XXX Olimpiada Colombiana

de Matemáticas y
XII Olimpiada Bolivariana
de Matemáticas
Solucionario

Nivel Superior

1. Let p(x) be a polynomial of degree 2 such that p(1) = 1, p(2) = 7 and

p(3) = 19. Show that for every natural n ≥ 1, the polynomial satisfies
p(1) + p(2) + · · · + p(n) = n3 .
Solution: Notice that the polynomial is determined in a unique way
by the values that it takes at the three given points. Taking p(x) =
ax2 + bx + c, we get the following three equations

a+b+c = 1
a + 2b + 4c = 7
a + 3b + 9c = 19

The solution to this system is a = 3, b = −3 aand c = 1. There are

several ways to end the problem from this point; two of the possible
solutions are shown below:
Solution 1: We may write the sum as follows:

1
 n
X
p(1) + p(2) + · · · + p(n) = p(i)
i=1
Xn
= 3i2 − 3i + 1
i=1
n
X n
X n
X
2
= 3 i −3 i+ 1
i=1 i=1 i=1
n(n + 1)(2n + 1) n(n + 1)
= 3 −3 +n
6 2
n(n + 1)(2n + 1) n(n + 1)
= 3 −3 +n
6 2
n + 3n2 + 2n3 3n2 + 3n
= − +n
2 2
= n3

Which is what we wanted to show.

Solution 2: We will prove the statement using induction. The base
case is n = 1, which can be easily verified. For the inductive step,
we assume that the statement is true for n − 1. Thus, we have that
p(1) + · · · + p(n − 1) + p(n) = (n − 1)3 + 3n2 − 3n + 1 = n3 .
2. Given a set S, we let PS be the product of all the elements in S.
Determine for which values of n the set {1, 2, . . . , 2n} can be divided
into two disjoint sets of size n, S1 y S2 , such that PS1 − PS2 = 1 or 2.
Solution: If PA − PB = 1, it means that one of the products is even
and the other one is odd. Since we have exactly n even numbers and n
odd numbers, then A must be the set with all the even numbers and B
the one with all the odd numbers, or vice versa. Notice that PA > PB
because for each odd k ∈ B, k + 1 ∈ A and comparing term by term
we get the desired inequality. It is clear that when n = 1 we can make
the difference between the product of the even numbers and the odd
numbers to be 1. But if n > 1 the product of the even numbers will
grow faster than the product of the odd numbers, and the difference
won’t be 1.
If PA − PB = 2, then both products are even, but they can’t be divis-
ible by 4 simultaneously. In consequence, we can assume without loss
of generality that the set B contains a number of the form 2k with k

2
 odd, and the rest of its elements are odd. This implies that there is
only one odd number in A, and the rest of its elements are even.

Since the difference is not divisible by 3, all the multiples of 3 must

be in one of the sets only. If n ≥ 6 the numbers 3, 6, 9, 12 must be in
the same set, which is imposible for the reasons given in the previous
paragragh.

If n = 5 then 3, 6, 9 must be in the same set, so 6 ∈ B; moreover 5, 10

are also in the same set, so 5 ∈ A. Therefore, A = {2, 4, 5, 8, 10} and
B = {1, 3, 6, 7, 9}. Hence PA = 3200 y PB = 1134, which don’t satisfy
the condition.
If n = 4, then 3 and 6 must be in the same set. There are only two
possibilities: if 6 ∈ A, A = {3, 4, 6, 8} and B = {1, 2, 5, 7}, we get
PA = 576, PB = 70 which don’t satisfy the condition. If 6 ∈ B, then A
contains one number among 1, 5, 7 and we check each case separately:

PA = 1 × 2 × 4 × 8 = 64 and PB = 3 × 5 × 6 × 7 = 630
PA = 2 × 4 × 5 × 8 = 320 and PB = 1 × 3 × 6 × 7 = 126
PA = 2 × 4 × 7 × 8 = 448 and PB = 1 × 3 × 5 × 6 = 90

None of them fulfills the conditions of the problem.

If n = 3, we have three cases:

PA = 3 × 4 × 6 = 72 and PB = 1 × 2 × 5 = 10
PA = 3 × 4 × 6 = 72 and PB = 1 × 2 × 5 = 10
PA = 1 × 2 × 4 = 8 and PB = 3 × 5 × 6 = 90

None of them satisfies the problem.

If n = 2 it is clearly possible with 2 × 3 − 1 × 4 = 2. If n = 1 it is not
possible.

3. Ivan and Alexander write lists of integers. Ivan writes all the lists of
length n with elements a1 , a2 , . . . , an such that |a1 |+|a2 |+· · ·+|an | ≤ k.
Alexander writes all the lists with length k with elements b1 , b2 , . . . , bk
such that |b1 | + |b2 | + · · · + |bk | ≤ n. Prove that Alexander and Ivan
wrote the same number of lists.
Solution 1: We will create a biyection between Ivan’s lists and Alexan-
der’s lists. In order to do that, we will define some terms that we will

3
use during the proof. A sublist of a list is a sequence of numbers ap-
pearing consecutively in the list. The weight of a sublist (list), is the
sum of the absolute value of the numbers of the sublist (list). The
length of a sublist (list) is the number of numbers in it. A block is
a sublist with the first element different from 0, and the rest of the
numbers equal to 0. In particular, a sublist of length one and weight
different from 0 is a block. The sign of a block corresponds to the sign
of the first element.
Thus, we can say that a list belongs to Ivan if it has lenght n and
weigth less or equal to k, and it belongs to Alexander if the length is
k and the weight is less than or equal to n. Consider a list belonging
to Ivan, and add a positive number at the beginning of it, so that the
weigth of the new list is k + 1; the number added must be different
from zero. The new list can be divided into blocks, since the initial
number is different from 0. To do this, consider the first number and
all the zeros that follow it. When we reach a number different from
0, we start a new block. Each block B will be changed by an inverse
block B 0 , with length equal to the weight of B and weight equal to
the length of B. Also, the sign of B must be equal to that of B 0 . This
process of changing a block by its inverse can only be made in a unique
way.
With this process, when we change all the blocks of a list that belongs
to Ivan, we will get a list with weight n+1 and length k +1, whose first
element is different from 0. When we delete this number, we get a list
with length k and weigth less than or equal to n, which corresponds to
a list that belongs to Alexander. When we make the inverse process,
adding a number at the beginning of a list belonging to Alexander in
such a way that the weight is k+1, invert the blocks and delete the first
number, we will get a list that belongs to Ivan. Hence, we have found an
invertible function that transforms Ivan’s lists into Alexander’s lists,
and therefore it’s a biyection, which concludes the problem.
Solution 2: Let nk be the number of lists of length k and weight


less than or equal to n. It is clear that n0 = 1 and n0 = 1. We will

 
n k
prove that k = n using induction on w when mı́n n, k ≤ w. The
base case, w = 0 is trivial. We will now show that the problem is true
when n ≥ k = w0 + 1, assuming that the property is true for w ≤ w0 .
We will show that:

4
          
n n n−1 n−2 0
= +2 +2 + ··· + 2 (1)
k k−1 k−1 k−1 k−1

The first term of the sum counts all the lists with first element equal to
0; counting these lists is equivalent to count the lists with length k − 1
and weight less than or equal to n. The second term counts the lists
with first element equal to ±1. Since the initial number has absolute
value 1, the remaining part of the list is a list of length k−1
and weigth
less than or equal to n − 1. In general, the term 2 n−r k−1 counts the
lists with first element equal to ±r. In this way, (1) is demonstrated.
Now, we will show another recurrence for the number of lists:

         
k k−1 k−1 k−1 k−1
= +2 +2 + ··· + 2 (2)
n n n−1 n−2 0

In this case, we will interpret the terms of the sum in a different way.
The first term represents those lists whose weigth is strictly less than
k. When the weigth is exactly k there are several cases to consider.
The first one is when the first term is different from 0. In such case,
we can complete a list of length n − 1 and weight less than or equal
to k − 1 by adding a correct initial number, without considering the
sign. In this way, the number of lists with weight exactly k whose first
k−1
number is different from 0 is 2 n−1 . If the first term is 0, but the
second isn’t, we take a sublist of length n − 2 and weight less than or
equal to k − 1, and complete it in two ways by adding 0, ±p, where p
isthe number required to complete the weight k. In general, the term
k−1
2 n−r counts the lists with weight exactly equal to k, and whose first
r − 1 elements are 0 and the r-th element is different from 0.
Combining (1) and (2), and considering k k − 1 = w0 , we can use
n that
the hypothesis of induction to get k = n , as desired.
Solution 3: We will count the number of lists a1 , a2 , . . . , an with |a1 |+
|a2 | + · · · + |an | ≤ k.
Let 0 ≤ r ≤ n be the number of terms in the list that are different
from 0. Considering their absolute values, we must find the number of
lists of positive integers b1 , · · · , br satisfying b1 + · · · + bn ≤ k.

5
 k


Lemma The number of such lists is r .

Proof It is well known that the number
of lists of positive integers
b1 , · · · , br with b1 + · · · + bn = k is k−1
r−1 . We must count the lists
whose sum is strictly less than k.
For each list with r terms with sum strictly less than k we can construct
a list with r + 1 terms that add exactly k by adding br+1 = k − (b1 +
· · · + br ), and it’s clear that the process is reversible: given a list with
r + 1 terms whose sum is exactly k we can delete br+1 to get a list of
r elements whose sum is strictly less than k. Therefore, the number
of lists with r terms whose sum is strictly less than k is the same as
the number of
lists with r + 1 terms with sum exactly k, which is
k−1 k−1


r+1−1 = r .

Hence, the total amount is k−1 k−1

= kr . 




r−1 + r

Returning to the problem, the number
of lists of nr
positive integers

k
with sum less than or equal to k is r . There are r possible subsets
of indexes I such that ai 6= 0 for i ∈ I and ai = 0 for i ∈ / I. Moreover,
we must choose a sign for the r terms different from 0, which can be
done in 2r ways. Thus, for a fixed
r,
the amount of lists with exactly
r terms different from zero is kr nr 2r . Since r varies from 0 to n we
have that the number of lists is
n   
X k n r
2
r r
r=0

Taking into account that ab = 0 si b > a, we see that it is enough

if we vary r from 0 to min(n, k), so n and k can be exchanged in the

formula.
4. Consider a regular polygon of n sides and all its diagonals drawn, and
light bulbs on each vertex. There are k bulbs on initially. In each step
of the game, the player erases a diagonal and changes the state of one
of the two bulbs in the ends of the diagonal. Determine all the possible
values of k for which it’s possible to turn on all the light bulbs after
having erased all the diagonals. Note: The sides of the polygon are also
considered to be diagonals.
Solution: The values of k that work are all those between 0 and n such
that n(n+1)
2 and k have the same parity. Let us call v1 , v2 , . . . , vk the

6
 vertices that have their bulbs turned on and vk+1 , . . . , vn the vertices
with their bulbs turned off, and for i between 1 and n let ai be the
number of times that vi changed its state after we have erased all the
diagonals.
The total number of changes of state equals to the number of diagonals,
so
n
n(n − 1) X
= ai
2
i=1
If all the lights are on in the end, it means that a1 , a2 , . . . , ak are all
even and all the others are odd, because a bulb will end in the same
state as the initial if and only
P if it changes an even number of times. It
follows that the parity of ni=k+1 ai is the same as that of n(n−1) 2 and
since all the terms are odd, the parity must be equal to that of n − k.
As a result, the parity of k and the parity of n(n+1)
2 must be the same.
Now we must show that if n and n(n+1) 2 have the same parity then it is
possible to turn on all the bulbs after having deleted all the diagonals.
It is easy to check that it’s possible for n = 1, 2, 3, 4. If we can turn on
all the bulbs for n then take a configuration for a polygon of n+4 sides.
Clearly n(n+1)
2 and (n+4)(n+5)
2 have the same parity and it is possible
to pick 4 points among which there is an even number of bulbs on. If
n is even we use the case n = 4 to leave the 4 bulbs on, and if n is
odd, we use the same case to turn them off. Then, for each of these
four bulbs we delete the n diagonals that come out from each of their
vertices and change the state of the bulbs. If n is odd they will change
an odd number of times, and since they were off, they will be on in
the end, and if n is even, they will change an even number of times
and since they were on, they will end on. Finally, we use the case n to
turn on the remaining n bulbs. This concludes the induction and the
problem.
5. Let p and q be fixed integers. We define:

a0 = 0, a1 = 1, and
an = pan−1 − an−2 for n ≥ 2 even,
an = qan−1 − an−2 for n ≥ 3 odd.

Find the values of p and q such that the sequence has negative values.
Solution: Let us write Xn = a2n y Yn = a2n+1 . The recurrence rela-
tions given in the problem can be rewritten as follows:

7
 X0 = 0, Y0 = 1, and
Xn = pYn−1 − Xn−1 for n ≥ 1,
Yn = qXn − Yn−1 for n ≥ 1.

It is clear that if any of the numbers p or q is non positive, then the

sequence takes negative values, and this can be checked in the first
three terms. So, we will consider p ≥ 1 y q ≥ 1.
When we multiply the first recurrence relation by q we get
qXn = pqYn−1 − qXn−1
for n ≥ 1. The terms X can be replaced using the second recurrence
relation, sinceYn + Yn−1 = qXn . Thus, we obtain
Yn + Yn−1 = pqYn−1 − (Yn−1 + Yn−2 )
as long as n − 1 ≥ 1 (because we used the second recurrence relation
with n − 1). This last equation can be simplified as
Yn = (pq − 2)Yn−1 − Yn−2 (3)

In a similar fashion we obtain that

Xn = (pq − 2)Xn−1 − Xn−2 (4)

In both cases, the relations are true for n ≥ 2. Notice that X1 = p

and Y1 = pq − 1. When pq = 1 or pq = 2 the recurrence relations (4)
and (2) have coefficients that make them negative. When pq = 3 we
get that Xn = Xn−1 − Xn−2 ; as a result, if we assume that they are
positive, Xn < Xn−1 . Then, Xn+1 = Xn − Xn−1 < 0. Hence, in order
to have non negative terms, we must have that pq ≥ 4. We will show
that in this case all the terms of sequence are positive, by showing
that the sequences Xn and Yn are increasing. We will use induction
on n. The base cases are trivial. Then, Xn = (pq − 2)Xn−1 − Xn−2 =
(pq − 3)Xn−1 + (Xn−1 − Xn−2 ) ≥ (pq − 3)Xn−1 ≥ Xn−1 . The induction
for Y is equivalent, which demonstrates the hypothesis, and the terms
increase and in consequence, they are always positive.
6. Let ABC be a triangle with ∠C = 90◦ . Let M be the midpoint of
the height CD. The circumferences Γ1 and Γ2 with diameters AD
and BD respectively, intersect AM and BM at X and Y respectively.
Moreover, Γ1 intersects AC at P , and Γ2 intersects BC at Q. Prove
that the cuadrilateral P XY Q is cyclic.

8
7. Notice that ∠DQB = 90 since Q is in a circle with diameter DB.
Then, ∠DQC = 90; in the same way, ∠CP D = 90. Therefore, DP CQ
is a rectangle. Since the diagonals of a rectangle intersect at their mid-
points, we conclude that M belongs to P Q. Then ∠QCD = ∠QP D,
and ∠QCD = ∠BCD = ∠CAD, since CD is the height. Hence,
∠QP D = ∠CAD = ∠P AD which implies that P Q is tangent to Γ1 ,
and in a similar way we conclude that it is also tangent to Γ2 . From
here we get that ∠QBY = ∠M QY . We also have that ∠CQM =
∠QCM = ∠CAD, and as a result ABQP is cyclic. Notice that DY
is perpendicular to M B since Y belongs to a circle with diameter
DB. In the same way, DX is perpendicular to M A, and we conclude
that DXM Y is cyclic. Therefore, ∠DBY = ∠M DY = ∠M XY . Al-
so, 4M XP ∼ 4M P A because they share the angle ∠P M X and
∠M P X = ∠M AP ; from this we get that ∠M XP = ∠M P A . Since
ABQP is cyclic, we have that ∠ABQ + ∠QP A = 180◦ which means
(∠DBY + ∠QBY ) + ∠M XP = 180◦ ; thus:

180◦ = (∠M XY + ∠M QY ) + ∠M XP
= ∠M QY + (∠M XY + ∠M XP )
= ∠P QY + ∠P XY

We conclude that P XY Q is cyclic.