Baltic Way 2016 – Solutions

1. Find all pairs of primes (p, q) such that

p3 − q 5 = (p + q)2 .

Solution. Assume first that neither of the numbers equals 3. Then, if p ≡ q mod 3, the
left hand side is divisible by 3, but the right hand side is not. But if p ≡ −q mod 3, the
left hand side is not divisible by 3, while the right hand side is. So this is not possible.
If p = 3, then q 5 < 27, which is impossible. Therefore q = 3, and the equation turns into
p3 − 243 = (p + 3)2 or
p(p2 − p − 6) = 252 = 7 · 36.
As p > 3 then p2 − p − 6 is positive and increases with p. So the equation has at most
one solution. It is easy to see that p = 7 is the one and (7, 3) is a solution to the given
equation.

2. Prove or disprove the following hypotheses.

a) For all k ≥ 2, each sequence of k consecutive positive integers contains a number
that is not divisible by any prime number less than k.
b) For all k ≥ 2, each sequence of k consecutive positive integers contains a number
that is relatively prime to all other members of the sequence.
Solution We give a counterexample to both claims. So neither of them is true.
For a), a counterexample is the sequence (2, 3, 4, 5, 6, 7, 8, 9) of eight consecutive integers
all of which are divisible by some prime less than 8.
To construct a counterexample to b), we notice that by the Chinese Remainder Theorem,
there exists an integer x such that x ≡ 0 mod 2, x ≡ 0 mod 5, x ≡ 0 mod 11, x ≡ 2 mod 3,
x ≡ 5 mod 7 and x ≡ 10 mod 13. The last three of these congruences mean that x + 16 is a
multiple of 3, 7, and 13. Now consider the sequence (x, x+1, . . . , x+16) of 17 consequtive
integers. Of these all numbers x + 2k, 0 ≤ k ≤ 8, are even and so have a common factor
with some other. Of the remaining, x + 1, x + 7 and x + 13 are divisible by 3, x + 3 is a
multiple of 13 as is x + 16, x + 5 is divisible by 5 as x, x + 9 is a multiple of 7 as x + 2,
x + 11 a multiple of 11 as is x, and finally x + 15 is a multiple of 5 as is x.
Remark . The counterexample given to either hypothesis is the shortest possible. The only
counterexamples of length 8 to the first hypothesis are those where numbers give remainders
2, 3, . . . , 9; 3, 4, . . . , 10; −2, −3, . . . , −9; or −3, −4, . . . , −10 modulo 210. The only
counterexamples of length 17 to the second hypothesis are those where the numbers give
remainders 2184, 2185, . . . , 2200 or −2184, −2185, . . . , −2200 modulo 30030.

3. For which integers n = 1, . . . , 6 does the equation

an + bn = cn + n

have a solution in integers?

 2

Solution. A solution clearly exists for n = 1, 2, 3:

11 + 01 = 01 + 1, 12 + 12 = 02 + 2, 13 + 13 = (−1)3 + 3.

We show that for n = 4, 5, 6 there is no solution.

For n = 4, the equation a4 + b4 = c4 + 4 may be considered modulo 8. Since each fourth
power x4 ≡ 0, 1 mod 8, the expression a4 + b4 − c4 can never be congruent to 4.
For n = 5, consider the equation a5 + b5 = c5 + 5 modulo 11. As x5 ≡ 0 or ≡ ±1 mod 11
(This can be seen by Fermat’s Little Theorem or by direct computation), a5 + b5 − c5
cannot be congruent to 5.
The case n = 6 is similarly dismissed by considering the equation modulo 13.

4. Let n be a positive integer and let a, b, c, d be integers such that n | a + b + c + d and

n | a2 + b2 + c2 + d2 . Show that

n | a4 + b4 + c4 + d4 + 4abcd.

Solution 1. Consider the polynomial

w(x) = (x − a)(x − b)(x − c)(x − d) = x4 + Ax3 + Bx2 + Cx + D.

It is clear that w(a) = w(b) = w(c) = w(d) = 0. By adding these values we get

w(a) + w(b) + w(c) + w(d) = a4 + b4 + c4 + d4 + A(a3 + b3 + c3 + d3 )+

+B(a2 + b2 + c2 + d2 ) + C(a + b + c + d) + 4D = 0.

Hence

a4 + b4 + c4 + d4 + 4D
= −A(a3 + b3 + c3 + d3 ) − B(a2 + b2 + c2 + d2 ) − C(a + b + c + d).

Using Vieta’s formulas, we can see that D = abcd and −A = a + b + c + d. Therefore the
right hand side of the equation above is divisible by n, and so is the left hand side.
Solution 2. Since the numbers (a + b + c + d)(a3 + b3 + c3 + d3 ), (a2 + b2 + c2 + d2 )(ab +
ac + ad + bc + bd + cd) and (a + b + c + d)(abc + acd + abd + bcd) are divisible by n, then
so is the number

(a + b + c + d)(a3 + b3 + c3 + d3 ) − (a2 + b2 + c2 + d2 )(ab + ac + ad + bc + bd + cd)+

+(a + b + c + d)(abc + acd + abd + bcd) = a4 + b4 + c4 + d4 + 4abcd.

(Heiki Niglas, Estonia)

 3

5. Let p > 3 be a prime such that p ≡ 3 (mod 4). Given a positive integer a0 , define the
n
sequence a0 , a1 , . . . of integers by an = a2n−1 for all n = 1, 2, . . .. Prove that it is possible
to choose a0 such that the subsequence aN , aN+1 , aN+2 , . . . is not constant modulo p for
any positive integer N .
Solution. Let p be a prime with residue 3 modulo 4 and p > 3. Then p − 1 = u · 2 where
u > 1 is odd. Choose a0 = 2. The order of 2 modulo p (that is, the smallest positive
integer t such that 2t ≡ 1 mod p) is a divisor of φ(p) = p − 1 = u · 2, but not a divisor of 2
since 1 < 22 < p. Hence the order of 2 modulo p is not a power of 2. By definition we see
1+2+···+n
that an = a20 . Since the order of a0 = 2 modulo p is not a power of 2, we know that
an ≡ 1 (mod p) for all n = 1, 2, 3, . . .. We proof the statement by contradiction. Assume
there exists a positive integer N such that an ≡ aN (mod p) for all n ≥ N . Let d > 1 be
n+1 n+1
the order of aN modulo p. Then aN ≡ an ≡ an+1 = a2n ≡ a2N (mod p), and hence
2n+1 −1 n+1
aN ≡ 1 (mod p) for all n ≥ N . Now d divides 2 − 1 for all n ≥ N , but this is a
contradiction since

gcd(2n+1 − 1, 2n+2 − 1) = gcd(2n+1 − 1, 2n+2 − 1 − 2(2n+1 − 1)) = gcd(2n+1 − 1, 1) = 1.

Hence there does not exist such an N .

6. The set {1, 2, . . . , 10} is partitioned into three subsets A, B and C. For each subset
the sum of its elements, the product of its elements and the sum of the digits of all its
elements are calculated. Is it possible that A alone has the largest sum of elements, B
alone has the largest product of elements, and C alone has the largest sum of digits?
Solution. It is indeed possible. Choose A = {1, 9, 10}, B = {3, 7, 8}, C = {2, 4, 5, 6}.
Then the sum of elements in A, B and C, respectively, is 20, 18 and 17, the sum of digits
11, 18 and 17, while the product of elements is 90, 168 and 240.

7. Find all positive integers n for which

3xn + n(x + 2) − 3 ≥ nx2

holds for all real numbers x.

Solution. We show that the inequality holds for even n and only for them.
If n is odd, the for x = −1 the left hand side of the inequality equals n − 6 while the right
hand side is n. So the inequality is not true for x = −1 for any odd n. So now assume
that n is even. Since |x| ≥ x, it is enough to prove 3xn + 2n − 3 ≥ nx2 + n|x| for all x or
equivalently that 3xn + (2n − 3) ≥ nx2 + nx for x ≥ 0. Now the AGM-inequality gives

1
2xn + (n − 2) = xn + xn + 1 + · · · + 1 ≥ n xn · xn · 1n−2 n = nx2 , (1)

and similarly

1
xn + (n − 1) ≥ n xn · 1n−1 n = nx. (2)
Adding (1) and (2) yields the claim.
 4

8. Find all real numbers a for which there exists a non-constant function f : R → R
satisfying the following two equations for all x ∈ R:
i) f (ax) = a2 f (x) and
ii) f (f (x)) = af (x).
Solution. The conditions of the problem give two representations for f (f (f (x))):

f (f (f (x))) = af (f (x)) = a2 f (x)

and
f (f (f (x))) = f (af (x)) = a2 f (f (x)) = a3 f (x).
So a2 f (x) = a3 f (x) for all x, and if there is an x such that f (x) = 0, the a = 0 or a = 1.
Otherwise f is the constant function f (x) = 0 for all x. If a = 1, the function f (x) = x
satisfies the conditions. For a = 0, one possible solution is the function f ,

1 for x < 0
f (x) = .
0 for x ≥ 0

9. Find all quadruples (a, b, c, d) of real numbers that simultaneously satisfy the following
equations: ⎧
⎪
⎪ a3 + c3 = 2
⎪
⎪
⎨ a2 b + c2 d = 0
⎪
⎪ b 3 + d3 = 1
⎪
⎪
⎩
ab2 + cd2 = −6

Solution 1. Consider the polynomial P (x) = (ax + b)3 + (cx + d)3 = (a3 + b3 )x3 + 3(a2 b +
c2 d)x2 + 3(ab2 + cd2 )x + b3 + d3 . By the conditions of the problem, P (x) = 2x3 − 18x + 1.
Clearly P (0) > 0, P (1) < 0 and P (3) > 0. Thus P has three distinct zeroes. But P (x) = 0
implies ax + b = −(cx + d) or (a + c)x + b + d = 0. This equation has only one solution,
unless a = −c and b = −d. But since the conditions of the problem do not allow this, we
infer that the system of equations in the problem has no solution.
Solution 2. If 0 ∈ {a, b}, then one easily gets that 0 ∈ {c, d}, which contradicts the
equation ab2 + cd2 = −6. Similarly, if 0 ∈ {c, d}, then 0 ∈ {a, b} and this contradicts
ab2 + cd2 = −6 again. Hence a, b, c, d = 0.
Let the four equations in the problem be (i), (ii), (iii) and (iv), respectively. Then
(i) + 3(ii) + 3(iii) + (iv) will give

(a + b)3 + (c + d)3 = −15. (1)

According to the equation (ii), b and d have different sign, and similarly (iv) yields that a
and c have different sign.
First, consider the case a > 0, b > 0. Then c < 0 and d < 0. By (i), we have a > −c (i.e.
|a| > |c|) and (iii) gives b > −d. Hence a + b > −(c + d) and so (a + b)3 > −(c + d)3 , thus
(a + b)3 + (c + d)3 > 0 which contradicts (1).
 5

Next, consider the case a > 0, b < 0. Then c < 0 and d > 0. By (i), we have a > −c and
by (iii), d > −b (i.e. b > −d). Thus a + b > −(c + d) and hence (a + b)3 + (c + d)3 > 0
which contradicts (1).
The case a < 0, b < 0 leads to c > 0, d > 0. By (i), we have c > −a and by (iii) d > −b.
So c + d > −(a + b) and hence (c + d)3 + (a + b)3 > 0 which contradicts (1) again.
Finally, consider the case a < 0, b > 0. Then c > 0 and d < 0. By (i), c > −a and by (iii)
b > −d which gives c + d > −(a + b) and hence (c + d)3 + (a + b)3 > 0 contradicting (1).
Hence there is no real solution to this system of equations. (Heiki Niglas)
Solution 3.  As in Solution 2, we conclude that a, b, c, d = 0. The equation a2 b + c2 d = 0
−d
yields a = ± c. On the other hand, we have a3 + c3 = 2 and ab2 + cd2 = −6 < 0
b 
−d
which implies that min{a, c} < 0 < max{a, c} and thus a = − c.
b

−d
Let x = − . Then a = xc and so
b

2 = a3 + c3 = c3 (1 + x3 ). (2)

Also −6 = ab2 + cd2 = cxb2 + cd2 , which, using (2), gives

3 −63
xb2 + d2 = = −108(x3 + 1).
c3
Thus 2

3
3
3
3 2 b 6 1 6 1 + x3
−108(1 + x ) = d x 2 +1 =d +1 =d . (3)
d x3 x3
If x3 + 1 = 0, then x = −1 and hence a = −c, which contradicts a3 + c3 = 2. So x3 + 1 = 0
and (3) gives
d6 (1 + x3 )2 = −108x9 . (4)
Now note that   3  
3 −d −d3 b3 − 1
x = − =− = −
b b3 b3
and hence (4) yields that
  2  3
b3 − 1 b3 − 1
(b3 − 1)2 1 − = 108 . (5)
b3 b3

b3 − 1 3 1
Let y = . Then b = and so (5) implies
b3 1 − y2

2
1
− 1 (1 − y 2 )2 = 108y 3 ,
1 − y2
 6

i.e.
y4
(1 − y)2 = 108y 3 .
(1 − y 2 )2
If y = 0, then b = 1 and so d = 0, a contradiction. So

y(1 − y)2 = 108(1 − y)2 (1 + y)2 .

Clearly y = 1 and hence y = 108 + 108y 2 + 216y, or 108y 2 + 215y + 108 = 0. The last
equation has no real solutions and thus the initial system of equations has no real solutions.
Remark 1. Note that this solution worked because RHS of a2 b + c2 d = 0 is zero. If
instead it was, e.g., a2 b + c2 d = 0.1 then this solution would not work out, but the first
solution still would.
Remark 2. The advantage of this solution is that solving the last equation 108y 2 +
215y + 108 = 0 one can find complex solutions of this system of equations. (Heiki Niglas)

10. Let a0,1 , a0,2 , . . . , a0, 2016 be positive real numbers. For n ≥ 0 and 1 ≤ k < 2016 set

1 1
an+1,k = an,k + and an+1,2016 = an,2016 + .
2an,k+1 2an,1

Show that max a2016,k > 44.

1≤k≤2016
Solution. We prove
m2n ≥ n (1)
for all n. The claim then follows from 442 = 1936 < 2016. To prove (1), first notice that
the inequality certainly holds for n = 0. Assume (1) is true for n. There is a k such that
an,k = mn . Also an,k+1 ≤ mn (or if k = 2016, an,1 ≤ mn ). Now (assuming k < 2016)

2
1 mn 1
a2n+1,k = mn + = m2n + + > n + 1.
2an,k+1 an,k+1 4a2n,k+1

Since m2n+1 ≥ a2n+1, k , we are done.

11. The set A consists of 2016 positive integers. All prime divisors of these numbers are
smaller than 30. Prove that there are four distinct numbers a, b, c and d in A such that
abcd is a perfect square.
Solution There are ten prime numbers ≤ 29. Let us denote them as p1 , p2 , ..., p10 . To
each number n in A we can assign a 10-element sequence (n1 , n2 , ..., n10 ) such that ni = 1
pi has an odd exponent in the prime factorization of n, and ni = 0 otherwise. Two numbers
to which identical sequences are assigned, multiply to a perfect square. There are only
1024 different 10-element {0, 1}-sequences so there exist some two numbers a and b with
identical sequencies, and after removing these from A certainly two other numbers c and d
with identical sequencies remain. These a, b, c and d satisfy the condition of the problem.
 7

12. Does there exist a hexagon (not necessarily convex) with side lengths 1, 2, 3, 4, 5, 6
(not necessarily in this order) that can be tiled with a) 31 b) 32 equilateral triangles with
side length 1?

Solution. The adjoining figure shows that question

a) can be answered positively.
For a negative answer to b), we show that the number
of triangles has to be odd. Assume there are x triangles
in the triangulation. They hav altogether 3x sides. Of
these, 1 + 2 + 3 + 4 + 6 = 21 are on the perimeter
of the hexagon. The remaining 3x − 21 sides are in
the interior, and they touch each other pairwise. So
3x − 21 has to be even, which is only possible, if x is
odd.

13. Let n numbers all equal to 1 be written on a black-

board. A move consists of replacing two numbers on
the board with two copies of their sum. It happens
that after h moves all n numbers on the blackboard
are equal to m. Prove that h ≤ 12 n log2 m.
Solution. Let the product of the numbers after the k-th move be ak . Suppose the
numbers involved in a move were a and b. By the arithmetic-geometric mean inequality,
(a + b)(a + b) ≥ 4ab. Therefore, regardless of the choice of the numbers in the move,
ak ≥ 4ak−1 , and since a0 = 1, ah = mn , we have mn ≥ 4h = 22h and h ≤ 12 n log2 m.

14. A cube consists of 43 unit cubes each containing an integer. At each move, you choose
a unit cube and increase by 1 all the integers in the neighbouring cubes having a face in
common with the chosen cube. Is it possible to reach a position where all the 43 integers
are divisible by 3, no matter what the starting position is?
Solution. Two unit cubes with a common face are called neighbours. Colour the cubes
either black or white in such a way that two neighbours always have different colours.
Notice that the integers in the white cubes only change when a black cube is chosen. Now
recolour the white cubes that have exactly 4 neighbours and make them green. If we look
at a random black cube it has either 0, 3 or 6 white neighbours. Hence if we look at the
sum of the integers in the white cubes, it changes by 0, 3 or 6 in each turn. From this it
follows that if this sum is not divisible by 3 at the beginning, it will never be, and none of
the integers in the white cubes is divisible by 3 at any state.

15. The Baltic Sea has 2016 harbours. There are two-way ferry connections between some
of them. It is impossible to make a sequence of direct voyages C1 − C2 − · · · − C1062 where
all the harbours C1 , . . . , C1062 are distinct. Prove that there exist two disjoint sets A and
B of 477 harbours each, such that there is no harbour in A with a direct ferry connection
to a harbour in B.
 8

Solution. Let V be the set of all harbours. Take any harbour C1 and set U = V \ {C1 },
W = ∅. If there is a ferry connection from C to another harbour, say C2 in V , consider the
route C1 C2 and remove C2 from U . Extend it as long as possible. Since there is no route
of length 1061, So we have a route from C1 to some Ck , k ≤ 1061, and no connection from
Ck to a harbor not already included in the route exists. There are at least 2016 − 1062
harbours in U . Now we move Ck from U to W and try to extend the route from Ck−1
onwards. The extension again terminates at some harbor, which we then move from U to
W . If no connection from C1 to any harbour exists, we move C1 to W and start the process
again from some other harbour. This algorithm produces two sets of harbours, W and U ,
between which there are no direct connections. During the process, the number of harbours
in U always decreases by 1 and the number of harbours in W increases by 1. So at some
1
point the number of harbours is the same, and it then is at least (2016 − 1062) = 477. By
2
removing, if necessary, some harbours fron U and W we get sets of exactly 477 harbours.

16. In triangle ABC, the points D and E are the intersections of the angular bisectors
from C and B with the sides AB and AC, respectively. Points F and G on the extensions
of AB and AC beyond B and C, respectively, satisfy BF = CG = BC. Prove that
F G  DE.

Solution. Since BE and CD are angle bisectors,

AD AC AE AB
= , = .
AB AC + BC AC AB + BC

So

AD AD AB AC · AB
= · =
AF AB AF (AC + BC)(AB + BC)

and

AE AE AC AB · AC
= · = .
AG AC AG (AB + AC)(AC + BC)

AD AE
Since = , DE and F G are parallel.
AF AG

17. Let ABCD be a convex quadrilateral with AB = AD. Let T be a point on the
diagonal AC such that ∠ABT + ∠ADT = ∠BCD. Prove that AT + AC ≥ AB + AD.
 9

Solution. On the segment AC, consider the unique

point T  such that AT  · AC = AB 2 . The trian-
gles ABC and AT  B are similar: they have the an-
gle at A common, and AT  : AB = AB : AC. So
∠ABT  = ∠ACB. Analogously, ∠ADT  = ∠ACD.
So ∠ABT  + ∠ADT  = ∠BCD. But ABT  + ADT 
increases strictly monotonously, as T  moves from A
towards C on AC. The assumption on T implies that
T  = T . So, by the arithmetic-geometric mean in-
equality,
√
AB + AD = 2AB = 2 AT · AC ≤ AT + AC.

18. Let ABCD be a parallelogram such that ∠BAD = 60◦ . Let K and L be the midpoints
of BC and CD, respectively. Assuming that ABKL is a cyclic quadrilateral, find ∠ABD.

Solution. Let ∠BAL = α. Since ABKL is cyclic,

∠KKC = α. Because LKDB and ABDC, we fur-
ther have ∠DBC = α and ∠ADB = α. Let BD and
AL intersect at P . The triangles ABP and DBA have
two equal angles, and hence ABP ∼ DBA. So

AB BP
= . (1)
DB AB

The triangles ABP and LDP are clearly similar with similarity ratio 2 : 1. Hence BP =
2
DB. Inserting this into (1) we get
3

2
AB = · DB.
3

The sine theorem applied to ABD (recall that ∠DAB = 60◦ ) immediately gives
 √ √
AB 2 3 2
sin α = sin 60◦ = · = = sin 45◦ .
BD 3 2 2

So ∠ABD = 180◦ − 60◦ − 45◦ = 75◦ .

19. Consider triangles in the plane where each vertex has integer coordinates. Such a
triangle can be legally transformed by moving one vertex parallel to the opposite side to
a different point with integer coordinates. Show that if two triangles have the same area,
then there exists a series of legal transformations that transforms one to the other.
 10

Solution. We will first show that any such triangle can be transformed to a special triangle
whose vertices are at (0, 0), (0, 1) and (n, 0). Since every transformation preserves the
triangle’s area, triangles with the same area will have the same value for n.
Define th y-span of a triangle to be the difference between the largest and the smallest y
coordinate of its vertices. First we show that a triangle with a y-span greater than one
can be transformed to a triangle with a strictly lower y-span.
Assume A has the highest and C the lowest y coordinate of ABC. Shifting C to C  by the
−→
vector BA results in the new triangle ABC  where C  has larger y coordinate than C baut
lower than A, and C  has integer coordinates. If AC is parallel to the x-axis, a horizontal
shift of B can be made to transform ABC into AB  C where B  C is vertical, and then
A can be vertically shifted so that the y coordinate of A is between those of B  and C.
Then the y-span of AB  C can be reduced in the manner described above. Continuing the
process, one necessarily arrives at a triangle with y-span equal to 1. Such a triangle then
necessarily has one side, say AC, horizontal. A legal horizontal move can take B to the
a position B  where AB  is horizontal and C has the highest x-coordinate. If B  is above
AC, perform a vertical and a horizontal legal move to take B  to the origin; the result is a
special triangle. If B  is below AC, legal transformation again can bring B  to the origin,
and a final horizontal transformation of one vertex produces the desired special triangle.
The inverse of a legal transformation is again a legal transformation. Hence any two
triangles having vertices with integer coordinates and same area can be legally transformed
into each other via a special triangle.

20. Let ABCD be a cyclic quadrilateral with AB and CD not parallel. Let M be the
midpoint of CD. Let P be a point inside ABCD such that P A = P B = CM . Prove that
AB, CD and the perpendicular bisector of M P are concurrent.

Solution. Let ω be the circumcircle of ABCD. Let

AB and CD intersect at X. Let ω1 and ω2 be the
circles with centers P and M and with equal radius
P B = M C = r. The power of X with respect to ω
and ω1 equals XA · XB and with respect to ω and
ω2 XD · XC. The latter power also equals (XM +
r)(XM −r) = XM 2 −r 2 . Analogously, the first power
is XP 2 − r 2 . But since XA · XB = XD · XC, we must
have XM 2 = XP 2 or XM = XP . X indeed is on the
perpendicular bisector of P M , and we are done.