Solutions of JBMO 2012

Wednesday, June 27, 2012

Problem 1
Let a, b and c be positive real numbers such that a + b + c = 1 . Prove that
a b b c c a ⎛ 1− a 1− b 1− c ⎞
+ + + + + + 6 ≥ 2 2 ⎜⎜ + + ⎟
b a c b a c ⎝ a b c ⎟⎠
When does equality hold?

Solution
Replacing 1 − a,1 − b,1 − c with b + c, c + a, a + b respectively on the right hand side,
the given inequality becomes
b+c c+a a+b ⎛ b+c c+a a+b ⎞
+ + + 6 ≥ 2 2 ⎜⎜ + + ⎟
a b c ⎝ a b c ⎟⎠
and equivalently

⎛b+c b+c ⎞ ⎛c+a c+a ⎞ ⎛ a+b a+b ⎞

⎜⎜ −2 2 + 2 ⎟⎟ + ⎜⎜ −2 2 + 2 ⎟⎟ + ⎜⎜ −2 2 + 2 ⎟⎟ ≥ 0
⎝ a a ⎠ ⎝ b b ⎠ ⎝ c c ⎠

which can be written as

2 2 2
⎛ b+c ⎞ ⎛ c+a ⎞ ⎛ a+b ⎞
⎜⎜ − 2 ⎟⎟ + ⎜⎜ − 2 ⎟⎟ + ⎜⎜ − 2 ⎟⎟ ≥ 0 ,
⎝ a ⎠ ⎝ b ⎠ ⎝ c ⎠
which is true.
The equality holds if and only if
b+c c+a a+b
= = ,
a b c
1
which together with the given condition a + b + c = 1 gives a = b = c = .
3
Problem 2
Let the circles k1 and k 2 intersect at two distinct points Α and Β , and let t be a
common tangent of k1 and k 2 , that touches k1 and k 2 at Μ and Ν , respectively. If
t ⊥ AM and ΜΝ = 2ΑΜ , evaluate ∠NMB .

Solution 1
Let P be the symmetric of A with respect to M (Figure 1). Then AM = MP and
t ⊥ AP , hence the triangle APN is isosceles with AP as its base, so ∠NAP = ∠NPA .
We have ∠BAP = ∠BAM = ∠BMN and ∠BAN = ∠BNM .
Thus we have
1800 − ∠NBM = ∠BNM + ∠BMN = ∠BAN + ∠BAP = ∠NAP = ∠NPA
so the quadrangle MBNP is cyclic (since the points B and P lie on different sides of
MN ). Hence ∠APB = ∠MPB = ∠MNB and the triangles APB and MNB are
congruent ( ΜΝ = 2ΑΜ = ΑΜ + ΜΡ = ΑΡ ). From that we get AB = MB , i.e. the
triangle AMB is isosceles, and since t is tangent to k1 and perpendicular to AM, the
centre of k1 is on AM, hence AMB is a right-angled triangle. From the last two
statements we infer ∠AMB = 450 , and so ∠NMB = 90D − ∠AMB = 450 .

Figure 1

Solution 2
Let C be the common point of MN, AB (Figure 2). Then CN 2 = CB ⋅ CA and
CM 2 = CB ⋅ CA . So CM = CN . But MN = 2 AM , so CM = CN = AM , thus the right
triangle ACM is isosceles, hence ∠NMB = ∠CMB = ∠BCM = 450 .

Figure 2
Problem 3
On a board there are n nails each two connected by a string. Each string is colored in
one of n given distinct colors. For each three distinct colors, there exist three nails
connected with strings in these three colors. Can n be
a) 6? b) 7?
Solution. (a) The answer is no.
Suppose it is possible. Consider some color, say blue. Each blue string is the side of 4
⎛ 5 ⎞ 5·4
triangles formed with vertices on the given points. As there exist ⎜ ⎟ = = 10 pairs
⎝2⎠ 2
of colors other than blue, and for any such pair of colors together with the blue color
there exists a triangle with strings in these colors, we conclude that there exist at least
3 blue strings (otherwise the number of triangles with a blue string as a side would be
at most 2·4 = 8 , a contradiction). The same is true for any color, so altogether there
⎛ 6 ⎞ 6·5
exist at least 6·3 = 18 strings, while we have just ⎜ ⎟ = = 15 of them.
⎝2⎠ 2
(b) The answer is yes
Put the nails at the vertices of a regular 7-gon and color each one of its sides in a
different color. Now color each diagonal in the color of the unique side parallel to it.
It can be checked directly that each triple of colors appears in some triangle (because
of symmetry, it is enough to check only the triples containing the first color).

Remark. The argument in (a) can be applied to any even n. The argument in (b) can
be applied to any odd n = 2k + 1 as follows: first number the nails as 0,1, 2 …, 2k
and similarly number the colors as 0,1, 2 …, 2k . Then connect nail x with nail y by a
string of color x + y (modn) . For each triple of colors ( p, q, r ) there are vertices
x, y, z connected by these three colors. Indeed, we need to solve (mod n) the system
(*) ( x + y ≡ p, x + z ≡ q, y + z ≡ r )
Adding all three, we get 2(x+ y + z) ≡ p + q + r and multiplying by k + 1 we get
x+ y + z ≡ (k +1)(p + q + r) . We can now find x, y, z from the identities (*) .
Problem 4
Find all positive integers x, y, z and t such that
2 x ·3y + 5z = 7t .

Solution
Reducing modulo 3 we get 5z ≡ 1 , therefore z is even, z = 2c, c ∈ ` .
Next we prove that t is even:
Obviously, t ≥ 2 . Let us suppose that t is odd, say t = 2d + 1, d ∈ ` . The equation
becomes 2 x ·3y + 25c = 7·49d. If x ≥ 2 , reducing modulo 4 we get 1 ≡ 3 , a
contradiction. And if x = 1 , we have 2·3 y + 25c = 7·49d and reducing modulo 24 we
obtain
2·3 y + 1 ≡ 7 ⇒ 24 | 2(3 y − 3) , i.e. 4 | 3 y −1 − 1
which means that y − 1 is even. Then y = 2b + 1, b ∈ ` . We obtain 6·9b + 25c = 7·49d ,
and reducing modulo 5 we get ( −1)b ≡ 2·( −1) d , which is false for all b, d ∈ ` . Hence
t is even, t = 2d , d ∈ ` , as claimed.
Now the equation can be written as
2 x ·3 y + 25d = 49d ⇔ 2 x ·3 y = ( 7d − 5c )( 7d + 5c ) .
As gcd ( 7d − 5c ,7d + 5c ) = 2 and 7d + 5c > 2 , there exist exactly three possibilities:
⎧⎪7d − 5d = 2 x −1 ⎧⎪7d − 5d = 2·3y ⎧⎪7d − 5d = 2
(1) ⎨ d d ; (2) ⎨ d d x −1
; (3) ⎨ d d x −1 y
⎪⎩7 + 5 = 2·3 ⎪⎩7 + 5 = 2 ⎩⎪7 + 5 = 2 ·3
y

Case (1)
We have 7d = 2 x −2 + 3 y and reducing modulo 3, we get 2 x −2 ≡ 1(mod 3) , hence x − 2
is even, i.e. x = 2a + 2, a ∈ ` , where a > 0 , since a = 0 would mean 3 y + 1 = 7d ,
which is impossible (even = odd).
We obtain
mod 4
7d − 5d = 2·4a ⇒ 7d ≡ 1(mod 4) ⇒ d = 2e, e ∈ ` .
Then we have
mod 8
49e − 5d = 2·4a ⇒ 5c ≡ 1(mod 8) ⇒ c = 2 f , f ∈ ` .
mod 3
We obtain 49e − 25 f = 2·4a ⇒ 0 ≡ 2( mod 3) , false. In conclusion, in this case there
are no solutions to the equation.

Case (2)
From 2 x −1 = 7d + 5c ≥ 12 we obtain x ≥ 5 . Then 7d + 5c ≡ 0(mod 4) , i.e.
3 + 1 ≡ 0(mod 4) , hence d is odd. As 7 = 5 + 2·3 ≥ 11 , we get d ≥ 2 , hence
d d c y

d = 2e + 1, e ∈ ` .
As in the previous case, from 7d = 2 x −2 + 3 y reducing modulo 3 we obtain x = 2a + 2
with a ≥ 2 (because x ≥ 5 ). We get 7d = 4a + 3 y i.e. 7·49e = 4a + 3 y , hence, reducing
modulo 8 we obtain 7 ≡ 3 y which is false, because 3 y is congruent either to 1 (if y is
even) or to 3 (if y is odd). In conclusion, in this case there are no solutions to the
equation.
Case (3)
From 7d = 5c + 2 it follows that the last digit of 7d is 7, hence d = 4k + 1, k ∈ ` .

If c ≥ 2 , from 74 k +1 = 5c + 2 reducing modulo 25 we obtain 7 ≡ 2( mod 25) which is

false. For c = 1 we get d = 1 and the solution x = 3, y = 1, z = t = 2 .
Problem 1.

Solution. As a3 b − 1 = b(a3 + 1) − (b + 1) and a + 1 | a3 + 1, we have a + 1 | b + 1.

As b3 a + 1 = a(b3 − 1) + (a + 1) and b − 1 | b3 − 1, we have b − 1 | a + 1.
So b − 1 | b + 1 and hence b − 1 | 2.

• If b = 2, then a + 1 | b + 1 = 3 gives a = 2. Hence (a, b) = (2, 2) is the only solution in

this case.
• If b = 3, then a + 1 | b + 1 = 4 gives a = 1 or a = 3. Hence (a, b) = (1, 3) and (3, 3) are
the only solutions in this case.

To summarize, (a, b) = (1, 3), (2, 2) and (3, 3) are the only solutions.
Problem 2.

Solution. We will show that M OP D is a parallelogram. From this it follows that M , N , P

are collinear.
Since ∠BAD = ∠CAO = 90◦ − ∠ABC , D is the foot of the perpendicular from A to side
BC . Since M is the midpoint of the line segment BE , we have BM = M E = M D and hence
∠M DE = ∠M ED = ∠ACB .
Let the line M D intersect the line AC at D1 . Since ∠ADD1 = ∠M DE = ∠ACD, M D is
perpendicular to AC . On the other hand, since O is the center of the circumcircle of triangle
ABC and P is the midpoint of the side AC , OP is perpendicular to AC . Therefore M D and
OP are parallel.
Similarly, since P is the midpoint of the side AC , we have AP = P C = DP and hence
∠P DC = ∠ACB . Let the line P D intersect the line BE at D2 . Since ∠BDD2 = ∠P DC =
∠ACB = ∠BED, we conclude that P D is perpendicular to BE . Since M is the midpoint of
the line segment BE , OM is perpendicular to BE and hence OM and P D are parallel.
Problem 3.

Solution 1. By the AM-GM Inequality we have:

a+1 2
+ ≥2
2 a+1
Therefore
2 a+3
a + 2b + ≥ + 2b .
a+1 2
and, similarly,
2 b+3
b + 2a + ≥ 2a + .
b+1 2

On the other hand,

√ √
(a + 4b + 3)(b + 4a + 3) ≥ ( ab + 4 ab + 3)2 ≥ 64

by the Cauchy-Schwarz Inequality as ab ≥ 1, and we are done.

Since ab ≥ 1, we have a + b ≥ a + 1/a ≥ 2 a · (1/a) = 2.

p
Solution 2.

Then
2 2
a + 2b + = b + (a + b) +
a+1 a+1
2
≥b+2+
a+1
b+1 b+1 2
= + +1+
2
s 2 a+1
(b + 1)2
≥44
2(a + 1)

by the AM-GM Inequality. Similarly,

s
2 (a + 1)2
b + 2a + ≥44 .
b+1 2(b + 1)
Now using these and applying the AM-GM Inequality another time we obtain:
   r
2 2 (a + 1)(b + 1)
4
a + 2b + b + 2a + ≥ 16
a+1 b+1 4
s
√ √
4 (2 a)(2 b)
≥ 16
4
√
8
= 16 ab
≥ 16

Solution 3. We have

     
2 2 2 2
a + 2b + b + 2a + = (a + b) + b + (a + b) + a +
a+1 b+1 a+1 b+1
!2
√ 2
≥ a+b+ ab + p
(a + 1)(b + 1)

by the Cauchy-Schwarz Inequality.

On the other hand,
2 4
p ≥
(a + 1)(b + 1) a+b+2

by the AM-GM Inequality and

√ 2 4 (a + b + 1)(a + b − 2)
a+b+ ab + p ≥a+b+1+ = +4≥4
(a + 1)(b + 1) a+b+2 a+b+2

√
as a + b ≥ 2 ab ≥ 2, nishing the proof.
Problem 4.

Solution. a. Yes. Let a ≤ b ≤ c ≤ d ≤ e be the numbers chosen by Alice. As each number

appears in a pairwise sum 4 times, by adding all 10 pairwise sums and dividing the result by
4, Bob obtains a + b + c + d + e. Subtracting the smallest and the largest pairwise sums a + b
and d + e from this he obtains c. Subtracting c from the second largest pairwise sum c + e he
obtains e. Subtracting e from the largest pairwise sum d + e he obtains d. He can similarly
determine a and b.

b. Yes. Let a ≤ b ≤ c ≤ d ≤ e ≤ f be the numbers chosen by Alice. As each number appears

in a pairwise sum 5 times, by adding all 15 pairwise sums and dividing the result by 5, Bob
obtains a + b + c + d + e + f . Subtracting the smallest and the largest pairwise sums a + b and
e + f from this he obtains c + d.
Subtracting the smallest and the second largest pairwise sums a + b and d + f from a + b + c +
d + e + f he obtains c + e. Similarly he can obtain b + d. He uses these to obtain a + f and
b + e.
Now a+d, a+e, b+c are the three smallest among the remaining six pairwise sums. If Bob adds
these up, subtracts the known sums c + d and b + e from the result and divides the dierence
by 2, he obtains a. Then he can determine the remaining numbers.

c. No. If Alice chooses the eight numbers 1, 5, 7, 9, 12, 14, 16, 20, then Bob cannot be sure to
guess these numbers correctly as the eight numbers 2, 4, 6, 10, 11, 15, 17, 19 also give exactly the
same 28 pairwise sums as these numbers.
 Problems and Solutions, JBMO 2014

Problem 1. Find all distinct prime numbers p , q and r such that

3p 4 - 5q 4 - 4r 2 = 26 .

Solution. First notice that if both primes q and r differ from 3 , then q 2 º r 2 º 1(mod 3) ,
hence the left hand side of the given equation is congruent to zero modulo 3, which is
impossible since 26 is not divisible by 3 . Thus, q = 3 or r = 3 . We consider two cases.
Case 1. q = 3 .
The equation reduces to 3p 4 - 4r 2 = 431 (1) .
If p ¹ 5, by Fermat’s little theorem, p 4 º 1 (mod 5) , which yields
3 - 4r 2 º 1 (mod 5) , or equivalently, r 2 + 2 º 0 (mod 5) . The last congruence is
impossible in view of the fact that a residue of a square of a positive integer belongs to the
set { 0, 1, 4 } . Therefore p = 5 and r = 19 .
Case 2. r = 3 .
The equation becomes 3p 4 - 5q 4 = 62 (2) .
Obviously p ¹ 5 . Hence, Fermat’s little theorem gives p 4 º 1 (mod 5) . But then
5q 4 º 1 (mod 5) , which is impossible .
Hence, the only solution of the given equation is p = 5 , q = 3 , r = 19 .
Problem 2. Consider an acute triangle ABC with area S. Let CD ^ AB ( D Î AB ),
DM ^ AC ( M Î AC ) and DN ^ BC ( N Î BC ). Denote by H 1 and H 2 the
orthocentres of the triangles MNC and MND respectively. Find the area of the
quadrilateral AH 1BH 2 in terms of S.

Solution 1. Let O, P, K, R and T be the C

mid-points of the segments CD, MN, R
CN, CH 1 and MH 1 , respectively. From H1 K
1 N
DMNC we have that PK = MC and O
2 T
P
PK  MC . Analogously, from DMH 1C
M
1
we have that TR = MC and
2 A B
D
TR  MC . Consequently, PK = TR and
H2
PK  TR . Also OK  DN (from
DCDN ) and since DN ^ BC and MH 1 ^ BC , it follows that TH 1  OK . Since O is the
circumcenter of DCMN , OP ^ MN . Thus, CH 1 ^ MN implies OP  CH 1 . We conclude
DTRH 1 @ DKPO (they have parallel sides and TR = PK ), hence RH 1 = PO , i.e.
CH 1 = 2PO and CH 1  PO .

Analogously, DH 2 = 2PO and DH 2  PO . From CH 1 = 2PO = DH 2 and

CH 1  PO  DH 2 the quadrilateral CH 1H 2D is a parallelogram, thus H 1H 2 = CD and

H 1H 2  CD . Therefore the area of the quadrilateral AH 1BH 2 is
AB ⋅ H 1H 2 AB ⋅ CD
= =S .
2 2
Solution 2. Since MH 1  DN and NH 1  DM , MDNH 1 is a parallelogram. Similarly,
NH 2  CM and MH 2  CN imply MCNH 2 is a parallelogram . Let P be the midpoint of
the segment MN . Then sP (D ) = H 1 and sP (C ) = H 2 , thus CD  H 1H 2 and CD = H 1H 2 .
1
From CD ^ AB we deduce AAH BH = AB ⋅ CD = S .
1 2
2
Problem 3. Let a, b, c be positive real numbers such that abc = 1 . Prove that
2 2 2
æ 1ö æ 1ö æ 1ö
ççça + ÷÷÷ + çççb + ÷÷÷ + çççc + ÷÷÷ ³ 3 (a + b + c + 1) .
è b ÷ø è c ÷ø è a ø÷
When does equality hold?
Solution 1. By using AM-GM ( x 2 + y 2 + z 2 ³ xy + yz + zx ) we have
2 2 2
æ ö æ ö æ ö æ öæ ö æ öæ ö æ öæ ö
çça + 1 ÷÷ + ççb + 1 ÷÷ + ççc + 1 ÷÷ ³ çça + 1 ÷÷ ççb + 1 ÷÷ + ççb + 1 ÷÷ ççc + 1 ÷÷ + ççc + 1 ÷÷ çça + 1 ÷÷
çè b ÷÷ø çè c ÷÷ø çè a ÷÷ø çè b ÷÷øèç c ÷÷ø èç ÷÷ ç
c øè a ø÷÷ èç ÷÷ ç
a øè b ø÷÷
æ a ö æ b ö æ c ö
= çççab + 1 + + a ÷÷÷ + çççbc + 1 + + b ÷÷÷ + çççca + 1 + + c ÷÷÷
è c ÷ø è a ÷ø è b ÷ø
a c b
= ab + bc + ca +
+ + + 3 +a +b +c.
c b a
b c a
Notice that by AM-GM we have ab + ³ 2b, bc + ³ 2c, and ca + ³ 2a .
a b c
Thus ,
2 2 2
æ ö æ ö æ ö æ ö æ ö æ ö
çça + 1 ÷÷ + ççb + 1 ÷÷ + ççc + 1 ÷÷ ³ ççab + b ÷÷ + ççbc + c ÷÷ + ççca + a ÷÷ + 3 + a + b + c ³ 3(a + b + c + 1).
çè b ÷ø÷ èç c ø÷÷ èç a ø÷÷ èç a ø÷÷ èç b ø÷÷ èç c ø÷÷

The equality holds if and only if a = b = c = 1.

Solution 2. From QM-AM we obtain

(a + ) + (b + c1 ) + (c + a1 )
2 2 2
1
b a + b1 + b + c1 + c + a1
³ 
3 3
(a + b1 + b + c1 + c + a1 )
2 2 2 2
æ ö æ ö æ ö÷
çça + 1 ÷÷ + ççb + 1 ÷÷ 1
+ çççc + ÷÷ ³ (1)
çè b ø÷÷ èç c ø÷÷ è a ø÷ 3
1 1 1 1
From AM-GM we have + + ³ 33 = 3 , and substituting in (1) we get
a b c abc

çça + 1 ÷÷ + ççb + 1 ÷÷ + ççc + 1 ÷÷ ³ (

a + b1 + b + c1 + c + a1 ) (a + b + c + 3)
2 2 2 2 2
æ ö æ ö æ ö
³ =
çè b ÷÷ø çè c ÷÷ø çè a ÷÷ø 3 3
(a + b + c )(a + b + c ) + 6 (a + b + c ) + 9 (a + b + c ) 3 3 abc + 6 (a + b + c ) + 9
= ³ =
3 3
9 (a + b + c ) + 9
= = 3 (a + b + c + 1).
3
The equality holds if and only if a = b = c = 1.

Solution 3.
By using x 2 + y 2 + z 2 ³ xy + yz + zx
 2 2 2
æ ö æ ö æ ö
çça + 1 ÷÷ + ççb + 1 ÷÷ + ççc + 1 ÷÷ = a 2 + b 2 + c 2 + 1 + 1 + 1 + 2a + 2b + 2c ³
çè b ÷÷ø çè c ÷÷ø çè a ÷÷ø b2 c2 a 2 b c a
1 1 1 2a 2b 2c
³ ab + ac + bc + + + + + + .
bc ca ab b c a

Clearly
1 1 1 abc abc abc
+ + = + + = a +b +c,
bc ca ab bc ca ab
a b c
ab + + bc + + ca + ³ 2a + 2b + 2c,
b c a
a b c a b c
+ + ³ 3 3 ⋅ ⋅ = 3.
b c a b c a
Hence
2 2 2
æ ö æ ö æ ö æ ö æ ö æ ö
çça + 1 ÷÷ + ççb + 1 ÷÷ + ççc + 1 ÷÷ ³ ççab + a ÷÷ + ççac + c ÷÷ + ççbc + b ÷÷ + a + b + c + a + b + c ³
çè b ÷ø÷ çè c ø÷÷ èç a ø÷÷ èç b ø÷÷ èç a ø÷÷ èç c ø÷÷ b c a
³ 2a + 2b + 2c + a + b + c + 3 = 3 (a + b + c + 1).

The equality holds if and only if a = b = c = 1.

x y z
Solution 4. a = ,b= ,c=
y z x
2 2 2
æx z ö æy x ö æz y ö æx y z ö
ççç + ÷÷÷ + ççç + ÷÷÷ + ççç + ÷÷÷ ³ 3 ççç + + + 1÷÷÷
è y y ÷ø è z z ø÷ è x x ø÷ èy z x ø÷
(x + z )2 x 2z 2 + (y + x )2 y 2x 2 + (z + y )2 z 2y 2 ³ 3xyz (x 2z + y 2x + z 2y + xyz )
x 4z 2 + 2x 3z 3 + x 2z 4 + x 2y 4 + 2x 3y 3 + x 4y 2 + y 2z 4 + 2y 3z 3 + y 4z 2 ³ 3x 3yz 2 + 3x 2y 3z + 3xy 2z 3 + 3x 2y 2z 2
1)x 3y 3 + y 3z 3 + z 3x 3 ³ 3x 2y 2z 2 .
2)x 4z 2 + z 4x 2 + x 3y 3 ³ 3x 3z 2y üïï
ï
3)x 4y 2 + y 4x 2 + y 3z 3 ³ 3y 3x 2z ïý
ï
4)z 4y 2 + y 4z 2 + x 3z 3 ³ 3z 3y 2x ïïï
þ
Equality holds when x = y = z , i.e., a = b = c = 1 .
1
Solution 5. å (a + b ) 2
³ 3å a + 3
cyc cyc

a æ 1 ö
 2å + å çça 2 + 2 - 3a - 1÷÷÷ ³ 0
cyc b
ç
cyc è a ÷ø

a abc
2å ³ 63 =6 (1)
cyc b bca
 1 3
"a > 0, a 2 +
2
- 3a ³ - 4
a a
4 3 2
 a - 3a + 4a - 3a + 1 ³ 0

(
 (a - 1) a 2 - a + 1 ³ 0 )
2

æ 1 ö÷ 1 1
å çèçça 2
+
a 2
- 3a - 1÷÷÷ ³ 3å a - 15 ³ 9 abc - 15 = -6
ø
3 (2)
cyc cyc

Using (1) and (2) we obtain

a æ 1 ö
2å + å ççça 2 + 2 - 3a - 1÷÷÷ ³ 6 - 6 = 0
cyc b è a ø÷
Equality holds when a = b = c = 1 .
Problem 4. For a positive integer n , two players A and B play the following game: Given
a pile of s stones, the players take turn alternatively with A going first. On each turn the
player is allowed to take either one stone, or a prime number of stones, or a multiple of n
stones. The winner is the one who takes the last stone. Assuming both A and B play
perfectly, for how many values of s the player A cannot win?

Solution. Denote by k the sought number and let {s1, s2 ,..., sk } be the corresponding values
for s . We call each si a losing number and every other nonnegative integer a winning
numbers.

Clearly every multiple of n is a winning number.

Suppose there are two different losing numbers si > s j , which are congruent modulo n .
Then, on his first turn of play, player A may remove si - s j stones (since n si - s j ),
leaving a pile with s j stones for B. This is in contradiction with both si and s j being
losing numbers.

Hence, there are at most n - 1 losing numbers, i.e. k £ n - 1 .

Suppose there exists an integer r Î {1,2,..., n - 1} , such that mn + r is a winning number

for every m Î  0 . Let us denote by u the greatest losing number (if k > 0 ) or 0 (if
k = 0 ), and let s = LCM (2, 3,..., u + n + 1) . Note that all the numbers s + 2 , s + 3 , …,
s +u +n +1 are composite. Let m ' Î 0, be such that
s + u + 2 £ m ' n + r £ s + u + n + 1 . In order for m ' n + r to be a winning number, there
must exist an integer p , which is either one, or prime, or a positive multiple of n , such
that m ' n + r - p is a losing number or 0, and hence lesser than or equal to u . Since
s + 2 £ m ' n + r - u £ p £ m ' n + r £ s + u + n + 1 , p must be a composite, hence p is a
multiple of n (say p = qn ). But then m ' n + r - p = (m '- q ) n + r must be a winning
number, according to our assumption. This contradicts our assumption that all numbers
mn + r , m Î  0 are winning.

Hence, each nonzero residue class modulo n contains a loosing number.

There are exactly n - 1 losing numbers .

Lemma: No pair (u, n ) of positive integers satisfies the following property:
(*) In  exists an arithmetic progression (at )t¥=1 with difference n such that each
segment
éa - u, a + u ù contains a prime.
êë i i úû

Proof of the lemma: Suppose such a pair (u, n ) and a corresponding arithmetic
¥
progression (a )
t t =1
exist. In  exist arbitrarily long patches of consecutive composites.
Take such a patch P of length 3un . Then, at least one segment éêëai - u, ai + u ùúû is fully
contained in P , a contradiction.

Suppose such a nonzero residue class modulo n exists (hence n > 1 ). Let u Î  be greater
than every loosing number. Consider the members of the supposed residue class which are
greater than u . They form an arithmetic progression with the property (*) , a
contradiction (by the lemma).
 19th Junior Balkan Mathematical Olympiad
June 24-29, 2015, Belgrade, Serbia

Problem 1. Find all prime numbers a, b, c and positive integers k which satisfy the
equation
a 2  b 2  16  c 2  9  k 2  1.
Solution:
The relation 9  k 2  1  1 mod 3 implies
a 2  b 2  16  c 2  1 mod 3  a 2  b 2  c 2  1 mod 3 .
Since a 2  0, 1 (mod 3), b 2  0, 1 (mod 3), c 2  0, 1 (mod 3) , we have:

a2 0 0 0 0 1 1 1 1
b2 0 0 1 1 0 0 1 1
c2 0 1 0 1 0 1 0 1
a2  b2  c2 0 1 1 2 1 2 2 0

From the previous table it follows that two of three prime numbers a, b, c are equal to 3.
Case 1. a  b  3
We have
a 2  b 2  16  c 2  9  k 2  1  9  k 2  16  c 2  17  3k  4c   3k  4c   17,
3k  4c  1, c  2,
  and (a, b, c, k )  (3,3,2,3).
3k  4c  17, k  3,
Case 2. c  3
If (3, b0 , c, k ) is a solution of the given equation, then (b0 ,3, c, k ) is a solution too.
Let a  3. We have
a 2  b 2  16  c 2  9  k 2  1  9  k 2  b 2  152  3k  b  3k  b  152 .
Both factors shall have the same parity and we obtain only 2 cases:
3k  b  2, b  37,
   and (a, b, c, k )  (3,37,3,13);
3k  b  76, k  13,
3k  b  4, b  17,
   and (a, b, c, k )  (3,17,3,7).
3k  b  38, k  7,

So, the given equation has 5 solutions:

(37,3,3,13), (17,3,3,7), (3,37,3,13), (3,17,3,7), (3,3,2,3).

Problem 2. Let a, b, c be positive real numbers such that a  b  c  3 . Find the minimum
value of the expression
2  a 3 2  b3 2  c 3
A   .
a b c
 19th Junior Balkan Mathematical Olympiad
June 24-29, 2015, Belgrade, Serbia

Solution:
We can rewrite A as follows:
2  a3 2  b3 2  c3  1 1 1
A    2     a 2  b 2  c 2 
a b c a b c
 ab  bc  ca   ab  bc  ca 
2   (a  b  c )  2   ((a  b  c)  2(ab  bc  ca )) 
2 2 2 2

 abc   abc 
 ab  bc  ca   ab  bc  ca 
2   (9  2(ab  bc  ca ))  2   2(ab  bc  ca )  9 
 abc   abc 
 1 
2(ab  bc  ca )  1  9.
 abc 
Recall now the well-known inequality ( x  y  z) 2  3( xy  yz  zx) and set
x  ab, y  bc, z  ca , to obtain (ab  bc  ca)  3abc(a  b  c)  9abc, where we have
2

used a  b  c  3 . By taking the square roots on both sides of the last one we obtain:
ab  bc  ca  3 abc . (1)
Also by using AM-GM inequality we get that
1 1
1  2 . (2)
abc abc
Multiplication of (1) and (2) gives:
 1  1
(ab  bc  ca )  1  3 abc  2  6.
 abc  abc
So A  2  6  9  3 and the equality holds if and only if a  b  c  1 , so the minimum value is
3.

Problem 3. Let ABC be an acute triangle. The lines l1, l2 are perpendicular to AB at the
points A, B respectively. The perpendicular lines from the midpoint M of AB to the sides of
the triangle AC, BC intersect the lines l1, l2 at the points E, F, respectively. If D is the
intersection point of EF and MC, prove that
ADB  EMF .

Solution:
Let H, G be the points of intersection of ME, MF with AC, BC respectively. From the
MH MA
similarity of triangles MHA and MAE we get  , thus
MA ME
MA2  MH  ME. (1)
MB MG
Similarly, from the similarity of triangles MBG and MFB we get  , thus
MF MB
MB 2  MF  MG. (2)
Since MA  MB , from (1), (2), we conclude that the points E, H, G, F are concyclic.
 19th Junior Balkan Mathematical Olympiad
June 24-29, 2015, Belgrade, Serbia

C
l1 F

l2
D

G
E
H

A B
M

Therefore, we get that FEH  FEM  HGM . Also, the quadrilateral CHMG is
cyclic, so CMH  HGC . We have
FEH  CMH  HGM  HGC  90 ,
thus CM  EF . Now, from the cyclic quadrilaterals FDMB and EAMD , we get that
DFM  DBM and DEM  DAM . Therefore, the triangles EMF and ADB are
similar, so ADB  EMF .

Problem 4.
An -figure is one of the following four pieces, each consisting of three unit squares:

A 5×5 board, consisting of 25 unit squares, a positive integer k ≤ 25 and an unlimited supply
L-figures are given. Two players, A and B, play the following game: starting with A they
alternatively mark a previously unmarked unit square until they marked a total of k unit
squares.
We say that a placement of L-figures on unmarked unit squares is called good if the L-figure
do not overlap and each of them covers exactly three unmarked unit squares of the board.
B wins if every good placement of L-figures leaves uncovered at least three unmarked unit
squares. Determine the minimum value of k for which B has a winning strategy.

Solution:
We will show that player A wins if k = 1, 2, 3, but player B wins if k  4 . Thus the smallest
k for which B has a winning strategy exists and is equal to 4.
If k  1 , player A marks the upper left corner of the square and then fills it as follows.
 19th Junior Balkan Mathematical Olympiad
June 24-29, 2015, Belgrade, Serbia

If k  2 , player A marks the upper left corner of the square. Whatever square player B marks,
then player A can fill in the square in exactly the same pattern as above except that he doesn't
put the L-figure which covers the marked square of B. Player A wins because he has left only
two unmarked squares uncovered.
For k  3 , player A wins by following the same strategy. When he has to mark a square for
the second time, he marks any yet unmarked square of the L-figure that covers the marked
square of B.
Let us now show that for k  4 player B has a winning strategy. Since there will be 21
unmarked squares, player A will need to cover all of them with seven L-figures. We can
assume that in his first move, player A does not mark any square in the bottom two rows of
the chessboard (otherwise just rotate the chessboard). In his first move player B marks the
square labeled 1 in the following figure.

If player A in his next move does not mark any of the squares labeled 2, 3 and 4 then player B
marks the square labeled 3. Player B wins as the square labeled 2 is left unmarked but cannot
be covered with an L-figure.
If player A in his next move marks the square labeled 2, then player B marks the square
labeled 5. Player B wins as the square labeled 3 is left unmarked but cannot be covered with
an L-figure.
Finally, if player A in his next move marks one of the squares labeled 3 or 4, player B marks
the other of these two squares. Player B wins as the square labeled 2 is left unmarked but
cannot be covered with an L-figure.
Since we have covered all possible cases, player B wins when k  4 .
 JBMO 2016
Problems and solutions

Problem 1. A trapezoid ABCD (AB k CD, AB > CD) is circumscribed. The incircle of the
triangle ABC touches the lines AB and AC at the points M and N , respectively. Prove that
the incenter of the trapezoid ABCD lies on the line M N .

D C

R I

A M B

Solution.
Version 1. Let I be the incenter of triangle ABC and R be the common point of the lines BI
and M N . Since
1 [ = 90◦ + 1 m(M
m(AN
\ M ) = 90◦ − m(M
\ AN ) and m(BIC) \ AN )
2 2
the quadrilateral IRNC is cyclic. (1)

[ = 90◦ and therefore

It follows that m(BRC)

[ = 90◦ − 1 180◦ − m(BCD)

[ = 90◦ − m(CBR) \ = 1 m(BCD).


m(BCR) \ (2)
2 2

So, (CR is the angle bisector of DCB

\ and R is the incenter of the trapezoid. (3)

Version 2. If R is the incentre of the trapezoid ABCD, then B, I and R are collinear, (1’)
[ = 90◦ .
and m(BRC) (2’)
The quadrilateral IRNC is cyclic. (3’)
Then m(M\ N C) = 90◦ + 12 · m(BAC)
[ (4’)
and m(RN
\ [ = 90◦ + 1 · m(BAC),
C) = m(BIC) [ (5’)
2
so that m(M
\ N C) = m(RN
\ C) and the points M , R and N are collinear. (6’)

Version 3. If R is the incentre of the trapezoid ABCD, let M 0 ∈ (AB) and N 0 ∈ (AC) be the
unique points, such that R ∈ M 0 N 0 and (AM 0 ) ≡ (AN 0 ). (1”)
Let S be the intersection point of CR and AB. Then CR = RS. (2”)
0 0 0
Consider K ∈ AC such that SK k M N . Then N is the midpoint of (CK). (3”)
We deduce
AK + AC AS + AC AB − BS + AC AB + AC − BC
AN 0 = = = = = AN. (400 )
2 2 2 2
We conclude that N = N 0 , hence M = M 0 , and R, M, N are collinear. (5”)

Problem 2. Let a, b and c be positive real numbers. Prove that
8 8 8 8 8 8
+ + + a2 + b 2 + c 2 ≥ + + .
(a + b)2 + 4abc (b + c)2 + 4abc (c + a)2 + 4abc a+3 b+3 c+3

Solution. Since 2ab ≤ a2 + b2 , it follows that (a + b)2 ≤ 2(a2 + b2 ) (1)

and 4abc ≤ 2c(a2 + b2 ), for any positive reals a, b, c. (2)
Adding these inequalities, we find

(a + b)2 + 4abc ≤ 2(a2 + b2 )(c + 1), (3)

so that
8 4
≥ . (4)
(a + b)2 + 4abc (a2 + b2 )(c + 1)
Using the AM-GM inequality, we have
r
4 a2 + b 2 2 4
2 2
+ ≥2 =p , (5)
(a + b )(c + 1) 2 c+1 2(c + 1)

respectively p
c+3 (c + 1) + 2 2(c + 1)
= ≥ . (6)
8 8 4
We conclude that
4 a2 + b 2 8
2 2
+ ≥ , (7)
(a + b )(c + 1) 2 c+3

and finally
8 8 8 8 8 8
+ + +a2 +b2 +c2 ≥ + + . (8)
(a + b)2 + 4abc (a + c)2 + 4abc (b + c)2 + 4abc a+3 b+3 c+3


Problem 3. Find all the triples of integers (a, b, c) such that the number

(a − b)(b − c)(c − a)
N= +2
2
is a power of 2016.
(A power of 2016 is an integer of the form 2016n , where n is a non-negative integer.)

Solution. Let a, b, c be integers and n be a positive integer such that

(a − b)(b − c)(c − a) + 4 = 2 · 2016n .

We set a − b = −x, b − c = −y and we rewrite the equation as

xy(x + y) + 4 = 2 · 2016n . (1)

If n > 0, then the right hand side is divisible by 7, so we have that

xy(x + y) + 4 ≡ 0 (mod 7) (2)

or
3xy(x + y) ≡ 2 (mod 7) (3)
or
(x + y)3 − x3 − y 3 ≡ 2 (mod 7). (4) (1)
Note that, by Fermat’s Little Theorem, for any integer k the cubic residues are k 3 ≡ −1, 0, 1
(mod 7). (5)
3 3 3
It follows that in (1) some of (x + y) , x and y should be divisible by 7.
But in this case, xy(x + y) is divisible by 7 and this is a contradiction. (6)

So, the only possibility is to have n = 0 and consequently, xy(x + y) + 4 = 2, or, equivalently,
xy(x + y) = −2. (7)
The solutions for this are (x, y) ∈ {(−1, −1), (2, −1), (−1, 2)}, (8)
so the required triples are (a, b, c) = (k +2, k +1, k), k ∈ Z, and all their cyclic permutations.(9)

Alternative version: If n > 0 then 9 divides (a − b)(b − c)(c − a) + 4, that is, the equation
xy(x + y) + 4 ≡ 0 (mod 9)) has the solution x = b − a, y = c − b. (1’)
But then x and y have to be 1 modulo 3, implying xy(x + y) ≡ 2 (mod 9), which is a contra-
diction. (2’)
We can continue now as in the first version.

Problem 4. A 5 × 5 table is called regular if each of its cells contains one of four pairwise
distinct real numbers, such that each of them occurs exactly once in every 2 × 2 subtable.
The sum of all numbers of a regular table is called the total sum of the table. With any four
numbers, one constructs all possible regular tables, computes their total sums and counts the
distinct outcomes. Determine the maximum possible count.

Solution. We will prove that the maximum number of total sums is 60.
The proof is based on the following claim.
Claim. In a regular table either each row contains exactly two of the numbers, or each column
contains exactly two of the numbers.
Proof of the Claim. Indeed, let R be a row containing at least three of the numbers. Then,
in row R we can find three of the numbers in consecutive positions, let x, y, z be the numbers
in consecutive positions(where {x, y, s, z} = {a, b, c, d}). Due to our hypothesis that in every
2 × 2 subarray each number is used exactly once, in the row above R(if there is such a row),
precisely above the numbers x, y, z will be the numbers z, t, x in this order. And above them
will be the numbers x, y, z in this order. The same happens in the rows below R (see at the
following figure).  
• x y z •
• z t x • 
 
• x y z • 
 
• z t x • 
 

• x y z •
Completing all the array, it easily follows that each column contains exactly two of the numbers
and our claim is proven. (1)
Rotating the matrix (if it is necessary), we may assume that each row contains exactly two of
the numbers. If we forget the first row and column from the array, we obtain a 4 × 4 array,
that can be divided into four 2 × 2 subarrays, containing thus each number exactly four times,
with a total sum of 4(a + b + c + d).
It suffices to find how many different ways are there to put the numbers in the first row R1 and
the first column C1 . (2)
Denoting by a1 , b1 , c1 , d1 the number of appearances of a, b, c, and respectively d in R1 and
C1 , the total sum of the numbers in the entire 5 × 5 array will be

S = 4(a + b + c + d) + a1 · a + b1 · b + c1 · c + d1 · d. (3)

If the first, the third and the fifth row contain the numbers x, y, with x denoting the number
at the entry (1, 1), then the second and the fourth row will contain only the numbers z, t, with
z denoting the number at the entry (2, 1). Then x1 + y1 = 7 and x1 > 3, y1 > 2, z1 + t1 = 2,
and z1 > t1 . Then {x1 , y1 } = {5, 2} or {x1 , y1 } = {4, 3}, respectively {z1 , t1 } = {2, 0} or
{z1 , t1 } = {1, 1}. (4)
Then (a1 , b1 , c1 , d1 ) is obtained by permuting one of the following quadruples:

(5, 2, 2, 0), (5, 2, 1, 1), (4, 3, 2, 0), (4, 3, 1, 1). (5)

There are a total of 4! 2!

= 12 permutations of (5, 2, 2, 0), also 12 permutations of (5, 2, 1, 1), 24
permutations of (4, 3, 2, 0) and finally, there are 12 permutations of (4, 3, 1, 1). Hence, there are
at most 60 different possible total sums. (6)
We can obtain indeed each of these 60 combinations: take three rows ababa alternating with
two rows cdcdc to get (5, 2, 2, 0); take three rows ababa alternating with one row cdcdc and
a row (dcdcd) to get (5, 2, 1, 1); take three rows ababc alternating with two rows cdcda to get
(4, 3, 2, 0); take three rows abcda alternating with two rows cdabc to get (4, 3, 1, 1). (7)

By choosing for example a = 103 , b = 102 , c = 10, d = 1, we can make all these sums different.
(8)
Hence, 60 is indeed the maximum possible number of different sums. (9)

Alternative Version: Consider a regular table containing the four distinct numbers a, b, c, d.
The four 2×2 corners contain each all the four numbers, so that, if a1 , b1 , c1 , d1 are the numbers
of appearances of a, b, c, and respectively d in the middle row and column, then

S = 4(a + b + c + d) + a1 · a + b1 · b + c1 · c + d1 · d. (10 )

Consider the numbers x in position (3, 3), y in position (3, 2), y 0 in position (3, 4), z in position
(2, 3) and z 0 in position (4, 3).
If z 6= z 0 = t, then y = y 0 , and in positions (3, 1) and (3, 5) there will be the number x. (2’)
The second and fourth row can only contain now the numbers z and t, respectively the first
and fifth row only x and y. (3’).
Then x1 + y1 = 7 and x1 > 3, y1 > 2, z1 + t1 = 2, and z1 > t1 . Then {x1 , y1 } = {5, 2} or
{x1 , y1 } = {4, 3}, respectively {z1 , t1 } = {2, 0} or {z1 , t1 } = {1, 1}. (4’)
One can continue now as in the first version.

 21th Junior Balkan Mathemati al Olympiad
Monday, June 26, 2017

Problem 1. Determine all the sets of six onse utive positive integers su h that the produ t of some
two of them, added to the produ t of some other two of them is equal to the produ t of the remaining
two numbers.

Solution. Exa tly two of the six numbers are multiples of 3 and these two need to be multiplied
together, otherwise two of the three terms of the equality are multiples of 3 but the third one is not.
Let n and n + 3 denote these multiples of 3. Two of the four remaining numbers give remainder 1
when divided by 3, while the other two give remainder 2, so the two other produ ts are either ≡ 1·1 = 1
(mod 3) and ≡ 2·2 ≡ 1 (mod 3), or they are both ≡ 1·2 ≡ 2 (mod 3). In on lusion, the term n(n+3)
needs to be on the right hand side of the equality.
Looking at parity, three of the numbers are odd, and three are even. One of n and n + 3 is odd,
the other even, so exa tly two of the other numbers are odd. As n(n + 3) is even, the two remaining
odd numbers need to appear in dierent terms.
We distinguish the following ases:
I. The numbers are n − 2, n − 1, n, n + 1, n + 2, n + 3.
The produ t of the two numbers on the RHS needs to be larger than n(n + 3). The only possibility is
(n − 2)(n − 1) + n(n + 3) = (n + 1)(n + 2) whi h leads to n = 3. Indeed, 1 · 2 + 3 · 6 = 4 · 5.
II. The numbers are n − 1, n, n + 1, n + 2, n + 3, n + 4.
As (n + 4)(n − 1) + n(n + 3) = (n + 1)(n + 2) has no solutions, n + 4 needs to be on the RHS, multiplied
with a number having a dierent parity, so n − 1 or n + 1.
(n + 2)(n − 1) + n(n + 3) = (n + 1)(n + 4) leads to n = 3. Indeed, 2 · 5 + 3 · 6 = 4 · 7.
(n + 2)(n + 1) + n(n + 3) = (n − 1)(n + 4) has no solution.
III. The numbers are n, n + 1, n + 2, n + 3, n + 4, n + 5.
We need to onsider the following situations:
(n + 1)(n + 2) + n(n + 3) = (n + 4)(n + 5) whi h leads to n = 6; indeed 7 · 8 + 6 · 9 = 10 · 11;
(n + 2)(n + 5) + n(n + 3) = (n + 1)(n + 4) obviously without solutions, and
(n + 1)(n + 4) + n(n + 3) = (n + 2)(n + 5) whi h leads to n = 2 (not a multiple of 3).
In on lusion, the problem has three solutions:
1 · 2 + 3 · 6 = 4 · 5, 2 · 5 + 3 · 6 = 4 · 7, and 7 · 8 + 6 · 9 = 10 · 11.

Problem 2. Let x, y, z be positive integers su h that x 6= y 6= z 6= x. Prove that

(x + y + z)(xy + yz + zx − 2) ≥ 9xyz.
When does the equality hold?
Solution. Sin e x, y, z are distin t positive integers, the required inequality is symmetri and
WLOG we an suppose that x ≥ y + 1 ≥ z + 2. We onsider 2 possible ases:
Case 1. y ≥ z + 2. Sin e x ≥ y + 1 ≥ z + 3 it follows that
(x − y)2 ≥ 1, (y − z)2 ≥ 4, (x − z)2 ≥ 9
whi h are equivalent to
x2 + y 2 ≥ 2xy + 1, y 2 + z 2 ≥ 2yz + 4, x2 + z 2 ≥ 2xz + 9
or otherwise
zx2 + zy 2 ≥ 2xyz + z, xy 2 + xz 2 ≥ 2xyz + 4x, yx2 + yz 2 ≥ 2xyz + 9y.
Adding up the last three inequalities we have
xy(x + y) + yz(y + z) + zx(z + x) ≥ 6xyz + 4x + 9y + z
whi h implies that (x + y + z)(xy + yz + zx − 2) ≥ 9xyz + 2x + 7y − z.
Sin e x ≥ z + 3 it follows that 2x + 7y − z ≥ 0 and our inequality follows.
Case 2. y = z + 1. Sin e x ≥ y + 1 = z + 2 it follows that x ≥ z + 2, and repla ing y = z + 1 in
the required inequality we have to prove
 (x + z + 1 + z)(x(z + 1) + (z + 1)z + zx − 2) ≥ 9x(z + 1)z
whi h is equivalent to
(x + 2z + 1)(z 2 + 2zx + z + x − 2) − 9x(z + 1)z ≥ 0
Doing easy algebrai manipulations, this is equivalent to prove
(x − z − 2)(x − z + 1)(2z + 1) ≥ 0
whi h is satised sin e x ≥ z + 2.
The equality is a hieved only in the Case 2 for x = z + 2, so we have equality when (x, y, z) =
(k + 2, k + 1, k) and all the permutations for any positive integer k.

Problem 3. Let ABC be an a ute triangle su h that AB 6= AC , with ir um ir le Γ and ir um enter

O. Let M be the midpoint of BC and D be a point on Γ su h that AD ⊥ BC . Let T be a point su h
that BDCT is a parallelogram and Q a point on the same side of BC as A su h that
∠BQM = ∠BCA and ∠CQM = ∠CBA.

Let the line AO interse t Γ at E , (E 6= A) and let the ir um ir le of △ET Q interse t Γ at point
X 6= E . Prove that the points A, M , and X are ollinear.
Let X ′ be symmetri point to Q in line BC . Now sin e ∠CBA = ∠CQM = ∠CX ′ M ,
Solution.
∠BCA = ∠BQM = ∠BX ′ M , we have

∠BX ′ C = ∠BX ′ M + ∠CX ′ M = ∠CBA + ∠BCA = 180◦ − ∠BAC

we have that X ′ ∈ Γ. Now sin e ∠AX ′ B = ∠ACB = ∠M X ′ B we have that A, M, X ′ are ollinear.
Note that sin e
∠DCB = ∠DAB = 90◦ − ∠ABC = ∠OAC = ∠EAC
we get that DBCE is an isos eles trapezoid.

A b

O
b Q
b

b T

M
b b b
C
B

b b

D E
X′

Sin e BDCT is a parallelogram we have M T = M D, with M, D, T being ollinear, BD = CT ,

and sin e BDEC is an isos eles trapezoid we have BD = CE and M E = M D. Sin e
∠BT C = ∠BDC = ∠BED, CE = BD = CT and M E = M T
we have that E and T are symmetri with respe t to the line BC . Now sin e Q and X ′ are symmetri
with respe t to the line BC as well, this means that QX ′ ET is an isos eles trapezoid whi h means that
Q, X ′ , E, T are on y li . Sin e X ′ ∈ Γ this means that X ≡ X ′ and therefore A, M, X are ollinear.
Alternative solution. Denote by H the ortho enter of △ABC . We use the following well known
properties:
 (i) Point D is the symmetri point of H with respe t to BC . Indeed, if H1 is the symmetri point
of H with respe t to BC then ∠BH1 C + ∠BAC = 180◦ and therefore H1 ≡ D.
(ii) The symmetri point of H with respe t to M is the point E . Indeed, if H2 is the symmetri
point of H with respe t to M then BH2CH is parallelogram, ∠BH2C + ∠BAC = 180◦ and sin e
EB k CH we have ∠EBA = 90◦ .
Sin e DET H is a parallelogram and M H = M D we have that DET H is a re tangle. Therefore
M T = M E and T E ⊥ BC implying that T and E are symmetri with respe t to BC . Denote by Q′
the symmetri point of Q with respe t to BC . Then Q′ ET Q is isos eles trapezoid, so Q′ is a point
on the ir um ir le of △ET Q. Moreover ∠BQ′C + ∠BAC = 180◦ and we on lude that Q′ ∈ Γ.
Therefore Q′ ≡ X .
It remains to observe that ∠CXM = ∠CQM = ∠CBA and ∠CXA = ∠CBA and we infer that
X , M and A are ollinear.

Problem 4. Consider a regular 2n-gon P , A1 A2 . . . A2n in the plane, where n is a positive integer.
We say that a point S on one of the sides of P an be seen from a point E that is external to P , if
the line segment SE ontains no other points that lie on the sides of P ex ept S . We olor the sides
of P in 3 dierent olors (ignore the verti es of P , we onsider them olorless), su h that every side
is olored in exa tly one olor, and ea h olor is used at least on e. Moreover, from every point in
the plane external to P , points of at most 2 dierent olors on P an be seen. Find the number of
distin t su h olorings of P (two olorings are onsidered distin t if at least one of the sides is olored
dierently).

Solution Answer: For n = 2, the answer is 36; for n = 3, the answer is 30 and for n ≥ 4, the
answer is 6n.
Lemma 1. Given a regular 2n-gon in the plane and a sequen e of n onse utive sides s1 , s2 , . . . , sn
there is an external point Q in the plane, su h that the olor of ea h si an be seen from Q, for
i = 1, 2, . . . , n.
Proof. It is obvious that for a semi- ir le S , there is a point R in the plane far enough on the
bise tor of its diameter su h that almost the entire semi- ir le an be seen from R.
Now, it is lear that looking at the ir ums ribed ir le around the 2n-gon, there is a semi- ir le S
su h that ea h si either has both endpoints on it, or has an endpoint that's on the semi- ir le, and is
not on the semi- ir le's end. So, take Q to be a point in the plane from whi h almost all of S an be
seen, learly, the olor of ea h si an be seen from Q.
Lemma 2. Given a regular 2n-gon in the plane, and a sequen e of n + 1 onse utive sides
s1 , s2 , . . . , sn+1 there is no external point Q in the plane, su h that the olor of ea h si an be seen
from Q, for i = 1, 2, . . . , n + 1.
Proof. Sin e s1 and sn+1 are parallel opposite sides of the 2n-gon, they annot be seen at the
same time from an external point.
For n = 2, we have a square, so all we have to do is make sure ea h olor is used. Two sides will
be of the same olor, and we have  to hoose whi h are these 2 sides, and then assign olors a ording
4
to this hoi e, so the answer is .3.2 = 36.
2
For n = 3, we have a hexagon. Denote the sides as a1 , a2 , . . . q6 , in that order. There must be 2
onse utive sides of dierent olors, say a1 is red, a2 is blue. We must have a green side, and only a4
and a5 an be green. We have 3 possibilities:
1) a4 is green, a5 is not. So, a3 must be blue and a5 must be blue (by elimination) and a6 must be
blue, so we get a valid oloring.
2) Both a4 and a5 are green, thus a6 must be red and a5 must be blue, and we get the oloring
rbbggr .
3) a5 is green, a4 is not. Then a6 must be red. Subsequently, a4 must be red (we assume it is not
green). It remains that a3 must be red, and the oloring is rbrrgr.
Thus, we have 2 kinds of ongurations:
i) 2 opposite sides have 2 opposite olors and all other sides are of the third olor. This an happen
in 3.(3.2.1) = 18 ways (rst hoosing the pair of opposite sides, then assigning olors),
ii) 3 pairs of onse utive sides, ea h pair in one of the 3 olors. This an happen in 2.6 = 12 ways
(2 partitioning into pairs of onse utive sides, for ea h partitioning, 6 ways to assign the olors).
 Thus, for n = 3, the answer is 18 + 12 = 30.
Finally, let's address the ase n ≥ 4. The important thing now is that any 4 onse utive sides an
be seen from an external point, by Lemma 1.
Denote the sides as a1 , a2 , . . . , a2n . Again, there must be 2 adja ent sides that are of dierent
olors, say a1 is blue and a2 is red. We must have a green side, and by Lemma 1, that an only be
an+1 or an+2 . So, we have 2 ases:
Case 1: an+1 is green, so an must be red ( annot be green due to Lemma 1 applied to a1 , a2 , . . . , an ,
annot be blue for the sake of a2 , . . . , an+1 . If an+2 is red, so are an+3 , . . . , a2n , and we get a valid
oloring: a1 is blue, an+1 is green, and all the others are red.
If an+2 is green:
a) an+3 annot be green, be ause of a2 , a1 , a2n . . . , an+3 .
b) an+3 annot be blue, be ause the 4 adja ent sides an , . . . , an+3 an be seen (this is the ase that
makes the separate treatment of n ≥ 4 ne essary)
) an+3 annot be red, be ause of a1 , a2n , . . . , an+2 .
So, in the ase that an+2 is also green, we annot get a valid oloring.
Case 2: an+2 is green is treated the same way as Case 1.
This means that the only valid onguration for n ≥ 4 is having 2 opposite sides olored in 2
dierent olors, and all other sides olored in the third olor. This an be done in n.3.2 = 6n ways.
  

22nd Junior Balkan Mathematical Olympiad

Rhodes 19-24 June 2018
Solutions
Problem 2. Let n three-digit numbers satisfy the following properties:

(1) No number contains the digit 0.

(2) The sum of the digits of each number is 9.

(3) The units digits of any two numbers are different.

(4) The tens digits of any two numbers are different.

(5) The hundreds digits of any two numbers are different.

Find the largest possible value of n.

Solution. Let S denote the set of three-digit numbers that have digit sum equal to 9 and no digit
equal to 0. We will first find the cardinality of S. We start from the number 111 and each element
of S can be obtained from 111 by a string of 6 A’s (which means that we add 1 to the current digit)
and 2 G’s (which means go to the next digit). Then for example 324 can be obtained from 111 by the
string AAGAGAAA. There are in total
8!
= 28
6! · 2!
such words, so S contains 28 numbers. Now, from the conditions (3), (4), (5), if abc is in T then each
of the other numbers of the form ∗ ∗ c cannot be in T , neither ∗b∗ can be, nor a ∗ ∗. Since there are
a + b − 2 numbers of the first category, a + c − 2 from the second and b + c − 2 from the third one. In
these three categories there are

(a + b − 2) + (b + c − 2) + (c + a − 2) = 2(a + b + c) − 6 = 2 · 9 − 6 = 12

distinct numbers that cannot be in T if abc is in T . So, if T has n numbers, then 12n are the forbidden
ones that are in S, but each number from S can be a forbidden number no more than three times,
once for each of its digits, so
12n 28
n+ ≤ 28 ⇐⇒ n ≤ ,
3 5
and since n is an integer, we get n ≤ 5. A possible example for n = 5 is

T = {144, 252, 315, 423, 531}.

Comment by PSC. It is classical to compute the cardinality of S and this can be done in many
ways. In general, the number of solutions of the equation

x1 + x2 + · · · + xk = n

in positive integers, where the order of xi matters, is well known that equals to n−1


k−1 . In our case,
we want to count the number of positive solutions to a + b + c = 9. By the above, this equals to
9−1


3−1 = 28. Using the general result above, we can also find that there are a + b − 2 numbers of the
form ∗ ∗ c.
Problem 3. Let k > 1 be a positive integer and n > 2018 be an odd positive integer. The nonzero
rational numbers x1 , x2 , . . . , xn are not all equal and satisfy
k k k k k
x1 + = x2 + = x3 + = · · · = xn−1 + = xn + ·
x2 x3 x4 xn x1
Find:

a) the product x1 x2 . . . xn as a function of k and n

b) the least value of k, such that there exist n, x1 , x2 , . . . , xn satisfying the given conditions.

a) If xi = xi+1 for some i (assuming xn+1 = x1 ), then by the given identity all xi will be equal, a
contradiction. Thus x1 6= x2 and
x2 − x3
x1 − x2 = k .
x2 x3
Analogously
x2 − x3 x3 − x4 x1 − x2
x1 − x2 = k = k2 = · · · = kn ·
x2 x3 (x2 x3 ) (x3 x4 ) (x2 x3 ) (x3 x4 ) . . . (x1 x2 )

Since x1 6= x2 we get √ n−1 √

x1 x2 . . . xn = ± k n = ±k 2 k.
If one among these two values, positive or negative, is obtained, then the other one will be also
obtained by changing the sign of all xi since n is odd.
b) From the above result, as n is odd, we conclude that k is a perfect square, so k ≥ 4. For k = 4
let n = 2019 and x3j = 4, x3j−1 = 1, x3j−2 = −2 for j = 1, 2, . . . , 673. So the required least value is
k = 4.

Comment by PSC. There are many ways to construct the example when k = 4 and n = 2019. Since
3 | 2019, the idea is to find three numbers x1 , x2 , x3 satisfying the given equations, not all equal, and
repeat them as values for the rest of the xi ’s. So, we want to find x1 , x2 , x3 such that
4 4 4
x1 + = x2 + = x3 + ·
x2 x3 x1
As above, x1 x2 x3 = ±8. Suppose without loss of generality that x1 x2 x3 = −8. Then, solving the
above system we see that if x1 6= 2, then
4 4
x2 = − and x3 = 2 − ,
x1 − 2 x1
leading to infinitely many solutions. The example in the official solution is obtained by choosing
x1 = −2.
Problem 4. Let ABC be an acute triangle, A0 , B 0 and C 0 be the reflections of the vertices A, B and
C with respect to BC, CA, and AB, respectively, and let the circumcircles of triangles ABB 0 and
ACC 0 meet again at A1 . Points B1 and C1 are defined similarly. Prove that the lines AA1 , BB1 and
CC1 have a common point.

Solution. Let O1 , O2 and O be the circumcenters of triangles ABB 0 , ACC 0 and ABC respectively. As
AB is the perpendicular bisector of the line segment CC 0 , O2 is the intersection of the perpendicular
bisector of AC with AB. Similarly, O1 is the intersection of the perpendicular bisector of AB with
AC. It follows that O is the orthocenter of triangle AO1 O2 . This means that AO is perpendicular
to O1 O2 . On the other hand, the segment AA1 is the common chord of the two circles, thus it is
perpendicular to O1 O2 . As a result, AA1 passes through O. Similarly, BB1 and CC1 pass through
O, so the three lines are concurrent at O.

B0
C0
A

O
B C

O1

O2

A1

Comment by PSC. We present here a different approach.

We first prove that A1 , B and C 0 are collinear. Indeed, since ∠BAB 0 = ∠CAC 0 = 2∠BAC, then from
the circles (ABB 0 ), (ACC 0 ) we get

∠BA1 B 0 180◦ − ∠BAB 0

∠AA1 B = = = 90◦ − ∠BAC = ∠AA1 C 0 .
2 2
It follows that
∠A1 AC = ∠A1 C 0 C = ∠BC 0 C = 90◦ − ∠ABC (1)
On the other hand, if O is the circumcenter of ABC, then

∠OAC = 90◦ − ∠ABC. (2)

From (1) and (2) we conclude that A1 , A and O are collinear. Similarly, BB1 and CC1 pass through
O, so the three lines are concurrent in O.
 Problems and Solutions

Problem 1. Find all prime numbers p for which there exist positive integers x, y and z such that
the number
xp + y p + z p − x − y − z
is a product of exactly three distinct prime numbers.
Solution. Let A = xp + y p + z p − x − y − z. For p = 2, we take x = y = 4 and z = 3. Then
A = 30 = 2 · 3 · 5. For p = 3 we can take x = 3 and y = 2 and z = 1. Then again A = 30 = 2 · 3 · 5.
For p = 5 we can take x = 2 and y = 1 and z = 1. Again A = 30 = 2 · 3 · 5.
Assume now that p 󰃍 7. Working modulo 2 and modulo 3 we see that A is divisible by both 2 and
3. Moreover, by Fermat’s Little Theorem, we have

xp + y p + z p − x − y − z ≡ x + y + z − x − y − z = 0 mod p .

Therefore, by the given condition, we have to solve the equation

xp + y p + z p − x − y − z = 6p .

If one of the numbers x, y and z is bigger than or equal to 2, let’s say x 󰃍 2, then

6p 󰃍 xp − x = x(xp−1 − 1) 󰃍 2(2p−1 − 1) = 2p − 2 .

It is easy to check by induction that 2n − 2 > 6n for all natural numbers n 󰃍 6. This contradiction
shows that there are no more values of p which satisfy the required property.
Remark. There are a couple of other ways to prove that 2p − 2 > 6p for p 󰃍 7. For example, we
can use the Binomial Theorem as follows:
p(p − 1) p(p − 1)(p − 2)
2p − 2 󰃍 1 + p + + − 2 󰃍 1 + p + 3p + 5p − 2 > 6p .
2 6
We can also use Bernoulli’s Inequality as follows:

2p − 2 = 8(1 + 1)p−3 − 2 󰃍 8(1 + (p − 3)) − 2 = 8p − 18 > 6p

The last inequality is true for p 󰃍 11. For p = 7 we can see directly that 2p − 2 > 6p.
One can also use calculus to show that f (x) = 2x − 6x is increasing for x 󰃍 5.
Problem 2. Let a, b be two distinct real numbers and let c be a positive real number such that

a4 − 2019a = b4 − 2019b = c .
√
Prove that − c < ab < 0.
Solution. Firstly, we see that

2019(a − b) = a4 − b4 = (a − b)(a + b)(a2 + b2 ) .

Since a ∕= b, we get (a + b)(a2 + b2 ) = 2019, so a + b ∕= 0. Thus

2c = a4 − 2019a + b4 − 2019b
= a4 + b4 − 2019(a + b)
= a4 + b4 − (a + b)2 (a2 + b2 )
= −2ab(a2 + ab + b2 ) .

Hence ab(a2 + ab + b2 ) = −c < 0. Note that

1Ä 2 ä
a2 + ab + b2 = a + b2 + (a + b)2 > 0 ,
2
thus ab < 0. Finally, a2 + ab + b2 = (a + b)2 − ab > −ab (the equality does not occur since a + b ∕= 0).
So √ √
−c = ab(a2 + ab + b2 ) < −(ab)2 =⇒ (ab)2 < c ⇒ − c < ab < c .
√
Therefore, we have − c < ab < 0.
Remark. We can get c = −ab(a2 + ab + b2 ) in several other ways. For example using that,

(a − b)c = a(b4 − 2019b) − b(a4 − 2019a) = ab(b3 − a3 ) = ab(b − a)(a2 + ab + b2 ) .

We can also divide f (x) = x4 − 2019x − c by (x − a)(x − b) and look at the constant term of the
remainder.
Alternative Solution. By Descartes’ Rule of Signs, the polynomial p(x) = x4 − 2019x − c has
exactly one positive root and exactly one negative root. So a, b must be its two real roots. Since
one of them is positive and the other is negative, then ab < 0. Let r ± is be the two non-real roots
of p(x).
By Vieta, we have

ab(r2 + s2 ) = −c , (1)
a + b + 2r = 0 , (2)
2 2
ab + 2ar + 2br + r + s = 0 . (3)

Using (2) and (3), we have

r2 + s2 = −2r(a + b) − ab = (a + b)2 − ab 󰃍 −ab . (4)

If in the last inequality we actually have an equality, then a + b = 0. Then (2) gives r = 0 and (3)
gives s2 = −ab. Thus the roots of p(x) are a, −a, ia, −ia. This would give that p(x) = x4 + a4 , a
contradiction.
So the inequality in (4) is strict and now from (1) we get

c = −(r2 + s2 )ab > (ab)2 ,

√
which gives that ab > − c.
Remark. One can get that x4 − 2019x − c has only two real roots by showing (e.g. by calculus)
that it is initially decreasing and then increasing.
Also, instead of Vieta one can also proceed by factorising the polynomial as:

x4 − 2019x − c = (x2 − (a + b)x + ab)(x2 + (a + b)x − c

ab ) .

Since the second quadratic has no real roots, its discriminant is negative which gives that c > (ab)2 .
Problem 3. Triangle ABC is such that AB < AC. The perpendicular bisector of side BC
intersects lines AB and AC at points P and Q, respectively. Let H be the orthocentre of triangle
ABC, and let M and N be the midpoints of segments BC and P Q, respectively. Prove that lines
HM and AN meet on the circumcircle of ABC.
Solution. We have
∠AP Q = ∠BP M = 90◦ − ∠M BP = 90◦ − ∠CBA = ∠HCB ,
and
∠AQP = ∠M QC = 90◦ − ∠QCM = 90◦ − ∠ACB = ∠CBH .
From these two equalities, we see that the triangles AP Q and HCB are similar. Moreover, since
M and N are the midpoints of the segments BC and P Q respectively, then the triangles AQN and
HBM are also similar. Therefore, we have ∠AN Q = ∠HM B.
P

Q
L

B M C

Let L be the intersection of AN and HM . We have

∠M LN = 180◦ − ∠LN M − ∠N M L = 180◦ − ∠LM B − ∠N M L = 180◦ − ∠N M B = 90◦ .
Now let D be the point on the circumcircle of ABC diametrically oposite to A. It is known that D
is also the relfection of point H over the point M . Therefore, we have that D belongs on M H and
that ∠DLA = ∠M LA = ∠M LN = 90◦ . But, as DA is the diameter of the circumcirle of ABC,
the condition that ∠DLA = 90◦ is enough to conclude that L belongs on the circumcircle of ABC.
Remark. There is a spiral similarity mapping AQP to HBC. Since the similarity maps AN to
HM , it also maps AH to N M , and since these two lines are parallel, the centre of the similarity
is L = AN ∩ HM . Since the similarity maps BC to QP , its centre belongs on the circumcircle of
BCX, where X = BQ ∩ P C. But X is the reflection of A on QM and so it must belong on the
circumcircle of ABC. Hence so must L.
Remark. Students have also submitted correct proofs using radical axes, harmonic quadruples,
coordinate geometry and complex numbers.
Problem 4. A 5 × 100 table is divided into 500 unit square cells, where n of them are coloured
black and the rest are coloured white. Two unit square cells are called adjacent if they share a
common side. Each of the unit square cells has at most two adjacent black unit square cells. Find
the largest possible value of n.
Solution. If we colour all the cells along all edges of the board together with the entire middle
row except the second and the last-but-one cell, the condition is satisfied and there are 302 black
cells. The figure below exhibits this colouring for the 5 × 8 case.

We can cover the table by one fragment like the first one on the figure below, 24 fragments like the
middle one, and one fragment like the third one.

a b a b h i h i
c a b f g h i m
c f g f g m
c d e f g j k m
d e d e j k j k

In each fragment, among the cells with the same letter, there are at most two coloured black, so
the total number of coloured cells is at most (5 + 24 · 6 + 1) · 2 + 2 = 302.
Alternative Solution. Consider the cells adjacent to all cells of the second and fourth row.
Counting multiplicity, each cell in the first and fifth row is counted once, each cell in the third row
twice, while each cell in the second and fourth row is also counted twice apart from their first and
last cells which are counted only once.
So there are 204 cells counted once and 296 cells counted twice. Those cells contain, counting
multiplicity, at most 400 black cells. Suppose a of the cells have multiplicity one and b of them
have multiplicity 2. Then a + 2b 󰃑 400 and a 󰃑 204. Thus

2a + 2b 󰃑 400 + a 󰃑 604 ,

and so a + b 󰃑 302 as required.

Remark. The alternative solution shows that if we have equality, then all cells in the perimeter
of the table except perhaps the two cells of the third row must be coloured black. No other cell
in the second or fourth row can be coloured black as this will give a cell in the first or fifth row
with at least three neighbouring black cells. For similar reasons we cannot colour black the second
and last-but-one cell of the third row. So we must colour black all other cells of the third row and
therefore the colouring is unique.
Alternative Solution. Suppose we have a black corner cells, b black side cells, and c black interior
cells. Let N be the number of pairs of cells (c1 , c2 ) such that c1 is black and c2 is neighbour of c1 .
Then N = 2a + 3b + 4c and N 󰃑 1000. Since also a 󰃑 4 and b 󰃑 202 we get

1000 󰃍 4(a + b + c) − 2a − b 󰃍 4(a + b + c) − 210 ,

giving a + b + c 󰃑 302.5.