Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
[x + 2, y + 2] − [x + 1, y + 1] = [x + 1, y + 1] − [x, y] .
Prove that one of the two numbers x and y divides the other.
(Here [a, b] denotes the least common multiple of a and b.)
Day 1. Problems
Problem 1. Solve the system of equations in real numbers:
(
(x − 1)(y − 1)(z − 1) = xyz − 1 ,
(x − 2)(y − 2)(z − 2) = xyz − 2 .
(Vladimir Bragin)
Problem 2. A convex quadrilateral ABCD is circumscribed about a circle ω. Let
P Q be the diameter of ω perpendicular to AC. Suppose lines BP and DQ intersect
at point X, and lines BQ and DP intersect at point Y . Show that the points X
and Y lie on the line AC. (Géza Kós)
Problem 3. Let k be a positive integer such that p = 8k + 5 is a prime number.
The integers r1 , r2 , . . . , r2k+1 are chosen so that the numbers 0, r14 , r24 , . . . , r2k+1
4
give
pairwise different remainders modulo p. Prove that the product
Y
ri4 + rj4
16i<j62k+1
1
The 3rd Olympiad of Metropolises
Day 2. Problems
Problem 4. Let 1 = d0 < d1 < . . . < dm = 4k be all positive divisors of 4k, where
k is a positive integer. Prove that there exists i ∈ {1, . . . , m} such that di −di−1 = 2.
(Ivan Mitrofanov)
Problem 5. Ann and Max play a game on a 100 × 100 board.
First, Ann writes an integer from 1 to 10 000 in each square of the board so that
each number is used exactly once.
Then Max chooses a square in the leftmost column and places a token on this square.
He makes a number of moves in order to reach the rightmost column. In each move
the token is moved to a square adjacent by side or by vertex. For each visited square
(including the starting one) Max pays Ann the number of coins equal to the number
written in that square.
Max wants to pay as little as possible, whereas Ann wants to write the numbers in
such a way to maximise the amount she will receive. How much money will Max
pay Ann if both players follow their best strategies? (Lev Shabanov)
Problem 6. The incircle of a triangle ABC touches the sides BC and AC at points
D and E, respectively. Suppose P is the point on the shorter arc DE of the incircle
such that ∠AP E = ∠DP B. The segments AP and BP meet the segment DE at
points K and L, respectively. Prove that 2KL = DE. (Dušan Djukić)
1
The 1st International Olympiad of Metropolises
September 2016
Solutions of day 1
Problem 1. Find all positive integers n such that there exist n consecutive positive
integers whose sum is a perfect square. (Pavel Kozhevnikov)
Every year, the government of Optimistica publishes its Annual Report with n eco-
nomic indicators. For each i = 1, . . . , n, the possible values of the i-th indicator are
1, 2, . . . , ai . The Annual Report is said to be optimistic if at least n − 1 indicators
have higher values than in the previous report. Prove that the government can pub-
lish optimistic Annual Reports in an infinitely long sequence.
(Ivan Mitrofanov, Fedor Petrov)
First we replace each ai by a power of 2. For every 1 ≤ i ≤ n, let ki be the
n
1
positive integer that satisfies 2ki ≤ ai < 2ki +1 . Notice that < a2i ≤ 1.
P
2ki
i=1
For every 1 ≤ i ≤ n, we will choose a residue class Ai modulo 2ki in such a way
that the classes A1 , . . . , An are pairwise disjoint. Without loss of generality we
can assume that k1 ≤ k2 ≤ . . . ≤ kn . We choose A1 , A2 , . . . , An in this order.
The residue class A1 can be chosen arbitrarily. Suppose that we have already
chosen the classes A1 , . . . , Ai−1 . In order to find the next class Ai , we require
a residue modulo 2ki which is not used in any of A1 , . . . , Ai−1 . Notice that for
each j < i, the set Aj is the union of 2ki −kj different residue classes modulo
i−1
P ki −kj n
2ki . As < 2ki 2−kj < 2ki , there are unused residues modulo 2ki
P
2
j=1 j=1
which makes it possible to choose the new class Ai .
Now let us turn to the solution of the problem. For every 1 ≤ i ≤ n, we
will use only the first 2ki values of the i-th indicator. In the beginning let all
indicators be equal to 1. In the y-th year, let the i-th indicator drop to 1 if
y ∈ Ai , otherwise let the indicator increase by 1. Notice that the i-th indicator
increases at most 2ki −1 times in a row, then drops to 1, so it never exceeds the
bound 2ki ≤ ai and therefore the values of the indicator form a valid report in
every year. Since the residue classes A1 , . . . , An are pairwise disjoint, at most
one indicator drops in the same year, the reports keep optimistic.
Problem 3. Let A1 A2 . . . An be a cyclic convex polygon whose circumcenter is
strictly in its interior. Let B1 , B2 , . . . , Bn be arbitrary points on the sides A1 A2 ,
A2 A3 , . . . , An A1 , respectively, other than the vertices. Prove that
B1 B2 B2 B3 Bn B1
+ + ... + > 1.
A1 A3 A2 A4 An A2
(Nairi Sedrakyan, David Harutyunyan)
Lemma 2 is proved.
Let us return to the problem. Let R denote the circumradius of the circle
A1 A2 . . . An , let Ri denote the circumradius of Bi Ai+1 Bi+1 (further we suppose
Bi Bi+1
An+1 ≡ A1 , An+2 ≡ A2 , Bn+1 ≡ B1 ). The sine law yields sin ∠Ai+1 = 2Ri ,
Ai Ai+2 Bi Bi+1 2Ri sin ∠Ai+1 Ri
sin ∠Ai+1 = 2R, thus Ai Ai+2 = 2R sin ∠Ai+1 = R .
B1 B2 B2 B3 Bn B1
+ + ... + >1
A1 A3 A2 A4 An A2
m
R1 R2 Rn
+ + ... + >1
R R R
m
R1 + R2 + . . . + Rn > R .
In the triangle Bi Ai+1 Bi+1 no side can be greater than the diameter of the
circumcircle, therefore Bi Ai+1 + Ai+1 Bi+1 ≤ 2Ri + 2Ri = 4Ri and Ri ≥
(Bi Ai+1 + Ai+1 Bi+1 )/4. Hence it suffices to prove that
B 1 A2 + A2 B 2 B2 A3 + A3 B3 B n A1 + A1 B 1 P
R< + + ... + = ,
4 4 4 4
but this follows from Lemma 2.
The 1st International Olympiad of Metropolises
September 2016
Solutions of day 2
Problem 4. A convex quadrilateral ABCD has right angles at A and C. A point E
lies on the extension of the side AD beyond D so that ∠ABE = ∠ADC. The point
K is symmetric to the point C with respect to point A. Prove that ∠ADB = ∠AKE.
(Boyan Obukhov and Fedor Petrov)
The quadrilateral ABCD is inscribed in the circle with diameter BD. Thus
∠ADB = ∠ACB since both angles are subtended by the same arc. So, we
have to prove that the angles BCA and AKE are equal, which in turn is
equivalent to the claim that the lines BC and KE are parallel.
Note that ∠BCD + ∠CDA = ∠BAD + ∠ABE < 180◦ . It implies that the
rays CB and DA have a common point which we denote by F . We have
∠BF A = 90◦ − ∠ADC = 90◦ − ∠ABE = ∠BEA. So BA is the altitude of the
isosceles triangle F BE, this yields F A = AE. On the other hand CA = AK.
So, the diagonals of the quadrilateral F CEK have a common midpoint, i.e.,
F CEK is a parallelogram. Therefore the lines F C and KE are indeed parallel
as desired.
Problem 5. Let r(x) be a polynomial of odd degree with real coefficients. Prove
that there exist only finitely many pairs of polynomials p(x) and q(x) with real
coefficients satisfying the equation (p(x))3 + q(x2 ) = r(x). (Fedor Petrov)
λ30 − λ3 2
λb2 − λ0 b20 = ta0 (t) .
3
Dividing by λ0 and factorizing the LHS as a difference of squares (which is
possible in real numbers since λ and λ0 have the same sign) we see that the
p λ3 −λ3
pair of polynomials λ/λ0 b(t) ± b0 (t) have the form f (t), 03λ0 g(t) with
f (t) · g(t) = ta20 (t). Again we may consider the case when f (t) and g(t) are
fixed up to a constant factor: f (t) = τ f0 (t), g(t) = τ −1 g0 (t). We get
λ30 − λ3 3
−1 λ0 − λ
3
2b0 (t) = f (t) − g(t) = τ f0 (t) − τ g0 (t) .
3λ0 3λ0
If this happens for two different pairs of values (τ, λ) and (τ 0 , λ0 ), we may take
the difference:
3 3 3 0 3
0 −1 λ0 − λ 0 −1 λ0 − (λ )
0 = (τ − τ )f0 (t) − τ − (τ ) g0 (t) . (1)
3λ0 3λ0
If τ 6= τ 0 , it follows that f0 (t) and g0 (t) are proportional; but this is impossi-
ble, since their product f0 (t) · g0 (t) = ta20 (t) has odd degree. Otherwise, the
coefficient of f (t) in (1) is zero, hence coefficient of g(t) is also zero, from which
we obtain (λ0 )3 = λ3 . It means that τ and λ are fixed, hence f (t) and g(t)
are fixed, and there is at most one solution. Since on each step we diverged
into finite number of cases, there is no more than a finite number of solutions
totally.
Problem 6. In a country with n cities, some pairs of cities are connected by one-
way flights operated by one of two companies A and B. Two cities can be connected
by more than one flight in either direction. An AB-word w is called implementable
if there is a sequence of connected flights whose companies’ names form the word
w. Given that every AB-word of length 2n is implementable, prove that every finite
AB-word is implementable. (An AB-word of length k is an arbitrary sequence of k
letters A or B; e. g. AABA is a word of length 4.) (Ivan Mitrofanov)
Day 1
F1
B C
L
K
A D
Answer: d = 200.
Example. Join 201 cities in a cycle and consider a pair of neighboring cities
in this cycle. Obviously, the shortest even path connecting these two cities
contains 200 flights.
Estimate. Let A and B be two arbitrary cities. We prove that there exists
an even path between A and B consisting of at most 200 flights. Consider
the smallest even number 2k of flights from city A to city B and assume that
k > 100. Let C be the k-th visited city after A. Note that one may reach
C from A by m ≤ 100 flights. If m and k have the same parity, this will
allow shortening the even path from A to B, which is a contradiction. Thus
m and k have different parities. Analogously, there is a route from C to B
containing n ≤ 100 flights, with n, k having different parity. These two routes
allow reaching B from A via C by an even number m + n ≤ 200 flights, a
contradiction.
Estimate, another way. Let A and B be two arbitrary cities. For a city X,
denote by f (X) the minimal number of flights needed to reach B from X.
Obviously |f (X) − f (Y )| ≤ 1 for any flight XY . If f (X) 6= f (Y ) for every
such flight, we see that flights are operated only between cities at which f
takes different parities. It would imply that any route between two directly
connected cities consists of an odd number of flights, a contradiction. Thus
there exists a flight XY such that f (X) = f (Y ) =: b ≤ 100. Take a route from
A to X containing a ≤ 100 flights. We may reach B from A by a + b flights,
and also by a + b + 1 flights. One of these routes is even and contains at most
200 flights.
Problem 3. Let Q(t) be a quadratic polynomial with two distinct real zeros. Prove
that there exists a non-constant polynomial P (x) with the leading coefficient 1 such
that the absolute values of all coefficients of the polynomial Q(P (x)) other than the
leading one are less than 0.001. (Dušan Djukić)
Let Q(t) = c (t − a)2 − D , where a, c and D are real constants with c 6= 0 and
D > 0. It suffices to construct a polynomial R(x) with the leading coefficient
equal to 1 such that all coefficients of the polynomial R(x)2 − D except for the
leading one are less than δ = 0.001
|c| in absolute value, and take P (x) = R(x)+a.
We are looking for the polynomial R in the form
m
X
R(x) = xn + ε · xn−ai + xai + b ,
i=1
where m and a1 < a2 < · · · < am < n are positive integers, b and ε are real
numbers. The monomials that occur in the expansion of R(x)2 , other than x2n ,
xn and the constant term, are x2n−ai , x2n−2ai , x2n−ai −aj , xn+ai , xn−ai +aj ,
xn−ai , xai +aj , x2ai and xai . Let us set the numbers ai in such a way that
each of these monomials occurs only once (or twice, as in xai +aj = xaj +ai ). If
n > 3am , this is equivalent to choosing numbers ai so that the numbers ai and
aj + ak are pairwise different for all i, j, k with j ≤ k. We may take ai = 3i−1
for i = 1, . . . , m and n = 3m + 1.
Note that the coefficient of the polynomial R(x)2 − D at xn is 2b + 2mε2 .
Thus each coefficient of R(x)2 − D except for the leading one equals one of the
numbers
0 , 2ε , ε2 , 2ε2 , 2bε , b2 − D , 2b + 2mε2 .
√ n » o
Taking b = − D, ε < min 2b , 2 , 2δ and m = b −b
δ δ
ε2 c, we achieve that all
these coefficients are less than δ in absolute value.
2nd Olympiad of Metropolises
Day 2
Problem 1. Find the largest positive integer N for which one can choose N distinct
numbers from the set {1, 2, 3, . . . , 100} such that neither the sum nor the product of
any two different chosen numbers is divisible by 100. (Mikhail Evdokimov)
Answer: 45.
Example. Let us choose all numbers from 1 to 49 except 20, 25, 30, and 40.
The sum of any two chosen numbers does not exceed 97, and thus cannot
be a multiple of 100. Moreover, since none of these numbers is a multiple
of 25 and the only multiple of 10 is 10 itself, the product of any two chosen
numbers cannot be a multiple of 100. Therefore this set of numbers satisfies
the condition.
Estimate. Consider the following 46 groups of numbers. The first group con-
tains all multiples of 10 (i. e. 10, 20, . . . , 100), whereas the other 45 groups
consist of the pairs of the remaining numbers summing up to 100 (i. e. 1 and
99, 2 and 98, . . . , 9 and 91, 11 and 89, . . . , 49 and 51). Since the product
of any two numbers from the first group is divisible by 100 and the sum of
the two numbers in each pair is 100, we can choose at most one number from
each group. Moreover, we must not choose numbers from both pairs (4, 96)
and (25, 75), because the product of these two numbers would be a multiple
of 100. Hence, we can choose at most 45 numbers.
Problem 2. Let x and y be positive integers greater than 1 such that
[x + 2, y + 2] − [x + 1, y + 1] = [x + 1, y + 1] − [x, y] .
Prove that one of the two numbers x and y divides the other.
(Here [a, b] denotes the least common multiple of a and b.) (Dušan Djukić)
Solution 1.
We will prove that the hexagon A1 B1 C1 D1 E1 F1 is centrally symmetric and
therefore the main diagonals A1 D1 , B1 E1 and C1 F1 pass through the symme-
try center of the hexagon.
We denote the centers of ωA , ωB , ωC , ωD , ωE , and ωF by IA , IB , IC , ID , IE ,
and IF , respectively.
Claim 1. `AB k CF k `DE , `BC k AD k `EF , and `CD k BE k `F A .
lAB
IA
IB
A B
M m
Claim 2. A1 B1 + C1 D1 + E1 F1 = B1 C1 + D1 E1 + F1 A1 .
Proof. Let TA , UA , VA , and WA be the points where ωA touches the lines AB,
F A, `AB , and `F A , respectively, and define the points TB , . . . , WF analogously
(Fig. 2). Since the hexagon ABCDEF is tangential, we have
AB + CD + EF = BC + DE + F A. (6)
Furthermore we have
ATA = AUA , . . . , F TF = F UF and A1 VA = A1 WA , . . . , F1 VF = F1 WF ,
(7)
because these pairs of segments are tangents drawn to the circles ωA , . . . , ωF .
Finally, from the symmetry about the lines IA IB , . . . , IF IA , we can see that
TA UB = VA WB , . . . , TF UA = VF WA . (8)
WA VB
UA
TB
A TA UB B
E1 D1
F1 X
Z C1
Y
A1 B1
F1 A1 + B1 C1 + D1 E1 = XB1 + Y D1 + ZF1
> Y B1 + ZD1 + XF1 = C1 D1 + E1 F1 + A1 B1 ,
contradicting Claim 2.
If XY Z is a triangle with the opposite orientation from the hexagon, we get
a contradiction in the same way, for then we have F1 A1 + B1 C1 + D1 E1 <
C1 D1 + E1 F1 + A1 B1 ,
Solution 2.
We present an alternative finishing of the solution after Claim 1. We denote
by I the incenter of the hexagon ABCDEF .
Claim 4. IIA · IA = IIB · IB = IIC · IC = IID · ID = IIE · IE = IIF · IF .
I C
IA
IB
A B
M
Proof. Again, let M be the midpoint of the arc AB of the circumcircle not
containing C. As was already mentioned in the proof of Claim 1, the points A,
B, IA , and IB lie on a circle centered at M (Fig. 4). Since AIA and BIB
are bisectors of the angles F AB and ABC, they meet at I. Thus IIA · IA =
IIB · IB is the power of I with respect to the circle mentioned above. The
other equalities can be proved in a similar way.
√
Claim 4 implies that the inversion f with center I and radius ρ = IIA · IA
maps the points A, . . . , F to IA , . . . , IF , respectively. Therefore, the hexagon
IA IB IC ID IE IF is cyclic.
It is well-known that the main diagonals of a cyclic hexagon X1 X2 X3 X4 X5 X6
are concurrent if and only if X1 X2 ·X3 X4 ·X5 X6 = X4 X5 ·X6 X1 ·X2 X3 (this fact
follows from the trigonometric form of Ceva’s theorem for triangle X1 X3 X5 ).
Thus, since ABCDEF is both cyclic and tangential, by the Brianchon theorem
we obtain
AB · CD · EF = BC · DE · F A. (9)
On the other hand, by means of the inversion f we obtain
IA · IB IF · IA
AB = IA IB · , ..., F A = IF IA · .
ρ2 ρ2
Plugging this into (9) we obtain that
IA IB · IC ID · IE IF = IB IC · ID IE · IF IA .
Proof. Let the angle bisectors of ∠F AB and ∠CDE meet the circumcircle
of ABCDEF again at NA and ND , respectively (Fig. 5); notice that
1 ˘ ˘
∠(NA ND , BE) = (ND E + NA B) =
2
1 ˘ ˘
= (CN D + F NA ) = ∠(CF, NA ND ) .
2
Thus NA ND is parallel to the bisector of the angle between BE and CF
(containing A). By Claim 1, NA ND is also parallel to the bisectors of an-
gles F1 A1 B1 and C1 D1 E1 .
Since the points A, D, NA , and ND are concyclic, lines NA ND and AD are
anti-parallel with respect to angle AID. On the other hand, by Claim 4 the
points A, D, IA , and ID are concyclic or collinear, which means that lines AD
and IA ID are antiparallel with respect to the same angle. The two observations
yield NA ND k IA ID .
Thus, the two angle bisectors are parallel to IA ID and pass through IA and ID ,
respectively. This proves the first part of the Claim; the proofs of the other
parts are similar.
E
NA
F ID
D1
I
C
A1
ND
IA
A B
E
D
A B
Claim 5, along with the above discussion, shows that the diagonals IA A1 D1 ID ,
IB B1 E1 IE , and IC C1 F1 IF are concurrent and bisect the angles of the hexagon
A1 B1 C1 D1 E1 F1 (Fig. 6). This shows another interesting fact: namely, this
hexagon is tangential and the diagonals connecting opposite points pass through
its incenter.
Solution 3.
Using Claim 4 from the previous solution, we shall prove a more general state-
ment:
Theorem. Two circles Ω and ω are given in the plane such that ω lies inside
Ω. Consider an arbitrary broken line ABCD inscribed in Ω whose segments
AB, BC and CD are tangent to ω. Let IB and IC be the incenters of the
triangles ABC and BCD, respectively. Let ` be the line symmetric to the
line BC about the line IB IC . Then ` touches a fixed circle independent of the
choice of the broken line ABCD.
Lemma. Let Ω and ω be two circles with centers at O and I, respectively,
such that ω lies inside Ω. Consider an arbitrary chord BC of the circle Ω
tangent to ω. Then the circumcenter X of the triangle BIC lies on a fixed
circle Γ centered at O, independent of the choice of BC.
O
L I
M K
B C
OB 2 − OI 2 = (OB 2 − XB 2 ) − (XI 2 − OI 2 )
= (OM 2 − XM 2 ) − (XL2 − OL2 )
= (OM + XM )(OM − XM ) − (XL + OL)(XL − OL)
= 2OX · LM = 2OX · IK.
The lengths of segments OB, OI and IK do not depend on the chord BC, so
neither does the length of OX.
G IC C
I
′
O T
X T
IB Y
A B
Proof of the theorem. Point X and circle Γ are defined as above. Consider the
inversion f centered at I that maps IB and IC to B and C, respectively. (Its
existence and independence from the broken line follows from the corollary of
Claim 4).
Denote γ = f (Γ). Let G be the center of γ and let T be the tangency point of
ω and BC (Fig. 8). It is well-known that f maps the circumcenter X of the
triangle BIC to the point Y symmetric to I about the line IB IC . We shall
prove that the distance from G to the variable line ` is constant.
We show first that GY ⊥ `. Points B and C are symmetric about OX, so
OX ⊥ BC. Also, since IT ⊥ BC, the lines IX and IT are symmetric about
the bisector of angle BIC, and the lines BC and IB IC are anti-parallel with
respect to the angle BIC, we have IX ⊥ IB IC . Now from OX ⊥ BC and
IX ⊥ IB IC we deduce that ∠(OX, IX) = ∠(BC, IB IC ). By properties of
inversion one can obtain ∠(OX, IX) = ∠(IY, GY ) (homothety of γ and Γ is
useful here). By the symmetry in IB IC we have ∠(BC, IB IC ) = ∠(IB IC , `).
Thus ∠(IY, GY ) = (IB IC , `), so IY ⊥ IB1 IC1 implies that GY ⊥ `.
Let T 0 be the reflection of T in IB IC . The segments IT and Y T 0 are symmetric
about the line IB IC and so are the lines BC and `. Since Y T 0 ⊥ ` it follows that
G lies on Y T 0 . Finally, the distance between G and ` is equal to GY − Y T 0 =
GY − IT = R(γ) − R(ω), which is constant. (Here R(s) denotes the radius of
a circle s.)
The problem statement easily follows from this theorem. Observe that all lines
`AB , `BC , `CD , `DE , `EF , `F A are tangent to the same circle, which implies
that the hexagon A1 B1 C1 D1 E1 F1 is tangential. By the Brianchon theorem its
diagonals are concurrent.
The 3rd Olympiad of Metropolises
Day 1. Solutions
Problem 1. Solve the system of equations in real numbers:
(
(x − 1)(y − 1)(z − 1) = xyz − 1 ,
(x − 2)(y − 2)(z − 2) = xyz − 2 .
(Vladimir Bragin)
Answer: x = 1, y = 1, z = 1.
Solution 1. By expanding the parentheses and reducing common terms we obtain
(
− (xy + yz + zx) + (x + y + z) = 0 ,
− 2(xy + yz + zx) + 4(x + y + z) = 6 .
From the first equation we can conclude that xy+yz+zx = x+y+z. By substituting
this into the second equation, we obtain that x + y + z = 3. We now have to solve
the system (
x + y + z = 3,
(1)
xy + yz + zx = 3 .
If we square the first equation, we get x2 + y 2 + z 2 + 2(xy + yz + zx) = 9. Hence
x2 + y 2 + z 2 = 3 = xy + yz + zx.
We will prove that if x2 + y 2 + z 2 = xy + yz + zx, than x = y = z:
x2 + y 2 + z 2 = xy + yz + zx ⇐⇒
2 2 2
2x + 2y + 2z = 2xy + 2yz + 2zx ⇐⇒
2 2 2 2 2 2
x − 2xy + y + x − 2xz + z + y − 2yz + z = 0 ⇐⇒
2 2 2
(x − y) + (x − z) + (y − z) = 0 .
The sum of three squares is 0, so all of them are zeroes, which implies x = y = z.
That means x = y = z = 1.
Solution 1’. We will show one more way to solve the system (1). Express z = 3−x−y
from first equation and substitute it into the second one:
xy + (y + x)(3 − x − y) = 3 ⇐⇒
2 2
xy + 3x + 3y − 2xy − x − y = 3 ⇐⇒
2 2
x + y + xy − 3x − 3y + 3 = 0 ⇐⇒
2 2
x + x(y − 3) + y − 3y + 3 = 0 .
1
Let us solve it as a quadratic equation over variable x:
p
(3 − y) ± (y − 3)2 − 4(y 2 − 3y + 3)
x= =
p 2
(3 − y) ± y 2 − 6y + 9 − 4y 2 + 12y − 12
= =
p 2
(3 − y) ± −3y 2 + 6y − 3
= =
p2
(3 − y) ± −3(y − 1)2
= .
2
We can conclude that y = 1, because otherwise the square root wouldn’t exist. It
follows that x = 3−1±0
2 = 1, and then z = 1.
Solution 2. Let’s make variable substitution u = x − 1, v = y − 1, w = z − 1. We
obtain the system (
(u + 1)(v + 1)(w + 1) = uvw + 1 ,
(u − 1)(v − 1)(w − 1) = uvw − 1 ,
(where the latter equation actually corresponds to the difference between two original
equations).
After expanding all parentheses and reducing common terms we have
(
uv + uw + vw + u + v + w = 0 ,
− (uv + uw + vw) + u + v + w = 0 .
Now consider the polynomial g(t) = f (t) − f (0) − t. Its main coefficient is 1, and 0,
1 and 2 are its roots. Hence g(t) = t(t − 1)(t − 2). It follows that
2
Observe that (t − 1)3 + f (0) + 1 is an increasing function, which means that different
real numbers cannot be its roots. So x = y = z and also x is also the root of the
derivative of f (t). But f 0 (t) = 3(t − 1)2 , hence x = y = z = 1.
Problem 2. A convex quadrilateral ABCD is circumscribed about a circle ω. Let
P Q be the diameter of ω perpendicular to AC. Suppose lines BP and DQ intersect
at point X, and lines BQ and DP intersect at point Y . Show that the points X
and Y lie on the line AC. (Géza Kós)
Solution. The role of points P and Q is symmetrical, so without loss of generality
we can assume that P lies inside triangle ACD and Q lies in triangle ABC.
Part 1. Denote the incircles of triangles of ABC and ACD by ω1 and ω2 and denote
their points of tangency on the diagonal AC by X1 and X2 , respectively. We will
show that line BP passes through X1 , DQ passes through X2 and X1 = X2 . Then
it follows that X = X1 = X2 is lying on AC (fig. 1).
P
ω
ω2
X1 = X2
A C
ω1
Q
B
Figure 1: for the solution of the problem 2.
As is well-known, the tangent segments AX1 and AX2 to the incircles can be ex-
pressed in terms of the side lengths as
1 1
AX1 = (AB + AC − BC) and AX2 = (AC + AD − CD) .
2 2
Since the quadrilateral ABCD has an incircle, we have AB + CD = BC + AD and
therefore
1
AX1 − AX2 = (AB − BC − AD + CD) = 0 ;
2
this proves X1 = X2 .
3
By having the common tangents BA and BC, the circles ω are ω1 are homothetic
with center B. The tangents to ω at X1 and to ω1 at P are parallel, so this homothety
maps P to X1 . Hence, the points B, P , X1 are collinear.
Similarly, from the homothety that maps ω to ω2 , one can see that D, Q, X2 are
collinear.
Part 2. Now let γ1 and γ2 be the excircles of triangles of ABC and ACD, opposite to
vertices B and D, respectively, and denote their points of tangency on the diagonal
AC by Y1 and Y2 , respectively. Analogously to the first part, we will show that line
BQ passes through Y1 , DP passes through Y2 and Y1 = Y2 (fig. 2).
P
γ1 ω
Y1 = Y2
A C
Q
γ2
B
The tangent segments CY1 and CY2 to the excircles can be expressed as
1 1
CY1 = (AB + AC − BC) and CY2 = (AC + AD − CD) ;
2 2
by AB + CD = BC + AD it follows that CY1 = CY2 , so Y1 = Y2 .
The circles ω and γ1 are homothetic with center B. The tangents to ω and γ1 at Q
and Y1 are parallel so this homothety maps Q to Y1 . Hence, the points B, Q, Y1 are
collinear.
Similarly, from the homothety that maps ω to γ2 , one can see that D, P , Y2 are
collinear.
Problem 3. Let k be a positive integer such that p = 8k + 5 is a prime number.
The integers r1 , r2 , . . . , r2k+1 are chosen so that the numbers 0, r14 , r24 , . . . , r2k+1
4
give
4
pairwise different remainders modulo p. Prove that the product
Y
ri4 + rj4
16i<j62k+1
We may write g 4f (j) − g 4f (i) = ±(g 4 max(f (j),f (i)) − g 4 min(f (j),f (i)) ), where the sign is
positive if f (j) > f (i) and negative if f (j) < f (i). Further, when the ordered pair
(i, j) runs over all k(2k + 1) ordered pairs satisfying 0 6 i < j 6 2k, the ordered pair
min(f (j), f (i)), max(f (j), f (i)) runs over the same set. Therefore the differences
Q g4f (j) −g4f (i)
cancel out and the above product of the ratios g 4j −g 4i equals (−1)N , where
N is the number of pairs i < j for which f (i) > f (j). This in turn happens when
i = 1, 2, . . . , k; j = k + 1, . . . , k + i, totally N = 1 + . . . + k = k(k + 1)/2. Thus the
result.
Solution 2. Denote ti = ri4 . Notice that the set T := {t1 , . . . , t2k+1 } consists of
distinct roots of the polynomial x2k+1 − 1 (over the field of residues modulo p). Let
us re-enumerate T so that tk+1 = 1, ti = 1/t2k+2−i for i = 1, 2, . . . , k. The map
2 k+1
t 7→
√ t is a bijection on T , the inverse map is s 7→ s √ and p we naturally
p denote
it s. For distinct Q elements t, s ∈ T we have t + s = st( s/t + t/s). In the
following formula denotes the product over all k(2k + 1) pairs of distinct elements
5
t, s ∈ T . We have
!k k
!2k+1
Y Y√ Y p p Y Y
(t + s) = st · s/t + t/s = t · (ti + 1/ti ) .
t∈T i=1
The first multiple equals 1 by Vieta’s formulas for x2k+1 − 1 = t∈T (x − t). As for
Q
the second multiple, note that there is a polynomial ψ(x) with integer coefficients
satisfying
1
ψ x+ = xk + xk−1 + . . . + 1 + . . . + x−k .
x
Obviously, the leading coefficient in ψ is 1. The constant term can be accessed by
substituting the complex unit x = i; the constant term is
k
(
1 X 1 if k ≡ 0, 1 (mod 4) ,
ψ(0) = ψ i + = ij =
i −1 if k ≡ 2, 3 (mod 4) .
j=−k
The roots of ψ in the modulo p field are exactly ti + 1/ti , i = 1, 2, . . . , k (they are
distinct). The product of the roots is
k
(
Y
k 1 if k ≡ 0, 3 (mod 4) ,
(ti + 1/ti ) = (−1) · ψ(0) =
i=1
−1 if k ≡ 1, 2 (mod 4) .
Finally, we conclude
(
Y 1 if k ≡ 0, 3 (mod 4) ,
(t + s) =
−1 if k ≡ 1, 2 (mod 4) .
6
The 3rd Olympiad of Metropolises
Day 2. Solutions
Problem 4. Let 1 = d0 < d1 < . . . < dm = 4k be all positive divisors of 4k, where
k is a positive integer. Prove that there exists i ∈ {1, . . . , m} such that di −di−1 = 2.
(Ivan Mitrofanov)
Solution 1. Assume the contrary. This means that if d and d + 2 both divide 4k,
then d + 1 also divides 4k. Note that if a divides 4k and a is not divisible by 4, then
2a also divides 4k. Using the properties above, we start from the pair (1, 2) and find
more pairs (a, a + 1) such that both a and a + 1 divide 4k and both a and a + 1 are
not divisible by 4.
Let (a, a + 1) be a pair of divisors of 4k such that 4 divides neither a nor a + 1. Then
2a and 2a + 2 divide 4k, hence 2a + 1 divides 4k. One of 2a or 2a + 2 is not divisible
by 4, hence in one of pairs (2a, 2a + 1), (2a + 1, 2a + 2) both numbers divide 4k, but
are not divisible by 4.
Apply this procedure to the new pair, and so on. Thus, starting from the pair (1, 2)
we obtain pairs (2, 3), (5, 6), (10, 11), etc. At each step the sum of numbers in pair
increases, hence we obtain an infinite set of divisors of 4k. A contradiction.
Solution 2. Assume the contrary. Let t be the minimal positive integer that does
not divide 4k. Then 1, 2, . . . , t − 1 divide 4k, while t and t + 1 do not (otherwise
t − 1 and t + 1 would be two consecutive divisors of 4k).
It follows that t and t + 1 are prime powers, otherwise one of them would be a
product of two coprime multiples less than t and would therefore divide 4k. One of
them is a power of 2 that we denote as 2m , m > 3. The other has a form of 2m + ε,
ε = ±1.
Observe that 2m−1 divides 4k, since 2m is the minimal even non-divisor of 4k, and
that 3 · 2m−2 + ε divides 4k, since it is odd and less than 2m + ε, which is the minimal
odd non-divisor of 4k. Also note that 3 divides 4k.
It follows that 3 · 2m−1 and 2 · (3 · 2m−2 + ε) also divide 4k. Hence, the number
3 · 2m−1 + ε between them divides 4k too. But now 4k is divided by 2 · (3 · 2m−1 + ε),
just as 4 · (3 · 2m−2 + ε). The number 3 · (2m + ε) is between them, and must be a
divisor of 4k as well. We conclude that 2m + ε divides 4k, contradiction.
Problem 5. Ann and Max play a game on a 100 × 100 board.
First, Ann writes an integer from 1 to 10 000 in each square of the board so that
each number is used exactly once.
1
Then Max chooses a square in the leftmost column and places a token on this square.
He makes a number of moves in order to reach the rightmost column. In each move
the token is moved to a square adjacent by side or by vertex. For each visited square
(including the starting one) Max pays Ann the number of coins equal to the number
written in that square.
Max wants to pay as little as possible, whereas Ann wants to write the numbers in
such a way to maximise the amount she will receive. How much money will Max
pay Ann if both players follow their best strategies? (Lev Shabanov)
Answer: 500 000 coins.
Solution. Lower bound / Ann’s strategy. First we will prove that Ann can get at
least 500 000 coins. Suppose Ann has arranged the numbers in the way depicted on
the fig. 1.
···
···
···
···
···
···
··
·
Consider a path constructed by Max. For every integer 1 6 n 6 50 there are two
squares of the path in the columns 2n − 1 and 2n, respectively, which are adjacent
by a move of the token. It is easy to see that the sum of the numbers in such
squares is at least 200(2n − 1). We obtain that the cost of this path is at least
200(1 + 3 + 5 + . . . + 99) = 500 000 coins.
Upper bound / Max’s strategy. Now consider an arbitrary arrangement of the num-
bers. Then exclude the square with the greatest number in each column. The least
of the numbers in excluded squares is not less than 100, the second least is not less
than 200, etc, the greatest number in excluded square is 10 000. Hence, the sum of
the excluded numbers is at least 100 + 200 + . . . + 10 000 = 505 000, while the sum of
2
numbers in the whole square is 50 005 000. It means that the sum of the remaining
numbers is at most 49 500 000.
The 9900 squares which are left can be split into 99 distinct paths. The first path
consists of the lowest squares in each column which are not excluded, the second
path consists of the lowest squares in each column which are not excluded and not
included into the first path, etc. The last path will include the uppermost squares
in each column which are not excluded. It is easy to see that all 99 paths are proper
paths that Max’s token can follow, because any two squares in two adjacent columns
either lie in the same row or in two adjacent rows.
The total cost of the constructed 99 paths is not greater than 49 500 000, therefore
one of these paths costs at most 500 000 coins. Thus in every arrangement of numbers
Max can pay less than or equal to 500 000 coins.
We proved that Max can pay 500 000 coins or less while Ann can make Max pay at
least 500 000 coins, so the answer is 500 000.
Alternate proof of the upper bound. Split our board into 50 horizontal rectangles
2 × 100. Since the sum of numbers in the whole board is 50 005 000, it is possible
to choose a rectangle with the sum of at most 1 000 100. Take a square with the
minimum number in each column of the chosen rectangle. These squares form a path
from the square in the left column to the square in the right column; denote the cost
of this path as S. In each column of the rectangle, the minimal number is less than
the other number by at least 1. Hence, the total sum of the numbers in the rectangle
is at least 2S + 100. It follows that 2S + 100 6 1 000 100 and S 6 500 000.
Problem 6. The incircle of a triangle ABC touches the sides BC and AC at points
D and E, respectively. Suppose P is the point on the shorter arc DE of the incircle
such that ∠AP E = ∠DP B. The segments AP and BP meet the segment DE at
points K and L, respectively. Prove that 2KL = DE. (Dušan Djukić)
A property of the symmedian. The symmedian of a triangle from one of its vertices
is defined as the reflection of the median from that vertex about the bisector from
the same vertex (fig. 2). In the following solutions we will use a well-known property
of the symmedian, namely that it passes through the intersection of the tangents to
the circumcircle of the triangle taken at the other two vertices.
Solution 1. Denote by F the tangency point of the incircle with the side AB, and
by M and N respectively the midpoints of the segments EF and DF (fig. 3). Since
P B is the symmedian in the triangle DP F , we have ∠KP E = ∠DP B = ∠N P F .
Moreover, ∠P EK = ∠P ED = ∠P F N , so 4P EK ∼ 4P F N . Analogously,
DF ·P E
4P DL ∼ 4P F M . Now we obtain EK = F N · P E
PF = 2P F and similarly
PD EF ·P D DF ·P E+EF ·P D 1
DL = F M · P F = 2P F , so EK + DL = 2P F = 2 DE by Ptolemy’s
theorem. Therefore, KL = DE − EK − DL = 12 DE.
Solution 2. Denote by F the tangency point of the incircle with the side AB. Con-
3
Figure 2: for the solution of problem 6.
P
E
K D
L
M
N
A F B
4
C
P
E
K S D
L
A F B
P
E
K L
D
Q
X
Y
A F B
Since the arcs EX and DY are congruent, we have XY k DE. Let the lines P F and
EY meet at point Q (fig. 5). The quadrilaterals EP F X and Y F P D are harmonic,
and the projection with the center Q from the incircle to itself maps points E, P, F
5
to Y, F, P , respectively. Therefore, this projection also sends X to D, i. e. the line
XD also passes through Q.
Now let ` be the line passing through the intersection point of XE and Y D, parallel
to DE. Consider the projective transformation preserving the incircle that takes
line ` to the infinity line. The lines DE, XY and ` will remain parallel under this
transformation and the ratio KL/DE will not change. Moreover, the quadrilaterals
XEP F and DY F P will remain harmonic. Thus we obtain fig. 6.
E K L D
X U V Y
F
Figure 6: for the solution of the problem 6.