STAR CIRCLE PROGRAM. Assignment 5. Solutions.

Problem 1.
Find all pairs m, p, where m is positive integer and p is prime, which satisfy
p4 = 23m + 22m + 1.
Solution.
1. For m = 2 we can check that there is no prime p such that p4 = 13 = 23 + 22 + 1, we can also check that
34 = 81 = 26 + 24 + 1 = 64 + 16 + 1 = 81, so the pair (2, 3) satisfy the equation, we will show that there is no
other pair.
2. Suppose there is a pair m, p satisfying the equation with m ≥ 3. Note that the equation is equivalent to
p4 − 1 = 23m + 22m , and factoring both sides we get that
(p2 + 1)(p + 1)(p − 1) = 22m (2m + 1).

3. Clearly p is an odd prime hence gcd(p + 1, p − 1) = 2, while p2 + 1 ≡ 2 (mod 4), so the maximum power of 2
dividing p2 + 1 is 2. Therefore we have that either 22m−2 | p − 1 or 22m−2 | p + 1.
4. • If 22m−2 | p − 1, then p > 22m−2 , it follows that p2 > 24m−4 , hence
(p2 + 1)(p + 1)(p − 1) > 24m−4 · 22m−2 · 22m−2 = 28m−8 > 23m+1 > 22m (2m + 1),
where the second to last inequality follows since 8m − 8 > 3m + 1, this is true since 5m > 9 for m ≥ 3.
5. • If 22m−2 | p + 1, then p + 1 ≥ 22m−2 , it follows that p − 1 ≥ 22m−2 − 2 ≥ 22m−3 (We use that 2k − 2 ≥ 2k−1
for k ≥ 2), and p2 + 1 > (p − 1)2 > 24m−6 , hence
(p2 + 1)(p + 1)(p − 1) > 24m−6 · 22m−3 · 22m−2 = 28m−11 > 23m+1 > 22m (2m + 1),
where the second to last inequality follows since 8m − 11 > 3m + 1, this is true since 5m > 12 for m ≥ 3.
6. We conclude that there is no pair that works for m ≥ 2, so the only pair that satisfy the equation is (2, 3).

Problem 2
A magic square is a 3 × 3 grid filled with numbers such that the sum of the numbers in each row, each column, and
both main diagonals is equal to S. Prove that the sum of the squares of the numbers in the top row is equal to the
sum of the squares of the numbers in the bottom row.
Solution. Suppose the numbers in the magic square are a, b, c, d, e, f , g h, i, arrearage like in the picture.

a b c
d e f
g h i

So we want to show that a2 + b2 + c2 = g 2 + h2 + i2 . To simplify some notation, let x = S − e, hence we would

have that a + i = x, b + h = x and c + g = x. Then we have that
a2 + b2 + c2 = g 2 + h2 + i2
⇔ a2 − i2 + b2 − h2 + c2 − g 2 = 0
⇔ (a + i)(a − i) + (b + h)(b − j) + (c + g)(c − g) = 0
⇔ x(a − i) + x(b − j) + x(c − g) = 0
⇔ x(a + b + c − i − h − g) = 0
⇔ x(S − S) = 0
Problem 3.
a) What is the largest number of integers from 1 to 2n that can be chosen such that among them, there are no
three numbers, one of which is equal to the sum of the other two?
b) Prove that for n ̸= 3 there exists exactly one optimal example..
Solution. a) the largest number is n+1. We can check that if x, y are two different integers among n, n+1, . . . , 2n,
then x + y ≥ 2n + 1 which is not in the list. So the numbers from n to 2n is a list of n + 1 with the given property.
Now suppose we have a set A with k integers from 1 to 2n, and let x be the minimum of A, then A − x =
{a − x : a ∈ A} is a set of k integers from 0 to 2n, we also have that A \ {x} has k − 1 integers from 0 to 2n, we
have then that if k + (k − 1) > 2n + 1, that is if k > n + 1, then (A − x) ∩ (A \ {x}) ̸= ∅, then there exists z in the
intersection, that means that z ∈ A and z ∈ A − x, the second one tell us that z = y − x for some y ∈ A, and also
note that y ̸= x since the z > 0, while z is bigger than both x and y. So z = x + y, where x, y, z are three different
integers in the set A.
b) We will show that the only example with n + 1 elements is the set {n, n + 1, . . . , 2n}. Suppose we have A
a set of n + 1 integers from 1 to 2n with the given property, and similarly as the previous proof, take x to be the
minimum of A.
It must hold that (A − x) ∩ (A \ {x}) = ∅, otherwise as in the previous case we would have a triple x, y, z such
that x + y = z. But since A − x has n + 1 elements and A \ {x} has n, it follows that every element from 0 to 2n is
in one of these sets.
We have that the numbers 1, 2, . . . , x must be in the set A − x, since no number smaller than x is in A \ {x}, by
definition we have that the numbers x + 1, x + 2, . . . , 2x must be in A. We will need to divide in three cases:
• If x = n, then A = {n, n + 1, . . . , 2n} which is the example we already know, so suppose x < n.
• If x = 1 (remember that n ≥ 3), then 2 ∈ A, so 3 ̸∈ A, we get that 3 ∈ A − 1, so 4 ∈ A, then it is not possible
5 to be in A, since 5 = 1 + 4, so 5 ∈ A − 1, but then 6 ∈ A, which is a contradiction since 2 + 4 = 6, with the
three elements in A.
• If 1 < x < n. Since x, x + 1 ∈ A, then 2x + 1 ̸∈ A, then (since 2x + 1 < 2n), 2x + 1 ∈ A − x, by
definition 3x + 1 ∈ A, this would be a contradiction since both x + 1 and 2x are in A, we would have that
3x + 1 = (x + 1) + 2x. Note that since x ̸= 1, we have that x + 1 ̸= 2x.

Note: For n = 1 indeed, {1, 2} is the unique example, while for n = 2 we have two examples {2, 3, 4} and {1, 2, 4}.

Problem 4.
Triangle ABC is inscribed in a circle ω. Tangents to ω are drawn through points A and B, intersecting at point
P . Points X and Y are the orthogonal projections of point P onto lines AC and BC, respectively. Prove that line
XY is perpendicular to the median of triangle ABC drawn from vertex C.
Solution. We will use the fact that the symmedian from C intersect the perpendicular bisector of AB at the
point of intersection of the tangents to the circumcircle through A and B, in this case P C is a symmedian.
Let E be the point of intersection of XY with CD
Note that P XCY is a cyclic quadrilateral. Then ∠P XY = ∠P CY = ∠DCA = ∠ECX. Then ∠XY C =
90◦ − ∠P XY = 90◦ − ∠ECX, therefore the triangle XEC is a right triangle with right angle at E, that is CD is
perpendicular to XY .

Problem 5.
In a right triangle ABC with hypotenuse AC, let K be the point of tangency of the incircle with side BC. Denote
by M and L the centers of the incircles of triangles ABK and ACK, respectively. The line AK intersects the
circumcircle of triangle ABC at point D, and the line M L intersects this same circle at points U and V . Let N be
the midpoint of U V , and let P be the point of intersection of M L with AK. Prove that DC = 2 P N .
Solution.
 1. Let E and F be the tangent points of the incircle of ABC to the sides AB and AC respectively. Let H, O, P1
be the tangent points of the incicrcle of ACK to the sides AC, CK and KA respectively. Let Q, G, P2 be the
tangent points of the incircle of ABK to the sides AB, BK and KA respectively.
2. We will show first that P1 = P2 , from this it will follow that the points M , P1 and L are colinear, so in fact
P1 = P and we would have that the line U V is perpendicular to the line AK.
3. Note that |CF | = |CK| and |CH| = |CO| so we have that |F H| = |CF | − |CH| = |CK| − |CO| = |KO|, and
similarly |EG| = |KQ|.
|AC| + |AK| − |KC| |AB| + |AK| − |KB|
4. We have that |AH| = , while |AG| = . (This is a property of the
2 2
tangent points of the incircle).
5. Note that |AH| = |AG|, this is equivalent as having |AC| − |KC| = |AB| − |KB|, which is also equivalent to
|AC| + |BK| = |AB| + |CK|, this is true since both numbers are the semiperimer of triangle ABC.
6. Since |AH| = |AG|, it follows that |HF | = |AF | − |AH| = |AG| − |AE| = |GE|, and from here we get that
|KQ| = |EG| = |HF | = |KO|.
7. We also have that |KO| = |KP1 | and |KQ| = |KP2 |, so in fact |KP1 | = |KP2 |, which means that P1 = P2 .
8. Now let X be the mid point of DC, then, since the chords U V and DC are parallel, both perpendicular to
AD, the line XN is a diameter of the circumcircle of ABC, hence XN is perpendicular to both CD and U V ,
in particular the quadrilateral XP N D is a rectangle, hence |XD| = |N P |, from here |CD| = 2|XD| = 2|N P |.

Problem 6.
Given coprime positive integers x and y. Find the largest integer n for which there are no such non-negative integers
a and b, that n = ax + by.
Solution. The largest number is ab − a − b.

1. First, we will show that there is no solution for the equation ax + by = ab − a − b for x, y non-negative integers.
2. Suppose otherwise, that a(x + 1) + b(y + 1) = ab, then a | b(y + 1) while a | a(x + 1), since a and b are coprime,
then a | y + 1 and b | x + 1, but then since both x + 1 and b + 1 are positive, we get that y + 1 ≥ a, and
x + 1 ≥ b, then ab = a(x + 1) + b(y + 1) ≥ ab + ab = 2ab, a contradiction.
3. Now we will show that if n > ab − a − b, then there exists non negative x, y such that ax + by = n.
4. By Bézout’s Identity (also known as Bachet-Bézout’s Theorem), there exists solutions to the equation ax+by =
n, with x and y integers, moreover, there are infinite solutions, since for a solution (x0 , y0 ) we have that
(x0 + kb, y0 − ka) is a solution for any integer k.
 5. Among the different solutions, choose a solution (x1 , y1 ), where x1 is non-negative and of minimal value, we
will show that y is no negative.

6. Note that x1 ≤ b−1, since otherwise we would (x1 −b, y1 +a), would be a solution with a smaller non-negativa
value x. Now, if y was negative, then y ≤ −1, it would follow that n = ax1 + by1 ≤ a(b − 1) − b = ab − a − b,
which contradincts the choice of n. Therefore y is non negative. Therefore there exists non-negative integers
x and y such that ax + by = n.

Solution 2.

1. Let us call an integer n good if it can be expressed as ax + by for non-negative integers x and y. Otherwise,
we call it bad.
2. It is obvious that if a number t is good, then t + a is also good.

3. Consider any remainder modulo a, denoted as r. We know that 0 ⩽ r ⩽ a − 1. Consider the set of integers

Sr = {ka + r | k ∈ Z}.

From the previous paragraph, we can conclude that some of the smallest numbers in Sr are bad, while all
larger numbers are good.

4. What is the smallest good number in Sr ? It must be equal to sb for some s < a, because if it is of the form
sb + ta with t > 0, then it is not the smallest, as sb would be smaller. If s ⩾ a, then (s − a)b is smaller than
sb.
5. Consider the numbers 0·b, 1·b, 2·b, . . . , (a−1)·b. These numbers yield distinct remainders modulo a. Therefore,
each of these numbers is the smallest good number in the corresponding Sr .

6. Thus, the smallest good numbers in the different sets Sr are

0 · b, 1 · b, 2 · b, . . . , (a − 1) · b

(though it is difficult to determine which number belongs to which set). Consequently, the largest bad numbers
are
0 · b − a, 1 · b − a, 2 · b − a, . . . , (a − 1) · b − a.
The largest among these numbers is
(a − 1)b − a = ab − a − b.

7. Therefore, ab − a − b is the largest bad number overall.

Problem 7.
Do there exist such two functions f : R −→ R and g : R −→ R, that f (g(x)) = x2 , а g(f (x)) = x3 for all real x?
Answer. No, there are no such functions.
Solution. Suppose by contradiction that there exist such functions.

1. f is an injection. Suppose f (a) = f (b) for some a ̸= b. Then a3 = g(f (a)) = g(f (b)) = b3 , but a3 ̸= b3 .

2. Consider h(x) = f (g(f (x))). On the one hand, h(x) = f (x3 ), on the other hand, h(x) = f (x)2 .
3. Thus, f (x3 ) = f (x)2 . Denote f (−1) = a, f (0) = b, f (1) = c. We know that all three numbers a, b, c are
the roots of equation t = t2 . But there are only two roots of this equation; hence, there should be two equal
numbers among a, b, c. But f is an injection, so a, b, c must be distinct.
Problem 8.
On the board, there are 2500 numbers. It is known that no matter how they are divided into 50 groups of 50
numbers each, there will always be a group whose sum of numbers is zero. What is the largest n for which it can
be guaranteed that there are at least n identical numbers among the numbers on the board?
Answer 2451.
Solution. An example for the number 2451 is obvious: simply take 2451 zeros and any additional 49 numbers.
Then at least one of the groups will consist entirely of zeros.
Now, let’s proceed to the estimate. Assume the contrary. First, consider the case where there are no negative
numbers or no positive numbers (without loss of generality, assume there are no negative numbers). If there are
at least 50 positive numbers, we can place one positive number in each group and distribute the rest arbitrarily,
ensuring that the sums in all groups are positive, which contradicts the condition. Thus, in these cases, the estimate
is proven.
Next, assume that there are both positive and negative numbers.
Zeros (if they exist) will be distributed evenly among the groups: we add one zero to each group, then another
zero to each group, and so on until the zeros are exhausted. Let the groups be denoted as S1 , S2 , . . . , S50 . Since no
single number occurs more than 2450 times, no group will consist entirely of zeros. Denote and order the nonzero
numbers as x1 ≤ x2 ≤ · · · ≤ xn . These numbers will be divided into groups in ascending order: starting from the
smallest number, we add it to the group with the smallest index that is not yet filled.
With this division, there can be at most one group Sk with a sum of zero: in such a group, there must be both
positive and negative numbers. This means that in the groups with smaller indices, there are no positive numbers,
and in the groups with larger indices, there are no negative numbers.
Assume k ̸= 50. Try to select a positive number from Sk and swap it with an arbitrary number from a group
with a larger index. The sum in the latter group will remain positive, so the sum in Sk will still equal zero. This
implies that all numbers in the groups with larger indices are positive and equal to each positive number in Sk .
Consequently, all positive numbers are equal, and there are no zeros in the groups with indices greater than k.
Similarly, Sk itself has no zeros, because otherwise, we could swap such a zero with any positive number from S50 .
Assume k ̸= 1. Using similar reasoning with the negative numbers in Sk and the groups with smaller indices,
we find that there are no zeros at all, all negative numbers are equal, and all positive numbers are equal. Let the
absolute value of the negative number (without loss of generality) be at least as large as the positive number.
Then we can swap a positive number from Sk with a negative number from S1 . The sum in Sk will change and no
longer equal zero. Meanwhile, in S1 , which now contains one positive number and the rest negative, the sum will
also no longer equal zero, since the absolute value of the negative number is at least as large as the positive. This
contradiction means that k = 1 or k = 50.
Assume, without loss of generality, that k = 1. Since 1 ̸= 50, the conclusion from the paragraph assuming k ̸= 50
still holds. Thus, in groups with indices greater than 1, all numbers are positive and identical. Furthermore, S1
contains a positive number, which is also equal to these positive numbers. Therefore, there are at least 50 · 49 + 1 =
2451 identical numbers.
The case where k = 50 is considered similarly, resulting in at least 2451 identical negative numbers.