10th European Mathematical

Cup MLADI NADARENI MATEMATIČARI

11th December 2021 - 19th December 2021 Marin Getaldic

Senior Category

Problems and Solutions

Problem 1. Alice drew a regular 2021-gon in the plane. Bob then labelled each vertex of the 2021-gon with a
real number, in such a way that the labels of consecutive vertices differ by at most 1. Then, for every pair of
non-consecutive vertices whose labels differ by at most 1, Alice drew a diagonal connecting them. Let d be the
number of diagonals Alice drew. Find the least possible value that d can obtain.
(Ivan Novak)

First Solution. Consider the following labelling of the vertices, where the i-th number of the 2021-tuple below is the
label of the i-th vertex:
(0.5, 1.5, 2.5, . . . , 1009.5, 1010.5, 1010, 1009, 1008, . . . , 2, 1).
It’s easy to see that in this case, Alice will draw 2018 diagonals, those connecting the vertices whose pairs of labels are
{1.5, 1}, {2.5, 2}, . . . , {1009.5, 1009} and {1.5, 2}, {2.5, 3}, . . . , {1009.5, 1010}.
3 points.
We now prove that 2018 is the minimum amount of diagonals Alice could have drawn. Call any labelling of a convex
n-gon which satisfies the condition that consecutive vertices have labels which differ by at most 1 a sweet labelling, and
also call the corresponding n-gon sweet.
We will prove by mathematical induction that for every n > 3, in any sweet labelling of an n-gon, there are at least n − 3
pairs of nonconsecutive vertices whose labels differ by at most 1. The claim is obvious for n = 3. Suppose that the claim
is true for some positive integer n.
Consider a sweet labelling of some n + 1-gon P . Consider a vertex v with the maximum label, L. Then both of its
neighbouring vertices have labels in the set [L − 1, L], which means that their labels differ by at most 1.
1 point.
Then the n-gon P obtained from P by erasing v and connecting its neighbouring vertices is also sweet.
0

2 points.
Applying the inductive hypothesis on it, there are at least n − 3 pairs of nonconsecutive vertices of P whose labels differ
0

by at most 1. Adding the pair of neighbours of v, we conclude that P has at most n − 2 pairs of such vertices. This
completes the step of the induction.
4 points.

Second Solution. The example which achieves the desired bound is the same as in the previous solution.
3 points.
Let the labels of the vertices of the 2021-gon be v1 , . . . , v2021 , where we assume the labels to be ordered so that
v1 6 . . . 6 v2021 . Note that this is not necessarily the order in which the values appear on the 2021-gon.

We claim that |vi − vi+2 | 6 1, for all i = 1, 2, . . . , 2019.

Assume for the sake of contradiction that there exists an i ∈ {1, 2, . . . , 2019} for which vi+2 −vi > 1. Start a circular walk
around the 2021-gon, going from the vertex which has the value v1 , visiting all of the vertices one by one, and returning
back to the starting vertex. Doing so visits the values v1 , . . . , v2021 in a certain permuted order, starting and ending on v1 .

1
We look at the first time during the walk when we step on a value whose index is greater than or equal to i + 2. Let
this index be j > i + 2. Let’s say that on the previous step, we were on value vb , where b 6 i + 1. Note that if b 6 i,
then vj − vb > vi+2 − vi > 1, so it must be the case that b = i + 1. Next, we look at the first time we return to an index
which is smaller than or equal to i. Such an index must exist since we eventually return back to v1 , and we’ll denote it
by k. A similar argument as for vj shows that in the step before reaching vk , we must have been on index b. This is a
contradiction as no vertex can be visited more than once, except for the one we started with.
5 points.
We now have |vi − vj | 6 1, for all i = 3, . . . , 2019 and j ∈ {i − 1, i − 2, i + 1, i + 2}, |v1 − vi | 6 1 for j ∈ {2, 3} and
|vj − v2021 | 6 1 for j ∈ {2019, 2020}, which gives at least 12 (2 + 3 + 4 · 2017 + 3 + 2) − 2021 = 2018 diagonals.
2 points.

Third Solution. The example which achieves the desired bound is the same as in the previous solution.
3 points.
Note that
We label the vertices v1 , v2 , . . ., v2021 , and, respectively, their labels x1 , x2 , . . ., x2021 and view the indices modulo 2021.
We’ll say vertices (vi , vj ) are a nigh pair if their labels differ by at most 1.
Without loss of generality, let 1 and g be the indices among {1, 2, . . . , 2021} of the vertices with the smallest and greatest
label, respectively. Without loss of generality we can also assume g 6 1011 since we can otherwise mirror the 2021-gon.
Additionally, we will assume that g > 4 , and the cases g = 3 and g = 2 are dealt with separately. We make use of the
following lemmas.
Lemma 1. For every 1 < k < g, there are at least two indices g 6 i, j 6 2022 such that (vk , vi ) and (vk , vj ) are pairs of
nigh vertices. Similarly, for every g < k < 2022, there are at least two indices 1 6 i, j 6 g such that (vk , vi ) and (vk , vj )
are pairs of nigh vertices.
Proof. As the statement is obviously symmetric, we will only prove the first half. Let m and M be the smallest and the
biggest label, respectively. If xk 6 m + 1, v2021 and v1 satisfy the condition of the lemma. Similarly, if xk > M − 1,
vg and vg+1 satisfy the condition of the lemma. Otherwise, there are at least two indices within the desired range with
labels in [x − 1, x + 1] due to the Intermediate Value Theorem. Namely, if we imagine jumping along the vertices from v1
do vg , at the vertex v1 the label is less than x − 1 and at the vertex vg the label is greater than x + 1. Then at some point
in between we must have been at a vertex whose label is from [x − 1, x) and at a vertex whose label is from [x, x + 1).
4 points.
Lemma 2. At most one of the vertices v2021 and v2 and at most one of the vertices vg−1 and vg+1 can be elements of
three nigh pairs.
Proof. Clearly, due to Lemma 1, both v2021 and v2 are elements of at least three nigh pairs. Assume both vertices are
elements of exactly three nigh pairs. Assume that x2021 6 x2 . It follows that x2 − 1 6 x1 6 x2020 6 x2021 + 1 6 x2 + 1,
therefore, v2020 , v2021 , v1 and v3 are four vertices forming nigh pairs with v2 , a contradiction. We treat the other cases
analogously.

2 points.

Finally, due to Lemma 1, each vertex not neighbouring with g or 1 forms at least 4 nigh pairs. The vertices v1 and vg
form at least 2 nigh pairs and at most two of their neighbouring vertices form 3 nigh pairs. The minimal number of nigh
pairs is therefore 12 · (2017 · 4 + 2 · 3 + 2 · 2) = 4039.
1 point.
Assume g 6 3. If g = 2, all the pairs of vertices are nigh as their labels are all in the set [x1 , x2 ]. If g = 3, each vertex
forms a nigh pair with at least one of v1 or v3 . Without loss of generality, more than half of
the remaining vertices form
nigh pairs with v1 . But then all of those vertices are nigh as well, so there are at least 1015
2
> 4039 nigh pairs, making
this case suboptimal as well.
0 points.
Notes on marking:
• Points from different solutions are not additive. Student’s score should be the maximum of points scored over all
solutions.
• Miscounting the number d in the optimal example or making a similar minor mistake in that part should be awarded
2 points out of possible 3 for that part of the solution.
• In the first solution, just stating the idea of induction is worth no points on its own.
• In the third solution, dealing with the case g 6 3 is worth no points on its own. However, a contestant who doesn’t
comment these cases can get at most 9 points for their solution.

2
Problem 2. Let ABC be a triangle and let D, E and F be the midpoints of sides BC, CA and AB, respectively.
Let X 6= A be the intersection of AD with the circumcircle of ABC. Let Ω be the circle through D and X,
tangent to the circumcircle of ABC. Let Y and Z be the intersections of the tangent to Ω at D with the
perpendicular bisectors of segments DE and DF , respectively. Let P be the intersection of Y E and ZF and
let G be the centroid of ABC. Show that the tangents at B and C to the circumcircle of ABC and the line
P G are concurrent.
(Jakob Jurij Snoj)

E
F

Z
D C

B
Y
X

First Solution. Due to the collinearity of A, X and D, there is a homothety at X sending the circumcircle of ABC to
Ω. This homothety also sends the tangent at A to the circumcircle of ABC to the tangent at D of Ω.
1 point.
The homothety in G with ratio − 12sends the circumcircle of ABC to its nine-point circle and the tangent at A to the
circumcircle to the tangent at D to the nine-point circle. These tangents are therefore parallel. It follows that Ω and
the nine-point circle of ABC share a tangent at D.
2 points.
As Y and Z lie on the perpendicular bisectors of DE and DF , respectively, it follows that |Y D| = |Y E| and Y E is also
tangent to the nine-point circle of ABC - similarly, ZF is also tangent to this circle. We conclude that the nine-point
circle of ABC is the incircle of P Y Z.
3 points.
Finally, the homothety at G sending the circumcircle of ABC to its nine-point circle sends the tangents through A, B
and C to the circumcircle of ABC respectively to the tangents at D, E and F to the nine-point circle of ABC. It,
therefore, sends the intersection of tangents at B and C to the circumcircle of ABC to the point P , thus proving the
desired collinearity.
4 points.

3
 A P

F E
G

Y
T
B D C
Ω

X
Z

Second Solution. As in the previous solution, we prove that the tangent at D to Ω is parallel to the tangent at A to
the circumcircle of ABC.
1 point.
Let ^BAC = α, ^ABC = β, ^ACB = γ. Let Q be the intersection of the tangents at B and C to the circumcircle of
ABC. It now suffices to prove that G lies on P Q.
0 points.
Let T be the point on BC such that the line T A is tangent to the circumcircle of ABC at A. By the tangent-chord
theorem, we have ^T AB = γ, which implies ^AT B = β − γ. Since ZY k T A and ^Y DC = ^AT B = β − γ.
1 point.
By definion of Y , we have |EY | = |DY |. Since DE is a midline of ABC, we have ^EDC = ^ABC = β. Thus
^DEY = ^EDY = ^EDC − ^Y DC = β − (β − γ) = γ.
2 points.
Furthermore, note that ^BED + ^DBE = ^EDC = β, and ^QBC = α, by the tangent-chord theorem. This implies
^QBE + ^BEY = (^QBC + ^DBE) + (^BED + ^DEY ) = α + β + γ = 180◦
so QB k P E. By analogous reasoning we conclude that QC k P F .
2 points.
Since EF is a midline of ABC, we have that the sides EP , P F , EF of triangle P EF are parallel to the sides QB, QC,
BC of triangle QBC, respectively. Those triangles are not congruent because EF = BC 2
, so there exists a homothety
which maps P EF to QBC. The centre of the homothety is the intersection of BE, CF , and P Q, which implies that G
lies on P Q and we are done.
4 points.
Notes on marking:
• In the second solution, once we obtain ^DEY = ^EDY = γ, we can conclude that Y E is tangent to the circumcircle
of DEF and finish as in the first solution.
• The final 4 points in either solution can only be awarded if the student correctly proves the other steps of the
problem. Otherwise, a contestant can only obtain up to 2 points for this part of the solution.
• No points are deducted if the student fails to argue that 4P Y Z and the triangle formed by the tangents through
A, B and C to the circumcircle of ABC are not congruent.
• Analytic approaches are only awarded points if their results are correctly interpreted by geometric means.

4
Problem 3. Let N denote the set of all positive integers. Find all functions f : N → N such that

x2 − y 2 + 2y(f (x) + f (y))

is a square of an integer for all positive integers x and y.

(Ivan Novak)

First Solution. Throughout the solution, let P (a, b) denote the assertion "a2 − b2 + 2b(f (a) + f (b)) is a perfect square".

Let p be a prime. Then P (p, p) implies 4pf (p) is a perfect square, which implies p | f (p).
1 point.
Let y be any positive integer, and let p be any prime. P (p, y) implies p + 2yf (p) + 2yf (y) − y is a perfect square.
2 2

Taking the assertion modulo p, it follows that 2yf (y) − y 2 is a quadratic residue modulo p. It is a well known fact that
if a positive integer is a quadratic residue modulo all primes, it must be a perfect square. We conclude that 2yf (y) − y 2
is a perfect square for all y ∈ N.
3 points.
Define g(y) to be 2yf (y) − y2 .
p

P (1, y) implies 1 − y 2 + 2yf (1) + 2yf (y) = g(y)2 + 2yf (1) + 1 is a perfect square, and since

g(y)2 + 2yf (1) + 1 > g(y)2 ,

we have a following chain of inequalities:

g(y)2 + 2yf (1) + 1 > (g(y) + 1)2 =⇒ 2yf (1) + 1 > 2g(y) + 1 =⇒

f (1)2 + 1
yf (1) > g(y) =⇒ y 2 f (1)2 > 2yf (y) − y 2 =⇒ y > f (y).
2
1 point.
2
Since p | f (p) and f (p)
p
6 f (1)2 +1 for any prime p, it follows from Pigeonhole principle that there exists a positive integer
a such that f (p) = ap for infinitely many primes p.
2 points.
Let p be any prime such that f (p) = ap and let n be any positive integer. P (p, n) implies

p2 − n2 + 2nap + 2nf (n) = (p + na)2 − n2 a2 − n2 + 2nf (n)

is a perfect square.

However, this means that 2nf (n) − n2 − n2 a2 can be written as a difference of squares in infinitely many ways, which is
2
only possible if it equals 0. Thus, 2nf (n) = n2 a2 + n2 , or, equivalently, f (n) = n(a 2+1) for all n ∈ N. This also implies
a2 +1
2
= a, which gives us a = 1.
Therefore, f (n) = n for all n ∈ N. It can be easily checked that the identity function is indeed a solution.
3 points.

Second Solution. Similarly as in the first solution, we conclude that p | f (p). Also, from P (4p, 4p), we also conclude
that p | f (4p) for every odd prime p.
1 point.
Fix an odd prime number p, and let A = f (p)
p
and B = f (4p)
p
.
Now, from P (4p, p), we have that 15p2 + 2p (A + B) is a square, which means 15 + 2(A + B) is also a square. Multiplying
2

by 4 yields the fact that 60 + 8(A + B) is a square.

From P (p, 4p), we have that −15p2 + 8p2 (A + B) is a square, which means −15 + 8(A + B) is a square.
We then have
75 = 60 + 8(A + B) − (−15 + 8(A + B)).
Since there are finitely many ways to write 75 as a difference of two squares, we conclude that A + B can take only
finitely many different values as p ranges over all primes. The same is true for A.
4 points.
Therefore, by Pigeonhole principle, there exists a positive integer a such that f (p) = ap for infinitely many primes p.
2 points.

5
The problem can now be finished the same way as in the first solution.
3 points.
Notes on marking:
• Both solutions follow a similar structure. In the first part, a constant a is found such that f (p) = ap for infinitely
many primes p, and in the second part the problem is completed using the first fact. The points from different
solutions are not additive.
• In the first solution, it is not expected from the students to prove the key lemma about quadratic residues that is
used; it suffices to state it. On the other hand, merely stating the lemma is not worth any points on its own.
• If a student solves the problem under the assumption that f (p) = ap for infinitely many p, but they don’t prove
this fact, they can get at most 1 point out of the last 3 points.
• In the second solution, if a student discusses P (1, 4) and P (4, 1) similarly to how P (p, 4p) and P (4p, p) are discussed,
and concludes that f (1) and f (4) can take on finitely many different values, they should get 1 point out of 4 for
that part of the solution.

6
Problem 4. Find all positive integers d for which there exist polynomials P (x) and Q(x) with real coefficients
such that degree of P equals d and
P (x)2 + 1 = (x2 + 1)Q(x)2 .

(Ivan Novak)

First Solution.
P (x)2 + 1 = (x2 + 1)Q(x)2 (1)
Let P and Q be polynomials satisfying the conditions. Note that the degree of the left hand side in (1) the equality is
2d, and the degree of the right hand side is 2 + 2 deg Q, which implies deg Q = d − 1.
Suppose that r is a real root of Q. Then P (r)2 + 1 = 0, which is clearly impossible. We conclude that Q has no real
roots. Since Q has real coefficients, we conclude that Q has even degree since its roots must come in conjugate pairs.
Thus, d must be odd.
1 point.
√
Now we prove that for any odd
√ d, there exist polynomials satisfying the conditions. Let R[x, + 1] be the set of all
x2
functions of the form A + B x2 + 1, where A and B are polynomials with real coefficients. Note that each
√ element of
R[x, x2 + 1] can be uniquely associated with a pair of polynomials (A, B). Consider a function n : R[x, x2 + 1] → R
defined by p
n(A + B x2 + 1) = A2 − (x2 + 1)B 2
for all real polynomials A and B. Note that the equality (1) is equivalent to the equality
p
n(P + x2 + 1Q) = −1.

1 point.
Note that
p p p
n((A+ x2 + 1B)(C+ x2 + 1D)) = n(AC+(x2 +1)BD+ x2 + 1(AD+BC)) = (AC+(x2 +1)BD)2 −(x2 +1)(AD+BC)2 .

On the other hand,

p p
n(A + x2 + 1B)n(C + x2 + 1D) = (A2 − (x2 + 1)B 2 )(C 2 − (x2 + 1)D2 ).

It can easily be checked that the two expressions are equal. Hence, the function n is multiplicative.
1 point.
√
Note that n(x + x2 + 1) = −1.
1 point.
√ √ √
Then, using the multiplicative property, n((x + x2 + 1) ) = −1 as well. Let (x + x2 + 1) = P + x2 + 1Q for some
d d

polynomials P and Q. By binomial theorem, we have

! !
p p d
X d d−j 2 d−j X d j 2 d−1−j p
2
P + x + 1Q = (x + x + 1) =2 x (x + 1) 2 + x (x + 1) 2 x2 + 1.
j j even
j
j odd

It’s now easy to see that P has degree d, since it is a sum of polynomials which have degree d and positive leading
coefficients, and P and Q satisfy the starting equality. Thus, all odd positive integers are solutions.
6 points.

Second Solution.
P (x)2 + 1 = (x2 + 1)Q(x)2 (2)
Similarly as in the first solution, we conclude that d needs to be odd.
1 point.
Let us now prove that for every odd d such polynomials P (x) and Q(x) exist. Fix an odd positive integer d. Observing
the roots of polynomials P (x) and Q(x), we can easily see from (2) that P (x) and Q(x) don’t have a common root.
Differentiating (2), we get :

P 0 (x)P (x) = x · Q(x)2 + (x2 + 1) · Q0 (x)Q(x) = Q(x) x · Q(x) + (x2 + 1) · Q0 (x) . (3)

Since P (x) and Q(x) don’t have common roots, from (3) we conclude that Q(x) must divide P 0 (x). Since they have the
same degree, there must exist a real number u such that uP 0 (x) = Q(x).
1 point.
Comparing coefficients in (2), we get that u must be 1
d
or − d1 .

7
 1 point.
We’ll take u = 1
d
. Plugging in Q(x) = P (x)/d in (3), we get
0

 
1 1 1
P (x) = x · P 0 (x) + (x2 + 1)P 00 (x) . (4)
d d d

We will now find all polynomials P of degree d which satisfy (4). Note that, by multiplying both sides with P 0 (x) and
0 2
integrating, each of these polynomials satisfies the equation P (x)2 + C = (x2 + 1) (P d(x)) 2 for some C ∈ R.
Denote P (x) = i=0 ai x . Then P (x) = i=1 iai x and P (x) = i=2 i(i − 1)ai x . Writing out the coefficients in
Pd i 0 P d i−1 00 P d i−2

(4), we get
d d d
!
X i 1 X 2 i
X i−2
ai x = 2 i ai x + i(i − 1)ai x
i=0
d i=1 i=2

Comparing the coefficients of xk for all k on the left hand side and the right hand side of the above equation for k > 0,
we get:
1
ad−1 = 2 (d − 1)2 ad−1 which implies ad−1 = 0,


d
and also
1
k2 ak + (k + 2)(k + 1)ak+2 which can be rewritten as


ak =
d2
2 2
d −k
ak+2 = · ak for all 0 6 k 6 d − 2.
(k + 2)(k + 1)

1 point.
From here, we now have that ak = 0 for all even 0 6 k 6 d − 2 and that for all odd 0 6 k 6 d − 2, we have ak = qk a1
for some nonzero real coefficient qk which is uniquely determined by the above recursion.
1 point.
It’s easy to see that any such choice of coefficients (ak )k with a1 6= 0 gives a solution to (4) which has degree d.
0 2
As we’ve already said, any solution to (4) is a solution to P (x)2 + C = (x2 + 1) (P d(x))
2 for some C ∈ R. Considering the
coefficient alongside x0 in both sides and noting a0 = 0, we get C = a21 /d2 . Thus, taking a solution with a1 = d, we get
0 2
the solution to P (x)2 + 1 = (x2 + 1) (P d(x))
2 , which proves that every odd d is a solution to the problem.
5 points.

Third Solution.
P (x)2 + 1 = (x2 + 1)Q(x)2 (5)
Similarly as in the first solution, we conclude that d needs to be odd.
1 point.
Note that d = 1 is a solution, taking P (x) = x and Q(x) = 1. Henceforth assume d > 3.
From x2 + 1 | P (x)2 + 1 we get x2 + 1 | P (x)2 − x2 =⇒ x2 + 1 | (P (x) − x)(P (x) + x). It is not hard to see that
the irreducible polynomial x2 + 1 divides exactly one of the two factors. We can replace P by −P , so without loss of
generality it is safe to assume that P (x) = A(x)(x2 + 1) + x for some real polynomial A. If we put this in (5), we obtain

(A(x) · x + 1)2 + A(x)2 = Q(x)2 .

We will find real polynomials α and β such that A(x) = 2α(x)β(x), Q(x) = α(x)2 + β(x)2 , xA(x) + 1 = α(x)2 − β(x)2
and deg α + deg β = d − 2. Note that then A(x) and Q(x) satisfy the conditions due to the identities for Pythagorean
triples.
2 points.
We thus need to find solutions to the equation

2xα(x)β(x) + 1 = α(x)2 − β(x)2 (6)

where α, β ∈ R[x] are polynomials with real coefficients. Notice that (α, β) = (1, −2x) is a pair of solutions.
1 point.

8
If we look at (6) as a quadratic equation in α we have

α2 − α · 2xβ + 1 − β 2 = 0.

Roots α1 , α2 must then satisfy α1 + α2 = 2xβ. It is now easily verified that if (α, β) is a pair of polynomials which satisfy
(6), then (β, 2xβ − α) is another such pair.
1 point.
Thus starting with solution (α0 , β0 ) = (1, −2x), we can recursively generate a sequence of solutions

(αi+1 , βi+1 ) = (βi , 2xβi − αi ).

The degrees of (αi , βi )i>0 now follow the pattern

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

More precisely, deg αi = i, deg βi = i + 1 for all i > 0.

But then, if d = 2i+1 for some i > 0, the pair (αi , βi ) gives a pair (A, Q) = (2αβ, α2 +β 2 ) such that (xA(x)+1)2 +A(x)2 =
Q(x)2 and deg A = i + (i − 1) = d − 2. Then, taking P (x) = (x2 + 1)A(x) + x yields a pair (P, Q) satisfying the original
equation such that deg P = d. We conclude that every odd d is a solution.
5 points.

Notes on marking:
• Points from different marking schemes are not additive.