Estonian math

ompetitions 2000/2001 2. Find the largest real number K having the following property: for any positive
real numbers a; b; satisfying the inequality a b 6 K , the inequality ab 6 K
also holds.
+ +

We thank the IMO ommunity for many of these problems whi h have been taken
from various materials distributed at the re ent IMO-s. 3. Prove that, for any integer n > , the number | {z: : : } is divisible by , but is
0 11 1 3
n

3 digits n

not divisible by +1 . 3
n

Autumn Open Contest: O tober 2000

4. The terms of the sequen e a1; a2; a3; : : : satisfy the ondition a a 1 a 2
for any n > . Find the sum of the rst terms of this sequen e, if the sum
n = n n

of the rst terms is and the sum of the rst terms is .

3 2000
Juniors (up to 10th grade)
1997 2002 2002 1997

1. How many positive integers less than and not ontaining digits other 5. On a plane n points are given, no three of them ollinear. At most how many line
than and are there? segments it is possible to draw between these points in su h a way that the line
20002001

segments form no triangle with verti es at the given points?

0 2

2. Find the two last digits of the number :::

1! + 2! + 3! + .
+ 2000!

3. Consider points C1 , C2 on the side AB of a triangle ABC , points A1 , A2 on Solutions of Autumn Open Contest
the side BC and points B1 , B2 on the side CA su h that these points divide the
orresponding sides to three equal parts. It is known that all the points A1 , A2 , J1. .
B1 , B2 , C1 and C2 are on y li . Prove that triangle ABC is equilateral. Answer: 136

The set of integers under onsideration onsists of all integers with up to digits
4. Real numbers x and y satisfy the system of equations ontaining only digits and , all1 -digit integers of the form  and the3
7

integer . There are integers with exa tly k digits and , and
0 2 8 20000
k

integers of the form  . So the required number of integers is

8 20002000 2 0 2 2
> x
<x+y +
>
> = 10
20000

y 0 + 21 + : : : + 26) + 8 + 1 = (27
: (2 1) + 9 = 136 :
>
> x(x + y)
>
:
y
= 20
J2. Answer: 13 .
Find the sum of all possible values of the expression x y . The produ t

: : :
has , and as fa tors, therefore being divisible by
. Hen e the last two digits of n are zeros for any n > and it su es to
1 2 10 2 5 10
+

nd two last digits of . The two last digits of the summands are
100 ! 10

5. Let m m for any odd integer m and m m for any even integer m .
= +3 =
01 02 06 24 20
:::
, , , , , , , and , yielding as the answer.
20
1! + 2! +

40 20
+ 9!

80 13

a) Find all integers k su h that k .

2

b) Prove that, for every odd integer K , there exist pre isely three dierent
= 1

C
integers k su h that k K .
) How many dierent integers k with the property k K exist for an even
=
B1 A2
integer K ?
= r r

B2 r
A1
r

r r

Seniors (grades 11 and 12) A C1 C2 B

Figure 1
1. Points A , B , C , D , E and F are given on a ir le in su h a way that the three
hords AB , CD and EF interse t in one point. Express angle EF A in terms of J3. Label the points on the sides of the triangle so that jAC1j jC1C2j jC2Bj ,
angles ABC and CDE (nd all possibilities). jBA1 j jA1 A2j jA2 C j and jCB1 j jB1 B2 j jB2 Aj (see Fig. 1). Then we have
= =

= = = =

1 2
 6 BA1 C2 = 6 BA2 C1 = 6 BCA and 6 BC2A1 = 6 BC1 A2 = 6 BAC . Sin e points In ases (a) and (b) similar arguments give 6 EF A 6 CDE 6 ABC and
A1 , A2 , C1 and C2 are on y li , we get 6 BA2 C1 = 180Æ 6 AC2A1 = 6 BC2A1 , 6 EF A 6 ABC 6 CDE , respe tively.
=

whi h gives 6 BCA = 6 BAC . The equality 6 BAC = 6 CBA follows by symme-
=

try. E F
J4. Answer: 10 . A C A C A C A C
By Viete's theorem, the possible values of x y are in luded in the set of roots
of the quadrati equation E F
+

F E
a2 10a + 20 = 0 : D B D F
B D B D E
B
This equation has two dierent roots be ause D 2
> . Viete's (a) (b) ( ) (d)
formulae give to be the sum of these roots. It remains to he k that is not
= 10 4 20 0

10 11

among the roots (as y x y

x y
from the rst equation, x y 6 enables
=
+
+ = 11 Figure 2
us to nd the orresponding values for x and y ).
11 ( + )

Consider ase (d). Sin e EF A and ADE are opposite angles of a y li quadri-
J5. Answer: a) , and ; ) .
1 2 8 5
lateral ADEF and 6 ADE 6 ADC 6 CDE 6 ABC 6 CDE , we have
= + = +

a), b) Observe that if m is odd, then both m and m are even. Hen e if 6 EF A = 180Æ 6 ADE = 180Æ 6 ABC 6 CDE :
K k is odd, then k K , and k and k are not both odd. This gives the p
following three possibilities.
= = 2
S2. Answer: 3 3 .
1) If both k and k are even, then k k  k  K . Let a b+ 6 K. By the AM-GM inequality we have
2) If k is odd and k is even, then k k K .
= 2 = 4 = 8 +

k
3) If k is even and k is odd, then k k
k K .
 a + b + 3  3 K2
= 3 = (2 3) = 4 3

K = 2 = 2 ( 3) = 2 (2 3) = 4 6
ab 6 6 K K
:
The numbers K , K and K are pairwise distin t sin e modulo they
=
3 3 27

are ongruent to , and , respe tively. For a), K gives k 2 f ; ; g .

8 4 3 4 6 4

Hen e if K 6 , or equivalently K 6 p , the required ondition is satis-

0 1 2 = 1 1 2 8 2
) Let now K be even. If k is even, then we get the same three possibilities for 1 3 3

k as above. If k is odd, then k is even and k an be either even or odd.

ed. However, if K > p and a b K , then a b K and
27

4) If k is even, then k k k K K .

5) If k is odd, then k k K .
= 2 = 4 = 4( 3) = 4 12 3 3 = = = + + =

k K
3

K2
> K , so the ondition is not satised.
= 3 = 2 3 = 2( 3) 3 = 2 9

Sin e K is even, the numbers K , K , K and K are ongruent to ab K

=

, , and , respe tively, modulo . Moreover, K is ongruent to either

8 4 3 4 6 4 12
27

or modulo . Hen e these ve numbers are pairwise distin t. S3. We use indu tion on n .
0 5 2 4 8 2 9

3 7 8

S1. Angle EF A is equal to either 6 ABC 6 CDE , or 6 ABC 6 CDE , or The proposition holds for n sin e is divisible by 0 and is not
divisible by .
Base: = 0 1 3 = 1
1
6 CDE 6 ABC , or Æ 6 ABC 6 CDE .
Answer: +
3 = 3
180

Given the hords AB and CD , the hord EF an be drawn in four essentially Step:Observe the equality
dierent ways  point E an lie on the ir le between points D and A , between : : : 1} = 11 : : : 1}
: : : 0} 1 00 : : : 0} 1 :
points A and C , between points C and B or between points B and D (see 11
| {z | {z 1 00
| {z | {z
Fig. 2). 3 +1
n 3n 3n 1 3n 1
Let us nd 6 EF A for ase ( ). 6 Sin e EF6 C and CDE are 6 angles subtended The rst fa tor here is divisibleby but not by +1 by the indu tion hypothesis,
n n

by the same hord EC , we have EF C CDE ; similarly CF A 6 ABC . and the se ond fa tor is divisible by+1 but not by . +2Sin e is prime, this implies
3 3

Hen e that the produ t is divisible by , but not by .

= = 3 9 3
n n
3 3

6 EF A = 6 CF A + 6 EF C = 6 ABC + 6 CDE : S4. Answer: . 2012

3 4
 Denote a1 =and a2 q . It is easy to see that
p = Spring Open Contest: Mar h 2001
a p if k ; ; ; : : :;
a q if k ; ; ; : : :;
k = = 1 7 13

a q p if k ; ; ; : : :;
k = = 2 8 14 Juniors (up to 10th grade)

p if k ; ; ; : : :;
k = = 3 9 15

a k =
q if k ; ; ; : : :;
= 4 10 16
1. Eight students, Anne, Mary, Cathy and Tina, Anthony, Mark, Carl and Tom have
a
a p q if k ; ; ; : : :.
k = = 5 11 17
to work in four pairs, one boy and one girl in ea h pair. They know ea h other,
k = = 6 12 18
with only these ex eptions: Anthony knows neither Anne nor Mary; Mark doesn't
Observe that the sum of any six onse utive members of the sequen e is equal to know Mary and both Carl and Tom know neither Cathy nor Tina. How many
zero. Denoting S a1 : : : a , we get ways are there to divide the students into pairs, so that ea h boy ould work with
a girl he knows?
k = + +

S p if k ; ; ; : : :;
k

S p q if k ; ; ; : : :;
k = = 1 7 13

S q if k ; ; ; : : :;
k = + = 2 8 14
2. In a triangle ABC , the lengths of the sides are onse utive integers and median
q p if k ; ; ; : : :;
k = 2 = 3 9 15
drawn from A is perpendi ular to the bise tor drawn from B . Find the lengths
S
S q p if k ; ; ; : : :;
k = 2 = 4 10 16
of the sides of triangle ABC .
if k ; ; ; : : :.
k = = 5 11 17

S k = 0 = 6 12 18
3. In a s hool lo ker room there are lo kers
Hen e q p S1997 and q p S2002 , whi h give q and in three rows. The lo kers in ea h row are la-
60
K
with S2000 p q . belled from left to right with numbers to
= = 2002 2 = = 1997 = 5

p L
in the top row, to in the middle row and M
= 2007 = + = 2012 1 20

to in the bottom row.

21 40
n2 2
S5. for even n and n for odd n . 1 41 60

Kate's, Lisa's and Mary's lo kers are lo ated as shown in the gure. Ea h of the
Answer:

Divide the points into two subsets with ardinalities as lose to ea h other as
4 4

possible, and draw a line segment between any two points from dierent subsets. three lo ker numbers is divisible by the number of Mary's house, whi h is not . 1

Then ea h losed line formed by these line segments ontains an even number a) What is the number of Mary's house?
of links and hen e the line segments do not form any triangles2 with verti es at b) What ould be the numbers on the girls' lo kers?
 n 2
the given points. The number of line segments is n
for even n and 4. Integers a , b , and d satisfy ja bdj jad b j . Prove that either
jaj jbj or j j jdj .
= + = + = 1
2 4
2
n

n n
for odd n .
= = 1 = = 1
1 + 1 1

5. A onvex hexagon is onstru ted from n pie es, ea h of whi h

=

Now prove that there annot be more line segments. Consider any olle tion of line
2 2 4

segments satisfying the onditions of the problem. Let m be the maximal number is an equilateral triangle (one example is given in the gure).
of line segments in ident to one point, and let X be any point in ident to m line a) Prove that the hexagon is equiangular.
segments. Let A be the set of the other endpoints of these m line segments, and b) Find all possible values of n .
B be the set of the other n m points (in luding X ). Ea h point of A an be
joined only to points of B be ause any two joined points from A together with
X would form a triangle. Hen e ea h of the m points of A o urs as an endpoint Seniors (11th and 12th grade)
for at most n m line segments. On the other hand, ea h of the n m points
from B o urs as an endpoint for at most m line segments by the hoi e of m . 1. The serial numbers of lottery ti kets are -digit integers. It is known that the
So there is at most m n m n m m m n m segment-endpoint pairs, serial number of a winning ti ket has seven distin t digits and is divisible by ea h
7

and sin e every line segment has two endpoints, we have at most m n m line
( )+( ) = 2 ( )

segments. It remains to noti e that this expression a hieves its maximum when
( )
of its digits.
the dieren e of m and n m is as small as possible, i.e. if m n m for even a) Prove that the serial numbers of all winning ti kets onsist of the same digits.
n and if jm n m j for odd n . b) Find the largest possible serial number of a winning ti ket.
=

( ) = 1

5 6
 2. Let us all a onvex hexagon ABCDEF if 6 A 6 C 6 E 6 B 6 D 6 F . From the gure we obtain L K and M L K . Sin e
K , L and M are all divisible by the number of Mary's house n , the dieren es
boring + + = + + = + 21 = + 14 = + 35

a) Is every y li hexagon boring? and M L are also divisible by n . It follows that the only
b) Is every boring hexagon y li ? L K
possible value of n is . Now, sin e 6 K 6 , we obtain K or K ,
= 21 = 14

3. Find all real-valued fun tions f x dened for all real numbers whi h satisfy the and hen e L and M or L and M .
7 1 20 = 7 = 14

ondition f x f
x2 for ea h real x .

( ) = 28 = 42 = 35 = 49

2001 + (0) = 2001

J4. If the numbers a bd and ad b have the same sign, then a bd ad b
and a bd ad b a b d . Hen e a b or d . If the numbers
+ + + = +

4. For some < x; y <  , two of the three expressions 2 x 2 y , 2 x y a bd and ad b have distin t signs, then
0 = + = ( )( ) = =

and have equal values and the third one is dierent.

0 sin + sin sin ( + )
a bd ad b a b d
and a b or d . In both ases jaj jbj or j j jdj . If jaj jbj , then
+ + 0 = + + + = ( + )( + )
1

a) Whi h of the three expressions has a dierent value? ja bdj is divisible by jaj , therefore jaj jbj . In ase j j jdj , we
= = = = =

b) Give an example of x and y for whi h su h a situation o urs. similarly obtain j j jdj .
1 = +

= = 1
= = 1 =

5. There are small boxes numbered from to , and one large box. John puts J5. b) all integers n > .
some balls in some (or all) of the small boxes, and starts relo ating them by the
10 1 10
Answer: 6

following rules: a) Let a vertex of the hexagon beÆ the vertex of k triangles. Then the interior
 during ea h move, John removes all balls from any small box numbered n
angle at this vertexÆ must be k
. Sin e the interior angles of a onvex hexagon
are less thanÆ , noneÆ of them an be larger than Æ . Sin e the sum of the
60

where the number of the balls equals n ; angles is
, it follows that all the angles are equal to Æ .
180 120

 he adds these balls into boxes to n (one ball into ea h box) and puts
720 = 6 120 120

the remaining ball into the large box.

1 1
b) There must be at least pie es, be ause there is at least one triangle on ea h
side of the hexagon, and sin e the hexagon has no angles equal to Æ , ea h
6

He ontinues this way until he annot make another move a ording to these rules. triangle an lie on only one side of the hexagon.
60

Find the largest possible total number of balls in the small boxes at the beginning
of the game, for whi h it is possible to put all balls in the large box by the end of
the game.
Solutions of Spring Open Contest
n=6 n=7 n=8
J1. Answer : . 4 Figure 3
It is lear that Carl and Tom an only work with Anne and Mary: we obtain two The onstru tions for n , n ja n are given in Figure 3. We an
ways to form two pairs. Now, Anthony and Mark have to work with Cathy and ontinue the same way, adding more large triangles in the middle.
= 6 = 7 = 8

Tina, whom they both know: there are also two ways to form the two remaining
pairs. Hen e, altogether there are four ways to form the pairs. S1. Answer: b) .
9867312

J2. , and . a) Obviously, the serial number annot ontain and must be even. Now, we
annot have among the digits, be ause every even number divisible by ends
Answer: 2 3 4 0

Let D be the midpointof BC , then the medianis AD . Sin e the bise tor of 6 B is with a . If the serial number didn't have as one of its digits, it would ontain
5 5

also an altitude in the triangle ABD , that triangle is equilateral, i.e. jBDj jBAj and should be divisible by , but the sum of the remaining seven digits is ,
0 9

and in the original triangle ABC we have jBC j jABj . Sin e the lengths of
=

ontradi tion. Hen e is one of the digits. Now, the serial number is divisible by
3 3 31

the sides of triangle ABC are onse utive integers, the dieren e jBC j jABj is
= 2

and the sum of its digits is between and . The only possible sum is
9

either or . In the rst ase jABj , jBC j and the length of the side and the serial number onsists of the digits ; ; ; ; ; ; .
9 32 39 36

AC must be either or , whi h is impossible. In the se ond ase we obtain

1 2 = 1 = 2
1 2 3 6 7 8 9

jAB j , jBC j and jAC j .

0 3
b) Any -digit number onsisting of these digits is divisible by , , and .
Now we must nd the largest possible even number among these whi h is divisible
= 2 = 4 = 3 7 1 3 6 9

J3. Answer: a) ; b) ; ; or ; ; .
7 7 28 42 14 35 49 by
7 . This number is
8 = 56 . 9867312

7 8
S2. Answer: a) yes, b) no. Sin e  x sin

x and
 
x
= os 

x sin
3
x , then

= sin

+
 
= os

in both2 ases 2 2 x 2 2 y 2 x 2 x .
2 2 2

B sin + sin = sin + os = 1

A ? Let x y x y . We shall prove that both of these are equal to .

Applying the formula for x y and squaring, we get
sin + sin = sin ( + ) 1

sin( + )

C sin
2 x + sin2 y = sin2 x os2 y + sin2 y os2 x + 2 sin x sin y os x os y ;

D sin
2 x (1 os
2y) + sin2 y (1 os
2x) 2 sin x sin y os x os y = 0 ;
2 x sin2 y x sin y os x os y = 0 :
E F 2 sin 2 sin

Figure 4 Figure 5 Sin e < x; y <  , neither x nor y equals . Hen e we must have
, i.e. x y and x y equals to either
0 sin sin 0

x y x y
or , when e . So the only expression that an have a dierent
sin sin os os = 0 os( + ) = 0 sin( + ) 1

a) Let ABCDEF be a y li hexagon. Sin e the quadrilaterals ABDF , CDF B 2x y

and EF BD are also y li (see Fig. 4), we obtain value is 2 x y .
1 sin ( + ) = 1

sin ( + )

6 BDF =  6 A; 6 DF B =  6 C; 6 F BD =  6 E: b) Taking < x <  and y x  , we get

0 = +

Now,  6 A  6 C  6 E  , and hen e 6 A 6 C 6 E  .

2 2

Thus 6 B 6 D 6 F    6 A 6 C 6 E , whi h proves that the 2 x + sin2 y = sin2 x + sin2 x + 

 
( ) + ( ) + ( ) = + + = 2
2 x + os2x = 1 :
hexagon ABCDEF is boring.
+ + = 4 2 = 2 = + + sin = sin
2

b) Let us ompress a regular hexagon along its two opposite sides (see Fig. 5). Sin e  < y < x y  x <  , we have 2 x y 6 .
The new hexagon is boring sin e all its angles are equal, but it is not y li sin e + = +2
3
sin ( + ) = 1

three of its verti es lie on one ir le and the rest on another ir le. 2 2 2

x2 2 S5. .
S3. and f x x .
Answer: 41

fx
( 2001)
We shall rst show that the total number of balls in the small boxes annot ex eed
. John an empty box number only on e, sin e no balls are put into it during
Answer. ( ) = ( ) =
2001 2001

the relo ations. He an also empty box only on e, sin e at most one ball is put
41 10

Sin e for any real number y there is an x y f su h that y = x f ,

2001
(0)
= 2001 + (0)
into it (from the tenth box). Also, boxes , and an be emptied only on e.
9

2
Box number an be emptied at most twi e (at most balls will be added to it
8 7 6

the equality f y
y f
( ) = 2001 holds for any real y . Taking y we
2001
(0)
= 0
from boxes with bigger numbers). Box an also be emptied at most twi e, box
5 5

at most times, box at most times and box at most times. John an
4

2
f
get f (0) = and hen e f
(0)
or f (0) = 0. Therefore the fun tion
(0) = 2001
therefore put no more than balls in the large box.
3 4 2

41
7 1 21

We shall now nd a way to pla e balls in the smallboxes, so that all boxes ould
2001
2 2
must be either f y y or f y y
( ) = . It is easy to he k that both
( ) =
( 2001)
be emptied. No balls are added to box , therefore it must ontain balls. One
41

of these satisfy the given onditions. ball will be added to box , therefore it must ontain balls in the beginning.
2001 2001 10 10

Similarly there must be , and balls in boxes , and respe tively. Sin e
9 8

S4. a) 2x y an be the only expression with a dierent value; b) balls are added to box , it must ontain balls in the beginning in order to
6 4 2 8 7 6

be emptied twi e. Box must ontain one ball, box three balls, boxes and
Answer: sin ( + ) 5 5 5

hoose < x < arbitrarily and take y x .

0
2

= +
2 must ontain ball. The number of balls in the small boxes is now . It is easy
4 3 2 1

a) Let 2 x y . We shall prove that in this ase also 2 x 2 y . to he k that if John always empties the box with the smallest possible number,
1 41

all balls will nally be in the large box.

sin ( + ) = 1 sin + sin = 1

From the equation 2 x y we have either x y  or x y  .

sin ( + ) = 1 + =
2
+ =
3

9 10
Final Round of National Olympiad: Mar h 2001 (1) A is a word;
(2) if w is a word, then ww and ww are also words, where w is obtained from
w by repla ing all letters A with B and all letters B with A ( xy denotes the
9th grade
on atenation of x and y );
1. John had to solve a math problem in the lass. While leaning the bla kboard, he (3) all words are reated by rules (1) and (2).
a identally erased a part of his problem as well: the text that remained on board Prove that any two words with the same number of letters dier exa tly in half of
was
x  , where  marks an erased digit. Show that John an their letters.
still solve his problem, knowing that x is an integer.
37 (72 + 3 ) = 14 45

2. Dividing a three-digit number by the number obtained from it by swapping its 11th grade
rst and last digit we get as the quotient and the sum of digits of the original
number as the remainder. Find all three-digit numbers with this property.
3

1. The angles of a onvex n -gon are ; ; : : : ; n . Find all possible values of n

and the orresponding values of  .
2

3. A ir le of radius is tangent to two adja ent sides of a square and interse ts

its two remaining sides at the endpoints of a diameter of the ir le. Find the side
10

length of the square. 2. A student wrote a orre t addition operation BA DC FE to the bla kboard, su h
+ =

4. It is known that the equation jx j jx j : : : jx j a has exa tly that both summands are irredu ible fra tions and F is the least ommon multiple
one solution. Find a . of B and D . After that, the student redu ed the obtained sum EF orre tly by
1 + 2 + + 2001 =

5. A table onsisting of rows and olumnsis lled with integers ; ; : : :; an integer d . Prove that d is a ommon divisor of B and D .
in su h a way that ea h of these integers o urs in the table exa tly times and
9 2001 1 2 2001

the integers in any olumn dier by no more than . Find the maximum possible
9
3. Points D , E and F are taken on the sides BC , CA , AB of a triangle ABC ,
value of the minimal olumn sum (sum of the numbers in one olumn).
3
respe tively, so that the segments AD , BE and CF have a ommon point O .
Prove that jjOD AOj jAE j jAF j
j jEC j jF B j .
= +

10th grade 4. Let2 2 x 2and 2y be non-negative real numbers su h that x y . Prove that
xy x y 6 .
+ = 2

1. A onvex n -gon has exa tly three obtuse interior angles. Find all possible values
( + ) 2

of n . 5. ConsiderÆall trapezoids in a oordinate plane with interior angles of Æ , Æ , Æ

and , su h that their bases are parallel to one of the oordinate axes and
90 90 45

2. Find the minimum value of n su h that, among any n integers, there are three all verti es have integer oordinates. Dene the of su h a trapezoid as the
135

whose sum is divisible by . total number of points with integer oordinates inside and on the boundary of the
size

trapezoid.
3

3. There are three squares in the pi ture. Find the sum of D

a) How many pairwise non- ongruent su h trapezoids of size are there?
angles ADC and BDC . b) Find all positive integers not greater than that do not appear as sizes of
2001

any su h trapezoid.
50

A B C
4. We all a triple of positive integers a; b;
( ) harmoni if a b . Prove that,
1
+
1
=
1

12th grade
for any given positive integer , the number of harmoni triples a; b; is equal
to the number of positive divisors of 2 .
( )

1. Solve the system of equations

5. A tribe alled Ababab uses only letters A and B, and they reate words a ording 
x y :
to the following rules:
sin =

sin y = x

11 12
 2. Find the maximum value of k for whi h one an hoose k integers out of 3) If , then 6 a b 6 363 whi h implies a > , a ontradi -
; ; : : :; n so that none of the hosen integers is divisible by any other hosen tion.
= 3 300 32 = 200 + 7 10

integer.
1 2 2

p
9-3. .
3. Let I and r be the midpoint and radius of the in ir le of a right-angled triangle
Answer: 10 + 5 2

ABC with the right angle at C . Rays AI and BI interse t the sides BC and
AC at points D and E , respe tively. Prove that 6B C
jAE j jBDj r .
1 1 1
+ =

4. Prove that, for any integer a > , there is a prime p su h that a a2 : : : a 1 p

O
q

is omposite.
1 1+ + + +

5. Consider a  table, lled with real numbers in su h a way that ea h number -

A
in the table is equal to the absolute value of the dieren e of the sum of numbers
3 3

in its row and the sum of numbers in its olumn. Figure 6

a) Prove that any number in this table an be expressed as a sum or a dieren e Introdu e a oordinate system where the sides of the square tangent to the ir le
of some two numbers in the table. are on the oordinate axesthen the entre of the ir le is O ; (see Fig. 6).
b) Show that there exists su h a table with numbers in it not all equal to 0. Let the side of the square be a (evidently a > ) and the interse tion points of
(10 10)

the ir le with its two other sides be A and B . As AB is the diameter of the
10

ir le, the ommon pointp C a; a of these two sides lies on the ir le. Sin e
p CO
is a radius,pwe obtain a 2 a 2 , giving a
( )
Solutions of Final Round and
.
( 10) + ( 10) = 10 10 = 5 2

a
9-1. x = 1271 .
= 10 + 5 2

9-4. .
Answer:

From the given equality we obtain x) =  . To nd the number

y = 24 + x , note that
Answer: 1001000

Note that if x is a solution of the equation, x is also a solution. For

111(24 + 14 45

uniqueness we have x x , or x . In this ase

2002

< 1445 < 222000 = 111
2000 ;

= 2002 = 1001
111 1000 = 111000

a : : : + 2 + 1 + 0 + 1 + 2 + : : : + 999 + 1000 =
therefore y is a 4-digit number, with as its rst digit. Evidently y must end = 1000 + 999 +

with . Let y ab , where 6 a; b 6 . Writing out the multipli ation we

1
= (1000 + 1) + (999 + 2) + : : : + (2 + 999) + (1 + 1000) =

see that b ends with , hen e b and there is a arry of at least from
5 = 1 5 0 9
= 1000
1001 = 1001000 :

the third position. Sin e there is no arry to the rst position, we have a 6 .
+ 5 4 = 9 1

If the arry from the third position were more than , we would have a > , a
2

ontradi tion. Hen e a and x .

1 8
Although this is not required in the problem, it an be veried that
is indeed the only solution of the equation for a .
Remark.
= 2 = 1295 24 = 1271
x = 1001 = 1001000

9-2. Answer: 441 and . 882

9-5. 24.
We look for a number ab su h that ab ba a b , or a b . Answer:

The numbers 1 an be in the same olumn only with numbers 2, 3 and 4. As there
Hen e 6 6 , and we have 3 ases.
= 3 +( + + ) 32 = 100 + 7

1) If , then 6 a
3

whi h implies 6 a 6 . If a , are altogether
of these, the 1-s an be at most in four olumns. If all 1-s are
b6 in the same olumn, the minimal olumn sum is 9. If the 1-s are in two olumns,
4 9

then b and b . If a , then b and b is not an

= 1 100 32 = 100 + 7 163 4 5 = 4

integer.
128 = 100 + 7 = 4 = 5 160 = 100 + 7
one of these must ontain at least 5 of them and the sum of this olumn is at most
2) If , then 6 a whi h implies 6 a 6 . If a ,

. If the 1-s are in four olumns, then the sum of all numbers in
these olumns is , hen e the minimal olumn sum is at most
5 1 + 4 4 = 21
b6

then b and b is not an integer. If a , then b , giving
= 2 200 32 = 200 + 7 263 7 8 = 7 9 (1 + 2 + 3 + 4) = 90
h i
b . = 8
224 = 200 + 7 = 8 256 = 200 + 7 90

4
. If the 1-s are in three olumns, we should have 3-s and 4-s in these
= 22

13 14
 olumns to obtain the largest olumn sum. In this ase the sum of numbers in the right-angled triangles, we have
three olumns is
and the minimal olumn sum is at most 24.  
From the table below we see that this value is indeed attainable.
9 (1 + 3 + 4) = 72
6 ADF = 6 ADC 6 F DG = 6 ADC 6 DF G =

1 1 1 2 2 6 7 ... 2001
2
 
6 ADC 6 BDC ;
1 1 1 2 2 6 7 ... 2001 =

1 1 1 2 2 6 7 ... 2001
2

3 3 3 2 2 6 7 ... 2001 that implies 6 ADC 6 BDC  6 ADF . The segments AF and DF are
3 3 3 2 5 6 7 ... 2001
+ = +

transformed into ea h other by a Æ rotation around F . Hen e AF D is an

2

3 3 3 5 5 6 7 ... 2001 90

4 4 4 5 5 6 7 ... 2001 isos eles right-angled triangle with 6 ADF , yielding 6 ADC 6 BDC .
 3

4 4 4 5 5 6 7 ... 2001
= + =
4 4

4 4 4 5 5 6 7 ... 2001 Remark: There are also solutions using the osine theorem or the identity
10-1. Answer: The possible values of n are , and . 4 5 6
tan(  
+

) =
 

tan
. + tan

The sum of the angles of a n -gon is n
 . Sin e three of these angles are 1 tan tan

10-4. As a and b are non-zero integers, we have

( 2)

greater than and less than  , and the remaining n angles are greater than
2
3

0and less or equal to  , we obtain n

 < n
 < n

 .

2
( 3) 0 + 3
2
( 2) ( 3)
2
+3
1

a
+
1

b
=() a ab
1+b
=
1

() ab = (a + b) ()
Dividing by  and transforming yields < n < . As n is an integer, we have
7
7 () ab a b = 0 () ab a b + 2 = 2 ()
6 n 6 , and it is easy to he k that all these three values are indeed possible.
4 6
2
() (a )(b ) = 2:
10-2. Answer: n . = 5 Now let a b . If a and b are positive, then a > and b > . On
1
+
1
=
1
0 0

The sum of any three integers ongruent to , and modulo is divisible the other hand, if a > and b > , then a and b are positive. Hen e the
by . Also, the sum of any three integers ongruent to ea h other modulo is
0 1 2 3

harmoni triples a; b; are 2in one-to-one orresponden e with pairs of positive

0 0

divisible by . Consequently, among any ve numbers there are three whose sum
3 3

integers r; s , where rs 2, and there are as many su h harmoni triples as

( )

is divisible by . On the other hand, among the numbers , , and there are
3

there are positive divisors of .

( ) =

no three with a sum divisible by .

3 1 3 4 6

10-5. We use indu tion on the length of a word. Let u1 and u2 be any dierent words
D of the same length, and suppose the laim holds for all shorter words. As there is
only one word of length 1, u1 and u2 are onstru ted by rule (2). This implies
that there exist words v1 and v2 so that u1 v1v1 or u1 v1v1 and u2 v2v2
B or u2 v2v2 . Note that v1 and v2 are of the same length. If v1 v2 v , then
= = =

A C
one of the words u1 and u2 is vv and the other vv , diering exa tly in half of
= = =

their letters. If v1 6 v2 , then v1 and v2 dier exa tly in half of their letters by
the indu tion hypothesis, and it remains to show that the latter halves of u1 and
=
F G
u2 also dier exa tly in half of their letters. If these halves are v1 and v2 or v1
Figure 7 and v2 , this is obviously true. The words v1 and v2 , as well as v2 and v1 , dier
exa tly in the letters where v1 and v2 oin idediering therefore also exa tly
10-3. Answer:
3
.  in half of their letters. Hen e in any ase u1 and u2 dier in half of their letters.
Consider points F and G as shown on Fig. 7. As BCD and DGF are ongruent
4
11-1. Answer: The only possibilities are n ,   and n ,   . = 3 =
6
= 4 =
5

15 16
 Obviously n > . As the sum of angles of the n -gon is n
 n  n ,
3
( + )
= ( 2)
11-4. Denote  = 1 x , then x = 1  and from x + y=2 we get y = 1+ . Now
x2 y2 (x2 + y2 ) )2 (1 + )2
((1 )2 + (1 + )2 ) =
2

we have  n nn . Be ause of onvexity, we have n nn < 

=
2 (

( + 1)
2)
=
2 (

+1
2) =

=
(1

((1 )(1 + ))2
(2 + 22) =

yielding n < . If n , we obtain   ; if n , then   .
5 = 3 =
6
= 4 =
5
= 2(1 2)2 (1 + 2) = 2(1 4)(1 2) :

11-2. Let D0 and B0 0be the0 multipliers of the0 rst and the se ond fra tion, respe tively. Sin e x; y > ,4we have2 jj 6 that implies 6 2 6 and 6 4 6 .
Hen e  6 .
0 1 0 1 1 0 1 1

Then E AD B C and F BD DB0 , with B0 and D0 oprime sin e F 

2(1 )(1 ) 2

is the least ommon multiple of the denominators. If, for a prime p , p divides d
= + = =

with k >0 , then p divides both E and0 F . Suppose p does not 0divide B . From0
k
11-5. Answer: a) ; b) , , , , , , , , ja .
7 1 2 3 4 6 8 10 16 28 32
k

F BD we obtain that p divides D , hen e p also divides B C E AD .

0
k
Consider a trapezoid of height h and the length of its shorter base a (see Fig. 9).
Therefore, p divides0 either B0 or C , and as B0 and D0 are oprime, p divides
= =
The longer base of the trapezoid is of length a h and thus there is a total of +

C . From F DB we get that D is divisible by p , hen e p is a ommon fa tor

=
k
N (a; h) = (a +1) + (a +2) + : : : + (a + h +1) =
(2 + a h +2)(h +1)
of C and D , ontradi ting the irredu ibility of DC . We on lude that p divides k 2

B , and similarly also D . Sin e this is true for any prime divisor p of d , then B points with integral oordinates inside and on the border of this trapezoid.
and D are both divisible by d .
11-3. Draw a line parallel to BC through A and denote its interse tion points with 6
rays BE and CF by L and M , respe tively (see Fig. 8). From similar triangles q q q
a q q q q q q

jAE j jALj . Analogously jAF j jAM j . Moreover,

q q q q q q q q q q

AEL and CEB we have

q q q q q q q q q q

jEC j jBC j =
jF B j jBC j = q
h q q q q q q q q

j AOj jALj
q q q q q q q q q q

from similar triangles AOL and DOB we get jODj jBDj , and analogously =
q

q
q

q
q

q
q

q
q

q
q

q
q

q
q

q
q

q
q

q -
jAOj jAM j Figure 9
jODj jDC j . Hen e
=

jAOj jALj jAM j jALj jAM j jALj jAM j jAE j jAF j a) We have to nd the number of distin t pairs a; h for whi h N a; h .
Taking into a ount that

, we onsider two ases:
( ) ( ) = 2001

jBC j jBC j jEC j jF B j :

+ +

jODj jBDj jDC j jBC j

2001 = 3 23 29

1) If h k is even, then N a; h a k
k where k > and

= = = + = +
+
= 2 ( ) = ( + + 1) (2 + 1) 2 +1 3

C
a k
+ >k
+ 1 >
k
+ 1. The fa tor k an be , or , yielding the
2 + 1
2 + 1 3 23 29

pairs ; , ; and ; .
2
(665 2) (75 22) (54 28)

D 2) If h k is odd, then N a; h a k
k , where h N a; h

k > and a k > k . The fa tor k an be , ,
= 2 1 ( ) = (2 +2 + 1) ( )
L a
or , yielding the pairs ; , ; , ; and ; .
1 2 +2 +1 2 +3 1 3 23 1 2 + 3
O 29 (999 1)
a (330 5) (20 45) (5 57)
E 2 3 + 6

F b) For h ; ; ; : : :; we express the size of a trapezoid in a

terms of a (see the table); if h > , then N a; h > for any
= 1 2 3 7 3 4 + 10

a
a > . It is easy to he k that numbers , , , , , , ,
A B 7 ( ) 50 4 5 + 15

a
, and are the only ones that annot be expressed by
1 1 2 3 4 6 8 10
5 6 + 21
M
any of the formulae in the table. a
16 28 32
6 7 + 28

Figure 8 12-1. Answer: The only solution is x y . a

= = 0
7 8 + 36

17 18
 Clearly x y is a solution. We know that j xj 6 jxj , where equality holds 12-4. If a , then p gives the desired result:
i x (this an be easily proved using derivatives). Now
= = 0 sin = 2 = 11

: : : + 210 = 211
= 0
1+ 2+4+ 1 = 2047 = 23
:
89
jxj > j xj = jyj > j sin yj = jxj ;
sin
If a > , then a > and there exists a prime2 p that divides a . Hen e a
and at least one of the inequalities is stri t if x 6 or y 6 . is ongruent to modulo p and M a a : : : a 1 is divisible by p .
2 1 1 1
p

We also have M > a > p , implying that M is omposite.

= 0 = 0 1 p = 1+ + + +

12-2. Answer: n.
p 1+ p

Let the hosen integers be a1; : : : ; a and, for ea h i ; : : :; k , let n be the 12-5. a) Let r1 , r2 , r3 be the sums of numbers in the rst, se ond and third row, and
exponent of in the prime fa torization of a , i.e. a
b with b odd. Sin e 1 , 2 , 3 be the sums of numbers in the rst, se ond and third olumn. Denote by
k = 1 i
ni

6 b 6 n , there are n possibilities for the numbers b . If k > n , then a the element in the i -th row and j -th olumn, and noti e that all the elements
2 i i = 2 i i

there exist indi es i and j su h that b b b and n > n . Then a of the table are non-negative.
1 i 2 1 i + 1 ij

b ni

is divisible by a
b. Sin e r1 r2 r3 1 2 3 , we have

i = j = i j i = 2
nj
j = 2

If k 6 n , then hoose any k numbers in the set f n ; : : :; n g . None of them

+ + = + +

is divisible by another sin e n <
n . jr1 1j j r2 r3 2 3 j j r2 2 r3 3)j =

+ 1 2

2 2 ( + 1)
a11 = = ( + ) ( + ) = ( ) +(

jr2 2j  jr3 3 j a22  a33 :

12-3. Let  6 IAE 6 BAI and  6 DBI 6 IBA , then 6 EIA 6 BID   = =

(see Fig. 10). Applying the sine rule for triangle AEI and the equality As all the elements are non-negative, a22 and a33 annot both have minus sign
= = = = = = +

r jAI j  , we obtain here and, onsequently, a11 is equal to the sum or dieren e of two numbers in
the table. The proof for all other elements of the table is similar.
= sin

jAE j jAI j r
sin(  + )
=
sin 6 AEI
=
sin  sin 6 AEI
: b) The tables below satisfy the required ondition for any real x > : 0

B x x x x
;
0 0

x 0 x x x x :
 0 x 0 2x 2x 2x

D I IMO Team Sele tion Test: April 2001

r  First Day



C E A 1. Consider on the oordinate plane all re tangles whose
Figure 10 (i) verti es have integer oordinates;
(ii) edges are parallel to oordinate axes;
From triangle BDI , we similarly get (iii) area is , where k ; ; : : :.
2
k
= 0 1 2

Is it possible to olor all points with integer oordinates in two olors so that no
jBDj jBI j = r
: su h re tangle has all its verti es of the same olor?
 +  ) sin 6 IDB sin  sin 6 IDB
=
sin(
2. Point X is taken inside a regular n -gon of side length a . Let h1; h2; : : : ; h be
Sin e sin 6 AEI = os  and sin 6 IDB = os  , we have the distan es from X to the lines dened by the sides of the n -gon. Prove that
n

sin  os  sin  os  sin( +  ) 

: :::+ > :
1 1 1 1 1 1 2

jAE j jBDj + = +
r sin( +  ) r sin( +  )
=
r sin( +  )
=
r h1
+
h2
+
hn a

19 20
3. Let k be a xed real number. Find all fun tions f R ! R su h that 3. If k then f x x or f x ; if k 6 then f x k or
f (x) = 0 .
: Answer: = 1 ( ) = ( ) = 0 = 1 ( ) = 1

f (x) + f (y) 2 = kf (x + y2 )


Substituting y = 0 in the original equation we get
for all real numbers x and y . k
( 1) ( f x) = f (0)2 : (1)
If k 6 , then substituting x in (1) we get f or f k .
So the solutions in the ase k 6 are tthe onstant fun tions f x and
= 1 = 0 (0) = 0 (0) = 1
Se ond Day
fx k .
= 1 ( ) = 0

4. Consider all produ ts by 2,o 4, 6, ..., 2000 of the elements of the set
( ) = 1

If k , then from (1)
2 we get f

= 1 . Substituting x in the original(0) = 0 = 0

A
n 1 1 1
; ; ; : : :; ;
1
. Find the sum of all these produ ts.
1
equation we get f y f y2 , and furhter substituting y we nd that
or f .
= ( ) = ( ) = 1
2 3 4 2000 2001
f(1) = 1 (1) = 0

5. Find the exponent of in the representation of the number | : :{z:: : : } as For any non-negative
real number z there is a real number y su h that y2 z , =

therefore from f y 2 f y2 we get that f z > for any z > . Also,

37 111 11

3
372000 digits
produ t of prime powers.
( ) = ( ) ( ) 0 0

substituting x y in the original
2 equation we get f y2
2f y
2 , so

2 = ( ) = ( )

6. Let C1 and C2 be the in ir le and0 the ir um ir le of the triangle ABC , respe - f z 6 for any z 6 . Sin e f y f y2 f y 2 f y , then we
must have f y f y , i.e. f is an odd fun tion. p
( ) 0 0 ( ) = ( ) = (( ) ) = ( )

tively. Prove that, for any point A on C2 , there exist points B0 0 0 and0 C 0 su h that ( ) = ( )

C1 and C2 are the in ir le and the ir um ir le of triangle A B C , respe tively. Now let x be any real number and z > , then denoting z y we get 0 =

f x z f x y2 f x f y 2 f x f y2 f x f z :
( + ) = ( +


) = ( )+ ( ) (2) = ( )+ ( ) = ( )+ ( )

Solutions of Sele tion Test

Hen e if a 6 b , then f b f a b a f a f b a > f a , i.e. f is
non-de reasing.
( ) = ( + ( )) = ( ) + ( ) ( )

1. Answer: Yes. Sin e f is an odd fun tion, (2) holds also when x and z are both negative. Now
Color the points with integer oordinates in three olors so that on ea h diagonal we show, using indu tion on n , that f nx nf x for any real x and integer n .
y x k all points are of the same olor and the olors hange in a y li manner
Indeed, this holds for n and if f nx nf x then
( ) = ( )

when k in reases. Sin e  for even m and  for odd

= +
m m = 0 ( ) = ( )

m , it is easy to understand that all three olors are present in verti es of ea h

2 1 (mod 3) 2 2 (mod 3)

f ((n +1)x) = f (nx + x) = f (nx)+ f (x) = nf (x)+ f (x) = (n +1)f (x) :

re tangle under onsideration. Now re oloring the verti es of some olor with one
of the remaining two olors, we obtain a oloring with the required properties. Sin e f is odd, we also have f nx f nx nf x , i.e. f nx nf x
holds for all integers n .
( ) = ( ) = ( ) ( ) = ( )

2. Let S be the area of the n -gon and r its inradius, then S n
ar . On the other =
Earlier we proved that f or f . If f then substituting
in f nx nf x we get that f n for all integers n , and sin e f is
(1) = 1 (1) = 0 (1) = 0
2
x
hand, S
a
h1 h2 : : : h . Using the AM-HM inequality, we get
=
1
( + + + n) non-de reasing, we have f x for all real x . We show now that if f
= 1 ( ) = ( ) ( ) = 0

then f x x for all x . For integers we get it from f nx nf x , substituting

( ) = 0 (1) = 1
2
( ) = ( ) = ( )

x . For a rational number we have

n a
6 h1 h2 n : : : h
+ + + n
=
2S
na
= r: = 1
b
:::+
1 1 1

h1 h2 hn
+ +
 a a
a = f (a) = f b
= b
f ;
Comparing the lengths of ir umferen es of the n -gon and its in ir le, we get b b
na > r . Hen e a
2
so f b
a
=
b
. Assume that for some real number x we have f x 6 x , then ( ) =
1
+
h1 h2
+ :::+
1
> n > 2 :
hn r a
1
f (x) = x + " where " 6= 0 . If " > , then let r be a rational number su h
0

21 22
 that x < r < x " , and if " < , then let r be a rational number su h that and note that
x > r > x " . In the rst ase we get r < x " f x 6 f r r , in the se ond
+ 0

37k+1 37k
37

ase r > x " f x > f r r , a ontradi tion.
+ + = ( ) ( ) =

+ = ( ) ( ) = 1000 1 = 1000 1 =
   
37k
37k + 37k
2 + : : : + 37k
36 :
4. Answer: 499 . 1001
= 1000 1 1 + 1000 1000 1000

The value of The exponent of in the representation of number 37 is k by the

2001
k

indu tion hypothesis. Hen e it su es to show that the exponent of in the
37 1000 1 + 1

     
representation of
37

1+
1

2

1+
1

3

:::
1+
2001
1
1

37k + 37k
2 + : : : + 37k
36

is equal to the sum of all produ ts of the elements of set A by 1, 2, 3, ..., 2000, 1 + 1000 1000 1000

and the value of is . Sin e  then 37k  1 (mod 37) . Let 37k = 37q + 1 ,
then
1 1000 1 (mod 37) 1000 1000
     
1
1

1
1

:::
1
1
1
2 3 2001
37k + 37k
2 + : : : + 37k
36=
is equal to a similar sum where the produ ts by 2, 4, 6, ..., 2000 are taken with
1 + 1000 1000 1000

q + 1) + (37q + 1)2 : : : + (37q + 1)36 

a plus sign but the produ ts by 1, 3, 5, ..., 1999 are taken with a minus sign. = 1 + (37

Denote the required sum by S , then  1 + (37q + 1) + (2
37q + 1) + : : : + (36
37q + 1) =

      =
37 36

q
37 + 37 = 37
2
18
q + 37  37 (mod 372) :
2 S = 1+
1

2

1+
1

3

:::
1+
2001
1
+
2

 1
 

1


:::

 1

So 1 + 1000
37 k
+
37
2 : : :
1000
37
36 is divisible by
k
+ + 1000
k
but not by 2 , 37 37
+ 1
2
1
3
1
2001
2 =
and the exponent of in the representation of 37 2000 is k .
37
+1
1000
k
1 +2

=
3


:::

4 2002
+
1


:::

2 2000
2 = Hen e the exponent of in the representation of 37
37 is . 1000 1 2001

6. Let I and O be the in enter and the ir um enter of the triangle ABC , respe -
2 3 2001 2 3 2001

; tively. We know by Euler's formula that jOI j2 R2 Rr , where r and R are

2002 1 1
= + 2 = 999

the radii of the in ir le and the ir um ir le, respe tively.

2 2001 2001 = 2

and S 1001
. Assume now that there exists a point A on the ir le C2 su h that it is impossible
0
= 499
2001 to onstru t the0 points B0 and C 0 as required 0in the problem. Let the tangents
5. . drawn from A to the ir le C1 tou h C1 0 in0 B and C 0 , hen e B0C 0 is not tan-
Answer: 2001

As and are relatively prime it is su ient to nd the exponent of in the gent to the ir le C1 . Suppose the line B C and the ir le C1 have no points in
representation of the number
37 9 37
ommon0 0(the ase of two ommon points is similar). Let the distan e between the
line B C and the ir le C1 be Æ0 > . 0 0

3
372000 372000 Now start moving the points B and C along the ir le C2 towards0 0 A0 in su h
|999 : :{z
:: : : 99} = 10 1 = 1000 1 :
a way that the distan es from the ir le C1 to the straight lines A B and A0C 0
3
372000 numbers remain equal (note that they are both equal to at the beginning)  denote
this distan e by  . The distan e Æ obviously de reases, whereas the distan e 
0

We show by indu tion on k that the exponent of 37 in the representation of in reases, hen e at some moment they must be ome equal. Now we0 an in rease
37
1000 is k . In the ase k we have
k
1 +1 = 0
the radius r by0 0Æ 0  > to make it the in ir le of the triangle A B0C 0 . Hen e
the triangle A B C has ir umradius R and inradius r Æ , but the distan e jOI j
= 0
370 3
1 = 999 = 3
37 ;
is the same as for the triangle ABC , hen e Euler's formula for triangle A0B0 C 0
1000 +

i.e. the exponent of is . Suppose now that for some k our assertion holds,
37 1 be omes violated.
23 24
Estonian math ompetitions 2001/2002 3. For any positive integer n , denote by S (n) the sum of its positive divisors (in-
luding 1 n ).
and

a) Prove that S (6n) 6 12S (n) for any n .

For whi h n does the equality S (6n) = 12S (n) hold?
We thank the IMO ommunity for many of these problems whi h have been taken
b)
from various materials distributed at the re ent IMO-s.

4. In a triangle ABC we have 6 B = 2
6 C and the angle bise tor drawn from A
interse ts BC in a point D su h that jAB j = jCDj . Find 6 A .
Autumn Open Contest: O tober 2001
5. Let b1 ; b2 ; : : :; b n be a rearrangement of positive real numbers a1 ; a2 ; : : :; a n.

Juniors (up to 10th grade) Prove that

a)

a1 +
1 
a + 1 
: : :
a + 1  > 2 n
;
b 1 b 2
2 b n

1
n
1. A gure onsisting of ve equal-sized squares is pla ed as 7
shown in a re tangle of size 78 units. Find the side length
b) if equality holds here for an odd n then at least one of the numbers a i is .

of the squares.

2. Find the remainder modulo 13 of the sum

8 Solutions of Autumn Open Contest
1 2001
+22001
+ 3 + : : : + 2000
2001 2001
+ 20012001
: p
J1. Answer: 5 .

3. Find all triples (x; y; z) of real numbers satisfying the system of equations (where Let a be the required side length, then the proje tions of ea h side of any square
[r℄ frg denote the integer and fra tional part of r , respe tively):
and to the sides of the re tangle are x and y where x2 + y2 = a2 . We have

8
< x + [y℄ + fz g = 200;2
> 8 = 2x + y + x + y = 3x + 2y
>
fxg + y + [z ℄ = 200;1 : and
:
[x℄ + fyg + z = 200;0
7 = 3x + y ;
4. Consider a point M inside triangle ABC su h that triangles ABM , BCM and p p
CAM have equal areas. Prove that M is the interse tion point of the medians of y=1 x=2
yielding , and a= x2 + y2 = 5 .
triangle ABC .
J2. Answer: 0 .
5. For any integer n> 1 onsider all squares with verti es in points having non- Arrange all terms of the sum ex ept 1001 2001
(whi h is divisible by 13 ) in pairs
negative integer oordinates not greater than n. (k ; (2002 k) ) . Sin e 2002 is divisible by 13 , k2001 + (2002 k)2001 is
2001 2001

a) How many su h squares are there for n = 4? ongruent to k

2001
+( k)2001 = 0 modulo 13 , i.e. the sum of ea h pair is divisible
b) Find a general formula for the number R of su h squares for any n .
n by 13 and hen e the required remainder is 0 .

J3. Answer: the only su h triple is x = 100;15 ; y = 100;95 ; z = 99;05 .

Seniors (grades 11 and 12) Adding all three equations and using [r℄+frg = r we have x+y +z = 300;15 . Sub-
tra ting from here the rst given equation, we obtain (y [y℄)+(z fzg) = 99;95
1. The sum of two distin t positive integers, obtainable from ea h other by rear- or fy g + [z ℄ = 99;95 , yielding fy g = 0;95 and [z ℄ = 99 . Similarly we get
rangement of digits, onsists of 2001 equal digits. Find all possible values of the [x℄+ fzg = 100;05 and fxg +[y℄ = 100;15 , i.e. [x℄ = 100 , fzg = 0;05 , fxg = 0;15
digits of the sum. and [y ℄ = 100 .

2. The side lengths of a triangle and the diameter of its in ir le, taken in some order, J4. It su es to prove that if triangles ABM and BCM have equal areas then M
form an arithmeti progression. Prove that the triangle is right-angled. lies on the median drawn from B. K and H be the perpendi ular proje tions
Let

1 2
 of A and C to BM , and let BM interse t AC in a point P (see Figure 1). Then Note. Using the identities R R 1 = n2 + (n
n n 1) + : : : + 1
2 2
and

jBM j
jAK j = S n(n + 1)(2n + 1)

2 ABM =S BC M = jBM j2
jCH j 1 +2 +:::+ n =
2 2 2
6
and hen e jAK j = jCH j . If AC is perpendi ular to BM , then K = H = P , we an show by indu tion that

jAP j = jP C j and BP is a median. If AC is not perpendi ular to BM , then (n + 1)
(n + 1) 1

2 2
R = :
learly one of K and H ABC and the other one outside of it.
lies inside triangle n
12
Hen e 6 AKP = 90Æ = 6
6 AP K = 6 CP H , i.e. triangles AKP and
CHP and
CHP are ongruent, whi h again yields jAP j = jP C j and BP being a median. S1. Answer: 1 2 3 4 5 6 7 8
, , , , , , , .

LetA 0 1 2 3
be any of the digits , , , . Taking

B n = A (A+5) : : : (A+5)(A+5)A(A+6); m = A (A+5) : : : (A+5)(A+6)A(A+5);

| {z } | {z }
1997 digits 1997 digits
H r r r r
M
r r r r r r r or

r r r r r r r r r
n =(A+1) : : : (A+1) A(A+2); m =(A+1) : : : (A+1)(A+2)A;
A P C r r r r r r r r r | {z } | {z }

K n=1 n=2 n=3
1999 digits 1999 digits

we have n and m obtainable from ea h other by rearrangement of digits and

Figure 1 Figure 2
n + m =(2A+1) : : : (2A+1); n + m =(2A+2) : : : (2A+2) ;
| {z } | {z }
J5. Answer: a) 50 ; b) 1
n + 2
(n 1) + : : : + (n 1)
2 + n
1
2 2 2 2
. 2001 digits 2001 digits

We rst study the possible squares for n = 1; 2; 3 .

18
n = 1 we have a single square of side length . 1
respe tively. Hen e to are possible digits.
For
m + n = 99 : : : 9}
For n = 2 we have 2
2
possible lo ations for the square of side length 1
Suppose now that . Moving from right to left it is easy to see
| {z
p and new
2001 digits
possible squares of side lengths 2 2 and , one of ea h. that there an be no arries during the addition. Hen e any digit A o urs in
For n = 3 , we have 3
3 p
possible lo ations for the square of side length 1 2
2
, n in these and only these positions where the digit 9 A o urs in m . Sin e n
possible lo ations for ea h of the squares of side lengths 2 2 and and three new and m are obtainable from ea h other by rearrangement of digits then any digit
types of squares, one of ea h (see Figure 2). A o urs in n the same number of times as 9 A 6= A , and the number of digits
in n, m and also n+m must be even  a ontradi tion.
We see that for ea h n we have 1=1 2
possible lo ation for ea h of the new
S2. Drawing a line through the in enter of a triangle parallel to ea h of its sides it is
squares (i.e. squares having all their verti es at the edges of the grid) and for the
next values of n we have 2 ; 3 ; 4 ; :::
2 2 2
possible lo ations for these squares. It
easy to see that the diameter of the in ir le is shorter than any of the sides. Let
x d>0
n and be the diameter of the in ir le and the dieren e of the arithmeti
x + d , x + 2d and x +3d .
remains to noti e that the number of the new squares is sin e we an pla e
one of its verti es either in a orner of the grid or in one of the n 1 points on
progression, then the side lengths are
S of the triangle in two ways we get
Finding the area

the side of the grid, thereby determining the lo ations of the other three verti es.
q
n x
Hen e for any we have
p

2 =S= p
p (x + d)

p (x + 2d)

p (x + 3d)
:
R n = 1
n + 2
(n 1) + 3
(n 2) + : : : + (n 1)
2 + n
1
2 2 2 2 2
; s

R = 50 p=
3(x + 2d) 3(x + 2d)x = 3(x + 2d)(x + 4d)(x + 2d)x
yielding 4 . Sin e
2 4 we have
16 , yield-

ing 3x = x + 4d x = 2d and x + d = 3d x + 2d = 4d
. Hen e the side lengths are ,

3 4
 and x + 3d = 5d , i.e. the triangle is right-angled. Sin e i1 b =a i2 and b2
i =a i1 , we an omit a1
i and a2
i and use the indu tion

S3. Answer: b) for n not divisible by 2 or 3 .

hypothesis.
1 1 1
n 6n If i3 6= i1 , we nd a 4 = b 3 = = , a = b 4 = a = a 1 et . Sooner or
Considering the representations of and as produ ts of primes we see that
a3 a1 5
i i i i i

any positive divisor of 6n is the produ t of a positive divisor of 6 and a positive later we must have i +1 = i1 for some even k (sin e a 1 6= 1 ). Similarly to the
i 4 i i

divisor of n. Hen e the positive divisors of 6n are numbers of the form d , 2d , 3d

k

previous ase we an now omit a 1 ; a 2 ; : : :; a k and use the indu tion hypothesis.
i

and 6d dwhere n
is a positive divisor of , and
i i i

S (6n) 6 S (n) + 2S (n) + 3S (n) + 6S (n) = 12S (n) :

Spring Open Contest: February 2002
Here equality holds if and only if the abovementioned four series of divisors do
not interse t, i.e. no divisor d of n an be represented as 2d0 or 3d0 where d0 is
n , or equivalently n is not divisible by 2 or 3 .
Juniors (up to 10th grade)
another divisor of

S4. Answer: 72Æ . 1. Is it possible to arrange the integers 1 to 16

Let jAB j = jCDj = a 6 C =  , and 6 A = 2
BAD =  ,
, then 6 CAD =6 a) on a straight line;
6 B = 2 6 BDA =  +  and . Applying the sine rule in triangles ACD and ABD b) on a ir le
we have
so that the sum of any two adja ent numbers is the square of an integer?
sin  = jADj = sin 2 ;
sin  a sin( + ) 2. Does there exist a re tangle with integer side lengths with the square of its diagonal
equal to 2002 ?
2 sin  os  =Æ sin( + ) tan  = tan  0Æ < ;  < 90Æ
ABC we have jAB j = jAC j and 6 BAC =  . Let P 6= B be a point
yielding , or . Sin e we have
= 180 = 2 +2 +  = 5
. Now from  = 36 6 A = 2 = 72Æ we have and . 3. In a triangle
on AB and Q a point on the altitude drawn from A su h that jP Qj = jQC j .
S5. a) Using the AM-GM inequality we have: Find 6 QP C .
 1 
a +
:::
a +
 1  r r
a
r
> 2
b
: : :
2
b = 2
ab

:: :: ::

ab = 2 :
a 1 n n 1 n n 4. Dene a1 ; a2 ; : : : ; a ; : : : as follows:
n
b1
1 b n
1 1
a1 = 0; a2 = 1; a = 5a n > 2.
n n n

a ;
n=1
1
a + =2 a =1
n n 1 n 2 for

b) If
a
then 1 and hen e 1 . Suppose now that the laim is true
For whi h n is a divisible by: a) 5 ; b) 15 ?
1 ra
1 n

for any odd integers less than n a + =2

. The equality holds if and only if
i
n n
b b i 5. For whi h positive integers is it possible to write real numbers, not all equal

1
i i
0
a = a =b a =b =1
to , on a ir le so that ea h of these numbers is equal to the absolute value of
for ea h i , i.e. i . If for some then and we are done.
b i i i i i the dieren e of its two neighbouring numbers?

a 6= b a 6= 1
i

If i for all i then onsider some b 1 su h that . Then equals to

1
i i i1 i1

some a i 6= i
where a =b = 1 , andb a . Also, equals to some where Seniors (11th and 12th grade)
i2 2
a i2 i1 i2 i3

1 =a
i1

i3 6= i a =b =
2 . Hen e
a i3 i2 i1 .
1. The sides a , b and of a right triangle form a geometri progression, and ab = 1 .
i2

If i =i
3 1 then Find a, b, .

a +
1 
a + 1  = a + 1 
a + 1  = a + 1 
a + 1  = 2. Let a; b be any real numbers su h that jaj 6= jbj . Prove that
b
i1
bi1 a
i2
i2 a a i1
a
i2
i2
i3
i1
i2
i2
i1

 1 1 
= (a + a )
+ = 2a
= 2 : 2 2
a
 + b  > 1 :
ab

i1 i1
a1 a1 i1
a1 a b
i i i

5 6
 3. Let ABCD be a rhombus with 6 DAB = 60Æ . Let K , L be points on its sides AD Sin e jAB j = jAC j then the altitude drawn from A is also an angle bise tor.
and DC and M a point on the diagonal AC su h that KDLM is a parallelogram. Note that jQB j = jQC j = jP Qj (see Figure 3), i.e. the triangles BQC , BQP
Prove that triangle BKL is equilateral. and P QC are isos eles. Denote 6 QBC = 6 QCB =  , 6 QBP = 6 QP B =
and 6 QP C = 6 QCP = Æ , then 6 QCA = . From triangle ABC we now have
4. Call a 10 -digit natural number magi if it onsists of 10 distin t digits and is  + 2 + 2 = 180Æ and from triangle P BC we have 2Æ +2 +2 = 180Æ . Hen e
divisible by 99999 . How many su h magi numbers are there (not starting with
6 QP C = Æ =  .
digit 0 )? 2
5. Find the maximum number of distin t four-digit positive integers onsisting only J4. Answer: a) for any odd n ; b) for n = 6k + 1 .
of digits 1 2
, and 3 su h that any two of these numbers have equal digits in at a) From the equality a = 5a 1 a 2 we see that a is divisible by 5 if and

2 is divisible by 5 . Sin e a1 = 0 is divisible by 5 but a2 = 1

n n n n
most one position? only if a is not
divisible by 5 , then a is divisible by 5 if and only if n is odd.
n

b) Taking n + 1 instead of n in the given equality we have

Solutions of Spring Open Contest
an+1 = 5a n an 1 = 5
(5a n 1 a n 2 ) a n 1 = 24a n 1 5an 2 :
J1. Answer: a) yes; b) no.
From here we see that a +1 is divisible by 3 if and only if a 2 is divisible by 3 .
4 = 16 < 16 + 1 6 = 36 > 16 + 15 9 16 a1 = 0 is divisible by 3 but a2 = 1 and a3 = 5
1 0 = 5 are not divisible
n n
2 2
Sin e and then only an be adja ent to Sin e
(giving 16 + 9 = 25 = 5 2
). Hen e it is impossible to arrange the numbers 1 to 16 by 3 , then a is divisible by 3 if and only if n = 3m + 1 for some m . Hen e a
is divisible by 15 if and only if n is both odd and of the form n = 3m + 1 , i.e.
n n
on a ir le in the required manner. A suitable arrangement on a straight line is:

n = 6k + 1 .
16; 9; 7; 2; 14; 11; 5; 4; 12; 13; 3; 6; 10; 15; 1; 8 :
J5. Answer: if and only if n is divisible by 3 .

J2. Answer: no. Sin e ea h number written on the ir le is equal to the absolute value of the dif-

We need to determine whether there exist positive integers a and b su h that feren e of two others, then all these numbers are non-negative. Leta be maximal
a 2
+ b = 2002
2
2002 7 among these numbers (then a > 0 ) and let b and be the numbers adja ent to
it, with b > > 0 . Sin e b 6 a then also b 6 a , and the equality a = b
. Note that is divisible by and the square of any integer
0 1 2 4 7 a b
holds only if b = a and = 0 . Hen e any number a on the ir le must have a
is ongruent to , , or modulo . Hen e and must both be divisible
7 a +b 2 2
49 2002 49
0 0
by , but then is divisible by , and is not divisible by .
and as its neighbours, and any number must have its both neighbours equal.
We see that the numbers on the ir le must be a , a , 0 , a , a , 0 , . . . (see Figure 4)
A and hen e n must be a multiple of 3 .

P 
22 ar
r

Æ
ar
Q 0r r
Æ r r

 a a
r
0
B C Figure 4

On the other hand it is easy to he k that for any n = 3k and a > 0 the numbers
Figure 3

 a; a; 0; a; a;{z0; : : :; a; a; 0} satisfy the required onditions.

6 QP C = |
J3. Answer:
2 .
k triples a; a; 0

7 8
 s p s p 3456
5 1 1 5+1 S4. Answer: .
S1. Answer:
2 , and
2 .
We an write any ten-digit number ab defghij as

b = 100000
ab de + fghij =
Assume w.l.o.g. that a < b < , then a = and = bq for some q > 1 . Hen e ab defghij
q
from ab = 1 we have b = 1 and b = 1 . From the Pythagorean Theorem we
3 = 99999
ab de + ab de + fghij :
 1 2
now have
q
+ 1 = q2 , or q4 q2 1 = 0 . Sin e the equation x2 x 1 has Hen e ab defghij 99999
is divisible by if and only if the sum ab de + fghij is

p s s p p divisible by 99999 . Sin e ea h summand here is positive and less than 99999 , we
5+1 q=
5+1 1= 5 1 must have ab de + fghij = 99999 , or equivalently
2 as its only positive solution, then
2 and
q 2 .

a+f = b+g = +h= d+i = e+j = 9

S2. For any real y and x>0 we have:

a) x > 1 , if x > 1
y
and y>0 or x<1 and y < 0; (sin e the sum ontains only digits 9 , no arries an o ur on addition). We see

x < 1 , if x > 1
y
y<0 x<1 y > 0; that magi numbers are in one-to-one orresponden e with numbers of the form
a 6= 0
b) and or and

x = 1 , if x = 1 y = 0. ab de where a; b; ; d; e are ve distin t digits su h that and the sum of

9 9
8
6
4
2 = 3456
y
) or
no two of them is . There are su h numbers ab de .
We shall now onsider the ases where ab is positive, negative or equal to zero.

If ab > 0 , then a and b have the same sign and ja + bj > ja bj > 0 , hen e S5. Answer: 9 .

a
 + b  > 1 and  a + b  > 1 .ab
Note that we annot have more than 3
3 = 9 integers with the required property
a b a b sin e the pairs of rst two digits of any two of them must be distin t. A suitable

If ab < 0 , then a and b have opposite signs and ja bj > ja + bj > 0 , hen e
set of 9 integers is 1111 1222 1333 2123 2231 2312 3132 3213 3321
, , , , , , , , .
   
a + b a + b ab

0 <  a b  < 1 and  a b  > 1 .

    Final Round of National Olympiad: Mar h 2002
a + b ab
a + b
If ab = 0 , then   = 1 sin e  a b  6= 0 .

a b 9th grade
S3. The rhombus ABCD onsists of two equilateral triangles ABD and BCD . We
shall prove that jKD j = jLC j (see Figure 5)  then triangles KBD and LBC are
K and L are taken on the sides BC CD ABCD
ongruent and jKB j = jLB j , 6 KBD = 6 LBC . Hen e 6 KBL = 6 DBC = 60 ,
Æ 1. Points
6 AKB = 6 AKL . Find 6 KAL .
and of a square so that

i.e. the triangle BKL is equilateral.

2. Do there exist distin t non-zero digits a, b and su h that the two-digit number

D L C ab is divisible by , the number b is divisible by a and a is divisible by b?

K 3. Let a1 ; a2 ; : : : ; a be pairwise distin t real numbers and m be the number of

a + a (where i 6= j ). Find the least possible value of m .
n
M distin t sums i j

4. Mary writes 5 numbers on the bla kboard. On ea h step John repla es one of the
A B numbers on the bla kboard by the number x + y z , where x , y and z are three
Figure 5 of the four other numbers on the bla kboard. Can John make all ve numbers on
the bla kboard equal, regardless of the numbers initially written by Mary?
To prove the equality jKDj = jLC j note that LM is parallel to AD and
6 LMC = 6 DAC = 6 DCA =6 LCM . Hen e the triangle MLC is isos eles, 5. There were n> 1 aborigines living on an island, ea h of them telling only the
i.e. jLC j = jLM j = jKD j . truth or only lying, and ea h having at least one friend among the others. The

9 10
 new governor asked ea h aborigine whether there are more truthful aborigines or 3. The tea her writes a 2002 -digit number onsisting only of digits 9 on the bla k-
liars among his friends, or an equal number of both. Ea h aborigine answered that board. The rst student fa tors this number as ab with a>1 and b>1 and
there are more liars than truthful aborigines among his friends. The governor then repla es it on the bla kboard by two numbers a0 and b0 su h that ja a0j = 2 and
ordered one of the aborigines to be exe uted for being a liar and asked ea h of jb b0 j = 2 . The se ond student hooses one of the numbers on the bla kboard,
the remaining n 1 aborigines the same question again. This time ea h aborigine fa tors it as d with >1 and d>1 and repla es the hosen number by two
answered that there are more truthful aborigines than liars among his friends. numbers 0 and d0 su h that j 0j = 2 and jd d0j = 2 . The third student again
Determine whether the exe uted aborigine was truthful or a liar, and whether hooses one of the numbers on the bla kboard and repla es it by two numbers

there are more truthful aborigines or liars remaining on the island. following a similar pro edure, et . Is it possible that after a ertain number of
students have been to the bla kboard all numbers written there are equal to 9
?

10th grade
4. Let a1; a2; a3; a4 ; a5 be real numbers su h that at least N of the sums a , i +a
j

where i < j , are integers. Find the greatest value of N for whi h it is possible
that not all of the sums a + a are integers.
i j

1. The greatest ommon divisor d and the least ommon multiple v of positive
integers m and n satisfy the equality 3m + n = 3v + d . Prove that m is divisible 5. John built a robot that moves along the border of a regular o tagon, passing ea h
by n. side of the o tagon in exa tly 1 minute. The robot begins its movement in some
vertex A of the o tagon, and further on rea hing ea h vertex an either ontinue
2. Let ABC be a non-right triangle with its altitudes interse ting in point H . Prove movement in the same dire tion, or turn around and ontinue in the opposite
that ABH is an a ute triangle if and only if 6 ACB is obtuse. dire tion. In how many dierent ways an the robot move so that after n minutes
it will be in the vertex B opposite to A?
3. John takes seven positive integers a1 ; a2 ; : : : ; a7 and writes the numbers aa ,
+a i 6= j
i j

a i j and ja i aj
j for all on the bla kboard. Find the greatest possible
number of distin t odd integers on the bla kboard.
12th grade
4. Find the maximum length of a broken line on the surfa e of a unit ube, su h that
its links are the ube's edges and diagonals of fa es, the line does not interse t itself 1. Peter, John, Kate and Mary are standing at the entran e of a dark tunnel. They
and passes no more than on e through any vertex of the ube, and its endpoints have one tor h and none of them dares to be in the tunnel without it, but the
are in two opposite verti es of the ube. tunnel is so narrow that at most two people an move together. It takes 1 minute
for Peter, 2 minutes for John, 5 minutes for Kate and 10 minutes for Mary to
5. The tea her writes numbers 1 at both ends of the bla kboard. The rst stu- pass the tunnel. Find the minimum time in whi h they an all get through the
dent adds a 2 in the middle between them; ea h next student adds the sum tunnel.
of ea h two adja ent numbers already on the bla kboard between them (hen e
there are numbers 1; 3; 2; 3; 1 on the bla kboard after the se ond student; 2. Does there exist an integer ontaining only digits 2 and 0 , whi h is a k -th power
1; 4; 3; 5; 2; 5; 3; 4; 1 after the third student et .) Find the sum of all numbers of a positive integer with k > 2?
on the bla kboard after the n -th student.
3. Prove that for positive real numbers a, b and the inequality

11th grade
2(a + b + ) < (a + b + )
4 4 4 2 2 2 2

holds if and only if there exists a triangle with side lengths a, b and .
1. Determine all real numbers a su h that the equation x8 + ax4 + 1 = 0 has four
real roots forming an arithmeti progression. 4. All verti es of a onvex quadrilateralABCD lie on a ir le ! . The rays AD ,
BC interse t in point K AB , DC interse t in point L . Prove that
and the rays
2. Inside an equilateral triangle there is a point su h that the distan es from it to the ir um ir le of triangle AKL is tangent to ! if and only if the ir um ir le
the sides of the triangle are 3 4 , and 5 . Find the area of the triangle. of triangle CKL is tangent to ! .

11 12
 5. There is a lottery at John's birthday party with a ertain number of identi al i.e. there are at least 2n 3 distin t sums. Taking a =i we have 1+2 = 3
(n 1) + n = 2n 1
i

prizes, whereas ea h of the guests an win at most one prize. It is known that as the minimal sum and as the maximal sum, so there are
if there was one prize less than there a tually is, then the number of possible exa tly 2n 3 distin t sums in this ase.
distributions of the prizes among the guests would be 50% less than it a tually
9-4. Answer: yes.
is, while if there was one prize more than there a tually is, then the number of
50% Denote the numbers written by Mary by a , b , , d and e (not ne essarily distin t).
a and b by x = + d e . Then he an repla e ea h
possible distributions of the prizes would be more than it a tually is. Find
First John an repla e ea h of
of and d by e + x x = e and nally repla e both numbers x by e + e e = e :
the number of possible distributions of the prizes.

Solutions of Final Round (a; b; ; d; e) ! (x; x; ; d; e) ! (x; x; e; e; e) ! (e; e; e; e; e) :

9-5. Answer: the exe uted aborigine was truthful and after the exe ution only liars
9-1. Answer: 45Æ . remained on the island.

Let AM be the perpendi ular drawn from A to KL (see Figure 6). Sin e ABK First note that there was a truthful aborigine on the island before the exe ution,
and AMK are ongruent right triangles then jAM j = jAB j = jADj , i.e. AML sin e otherwise all friends of ea h aborigine would have been liars, and hen e their
and ADL are also ongruent right triangles, and answers would have been true  a ontradi tion.

6 KAL = 6 KAM + 6 LAM = 6 KAB + 6 LAD ; Suppose now there was a truthful aborigine on the island after the exe ution.
Then both his answers must have been true  but this is impossible sin e the
when e exe ution of one aborigine ould not hange the dieren e of the numbers of liars

26 KAL = 6 KAM + 6 LAM + 6 KAB + 6 LAD = 90Æ

and truthful aborigines among his friends from positive to negative.

m = dm0 and n = dn0 where g d (m0 ; n0) = 1 . Then v = m0 n0d and we have
6 KAL = 45Æ .
10-1. Let
and 3m d + n0d0 = 3m0n0d + d , yielding
0 3m0 + n0 = 3m0n0 +1 and (3m0 1)(n0 1) = 0 .
Sin e 3m 1 6= 0 , then n 1 = 0 and hen e n = d divides m .
0
D L C 10-2. If H is the ortho enter of triangle ABC , then C is the ortho enter of triangle

M ABH . We'll onsider three possible ases.
(1) If ABC is an a ute triangle (see Figure 7), then H lies inside triangle ABC
K
and C lies outside triangle ABH , when e ABH is an obtuse triangle.

(2) If ABC is an obtuse triangle with 6 ACB a ute, then assume w.l.o.g. that
A B 6 B is obtuse (see Figure 8). Then C and H lie on opposite sides of AB . Hen e
Figure 6 C is outside triangle ABH , and ABH is an obtuse triangle.
(3) If 6 ACB is obtuse (see Figure 9), then C lies inside triangle ABH , when e
9-2. Answer: no.
ABH is an a ute triangle.
Note that if a, b and satisfy the required onditions and one of them is even,
a b
then all three are even. Then
2 2
, and
2 also satisfy the required onditions.
C H H
Hen e we an assume w.l.o.g. that a, b and are all odd. Also note that none







of these numbers an be 5 , sin e then the other two should also be 5 . Hen e it
H B C
su es to onsider 1 3 7
, , and 9 and one of a , b and must be 3 or 9  let
this be a. Then b is divisible by 3 , whi h gives fb; g = f3; 9g , a ontradi tion.



9-3. Answer: 2n 3 . A B A C A B
We an assume w.l.o.g. that a1 < a2 < : : : < a . Then n Figure 7 Figure 8 Figure 9

a1 + a2 < a1 + a3 < : : : < a1 + a < a2 + a < : : : < a

n n n 1 +a n ; 10-3. Answer: 30 .

13 14
 If there are m odd integers among a i, then the maximum number of odd integers a=
82
written on the bla kboard is
11-1. Answer:
9 .

t=x 4
t , and the equation x4 = t0
m(m 1 1) + 2
m(7 Substituting we get a quadrati equation in

f (m) = m) = (m m + 28m 4m ) = 2 2
2 2 
has at most two roots whi h in this ase have equal absolute values. Hen e the

 four roots of the equation x8 + ax4 + 1 = 0 have to be of the form x0 and
3 3 81
= 2 m(9 m) = 2
4 2 m
9  2 x1 . Assuming w.l.o.g. that x1 > x0 we see that these roots form an arithmeti
progression if and only if x1 = 3x0 . Sin e the roots of the equation t + at +1 are
2

then x0 and 81x0 , we have 81x0 = 1 , yielding x0 =

4 4 8 4 1 and a = 82x4 = 82 .
f (m) f (4) = f (5) = 30
and the maximum value of is .
0
9 9
a1 = 2 , a2 = 4 , p
It remains to he k that these numbers an all be distin t, e.g. for
a3 = 6 a4 = 25 = 52 , a5 = 125 = 53 , a6 = 15625 = 56 , a7 = 9765625 = 510 . 36 + 25 3
and 11-2. Answer:
4 .

Consider a triangle ABC with a point jP Aj = 3 jP B j = 4

P inside su h that ,
and jP C j = 5 . 60
Rotating the triangle by
Æ C A B aroundB , we map to and
s
to some B0 (see Figure 12). P P0 Then jP 0B j = jP Aj = 3
maps to , where ,
s
jP 0B 0 j = jP B j = 4 jP 0C j = jP C j = 5
and CP P 0 . Sin e is an equilateral
triangle then jP P j = 5
0 jP B j + jP B j = 4 + 3 = 5 = jP P 0j
.
0Hen e
2 2 2 2 2 2
,

s and 6 P BP = 90
0 Æ AP B
. From triangles BP 0 B 0 and we have

s 6 ABP + 6 BAP = 6 ABP + 6 B 0 BP 0 = 120Æ 90Æ = 30Æ

Figure 10 Figure 11 and 6 AP B = 180Æ 30Æ = 150Æ AP B
. The osine law in now gives
p
p jAB j = jAP j + jBP j 2
jAP j
jBP j
os 6 AP B = 25 + 12 3
2 2 2

10-4. Answer: 3+4 2 .

p p
p
The links of su h a broken line are the edges of the ube (of length 1 ) and the
S=
3
j AB j =
36 + 25 3 2

diagonals of its fa es (of length 2 ). Sin e the line passes ea h vertex at most
and the area is
4 4 .

on e, it an have at most 7 links. Coloring the verti es as shown on Figure 10 we

see that opposite verti es are of dierent olour and the endpoints of the diagonal B0 C
of ea h fa e are of the same olour  hen e an odd number of links have to be
4 5
P0
edges of the ube. Also, it is lear from this olouring that no more than three
onse utive links an be diagonals (sin e the ube has only 4 verti es of ea h
5
olour). It is now easy to he k that a broken line with
p 1 edge and 6 diagonals is

impossible, hen e its length annot ex eed 3+4 2 ; a suitable broken line of this 3 4 3
length is shown on Figure 11.
P
10-5. Answer: 3 +1
n
.
B A
Let S be the sum of the numbers on the bla kboard after the n -th student. We Figure 12
= 3 +1 . Indeed, S0 = 2 = 3 + 1 and ea h
n

shall prove by indu tion that Sn

n 0

k -th student is ounted in exa tly two 11-3. Answer: no.

2002 999 : : : 9 3 4 = ab
number that is on the bla kboard after the
(k+1) 1 The initial -digit number is ongruent to modulo . If N
3 4 3
of the sums written by the -th student, with the ex eption of the two -s
written by the tea her whi h are ounted only on e  hen e and N is ongruent to modulo then one of a and b is ongruent to and
the other is ongruent to 1 4
modulo and the same is true for a0 and b0 . Hen e

S = S + 2S 2 = 3(3 + 1) 2 = 3
k k +1
+1: at all times there is a number on the bla kboard whi h is ongruent to 3 modulo
4 9 1 4
k +1 k k
, while is ongruent to modulo .

15 16
11-4. Answer: 6 . Note that it is possible to get everyone through the tunnel in 17 minutes:

If there are four integers and one non-integer among a i then N =6 . To prove 1) Peter and John go through the tunnel (2 minutes);
the maximality we denote the fra tional part of x by fxg and note that: 2) Peter brings ba k the tor h (1 minute);

(a) if fag =6 fbg and is any real number then at most one of + a and + b 3) Kate and Mary go through the tunnel (10 minutes);
4) John brings ba k the tor h (2 minutes);
is an integer;

(b) if a = b then a + b is an integer if and only if fag = 0 or fag = 0;5 ; 5) Peter and John go through the tunnel (2 minutes).

( ) if fag 6= fbg and a + b is an integer then neither fag nor fbg is 0 or 0;5 .
It remains to show the minimality of this total. Clearly they have to go through
the tunnel an odd number of times and bring ba k the tor h at least twi e, hen e
Considering now the possible partitions of the set fa1 ; a2 ; a3 ; a4 ; a5 g to subsets they have to go through the tunnel at least 3 times in one dire tion and 2 times
(of elements with equal fra tional parts) and keeping in mind the above remarks 7
6
in another dire tion. If they go through the tunnel or more times then the total
(a), (b) and ( ) we see that the only ase when there an be more than integer 10+2+5
1 = 17
+a fa1 g = fa2g = fa3g = fa4 g = fa5g , and in this ase all
time annot be less than minutes. If they go through the tunnel
sums a i j is when 5 times then ea h pass in the initial dire tion takes at least 2 minutes and one
these sums are integers. of these (with Mary) takes 10 minutes. If Peter brings ba k the tor h both times

11-5. Answer: 2 (2k 1 k 1

1) for n = 2k , and 0 for n = 2k + 1 .
then Kate and Mary annot go through the tunnel together and they need at least
10 + 5 + 2 + 2
1 = 17 minutes. If someone else brings ba k the tor h at one time
Colour the verti es of the o tagon alternately bla k and white. Sin e A and B then they need at least 10 + 2 + 2 + 2 + 1 = 17 minutes.
are of the same olour and in ea h minute the robot moves from a vertex of one
olour to a vertex of the opposite olour, then n must be even. 12-2. Answer: no.

We now label the verti es by 1 to 8 so that A = 1 and B = 5 , and denote by Consider an integer N ontaining only digits 2 and 0 and ending in t zeroes

a(k )
= (a ; a ; : : :; a )
(k ) (k ) (k )
the numbers of possibilities, for the robot to rea h ( t > 0 ), then
1; 2; : : : ; 8 A=1 k
1 2 8
verti es from in minutes. Using indu tion by m we shall
N = 2 : : : 2
10 = 1 : : : 1
2
5 ;
m>1
t t+1 t
prove that for any

a = (2 + 2 ; 0; 2 ; 0; 2 2 ; 0; 22 ; 0) :
where the dotted part in 2 : : :2 an ontain both 2 -s and 0 -s (in 1 : : :1 both 1 -s
0 1 : : :1 2 5
(2m) 2m 2 m 1 2m 2 2m 2 m 1 m 2
and -s respe tively). Sin e is not divisible by or then in the ase

Obviously we have a = (2; 0; 1; 0; 0; 0; 1; 0) (2)

. Suppose now that the laim is true
when N =n k
botht+1 and t have to be multiples of k , yielding k = 1 .
for m=k 2 =s
and denote
k 1
, then 12-3. The given inequality is equivalent to

a = (s + s; 0; s ; 0; s s; 0; s ; 0) :
(2k ) 2 2 2 2 a4 + b4 + 4 2a b 2b 2 a
2 2 2 2 2 2
<0:
Sin e the robot an move to any vertex M from either of its neighbouring verti es, Transforming the left side of this inequality we have

we have
a4 + b4 + 4 2a2b2 2b2 2 2 2a2 = (a2 + b2 2)2 4a b =
2 2

a(2 k +1)
= (0; 2s + s; 0; 2s 2 2
s; 0; 2s2 s; 0; 2s2 + s) = (a2 + b2 2
2ab)(a2 + b2
2 + 2ab) =
= (a b)2 2) (a + b)2 2 =
and similarly
= (a b + )(a b )(a + b + )(a + b ) :
a(2 = (4s + 2s; 0; 4s ; 0; 4s
k +2) 2 2 2
2s; 0; 4s ; 0) :
2
Hen e the given inequality is equivalent to

Sin e4s = 2 2s = 2
2 2k
and
k
, we have proved the laim for m = k + 1 . Hen e (a + b + )(a + b )(b + a)( + a b) > 0 : (1)
the number of possibilities to rea h B =5 from A =1 in n = 2k minutes is
22k
2 = 2 (2
2 k
1) 1 k 1 k 1
. Here the rst term is positive and at most one of the other three an be negative
(e.g. a + b < 0 and b + a < 0 would give 2b < 0  a ontradi tion). Hen e
12-1. Answer: 17 minutes. (1) holds if and only if a , b and satisfy the triangle inequalities.

17 18
  
!1 !2 be the ir um ir les of triangles AKL and CKL (see Figure 13). n
12-4. Let and
We have possible distributions of k prizes among n guests, and
Suppose that ! and !2 are tangent to ea h other in point C , and let l2 be their k
ommon tangent. Then    
n
6 KLC = 6 KCl2 = 6 BCl2 = 6 BDC : k+1
= nk+1k
n
k
: (2)

Hen e KL k BD and 6 ADB = 6 AKL  therefore then angle between AB and

the tangent to ! in A is equal to the angle between AL and the tangent to !1
Let nbethe number of guests and m the a tual number of prizes, then we

in A . Sin e the points A , B , L are ollinear then the tangents to ! and !1 in

n
= 2


n
 
n

= 3
 n  . Substituting from (2) we
m 1 m+1 2 m  
have and
A oin ide, i.e. these ir les are tangent to ea h other. m
   
This argument an be reversed to show that ! and !1 being tangent to ea h other n m+1

n
= 2
n n m n

= 3
 n  . Hen e
m 1 m 1 m+1 m 2 m
have and
in A implies ! and !2 being tangent to ea h other in C . m
n m + 1 = 2m , yielding n = 3m 1 , and 2(n m) = 3(m + 1) . Plugging in
n = 3m 1 here wehave  4m 2 = 3m + 3 , when e m = 5 and n = 14 . It
14 = 2002
remains to al ulate
5 .

l2
Kq
IMO Team Sele tion Test: May 2002
Dq
First Day
qC

q
!2 1. The prin ess wishes to have a bra elet with r rubies and s emeralds arranged
q in su h order that there exist two jewels on the bra elet su h that starting with
q B L
A these and enumerating the jewels in the same dire tion she would obtain identi al

! sequen es of jewels. Prove that it is possible to fulll the prin ess's wish if and

l1 !1 only if r and s have a ommon divisor.

2. Consider an isos eles triangle KL1 L2 with jKL1j = jKL2 j , and let KA , L1 B1 ,
Figure 13
L2 B2 os 6 B1 AB2 <
3
Alternative solution. Let !1 and !2 be the ir um ir les of triangles AKL
be its angle bise tors. Prove that
5 .

and CKL . ! and !1 are tangent to ea h other in A then some homoth-

If 3. In a ertain ountry there are 10 ities onne ted by a network of one-way nonstop
ety relative to A maps ! to !1 . Sin e K is the interse tion point of AD with ights so that it is possible to y (using one or more ights) from any ity to any
!1 and L is the interse tion point of AB with !1 , and points B and D lie on ! , other. Let n be the least number of ights needed to omplete a trip starting
then this homothety takes D to K and B to L , when e KL k BD . Sin e BK
from one of the ities, visiting all others and returning to the starting point. Find
and DL interse t in C then some homothety relative to C maps B to K and the greatest possible value of n.
D to L . This homothety then maps the ir um ir le ! of triangle CDB to the
ir um ir le !2 of triangle CKL . Hen e ! and !2 are tangent to ea h other
in C .
Se ond Day
Similarly we an show that ! and !2 being tangent to ea h other in C implies
! and !1 being tangent to ea h other in A .
4. Let ABCD be a y li quadrilateral su h that 6 ACB = 26 CAD and
12-5. Answer: 2002 . 6 ACD = 26 BAC . Prove that jCAj = jCB j + jCDj .

19 20
5. Let 0 <  < 2 and x1 ; x2; : : :; xn be real numbers su h that The sine rule in triangle AL1 B2 yields

sin x + sin x + : : : + sin x > n
sin  :

1 2 n
s
t
= sin 6 sinB AL = os 2 =2
os =2
1

= 11 ++ os
os 
Prove that s 2
and AB B
. The osine rule in triangle 2 yields
sin(x ) + sin(x2 ) + : : : + sin(x ) > 0 : t  1
1 n

6. Pla e a pebble at ea h non-positive integer point on the real line, and let n be a
s = t + t 2t os  = 2t (1 os  );
2 2 2 2 2

xed positive integer. At ea h step we hoose some n onse utive integer points, s
= 2(1
2
os )
1 + os  = 2(1 os )
remove one of the pebbles lo ated at these points and rearrange all others arbi-
trarily within these points (pla ing at most one pebble at ea h point).
and
t 1 + os  . Hen e and

Determine whether there exists a positive integer

we an pla e a pebble at a point with oordinate greater than
n su h that for any given N > 0
N in a nite number os  = 13 ++ 2 os  2 2 3
2 os  = 1 3 + 2 os  < 1 5 = 5 :
of steps des ribed above.
K

Solutions of Sele tion Test

r B1 s D B2
1. Note that if g d (r; s) = d > 1 then the prin ess's wish an be fullled. Let
d
r0 =
0 s  we pla e on the bra elet r0 rubies and s0 emeralds, then again r0 s t t s
and s =  
d
rubies and s emeralds, et . ( d times)  now any two jewels at distan e r + s
0 0 0 22
have the required property.
L2 A L1
Suppose now that the required ordering exists. Label the positions on the bra elet
by 0; : : :; n 1 n = r + s (thinking of them as modulo n ) and denote by P (i)
where
i for an ordering P . It su es to show that g d (r; n) > 1 .
Figure 14
the jewel at position

Let P be the required ordering, and let a , a + i (where 0 < i < n ) be the positions 3. Answer: 30 .
of the two jewels mentioned in the ondition. Then P (a + j ) = P (a + i + j ) for L1 ; : : : ; L10 x
any j > 0 and hen e P (b) = P (b + i) = P (b + 2i) = : : : for any position b . Let
Let be the ities and denote by ij the minimum number of ights
required to rea h L from L . Let
k be the smallest positive integer su h that n divides ki , and let R be the set of
j i

all positions with rubies. For any b in R we have k distin t rubies at positions m = max x ;
b; b + i; : : :; b + (k 1)i : denote the set of these rubies O(b) and all the orbit 6= ij
i j

of b . A standard argument shows that the set of all rubies on the bra elet is
we an assume w.l.o.g. that i = 1, j = m + 1 and the shortest path from L1 to
the disjoint union of some number of orbits, with ea h orbit ontaining k rubies.
L
Hen e k divides r and sin e n divides ki with 0 6 i < n then g d (r; n) > 1 .
m+1 is

L1 ; L2 ; : : :; L ; L :
2. Denote  =6
L1 KL2 ,  = 6 KL1 L2 = 6 KL2 L1 and  = 6 B1 AB2 (see Fig-
m m+1

ure 14). Sin e B1 and B2 are symmetri relative to KA , we have B1 B2 ? KA We ontinue this path, ying from L to L , then to L et . and nally
and B1 B2 k L1 L2 , i.e. triangle AB1 B2 is isos eles and AK is its angle bise tor.
m+1 m+2 m+3

from L ba k to L1 :
Sin e 6 B2 B1 L1 = 6 L2 L1 B1 = 6 B2 L1 B1 , then B2 L1 B1 is also isos eles. Denote
n

s = jB1 B2 j = jB2 L1 j and t = jAB1 j = jAB2 j . L1 ; L2 ; : : :; L ; L m m+1 !L m+2 ! : : : ! L ! L1 :

n

21 22
 Here ea h of the 10 (m +1) + 1 = 10 m se tions denoted by arrows ontains On the other hand,
at most m ights, hen e the entire round-trip ontains at most
X
n
X
n

+ 11 (sin x + : : : + sin x ) + ( os x + : : : + os x ) = os(x x ) 6 n2 ;

= 121
m m 2
2 2

m + m
(10 m) = m
(11 m) 6
1 n 1 n i j

2 4 i=1 j =1

ights, i.e. no more than 30 ights.

a ontradi tion.

An example of a network requiring exa tly 30 ights is shown on Figure 15. 6. Answer: there is no su h n.
For n =1 and n =2 we annot pla e a pebble in any point with a positive
r
D oordinate  hen e let n > 3. Consider the innite sum
C
r 2 S =a x1
+a + a + :::
x2 x3

2 where x1; x2; x3 ; : : : are the oordinates of the points with pebbles at some
r r r r r r
given moment and a a positive real number to be determined later. We
 a n
r B 
show that it is possible to hoose a ording to so that the initial sum
S0 = a + a + a + a + :::
0 1 2 3
onverges (for this it su es to have a > 1)
r A and at ea h step the sum S an only de rease. Hen e we always have S 6 S0
and it is impossible to pla e a pebble at a point with an arbitrarily large positive
Figure 15 Figure 16 oordinateN , sin e for su iently large N we have a > S0 . N

To show this onsider for n = 2k 1 the equation

4. Denote 6 CAD = and 6 BAC = , then 6 ACB = 2
and 6 ACD = 2 (see
1+ x+x +:::+x = x +:::+x
Figure 16). Sin e ABCD is y li then 3 + 3 = 6 BCD + 6 BAD = 180Æ and
2 k 1 k 2k 2
(3)

 +  = 60Æ . Applying the sine rule to triangles ABC and ACD gives and for n = 2k the equation

jCB j = 2R
sin  ; jCDj = 2R
sin  ; jCAj = 2R
sin( + 2 ) ; 1+ x+x +:::+x

2 k 1
=x k +1
+:::+x 2k 1
: (4)

where R is the ir umradius of ABCD . Hen e it is su ient to show that

06x61
sin  + sin  = sin( + 2)  +  = 60Æ if . Indeed:
For
x
the left side ex eeds the right side but for su iently large positive
the right side ex eeds the left side. Hen e the equation has a root a > i1 . It
hn +1
sin  + sin  = 2 sin  +2  os  2  = 2
12
os  +2   = m and t su h that 1 6 t 6
 
remains to show that for any integers the
2
= os(30Æ ) = sin(60Æ + ) = sin( + 2) : sum of any t 1 elements of A = fa ; a ; : : :; a
+1 + 1
g does not ex eed the
m m m n

sum of any t elements of A (here m; : : :; m + n 1 are the hosen n onse utive

integer points and t is the number of points having a pebble before this step and
5. Suppose the laim does not hold, i.e. no pebble after this step  hen e t 1 points have no pebble before this step and
a pebble after this step). Note that it su es to have m = 0 and prove that the
sin(x 1 ) + sin(x2 ) + : : : + sin(x n ) < 0 sum of t 1 largest elements of A does not ex eed the sum of t smallest elements,
i.e.
whi h gives

os 
(sin x + : : : + sin x ) > n
os  1+ a+a + :::+ a > a

2 t 1 n t+1
+ :::+a n 1
;
os x + : : : + os x >
sin 
1 6 t 6 n +2 1 = k
1 1
n n
h i
where . This dire tly follows from (3) or (4) for x=a and

a>1
and hen e
from the fa t that sin e then deleting an equal number of terms from ea h

(sin x + : : : + sin x ) + ( os x + : : : + os x )
1 n
2
1 n
2
>n : 2 side makes the left side larger than the right side.

23 24
Estonian math ompetitions 2002/2003 Seniors (grades 11 and 12)

1. Find all quadruples of integer numbers ( a; b; ; d) for whi h

We thank the IMO ommunity for many of these problems whi h have been taken
52
a
77b
88
91d = 2002 :
from various materials distributed at the re ent IMO-s.

2. Four rays spread out from point O in a 3-dimensional spa e in a way that the
angle between every two rays is  . Find os  .
3. Three onse utive positive integers ea h happen to be a power of some prime
Autumn Open Contest: O tober 2002 number. Find all triples of integers with this property.

4. n points of integral oordinates in the plane have been painted white. If some
Juniors (up to 10th grade) points A and B are white, we may paint white the point symmetri al to A in
respe t to B (at every step we paint only one new point). For whi h smallest
natural number n an the initial n points be hosen in a way that any point of
All inner angles of a 7-gon are obtuse, their sizes in degrees being pairwise dierent integral oordinates ould be painted white in a nite number of steps?
integer numbers divisible by 9. Find the sum of the two biggest angles of this 7-
gon. 5. During an ele tion ampaign K tabloid newspapers published ompromising ma-
terial about P politi ians; every politi ian being written about in an odd number
2. Cir les with entres O1 and O2 interse t in two points, let one of whi h be A. of newspapers and every newspaper writing about an odd number of politi ians.
The ommon tangent of these ir les tou hes them respe tively in points P and a) Prove that the number of newspapers K and the number of politi ians P are
Q. It is known that points O1 , A and Q are on a ommon straight line and either both odd or both even numbers.
points O2 , A and P are on a ommon straight line. Prove that the radii of the b) Find the total number of all possibilities whi h newspaper wrote about whi h
ir les are equal. politi ians.

3. A 4-digit number ab d is divisible by the produ t of 2-digit numbers ab and d .

Find all 4-digit numbers with su h property. Solutions of Autumn Open Contest
4. Mari and Jüri ordered a round pizza. Jüri ut the pizza into four pie es by two J1. Answer: 315 .
Æ
straight uts, none of whi h passed through the entre point of the pizza. Mari
Let the interior angles of the 7-gon in degrees be a ; a ; : : :; 9a7 ,
9 1 9 2 where
an hoose two pie es not aside of these four, and Jüri gets the rest two pie es.
a1; a2; : : :; a7 are integers and a1 < a2 < : : : < a7 . As all the interior angles
Prove that if Mari hooses the pie e that overs the entre point of the pizza, she
are obtuse, we have a1 > 11 and
will get more pizza than Jüri.
a1 + a2 + : : : + a7 > 11 + 12 + 13 + 14 + 15 + 16 + 17 = 98 :
5. The lo k on the wall shows four numbers (possible times from 00:00 to 23:59),
On the other hand a1 + a2 + : : : + a7 = 100 , be ause the sum of the interior angles

180Æ = 900Æ . Bearing in mind that a1 < a2 < : : : < a7
the shapes of the numbers being as shown in the pi ture.
of a 7-gon is (7 2)
we nd that the only possibilities are (a1 ; a2 ; : : :; a7 ) = (11; 12; 13; 14; 15; 17; 18)
a) How many times a day does the mirror image of the lo k on the wall or-

and (a1 ; a2 ; : : :; a7 ) = (11; 12; 13; 14; 15; 16; 19) . In either ase the sum of the two
respond to some time (the mirror image of number 1 is equal to number
1)?
biggest angles is 9
(a6 + a7 ) = 9
35 = 315 .
Æ
b) How many times a day does the mirror image of the lo k on the wall orre-

spond to the same time as the lo k? J2. Solution 1. By onstru tion 6 O1 P Q = = 6 P QO2 . We also have
2
jO1P j = jO1Aj and jO2Qj = jO2Aj that implies
6 O1P A = 6 P AO1 = 6 O2AQ = 6 AQO2:

1 2
 Consequently 6 AO1P = 6 QO2A and is that 100 + n is divisible by d . We have obtained 100 + n = k
d = k
n
ab
(where k > 1 , be ause 100 + n is a 3-digit and d is a 2-digit number). Therefore
6 AP Q =  6 O1P A =  6
AQO2 = 6 P QA : 100 must be divisible by a one-digit number n , implying n equals 1, 2, 4 or 5.
2 2
Consider these four ases separately.

Thus the triangles O1 P Q and O2 QP are similar be ause the respe tive angles
1) Case n=1 gives the equation 101 = k
ab . As 101 is prime and k > 1 , this
are equal. As PQ is their ommon side, these triangles are ongruent, giving
jO1P j = jO2Qj .
ase has no solutions.

2) Case n=2 gives 102 = k
2
ab that implies 51 = k
ab . As 51 = 3
17 , the

Solution 2. Let 6 O1 QP = 6 AQP =  . As 6 O1 P Q = we have
Æ
90 , only possibility is k = 3 , ab = 17 , and we obtain one 4-digit answer 1734 .
6 P O1A = 6 P O1Q = 90Æ  . Be ause the triangle P O1A is isos eles, we obtain 3) Case n = 4 gives 104 = k
4
ab that implies 26 = k
ab . As 26 = 2
13 , the
only possibility is k = 2 , ab = 13 , and we obtain one 4-digit answer 1352 .

6 O1P A = 6 O1 AP 1 Æ Æ
Æ  4) Case n = 5 gives 105 = k
5
ab that implies 21 = k
ab . As 21 = 3
7 ,

= 180 (90 ) = 45 + :
2 2 this ase has no solutions (the number 21 annot be expressed as a produ t of a

Æ
On the other hand 6 P QO2 = 90 implies 6 AQO2 = 90
Æ  . The triangle AQO2 2-digit number and a number ex eeding 1).

is isos eles, onsequently 6 AQO2 = 90

Æ  . The onditions of the problem imply J4. Constru t a straight line parallel to one of the uts passing through the entre
6 O2AQ = 6 O1AP , therefore point of the pizza O  this line passes through two pie es of all four, dividing
a, b, d
Æ  = 90Æ  ;
both into two parts. Let and be areas as shown in the gure 2. If Mari
45 + hooses the pie e having the entre point of the pizza, the total area of her share
2
is (a + b) + ( d) and the total share of Jüri is ( a b) + ( + d) . To omplete
giving  Æ
= 30 . Thus 6 P O1A = 60Æ and 6 O1AP = 6 O2 AQ = 60Æ , i.e. the the solution, one has to prove that a + b + d > a b + + d , or b d > d b .
triangles O1AP and O2 AQ are equilateral. As 6 O1 P Q = 6 P QO2 = 90 , the
Æ This is similar to b>d that evidently holds.

distan e of the point A to the straight line P Q is equal to

1
jO1P j . On the other 2

hand the distan e equals

1
2
jO2Qj , giving us jO1P j = jO2Qj .
PSfrag repla ements
O1 a
0
0
1
d b

PSfrag repla ements O1 O2

Figure 2

A
J5. Answer: a) 121 ; b) 11 .
P Q a) The numbers having a number as its mirror image are 0, 1, 2, 5 and 8. The
number 8 an only be the last digit of hours or minutes, the mirror image of whi h
Figure 1
is respe tively the rst digit of minuts or hours, implying that 8 annot be in any
time onsidered in the problem. Thus the rst digit of hours an be 0 , 1 or 2 and
J3. Answer: 1352 , 1734 .
the last digit of hours an be 0 , 1 , 2 or 5 , totally giving 3
4 1 = 11 numbers
We need the integer ab d = 100
ab + d be divisible by ab , therefore d must ( 25 is not an hour) as hour time. Sin e the mirror image of minutes is hours, the
be divisible by ab . Let d = n
ab , then n is a one-digit number (sin e a 6= 0 ). minutes an be any of these 11 numbers. Consequently the mirror image of the
The ne essary ondition that ab d = (100 + n)
ab would be divisible by ab
d lo k orresponds to some time 11
11 = 121 times a day.
3 4
 b) For having the mirror image time equal to the real image time, the minutes Hen e the entre point K of the segment A3 A4 is also on the plane ! , having
must equal the mirror of the hours, i.e. any of the hours possible found in a) gives jKA1j = jKA2 j . As A1 M1 ? KA2 and A2 M2 ? KA1 , we have that KM1 OM2
exa tly one su h time. Therefore the mirror image time equals the real image time is a y li quadrangle that implies
11 times a day.
 = 6 A1OA2 = 6 M1 OM2 = 180Æ 6 M1 KM2 :
S1. Answer: the only su h quadruple is (2 2 ; ; 1; 1) .
After fa torisation of the bases of powers and 2002 , we may rewrite the given
As KA2 is a median of A2 A3A4 , the ratio jKM1j : jKA2j = 1 : 3,

equation as 6 M1 KM2 jKM1j jKM1 j 1

therefore os =
jKA1j =
jKA2j =
3
. Finally we obtain

2a
13a)
(7b
11b)
(23
11 )
(7d
13d ) = 2
7
11
13 6 M1 KM2 = 1

(2 os = os .
3
or in the form
A3
2
a 7b+d 11b+ 13a+d = 2
7
11
13:
2 +3

For the last equation to hold, the powers of respe tive primes on the left and right
side must be equal. This implies a system of equations

8 1111111111
0000000000 K 11111111
00000000
> 2a + 3 = 1
PSfrag repla ements
> 000000000011111111
1111111111 00000000
< 000000000011111111
1111111111 00000000
b+d = 1
:
000000000011111111
1111111111
M
0000000000
1111111111
00000000
00000000
11111111
>
> b+ = 1 0000000000
1111111111
000000000000
111111111111
1
0000000000
1111111111 O
000000000000
111111111111
00000000
11111111
00000000000
11111111111
00000000
11111111
:
a+d = 1 0000000000
1111111111
000000000000
111111111111 M2
00000000000
11111111111
00000000
11111111
00000000000
11111111111
0000000000
1111111111 00000000
11111111
000000000000
111111111111 A4
00000000000
11111111111
0000000000
1111111111 00000000
11111111
000000000000
111111111111
00000000000
11111111111
0000000000
1111111111 00000000
11111111
000000000000
111111111111
00000000000
11111111111
0000000000
1111111111 00000000
11111111
The se ond and the third equation give = d , and the se ond and the fourth
000000000000
111111111111
00000000000
11111111111
0000000000
1111111111 00000000
11111111
000000000000
111111111111
00000000000
11111111111
0000000000
1111111111 00000000
11111111
a = b . Multiplying the third equation by 2 and subtra ting from the rst gives 000000000000
111111111111
00000000000
11111111111
0000000000
1111111111 00000000
11111111
000000000000
111111111111
00000000000
11111111111
0000000000
1111111111 00000000
11111111
000000000000
111111111111
00000000000
11111111111
0000000000
1111111111 00000000
11111111
= 1 and b = 2 . The rest of the unknown variables must then be a = 2 000000000000
111111111111
00000000000
11111111111
0000000000
1111111111
000000000000
111111111111 00000000
11111111
00000000000
11111111111
and d = 1 . We have obtained the only quadruple satisfying the onditions A21111111111
0000000000
000000000000
111111111111
0000000000
1111111111
00000000
11111111
00000000A1
11111111111
00000000000
11111111
(2; 2; 1; 1) .
Figure 3
1
S2. Answer:
3
.
; ; ; ; ; ;
S3. Answer: (1 2 3) , (2 3 4) , (3 4 5) and (7 8 9) . ; ;
At least one of the three onse utive integers is divisible by 2 and hen e must be
Consider the points A1 , A2 , A3 and A4 on the respe tive rays equidistant from 1 2 3
O . Consider any pair of points Ai1 ; Aj1 and Ai2 ; Aj2 , where i1 6= j1 and i2 6= j2 . a power of 2 (with positive exponent). Considering 2 = 2 , 2 = 4 and 2 = 8 ,
k
As jOAi1 j = jOAi2 j and jOAj1 j = jOAj2 j , also 6 Ai1 OAj1 =  = 6 Ai2 OAj2 ,
we nd the solutions. We prove that no more solutions exist. Say 2 belongs to

we obtain that the triangles OAi1 Aj1 and OAi2 Aj2 are ongruent, implying
a triple in question, where k > 4. Sin e the only onse utive even numbers that
k k k
jAi1 Aj1 j = jAi2 Aj2 j . Therefore all segments Ai Aj , where i 6= j are of equal are powers of 2 are 2 and 4 , the triple must be in the form 2 1, 2 and 2 + 1 .

lengths. Consequently all the fa es of Ai Aj Ak , having i; j; k pairwise dierent,

As one of three onse utive integers is always divisible by 3, one of the numbers
k k
are equilateral triangles, and A1 A2 A3 A4 is a regular tetrahedron.
2 1 and 2 + 1 must be a power of 3 . Consider these two ases separately.
n k 1 ; onsider the sides of the equation modulo 8 : 2k 1  7 , but
Cut this tetrahedron with a plane ! ontaining the points O , A1 and A2 . The
1) Say 3 = 2
2x 2x+1
straight line A1 O passes through the triangle A2 A3 A4 in its entre M1 and the
3  1 and 3  3 , a ontradi tion.
straight line A2 O passes through the triangle A1 A3 A4 in its entre M2 (see gure
n k
2) Say 3 = 2 + 1 ; onsider the sides of the equation modulo 4 : 2 + 1  1
k
2x 2x+1
3). By onstru tion M1 and M2 are on the plane ! . The straight line A2 M1 and 3  1 , 3  3 implying that n is even. Now onsider the sides of the
3y 3y +1 3y +2 6z
as a median of the triangle A2 A3 A4 bise ts the segment A3 A4 and the straight equation modulo 7 : sin e 2 + 1  2 , 2 + 1  3, 2 + 1  5 and 3  1 ,
6z +2 6z +4 k n
line A1 M2 as a median of the triangle A1 A3 A4 bise ts the segment A3 A4 as well. 3  2 , 3  4 at the same time, we must have 2 + 1 = 3  2 (mod 7) .
5 6
 Consequently the rst number of the triple 2
k 1 is divisible by 7 , therefore it Remark. The onstru tion des ribed above an be presented more onveniently if

is a power of 7 : 2
k m
1 = 7 . Consider the sides of the equation modulo 16 : one performs su h ane transdu tion that takes the three verti es of the parallel-
2
k 1  15 , but 72w  1 , 72w+1  7  a ontradi tion. ; ; ;
ogram to points (0 0) , (0 1) and (1 0)  then the points to be hosen into the
subset are exa tly the points at least one of the oordinates of whi h is an even
1
0
0
1 number.
1 0
0 1
11
00
00 0
11 Solution 2. Similarly to the previous solution we noti e that it su es to nd su h
00 1 0
11 1 0
1
1 0
0 1 0
1 1 0
0 1 a set of points of integral oordinates on the plane that:
1
0 1 0
0 1 0
1 1 0
0 0 1
1 1 0 (i) is symmetri al in respe t of every its point (i.e. any point that is symmetri al
1
0 1
0
1 0
0 1 1 0
0 1 0
1 1 0
0 1 to any point in the set in respe t to any other point in the set also belongs to

1 0
0 1 0
1 0 1 1 0
0 1 0
1 the set);
0
1
1 0
0 1 1 0
0 1 0
1 1 0
0 1 (ii) ontains all three initial white points;
1
0 1 0
0 1 0
1 1 0
0 1 0 1
0
1 0 (iii) does not ontain all points of integral oordinates on the plane.
1
0 1
1 0
0 1 1 0
0 1 0
1 1 0
0 1
In order to onstru t su h a set we onsider the parities of the three initial points
1 0
0 0 1
1 1 0
0
1 1 0
0 1 0
1 and see that one an (if ne essary, by shifting the oordinate system by one unit)
1 0
0 1 1 0
0 1 0
1 always obtain the situation where all of these three points have at least one of
1
0
0 1 0
0 1 0
1 the oordinates an even number. Now the set in question an be expressed by all
1
1 0
0 1 of these points of integral oordinates on the plane that have at least one of the
oordinates an even number (if two points of integral oordinates are symmetri al
in respe t to some third point of integral oordinates, then the respe tive oor-
Figure 4 dinates of the two points are of the same parity; the onditions (ii) and (iii) are
evidently satised).
S4. Answer: n = 4.
Solution 1. It is evident that one annot nd suitable olouring in the ase n=2 S5. Answer: b) 2
( K 1)( P 1)
.
(one an only olour the points on the straight line determined by the two initial Solution 1. a) Consider the situation des ribed in the problem as a table onsisting
points) and that suitable olouring exists in the ase n= 4 (e.g. olour points of K rows and P olumns, orresponding respe tively to the newspapers and the
; ; ; ;
(0 0) , (0 1) , (1 0) and (1 1) ). politi ians, and olour the square in the i th
j th olumn bla k in
row and the
It remains to prove that no su h olouring exists in the ase n = 3. If one the ase the i th j th politi ian. Let ai and bj be
newspaper wrote about the
hooses three points ollinear, then no points outside the ommon straight line respe tively the number of oloured squares in the i th row and in the j th olumn.
an be oloured white. Now we prove that if one hooses three points at three For a1 + : : :+ aK = b1 + : : :+ bP (either sum represents the total number of oloured
verti es of a parallelogram (su h parallelogram always exists if the points are not squares) and a ording to the onditions of the problem all the addends are odd,
hosen ollinear), one annot olour the fourth vertex (having evidently integral the ounts of the addends K and P must be of the same parity.
oordinates) of this parallelogram. For this it su es to show su h subset of all b) The number of the squares not being in the last row and in the last olumn
points of integral oordinates that is symmetri in respe t of any of its point and is ( K 1)(P 1) , and as we an either olour or leave un oloured any of these
ontains the three verti es of the parallelogram, but does not in lude the fourth squares, the total number of possibilities to olour these squares is 2
(K P
1)( 1)
.
vertex. We obtain su h set of points in the following way: Show that any su h olouring an be ompleted exa tly uniquely to the olouring
(1) Consider a latti e on the plane, the axes of the latti e being determined by of the whole table satisfying the onditions of the problem (i.e. the number of
the edges of the parallelogram, one latti e-point in the opposite vertex of the oloured squares in every row and in every olumn is odd). Indeed: the square
un- oloured vertex and the distan e between latti e points being equal to in the last olumn must be oloured exa tly in these rows (from the 1 st until the
twi e the respe tive side-length of the parallelogram. ( K 1 )th) that have an even number of squares oloured. After that the square
(2) Choose the latti e-points and the midpoints of segments that onne t the in the last row must be oloured exa tly in these olumns (from the 1 st until the
neighbouring latti e points and are parallel to the axes of the latti e (these P th) that have an even number of squares oloured. It remains to he k that the
points are shown white in gure 4). last row has an odd number of squares oloured as well  this is proved by the

7 8
 fa t that the numbers of rows and olumns are by a) odd, all the rows ontain
altogether as many oloured squares as do all the olumns, and all the rows ex ept
2. The shape of a dog kennel from above is an equilat-

C
q
eral triangle with side length 1 m and its orners in
the last and all the olumns ontain an odd number of oloured squares. points A, B and C , as shown in the pi ture. The
K
q

A B
q q

hain of the dog is of length 6 m and its end is xed

Solution 2. a) Consider the situation des ribed in the problem as a graph of K to the orner in point A . The dog himself is in point K in a way that the hain
yellow and P brown verti es, where one has edges only between verti es of dier- is tight and points K , A and B are on the same straight line. The dog starts to
ent olours; an edge between the i th yellow and the j th brown vertex shows that move lo kwise around the kennel, holding the hain tight all the time. How long
the i th newspaper wrote about the j th politi ian. A ording to the onditions of is the walk of the dog until the moment when the hain is tied round the kennel
the problem an odd number of edges emerges from every vertex of the graph, the at full?
sum of all these numbers equals twi e the number of edges (as every edge ounts
3. A ar, a motor y le, a moped and a bi y le were driving at unvaried speeds in a
at its both vertex) and is thus an even number. Hen e the number of verti es
K+P is an even number, implying that K and P are of the same parity.
straight road. At 12:00 the ar passed the moped driving in the same dire tion and

b) Let the yellow verti es be A1 ; : : :; AK and the brown verti es B1 ; : : : ; BP . met the bi y le and the motor y le driving in the opposite dire tion respe tively

One has 2
( K 1)(P 1)
possibilities for joining the verti es A1 ; : : :; AK 1 and
at 14:00 and 16:00. The motor y le met the moped at 17:00 and passed the bi y le
at 18:00. At what time did the moped meet the bi y le?
B1 ; : : :; BP 1 (one may or may not have an edge between the verti es in any
of the pairs ( Ai ; Bj ) ). We shall prove that every su h graph an be ompleted 4. Consider the points A1 and A2 on the side AB of the square ABCD taken in
to a graph satisfying the onditions of the problem (i.e. an odd number of edges su h a way that jAB j = 3 jAA1j and jAB j = 4 jA2B j , similarly onsider points B1
emerging from every vertex) by adding edges that emerge from AK and BK in an and B2 , C1 and C2 , D1 and D2 BC , CD and DA .
respe tively on the sides
exa tly unique way. Indeed, one must draw edges from the vertex AK exa tly into The interse tion point of straight lines D2 A1 and A2B1 is E , the interse tion
the verti es Bj ( j = 1 ; : : :; P 1 ) that had an even number of edges emerging. point of straight lines A2 B1 and B2 C1 is F , the interse tion point of straight
After that one must draw edges from the vertex BP exa tly into the verti es Ai lines B2 C1 and C2 D1 is G and the interse tion point of straight lines C2 D1 and
(i = 1; : : :; K ) that had an even number of edges emerging. It remains to he k D2 A1 is H . Find the area of the square EF GH , knowing that the area of ABCD
that the vertex BP now also has an odd number of edges  this is proved by the is 1 .
fa t that a ording to a) the total number of verti es is even, all the rest of the
verti es has an odd number of edges emerging and the total sum of edges emerging 5. Is it possible to write one of the letters A, B , C or D in every square of an

from any vertex is twi e the number of edges, thus an even number. innite he ked paper in su h a way that every 2 2 square ontains all four
letters?

Seniors (11th and 12th grade)

1. Find the values of sharp angles  and  that satisfy

8
> 3
< os  + os  =
> 2 2

Spring Open Contest: February 2003 2

:
>

sin 
> 1
: sin =
4

Juniors (up to 10th grade) 2. Juku has 2003 wooden sti ks (nothing is known about their lengths). Juku on-
stru ts a re tangle of all these sti ks, breaking some sti ks into two parts. Find
the least number of breakings that would be su ient for any lengths of the sti ks.

1. A four-digit number a not ontaining digit 9 is a square of an integer. If we

3. The sequen e fFng is dened as follows:
in rease every digit of a by 1 , we obtain a square of an integer again. Find all
4-digit numbers with su h property. F1 = 1 ; F2 = a; Fn = Fn 1 + Fn 2 ; if n > 3:

9 10
 a) Do there exist integers a and N> 1 that no member of the sequen e fFng
would be divisible by N?
b) Do there exist integers a and N >1 that some two onse utive members of
the sequen e fFng would be divisible by N ?
4. Consider the points D , E and F on the respe tive sides BC , CA and AB of the
triangle ABC in a way that the segments AD , BE and CF have a ommon point C
jAP j jBP j jCP j
q

P. B
jP Dj = x , jP E j = y and jP F j = z . Prove that xyz (x + y + z ) = 2 .
q q q
Let
K A
5. Consider nn squares painted bla k and white as a hessboard in a omputer
monitor. One an mark any re tangle onsisting of whole squares by mouse-
dragging and invert the olours of the squares in a marked re tangle by mouse-
li king. Find the smallest number of mouse- li ks by what one an onvert all
the squares mono hrome.
Figure 5

J3. Answer: At 15.20.

Solution 1. Constru t a diagram with the graphs of movement of the ar, the
motor y le, the moped and the bi y le, and have one axis for time in hours and
the other for distan e in kilometers. Let A, B , C , D, E and F be points in this
diagram that respe tively orrespond to the meeting of the ar and the moped, the
ar and the bi y le, the ar and the motor y le, the motor y le and the moped,
the motor y le and the bi y le, and the moped and the bi y le (see gure 6). In
Solutions of Spring Open Contest this ase B is the midpoint of segment AC and D is the midpoint of segment CE .
Hen e F is the interse tion point of the medians AD and EB in triangle ACE ,
implying that jAF j = 23 jADj and the dieren e in time oordinates between

J1. Answer: 2025 is the only su h number.

Solution. Let a=x 2

and a + 1111 = y 2
. Then
points A and F is
2
3

(17 12) =
10
3
hours. Therefore the moped met the bi y le

10


1111 = 101 11 = y2 x2 = (y + x)(y x) 3
hours, or 3 hours and 20 minutes later than it met the ar, i.e. at 15.20.

that implies y + x = 101 , y x = 11 or y + x = 1111 , y x = 1 . The rst ase

leads to x = 45 , y = 56 and x2 = 2025 . The se ond ase leads to x = 555 and
y = 556 , but the squares of these numbers are not 4-digit numbers.
J2. Answer: 14  m.
Solution: The dog's path onsists of six 120
Æ ar s of a ir le, the radii of whi h
are 6 , 5 , 4 , 3 , 2 ja 1 metres (see gure 5). Hen e the total length of the dog's
path is

2
1
3

2
(6 + 5 + 4 + 3 + 2 + 1) =
3

21 = 14 (m):

11 12
 x the ar
G
6
the motor y le
C D C2 C1 C
the moped
D B2
the bi y le
B D1
F E F
H
A B1
D2
-

12.00 14.00 16.00 17.00 18.00 t A A1 A2 B

Figure 6
E
Solution 2. Say zero time and zero point the time and pla e of the meeting of Figure 7
the ar and the moped, and let the positive dire tion of the axis be the dire tion
where the ar and the moped move. Let the speeds of the ar, the motor y le, J4. Right triangles AA1 D2 and EA1 A2 are similar, sin e 6 AA1 D2 = 6 EA1 A2 (see
the moped and the bi y le be respe tively a, b, and d, then at time t their gure 7). We shall show that jD2A1 j = jA1A2 j , i.e. these triangles are ongruent.
respe tive lo ations areat , 4a b(t 4) (sin e the ar and the motor y le met at Indeed,

time t = 4 ), t and 2a d(t 2) (sin e the ar and the bi y le met at time t = 2 ). s

As the motor y le and the moped met at time t = 5 , we obtain an equation p
jD2 A1j = jD2Aj 2 + jAA1 j 2 =
1
16
j AB j2 + jAB j2 = jAB j
1
9
5
12
4 a b = 5 ; (1)

and
and from the meeting of the motor y le and the bi y le at time t=6 we get
 
4 a b
2 =2 a 4 d: (2)
jA1A2 j = jAB j jAA1j jA2 B j = 1
1
3
1
4
jAB j = 125 jAB j :
What should be omputed is the meeting time of the moped and the bi y le, Therefore jD2A1 j = jA1A2j , jAA1 j = jA1E j , jA2 B j = jD2Aj = jEA2j = jHD2j and
a+d
i.e. the time x that satises x = 2 a d(x 2) , or x = 2 . Having
jHE j = jHD2j + jD2 A1j + jA1E j = jA2B j + jA1A2 j + jAA1 j = jAB j :
+d
expressed b from the equation (1) and substituted it into the equation (2), we
obtain 4 a
a 5 ) = 2a 4d , or 6a = 10 + 4d . Adding to both sides 6d gives
2(4 Consequently the squares ABCD and EF GH are of the same side length, i.e.

a+d EF GH is 1 as well.
= 2

10 10 the area of the square
6(a + d) = 10( + d) , implying x = 2 = . Therefore the moped
+d 6 3
10 J5. Answer: no.
met the bi y le hours or 3 hours 20 minutes later than it met the ar, i.e. at
3 Solution. Assume the required onguration of the letters exists. Then no row an

15.20. ontain two similar letters onse utively, therefore one must have three dierent
letters onse utively  without loss of generality assume they be A , B and C .
The next row must ontain aligned to them the letters C, D and A , the row
after the next again has A, B and C , et . We see that any of the three olumns
ontains only two letters, a ontradi tion with the assumption made above.

S1. Answer:  =  = 30Æ .

13 14
 Substitutions os
2
=1 sin
2
 and os2  = 1 sin
2
 give a system of equation S3. Answer: a) yes; b) no.
similar to the given system a) Having a = 3 , the residues of Fn modulo N = 5 are the following
8
1
>
< sin  + sin  =
> 2 2 ; ; ; ; ; ; ::: :
1 3 4 2 1 3
2
:
>

sin 
1
>
: sin =
Sin e every term in the sequen e is determined by two previous terms, the residues
4 modulo 5 are also determined by two previous residues, hen e the quadruple
; ; ;
of residues 1 3 4 2 will repeat innitely and onsequently no member of the
By letting x = sin  and y = sin  , one has a system of equations in x and y sequen e is divisible by 5 .
8 Fn = Fn Fn Fn Fn Fn
1 b) Having written 1+ 2 in the form 2 = 1 , we see that
>
<x +y =
2 2
> every term of this sequen e equals the dieren e of two next onse utive terms.
2
>
: Therefore if some two onse utive terms of the sequen e were divisible by N > 1,
x
y
> 1
: =
4
all the terms pre eeding them would also be divisible by N, in luding the rst
term F1 = 1  a ontradi tion. Consequently for any hoi e of a two onse utive
Having added two times the se ond equation to the rst, we obtain terms of the sequen e are pairwise primes.

(x + y) 2
=
1
2
+2
4 1
= 1 . Sin e  and  are a ute angles, x and y are pos-
S4. Let the base points of heights drawn to the segment AB in triangles ABC and
itive numbers, thereforex + y = 1 , or y = 1 x . Substituting it into the se ond ABP be respe tively K
L (one or both of these points may lie on the
and
extension of side AB ) and let Q be the interse tion point of the segment CK
equation in the system, we obtain x
(1
1
x) = or
with straight line s , passing through point P and being parallel to side AB (see
4
gure 8). Let SXY Z be the area of triangle XY Z , then
1
 
1 2
0= x2 x + = x ; jP F j = jQK j = jP Lj = SABP
4 2
jCF j jCK j jCK j SABC
 = 30Æ .
1 1
x= x+y = 1 y=
from whi h
2
and The equation now gives
2
and
jP Dj = SBCP and jP E j = SACP . Sin e
 = 30Æ . and similarly
jADj SABC jBE j SABC
S2. Answer: 2 . jP Dj = 1 = 1
Solution. Juku may ut one sti k into two halves; onsider these halves one pair jADj 1 + jAP j 1 + x
of sides of a re tangle and divide the rest of the sti ks into two sets A and B. If
jP Dj
the total lengths of the sti ks in sets A and B are equal, no more sti ks need to
be broken  but if, for instan e, the total length of the sti ks in A is bigger, we jP E j 1 jP F j 1
put one by one sti ks from A to B until the inequality of total lengths reverses. and similarly
jBE j = 1 + y and jCF j = 1 + z , we have
Now it su es to break the last sti k that was put from A to B (in the ase the
total lengths equalized no breaking is ne essary). 1 1 1 SBCP + SACP + SABP
+ + = =1 ;
One does not nd su ient to break only 1 sti k e.g. in the ase where the 1+ x 1+ y 1+ z SABC
2002
lengths are 1 , 2 , . . . , 2 . If it would be enough to break only one sti k in
order to onstru t a re tangle, one had its two sides of integer and two sides of y z x z x y x
or (1+ )(1+ )+(1+ )(1+ )+(1+ )(1+ ) = (1+ )(1+ )(1+ ) , giving after mul- y z
fra tional lengths (if all four sides were of integral length, their sum would be an tipli ation and olle ting similar terms the required equation xyz x y z
( + + ) = 2.
even number, but 1 + 2 + : : : + 22002 is odd). Hen e one pair of opposite sides
should be onstru ted only of initial sti ks, but it is not possible, sin e one an
nd a sti k longer than the total sum of the others in every subset of the initial
sti ks.

15 16
 C 4. Ella the Wit h was mixing a magi elixir whi h onsisted of three omponents:
140 ml of reindeer moss tea, 160 ml of y agari extra t, and 50 ml of moonshine.
She took an empty 350 ml bottle, poured 140 ml of reindeer moss tea into it and
E D
started adding y agari extra t when she was disturbed by its bla k at Mesto.

Q P s So she mistakenly poured too mu h y agari extra t into the bottle and noti ed
her fault only later when the bottle lled before all 50 ml of moonshine was added.
Ella made qui k al ulations, arefully shaked up the ontents of the bottle, poured
out some part of liquid and added some amount of mixture of reindeer moss tea

and y agari extra t taken in a ertain proportion until the bottle was full again
and the elixir had exa tly the right ompositsion. Whi h was the proportion of
A K L F B reindeer moss tea and y agari extra t in the mixture that Ella added into the

Figure 8 bottle?

S5. Answer: n , if n is even; n 1 , if n is odd.

5. Is it possible to over an nn hessboard whi h has its enter square
ut out with tiles shown in the pi ture (ea h tile overs exa tly 4
Solution. Every mouse- li k an onvert mono hrome at most 4 pairs of squares
of dierent olours on the outer edge of the re tangle. As one has 4( n 1) of
squares; tiles an be rotated and turned around) if a) n = 5 ; b)

su h pairs initially, one requires at least n 1 mouse- li ks. For even n one needs n = 2003 ?
at least n mouse- li ks, sin e the orner squares are not of same olour and by
marking any re tangle ontaining a orner square one onverts mono hrome only 10th grade
2 pairs of squares of dierent olours on the outer edge of the re tangle.
These numbers prove to be su ient. We may hoose subsequently the 2nd, 4th, 1. The pi ture shows 10 equal regular pentagons where ea h
6th, . . . row and then the 2nd, 4th, 6th, . . . olumn  for odd n we have altogether two neighbouring pentagons have a ommon side. The

2
n 2 1 = n 1 mouse- li ks and for even

n
n we have altogether 2
= n mouse-
2
smaller ir le is tangent to one side of ea h pentagon and
the larger ir le passes through the opposite verti es of
li ks.
these sides. Find the area of the larger ir le if the area of
the smaller ir le is 1 .

Final Round of National Olympiad: Mar h 2003

m2 + n2
9th grade 2. Find all possible integer values of where m and n are integers.
mn
1. Let A1 , A2 , . . . , Am and B2 , B3 , . . . , Bn be the points on a ir le su h that 3. In the a ute-angled triangle ABC AM and Æ
all angles are greater than 45 . Let
A1A2 : : :An is a regular m -gon and A1 B2 : : : Bn is a regular n -gon whereby BN be the heights of this triangle and let X and Y be the points on MA and
n > m and the point B2 lies between A1 and A2 . Find 6 B2 A1 A2 . NB , respe ively, su h that jMX j = jMB j and jNY j = jNAj . Prove that MN
and XY are parallel.
2. Find all positive integers n su h that

hni h i h i a , b , and
6= n2 + 23n :
4. Let be positive real numbers not greater than 2 . Prove the inequality
n+
6
ab
a+b+
6 4
:
x
Here [ ℄ denotes the largest integer not greater than x. 3

3. In the re tangle ABCD with jAB j < 2jADj , let E beÆ the midpoint of AB and 5. The game Clobber is played by two on a strip of 2 k squares. At the beginning
F a point on the hord CE su h that 6 CF D = 90 . Prove that F AD is an there is a pie e on ea h square, the pie es of both players stand alternatingly. At
isos eles triangle. ea h move the player shifts one of his pie es to the neighbouring square that holds

17 18
 a pie e of his opponent and removes his opponent's pie e from the table. The 12th grade
moves are made in turn, the player whose opponent annot move anymore is the
winner.
1. Jüri and Mari both wish to tile an nn hessboard with ards shown
in the pi ture (ea h ard overs exa tly one square). Jüri wants that for
Prove that if for some k the player who does not start the game has the winning ea h two ards that have a ommon edge, the neighbouring parts are of
strategy, then for k + 1 and k + 2 the player who makes the rst move has the dierent olor, and Mari wants that the neighbouring parts are always of the same
winning strategy.
olor. How many possibilities does Jüri have to tile the hessboard and how many
possibilities does Mari have?

px = log x .
2. Solve the equation 2

ABC be a triangle with 6 C = 90Æ and D a point on the ray CB su h that

11th grade
3. Let
jAC j
jCDj = jBC j2 . A parallel line to AB through D interse ts the ray CA at
1. E . Find 6 BEC .
Juhan is touring in Europe. He stands on a highway and wat hes ars. There are 4. Call a positive integer lonely if the sum of re ipro als of its divisors (in luding

three ars driving along the highway at onstant speeds: an Opel and a Trabant in 1 and the integer itself ) is not equal to the sum of re ipro als of divisors of any
one dire tion and a Mer edes in the opposite dire tion. At the moment when the other positive integer. Prove that a) all primes are lonely; b) there exist innitely

Trabant passes Juhan, the Opel and the Mer edes lie at equal distan es from him many non-lonely positive integers.

in opposite dire tions. At the moment when the Mer edes passes Juhan, the Opel
and the Trabant lie at equal distan es from him in opposite dire tions. Prove that 5. On a lottery ti ket a player has to mark 6 numbers from 36. Then 6 numbers from

at the moment when the Opel passes Juhan, also the Mer edes and the Trabant these 36 are drawn randomly and the ti ket wins if none of the numbers that ame

lie at equal distan es from him in opposite dire tions. out is marked on the ti ket. Prove that a) it is possible to mark the numbers on
9 ti kets so that one of these ti kets always wins; b) it is not possible to mark the

2. Prove that for all positive real numbers a , b , and numbers on 8 ti kets so that one of ti kets always wins.

p3 p
ab +
1
+
a b
> 1
2 3:+
1

Solutions of Final Round

When does the equality o ur?
 
9-1. Answer: .
3. Let ABC A1 , B1 , C1 points on BC , CA , AB , respe tively,
be a triangle and
m n
su h that the lines AA1 , BB1 , CC1 meet at a single point. It is known that A , O
B1 , A1 , B are on y li and B , C1 , B1 , C are on y li . Prove that a) C , A1 , Let be the enter of the ir le.
6 B2 A1 A2 = 6 B2 OA2=2 = (6 A1OA2 6 A1 OB2 )=2 = =m =n .
Then

C1 , A are on y li ; b) AA1 , BB1 , CC1 are the heights of ABC .

n
pn 9-2. Answer: n = 6k + 1 where k = 0, 1, . . .
p
4. Prove that there exist innitely many positive integers
an integer and n is divisible by [ n℄ . x
su h that is not
(Here [ ℄ denotes the largest integer not
If the equality holds (does not hold) for some n, then it holds (does
greater than x .) hn + 6i hni
not hold) for n + 6. Indeed, ( n + 6) + = n + + 7 and
6 6
hn + 6i h 2(n + 6) i hni h 2n i
5. For whi h positive integers n is it possible to over a (2n+1)  (2n+1) + = +
+ 7 . Therefore it su es to he k the values
hessboard whi h has one of its orner squares ut out with tiles shown 2 3 2 3
in the gure (ea h tile overs exa tly 4 squares; tiles an be rotated n = 0 , 1 , 2 , 3 , 4 , 5 . Doing so, we nd that n
= 1 is the only ase when the
and turned around)? equality is violated.

19 20
 9-3. Sin e E AB , the right-angled triangles EAD and EBC
is the midpoint of Let O be the ommon enter of the ir les and ABCDE one of the pentagons
are ongruent. 6 DAE = 6 DF E = 90Æ , the quadrilateral AEF D
Also, sin e whereby the inner ir le tou hes the side AB and the outer ir le passes through
is y li . Therefore 6 AF D = 6 AED = 6 BEC . On the other hand, D . Let Q be the enter of ABCDE . Then 6 AOB = 360Æ=10 = 36Æ and
6 ADF = 180Æ 6 AEF = 6 BEC . Hen e 6 AF D = 6 ADF . 6 ADB = 6 AQB=2 = 360Æ=10 = 36Æ . So the isos eles triangles AOB and ADB
are ongruent and the radius of the inner ir le is exa tly half of the radius of the
Solution 2. Let K CE and DA . Then EA is the midline
be the interse tion of
Æ
of KCD , therefore AK = AD . Sin e 6 KF D = 90 , the points K , F , and D
outer ir le.

lie on the ir le entered at A . This means AF = AD . 10-2. Answer: 2 and 2.

9-4. Answer: 2 : 3 . Using the fa t that mn divides m2 + n2 , we on lude that m divides n2 and n
divides m . Therefore m and n have the same prime divisors. Now let m = p m
2 a 0
When the bottle got lled rst time, the reindeer moss tea made up 140 ml of b 0 0 0
and n = p n where the prime p is not a divisor of m and n . If for example
its 350 ml volume. Therefore the mixture that was poured out onsisted of 2 5 = a > b , then mn and m2 are both divisible by pa+b but n2 is not (it is only
=
of reindeer moss tea and 3 5 of other substan es. Sin e the amount of liquid
divisible by p
2b
where 2b < a + b ). That is, (m + n )=(mn) is not an inte-
2 2

ger, a ontradi tion. So it must be a = b and onsequently m = n . Then

that was poured out of the bottle and the amount that was added afterwards are
equal and the bottle nally ontained 140 ml of reindeer moss tea again, also the
m2 + n2 2n2
=
mixture that was added was made up of exa tly 2 5 of reindeer moss tea. So the
amount of moonshine in that mixture was 3 5 . = mn
=
 n2 = 2 .
10-3. The quadilateral AXY B is y li sin e 6 AXB = 180
Æ 6 MXB = 135Æ and
9-5. Answer: a) yes; b) yes.
analogously 6 Æ
AY B = 135 . Also, the quadrilateral ANMB is y li . Therefore
a) See gure 9.
6 MXY = 180Æ 6 AXY = 6 ABN = 6 AMN whi h means that MN and XY
are parallel.
b) Figure 10 shows how to tile the border of width 2 around the (4 k 1)  (4k 1)
square (we use the fa t that 2  4 re tangle an be overed with two tiles). Starting p
10-4. First, from the inequality between arithmeti mean and geometri mean we get

from the tiling forn = 3 (its existen e an be seen from the gure 9) and applying a b >
+ + 3
3
ab
. Se ond, sin e 2, 2 , and a6
2 , we have b6 8 . Now 6 ab 6
this onstru tion 500 times, we get the tiling for n = 2003 . p p3
ab ab 3
ab )2
6 p
a + b + 3 3 ab
=
(
3
6 3
64
=
4
3
:

2  4k 10-5. If the length of the strip is 2( k + 1) , then at his rst move, the rst player beats
his opponent's pie e that stands at the end of the strip. This divides the pie es
into two se tions: one of length 2 k and the other of length 1 . Sin e the se ond
se tion annot hange, the situation is equivalent to playing the game on a strip
of length 2 k where the se ond player makes the rst move. So the rst player an

2  4k - follow the winning strategy.

k
If the length of the strip is 2( +2) , then at his rst move, the rst player beats his
opponent's pie e that stands on the third square from the end with his pie e that
stands on the fourth square from the end. After that the board again ontains two
se tions: one of length 2 k and the other of length 3 . If the se ond player makes
2  4k a move in the rst se tion, the rst player responds a ording to the winning
strategy. If the se ond player makes a move in the se ond se tion, the rst player
also makes his move in the se ond se tion whi h thereafter has only one pie e left.
Figure 9 Figure 10
11-1. Let t1 and t2 be the time instants when the Trabant and the Mer edes passed
10-1. Answer: 4. Juhan, respe tively. The Mer edes was at t1 some distan e d away from Juhan

21 22
 and by t2 it rea hed him. The Opel also was at t1 the distan e d away from length 4 m+2 whi h ontains 2 m+1 bla k squares and 2 m+1 white squares,
Juhan (on the other side) and at t2 its distan e to Juhan was the same as the the board altogether has an odd number of bla k squares and an odd number of
distan e the Trabant overed between t1 and t2 . This means that the velo ity of white squares. Ea h tile, not depending of its position, always overs 3 squares of
the Mer edes equals the sum of velo ities of the Opel and the Trabant. If t3 is the one olor and 1 square of other olor, hen e the total number of tiles must be odd.
time instant when the Opel rea hes Juhan, having overed the distan e d sin e On the other hand, the board has (4 m + 3)2 1 = 16 m2 + 24m + 8 squares. This
t1 , the Mer edes has overed the distan e d , passing Juhan at t2 in the opposite number is divisible by 8, so the number of tiles must be even, a ontradi tion.
dire tion, plus the distan e that the Trabant has overed between t1 and t3 .
p
11-2. Answer: equality o urs when a=b= = 3.
2  (4m 4)
Using twi e the inequality between arithmeti and geometri means, we get

s v s
u
p3 p3 u3 p
ab +
1
+
a b
> ab + 3 3 ab >
1
+
1 1
2 t ab 3 3
1
=
ab
p 
2  4m -

= 2 3 :
s
p3 1
The equality holds if and only if a = b = and ab = 3 3 . These onditions
ab
s
p3 1 3 p 2  (4m 4)
give a 3 =33 or a= . Hen e a=b= = 3.
a 3 a
11-3. a) LetAA1 , BB1 , and CC1 interse t at P . From the property of interse ting Figure 11 Figure 12

hords we have jP Aj
jP A1j = jP B j
jP B1j and also jP C j
jP C1j = jP B j
jP B1j . n

So jP Aj
jP A1 j = jP C j
jP C1j from whi h it follows that the points C , A1 , C1 ,
12-1. Answer: both Jüri and Mari have 4 possibilities.

and A are on y li . Let us pla e n ards on the n squares of main diagonal. This an be done in 4
n
b) We know that AB1 A1 B , BC1 B1 C , and CA1C1A are y li quadrilaterals. ways, sin e ea h ard an be in one of 4 positions. After that, the position of

Therefore 6 AB1 B = 6 AC1 C be ause 6 CB1 B and 6 CC1 B are equal. Now
other ards is uniquely determined for both Jüri and Mari: if we know the olors

6 AA1 B = 6 AB1 B = 6 AC1C = 6 AA1 C . Hen e 6 AA1 B = 90Æ . Analogously of two neighbouring edges of a square, the ard an be pla ed on that square only

6 BB1 C = 90Æ and 6 CC1A = 90Æ . in one way. So we an ll the whole board diagonal by diagonal.

11-4. The numbers m2 + m and m2 + 2m all have the desired property where m is
12-2. Answer: x=4 and x = 16 .
p p
an arbitrary positive integer. m2 + m and m2 + 2m
Indeed, the numbers
p
First note that 4 and 16 satisfy the equation. To show that there are no more
are not integers, sin e m < m + m < m + 2m < (m + 1) . In addition,
2 2 2 2
px ln 2
solutions, onsider the fun tion f (x) = x log2 x . Finding the derivative
p  p   2 2
m2 + m = m2 + 2m = m whi h is a divisor of both m2 + m and m2 +2m . f 0 (x) =
2
, we see that f (x) has only one extremum point x = .
2x ln 2 ln 2
11-5. Answer: all even numbers. Be ause f (x) is ontinuously dierentiable in its domain (0; 1) , it must have a
lo al extremum between ea h two zero points. Therefore f (x) annot have more
Figure 11 gives the tiling for n = 2 , gure 12 shows how to extend the tiling for
n = 2(m + 1) (we use the fa t that 2  4 re tangle an be
than 2 zero points.
n = 2m to the ase
overed with two tiles). 12-3. Answer: 45 .
Æ
Let now n = 2m + 1 . Let us olor the rows alternatingly bla k and white. Sin e Triangles EDC and ABC are similar with the similarity ratio
there are an even number of olumns of full length 4 m+3 and one olumn of k = jDC j=jBC j . From the given fa ts we get jDC j=jBC j = jBC j=jAC j .
23 24
 Now jEC j = jAC j
k = jAC j
jBC j=jAC j = jBC j . Triangle BCE is right-angled of remaining six ti kets, say on C and D. Now the drawn set an ontain 1 , 12
and isos eles, so 6 CEB = 45Æ . and one number from ea h of 4 remaining ti kets (not A , B , C , D ), and again,
none of the 8 ti kets wins.
12-4. a) For a primep , the sum of re ipro als of its divisors is (p + 1)=p . If there exists
another positive integer a 6= p whose sum of re ipro als of divisors is (p + 1)=p ,
then p must be one of a 's divisors. Then a is divisible at least by 1 , p , and a , IMO Team Sele tion Test: May 2003
and the sum of their re ipro als is greater than 1 + 1=p .

b) Suppose that there exist an integer a d1 , . . . , dk and a se ond

with divisors First Day
integer b with divisors e1 , . . . , el su h that 1=d1 + : : : + 1=dk = 1=e1 + : : : + 1=el .
Let p be a prime that is not a divisor of neither a nor b . Then ap has divisors d1 , 1. Two treasure-hunters found a treasure ontaining oins of value
. . . , dk , d1 p , . . . , dk p , and bp has divisors e1 , . . . , el , e1 p , . . . , el p . Computing a1 < a2 < : : : < a2003 (the quantity of oins of ea h value is unlimited).
their re ipro al sum, we get for ap and bp , respe tively: The rst treasure-hunter forms all the possible sets of dierent oins ontaining
   odd number of elements, and takes the most valuable oin of ea h su h set. The
1 1 1 1 1 1 1
+ :::+ + + :::+ = 1+ + :::+ se ond treasure-hunter forms all the possible sets of dierent oins ontaining
d1 dk d1p dk p p d1 dk even number of elements, and takes the most valuable oin of ea h su h set.
  
1 1 1 1 1 1 1 Whi h one of them is going to have more money and how mu h more? (H.
+ :::+ + + :::+ = 1+ + :::+
e1 el e1 p el p p e1 el Nestra )

The results are equal. So if there exist non-lonely integers a and b , then all 2. Let n be a positive integer. Prove that if the number 99 : : : 9}
| {z is divisible by n,
integers ap
bp where p is a prime not dividing a and b are also non-lonely.
and n
Now we an take a = 6 and b = 28 (in general, any two perfe t numbers), then then the number 11 : : : 1}
| {z is also divisible by n. (H. Nestra )

1 + 1=2 + 1=3 + 1=6 = 2 and 1 + 1=2 + 1=4 + 1=7 + 1=14 + 1=28 = 2 . n

12-5. a) Mark the numbers on 9 ti kets as follows:
3. Let N be the set of all non-negative integers and for ea h n 2 N denote n0 = n +1 .
The fun tion A : N ! N is dened as follows:
3

; ; ; ; ;
(1 2 3 4 5 6) ; ; ; ; ;
(10 11 12 13 14 15) ; ; ; ; ;
(19 20 21 22 23 24)
0
(i) A(0; m; n) = m for all m; n 2 N ;
; ; ; ; ; ; ; ; ; ; ; ; ; ; ; (
(1 2 3 7 8 9) (10 11 12 16 17 18) (25 26 27 28 29 30) n; if k = 0,
(4; 5; 6; 7; 8; 9) (13; 14; 15; 16; 17; 18) (31; 32; 33; 34; 35; 36) (ii) A(k0 ; 0; n) = 0; if k = 1, for all k; n 2 N ;
1; if k > 1
If none of the ti kets in the rst olumn wins, two of the six drawn numbers must
A(k ; m ; n) = A(k; A(k0 ; m; n); n) for all k; m; n 2 N .
0 0
belong to f1; 2; : : :; 9g . If none of the ti kets in the se ond olumn wins, two of
(iii)

the six drawn numbers must belong to f10; 11; : : :; 18g . If none of the ti kets in Compute A(5; 3; 2) . (H. Nestra )
the last olumn wins, three of the six drawn numbers must belong to f19; : : :; 36g .
Sin e only 6 numbers are drawn, all these onditions annot be satised at the
same time. Se ond Day

b) If some number is marked on 3 ti kets form 8, then the drawn set of numbers
4. A de k onsists of 2
n ards. The de k is shued using the following operation: if
an ontain this number and one number from ea h of 5 remaining ti kets, so none
the ards are initially in the order
of 8 ti kets wins. Now suppose that ea h number is marked on at most 2 ti kets.
All 8 ti kets together have 48 numbers marked but sin e there are only 36 dierent
a1 ; a2 ; a3 ; a4 ; : : :; a2n 1 ; a2n ;
numbers, at least 12 numbers must o ur twi e. Without loss of generality, let
they be 1 , 2 , . . . , 12 . Let us take two ti kets, say A and B, whi h both have then after shuing the order be omes
1 marked. They ontain 10 more numbers, so one of the numbers 2 , . . . , 12 is
not marked on either of them. Let this number be 12 . Then 12 is marked on two a2n 1 +1 ; a1 ; a2n 1 +2 ; a2 ; : : :; a2n ; a2n 1 :

25 26
 Find the smallest number of su h operations after whi h the original order of the b  1 mod p and 1 + b + : : : + bp 1
 1+1+
| {z
: : : + 1}  p
0 mod . Hen e,
ards is restored. (R. Palm ) p
(bp m = (b 1)m
(1 + b + : : : + bp
1)
1
) is divisible by ( b 1)mp = (b 1) n
5. Let a; b; be positive real numbers satisfying the ondition
1
+
1
+
1
= 1. and thus (b 1)n j b
n 1 as required.
ab a b
Prove the inequality
3. Answer: 65536 .
p k that for all integers k > 1 n
p a p b p
Solution. First we show by indu tion on and any
+ + 6 323 : the equality A(k; 1; n) = n holds. Indeed, if k = 2 , we have
1+ 2a 1+ 2 b 1+ 2

A(2; 1; n) = A(10 ; 00; n) = A(1; A(10; 0; n); n) =

When does the equality hold? (L. Parts )
0
= A(1; 0; n) = A(0 ; 0; n) = n :
6. Let ABC be an a ute-angled triangle, O its ir um enter and H its ortho enter.
The orthogonal proje tion of the vertex A to the line BC lies on the perpendi ular Assuming now that A(k; 1; n) = n holds for some k > 1 . Then
AC . Compute
jCH j . (J. Willemson ) A(k0 ; 1; n) = A(k0 ; 00; n) = A(k; A(k0; 0; n); n) = A(k; 1; n) = n ;
bise tor of the segment
jBOj
hen e we have the required equality for k0 = k + 1 and the indu tion is omplete.
In a similar way we an use indu tion on m to prove that for any natural numbers
Solutions of Sele tion Test m and n the equalities A(1; m; n) = m+n , A(2; m; n) = mn and A(3; m; n) = nm
hold.
1. Answer: The rst treasure-hunter gets one heapest oin more than the se ond Next we use indu tion on k to show that for any integer k > 0 the equality
one. A(k; 2; 2) = 4 holds. Indeed: for k = 1 we have A(1; 2; 2) = 2 + 2 = 4 . If
Solution. Take all the odd oin sets of the rst treasure-hunter and form an even A(k; 2; 2) = 4 for some k > 0 we get k0 > 1 and
A(k0 ; 2; 2) = A(k0 ; 10; 2) = A(k; A(k0 ; 1; 2); 2) = A(k; 2; 2) = 4 :
oin set from ea h one of them by the following rule. If the odd set ontains the
heapest oin then leave it out, otherwise add the heapest oin to the set. This
way we obtain exa tly all the sets of the se ond treasure-hunter plus one empty
oin set. It is lear that the des ribed operation does not hange the most valuable
We also note that for any natural k
oins of the sets ex ept for the set having originally only one heapest oin. Hen e,
A(k0 ; 3; 2) = A(k0 ; 20; 2) = A(k; A(k0 ; 2; 2); 2) = A(k; 4; 2) :
the rst treasure-hunter gets one heapest oin more than the se ond one.

n j 11 n j 99
Finally we ompute
2. The ondition of : : : 1}
| {z is equivalent to 9 : : : 9} .
| {z We will prove a more
n n A(5; 3; 2) = A(40 ; 3; 2) = A(4; 4; 2) = A(30; 30; 2) = A(3; A(30; 3; 2); 2) =
general result for any positional number system. Namely, we will show by indu -
24
= A(3; A(3; 4; 2); 2) = 2 = 65536 :
tion on n that for any positive integers n and b the ondition n j bn 1 implies
(b 1)n j b
n 1 . Taking b = 10 gives the desired result.
For n = 1 the laim is true as b
n 1 = (b 1)(bn 1 + : : : + b + 1) . Now assume 4. Answer: At least 2 n shuings are needed.
that for all numbers less than n and any b the laim holds. Take any positive
n 1 . Consider two possible ases.
Solution. Assume the ards are labeled by numbers ; ; : : : ; 2n . We note that
1 2
integer b and assume n j b n
the position of the ard number x after the rst shuing is f (x) = 2x mod 2 + 1 .
a) If n and b 1 are oprime, the onditions n j b
n 1 and b 1 j bn 1 imply k k n
After k th shuing its position be omes f (x) = 2 x mod 2 + 1 . Our task is to
(b 1)n j b
n 1. k
nd the least number k su h that for ea h x the equality f (x) = x holds, or
b) If n and b 1 have a ommon prime fa tor p , let n = mp ; then equivalently, 2
k  1 mod 2n + 1 .
bn 1 = bmp 1 = (bp )m 1 . As m < n , we an use the indu tion hy- Let k = 2n . Sin e
pothesis for m : as (b )
p m 1 = bn 1 is divisible by n = mp and hen e
also by m , we have (b
p 1)m j (bp)m 1 = bn 1 . Sin e p j b 1 we get 2
2 n  (2n + 1)2 n
2(2 + 1) + 1  1 mod 2n + 1 ;
27 28
 n th shuing the order of the ards is restored. Assume that for some
then after 2 ongruent. Indeed, 6 AB 0 B = 90Æ = 6 HB 0 C , jBB 0 j = jB 0 C j and jAB 0j = jB 0H j
m < 2n the order is restored after m th shuing. In this ase we must have (sin e AB 0 H is also a right isos eles triangle). This ompletes the proof.
m > n , sin e the ard number 1 only rea hes the position 2n after n th shuf-
ing. Subtra ting now the ongruen e 2
m  1 mod 2n + 1 from the ongruen e
2
2 n  1 mod 2n + 1 , we obtain
m
2 (2
2n m 1)  0 mod 2n + 1 :
This is the ontradi tion as 2
m n
and the modulus 2 + 1 are oprime and the value
of 2
2n m 1 is less than the same modulus be ause of m > n. Hen e, k = 2n is
the least number of shuings restoring the original order of ards

p
5. Answer: the equality holds i a=b= = 3.
Solution. The given equality is equivalent to ab = a + b + . Sin e a; b; > 0 we

an nd su h 0 < ; ; < that a = tan  , b = tan  , = tan . Using the
2
tan x + tan y
equality tan(x + y ) = repeatedly, we obtain tan( +  + ) = 0
1 tan x tan y
tan2 x
or  +  + =  . Using now the equality = sin x and taking into
2
1 + tan x
2

a ount that the angles  ,  and are a ute, we an transform the required
inequality to the form

p
 + sin  + sin 6
3 3
sin :
2

Sin e + + =  and the sine fun tion is on ave in the segment [0 ; ℄ , we an

use Jensen's inequality to obtain

p

6 sin ( +  + ) = sin
1 1 1 1 3
sin + sin + sin = ;
3 3 3 3 3 2

whi h implies the required inequality.

 p We also see that the equality holds i

== = , or a=b= = 3.
3
p
6. Answer: 2.
Solution. Let A0 and B0 be the base points of the heights drawn from the verti es
A and B , respe tively. From the onditions of the problem we on lude that
0
triangle AA C is an isos eles right triangle with the right angle at the vertex
A . Thus, A0 AC = 6 A0 CA = 45Æ . Hen e the triangle BB 0 C is also right and
0 6
Æ Æ
isos eles. We have 6 AOB = 26 ACB = 2
45 = 90 and the triangle AOB is
j CH j p
right and isos eles as well. In order to prove that
jBOj = 20 it is enough
0
to

prove that jAB j = jCH j . We will show that the triangles ABB and HCB are

29 30
Estonian Math Competitions
2003/2004

The Gifted and Talented Development Centre

Tartu 2004
 second day, the first Liarian answered that he lied on exactly 2 days a year, the second,
that he lied on at least 2 days a year, and the third, that he lied on at most 2 days a year,
Selected Problems from Open Contests and so on, until the last, 2004th day of the year, when the first answered that he lied
on exactly 2004 days, the second said that he lied on at least 2004 days, and the third
replied that he lied on at most 2004 days a year. How many days a year does each of the
1. Diameter AB is drawn to a circle with radius 1. Two straight lines s and t touch the three Liarians actually lie? (Juniors.)
circle at points A and B , respectively. Points P and Q are chosen on the lines s and t,
Answer: the first Liarian lies on days, the second on days and the third on
respectively, so that the line P Q touches the circle. Find the smallest possible area of the
2003 1002 0

days a year.
quadrangle AP QB . (Juniors.)
Solution. Since the answers of the first Liarian are all mutually exclusive, he could tell
Answer: 2. the truth on at most one day. If he had lied on all days, the answer he gave on the 2004th
Solution. As ABCD is a trapezium, its area can be found as follows: day would have been true — a contradiction. Thus, the first Liarian tells the truth on
one day (the 2003th day of the year) and lies on the remaining 2003 days of the year.
SAP QB =
jAP j jBQj
jAB j jAP j jBQj :
+
= + Suppose that the second Liarian lies on n days of a year, then his answers are true on the
2
first n days and false on the remaining 2004 n days. Hence n = 2004 n, and n = 1002
On the other hand, jAP j = jP K j and jBQj = jQK j, thus SAP QB = jP K j jQK j jP Qj.
+ = (the second Liarian lies on the last 1002 days of the year).
The smallest possible length of the segment P Q is 2. Suppose now that the third Liarian lies on n days of a year, where n  1. Then the
answers he gave on the first n 1 days are false and the remaining answers (starting
2. Find all pairs of positive integers (m; n), m  n, such that a rectangle of size m  n from the nth day) are true. Thus he lies on n 1 days a year — a contradiction. However,
can not be divided into rectangles of sizes 2  5 and 1  3. (Juniors.) if he lies on 0 days a year, all his answers are true and there is no contradiction. Hence
Answer: (2; 2), (2; 4), (4; 4), (2; 7) and (1; 3k ,
1) (1 3 ; k 2) , where k is any positive integer. the third Liarian lies on 0 days a year.
Solution. The answer is obvious for rectangles with one side of length 1. Hence, we need 4. Circles 1 and 2 with centres O1 and O2 , respectively, intersect at points A and B
to consider the rectangles with both of the sides longer than 1. so that the centre of each circle lies outside the other circle. Line O1 A intersects circle 2
All the rectangles 3  m can be divided into 1  3 pieces. All the rectangles 5  (2k ) can be again at point P2 and line O2 A intersects circle 1 again at point P1 . Prove that the points
divided into 2  5 pieces. All the rectangles 5  (2k + 1) can be divided into k 1 pieces O1 , O2 , P1 , P2 and B are concyclic. (Juniors.)
of 2  5 and 5 pieces of 1  3. Thus, all the rectangles with one side having the form
3s + 5t, where s; t  0, can be divided in the required way. Since 6 = 3
2, 8 = 3
1 + 5
1,
Solution 1. Triangles AO1 P1 and AO2 P2 are isosceles and have equal base angles, thus
9 = 3
3 and 10 = 5
2, every positive integer n > 7 can be represented as 8 + 3k , 9 + 3k
also their vertex angles are equal. Now
or 10 + 3k for some integer k  0. Hence we are done, if one of the sides of the original \P O P
1 1 2 = \P O A \AO P
1 1 = 2 2 = \P O P 1 2 2
rectangle is 3, 5, 6 or n > 7 and the length of the other side is at least 2.
(See figures 2 and 3 for the possible positions of points P1 and P2 , according to the angle
We still need to consider the rectangles with both sides from \O1P1O2 = \O1P2 O2 being smaller or larger than a right angle). Also,
the set f2; 4; 7g. The rectangles 2  2, 2  4 and 4  4 cannot
be divided, as they cannot contain a 2  5 piece, but their \P BP \P BA \ABP \P O A \AO P \P O P : 1 1
1 2 = 1 + 2 = 1 1 + 2 2 = 1 2 2
area is not divisible by 3. A similar argument holds for 2  7 2 2

rectangle — cutting away either one or zero 25 pieces leaves Hence the angles \P O P , \P O P and \P BP lying on segment P P are equal, and
1 1 2 1 2 2 1 2 1 2

a part with area not divisible by 3. the points O1 , O2 , P1 , P2 and B are concyclic.
Figure 1 It is easy to construct a required division for 4  7 rectangle
P1 P2
(see Figure 1), hence 7  7 rectangle can be divided as well. A
1 A 2 1 2
3. On Liarians planet, a year consists of 2004 days and for each inhabitant on the P2 P1
planet, each day of the year is either a truth-day or a lie-day: on truth-days a Liarian
speaks only the truth and on lie-days the Liarian always lies (the number of a Liarian’s O1 O2 O1 O2
truth-days or lie-days can also be 0). On each day of a given year, three Liarians were
B
asked, “How many days a year do you lie?” On the first day of the year, the first Liarian B
answered that he lied on exactly 1 day a year, the second Liarian said that he lied on
Figure 2 Figure 3
at least 1 day a year, and the third replied that he lied on at most 1 day a year. On the

1 2
Solution 2. Since (1) ABCD is not cyclic;
(2) the sides AB , BC , CD and DA have pairwise different lengths;
\O P A \O AP
1 1 = 1 1 = \O AP2 2 = \O P A ;2 2
(3) the circumradii of the triangles ABC , ADC , BAD and BCD are equal?
points O1 , P1 , P2 and O2 lie on the same circle. On the other hand, since
b) Does there exist such a non-convex quadrangle? (Seniors.)
\O P O \O BO \O AP \O AO Æ
1 1 2 + 1 2 = 1 1 + 1 2 = 180
Answer: a) no; b) yes.
(for the case shown on figure 2) or Solution. a) Assume the quadrangle ABCD satisfies the conditions of the problem. Let
\O P O \O BO \180Æ \O P A \O AO R be the common circumradius of triangles ABC , ADC , BAD and BCD. The Sine law
2 + = + 2 =
for triangles ABC and BAD gives
1 1 1 2 1 1 1

= \180Æ \O AO \O AO
1 2 + 1 2 = 180
Æ
jAB j R=
jAB j
(for the case shown on figure 3), also points O1 , P1 , O2 and B lie on the same circle.
These two circles coincide, since both of them are the circumcircle of triangle O1 P1 O2 .
sin \ADB ;
\ACB = 2
sin

Solution 3. We note that jO1 B j = jO1 P1 j, jO2 B j = jO2 P2 j, \AO1 O2 = \BO1 O2 ja \AO2 O1 =
or \ACB
sin \ADB . If \ACB \ADB , the quadrangle ABCD would be cyclic,
= sin =

hence \ACB \ADB  (see Figure 4). Similarly we find \DAC \DBC  , but
\BO2O1. Letting R0 , R1 and R2 be the radii of the circumcircles of triangles O1O2B , + =

then the sum of interior angles of ABCD would be greater than 2 , a contradiction.
+ =

O1 O2 P1 and O1 O2 P2 , and applying the Sine law, we get

b) Let ABD be an isosceles triangle with jAB j = a and jBD j = jDAj > 2a. Let C be
R1
jO P j
1 1 jO P j
1 1 jO B j
1
R0 = the orthocenter of the triangle ABD and let B1 and D1 be the feet of the perpendiculars
2 =
sin \P O O1 2 1
=
sin \AO O 2 1
=
sin \BO O 2 1
= 2
drawn from the vertices B and D to the opposing sides (see Figure 5).
jO B j
2 jO P j
2 2 jO P j
2 2
R2 : Denote \ADC = . As the triangles ADD1 and ABB1 are right with a common acute
=
sin \BO O 1 2
=
sin \AO O 1 2
=
sin \P O O 2 1 2
= 2
angle, they are similar and \ABC = \ABB1 = \ADD1 = \ADC = . Due to symme-
try we also have \BAC = \BDC = . As the triangles ADC and BDC are congruent,
Hence triangles O1 O2 B , O1 O2 P1 and O1 O2 P2 share a common side O1 O2 and have equal
their circumradii are equal — let it be R. The Sine law for triangles ABC and BDC gives
radii of circumcircles.
If \O1 P1 O2 is larger than a right angle, then all the midperpendiculars of the triangles jCB j jCB j jCB j
intersect at the same side of line O1 O2 as point B . If \O1 P1 O2 is smaller than a right RABC = R;
2 sin \CAB
 2 sin \CDB
=
2 sin
= =

angle, then all the centres of the circumcircles lie on the other side of line O1 O2 from
point B . Thus, in both cases the triangles share a common circumcircle. and the Sine law for triangles BAD and ABC gives
Remark. The claim of the problem is actually true without assuming that the centres of
RBAD =
jAB j jAB j jAB j jAB j R:
the circles lie outside the other circle. 2 sin \ADB =
2 sin 2 
=
2 sin(  2)
=
2 sin \ACB =

5. Let R + be the set of all positive real numbers. Find all functions f : R + !R +
, such D
B
that for every x; y 2 R + the equality
 
x C
y
f (x) = f
2  C 
y D0 
 D
holds. (Seniors.)
a B0
Answer: All functions of the form f (x) = , where a 2 R + . A
x2 B A
1
Solution. Taking x = 1, we obtain y 2 f (1) = f . Substitution z =
1
gives f (z ) = Figure 4 Figure 5
y y
f (1) Thus we have shown that the circumradii of the triangles ABC , ADC , BAD and BCD
. Substituting this to the original equation we see that f (1) can be any positive real
z2 are equal. We still have to prove that the lengths of the sides of ABCD are all different.
number.
Indeed, as AB is the hypotenuse of ABB1 , we have jAB j > jBB1 j > jBC j. Similarly,
6. a) Does there exist a convex quadrangle ABCD satisfying the following conditions

3 4
 ADD we have jADj > jDD j > jCDj. Since jBDj > a, jBC j < a and
in triangle 1 1 2 hence y < . On the other hand, if ; x; y are real numbers satisfying x; y  0 and y < ,
jBDj < jBC j jCDj, we get
+ there exists a triangle with sides + x + y , + x and (since all the triangle inequalities
are satisfied). We will show that numbers less than 63 can not be represented in two
jCDj > jBDj jBC j > a a a jAB j : 2 = =
different ways in the form s( +x+y; +x; ), where x; y; are integers and x; y  0,
y < (disregarding the order of the numbers). Note that when x increases, the value
Thus jDAj > jCD j > jAB j > jBC j as required. As a non-convex quadrangle can not be
of the expression s( +x+y; +x; ) increases as well. Taking this into account, we will
cyclic, all the conditions for the quadrangle ABCD are satisfied.
consider all the possibilities for .
7. Find all positive integers n such that the number (1) If = 1, we have y = 0 and for x = 0; 1; 2; 3, we have 3; ; 35; 63 as the values for
15

the expression s( +x+y; +x; ), respectively.

n1 n2 nn 1
nn (2) Let = 2. If y = 0, then for x = 0; 1 the expression s( +x+y; +x; ) takes the values
+ + :::+ +
(n n! 48 and 128, respectively. If y = 1, then for x = 0 we have s( +x+y; +x; ) = 63.
1! 2! 1)!

is an integer. (Seniors.) (3) Let  3 . As x; y  0 and y < , we obtain s( +x+y; +x; )  3
1

= 3 3 

81 > 63.
Answer: n = 1; 2; 3.
Hence we have covered all the cases when s( +x+y; +x; ) can be less than 63. We have
Solution. Suitability of the solutions 1, 2 and 3 can be verified directly. We will prove also seen that
that there are no other solutions.
Let n  2, then the given expression can be written as: s(4; 4; 1) = s(3; 2; 2) = 63
p p
n1 n2 nn 2 nn 1 nn 63 3 7
+ + ::: + + + = and the respective area is = .
1! 2! (n 2)! (n 1)! n! 4 4

n
(n 1)! n
3
4
: : :
(n 1)
2
= + + ::: + 9. Find all functions f that are defined on the set of positive real numbers, have real
(n (n
1)! 1)!
values and satisfy for all positive real numbers x and y the equation
nn 2
(n 1) nn 1 nn 1
+ + + =
(n (n (n
1 1
1)! 1)! 1)!
f (x)f (y ) = f (xy ) + + :
n
(n 1)! + n2
3
4
: : :
(n 1) + : : : + nn 2

n
( 1) + 2
n n 1 x y
= :
(n 1)! (Seniors.)
As the first n 1 terms in the numerator and the denominator are divisible by n 1, the Answer: f (x) = 1 + .
1

last summand 2
nn 1 has to be divisible by n 1 as well. As the numbers n 1 and x

nn 1 are coprime, n 1 must divide 2, hence n = 2 or n = 3. Solution. Taking y = 1, we get for each x

8. Find the smallest real number x for which there exist two non-congruent triangles f (x)f (1) = f (x) +
1
+ 1 : (1)
with integral side lengths having area x. (Seniors.) x
p Taking also x = 1, we obtain a quadratic equation f (1)2 f (1) 2 = 0, that has solutions
f (1) = 1 and f (1) = 2. Thus f (1) 1 6= 0 and from the equation (1) we get
3 7
Answer: .
4

Solution. Denote    
f (x) =
1

f (1)

1 +
1

x
:
s(a; b; ) = (a + b + )(b + a)( + a b)(a + b ) : 1

 
a triangle with sides a, b and can be found from the Heron’s formula and
The area ofp Now either f (x) =
1

2

1+
1

x
1
or f (x) = 1 + . The second solution satisfies the initial
x
s(a; b; ) 1
is equal to . W.L.O.G. we may assume that b = + x and a = b + y = + x + y, condition, whereas the first one does not for e.g. x = y = .

4 2
where x; y 0 . Then
10. There are N lightbulbs on a circle, labelled clockwise with numbers 1 to N . Ini-
s(a; b; ) = s( + x + y; + x; ) = tially, none of the lightbulbs are lit. Then the following operation is performed for each
= (3 + 2x + y )( y )( + y )( + 2x + y ) ; positive divisor d of the number N (1 and N included): starting from bulb number 1

5 6
and moving clockwise, the state of each d-th bulb is changed i.e. it is lit, if it is off, and Selected Problems from the Final Round of National
switched off, if it is lit, and this is repeated exactly N times. (E.g. for N = 6 and d = 3,
the bulbs are lit or switched off in the following order: 3; 6; 3; 6; 3; 6.) Olympiad
For which values of N will all bulbs be lit, after the procedure has been completed for
all divisors of N ? (Seniors.)
1. Find all pairs of real numbers (x; y ) that satisfy the equation
Answer: N = 2
k
, where k is any non-negative integer.
x + 6 13 4 y
Solution. We note that for a given divisor d of the number N , all states of bulbs having + = :
N y xy x
number divisible by d are changed, and only these. Since there are such bulbs and
d (Grade 9.)
exactly N changes are made, the state of each bulb is changed exactly d times. In partic-
ular, for d = 1 the state of each of the N bulbs is changed once. Answer: x= 3 ,y = 2 .
To have all bulbs lit by the end of the procedure, the state of each bulb must be changed Solution. Multiply the equation by xy to obtain
an odd number of times. As we showed, the state of bulb m is changed once for divisor
1, and additionally, for each such divisor d > 1, that is also a divisor of m, the state (x + 6)
x + 13 = (4 y )
y
is changed d times. Obviously, the condition is satisfied when all divisors d > 1 of N
are even, i.e. N = 2k , where k  0. Suppose now that the number N has odd divisors or equivalently
greater than 1, and let d be the smallest of them. Then the state of bulb number d is x2 + 6x + y 2 4 y + 13 = 0 :
changed once for divisor 1, d times for divisor d, and 0 times for all other divisors of N ,
since d is not divisible by them. Hence the state of bulb d is changed an even number of Notice that the left side can be expressed as sum of squares:
times and the bulb is not lit by the end of the procedure.
(x + 3)2 + (y 2)
2
= 0 ;
11. On the circumcircle of triangle ABC , point P is chosen, such that the perpendic-
ular drawn from point P to line AC intersects the circle again at a point Q, the perpen- giving x + 3 = 0 and y 2 = 0 , or x = 3 and y = 2 .
dicular drawn from point Q to line AB intersects the circle again at a point R and the
perpendicular drawn from point R to line BC intersects the circle again at the initial 2. The positive differences ai aj of five different positive integers a1 , a2 , a3 , a4 , a5 are
point P . Let O be the centre of this circle. Prove that \P OC = 90Æ . (Seniors.) all different (there are altogether 10 such differences). Find the least possible value of
the largest number among the ai . (Grade 9.)
Solution 1. Rotate triangle P QR clockwise 90Æ and denote
the triangle obtained P 0 Q0 R0 . Since P Q ? AC , QR ? AB A Answer: 12 .

q
P
and RP ? BC (see figure 6), P 0 Q0 k AC , Q0 R0 k AB and
Solution 1. Say a1 < a2 < a3 < a4 < a5 . Since the 10 positive differences ai aj are all
R0 P 0 k BC . Hence triangle Q0 R0 P 0 is similar to triangle different, the largest a5 a1 must be greater than or equal to 10, giving a5  . In the
11

ABC . Since the triangles ABC and Q0 R0 P 0 share a com- case a5 = 11 the differences must be exactly 1; 2; : : : ; 10, hence
mon circumcircle, they are actually congruent and either Q
O : : : + 10 =
coincide or are rotated 180Æ from the midpoint of the cir-

55 = 1 + 2 +

B C a a4 ) + (a5 a3 ) + (a5 a2 ) + (a5 a1 ) +

cle. In the first case, C = P 0 , and in the second case, points
= ( 5

R + (a4 a3 ) + (a4 a2 ) + (a4 a1 ) +

C and P 0 are endpoints of the same diameter of the tri- + (a3 a2 ) + (a3 a1 ) +
angle. Rotating triangle P 0 Q0 R0 back to triangle P QR, we
+ (a2 a1 ) =
obtain in both cases that \P OC = 90Æ .
_ _
Figure 6
= 4a5 + 2a4 2a2 4a1 :

P C AQ Æ . Analo-
_ _ _ _
+
Solution 2. Since chords AC and P Q intersect at a right angle, This is impossible, because 4a5 + 2a4 2a2 4a1 is even.

_ _ _
= 90
2
We have obtained that the largest of the numbers must be larger than or equal to 12. For
AQ + RB Æ and RB + P C Æ . These equations give instance, we may choose the numbers 1; 3; 8; 11; 12 (the differences ai aj are therefore
gously, = 90 = 90 P C AQ RB = = =
Æ
90 .
2 2 1 = 12 11, 2 = 3 1, 3 = 11 8, 4 = 12 8, 5 = 8 3, 7 = 8 1, 8 = 11 3, 9 = 12 3,

10 = 11 1 and 11 = 12 1).

Solution 2. Construct the example for the largest number 12 similarly to the previous
solution and notice that if the largest of numbers ai equals 11, the differences in question

7 8
equal 1; 2; : : : ; 10. Therefore exactly 5 of these differences are odd. Denote by x and y Therefore x0 y 0 z 0 = 20 and, since x0 , y 0 and z 0 are pairwise coprime, they are equal to (in
the number of odd, resp. even, numbers among the ai . The difference of two integers some order) 1; 1; 20 or 1; 4; 5. Requiring x < y < z for the numbers x = 24x0 , y = 30y 0
is odd iff the numbers are of different parity. There are xy possibilities to choose a pair and z = 40z 0 , the first case implies x0 = 1, y 0 = 1 and z 0 = 20, while the second case gives
of integers of different parity, hence there are xy such odd differences. So x and y must two possibilities: x0 = 1, y 0 = 4, z 0 = 5, or x0 = 1, y 0 = 5, z 0 = 4. The triples (x; y; z )
satisfy the system of equations are (24; 30; 800), (24; 120; 200) and (24; 150; 160), respectively. The first and the third
 triple satisfy the conditions of the problem, but the second triple does not, because the
x+y =5 greatest common divisor of y = 120 and z = 200 is 40 rather than 10.
;
xy = 5 Solution 2. Let p . a denote the exponent of prime p in the canonical form of number a.
which has no positive integral solutions. First, consider the exponents of 2: the conditions of the problem imply

3. Three different circles of equal radii intersect in point Q. The circle C touches all of min(2 . x; 2 . y ) = 2 .6=1; (2)
them. Prove that Q is the center of C . (Grade 9.) min(2 . x; 2 . z ) = 2 .8=3; (3)


max(2 . x; 2 . y; 2 . z ) 2 . 2400 = 5 : (4)
Solution. Consider an arbitrary point P inside the circle C
=

qq q
different from the center point of C . Fix r > 0. We show that The equation (3) gives us 2 . z  3 and 2 . x  3. The equation (2) now implies that


there exist at most two circles of radius r that touch C from 2 . y = 1, and from (3) and (4) we obtain that one of the exponents 2 . x and 2 . z equals
inside and pass through P . For this roll the circle of radius 3 and the other is 5. Altogether we now have two possibilities:
r inside C ; evidently the circle being rolled passes through (2.x = 3 (2 . x = 5
point P in at most two positions (or equivalently: any fixed P r
circle that touches C from inside has at most two common 2 .y =1 ; 2 .y =1 : (5)
points with such a circle that passes through P and has the 2 .z =5 2 .z =3
centre point in the centre point of C — see figure 7). Hence no Consider now the exponents of 3:
such point P can be the intersection point of three different Figure 7
circles of different radii, all touching the circle C . min(3 . x; 3 . y ) = 3 .6=1; (6)
min(3 . x; 3 . z ) = 3 .8=0; (7)
4. Find all triples of positive integers (x; y; z ) satisfying x < y < z, x; y ) ,
max(3 . x; 3 . y; 3 . z ) 3 . 2400 = 1 :
g d( = 6
(8)
y; z ) = 10, g d(z; x) = 8 and l m(x; y; z ) = 2400. (Grade 10.)
g d(
=

Answer: ; 30; 800) and (24; 150; 160).

(24 The equations (6) and (8) imply that 3 .x = 3 .y = 1 , and the equation (7) gives that
3 . z = 0. Therefore
Solution 1. As 6 and 8 both divide x, l m(6; 8) = 24 divides x. Similarly l m(6; 10) = 30 (3.x = 1
divides y and l m(10; 8) = 40 divides z . Hence there exist positive integers x0 , y 0 , z 0 , that
x = 24x0 , y = 30y 0, z = 40z 0 , and 3 .y =1 : (9)
3 .z =0
6 = g d(x; y ) = g d(24x0 ; 30y 0) = 6
g d(4x0 ; 5y 0) ; Finally, consider the exponents of 5:
g d(y; z ) = g d(30y ; 40z ) = 10
g d(3y ; 4z ) ;
10 =
0 0 0 0
g d(z; x) = g d(40z ; 24x ) = 8
g d(5z ; 3x ) :
8 =
0 0 0 0 min(5 . x; 5 . y ) = 5 .6=0; (10)
min(5 . y; 5 . z ) 5 . 10 = 1 ; (11)
These equations imply g d(4x0 ; 5y 0 ) = 1, g d(3y 0 ; 4z 0 ) = 1 and g d(5z 0 ; 3x0 ) = 1. Hence
=

max(5 . x; 5 . y; 5 . z ) 5 . 2400 = 2 :
x0 ; y 0; z 0 are pairwise coprime and g d(x0 ; 5) = 1, g d(y 0 ; 4) = 1 and g d(z 0 ; 3) = 1. = (12)
We show now that l m(24x0 ; 30y 0 ; 40z 0 ) = l m(120x0 ; 120y 0 ; 120z 0 ). For this it suffices to The equation (11) gives us 5 .y  1 and 5 .z  1. The equation (10) now implies 5 .x = 0
prove that either side of the equation divides the other. Evidently lhs divides rhs. For and from (11) and (12)we obtain that one of the exponents 5 . y and 5 . z equals 1 and
the other direction, prove that 120x0 , 120y 0 , 120z 0 all divide lhs. Indeed, 24x0 and 5 divide the other is 2. Altogether there are again two possibilities:
lhs and, since g d(x0 ; 5) = 1, we have g d(24x0 ; 5) = 1 giving that 24x0
5 = 120x0 divides (5.x = 0 (5 . x = 0
lhs. Similarly prove that 120y 0 and 120z 0 divide lhs as well. Now
5 .y =1 ; 5 .y =2 : (13)
2400 = l m(x; y; z ) = l m(24x0 ; 30y 0; 40z 0 ) = 5 .z =2 5 .z =1
l m(120x ; 120y ; 120z ) = 120
l m(x ; y ; z ) = 120
x y z :
=
0 0 0 0 0 0 0 0 0

9 10
Since no other prime divides l m(x; y; z ) = 2400, no other prime divides x, y , z either. (1) invert the order of the letters in the word;
The conditions (5), (9) ja (13) now give altogether four possibilities: (2) replace two consecutive letters: BA ! UU, AU ! BB, UB ! AA, UU ! BA, BB !
( ( AU or AA ! UB.
x = 2
3
5 = 24
3 1 0
x = 2
3
5 = 24
3 1 0

y = 21
31
51 = 30 ; y = 21
31
52 = 150 ;
( z = 2 5
3 1
5 0 = 800 ( z = 2 5
3 1
5 0 = 160
5 0 2 5 0 1 Given that BBAUABAUUABAUUUABAUUUUABB is a BAU word, does BAU have
x = 2
3
5 = 96 x = 2
3
5 = 96 a) the word BUABUABUABUABAUBAUBAUBAUB?
y = 21
31
51 = 30 ; y = 21
31
52 = 150 :
z = 23
30
52 = 200 z = 23
30
51 = 40 b) the word ABUABUABUABUAUBAUBAUBAUBA?

We may check that all the conditions of the problem, except x < y < z , are satisfied, (Grade 11.)
which leaves us the two former possibilities.
Answer: a) no; b) yes.
5. In the beginning, number 1 has been written to point (0; 0) and 0 has been written Solution 1. a) Define the value of a word w as the number b a where b and a are the
to any other point of integral coordinates. After every second, all numbers are replaced numbers of B s and As, respectively, in w . Note that any allowed operation preserves
with the sum of the numbers in four neighbouring points at the previous second. Find the value of word modulo 3. The value of the known BAU word is 7 8 = 1 and the
the sum of numbers in all points of integral coordinates after n seconds. (Grade 10.) value of the word under consideration is 9 8 = 1, so the latter cannot belong to BAU.
Answer: 4
n
. b) At first, note that the allowed operations enable to interchange any two consecutive
letters of a word. Indeed, if these letters are BA, AU or UB then replace them with UU,
Solution. At the initial moment, the sum of numbers in points of integral coordinates is BB or AA, respectively, then invert the order of letters, then replace the letters UU, BB
1 = 4 . After every second, any number x on the plane contributes to the sum of the
0
or AA back, and finally invert the order once more. If the letters to be interchanged are
next second as 4x, hence the total sum of numbers increases 4 times at every step, giving AB, UA or BU then perform all these steps in reverse order.
4 after n seconds.
n
Iteration of the interchanging operation enables to interchange any two letters (move
the second letter side by side with the first, interchange them, and finally move the first
6. Real numbers a, b and satisfy a2 + b2 + 2 = 1 and a3 + b3 + 3 . Find a + b + .
= 1
letter to the original place of the second). Note that the known BAU word contains 7
(Grade 10.)
letters B, 8 letters A and 10 letters U, the word under consideration has 8 letters B, 9
Answer: 1. letters A and 8 letters U. Thus we can replace some consecutive letters UU with BA and
Solution. As a2 + b2 + 2 = 1, we have 1  a; b;  1, implying a2  a3 , b2  b3 and then reorder the letters in the word, obtaining the required word.
2  3 . Since a2 + b2 + 2 = a3 + b3 + 3 , all three inequalities must have equality: a2 = a3 , Solution 2. a) Part b) of the previous solution implies that a word w belongs to BAU if
b2 = b3 and 2 = 3 . Therefore a; b; can only be equal to 0 or 1 and exactly two of them and only if there is a word w 0 in BAU containing any letter the same number of times as
must be 0 and the third one 1 because a2 + b2 + 2 = 1. All cases imply a + b + = 1. w. Suppose the right numbers of letters can be obtained from the known BAU word by a
sequence of the allowed operations. Let x be the difference of the number of performed
7. Find all functions f which are defined on all non-negative real numbers, take non- replacements BA ! UU and the number of the performed replacements UU ! BA;
negative real values only, and satisfy the condition analogously define y and z for the rules AU ! BB and UB ! AA, respectively. The
known word contains 7 letters B, 8 letters A ja 10 letters U, the word under consideration
x
f (y ) + y
f (x) = f (x)
f (y )
(f (x) + f (y )) has 9 letters B, 8 letters A and 8 letters U. This leads to the system of equations
for all non-negative real numbers x, y . (Grade 11.) 8
p < 7 x + 2y z = 9

Answer: f (x)  0 and f (x)  x. 8 x y + 2z

: 10 + 2x y z
= 8

Solution. Taking x = y = 1, one gets 2f (1) = 2f (1)3 , so either f (1) = 1 or f (1) = 0. If

= 8

f (1) = 1 then, taking y = 1 in the

p original equality, one gets x + f (x) = f (x)(f (x) + 1) for (each equation describes the number of one letter). This system has no integral solu-
all x. From this we get f (x)  x as one solution. If f (1) = 0 then the same substitution tions, so the answer must be “no”.
leads to f (x)  0. b) Here the word under consideration has 8 letters B, 9 letters A and 9 letters U. The
It remains to check that both functions satisfy the conditions of the problem. corresponding system of equations is
8
8. The alphabet of language BAU consists of letters B, A, and U. Independently of the < 7 x + 2y z = 8
choice of the BAU word of length n from which to start, one can construct all the BAU 8 x y + 2z = 9 :
words with length n using iteratively the following rules: : 10 + 2x y z = 8

11 12
 r
One of the solutions of it is x = 0, y = z = 1. Hence one can construct the word under 2 S
Analogously we find b = = , i.e. triangle ABC is equilateral.
consideration from the known word, performing both the replacements AU ! BB and k
UB ! AA once and then reordering the letters.
11. Let a, b, be positive real numbers such that a2 + b2 + 2 = 3 . Prove that
9. Inside a circle, point K is taken such that the ray drawn from K through the centre
O of the circle and the chord perpendicular to this ray passing through K divide the 1

ab
+
1

b
+
1

a
 : 1
circle into three pieces with equal area. Let L be one of the endpoints of the chord 1 + 2 1 + 2 1+ 2

mentioned. Does the inequality \KOL < 75Æ hold? (Grade 12.) A (Grade 12.)
Answer: yes.
L K M
Solution 1. Applying the AM-GM inequality to each denominator, one obtains
Solution. Let AB be the diameter containing K (see figure 8).
Moving point K together with chord LM along this diame-
O
1+2
1

ab
+
1+2
1

b
+
1+2
1

a
 1+ a
1
2
+ b 2
+
1+ b
1
2
+ 2
+
1+
1

2 + a2
:
ter away from O , the angle KOL decreases, so the segment
bounded by chord LM and arc LAM decreases, too. Hence it Applying now the AM-HM inequality to the whole expression, one obtains
suffices to show that, if \KOL = 75Æ , the area of this segment is
B
more than one third of the area of the circle. Figure 8 1

a 2
b 2
+
1

b 2
2
+
1

2 + a2

Æ
r
1 + + 1+ + 1 +
r be the radius and \KOL
150
75 ; then the area of the sector LOM is
2
Let


= = 3
360 3 =
5

r 2
and the area of triangle LOM is
(1 + a2 + b2 ) + (1 + b2 + 2 ) + (1 + 2 + a2 )
12 9 9
:
SLOM = 2 SLOK = jLK j
jOK j = r 2

sin \KOL
os \KOL =
=
3 + 2( a 2 + b2 + 2 )
=
3 + 2
3
= 1

r2
sin 2\KOL = r2 :
1 1
=
2 4
Solution 2. Applying the AM-HM inequality, one obtains

Hence the area of the segment bounded by chord LM and arc LAM is 1
+
1
+
1
 9
=
  1 + 2 ab 1 + 2 b 1+ 2 a 3 + 2 ab + 2b + 2 a
5

r 2 1
r2 =
5 1


r2 : =
9
=
9
:
12 4 12 4 a2 + b2 + 2 + 2ab + 2b + 2 a (a + b + )2
1 1 5 1 4 1
As  > 3 implies < , we have > = . Consequently, the area of the The Jensen inequality for the square function establishes
4 12 12 4 12 3
segment under consideration is indeed greater than one third of the area of the circle.
a + b + )2
a b 2 + 2
2
( + 3
= = 1 :
10. Let K , L, M be the basepoints of the altitudes drawn from the vertices A, B , C 9 3 3

! ! ! !
of triangle ABC , respectively. Prove that AK + BL + CM = 0 if and only if ABC is Thus
equilateral. (Grade 12.)
! !
Solution. If triangle ABC is equilateral then vectors AK , BL and CM have equal lengths
!
1

ab
+
1

b
+
1

a
 a b  : 9
2
1

! ! ! !
1 + 2 1 + 2 1+ 2 ( + + )

and the sizes of the angles between them are equal to 120Æ . So AK + BL + CM = 0 .
! ! ! !
Assume now that AK +BL+CM = 0 . Let a, b, be the lengths of the sides BC , CA, AB , Solution 3. Multiplying both sides of the desired inequality by (1+2ab)(1+2b )(1+2 a),
1

respectively, and let

!j
S be the area of triangle ABC . Then jAK =
S !
2
, jBLj =
2 S
and
one gets the equivalent inequality
8

a b
!j
jCM =
2S ! ! !
. Rotating vectors AK , BL and CM counterclockwise by 90Æ , they become
1 + ab + b + a
a b : 2 2 2
4
parallel to the corresponding sides of the triangle and, if we draw them one by one onto
the plane, every vector starting from the endpoint of the previous, we obtain a triangle Note that
similar to ABC because the sum of the vectors is zero. Let k be the similarity coefficient;
r a2 + b2 + 2 p
! ! ! S !j ka implying a S
3
 ab ;3 2 2 2

then jAK j ka, jBLj kb and jCM j k . Hence jAK 2 2 1 = =

= = = = = = . 3 3
a k
13 14
giving ab  1; thus Solution 2. b) Note at first that the size of the angle between two arbitrary line segments
p  
connecting two vertices is an integral multiple of , and so the angle can occur in case
a2 b2 2 = (ab )2  ab : n n
n
2

of even n only. So must be odd since and n are coprime, which gives
(mod n).
2 2
So indeed one has n n n
Thus (i + ) = i +  i + n), i.e. any diameter transforms to a diameter.
ab + b + a p p (mod

 ab ab  a2 b2 2 :
1 + 4 2 2 2

q q
2 2 2
= Now consider two cases.
4
If line segment Ai Aj is a diameter then j  Ak Ak
n
Aj
12. Let n and be coprime positive integers. For any integer i, denote by i0 the re-
i + (mod n) and Ai Aj ? Ak Al if and only Aj
2
if j k  l j (mod n) (see figure 11). Ai Am


mainder of division of product i by n. Let A0 A1 : : : An 1 be a regular n-gon. Prove
that Then j k = (j k)   (l j ) = Ai
l j (mod n), i.e. j 0 k0  l0 j 0 (mod n) Al Al
a) if Ai Aj k Ak Al then Ai Aj k Ak Al ;
0 0 0 0 and hence Ai Aj ? Ak Al .
0 0 0 0

b) if Ai Aj ? Ak Al then Ai Aj ? Ak Al .
0 0 0 0
If line segment Ai Aj is not a diameter then Figure 11 Figure 12
consider the vertex Am such that Ai Am is
(Grade 12.) a diameter (see figure 12). Then Ai Aj ? Aj Am because the angle Ai Aj Am is supported
by the diameter Ai Am . As diameters transform to diameters, also angle Ai Aj Am is
Solution 1. a) Clearly Ai Aj k Ak Al iff i k  l j n (mod ) (see figure 9). Consequently,
supported by a diameter giving Ai Aj ? Aj Am . Since Ai Aj ? Aj Am and Ai Aj ? Ak Al ,
0 0 0 0
0 0 0

Ai Aj k A k Al ) i kl j n ) (mod )
we get Aj Am k Ak Al and, by a), also Aj Am k Ak Al . Hence Ai Aj ? Ak Al .
0 0 0 0 0 0 0 0

) i k  l j
( n )) ( ) (mod )

) i k  l j n ) (mod ) IMO Team Selection Contest

) i0 k 0  l 0 j 0 n ) (mod )

) Ai Aj k Ak Al :
0 0 0 0 First Day
Ak
1. Let k > 1 be a fixed natural number. Find all polynomials P (x) satisfying the
Aj
Aj
condition P (xk ) = (P (x))k for all real numbers x.
Ai Ai Ak+ n4 Answer: P (x) = 0 and P (x) = xn , where n is an arbitrary non-negative integer; in the
Al case of odd k also P (x) = xn .
Ak Al+ n4 Al
Solution. Let the degree of a polynomial P (x) be n > 0, then
Figure 9 Figure 10
P (x) = an xn + an 1 xn 1
+ : : : + a 1 x + a0 ;

_ _ _ _
b) Using a well-known theorem of geometry, we obtain
where an 6= 0.
Ai Aj ? Ak Al () i k j l k j l i A A A A A A A A () + = +
Let i be the largest index smaller than n for which ai 6
= 0 (suppose that such an index i
exists), then
() k i l j  j k i l
( ) + ( ) ( ) + ( ) (mod n) ()
() k l i j  n 2( + ) 0 (mod ) P (xk ) = an xkn + ai xki + ai 1 xki k
+ : : : + a1 xk + a0
where the arcs considered are taken on the circumcircle of the polygon. Thus and

Ai Aj ? Ak Al ) k l i j 
2( + ) 0 (mod n) ) (P (x))k = (an xn + ai xi + ai 1 xi 1
+ : : : + a 1 x + a 0 )k :
) k l i j 
2 ( + ) 0 (mod n) )
Find next the coefficient of the term x(k 1)n+i in both polynomials. As i < n, we get
) k 0 l 0 i0 j 0 
2( + ) 0 (mod n) )
kn > (k 1)n + i > ki, and the coefficient of this term in the polynomial P (xk ) is
) A i Aj ? Ak Al :
0 0 0 0 therefore 0. On the other hand, we get the term an xn in (P (x))k iff we take the term

15 16
an xn from k 1 factors and ai xi from one factor; therefore in the polynomial (P (x))k , Construction for n = 4k :
the coefficient of this term is kakn 1 ai 6= 0. This contradiction shows that there is no such
index i and the polynomial P (x) has the form P (x) = an xn . Also notice that if the degree
; k
(0 4 ) 4k 1

n of the polynomial P (x) is 0, then also P (x) = a0 = an xn . 9 9

k ; k >
> (6k ; k + 1)
1 6 2 >
>
From the equality P (xk ) = (P (x))k , we now get an xnk = akn xnk for all real x, i.e. an = akn
(2 1 2 ) 1
= (6k 3; 6k + 3)
=
(2k 3; 2k + 2)
or an (akn 1 1) = 0. Hence an 2 f 1; 0; 1g for odd k (an = 0 is possible only if n = 0) and k 5

> k 6

>
> >
;
an 2 f0; 1g for even k. (1 4 ; k 2)
; 4 k 3 (4 k + 1; 8k 1) k
4 2

9 9
2. Let O be the circumcentre of the acute triangle ABC and let lines AO and BC inter- k ; k + 2) 3 > ; k >
>
(4 1 4
> (2 8 2) 4
=
sect at point K . On sides AB and AC , points L and M are chosen such that jKLj = jKB j (4k 3; 4k + 4)
= (4; 8k

and jKM j = jKC j. Prove that segments LM and BC are parallel. k 7 4)

> k 8

>
>
; >
;
1

(2 k + 1; 6k) 4 k 1 (2k ; k + 2) 4k
2 6 4
Solution. Draw heights for triangles KBL and KCM from
A

q
the vertex K and let their bases be S and T , respectively.
Also lengthen the segment AK until it intersects the cir- Construction for n = 4k + 1:
cumcircle of ABC at point P (see Figure 13). As segment
L M (0 4 ; k + 1) 4k + 1 (5 k + 1; 7k + 1) 2k
AP is a diameter of the circumcircle of ABP C , the trian-
1 1

S
O
T 9 9
gles ABP and ACP are right-angled. Triangle ASK is k + 1; k) 1 > k + 1; 8k + 1) 2k + 2 >
similar to triangle ABP and triangle AT K is similar to B


C
(

(k + 2; k
>
=
(2

(2k + 2; 8k ) 2k + 4
>
=
triangle ACP (their corresponding sides are parallel), so
K 1)
k 3
k

























>
>
























>
>
jAS j jAK j jAT j P k; 1) ; k k; 7k + 2) ; k
jAB j jAP j jAC j . Hence triangle AST is similar to
= = (2 2 1 (3 4

9 9
triangle ABC and therefore ST k BC . As jLS j jSB j and Figure 13 k; 5k + 2) 2 > k; 6k + 1) 2k + 1 >
= (5
> (4
>
jMT j jT C j, we obtain LM k BC .
= (5k 1; 5k + 3)
= (4k 1; 6k + 2) 2k + 3
=

























> k 4
1

























>k
3. For which natural number n is it possible to draw n line segments between vertices
>
; >
;
(4 k + 2; 6k) 2 k 2 (3 k + 1; 7k) k
4 1
of a regular 2n-gon so that every vertex is an endpoint for exactly one segment and these
segments have pairwise different lengths? (There are actually many other constructions for both cases.)
Answer: n = 4k and n = 4k + 1, where k is an arbitrary positive integer.
Solution. Colour the vertices of the 2n-gon alternately black and white. Call the smallest Second Day
number of sides needed to pass when moving from one vertex to another the weight of
the segment with endpoints at these vertices. We see that segments with odd weigths 4. Denote
connect vertices with different colours, but segments with even weights connect vertices
with the same colour. X
m
k
Suppose that the required construction exists for a given n. As there is an equal number
f (m) = ( 1)
k
os :
2m + 1
k=1
of vertices of both colours and all segments with odd weights occupy an equal number
of vertices of both colours, also all segments with even weights must take occupy an For which positive integers m is f (m) rational?
equal number of vertices of both colours. Therefore the number of segments connecting
two white vertices equals the number of segments connecting two black vertices, and Answer: for all positive integers m.
the number of segments with even weights is an even number. Therefore there must be Solution Fix a positive integer m arbitrarily and take
an even number of even numbers among numbers 1; 2; : : : ; n, which is possible only if

n  0 (mod 4) or n  1 (mod 4). a = os
2(2m + 1)
6
= 0 :
We show next that such sets exist for n = 4k and n = 4k + 1. In the following tables,
the segments are grouped into blocks of parallel segments. In each row of a block, (x; y ) Using the formula
shows that a segment is drawn between vertices x and y , next comes the weight of this
1
segment and, after the end of a block, the number of segments in this block is shown. os x os y = ( os( x y ) + os(x + y )) ;
2

17 18
we get the chain of equations Let now be k > l. Then blog2 n + 1  2(blog3 n + 1) or
X
m
k  b n  2blog3 n + 1 ; (15)
f (m)
a
a=
log2
k
= ( 1) os
2m + 1
k=1 this holds for n = 2 and n = 8. With immediate check we see that for 4 n 7 and
X 9  n < 27 the inequality (15) does not hold. If n  27 then
m
k k 
= ( 1) os os =

k=1
2m + 1 2(2m + 1) log2 n log2 27 = log2 3(log3 n log3 27) < 2(log3 n 3) = 2 log3 n 6 ;
mX   k  
2 log3 n log2 n > 6 log2 27 > 6 5 = 1. Therefore blog2 n < b2 log3 n 
1
k hence
b n + 1 for all n  27 . Hence (15) holds iff n = 2 or n = 8.
= ( 1) os +
2
k=1
m+1 2 2(2m + 1) 2 log3
 k   Therefore the only possible values for n are 1, 2, 3 and 8. For them, we get A(1) = 1 = 20 ,
+ os
2m + 1
+
2(2m + 1)
=
A(2) = 2 = 21 , A(3) = 4 = 22 and A(8) = 32 = 25 . Hence all these values suit.
1
Xm  (2k 1) (2k + 1)

k 6. Call a convex polyhedron a footballoid if it has the following properties.
= ( 1) os + os =
2 2(2m + 1) 2(2m + 1)
k=1
1
  (2m + 1)
 (1) Any face is either a regular pentagon or a regular hexagon.
m
= os + ( 1) os =
2 2(2m + 1) 2(2m + 1) (2) All neighbours of a pentagonal face are hexagonal (a neighbour of a face is a face that
=
1

2
( a+( 1)
m

0) =
1

2
a:
has a common edge with it).
Find all possibilities for the number of pentagonal and hexagonal faces of a footballoid.
1
Hence f (m) = for all m. Answer: there are 12 pentagonal and 20 hexagonal faces.
2

Solution. We show first that there exists a footballoid with 12 pentagonal and 20 hexag-
5. Find all natural numbers n for which the number of all positive divisors of the onal faces. Start with a regular icosahedron and abstract from every vertex a regular
number lcm (1; 2; : : : ; n) is equal to 2k for some non-negative integer k . 1
pyramid with lateral edge of the edge of the icosahedron. In such a way, we get 12
3
Answer: 1, 2, 3 and 8.
regular pentagons instead of 12 vertices of icosahedron and we get 20 regular hexagons
Solution. Let P be the set of all prime numbers. Let Æ (m) denote the number of positive instead of 20 faces of icosahedron. All neighbours of any pentagonal face are hexagonal.
divisors of natural number m and let A(n) = Æ (lcm (1; : : : ; n)). Denote by p . m the Now show that it is the only possibility. Let B be a footballoid. Consider an arbitrary
exponent of prime number p in the canonical representation of m. Notice that vertex of B ; let it belong to x pentagonal and y hexagonal faces. Then x + y  3 as every
Y  vertex of a polyhedron belongs to at least 3 faces. As the sizes of the interior angles of
A(n) = Æ (lcm (1; : : : ; n)) = Æ pmax(p.1;:::;p.n) = pentagonal and hexagonal faces are 108Æ and 120Æ , respectively, we get x
108Æ + y
120Æ <
360 . Hence x + y  3 and x > 0. Therefore x + y = 3, which means that every vertex of
p2P Æ
Y  Y
= Æ pblogp n = (blogp n + 1) : a footballoid belongs to exactly 3 faces, at least one of which is pentagonal. As these 3
p2P p2P faces are pairwise neighbours and pentagonal faces cannot be neighbours, every vertex
must belong to exactly one pentagonal and two hexagonal faces.
Therefore A(n) is a power of 2 iff all numbers in the form blogp n + 1, where p 2 P, are Consider an arbitrary hexagonal face. All its vertices belong to one pentagonal and one
powers of 2. Let blog2 n + 1 = 2k and blog3 n + 1 = 2l . As log2 n  log3 n, we get k  l. hexagonal face. Therefore the neighbours of a hexagonal face are alternately pentagonal
Consider two cases. and hexagonal, so there are exactly three of both kinds.
If k = l then Now cover every pentagonal face with a regular pentagonal pyramid, whose lateral
blog2 n = blog3 n ; (14) edges are continuations of the (hexagonal) neighbours of this face. In this way, hexag-
onal faces become equilateral triangles and pentagonal faces are replaced with vertices
this holds for n = 1 and n = 3. With immediate check we see that, for n = 2 and in which five edges meet. As any two neighbours of a pentagonal face which are neigh-
4  n < 8, the equation (14) does not hold. If n  8 then bouring themselves meet under the same angle (two regular hexagons and one regular
pentagon can meet in one vertex in principle in only one way), also the triangles meet-
log2 n 3 = log2 n log2 8 = log2 3(log3 n log3 8) > log3 n log3 8 ; ing in a vertex of our new polyhedron meet under the same angle. Therefore the new
hence log2 n log3 n > 3 log3 8 > 3 2 = 1 . Therefore blog2 n > blog3 n for all n  8. polyhedron is a regular icosahedron. It has 12 vertices and 20 faces, so the footballoid
Hence (14) holds iff n = 1 or n = 3. B had to have 12 pentagonal and 20 hexagonal faces.

19 20
Estonian Math Competitions
2004/2005

The Gifted and Talented Development Centre

Tartu 2005

Selected Problems from Open Contests
1. In the final tournament of football championship, the teams are divided into groups
of four. Each team plays one game with every other team in its group. A win gives 3
points, a draw 1 point and a loss 0 points. From each group, two teams advance so that
each advancing team gets at least as many points as each non-advancing team.
a) What is the smallest possible score of an advancing team?
b) What is the biggest possible score of a non-advancing team? (Juniors.)
Answer: a) 2; b) 6.
Solution. a) If one team in a group wins all matches and the other three draw all matches,
then the second advancing team collects just 2 points. Let us prove that a team cannot
advance with less than 2 points. Indeed, if team A gets at most 1 point, it must lose at
least two games. Therefore two teams get at least 3 points and outperform A, so that A
cannot be in top two.
b) If one team in a group loses all matches and the other three win cyclically (X wins
Y , Y wins Z and Z wins X ), then three teams collect 6 points and one of them does
not advance. Let us prove that a team with more than 6 points must advance. If team
C gets at least 7 points, it must win at least two games. As the losing teams can get at
most 6 points, team C must be in top two.
2. In triangle ABC let D, E be the midpoints of AB and AC , respectively. Prove that
the intersection point of the bisectors of angles BDE and CED lies on AB if and only if
the length of BC is equal to the arithmetic mean of the lengths of AB and AC . (Juniors.)
Solution. Let K and L be the intersection points of BC with the bisectors of
angles BDE and CED, respectively (see Figure 1). As DE k BC , we have
|AB| |AC|
∠BDK = ∠EDK = ∠BKD, and |BK| = |BD| = . Similarily |CL| = . cycle, we see that, after any odd number of repetitions, the white face of the cube is at
2 2
Therefore
|AB| + |AC|
|BK| + |CL| = .
2 b, c, and d be the numbers of steps made during this circuit respectively to the right,
1 2
As rays BK and CL are opposite-directed, we have
|AB| + |AC|
|BC| = ⇐⇒ |BC| = |BK| + |CL| ⇐⇒ K = L .
2
3. On some square of an infinite squared plane, there is a cube which covers the square
exactly. The top face of the cube is white, the other faces are black. With one step, one
can turn the cube over any edge so that it starts covering a neighbouring square. Is it
possible to achieve a situation where the cube lies on the initial square with the white
face at the bottom, making exactly
a) 2004 steps; b) 2005 steps? (Juniors.)
A
D E
B KL C
Figure 1 Figure 2 Figure 3
Answer: a) yes; b) no.
Solution 1. a) Turn the cube two steps forward, one step to the right, two steps backward,
one step to the left (see Figure 2). After these six steps, the cube gets back onto the initial
square but the white face is now at the bottom. The rest 1998 steps are made in pairs:
turn the cube onto arbitrary neighbouring square and then turn it back onto the initial.
b) We colour squares dark and light by diagonals so that the cube lies on a light-coloured
square at the beginning (see Figure 3). Since, from any square, the cube can move only
to squares of the opposite colour, the cube lies on a light-coloured square after any even
number of steps and on a dark-coloured square after any odd number of steps. Thus
after 2005 steps, the cube lies on a square different from the initial.
Solution 2. a) Turn the cube two steps forward, one step to the left, two steps forward,
two steps to the right, four steps backward, one step to the left. After these 12 steps, the
cube is back on the initial square but the white face is now at the bottom. Repeating this
the bottom and, after any even number of repetitions, the white face is at the top. Since
2004 = 12 · 167, we obtain the desired result by repeating this cycle for 167 times.
b) Assume that the cube makes a circuit and gets back onto the initial square. Let a,
to the left, up, and down. Then a = b and c = d. Thus the cube makes altogether
 K

a + b + c + d = 2(a + c) steps which is even number. Hence the cube cannot get back to Solution. Let ∠O1AG1 = α and ∠O1 G1 A = β H1 H2
the initial square after 2005 steps. (see Figure 4). If line G1 K touches circle C1 , then β P
∠AH1 G1 = 90◦ and ∠H1 G1 O1 = ∠AG1 O1 = β. β
G1 O1 O2 G2
Also, ∠G1 P A = 90◦ and ∠P AO1 = G1 AO1 = α.
a+b
4. Relatively prime positive integers a and b are chosen in such a way that is From the right-angled triangles AH1 G1 and P1 αα P2
a−b G1 P A, we get
also a positive integer. Prove that at least one of the numbers ab + 1 and 4ab + 1 is a A
perfect square. (Juniors.) α + 2β = 90◦ , Figure 4
a+b a m+1 β + 2α = 90◦ .
Solution 1. Let = m. Then a + b = ma − mb implying = . As a and b
a−b b m−1
are relatively prime, there exists an integer k such that m + 1 = ka and m − 1 = kb. By Solving the system gives α = β = 30◦ . Therefore ∠P AG1 = 60◦ and
multiplying these equalities, we get m2 − 1 = k 2 ab implying k 2 ab + 1 = m2 . Number k ◦ ◦ ◦
∠P AG2 = 120 − 60 = 60 . Thus we have ∠P AG1 = ∠P AG2 . As circles C1 and
as a common divisor of numbers m − 1 and m + 1 must be a divisor of their difference C2 lie in equal angles and touch at P , their radii must be equal. Therefore AK is the
2. Hence k can only be 1 or 2 and we are done. symmetry axis. By symmetry, line G2 K is a tangent to C2 and touches the circle at H2 .
Solution 2. As
a+b a−b+b+b 2b 7. A king wants to connect n towns of his kingdom with one-directional airways so
= =1+ ,
a−b a−b a−b that, from each town, exactly two airlines depart. From each town, it must be possible
2b to fly to every other town with at most one change. Find the biggest n for which this
we see that must be an integer. Numbers b and a − b are relatively prime because plan is feasible. (Seniors.)
a−b
a and b are relatively prime. Hence 2 must be divisible by a − b. Therefore a − b = 1 or
Answer: 6.
a − b = 2. The former case implies 4ab + 1 = 4(b + 1)b + 1 = (2b + 1)2 , the latter case
implies ab + 1 = (b + 2)b + 1 = (b + 1)2 . Solution. From a fixed town, one can get directly to two towns and further to at most
four more towns. Thus the number of towns cannot exceed 1 + 2 + 4 = 7.
a √ 2 Let us assume that a suitable airway plan for 7 towns exists. From each town, there
5. The teacher has chosen positive integers a and b such that · a + b2 is an integer. must be a unique way to every other town (either direct or with one change), otherwise
b
a) Silly-Sam claims that a is divisible by every prime factor of b. Prove that he is right. some town would have less than 6 possible final destinations. Without loss of generality
assume that there is a direct flight from town L1 to towns L2 and L3 , from town L2 to
b) Silly-Sam claims that actually b 6 a. Is he right this time? (Seniors.)
towns L4 and L5 , and from town L3 to towns L6 and L7 (see Figure 5).
Answer: b) no.
Solution. a) Let p be √ an arbitrary prime factor of b. If the observed expression is an L4
integer,√the number a a2 + b2 must be divisible by p. As p is prime, either a is divisible L5
L2
by p or a2 + b2 is divisible by p. In the latter case, squaring gives that a2 +b2 is divisible
by p2 . By the initial assumption, b2 is divisible by p2 , hence also a2 is divisible by p2 . L1 L6
Therefore a is divisible by p in both cases. L3
a √ L7
b) The teacher may choose a = 12 and b = 16. In this case, · a2 + b2 = 15 is an
b
integer. Therefore the inequality b 6 a might be wrong. Figure 5 Figure 6 Figure 7

6. Two circles C1 and C2 with centres O1 and O2 , respectively, are touching exter-
From L2 , there must be a way to all towns in list L1 , L3 , L6 and L7 . As the direct flights
nally at P . On their common tangent at P , point A is chosen, rays drawn from which
from L2 take to towns L4 and L5 , there must be a direct flight from L4 to two towns in
touch the circles C1 and C2 at points P1 and P2 both different from P . It is known that
the list and from L5 to the other two. Without loss of generality assume that there is a
∠P1 AP2 = 120◦ and angles P1 AP and P2 AP are both acute. Rays AP1 and AP2 intersect
direct flight from L4 to L3 . But now there can be a flight from L4 to none of L1 , L6 , L7
line O1 O2 at points G1 and G2 , respectively. The second intersection between ray AO1
because otherwise there would be two ways to get from town L4 to town L3 , L6 , L7 ,
and C1 is H1 , the second intersection between ray AO2 and C2 is H2 . Lines G1 H1 and
respectively.
AP intersect at K . Prove that if G1 K is a tangent to circle C1 , then line G2 K is tangent
to circle C2 with tangency point H2 . (Seniors.) A suitable airway plan for 6 towns exists, as shown on Figure 6 or Figure 7.

3 4
8. For which integers a does there exist two different finite sequences of positive inte- 2004 2005
gers i1 < i2 < · · · < ik and j1 < j2 < · · · < jl such that

(ai1 + 1)(ai2 + 1) · · · (aik + 1) = (aj1 + 1)(aj2 + 1) · · · (ajl + 1) ?

2004

(Seniors.)
Answer: −1, 0, and 1.
Solution. In the case a = −1, both sides of the equation equal to zero whenever both
sequences contain an odd number. In the case a = 0, both sides equal to one irrespective
of the sequences. In the case a = 1, all factors of the products are equal to 2, so the
products are equal whenever the sequences have the same length. Therefore suitable
sequences exist in these three cases.
Let us prove that if |a| > 1, then such sequences do not exist. Suppose the contrary,
i.e. i1 , . . . , ik and j1 , . . . , jl are different sequences that lead to equal products. We
may assume that no integer is in both sequences or else the respective terms can be
cancelled in the products. After deletions, both sequences are still nonempty or else we
get an equation between 1 and the product of integers not being equal to 1. We can also
assume that i1 < j1 .
Multiplying and then removing the parentheses on both sides gives an equation be-
tween sums of powers of a. Both sides contain term 1 and we can reduce that. This
ends up in the equation of form
ai1 +···+ik + · · · + ai1 = aj1 +···+jl + · · · + aj1 .
The smallest exponent is i1 on the left-hand side and j1 on the right-hand side. As
j1 > i1 + 1, the right-hand side is divisible by ai1 +1 . On the left-hand side, all terms
except the last one are divisible by ai1 +1 . Hence the left-hand side is not divisible by
ai1 +1 , a contradiction.
Comment. From the solution, we get that for each a (|a| > 1), any finite subset I of
natural numbers can be assigned a unique number
Y Answer: f (x) = x + 1 is the only such function.
f (I) = (ai + 1).
i∈I
The number determines the subset uniquely. This can be used in proving that the num-
ber of finite subsets of natural numbers is countable.
9. Mother has baked a platecake and cut it into m×n square pieces of equal size. Kalle
and Juku play the following game. Each player at his move eats two pieces having a
common side. Moves are made by turns, Juku starts. A player who cannot move loses.
Who wins the game if
a) m = 3, n = 3;
b) m = 2004, n = 2004;
c) m = 2004, n = 2005? (Seniors.)
2004
Figure 8 Figure 9 Figure 10
Answer: a) Kalle; b) Kalle; c) Juku.
Solution. a) To Juku’s first move, Kalle can reply with a move after which a part of shape
2 × 2 has been eaten up (see Figure 8). Irrespectively of Juku’s second move, Kalle can
make one more move. After that, Juku has no move.
b) When Juku eats up some two pieces, Kalle replies by eating two pieces which lie
symmetrically with respect to the midpoint of the cake (see Figure 9). This guarantees
that, after every move by Kalle, pieces not yet eaten up are situated symmetrically to
the midpoint of the cake and, as it is not possible to eat two symmetric pieces at one
move by the same player, Kalle can always follow the strategy described. Hence Juku’s
moves come first to the end.
c) On his first move, Juku can eat two pieces between which the midpoint of the cake
lies and later use the strategy of Kalle from part b) (see Figure 10).
Remark. This game is called Cram, its full analysis for the case where length and width
are both odd numbers seems to be quite complicated and is not completed yet.
10. Find all functions f : R → R satisfying
f (x + f (y)) = x + f (f (y))
for all real numbers x and y whereby f (2004) = 2005. (Seniors.)
Solution 1. Taking y = 0, we get the equality f (x + f (0)) = x + f (f (0)). Making the
substitution x + f (0) = z, we obtain f (z) = z − f (0) + f (f (0)) for every real number z.
Hence f is a linear function f (x) = x + a. To find a, take x = 2004 in the last expression
and, by using the known value of the function, obtain a = 1. A quick checking shows
that the function f (x) = x + 1 satisfies the conditions of the problem.
Solution 2. Taking x = −f (f (y)), we see that f (y) = 0 for some y. Then f (x) = x + f (0).
From the condition f (2004) = 2005, we find f (0) = 1. Thus f (x) = x + 1.
11. Three rays are going out from point O in space, forming pairwise angles α, β and
γ with 0◦ < α 6 β 6 γ 6 180◦ . Prove that
α β γ
sin + sin > sin .
2 2 2
(Seniors.)

5 6
 O Solution 1. Choose points A, B, and C on the three Solution. Let a1 , a2 , a3 , a4 be the smallest numbers
10 120 5 15
α α rays, respectively, so that |OA| = |OB| = |OC| = d and b1 , b2 , b3 , b4 be the biggest numbers of the first,
d d 12 24 3
2 2 and ∠BOC = α, ∠COA = β, ∠AOB = γ . These second, third and fourth rows, respectively. Obvi-
2 8
three points must be different and do not lie on the 4 ously b4 = a4 and b3 = a3 b4 . In the second row,
same line. From isosceles triangle BOC with vertex there exists a number which is the product of the
B C Figure 12
angle α and side length d (see Figure 11), we obtain biggest element in the third row and some other
α element in the second. Thus b2 > a2 b3 = a2 a3 a4 .
Figure 11 |BC| = 2d sin . Analogously from triangles COA
2 Finally for the first row, we obtain similarily b1 > a1 b2 > a1 a2 a3 a4 . All numbers in the
β γ
and AOB, find |CA| = 2d sin and |AB| = 2d sin . As triangle are greater than 1, otherwise we could find two equal numbers in it. Since all
2 2 numbers are different, we have a1 a2 a3 a4 > 2 · 3 · 4 · 5 = 120. Hence b1 > 120. Number
|BC| + |CA| > |AB|, 120 is achievable as follows from Figure 12.

we obtain
α β γ
Selected Problems from the Final Round of National
2d sin + 2d sin > 2d sin ,
2 2 2 Olympiad
giving the desired inequality.
Solution 2. At first, we show that γ 6 α + β. Consider the two of the given three rays
which form angle of size γ , and build two cones by moving the third ray around both 1. Rein solved a test on mathematics that consisted of questions on algebra, geometry
rays. On the plane defined by the axes, the first cone cuts angle α off from angle γ and and logic. After checking the results, it occurred that Rein had answered correctly 50%
the second cone cuts angle β. Assume γ > α+β, then the last two angles do not overlap, of questions on algebra, 70% of questions on geometry and 80% of questions on logic.
therefore the cones have no common points except the vertex O, a contradiction. Thus Thereby, Rein had answered correctly altogether 62% of questions on algebra and logic,
γ 6 α + β. On the other hand, α + β + γ 6 360◦ , giving and altogether 74% of questions on geometry and logic. What was the percentage of
correctly answered questions throughout all the test by Rein? (Grade 9.)
γ α β γ
6 + 6 180◦ − . Answer: 65%.
2 2 2 2
Solution: Let a, g, and l be the numbers of correctly answered questions on algebra,
Therefore geometry and logic, and A, G, and L be the total number of questions on algebra, ge-
γ

α β

α β β α α β ometry and logic, respectively. The conditions of the problem imply a = 0.5A, g = 0.7G,
sin 6 sin + = sin cos + sin cos < sin + sin l = 0.8L, a + l = 0.62(A + L), g + l = 0.74(G + L). After substituting to the fourth and
2 2 2 2 2 2 2 2 2
fifth equation, we obtain 0.5A + 0.8L = 0.62A + 0.62L, or equivalently 0.12A = 0.18L,
α β γ β giving
because , and are grater than 0◦ but do not exceed 90◦ , and also cos < 1 and
2 2 2 2 A = 1.5L,
α
cos < 1.
2 and 0.7G + 0.8L = 0.74G + 0.74L, or equivalently 0.04G = 0.06L, giving
G = 1.5L.
12. We call a number triangle amazing if all its elements are 21 84 7
different positive integers and, under every two neighbouring 4 12 Now
numbers, the quotient by division of the greater of two by the 3
smaller is written. In the figure, one amazing triangle with side a + g + l = 0.5A + 0.7G + 0.8L = 0.75L + 1.05L + 0.8L = 2.6L
length 3 is shown. Find the smallest number which can occur as the greatest element in and
an amazing triangle with side length 4. (Seniors.)
A + G + L = 1.5L + 1.5L + L = 4L.
Thus the percentage of correct answers was
a+g+l 2.6
= = 65%.
A+G+L 4

7 8
2. Represent the number
√
q
3
1342 167 + 2005

in the form where it contains only addition, subtraction, multiplication, division and
square roots. (Grade 10.)
√ Figure 13 Figure 14
Answer: 2 167 + 1.
Solution 1. First, isolate the terms divisible by 167:
√ √ √ Solution. Suppose all dark squares are covered. Since one hook cannot cover more than
1342 167 + 2005 = 1336 167 + 2004 + 6 167 + 1 = one dark square, in total at least nine hooks are needed. As only eight of these are avail-
√ √
= 8 · 167 167 + 12 · 167 + 6 167 + 1. able, one of the dark squares must remain uncovered. There are three fundamentally
different possibilities: the free square lies in the corner, in the middle of a side or in the
Second, represent the result in the form centre of the board. The corresponding tilings are shown in the Figure 14.
√ √ √ √
1342 167 + 2005 = (2 167)3 + 3 · (2 167)2 + 3 · (2 167) + 1 =
√ 4. Real numbers x and y satisfy the system of equalities
= (2 167 + 1)3 .
(
sin x + cos y = 1
Therefore .
cos x + sin y = −1
√ √
q
3
1342 167 + 2005 = 2 167 + 1.
Prove that cos 2x = cos 2y. (Grade 11.)
√
Solution 2. Search the answer in the form a 167 + b, where a and b are integers. Then Solution 1. After squaring both sides, we obtain
we must have
sin2 x + 2 sin x cos y + cos2 y = 1,
√ √
(a 167 + b)3 = 1342 167 + 2005, cos2 x + 2 cos x sin y + sin2 y = 1.

or equivalently After adding the equations and dividing by 2, we obtain

√ √ √
167a3 167 + 3 · 167a2 b + 3ab2 167 + b3 = 1342 167 + 2005. sin x cos y + sin y cos x = 0,

Thus a and b must satisfy the system or equivalently

(
167a3 + 3ab2 = 1342 sin(x + y) = 0.
.
501a2 b + b3 = 2005 Hence x + y = kπ, where k is integral. Therefore 2x = 2kπ − 2y, giving cos 2x = cos 2y.
Solution 2. After adding the equations, we obtain
The second equation can be rewritten in the form (501a2 +b2 )b = 2005. Since a and b dif-
fer from 0 and are integers, 501a2 + b2 must be a divisor of 2005 that is greater than 501. sin x + sin y + cos x + cos y = 0,
The only possibility is now 501a2 + b2 = 2005, giving b = √ 1, a = 2. Simple check
√ shows
that a = 2, b = 1 satisfy the first equation as well. Hence (2 167+1)3 = 1342 167+2005. which is equivalent to
x+y x−y x+y x−y
2 sin cos + 2 cos cos = 0,
3. A 5×5 board is covered by eight hooks (a three unit square figure, shown 2 2 2 2
in the picture) so that one unit square remains free. Determine all squares of implying
the board that can remain free after such covering. (Grade 10.)  
x−y x+y x+y
Answer: All the squares that are colored dark in the Figure 13. cos sin + cos = 0.
2 2 2

9 10
 x−y x−y π assumption that C is more important than A. Analogously, if C comes up first, then
If cos = 0, then = (2k − 1) · , giving 2x − 2y = (2k − 1) · 2π.
2 2 2 we get a contradiction with the assumption that B is more important than A. Thus our
x+y x+y x+y
Hence cos 2x = cos 2y. If sin + cos = 0, then tan = −1, original assumption was false.
2 2 2
x+y π π Remark. In graph theory, the relation “is more important than” of this problem is called
implying = kπ − and x + y = 2kπ − . Therefore
2 π  π 4  2 postdominance.
cos y = cos − − x = cos + x = − sin x, leading to sin x + cos y = 0 that
2 2
contradicts the first equation of the initial system. 7. In a fixed plane, consider a convex quadrilateral ABCD. Choose a point O in the
plane and let K , L, M , and N be the circumcentres of triangles AOB, BOC , COD, and
5. Let a, b, and n be integers such that a + b is divisible by n and a2 + b2 is divisible DOA, respectively. Prove that there exists exactly one point O in the plane such that
by n2 . Prove that am + bm is divisible by nm for all positive integers m. (Grade 11.) KLMN is a parallelogram. (Grade 11.)

Solution 1. We prove that a and b are divisible by n, then the claim immediately follows. Solution. If O is the point described in the problem, then D
As M
we must have KL ⊥ BO because K and L both lie on
the perpendicular bisector of BO. Similarly LM ⊥ CO, C
2ab = (a + b)2 − (a2 + b2 ), MN ⊥ DO, and NK ⊥ AO. Let O be the intersec-
tion point of the diagonals of ABCD (see Figure 15). O
2ab is divisible by n2 . Let p be any prime in the prime decomposition of n and let α be N
Then both KL and MN are perpendicular to BD, giving
its exponent. Then the exponent of p is at least 2α in the prime decomposition of 2ab, L
KL k MN . Similarly LM k NK . Therefore the oppo-
and at least 2α − 1 in the prime decomposition of ab. Therefore at least one of numbers
site sides of KLMN are parallel, meaning that KLMN
a and b must be divisible by pα . As a + b is divisible by n and hence by pα , also the other
is a parallelogram. On the other hand, if O is a point for A B
of the numbers a and b must be divisible by pα . Altogether, this means that both a and K
which KLMN is a parallelogram, we have KL k MN .
b are divisible by n.
Then also BO k DO, giving that O lies on the line BD. Figure 15
We can show similarily that O lies also on the line AC .
6. A post service of some country uses carriers to transport the mail; each carrier’s Therefore O is the intersection point of the diagonals.
task is to bring the mail from one city to a neighbouring city. It is known that it is
possible to send mail from any city to the capital P . For any two cities A and B, call
8. Does there exist an integer n > 1 such that
B more important than A, if every possible route of mail from A to the capital P goes
through B. n −1
22 −7
a) Prove that, for any three different cities A, B, and C , if B is more important than A
is not a perfect square? (Grade 11.)
and C is more important than B, then C is more important than A.
b) Prove that, for any three different cities A, B, and C , if both B and C are more Answer: Yes.
n
important than A, then either C is more important than B or B is more important Solution 1. Let us show that if n = 5, then the number 22 −1 − 7 is not a perfect square.
than C . (Grade 11.) 10 31 10 3
Note that 2 = 1024 ≡ 1 (mod 11), giving 2 = 2 · (2 ) ≡ 2 (mod 11). Hence the
remainder of division of 231 − 7 by 11 is 6. On the other hand, squares of integers can
Solution. a) Let t be any possible mail route from A to P . Since B is more important have remainders 0, 1, 4, 9, 5, and 3 in division by 11.
than A, the route t goes through B. The end part of t from B to P is a mail route from Solution 2. Computation gives 231 − 7 = 32768 · 65536 − 7 = 2147483641 but
B to P . Since C is more important than B, this route goes through C . Therefore, t goes 463402 = 2147395600 and 463412 = 463402 + 2 · 46340 + 1 = 2147488281. So
through C . 463402 < 231 − 7 < 463412 .
5
b) Assume that the claim doesn’t hold, that is, C is not more important than B and B Remark. The number 11 is the least modulus with respect to which 22 −1 − 7 is not a
is not more important than C . Then there exist a route from B to P not going through quadratic residue. There exist greater such numbers, e.g. 31.
C and a route from C to P not going through B. Consider any route from A to P .
Since B and C are more important than A, this route goes through both B and C . Start
moving from A along this route and find out which of the cities B and C comes up
first. If it is B, then continue along the route to P that doesn’t pass through C . So we
have found a route from A to P that doesn’t go through C , a contradiction with the

11 12
 9. Punches in the buses of a certain bus company always cut exactly six
holes into the ticket. The possible locations of the holes form a 3 × 3 table Ai+1
as shown in the figure. Mr. Freerider wants to put together a collection
of tickets such that, for any combination of punch holes, he would have a Q
ticket with the same combination in his collection. The ticket can be viewed P
both from the front and from the back. Find the smallest number of tickets in such a Ai
collection. (Grade 12.)
Figure 16 Figure 17
Answer: 47.
Solution. Instead of holes, we can deal with non-holes — the locations that are not cut
through during punching. The number of possibilities to choose 3 locations for non- As P and Q are different, there exist such vertices of the polygon Ai and Ai+1 that the
holes from 9 locations is point Q is inside or on the side of Ai P Ai+1 , not coinciding with the vertex P (see Figure
  17). But this implies ∠Ai QAi+1 > ∠Ai P Ai+1 , a contradiction. Hence there exists only
9
= 84. one point satisfying the conditions of the problem.
3
Remark. For n = 4, 6 one may find several different points satisfying the conditions of
One ticket can represent either one punch combination that is symmetric with respect the problem.
to the central axis parallel to the longer sides of the ticket or two different combinations
that are mirror images of each other with respect to this axis. In the case of symmetric
combinations, either all three non-holes lie in the second column (1 possibility) or one 11. A string having a small loop in one end is set over a horizontal pipe so
non-hole lies in the second column and other two lie in the same rows, one in the first that the ends hang loosely. After that, the other end is put through the loop,
column and the other in the second (3 · 3 = 9 possibilities). So there are 1 + 9 = 10 pulled as far as possible from the pipe and fixed in that position whereby α
symmetric combinations and 84 − 10 = 74 non-symmetric combinations. The number this end of the string is farther from the pipe than the loop. Let α be the
of tickets needed to cover these combinations is angle by which the string turns at the point where it passes through the loop
(see picture). Find α. (Grade 12.)
74
10 + = 47. π
2 Answer: .
3
Solution. Let O and r be the centre point and the radius of the pipe, respectively. Let
l be the length of the string, A and B the loose end and the end with the loop, re-
10. Consider a convex n-gon in the plane with n being odd. Prove that if one may find spectively, and let C and D be the first and the last tangent point with the surface of
a point in the plane from which all the sides of the n-gon are viewed at equal angles, the pipe (see Figure 18). At first, find the length of AO, denote by d(α). Obviously
then this point is unique. (We say that segment AB is viewed at angle γ from point O iff |BC| = |BD| = r cot α. As ∠COD = π − 2α, the string touches the pipe along the arc
∠AOB = γ .) (Grade 12.) CD at angle π + 2α and thus the length of the string along the arc CD is r(π + 2α). After
Solution. Draw the rays from the point described in the problem through the vertices of subtracting the lengths of segments BC and BD and the arc CD from the total length
of the string, we obtain
the polygon. The point can lie either inside or outside the polygon, therefore there are
two possibilities for the rays: they divide either all the plane or only an angle into equal |AB| = l − 2r cot α − r(π + 2α).
angles (see Figure 16). The latter case would imply that the outermost rays were both
We also have
incident to one vertex and all the others were incident to two vertices of the polygon,
r
giving that n is even. This contradiction shows that the point satisfying the conditions |BO| = .
of the problem lies inside the polygon. sin α
Assume now that there are two different points P and Q inside A1 A2 . . . An , from Altogether d(α) = |AB| + |BO|, or equivalently
which all the sides are viewed at equal angles. Then for every i = 1, 2, . . . , n (tak- r
d(α) = l − 2r cot α − r(π + 2α) + .
ing An+1 = A1 ), sin α
Now we find α for which the value of d(α) is the greatest. The derivative equals
2π 2π
∠Ai P Ai+1 = , ∠Ai QAi+1 = . 2r r cos α
n n d0 (α) = − 2r − ,
sin2 α sin2 α

13 14
 be periodic modulo 2N . On the other hand, if ai and aj are congruent modulo 2N ,
they are also congruent modulo N . Therefore, the sequence is periodic modulo N , a
contradiction.
O b) We are going to prove that the sequence ai = 2i−1 is periodic modulo no even num-
bers and all odd numbers. Let first the modulus N be even. If the sequence were
periodic modulo N , then, using a similar argument as in a), we get that the sequence
r r
would be periodic modulo any factor of N . But the sequence 1, 0, 0, 0, . . . of remainders
modulo 2 is not periodic, so our sequence cannot be periodic modulo N . Now let the
α α modulus N be odd. As the set of possible remainders in division by N is finite, there
D C
exist two indices i and j with i < j such that ai = 2i−1 and aj = 2j−1 are congruent
B
modulo N . Then the difference 2j−1 − 2i−1 = 2i−1 (2j−i − 1) is divisible by N . Since N
A is odd, 2j−i − 1 must be divisible by N . Hence a1 = 1 and aj−i+1 = 2j−i are congruent
modulo N , implying also that a2 = 2a1 and aj−i+2 = 2aj−i+1 are congruent modulo N ,
Figure 18 that a3 = 2a2 and aj−i+3 = 2aj−i+2 are congruent modulo N etc, that is, the sequence is
periodic modulo N .
or equivalently,
13. A crymble is a solid consisting of four white and one black unit cubes
2 − 2 sin2 α − cos α 2 cos2 α − cos α
   
0
d (α) = r =r . as shown in the picture. Find the side length of the smallest cube that can
sin2 α sin2 α
be exactly filled up with crymbles. (Grade 12.)
The condition d0 (α) = 0 gives the equation 2 cos2 α − cos α = 0, implying cos α = 0 or Answer: 10.
1 π π π
cos α = . Thus α = or α = , bearing in mind that 0 < α 6 . In order to find Solution. Since a crymble consists of 5 unit cubes, the volume of the cube made up from
2 2 3 2
the maximal value, we consider d0 (α) in the neighbourhoods of α found out before. The crymbles and hence also the length of its side must be divisible by 5. A cube with the
denominator sin2 α is positive in these neighbourhoods, the numerator 2 cos2 α − cos α side length 5 cannot be filled up with crymbles. To prove this, colour 27 unit cubes as
1 shown in the Figure 19. One crymble cannot fill more than one coloured cube, therefore
is negative iff 0 < cos α < . Therefore d0 (α) is positive and d(α) is increasing in the at least 27 crymbles are needed. But their volume 27 · 5 = 135 is larger than the volume
2
π π π 53 = 125 of the cube.
interval 0 < α < , d0 (α) is negative and d(α) is decreasing in the interval < α < .
3 3 2 A cube with side length 10 can be filled up with crymbles. By putting together two
π
Hence the function d(α) obtains its maximal value at α = . crymbles, construct the solid that is in the Figure 20. From two such solids, make a
3
Remark. The value for α found out in the solution is attainable iff√the ratio between the 2 × 2 × 5 cuboid. From such cuboids, it is possible to put together a 10 × 10 × 10 cube.
5π + 2 3
length of the string and the diameter of the pipe is at least .
6

12. A sequence of natural numbers a1 , a2 , a3 , . . . is called periodic modulo n if there

exists a positive integer k such that, for any positive integer i, the terms ai and ai+k are
equal modulo n. Does there exist a strictly increasing sequence of natural numbers that

a) is not periodic modulo finitely many positive integers and is periodic modulo all the
other positive integers;
b) is not periodic modulo infinitely many positive integers and is periodic modulo in- Figure 19 Figure 20
finitely many positive integers? (Grade 12.)

Answer: a) no; b) yes.

Solution. a) Suppose that the sequence a1 , a2 , a3 , . . . is not periodic modulo finitely
many positive integers, let N be the largest of these. Since 2N > N , this sequence must

15 16
 IMO Team Selection Contest c1
c1

First Day Q1
c2 O1
l
Q1
O2 Q2 P M
1. On a plane, a line l and two circles c1 and c2 of different radii are given such that l
touches both circles at point P . Point M 6= P on l is chosen so that the angle Q1 MQ2 is l
as large as possible where Q1 and Q2 are the tangency points of the tangent lines drawn Q2
c2
P M
from M to c1 and c2 , respectively, differing from l. Find ∠P MQ1 + ∠P MQ2 .
Figure 21 Figure 22
Answer: π.
Solution. Consider first the case where c1 and c2 are on the same side from l (see Figure
21). Let O1 and O2 be the circumcentres and r1 and r2 the radii of c1 and c2 , respec- • each Automorian is loved by some Automorian;
tively. Without loss of generality, assume r1 > r2 . Denote ∠P MQ1 = α1 , ∠P MQ2 = α2 • if Automorian A loves Automorian B, then also all Automorians honouring A
1 1 love B;
and |P M| = d. As ∠P MO1 = ∠P MQ1 and ∠P MO2 = ∠P MQ2 , we see that
2 2 • if Automorian A honours Automorian B, then also all Automorians loving A hon-
α1 α2
∠Q1 MQ2 = α1 − α2 is maximal if and only if ∠O1 MO2 = − is maximal; the our B.
α 2 2
1 α2

latter holds if and only if tan − is maximal because the angle is in the first Is it correct to claim that every Automorian honours and loves the same Automorian?
2 2
quadrant. The formula of tangent of difference gives
Answer: Yes.
α1 α2 r1 r2
α tan − tan − Solution. Denote by f (A) the Automorian loved by A and by g(A) the Automorian
1 α2  2 2 d d .
tan − = α1 α2 = r1 r2 honoured by A. The conditions of the problem imply the following:
2 2 1 + tan · tan 1+ ·
2 2 d d • for every Automorian A, there exists an Automorian C such that f (C) = A;
Representing the result in the form • for every Automorian C , f (g(C)) = f (C);
α α2  r1 − r2 r − r2 • for every Automorian C , g(f (C)) = g(C).
1
tan − = r1 r2 = √  1 √ ,
2 2 d+ d r1 r2 We will show that f (A) = g(A) for every A. Applying f to both sides of the third
d r r
1 2 √ +
r1 r2 d condition, we get
we obtain that the value of the denominator of the last expression is minimal in the case f (g(f (C))) = f (g(C)).
√
d = r1 r2 . Now
Using the second condition in both sides, we get
√ √
α1 r1 r1 α2 r2 r2
tan = =√ , tan = =√ , f (f (C)) = f (C).
2 d r2 2 d r1
α1 α2 α1 α2 π Using the first condition, this implies
i.e. tan and tan are reciprocals of each other. Therefore + = and
2 2 2 2 2 f (A) = A
∠P MQ1 + ∠P MQ2 = α1 + α2 = π.
In the other case when c1 and c2 are on the different sides from l (see Figure 22), the for all A. Using the second condition again, we obtain the desired result:
maximal size of the angle Q1 MQ2 is π which is the greatest size an angle can have. In
this case, the point M lies on the common tangent to c1 and c2 intersecting l. Then f (A) = f (g(A)) = g(A).
∠P MQ1 + ∠P MQ2 = ∠Q1 MQ2 = π. Thus, every Automorian loves and honours himself.
2. On the planet Automory, there are infinitely many inhabitants. Every Automorian
3. Find all pairs (x, y) of positive integers satisfying the equation
loves exactly one Automorian and honours exactly one Automorian. Additionally, the
following can be noticed: (x + y)x = xy .

17 18
Answer: (2, 6) and (3, 6). We see that in both inequalities, equality actually holds. Consequently, x1 = x2 ,
Solution. We have xy = (x + y)x > xx implying y > x. Let y = nx where n > 1 is a 2
x01 = x02 = x03 , and − a = 4. From the latter, we find a = −6. Thus x01 x02 x03 = 8
rational number. From the equality given in the problem, we get 3
leading to x01 = x02 = x03 = 2 which gives b = 12.
(x + nx)x = xnx . On the other hand, taking a = −6, b = 12 gives 6x2 − 24x − 4a = 6(x − 2)2 and
x3 + ax2 + bx − 8 = (x − 2)3 satisfying the conditions of the problem.
1
Raise both sides to the power of and then divide them by x; we obtain
x 5. On a horizontal line, 2005 points are marked, each of which is either white or black.
1+n=x n−1
. (1) For every point, one finds the sum of the number of white points on the right of it and
the number of black points on the left of it. Among the 2005 sums, exactly one number
p occurs an odd number of times. Find all possible values of this number.
On the right hand side, the exponent n − 1 can be represented as a reduced fraction ,
q
therefore the number Answer: 1002.
p √ Solution. It is easy to see that the sums computed for a white point V and a black point
xn−1 = x q = q xp M immediately following V on its right are equal. Note also that the sums are equal
also if the two points of different colour lie in the opposite order. If one interchanges
is either natural or irrational. The left hand side of (1) cannot be irrational, thus it is
two consecutive points of different colour, only the two equal sums corresponding to
natural. Hence n is natural. By choice, n > 1.
these two points change giving rise to two new equal sums. Hence, for any k, such
If n = 2, (1) gives x = 3 implying y = 2x = 6. If n = 3, (1) gives x = 2 implying transitions preserve parity of the number of occurrences of k among the 2005 sums.
y = 3x = 6. Note that (1) implies x > 2, thus 1 + n > 2n−1 . The latter inequality does
Assume there are n white and 2005 − n black points on the line. By a sequence of
not hold in the case n > 4. Hence no more solutions exist.
transitions described, collect all white points to the left. Then, going from left to right,
the corresponding sums are
Second Day
n − 1, n − 2, . . . , 1, 0, 0, 1, . . . , 2003 − n, 2004 − n.

4. Find all pairs (a, b) of real numbers such that the roots of polynomials 6x2 −24x−4a According to the invariant discovered in the first paragraph, exactly one number must
and x3 + ax2 + bx − 8 are all non-negative real numbers. occur an odd number of times also in this sequence. As the middle numbers occur in
pairs, the single number occurring an odd number of times is either the leftmost n − 1
Answer: (−6, 12). or the rightmost 2004 − n. If the former case takes place, then n − 2 = 2004 − n leading
Solution. Let x1 , x2 be the roots of the first polynomial and x01 , x02 , x03 be the roots of the to n = 1003 and n − 1 = 1002. If the other case takes place, then n − 1 = 2003 − n, giving
2 n = 1002 and 2004 − n = 1002 just like in the first case.
other polynomial. Division of the first polynomial by 6 gives x2 − 4x − a whose roots
3
are x1 and x2 , too. By Viète’s formulae,
6. On a plane, line l and a circle having no common points are given. Let AB be the
2 diameter of the circle being perpendicular to l whereby B is nearer to l than A. Let C
x1 + x2 = 4, x1 x2 = − a be a point on the circle different from both A and B. Line AC intersects l at point D.
3
Points B and E, the latter obtained as the tangency point of a line drawn from D to the
and circle, lie on the same side from AC . Line EB intersects l at point F ; line F A intersects
the circle second time at point G. Prove that the point symmetric to G with respect to
x01 + x02 + x03 = −a, x01 x02 + x02 x03 + x03 x01 = b, x01 x02 x03 = 8. AB lies on F C .
Now Solution. See IMO-2004 Shortlist.
 2  2
4 x1 + x2 2
4= = > x1 x2 = − a
2 2 3

and
2 2 p
− a = (x01 + x02 + x03 ) > 2 3 x01 x02 x03 = 4.
3 3

19 20
 WE THANK:

Estonian Math Competitions

2005/2006
Tallinn City Government

Tartu City Government

Narva City Government

The Gifted and Talented Development Centre

Tartu 2006

Selected Problems from Open Contests
1. A farmer noticed that, during the last year, there were exactly as many calves born
as during the two preceding years together. Even better, the number of pigs born during
the last year was one larger than the number of pigs born during the two preceding
years together. The farmer promised that if such a trend will continue then, after some
years, at least twice as many pigs as calves will be born in his cattle, even though this far
this target has not yet ever been reached. Will the farmer be able to keep his promise?
(Juniors.)
Answer: no.
Solution. Let Fn and Gn be the numbers of calves and pigs born during the nth year,
respectively. We will prove that if the farmer’s promise has been false during the pre-
vious years, it will remain so after the nth year as well. From the problem statement,
we have Fn = Fn−1 + Fn−2 and Gn = Gn−1 + Gn−2 + 1. If the number of born pigs
was less than twice the number of born calves during the previous years, we must have
Gn−1 6 2Fn−1 − 1 and Gn−2 6 2Fn−2 − 1. Consequently,
Gn = Gn−1 + Gn−2 + 1 6 2Fn−1 − 1 + 2Fn−2 − 1 + 1 =
= 2( Fn−1 + Fn−2 ) − 1 = 2Fn − 1
or Gn 6 2Fn − 1, i.e. the promise will not be true during the nth year. Since the farmer
has kept the cattle for at least two years, we can claim by induction that the promise will
never come true.
2. Let ABCD be a parallelogram, M the midpoint of AB and N the intersection of CD
and the angle bisector of ABC. Prove that CM and BN are perpendicular iff AN is the
angle bisector of DAB. (Juniors.)
Solution. The triangle CNB is isosceles since ∠CNB = ∠ MBN = ∠CBN (see Figure 1).
Thus we have | NC| = | BC|.
Assume first CM ⊥ BN. Since BN is both bisector and altitude for triangle BMC, we
have | BM| = | BC|. Consequently | BM| = |CN |, implying that N is the midpoint of CD
and MN is a segment joining the midpoints of the sides of the parallelogram. Then we
must have | AM| = | DN | = | NC| and | AD | = | MN | = | BC|. Thus the sides of AMND
are equal and we have a rhombus. Its diagonal AN bisects DAM.
Assume now that AN bisects DAB. Then ∠ DN A = ∠ BAN = ∠ DAN, which implies
| DN | = | DA|. On the other hand, | DA| = |CB| = | NC|. Thus N is the midpoint of CD.
Since M is the midpoint of AB, we have that MBCN is a rhombus with the diagonals
CM and BN being perpendicular.
3. Does there exist a natural number with the sum of digits of its kth power being
equal to k, if a) k = 2004; b) k = 2006? (Juniors.)
Answer: a) no; b) no.
Solution. a) Assume there exists such a number a. Since the sum of digits of a2004 is
1 2
D N C
A M B
Figure 1
divisible by 3, a2004 is divisible by 3 and so is a. But then a2004 is also divisible by 9
making the sum of its digits divisible by 9. Since 2004 is not divisible by 9, the sum of
the digits of a2004 can not be equal to 2004, a contradiction.
b) Assume there exists such a number a. We will use the fact that a natural number and
its sum of digits give the same remainder when divided by 3. It follows that a2006 ≡
2006 ≡ 2 mod 3. On the other hand, the number a2006 = (a1003 )2 is a perfect square that
can not give the remainder 2 when divided by 3, a contradiction.
4. A 9 × 9 square is divided into unit squares. Is it possible to fill
each unit square with a number 1, 2, . . . , 9 in such a way that, when-
ever one places the tile so that it fully covers nine unit squares, the
tile will cover nine different numbers? (Juniors.)
Answer: no.
Solution 1. Assume that the numbers can be written in the required way. Put the tile over
the central square; w.l.o.g. we can assume that the numbers are placed like in Figure 2,
left. Next move the tile like in Figure 2, middle. Two upper left vacant squares can have
neither 6 nor 7. Thus we must have 8 and 9 there, in some order. Now place the tile like
in Figure 2, right. We can see that either way we must cover number 8 twice, hence the
required configuration of numbers does not exist.
8, 9
8, 9
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
6 8 6 8 6 8
7 9 7 9 7 9
Figure 2
Solution 2. Put the tile over the central square; w.l.o.g. we can assume that the numbers
are written like in Figure 3, left. We will analyse which number can be written into the
gray central square. Moving the tile one position left, we see that the central square can
not contain 1, 2, 3 or 4. Moving the tile one position right, we see that the central square
can not contain 5. Moving the tile one position down, we see that the central square can
not contain 6, 7, 8 or 9. Thus the required numbering is not possible.
5. Find all real numbers with the following property: the difference of its cube and
its square is equal to the square of the difference of its square and the number itself.
(Juniors.)
 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 A′ D′
6 8 6 8 6 8 6 8
7 9 7 9 7 9 7 9
A D

Figure 3 P
C
B

′
C
B′

Figure 5

Figure 4 Solution 2. Independently of the location of P, the equalities

| PA| · | PA′ | = | PC| · | PC′ |, | PB| · | PB′ | = | PD | · | PD ′ |

Answer: 0, 1 and 2.
Solution 1. Let x be a number with this property. Then x3 − x2 = ( x2 − x )2 , which are valid. Multiplying these, we get
leads to the equation x4 − 3x3 + 2x2 = 0 or x2 ( x2 − 3x + 2) = 0. Hence x2 = 0 or
x2 − 3x + 2 = 0. Solving these quadratic equations, we get x0 = 0, x1 = 1, x2 = 2. | PA| · | PB| · | PA′ | · | PB′ | = | PC| · | PD | · | PC′ | · | PD ′ |.
Solution 2. Transform the equation x3 − x2 = ( x2 − x )2 to obtain x2 ( x − 1) = x2 ( x − 1)2 . Points A, B, C, D are concyclic if and only if
Therefore x2 ( x − 1)2 − x2 ( x − 1) = 0 or x2 ( x − 1)( x − 2) = 0. Clearly, the solutions of
the last equation are 0, 1, 2. | PA| · | PB| = | PC| · | PD |

6. A solid figure consisting of unit cubes is shown in the picture. Is it
possible to exactly fill a cube with these figures if the side length of the
cube is a) 15; b) 30? (Juniors.)
Answer: a) no; b) yes.
Solution. a) Since the figure consists of four unit cubes, the number of unit cubes in every
solid composable from these figures is divisible by 4. Since the cube with side length 15
contains an odd number of unit cubes, this cube is not among these solids.
b) From two figures, it is possible to assemble a cube with side length 2 (see Figure 4).
From these cubes, it is possible to assemble a cube with side length 30.
7. Two non-intersecting circles, not lying inside each other, are drawn in the plane.
Two lines pass through a point P which lies outside each circle. The first line intersects
the first circle at A and A′ and the second circle at B and B′ ; here A and B are closer
to P than A′ and B′ , respectively, and P lies on segment AB. Analogously, the second
line intersects the first circle at C and C′ and the second circle at D and D ′ . Prove that
the points A, B, C, D are concyclic if and only if the points A′ , B′ , C′ , D ′ are concyclic.
(Juniors.)
Solution 1. Since A, A′ , C′ , C are concyclic (see Figure 5), we have ∠ AA′ C′ + ∠ ACC′ =
180◦ , hence ∠ B′ A′ C′ = ∠ ACD. Analogously ∠C′ D ′ B′ = ∠ DBA. Points A, B, C, D are
concyclic if and only if ∠ ACD = ∠ DBA, which is equivalent to ∠ B′ A′ C′ = ∠C′ D ′ B′ ,
the last equality holds if and only if points A′ , B′ , C′ , D ′ are concyclic.
or, taking into account the previous equality, if and only if
| PA′ | · | PB′ | = | PC′ | · | PD ′ |,
which holds if and only if A′ , B′ , C′ , D ′ are concyclic.
Note. As can be seen from solution 2, the assertion of the problem holds regardless of
positions of the circles.
8. A computer outputs the values of the expression (n + 1) · 2n for n = 1, n = 2, n = 3,
etc. What is the largest number of consecutive values that are perfect squares? (Juniors.)
Answer: 2.
Solution. Two consecutive values can be perfect squares, for example, for n = 7 and
n = 8 we get 8 · 27 = (25 )2 and 9 · 28 = (3 · 24 )2 .
Now prove that three consecutive values cannot be perfect squares. Assume that (n +
1) · 2n and (n + 3) · 2n+2 are both perfect squares. If n is even then both 2n and 2n+2
are perfect squares. Therefore also n + 1 and n + 3 must be perfect squares, which is
impossible. If n is odd, i.e. n = 2k + 1 for some k > 0, then (n + 1) · 2n = (2k + 2) ·
22k+1 = (k + 1) · 22k+2 and (n + 3) · 2n+2 = (2k + 4) · 22k+3 = (k + 2) · 22k+4 . Here, the
factors 22k+2 and 22k+4 are perfect squares, therefore also k + 1 and k + 2 must be perfect
squares, which is impossible for non-negative k.
9. Let a, b, c be positive integers. Prove that the inequality

( x − y ) a ( x − z)b (y − z)c > 0

3 4

Figure 6 Figure 7
holds for all reals x, y, z if and only if a, b, c are even. (Juniors.)
Solution. If a, b, c are positive then the inequality holds. On the other hand, assume that
the inequality holds for all reals x, y, z. Choosing z < x < y and dividing the given
inequality by a positive number ( x − z)b (y − z)c , we get the inequality ( x − y)a > 0,
from which we conclude due to x − y < 0 that a is even. Analogously, choosing y <
z < x, we conclude that c is even. Finally, choosing x < y < z, we get after dividing
the inequality by a positive number ( x − y)a (y − z)c that ( x − z)b > 0 from which we
conclude due to x − z < 0 that b is even.
10. All the streets in a city run in one of two perpendicular directions, forming unit
squares. Organizers of a car race want to mark down a closed race track in the city in
such a way that it would not go through any of the crossings twice and that the track
would turn 90◦ right or left at every crossing. Find all possible values of the length of
the track. (Seniors.)
Answer: all positive integers divisible by 4, except for 8.
Solution. Define natural coordinates with the origin at some crossing and consider two
consecutive track fragments of length 1. One of them is parallel to x-axis and the other
one to y-axis; moving along the first one, parity of the x-coordinate changes, and mov-
ing along the second one, parity of the y-coordinate changes. Moving along the track,
parities of x- and y-coordinates change alternately, but when returning to the initial
point, both parities must be the same as in the beginning. Since the pairs of parities
repeat after every four track fragments, the length of the track must be divisible by 4.
There exists a suitable track of length 4 going around one block. There is no track of
length 8. If there were one, it would have four fragments parallel to x-axis and four
fragments parallel to y-axis. Hence, we could not deviate more than 2 units in either
direction and the whole track should fit into 2 × 2 square. It is easy to see that the track
can not contain three corners of the square, but then there will be less that 8 possible
turning points left.
All the other positive integers divisible by 4 are attainable. Figure 6 shows a track of
length 12 and we can increase this length repeatedly by 4 units using the operation in
Figure 7.
11. After the schoolday is over, Juku must attend an extra math class. The teacher
writes a quadratic equation x2 + p1 x + q1 = 0 with integer coefficients on the blackboard
and Juku has to find its solutions. If they are not both integers, Juku may go home. If the
solutions are integers, then the teacher writes a new equation x2 + p2 x + q2 = 0, where
p2 and q2 are the solutions of the previous equation taken in some order, and everything
5 6
starts all over. Find all possible values for p1 and q1 such that the teacher can hold Juku
at school forever. (Seniors.)
Answer: either p1 is any integer and q1 = 0 or p1 = 1 and q1 = −2.
Solution. If q1 = 0 then we have the equation x2 + p1 x = 0 with solutions − p1 and 0.
The teacher can write another equation x2 − p1 x = 0 with solutions p1 ja 0, then again
x2 + p1 x = 0, etc. Thus all pairs ( p1 , 0) satisfy the conditions of the problem.
If q1 = −1 then the product of the solutions must be −1 and the solutions −1 and 1
in some order. Since the equations x2 − x + 1 = 0 and x2 + x − 1 = 0 have no integer
solutions, no pairs of the form ( p1 , −1) satisfy the conditions of the problem.
If q1 = −2 then the product of the solutions must be −2 and the solutions are either
2 and −1 or 1 and −2. In the first case, the teacher can choose between the equations
x2 + 2x − 1 = 0 and x2 − x + 2 = 0, none of them having integer solutions. In the second
case, the teacher can write the equation x2 + x − 2 = 0 with solutions 1 and −2. Thus
we see that the pair (1, −2) staisfies the conditions of the problem.
Now let q1 be an integer not equal to 0, −1 nor −2. If x1 and x2 are the solutions of
x2 + p1 x + q1 = 0, Viète formulae imply that x1 + x2 = − p1 , x1 x2 = q1 and
x12 + x22 = ( x1 + x2 )2 − 2x1 x2 = p21 − 2q1 < p21 + q21 .
Thus the sum of squares of the coefficients of the equation strictly decreases for q1 6∈
[−2; 0]. Since sum of squares is non-negative, we must sooner or later reach one of the
two situations: the solutions are not integers or the constant term belongs to the interval
[−2; 0]. The latter case is impossible, since every equation x2 + px + q = 0 uniquely
determines its predecessor x2 − ( p + q) x + pq = 0, hence the pairs ( p1 , 0) ja (1, −2) can
only arise from equations with constant terms 0 or −2, respectively. Thus there are no
other pairs of numbers satisfying the conditions of the problem.
12. Let ABC be an acute triangle and choose points A1 , B1 and C1 on sides BC, CA
and AB, respectively. Prove that if the quadrilaterals ABA1 B1 , BCB1 C1 and CAC1 A1 are
cyclic then their circumcentres lie on the sides of ABC. (Seniors.)
Solution. Since BCB1 C1 is cyclic (see Figure 8), we have ∠ BB1 C = ∠ BC1 C = α. Similar-
ily, let ∠CC1 A = ∠CA1 A = β and ∠ AA1 B = ∠ AB1 B = γ. Considering the angles with
vertices at points A1 , B1 and C1 , we get the following system of equations:
 β + γ = 180◦

γ + α = 180◦
α + β = 180◦ .

Adding the equalities and dividing by 2 gives α + β + γ = 270◦ , implying α = β = γ =
90◦ . Thus the segments BC, CA and AB are the diameters of the circles and contain their
circumcentres.
s1
13. Martin invented the following algorithm. Let two irreducible fractions and
t1
s2
be given as inputs, with the numerators and denominators being positive integers.
t2
Divide s1 and s2 by their greatest common divisor c and obtain a1 and a2 , respectively.
Similarily, divide t1 and t2 by their greatest common divisor d and obtain b1 and b2 ,
 A smallest on the board and strictly less than the other number in the pair. The situation
with all numbers being equal to zero can this way only occur after B’s move.
B1
Now let m be odd and n even. If none of the numbers is yet zero, B can ensure that after
C1
his move the following conditions hold: the smallest number on the board is even, the
quantity of the smallest numbers is odd and the number of occurrencies of every other
number is even. Indeed, if A changes the smallest number then B can change it again,
but if A changes some other number then B can change another number equal to the
B A1 C
one A changed. It goes on until some number becomes zero, afterwards B can divide
all the remaining numbers into pairs and use the strategy described above.
Finally, let m and n be odd. Then after A’s first move, there is a position described in
the previous paragraph on the board. Thus A can use the strategy of B and win.

Figure 8 15. Kati cut two equal regular n-gons out of paper. To the vertices of both n-gons, she
wrote the numbers 1 to n in some order. Then she stabbed a needle through the centres
a1 b2 + a2 b1 of these n-gons so that they could be rotated with respect to each other. Kati noticed
respectively. After that, form a new fraction , reduce it, and multiply the that there is a position where the numbers at each pair of aligned vertices are different.
t1 b2
numerator of the result by c. Martin claims that this algorithm always finds the sum of Prove that the n-gons can be rotated to a position where at least two pairs of aligned
the original fractions as an irreducible fraction. Is his claim correct? (Seniors.) vertices contain equal numbers. (Seniors.)
Answer: yes. Solution. Assume that the lower n-gon is fixed and move the upper n-gon. Let the initial
Solution. Since position of n-gons be the one found by Kati. For each vertex of the upper n-gon, there
  is an angle by which rotating clockwise the upper n-gon brings this vertex atop of the
s 1 t2 s 2 t1 vertex of the lower n-gon with the same number. There are n different vertices, but only
c· · + ·
c · (a1 b2 + a2 b1 ) c d c d s 1 t2 + s 2 t1 s s n − 1 different rotation angles since the angle 0◦ is excluded by conditions. Hence for
= = = 1 + 2,
t1 b2 t2 t1 t2 t1 t2 two vertices of the upper n-gon, the rotation angles are equal.
t1 ·
d
16. A real-valued function f satisfies for all reals x and y the equality
the resulting fraction has correct value. We still need to prove that it is irreducible. For
that, it is enough to show that the numbers c and t1 b2 are relatively prime. f ( xy) = f ( x )y + x f (y).
Suppose there exists a prime p dividing both c and t1 b2 . Since c = gcd(s1 , s2 ), we have
Prove that this function satisfies for all reals x and y 6= 0 the equality
that p also divides both s1 and s2 . Consequently, t1 and t2 can not be divisible by p,
s s2  
f ( x )y − x f (y )
because the fractions 1 and are irreducible. Thus p does not divide t1 t2 , implying f
x
= .
t1 t2 y y2
t2
that p can not divide t1 b2 = t1 · either, a contradiction.
d (Seniors.)
f ( xy) − x f (y)
14. Two players A and B play the following game. Initially, there are m equal positive Solution. From the given expression we obtain f ( x ) = , this equality
integers n written on a blackboard. A begins and the players move alternately. The y
x
player to move chooses one of the non-zero numbers on the board. If this number k is holds for any reals x and y 6= 0. Taking at place of x, we get
y
the smallest among all positive integers on the board, the player replaces it with k − 1;  
if not, the player replaces it with the smallest positive number on the board. The player x x
f · y − f (y )
who first turns all the numbers into zeroes, wins. Who wins if both players use their f ( x )y − x f (y )
 
x y y
best strategies? (Seniors.) f = = .
y y y2
Answer: A wins if mn is odd; B wins if mn is even. Note. It is possible to prove (for example, using reduction to the Cauchy equation), that
Solution. If the quantity m of numbers is even then B has the following winning strategy. the only continuous functions satisfying the given conditions are
B divides all the numbers into pairs and if A makes a move and changes some number, (
B changes the other number in the pair, ensuring that after his move all pairs contain ax ln | x |, if x 6= 0,
f (x) =
equal numbers. This is possible, since after A’s move the number A wrote must be the 0, if x = 0.

7 8
 F
O
E D
A
B C
G ( xi − x j )ai,j where 1 6 i < j 6 n. Prove that if the inequality K ( x1 , . . . , xn ) > 0 holds
Figure 9
17. Four points A, B, C, D are chosen on a circle in such a way that arcs AB, BC, and
CD are of the same length and the arc DA is longer than these three. Line AD and the
line tangent to the circle at B intersect at E. Let F be the other endpoint of the diameter
starting at C of the circle. Prove that triangle DEF is equilateral. (Seniors.)
Solution 1. Let O be the centre of the circle and let G be the point where BE intersects
the tangent drawn from D to the circle. Since the total length of arcs DA and AB is
larger than the total length of arcs BC and CD, the points G and E lie on different sides
from B (see Figure 9). Points B and D are symmetric with respect to the line CF, hence
1 which has negative base and odd exponent. So the whole product is negative.
G lies on the line CF and ∠ BGF = ∠ DGF. Now ∠ BED = (∠ BOD − ∠ AOB) =
2
1 in such a way that they are pairwise different and consider the product K ( x1 , . . . , xn ) as
∠COD = ∠CFD, therefore ∠GED = ∠GFD. Consequently G, E, F, D are concyclic.
2
Since ∠EGF = ∠ DGF, chords EF and DF of the circumcircle of quadrangle GEFD are
equal.
Solution 2. Since the lengths of the arcs BA and BC are equal, the lines EB and AC
are parallel. Therefore ∠ BED = ∠CAD = ∠ BDA. Thus triangle BED is equilateral.
Further, ∠CBF = 90◦ . By symmetry, the lines BC and AD are parallel, hence BF is
perpendicular to AD. Segment BF is the altitude of the equilateral triangle BED, it
bisects its base ED. This segment is also the altitude of the triangle DEF and it bisects
its base. This is possible only when the triangle DEF is equilateral.
Note. The assertion of the problem holds also in the case when the arc DA is shorter
than the other three, the point G then lies on the other side.
18. In the sequence (an ) with general term an = n3 − (2n + 1)2 , does there exist a term
that is divisible by 2006? (Seniors.)
Answer: yes.
Solution. First, a4 = 43 − 92 = −17 and a7 = 73 − 152 = 118. Since n3 − (2n + 1)2
is a polynomial, a4+17k is divisible by 17 and a7+118l is divisible by 118 for all natural
numbers k and l. Since 119 is divisible by 17, adding 118 decreases the remainder by 1
on division by 17. Therefore, when 361 = 7 + 118 · 3 is divided by 17, the remainder
9 10
is 7 − 1 · 3 = 4. This means that the term a361 is divisible by both 17 and 118, i.e. it is
divisible by 2006.
Note 1. Since 17 and 118 are relatively prime, the existence of the suitable n follows from
the Chinese Remainder Theorem: the remainders of n upon division by 17 and 118 must
be 4 and 7, respectively.
Note 2. The least number satisfying the given conditions corresponds to n = 87, in this
case a87 = 627878 = 313 · 2006.
19. Let n > 2 be a fixed integer and let ai,j (1 6 i < j 6 n) be some positive integers.
For a sequence x1 , . . . , xn of reals, let K ( x1 , . . . , xn ) be the product of all expressions
independently of the choice of the sequence x1 , . . . , xn then all integers ai,j are even.
(Seniors.)
Solution 1. Suppose the contrary: some of the numbers ai,j are odd. Let l be the smallest
index for which there are odd numbers among the numbers ai,l (1 6 i < l); also let k
be the largest index for which ak,l is odd. Then ak,l is the only odd number among the
numbers ai,j where k 6 i < j 6 l. Now choose x1 , . . . , xn as follows:
x1 > x2 > . . . > x k −1 > x l > x k +1 > . . . > x l −1 > x k > x l +1 > . . . > x n ;
i.e., choose some n numbers in decreasing order and swap the positions of the k-th and
the l-th number. Then in the given expression, the factors ( xi − x j )ai,j , where i < k or
j > l, are positive, since the bases of the power are positive. All the remaining factors
( xi − x j )ai,j where k 6 i < j 6 l have even exponents with the exception of ( xk − xl )ak,l ,
Solution 2. Assume that ak,l is odd for some indices k and l. Fix x1 , . . . , xk−1 , xk+1 , . . . , xn
a polynomial of one variable K ( xk ). The root xl of this polynomial has odd multiplicity.
Therefore the graph of K ( xk ) intersects the x-axis at xl and we can choose xk such that
K ( xk ) is negative, a contradiction.
Selected Problems from the Final Round of National
Olympiad
1. Find all pairs of positive integers (a, b) such that
ab = gcd(a, b) + lcm(a, b).
(Grade 9.)
Answer: (2, 2).
Solution. As the left-hand side and summand lcm(a, b) on the right-hand side are both
divisible by a, also gcd(a, b) has to be divisible by a. On the other hand, gcd(a, b) 6 a as
a is positive. Thus gcd(a, b) = a. Analogously we obtain that gcd(a, b) = b. Therefore
a = b and the equation has the form a2 = a + a or a(a − 2) = 0. The only positive
solution of the equation is a = 2 and thus also b = 2.
 C
√
3c 1 3c  c 2 3c2 c2
h: = tan 30◦ = √ , therefore h = . Hence b2 = h2 + = + = c2 ,
2 3 2 2 4 4
a
b b
a giving b = c. We now have that CDE is an equilateral triangle and its angles are 60◦ .
M h |CH | 1 | DH | 1
Solution 3. As = sin 30◦ = and = , segment CD satisfies the Angle
30 ◦ 30 ◦ |CA| 2 | DA| 2
A c D c H c E c B Bisector Property. Hence CD bisects angle ACH whose size is 60◦ and ∠ DCH = 30◦ .
2 2 Then the vertex angle of the isosceles triangle CDE is 60◦ and the base angles are also
60◦ .
Solution 4. Let the medians of triangle CDE meet at M. By the Ray Property, | DM| =
C′ 1 1 1
| AC| = | BC| = | EM|. We obtain |CH | = | AC| as in the previous solution and thus
3 3 2
Figure 10 1
|CM| = | AC|. So |CM| = | DM| = | EM|, i.e., the intersection point M of medians of
3
CDE is also the circumcentre of CDE. It follows that CDE is an equilateral triangle and
Solution 2. It is known that ab = gcd(a, b) lcm (a, b) for positive integers a and b. Thus all its angles are of size 60◦ .
we obtain the equality gcd(a, b) lcm (a, b) = gcd(a, b) + lcm(a, b) which is equivalent to Solution 5. Assume that ∠ DCE > 60◦ . Then ∠CED < 60◦ . Hence | DE| > |CD | giving
(gcd(a, b) − 1) (lcm(a, b) − 1) = 1. The last equality expresses 1 as a product of two non- also | AD | > |CD |. Then in triangle ACD, we have ∠ ACD > 30◦ , and in triangle
negative integers; this is only possible if both of them are equal to 1. Thus gcd(a, b) − BCE by symmetry, ∠ BCE > 30◦ . Adding the inequalities, we get ∠ ACB > 120◦ —
1 = 1 and lcm(a, b) − 1 = 1 or gcd(a, b) = 2 and lcm(a, b) = 2. Hence a = b = 2. a contradiction. Analogously we obtain a contradiction by starting with assumption
Solution 3. Let gcd(a, b) = d. We represent integers a and b as a = da′ and b = db′ ∠DCE < 60◦ . Hence ∠DCE = 60◦ and thus ∠CDE = ∠CED = 60◦ .
where a′ and b′ are relatively prime. Then lcm(a, b) = da′ b′ . The equation gets the
1 4. Consider a rectangular grid of 10 × 10 unit squares. We call a ship a figure made
form d2 a′ b′ = d + da′ b′ , or a′ b′ = . As a′ b′ is an integer, we must have d − 1 = 1
d−1 up of unit squares connected by common edges. We call a fleet a set of ships where no
and a′ b′ = 1. Hence d = 2 and a′ = b′ = 1. Therefore the solution of the equation is two ships contain squares that share a common vertex (i.e. all ships are vertex-disjoint).
a = da′ = 2 and b = db′ = 2. Find the least number of squares in a fleet to which no new ship can be added. (Grade
9.)
2. Let there be n > 2 real numbers such that none of them is greater than the arith-
metic mean of the other numbers. Prove that all the numbers are equal. (Grade 9.) Answer: 16.
Solution. Call a fleet full if no new ships can be added. We have to find the least number
Solution. Let a be the greatest among the numbers. Suppose, by contradiction, that not
of squares in a full fleet.
all the numbers are equal. Then there must be some numbers less than a. Considering
First we show that a full fleet covering 16 unit squares exists. Put on the grid 16 one-
numbers other than a, we therefore know that their arithmetic mean is less than a. But
square ships as shown in Figure 11. Note that then each square of the grid has a common
this contradicts the conditions of the problem. Hence all the numbers are equal.
vertex with one of those ships and thus no ship can be added.
3. Triangle ABC is isosceles with AC = BC and ∠C = 120◦ . Points D and E are Second we prove that there can not be fewer than 16 unit squares in a full fleet. Suppose
chosen on segment AB so that | AD | = | DE| = | EB|. Find the sizes of the angles of a full fleet is fixed. Consider the set of 16 unit squares painted gray in Figure 11. For
triangle CDE. (Grade 9.) each of these 16 squares, there is a square of the full fleet that shares (at least) a common

Answer: all the angles are 60◦ .

Solution 1. The base angle of the isosceles triangle ABC is (180◦ − 120◦ ) : 2 = 30◦ . Let
H be the foot of the altitude drawn from vertex C (see Figure 10). Reflect the triangle
ABC with respect to side AB, the point C going to C′ . As ∠CAC′ = 60◦ and | AC| =
| AC′ |, triangle ACC′ is equilateral and AH is its median. Moreover, point D divides the
median with ratio 2 : 1. Thus the medians of ACC′ meet at D, and CD is both a median
and an angle bisector to ACC′ . We obtain ∠ DCH = 30◦ and ∠ DCE = 60◦ . As CDE is
isosceles, ∠CDE = ∠CED = 60◦ .
Solution 2. Denote |CA| = |CB| = a, |CD | = |CE| = b and | AD | = | DE| = | EB| = c for
brevity. Let h be the altitude drawn from vertex C to side AB. In triangle CH A, we have Figure 11

11 12

Figure 12 Figure 13
vertex with it. All these squares of the fleet must be different. Hence there are at least
16 squares in the fleet.
5. Consider a rectangular grid of 10 × 10 unit squares. We call a ship a figure made
up of unit squares connected by common edges. We call a fleet a set of ships where no
two ships contain squares that share a common vertex (i.e. all ships are vertex-disjoint).
Find the greatest natural number that, for each its representation as a sum of positive
integers, there exists a fleet such that the summands are exactly the numbers of squares
contained in individual ships. (Grade 10.)
Answer: 25.
Solution. First we prove that, for all n > 25, we can divide n into summands so that a
fleet with respective ship sizes can not be put on the grid. In particular, we prove that
one can not put more than 25 ships of size 1 on the grid. Let us divide the grid into
squares of 2 × 2 (as in Figure 12), there are 25 of them. As each 2 × 2 square can contain
at most one ship of size 1, then the total number of such ships is at most 25.
Second we show that, for any representation of 25 as a sum of positive integers, there is a
fleet with respective ship sizes. Let us initially put 25 ships of size 1 on the grid as shown
in Figure 13. Then, starting from the upper left corner, shift along the line together
as many ships as the first summand of the representation tells; then shift together as
many ships as the second summand tells etc. The fleet obtained this way satisfies the
conditions of the problem.
6. Find the greatest possible value of sin(cos x ) + cos(sin x ) and determine all real
numbers x, for which this value is achieved. (Grade 11.)
Answer: the greatest possible value is sin 1 + 1, which is achieved iff x = 2kπ, where k
is an arbitrary integer.
Solution. Since the value of cos x is in the interval [−1; 1] and since sin x is increasing
in this interval, the greatest possible value of the first addend is sin 1, and the value is
achieved iff cos x = 1, or x = 2kπ, where k is an arbitrary integer. The greatest possible
value of the second addend is 1, which is achieved iff sin x = 0, or x = lπ, where l is
an arbitrary integer. Both terms achieve the maximal value simultaneously iff x = 2kπ,
where k is an arbitrary integer; the value of the expression is then sin 1 + 1.
7. In a right triangle, the length of one side is a prime and the lengths of the other
side and the hypotenuse are integral. The ratio of the triangle perimeter and the incircle
13
diameter is also an integer. Find all possible side lengths of the triangle. (Grade 11.)
Answer: 3, 4 and 5.
Solution. Let p and m be the lengths of the sides of the triangle and let n be the length of
the hypotenuse, where p is a prime (Figure 14). Then p2 + m2 = n2 , implying p2 = (n −
m)(n + m). Since p is prime, the terms of the right-hand side must satisfy n − m = 1,
n + m = p2 .
The perimeter of the triangle is p + m + n. In order to find the incircle diameter d, we
note that the total length of the segments tangent to the incircle, originating from the
vertex of the right angle, equals the length of the incircle diameter, whereas the total
length of the four tangents originating from the other two vertices is 2n. Thus, d + 2n =
p + m + n. According to the assumption, p + m + n is divisible by d = p + m − n.
Substituting m + n and m − n from above, we get that p + p2 = p( p + 1) is divisible by
p − 1. Since p and p − 1 are coprime, it must be that p + 1 is divisible by p − 1. Hence,
p − 1 = 1 or p − 1 = 2. We see that p = 2 is impossible, since n − m and n + m cannot
be of different parity. Thus, p = 3, m = 4 and n = 5. A triangle with side lengths 3, 4, 5
is clearly a right triangle.
p m
n
Figure 14
8. The sequence ( Fn ) of Fibonacci numbers satisfies F1 = 1, F2 = 1 and Fn = Fn−1 +
Fn−2 for all n > 3. Find all pairs of positive integers (m, n), such that Fm · Fn = mn.
(Grade 11.)
Answer: (1, 1), (1, 5), (4, 6), (5, 1), (5, 5) and (6, 4).
Solution. By induction on n, it is easy to prove that Fn > n for all n > 6 and Fn > 2n for
all n > 8. Thus, if m > 6 and n > 6, then Fm · Fn > m · n. W.l.o.g., we may now assume
m 6 5 (the remaining solutions can be obtained by exchanging m and n).
• If m = 1, then 1 · Fn = 1 · n, or Fn = n. From above, the only solutions are n = 1
and n = 5 and the suitable pairs are (1, 1), (1, 5), (5, 1) and (5, 5).
• If m = 2, we get 1 · Fn = 2 · n, or Fn = 2n. Since there are no solutions for n < 8,
there are no solutions at all.
3 3
• If m = 3, we get 2 · Fn = 3 · n, or Fn = n. Since n < 2n, there are no solutions, as
2 2
in the previous case.
4
• If m = 4, then 3 · Fn = 4 · n, or Fn = n. Here, the only solution is n = 6, giving
3
pairs (4, 6) and (6, 4).
14
 • If m = 5, then 5 · Fn = 5 · n, or Fn = n; this case is analysed above.
9. In a triangle ABC with circumcentre O and centroid M, lines OM and AM are
perpendicular. Let AM intersect the circumcircle of ABC again at A′ . Let lines BA′ and
AC intersect at D and let lines CA′ and AB intersect at E. Prove that the circumcentre
of triangle ADE lies on the circumcircle of ABC. (Grade 11.)
A
H G
O
M
B C
F
A′
D
E
Figure 15
Solution. Let F, G and H be the base points of the medians drawn from vertices A, B
and C, respectively (see Figure 15). Then, triangle A′ OA is isosceles with height OM
and | A′ M| = | MA|. Since the centroid divides a median in ratio 2 : 1, we get | FM| =
1 n2 − n
| MA|, and | A′ F| = | FM|. On the other hand, | BF| = | FC|. Hence, A′ BMC is a
2 2
parallelogram. Parallel sides then imply that triangles ABD and AHC are similar with
similarity ratio 2 — the ratio of the lengths of AB and AH. Analogously, triangles ACE
and AGB are similar with the same ratio. Homothecy with centre A and ratio 2 brings
triangle ABC to triangle AED, while the circumcentre O of triangle ABC is transformed
to the second intersection point of AO and the said circumcircle.
Note. The use of homothecy can be avoided by finding the second intersection point
P of line AO and the circle and by proving that | AP| = 2| AO|, | DP| = 2|CO| and
| EP| = 2| BO|.
10. A pawn is placed on a square of a n × n board. There are two types of legal
moves: (a) the pawn can be moved to a neighbouring square, which shares a common
side with the current square; or (b) the pawn can be moved to a neighbouring square,
which shares a common vertex, but not a common side with the current square. Any
two consecutive moves must be of different type. Find all integers n > 2, for which it is
possible to choose an initial square and a sequence of moves such that the pawn visits
each square exactly once (it is not required that the pawn returns to the initial square).
(Grade 11.)
Answer: n = 2k, where k is an arbitrary positive integer.
15
Solution. First, we prove that a suitable sequence of moves exists for even n. Divide the
board into blocks of 2 × 2 squares (see Figure 16) and place the pawn on the upper left
corner square. To move through the first block, take the following moves: down-right,
up, down-left, down. Repeat the same combination of moves until the pawn reaches the
bottom-most block in a column of blocks. In the bottom-most block, move down-right,
left, up-right, and right; the first column of blocks is passed. In the bottom-most block
of the second column, move down-right, left, up-right, up, and continue by moving
upwards block by block. By passing the columns alternatingly up and down, the pawn
visits each square exactly once.
Figure 16 Figure 17
Next, we prove that a sequence does not exist for odd n. Colour the second, fourth,
n2 − n
sixth, etc. row dark (see Figure 17), then there are dark squares. Note that every
2
diagonal move starts from a dark square or ends on a dark square. Since a suitable
sequence does not visit a square twice and does not contain two consecutive diagonal
moves, each diagonal move corresponds to a different dark square. Thus, we can make
n2 − n
at most diagonal moves and, consequently, at most + 1 non-diagonal
2
moves, or n − n + 1 moves altogether. For n > 3, this number is smaller than n2 − 1,
2
the number of moves required to visit all squares.
11. We call a ship a figure made up of unit squares connected by common edges.
Prove that if there is an odd number of possible different ships consisting of n unit
squares on a 10 × 10 board, then n is divisible by 4. (Grade 12.)
Solution. Let n be such that the number of possible different ships of n squares is odd.
Divide all ships in classes, such that all ships in the same class are precisely those that
can be obtained from one another by shifts, vertical and horizontal reflections. Then
there must exist a class with an odd number of ships.
Let L be a ship in such a class. Assume that L is not symmetrical w.r.t. either the vertical
or horizontal axis of symmetry of its tight bounding box. Then no ship in this class is
symmetrical w.r.t. this axis. Thus, we can divide all ships in this class into pairs: a ship
and its reflection from this axis; a contradiction with the odd cardinality of this class.
Therefore, L must be symmetrical w.r.t. both the vertical and the horizontal axis.
The side lengths of the rectangle bounding L must be even, for there is an even number
of ways to place a rectangle with an odd side on a board with an even side length
10; again a contradiction in parity. Thus, the vertical and horizontal axes of symmetry
16
divide the squares of L into four disjoint reflection symmetrical sets. Since there is an b) Assume that for some n, there exist n suitable integers a1 < a2 < . . . < an . Then, on
equal number of squares in each set, the total number of squares of the ship is divisible the one hand,
by 4.
a1 + a2 + . . . + a n − 2 < a n − 2 + a n − 2 + . . . + a n − 2 = ( n − 2 ) a n − 2
12. Find the smallest possible distance of points P and Q on a xy-plane, if P lies on
but, on the other hand,
the line y = x and Q lies on the curve y = 2x . (Grade 12.)
1 + ln ln 2 a n −1 a n > ( n − 1) a n > ( n − 2) a n −2 .
Answer: √ .
2 ln 2 Thus, a1 + a2 + . . . + an−2 < an−1 an , and the sum of a1 , a2 , . . . , an−2 can not be divisible
Solution. We find the minimum point of h( x ) = 2x − x. Since h′ ( x ) = 2x ln 2 − 1, we get by the product of an−1 and an .
2x ln 2 − 1 = 0, and Solution 2. a) Choose arbitrary pairwise distinct numbers b1 , b2 , . . . , bn and let m be
1 ln ln 2 the least common multiple of all sums of (n − 2) terms. For each i = 1, 2, . . . , n, take
2x = and x=− . ai = mbi . Then the product of any two numbers ak al is ak al = (mbk ) · (mbl ), which is
ln 2 ln 2
divisible by the sum of the remaining numbers, since m2 is divisible by this sum.
Since h′ ( x ) is increasing, this is indeed a minimum. The value of h at this point is
14. Let O be the circumcentre of an acute triangle ABC and let A′ , B′ and C′ be the
1 ln ln 2 1 + ln ln 2 circumcentres of triangles BCO, CAO and ABO, respectively. Prove that the area of
h( x ) = + = .
ln 2 ln 2 ln 2 triangle ABC does not exceed the area of triangle A′ B′ C′ . (Grade 12.)
Here 1 + ln ln 2 = ln(e ln 2). Since 2 < e < 4, we get ln(e ln 2) > ln(2 ln 2) = ln ln 4 > Solution 1. First, we prove that for a fixed circumcircle, a triangle with maximal area
> ln ln e = 0. Thus, the value of h at the minimum is positive, so the graph of g( x ) = 2x is equilateral. Assume that a triangle KLM with maximal area has two sides of un-
is always higher than the graph of f ( x ) = x. Consider points A( x, f ( x )) and B( x, g( x )) equal lengths, say, KM and LM. Take a point M ′ on the circumcircle of KLM such that
and let C be the projection of B to the graph of f . Then, triangle ABC is a right isosceles |KM′ | = | LM′ | (Figure 18). Triangles KLM and KLM′ have a common base but the
| AB| h( x ) altitude of the first triangle is smaller, a contradiction.
triangle, since ∠ BAC = 45◦ . Consequently, | BC| = √ = √ and the distance | BC|
2 2
is minimal iff h( x ) is minimal. The sought distance is thus M
M′
M
Q
h( x ) 1 + ln ln 2
√ = √ . M′
2 2 ln 2
log2 e − log2 log2 e
Note. The answer can be expressed in many different ways, e.g., by √ .
2 X

13. Prove or disprove the following statements.
a) For every integer n > 3, there exist n pairwise distinct positive integers such that
the product of any two of them is divisible by the sum of the remaining n − 2 num-
bers.
b) For some integer n > 3, there exist n pairwise distinct positive integers, such that
the sum of any n − 2 of them is divisible by the product of the remaining two num-
bers.
(Grade 12.)
Answer: a) true; b) false.
Solution 1. a) Take n numbers (n2 )!, 2(n2 )!, 3(n2 )!, . . . , n(n2 )!. The product of any two
of these numbers is divisible by (n2 )!(n2 )!, whereas the sum of the remaining numbers
is k(n2 )!, where k is some positive integer smaller than 1 + 2 + . . . + n, which in turn is
smaller than n2 .
P
K L K L L′
Figure 18 Figure 19
Next, we prove that for a fixed incircle, a triangle with minimal area is equilateral.
Assume w.l.o.g. that in a triangle KLM with minimal area, ∠KLM > ∠KML. Con-
sider a triangle KL′ M′ , where points L′ and M′ lie on lines KL and KM such that
∠KL′ M′ = ∠KM′ L′ and L′ M′ is tangent to the incircle (Figure 19). Let X be the in-
tersection point of lines LM and L′ M′ . Draw perpendicular segments LP and MQ from
points L and M to line L′ M′ . Then | XM′ | > | XL′ | and | MQ| > | LP|, since | XQ| > | XP|
and right triangles MQX and LPX are similar. Thus, the area of triangle MM ′ X is
greater than the area of triangle LL′ X and consequently, the area of KLM is greater than
the area of KL′ M′ , a contradiction.
Now, let R be the circumradius of ABC. Since the sides of triangle A′ B′ C′ are perpen-
dicular to OA, OB and OC (see Figure 20) and bisect these segments, point O is the

17 18
 A
15. The Ababi alphabet consists of letters A and B, and the words in the Ababi lan-
C ′ guage are precisely those that can be formed by the following two rules:

B′
1) A is a word.

2) If s is a word, then s ⊕ s and s ⊕ s̄ are words, where s̄ denotes a word that is obtained
β
O β by replacing all letters A in s with letters B, and vice versa; and x ⊕ y denotes the
αα concatenation of x and y.

The Ululu alphabet consists also of letters A and B and the words in the Ululu language
B C are precisely those that can be formed by the following two rules:
A′
1) A is a word.
Figure 20 2) If s is a word, then s ⊗ s and s ⊗ s̄ are words, where s̄ is defined as above and x ⊗ y
is a word obtained from words x and y of equal length by writing the letters of x
R and y alternatingly, starting from the first letter of x.
incentre of A′ B′ C′ and the inradius is . An equilateral triangle with circumradius R
2 R
and an equilateral triangle with inradius have equal area S by the property of the Prove that the two languages consist of the same words. (Grade 12.)
2
centroid. From above, we now get S ABC 6 S 6 S A′ B′ C ′ . Solution. Since each step doubles the length of a word, both languages contain only
Solution 2. Let R be the circumradius of ABC and let α, β and γ be the angles at the words of length 2n , where each such word has been obtained in exactly n steps.
vertices of the triangle. Then ∠ BOC = 2α, ∠COA = 2β, and ∠ AOB = 2γ. Thus, First, we show that each language contains 2n words that can be obtained in exactly n
steps. Indeed, in 0 steps, we obtain only the word A in both languages. Every k-step
R2 word gives two different (k + 1)-step words and any two different k-step words give
S ABC = SBOC + SCOA + S AOB = (sin 2α + sin 2β + sin 2γ).
2 different (k + 1)-step words, since the initial word is always a part of the new word.
Since triangle BOC is isosceles, the midperpendicular OA′ of side BC is also an an- Thus, the number of k + 1-step words is twice the number of k-step words. Now, it
gle bisector, and ∠ BOA′ = ∠COA′ = α. Similarly, ∠COB′ = ∠ AOB′ = β and suffices to prove that every Ababi word is an Ululu word.
R Any 0-step Ababi word is clearly an Ululu word. Assume that the claim holds for all
∠ AOC′ = ∠BOC′ = γ. Consider triangle B′ OC′ . The height drawn to side B′ C′ is , k-step Ababi words and consider a k + 1-step word t. Then, for some word s in the
2
R R2 Ababi language, t = s ⊕ s or t = s ⊕ s̄. By the induction hypothesis, there must exist a
so | B′ C′ | = (tan β + tan γ) and the area of this triangle is SB′ OC ′ = (tan β + tan γ). sequence of k operations by the Ululu rules that allows to construct the word s from the
2 8
R2 R2 word A.
Analogously, SC ′ OA′ = (tan γ + tan α) and S A′ OB′ = (tan α + tan β). Thus,
8 8
• If the word t is obtained in the Ababi language by the rule t = s ⊕ s, apply the
R2 aforementioned sequence of Ululu rules to the word AA. It is easy to see that after
S A′ B′ C ′ = SB′ OC ′ + SC ′ OA′ + S A′ OC ′ = (tan α + tan β + tan γ).
4 each step, the new word is of the form a ⊕ a, where a is the corresponding inter-
h πi mediate word in the construction process of s, since each Ululu operation has the
Function f ( x ) = tan x − 2 sin 2x is concave in the interval 0; , since the second same effect on the two equal halves of a word.
2
derivative
• If the word t is obtained by the rule t = s ⊕ s̄, apply the aforementioned sequence of
2 sin x
f ′′ ( x ) = + 8 sin 2x Ululu rules to the Ululu word AB. In this case, after each step, the obtained word is
cos3 x a ⊕ ā, where a is the corresponding intermediate word in the construction process
is non-negative in this interval. Jensen’s inequality now gives of s, since if the two halves of a word are “negations” of each other, any Ululu
    operation preserves this property.
α+β+γ π 2π
f (α ) + f ( β) + f (γ ) > 3 f = 3 tan − 2 sin = 0, Consequently, t is a word in the Ululu language.
3 3 3
so tan α + tan β + tan γ > 2 sin 2α + 2 sin 2β + 2 sin 2γ. The final equality yields S A′ B′ C ′ >
S ABC .

19 20
 WE THANK:

Estonian Math Competitions

2006/2007

The Gifted and Talented Development Centre

Tartu 2007

Selected Problems from Open Contests
1. Find all positive integers N with at most 4 digits such that the number obtained by
reversing the order of digits of N is divisible by N and differs from N. (Juniors.)
Answer: 1089 and 2178.
Solution. Let N be any number satisfying the conditions. Let N ′ be the number with
digits in reversed order. Then N ′ = k · N for some integer k ≥ 2. Clearly, N must have
at least 2 digits. It cannot end with 0 as otherwise N ′ < N, which is impossible. Let
N = a . . . b and N ′ = b . . . a where a and b are digits. Then b ≥ ka ≥ 2a. On the other
hand, a multiple of b ends with a. The pairs (a, b) that meet all conditions imposed by
the observations made so far are the following:
(1, 3), (1, 7), (1, 9), (2, 4), (2, 6), (2, 7), (2, 8), (2, 9), (3, 7), (3, 9), (4, 8), (4, 9).
For pairs (1, 3), (2, 4), (2, 7), (2, 9), (3, 7), (3, 9), (4, 8), (4, 9), the conditions ka ≤ b and
kb ≡ a (mod 10) are contradictory. Among the remaining pairs, (1, 7) gives k = 3,
while the first digit of 3 · 1 . . . 7 cannot be 7. Analogously, (2, 6) gives k = 2, while the
first digit of 2 · 2 . . . 6 cannot be 6. Hence only the pairs (1, 9) and (2, 8) are possible.
As 91 is not divisible by 19 and 82 is not divisible by 28, no two-digit numbers satisfy
the conditions. For three-digit numbers, there must be either N = 1c9 and N ′ = 9c1
with k = 9 or N = 2c8 and N ′ = 8c2 with k = 4. The first case leads to equality
900 + 10c + 1 = 9 · (100 + 10c + 9) which gives 80c + 80 = 0, so no suitable c exists. The
second case fails analogously.
In the case of four-digit numbers, we have either N = 1cd9 and N ′ = 9dc1 with k = 9
or N = 2cd8 and N ′ = 8dc2 with k = 4. In the first case, the equality 9000 + 100d +
10c + 1 = 9 · (1000 + 100c + 10d + 9) leads to d = 89c + 8. The only solution in digits is
c = 0, d = 8. In the second case, the equality 8000 + 100d + 10c + 2 = 4 · (2000 + 100c +
10d + 8) leads to 2d = 13c + 1. The only solution in digits is c = 1, d = 7.
2. Call a scalene triangle K disguisable if there exists a triangle K ′ similar to K with
two shorter sides precisely as long as the two longer sides of K, respectively. Call a
disguisable triangle integral if the lengths of all its sides are integers.
(a) Find the side lengths of the integral disguisable triangle with the smallest possible
perimeter.
(b) Let K be an arbitrary integral disguisable triangle for which no smaller integral
disguisable triangle similar to it exists. Prove that at least two side lengths of K are
perfect squares.
(Juniors.)
Answer: (a) 4, 6, 9.
Solution. (a) Let K be a triangle with side lengths a, b, c, where a < b < c. We show that
K is disguisable iff b2 = ac. First let K be disguisable and let K ′ be a similar triangle to K
1 2
with side lengths a′ , b′ , c′ where a′ = b and b′ = c. Then
b a′ b′ c
= = = ,
a a b b
c2
hence b2 = ac. Second let K satisfy b2 = ac. Taking a′ = b, b′ = c and c′ = , we see
b
that
a′ b′ c′
= = ,
a b c
i.e., triangle K ′ with side lengths a′ , b′ , c′ is just like required by the definition of disguis-
ability.
If a, b, c are integers then obviously b ≥ 2. A case study of possible values of b shows
that, for b = 2, . . . , 5, there exist no integers a and c such that a < b < c and ac = b2 and
c < a + b. For b = 6, we can take a = 4 and c = 9, giving perimeter 19. Thus we have
one more condition: a + b + c ≤ 19. As a ≥ 1 and c ≥ b + 1, this implies 2b + 2 ≤ 19
and b ≤ 8. So it suffices to check that for b = 7 and b = 8, no integers a and c such that
a < b < c and ac = b2 and c < a + b exist.
(b) Let K be a triangle satisfying the conditions of the problem. Then gcd(a, b, c) = 1
and b2 = ac. This implies that gcd(a, c) = 1 (as each common prime divisor of a and c
would also divide b). Thus, both a and c are perfect squares.
3. In a school tennis tournament with m ≥ 2 participants, each match consists of 4
sets. A player who wins more than half of all sets during a match gets 2 points for
this match. A player who wins exactly half of all sets during the match gets 1 point,
and a player who wins less than half of all sets gets 0 points. During the tournament,
each participant plays exactly one match against each remaining player. Find the least
number of participants m for which it is possible that some participant wins more sets
than any other participant but obtains less points than any other participant. (Juniors.)
Answer: 6.
Solution. Let m = 5. A participant who wins more sets than any other during the
tournament must win more than half of all sets he plays. This implies that he must win
more sets than his opponent in at least one match, i.e., he must win at least one match.
But in order to obtain less points than anyone else, he must lose more matches than
win. As each participant plays 4 matches, this special participant must win exactly one
match and lose at least two. Under such conditions, he can win at most 8 sets during
the tournament but this is not more than half of the number 16 of all sets.
Thus, for m = 5, the described situation is impossible. If it were possible for some m
such that m < 5, we could obtain a suitable tournament table also for m = 5 by adding
an appropriate number of players whose matches all end in draw.
The following table shows a situation for m = 6 where all conditions are fulfilled:
Player Marks Sets won
1. 4:0 4:0 1:3 1:3 1:3 4 11
2. 0:4 2:2 3:1 2:2 3:1 6 10
 n
3. 0:4 2:2 2:2 3:1 2:2 5 9 by exactly students. Thus k and n must be even.
4. 3:1 1:3 2:2 2:2 2:2 5 10 2
It remains to show that the situation is possible for arbitrary even k and n. For this,
5. 3:1 2:2 1:3 2:2 2:2 5 10 enumerate the questions by numbers from 1 to k and the students by numbers 1 to
6. 3:1 1:3 2:2 2:2 2:2 5 10 n. Let every student with an odd number answer correctly exactly the questions with
an odd number and every student with an even number answer correctly exactly the
4. Call a k-digit positive integer a hyperprime if all its segments consisting of 1, 2, . . . , questions with an even number. The requirements are fulfilled.
k consecutive digits are prime. Find all hyperprimes. (Juniors.)
6. Let an = 1 + 2 + . . . + n for every n ≥ 1; the numbers an are called triangular. Prove
Answer: 2, 3, 5, 7, 23, 37, 53, 73, 373. that if 2am = an then a2m−n is a perfect square. (Seniors.)
Solution. One-digit hyperprimes are precisely the one-digit primes 2, 3, 5, 7.
In a larger hyperprime, all digits must be prime. The last digit can be neither 2 nor Solution. We depict an as a set of points orga- (n − m) 2
5, and no two consecutive digits can be equal (this would form a composite segment). nized triangularly as shown in Fig. 1. From
Adding one digit to all one-digit primes, while following these requirements, we obtain two ends of the base, separate two triangles

|
numbers 23, 27, 37, 53, 57, 73. Among these, only 23, 37, 53, 73 are primes. both containing am points. For counting 2am
Note that all segments of any hyperprime are hyperprime. Thus all three-digit hyper- points, we count the points in the intersec-
primes can be obtained from two-digit hyperprimes by adding one digit to the end. tion of the two triangles twice, while leav-
Following the requirements above, we get hyperprime candidates 237, 373, 537, 737, ing the points in the upper rhomboid un- am

{z
among which only 373 is really a prime and a hyperprime. counted; for counting an points, every point
is taken into account once. Thus if 2am = an
Hyperprimes with more than 3 digits are impossible since their every segment of 3
then the intersection contains as many points
digits should be 373.
as the rhomboid. The former contains a2m−n
points while the latter contains (n − m)2 .

}
5. In an exam with k questions, n students are taking part. A student fails the exam
if he answers correctly less than half of all questions. Call a question easy if more than Remark. One can also prove the claim al- | {z
a 2m −n
}

half of all students answer it correctly. For which pairs (k, n) of positive integers is it gebraically using the formula of the sum of
possible that arithmetic progression. Figure 1
(a) all students fail the exam although all questions are easy; 7. Three circles with centres A, B, C touch each other pairwise externally, and touch
(b) no student fails the exam although no question is easy? circle c from inside. Prove that if the centre of c coincides with the orthocentre of triangle
ABC, then ABC is equilateral. (Seniors.)
(Juniors.)
Solution 1. Let the tangent point of circles with centres A and B be C′ , the tangent point
Answer: (a) there are no such pairs; (b) all pairs (k, n) with both k and n even.
of circles with centres B and C be A′ and the tangent point of circles with centres C and
Solution. Let v be the total number of correct answers given by all students.
A be B′ (see Fig. ??). Let A′′ , B′′ , C′′ be the tangent points of circle c with the circles with
k
(a) If all students fail then each of them gives less than correct answers, i.e., v < centre A, B, C, respectively. Let H and I be the orthocentre and the incentre of ABC,
2 respectively. Assume that H is the centre of c.
k nk n
n· =
2 2
. If all questions are easy then, for each of them, more than correct answers
2 We prove first that triangle ABC is acute. Line H A′′ passes through A and contains both
n nk a radius of c and the altitude of ABC drawn from A. If angle BAC were not acute then
are given, i.e., v > k · = , contradiction. the orthocentre of ABC would be on ray AA′′ while the centre of c would be outside
2 2
k k this ray since | AA′′ | < | H A′′ |. Analogously, the other angles of ABC must be acute.
(b) If no student fails then each of them gives at least correct answers, i.e., v ≥ n · =
2 2 Draw tangents to c from points A′′ , B′′ , C′′ . As ABC is acute, point H lies inside it. Hence
nk k each of the three arcs of c with endpoints A′′ , B′′ , C′′ is less than 180◦ . Consequently,
; the equality holds iff each student gives exactly correct answers. On the other
2 n 2 these tangents intersect each other pairwise, forming a triangle DEF whose incircle is c.
hand, if no question is easy then no more than correct answers are given to each of
n nk 2 n As both BC and EF are perpendicular to H A′′ , they are parallel. Analogously, CA and
them, i.e., v ≤ k · = , whereby equality holds iff each question gets exactly correct FD are parallel, and AB and DE, too. Thus triangles ABC and DEF are similar.
2 2 2
nk Prove now that the orthocentre of DEF is I. Points A′ , B′ , C′ lie on the sides of ABC;
answers. These two inequalities can both hold only if v = . Consequently, each it is known that they are also the points where the incircle of ABC touches the sides.
2
k Thus I A′ , C′′ D and B′′ D are the radical axes of c and two circles touching it and each
student answers exactly questions correctly and each question is answered correctly
2

3 4
other. The radical axes meet at D. Thus ID, IE and IF are perpendicular to BC, CA, AB, these into (??), we obtain
respectively, and consequently also to the corresponding sides of DEF.
From all this, we get that the distance between the orthocentre and the incentre is the r cos ∠CAB + | AC| = r cos ∠CBA + | BC|. (2)
same in triangles ABC and DEF. As these triangles are similar but not equal, this can
happen only if the distance is zero, i.e., the orthocentre and incentre coincide. This As the opposite angle of a bigger side is bigger in every triangle, | AC| < | BC| would
implies that ABC is equilateral. imply ∠CBA < ∠CAB and cos ∠CBA > cos ∠CAB, leading to r cos ∠CAB + | AC| <
F r cos ∠CBA + | BC| which contradicts (2). Analogously, also | AC| > | BC| cannot be.
Consequently, | AC| = | BC|. In the same way, we get | AB| = | AC|, i.e., triangle ABC is
equilateral.
M Remark. The claim of the problem holds also without the assumption that the tangency
of c with the three circles is inner.
B ′′ A′′ B ′′ A′′
B A B A
8. Let b be an even positive integer for which there exists a natural number n such
C′ bn − 1
H H that n > 1 and is a perfect square. Prove that b is divisible by 8. (Seniors.)
A′ B′ b−1
C K C L bn − 1
Solution. As b is even, the perfect square is odd. Hence it is congruent to 8
b−1
D C ′′ E C ′′ modulo 1, i.e., the number

Figure 2 Figure 3
Solution 2. Let r be the radius of c and let r A , r B , rC be the radii of circles with centre A,
B, C, respectively. Let h A be the length of the altitude of ABC drawn from A. Let A′ , B′ ,
C′ be the feet of altitudes of triangle ABC drawn from vertices A, B, C, respectively.
From triangle H A′ C, we get | H A′ |2 = |CH |2 − |CA′ |2 ; from triangle AA′ C, we get
|CA′ |2 = | AC|2 − | AA′ |2 . Here, |CH | = r − rC , | AC| = r A + rC and | AA′ | = h A . Thus
| H A′ |2 = (r − rC )2 − (r A + rC )2 + h2A = r2 − 2(r + r A )rC − r2A + h2A .
Analogously, we obtain
bn − 1  
− 1 = b + b2 + . . . + b n − 1 = b 1 + b + . . . + b n − 2
b−1
is divisible by 8. As the factor 1 + b + . . . + bn−2 is odd, it is relatively prime to 8 and
hence b is divisible by 8.
Remark. One can also prove the claim by considering b modulo 8.
9. The Fibonacci sequence is determined by conditions F0 = 0, F1 = 1, and Fk =
Fk−1 + Fk−2 for all k ≥ 2. Let n be a positive integer and let P( x ) = am x m + . . . + a1 x + a0
be a polynomial that satisfies the following two conditions:

| H A′ |2 = r2 − 2(r + r A )r B − r2A + h2A . (1) P( Fn ) = Fn2 ; (2) P( Fk ) = P( Fk−1 ) + P( Fk−2 ) for all k ≥ 2.

These two equalities together give r B = rC . Analogously, r B = r A . Thus the radii of the Find the sum of the coefficients of P. (Seniors.)
circles drawn around A, B, C touching pairwise each other are equal. This can be only
Answer. Fn .
if the sides of ABC are all equal.
Solution. We are asked to find P(1). If n = 1 then P(1) = P( F1 ) = F12 = 1, giving
Solution 3. Let r A , r B , rC be defined as in Solution 2. Like in Solution 1, note that H lies
P(1) = F1 . If n ≥ 2 then using condition P( Fk ) = P( Fk−1 ) + P( Fk−2 ), 2 ≤ k ≤ n,
inside triangle ABC. As H is the centre of c, radii H A′′ and HB′′ are equal (see Fig. ??)
repeatedly, we get
which means that | H A| + r A = | HB| + r B . Adding rC to both sides of this equality, we
get P( Fn ) = P( Fn−1 ) + P( Fn−2 ) = 2P( Fn−2 ) + P( Fn−3 ) = 3P( Fn−3 ) + 2P( Fn−4 ) = . . .
| H A| + | AC| = | HB| + | BC|. (1) = Fn P( F1 ) + Fn−1 P( F0 ) = Fn P(1) + Fn−1 P(0).

Take triangle KLM whose midlines are the sides of ABC; then triangles KLM and ABC
are similar. Thereby, AH is the perpendicular bisector of side LM as LM k BC and
| LA| = | AM|. Also, BH is the perpendicular bisector of side MK. Thus H is the circum-
1
centre of triangle KLM and ∠ MH A = ∠ MHL = ∠ MKL = ∠CAB.
2
From the right triangle MAH, we get | H A| = r cos ∠ MH A = r cos ∠CAB where r is
the radius of the circumcircle of KLM. Analogously, | HB| = r cos ∠CBA. Substituting
Using the given condition again for k = 2, we obtain
P(1) = P( F2 ) = P( F1 ) + P( F0 ) = P(1) + P(0),
which gives P(0) = 0. Altogether, Fn2 = P( Fn ) = Fn P(1), showing that P(1) = Fn .
Remark. We could ask whether there exist polynomials for every n satisfying the con-
ditions of the problem. Using the condition P( Fk ) = P( Fk−1 ) + P( Fk−2 ), 2 ≤ k ≤ n,

5 6
 T
for finding the other values P( Fk ), 2 ≤ k ≤ n, we get P( Fk ) = Fn Fk , for all k such that
0 ≤ k ≤ n. Elsewhere, the values of the polynomial are not determined. One suitable
polynomial is P( x ) = Fn x.

10. Does there exist a natural number n such that n > 2 and the sum of squares of
some n consecutive integers is a perfect square? (Seniors.)
Answer: yes.
Solution 1. For n = 11, we can construct the following example: D1
A2
(−4)2 + (−3)2 + (−2)2 + (−1)2 + 02 + 12 + 22 + 32 + 42 + 52 + 62 = 112 . C1
B2
n
Solution 2. It is easy to prove by induction that 12 + 22 + . . . + n2 = (n + 1)(2n + 1). P
6
If n = 24 then all three factors in the last product are perfect squares (4, 25 and 49, B1
C2
respectively). Thus the product is also a perfect square.
l1
Remark 1. For n = 3, . . . , 10, no suitable examples exist because there is a number mod- A1
ulo which no sum of squares of n consecutive integers is a quadratic residue. l2
D2
n sum expression bad modulus
3 ( a − 1 )2 + . . . + ( a + 1 )2 3a2 + 2 3 Figure 4
4 ( a − 1 )2 + . . . + ( a + 2 )2 4a2 + 4a + 6 4
5 ( a − 2 )2 + . . . + ( a + 2 )2 5a2 + 10 25 Solution 1. Let r1 and r2 be the radii of c1 and c2 , respectively. Since quadrangle
6 ( a − 2 )2 + . . . + ( a + 3 )2 6a2 + 6a + 19 4 A1 A2 D1 D2 is cyclic, ∠ A2 A1 D1 = ∠ D1 D2 A2 . In addition, ∠ A1 PA2 = ∠ D2 PD1 (see
7 ( a − 3 )2 + . . . + ( a + 3 )2 7a2 + 28 49 Fig. 4). Thus triangles A1 A2 P and D2 D1 P are similar. Their incircles are c1 and c2 , hence
8 ( a − 3 )2 + . . . + ( a + 4 )2 8a2 + 8a + 44 16
| A1 P | r
9 ( a − 4 )2 + . . . + ( a + 4 )2 9a2 + 60 9 = 1.
| D2 P | r2
10 ( a − 4 )2 + . . . + ( a + 5 )2 10a2 + 10a + 85 25
Since quadrangle B1 B2 C1 C2 is cyclic, ∠ B2 B1 C1 = ∠C1 C2 B2 , whence ∠TB1 A1 =
Remark 2. Using diophantine equation theory, it has been proven that the sum of squares ∠TC2 D2 . As also ∠TA1 B1 = ∠TD2 C2 , triangles A1 B1 T and D2 C2 T are similar. Their
of numbers from 1 to n is a perfect square only for n = 0, n = 1 and n = 24. Thus the incircles are c1 and c2 again, hence
choice n = 24 in Solution 2 is the only possibility to succeed.
Remark 3. Sloane’s Encyclopedia of Integer Sequences contains the sequence A001032 with | A1 T | r
= 1.
description “Numbers n such that the sum of squares of n consecutive positive integers | D2 T | r2
can be a perfect square” (i.e., in addition to the conditions of our problem, the numbers | A1 P | | A T|
whose squares are considered must be positive). The sequence starts with numbers 1, Now consider triangles A1 PT and D2 PT. We have = 1 and ∠TA1 P =
| D2 P | | D2 T |
2, 11, 23, 24, 26, 33, 47, 49, 50, 59, 73, 74, 88, 96, 97, 107, 121, 122, 146, 169, 177, 184, 191, ∠TD2 P. Thus these triangles are similar and
193, 194, 218, 239, 241, 242, 249, 289, 297, 299, 311, 312, 313, 337, 338, 347, 352, 361, 362,
376, 383, 393, 407, 409, 431, 443, 457, 458, 479, 481, 491. r1 | A P| | PT |
= 1 = = 1.
r2 | D2 P | | PT |
11. Tangents l1 and l2 common to circles c1 and c2 intersect at point P, whereby tangent
points remain to different sides from P on both tangent lines. Through some point Solution 2. Let r1 and r2 be the radii of c1 and c2 , respectively. Let s be the line that passes
T, tangents p1 and p2 to circle c1 and tangents p3 and p4 to circle c2 are drawn. The through P and is perpendicular to the line joining the centres of c1 and c2 . Consider the
intersection points of l1 with lines p1 , p2 , p3 , p4 are A1 , B1 , C1 , D1 , respectively, whereby composition of two plane transformations: reflection w.r.t. s and homothety w.r.t. P
the order of points on l1 is: A1 , B1 , P, C1 , D1 . Analogously, the intersection points of l2 r2
with factor . This composite transformation takes c1 to c2 and P to P.
with lines p1 , p2 , p3 , p4 are A2 , B2 , C2 , D2 , respectively. Prove that if both quadrangles r1
A1 A2 D1 D2 and B1 B2 C1 C2 are cyclic then radii of c1 and c2 are equal. (Seniors.) Denote the image of any point X under this transformation by X ′ . As the transformation
respects all angles between lines, the equality of angles in cyclic quadrilateral B1 B2 C1 C2

7 8
 √ √
implies ∠ B2′ B1′ P = ∠ B2 B1 P = ∠C1 C2 P. As lines B1′ B2′ and C2 C1 both touch circle c2 and 2+ 2 1 2+ 2 1
inradius and the perimeter gives r. Thus = r, whence r = √ =
intersect PC2 under the same angle, line B1′ B2′ coincides with line C2 C1 . Analogously, 2 2 2 2 + 2
√ √
A1′ A2′ coincides with line D2 D1 . 2− 2 2 2 1 √ 3 √ 2
Hence the transformation takes the intersection point of lines A1 A2 and B1 B2 to the = 1− . Therefore r = 1 + − 2 = − 2, so r is not rational.
2 2 2 2
intersection point of lines C2 C1 and D2 D1 , i.e., T is taken to T. But if r1 6= r2 then the abc a2 b2 c 2
transformation obviously can have only one fixpoint. Consequently, r1 = r2 . Solution 2. (a) By the formula S = , we get R2 = . Thus it suffices to prove
4R 16S2
2
that S is rational. By Heron’s formula,
12. Find all positive integers n such that one can write an integer 1 to n2 into each unit
square of a n2 × n2 table in such a way that, in each row, each column and each n × n 1  2 2 
S2 = p( p − a)( p − b)( p − c) = 2a (b + c2 ) − (b2 − c2 )2 − (a2 )2 .
block of unit squares, each number 1 to n2 occurs exactly once. (Seniors.) 16
The last expression clearly evaluates to a rational number.
Answer: 1.
(b) In part (a), it was proven that S2 is rational. For the triangle with side lengths 1, 1
Solution. In the case n = 1, the conditions can be fulfilled trivially. Assume n ≥ 2. √ !2 √
√ 2+ 2 6+4 2 3 √
Denote the unit square in the ith row and jth column by 1 2 ... n n +1 and 2, we get p2 = = = + 2, hence p2 is not rational and
(i, j). Let A be the n × n block containing both (1, 1) and 1
2 4 2
(n, n). Let B and C be the n × n blocks obtained from A by S2
2 neither is r2 = .
shifting it by one unit down and right, respectively (see .. p2
Fig. 5). .
The numbers in the bottommost row of B must be the n
same as the numbers in the topmost row of A in some or- Selected Problems from the Final Round of National
n +1
der since both blocks must contain each number 1, . . . , n2 Olympiad
exactly once. Analogously, the rightmost column of C
must contain the same numbers as the leftmost column Figure 5
of A in some order. Now, the number in (n + 1, n + 1) 1. Two medians drawn from vertices A and B of triangle ABC are perpendicular.
cannot occur in the topmost row of A since all these occur already in the row number Prove that side AB is the shortest side of ABC. (Grade 9.)
n + 1 left from the square under consideration. Analogously, this number cannot occur
in the leftmost column of A. As A contains all numbers 1 to n2 , this number must occur Solution. Let the medians intersect in point M and let the median drawn from vertex
elsewhere in A. But then it occurs twice in the n × n block that contains squares (2, 2) C intersect AB in point F (see Fig. 6). Then, F is the midpoint of the hypotenuse of
and (n + 1, n + 1) which is prohibited. the right triangle ABM and thus the midpoint of the circumcircle of ABM, so we get
| AB| = 2| FM|. Since M divides median CF in ratio 2 : 1, we have | AB| = |CM|. The
13. Consider triangles whose each side length squared is a rational number. Is it true largest angle of triangle AMC is the obtuse angle AMC, therefore AC is the longest side
that of this triangle. We get | AC| > | MC| = | AB|. The proof of | BC| > | AB| is analogous.
(a) the square of the circumradius of every such triangle is rational; Remark. One can also solve the problem using the Pythagorean theorem and the fact
(b) the square of the inradius of every such triangle is rational? that the centroid divides the medians in ratio 2 : 1.

(Seniors.) 2. Juhan wants to order by weight five balls of pairwise different weight, using only
Answer: (a) yes; (b) no. a balance scale. First, he labels the balls with numbers 1 to 5 and creates a list of weigh-
ings, such that each element in the list is a pair of two balls. Then, for every pair in
Solution 1. (a) Fix a triangle from the family under consideration. Let a, b, c be its side
the list, he weighs the two balls against each other. Can Juhan sort the balls by weight,
lengths, γ the size of the angle opposite to the last side and R the circumradius. The
using a list with less than 10 pairs? (Grade 9.)
a2 + b2 − c 2 ( a2 + b 2 − c 2 )2
cosine law gives cos γ = , whence cos2 γ = . As a2 , b2 , c2 are Answer: no.
2ab 4a2 b2
rational, also cos2 γ is rational. Therefore 1 − cos2 γ, i.e., sin2 γ, is rational. The sine law Solution. There are 10 possible pairs of 5 balls. Suppose w.l.o.g. that Juhan does not
c c2 weigh the pair (1, 2). Then, it is not possible to distinguish orderings 1, 2, 3, 4, 5 and
gives R = , hence R2 = . Thus R2 is rational.
2 sin γ 4 sin2 γ 2, 1, 3, 4, 5, since the remaining 9 weighings give the same result. Thus, Juhan’s list must
√
(b) Consider the right isosceles triangle with side lengths 1, 1, 2. Let r be its inradius. contain all 10 pairs.
1
The area of this triangle, computed via leg lengths, is ; computing the area via the 3. Two radii OA and OB of a circle c with midpoint O are perpendicular. Another
2

9 10
circle touches c in point Q and the radii in points C and D, respectively. Determine
∠ AQC. (Grade 10.)
Answer: 45◦ .
C 3 vertices 4 vertices 5 vertices 6 vertices 7 vertices

b a
2 2 B
8 vertices 9 vertices 10 vertices 12 vertices

D Q Figure 8
E
y x

M D union is an n-gon (not necessarily convex). Find all possible values of the number of
b a vertices n. (Grade 10.)
O′
2 2x 2y 2
Answer: all integers from 3 to 12, except 11.
A′ A
Solution. The n-gon must have at least 3 vertices. We show first that the number of
A B
O C vertices is at most 12. Indeed, each vertex of the n-gon is either a vertex of one of the
F
triangles or an intersection point of some two sides of the two triangles. There are 6
Figure 6 Figure 7 triangle vertices and 6 possible intersection points, since every side of the first triangle
can intersect at most two sides of the second triangle. Thus, n ≤ 12.
Solution. By symmetry, ∠ AQC = ∠ BQD (see Fig. 7). Since AQB is an internal angle
of a regular octagon, we have ∠ AQB = 135◦ . Let now O′ be the midpoint of the circle Next, suppose that the n-gon has 11 vertices. If 6 of those vertices are vertices of the two
through C, D and Q. The quadrilateral OCO′ D has three right angles: ∠COD = 90◦ by triangles, then neither triangle can contain a vertex of the other triangle. Thus, each side
assumption, while angles OCO′ and ODO′ are angles between a radius and a tangent. of each triangle intersects the other triangle either never or twice, so we cannot have an
1 odd number of intersection points. On the contrary, if the n-gon has 6 intersection points
Thus, CO′ D is also a right angle, so ∠CQD = ∠CO′ D = 45◦ and as vertices, every side of each triangle must intersect the second triangle twice, and thus
2
all vertices of one triangle must be outside the other triangle. Thus, all 6 triangle vertices
1 1 are also vertices of the n-gon, and n = 12.
∠ AQC = (∠ AQB − ∠CQD ) = (135◦ − 45◦ ) = 45◦ .
2 2 The remaining configurations from 3 to 12 are all possible (see Fig. 8).
Remark. One can also make use of the similarity of triangles QO′ D and QOA′ (where A′
6. The identifier of a book is an n-tuple of numbers 0, 1, . . . , 9, followed by a check-
is the other endpoint of the diameter of c drawn through A), yielding that angle AQA′
sum. The checksum is computed by a fixed rule that satisfies the following property:
subtends the diameter. Another approach is to apply the tangent chord property on
whenever one increases a single number in the n-tuple (without modifying the other
tangent AC and secant CD to obtain that triangles AQC and CQD are similar.
numbers), the checksum also increases. Find the smallest possible number of required
4. Prove that the sum of the squares of any three pairwise different positive odd in- checksums if all possible n-tuples are in use. (Grade 10.)
tegers can be represented as the sum of the squares of six (not necessarily different) Answer: 9n + 1.
positive integers. (Grade 10.) Solution. Consider the checksum of (0, 0, . . . , 0). Increasing the first number, we get n-
Solution. Let a > b > c be positive integers. Then tuples (1, 0, . . . , 0), . . . , (9, 0, . . . , 0) with 9 new checksums. Increasing the second num-
ber in the last n-tuple, we get (9, 1, . . . , 0), . . . , (9, 9, . . . , 0), and again obtain 9 new val-
a2 b2 b2 c 2 c 2 a2 ues. Continuing like this, we see that the number of different check values is at least
a2 + b2 + c 2 = + + + + + = 9n + 1.
2 2 2 2 2 2

a+b
2 
a−b
2 
b+c
2  
b−c 2
 
a+c 2
 
a−c 2 It is easy to see that the sum of all numbers in the n-tuple is a valid checksum. On the
= + + + + + . other hand, the sum of n numbers 0, 1, . . . , 9 is at least 0 and at most 9n, so with this
2 2 2 2 2 2
rule, we have exactly 9n + 1 different checksums.
Since a, b and c are all odd, the latter is a sum of squares of six positive integers.
7. Find all real numbers a such that all solutions to the quadratic equation x2 − ax +
5. Two triangles are drawn on a plane in such a way that the area covered by their a = 0 are integers. (Grade 11.)

11 12
Answer: a = 0 and a = 4. ∠CAR = ∠CBQ = ∠PAB, so triangles ACR and ABP are similar. If AQ ⊥ BC, then AR
Solution. Let x and y be the solutions of the quadratic equation. Viète formulae give is the height of ABC and ∠ ARC = ∠ APB = 90◦ , so P = R and thus P lies on AQ. If P
x + y = xy = a. The case y = 1 gives a contradiction 1 + x = x, while y 6= 1 gives lies on AQ, then again P = R and ∠CRA = ∠CPA = 180◦ − ∠ BPA = 180◦ − ∠CRA, or
x=
y
. Thus, x is an integer if and only if y = 2 or y = 0. Now, y = 2 gives x = 2 ∠CRA = 90◦ , so AQ ⊥ BC.
y−1 A
Solution 2. First consider the case where AB
and a = x + y = 4; y = 0 gives x = 0 and a = x + y = 0.
is the diameter of the circle through A, B and
Remark. Oneqcan also proceed from the fact that the sum a as well as the difference
p P. Then ∠ APB = 90◦ and ∠QBA = 90◦ giv-
a2 − 4a = (a − 2)2 − 4 of the solutions is integral. ing that AQ is the diameter of circumcircle
of ABC. Thus AQ ⊥ BC and P lies on line
8. A 3-dimensional chess board consists of 4 × 4 × 4 unit cubes. A rook can step from AQ.
any unit cube K to any other unit cube that has a common face with K. A bishop can On the other hand, when the centre of
step from any unit cube K to any other unit cube that has a common edge with K, but the circle passing through points A and B
does not have a common face. One move of both a rook and a bishop consists of an moves away from the leg AB on the half-
arbitrary positive number of consecutive steps in the same direction. Find the average plane containing point C, points Q and P R
number of possible moves for either piece, where the average is taken over all possible will move toward points B and C, respec- B P C
starting cubes K. (Grade 11.) tively. When the centre of the circle moves
away from AB on the other halfplane, points
Answer: the rook has on average 9 moves, the bishop has 10.5. Q
Q and P will move toward points C and B,
Solution. The rook has always 3 pos- respectively. Therefore in either case point P Figure 10
R R sible moves in the direction of one neither lies on line AQ nor AQ is perpendic-
axis, regardless of the choice of the ular to BC.
starting cube K (see Fig. 9), and thus
9 possible moves on average. 10. Find all pairs (m, n) of positive integers such that mn − nm = 3. (Grade 11.)
In any of the 8 “middle” cubes of the Answer: (4, 1).
B chess board, the bishop has 5 possi-
Solution. First, m and n must have different parity, for otherwise the lhs is even.
B B ble moves in the direction of every
axis, and thus 15 moves in total. In • If m is odd and n even, then mn ≡ 1 (mod 4), so nm ≡ 2 (mod 4), which is possible
Figure 9 any of the 24 cubes that have a com- only for m = 1. But then mn = 1 and the lhs is smaller than 3. Thus, in this case
mon face with a middle cube, the there are no solutions.
bishop has 5 moves on the plane parallel to the common face, and 3 moves on each • If now m is even and n odd, then nm ≡ 1 (mod 8), so mn ≡ 4 (mod 8). Thus,
of the remaining two planes, or 11 moves in total. For the remaining 32 cubes on the n ≤ 2, and n odd gives n = 1, m = 4 as the only solution.
edge of the board, the bishop has 3 moves in the direction of every axis, and 9 moves in
total. Averaging over all cubes, we get the result.
11. Some circles of radius 2 are drawn on the plane. Prove that the numerical value of
Remark. The number of possible moves shows to some extent the strength of every piece. the total area covered by these circles is at least as big as the total length of arcs bounding
The solution to this problem implies that in 3-dimensional chess, a bishop is perhaps the area. (Grade 11.)
stronger than a rook (as opposed to regular chess). On the other hand, the bishop is
weakened by the fact that it can always reach only half of the cubes or squares. Solution 1. The boundary line of the area consists of circular arcs, each corresponding to
While a 3-dimensional 4 × 4 × 4 board and a regular 8 × 8 board have the same number rl
a circular sector. According to the formula S = , where r is the radius and l the length
of cells, a regular rook and bishop have 14 and 8,75 moves on average, respectively. 2
of arc, we obtain that in the case r = 2 the area of circular sector equals numerically the
length of the circular arc on its boundary.
9. A circle passing through the endpoints of the leg AB of an isosceles triangle ABC
intersects the base BC in point P. A line tangent to the circle in point B intersects the We prove that no two such sectors have common interior points. First, the sectors from
circumcircle of ABC in point Q. Prove that P lies on line AQ if and only if AQ and BC the same circle share only the centre of the circle. Now let one sector be part of circle w1
are perpendicular. (Grade 11.) with centre O1 and the other be part of circle w2 with centre O2 . Assuming that the sec-
tors have a common interior point C, we draw the radii O1 A and O2 B of circles w1 and
Solution 1. Let line AQ intersect base BC in point R (see Fig. 10). On one hand, ABC w2 through C, respectively. Obviously A and B lie on the boundary of the area covered
is isosceles, so ∠ ACR = ∠ ABP. On the other hand, the tangent chord property gives by circles, meaning that B and A do not lie in the interior of w1 and w2 , respectively.

13 14
  n
Therefore |O1 B| ≥ |O1 A| and |O2 A| ≥ |O2 B|. We draw the mediator of segment AB x x z
Since < 1, + 1 < 2 holds for all n. On the other hand, since > 1, there exists
(see Fig. 11). The inequalities imply that O1 and A lie on the one side of the mediator y y y
 N
and O2 and B lie on the other side. Hence segments O1 A and O2 B cannot have common z
points, a contradiction. A
an integer N such that > 2, contradiction.
y
We have obtained that the total length of the bound- B

ary line of the area equals the sum of the lengths of 14. Does there exist an equilateral triangle
circular arcs. But the total area is greater or equal (a) on a plane; (b) in a 3-dimensional space;
than the sum of the areas of the sectors.
Solution 2. We shall use induction on the number of such that all its three vertices have integral coordinates? (Grade 12.)
O1
circles. For one circle, the circumference and the area Answer: (a) no; (b) yes.
O2
are equal (4π). We prove that whenever we add a −→ −→
circle, the area a covered by the intersection of the Solution 1. (a) Suppose that such a triangle ABC exists. Then vectors AB and AC have
−→
new circle with the old area is at most as big as the Figure 11 integral coordinates.
√ Denote AB = ( x, y), then the vector of the height drawn from
perimeter p of that intersection. 3 −→
vertex C is (y, − x ). The coordinates of AC are then either
First, notice that a ≤ 4π. Assume now that also p ≤ 4π. We write p = 2πr, so r ≤ 2
2. Since a circle maximizes the area for a fixed perimeter, we see that the area of the √ √ √ !
−→ 1 3 x + 3y y − 3x
intersection is a ≤ πr2 ≤ 2πr = p as desired. AC = ( x, y) + (y, − x ) = ,
2 2 2 2
12. Consider a cylinder and a cone with a common base such that the volume of the or
part of the cylinder enclosed in the cone equals the volume of the part of the cylinder √ √ √ !
outside the cone. Find the ratio of the height of the cone to the height of the cylinder. −→ 1 3 x− 3y y + 3x
AC = ( x, y) − (y, − x ) = , .
(Grade 12.) 2 2 2 2
1
Answer: 1 + √ . In either case, the coordinates are integral only for x = y = 0.
3 (b) Triangle ABC with A = (1, 0, 0), B = (0, 1, 0) and C = (0, 0, 1) is equilateral.
Solution. Denote by v and V, h and H the volume and the height of the cylinder and the
Solution 2. (a) W.l.o.g. assume that one vertex of the triangle is A(0, 0). We also assume
cone, respectively, and denote by S the area of the common base. The vertex of the cone
w.l.o.g. that the ordinates of B and C are non-negative and these vertices do not lie on
must lie outside the cylinder, for otherwise the volume of the intersection would be at
1 the y-axis.
most of the total volume of the cylinder. Since all the coordinates are integral, the slopes of AB and AC are rational. Denote the
3 h
Denote = x. The part of the cone that lies outside the cylinder is a cone similar to slope angles of lines AB and AC by β and γ, respectively, then γ = β ± 60◦ . Denote
H H−h tan β = k. We have
the original cone with scale factor = 1 − x and volume (1 − x )3 V. The volume √ √ √
H
v tan β ± tan 60◦ k± 3 (k ± 3)(1 ± k 3)
of the part of the cylinder inside the cone is thus V − (1 − x )3 V = . From v = Sh and tan γ = = √ = .
√ 2 1 ∓ tan β tan 60◦ 1∓k 3 1 − 3k2
1 3 3± 3 √ √ √ √ √
V = SH we get ( x3 − 3x2 + 3x ) = x, so x = . Since h < H implies x < 1, the But (k ± 3)(1 ± k 3) = k ± k2 3 ± 3 + 3k = 4k ± (k2 + 1) 3. As k2 + 1 > 0, it is
3 √ 2 2
3− 3 impossible that tan γ and k would be simultaneously rational.
only possible solution is x = and we get the desired ratio.
2 Remark. There are several other solutions to this problem. One may w.l.o.g. denote
the vertices A(0, 0), B( x1 , y1 ) and C( x1 , y1 ), where x1 , y1 , x2 and y2 are relatively prime,
13. Let x, y, z be positive real numbers such that x n , yn and zn are side lengths of some derive the equalities x12 + y21 = x22 + y22 = 2( x1 x2 + y1 y2 ) and consider all cases of re-
triangle for all positive integers n. Prove that at least two of x, y and z are equal. (Grade mainders of x1 , y1 , x2 and y2 modulo 2.
12.) Another approach would be to notice that ( x1 + x2 )2 + (y1 + y2 )2 is divisible by 3 and
Solution. Assume that x, y, z are all different and assume w.l.o.g. x < y < z. For any n, use the fact that perfect squares have only remainders 0 and 1 modulo 3.
the triangle inequality implies x n + yn > zn , or The most straightforward way is to compute the area S of the  triangle with vertices
 x1 x2 x3 
 n  n  
x z A( x1 , y1 ), B( x2 , y2 ) and C( x3 , y3 ) in two different ways: 2S =  y1 y2 y3  and 2S =
+1 > .  1 1 1 
y y

15 16
 √
a2 3
(where a is the side of the triangle). Since all the coordinates as well as the side
2
length are integral, such a triangle cannot exist.

15. Let a, b, c be positive integers such that gcd(a, b, c) = 1 and the product of every
two of these integers is divisible by the third one.
(a) Prove that every one of these integers equals the least common multiple of the re-
maining two integers divided by the greatest common divisor of these two integers.
n = 4q n = 4q + 1 n = 4q + 2 n = 4q + 3
(b) Give an example of such integers a > 1, b > 1 and c > 1.
Figure 12
(Grade 12.)
Answer: (b) For example, a = 6, b = 10, c = 15. r 4
Solution 1. (a) Let d = gcd(a, b), a = a′ d and b = b′ d, where gcd(a′ , b′ ) = 1. Then
4
lcm(a, b) ..
lcm(a, b) = a′ b′ d and = a′ b′ . We prove that a′ b′ = c. .
gcd(a, b) ..
.
On the one hand, since ab = a′ b′ d2 is divisible by c and gcd(d, c) = 1 because of 4
gcd(a, b, c) = 1, a′ b′ must be divisible by c. On the other hand, the conditions of the 4
problem imply that ca = ca′ d is divisible by b = b′ d, i.e. ca′ is divisible by b′ . As 4
gcd(a′ , b′ ) = 1, c is divisible by b′ . Analogously, c is divisible by a′ . Since a′ and b′ are 4 r
coprime, c is divisibly by a′ b′ . Altogether we have c = a′ b′ as desired. The claims about
4 4 ··· 4 r r 4 4 ··· 4
a and b are proven analogously.
(b) Let x, y and z be different numbers that are pairwise coprime, e.g. different primes. Figure 13
Having a = xy, b = yz and c = xz, the numbers a, b and c satisfy the conditions of the
problem. Indeed,
• If r = 1, we mark groups of 8 squares. Since there are q2 such groups, we again get
lcm(a, b) xyz A(4q + 1) ≤ 8q2 .
= = xz = c,
gcd(a, b) y • If r = 2, we mark as in the first case, but leave the last two rows empty. This way,
we mark 2q rows with n squares, or 2q(4q + 2) squares in total. Thus, A(4q + 2) ≤
analogously for other cases. For an example we may take x = 2, y = 3, z = 5 that yields 8q2 + 4q.
6, 15 and 10.
• If r = 3 and q > 0, we mark every fourth row and column and the middle squares
Solution 2. (a) Let p be some prime factor of at least one of a, b and c. Since gcd(a, b, c) =
of the remaining 3 × 3 groups. There are (q + 1)2 marked middle squares and q
1, we may assume w.l.o.g. that p does not divide c. At the same time, since ca is divisible
marked rows and columns, giving (q + 1)2 + 2(4q + 3)q − q2 marked squares in
by b and cb is divisible by a, the factor p must have the same exponent α in both a and
total. Thus, A(4q + 3) ≤ 8q2 + 8q + 1. For q = 0, we get n = 3 and A(3) = 0.
b. Similarly, the exponents of any prime factor p′ in the prime factorization of a, b and c
are α′ , α′ and 0 in some order. Since every exponent in such a triple equals the difference We now show that these bounds are tight.
of the maximum and the minimum of the remaining two, the result follows.
• For r = 0 or r = 1, we can divide a 4q × 4q grid into q2 squares of size 4 × 4. In
16. Some squares of an n × n grid are marked in such a way that in every 4 × 4 square, every such square, at least 8 squares must be marked, so A(n) ≥ 8q2 .
at least half of the squares are marked. Find the smallest possible number of marked • For r = 2 or r = 3, cut an r × r square from the lower left corner and divide the
squares in the grid. (Grade 12.) remaining squares into q L-shaped strips of width 4 (see Fig. 13, right). The ith strip
then contains 2i 4 × 4 squares. The last two squares intersect in the corner of the L
Answer: 8q2 for n = 4q or n = 4q + 1; 8q2 + 4q for n = 4q + 2; 8q2 + 8q + 1 for n = 4q + 3
and the intersection is a square of size (4 − r ) × (4 − r ). Every strip also contains
and q ≥ 1; 0 for n = 3.
an r × r square in the corner.
Solution. Let A(n) be the smallest possible number of marked squares. Write n = 4q + r,
where 0 ≤ r < 4. First, we bound A(n) from above. In order to bound A(n) from below, we add the smallest possible number of
marked squares in the 4 × 4 squares and the r × r squares and subtract the largest
• If r = 0, we mark half of the squares in groups of two rows as shown in Fig. 12, so possible number of marked squares in the (4 − r ) × (4 − r ) squares.
A(4q) ≤ 8q2 .

17 18
 The case r = 2 gives Solution 2. Let ai be the number of times the ith switch changes its position during the
whole process. According to the conditions of the problem, each switch moves either
(2q + 2)q when it is moved directly by the operator or its left neighbour
A(n) ≥ 8 · (2 + 4 + . . . + 2q) + 0 · (q + 1) − 4 · q = 8 · − 4q = 8q2 + 4q. j amoves
k down. As all
2 switches are down at the beginning, the ith switch moves down i times. Thus a1 = 1
ja k 2
If r = 3 every 3 × 3 square contains at least one marked square, so
and ai = i + i −1 for all i ≥ 2.
2  
(2q + 2)q ja k 1
A(n) ≥ 8 · (2 + 4 + . . . + 2q) + 1 · (q + 1) − 1 · q = 8 · + 1 = 8q2 + 8q + 1. We prove by induction that ai = 2(i − 1) for all i ≥ 2. As a2 = 2 +
1
= 2+ = 2,
2 2 2
this claim holds for i = 2. Assuming that it holds for i, we obtain
 
IMO team selection contest ja k
ai + 1 = i + 1 + i = i + 1 +
2 (i − 1 )
= 2i,
2 2
First day i.e., the claim holds also for i + 1.
Altogehter, this shows that a1 is odd and ai is even for all i ≥ 2. Hence, after the process,
1. On the control board of a nuclear station, there are n electric switches (n > 0), all the first switch is up and all the others are down.
in one row. Each switch has two possible positions: up and down. The switches are Solution 3. Interpret the position of switches on the board as binary numbers so that the
connected to each other in such a way that, whenever a switch moves down from its ith switch from the left corresponds to the ith lowest binary digit: being down encodes
upper position, its right neighbour (if it exists) automatically changes position. At the 0 and being up encodes 1. Changing the ith switch then works like addition of 2i −1
beginning, all switches are down. The operator of the board first changes the position modulo 2n . The initial position encodes number 0 and the final position encodes 1 · 20 +
of the leftmost switch once, then the position of the second leftmost switch twice etc., 2 · 21 + . . . + n · 2n−1 modulo 2n .
until eventually he changes the position of the rightmost switch n times. How many
We prove by induction that 1 · 20 + 2 · 21 + . . . + n · 2n−1 ≡ 1 (mod 2n ). If n = 1 then
switches are up after all these operations?
this holds. Assume that the claim holds for n = k. Multiplying this congruence by 2
Answer: 1. gives
Solution 1. Enumerate the switches with numbers 1 to n from left to right. We prove first
that the result of two consecutive changes does not depend on the order of the changes. 1 · 21 + 2 · 22 + . . . + k · 2 k ≡ 2 (mod 2k+1 ).
Let x and y be the numbers of the switches changed, x < y.
Adding 20 + 21 + . . . + 2k to both sides gives
• If there exists a number z such that x ≤ z < y and switch number z is down then
changing the position of x can influence only switches from x to z, changing the 1 · 20 + 2 · 21 + 3 · 22 + . . . + ( k + 1 ) · 2 k ≡ 2 + 2 k + 1 − 1 ≡ 1 (mod 2k+1 ),
position of switch y can influence only this switch and switches right from y. Thus
i.e., the claim holds for n = k + 1.
the results of the changes are independent of each other.
Remark. In Solution 3, one could prove by induction a stronger claim: 1 · 20 + 2 · 21 +
• If no such z exists then changing switch number x causes a change of switch number
. . . + n · 2n−1 = (n − 1) · 2n + 1.
y. After that, switches x to y − 1 are all down while all switches in the right from
them are in the same position as if switch number y were changed. Thus after 2. Let D be the foot of the altitude of triangle ABC drawn from vertex A. Let E and
moving both x and y in either order, switches from x to y − 1 are down and the F be points symmetric to D w.r.t. lines AB and AC, respectively. Let R1 and R2 be the
switches with larger number are in the position as when switch y were moved circumradii of triangles BDE and CDF, respectively, and let r1 and r2 be the inradii of
twice. the same triangles. Prove that
We prove now that, after all operations, precisely the leftmost switch is up. This claim |S ABD − S ACD | ≥ | R1 r1 − R2 r2 |
holds trivially for n = 1. Assume the claim holding for n switches and consider a board
with n + 1 switches. According to what was proven above, the moves can be performed where SK denotes the area of figure K.
in arbitrary order. Therefore, first change switch number 2 once, then switch number 3
Solution 1. Consider first the case where D lies between points B and C (see Fig. 14). As
twice etc., until the last switch n times. By the induction hypothesis, switch number 2 1 1
is up and all the others are down. Each switch has to be moved once more; if we do it S ABD = · | AD | · | BD | and S ACD = · | AD | · |CD |, we have
2 2
from right to left then switches n + 1 to 3 go up, then moving switch 2 down brings all
them down and finally switch 1 is moved up. Thus 1 is the only switch remaining up. 1
S ABD − S ACD = · | AD | · (| BD | − |CD |).
2

19 20
Let G be the incentre of triangle BDE and let G ′ be the projection of G to line BD. Then Analogously we obtain
| GG′ | = r1 . By symmetry, ∠BEA = ∠BDA = 90◦ , hence quadrangle BEAD is cyclic
and line segment AB is its circumdiameter. Thus | AB| = 2R1 . As triangles ADB and 1 1 γ
S ACD = | AD |2 tan γ, R2 r 2 = | AD |2 tan γ tan .
| AB| | GB| 2 2 2
GG ′ B are similar, we have = , implying 2R1 r1 = | AD | · | GB|. Let H be the
| AD | | GG′ | From these equalities, we can conclude that S ABD − S ACD and R1 r1 − R2 r2 have the
incentre of triangle CDF; then analogously 2R2 r2 = | AD | · | HC|. Hence same sign since β and γ belong to the first quarter where tan is increasing. W.l.o.g.,
1 assume that both are non-negative (otherwise interchange B and C). Then β ≥ γ and
R1 r 1 − R2 r 2 = · | AD | · (| GB| − | HC|). the desired inequality is equivalent to S ABD − R1 r1 ≥ S ACD − R2 r2 . Now
2
1 1 1  β
Triangle ADG is isosceles because ∠ ADG = 90◦ − ∠ BDE = 90◦ − ∠ DAG. Thus S ABD − R1 r1 = | AD |2 tan β 1 − tan =
2 2 2 2
| AD | = | AG|. Analogously, | AD | = | AH |. Thus | AG| = | AH |. β β
2 tan tan
Subtracting equality | AD |2 + |CD |2 = | AC|2 from | AD |2 + |
BD |2 = | AB|2 gives 2 1  β

| BD | − = | AD |2 2 1 − tan = | AD |2 2 ,
|CD |2 = | AB|2 − | AC|2 which is equivalent to | BD | − |CD | · | BD | + |CD | = | AB| − 2 β 2 β


1 − tan2 1 + tan
| AC| · | AB| + | AC| . Consequently, 2 2
   
whence
| BD | − |CD | · | BC| = | GB| − | HC| · | AB| + | AC| .
 
   
As | BC| < | AB| + | AC|, we must have | BD | − |CD | ≥ | GB| − | HC|, which gives the 1
S ABD − R1 r1 = | AD |2 1 −
 
desired inequality. β

E 1 + tan
E 2
A A
and, analogously,
F
F  
K 2 1
S ACD − R2 r2 = | AD | 1 − γ
.
G L
H G
1 + tan
2
H H′
By β ≥ γ and tan being increasing, the inequality S ABD − R1 r1 ≥ S ACD − R2 r2 follows.
B G′ D C B C D C′
bn − 1
Figure 14 Figure 15 3. Let n be a natural number, n ≥ 2. Prove that if is a prime power for some
b−1
If D does not lie between B and C (see Fig. 15) then assume w.l.o.g. that it is on ray BC. positive integer b then n is prime.
Reflect line segment AC w.r.t. line AD; points C and H transform to some points C′ and bn − 1
H ′ , respectively. Now apply the solution above for triangle ABC′ . The desired claim Solution. Clearly b ≥ 2. Assume that = pl where p is prime, then n ≥ 2 implies
b−1
follows then by using |C′ D | = |CD | and | H ′ C′ | = | HC|. l ≥ 1. If n = xy where both x and y are greater than 1 then consider the representation
Solution 2. Denote ∠ BAD = β and ∠CAD = γ. Then
bxy − 1 bxy − 1 by − 1 by − 1
1 1 = y · = (1 + b y + . . . + b y ( x − 1) ) · .
S ABD = · | AD | · | BD | = | AD |2 tan β. b−1 b −1 b−1 b−1
2 2
As the product is a power of p, both factors must be powers of p. As x > 1 and y > 1,
As in Solution 1, show that quadrangle BEAD is cyclic. Let K be the point of intersec- both factors are multiples of p. Then by − 1 is a multiple of p. Thus all addends in the
| AB| | AD | first factor are congruent to 1 modulo p which implies that the first factor is congruent
tion of its diagonals. As R1 = , we get R1 = . Furthermore, r1 = | GK | and
2 2 cos β to x modulo p. Hence x is divisible by p. As x was an arbitrary non-trivial factor of n,
∠BDE ∠BAD β β β this shows that n = pm for a positive integer m.
∠GDK = = = . Thus r1 = | DK | tan = | AD | sin β tan . Conse-
2 2 2 2 2 Now consider the representation
quently,
m m 2
| AD | β 1 β bp − 1 bp − 1 bp − 1 bp − 1
R1 r 1 = · | AD | sin β tan = | AD |2 tan β tan . = p m −1 ·...· p · .
2 cos β 2 2 2 b−1 b −1 b −1 b−1

21 22
 bp − 1 Taking y = a in the initial relation, we get f ( x ) = a + f ( x + 1), and so g( x − 1) − g( x ) =
Each factor is both greater than 1 and a power of p. As is a positive integral
b−1 a − 1 for all x. Since a 6= 1, g is unbounded and by continuity, takes all real values, so
power of p, the numerator is divisible by p, i.e., b p ≡ 1 (mod p). By Fermat’s little f (z) = 1 + z for all z.
theorem, b p ≡ b (mod p). Thus b ≡ 1 (mod p) and b − 1 is divisible by p. But then Let now a = 1, i.e., f (1) = 0. Then x = 0 yields f ( f (y)) = y for all reals y. Taking now
the numerator b p − 1 must be divisible by p2 , i.e., b p ≡ 1 (mod p2 ). If m ≥ 2 then the
2 y = f (1 − x ) in the initial relation, we get f ( x + f ( f (1 − x ))) = f (1 − x ) + f ( x + 1),
bp − 1 or 0 = f (1 − x ) + f ( x + 1). Finally, taking y = 1 − x yields f ( x + f (1 − x )) = 1 −
representation above contains factor p = 1 + b p + . . . + b p( p−1) . On one hand,
b −1 x + f ( x + 1), so f ( x + f (1 − x )) = 1 − x − f (1 − x ). Let h( x ) = x + f (1 − x ), then
this is congruent to p modulo p2 as all addends are congruent to 1. On the other hand, f (h( x )) = 1 − h( x ) holds for all x. Replacing x with − x and taking y = 1 in the initial
this factor is a power of p while being greater than p, hence it is a multiple of p2 . This relation, we get f (− x ) = 1 + f (1 − x ), so h( x + 1) − h( x ) = 2. Again, h is continuous
contradiction shows that m = 1, qed. and must take all real values, so f (z) = 1 − z for all z.
Remark 1. Fermat’s little theorem can easily be avoided in the solution. Cutting this It is straightforward to verify that both solutions indeed satisfy the initial relation.
out from the solution above, it still shows that if m ≥ 2 then b p − 1 is not divisible by
bp − 1 6. Consider a 10 × 10 grid. On every move, we colour 4 unit squares that lie in the
p2 . Continuing from this, we see that is divisible by p but not by p2 . Hence this
b−1 intersection of some two rows and two columns. A move is allowed if at least one of the
factor must be p. Now 4 squares is previously uncoloured. What is the largest possible number of moves that
bp − 1 can be taken to colour the whole grid?
= 1 + b + . . . + b p−1 > b p−1 ≥ 2 p−1 ≥ p
b−1 Answer: 81.
gives a contradiction. Solution. By always choosing the first line, the first column and a square of the remaining
Remark 2. In the special case b = 2, l = 1, the problem reduces to the well-known fact 9 × 9 grid as the lower right square, the whole grid can be coloured in 81 moves.
that a Mersenne’s number Mn can be prime only if n is prime. We now prove that it is not possible to make more than 81 moves. Consider a sequence
of moves. Select for each move one square that is chosen for the first time during this
Second day move and colour the remaining squares already before starting the sequence. Then,
take all squares that were not selected and colour them in advance, i.e., already before
starting the sequence of moves. Since all selected squares must be different, every move
4. In square ABCD, points E and F are chosen in the in- D F C in the sequence now colours exactly one square.
terior of sides BC and CD, respectively. The line drawn Next, consider a bipartite graph with the 10 rows and 10 columns as vertices. Every time
from F perpendicular to AE passes through the intersec- a square is coloured, draw an edge between the row and the column corresponding to
tion point G of AE and diagonal BD. A point K is chosen this square. We claim that the graph is connected before we start the sequence of moves.
on FG such that | AK | = | EF|. Find ∠EKF. Indeed, suppose that during some move, we pick rows (a, b) and columns (c, d), such
that only the square (b, d) is coloured for the first time, i.e., we add the edge (b, d). But
Answer: 135◦ . K
then b is already connected with d through b − c − a − d, so the number of connected
Solution. Since AGFD is a cyclic quadrilateral (see Fig. 16), E components does not decrease. Since the graph of a fully coloured grid is connected, it
∠GAF = ∠GDF = 45◦ and ∠GFA = ∠GDA = 45◦ , so G
must also be connected in the beginning. But a connected graph with 20 vertices must
triangle AGF is isosceles and | GA| = | GF|. Now, right A B have at least 19 edges, so we can add only 100 − 19 = 81 new edges, and hence any
triangles AGK and FGE are congruent, and | GK | = | GE|,
sequence can have at most 81 moves.
so triangle GKE is also isosceles. Finally, ∠GKE = 45◦ and Figure 16
∠EKF = 180◦ − ∠GKE = 135◦ .

5. Find all continuous functions f : R → R such that for all reals x and y
f ( x + f (y)) = y + f ( x + 1).

Answer: f ( x ) = 1 + x and f ( x ) = 1 − x.
Solution. Taking y = − f ( x + 1), we see that there is a value a such that f (a) = 0. We
consider two cases.
Let first a 6= 1. Taking y = x + 1, we get f ( x + f ( x + 1)) = x + 1 + f ( x + 1). Let
g( x ) = x + f ( x + 1), then f ( g( x )) = 1 + g( x ) for all x. Since f is continuous, so is g.

23 24