8th Benelux Mathematical Olympiad

Soest, 29 April – 1 May 2016

Solutions

Problem 1. Find the greatest positive integer N with the following property: there exist
integers x1 , . . . , xN such that x2i − xi xj is not divisible by 1111 for any i 6= j.

Solution. We prove that the greatest N with the required property is N = 1000. First
note that x2i − xi xj = xi (xi − xj ), and that the prime factorisation of 1111 is 11 · 101.

We first show that we can find 1000 integers x1 , x2 , . . . , x1000 such that x2i − xi xj is not
divisible by 1111 for any i 6= j. Consider the set {1, 2, . . . , 1110}. This set contains 10
integers divisible by 101, and it contains 100 integers divisible by 11. None of the integers
in the set are divisible by both 11 and 101. If we delete all of these 10 + 100 integers from
the set, we are left with 1000 integers. Call these x1 , x2 , . . . , x1000 . Now we have 11 - xi
and 101 - xi for all i. Suppose there are i 6= j with 1111 | xi (xi − xj ), then we must have
1111 | xi − xj , which is a contradiction, since xi , xj ∈ {1, 2, . . . , 1110}. So this set satisfies
the requirement.

We now prove that given 1001 (or more) integers x1 , x2 , . . . , x1001 there are i 6= j with
1111 | xi (xi − xj ). Suppose for a contradiction that for all indices i 6= j, we have that
xi (xi − xj ) is not divisible by 1111, and write X = {x1 , . . . , x1001 }. We may (after reducing
modulo 1111 if necessary) assume that xi ∈ {0, 1, . . . , 1110} for all i. Then we know that
xi 6= 0 for all i, and xi 6= xj for all i 6= j. Suppose for some i we have 11 | xi . (Since
xi 6= 0, we know that 101 - xi .) Then any integer a 6= xi with a ≡ xi mod 101 cannot be
an element of X, since 1111 | xi (xi − a). In {1, 2, . . . , 1110} there are 10 such integers, all
of them coprime with 11 · 101. If there are exactly k different values of i such that 11 | xi ,
there are 10k different integers from {1, 2, . . . , 1110} that cannot be elements of X, all of
them coprime with 11 · 101. Similarly, if there are exactly m different values of i such
that 101 | xi , then there are 100m different integers from {1, 2, . . . , 1110} that cannot be
elements of X, all of them coprime with 11 · 101. (Note that those 10k and 100m integers
can overlap.)

In {1, 2, . . . , 1110} there are 100 multiples of 11, there are 10 multiples of 101 and there
is no multiple of 11 · 101, so there are 1000 integers that are coprime with 11 · 101. In X
we have 1001 − k − m integers that are coprime with 11 · 101, so exactly k + m − 1 of
the coprime integers in {1, 2, . . . , 1110} are not in X. This implies that 10k 6 k + m − 1
and 100m 6 k + m − 1. Adding these two inequalities we find 8k + 98m 6 −2, a clear
contradiction. So N < 1001. 
Problem 2. Let n be a positive integer. Suppose that its positive divisors can be parti-
tioned into pairs (i.e. can be split in groups of two) in such a way that the sum of each pair
is a prime number. Prove that these prime numbers are distinct and that none of these
are a divisor of n.

Solution. Let d1 and d2 be positive divisors of n that form a pair as given in the problem. If
d1 and d2 have a non-trivial prime divisor p in common, then p | d1 +d2 and p 6 d1 < d1 +d2 ,
so d1 + d2 cannot be prime. Hence gcd(d1 , d2 ) = 1, which implies that d1 d2 | n. Suppose
the number of positive divisors of n is 2t (it is even since the divisors can be split into
pairs). If we now multiply all divisors, then on one hand we have the product of all d1 d2
where {d1 , d2 } is a pair, so that product is at most nt . On the other hand for every divisor
d there is another divisor nd (since the number of divisors is even, the case d = nd does not
occur), and the product of all those is equal to nt . Hence there must be equality in every
inequality d1 d2 6 n. So the pairs of divisors given in the problem are all of the form {d, nd }.

Now we prove the two statements in the problem. Suppose that d, d0 are positive divisors
of n such that d + nd = d0 + dn0 . Then d2 d0 + nd0 = d(d0 )2 + nd, so dd0 (d − d0 ) = n(d − d0 )
and hence (dd0 − n)(d − d0 ) = 0. Therefore either d = d0 or dd0 = n, which implies that
{d, nd } = {d0 , dn0 }, as required.

Now let d be a positive divisor of n. Every prime divisor p of n divides precisely one of d
and nd , since d nd = n and gcd(d, nd ) = 1; so p - d + nd . Therefore gcd(n, d + nd ) = 1. Since
d + nd > 1 we conclude that d + nd cannot be a divisor of n. 
Problem 3. Find all functions f : R → Z such that

2
   
2 2


f f (y) − x + f (x) + f (y) = f (y) · 1 + 2f f (y)

for all x, y ∈ R.

Solution I. Take x = y = 0 and write c = f (0), then we find f (c)2 + c2 + c2 = c + 2cf (c),
so (f (c) − c)2 = c − c2 . The left-hand side is non-negative, so the right-hand side must be
non-negative as well, hence c − c2 > 0, so c(1 − c) > 0. This implies 0 6 c 6 1, and since
c ∈ Z, we get c = 0 or c = 1. In both cases we have c − c2 = 0, so f (c) − c = 0, hence
f (c) = c. Now taking y = 0 we find for all x ∈ R that

2
f (c − x) + f (x)2 + c2 = c + 2c2 . (1)
If c = 0, then this equation reduces to f (−x)2 + f (x)2 = 0, and since the left-hand side
consists of the sum of two squares, both squares must be zero. Therefore f (x) = 0 for all
x. This function is indeed a solution of the given functional equation.

Now we consider the other case: c = 1. Then (1) reduces to f (1 − x)2 + f (x)2 = 2. The
left-hand side consists of two squares of integers, so they must both be 1. Therefore for all
x we have f (x) = 1 or f (x) = −1.

Suppose there is an a with f (a) = −1. We take y = a in the functional equation; using
that f (w)2 = 1 for any w, we find 1 + 1 + 1 = −1 · (1 + 2f (−1)). Both with f (−1) = 1 and
with f (−1) = −1 this gives a contradiction. So there exists no such a, and we conclude
that f (x) = 1 for all x ∈ R. This is also a solution of the given functional equation.

We conclude that there are two solutions: f (x) = 0 for all x, and f (x) = 1 for all x. 

Solution II. We proceed as in the first solution up to the point of deriving f (x) = ±1 for
all x ∈ R. Now we take x = 0 in the original functional equation, and we find

2
   
2 2


f f (y) + f (0) + f (y) = f (y) · 1 + 2f f (y) .

We can rewrite this as

2
 
f (y) − f f (y) + f (0)2 = f (y).

The left-hand side is non-negative, so f (y) > 0 for all y. If we combine this with f (x) = ±1
for all x, we can conclude that f (x) = 1 for all x ∈ R. And this is a solution. 
Problem 4. A circle ω passes through the two vertices B and C of a triangle ABC.
Furthermore, ω intersects segment AC in D 6= C and segment AB in E 6= B. On the ray
from B through D lies a point K such that |BK| = |AC|, and on the ray from C through
E lies a point L such that |CL| = |AB|. Show that the circumcentre O of triangle AKL
lies on ω.

Solution I. Let M be the midpoint of the arc BC of ω that is on the same side of BC as
A. Then |BM | = |CM |. We also have |BA| = |CL| and ∠ABM = ∠EBM = ∠ECM =
∠LCM . Hence 4ABM ∼ = 4LCM . So |AM | = |LM |. (In case M = E the triangles ABM
and LCM are degenerate, but then the proof still works, since |AM | = |AB| − |BM | =
|LC| − |CM | = |LM |.) Similarly, we prove that |AM | = |KM |, so M is the circumcentre
of 4AKL. This means that M = O, and hence we are done as M was defined to be on ω. 

Solution II. We consider the configuration where O is in the interior of triangle ABC.
The proof for other configurations is similar. Furthermore, we exclude the case that O = D
or O = E; in those cases it is immediate that O is on ω.

We have ∠ABK = ∠EBD = ∠ECD = ∠LCA. Together with |AB| = |CL| and |AC| =
|BK| this implies 4ABK ∼ = 4LCA. Hence |AK| = |AL|, ∠AKB = ∠LAC (denote this
angle by α) and ∠BAK = ∠CLA (denote this angle by β). Furthermore, let γ = ∠BEC =
∠BDC. Then ∠BAC = 180◦ − ∠BDA − ∠ABD = ∠BDC − ∠ABD = γ − ∠ABD.
Therefore ∠KAL = ∠BAK +∠LAC −∠BAC = α+β −(γ −∠ABD) = α+β +∠ABD−γ.
Note that in triangle ABK we have α + β + ∠ABD = 180◦ , hence ∠KAL = 180◦ − γ.

We have ∠KOL = 2(180◦ − ∠KAL) = 2γ. Since |AK| = |AL|, this implies ∠AOL = γ.
As we also have ∠AEL = ∠BEC = γ, the quadrilateral OELA is cyclic. Analogously,
ODKA is cyclic. Now we have

∠DOE = 360◦ − ∠DOA − ∠AOE = 180◦ − ∠DOA + 180◦ − ∠AOE

= ∠DKA + ∠ALE = ∠BKA + ∠ALC = α + β.

On the other hand

∠DBE = ∠KBA = 180◦ − ∠BAK − ∠AKB = 180◦ − α − β.

Hence ∠DOE +∠DBE = 180◦ , so O is on the circle containing D, B and E, which is ω.