand, as 'BNP + 'BMP = 180, points B, N, S, M are concyclic, thus P, S, N, B, M are
points on the circle C
1
(O
1
) of diameter BP. Likewise, points P, S, Q, D, R he on the circle C
2
(O
2
) of
diameter DP.
Since PS is the common chord of the circles C
1
and C
2
, lines PS and O
1
O
2
are perpendicular. As O
1

and O
2
are the midpoints of the segments BP and DP, lines O
1
O
2
and BD are parallel, so PS BD and
B
A
R
M
N
C
Q
D
S O
T
P
A
E
B
C
F
D

55 55 55 55
then PS || AC. Likewise, PT || BD and consequently PS PT.
Furthermore, because O
1
O
2
is middle line in the triangle PBD one find that S lies on the segment
BD. Analogously, T (AC). Thus, PSOT is a rectangle.

We have to prove that 2'O
1
MA = 2'O
2
BM, which is equivalent to
(1)
BM
B O
AM
A O
2 1
= .

The lenght of the common chord AB is equal to 2O
1
A sin
2
1
AB = 2O
1
A sin'BAM, regardless if
AB is the small arc or the great arc AB. Similarly, AB = 2O
2
B sin'ABM, hence
(2)
BAM
B O
ABM
A O
' ' sin sin
2 1
= .
By the Law of Sines in the triangle ABM we derive that
(3)
BAM
MB
ABM
MA
' ' sin sin
= .
From the relations (2) and (3) we obtain
BM
B O
AM
A O
2 1
= , as desired.

Observe that 'MAB + 'MBA = 'MBC + 'MBA = 90, hence
'AMB = 90.
Let F be the midpoint of the side AB. Then MF = FA = FB =
2
1
AB, so 'MBF = 'MBF. It follows
that
'EMF = 'EMB + 'BMF = 'MAB + 'MBA = 90.
A
B
M
O
1 O
2

56 56 56 56

In the right triangle MEF, the leg MF is equal to
2
1
EF. hence 'MEF = 30. We obtain 'MBF =
=
2
1
'MFA =
2
1
MEF = 15 and x = 75.

Lines CD and FG meet at M and lines BC and EF meet at N. As DHEA and FHCT are cyclic
quadrilaterals, it follows that
'FTC = 180 'FHC = 'DAE and 'DAH = 'DEH.

Since 'DMG = 90 'FTC = 90 'DAE = 'DAG, it follows that the quadrilateral ADGM is
cyclic. Hence 'DAM = 'FGE and consequently 'MAH = 'DAM + 'DAH = 'FGE + 'DEH = 90.
Likewise, 'NAH = 90 and therefore points M, A, N are collinear.
In the triangle TMN, point H is the orthocenter. Thus A, H, T lie on the altitude of the triangle, as
desired.

Let MNPQ be the rectangle at the intersection of the unit squares with
centers A and B. Set MN = x and PQ = y, hence
8
1
= xy , x, y [0, 1].
The parallel from A to MN intersects the parallel from B to NP at point C.
It is easy to observe that AC = 1 x and BC = 1 y, so

T
H
F
G
M
A
N
B
C
E
D
A
B
C
D
M
A
B
C
Q
P
M
N

57 57 57 57
AB
2
= (1 x)
2
+ (1 y)
2
= x
2
+ y
2
2(x + y) + 2 =
= x
2
+ 2xy + y
2
2(x + y)
4
1
+ 2 + (x + y)
2
2(x + y) +
4
7
=
=
4
3
) 1 (
2
+ + y x .
It follows that the minimal value of the distance between the centers A and B is equal to
2
3
and it
is obtained for x + y = 1, xy =
8
1
; i.e. x =
4
2 2+
, y =
4
2 2
or x =
4
2 2
, y =
4
2 2+
.
To find the maximal value of AB observe that 0 (1 x)(1 y) = 1 x y + xy =
8
9
(x + y), so
x + y
8
9
. On the other hand, we have x + y
2
1
2 = xy , therefore
2
1
1 x + y 1
8
1
. As
2 2
1
2
1
8
1
|
\
|
|
\
|
, we find that AB
2
2
2
1
2 2
4
9
4
3
1 2
2
1
|
\
|
= = + . Thus AB 2
2
1
, with
equality when x = y =
2 2
1
.
Consequently,
2
1
2
2
3
AB .

The approach of the problem is to see no circles in the figure. Instead, recall that a quadrilateral
ABCD is orthogonal if and only if
(1) AB
2
CD
2
= AD
2
+ BC
2
.
Using successively the Pytagoras theorem we have
BA
2

2
2
BT = BA
2
(BO
2

2
2
OT ) = BA
2
(BO
2

2
1
OT ) =
= BA
2
(BA
2

2
1
AT ) =
2
1
AT = AO
2

2
1
OT = OA
2

2
2
OT .
so the conclusion follows from the relation (1).

Let a, b, c, x, y be the lengths of the sides BC, CA, AB, BD, DC, respectively and let A be the foot
of the altitude from A in the triangle ABC. Notice that x + y = a. Due to Ceva theorem the claim is
equivalent to
DC
BD
AC
AB
EB
AE
FA
CF
DC
BD
= = 1 .
As CF = y cos C, FA = b y cos C, BE = x cos B, AE = c x cos B, BA = c cos B and AC = b cos C,
the equivalence rewrites
cy(c x cos B) = bx(b y cos C) xb = cy.

58 58 58 58
Indeed, we have
ab
c b a
bxy x b
ac
b c a
cxy y c
2 2
2 2 2
2
2 2 2
2
+
=
+

a(c
2
y b
2
x) = xy(c
2
b
2
) c
2
y(a x) = b
2
x(a y)
c
2
y
2
= b
2
x
2
cy = bx,
as claimed.

Squaring both sides of the equality yields
2 2 2 2 2
) ( ) )( ( 2 c b a c a a b a a + + = + + .
It is easy to observe that the equality holds if a
2
= b
2
+ c
2
. To prove the converse statement, assume
that a
2
> b
2
+ c
2
. Then
2 2
b a > c and
2 2
c a > b, hence
) )( ( 2
2 2 2 2
c a a b a a + + > 2(a + b)(a + c) =
= 2a
2
+ 2(ab + bc + ca) > a
2
+ b
2
+ c
2
+ 2(ab + bc + ca) = (a + b + c)
2
,
false.
The case a
2
b
2
+ c
2
leads similarly to a contradiction and we are done.

The idea is to prove that the sum of the radii is equal to the altitude h from A of the triangle ABC; a
hint is to assume that line MA is parallel to BC.
The Stewart relation gives AM
2
BC + AC
2
MB = AB
2
MC + MB MC BC, so AM
2
BC =
= AB
2
(MC MB) + MB MC BC, hence AM
2
= AB
2
+ MB MC. Let r be the inradius of triangle
AMB and R the exradius of triangle AMC corresponding to the angle 'M.
Since
AB MB AM
AMB
r
+ +
=
area 2
,
and
AB MB MA
AMC
R
+
=
area 2
,
then
AB MB AM
MB
h
r
+ +
= ,
and
AB MB AM
MC
h
R
+
= .
Thus
r + R = h MB(MA + MB AB) + MC(MA + MB + AB) = (MA + MB + AB)(MA + MB AB)
MB(MA + MB AB) = (MA + MB + AB)(MA AB)
MB MC + MB(MA AB) = MA
2
AB
2
+ MB(MA AB) MA
2
= AB
2
+ MB MC,
so the claim holds.

59 59 59 59
Let Q be the intersection point of the line segments AB and MP. The tangents from A and M to the
incircle are equal (to r cot
2
A
). Moreover, the tangents from Q to the incircle are equal, so AQ = MQ.
This implies 'QMA = 'QAM, so the arcs AP and BM are equal. In the trapezoid APBM, the diagonals
AB and MP are equal, and likewise AC = MN. This concludes the proof.

By the Power of a point theorem we have AG AM = AP AB, so 4MA
2
= 3AB
2
and thus AM =
=
2
3
AB = BN. Then AG = GB, so the median GP is also an altitude in the triangle AGB. This implies
'BPG = 90, and since BMGP is cyclic, 'GMA = 90. It follows that BC = CA and AB = AC, so the
triangle is equilateral.

Consider the case when the ray is interior to the angle 'BAD. Then
'CMP = 'MCA + 'CAM = 'MAD + 'CAM = 'CAB.
On the other side,
'CPM = 'CDA,
since both subtend the chord AC in the circumcircle ACD.
From the similarity of the triangles ACD and MCP we derive that
AD
MP
AC
MC
= .
Furthermore, 'ACM = 'NAD and 'CAM = 'ADN, so 'ACM ~ 'DAN.
Hence
AC
MC
AC
AN
= ,
thus MP = AN and AM = NP, as claimed.

The distance from M to the line BC is equal to the distance from Q to the line DC, since ABCD is a
rhombus. Hence it suffices to prove that CP = CN or DP = BN.
Using Thales theorem we obtain
UM
NU
UC
QU
UD
BU
= = , hence NQ || MC and likewise MP |, CQ.
Triangles MDP and BQC are similar, so
BT
DT
BC
MD
BQ
DP
= = . Analogously,
DC
BQ
MD
BN
= . It follows that
MD
BN
MD
DP
= , therefore DP = BN.

60 60 60 60
Denote AB = x
1
, BC = y
1
, CA = z
1
, AB = x
2
, BC = y
2
, CA = z
2
. From the Ceva theorem we have
x
1
y
1
z
1
= x
2
y
2
z
2
.
As AB
2
= AP
2
PB
2
(and the analogous relations) we obtain by summation
2
2
2
2
2
2
2
1
2
1
2
1
z y x z y x + + = + + .
Since x
1
+ y
2
= x
2
+ z
1
= y
1
+ z
2
= AB and x
1
+ x
2
+ y
1
+ y
2
+ z
1
+ z
2
= 3 AB, it follows that
x
1
+ y
1
+ z
1
= x
2
+ y
2
+ z
2
.
Thus
x
1
+ y
1
+ z
1
= = x
2
+ y
2
+ z
2
;
2 2 2 2 2 2 1 1 1 1 1 1
z y z x y x z y z x y x + + = + + ; x
1
y
1
z
1
= x
2
y
2
z
2
,
so {x
1
, y
1
, z
1
} = {x
2
, y
2
, z
2
}.
An easy check shows that the claim holds.

It suffices to prove that 'GAC < 'GBA and 'GBC < 'GAC, given that CA > AB and BC > CA.
Since the two claims are equivalent, we will prove the first inequality.
First approach
The reflection B of the point B across AG lies on the parallel line through C at AG. Observe that
point C is further than B from the perpendicular bisector d of the line segment AG, since d intersects
line BC at a point located at the left of point B (with respect to point C) and we are done.
Second approach
As area[GBA] = area[GAC], the inequality 'GAC < 'GBA (< 90!) is equivalent to c m
b
> b m
c
.
Squaring both sides yields b
2
(2a
2
+ 2c
2
b
2
) > c
2
(2a
2
+ 2b
2
c
2
), hence (b
2
c
2
)(b
2
+ c
2
2a
2
) > 0,
which is implied by a > b > c.

Set Q
1
,

Q
2
, Q
3
the mirror images of Q across O
1,
O
2
, O
3
, in other words consider the diametrically
opposite points of Q in the circles C
1
(O
1
), C
2
(O
2
), C
3
(O
3
).
It is easy to check that the claim is equivalent to the fact that points Q, Q
1
, Q
2
, Q
3
lie on the same
circle.
An inversion of pole Q maps the line A B C to a circle A B C passing through Q, while points
Q
1
,

Q
2
, Q
3
map to points
3 2 1
, , Q Q Q , which are the projection of Q to lines AB, BC, CA. By Simpson
theorem, points
3 2 1
, , Q Q Q are collinear, thus points Q, Q
1
, Q
2
, Q
3
lie on the same circle, as needed.

Let P be the intersection point of the circle with the radius ON. Triangle AOP is isosceles at A and
'AOP = 60, so 'ANO = 90. On the other side 'AMO = 90, so AMNO is cyclic. Hence
'OMN = 'OAN = 30.

The first decomposition is obtain as follows:
Consider the midpoint M of the side DE and C the intersection point of the lines BE and CM. The
right isosceles triangle obtain by reassembling is CAC, since ABC = AEC and CMD = MEC.

61 61 61 61
For the second decomposition consider N the midpoint of DE and let A be the intersection point of
the lines BD and AM. The right isosceles triangle obtain by reassembling is CAA, as ANE = NDA
and ABC = CDA.

Consider the rectangle ACDP. The hypothesis rewrites as k
AP
AE
DP
BD
= = , so 'APE = 'BPD and
'APD = 'EPB. Moreover,
PB
PD
PE
AP
= , hence PAD ~ PEB.
It follows that 'DAP = 'PEB, so APOE is cyclic and hence 'BOD = 'AOE = 'APE.
The claim is proved by the following chain of equivalences:
'BOD = 60 tan'BOD = 3 tan'APE = 3 =
AP
AE
k = 3 .

Denote A, B, C, D, E the given points and suppose ABC is the triangle having the maximal area.
The distance from D to BC is not greater than the distance from A to BC, hence D and similarly E
are located between the parallel line from A to BC and its mirror image across BC. Applying the same
reasoning for AB and AC, one obtains a triangular (bounded) region A
1
B
1
C
1
with ABC as median
triangle in which all points must lie.
Points D and E are located in at most 2 of the triangles A
1
BC, AB
1
C, ABC
1
, hence one of the
trapezoids APA
1
B
1
, BCB
1
C
1
or ABA
1
B
1
contains all points. Since the area of the trapezoids is 3 times
the area of ABC, hence not greater than 3, the conclusion is proved.

Let R, r
1
, r
2
be the radii of the circles C, C
1
, C
2
and let r = R r
1
r
2
. Consider the point P so that
OO
1
PO
2
is a rectangle. From the tangency conditions we get OO
1
= R r
1
, OO
2
= R r
2
and O
1
O
2
=
= r
1
+ r
2
= R r. It is sufficient to prove that the circle centered at P with radius r is C
3
.
To prove this, notice that O
1
P = OO
2
= R r
2
= r + r
1
, O
2
P = OO
1
= R r
1
= r + r
2
, and OP =
= O
1
O
2
= R r, so the 3 tangency conditions are fulfilled.

Let be the measure of the angle determined by the diagonals.
Since 8 = OA + OB + OC + OD AC + BD 2 = = S BD AC BD AC 2 2 sin 2 8, we
get AC = BD = 1 and = 90. The claim follows from a simple arch subtraction.

Solution t Solution t Solution t Solution to o o o Problem Problem Problem Problem 7 77 72 22 2
Let A
2
, B
2
, C
2
be the mirror images of the point M with respect to points A
1
, B
1
, C
1
. The given
condition shows that MA = MA
2
, MB = MB
2
, MC = MC
2
. From the parallelograms AMBC
2
, BMCA
2
,
AMCB
2
we derive that MA = MA
2
= BC
2
= B
2
C, MB = MB
2
= AC
2
= CA, and MC = MC
2
= PA
2
= AP
2
.
It follows that MA
2
BC
2
, MA
2
CB
2
and MB
2
AC
2
are also parallelograms, therefore A, M and A
2
are
collinear. The conclusion is now obvious.

62 62 62 62
Let E be the mirror image of D across the midpoint of the side BC. We notice that DBEC is a
parallelogram and ABEC is cyclic. The equality of the areas of triangles ABE and ACE implies
AB BE = AC CE. We are left only to notice that CE = BD and BE = CD.

Let Q be the intersection point of the diagonals, T the second point of intersection of the line AC
with the circle C
1
of diameter AD and S the second point of intersection of the line BD with the circle
C
2
of diameter BC. Then 'ATD = 'BSC = 90, so DT and SC meet in the orthocenter H of the triangle
DQC. Denote by C
3
the circle DCTS.
The radical axis of the circles C
1
, C
2
is MN, the radical axis of the circles C
1
, C
3
is DT, while the
pair of circles C
2
, C
3
has SC as radical axis, hence the radical center of the three circles is H.
The line segment MN is the common chord of the circles C
1
and C
2
, thus perpendicular to the line
passing through the centers, which is in fact the middle line of the trapezoid. As H MN, then
MH ,, DC, and since QH ,, DC the conclusion follows.

Let T BD be the tangency point of incircles of the triangles ABD and CBD. Notice that BM =
= BT = BN and DN = DT = DM.
a) We have AB + CD = AM + MB + CP + PD = AN + BQ + CQ + DN = AD + BC, so the
quadrilateral ABCD is circumscriptible.
b) Triangles AMN, DNP, CQP, BQM are isosceles, so
'QMN + 'NPQ = 360 ('AMN + 'BMQ + 'QPC + 'NPC) =
= 360
2
1
(4 180 A B C D) = 180,
hence the quadrilateral MNPQ is cyclic.
c) Let U be the point where the side AC touches the incircle of triangle ABC. Since AB BC =
= AD DC, then AU =
2 2
DC AC AD BC AC AB +
=
+
, so U is also the tangency point of the side
AC with the incircle of triangle ADC, as needed.

For a fixed point P inside the given triangle consider the point Q on bisector line of BC so that
AQ = AP. The parallel line d from Q to BC separates the arc MN and the side BC, so d meets the line
segment [BP] at a point, say S. The triangles inequality gives SP PC SC, so BP + PC BS + SC.
On the other hand, with an argument frequently refer to as Herons problem we have BS + SC BQ +
+ QC, so BP + PC is minimum if P = Q.
Let T be the midpoint of the segment MS. Notice that triangle AMQ is isosceles and MT is an
altitude in this triangle, hence MT = QZ, where Z is the foot of the altitude from Q onto AC. Then
MN + BQ + QC = 2(MT + CQ) = 2(CQ + QZ) is minimum when CZ AC. Consequently, the
required point is the orthocenter of the triangle ABC, which belongs to the interior of the triangle,
since it is an acute-angled one.

63 63 63 63
Let
AB
AM
k = . Using Thales Theorem, MP ,, BC yields k
AC
AP
= ,while MN || AC implies k
MN
MS
= .
On the other hand, k
PM
PR
BC
CN
= = .
Setting S = area[MNP], we have
S
NPR
PM
RP
MN
MS
S
MSP ] [ area ] [ area
= = = ,
hence area[MSP] = area[NPR]. Subtracting area[RPQ] from both sides of the latter equality we get the
conclusion.

Line AB is the radical axis of the circles
1
and
2
, and line DE is the radical axis of the circles
1

and
3
, hence point G is the radical center of the three circles. Since the radical axis of the circles
3

and
2
is the tangent line at C to these circles, it follows that the tangents from G to 3 are GF and
GC. Then GF = GC and GH = GF, so the triangle HCF is right-angled at C. Therefore 'HCF = 90.

A special position occurs when 'CAB = 60, when C = E = F. In this case the claim is obvious.
Consider the case 'CAB > 60, where E belongs to the small arc CD and F lie on the segment CD.
Notice that 'PFC = 'ADC = 'BCD, hence the trapezoid PBFC is isosceles. On the other hand, as
CD ,, AB and CE ,, AD, it follows that the arcs AC, BD, DE are equal. Then 'EFC = CE + BD = CE +
+ DE = CD = 'P'CD, where P' PC, C (PP'). The last equality proves the conclusion.
Slight changes in notations are required for the case 'CAB < 60.

Line BC meet DF, AE at points T, G respectively. Using Cevas theorem, it suffices to prove that
1 =
DC
AD
GA
EG
FE
CF
,
since only one or all points D, F, G lies on the sides of the triangle AEC.
Observe that 'BAD = 'BED = 'BTD = 90, so points A, C, E, T, D lies on the circle of diameter
BD. Then 'FDE = 'TBE
not
= and 'TDC = 'ABC
not
= . Moreover, DE = DA and AB = BE. The law
of sinuses gives
= =
sin sin
,
sin sin
FC
CFD
DC
EFD
DE FE
' '
.
and since sin'EFD = sin'CFD we have
(1)
=
sin
sin
DA
DC
FE
CF
.
On the other hand,

64 64 64 64
= =
sin sin
,
sin sin
AG
AGB
BA
EGB
BE EG
' '
.
hence
(2)
=
sin
sin
GA
EG
.
Multiplying the inequalities (1) and (2) concludes the proof.
Alternative solution. Let H be the intersection point of the lines DF and AE. The claim is equivalent
to
HE
AH
GE
AG
= , in other words the pairs of points A, E and G, H are harmonical conjugates.
Since I is the midpoint of the segment AE, the claim reduces to IG IH = IA
2
.
The segment AI is an altitude in the right-angled triangle ABD, so AI
2
= ID IB.
Angles 'HTB and 'HIB are right, so the points H, T, I, B are cocyclic. It follows that 'DHI = 'IBG
and further, DHI ~ IBG. Hence
IB
IG
IH
ID
= , so ID IB = IH IG, the conclusion is now obvious.

Consider the equilateral triangle BCA
1
, constructed in the exterior of triangle ABC. Then points A,
A, A
1
are collinear through the homothety of center A which map points U, V in B, C, respectively
Since AA
1
, BB
1
, CC
1
concur in the Fermat-Torricelli point of the triangle ABC, the claim is proved.

Let E and F be the midpoint of the sides AC and AB and let P be the mirror image of D across E.
The relation AM
2
+ AN
2
= BM
2
+ CN
2
gives
2 2 2 2
2 2 2 2
|
\
|
+ + |
\
|
= |
\
|
+ |
\
|
+ NE
b
FM
c
NE
b
FM
c
,
hence c FM = b NE. Then
AB
NE
AC
FM
= , so
EP
NE
FD
FM
= . Since 'MFD = 'NEP, we get MFD ~
~ NEP, which implies 'MDF = 'NPE. On the other hand 'MDN = 'BAC = 'FDE, so 'MDF =
= 'NDE.
Now the triangle NPD is isosceles and NE is a median in this triangle, so NE DP, in other words
A = 90.

Observe that AD = AN = AM and 'AND = 'AMC = 90, due to the reflections across AB and AC. It
is known that AD and AE are isogonal cevians, that is 'BAD = 'EAC. Then 'NAE = 'NAB + 'BAE =
= 'BAD + 'BAE = 'EAC + 'DAC = 'EAC + 'CAM = 'EAM and consequently NAE = EAM. It
follows that 'ENA = 'EMA, so 'BNE = 90 'ENA = 90 'EMA = 'EMC, as desired.

Let a, b, c, d be the radii of the circles , , , . It suffices to prove that
d
c
b
a
= , in other words the

65 65 65 65
ratio
b
a
is constant while point M varies on line d.
Let R and S be the midpoints of the arcs determined by d on the fifth circle K, the one tangent
simultaneously to , , , , and let N be on the same side of d as A. Denote by A
1
and B
1
the tangency
point of and to K, respectively. Observe that points A
1
, M, R are collinear via the homothety
which maps circle a onto circle K and similarly points B
1
, M, S are collinear. Since RS is a diameter
of K, angles 'RA
1
S and 'SB
1
R are right. If lines B
1
R and A
1
S meet as point V, then M is the orthocenter
of the triangle VRS. Notice that d RS. hence V d; denote by O the intersection point of d and RS.
Lines A
1
S and B
1
R intersect the circles and at points U and Z respectively. Since 'RA
1
S =
= 'SB
1
R = 90, the segments UM and ZM are diameters in circles , , so
SO
RO
ZM
UM
b
a
= = . The latter
ratio is constant, as claimed.

Let O be the midpoint of AD, R be the intersection point of lines AC and BD and S be the
intersection point of lines AF and DE. Since N and Q are the midpoints of the sides DB and DE of the
triangle DBE, we have O NQ and similarly O MP. Moreover, as DRAS is a parallelogram, the
diagonal RS passes through the midpoint O of the other diagonal, AD. Now apply Desargues Theorem
for triangles NRM and SPQ, given that O lies simultaneously on lines NQ, MP, RS and we are done.

The diagonal AC is the internal bisector for the angles 'BAD and 'BCD, so 'BAC = 'CAD =
= 'BCA = 'ACD. Observe that the triangles CMN and CMD have 'MCN = 'MCD and MN = MD,
hence their circumradius are equal and consequently CNMD is cyclic. It follows that 'MDP = 'MCN =
= 'RAP, implying that ARPD is cyclic. Then 'DRP = 'PAD = 'RAP = 'MDP, hence RP = PD, as
needed.

A well-known result states that the quadrilateral MNPQ is cyclic. For this, notice that the
quadrilaterals BMON, CNOP, DPOQ and AQOP have a pair of opposite right angles, hence all of
them are cyclic. It follows that 'OMN = 'OBN, 'OMQ = 'OAQ, 'OPQ = 'ODQ and 'OPN = 'OCN.
Summing up yields 'QMN = 'QPN, so MNPQ is cyclic, as claimed.
Next well show that the points M, N, P, Q lie on the circumcircle K of MNPQ. As an external
angle, we get 'NQO = 'QCO + 'COQ. Since CNOP is cyclic, then 'QCO = 'NPO. On the other
hand we have 'QOP = 'ODQ, because the sides are perpendicular, and 'ODQ = 'OPQ, as DPOQ is
cyclic. Therefore 'NQO = 'NPO + 'OPQ = 'NPQ, implying that Q belongs to the circumcircle of
MNPQ. In the same manner we find that N, P, Q he on K, which concludes the proof.

Since m
2
n
, no segment A
i
A
r(i)
contain the center O of the polygon A
0
A
1
A
n1
.

66 66 66 66
The chords A
i
A
r(i)
are equal, since all subtend m from the n equal arcs in which the vertices A
0
, A
1
,
, A
n1
divide the circumcircle of the polygon.
Consequently, the distances from O to each of the segments A
i
A
r(i)
are equal, hence these segments
are all tangent to a non-degenerate circle centered at O.
Since through a point one cannot draw more than two tangents to a circle, it follows that there are
no three concurrent segments of the form A
i
A
r(i)
.

Since the points X and Y are inside the polygon K, the line XY with the border of K contains two
points Z and Z, such that X, Y (ZZ).
The diameter of the polygon K is not greater than 1, hence ,ZZ, 1. To prove this, let AB the side
of K containing Z and let CD the side containing Z, where A, B, C, D are vertices of the polygon, not
necessarily distinct. Notice that AC, AD, BC, BD 1. In a triangle MNP, a cevian MS, S (NP)
satisfies the inequality MS < max(MN, MP). Therefore
ZZ < max(ZC, ZD) < max(max(CA, CB), max(DA, DB)) 1,
as claimed.
To conclude, observe that (XZ + YZ) + (XZ + YZ) = (XZ + XZ) + (YZ + YZ) = 2ZZ 2, hence
at least one of the sums XZ + YZ, XZ+YZ is less than or equal to 1.

Let a =
BC
AB
and
BC
AC
b = .

Then a
2
+ b
2
= 1 and the inequality rewrites as (a b)
2
(1 + 4ab)
2
2 and furthermore
(1 2ab)(1 + 4ab)
2
2.
Put x = 2ab to obtain (1 x)(1 + 2x)
2
2 or 4x
3
3x + 1 0, which is equivalent to
(2x 1)
2
(x + 1) > 0.
The latter is obvious, so we are done.

If the triangle ABC is isosceles, the claim is obvious. To prove the converse, observe that
AB
2
+ BC
2
+ CA
2
= AC
2
+ BA
2
+ CB
2
.
Using the standard notation one has
AB =
c b
ac
+
, AC =
c b
ab
+

and the analogous formulas. The above equality rewrites as

+
=
+
2
2 2
2
2 2
) ( ) ( c b
b a
c b
c a
,
hence
0
) (
2
=
+
c b
b c a
.
Clearing the denominators one obtains (a b)(b c)(c a)(a + b + c)
2
= 0, implying that ABC has
at least two congruent sides.

67 67 67 67
Chapter III

NUMBER THEORY
Suppose
2
2
5
2 4
b
a
n
n
=
+
, where a and b are coprime integers. One obtains

n =
2 2
2
2 2
2 2
4
22
5
4
5 2
a b
b
a b
a b
+ =
+
.
Since gcd(b
2
, 4b
2
a
2
) = 1, it follows that 4b
2
a
2
divides 22.
Observe that 4b
2
a
2
0 or 4b
2
a
2
3 (mod 4), hence we have either 4b
2
a
2
= 1, or 4b
2
a
2
=
= 1. The first case leads to b = 0, which is impossible. In the second case, we obtain 2b a = 1 and
2b + a = 11, hence a = 5, b = 3 and n = 13.

Suppose that the number obtained is n
2
= abcdef , where ef cd ab , , {16, 25, 36, 49, 64, 81}.
Since 161616 n
2
818181, it follows that 402 n
2
904. Thus, n = 100x + 10y + z and x 4, z 1.
By the squaring algorithm we obtain x
2
= ab . Also:
(l00x + 10y + z)
2
= 10
4
ab + 10
3
c + 10
2
d + 10e + f,
hence
2 10
3
xy + 2 10
2
xz + 10
2
y
2
+ 2 10yz + z
2
= 10
3
c + 10
2
d + 10e + f.
Two cases arise:
a) x = 4, y {0, l};
For y = 1, ab = 16, cd = 81, by using the above equality we get a contradiction.
For y = 0, ab = 16 and 8z 10
2
+ z
2
= 10
2
ef cd + . Since the representation of a number is unique
we get cd = 8z and since ef is a two digit number, it follows that 8z {16, 64}. Therefore, z = 8 and
n = 408.
b) x > 4 and y = 0. We obtain (200x + z)z = 10
2
ef cd + . As before we obtain cd = 2xz, ef = z
2
,
z 4. It follows that x = 8, z = 4, hence n = 804.
In conclusion, the students can obtain the numbers 408
2
or 804
2
.

Solution to Solution to Solution to Solution to Pr Pr Pr Problem oblem oblem oblem 94 94 94 94
Since a + b + c < 3d and d | a + b + c, it follows that a + b + c = d or a + b + c = 2d. Suppose that
a + b + c = d. Since a | b + c + d = 2d a, it follows that a | 2d and, similarly, b , 2d, c , 2d. Let

68 68 68 68
2d = ax = by = cz, where x > y > z > 2. We obtain
2
1 1 1 1
= + +
z y x
.
If z 6, then
2
1
6
1
6
1
6
1 1 1 1
= + + < + +
z y x
, so there are no solutions.
If z = 5, then
5
3 1 1
= +
y x
, and we obtain y = 3, again not possible.
If z = 4, then
4
1 1 1
= +
y x
, and we obtain the solutions (k, 4k, 5k, 10k) and (k, 2k, 3k, 6k), with k N.
If z = 3, then
6
1 1 1
= +
y x
, and we obtain the solutions (k, 6k, 14k, 21k), (k, 3k, 8k, 12k), (k, 2k, 6k, 9k)
and (2k, 3k, 10k, 15k), with k N.
Now, suppose that a + b + c = 2d. Analogously, we obtain that a, b, c | 3d, hence 3d = ax = by = cz
with x > y > z > 3 and
3
2 1 1 1
= + +
z y x
. Then z 4, y 5, x 6, thus
3
2
60
37
4
1
5
1
6
1 1 1 1
< = + + + +
z y x
, so
there are no solutions in this case.

For n = 0, we have 2
2
+ 1
2
+ 1
2
+ 1
2
= 7, hence (a, b, c, d) = {2, 1, 1, 1) and all permutations.
If n 1, then a
2
+ b
2
+ c
2
+ d
2
= 0 (mod 4), hence the numbers have the same parity. We analyse
two cases.
a) The numbers a, b, c, d are odd. We write a = 2a+ 1 etc. We obtain:
4a(a + 1) + 4b(b+ 1) + 4c(c + 1) + 4d(d+ 1) = 4(7 4
n1
1).
The left hand side of the equality is divisible by 8, hence 7 4
n1
must be even. This happens only
for n = 1. We obtain a
2
+ b
2
+ c
2
+ d
2
= 28, with the solutions (3, 3, 3, 1) and (1, 1, 1, 5).
b) The numbers a, b, c, d are even. Write a = 2a etc. We obtain
a
2
+ b
2
+ c
2
+ d
2
= 7 4
n1

so we proceed recursively.
Finally, we obtain the solutions (2
n+1
, 2
n
, 2
n
, 2
n
), (3 2
n
, 3 2
n
, 3 2
n
, 2
n
), (2
n
, 2
n
, 2
n
, 5 2
n
), n N,
and all permutations.

S SS Solution to olution to olution to olution to Problem Problem Problem Problem 96 96 96 96
Clearly x n
2
, so let x = n
2
p, with p > 0. If the number of radicals is 2, we obtain that x n
2
n.
It is easy to check using induction that all x n
2
n verify the inequality regardless the number of
radicals.

Solution to Solution to Solution to Solution to Pr Pr Pr Problem oblem oblem oblem 97 97 97 97
a) By squaring the members of the equation we get k + 100x = n
2
p + 2nxp + x
2
p, or 100 = p(2np + x).
The conclusion follows from the fact that p is a prime number.
b) If p = 2, then 50 = 2n + x and 0 n
2
25. Since n
2
=
2
k
p
k
= < 500, it follows that n 22 and we

69 69 69 69
have 23 solutions.
If p = 5, then 20 = 2n + x and 0 n 10. Notice that n
2
=
5
k
p
k
= < 200 for any n 10, therefore we
have other 11 solutions. We have 34 solutions in all.

We have ab + ad = 2(a + b + c + d) 6, so (a 2)(b 2) + (c 2)(d 2) = 2.
Assuming that a is the smallest number among a, b, c, d, we get 1 a 2 1.
If a 2 = 1 then b 2 = c 2 = d 2 = 1 and a = b = c = d = 3.
If a 2 = 0, then c 2 = 1 and d 2 = 2 (or c 2 = 2 and d 2 = 1). It follows that c d = 12, a = 2,
that is b = 6.
If a 2 = 1, then a = 1 and b + c + d 2 = b = cd. Hence c + d = 2, implying c = d = 1 and b = 1.
The solutions are (a, b, c, d) = (3, 3, 3, 3); (1, 1, 1, 1); (2, 6, 3, 4); (6, 2, 3, 4); (2, 6, 4, 3); (6, 2, 4, 3);
(3, 4, 2, 6); (3, 4, 6, 2); (4, 3, 2, 6); (4, 3, 6, 2).

As n is even, we have a
n
b
n
= (a
2
b
2
)(a
n2
a
n4
b
2
+ + b
n2
).
Since a + b is a divisor of a
2
b
2
, it follows that a + b is a divisor of a
n
b
n
. In turn, a + b divides
2a
n
= (a
n
+ b
n
) + (a
n
b
n
), and 2b
n
= (a
n
+ b
n
) (a
n
b
n
). But a and b are coprime numbers, and so
g.c.d.(2a
n
, 2b
n
) = 2. Therefore a + b is a divisor of 2, hence a = b = 1.

Solution to Solution to Solution to Solution to Problem Problem Problem Problem 10 10 10 100 00 0
Suppose, by way of contradiction, that u =
2
3 x and v =
3 3
x a are rational numbers. It
follows that x
2
= 3 u
2
and x
3
= a v
3
, that is a v
3
= (3 u
2
)
2
3 u . It follows that
2
3 u = q has
to be rational, and 3 = u
2
+ q
2
, both u and q being rationals.
Let m, n, p be integers with g.c.d.(m, n, p) = 1, such that u =
p
m
and v =
p
n
. Then 3p
2
= m
2
+ n
2
,
that is 3 is a divisor of m
2
+ n
2
. It is easy to see that 3 has to be a divisor of both m and n. Furthermore
9 is a divisor of 3p
2
, implying that 3 divides p. Since g.c.d.(m, n, p) = 1 we get a contradiction.

Denote by n
2
the perfect square and by a the digit that appears in the last four positions. It easily
follows that a is one of the numbers 0, 1, 4, 5, 6 or 9. It follows n
2
a 1111(mod 10
4
) and
consequently n
2
= a 1111(mod 16).
When a = 0 we are done. Suppose that a is 1, 5 or 9. Since n
2
0 or 1 or 4 (mod 8) and 1111
7 (mod 8), we obtain 1 1111 7 (mod 8), 5 1111 3 (mod 8) and 9 1111 = 7 (mod 8). Thus the
congruence n
2
a 1111 (mod 16) cannot hold.
Suppose a is 4 or 6. As 1111 7 (mod 16), 4 1111 12 (mod 16) and 6 1111 10 (mod 16). We
conclude that in this case the congruence n
2
a 1111 (mod 16) cannot hold either.

70 70 70 70
Solution to Solution to Solution to Solution to Problem Problem Problem Problem 1 11 10 00 02 22 2
Let d = g.c.d.(x, y) and x = da, y = db, where (a, b) = 1. It is easy to see that a and b are both odd
numbers and a
n
+ b
n
= 2
k
for some integer k.
Suppose that n is even. As a
2
b
2
1 modulo 8 we have also a
n
b
n
1 modulo 8. As 2
k
= a
n
+ b
n

2 (mod 8), we conclude k = 1 and a = b = 1, thus x = y = d.
The equation becomes x
n
= 2
m1
. It has an integer solution if and only if n is a divisor of m 1 and
x = y =
n
m 1
2

.
Consider the case when n is odd. From the decomposition a
n
+ b
n
= (a + b)(a
n1
a
n2
b + + b
n1
),
we easily get a + b = 2
k
= a
n
+ b
n
. In this case a = b = 1, and the proof goes on the lines of the
previous case.
To conclude, the given equation has solutions if and only if
n
m 1
is an integer and in that case
x = y = 2
p
.

The sum of all divisors of n is given by the formula
(1 + p + p
2
+ + p
a
)(1 + q + q
2
+ +q
b
).
The number n has (a + 1)(b + 1) positive divisors and their arithmetic mean is
) 1 )( 1 (
) ... 1 )( ... 1 (
2 2
+ +
+ + + + + + + +
=
b a
q q q p p p
m
b a
.

If p and q are both odd numbers, we can take a = p and b = q and it is easy to see that m is an
integer.
If p = 2 and q is odd, one can choose again b = q and a + 1 = 1 + q
2
+ + q
q1
. Then m = 1 + 2 +
+ 2
2
+ + 2
a
, and it is an integer.
For p odd and q = 2, we choose a = p and b = p
2
+ + p
p1
, concluding the proof.

Denote S =
4
31
4
2
4
1
... n n n + + + and A = {n
1
, n
2
, , n
31
}.
Firstly, observe that 2 A, otherwise all numbers n
i
, i = 31 , 1 are odd and consequently S must be
odd, a contradiction.
Then, 3 A, else n
i
1(mod 3) and
4
i
n 1(mod 3) for all i = 31 , 1 . It follows that S 31 1 (mod 3),
a contradiction.
Finally, we prove that 5 A. Indeed, if else, then n
i
1 (mod 5) or n
i
2 (mod 5) for all i = 31 , 1 .
Consequently,
2
i
n 1 (mod 5) and
4
i
n 1 (mod 5) for all i = 31 , 1 . Thus, S 31 1 (mod 5), a
contradiction.

Assume that such numbers exist. By squaring,
(1) 2n + 1 + 2 4 2 2
2
+ < + + < + n xy y x n n .

71 71 71 71
Since 4n + 1 < x + y + xy 2 2(x + y), we obtain x + y > 2n +
2
1
. Numbers x and y are integers,
so
x + y 2n + 1.
Set a = x y (2n + 1) 0, where a is an integer. The second inequality from (1) gives xy 2 <
< 2n + 1 a, hence 4xy < (2n + 1 a)
2
. Numbers 4xy and 2n + 1 a are also integers, therefore 4xy
(2n + 1 a)
2
1 and then 1 ) 1 2 ( 2
2
+ a n xy . From (1) we have
+ < + a n n
2
2 1 ) 1 2 ( 2
2
+ + a n a xy ,
hence
(2) 1 ) 1 2 ( 1 ) 1 2 (
2 2
+ + a n a n .
As x + y < 4n + 2, then a = x + y (2n + 1) 2n and so a < 1 ) 1 2 (
2
+ n . By squaring both sides
of the relation (2) we obtain
(2n + 1)
2
1 + a
2
1 ) 1 2 ( 2
2
+ n a < (2n + 1 a)
2
1
1 ) 1 2 ( 2
2
+ n a < 2a(2n + 1),
a contradiction.

Number a has an odd number of digits, hence 10
2k
a < 10
2k+1
for some integer k > 0. It suffices to
prove that a = 10
2k
.
Firstly, observe that 10
4k
a
2
< 10
4k+2
. Number a
2
has also an odd number of digits, hence 10
2k

a <
2
1
2
10
+ k
. Next, 10
8k
a
4
< 10
8k+2
and consequently 10
2k
a <
4
1
2
10
+ k
. Inducting on n we obtain
10
2k
a <
n
k
2
1
2
10
+
for all n > 0.
Assume by contradiction that a 10
2k
+ 1. Then
n
k
2
1
2
10
+
> 10
2k
+ 1 1 1 10 10
2
1
2
>
|
|
\
|
+
n
k

n
k k
n
2
2 2
2
1
10
1
1 10
10
1
1 10 |
\
|
+ > + > , for all n > 0.
On the other hand, using Bernoulli inequality we find that
k
n
n
k 2
2
2
10
2
1
10
1
1 + |
\
|
+ for all n > 0.
For sufficiently large n we have
k
n
2
10
2
1+ > 10, a contradiction.

72 72 72 72
Rearrange the numbers a
1
, a
2
, , a
n
in ascending order: b
1
< b
2
< < b
n
. Obviously, k b
k
, and
substituting b
k
with k, the left-hand side term increases. Furthermore, by the Rearrangements
inequality we infer that the maximum value of the left-hand side term is
1
...
1
2 1 n
n n
+ +
+ .
On the other side, the right-hand side term is greater than or equal to
4
) 1 (
2
... 2 1 +
=
+ + + n n n
.
We have
=
+
= + +
+
n
k
k
k n n
n n
1
1
1
...
1
2 1
=
=

+
= = =
+ = + + = +
1
2 2 1
1
) 1 (
1
) 1 ( 1
1
) 1 (
n
k
n
k
n
k
k
n
k
n n
k
n .
For n > 6 we prove by induction on n that
+
=
1
2
1
4
n
k
k
n
,
which implies that the given equality cannot hold. Indeed, for n = 7 we have
4
7
= 1.75 + +
3
1
2
1
+
+
8
1
= 1.51
If the inequality holds for n > 7 then it is true for n + 1, as
1
1
4
1
+
n
.
We are left with the cases when n = 2, 3, 4, 5, 6. Clearly, the case case n = 2 is impossible.
For n = 3 we have the numbers a
1
= 1, a
2
= 2 and a
3
= 3, so n = 3 is a solution.
If n = 4, then
a
1
+ a
2
+ a
3
+ a
4
= 13
1
4
2
3
3
2
4
1
2
4 3 2 1
2
4 3 2 1
< |
\
|
+ + +
|
|
\
|
+ + +
a a a a
,
so a
1
+ a
2
+ a
3
+ a
4
12. By inspection, all the cases: {a
1
, a
2
, a
3
, a
4
} = {1, 2, 3, 4}, {1, 2, 3, 5}, {1, 2,
4, 5} and {1, 2, 3, 6} fail to satisfy the required relation.
If n = 5, then
a
1
+ a
2
+ a
3
+ a
4
+ a
5
= 4 . 17
1
5
2
4
3
3
4
2
5
1
2
5 4 3 2 1
2
5 4 3 2 1
< |
\
|
+ + + +
|
|
\
|
+ + + +
a a a a a
,
so a
1
+ a
2
+ a
3
+ a
4
+ a
5
17. We study the cases {a
1
, a
2
, a
3
, a
4
, a
5
} = {1, 2, 3, 4, 5}, {1, 2, 3, 4, 6} and
{1, 2, 3, 5, 6} with no success (for an easy argument, observe that 5 must be a
5
and so on).
Finally, for n = 6 we obtain similarly a
1
+ a
2
+ + a
6
22, thus {a
1
, a
2
, a
3
, a
4
, a
5
, a
6
} can be {1, 2,
3, 4, 5, 6} or {1, 2, 3, 4, 5, 7}. The last case fails immediately because of 7, and the same outcame is
for the first one.
Therefore n = 3.

73 73 73 73
Clearly one of the primes p, q or r is equal to 2. If r = 2 then p
n
+ q
n
= 4, false, so assume that
p > q = 2.
Consider the case when n > 1 is odd; we have
2 1 3 2 2 1
) 2 ... 2 2 )( 2 ( r p p p p
n n n n
= + + +

.

Notice that ...) )( 2 ( 2 2 ... 2 2
4 2 1 1 3 2 2 1
+ + + = + +
n n n n n n n
p p p p p p > 1 and p + 2 > 1 hence
both factors are equal to r. This rewrites as p
n
+ 2
n
= (p + 2)
2
= p
2
+ 4p + 4, which is false for n 3.
Consider the case when n > 1 is even and let n = 2m. It follows that p
m
= a
2
b
2
, 2
m
= 2ab and r =
a
2
+ b
2
, for some integers a, b with (a, b) = 1. Therefore a and b are powers of 2, so b = 1 and a = 2
m1
.
This implies p
m
= 4
m1
1 < 4
m
, so p must be equal to 3. The equality 3
m
= 4
m1
1 fails for m = 1 and
also for m 2, as 4
m1
> 3
m
+ 1, by induction.
Consequently n = 1 take for example p = 23, q = 2 and r = 5.

The sum of the elements of the set A is S = na
2
n(n + 1)b +
2
6
) 1 2 )( 1 (
b
n n n + +
. Assume that n has
a prime divisor p > 3. Then p , S and p , (a + b)
2
+ (a + 2b)
2
+ + (a + pb)
2
= pa
2
+ p(p + 1)b +
+
2
6
) 1 2 )( 1 (
b
p p p + +
, a contradiction. It remains n = 2
k
3
l
, for some integers k, l. Suppose that k > 1.
Then 2 , S and so all elements of A must be odd. Taking the subset given by any pair we reach the con-
tradiction. Finally, suppose that l > 1, so 3 | S. If one of the numbers a + b, a + 2b, a + 3b is divisible
by 3, then we have a contradiction; if not, then 3 | b. Then 3 | (a + b)
2
+ (a + 2b)
2
+ (a + 3b)
2
= 3a
2
+
+ 12ab + 14b
2
, again a contradiction.
We are left with n = 6, which satisfies the claim: the set A = {4, 9, 16, 25, 36, 49} is isolated,
because the sum of its elements is a prime number (139).

S SS Solution to olution to olution to olution to Problem Problem Problem Problem 11 11 11 110 00 0
The required number is 505.
At start, note that the remainder of n when divided by 4 is odd, hence n is odd.
Furthermore, observe that the quotient of n when divided to a square less then
2
n
is greater than or
equal to 2. On the other side, the quotient at a division by an odd square cannot equal 3, as the
remainder would be even. Consequently, there are no positive integers k so that 3
2
) 1 2 ( k
n
< 4, in
other words there is no k N with
3
) 1 2 (
4
2
n
k
n
< .
Let m N
*
so that (2m 1)
2

3 4
n n
< < (2m + 1)
2
. Then (2m + 1)
2
(2m 1)
2
>
4 3
n n
, hence
8m >
12
n
. It follows that

74 74 74 74
(2m 1)
2

4
n
< 24 m,
so m {1, 2, , 6}. Since n < 96 m 576, then the odd squares less then
2
n
< 288 are 9, 25 , 225.
Recall that the quotients at the division by 9, 25, , 225 are even, so the quotients at the division by
225 and 169 are both 2 (else 4 169 > 576).
Thus n = 450 + a = 338 + b with 0 < a < 225, 0 < b 137 and a, b are odd, so n 338 + 137 = 505.
For n = 505 one can easy check the claim.

At first, note that if r = 1, then the new number is obviously rational.
Furthermore, if x is a simple rational number (that is with no period), then again the new number is
rational.
Now assume that x has a period and let p 1 the number of digits of the period.
Let m > 0 be the rank of the last decimal which do not belong to the period, thus
x = 0,a
1
a
2
a
m
(a
m+1
a
m+p
).
Let l = k + ir > m, i N. Then
x = 0,a
1
a
2
a
l
(a
l+1
a
l+r
a
l+pr
).
The number obtained by canceling the decimals a
k
, a
k+r
, a
k+2r
, a
k+3r
, will have the period
a
l
a
l+r1
a
l+r+1
a
l+pr1
,
therefore it will be a rational number.

Solu Solu Solu Solution to tion to tion to tion to Problem Problem Problem Problem 11 11 11 112 22 2
Suppose that n p. Then n
8
n
2
= n
2
(n
3
1)(n
3
+ 1) n
2
n
2
(n
3
+ 1) = n
7
+ n
4
p
7
+ p
5
> p
5
+ p
2
, a
contradiction. Therefore n < p. As p is prime and p
2
, n
2
(n
3
1)(n
3
+ 1), it follows that p
2
, n
3
1 or
p
2
, n
3
+ 1, so p
2
n
3
+ 1. Next we have n
2
(n
3
1)(n
3
+ 1) = p
5
+ p
2

5 3
) 1 ( + n + (n
3
+ 1), hence
(1) 1 1 ) 1 ( ) 1 (
3 3 3 2
+ + + n n n n ,
On the other hand n
3
+ 1 <
2 2
) 1 ( n , since this rewrites as n
2
n + 2. The relation (1) yields
n
2
(n
3
1) (n
3
+ 1)(n
3
1) + 1 = n
5
n
3
+ n
2
, then n
3
2n
2
. Consequently n = 2 and p
2
n
3
+ 1 = 9,
thus p = 3.

The numbers are 18 and 27.
Let k be the number of digits of n in decimal representation. Notice that:
(1) n = p s(n), where p is prime so that any prime divisor of s(n) is greater than or equal to p;
(2) s(n)
2
n, so 10
k1
n s(n)
2
(9k)
2
, hence k 4.
We study the following cases:
a) If k = 4, then n = abcd , n s(n)
2
36
2
= 1296, so a = 1. Then s(n) 28, thus n 28
2
< 1000,
false.

b) If k 3, then abc , so 9(11 a + b) = (p 1)(a + b + c).

75 75 75 75
If 9 divides p 1, since p < a + b + c = 27 we get p = 19. Next 9a = b + 2c, hence a 3. As a + b +
+ c 23 see (1) we have no solution.
If 9 do not divide p 1, from 3 | a + b + c and (1) we get p = 2 or p = 3.
For p = 3 we have n = 3(a + b + c), so a = 0 and 10 b + c = 3(b + c). Consequently, 7b = 2c and
n = 27.
For p = 2 we get n = 2(a + b + c), so a = 0 and 8b = c, hence n = 18.

Without loss of generality, assume a
k
> b
k
, k = 4 , 1 . Then a
1
+ a
2
= a
3
+ a
4
and b
1
+ b
2
= b
3
+ b
4
. We
analyze 2 cases:
i) a
1
+ a
2
= a
3
+ a
4
and a
2
+ b
1
= a
4
+ b
3
. Subtracting we get ,a
2
b
2
| = |a
4
b
4
,, |a
1
b
1
| = |a
3
b
3
,
and the claim is obvious.
ii) a
1
+ b
2
= a
4
+ b
3
and a
2
+ b
1
= a
3
+ b
4
. By subtraction we obtain |a
2
b
2
| = ,a
3
b
3
|, |a
1
b
1
, =
= ,a
4
b
4
|, as needed.

Solution to Solution to Solution to Solution to Problem Problem Problem Problem 1 11 11 11 15 55 5
Assume by contradiction that the claim holds and let b a. The number r(b) has at most as many
digits as b, so r(b) < 10b 10a. It follows that
(2a)
2
< 4a
2
+ 10a < (2a+ 3)
2
,
hence 4a
2
+ r(b) = (2a + 1)
2
or (2a + 2)
2
, thus r(b) = 4a + 1 or 8a + 4. Notice that r(b) > a b,
implying that a and b have the same number of digits. Then, as above, we get r(a) {4b + 1, 8b + 4}.
We analyze 3 cases:
1. r(a) = 4b + 1 and r(b) = 4a + 1. Subtracting we get (r(a) a) + (r(b) b) = 3(b a) + 2, which is
false since 9 divides r(n) n for any positive integer n.
2. r(a) = 8b + 4 and r(b) = 4a + 1 (the same reasoning is to be applied for r(b) = 8a + 4 and r(a) =
= 4b + 1). Subtracting we obtain (r(a) a) + (r(b) b) = 7b + 3a + 3, so 3 divides b. Then 3 divides
also r(b) = 4a + 1, so a and r(a) have the remainder 2 when divided at 3. This leads to a contradiction
with r(a) = (8b + 3) + 1.
3. r(a) = 8b + 4 and r(b) = 8a + 4. Both r(a) and r(b) have the last digit even, so at least 2. Then a
and b have the first digit greater than or equal to 2, so 8a + 4 and 8b + 4 have more digits than a and b.
It follows that r(a) < 8b + 4 and r(b) < 8a + 4, a contradiction.

Let m = ] [ n . Since n 4, then m 2 is an integer. We have m
2
n < (m + 1)
2
, so
m
2
n m
2
+ 2m.
Set n = m
2
+ k, k = 0, 1, 2, , 2m. From m 1 | m
2
+ k 1 we get m 1 | k. On the other hand
k < 2(m + 1), thus k = 0 or k = m + 1.
If k = 0, from m 1 | m
2
+ 1 follows m 1 | 2, so m = 2 or m = 3, hence n = 4 or n = 9.
If k = m + 1, then m 1 | m
2
+ m + 2 = m
2
1 + m 1 + 4, so m 1 | 4. We obtain m = 2, 3 or 5,
hence n = 7, 13 or 31.
Therefore, n {4, 7, 9, 13, 31}.

76 76 76 76
At first, notice that (x 1)(x 2) 0 for all x N, (y 1)(y 3) 0 if y N \ {2} and (z 1)(z 4)
0 when z N \ {2, 3}. In other words, if y 2 and z {2, 3}, then x
2
+ 2 2x, y
2
+ 3 3y and
z
2
+ 4 4z. Multiplying the above inequalities yields (x
2
+ 2)(y
2
+ 3)(z
2
+ 4) 60xyz, so in all three
inequalities the equality must occur. Until now we have the solutions:
(x, y, z) = (1, 1, 1), (1, 1, 4), (1, 3, 1), (2, 1, 1), (2, 3, 1), (2, 1, 4), (1, 3, 4), (2, 3, 4).
We claim the there are no more solutions. For this, we will show that if z = 2 or z = 3 or y = 2, there
are no integers satisfying the given equation.
The quadratic residues modulo 5 are 0, 1, 4, so 5 do not divide neither x
2
+ 2 nor y
2
+ 3. Since 5
divides 60xyz, it follows that 5 divides z
2
+ 4, hence z {5k 1 , k Z}. As a consequence, z 2 and
z 3.
If y = 2, the equation rewrites as 120xz = 7(x
2
+ 2)(z
2
+ 4), from which we may notice that 8 divides
(x
2
+ 2)(z
2
+ 4). If x, z are even integers, then x
2
+ 2 is even and z
2
+ 4 is divisible by 4, but 4 P x
2
+ 2
and 16 P z
2
+ 4, so the power of 2 in the right-hand side is at most 4, while in the left hand-side is at
least 5, a contradiction. If only one of the numbers x and z is even, the contradiction is reached
similarly. Hence y 2 and the only solutions of the equation are the ones previously obtained.

Any integer which can be represented as described in the problem will be called good.
Setting b = c = 1 yields [a, b] + [b, c] + [c, a] = a + 1 + a = 2a + 1, hence any odd integer is good.
Notice that [2x, 2y] = 2 [x, y]. Therefore, if n can be represented as [a, b] + [b, c] + [c, a], then 2n
writes as [2a, 2b] + [2b, 2c] + [2c, 2a] = 2([a, b] + [b, c] + [c, a]), thus all integers with are not powers
of 2 are good.
We claim that all numbers of the form 2
k
, k N are not good. For k = 0 and k = 1 this is obvious, as
[a, b] + [b, c] + [c, a] 1 + 1 + 1 = 3. If k 2, suppose by contradiction that there exist a, b, c as
needed. Let a = 2
A
a
1
, b = 2
B
b
1
, c = 2
C
c
1
, where a
1
, b
1
, c
1
are odd. Without loss of generality,
assume that A B C. Then 2
k
= [a, b] + [b, c] + [c, a] = 2
A
([a
l
, b
1
] + [a
1
, c
1
]) + 2
B
[b
1
, c
1
]. Dividing
by 2
B
, k > B yields 2
kB
= 2
AB
([a
l
, b
1
] + [a
1
, c
1
]) + [b
1
, c
1
]. But [a
l
, b
1
] + [a
1
, c
1
] is even and [b
1
, c
1
] is
odd, contradiction.

Suppose that p
2
P a + b. It suffices to prove that p
3
, a
3
+ b
3
. Indeed, if p
2
, (a + b)
3
3ab(a + b), we
have p | 3ab. As p 3 is prime, it follows that p , a or p , b. Since p , a + b, we get p | a and p | b. As a
consequence, p
3
| a
3
and p
3
, b
3
, implying p
3
| a
3
+ b
3
.

Let n 1 and let 10
k
, k N. Consider all the remainders of the numbers 10
k
at the division by n.
Since there are only finitely many residues, there exists a = 0, 1, , n 1 so that 10
m
a (mod n) for
infinitely many values of m N. Among them select only n, namely
n
m m m
10 ..., , 10 , 10
2 1
, with m
1
>
> m
2
> > m
n
. The number A =
n
m m m
10 ... 10 10
2 1
+ + + has n digits equal to 1 and the rest equal to 0,
has the sum of all digits is n. Moreover, A na 0 (mod n), so n | A, which concludes the proof.

77 77 77 77
Consider A
n1
= a
n
a
n1
. Since a
n
n and a
n1
1, we have A
n1
n 1.
If A
n1
= 0, that is a
n1
= a
n
, then a
1
+ a
2
+ + a
n2
is even and the claim reduces to the case of n 2
numbers.
If A
n1
> 0, then a
1
+ a
2
+ + a
n2
is even and the claim reduces to the case of n 1 numbers.

Notice that a
1
= n, as 1 divides any a
j
, and a
j
n, for any j = 1, 2, , n.
Consider an array with rows corresponding to 1, 2., n and columns corresponding to the numbers
a
1
, a
2
, , a
n
. Define the entries as follows, on the position (i, a
j
) put 1 if i divides a
j
and 0 otherwise.
Now observe that the sum of the numbers placed of the i
th
row is equal to a
i
, as we see the entries 1 for
each multiple of i in the sequence a
1
, a
2
, , a
n
. Hence the sum of all the entries in the array is
j
a .
On the other hand, on the
th
j
a column we get and entry 1 as long as we count the divisors of a
j
, so
the sum of the numbers on the j
th
columns is the number of divisors of a
j
. This implies that the sum of
all the entries in the array is the sum of all divisors of the numbers a
j
.
Using this double-counting, together with the obvious fact that the number of divisor of a is less
than a unless a is 1 or 2, show that n = 1 or n = 2.
Alternative solution. Recall that a
1
= n and a
i
n, for all i = 1, 2, , n.
Assume that n > 3. As a
n1
1, there exists a multiple of n 1, where n 1 1, in the given
sequence; let a
k
, k > 1 be such a multiple. The condition a
i
n shows that a
k
= n 1, in other words
there are n 1 multiples of k in the sequence. As n and n 1 are coprime, k does not divide a
1
= n, so k
divides a
2
, , a
n
. But k 2 and k | a
n
, therefore a
n
> 1. Thus n must occur at least twice in the
sequence, so, beside a
1
we have a
j
= n, j > 1. Hence k | n, a contradiction. As before, n = 1 or n = 2 are
the only possible values.

Let [an + b] = 2x
n
, for all integers n > 0. Then
(1) 2x
n
an + b < 2x
n
+ 1,
(2) 2x
n+1
a(n + 1) + b < 2x
n+1
+ 1.
Subtracting (1) from (2) we get 2(x
n+1
x
n
) 1 < a < 2(x
n+1
x
n
) + 1, for all n > 0. As 2(x
n+1
x
n
) 1
is an odd integer, it follows that all numbers 2(x
n+1
x
n
) 1 must be equal, as otherwise a belongs to
two open intervals of lengths 2 having the left margins at a difference of at least 2, which is
impossible. Hence 2(x
n+1
x
n
) 1 = 2s 1, s Z, so x
n+1
x
n
= s and then x
n
= ns + p, p Z, n > 0.
Plugging in (1) we get 2p b (a 2s)n < 2p b + 1, n > 0, so a = 2s, since otherwise the set of
the positive integers will have an upper margin. Observe that s is an integer, so a is an even integer, as
needed.

It is easy to observe that p is odd and p q. in other words p 3 and (p, q) = 1.
If q = 2, then 2
p+1
= 7 + p
2
. The only solution is p = 3, as 2
n+1
> 7 + n
2
, for all n 4.
For q 3, by Little Fermats Theorem we get p , 2q 7 and q , p + 7. Set p + 7 = kq, k N
*
.
If 2q 7 0, we have q = 3 and p | 1, false.

78 78 78 78
If 2q 7 > 0, then 2q 7 p, so 2q p + 7 kq, therefore k = 1 or k = 2. For k = 1 we obtain
p + 7 = q, so p , 2p + 7. This implies p = 7 and then q = 14, false. Hence k = 2 and p + 7 = 2q. Suppose
p > q; as p, q 3 we get q
p
q
p
and then 7 = 2q
p
p
q
q
p
3
3
= 27, a contradiction. Thus q > p and
then p + 7 = 2q > 2p, which yields p = 3 or p = 5. For p = 3 we have q = 5, while p = 5 gives q | 12,
with no solution.
To conclude, the solutions are (p, q) = (3, 2), (3, 5).

The solutions are (k, k
2
) and (k
2
, k), with k > 1.
We have mn 1 | (n
3
1)m n
2
(mn 1) = n
2
m. On the other hand, mn 1 , m(n
2
m) (mn
1)n = n m
2
.
If n > m
2
, then mn 1 n m
2
n 1, so mn n, false.
If n = m
2
, then obviously m
3
1 , m
6
1, so all pairs (m, m
2
), m > 1 are solutions.
If n < m
2
, from mn 1 n
3
1 we derive that
2
n m n < . Then mn 1 m
2
n < m
2
1, so
n < m. If n
2
m > 0, we obtain mn 1 n
2
m < n
2
1, so m < n, a contradiction. Hence n = m
2
,
which holds, since m
3
1 , m
3
1, so all pairs (n
2
, n), n > 1 are also solutions.

Since is not divisible by 2 nor by 5, the number a
2009
has a multiple M with all digits equal to 1.
Let m > 0 be the number of digits of M, that is: M =

times
1 ... 11
m
. As a
2009+m
= a
2009
10
m
+ 3M, we get
a
2009
| a
2009+m
, as needed.

Considering the remainders modulo 3 we get d = 0, otherwise 1 = 1 + (1)
c
, false. The equation
rewrites as 7
a
= 4
b
+ 5
c
+ 1.
Assuming b 0, the left-hand side term is odd while the right hand-side is even, so b = 0 and then
7
a
= 5
c
+ 2.
It is obvious that c 1. Taking the remainders modulo 5 we obtain 2
a
= 2, hence a = 1 + 4k and
7(49)
2k
= 5
c
+ 2. If k 1 then c 2. Modulo 25 we get 7(1)
2k
= 0 + 2, false. It follows that k = 0, then
a = 1 and c = 1.
Therefore a = 1, b = 0, c = 1 and d = 0.

Suppose n and n + 1 are saturated. Then the numbers 4n(n + 1) and 4n(n + 1) + 1 = (2n + 1)
2
are
also saturated, q.e.d.

79 79 79 79
Chapter IV

COMBINATORICS
a) If n = 2k, there are no such symmetrical tilings (otherwise the k and k + 1 tiles must have the
same colour).
If n = 2k + 1, the problem is to count the possible tilings for k + 1 squares. There are 4

3 ... 3 3 =
= 4 3
k
such tilings.
b) There are 4 3

2 ... 2 2 = 4 3 2
n2
tilings.

The first step give rise to one black square of area
9
1
3
1
2
= |
\
|
. After the second step we obtain eight
more squares of side
9
1
, the black region increasing by
2
9
8
. In the same manner, the third step
increases the black area by 8
2
= 64 black squares, each of area
27
1
, that is at this stage the black area
becomes
3
2
2
9
8
9
8
9
1
+ + .
We conclude that after 1000 steps, the area of the black region is
=
|
|
\
|
|
\
|
+ + |
\
|
+ + = + + + +
999 2
1000
999
3
2
2
9
8
...
9
8
9
8
1
9
1
9
8
...
9
8
9
8
9
1

1000
1000
9
8
1
9
8
1
9
8
1
9
1
|
\
|
=
\
|
= .
It is left to prove that the last number is greater than 0.001. This follows using a binomial
expansion evaluation, namely
1000
64 2
999 1000
8
1
2
1000
8
1
1
8
9
2
1000 1000
>
=
|
|
\
|
> |
\
|
+ = |
\
|
.
Consequently, 9999 , 0
1000
1
1
9
8
1
1000
= > |
\
|
.

80 80 80 80
Let us consider the general case, that is to consider the number a
n
of equilateral triangles formed by
division in n segments. We shall find a recurrence relation.
Consider an equilateral triangle with sides partitioned into n + 1 equal segments and draw the n
parallels to each side of the given triangle. We will count all triangles with at least one vertex on BC;
the remaining ones are triangles counted in a
n
.
Consider first the triangles that have two vertices on BC. When choosing two division points on
BC, say M and N with M (BN), one counts exactly one triangle, namely that one obtained by
drawing parallels from M, N to AB, AC respectively. Hence we add
2
) 1 )( 2 ( + + n n
new triangles.
Consider the triangles with only one vertex on (BC). For each of the n division points we count one
unit triangle. Except for the two points closer to B and C respectively, we count n 2 triangles of side
2, and so on. Therefore,
+ + + +
+ +
+ =
+
) 4 ( ) 2 (
2
) 1 )( 2 (
1
n n n
n n
a a
n n

Changing n in n + 1, we get
= + + + +
+ +
+ =
+ +
) 3 ( ) 1 ( ) 1 (
2
) 2 )( 3 (
1 2
n n n
n n
a a
n n

Summing up, we obtain
5
) 5 3 )( 2 (
2
) 2 )( 1 (
2
) 2 )( 3 (
2
) 1 )( 2 (
2
+ +
+ =
+ +
+
+ +
+
+ +
+ =
+
n n
a
n n n n n n
a a
n n n
.
It follows that
315 315 ... 237 145
2
) 5 8 3 ( 10
0 6 8 8 10
= + = = + = + =
+
+ = a a a a a .
Therefore, the number of triangles is 315.

Choose an arbitrary point A in the plane. Points located in the plane at a rational distance from A
are colored in red, while the others are colored in blue. Consider an arbitrary segment PQ. We may
assume that AP < AQ; if not, take instead P another point of the line segment (PQ).
Recall that between two real numbers one can find a rational number q and an irrational number r.
The circles centered at A and having the radii q and r intersect the segment PQ at the points M and N
respectively. It is obvious that M is colored in red and N in blue, so the claim is proved.

Color the 8 vertices of the cube in black or white such that the 4 vertices of the 2 regular
tetrahedrons have the same color; notice that the 3 vertices adjacent to a vertex have its opposite color.
Therefore, each movement increases the sum of the numbers assigned to the vertices sharing the same
color by 3. Consider the cases:
1) MN is a diagonal of a face of the cube. Then M and N have the same color, say black. Assume
by contradiction that there is a sequence of movements after which the same number n is assigned to
all vertices. Let k
1
and k
2
be the number of white, respectively black vertices that were selected to

81 81 81 81
perform the movements. Then
4n = 3k
1
+ 2 = 3k
2
,
a contradiction.
2) MN is a diagonal of the cube. Selecting the vertices M, then N, and performing these 2
movements, the number 1 will be assigned to all the vertices, as needed.
3) MN is a side of the cube. The same outcome as in the previous case will occur after 2 movements
when selecting the diagonally opposite vertices of M and N.
This provides us with a full solution.

The main idea is to observe that two consecutive rows have exactly 4 equal elements, namely those
lying on the columns 1, 3, 5, 7 or 2, 4, 6, 8. Moreover, on the other 4 columns the elements are
different. Wlog, assume that rows 1 and 2 are equal with respect to the columns 1, 3, 5, 7 and different
on the column 2, 4, 6, 8; we call these rows odd equal. If rows 2 and 3 are also odd equal, then rows 1
and 3 are equal, as needed. If not, then rows 2 and 3 are even equal. Now consider the rows 3 and 4;
we are done if the rows are even equal, so assume that they are odd equal. Finally, if rows 4 and 5 are
odd equal, then rows 3 and 5 are equal, and if rows 4 and 5 are even equal, then rows 1 and 5 are
equal. This concludes the proof.

Let n be the number of players in the tournament. The total numbers of matches is n(n 1), hence
each player end up with n 1 points.
a) Assume by contradiction that each player has a different number of draws. As a draw gives 0.5
points, it follows that each player has an odd number of draws. Since the possible cases are: 0, 2, 4, ,
2(n 1), we infer that each of these numbers is assigned to each of the players. Consider A the player
with 0 draws and B the player with 2(n 1) draws. Each player has played 2(n 1) matches, hence B
obtained a draw in each match played. The match A B thus ended with a draw, a contradiction, since
A has no draws.
b) Suppose the contrary. Then each of the n players has 0, 1, , n 1 losses when playing the
white. Let X and Y be the players with 0 and n 1 losses, respectively. The player Y has no points
when playing the white and n 1 points, so he won all the matches with the black pieces. This implies
that the match X Y is won by Y, so is lost by X, a contradiction, since X has 0 losses with the white
pieces.

Let A
1
, A
2
, , A
1000
be the vertices of the polygon. We start with two lemmas.
Lemma 1. Three of four consecutive vertices have the same color. Then after a sequence of moves
all vertices will have the color of the fourth vertex.
Proof. Let the colors be 0, 1 and 2. We have two cases:
a) 1110 1122 1002 2202 2112 0000.
b) 1011 1221 0000.
Lemma 2. Any 4 consecutive vertices will turn after several moves in the same color.
Proof. Form two pairs of consecutive vertices and change them in the same color if they do not
already have it. Then follow the sequence 1122 1002 2202 2112 0000.

82 82 82 82
By the second lemma, after several moves the vertices A
1
, A
2
, A
3
, A
4
will have the same color, say
red. Likewise, A
5
, A
6
, A
7
, A
8
will have the same color. Consider now the vertices A
4
, A
5
, A
6
, A
7
; the first
is red and the other three have the same color. By the first lemma they all will turn red of course, we
do nothing if they were already red. We move on with this procedure until A
1
, A
2
, , A
997
turn red
(note that 997 = 4 + 3 332, so this requires 332 steps). Now consider the vertices A
998
, A
999
, A
1000
, A
1
;
by the second lemma they all will share the same color. If this is red, we are done. If not, say that they
are blue, and taking the vertices A
997
, A
998
, A
999
, A
1000
we obtain using the first lemma all vertices to
be red, except for A
1
, which is blue. Now A
1
, A
2
, A
3
, A
4
turn blue, then A
5
, A
6
, A
7
, A
8
and so on. This
time, after 333 steps, all the 1000 vertices (1000 = 1 + 3 333) will be colored in blue.
Comment. Substituting colors with digits, notice that all moves: 01 22, 02 11 and 12 00
preserve the sum (mod 3). This means that the final color is unique and, of course, is given by the sum
of the digits assigned to the vertices of the initial configuration.

The number of routes between 2 cities is
|
|
\
|
2
6
= 15. By the pigeonhole principle one can find that a
company say M operates at least 8 routes.
There are
|
|
\
|
4
6
= 15 subsets of 4 cities and each pair of cities occurs in
|
|
\
|
2
4
= 6 such subsets. Using
again the pigeonhole principle follows that there exists a subset of 4 cities among which at least 4
routes are operated by the company M (3 15 < 6 8).
If those routes form a cycle, we are done. If else, then one can easy observe that among these 4
cities X, Y, Z, T, the routes X Y, Y Z, Z X and X T are operated by the company M (set the
notation accordingly).
The other 2 cities, say P and Q, are connected by at least 4 routes operated by M. Even if P Q is
one of them, from P or Q say P at least 2 routes to X, Y, Z, T are operated by M. If both
destinations are from X, Y, Z, a cycle is obtained, so assume that one of the routes is P T. In this
case we are done if the second route is P Y or P Z, hence we are left with the case P X.
From those above, the existence of a third route from P to X, Y, Z or T will provide a cycle. If not,
from Q exist 2 routes to X, Y, Z, T. The same line of reasoning shows that we have to consider that on
the route Q T is operated by M.
Any of the routes Q X, Q Y, Q Z will close a cycle. To conclude, if the route P Q is
operated by M, then any route from Q to X, Y, Z or T will provide us with the desired cycle.

1) A person can speak free of charge with at most 2k persons k that he chooses and other k (at
most) that select his numbers among their free calls. Since n 2k + 2, each person will be charged
when speaking to (at least) another one.
2) Assume that all 2k + 1 persons are arranged in a circle. Each person will choose to speak free of
charge when calling any of the k persons located consecutively at his right. Then any person will
speak free of charge with the k persons located at his left, as all of them will choose him among their
favorite numbers.

83 83 83 83
The sum of all integers is 1 + 2 + + n
2
=
2
) 1 (
2 2
+ n n
, so there is a row r having the sum of the
numbers assigned to its squares at least equal to
2
) 1 (
2 2
+ n n
. Consider the 2n 2 numbers written on
the first and last column except for the two numbers which belong to the row r and observe that
their sum is at least 1 + 2 + + (2n 2) = (n 1)(2n 1).
Now we can select two complementary parts of these columns in order to complete the row r to
a path so that the sum of the numbers placed on these n 1 squares is at least
2
) 1 2 )( 1 ( n n
. Since
2
) 1 (
2
+ n n
+
2
1
2 2
) 1 2 )( 1 (
2
3
+ + =

n n
n n n
, to conclude we only have to notice that
(
2
3
n
+ 1 =
=
2
1
2
3
+
n
if n odd and that
(
2
3
n
+ n
2
n +1 is the smallest integer greater than
2
1
2
2
3
+ + n n
n

when n is even.

One can place 4, 6, , 40, 42 checkers under the given conditions.
We start by noticing that n is the sum of 7 even numbers, hence n is also even. On a row one can
place at most 6 checkers, hence n 6 7 = 42.
The key step is to use 2k 2k squares filled completely with checkers and 2k + 1 2k + 1 squares
having checkers on each unit square except for one diagonal. Notice that these types of squares
satisfies the conditions, and moreover, we can glue together several such squares within the problem
conditions.
Below we describe the disposal of n checkers for any even n between 4 and 42.
For 4, 8, 12, 16, 20, 24, 28, 32 or 36 checkers 1, 2, 3, 4, 5, 6, 7, 8 or 9 2 2 squares; notice that all
fit inside the 7 7 array!
For 6 checkers consider a 3 3 square except for one diagonal; then adding 2 2 squares we get
the disposal of 10, 14, 18, 22, , 38 checkers.
For 40 checkers we use a 5 5 and five 2 2 squares.

Let n be the required number. We claim that n = 2007.
With 7 cuts, from the given rectangular piece one can obtain an 11-sided polygon and some
triangles. From a triangle, with 8 cuts one can get an 11-sided polygon and some extra pieces,
sufficiently enough to continue the same procedure. Hence, using 7 + 8 250 = 2007 cuts one can
obtain the 251 requested 11-sided polygons.
Denote by k the number of pieces left at the end which are not 11-sided polygons and notice that
each has at least 3 sides. Now, observe that with each cut the number of pieces increases by 1 and total
the number of vertices increases with at most 4 actually, with 2, 3 or 4, according to the number of
existent vertices through which the cutting line passes.

84 84 84 84
Then n = k + 250 and 4n + 4 v 11 251 + 3k, where v is the total number of vertices of all
polygons at the end. Hence 4n + 4 11 251 + 3(n 250) = 2011 + 3n, so n 2007, as claimed.

Assign the number 0 to each white square and the number 1 to each black square. The claim is
achieved if we prove the existence of a 2 2 square with an odd sum of the 4 numbers inside.
Assume the contrary, so each sum of the 4 numbers inside a 2 2 square is even. Summing over all
squares we get an even number S. Notice that each square not sharing a common side with the given
array occurs 4 times in 5, the squares with only a common side occurs twice, while the 4 squares in the
corners only once. But in the four corners there are three 0s and one 1, so the sum S is even, a
contradiction.
Remark. The given array may have a rectangular form, and the above solution requires no
alteration. However, this remark can easily lead to alternative solutions using induction. Here is a
sketch: choose a row with 0 and 1 at endpoints and call it the first row. Suppose that the number below
0 is also 0; arguing by contradiction, we notice that all doubletons formed vertically from the first
two rows have equal numbers inside, so the second row which starts with 0 ends with 1. Deleting
the first row of the given rectangular array, the claim is reached by induction. The same line of
reasoning is applied to the case when below 0 the number is 1.

Let a b be the greatest prime divisor of a + b.
At first, notice that from the initial 16 numbers we obtain 8 primes. The largest prime that can be
obtained is 31 = 15 16; if this number occurs, the second largest can be 23 = 11 12. Otherwise, 29
may occurs twice, from 16 13 and 15 14, followed by 19 or lower.
From the stage when we are left with 8 primes, and after pairing them we get 4 primes. If a prime is
obtained from two odd primes a and b, then a b
2
b a +
.
If else, at least one is 2 and let p be the other. The number p + 2 is prime only if p {3, 11, 17,
29}. Therefore, if p and q are prime with p q, then p g q + 2.
We will prove that the largest number which can end the game is 19. One example to obtain it is
exhibit below:
(1, 8); (2, 7); (9, 16); (10, 15); (3, 14); (4, 13); (5, 12); (6, 11) 3, 3, 5, 5, 17, 17, 17, 17
(3, 3); (5, 5); (17, 17); (17, 17) 3, 5, 17, 17
(3, 5); (17, 17) 2, 17
(2, 17) 19.
Now, we have to show that the game cannot end with a number strictly greater than 19. Since from
the second stage the number cannot increase with more than 2, and since 31 2 = 11, we derive that
the game will end with a prime p 31. Suppose by contradiction that p {23, 29, 31}.
If p = 29, as 29 is not sum of two primes, then p is obtained from two of 29. In the previous stage
four 29s are needed, then in the second stage eight 29s are required, in contradiction with an initial
observation. Moreover, we have obtained a stronger result: 29 cannot end the game and cannot occur
even in the last pair, since after 2 steps at most one 29 may occur.
Suppose that p = 31. Two cases are possible: 31 = 2 29 or 31 = 31 31. The latter result forbids

85 85 85 85
the first case, while the second case requires that the last four numbers are 31, 31, 31, 31 or 31, 31, 29,
2. But among the 8 primes obtained after the first step we have at most two 29s or one 31, not enough
to produce three 31s or two 31s and one 29.
Assume that p = 23. Again two cases are possible: 23 = 29 17 or 23 = 23 23. The first case is
impossible as shown above, while the second case is allowed if the last four primes are 23, 23, 23, 23
or 29, 17, 23, 23. If all primes are 23, the previous step has eight numbers with the average of 23,
which is a contradiction with
8 23 < 1 + 2 + 3 + + 16 = 8 17.
The second case lead similarly to contradiction, since 29 requires two 29s and the pair of 23s are
given by four numbers with the sum 4 23:
2 29 + 4 23 = 150 < 1 + 2 + + 16 = 136.
The solution is now complete.

Consider an arbitrary grouping in pairs. A pair in which the persons cannot speak will be called
bad. If there are bad pairs, we prove that some changes can be made to decrease to number of bad
pairs. Applying this at most 4 times exhibit a grouping with no bad pairs, and we are done.
Label the pairs A, B, C, D and the persons in pair X by X
1
and X
2
. Two persons that cannot converse
are called enemies, otherwise friends. Assume that A is a bad pair. Beside A
1
, the person A
2
has at
most two other enemies. Two cases arise:
a) If the other enemies of A
2
belong to the same pair call it B, then A
1
has at least a friend among
C
1
, C
2
, D
1
, D
2
.
Choose C
1
as a friend of A
1
and swap A
1
with C
2
. The new pairs A and C are good, and the claim is
satisfied.
b) If not, in at least one of the pairs B, C, D there are only friends of A
1
. Wlog, say that this pair is
B. One of the persons in this pair must be a friend of A
2
; call this person B
1
. Now swap A
1
with B
1
and
the new pairs A and B are good, as desired.
Remark. Consider the graph with vertices in the eight persons and edges corresponding to each pair
of friends. The degree of each vertex is at least 4, so, according to Diracs theorem there exists a
hamiltonian cycle. Taking 4 edges with ho common endpoint from this cycle, we get 4 good pairs, as
needed.

Given a set X of n points in the plane, consider a maximal free subset Y made of m elements, hence
such that any point in X \ Y completes an equilateral triangle with (at least) a pair of points from Y.
(Any X contains free subsets, since any subset with 1 or 2 elements is obviously free.) But for any pair
of points from Y there exist only two points in the plane which complete an equilateral triangle, so
n m
|
|
\
|
2
2
m
, that is n m
2
, or m n .
One checks the validity of this result for small values (1, 2, 3) of n, too (while the coplanarity
restriction is obvious).

86 86 86 86
Label the rows from 1 to 8 and the columns from 1 to 8. The unit square which lies on the row i
and the column j will be referred as (i, j).
On the skew-diagonal {(i, i) , i = 1, 2, , 8} there are exactly 2 squares in which checkers were
placed; wlog, assume that the squares are (1, 1), (2, 2). Looking at the 6 6 sub-array Q determined by
the rows 3-8 and the columns 3-8, we see that any skew-diagonal of Q, togheter with (1, 1), (2, 2), is
a skew-diagonal of the initial array. In view of the given conditions, no checkers are placed in the
squares of Q. Now take any skew-diagonal of Q with the squares (2, 1), (1, 2); this is a skew-diagonal
of the initial array, and the two checkers are placed inside (2, 1), (1, 2).
Up to this point, we know that checkers are placed in the squares on the rows 1-2 or on the columns
1-2. Suppose by way of contradiction that there exist a square located on the first two rows say
(i, m), i = 1, 2, m 3 and a square on the first two columns say (s, j), j = 1, 2, s 3 that hold
checkers. Then squares (i, m), (s, j), (3 i, 3 j) belongs to a skew-diagonal, contradiction.

For n 3, it suffices to draw 2n 2 squares, as below:
if n is odd, from each vertex V of P draw squares of sides
2
1 n
squares n 1, n 2, ,
2
1 + n
,
with V as vertex in each square.
if n is even, from each vertex V of P draw
2
n
1 squares of sides n 1, n 2,
2
n
+ 1, with V as
vertex in each square and add 2 more squares of side
2
n
with vertices in 2 opposite vertices of P.
To show that 2n 2 square are needed, divide all four sides in unit segments with 4n 4 points
other than the vertices of the square and consider for each point the unit segments perpendicular to
the border of the square. This 4n 4 segments can be covered by no less than 2n 2 squares, since a
square cannot cover 3 such segments, so we are done.

Two types of triangles arise after such a partition: < and =. Suppose that the initial triangle
ABC is of type < and notice that all three adjacent triangles the ones sharing a common side are of
type =. Call A-stripe the portion of plane containing the triangle ABC and bounded by the line BC and
the parallel from A to BC. (It is easy to observe that the plane is divided in stripes parallel to this
A-stripe.) Two =-triangles adjacent to ABC lie in this A-stripe. We claim that no path from ABC to
one of these triangles exists.
Indeed, observe that a move changes the type of the triangle hosting the checker, implying that if a
path must exists, then it has an odd number of moves. On the other hand, a move changes the position
of the checker from a stripe to an adjacent stripe located upwards or downwards. To reach in the end
the same A-stripe, such a path must consists in an even number of moves, a contradiction.
To conclude the proof, consider the B-stripe and the two =-triangles adjacent to ABC. Repeating
the argument, we are done.
Alternative Solution. Use four colors one for ABC and the three adjacent triangles to indicate
the triangles that can be visited starting from any of these triangles. Justify.

87 87 87 87
Suppose A = a
0
, a
1
, ,
1 2
n
a is a rearrangement of the numbers 0, 1, 2, , 2
n
1 which satisfies
the required condition. Then (2A), (2A + 1) is a rearrangement of the numbers 0, 1, 2, , 2
n+1
1,
where (2A) = 2a
0
, 2a
1
, , 2
1 2
n
a and (2A + 1) = 2a
0
+ 1, 2a
1
+1, , 2
1 2
n
a + 1. It is obvious that the
rearrangement (2A), (2A + 1) satisfies the claim. Indeed, no triples b
i
, b
j
, b
k
, with b
i
b
j
= b
j
b
k
,
i < j < k may occur in (2A) nor in (2A + 1), since either
2
,
2
,
2
k
j
i
b
b
b
or
2
1
,
2
1
,
2
1
k
j
i
b
b
b
belongs to A,
contrary to the fact that A is free of triples in arithmetic progressions.
Starting with the arrangement 2, 0, 3, 1 for the numbers 0, 1, 2, 3 and applying the above
procedure, in 5 steps one has the required rearrangement.

The required number is 2006, the number of the subsets having 2005 elements.
For start, notice that each subset must have at least 2004 elements. If there exist a set with exactly
2004 elements, then this is unique and moreover, only 2 other subsets may be chosen.
If no set has 2004 elements, then we can choose among the 2006 subsets, with 2005 elements and
the set A with 2006 elements. But if A is among the chosen subsets, then any intersection will have
more than 2004 elements, false. The claim is now justified.

We claim v(m, n) = mn + 1.
Partition M into n sets P
k
= {(x, y); n , x + y k}, k = 1, 2, , n. The pigeonhole principle now
forces (at least) m + 1 elements from P, be them A
i
= (x
i
, y
i
), to belong to a same P
k
. Now, if we
assume x
i
= x
j
, then from x
i
+ y
i
k x
i
+ y
i
k(mod n) it follows n | y
i
y
j
, but as y
i
, y
j
{1, 2, , n},
it follows y
i
= y
j
, i.e. A
i
= A
j
.
Conversely, mn + 1 is the least cardinality of P to warrant the claimed result; for |P| = mn, one can
pick P = {{x, y); 1 x m, 1 y n}; then any m + 1 elements from P, be them A
i
= (x
i
, y
i
), will share
at least one x
i
= x
j
(pigeonhole principle again).

Consider all positive differences a b among all 10 numbers. Since there are
2
10
C = 45 positive
differences and all belong to the set 1, 2, 3, , 36, at least two of them are equal. Let them be a b
and c d, with a > c. If a, b, c, d are all distinct, we are done; if not, then b = c, so b = c is one of the
8 numbers which are neither the lowest nor the greatest number from the initial ones.
Now observe that we have 45 positive differences and 36 possible values for them, so either 3
positive differences are equal or there are 9 pairs of equal positive differences.
The first case, gives a b = c d = e f, with a > c > e. Since we cannot have b = c, b = e and d =
e, we are done.
The second case gives at least one pair of positive differences in which case b = c is excluded, as
only 8 candidates for b = c exist, so we are done.

88 88 88 88
Consider a set S which satisfies all requirements. For each i A = {1, 2, , n}, define B
i
B the
set of all elements j B for which the pair belongs to the set S notice that some subsets B
i
can be
empty. Counting all pairs in S over all second element in each pair, we have ,S, = |B
1
| + |B
2
| + + |B
n
|.
The main idea is to observe the chain of inequalities
B
1
B
2
B
n
,
where by X Y we mean that x y, for any x X and y Y, X, Y being sets of integers. (This
definition allows the empty set to occupy any position in this chain.)
Since B
1
B
2
B
n
= B = {1, 2, , m} and any two consecutive subsets B
i
share in common
at most one element, we get by sieve theorem that ,S, m + n 1, as claimed.

Divide the interior of the circle into 12 congruent sectors such that each marked point lies in the
interior of some sector. If one of them contains exactly 100 marked points, we are done. If not, we can
find a sector A containing less than 100 points and a sector B containing more than 100 points (observe
that it is not possible that all sectors contain less than 100 points or more than 100 points).
Rotate sector A towards sector B. At each moment at most one marked point gets in or out sector A.
It follows that there exists a moment in which the rotating sector contains exactly 100 marked points.

Let A
1
, A
2
, , A
n
be the vertices of the polygon. We start with the following
Lemma: Each segment A
i
A
j
belongs to at most 2 triangle of area 1 located on the same side of the
line A
i
A
j
.
Proof of the lemma. Indeed, suppose that on the same side of the line A
i
A
j
exist the vertices A
m
, A
n
,
A
p
, so that the triangles A
i
A
j
A
m
, A
i
A
j
A
n
and A
i
A
j
A
p
have the area 1. Then the points A
m
, A
n
, A
p
will be at
the same distance to the line A
i
A
j
, hence colinear. This is a contradiction, since the polygon is convex.
Consider first the n sides of the polygon. Each of them can form at most 2 triangles of area 1, as all
the vertices lie on the same side, hence we have by now at most 2n such triangles.
Consider now the n diagonals A
i
A
i+2
with the cyclic notations: A
n+j
= A
j
. Each of them can form at
most 3 triangles of area 1, one with A
i+1
and two with the vertices lying on the other side. Thus we
have at most 5n = 2n + 3n triangles.
Finally, consider the other diagonals of the polygon. They are
2
) 5 ( n n
, and each of them can form
at most 4 triangles. The final counting is 5n + 4
2
) 5 ( n n
= n(2n 5), except that we have counted each
triangle three times, one time for each side. Therefore, there are at most
2
) 5 2 ( n n
triangles, as
claimed.

Table of Contents

FOREWORD ............................................................................................................................. 5

PROBLEMS
CHAPTER I. ALGEBRA .......................................................................................................................... 9
CHAPTER II. GEOMETRY .................................................................................................................... 16
CHAPTER III. NUMBER THEORY ........................................................................................................ 24
CHAPTER IV. COMBINATORICS ......................................................................................................... 29

FORMAL SOLUTIONS35
CHAPTER I. ALGEBRA ....................................................................................................................... 37
CHAPTER II. GEOMETRY .................................................................................................................. 49
CHAPTER III. NUMBER THEORY ...................................................................................................... 67
CHAPTER IV. COMBINATORICS .............................................................................................. 79

Tiparul executat la
Tipografia Editurii Paralela 45
COMENZI CARTEA PRIN POT

EDITURA PARALELA 45
Piteti, jud. Arge, cod 110174, str. Fraii Goleti 130
Tel./fax: 0248 214 533; 0248 631 439; 0248 631 492
Tel.: 0753 040 444;
0721 247 918.
E-mail: comenzi@edituraparalela45.ro
sau accesai www.edituraparalela45.ro

Carte Olimpiada - Balcanica-2010

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Carte Olimpiada - Balcanica-2010

Загружено:

Авторское право:

Доступные форматы

c o l e c i a

Aceast lucrare este rezultatul colaborrii dintre

, where a and b are coprime integers. One obtains

Вам также может понравиться