53rd International Mathematical

Olympiad

Solutions

Author:
maths1900

September 30, 2012

 Problem 1: ( Evangelos Psychas, Greece ) Given a triangle ABC, let
J be the center of the excircle opposite to the vertex A. This circle is tangent
to lines AB, AC, and BC at K, L, and M , respectively. The lines BM and JF
meet at F , and the lines KM and CJ meet at G. Let S be the intersection of
AF and BC, and let T be the intersection of AG and BC. Prove that M is the
midpoint of BC.

Solution: We have KM ⊥ BJ hence BM is parallel to the bisector of ∠ABC.

Therefore ∠BM K = 21 ∠ABC and ∠F M B = 12 ∠ACB. Denote by X the inter-
section of KM and F J. From the triangle F XM we derive:
1
∠XF M = 90◦ − ∠F M B − ∠BM K = ∠BAC.
2
The points K and M are symmetric with respect to the line F J hence
∠KF J = ∠JF M = 21 ∠BAC = ∠KAJ, therefore K, J, A, and F belong
to a circle which implies that ∠JF A = ∠JKA = 90◦ and KM kAS. The
quadrilateral SKM A is a trapezoid and from ∠SM K = ∠AKM we obtain
SM = AK. Similarly, we get AL = T M . Since AK = AL (tangents from A to
the excircle) we get SM = T M .

1
 Problem 2: ( Angelo di Pasquale, Australia ) Let a2 , a3 , . . ., an be
positive real numbers that satisfy a2 · a3 · · · an = 1. Prove that
2 3 n
(a2 + 1) · (a3 + 1) · · · (an + n) nn .

Solution: The inequality between arithmetic and geometric mean implies

k
kk

k 1 1 1 1
(ak + 1) = ak + + + ··· + ≥ k k ·ak · = ·ak .
k−1 k−1 k−1 (k − 1)k−1 (k − 1)k−1
1
The inequality is strict unless ak = k−1 .
Multiplying analogous inequalities for k = 2, 3, . . ., n yields

2 3 n 33 44 nn
(a2 + 1) · (a3 + 1) · · · (an + n) · · · · · a2 a3 · · · an = nn .
22 33 (n − 1)n−1

The inequality is strict because at least one of a2 , . . ., an has to be greater

1
than or equal than 1 and thus ak k−1 holds for at least one integer k ∈
{2, . . . , n}.

2
 Problem 3: (David Arthur, Canada) The guessing game is a game
played between two players A and B. The rules of the game depend on two
positive integers k and n which are known to both players.
At the start of the game the player A chooses integers x and N with 1 ≤ x ≤
N . Player A keeps x secret, and truthfully tells N to the player B. The player B
now tries to obtain information about x by asking player A questions as follows:
each question consists of B specifying an arbitrary set S of positive integers
(possibly one specified in some previuos question), and asking A whether x
belongs to S. Player B may ask as many questions as he wishes. After each
question, player A must immediately answer it with yes or no, but is allowed
to lie as many times as she wants; the only restriction is that, among any k + 1
consecutive answers, at least one answer must be truthful.
After B has asked as many questions as he wants, he must specify a set X of
at most n positive integers. If x ∈ X, then B wins; otherwise, he loses. Prove
that:
(a) If n ≥ 2k then B has a winning strategy.
(b) There exists a positive integer k0 such that for every k ≥ k0 there exists
an integer n ≥ 1.99k for which B cannot guarantee a victory.

Solution:
The game can be re-formulated in an equivalent one: The player A chooses
an element x from the set S (with |S| = N ) and the player B asks the sequence
of questions. The j-th question consists of B choosing a set Dj ⊆ S and player
A selecting a set Pj ∈ Qj , QC j . For every j ≥ 1 the following relation holds:

x ∈ Pj ∪ Pj+1 ∪ · · · ∪ Pj+k .

The player B wins if after a finite number of steps he can choose a set X
with |X| ≤ n such that x ∈ X.
(a) It suffices to prove that if N ≥ 2k + 1 then the player B can determine
a set S 0 ⊆ S with |S 0 | ≤ N − 1 such that x ∈ S 0 .
Assume that N ≥ 2n + 1.
In the first move B selects any set D1 ⊆ S such that |D1 | ≥ 2k−1 and
|D1 | ≥ 2k−1 . After receiving the set P1 from A, B makes the second move. The
C

player B selects a set D2 ⊆ S such that |D2 ∩P1C | ≥ 2k−2 and |D2C ∩P1C | ≥ 2k−2 .
The player B continues this way: in the move j he/she chooses a set Dj such
that |Dj ∩ PjC | ≥ 2k−j and |DjC ∩ PjC | ≥ 2k−j .
In this way the player B has obtained the sets P1 , P2 , . . ., Pk such that
C
(P1 ∪ · · · ∪ Pk ) ≥ 1. Then B chooses the set Dk+1 to be a singleton containing
any element of P1 ∪ · · · ∪ Pk . There are two cases now:
1◦ The player A selects Pk+1 = Dk+1 C
. Then B can take S 0 = S \ Dk+1 and
the statement is proved.
2◦ The player A selects Pk+1 = Dk+1 . Now the player B repeats the previous
procedure on the set S1 = S \ Dk+1 to obtain the sequence of sets Pk+2 , Pk+3 ,

3
. . ., P2k+1 . The following inequality holds:

|S1 \ (Pk+2 · · · P2k+1 )| ≥ 1,

since |S1 | ≥ 2k . However, now we have

C
(Pk+1 ∪ Pk+2 ∪ · · · ∪ P2k+1 ) ≥ 1,

and we may take S 0 = Pk+1 ∪ · · · ∪ P2k+1 .

(b)
Let p and q be two positive integers such that 1.99 p q 2. Let us
choose k0 such that
 k0
p  q
≤2· 1− and pk − 1.99k 1.
q 2


We will prove that for every k ≥ k0 if |S| ∈ 1.99k , pk then
there is a strategy for the player A to select sets P1 , P2 , . . . (based on sets
D1 , D2 , . . . provided by B) such that for each j the following relation holds:

Pj ∪ Pj+1 ∪ · · · ∪ Pj+k = S.

Assuming that S = {1, 2, . . . ,N }, the playerA will maintain the following
j j j
sequence of N -tuples: (x)∞ 0 0
j=0 = x1 , x2 , . . . , xN . Initially we set x1 = x2 =
· · · = x0N = 1. After the set Pj is selected then we define xj+1 based on xj as
follows:

j+1 1, if i ∈ S
xi =
q · xji , if i 6∈ S.

The player A can keep B from winning if xji ≤ q k for each pair (i, j). For a
PN
sequence x,
let us define T (x) = i=1 xi . It suffices for player A to make sure
that T xj ≤ q k for each

j.
Notice that T x0 = N ≤ pk q k .

We will now prove that given xj such that T xj ≤ q k , and
a set Dj+1 the
C
player A can choose Pj+1 ∈ Dj+1 , Dj+1 such that T xj+1 ≤ q k .
Let y be the sequence that would be obtained if Pj+1 = Dj+1 , and let z be
C
the sequence that would be obtained if Pj+1 = Dj+1 . Then we have
X
T (y) = qxji + |Dj+1 |
C
i∈Dj+1
X
T (z) = qxji + Dj+1
C
.
i∈Dj+1

Summing up the previous two equalities gives:

4
 T (y) + T (z) = q · T xj + N ≤ q k+1 + pk , hence

q k pk
min {T (y) , T (z)} ≤ ·q + ≤ qk ,
2 2
because of our choice of k0 .

5
 Problem 4: (Liam Baker, South Africa ) Find all functions f : Z → Z
such that, for all integers a, b, c with a + b + c = 0 the following inequality holds:

f (a)2 + f (b)2 + f (c)2 = 2f (a)f (b) + 2f (b)f (c) + 2f (c)f (a).

Solution:
Placing a = b = c = 0 yields 3f (0)2 = 6f (0)2 which implies f (0) = 0.
Now we can place b = −a, c = 0 to obtain f (a)2 + f (−a)2 = 2f (a)f (−a), or,
2
equivalently (f (a) − f (−a)) = 0 which implies f (a) = f (−a).
Assume now that f (a) = 0 for some a ∈ Z. Then for any b we have a + b +
(−a − b) = 0 hence f (a)2 + f (b)2 + f (a + b)2 = 2f (b)f (a + b), which is equivalent
2
to (f (b) − f (a + b)) = 0, or f (a + b) = f (b). Therefore if f (a) = 0 for some
a 6= 0, then f is a periodic function with period a.
Placing b = a and c = 2a in the original equation yields f (2a)·(f (2a) − 4f (a)) =
0. Choosing a = 1 we get f (2) = 0 or f (2) = 4f (1).
If f (2) = 0, then f is periodic with period 2 and we must have f (n) = f (1)
for all odd n. It is easy to verify that for each c ∈ Z the function

0, 2 | n,
f (x) =
c, 2 6| n

satisfies the conditions of the problem.

Assume now that f (2) = 4f (1) and that f (1) 6= 0. We will now prove by
induction that f (n) = n2 · f (1). The statement holds for n ∈ {0, 1, 2}. Assume
that the statement holds for all i ∈ {1, 2, . . . , n} and let us prove it for n + 1.
We place a = 1, b = n, c = −n − 1 in the original equation to obtain:

f (1)2 + n4 f (1)2 + f (n + 1)2 = 2n2 f (1)2 + 2(n2 + 1)f (n + 1)f (1)

f (n + 1) − (n + 1)2 f (1) · f (n + 1) − (n − 1)2 f (1) = 0.

⇔

If f (n + 1) = (n − 1)2 f (1) then setting a = n + 1, b = 1 − n, and c = −2 in the

original equation yields

2(n − 1)4 f (1)2 + 16f (1)2 = 2 · 4 · 2(n − 1)2 f (1)2 + 2 · (n − 1)4 f (1)

which implies (n − 1)2 = 1 hence n = 2. Therefore f (3) = f (1). Placing

a = 1, b = 3, and c = 4 into the original equation implies that f (4) = 0 or
f (4) = 4f (1) = f (2). If f (4) 6= 0 we get

f (2)2 + f (2)2 + f (4)2 = 2f (2)2 + 4f (2)f (4)

hence f (4) = 4f (2). We already have that f (4) = f (2) and this implies that
f (2) = 0, which is impossible according to our assumption. Therefore f (4) = 0
and the function f has period 4. Then f (4k) = 0, f (4k +1) = f (4k +3) = c, and
f (4k + 2) = 4c. It is easy to verify that this function satisfies the requirements
of the problem.

6
 
0, 2 | n,
Thus the solutions are: f (x) = cx2 for some c ∈ Z; f (x) = for
c, 2 |6 n
some c ∈ Z;and
 0, 4 | n,
f (x) = c, 2 - n, for some c ∈ Z.
4c, n ≡ 2 (mod 4)


7
 Problem 5: (Josef Tkadlec, Czech Republic) Given a triangle ABC,
assume that ∠C = 90◦ . Let D be the foot of the perpendicular from C to AB,
and let X be any point of the segment CD. Let K and L be the points on the
segments AX and BX such that BK = BC and AL = AC, respectively. Let
M be the intersection of AL and BK. Prove that M K = M L.

Solution: Since AL2 = AC 2 = AD · AB the triangles ALD and ABL are

similar hence ∠ALD = ∠XBA.
Let R be the point on the extension of DC over point C such that DX ·DR =
BD · AD. Since ∠BDX = ∠RDA = 90◦ we conclude 4RAD ∼ 4BXD hence
∠XBD = ∠ARD, therefore ∠ALD = ∠ARD and the points R, A, D, and L
belong to a circle. This implies that ∠RLA = 90◦ hence RL2 = AR2 − AL2 =
AR2 − AC 2 .
Analogously we prove that RK 2 = BR2 − BC 2 and ∠RKB = 90◦ . Since
RC ⊥ AB we have AR2 − AC 2 = BR2 − BC 2 , therefore RL2 = RK 2 hence
RL = RK. Together with ∠RLM = ∠RKM = 90◦ we conclude 4RLM ∼ =
4LKM hence M K = M L.

8
 Problem 6: (Duan Djuki, Serbia ) Find all positive integers n for which
there exist non-negative integers a1 , a2 , . . ., an such that
1 1 1 1 2 n
+ a2 + · · · + an = a1 + a2 + · · · + an = 1.
2a1 2 2 3 3 3

Solution:
Let M = max{a1 , . . . , an }. Then we have 3M = 1 · 3M −a1 + 2 · 3M −a2 + · · · +
n · 3M −an ≡ 1 + 2 + · · · + n = n(n+1) 2 (mod n). Therefore, the number n(n+1)
2
must be odd and hence n ≡ 1 (mod 4) or n ≡ 2 (mod 4).
We will now prove that each n ∈ N of the form 4k + 1 or 4k + 2 for some
k ∈ N there exist integers a1 , . . . , an with the described property.
For a sequence a = (a1 , a2 , . . . , an ) let us introduce the following notation:

1 1 1 1 2 n
L(a) = + + · · · + an and R(a) = + + · · · + an .
2a1 2a2 2 3a1 3a2 3
Assume that for n = 2m + 1 there exists a sequence a = (a1 , . . . , an ) of
non-negative integers with L(a) = R(a) = 1. Consider the sequence a0 =
(a01 , . . . , a0n+1 ) defined in the following way:

0 aj , if j 6∈ {m + 1, 2m + 2}
aj =
am+1 + 1, if j ∈ {m + 1, 2m + 2}.
Then we have
1 1
L(a0 ) = L(a) − + 2 · am+1 +1 = 1
2am+1 2
m+1 m+1 2m + 2
R(a0 ) = R(a) − + am+1 +1 + am+1 +1 = 1.
3am+1 3 3
This implies that if the statement holds for 2m + 1, then it holds for 2m + 2.
Assume now that the statement holds for n = 4m + 2 for some m ≥ 2, and
assume that a = (a1 , . . . , a4m+2 ) is the corresponding sequence of n non-negative
integers. We will construct a following sequence a0 = a01 , a02 , . . . , a04m+13 that

satisfies L (a0 ) = R (a0 ) = 1 thus proving that the statement holds for 4m + 13.
Define:



 am+2 + 2, if j = m + 2


 aj + 1, if j ∈ {2m + 2, 2m + 3, 2m + 4, 2m + 5, 2m + 6}
a0j = a j + 1,
2
if j ∈ {4m + 4, 4m + 6, 4m + 8, 4m + 10, 4m + 12}
 am+2 + 3, if j ∈ {4m + 3, 4m + 5, 4m + 7, 4m + 9, 4m + 11, 4m + 13}



aj , otherwise.


We now have

9
 6 6 6
1 X 1 1 X 1 X 1 1
L (a0 ) = L(a)− − + + + +6· am+2 +3 = 1.
2am+2 j=2 2a2m+j 2am+2 +2 j=2 2a2m+j +1 j=2 2a2m+j +1 2

It remains to verify that R (a0 ) = R(a) = 1. We write

 0
am+2 , a04m+3 , a04m+5 , a04m+7 , a04m+9 , a04m+11 , a04m+13
  
0 am+2
R (a )−R(a) = R −R
m + 2, 4m + 3, 4m + 5, 4m + 7, 4m + 9, 4m + 11, 4m + 13 m+2

6  
a02m+j , a04m+2j
  
X a2m+j
+ R −R ,
2m + j, 4m + j 2m + j
j=2

where  
c1 , . . . , ck d1 dk
R = + ··· + c .
d1 , . . . , dk 3 c1 3k
For each j ∈ {2, 3, 4, 5, 6} we have
 0
a2m+j , a04m+2j
  
a2m+j 2m + j 4m + 2j 2m + j
R −R = a2m+j +1 + a2m+j +1 − a2m+j = 0.
2m + j, 4m + j 2m + j 3 3 3

The first term in the expression for R (a0 ) − R(a) is also equal to 0 because

a0m+2 , a04m+3 , a04m+5 , a04m+7 , a04m+9 , a04m+11 , a04m+13

   
am+2
R −R
m + 2, 4m + 3, 4m + 5, 4m + 7, 4m + 9, 4m + 11, 4m + 13 m+2

6
m+2 X 4m + 2j + 1 m + 2
= am+2 +2 + − am+2 = 0.
3 j=1
3am+2 +3 3

Thus R (a0 ) = 0 and the statement holds for 4m + 13. It remains to verify
that there are sequences of lengths 1, 5, 9, 13, and 17. One way to choose these
sequences is:

(1), (2, 1, 3, 4, 4), (2, 3, 3, 3, 3, 4, 4, 4, 4), (2, 3, 3, 4, 4, 4, 5, 4, 4, 5, 4, 5, 5),

(3, 2, 2, 4, 4, 5, 5, 6, 5, 6, 6, 6, 6, 6, 6, 6, 5).

10