PROBLEMS FOR THE 8th INTERNATIONAL TOURNAMENT OF YOUNG

MATHEMATICIANS

JULY 2016, SAINT PETERSBURG (RUSSIA)

Contents
Notation 1
1. Rankings of Teams 1
2. Cutting Segments 2
3. Loose Number Sets 3
4. Suprema of Integer Polynomials 4
5. Composite Polygons 4
6. Different Means 5
7. A Colouring Game 6
8. Functional Equations 7
9. Mathematics of Electric Circuits 8
10. Rich Sequences 9

Keywords: 1. combinatorics, probability, series – 2. geometry, intersection – 3. additive

number theory – 4. integer polynomial, approximation – 5. combinatorial geometry, partition
– 6. mean, limit – 7. game, strategy – 8. analysis, functional equation – 9. linear algebra,
circuit, Cayley graph – 10. sequence, Hankel matrix

Notation
N = {1, 2, 3, . . .} set of natural numbers (positive integers)
Zn = {0, 1, . . . , n − 1} ring of residue classes modulo n
Z, Q, R, C sets of integer, rational, real and complex numbers
R+ set of positive real numbers
Fq finite field of size q = pk
M ⊗N mean-product of two functions M and N
gcd(x1 , . . . , xn ) greatest common divisor of x1 , . . . , xn
P [.] and E [.] probability and expectation
det(A) or |A| determinant of a matrix A
|x| absolute value of x
#M or |M | cardinality of a set M
sup f (x) supremum of a set {f (x) | x ∈ M }
x∈M

1. Rankings of Teams
A mathematics tournament is being held in Saint Petersburg with n ∈ N participating teams
T1 , . . . , Tn . In the end of the tournament, each team will be ranked according to its rating
(teams can have the same rating). The teams with the highest rating will have rank 1, the
teams with the second highest rating will have rank 2, and so on. Denote by Rn the set of all
possible rankings. In other words,
  
Rn = (r1 , . . . , rn ) ∈ N
n 
for any integer 1 ≤ t ≤ max(rk ), there exists i such that ri = t ,
k


Date: March 22, 2016.

1
2 JULY 2016, SAINT PETERSBURG (RUSSIA)

where ri is the rank of the team Ti .

1. Let an be the number of elements in Rn and put a0 = 1. Find a recurrent formula for an in
terms of a0 , a1 , ..., an−1 .
2. Let σ = (s1 , . . . , sn ) be a permutation of {1, . . . , n}. Find the cardinality of the subset
P(σ) := {(r1 , . . . , rn ) ∈ Rn | for any 1 ≤ i < j ≤ n, one has ri < rj if and only if si < sj } .

3. Show that for every real number x with |x| sufficiently small, one has
∞
X an n 1
x = .
n=0 n! 2 − ex

4. Denote by Mn the sum of all possible last ranks, that is

X
Mn = max(r1 , . . . , rn )
(r1 ,...,rn )∈Rn
∞
P Mn n
and let M0 = 0. Compute the series x .
n!
n=0

5. Let ρn = (X1 , . . . , Xn ) be a random element of Rn chosen uniformly. Study the behaviour of

E [max(ρn )], that is the expected value of the rank of a team with the lowest rating, as n → ∞.
6. Study the behaviour of E [X1 ], that is the expected value of the rank of the team T1 , as
n → ∞.
7. Denote by ft (ρn ) the number of teams with rank t: ft (ρn ) = # {1 ≤ i ≤ n | Xi = t }.
a) Let k be a positive integer. Find lim P [f1 (ρn ) = k], that is the limit of the probability
n→∞
that exactly k teams have rank 1.
b) How does E [f1 (ρn )] behave as n → ∞ (the expected number of teams ranked first)?
c) How does P [X1 = 1] behave as n → ∞ (the probability that the team T1 is first)?

8. Same questions for t = 2, 3, etc.

9. Denote by (X10 , . . . , Xn0 ) the coordinates of ρn ordered in non-increasing order. Let s ∈]0, 1[
be a fixed real number. Study the behaviour of the random variable Xbsnc 0 as n → ∞, where
byc denotes the integer part of a real number y.
10. Study other properties of ρn . Suggest and investigate additional directions of research.

2. Cutting Segments
Let A and B be two bounded domains in the real plane. We suppose that either B ⊆ A or
B ∩ A = ∅. For positive integers m and n, let A1 , . . . , Am and B1 , . . . , Bn be some points on
the boundaries of A and B respectively. We say that a segment Ai Bj is cutting if it contains
an inner point of the domain B (in other words, it is dividing B into parts, at least two of
which have nonzero area). Denote by CutB (A1 . . . Am , B1 . . . Bn ) the set of cutting segments.
We would like to find all possible values of the cardinality of this set, and in particular its
minimum and maximum, in the following cases:
(i) the points (Ai ) and (Bj ) are fixed on the respective boundaries, and the domain B can
float (move and rotate) strictly inside the domain A;
(ii) the points (Ai ) and (Bj ) are fixed on the respective boundaries, and the domain B can
float (move and rotate) strictly outside the domain A;
(iii) the domains A and B are fixed as well as the points (Ai ) on the boundary of A, and
the points (Bj ) can float on the boundary of B.
 PROBLEMS FOR THE 8th INTERNATIONAL TOURNAMENT OF YOUNG MATHEMATICIANS 3

1. Let A = A1 . . . Am be a regular m-gon and B = B1 . . . Bn be a regular n-gon.

a) What area may the polygon B have so that it could be placed strictly inside A?
b) Suppose B lies strictly inside A. What minimum and maximum number of cutting
segments can there be (cf. Figure 1)?
c) Determine CutB (A1 . . . Am , B1 . . . Bn ) in the case (i), depending on the size and the
position of B within A.
d) Consider the case (ii).
e) Consider the case (iii).

A2 A2

B2
B3 B1
B1 B3

B2
A1 A3 A1 A3

Figure 1. Two examples where A and B are triangles in the case (i), and
|CutB (A1 A2 A3 , B1 B2 B3 )| is equal to 0 and 3.

2. Study the problem for other convex polygons.

3. Study the problem for non-convex polygons. Start with A convex and B non-convex.
4. Study the problem when A and B are circles. Start with the case (i) and suppose that the
points divide the respective circumferences into arcs of equal length (equidistribution).
5. Study the problem when A and B are ellipses.
6. Generalise to higher dimensions N ≥ 3. In particular, study the problem when A and B are
a) N -dimensional simplexes;
b) N -dimensional balls.

7. Suggest and investigate other directions of research.

3. Loose Number Sets

We say that a set of integers is loose, if it does not contain four different numbers a < b <
c < d such that a + d = b + c. Denote by P (n) the maximal number of elements in a loose
subset of the set {1, 2, . . . , n}. We are interested in calculating P (n) for large n or finding lower
and upper bounds.
1. Compute P (n) for small n up to 10.
P (n)
2. Prove that lim n = 0.
n→∞

3. Try to find best possible upper and lower bounds for P (100) and P (2016).
lim Pn(n)
4. Is it true that n→∞ ε = 0 for

a) ε = 1/2,
4 JULY 2016, SAINT PETERSBURG (RUSSIA)

b) ε = 1/3,
c) any ε > 0?

5. We call a sequence very loose if it contains neither four different numbers a < b < c < d
such that a + d = b + c, nor triples a < b < c such that a + c = 2b. Denote by Q(n) the
maximal number of elements in a very loose subset of the set {1, 2, . . . , n}. How the answers
to the above questions change for Q(n)?
6. Suggest and investigate generalisations.
Note. If you use computer programs, then please provide the full source codes and all necessary
instructions so that other participants and the jury members are able to reproduce your results.

4. Suprema of Integer Polynomials

Let D(a, r) := {z ∈ C | | z − a |≤ r } be the closed disk of radius r ∈ R+ centred at a ∈ C
in the complex plane. Denote by c = c(a, r) the largest real number such that for any nonzero
polynomial P ∈ Z[z] with integer coefficients, one has
sup |P (z)| ≥ cn ,
z∈D(a,r)

where n is the degree of P .

1. Find c( 21 , 21 ) or give lower and upper bounds.
2. Find c(a, r) or give lower and upper bounds, when a ∈ Q.
3. Estimate c(a, r) when a is a quadratic number, that is ua2 + va + w = 0 for some u, v, w ∈ Z
and u 6= 0.
4. Study the general case.
5. Suggest and study additional directions of research.

5. Composite Polygons
Let ` be a line in the real plane. We will say that a polygon P is divisible by ` if there exist
two finite families of lines F1 and F2 in the plane such that
– the lines in F1 are parallel to `,
– the lines in F2 are orthogonal to `,
– the polygon P is divided by F1 and F2 into parts – each part is either a rectangular
triangle or a rectangle.
We will also say that P is divisible by a segment s if it is divisible by the line containing s.
Finally, a polygon will be called composite if it is divisible by at least one line.
1. Given a triangle, describe all lines by which it is divisible.
2. Given a convex quadrilateral, describe all lines by which it is divisible.
3. Give a criterion for a non-convex quadrilateral to be composite.
4. A polygon is obtuse if all its angles are greater than π/2. Does there exist an obtuse convex
pentagon which is
a) not divisible by at least one of its sides?
b) not divisible by all its sides?
c) not composite?
 PROBLEMS FOR THE 8th INTERNATIONAL TOURNAMENT OF YOUNG MATHEMATICIANS 5

A D

Figure 2. A convex quadrilateral ABCD divisible by its side AD.

5. In how many parts can a composite obtuse convex pentagon be divided by a line `, if
a) ` contains a side?
b) ` is neither parallel nor orthogonal to any of the sides?

6. Study the divisibility of non-convex and non-obtuse pentagons.

7. Consider other polygons. Describe composite n-gons when n > 5.
8. Suggest and study additional directions of research. One could also try to extend the notion
of divisibility to α-divisibility where α is the angle between families F1 and F2 .

6. Different Means
Let M and N be two real-valued functions defined in the domain R+ × R+ . Given any
positive real numbers a and b, one can construct two sequences (an ) and (bn ) satisfying the
following recurrence relations
an = M (an−1 , bn−1 ) and bn = N (an−1 , bn−1 ),
where a0 = a, b0 = b and n ∈ N. If the limits n→∞
lim an and n→∞
lim bn exist and coincide for all
a, b ∈ R+ , then the function
M ⊗ N (a, b) := lim an = lim bn
n→∞ n→∞
is well-defined in R+ × R+ . We will call it the mean-product of M and N .
1. Consider two classical functions – the arithmetic mean A(a, b) = a+b2 and the harmonic mean
2
H(a, b) = 1 + 1 . For any a, b ∈ R + and n ∈ N , take a n = A(a n−1 n−1 ) and bn = H(an−1 , bn−1 )
, b
a b
with a0 = a and b0 = b.
Prove that the sequences (an ) and (bn ) converge to the same limit. Determine A ⊗ H. Is it
a symmetric function, that is A ⊗ H(a, b) = A ⊗ H(b, a)?
√
2. Let G(a, b) = ab be the geometric mean.
a) Show that the functions A ⊗ G and H ⊗ G are well-defined in R+ × R+ .
b) Prove that
−1
Z∞

1
A ⊗ G(a, b) = π 2 q dx .
0 (x2 + a2 )(x2 + b2 )
c) Find a formula for H ⊗ G.

3. Consider a Hoelder mean s

p ap + b p
Hp (a, b) = ,
2
6 JULY 2016, SAINT PETERSBURG (RUSSIA)

where p 6= 0 is an arbitrary real number. Investigate the following functions (if they are
well-defined, then find a formula or a Taylor expansion):
a) Hp ⊗ G;
b) Hp ⊗ Hq , for any real nonzero p and q; q
p t·ap +(2−t)·bp
c) Hp [λ] ⊗ Hq [µ], where 0 < λ, µ < 2 and Hp [t](a, b) = 2 .
4. One can instead construct the sequences (an ) and (bn ) as follows:
an = M (an−1 , bn−1 ) and bn = N (an , bn−1 ),
where n ∈ N, a0 = a and b0 = b. If the limits lim an and lim bn exist and coincide for all
n→∞ n→∞
positive a and b, then one can define the Archimedes mean-product of M and N ,
M ⊗A N (a, b) := n→∞
lim an = n→∞
lim bn .
Investigate the following functions (if well-defined, find a formula or a Taylor expansion):
a) Hp ⊗A Hp ;
b) Hp ⊗A G;
c) Hp ⊗A Hq ;
d) Hp [λ] ⊗A Hq [µ], where 0 < λ, µ < 2.
p p 1 1
ap −bp
  
a +b p p−1
5. Consider other means: Lehmer means ap−1 +bp−1
, Stolarsky means p(a−b) , weighted
versions of means and their convex combinations.
6. Find criteria for a mean-product to be continuous, symmetric, smooth, analytic, homoge-
neous, etc.
7. Try to generalise results to means of more than two variables.

7. A Colouring Game
Alice and Bob play a game on a circle of circumference 1. Initially, the circle is completely
white. In each step, one player picks a real number t ∈ [0, 1] and the other chooses an arc of
length t. All white points in the chosen arc become black and all black points become white.
The players pick numbers alternately, starting with Alice. The game stops after n steps, where
n ∈ N is given in advance. In the end, Alice’s gain, denoted by GA , is the proportion of points
which are white. Bob’s gain, denoted by GB , is the proportion of points which are black.

Figure 3. An example with n = 3. Alice picks t = 1/6 and Bob chooses an arc
on the right. Then he picks t = 1/2 and Alice chooses an arc on the top. Finally
Alice picks t = 1 and Bob chooses the entire circle. Here GA = GB = 12 .

1. Depending on n, does there exist a strategy allowing Alice to ensure a gain at least 12 ? And
what about Bob?
2. What maximal gain can Alice ensure? Start with n = 2, 3, 4.
In the next two questions, the game is on the segment [0, 1] which is initially white. In each
step, a player picks a real 0 ≤ t ≤ 1 and the other player chooses a subinterval of length t
 PROBLEMS FOR THE 8th INTERNATIONAL TOURNAMENT OF YOUNG MATHEMATICIANS 7

within [0, 1]. The white points in the subinterval become black and vice-versa. The gains are
defined in a similar manner.

0 1 0 1 0 1 0 1

Figure 4. An example with n = 3. Alice picks t = 1/3 and Bob chooses a

subinterval on the left. Then he picks t = 1/2 and Alice chooses a subinterval on
the right. Finally, Alice chooses t = 1 and Bob chooses the entire segment [0, 1].
As a result GA = GB = 21 .

3. What is the maximal gain that Alice can ensure (as a function of n)?
4. Find the maximal gains that Alice and Bob can ensure if they are only allowed to pick real
numbers t which are
a) of the form k1 with k ∈ N ;
b) not greater than 12 .
5. Now, the game is on the segment [0, 1] with 3 colours: white, blue and red. All white points
in a chosen subinterval become blue, all blue points become red and all red points become
white. Bob’s gain is the proportion of white and blue points, Alice’s gain is the proportion of
red points. Find the maximal gain that Alice can ensure after n steps.
6. Suggest and study additional directions of research.

8. Functional Equations
Let ]a, b[ be a non-empty open interval of the real line with a, b ∈ R ∪ {−∞, +∞}. We would
like to investigate equations of the form
f (f (x)) = g(x),
where f and g are two real-valued functions defined on the interval ]a, b[.
1. Given that g(x) = ex , does the equation have a solution f which is
a) continuous?
b) differentiable (twice differentiable, etc.)?
c) analytic?
d) presentable as a power series convergent in ]a, b[?
Start with the case ]a, b[= R.
2. Analoguous questions when
a) g(x) = x2 + px + q with p, q ∈ R;
b) g(x) is a polynomial of degree n > 2.
3. Solve the equation f (f (x)) = f 0 (x) for a continuously differentiable function f : ]a, b[→ R.
4. Investigate the equations f (f (x)) = ab f (t)dt and f (f (x)) = f (t)dt, where f : [a, b] → R
R Rx
a
is a) continuous; b) Riemann integrable; c) Lebesgue integrable.
5. Let M be a bounded closed interval [a, b] or the entire R. We say that a continuous
function g : M → R is a k-square if there exists a continuous function f : M → R such
that g(x) = f (f (x)) for all x ∈ M . Determine the density of k-squares in the set C(M ) of
continuous functions on M for different topologies.
6. Suggest and study other directions of research.
8 JULY 2016, SAINT PETERSBURG (RUSSIA)

9. Mathematics of Electric Circuits

An electric circuit can be viewed as an undirected weighted graph without loops and with
multiple edges, where the weight of each edge is the resistance of the corresponding circuit
element. Given such graph Γ, we are interested in measuring the resistance Rij between any
pair of vertices vi and vj of the graph. A well-known method consists in connecting a unit
current source between the vertices, and applying Kirchhoff’s circuit laws and Ohm’s law in
order to find the potentials at all vertices (see Figure 5). The resistance Rij will then be the
difference of potentials at vi and vj .

v1 v2
r12

r23
r14

4
r2
1

r34
v4 v3

Figure 5. Connecting a unit current source between the vertices v1 and v3 .

1. Let rij be the resistance (weight) of the edge connecting vertices vi , vj ∈ Γ. We denote
rij = ∞ (or rather 1/rij = 0) if there is no edge between these two vertices. Using Kirchhoff’s
laws define a system of linear equations that allows to compute the resistance Rij between any
two vertices vi , vj in terms of the weights rij .
2. Assume that the graph is built from the edges of a polyhedron in an N -dimensional Euclidean
space and the resistance of any edge is exactly 1 ohm. Compute the resistance between opposite
vertices for the following polyhedrons:
a) a regular polygon in the real plane;
b) a cube in the 3-dimensional Euclidean space;
c) an octahedron in the 3-dimensional Euclidean space;
d) an icosahedron in the 3-dimensional Euclidean space;
e) a cube in the N -dimensional Euclidean space.
3. Assume that you can vary the resistance of each edge in a given interval [a, b]. What are
the maximal and minimal resistances that you can get between two opposite vertices for the
above graphs?
4. Let Sn be the symmetric group of degree n, and let T ⊂ Sn be the set of transpositions
(1 2), (2 3), . . . , (n−1 n). The Cayley graph associated with (Sn , T ) is defined as the undirected
graph having n! vertices corresponding to the permutations in Sn and such that there is an
edge between vertices σ, τ ∈ Sn if and only if τ σ −1 ∈ T .
Assuming that each edge of this graph has resistance 1 ohm, compute the resistance between
the following two “opposite” permutations:
! !
1 2 3 ··· n − 1 n 1 2 3 ··· n − 1 n
and .
1 2 3 ··· n − 1 n n n − 1 n − 2 ··· 2 1

5. Consider other groups G and their generators.

6. Suggest and study additional questions.
 PROBLEMS FOR THE 8th INTERNATIONAL TOURNAMENT OF YOUNG MATHEMATICIANS 9

10. Rich Sequences

Let (ai ) be a sequence over a set S, that is the terms of the sequence are elements of S. An
(i, n)-order Hankel matrix of the sequence is the following n × n matrix:
 
ai ai+1 . . . ai+n−1

 ai+1 ai+2 . . . ai+n 
Hi,n =  .. .. .. .. .

 . . . . 
ai+n−1 ai+n . . . ai+2n−2
An infinite sequence (ai )i≥1 is said to be rich if, for all positive integers i and n, the determi-
nant |Hi,n | is nonzero. A finite sequence (ai )1≤i≤N is rich if the determinant |Hi,n | is nonzero
for all positive integers i and n such that i + 2n − 2 ≤ N .
1. Let S be a subset of Z of cardinality 2.
a) Show that there exist no infinite rich sequences over S.
b) What is the maximum length of a rich sequence over S?
c) Describe all rich sequences when S = {1, 2}.
2. Let S be a subset of Z of cardinality 3.
a) Construct finite rich sequences when S = {1, 2, 3}.
b) Is there an infinite rich sequence over S?
3. Investigate the existence of infinite rich sequences when S ⊂ Z is a subset of cardinality
more than 3.
4. Consider the infinite sequence (ai )i≥1 over the set {1, 2, 3, 4} such that the infinite word
a1 a2 a3 . . . is invariant under the transformation 1 7→ 12, 2 7→ 23, 3 7→ 14, 4 7→ 32. The first
sixteen terms of the sequence are 1, 2, 2, 3, 2, 3, 1, 4, 2, 3, 1, 4, 1, 2, 3, 2. Is the sequence
(ai )i≥1 rich?
5. Investigate the existence of infinite rich sequences and construct finite rich sequences, when
S is a subset of a ring Zm of residues modulo m.
6. Investigate the existence of infinite rich sequences and construct finite rich sequences, when
S is a subset of a finite field Fq of size q = pk , where p is prime.
7. Suggest and study additional questions.

E-mail address: oc@itym.org

URL: http://www.itym.org/