Question 1:

4 parallel planes are given in a 3-dimensional space.

Prove that there exist a regular tetrahedron such
that every vertex is on exactly one of those planes.
Solution 1:
Let’s look at a different problem:
Suppose there are 3 parallel lines in a plane.
Show that there exists an equilateral triangle such
that every vertex is on exactly one of those lines.
Take a point P on the middle line.
Construct points B and C such that
∠BPX=∠CPX =60o

Now, construct the circumcircle of ΔPBC. Let it

intersect the middle line at A (other than P).

ΔABC is equilateral because ∠BCA = ∠BPA =60o

and ∠CBA = ∠CPA = 60o.
Now, let the 4 parallel planes be P1, P2, P3 and P4.
Consider a line L1 on P1 perpendicular to the plane of
this paper. Consider a plane T1 which is at an angle β1
with P1 as shown below, where β1 is sufficiently close
to 180o.

Let the intersection lines of T1 with P3 and P4 be C1 and

C2 respectively.
So, L1, C1 and C2 are parallel lines on the plane T1.
Therefore, we can construct an equilateral triangle with
each vertex on each of these lines. Therefore, these
vertices lie on P1, P3 and P4. Now, considering this
triangle as the base of a pyramid, construct a point F1
on one side of the plane T1 such that F1 along with
those 3 points form a regular tetrahedron.
Since, β1 is close to 180o, F1 must lie below P2.
Now, rotate this plane about L1 in anti-clockwise sense.
Suppose we have reached a plane T2 passing through L1
at some sufficiently small angle α1 as shown below:

Let the intersection lines of T2 with P3 and P4 be D1 and

D2 respectively.
We perform the similar construction like last time.
L1, D1 and D2 are parallel lines on the plane T2.
Therefore, we can construct an equilateral triangle with
each vertex on each of these lines. Therefore, these
vertices lie on P1, P3 and P4. Now, considering this
triangle as the base of a pyramid, construct a point F2
on the same side of the plane T2 such that F2 along with
those 3 points form a regular tetrahedron.
F2 must lie above P2 as α1 is close to 0o.
Since, this rotation is continuous in nature, therefore
the locus of the fourth vertex must form a
continuous path.
Therefore, there exists a stage where the fourth
vertex lies on P2.
So, we have a regular tetrahedron with each vertex
on exactly one plane.
Question 2:
Let z1, z2, …, zn ∈ ℂ such that |zk| ≤ 1 for k=1, 2, …, n.
Show that there exist e1, e2, …, en ∈ {1, –1} such that
|e1z1 + e2z2 + … + emzm| ≤ 2
for any m ≤ n.
Solution 2:
Call a finite sequence of complex numbers {zk}𝑛𝑘=1 a
green sequence if |zk| ≤ 1 for all k=1, 2, …, n.
Call a green sequence happy if it has a friend
sequence {ek}𝑛𝑘=1 of 1’s and –1’s such that
|e1z1 + e2z2 + … + emzm| ≤ 2
for any m ≤ n.

We will show by induction on n that all green

sequences are happy.
For n=1, this is true.
For n=2, choose e1 = e2 = 1.
Then, |z1| ≤ 1 < 2 and |z1 + z2| ≤ |z1| +|z2| ≤ 2.
Now assume, every green sequence of length n is
happy.
Consider a green sequence {zk}𝑛+1
𝑘=1 of length n+1.
Mark the complex numbers zk in the complex plane.
For each k, let lk be the line passing through O and zk.
Consider the lines l1, l2 and l3.

Each of the six angles cannot be more than 60o. So,

we have two lines among l1, l2 and l3 such that the
acute angle between them ≤ 60o.
WLOG, assume they are l1 and l2.
Note that one of z1 + z2 and z1 – z2 has modulus at
most 1 because the angle between l1 and l2 ≤ 60o.

Call it z’= z1 + τ z2 where τ is either +1 or –1.

Now, consider the green sequence {z’, z3, …, zn+1}.
This is of length n. Hence, it must be happy and thus
has a friend sequence. Let the friend sequence be {e’,
e3, …, en+1}. Take e1 = e’ and e2 = τ e’.
Consider the sequence {e1, e2, e3, …, en+1}
Is it a friend of {zk}𝑛+1
𝑘=1 ?
We have:
|e1z1| = |z1| ≤ 1 < 2
|e1z1 + e2z2| = |e’ (z1 + τ z2) | = |e’z’| = |z’| ≤ 1 < 2
For all m>2, m ≤ n+1:
| e1z1 + e2z2 + e3z3 + … + emzm|
= | e’z’ + e3z3 + … + emzm| ≤ 2
because {e’, e3, …, en+1} is a friend of { z’, z3, …, zn+1}.

Hence, {zk}𝑛+1
𝑘=1 is happy :)
Hence proved.
Question 3:
A positive integer n is given. Two players take turns
marking points on a circle with circumference 1 unit.
The first player uses the red colour while the second
player uses the blue colour. When n points of each
colour have been marked, the game is over, and the
circle has been divided into 2n arcs. The winner is
the player who has the longest arc among these 2n
arcs both endpoints of which are of this player’s
colour. If no such arc exists, the game ends in a draw.
If both the players are playing intelligently, who has
a winning strategy?
Solution 3:

The first observation is that player 2 can ensure that

player 1 doesn’t win the game by marking the blue
points in such a way so that the red and blue points
appear alternately. Here, the game ends in a draw.
So, player 1 doesn’t have a winning strategy. Now,
we need to check whether player 2 can ensure
his/her victory.
Suppose the second player uses the following
strategy:
After the first player has marked the initial red
point, the second player defines as principal points
the remaining vertices of a regular n-gon inscribed
in the circle, with this red point as one of the
vertices.
The second player marks principal points whenever
possible, until all have been marked. Since the
second player has n moves and there are only n−1
unmarked principal points initially, this stage ends
before the second player’s last move.
Once all the principal points have been marked, the
second player find pairs of adjacent red principal
points. For each such pair, the second player marks a
blue point between the two red points. Suppose the
first player has marked k principal points red while
the second player has marked the remaining n − k
principal points blue. There are at most k − 1 pairs
of adjacent red principal points. Hence this stage
also ends before the second player’s last move.
When the second player is ready to make the last
move, all n principal points have been marked.
There are n−1 other marked points. Hence there
exist two adjacent principal points with no other
points in between because there are 2n–1 marked
points on the circle. At least one of them is blue since
the second player ensured there is a blue point
between any two adjacent red principal points.
The second player’s final marked point is on this arc,
arbitrarily close to a principal point where the other
principal is blue.

The longest arc for the second player can be made

1
arbitrarily close to , while all arcs for the first
𝑛
1
player are < . Hence the second player can be
𝑛
assured of a win regardless of any action by the first
player.
Question 1:
In the plane are marked all points with integer
coordinates (x, y), 0 ≤ y ≤ 10. Consider a polynomial
of degree 20 with integer coefficients. Find the
maximal possible number of marked points which
can lie on its graph.
Solution 1:
We need to consider integer solutions of the system
of inequalities:
0 ≤ P(x) ≤ 10 … (∗)
Let us prove by contradiction that there are no more
than 20 integer solutions to (∗).
Assume that x1 < x2 < . . . < x21 satisfy (∗); denote
a=x1, b = x21. So, b − a ≥ 20.
Consider P(b) − P(a). Since both a and b satisfy (∗)
we conclude that |P(b) − P(a)| ≤ 10. However, since
P(x) has integer coefficient, the number P(b)−P(a)
must be divisible by (b − a). Since |P(b) − P(a)| ≤
10 and b − a ≥ 20, it implies that P(b) − P(a) = 0.
So, P(a) = P(b) = c with 0 ≤ c ≤ 10. Then P(x) =
(x − a)(b − x)Q(x) + c
where Q(x) is a polynomial of degree 18.
Observe that (x − a)(b − x) ≥ 19 for a+1 ≤ x ≤ b−1.
Then P(x) cannot satisfy (∗) unless Q(x) = 0. This is
because if Q(x) ≠ 0, then either P(xk ) ≤ −19 + c < 0
or P(xk ) ≥ 19 + c > 10 for some k.
Therefore Q(xk) = 0, k = 2, . . . , 20 but polynomial
Q(x) of degree 18 cannot have more than 18 roots.
(Contradiction!)

On the other hand, for P(x)=(x−1)(x−2)…(x−20)

passes through 20 marked points (k, 0), k=0,…,20.
So, 20 is the required answer.
Question 2:
Does there exist 2020 consecutive natural numbers
such that none of them is a perfect power?
Solution 2:

Let us look at a different problem:

Does there exist 2020 consecutive natural numbers
such that none of them is a power of some prime?
The answer to this question is yes. It follows from
Chinese Remainder Theorem. Take 4040 distinct
primes p, p2, p3, p4, … , p4039, p4040.
Consider the following system of congruences:
N≡0 (mod p1p2 )
N≡–1 (mod p3p4)
N≡–2 (mod p5p6)
…
N≡–2019 (mod p4039p4040)
By Chinese Remainder Theorem, this system has
infinitely many solutions N ∈ ℕ. Let N0 be one such
solution.
Note that each of N0, N0+1, N0+2, …, N0+2019 is
NOT power of some prime because each of them has
at least 2 distinct prime factors.
However, this only guarantees that each of these
numbers are not power of some prime as each
contains at least two prime factors.
It might happen that some of these numbers are
perfect powers.
For the given problem:
Method I:
Take n distinct primes p1, p2, …, pn.
Consider the following system of congruences:
N≡p1 – 1 (mod p12 )
N≡p2 – 2 (mod p22 )
…
N≡pn – n (mod pn2 )
Since pi2 are mutually co-prime, this system of
congruence has infinitely many solutions N ∈ ℕ.
Let N0 be one such solution.
Now, N0+1 ≡p1(mod p12 ). So, p1 divides N0+1 but
p12 doesn’t.
In general, N0+i ≡pi (mod pi2 ) for all i=1, 2, …, n.
So, pi divides N0+i but pi2 doesn’t.
Therefore, N0+i cannot be a perfect power for all
i=1,2,…,n.
This is true for all n ∈ ℕ.
Method II:

Let us denote the kth perfect power by nk.

Let dk = nk+1 – nk, k≥1.
If we show that this sequence {dk} is unbounded,
then the problem is solved.
So, assume that this sequence is bounded above B.
This is equivalent to:
If we take any consecutive B natural numbers, there
is at least one perfect power among them.
Now, what is the minimum perfect tth power >1?
The answer is 2t.
Now, let us take N=22n.
Number of perfect squares ≤N = 2n
2𝑛
Number of perfect cubes ≤N = ⌊ 2 3 ⌋ < 2n
2𝑛
Number of perfect fourth powers ≤N = ⌊ 2 ⌋ < 2n 3

…
Number of perfect 2nth powers ≤N = 2 < 2n
____________________________________________________________
Therefore, number of perfect powers ≤N < (2n).2n

Now, if f(z) denotes the number of perfect powers

≤z, then:
𝑓(22𝑛 ) 2𝑛. 2𝑛 𝑛
< =
22𝑛 22𝑛 2𝑛−1
In other words, the proportion of perfect powers
𝑛
among 1, 2, …, 22n is bounded above by 𝑛−1.
2

By Sandwich Theorem,
𝑓(22𝑛 )
lim =0
𝑛→∞ 22𝑛
Hence, there exists K ∈ ℕ such that the proportion of
1
perfect powers among 1, 2, …, 22k is < ……………..(*)
𝐵
Let us partition these numbers into blocks of length
B containing consecutive natural numbers starting
from 1.
1,
⏟ 2, … , 𝐵 𝐵
⏟ + 1, 𝐵 + 2, … , 2𝐵 , … … …
𝑏𝑙𝑜𝑐𝑘 1 𝑏𝑙𝑜𝑐𝑘 2

There are finitely many blocks and the final block

may not contain exactly B elements.
Recall that:
If we take any consecutive B natural numbers, there
is at least one perfect power among them.
So, each block contains at least one perfect power.
The last block contains at most B elements and 22n, a
perfect power is in it.
So, the proportion of perfect powers among 1, 2, …,
1
22k is at least .
𝐵
This contradicts (*).
Hence, {dk} is unbounded.
Therefore, there exists k ∈ ℕ such that dk > 2020.
Hence, for this particular k, nk+1, nk+2, …, nk+2020
are 2020 consecutive non-perfect powers.
Question 3:
a) Show that the numbers 1,2, …, 16 can be arranged
in a row such that sum of any two consecutive
numbers is a perfect square.
b) Determine all positive integers d such that there
exists an integer k≥3 such that one can arrange the
numbers d, 2d, …, kd in a row, such that the sum of
every two consecutive of them is a perfect square.
Solution 3:
a)
8, 1, 15, 10, 6, 3, 13, 12, 4, 5, 11, 14, 2, 7, 9, 16
b) For this part, d=1 is a solution.
Note that d is a solution if and only if dx2 is a
solution for some x∈ℕ.
So, we conclude that every perfect square is a
solution. Indeed,
8k2, k2, 15k2, 10 k2, 6k2, 3k2, 13k2, 12k2, 4k2, 5k2,
11k2, 14k2, 2k2, 7k2, 9k2, 16k2
clearly satisfies the conditions for some perfect
square k2.
Let us consider some N, a non-perfect square.
We remove every even prime powers to make it
square-free. Let the number obtained be d.
Is d=1?
Of course not, because otherwise N must have a
perfect square.
Since d>1, it is divisible by some p≥2. Also, since d
is square-free, p2∤d.
Let (a1d, a2d, …, akd) be a permutation of (d, 2d, …,
kd) such that (ai+ai+1)d is a perfect square for all i,
for some k≥3.
If p=2, then 4|(ai+ai+1)d for all i. Now, 2|d and 4∤d.
Hence, 2|ai+ai+1 for all i. Therefore, ai’s are of same
parity for all i which is clearly impossible as (a1, a2,
…, ak) is a permutation of (1, 2, …, k).
Now, for p≥3:
By similar argument,
p|ai+ai+1 for all i.
Therefore, if p|a1, then it implies p|a2 which is turn
implies p|a3 and so on.
This is definitely not possible as (a1, a2, …, ak) is a
permutation of (1, 2, …, k).
So, p∤a1 which in turn implies p∤ai for all i.
Therefore, k<p.
Hence, p>3 as k≥3.
p≠4 ⇒ p≥5.
Let at = 1.
Then at+1 ≡ –1 (mod p) and at-1 ≡ –1 (mod p). So,
at+2 ≡ +1 (mod p) and at-2 ≡ +1 (mod p) and so on.
In general, we have ai ≡ +1 or –1 (mod p) which is
again not possible.
Hence, d cannot have any factor ≥2.
Therefore, d=1 is the only possibility.
So, the solutions are every perfect square.
Question 4:
Find all continuous functions f: ℝ→ℝ such that
f(x) = f(x2 + ¼)
for all x ∈ ℝ.
Solution 4:
First of all, note that f is even.
So, it is sufficient to look at x≥0.
Let x0≥0. Consider two cases:
Case I: 0 ≤ x0 ≤ ½:
Consider the sequence {xn} given by
xn+1 = xn2 + ¼
Since, x0 ≤ ½, xn ≤ ½ implies xn+1 ≤ ¼ + ¼ = ½. So,
by induction, we conclude the {xn} is bounded above
by ½.
Also, xn+1 – xn = xn2 – xn + ¼ = ( xn – ½)2 ≥ 0.
So, {xn} is increasing as well as bounded.
Therefore {xn} is convergent. The limit L = ½.
Now, f(x0) = f(x1) = … . Since f is continuous, we
have, f(x0) = lim f(xn) = f(½).
𝑛→∞

Therefore, f(x) = f(½) for all x ∈ [0, ½]

Case II: x0 > ½:
Define {xn} given by
1
xn+1 = √𝑥𝑛 2 −
4

By similar argument like in the previous case, {xn} is

bounded below by ½ and is monotonically
decreasing.
Therefore, it is convergent and its limit is ½.
So, by continuity of f, we conclude:
f(x0) = lim f(xn) = f(½).
𝑛→∞
Hence, f(x) = f(½) for all x > ½.

Overall, we have f(x)=f(½) for all x ∈ ℝ.

 7. We may assume, without loss of generality, that a ≤ b. We differentiate
the given identity to get

f (x) = af (ax) + bf (bx) (1)

and thus

f (x) = a af a2 x + bf (abx) + b af (abx) + bf b2 x

= a2 f a2 x + 2abf (abx) + b2 f b2 x .

An easy induction shows that for any integer n ≥ 1 we have

n  
X n
ak bn−k f ak bn−k x


f (x) = (2)
k
k=0

Let ϵ > 0 be given. Since f is continuous, there exists δ > 0 such that
|f (x) − f (0)| < ϵ if 0 ≤ x < δ.
Let x0 ∈ [0, 1]. Since 0 < b < 1, limn→∞ bn x0 = 0 and so there exists N > 0
such that 0 ≤ bn x0 < δ if n ≥ N .
Thus for n ≥
N and 0 ≤ k ≤ n we have 0 ≤ ak bn−k x0 ≤ bn x0 < δ and so
f ak bn−k x0 − f (0) < ϵ.
Therefore if n ≥ N , then, by (2),

n   n  
X n k n−k k n−k


X n k n−k
|f (x0 ) − f (0)| = a b f a b x0 − f (0) + f (0) a b − f (0)
k k
k=0 k=0
n  
X n
ak bn−k f ak bn−k x0 − f (0) + |f (0) ((a + b)n − 1)|


≤
k
k=0
< (a + b)n ϵ + |f (0) ((a + b)n − 1)| . (3)

f a + b = 1, then (3) gives |f (x0 ) − f (0)| < ϵ and thus f (x0 ) = f (0), i.e. f
is constant.
If a + b < 1, then (1) with x = 0 gives f (0) = 0. Hence, by (3), |f (x0 )| <
(a + b)n ϵ < ϵ and so f (x0 ) = 0, i.e. f ≡ 0.

8. (a) Since, f is monotone decreasing, the function F (x) = f (x) − x is 0

for some α as f (0) > 0 ⇒ f (f (α)) = α
For, y > x we have, f (f (x)) − f (f (y)) ≥ f (y) − f (x) > x − y Thus, g(x) =
f (f (x)) − x is monotone, so α is unique.

1
(b) WLOG, we assume, x2 ≥ x1 Then, x1 ≤ x3 ≤ x5 . . . which is bounded above
by x2 and x2 ≥ x4 ≥ . . . which is bounded below by 0 .
Hence the even and odd subsequences are convergent. Let, {x2n } → l, {x2n−1 } →
m.
So, l = f (m), m = f (l). Thus, f (f (l)) = l, f (f (m)) = m.
But α is the unique solution of f (f (x)) = x. Hence we have, l = m.