STAR CIRCLE PROGRAM. Assignment 4.

Problem 1
Given three positive real numbers a, b, c such that following holds a2 = b2 + bc, b2 = c2 + ac Prove that 1
c = 1
a + 1b .
Solution:
We are given the equations:
a2 = b2 + bc and b2 = c2 + ac.
Starting from the first equation:

a2 = b2 + bc =⇒ a2 · c = b2 · c + b · c2 .

From the second equation, we also have:

a2 = b2 + bc = c2 + ac + bc =⇒ a2 = c(a + b + c).

a2 · c = b2 · c + b · c2

a2 · c + abc = b2 · c + b · c2 + abc.
Rearranging terms, we have:
ac(a + b) = b · c(a + b + c).
This implies:
ac(a + b) = ba2 .
Then,
1 1 1
c(a + b) = ab ⇒ = + .
c a b

Problem 2
On the sides AB and BC of an acute-angled triangle ABC, squares ABC1 D1 and A2 BCD2 are constructed
externally. Lines AD2 and CD1 intersect at D. Prove that BD ⊥ AC.
Solution
Step 1: Showing that CC1 ⊥ AA2
To prove that CC1 is perpendicular to AA2 , we draw the circumcircles of squares ABC1 D1 and A2 BCD2 . These
circles intersect at points B and E.
We aim to show that line AE passes through A2 , and similarly, CE passes through C1 .
Consider the angles:
270◦
∠AEB = = 135◦ .
2
We also have ∠BEA2 = ∠BCA2 = 45◦ . Since

∠AEB + ∠A2 EB = 180◦ ,

we conclude that AE indeed passes through A2 . By a similar argument, CE passes through C1 .

270◦ 270◦
 
∠AEC = 360◦ − (∠CEB + ∠AEB) = 360◦ − + = 90◦ .
2 2

Therefore, CC1 ⊥ AA2 .

Step 2: Constructing Parallelogram C1 BA2 H

To proceed, draw a line parallel to A2 B passing through C1 , and another line parallel to C1 B passing through
A2 . Let these lines intersect at H, forming the parallelogram C1 BA2 H.
In C1 BA2 H, we have:
∠D1 C1 H = ∠ABA2 and AA2 = D1 H.
Since △D1 C1 H ∼= △ABA2 by the Side-Angle-Side (SAS), we conclude that AA2 = D1 H and AA2 ∥ D1 H. Similarly,
△CBC1 ∼ = △D2 A2 H, giving CC1 = D2 H and CC1 ∥ D2 H.
Finally, we have HD1 = HD2 , which shows that △HCD1 ∼ = △HAD2 .

Step 3: Proving AD2 ⊥ CH and CD1 ⊥ AH

We claim that AD2 ⊥ CH and CD1 ⊥ AH. This follows because △HCD1 ∼ = △HAD2 , where △HCD1 is a 90◦
rotation of △HAD2 .
Since CC1 ⊥ AA2 , with CC1 ∥ D2 H and AA2 ∥ D1 H, it follows that D2 H ⊥ D1 H. Consequently, AD2 ⊥ CH
and CD1 ⊥ AH. The intersection point D serves as the orthocenter of △AHC, implying:

HD ⊥ AC.

Conclusion
Since ∠C1 BH = ∠A, then ∠ABF = 90◦ − ∠A, thus we conclude that HB ⊥ AC. Therefore, points H, B, and
D are collinear, and this line is perpendicular to AC, as required.

Problem 3
Find all such polynomials P (x) with real coefficients which satisfy

(x2 − 6x + 8)P (x) = (x2 + 2x)P (x − 2)

for all real x.

Answer: The polynomials that satisfy the identity are the polynomials of the form P (x) = cx2 (x − 2)(x + 2),
with c ∈ R.

1. First we can factor the polynomials x2 − 6x + 8 = (x − 2)(x − 4) and x2 + 2x = x(x + 2), so we have the
equality
(x − 2)(x − 4)P (x) = x(x + 2)P (x − 2), (1)
 2. Evaluating in x = 0, x = −2 and x = 4 we conclude that P (0) = P (−2) = P (2) = 0, therefore there exists a
polynomial Q(x) such that P (x) = x(x−2)(x+2)Q(x), so P (x−2) = (x−2)(x−4)xQ(x−2) and substituting
in (1) we get that
(x − 2)2 (x − 4)x(x + 2)Q(x) = x2 (x + 2)(x − 2)(x − 4)Q(x − 2)
holds all real x, or canceling the common factors we get that

(x − 2)Q(x) = xQ(x − 2) (2)

holds for every real x.

3. In fact we get that (2) holds for every x distinct to −2, 0, 2, 4, however since the polynomial identity holds for
infinitely many values, it holds for every real value x.
4. Evaluating in x = 0 we get that Q(x) = 0, so there exists a polynomial R such that Q(x) = xR(x), so
Q(x − 2) = (x − 2)R(x − 2), so we get that x(x − 2)R(x) = x(x − 2)R(x − 2), or R(x) = R(x − 2), from here
it follows that R(x) is in fact a constant since the equality implies that R(0) = R(2) = R(4) = · · · . Therefore
R(x) = c for some c ∈ R and we conclude that P (x) = cx2 (x − 2)(x + 2).
5. Replacing P (x) = cx2 (x − 2)(x + 2) in the original equation we can verify that the identity indeed holds.

Problem 4
Given positive integers numbers n > k > 1. Andrey and Kirill take turns (Andrey starts) coloring the cells of an
n × n square black. The player who, after their move, ensures that every k × k square contains at least one black
cell wins. Who wins with optimal play?
Comment. If k > n/2 then a cell in the middle would intersect every k × k square, so Andrey could win in a
single move.
Solution
Case 1: n is even

• If k > n2 , Andrey wins.

Since k is greater than half of n, Andrey can win on his first move by coloring one of the four center cells.
This ensures that each k × k square intersects with at least one black cell (the one he colored), so he wins
immediately.
• If k ≤ n2 , Kirill wins.
Andrey cannot ensure a win on his first move in this case, so the game proceeds with each player coloring one
cell per turn.
To see why Kirill wins, observe that:
– After Andrey’s first move, it is Kirill’s turn, and he can play optimally by ensuring that he never allows
Andrey to complete a k × k square before him.
– Specifically, either there exists a move for Kirill to win directly by covering all uncolored k × k squares
with his choice, or there are at least two non-intersecting k × k squares remaining.
– Suppose, for contradiction, that there is not at least two non-intersecting k × k squares left. Then
there would exist a move where Kirill could complete all k × k squares by coloring one cell, thus winning.
Otherwise, Kirill can always make a move that avoids covering both of these two non-intersecting squares
in a single turn, there is always such a move because in his turn there will be always an odd number of
cells.
You can prove the part as follows: if there are m pairwise intersecting k × k squares, then there is always a
square that is contained in all of these m k × k squares. To prove this, we use the fact that the intersection of
two rectangles is itself a rectangle, so the intersection of m k × k squares is also a rectangle. We can proceed
by induction:

– The base case is that the intersection of two k × k squares is a rectangle, which holds by definition.
 – Now, assume that the intersection of (m − 1) k × k squares is a rectangle. We add the m-th k × k square,
assuming it intersects all of the other (m − 1) squares without containing any of these squares in the
rectangle formed by their intersection.
– Consider the rectangle formed by the intersection of the (m − 1) k × k squares. The leftmost border
of this rectangle is the rightmost left boundary among the (m − 1) squares, and similarly for the top,
bottom, and right boundaries.

Thus, if the m-th square does not intersect any cells of this rectangle, it would lie entirely to the left, right,
above, or below the intersection, which would mean that it does not fill at least one of the squares. This
contradiction completes the proof.
So by playing in this way, Kirill will either win during the sequence of moves or eventually reach a position
where only two k × k squares remain uncolored. Then, on Andrey’s turn, he will be forced to color one of the
two remaining squares, and Kirill can color the other on his next turn, winning the game.

Case 2: n is odd
In this case, Andrey always wins. Here’s why:
• Andrey starts by coloring the central square of the n × n board.
• From this point on, Andrey can mirror Kirill’s moves along the central square, which serves as the axis of
symmetry.

• Each time Kirill covers an uncolored k × k square, there will be a corresponding k × k square on the opposite
side of the board that Andrey can cover on his next move.
• Since the central cell is already colored, no k×k square that contains this cell can become completely uncolored,
ensuring that every uncolored k × k square has a corresponding k × k square that doesn’t intersect it at all
as its mirror. This strategy guarantees that each of Kirill’s moves will have a symmetric counter-move from
Andrey, eventually forcing Kirill to lose.
Conclusion
• If n is even:

– If k > 2,
n
Andrey wins.
– If k ≤ 2,
n
Kirill wins.
• If n is odd, Andrey always wins by mirroring Kirill’s moves along the center.

Problem 5
There are 10 students in the class. Every day 4 of them go for lunch. Prove that after 100 days it is possible to find
two people who had lunch together at least 14 times.
Solution. Suppose otherwise, that no two people have gone to lunch at least 14 times. There are 10 2 = 45 pairs

of students, so we counting with multiplicity,
there ar at most 13 · 45 = 585 pairs students that have go for lunch
in the different days, but every day we have 42 = 6 pairs of students, hence we would have in total of 6 · 100 = 600
pairs of students that have lunch in the same day, a contradiction.

Problem 6
The sequence an is defined by the conditions a1 = 2 and an = 2an−1 + 2 for all n ⩾ 2. Prove that an is divisible by
an−1 for all n ⩾ 2.
an+1
Solution. We will prove, that is integer and odd using induction on n.
an
Base case. As far as a1 = 2 and a2 = 6, a3 = 66 the base case is clear.
 Step of Induction n − 1 → n + 1. We need to prove that 2an+1 + 2 is divisible by 2an + 2.

2an + 2|2an+1 + 2 =⇒
2an −1 + 1|2an+1 −1 + 1 =⇒
an − 1|an+1 − 1 as far as an − 1 and an+1 − 1 are odd =⇒
2an−1 + 1|2an + 1 =⇒
an /an−1 is integer and odd by induction hypothesis

Problem 7
A polygon P is inscribed in a circle ω. A triangulation of the polygon P is defined as its subdivision into triangles
by non-intersecting diagonals. For each triangulation, compute the sum of the inradii of the triangles formed by the
triangulation. Prove that the resulting sum does not depend on the chosen triangulation.

Solution
1. We start with the case of a polygon with only 4 vertices. Denote the 4 vertices as A, B, C, D, and let IA ,
IB , IC , ID be the incenters of triangles BCD, ACD, ABD, and ABC, respectively. Denote F , E, H, G as
the midpoints of the smaller arcs. From one of the problems in the previous assignments, we know that IA ,
IB , IC , and ID are the vertices of a rectangle with sides parallel to HF and GE. It is not difficult to see that
HF and GE form equal angles with AB and CD, as well as with BC and AD.

2. Let us consider the following construction. We know that points on the bisector of an angle are equidistant
from the sides of the angle. Consider an arbitrary angle ∠P OQ, its bisector ℓ, and an arbitrary line ℓ1 parallel
to ℓ. Choose an arbitrary point F on ℓ1 , and let G be the point on ℓ such that GF ⊥ ℓ. The length of GF
does not depend on the choice of F . We will show that the difference ρ(F, OP ) − ρ(F, OQ) is the same for all
possible points F .
 Denote A and B as the feet of the perpendiculars from F to OP and OQ, and C, H as the feet of the
perpendiculars from G to OP and OQ. Let I and J be the intersections of F G with OP and OQ. Since
GJ = GI, denote ∠P OQ = 2α. Then:
F B − F A = F I cos α − F J cos α = (F I − F J) cos α = 2F G cos α.
Since F G does not depend on the choice of F , we conclude that ρ(F, OP ) − ρ(F, OQ) is independent of the
choice of F .

3. Let us apply the previous lemma to our construction.

Denote x1 = ρ(IA , BC), x2 = ρ(IA , AD), x3 = ρ(IB , BC), x4 = ρ(IB , DA), x5 = ρ(IC , BC), x6 = ρ(IC , DA),
x7 = ρ(ID , BC), and x8 = ρ(ID , DA).
We need to prove that x1 + x6 = x4 + x7 .
 We know that:

x2 − x1 = x8 − x7 , x 4 − x3 = x6 − x5 , x1 + x5 = x3 + x7 (because IA IB IC ID is a rectangle).

Summing up the last two equations, we get:

x1 + x5 + x6 − x5 = x3 + x7 + x4 − x7 , so x1 + x6 = x4 + x7 .

4. Now, let us prove the statement of the problem for polygons with n > 4 vertices. Consider an arbitrary
inscribed polygon A1 A2 . . . An . We will prove that for any triangulation, the sum of the inradii is the same.
5. Consider an arbitrary triangulation. Define an operation called a swap: take two triangles with a common
side, Ai Aj Ak and Aj Ak Aℓ , and replace them with Ai Aℓ Aj and Ai Aℓ Ak .
6. We have proved that if we take any triangulation T1 and perform a swap to obtain a new triangulation T2 ,
the sum of inradii in T1 is equal to the sum of inradii in T2 .
7. Thus, it is sufficient to prove that we can transform any triangulation into any other triangulation using a
sequence of swaps. Starting from an arbitrary triangulation, we will show that it is possible to achieve the
triangulation {An Ai Ai+1 | i = 1, . . . , n − 2}. This is an important step. In this triangulation, there are n − 3
diagonals originating from An . To achieve this from any other triangulation, we only need to increase the
number of diagonals originating from An . I mean, every time if not all n − 3 diagonals from An are drawn,
there must be a triangle with vertex An and two other vertices not consequent. So the opposite side to An is
a diagonal and there is another triangle which contains this diagonal. So we can make a swap with these two
triangles and increase the number of diagonals from An .

Problem 8
For the largest possible value of λ, that for any positive a, b, c, d such that their product equals 1, the following
inequality holds:
1 1 1 1
q +q +q +q ⩾λ
1 1 1 1
2 + a + ab + abc 2 + b + bc + bcd 2 + c + cd + cda 2 + d + da + dab
√
Answer: 2
Solution.

1. We’ll start with the following substitution.

y z t x
a= ,b= ,c= ,d=
x y z t
where x, y, z, t are positive.
2. In terms of x, y, z, t LHS will look like
1 1 1 1
+q +q +p =
x
+ yt + z
q
1/2 + y
+ z
+ xt 1/2 + z
+ t
+ x
1/2 + t
+ xz + y 1/2 + t t
x x y y y z z
√ √ √ √
x y z t
=p +p +p +p
x/2 + y + z + t y/2 + z + t + x z/2 + t + x + y t/2 + x + y + z

3. The value of this expression won’t change if we multiply all numbers x, y, z, t by the same number. Hence, we
can assume that x + y + z + t = 1. So we can rewrite the expression as
√ √ √ √
x y z t
p + p + p + p
1 − x/2 1 − y/2 1 − z/2 1 − t/2
 √
α
4. Define f (α) = √ .
1−α/2
√
5. Let’s prove that f (α) ⩾ 2α.

√
α √
p 2α⩾
1 − α/2
α
⩾ 2α2
1 − α/2
1 ⩾ 2α(1 − α/2)
α2 + 1 ⩾ 2

6. Hence, we have an inequality for LHS

√ √ √ √ √
f (x) + f (y) + f (y) + f (z) ⩾ 2x + 2y + 2z + 2t = 2.
√
7. As far as we claim that the answer is√ 2 we need to prove that we cannot replace λ by bigger number.
Suppose the inequality is true for λ = 2 + ϵ, where 0 < ϵ < 1/10 Let’s pick up numbers x, y, z, t which will
give counterexample.
8. We will pick x = y = z, where x is small and t = 1 − 3x.
√
9. If x < 1/10, then f (x) < 2 x. That is obvious, since 1 − x/2 > 1/2.
p

√ √
10. If x < 1/10, then √ 1−3x < 2(1 − x). Let’s prove it.
1/2+3x/2

√
1 − 3x √ √
p <1 − 3x 2·
1/2 + 3x/2
√
It is sufficient to prove that 1 − 3x < 1 − x
1 − 3x < 1 − 2x + x2 this is obvious

√ √
for small x we have f (x) + f (y) + f (z) + f (t) < 6 x + 2 − x. Now we need to select x in such a way
Thus, √
that 6 x − x < ϵ. For x = (ϵ/6)2 it is obviously true.
√
11. So we have proved that the inequality is not always true for values of λ larger than 2.

Problem 9
Husam has an unusual calculator with three buttons: φ, τ , σ.

• Pressing the first button changes the number n on the calculator screen to the number of natural numbers
less than or equal to n that are co-prime with n.
• Pressing the second button changes the number n on the screen to the number of divisors of n.
• Pressing the third button changes the number n to the sum of its divisors.
Prove that for any integers a, b > 1, it is possible to obtain b from a on the screen after a finite number of button
presses.
Solution.
 1 p2 . . . pk , then:
1. We know that if n = pα1 α2 αk

φ(n) = (p1 − 1)p1α1 −1 (p2 − 1)pα

2
2 −1
. . . (pk − 1)pkαk −1 ,

τ (n) = (α1 + 1)(α2 + 1) . . . (αk + 1),

pα
1
1 +1
− 1 pα 2 +1
−1 pαk +1 − 1
σ(n) = · 2 ··· k .
p1 − 1 p2 − 1 pk − 1

2. For n = p1α1 pα
2 . . . pk , define β(n) = α1 + α2 + . . . + αk . Note that for any positive integers m and n, we have
2 αk

β(mn) = β(m) + β(n).

3. Let us now derive some further conclusions based on these formulas.
4. σ(n) > n. This is straightforward and follows directly from the formula for σ(n).

5. For any prime p and positive integer n, vp (φ(n)) ≥ vp (n) − 1.

6. If n is not a power of 2, then v2 (φ(n)) ≥ v2 (n).
φ(n)
7. The ratio n can be expressed as:
    
φ(n) 1 1 1
= 1− 1− ··· 1 − .
n p1 p2 pk

b
8. Start with the number a and repeatedly press the third button until obtaining a number greater than 2b · 32 ·
3 b
(2b )b ·2 +b . Denote this number by n0 . After this step, press only the first button to generate a sequence of
numbers n1 , n2 , . . ., aiming to reach specific cases.

9. If at any point, a number nt divisible by 2b−1 is reached, then for some nt+s , we will have nt+s = 2b−1 and
τ (nt+s ) = b.
10. If a number with at least b − 1 odd prime factors is obtained, then φ(n) will be divisible by 2b−1 .
11. If no number in the sequence achieves b distinct prime factors, all numbers will have at most b − 1 odd prime
factors. From this and the condition v2 (φ(n)) ≥ v2 (n), it follows that ni > n2b−1
i−1
. Consequently, if t < b3 ·2b +b,
b
then nt > 2b · 32 .
b
12. If a number nt divisible by p2 −1 for some prime p is reached, then for some nt+s , vp (nt+s ) = 2b − 1 and
τ (nt+s ) will be divisible by 2b . Using the method outlined in step 9, we can then achieve 2b , 2b−1 , and finally
b.
13. If a number nt divisible by a prime p ≡ −1 (mod 2b−1 ) is encountered, pressing the first button repeatedly
will result in a number nt+s where p appears in exactly power 1. At this point, σ(nt+s ) will be divisible by
both p + 1 and 2b−1 . Using the approach from step 9, b can be achieved.
14. Suppose the sequence {ni } contains no numbers divisible by 2b−1 , no numbers with b distinct prime factors,
and no numbers divisible by a large power of any single prime. Then, for any i, we have β(ni ) < b · 2b .
15. We will now prove that for any t < b3 · 2b , among the numbers nt+1 , nt+2 , . . . , nt+b , there is either:
• A number with a higher power of 2 than nt , or
• A number with a higher value of β than β(nt ), or
• A specific condition (as described above) is achieved.
16. To establish this, we analyze the behavior of the sequence:
• If ni has at least two odd prime divisors, then v2 (ni+1 ) > v2 (ni ).
 • If ni = 2k · pℓ with only one odd prime p, then:

ni+1 = 2k−1 · (p − 1) · pℓ−1 .

In this case, β(ni ) = k + ℓ, and:

β(ni+1 ) = β(p − 1) + k − 1 + ℓ − 1.

Since p > 3, we have β(p − 1) ≥ 2, so β(ni+1 ) ≥ β(ni ), with strict inequality if (p − 1)/2 is not prime.

17. Now, we prove the claim from step 15. Assume the contrary, that for all 1 ≤ i ≤ b, v2 (nt+i ) ≤ v2 (nt ) and
β(nt+i ) ≤ β(nt ). As far as all ni have prime divisors greater than 3, we can apply previous step. Hence, from
this assumption:
v2 (nt ) = v2 (nt+1 ) = . . . = v2 (nt+b ),
and
β(nt ) = β(nt+1 ) = . . . = β(nt+b ).
This can only happen if nt+i = 2 · pi , where pi is a single odd prime.
k

18. Using induction on i, it can be shown that:

p0 = 2i · pi + 2i − 1.

The base case is straightforward. For the inductive step, assume p0 = 2i · pi + 2i − 1 and pi = 2pi+1 + 1. Then:

p0 = 2i (2pi+1 + 1) + 2i − 1 = 2i+1 pi+1 + 2i+1 − 1.

19. Therefore, p0 = 2b−1 · pb−1 + 2b−1 − 1, which implies p0 ≡ −1 (mod 2b−1 ). This creates an exceptional case.

20. Finally, consider the sequence {ni } for i < b3 · 2b + b. If none of the earlier conditions are met, then ni > n2b−1
i−1
,
so all terms are greater than 2 , and none are powers of 2. Thus, v2 (ni ) is non-decreasing. However, as shown
b

in step 15, v2 (ni ) cannot remain constant across b · 2b terms without β(ni ) increasing at least b · 2b times,
which contradicts step 14. This leads to the conclusion that v2 (ni ) increases at least b times, which is also
forbidden. Hence, we reach a contradiction.