IMC 2018, Blagoevgrad, Bulgaria

Day 1, July 24, 2018

Problem 1. Let (an )∞

n=1 and (bn )n=1 be two sequences of positive numbers. Show that the
∞

following statements are equivalent:

∞ ∞
an cn
(1) There is a sequence of positive numbers such that and both converge;
X X
(cn )∞
n=1
n=1
cn n=1
bn
∞ r
an
(2) converges.
X

n=1
bn
(Proposed by Tomá² Bárta, Charles University, Prague)
Solution. Note that the sum of a series with positive terms can be either nite or +∞, so for
such a series, "converges" is equivalent to "is nite".
Proof for (1) =⇒ (2): By the AM-GM inequality,

r r  
an an c n 1 an c n
= · ≤ + ,
bn c n bn 2 c n bn
so ∞ r ∞ ∞
X an 1 X an 1 X c n
≤ + < +∞.
n=1
bn 2 n=1 cn 2 n=1 bn
∞ r
an
Hence, is nite and therefore convergent.
X

n=1
bn
√
Proof for (2) =⇒ (1): Choose cn = an bn . Then
r
an cn an
= = .
cn bn bn
∞ r ∞ ∞
an an cn
By the condition converges, therefore and converge, too.
X X X

n=1
bn n=1
cn n=1
bn

Problem 2. Does there exist a eld such that its multiplicative group is isomorphic to its
additive group?
(Proposed by Alexandre Chapovalov, New York University, Abu Dhabi)
Solution. There exist no such eld.
Suppose that F is such a eld and g : F ∗ → F + is a group isomorphism. Then g(1) = 0.
Let a = g(−1). Then 2a = 2 · g(−1) = g((−1)2 ) = g(1) = 0; so either a = 0 or char F = 2.
If a = 0 then −1 = g −1 (a) = g −1 (0) = 1; we have char F = 2 in any case.
For every x ∈ F , we have g(x2 ) = 2g(x) = 0 = g(1), so x2 = 1. But this equation has only
one or two solutions. Hence F is the 2-element eld; but its additive and multiplicative groups
have dierent numbers of elements and are not isomorphic.

1
Problem 3. Determine all rational numbers a for which the matrix
 
a −a −1 0
a −a 0 −1
 
1 0 a −a
0 1 a −a

is the square of a matrix with all rational entries.

(Proposed by Daniël Kroes, University of California, San Diego)
Solution. We
 will show that the
 only such number is a = 0.
a −a −1 0
a −a 0 −1
Let A =   and suppose that A = B 2 . It is easy to compute the charac-
1 0 a −a
0 1 a −a
teristic polynomial of A, which is
pA (x) = det(A − xI) = (x2 + 1)2 .

By the Cayley-Hamilton theorem we have pA (B 2 ) = pA (A) = 0.

Let µB (x) be the minimal polynomial of B . The minimal polynomial divides all polynomials
that vanish at B ; in particular µB (x) must be a divisor of the polynomial pA (x2 ) = (x4 + 1)2 .
The polynomial µB (x) has rational coecients and degree at most 4. On the other hand, the
polynomial x4 + 1, being the 8th cyclotomic polynomial, is irreducible in Q[x]. Hence the only
possibility for µB is µB (x) = x4 + 1. Therefore,
A2 + I = µB (B) = 0. (1)

Since we have  
0 0 −2a 2a
0 0 −2a 2a
A2 + I = 
2a −2a 0
,
0
2a −2a 0 0
the relation (1) forces a = 0.
In case a = 0 we have
   2
0 0 −1 0 0 0 0 −1
0 0 0 −1 1
  0 0 0
A=
1 =  ,
0 0 0  0 1 0 0
0 1 0 0 0 0 1 0

hence a = 0 satises the condition.

Problem 4. Find all dierentiable functions f : (0, ∞) → R such that
√ 
f (b) − f (a) = (b − a)f ab0
for all a, b > 0. (2)

(Proposed by Orif Ibrogimov, National University of Uzbekistan)

Solution. First we show that f is innitely many times dierentiable. By substituting a = 21 t
and b = 2t in (2),
f (2t) − f ( 12 t)
f 0 (t) = 3 . (3)
2
t

2
Inductively, if f is k times dierentiable then the right-hand side of (3) is k times dierentiable,
so the f 0 (t) on the left-hand-side is k times dierentiable as well; hence f is k + 1 times
dierentiable.
Now substitute b = eh t and a = e−h t in (2), dierentiate three times with respect to h then
take limits with h → 0:
f (eh t) − f (e−h t) − (eh t − e−h t)f (t) = 0
 3 
∂ h −h h −h

f (e t) − f (e t) − (e t − e t)f (t) = 0
∂h
e3h t3 f 000 (eh t)+3e2h t2 f 00 (eh t)+eh tf 0 (eh t)+e−3h t3 f 000 (e−h t)+3e−2h t2 f 00 (e−h t)+e−h tf 0 (e−h t)−
−(eh t + e−h t)f 0 (t) = 0
2t3 f 000 (t) + 6t2 f 00 (t) = 0
tf 000 (t) + 3f 00 (t) = 0
(t f (t))000 = 0.
Consequently, tf (t) is an at most quadratic polynomial of t, and therefore
C2
f (t) = C1 t + + C3 (4)
t
with some constants C1 , C2 and C3 .
It is easy to verify that all functions of the form (4) satisfy the equation (1).
Problem 5. Let p and q be prime numbers with p < q . Suppose that in a convex polygon
P1 P2 . . . Ppq all angles are equal and the side lengths are distinct positive integers. Prove that
k3 + k
P1 P2 + P2 P3 + · · · + Pk Pk+1 ≥
2
holds for every integer k with 1 ≤ k ≤ p.
(Proposed by Ander Lamaison Vidarte, Berlin Mathematical School, Berlin)
Solution. Place the polygon in the complex plane counterclockwise, so that P2 − P1 is a
positive real number. Let ai = |Pi+2 − Pi+1 |, which is an integer, and dene the polynomial
f (x) = apq−1 xpq−1 + · · · + a1 x + a0 . Let ω = e pq ; then Pi+1 − Pi = ai−1 ω i−1 , so f (ω) = 0.
2πi

The minimal polynomial of ω over Q[x] is the cyclotomic polynomial Φpq (x) = (x p −1)(xq −1) ,
(xpq −1)(x−1)

so Φpq (x) divides f (x). At the same time, Φpq (x) is the greatest common divisor of s(x) =
xpq −1
= Φq (xp ) and t(x) = xxq −1 = Φp (xq ), so by Bézout's identity (for real polynomials), we
pq −1
xp −1
can write f (x) = s(x)u(x) + t(x)v(x), with some polynomials u(x), v(x). These polynomials
can be replaced by u∗ (x) = u(x) + w(x) xx−1 and v ∗ (x) = v(x) − w(x) xx−1 , so without loss of
p −1 q −1

generality we may assume that deg u ≤ p − 1. Since deg a = pq − 1, this forces deg v ≤ q − 1.
Let u(x) = up−1 xp−1 + · · · + u1 x + u0 and v(x) = vq−1 xq−1 + · · · + v1 x + v0 . Denote by (i, j)
the unique integer n ∈ {0, 1, . . . , pq − 1} with n ≡ i (mod p) and n ≡ j (mod q). By the choice
of s and t, we have a(i,j) = ui + vj . Then
k−1 k−1 k−1 k−1
X X 1 XX
P1 P2 + · · · + Pk Pk+1 = a(i,i) = ui + vi = (ui + vj )
i=0 i=0
k i=0 j=0
k−1 X
k−1
1 X (∗) 1
k3 + k
= a(i,j) ≥ 1 + 2 + · · · + k2 =
k i=0 j=0
k 2

where (∗) uses the fact that the numbers (i, j) are pairwise dierent.