STAR CIRCLE PROGRAM. Assignment 1. Solutions.

Problem 1. Do there exist 4 distinct positive integers a, b, c, d which satisfy

a! · b! = c! · d!?

Solution. We can take for example a = 3, b = 5, c = 1 and d = 6, and note that

3! · 5! = 6 · 5! = 6! = 1! · 6!.

Problem 2. Real a, b, c, d satisfy

a2 + b2 + (a + b)2 = c2 + d2 + (c + d)2 .

Prove that
a4 + b4 + (a + b)4 = c4 + d4 + (c + d)4
Solution. You can check that a4 + b4 + (a + b)4 = 2(a2 + ab + b2 )2 after expanding it manually. Similarly, for c
and d, we have: c4 + d4 + (c + d)4 = 2(c2 + cd + d2 )2 . Now, given that: a2 + b2 + (a + b)2 = c2 + d2 + (c + d)2 , it
follows that: a2 + ab + b2 = c2 + cd + d2 . Thus, we conclude that: a4 + b4 + (a + b)4 = c4 + d4 + (c + d)4 .

Problem 3. Triangle ABC is inscribed in a circle ω with center O and radius r. Circles ω1 , ω2 , ω3 have centers
A, B and C respectively and radius r. It is given that ω1 and ω2 intersect at O and X, ω2 and ω3 intersect at O
and Y , ω3 and ω1 intersect at O and Z. Prove that point X, Y and Z form a triangle which is congruent to ABC.
Solution. We will show first that |AB| = |ZY |. Note that both BOCY and OCZA are rhombus, as each side
of these quadrilaterals is of length r, from here we have that BY ∥ OC ∥ AZ, hence BY and AZ are both parallel
of length r. Therefore BY ZA is a parallelogram, and we conclude that |AB| = |Y Z|.
Analogously, we conclude that |BC| = |ZY | and |CA| = |XY |. Hence, the triangle Y ZX is congruent to ABC.

Problem 4. Two graphs G1 and G2 of quadratic polynomials intersect at points A and B. Let O be the vertex
of G1 . Lines OA and OB intersect G2 again at points C and D. Prove that CD is parallel to the x-axis.
Solution: First, shift the entire picture so that point O coincides with the origin of the coordinate plane.
Let the function representing the graph G1 be given by:

y = tx2 ,
where t is a constant. Let point A have coordinates (a, ta2 ) and point B have coordinates (b, tb2 ). The difference
between the quadratic functions defining the graphs G1 and G2 is a quadratic function that vanishes at points a
and b. This implies that the graph of G2 is described by the equation:

y = tx2 + k(x − a)(x − b),

for some real constant k ̸= 0.
ta2
The line OA is given by the equation y = a x = tax, so point C can be found as the solution to the system:

y = tax, y = tx2 + k(x − a)(x − b).

Equating the right-hand sides, we get the equation:

tx2 + k(x − a)(x − b) = tax.

One solution to this equation is x = a, and the second solution is:
 kb
x= .
t+k
Substituting this value into the equation y = tax, we get the y-coordinate of point C:
kb kabt
y = ta · = .
t+k t+k
Notice that this expression is symmetric with respect to a and b. Therefore, if we perform the same steps for
points B and D, we will obtain the same y-coordinate for point D:
kabt
y= .
t+k
Thus, the y-coordinates of points C and D are equal, which implies that the line CD is parallel to the x-axis,
as required.

Problem 5. What is the smallest n such that for any collection A of 2016 sets, there exists a collection B of n
sets such that each set in collection A is the intersection of two distinct sets from collection B?
Solution. The smallest value of n is 2017.
Firs we will see there is always a collection of at most 2017 sets. Suppose X1 , X2 , . . . , X2016 are the sets in
A, then let X be the union of all these sets and let B be the collection of the 2016 sets in A adding X, since
Xi = Xi ∩ X, every set in A is the intersection of two sets in B.
Now, to see that 2017 is the smallest number, consider the collection of sets Xi = {1, 2, . . . , 2016} \ {i}, that is,
Xi is the set whose elements are the integer from 1 to 2016 except i.
Since for each i, the set Xi is the intersection of two sets in B, we conclude that Xi is contain in at least two
sets of B, more over, one of these sets must not contain the element i, since otherwise the intersection would also
contain i, hence there is a set Yi in B such that Yi contains every integer from 1 to 2016 except i (possible containing
more elements), this means that B contains at least 2016 different sets.
But note that Yi ∩ Yj is not a set of A, since this set is missing both i and j as elements, hence there must be
at least one more set in B. Therefore B has at least 2017 sets.

Problem 6. On sides BC, CA, AB of a triangle ABC points K, L, M are chosen, respectively, and a point
P is inside ABC is chosen so that P L ∥ BC, P M ∥ CA, P K ∥ AB. Determine if it is possible that each of three
trapezoids AM P L, BKP M , CLP K has an inscribed circle.
Solution. It is not possible. Suppose by contradiction each of the trapezoids has a circle inscribed. If a
quadrilateral has a circle inscribed, then the sum of opposite sides is the same, hence |AM | + |P L| = |AL| + |P M |,
|BK| + |P M | = |BM | + |P K| and |CL| + |P K| = |CK| + |P L|. Adding this equalities we have that

|AM | + |P L| + |BK| + |P M | + |CL| + |P K| = |AL| + |P M | + |BM | + |P K| + |CK| + |P L|,

or
|AM | + |BK| + |CL| = |AL| + |BM | + |CK|. (1)
 Let K ′ , L′ and M ′ the intersection of KP with AC, the intersection of LP with AB and the intersection of
M P with BC respectively. Since AM P K ′ , BKP L′ and KP LC are parallelograms we have that |AM | = |K ′ P |,
|BK| = |L′ P | and |CL| = |M ′ P |, also |AL| = |AK ′ |+|K ′ L| = |M P |+|K ′ L|, |BM | = |BL′ |+|L′ M | = |KP |+|L′ K|
and |CK| = |CM ′ | + |M ′ K| = |LP | + |M ′ K|, substituting in equation (1) we have that

|K ′ P | + |L′ P | + |M ′ P | = |M P | + |K ′ L| + |KP | + |L′ M | + |LP | + |M ′ K|,

but this is not possible, since by the triangle inequality on the triangles P KM ′ , P LK ′ and P M L′ we have that
|M ′ P | < |KP | + |M ′ K|, |K ′ P | < |LP | + |K ′ L| and |L′ P | < |M P | + |L′ M |, so in fact we should have

|K ′ P | + |L′ P | + |M ′ P | < |M P | + |K ′ L| + |KP | + |L′ M | + |LP | + |M ′ K|.

Solution 2. Suppose the opposite and we have a construction with all three trapezoids have inscribed circles.
Consider the trapezoid with the largest radius (or one of them). WLOG we can claim that ALP M has the largest
radius. This radius rA is equal to 12 d(M P, AL). Denote the radius of CLP K as rC . Since rC ⩽ rA any circle with
radius rC tangent to AC lies between parallel lines AC and M M ′ . So it cannot be tangent to segment KP .

Problem 7. For any positive integer n define S(n) the sum of decimal digits of the number n. Prove that:
a) S(a + b) ⩽ S(a) + S(b)
b) S(a · b) ⩽ S(a) · S(b)
c) 8 · S(8n) ⩾ S(n).
a) Proof. We will prove that S(a) + S(b) ≥ S(a + b) using induction on the number of digits in max(a, b). First,
both a and b are non-zero; if one is zero, then the left-hand side equals the right-hand side. We have:
j a k
S(a) = a0 + S
10
where a = a0 + 10(a1 + 10(a2 + · · · )).
Induction : Assume the statement holds for any a, b such that max(a, b) has n digits. Without loss of generality,
let a ≥ b and a have n + 1 digits, written as a0 , a1 , . . . , an and b0 , b1 , . . . , bk .
Define:
a − a0 b − b0
aa = , bb =
10 10
If a0 + b0 ≤ 9, then:

S(a) + S(b) = a0 + b0 + S(aa) + S(bb), S(a + b) = a0 + b0 + S(aa + bb).

By induction, S(aa) + S(bb) ≥ S(aa + bb), hence:

S(a) + S(b) ≥ S(a + b).

Now assume a0 + b0 > 9. Then:

S(a) + S(b) = S(aa) + S(bb) + a0 + b0 ,

and
S(a + b) = a0 + b0 − 10 + S(aa + bb + 1).
Thus, we need to show:
S(aa) + S(bb) + a0 + b0 ≥ a0 + b0 − 10 + S(aa + bb + 1),
which simplifies to:
S(aa) + S(bb) + 10 ≥ S(aa + bb + 1).
Using induction on n:

S(aa + bb + 1) ≤ S(aa) + S(bb + 1) ≤ S(aa) + S(bb) + 1 < S(aa) + S(bb) + 10.

Otherwise, if bb + 1 has n + 1 digits, then bb = 10n − 1, which is an edge case (you can check it on your own use
that (a ≥ b)).
b) Proof: To prove the result S(a) · S(b) ≤ S(a · b), we will use the same symbols and induction argument as
before, with a few key changes. We have S(a) = a0 + a1 + · · · + an and S(b) = b0 + · · · + bk . Then,

S(a) · S(b) = a0 b0 + a0 (b1 + · · · + bk ) + b0 (a1 + · · · + an ) + (a1 + · · · + an )(b1 + · · · + bk ).

This can be rewritten as:

S(a) · S(b) = S(a0 ) · S(b0 ) + S(a0 ) · S(bb) + S(b0 ) · S(aa) + S(aa) · S(bb).

Now, we have S(a · b) = S(a0 · b0 + a0 · bb + b0 · aa + aa · bb) ≤ S(a0 · b0 ) + S(a0 · bb) + S(b0 · aa) + S(aa · bb). Using
the result from part a) that S(a) + S(b) ≥ S(a + b), we notice that a0 , aa, bb, b0 each have fewer than n + 1 digits,
so we can apply the induction on them.
c) Proof: To prove that 8 · S(8n) ≥ S(n), we start with S(n) = S(1000 · n). Thus, we need to show that
8 · S(8n) ≥ S(1000n). This can be established by proving the stronger version: 8 · S(m) ≥ S(125m). We have:

S(125m) = S(100m+20m+5m) ≤ S(100m)+S(20m)+S(5m), S(100m) = S(m), S(20m) = S(2m), S(5m) = S(5m).

Thus:
S(125m) ≤ S(m) + S(2m) + S(5m) ≤ S(m) + 2 · S(m) + 5 · S(m) = 8 · S(m).
Therefore, we conclude:
8 · S(m) ≥ S(125m),
which implies:
8 · S(8n) ≥ S(1000n) = S(n).
Solution 2. Parts a and b can be solved using a slightly different approach.
Definition 1: Define a decimal partition of a positive integer n as any representation of n as a sum of powers
of 10.
Definition 2: A decimal partition is called correct if it contains no set of 10 identical powers of 10.
Statement 1: The correct decimal partition is unique.
Proof: Suppose the contrary. Let n be the smallest number with two distinct correct decimal representations:

n = 10a1 + 10a2 + . . . + 10as = 10b1 + 10b2 + . . . + 10bt .

Without loss of generality, assume that a1 is the largest power on both sides. If any term on the right-hand side
(RHS) is equal to 10a1 , then the number n − 10a1 also has two distinct correct partitions.
 If all terms on the RHS are smaller than 10a1 , then the RHS is at most:
9 · 10a1 −1 + 9 · 10a1 −2 + . . . + 9 · 10 + 9 = 10a1 − 1,
which contradicts the equality. Therefore, the correct decimal partition must be unique.
Definition 3: Given any decimal representation, define the operation of merging: we take any set of 10 identical
powers of 10 and replace them with one larger power.
Statement 2: Any decimal partition can be transformed into a correct one after a finite number of merging
operations.
Proof: Perform merging operations arbitrarily. Each time, the number of terms in the partition decreases, so
this process cannot continue indefinitely. In the final partition, there cannot be 10 identical terms, ensuring the
partition is correct.
While these results may seem rather elementary and obvious, the previous parts of the solution may appear
somewhat tedious. However, they are necessary for the final conclusion.
Statement 3: The sum of the decimal digits of a positive integer n is equal to the smallest size of decimal
partition of n.
After this observation parts a and b are rather easy.
a) We know that there is decimal partition of a with S(a) terms and there is a partition of b with S(b) terms.
Hence there is a partition of a + b with S(a) + S(b) terms. Thus S(a + b) ⩽ S(a) + S(b).
b) We know that there is decimal partition of a with S(a) terms and there is a partition of b with S(b) terms.
If a = 10c1 + 10c2 + . . . + 10cS(a) and b = 10d1 + 10d2 + . . . + 10dS(b) then there is a partition of a · b with S(a)S(b)
terms (we can take 10ai +bj for all 1 ⩽ i ⩽ S(a) and 1 ⩽ j ⩽ S(b)). Thus S(a · b) ⩽ S(a) · S(b).

Problem 8. In the cells of an m × n table, distinct real numbers are written. Let l ⩽ m and k ⩽ n. In each
column, the l largest numbers are underlined, and in each row, the k largest numbers are underlined. Prove that at
least kl numbers are underlined twice.
Solution. We will prove the following statement by induction on A:
For any non-negative integers k and ℓ such that k + ℓ = A, and for any n ≥ k and m ≥ ℓ, at least kℓ numbers
will be underlined twice.
Base Case: For A = 0, 1, at least one of the numbers k or ℓ is zero, so the statement is obvious.
Inductive Step A → A + 1: We will prove the necessary statement for k + ℓ = A + 1. If either k or ℓ is zero,
the statement is obvious.
1. Case when both k and ℓ are positive:
We have a table T where numbers are underlined according to the conditions. We will show that in table T ,
at least k numbers are underlined twice.
2. Selecting the Critical Number:
Circle the numbers that are the k-th largest in their row or the ℓ-th largest in their column. Among all the
circled numbers, find the largest one. Without loss of generality, assume this number is aij (the number in
the i-th row and j-th column), which is the k-th largest in row i.
3. Proving All k Numbers in Row i are Underlined Twice:
We will prove that all k largest numbers in row i are underlined twice. Suppose not, and there exists a number
ais , which is among the k largest in row i but is not among the ℓ largest in column s. Let ats be the ℓ-th
largest number in column s. Then we have the chain of inequalities:
aij ≤ ais < ats .
However, ats is also circled due to column s. This contradicts the assumption that aij is the largest circled
number. Therefore, all k largest numbers in row i are underlined twice.
4. Constructing the Reduced Table T ′ :
Now, consider a new table T ′ obtained from T by removing row i. In each remaining row of T ′ , underline the
k largest numbers, and in each column, underline the ℓ − 1 largest numbers. It’s clear that numbers underlined
twice in T ′ were also underlined twice in T .
 5. Applying the Induction Hypothesis:
By the induction hypothesis, in table T ′ , at least k(ℓ − 1) numbers are underlined twice. In the removed row
i of table T , there are k numbers underlined twice. Therefore, in table T , there are at least:

k + k(ℓ − 1) = kℓ

numbers underlined twice. This completes the induction step.

Conclusion:
By induction, the statement holds for all non-negative integers k and ℓ satisfying k + ℓ = A. Therefore, in any
m × n table filled with distinct real numbers, where in each column the ℓ largest numbers are underlined and in
each row the k largest numbers are underlined, at least kℓ numbers are underlined twice.

Problem 9. Do there exist such 22021 distinct pairs of positive integers (ai , bi ), that
1 1 1
+ + ... + =1
a1 b1 a2 b2 a22021 b22021
and

a1 + a2 + . . . + a22021 + b1 + b2 + . . . + b22021 = 32022 ?

We will prove that there exist k = 2n distinct pairs (a1 , b1 ), (a2 , b2 ), . . . , (ak , bk ) such that

a1 + a2 + . . . + ak + b1 + b2 + . . . + bk = 32022

and
1 1 1
+ + ... + = 1.
a1 b1 a2 b2 ak bk
Base case: n = 2. We choose (1 + 2 + 1 + 1) + (2 + 2 + 6 + 12) = 27, and
1 1 1 1
+ + + = 1.
1 · 2 2 · 2 1 × 6 1 × 12
Inductive step n → n + 1. We replace each pair (a, b) with two pairs (a + b, b) and (a, a + b).
1. Proof that all pairs are distinct. Suppose there are two identical pairs. For both, either the first numbers
are larger than the second, or the second numbers are larger than the first. If the first numbers are larger, then the
pairs have the form (a + b, b) and (c + d, d). If these pairs are equal, then the original pairs (a, b) and (c, d) must
also be equal, which contradicts the inductive hypothesis.
2. Proof that the sum triples. This is obvious because the sum of the two new pairs is 3a + 3b, which is three
times the sum of the original parent pair.
3. Proof that the sum of the reciprocals remains unchanged. Indeed,
1 1 1
= + .
ab (a + b)b a(a + b)

Thus, all necessary conditions for the new pairs are satisfied. The inductive step is complete.