ISI-CMI Book

SciAstra

December 16, 2024

1
1 Number Systems
1. If x is irrational, then choose ξ = x2 . Clearly 0 < ξ < x.
√
If x is rational, then choose ξ = √x2 . Since 2 > 1, we have 0 < ξ < x.
(In fact there are infinitely many irrational numbers between any two real numbers.)

√ √
2. Suppose 2 + 5 = x = pq is a rational number with p, q ∈ Z
√ √
Then (x − 2)2 = 5 i.e. x2 − 2 2x + 2 = 5.
√ √ 2
Hence x2 − 3 = 2 2x, which gives 2 = x 2x−3 , a rational number.
√ √ √
This contradicts the fact that 2 is rational. So 2 + 5 is irrational.

√ √
3. Let xn = n−1+ n + 1 be rational. Then 1/xn is also rational.
But √ √ √ √ √ √
1 ( n + 1 − n − 1)( n − 1 + n + 1) n+1+ n−1
1/xn = √ √ = √ √ = .
n−1+ n+1 n+1− n−1 2
√ √
This means n + 1 − n − 1 is also rational.
√ √
So n − 1 and n + 1 are also rational,

i.e., (n − 1) and (n + 1) are perfect squares.

This
√ is not√possible as any two perfect squares differ at least by 3. Hence there is no positive integer n such that
n − 1 + n + 1 is rational.

√ √ √
4. Suppose a ̸= c, let a = c + x. Then a + b=c+x+ b=c+ d.
√ √
So x + b = d. Squaring we get, √
d − x2 − b = 2x b.
√ √
This implies that b is rational, hence d is also rational. Thus b and d are squares of rationals. Hence the
result.

5. We have
√ p
a + b( 3 p) + c( 3 p2 ) = 0 (1)

Therefore,
√ p
a( 3 p) + b( 3 p2 ) + cp = 0 (2)

Now b × (1) − c × (2) gives

√
(b2 − ac) 3 p + ab − c2 p = 0 (3)
√
3 p is irrational and therefore from (3) we get b2 − ac = 0 and ab = c2 p.
Thus,
c4 p2 = a2 b2 = a3 c.
If c ̸= 0, then we get
a3
p2 =
c3

2
 √
which is not true as 3 p is irrational. Therefore, c = 0 which in turn implies that a = b = 0.

3
2 Working with Integers
2.1 Exercise 2.1
1. The base case is trivial.
Now assume that in a knockout tournament with k, k ≤ n people, k − 1 matches are needed to declare a champion.
We now prove that if the tournament has n + 1 people, n matches are needed:
Say you are about to dispute the final. That is, 2 players, A and B are about to play the last match to determine
the winner. Then we know that A was the winner of some sub-tournament with a, a ≤ n people and B won a
sub-tournament with b, b ≤ n people. We also know that

a + b = n + 1 ⇐⇒ b = n + 1 − a.

But A was declared victorious of his sub-tournament after a − 1 games and B of his after b − 1 = n + 1 − a − 1
games. That is, it took a total of

a−1+n−a=n−1

games to determine the two contenders. We play one final match, and we get n − 1 + 1 = n matches to determine
the winner.
2. The base case is true. Let
n
X
Sn = k2
k=1

Assume that the statement is true for some n ∈ N. Then we can say that

n(n + 1)(2n + 1) (n + 1)(n + 2)(2n + 3)

Sn+1 = Sn + (n + 1)2 = =
6 6
Hence we can that the statement is true for all n ∈ N
3. The base case is true. Let
n
X
Sn = k3
k=1

Assume that the statement is true for some n ∈ N. Then we can say that

n( n + 1)2 (n + 1)2 (n + 2)2

Sn+1 = Sn + (n + 1)3 = =
4 4
Hence we can that the statement is true for all n ∈ N
4. The base case is true. Let
n
X
Sn = (2k − 1)
k=1

Assume that the statement is true for some n ∈ N. Then we can say that

Sn+1 = Sn + (2n + 1) = n2 + (2n + 1) = (n + 1)2

Hence we can that the statement is true for all n ∈ N

5. Note that when n = 10, 2n = 1024 > 1000 = n3 . Now suppose that 2n > n3 for n > 9. Then,

4
 2n+1 = 2 · 2n
> 2n3
= n3 + n3
> n3 + 9n2
= n3 + 3n2 + 6n2
> n3 + 3n2 + 54n
= n3 + 3n2 + 3n + 51n
> n3 + 3n2 + 3n + 1
= (n + 1)3 .

6. For the base case we have 2.71 + 3.51 − 5 = 24 which is divisible by 24. Now assume that for some n, 24 divides
2.7n + 3.5n − 5. Hence we can write

2.7n + 3.5n − 5 = 24λ =⇒ 2.7n = 24λ − 3.5n + 5

√ √
7. Observe that the base case is true as (3+ 5)+(3− 5) = 6 is divisible by 6. Observe we have the identity

an+1 + bn+1 = (a + b)(an + bn ) − ab(an−1 + bn−1 )

Using this we can say that

√ √ √ √ √ √
(3 + 5)n+1 + (3 − 5)n+1 = 6(3 + 5)n + (3 − 5)n ) − 4(3 + 5)n−1 + (3 − 5)n−1
√ √ n
Now observe that 2n divides (3+ √5)n +(3− √ 5) ) and 2 divides 6 and hence 2n+1 divides the first term. Similarly
we can say that 2n−1 divides (3 + 5)n + (3 − 5)n ) and 22 divides 4 and hence we have that 2n+1 dives the entire
thing. This proves the required statement.
1 n n+1
10. Let A = 2 · 34 · · · 2n−1
2n and B =
2
3 · 45 · · · 2n+1
2n
Then, we have B > A as < and AB = 1
2n+1 . From
n+1 n+2
this two relation we can conclude A < √ 1 .
2n+1

5
2.2 Exercise 2.2
1. A typical term 11 . . . 11 is of the type 4k + 3. The perfect squares are either of the form 4k or of the form
4k + 1. Hence, the number 11 . . . 111 cannot be a perfect square.
2. Note that

(ma, mb) = least positive value of max + mby = m(least positive value of ax + by) = m(a, b).

a b


3. Note that d · d, d = (a, b). Hence, we get the required result.
4. Suppose 2d − 1 = x2 , 5d − 1 = y 2 and 13d − 1 = z 2 . Since 2d − 1 = x2 we get that d must be odd. Hence y
and z must be even. Also, x2 − y 2 = (x − y)(x + y) = 8d. Hence,

x2 − y 2 (x + y)(x − y)
= = 2d.
2 2
Now either x+y x−y
2 or 2 must be even. Moreover, these two numbers differ by y, an even number. Hence their
product 2d must be divisible by 4. Hence, d must be even, a contradiction. Hence, one can find distinct a, b ∈
{2, 5, 13} such that ab − 1 is not a perfect square.
5. (7645, 2872) = 1 since 3543(2872) − 1331(7645) and (3645, 2357) = 1.
6. Note that x + 1 | 2k x − 1 if k is a positive integer. We may assume that m < n. Put x = a2m and k = 2n−m .
n−m
Hence, a2m + 1 | (a2m )2 − 1. Hence, we get that (a2m + 1, a2n + 1) =
• 1 if a is even,
• 2 if a is odd.
7. Let a, b, c be integers such that (a, b) = 1, c > 0. Let P denote the product of primes common to both a and
c and P = 1 if there are no primes common to both. Let Q denote the product of primes common to both b and
c and Q = 1 if there are no primes common to both. Let R denote the product of primes which divide c but do
not divide ab and R = 1 if there are no such primes. Note that (P, Q) = (P, R) = (Q, R) = 1. It is now easy to
see that (a + bQR, c) = 1.
8. Assume that there are finitely many primes of the type 6n − 1, say p1 = 5, p2 , . . . , pr . Let N = 6p1 p2 · · · pr − 1.
Clearly N > 2 and 2, 3 and none of the pj ’s divide N . Moreover, since N is of the type 6n − 1, N must have a
prime factor of the type 6n − 1. Hence, this prime is a new prime of the type 6n − 1, a contradiction. Hence, there
are infinitely many primes of the type 6n − 1.
9. If k, k + 1, k + 2 are consecutive integers, then at least one of them is even and one of them is divisible by 3. If
there are four consecutive integers, then there are two even integers and one of them is divisible by 4 as well.
10. Since m = n2 − n = n(n − 1) and m − 2 = (n − 2)(n + 1) and m2 − 2m = (n − 2)(n − 1)n(n + 1). Hence,
m2 − 2m is the product of 4 consecutive integers and hence 24 | m(m − 2).
11. It can be easily seen that the book has more than 1000 pages. Suppose the book has 999 + x pages. Then,
we get the equation
9 + 2(90) + 3(900) + 4x = 3189.
Hence, x = 75. Hence, the book has 1074 pages.
12. Note that

6k = (k + 1)3 + (k − 1)3 − k 3 − k 3 and 6k − 15 = (2k)3 − (2k + 1)3 + (k − 2)3 − (k + 2)3 .

13. Let p > 3 be an odd prime. Note that

p−1 (p−1)/2
X 1 X p
= .
k k(p − k)
k=1 k=1

6
Hence, the numerator is divisible by p.
14. Note that if n ≥ 4, then 3 | n(n + 2)(n + 4).
15. Since 2 is the only even prime, the product a = p1 p2 · · · pn is of the form a = 2k where k is odd. But if
a + 1 = q 2 , then q 2 and hence q is odd and so a = (q + 1)(q − 1) is divisible by 4. This is a contradiction. So a + 1
cannot be a square.
16. The units digit of the square of an integer must be one of 1, 4, 5, 6, 9. But when n > 4, the number
a = 5! + · · · + n! is divisible by 10 and so the units digit of

b = 1! + 2! + 3! + 4! + 6! + g = 33 + a = 3.

Hence, b is not a square.

17. Let a < b and a = 81m, b = 81n so that by data (m, n) = 1. Also, then the l.c.m. of a, b = 81mn = 5103 =
81 × 63, so that mn = 63 = 7 × 9. But (m, n) = 1. Hence, we may take m = 7, n = 9. Thus, a = 81m = 567,
b = 81n = 729.
18. Let a = 22 39 52 b where b is coprime to 2, 3, and 5. Then given conditions imply that
• (i) x + 1, y, z are divisible by 2 and
• (ii) x, y + 1, z are divisible by 3 and
• (iii) x, y, z + 1 are divisible by 5.
So by trial, the smallest values of x, y, z are 15, 20, 24 respectively. Also, for any integer n, b = n30 is a square,
cube, and a fifth power. Hence, we may take a = 215 320 524 n30 .
19. Let (a, b) = d so that a = dm, b = dn and (m, n) = 1. Then

a+1 b+1
+ = c, an integer.
b a
This implies that a2 + a + b2 + b = abc.
20. If n is composite then n = n1 n2 where n1 > 1 and n2 > 1. Hence, 2n1 − 1 > 1 and 2n1 − 1 | 2n − 1.
21. If n = 2α n1 , where n1 is an odd integer greater than 1, then 2α + 1 | 2n + 1, a contradiction.
22. Just take n = 2k +1 and expand it. For the second part, we can see that the right hand side leaves a remainder
9 when divided by 16 and the left hand side can leave a maximum remainder of 8 when divided by 16. Hence this
equation does not have any integer solutions.
23. The only possibilities are

12k + 1, 12k + 5, 12k + 7, 12k + 11

Since 4 divides p − 1, only 12k + 1, and 12 + 5 are possible and (p − 1)/4 = 3k or 3k + 1. If k ̸= 1 then 12k + 1 is
not possible.
Also (p + 1)/2 = 6k + 1 or 6k + 3. This shows that 6k + 5 is not possible.
So p = 12 + 1 = 13.
24. Any number is of the form 30k + c where k ∈ N and

c ∈ {0, 1, 2, . . . , 29}

Now we can easily see that as the number is prime, so c ∈ / {0, 2, 4, . . . , 28}, c ∈
/ {0, 3, 6, . . . , 27}, c ∈
/ {0, 5, 10, . . . , 25},
c∈
/ {0, 3, 6, . . . , 27}. So the only numbers that remain are the primes and 1.

7
25. Call the middle number n. Now, if n is odd, then n is co-prime with n − 1 and n + 1 since the difference
is only 1. But n is coprime with n − 2 and n + 2 as well, since all of them being odd means they don’t have a
common prime factor of 2, and since the difference between n and the other two is only 2, there cannot be any
other common prime factor.
If n is even, then n − 1 or n + 1 are both odd, and at least one of them is not divisible by 3, since with a difference
of 2, they can’t both be divisible by 3. So the one that is not divisible by 3 has no prime factors smaller or equal
to 3, and since the difference with all the others is at most 3, that one shares no common prime factors with any
of the others, and is therefore coprime with all the others. So: if n is even then either n − 1 or n + 1 (or both) are
coprime with all the others.

8
3 Congruences
3.1 Exercise 3.1
1. Note that b = 2q + 1 for some integer q. Hence, b2 = 4q(q + 1) + 1. Since 2 | q(q + 1), we get that b2 ≡ 1
(mod 8).
2. An integer b can be written as 3q or 3q ± 1. If b = 3q, then b2 ≡ 0 (mod 3). If b = 3q ± 1, then b2 ≡ 1 (mod 3).
Hence, the square of an integer is ≡ 0, 1 (mod 3).
3. If p ̸= 3, then 3 ∤ p, then p2 ≡ 1 (mod 3). Thus, 3 | p2 + 8. Hence, p = 3 is the only prime.
4. An integer b can be written as 5q or 5q ± 1 or 5q ± 2. If b = 5q, then b2 ≡ 0 (mod 5). If b = 5q ± 1, then b2 ≡ 1
(mod 5). If b = 5q ± 2, then b2 ≡ −1 (mod 5). Hence, the square of an integer is ≡ 0, 1, −1 (mod 5).
5. Put 2n + 1 = a2 and 3n + 1 = b2 . Since 2n + 1 is a perfect square, then n is divisible by 4. Hence, 3n + 1 ≡ 1
(mod 8). Hence, 8 | n. Since the square of an integer is ≡ 0, 1, −1 (mod 5). We get 2n + 1 ≡ 0, 1, −1 (mod 5)
and 3n + 1 ≡ 0, 1, −1 (mod 5). If 2n + 1 ≡ 0 (mod 5), then 3n + 1 ≡ 2 (mod 5), a contradiction. If 2n + 1 ≡ −1
(mod 5), then 3n + 1 ≡ 3 (mod 5), a contradiction. Hence, 2n + 1 ≡ 1 (mod 5) and 3n + 1 ≡ 1 (mod 5). Since
(8, 5) = 1 we get 40 | n.
6. Since (n, 6) = 1, n2 ≡ 1 (mod 8) and n2 ≡ 1 (mod 3).
√
7. We note that a2 + 2 28n2 + 1 is an even integer. Hence, 28n2 + 1 is a perfect square of an odd integer, say m.
Now 28n2 = m2 − 1 = (m − 1)(m + 1) and
  
m−1 m+1
(m − 1)(m + 1) = .
2 2

Hence,
m+1 m−1
= 7a2 and = b2 .
2 2
If m+1 2
2 = 7a and
m−1 2 2
2 = b , then b ≡ −1 (mod 7), a contradiction. Hence,
m−1
2 = 7a2 and m+1
2 = b2 . Hence,
2 2
2 + 2m = 2 + 2(2b − 1) = 4b , a perfect square.
8. Note that am − bm = (a − b)(am−1 + am−2 b + · · · + abm−2 + bm−1 ). Since a ≡ b (mod mn ), a ≡ b (mod m).
Hence, am−i bi = bm (mod m) and am−1 + am−2 b + · · · + abm−2 + bm−1 = mbm ≡ 0 (mod m). Hence, am − bm is
divisible by mn+1 .
9. Note that
p−1
ap − bp X i p−1−i
= ab ≡ pbp−1 ≡ pap−1 (mod a − b).
a−b
i=0
 p p 
−b
It follows that the gcd aa−b , a − b divides pap−1 and pbp−1 . Since a and b are coprime, the gcd is 1 or p.

9
3.2 Exercise 3.2
1. Let n = m − 1, so that m is odd. We must show that m(m − 2) divides 2(m−1)! − 1. Because ϕ(m) < m, ϕ(m)
divides (m − 1)!, so 2ϕ(m) − 1 divides 2(m−1)! − 1. Euler’s theorem implies that m divides 2ϕ(m) − 1. Therefore, m
divides 2(m−1)! − 1. Arguing similarly for m − 2, we see that m − 2 divides 2(m−1)! − 1 as well. The numbers m
and m − 2 are relatively prime, so m(m − 2) divides 2(m−1)! − 1, as desired.
2. We show that any subset of S having n elements that are pairwise coprime can be extended to a set with n + 1
elements. Indeed, if N is the product of the elements of the subset, then since the elements of S are coprime to a,
so must be N . By Euler’s theorem,

aϕ(N )+1 + aϕ(N ) − 1 ≡ a + 1 − 1 ≡ a (mod N ).

It follows that aϕ(N )+1 + aϕ(N ) − 1 is coprime to N and can be added to S. We are done.
3. Add the two equations, then add 1 to each side to obtain the Diophantine equation

(x3 + y + 1)2 + z 9 = 147157 + 157147 + 1.

The right-hand side is rather large, and it is natural to reduce modulo some number. And since the left-hand side
is a sum of a square and a ninth power, it is natural to reduce modulo 19 because 2 × 9 + 1 = 19. By Fermat’s
little theorem, a18 ≡ 1 (mod 19) whenever a is not a multiple of 19, and so the order of a square is either 1, 3, or
9, while the order of a ninth-power is either 1 or 2.
Computed by hand, the quadratic residues mod 19 are −8, −3, −2, 0, 1, 4, 5, 6, 7, 9, while the residues of ninth
powers are −1, 0, 1. Also, applying Fermat’s little theorem we see that

147157 + 157147 + 1 ≡ 1413 + 53 + 1 ≡ 14 (mod 19).

An easy verification shows that 14 cannot be obtained as a sum of a quadratic residue and a ninth-power residue.
Thus the original system has no solution in integers x, y, and z.
4. The property is true for p = 2 and p = 3, since 22 + 32 + 62 − 1 = 48. Let p be a prime greater than 3. By
Fermat’s little theorem, 2p−1 , 3p−1 , and 6p−1 are all congruent to 1 modulo p. Hence

3 · 2p−1 + 2 · 3p−1 + 6p−1 ≡ 3 + 2 + 1 = 6 (mod p).

It follows that
6 · 2p−2 + 6 · 3p−2 + 6 · 6p−2 ≡ 6 (mod p).

Dividing by 6, we find that 2p−2 + 3p−2 + 6p−2 − 1 is divisible by p, and we are done.
5. If x is a solution to the equation from the statement, then using Fermat’s little theorem, we obtain
p−1
1 ≡ xp−1 ≡ a 2 (mod p).

If m is an integer, then every odd prime factor p of m2 + 1 must be of the form 4m + 1, with m an integer. Indeed,
in this case because m2 ≡ −1 (mod p), and by what we just proved,
p−1
(−1) 2 = 1,

which means that p − 1 is divisible by 4.

Now assume that there are only finitely many primes of the form 4m + 1, m an integer, say p1 , p2 , . . . , pn . The
number (2p1 p2 · · · pn )2 + 1 has only odd prime factors, and these must be of the form 4m + 1, m an integer. Yet
these are none of p1 , p2 , . . . , pn , a contradiction. Hence the conclusion.
6. We have the factorization
16320 = 26 · 3 · 5 · 17.

10
First, note that pab − 1 = (pa )b − 1 is divisible by pa − 1. Hence p32 − 1 is divisible by p2 − 1, p4 − 1, and p16 − 1. By
Fermat’s little theorem, p2 −1 = p3−1 −1 is divisible by 3, p4 −1 = p5−1 −1 is divisible by 5, and p16 −1 = p17−1 −1
is divisible by 17. Here we used the fact that p, being prime and greater than 17, is coprime to 3, 5, and 17.
We are left to show that p32 − 1 is divisible by 26 . Of course, p is odd, say p = 2m + 1, m an integer. Then
p32 − 1 = (2m + 1)32 − 1. Expanding with Newton’s binomial formula, we get
     
32 32 31 32 2 32
(2m) + (2m) + · · · + (2m) + (2m).
1 2 1

In this sum all but the last five terms contain a power of two greater than or equal to 6. On the other hand, it is
easy to check that in
         
32 32 32 32 32
(2m)5 + (2m)4 + (2m)3 + (2m)2 + (2m),
5 4 3 2 1

the first binomial coefficient is divisible by 22 , the second by 22 , the third by 23 , the fourth by 24 , and the fifth by
25 . So this sum is divisible by 26 , and hence (2m + 1)32 − 1 = p32 − 1 is itself divisible by 26 . This completes the
solution.
8. Assume a solution (x, y) exists. If y were even, then y 3 + 7 would be congruent to 3 modulo 4. But a square
cannot be congruent to 3 modulo 4. Hence y must be odd, say y = 2k + 1. We have

x2 + 1 = y 3 + 23 = (y + 2) (y − 1)2 + 3 = (y + 2)(4k 2 + 3).

 

We deduce that x2 + 1 is divisible by a number of the form 4m + 3, namely, 4k 2 + 3. It must therefore be divisible
by a prime number of this form. But we have seen in the previous problem that this is impossible. Hence the
equation has no solutions.
9. Assume that the equation admits a solution (x, y). Let p be the smallest prime number that divides n. Because
(x + 1)n − xn is divisible by p, and x and x + 1 cannot both be divisible by p, it follows that x and x + 1 are
relatively prime to p. By Fermat’s little theorem, (x + 1)p−1 ≡ 1 ≡ xp−1 (mod p). Also, (x + 1)n ≡ xn (mod p)
by hypothesis.
Additionally, because p is the smallest prime dividing n, the numbers p − 1 and n are coprime. By the fundamental
theorem of arithmetic, there exist integers a and b such that a(p − 1) + bn = 1. It follows that

x + 1 = (x + 1)a(p−1)+bn ≡ xa(p−1)+bn ≡ x (mod p),

which is impossible. Hence the equation has no solutions.

11
4 Greatest Integer Function
1. Define f : R → N,    
1 n−1
f (x) = ⌊x⌋ + x + + ··· + x + − ⌊nx⌋ .
n n
We have       j
1 1 n−1 nk
f x+ = x+ + ··· + x + + x+ − ⌊nx + 1⌋ = f (x).
n n n n
1
Therefore, f is periodic, with period n. Also, since f (x) = 0 for x ∈ [0, n1 ), it follows that f is identically 0, and
the identity is proved.
2. Ignoring the “brackets” we have

p 2p (q − 1)p (q − 1)p
+ + ··· + = .
q q q 2

The difference between kp/q and ⌊kp/q⌋ is r/q, where r is the remainder obtained on dividing kp by q. Since p
and q are coprime, p, 2p, . . . , (q − 1)p form a complete set of residues modulo q. So for k = 1, 2, . . . , q − 1, the
numbers k/p − ⌊kp/q⌋ are a permutation of 1, 2, . . . , q − 1. Therefore,
q−1   q−1 q−1
X kp X kp X k (q − 1)p q − 1 (p − 1)(q − 1)
= − = − = ,
q q q 2 2 2
k=1 k=1 k=1

and the reciprocity law follows.

3. The function
⌊x⌋ ⌊2x⌋ ⌊3x⌋ ⌊nx⌋
f (x) = ⌊nx⌋ − − − − ··· −
1 2 3 n
satisfies f (x) = f (x + 1) for all x and f (0) = 0. Moreover, the function is constant on subintervals of [0, 1) that
do not contain numbers of the form p/q, 2 ≤ q ≤ n and 1 ≤ p ≤ q − 1. Thus it suffices to verify the inequality
for x = p/q, where p and q are coprime positive integers, 2 ≤ q ≤ n, 1 ≤ p ≤ q − 1. Subtracting the inequality
from
x 2x nx
x= + + ··· + ,
1 2 n
we obtain the equivalent inequality for the fractional part {} ({x} = x − ⌊x⌋),

{x} {2x} {3x} {nx}

{nx} ≤ + + + ··· + ,
1 2 3 n
which we prove for the particular values of x mentioned above. If rk is the remainder obtained on dividing kp by
q, then {kx} = rqk , and so the inequality can be written as

rn r1 /q r2 /q r3 /q rn /q
≤ + + + ··· + ,
q 1 2 3 n
or
r1 r2 r3 rn
rn ≤
+ + + ··· + .
1 2 3 n
Truncate the sum on the right to the (q − 1)st term. Since p and q are coprime, the numbers r1 , r2 , . . . , rq−1 are a
permutation of 1, 2, . . . , q − 1. Applying this fact and the AM–GM inequality, we obtain
 1/(q−1)
r1 r2 r3 rq−1 r1 r2 r3 rq−1
+ + + ··· + ≥ (q − 1) · · ··· = (q − 1) ≥ rn .
1 2 3 q−1 1 2 3 q−1

This proves the (weaker) inequality

r1 r2 r3 rn
+ + + ··· + ≥ rn ,
1 2 3 n

12
and consequently the inequality from the statement of the problem.
5. There are clearly more 2’s than 5’s in the prime factorization of n!, so it suffices to solve the equation
jnk j n k j n k
+ 2 + 3 + · · · = 1000.
5 5 5
On the one hand, jnk jnk jnk n n n n 1 n
+ + + ··· < + 2 + 3 + ··· = · 1 = ,
5 52 53 5 5 5 5 1− 5
4
and hence n > 4000. On the other hand, using the inequality ⌊a⌋ > a − 1, we have
n  n  n  n  n 
1000 > −1 + 2 −1 + 3 −1 + 4 −1 + 5 −1 ,
5 5 5 5 5
5
n 1 − 15
 

n 1 1 1 1
= 1+ + 2 + 3 + 4 −5= · − 5,
5 5 5 5 5 5 1 − 51
so
1005 · 4 · 3125
n< < 4022.
3124
We have narrowed down our search to {4001, 4002, . . . , 4021}. Checking each case with Polignac’s formula, we find
that the only solutions are n = 4005, 4006, 4007, 4008, and 4009.
6. Polignac’s formula implies that the exponent of the number 2 in n! is
jnk j n k j n k
+ 2 + 3 + ··· .
2 2 2
Because
n n n
+ + + ··· = n
2 22 23
and not all terms in this infinite sum are integers, it follows that n is strictly greater than the exponent of 2 in n!,
and the claim is proved.
7. Let p be a prime number. The power of p in lcm(1, 2, . . . , ni ) is equal to k if and only if
 

   
n n
< i ≤ .
pk+1 pk

Hence the power of p in the expression on the right-hand side is

X  n   n  X   X 
n n
k − = (k − (k − 1)) = .
pk pk+1 pk pk
k≥1 k≥1 k≥1

By Polignac’s formula, this is the exponent of p in n! and we are done.

13
5 Some Harder Problems
2. For any prime p, let νp (n) be the maximum power of p dividing n; i.e. pνp (n) divides n but not higher power.
Let r be the length of the non-periodic part of the infinite decimal expansion of 1/n.
Write
1
= 0.a1 a2 · · · ar b1 b2 · · · bs .
n
We show that r = max(ν2 (n), ν5 (n)).
Let a and b be the numbers a1 a2 · · · ar and b = b1 b2 · · · bs respectively. (Here a and b can be both 0.) Then
 
 
1 1  X b  1 b
= r a+ = r a+ s .
n 10 (10s )k 10 10 − 1
k≥1

Thus we get 10r (10s − 1) = n((10s − 1)a + b). It shows that r ≥ max(ν2 (n), ν5 (n)). Suppose r > max(ν2 (n), ν5 (n)).
Then 10 divides b − a. Hence the last digits of a and b are equal: ar = bs . This means
1
= 0.a1 a2 · · · ar−1 bs b1 b2 · · · bs−1 .
n
This contradicts the definition of r. Therefore r = max(ν2 (n), ν5 (n)).
4. Choose x = 24r and y = 23r . Then the left side is 212r+1 . If we take z = 2k , then we get 212r+1 = 231k . Thus
it is sufficient to prove that the equation 12r + 1 = 31k has infinitely many solutions in integers. Observe that
(12 × 18) + 1 = 31 × 7. If we choose r = 31l + 18 and k = 12l + 7, we get

12(31l + 18) + 1 = 31(12l + 7),

for all l. Choosing l ∈ N, we get infinitely many r = 31l + 18 and k = 12l + 7 such that 12r + 1 = 31k. Going
back we have infinitely many (x, y, z) of integers satisfying the given equation.
5. Let a = 3 + f , where 0 < f < 1. We are given that (3 + f )(3 − 2f ) is an integer. This implies that 2f 2 + 3f
is an integer. Since 0 < f < 1, we have 0 < 2f 2 + 3f < 5. Therefore 2f 2 + 3f can take 1, 2, 3 or 4. Equating
2f 2 + 3f to each one of them and using f > 0, we get
√ √ √
−3 + 17 1 −3 + 33 −3 + 41
f= , , , .
4 2 4 4

Therefore a takes the values:

√ √ √
−3 + 17 1 −3 + 33 −3 + 41
a=3+ , 3+ , 3+ , 3+ .
4 2 4 4

6. Suppose a = b. Then we get one equation: a2 = ac + 1. This reduces to a(a − c) = 1. Therefore a = 1, a − c = 1;

and a = −1, a − c = −1. Thus we get (a, b, c) = (1, 1, 0) and (−1, −1, 0).
If a ̸= b, subtracting the second relation from the first we get

a2 − b2 = c(b − a).

This gives a + b = −c. Substituting this in the first equation, we get

a2 = b(−a − b) + 1.

Thus a2 + b2 + ab = 1. Multiplication by 2 gives

(a + b)2 + a2 + b2 = 2.

14
Thus (a, b) = (1, −1), (−1, 1), (1, 0), (−1, 0), (0, 1), (0, −1). We get respectively c = 0, 0, −1, 1, −1, 1. Thus we get
the triples:

(a, b, c) = (1, 1, 0), (−1, −1, 0), (1, −1, 0), (−1, 1, 0), (1, 0, −1), (−1, 0, 1), (0, 1, −1), (0, −1, 1).

7. Let p and q be two consecutive primes, p < q. If we take any n such that p ≤ n < q, we see that P (n) = p and
N (n) = q. Hence the term pq1
occurs in the sum q − p times. The contribution from such terms is q−p 1 1
pq = p − q .
Since n + 1 is prime, we obtain
1 1 1 1
+ + + ··· +
P (2)N (2) P (3)N (3) P (4)N (4) P (n)N (n)
     
1 1 1 1 1 1 1 1 n−1
= − + − + ··· + − = − = .
2 3 3 5 p n+1 2 n+1 2n + 2

Here p is used for the prime preceding n + 1.

8. Rewriting the given equation, we have

4m3 + m2 + 12m + 3 = 3pn .

The left-hand side equals (4m + 1)(m2 + 3).

Suppose that (4m + 1, m2 + 3) = 1. Then (4m + 1, m2 + 3) = (3pn , 1), (3, pn ), (pn , 3) or (1, 3pn ), a contradiction
since 4m + 1, m2 + 3 ≥ 4. Therefore (4m + 1, m2 + 3) > 1.
Since 4m + 1 is odd, we have (4m + 1, m2 + 3) = (4m + 1, 49) = 7 or 49. This proves that p = 7, and 4m + 1 = 3 · 7k
or 7k for some natural number k. If (4m + 1, 49) = 7, then we have k = 1 and 4m + 1 = 21, which does not lead
to a solution. Therefore (4m + 1, m2 + 3) = 49.
If 73 divides 4m + 1, then it does not divide m2 + 3, so we get m2 + 3 ≤ 3 · 72 < 4m + 1. This implies (m − 2)2 < 2,
so m ≤ 3, which does not lead to a solution. Therefore we have 4m + 1 = 49, which implies m = 12 and n = 4.
Thus (m, n, p) = (12, 4, 7) is the only solution.
9. Since p4 − p1 = 8, and no prime is even, we observe that {p1 , p2 , p3 , p4 } is a subset of {p1 , p1 + 2, p1 + 4, p1 +
6, p1 + 8}. Moreover, p1 is larger than 3. If p1 ≡ 1 (mod 3), then p1 + 2 and p1 + 8 are divisible by 3. Hence we
do not get 4 primes in the set {p1 , p1 + 2, p1 + 4, p1 + 6, p1 + 8}. Thus p1 ≡ 2 (mod 3) and p1 + 4 is not a prime.
We get p2 = p1 + 2, p3 = p1 + 6, and p4 = p1 + 8.
Consider the remainders of p1 , p1 + 2, p1 + 6, p1 + 8 when divided by 5. If p1 ≡ 2 (mod 5), then p1 + 8 is divisible
by 5 and hence is not a prime. If p1 ≡ 3 (mod 5), then p1 + 2 is divisible by 5. If p1 ≡ 4 (mod 5), then p1 + 6 is
divisible by 5. Hence the only possibility is p1 ≡ 1 (mod 5).
Thus we see that p1 ≡ 1 (mod 2), p1 ≡ 2 (mod 3), and p1 ≡ 1 (mod 5). We conclude that p1 ≡ 11 (mod 30).
Similarly, q1 ≡ 11 (mod 30). It follows that 30 divides p1 − q1 .
10. Since xyz ̸= 0, we can divide the second relation by the first. Observe that

x4 + x2 y 2 + y 4 = (x2 + xy + y 2 )(x2 − xy + y 2 ),

holds for any x, y. Thus we get

(x2 − xy + y 2 )(y 2 − yz + z 2 )(z 2 − zx + x2 ) = x2 y 2 z 2 .

However, for any real numbers x, y, we have

x2 − xy + y 2 ≥ |xy|.

15
Since x2 y 2 z 2 = |xy||yz||zx|, we get

|xy||yz||zx| = (x2 − xy + y 2 )(y 2 − yz + z 2 )(z 2 − zx + x2 ) ≥ |xy||yz||zx|.

This is possible only if

x2 − xy + y 2 = |xy|, y 2 − yz + z 2 = |yz|, z 2 − zx + x2 = |zx|,

hold simultaneously. However, |xy| = ±xy. If x2 − xy + y 2 = −xy, then x2 + y 2 = 0, giving x = y = 0. Since

we are looking for nonzero x, y, z, we conclude that x2 − xy + y 2 = xy, which is the same as x = y. Using the
other two relations, we also get y = z and z = x. The first equation now gives 27x6 = x3 . This gives x3 = 1/27
(since x ̸= 0), or x = 1/3. We thus have x = y = z = 1/3. These also satisfy the second relation, as may be
verified.
11. If n > 1 is such that

2p = (n − 1)2 + n2 + (n + 1)2 + (n + 2)2 = 4n2 + 4n + 6

then
p = 2n(n + 1) + 3 > 3.

Observe that if n ≡ 0 (mod 3) or n ≡ 2 (mod 3) then p ≡ 0 (mod 3) and hence can’t be a prime. Therefore n ≡ 1
(mod 3). Write n = 3k + 1 for some positive integer k. Observe that

p − 7 = 2(n2 + n − 2) = 2(n − 1)(n + 2) = 18k(k + 1) ≡ 0 (mod 36).

12. Since x5 is rational, we see that (20x)5 and (x/19)5 are rational numbers. But
5
194
   
5 19 19 4 3 2 2 2 3 1
(20x) − = 20x − (20x) + (20 · 19)x + 20 · 19 + (20 · 19 ) 2 + 4 .
x x x x

Consider
194
 
4 3 2 2 2 3 1
T = (20x) + (20 · 19)x + 20 · 19 + (20 · 19 ) 2 + 4
x x
194 192
   
= (20x)4 + 4 + 20 · 19 (20x)2 + 2 + (202 · 192 ).
x x

Using 20x + (19)/x is rational, we get

2
192

2 19
(20x) + 2 = 20x + − 2 · 20 · 19
x x
is rational. This leads to 2
194 192

4 2
(20x) + 4 = (20x) + 2 − 2 · 202 · 192
x x
is also rational. Thus T is a rational number and T ̸= 0. We conclude that 20x − (19/x) is a rational number.
This combined with the given condition that 20x + (19/x) is rational shows 2 · 20 · x is rational. Therefore x is
rational.
√
13. Let ⌊ 2n⌋ = k. We observe that x − 1 < ⌊x⌋ ≤ x. Hence
√ √
2n < 1 + k ≤ 1 + 2n.

Divisibility gives (1 + k)d = 2n for some positive integer d. Therefore we obtain

√ 2n √
2n < ≤ 1 + 2n.
d

16
 √
The first inequality gives d < 2n < 1 + k. But then
√
2n ( 2n)2 k2 1
d= = ≥ = (k − 1) + > k − 1.
1+k 1+k 1+k k+1
We thus obtain k − 1 < d < k + 1. Since d is an integer, it follows that d = k. This implies that n = k(k + 1)/2.
Thus n is a triangular number. It is easy to check that every triangular number is a solution.
14. Observe that lcm(3, 4, 5) = 60. We look for a solution in the form a3 = b4 = c5 = l60k . We choose l, k such
that the condition is satisfied. The given equation gives

3l60k = d7 .

This suggests choosing l = 3 so that 360k+1 = d7 . Now we take care of k by choosing k such that 7 divides 60k + 1.
For example, we can take k = 5 so that 301 = 7 × 43. Thus we get

3300 + 3300 + 3300 = 3301 .

Choose a = 3100 , b = 375 , c = 360 , and d = 343 . Then we get

a3 + b4 + c5 = d7 .

This gives one solution. This suggests choosing a = 3100 · m140 , b = 375 · m105 , c = 360 · m84 , and d = 343 · m60 .
We see that

a3 + b4 + c5 = 3300 · m420 + 3300 · m420 + 3300 · m420 = m420 3301 = (343 · m60 )7 = d7 .

We can give different values for m and get infinitely many solutions of the equation.
15. We use divisibility argument by 7. Observe that the remainders of seven consecutive cubes modulo 7 are
0, 1, 1, 6, 1, 6, 6 in some (cyclic) order. Hence the sum of seven consecutive cubes is 0 modulo 7. On the other
hand, the remainders of two consecutive fourth powers modulo 7 is one of the sets {0, 1}, {1, 2}, {2, 4}, {4, 4}.
Hence the sum of two fourth powers is never divisible by 7. It follows that the given equation has no solution in
integers.
16. Firstly we show that if m > 1 is good, then so is 2m. This is true since some proper divisors of m, including
1 (and hence not including m itself), sum to m; if we consider all these together with m, they will all be factors
of 2m which sum to 2m.
This means that it suffices to prove the claim for odd numbers. The claim holds for 1 since good numbers exist
(such as 6, which is 1 + 2 + 3, for example).
If a > 1 is odd and n = 2k a for some k, then

a + 2a + 4a + · · · + 2k−1 a = (2k − 1)a = n − a

which is close to n. This value of n will be good if we can find some other factors of n, including 1, which sum to
a.
To do this, we write a as a sum of powers of 2, including 1, by writing a in binary, and then choose k to be large
enough for all those powers of 2 to be factors of n. (We may take k to be ⌈log2 (a)⌉, the smallest integer greater
than or equal to the base-2 logarithm of a.) None of these powers of 2 are multiples of a so there is no risk that
we are using the same factor of n twice.
19. Since q(p − 1) is a perfect square and q is prime, we should have p − 1 = qb2 for some positive integer b. Let
a2 = p − q. Therefore, q = p − a2 , and substituting that into the p − 1 = qb2 and solving for p gives

a2 b2 − 1 (ab − 1)(ab + 1)
p= 2
= .
b −1 b2 − 1

17
Notice that we also have
a2 b2 − 1 2 a2 − 1
p= = a +
b2 − 1 b2 − 1
and so b2 − 1 | a2 − 1. We run through the cases:
• a = 1: Then p − q = 1 so (p, q) = (3, 2), which works.
• a = b: This means p = a2 + 1, so q = 1, a contradiction.
• a > b: This means that b2 − 1 < ab − 1. Since b2 − 1 can be split up into two factors F1 , F2 such that
F1 | ab − 1 and F2 | ab + 1, we get
ab − 1 ab + 1
p= ·
F1 F2
and each factor is greater than 1, contradicting the primality of p.
Thus, the only solution is (p, q) = (3, 2).

18
6 Polynomials
1. We know from Vieta that a + b = −a and ab = b. Now from the second equation if b = 0, we get a = 0. If not,
we can say that a = 1 and b = −2. Thus the solutions are

(a, b) = (0, 0) or (1, −2)

2. If u = α and v are the roots

x2 + ax + b + 1 = (x − u)(x − v)
= x2 − (u + v)x + uv
Then a = −(u + v) and b = uv − 1. Since u = α ∈ Z, and a ∈ Z, we know that v ∈ Z. Therefore,

a2 + b2 = (u + v)2 + (uv − 1)2

= u2 + v 2 + u2 v 2 + 1
= (u2 + 1)(v 2 + 1)
Which is a composite number.
3. Since β and γ are the roots of the first equation, we have β + γ = −α and βγ = −a. Similarly, from the second
equation we have γ + α = −β and γα = −b. Thus, we have α + β + γ = 0 and βγ = −a and γα = −b.

(α + β + γ)2 = 0 =⇒ (α2 + β 2 + γ 2 ) + 2(αβ + βγ + γα) = 0.

From this we obtain
2(a + b + c) + 2(αβ − a − b) = 0 =⇒ αβ = −c.

Since we have αβ = −c and α+β = −γ, we can say that α and β are the roots of the equation x2 +γx−c = 0.
4. The cubic vanishes at x = 1 and x = −2, thus we can write it as

p(x) = k(x − 1)(x + 2)(x − r)

where r is the third root. Now, we know that

p(−1) = k(−2)(1)(−1 − r) = 4

p(2) = k(1)(4)(2 − r) = 8
On solving these equations, we get that the polynomial should be
 1
p(x) = 3(x − 1)(x + 2) x −
3

5. The rational root theorem.

6. Of the two integers −a and 1 − a, exactly one is even. If f (a) = 0, then

f (x) = (x − a)g(x).

But f (0) = −ag(a), and f (1) = (1 − a)g(a). Both f (0) and f (1) cannot be odd.
7. We are given that x = 2 is a root of the cubic equation:

84x3 − 157x2 − kx + 78 = 0.

To find k, substitute x = 2 into the equation:

84(23 ) − 157(22 ) − k(2) + 78 = 0.

19
Calculating each term:
23 = 8, 84 · 8 = 672, 22 = 4, −157 · 4 = −628.
Thus:
672 − 628 − 2k + 78 = 0.
Combine like terms:
672 − 628 = 44, 44 + 78 = 122.
We get:
122
122 − 2k = 0 =⇒ k = = 61.
2
With k = 61, the equation becomes:
84x3 − 157x2 − 61x + 78 = 0.
Since x = 2 is a root, we divide the polynomial by (x − 2) using synthetic division:

84 −157 −61 78
2 168 22 −78
84 11 −39 0
The quotient is:
84x2 + 11x − 39.

To find the remaining roots, solve the quadratic equation:

84x2 + 11x − 39 = 0.

Using the quadratic formula:

√
−b ± b2 − 4ac
x= , where a = 84, b = 11, c = −39.
2a
First, compute the discriminant:

b2 − 4ac = 112 − 4(84)(−39) = 121 + 13104 = 13225.

√
Since 13225 = 115, we substitute into the formula:
−11 ± 115 −11 ± 115
x= = .
2 · 84 168
Consider the two cases:
−11 + 115 104 13 −11 − 115 −126 3
x= = = , x= = =− .
168 168 21 168 168 4

The roots of the equation are therefore:

13 3
x = 2, x= , x=− .
21 4

8. Observe that the roots of x2 + x + 1 are of ω, ω 2 . Plugging these into the given polynomials and equating it
with 0, we get the values of a = 4, b = 1.
9. 8x2 − 10x − 61
10. We are given a cubic equation:
x3 + px − q = 0
with roots α, β, γ.
From this, we know the standard relationships between the roots and the coefficients using Vieta’s theorem:

20
 1. α + β + γ = 0 (since the x2 term is missing),
2. αβ + βγ + γα = p,
3. αβγ = q.
We need to find the cubic equation whose roots are:

α + β, β + γ, γ + α.

Given α + β + γ = 0, we can rewrite the new roots in terms of the original ones:

α + β = −γ, β + γ = −α, γ + α = −β.

Thus, the new roots α + β, β + γ, and γ + α are simply −γ, −α, and −β, respectively. In other words:

{α + β, β + γ, γ + α} = {−α, −β, −γ}.

Since α, β, γ are roots of x3 + px − q = 0, the polynomial can be written as:

(x − α)(x − β)(x − γ) = x3 + px − q.

The polynomial whose roots are −α, −β, and −γ is:

(x − (−α))(x − (−β))(x − (−γ)) = (x + α)(x + β)(x + γ).

Now, consider the transformation x 7→ −x. If we start from f (x) = x3 + px − q, then:

f (−x) = (−x)3 + p(−x) − q = −x3 − px − q.

Multiplying by −1 gives:
−f (−x) = x3 + px + q.

Thus, the polynomial with roots −α, −β, and −γ is:

x3 + px + q.

11. x3 − 2x2 + 5x − 11 = 0.
12.Just use Vieta’s theorem. If originally, the roots were a,b,c, then the new equation we are seeking will have
roots a3 , b3 , c3 .
We can use simple algebraic manipulations to find that sum of roots is 0, sum of roots taken two at a time is 2,
and the product of roots is 1. Thus, the equation is x3 + 2x − 1 = 0
13.The polynomial f (x) + 1 has x − a, x − b, x − c as factors, and so:

f (x) = (x − a)(x − b)(x − c)g(x) − 1

for some polynomial g(x) with integer coefficients.

But then, if f (d) = 0 for some integer d, it follows that:

(d − a)(d − b)(d − c)g(d) = 1.

Thus, d − a, d − b, d − c ∈ {+1, −1}, and so two of them must be equal. This contradicts the assumption that
a, b, c are distinct.

21
Hence, g(d) ̸= 0 for d ∈ Z.
14. If a is negative or zero, then the quadratic has two real roots.
But we can easily check that the other polynomial has derivative everywhere positive and hence only one real
root.
So a must be positive. If
x2 − x + a
divides x13 + x + 90, then x13 + x + 90 = f (x)(x2 − x + a),
where f (x) is a polynomial with integer coefficients.
Let x = 0 we see that a must divide 90. Let x = 1 we see that it must divide 92.
Hence it must divide 92 - 90 = 2. So the only possibilities are 1 and 2.
Suppose a = 1 then putting x = 2 we have that 3 divides 213 + 92 but 2odd 2 mod 3, so 213 + 92 is congruent to
1 mod 3. is congruent to
So a cannot be 1.
To see that a = 2 is possible, we write

(x2 − x + 2)(x11 + x10 − x9 − 3x8 − x7 + 5x6 + 7x5 − 3x4 − 17x3 − 11x2 + 23x + 45)
= x13 + x + 90

15 We can solve this using the idea if extremal principle. If f (x) had odd degree, then it would have real roots.
Let x0 be the greatest real root. Then

0 = f (x0 )g(x0 ) = f (x0 2 + x0 + 1).

So, x0 2 + x0 + 1 is another real root of f (x), which is impossible, since x0 2 + x0 + 1 > x0 and x0 is the greatest
real root.
16. X
f (x, y) = cij xi y j .
i,j

Since f is antisymmetric, we find cij = −cji , so we can write

X
f (x, y) = cij (xi y j − y i xj ).
i>j

Now we’ve found out that f is some linear combination of terms xi y j − y i xj . If we show that each of these terms
is divisible by x − y, then so is f , and we’re done.
So suppose i > j, and let k = i − j. Since

xk − y k = (x − y)(xk−1 + xk−2 y + xk−3 y 2 + · · · + y k−1 )

is divisible by x − y, so is xi y j − y i xj = xj y j (xk − y k ).
17.Use similar idea as the previous problem.
18.The claim is that if a symmetric polynomial f (x, y) is divisible by (x − y), then it is actually divisible by
(x − y)2 .
We use the fundamental fact that every symmetric polynomial in x and y can be expressed as a polynomial in the
elementary symmetric polynomials:
e1 = x + y, e2 = xy.

22
Thus, there exists a polynomial F (e1 , e2 ) such that:

f (x, y) = F (x + y, xy).

Since f (x, y) is divisible by (x − y), it must vanish whenever x = y. Substituting x = y into f , we find:

f (x, x) = F (2x, x2 ) = 0 for all x.

This condition tells us that F (e1 , e2 ) vanishes on the infinite set of points (e1 , e2 ) = (2x, x2 ) for all x.
Consider the curve defined by e1 = 2x and e2 = x2 . Eliminating x gives the relation:
 e 2
1
e2 = .
2

e21
The polynomial F (e1 , e2 ) vanishes for every point on the parabola e2 = 4. This implies that F (e1 , e2 ) must be
divisible by the polynomial defining this parabola:

e21
e2 − = 0.
4

Thus, there exists a polynomial G(e1 , e2 ) such that:

e2
 
F (e1 , e2 ) = e2 − 1 G(e1 , e2 ).
4

Clearing the fraction by multiplying both sides by 4 gives:

4F (e1 , e2 ) = (4e2 − e21 )G(e1 , e2 ).

Hence:
F (e1 , e2 ) is divisible by (e21 − 4e2 ).

Now relate this back to x and y. Recall the well-known identity:

(x − y)2 = x2 − 2xy + y 2 = (x + y)2 − 4xy = e21 − 4e2 .

We have shown that e21 − 4e2 divides F (e1 , e2 ). Since:

f (x, y) = F (x + y, xy),

it follows that:
(x − y)2 = e21 − 4e2 | f (x, y).

Therefore, we conclude that if f (x, y) is symmetric and (x − y) divides f (x, y), then (x − y)2 also divides
f (x, y).
19. Let p(x) = nxn+1 − (n + 1)xn + 1.
Then (x − 1)2 is a factor of p(x) iff 1 is a double root.
Clearly, p(1) = 0.
p′ (x) = n(n + 1)xn − (n + 1)nxn−1 .
Clearly, p′ (1) = 0.
So, 1 indeed is a double root.
20.
Using the quadratic formula, the roots of the polynomial x2 − 6x + 1 are:

23
 √
x1 = 3 + 2 2
√
x2 = 3 − 2 2

Let Sn = xn1 + xn2

• S0 = x01 + x02 = 1 + 1 = 2
√ √
• S1 = x11 + x12 = 3 + 2 2 + 3 − 2 2 = 6
Now, multiply the quadratic equation (x2 − 6x + 1 = 0) by xn−2 :
xn − 6xn−1 + xn−2 = 0
Substitute x1 and x2 into this equation and add the two resulting equations:

xn1 − 6x1n−1 + x1n−2 = 0

xn2 − 6x2n−1 + x2n−2 = 0

(xn1 + xn2 ) − 6(xn−1

1 + xn−1
2 ) + (xn−2
1 + xn−2
2 )=0
This gives us the recursive relation:
Sn − 6Sn−1 + Sn−2 = 0 or Sn = 6Sn−1 − Sn−2
we prove Sn is an integer by induction
• Base Case: We’ve shown S0 = 2 and S1 = 6 are integers.
• Inductive Hypothesis: Assume Sk and Sk−1 are integers for some arbitrary integer k ≥ 1.
• Inductive Step: We need to show that Sk+1 is an integer. Using the recursive relation: Sk+1 = 6Sk − Sk−1 .
Since Sk and Sk−1 are integers (by the inductive hypothesis), and 6 is an integer, Sk+1 is also an integer.
Therefore, Sn is an integer for all nonnegative integers n.
We now prove Sn is not divisible by 5 by induction
• Base Case: S0 = 2 and S1 = 6 are not divisible by 5.
• Inductive Hypothesis: Assume Sk and Sk−1 are not divisible by 5 for some arbitrary integer k ≥ 1.
• Inductive Step: We need to show that Sk+1 is not divisible by 5. Sk+1 = 6Sk − Sk−1 .
Consider the possible remainders when an integer is divided by 5 (0, 1, 2, 3, 4).
– If Sk leaves a remainder of 1 when divided by 5, and Sk−1 leaves a remainder of 2, then 6Sk leaves a
remainder of 1 and 6Sk − Sk−1 leaves a remainder of 4.
– Continue checking all possible combinations of remainders for Sk and Sk−1 . You’ll find that Sk+1 never
leaves a remainder of 0, meaning it’s never divisible by 5.
Therefore, Sn is not divisible by 5 for all nonnegative integers n.
We have proven that for every nonnegative integer n, xn1 + xn2 is an integer and not divisible by 5.
21. If f (x) = 0 has a rational root, then this root is an integer. Suppose that f (x) has the integral root x0 = m,
that is f (m) = 0. Then f (x) = (x − m)g(x), where g(x) has integral coefficients. By setting x = k, k + 1, ..., k + p
in the last equation, we get f (k) = (k − m)g(k), f (k + 1) = (k + 1 − m)g(k + 1), ..., f (k + p) = (k − p − m)g(k + p).
One of the p + 1 successive integers k − m, ..., k + p − m is divisible by p + 1. This proves the contrapositive
statement which is equivalent to the original statement.

24
22. For x ≤ 0 we have obviously p(x) > 0. Let x > 0. We transform the polynomial in the same way as a
geometric series:

p(x) = x2n − 2x2n−1 + 3x2n−2 − · · · − 2nx + 2n + 1

xp(x) = x2n+1 − 2x2n + 3x2n−1 − 4x2n−2 + · · · + (2n + 1)x.

Adding, we get

xp(x) + p(x) = x2n+1 − x2n + x2n−1 − x2n−2 + · · · + x + 2n + 1

1 + x2n+1
(1 + x)p(x) = x · + 2n + 1.
1+x
From here we see that p(x) > 0 for x > 0.
1. For i ≥ 0, let fi (x) denote the polynomial on the blackboard after Hobbes’ i-th turn. We let Calvin decrease
the coefficient of x by 1. Therefore fi+1 (2) = fi (2) − 1 or fi+1 (2) = fi (2) − 3 (depending on whether Hobbes
increases or decreases the constant term). So for some i, we have 0 ≤ fi (2) ≤ 2. If fi (2) = 0 then Calvin has
won the game. If fi (2) = 2 then Calvin wins the game by reducing the coefficient of x by 1. If fi (2) = 1 then
fi+1 (2) = 0 or fi+1 (2) = −2. In the former case, Calvin has won the game and in the latter case Calvin wins the
game by increasing the coefficient of x by 1.
2. Suppose that m is an integer root of x4 − ax3 − bx2 − cx − d = 0. As d ̸= 0, we have m ̸= 0. Suppose now that
m > 0. Then m4 −am3 = bm2 +cm+d > 0 and hence m > a ≥ d. On the other hand d = m(m3 −am2 −bm−c) and
hence m divides d, so m ≤ d, a contradiction. If m < 0, then writing n = −m > 0 we have n4 +an3 −bn2 +cn−d =
n4 + n2 (an − b) + (cn − d) > 0, a contradiction. This proves that the given polynomial has no integer roots.
3. Let r = u/v where gcd(u, v) = 1. Then we get

an un + an−1 un−1 v + · · · + a1 uv n−1 + a0 v n = 0,

bn un + bn−1 un−1 v + · · · + b1 uv n−1 + b0 v n = 0.

Subtraction gives

(an − bn )un + (an−2 − bn−2 )un−2 v 2 + · · · + (a1 − b1 )uv n−1 + (a0 − b0 )v n = 0,

since an−1 = bn−1 . This shows that v divides (an − bn )un and hence it divides an − bn . Since an − bn is a prime,
either v = 1 or v = an − bn . Suppose the latter holds. The relation takes the form

un + (an−2 − bn−2 )un−2 v + · · · + (a1 − b1 )uv n−2 + (a0 − b0 )v n−1 = 0.

(Here we have divided through-out by v.) If n > 1, this forces v | u, which is impossible since gcd(v, u) = 1 (v > 1
since it is equal to the prime an − bn ). If n = 1, then we get two equations:

a1 u + a0 v = 0,

b1 u + b0 v = 0.

This forces a1 b0 − a0 b1 = 0 contradicting an b0 − a0 bn ̸= 0. (Note: The condition an b0 − a0 bn ̸= 0 is extraneous.

The condition an−1 = bn−1 forces that for n = 1, we have a0 = b0 . Thus we obtain, after subtraction

(a1 − b1 )u = 0.

This implies that u = 0 and hence r = 0 is an integer.)

4. Consider the discriminants of the three equations

px2 + qr + r = 0 (1)

25
 qx2 + rx + p = 0 (2)
rx2 + px + q = 0. (3)

Let us denote them by D1 , D2 , D3 respectively. Then we have

D1 = 4(q 2 − rp), D2 = 4(r2 − pq), D3 = 4(p2 − qr).

We observe that

D1 + D2 + D3 = 4(p2 + q 2 + r2 − pq − qr − rp) = 2 (p − q)2 + (q − r)2 + (r − p)2 > 0.



since p, q, r are not all equal. Hence at least one of D1 , D2 , D3 must be positive. We may assume D1 > 0.
Suppose D2 < 0 and D3 < 0. In this case both the equations (2) and (3) have only non-real roots and equation
(1) has only real roots. Hence the common root α must be between (2) and (3). But then ᾱ is the other root of
both (2) and (3). Hence it follows that (2) and (3) have the same set of roots. This implies that
q r p
= = .
r p q

Thus p = q = r contradicting the given condition. Hence both D2 and D3 cannot be negative. We may assume
D2 ≥ 0. Thus we have
q 2 − rp > 0, r2 − pq ≥ 0.

These two give

q 2 r2 > p2 qr
since p, q, r are all positive. Hence we obtain qr > p2 or D3 < 0. We conclude that the common root must be
between equations (1) and (2). Thus
pα2 + qα + r = 0,
qα2 + rα + p = 0.

Eliminating α2 , we obtain
2(q 2 − pr)α = p2 − qr.

Since q 2 − pr > 0 and p2 − qr < 0, we conclude that α < 0. The condition p2 − qr < 0 implies that the equation
(3) has only non-real roots.
Alternatively one can argue as follows. Suppose α is a common root of two equations, say, (1) and (2). If α is
non-real, then ᾱ is also a root of both (1) and (2). Hence the coefficients of (1) and (2) are proportional. This
forces p = q = r, a contradiction. Hence the common root between any two equations cannot be non-real. Looking
at the coefficients, we conclude that the common root α must be negative. If (1) and (2) have common root α,
then q 2 ≥ rp and r2 ≥ pq. Here at least one inequality is strict for q 2 = pr and r2 = pq forces p = q = r. Hence
q 2 r2 > p2 qr. This gives p2 < qr and hence (3) has nonreal roots.
5. Suppose α is a real root of the given equation. Then

α5 − α3 + α − 2 = 0. (1)

This gives α5 − α3 + α − 1 = 1 and hence (α − 1)(α4 + α3 + 1) = 1. Observe that α4 + α3 + 1 ≥ 2α2 + α3 = α2 (α + 2).

If −1 ≤ α < 0, then α + 2 > 0, giving α2 (α + 2) > 0 and hence (α − 1)(α4 + α3 + 1) < 0. If α < −1, then
α4 + α3 = α3 (α + 1) > 0 and hence α4 + α3 + 1 > 0. This again gives (α − 1)(α4 + α3 + 1) < 0.
The above reasoning shows that for α < 0, we have α5 − α3 + α − 1 < 0 and hence cannot be equal to 1. We
conclude that a real root α of x5 − x3 + x − 2 = 0 is positive (obviously α ̸= 0).

26
Now using α5 − α3 + α − 2 = 0, we get
α6 = α4 − α2 + 2α.

The statement ⌊α6 ⌋ = 3 is equivalent to 3 ≤ α6 < 4.

Consider α4 − α2 + 2α < 4. Since α > 0, this is equivalent to α5 − α3 + 2α2 < 4α. Using the relation (1),
we can write 2α2 − α + 2 < 4α or 2α2 − 5α + 2 < 0. Treating this as a quadratic, we get this is equivalent to
1 4 3 1
2 < α < 2. Now observe that if α ≥ 2 then 1 = (α − 1)(α + α + 1) ≥ 25, which is impossible. If 0 < α ≤ 2 , then
1 = (α − 1)(α4 + α3 + 1) < 0 which again is impossible.
We conclude that 12 < α < 2. Similarly α4 − α2 + 2α ≥ 3 is equivalent to α5 − α3 + 2α2 − 3α ≥ 0 which is equivalent
to 2α2 − 4α + 2 ≥ 0. But this is 2(α − 1)2 ≥ 0 which is valid. Hence 3 ≤ α6 < 4 and we get ⌊α6 ⌋ = 3.
6. Since λ is a root of the equation x3 + ax2 + bx + c = 0, we have

λ3 = −aλ2 − bλ − c.

This implies that

λ4 = −aλ3 − bλ2 − cλ = (1 − a)λ3 + (a − b)λ2 + (b − c)λ + c
where we have used again
−λ3 − aλ2 − bλ − c = 0.

Suppose |λ| ≥ 1. Then we obtain

|λ|4 ≤ (1 − a)|λ|3 + (a − b)|λ|2 + (b − c)|λ| + c

≤ (1 − a)|λ|3 + (a − b)|λ|3 + (b − c)|λ|3 + c|λ|3

≤ |λ|3 .

This shows that |λ| ≤ 1. Hence the only possibility in this case is |λ| = 1. We conclude that |λ| ≤ 1 is always
true.
7. Let ω = e2πi/5 , so that ω 5 = 1. We set for x in (*), ω, ω 2 , ω 3 , ω 4 successively, and get the following equations 1
to 4. If we multiply 1 to 4 by −ω, −ω 2 , −ω 3 , −ω 4 , then we get the last 4 equations.

P (1) + ωQ(1) + ω 2 R(1) = 0,

P (1) + ω 2 Q(1) + ω 4 R(1) = 0,
P (1) + ω 3 Q(1) + ωR(1) = 1,
P (1) + ω 4 Q(1) + ω 3 R(1) = 0,
−ωP (1) − ω 2 Q(1) − ω 3 R(1) = 0,
−ω 2 P (1) − ω 4 Q(1) − R(1) = 0,
−ω 3 P (1) − ωQ(1) − ω 4 R(1) = 0,
−ω 4 P (1) − ω 3 Q(1) − ω 2 R(1) = 0.

Using 1 + ω + ω 2 + ω 3 + ω 4 = 0, we get the sum 5P (1) = 0, that is, x − 1 | P (x).

8. Let Q(x) = (x + 1)P (x) − x. Then the polynomial Q(x) vanishes for k = 0, . . . , n, that is,

(x + 1)P (x) − x = a · x · (x − 1)(x − 2) · · · (x − n).

27
To find a we set x = −1 and get 1 = a(−1)n+1 (n + 1)!. Thus,

(−1)n+1 x(x − 1) · · · (x − n)
P (x) = + x,
(n + 1)!

and (
1 for odd n,
P (n + 1) = n
n+2 for even n.

28
7 Inequalities
7.1 10.1
1. By AM-HM inequality,
 
1 1 1
((b + c) + (c + a) + (a + b)) + + ≥ 32
b+c c+a a+b
a+b+c a+b+c a+b+c 9
⇒ + + ≥
b+c c+a a+b 2
a b c 9 3
⇒ + + ≥ −3= .
b+c c+a a+b 2 2

2. Solution: By AM ≥ HM,
1 1 4 2
+ ≥ =
b+c−a c+a−b b+c−a+c+a−b c
Similarly we can obtain
1 1 2 1 1 2
+ ≥ , + ≥
c+a−b a+b−c a a+b−c b+c−a b

Adding these together will give us the required inequality.

3. If we add 1 to each term on the LHS, we get
s s s
LHS + n = + + ··· +
s − a1 s − a2 s − an
Now, we can use AM ≥ HM to get
1 1 1 n2 n2
+ + ··· + ≥ =
s − a1 s − a2 s − an s − a1 + s − a2 + · · · + s − an ns − s
Therefore,
s s s n2 s n
LHS = + + ··· + −n≥ −n= .
s − a1 s − a2 s − an (n − 1)s n−1
 
1 1 1 1 1
4. Solution: = ≤ + , last inequality due to HM ≤ AM. Hence, we
a + b + 2c (a + c) + (b + c) 4 a+c b+c
obtain
X ab X ab  1 1
 
1 ab + cb ba + ca bc + ac

b+a+c
≤ + = + + = .
cyc
a + b + 2c cyc
4 a+c b+c 4 a+c b+c a+b 4

ab2 1
5. Solution: Using HM ≤ AM, a+b = 1/b2 +1/ab
≤ 41 (b2 + ab). Hence,

X ab2 X 1 1X 2 1
= 2
≤ (b + ab) ≤ (a2 + b2 + c2 ),
cyc
a+b cyc
1/b + 1/ab 4 cyc 2

last inequality due to ab + bc + ca ≤ a2 + b2 + c2 . Next,

ab2 bc2 ca2 AM-GM 3abc
+ + ≥ p
a+b b+c c+a 3
(a + b)(b + c)(c + a)
p
3 (a+b)+(b+c)+(c+a) 2(a+b+c)
Again, by AM-GM, (a + b)(b + c)(c + a) ≤ 3 = 3 . Hence we get

ab2 bc2 ca2 3abc

+ + ≥ .
a+b b+c c+a 2(a + b + c)/3

29
7.2 10.2
1. Using C-S inequality,
 √  √ 
x2 + ( 5y)2 + z 2 22 + ( 5)2 + 42 ≥ (2x + 5y + 4z)2

This gives 2x + 5y + 4z ≤ 5.
2. Using C-S inequality, we get
 
2 2 2 1 1 1
≥ (x + y + z)2


2x + 3y + 6z + +
2 3 6

which gives 2x2 + 3y 2 + 6z 2 ≥ 7.

3.. By Cauchy-Schwarz inequality,

x2 y 2 z 2
x   
y z 2 1 1 1
+ + ≤ + + + +
2 3 6 2 3 6 2 3 6

1
4. Let P(x) = an xn +an−1 xn−1 +· · ·+a = P (1)2 ≥ 1. Now, by Cauchy-Schwarz inequality


1 x+a 0 . It is given that P (1)P 1
an xn + an−1 xn−1 + · · · + a1 x + a0 an x1n + an−1 xn−1 1
+ · · · + a1 x1 + a0

≥ (an + an−1 + · · · + a1 + a0 )2 = P (1)2 ≥ 1.

5. By Cauchy-Schwarz inequality,
p p p 2
x(3x + y) + y(3y + z) + z(3z + x) ≤ (x + y + z)(3x + y + 3y + z + 3z + x) = 4(x + y + z)2 .

6.
1 1 1 x−1 y−1 z−1
+ + = 2 =⇒ + + = 1.
x y z x y z

Then,
!2
X X 
1 C-S X√
x+y+z = x· x− ≥ x−1 .
cyc cyc
x cyc

7. Solution: This problem is special in appearance, because given condition looks heavier than what to prove! Anyways,
By CS inequality,
(a + b + 1)(a + b + c2 ) ≥ (a + b + c)2

Hence,
1 1 1 X a + b + c2 2(a + b + c) + a2 + b2 + c2
1≤ + + ≤ =
a+b+1 b+c+1 c+a+1 cyc
(a + b + c)2 (a + b + c)2

Hence (a + b + c)2 ≤ 2(a + b + c) + (a2 + b2 + c2 ) ⇒ ab + bc + ca ≤ a + b + c.

8.
C-S
(a + b4 + c4 )(a3 + 1 + 1) ≥ (a2 + b2 + c2 )2 gives
X a a4 + b4 + c4 + 2(a + b + c) a4 + b4 + c4 + 2abc(a + b + c)
≤ = ≤ 1,
cyc
a + b4 + c4 (a2 + b2 + c2 )2 (a2 + b2 + c2 )2

Since a2 b2 + b2 c2 + c2 a2 ≥ abc(a + b + c). (By C-S or AM-GM.)

30
7.3 10.3
1. Assume that all of them are true. Multiplying the first two, we get

ab + cd > (a + b)2 ≥ 4ab ⇒ cd ≥ 3ab . . . (1).

Multiplying the last two, we get

ab(ab + cd) > cd(a + b)2 ≥ cd · 4ab ⇒ ab > 3cd . . . (2).

Adding (1) and (2) we get

ab + cd > 3(ab + cd)– not possible for a, b, c, d > 0.

2. One side is easier:

1 1 1 1 1 1 1 n
+ + ··· + ≥ + + + ··· + = .
n+1 n+2 2n n+1 n+n n+n 2n 2n

The other side is a little trickier:

     
1 1 1 1 1 1 1 1 1 1 1
+ + + ··· + = + + + + ··· + +
n n+1 n+2 2n 2 n 2n n + 1 2n − 1 2n n
   
3n 1 1 1 3n 1 3 1
= 2
+ 2 + ··· + 2 ≤ (n + 1) 2 = + .
2 2n 2n + (n − 1) 2n 2 2n 4 n
√
3. Using x + y ≥ 2 xy for x, y > 0, we get

1 1 1 1
+ ≤ √ + √ .
a+b+2 c+d+2 2 ab + 2 2 cd + 2

Now using abcd = 1,

√ !
1 1 1 1 ab 1
√ + √ = √ + √ = .
2 ab + 2 2 cd + 2 2 ab + 1 1 + ab 2

And in a similar manner, we can show

1 1 1
+ ≤ .
b+c+2 d+a+2 2

4.
5.
6. r
a p p
ab (b + c)(c + a) = a3 b(b + c)(c + a) = a2 (b + c) · ab(c + a).
b
Now, by AM-GM, !
Xp 1 X X
a2 (b + c) · ab(c + a) ≤ a2 (b + c) + ab(c + a) .
cyc
2 cyc cyc

Note that, a2 (b + c) + b2 (c + a) + c2 (a + b) = ab(a + b) + bc(b + c) + ca(c + a) and hence we are left to show

ab(c + a) + bc(a + b) + ca(b + c) ≤ ab(a + b) + bc(b + c) + ca(c + a),

which reduces to 3abc ≤ ab2 + bc2 + ca2 . Last inequality is implied by AM-GM.

31
7. Let a = a21 + · · · + a2n and b = b21 + · · · + b2n . If A2 = a or B 2 = b then the given inequality holds trivially. So
assume now A2 > a, B 2 > b. Then,
p C-S p C-S p
(A2 − a)(B 2 − b) + a1 b1 + · · · + an bn ≤ (A2 − a)(B 2 − b) + ab ≤ (A2 − a + a)(B 2 − b + b) = AB.

8. Let us denote A = a2 x2 + b2 y 2 + c2 z 2 , B = b2 x2 + c2 y 2 + a2 z 2 , and C = c2 x2 + a2 y 2 + b2 z 2 . Note that

A + B + C = (a2 + b2 + c2 )(x2 + y 2 + z 2 ) = a2 + b2 + c2 .
Now,
√ √ √ C-S p p
A+ B+ C ≤ (1 + 1 + 1)(A + B + C) = 3(a2 + b2 + c2 ).

And equality is achieved when x = y = z. So it is indeed the maximum value of the given expression. On the
other hand, observe that,
C-S
AB = (a2 x2 + b2 y 2 + c2 z 2 )(b2 x2 + c2 y 2 + a2 z 2 ) ≥ (abx2 + bcy 2 + caz 2 )2 .

So that, √ √ √
AB + BC + CA ≥ (ab + bc + ca)(x2 + y 2 + z 2 ) = ab + bc + ca.

Hence,
√ √ √ 2 √ √ √ 
A+ B+ C =A+B+C +2 AB + BC + CA ≥ a2 + b2 + c2 + 2(ab + bc + ca) = (a + b + c)2 .

Again, it’s easy to check that this minimum is achieved; e.g. when x = 1, y = 0, z = 0.
√
9.
√ Since most√ of the inequalities we know are true for non-negative real numbers, so let us denote a − 1 = x,
b − 1 = y, c − 1 = z. Then we have to show

(x + y + z)2 ≤ (x2 + 1) (y 2 + 1)(z 2 + 1) + 1 .

This hints at using Cauchy-Schwarz! Indeed, by using C-S inequality, we get that

(x + (y + z))2 ≤ (x2 + 1)(1 + (y + z)2 ) and (y + z)2 ≤ (y 2 + 1)(1 + z 2 ).

Hence done.
10. By AM-GM,
p p (a + 1) + (a2 − a + 1) a2 + 2
a3 + 1 = (a + 1)(a2 − a + 1) ≤ = .
2 2
Hence,
2 2 2 1 1 1
+ 2 + 2 ≤√ +√ +√ ≤ 1.
a2 +2 b +2 c +2 a3 + 1 b3 + 1 c3 + 1

Now, by AM-HM/Titu’s lemma,

X 1 2·3
2 ≥ 2 ⇒ a2 + b2 + c2 ≥ 12.
cyc
a2 +2 a + b2 + c2 + 6

11. Since abc = 1, we can substitute a = yz , b = xz , c = xy where x, y, z are positive reals. In terms of x, y, z the
given inequality takes the form
(x + y − z)(y + z − x)(z + x − y) ≤ xyz.

32
Let x + y − z = u, y + z − x = v, z + x − y = w. If LHS is negative then it is trivial. And LHS is non-negative if
either u, v, w are all non-negative, in which case it becomes 8uvw ≤ (u + v)(v + w)(w + u), an easy application of
AM-GM.
Or exactly two among u, v, w are non-positive, say for instance u ≤ 0, v ≤ 0 and w ≥ 0. But then we observe
that, u + v = 2y ≤ 0 and we started with y > 0, so this case cannot occur.
12.
4(x2 + xy + y 2 ) = 3(x + y)2 + (x − y)2 ≥ 3(x + y)2 .
Hence,
2
34 34

8
LHS ≥ 3 (x + y)(y + z)(z + x) ≥ 3 · (x + y + z)(xy + yz + zx) .
4 4 9

And last quantity is same as (x + y + z)2 (xy + yz + zx)2 .

13. Difference technique!
X ab2 c ab2 c
  
X a AM-GM X
= a− ≥ a− √
cyc
1 + b2 c cyc
1 + b2 c cyc
2b c

X b√a · ac AM-GM X b(a + ac) 1X 1X

=4− ≥ 4− =4− ab − abc.
cyc
2 cyc
4 4 cyc 4 cyc

Now, we have a + b + c + d = 4. So using AM-GM again,

!2 !3
X 1 X X 1 X
ab ≤ a = 4, abc ≤ a = 4.
cyc
4 cyc cyc
16 cyc

Hence, LHS ≥ 4 − 1 − 1 = 2.
14. Use (x3 + y 3 ) = 8 − 6xy. So, if t = xy, we need to show t3 (4 − 3t) ≤ 1 where t ≤ 1.
1
15. Take xk = k and λk = k to apply weighted AM-GM. You may also do it by weighted GM-HM.
16.
17.
18.
19. By weighted AM-GM (with 3 and 1 as weights)
4
8(a + b + c)3

a+b+c 8abc
3 √3
+ ≥4 .
3 abc (a + b)(b + c)(c + a) 27(a + b)(b + c)(c + a)

And,
AM-GM
8(a + b + c)3 = [(b + c) + (c + a) + (a + b)]3 ≥ 27(a + b)(b + c)(c + a).

20. AM-GM twice!

 
X 1 AM-GM X 1 Weighted AM-GM 1 X1 3 1
√ ≤ √ ≤ √ + .
cyc a3 + b cyc 2 a3/4 b1/4 2 cyc 4 a b

21. By weighted AM-GM (using 1 − x and x as weights), we have

 x
1−x 1 1 1
1 ≤ (1 − x) · 1 + x · ⇒ x ≤ 2 − x.
x x x

33
22. By weighted AM-GM,
1
kk

1 ak k
(k − 1) + ak ≥ k ⇒ (1 + ak )k ≥ · ak .
k−1 (k − 1)k−1 (k − 1)k−1

So,
n n
Y Y kk
(1 + ak )k ≥ · ak = nn · a2 a3 · · · an = nn .
(k − 1)k−1
k=2 k=2

Easy to see that equality does not hold here.

23.      
b c d a a b c d b c d a
3(a + b + c + d) + + + + >3 + + + + + + +
a b c d b c d a a b c d
r
X  a a b a  X 4 a3
= + + + ≥ 4 = 4(a + b + c + d). (using abcd = 1)
cyc
b b c d cyc
bcd

34
8 Complex Numbers
8.1 11.1
1. 4z 2 + 8|z|2 = 8 =⇒ z 2 + 2|z|2 = 2
Now, 2 and 2|z|2 are integers so z must be integer . Thus , z can either be completely real or completely imaginary.
If z is completely real, √
2 2 2 2
x + 2x = 2 =⇒ 3x = 2 =⇒ x = ± √
3 √
If z is completely imaginary, −x 2 + 2x2 = 2 =⇒ x = ± 2
√
So, z = ± 2i
2 √
r
Thus solutions are ± , ± 2i
3

2. z 3 = z̄ =⇒ z 4 = |z|2
Taking modulus on both sides for z = reiθ
We get r4 = r2 =⇒ r = 0 or 1.
If r = 1,
|z|2 = 1
So, z 4 = 1 which has 4 roots namely −1, 1, i, −i .
So, all possible roots are ±1, ±i, 0

1 1 z̄ 1
3. | − | = | 2 − |
z 2 |z| 2
1 1 1 1 1 1
Now, showing | − | < is same as showing | − |2 <
z 2 2 z 2 4
1 1 1 1 1 1 z + z̄
LHS=( − )( − ) = + 2 −
z 2 z̄ 2 4 |z| 2|z|2
Now, Re(Z) > 1
1 z + z̄ 2 − (z + z̄)
So, 2
− 2
= < 0 as z + z̄ = 2Re(z) < 2
|z| 2|z| 2|z|2
1 1 z + z̄ 1
Thus , + 2 − 2
<
4 |z| 2|z| 4
Hence, proved.
4. z ′ = (z − 2)(z̄ + i) is real.
So, (z − 2)(z̄ + i) = |z|2 − 2i − 2z̄ + iz is real.
z ′ = (z − 2)(z̄ + i) is real so z¯′ is also real .
z ′ = (z − 2)(z̄ + i) = (z̄ − 2)(z − i) = |z|2 − 2z − iz̄ + 2i
As both are equal,
|z|2 − 2z − iz̄ + 2i = |z|2 − 2i − 2z̄ + iz
=⇒ 2z + iz̄ − 2i = 2z̄ − iz + 2i =⇒ 2(z − z̄) + i(z + z̄) = 4i =⇒ 2Im(z) + iRe(z) = 4i
=⇒ 2|Im(z)| = |4 − Re(z)|
Taking Re(z) = a,
|4 − a|
|Im(z)| = |4−Re(z)|
2 = |4−a|
2 =⇒ Im(z) = ±
2
5 |z| = | z1 | =⇒ |z| = | |z|z̄ 2 | =⇒ |z|3 = |z̄| = |z| =⇒ |z| = 0 or 1
So, solutions are {0, a + ib( such that a2 + b2 = 1)}
6 |z1 + z2 |2 = 3 =⇒ (z1 + z2 )(z1 + z2 ) = |z1 |2 + |z2 |2 + z1 z¯2 + z¯1 z2 = 3 =⇒ z1 z¯2 + z¯1 z2 = 1
|z1 − z2 |2 = (z1 − z2 )(z1 − z2 ) = |z1 |2 + |z2 |2 − z1 z¯2 − z¯1 z2 = 1 + 1 − 1 = 1 =⇒ |z1 − z2 | = 1
7 z n−1 = iz̄ =⇒ |z|n−1 = |z̄| =⇒ |z|(|z|n−2 − 1) = 0 So, |z| = 0 or 1
If |z| = 0 =⇒ z = 0
If |z| = 1, z n = i we have n solutions.

35
So, we have a total of n+1 solutions. (1 from 0 and other n from z n = i)
8 |z1 | = |z2 | = |z3 | = R > 0
|z1 − z2 |2 + |z2 − z3 |2 1
|z1 − z2 ||z2 − z3 | ≤ ≤ (|z1 |2 + |z2 |2 + |z2 |2 + |z3 |2 ) = 2R2
2 2
Thus, |z1 − z2 ||z3 − z1 | ≤ 2R2
|z2 − z3 ||z3 − z1 | ≤ 2R2

Therefore, we have |z1 − z2 ||z2 − z3 | + |z1 − z2 ||z3 − z1 | + |z2 − z3 ||z3 − z1 | ≤ 6R2 ≤ 9R2 9

|w| = |v| = |u − z||uz − 1| = |u − z||uz − 1|

Let
|w| ≤ 1

|u − z| ≤ |uz − 1|

(u − z)(u − z) ≤ (uz − 1)(uz − 1)

(|u|2 − 1)(|z|2 − 1) ≥ 0

|z|2 − 1 ≤ 0

Similarly, for |w| ≥ 1,

|z| ≥ 1.

1 1 1 z1 z2 + z1 z3 + z2 z3
10 z1 +z2 +z3 = 0 =⇒ z¯1 + z¯2 + z¯3 = = 0 =⇒ z1 z2 + z1 z3 + z3 z1 = 0 =⇒ z1 z2 +z1 z3 +z3 z1 =
z1 z 2 z3
0
Now, (z1 + z2 + z3 )2 = 02 = 0 and z1 z2 + z1 z3 + z3 z1 = 0

This implies z12 + z22 + z32 = 0

11 Since |zk | = r, we write zk = reiθk for k = 1, 2, . . . , n. Each term in the numerator is:
  θ +θ
  θk − θk+1 k k+1
zk + zk+1 = r eiθk + eiθk+1 = r · 2 cos ei 2 .
2

The numerator becomes:

n n   Yn
Y
n
Y θk − θk+1 θk +θk+1
(zk + zk+1 ) = (2r) cos · ei 2 .
2
k=1 k=1 k=1

The denominator is:

z1 z2 · · · zn = rn ei(θ1 +θ2 +···+θn ) .

Combining, we have:  Q θk +θk+1

Qn θk −θk+1 n i
(2r)n k=1 cos 2 k=1 e
2

E= .
rn ei(θ1 +θ2 +···+θn )

36
The exponential terms simplify because:
n
X θk + θk+1
= θ1 + θ2 + · · · + θn .
2
k=1

Thus, the exponential factor reduces to 1, and we have:

n  
n
Y θk − θk+1
E=2 cos .
2
k=1

 
Qn θk −θk+1
Since cos(x) is real for all real x, the product k=1 cos 2 is real, and E is real.

12 a Let z1 , z2 be the roots of the equation with |z1 | = 1 From z2 = ac . z11

we get |z2 | = | ac | |z11 | = 1 As z1 + z2 = − ab , |a| = |b|

we get |z1 + z2 |2 = 1
⇒ (z1 + z2 )2 = z1 .z2 = (− ab )2 = c
a

⇒ b2 = ac
b From part (a), if az 2 + bz + c = 0 has a root on the unit circle, we have b2 = ac.
Similarly, if bz 2 + cz + a = 0 has a root on the unit circle and |a| = |b| = |c|, applying the same reasoning
gives:
c2 = ab.

We now have:
b2 = ac, c2 = ab.

The conditions b2 = ac and c2 = ab imply a symmetrical relationship among a, b, c. Geometrically, these relations
mean that the arguments of a, b, c are arranged so that the points representing a, b, c in the complex plane form an
equilateral triangle. If three complex numbers of equal magnitude form an equilateral triangle, then the distances
between them are all equal. Thus:
|a − b| = |b − c| = |c − a|.

8.2 Hard Exercise

1. The equation z 12 − 236 = 0 can be factored as follows:

z 6 − 218 z 6 + 218 = 0

=⇒ z 2 − 26 z 2 + 26 (z 2 + 26 )2 − z 2 · 26 (z 2 − 26 )2 + z 2 · 26 = 0

=⇒ z 2 − 26 z 2 + 26 z 2 + 26 − z · 23 z 2 + 26 + z · 23 × z 2 − 26 − iz · 23 z 2 − 26 + iz · 23 = 0.

Since this is a 12th degree equation, there are 12 roots. Also, since each term in the equation is even, the positive
or negative value of each root is another root. That would mean there are 6 roots that can be multiplied by −1
and since we have 6 factors, that’s 1 root per factor. We just need to solve for z in each factor and pick whether
or not to multiply by i and√−1 for each √ one depending on the one that yields the highest
√ real value. After that
process, we get 8 + 8 + 2((4 3 + 4) + (4 3 − 4)) Adding the values up yields 16 + 16 3
√
2. Take z = a + ib. So, |z| = a2 + b2  
b
Then z can be written as z = |z|(cos θ + i sin θ) where θ = tan−1
a
∴ We have |z|2002 (cos(2002θ) + i sin(2002θ)) = |z|(cos(−θ) + i sin(−θ))

37
We know sin2 θ + cos2 θ = 1.
So, taking modulus on both sides in the above equation , we get |z|2002 = |z| =⇒ |z| = 0 or 1
Now, if |z| = 0 we get z = 0
If z = 1, z 2002 = z̄ ⇒ z 2003 = |z|2 = 12 = 1 =⇒ z 2003 = 1 which has 2003 solutions.
So, total number of solutions of z = 2003 + 1 = 2004
Hence, we have 2004 ordered pairs. 3 |z − αk | ≤ 1 where αk is a primitive root of unity k ∈ {0, 1, 2, 3..., (n − 1)}
Now, if we observe in a 2 − D plane , αk is the center and we are looking at circle with radius 1 for |z − αk | = 1.
Now, as we need |z − αk | ≤ 1
We have to find the intersection region of all circles |z − αk | = 1 to get the required values.
Now, 0 is in the intersection region as for all αk |0 − αk | = 1 ≤ 1

Now, if n is even , we have one circle with center at 1 and one circle with center at −1 and they intersect at
exactly one point 0 as both of them has radius 1. So, in that case 0 is the only intersection point of all the circles.
If n is odd, then for n ≥ 3 we will try to find the number of points such that |z − αk | ≤ 1 k ∈ {0, 1, 2, ..., (n − 1)
Now, say there exist x ∈ C such that x in also in the intersection region.
Then, as n ≥ 3 and no other root lies on the axis except for α = 0 then we have at least 2 roots in 2 different
quadrants.
Now, x ∈ some quadrant. Denote point x by A. Center is O and choose αk for some k such that If we call αk
as B , then ∠AOB > 90◦ , Then , we have AO2 + BO2 > AB 2 ⇒ AB 2 > 1 + AO2 . So, we get |x − αk | > 1.
Contradiction as |αk − 1| ≤ 1. So, there cannot be any other element other than 0 so 0 is the only point within
the intersection region.
Hence, z = 0
|z1 z2 + z1 z3 + z2 z3 |
4
|z1 + z2 + z3 |
Take zj = reiθj j ∈ {1, 2, 3}
|z1 z2 + z1 z3 + z2 z3 | |ei(θ1 +θ2 ) + ei(θ3 +θ2 ) + ei(θ1 +θ3 )|
Then, =r
|z1 + z2 + z3 | |eiθ1 + eiθ2 + eiθ3 |
!2
|ei(θ1 +θ2 ) + ei(θ3 +θ2 ) + ei(θ1 +θ3 ) | i(θ1 +θ2 ) + ei(θ3 +θ2 ) + ei(θ1 +θ3 ) )(e−i(θ1 +θ2 ) + e−i(θ3 +θ2 ) + e−i(θ1 +θ3 ) )
Now, r = r 2 (e
|eiθ1 + eiθ2 + eiθ3 | |eiθ1 + eiθ2 + eiθ3 |
3 + ei(θ1 −θ2 ) + ei(θ1 −θ3 ) + ei(θ2 −θ3 ) + ei(θ2 −θ1 ) + ei(θ3 −θ1 ) + ei(θ3 −θ2 )
= r2
(eiθ1 + eiθ2 + eiθ3 )(e−iθ1 + e−iθ2 + e−iθ3 )
3 + e 1 2 + ei(θ1 −θ3 ) + ei(θ2 −θ3 ) + ei(θ2 −θ1 ) + ei(θ3 −θ1 ) + ei(θ3 −θ2 )
i(θ −θ )
= r2 = r2
3 + ei(θ1 −θ2 ) + ei(θ1 −θ3 ) + ei(θ2 −θ3 ) + ei(θ2 −θ1 ) + ei(θ3 −θ1 ) + ei(θ3 −θ2 )
|z1 z2 + z1 z3 + z2 z3 |
As r > 0, So, taking square roots we get = r 5 |z1 | = |z2 | = r
|z1 + z2 + z3 |
Take z1 = reiθ1 and z2 = reiθ2
Then ,
2  iθ1 2 !
z1 + z2 2 z 1 − z2 2
 iθ1
e + eiθ2 e − eiθ2
   
1
+ = 2 +
r 2 + z 1 z2 r 2 − z1 z2 r 1 + ei(θ1 +θ2 ) 1 − ei(θ1 +θ2 )
 iθ1 iθ2
2  iθ1 iθ2
2 !
1 e + e e − e
Here r2 got cancelled for both sides so we end up with 2 +
r 1 + ei(θ1 +θ2 ) 1 − ei(θ1 +θ2 )
 iθ1
e + eiθ2
  iθ1
e + eiθ2

For first term , observe =
1 + ei(θ1 +θ2 ) 1 + ei(θ1 +θ2 )
So, it is real .
 iθ1 2  iθ1
e + eiθ2 e + eiθ2
  iθ1
e + eiθ2

Thus , =
1 + ei(θ1 +θ2 ) 1 + ei(θ1 +θ2 ) 1 + ei(θ1 +θ2 )
 iθ1
e + eiθ2 eiθ1 + eiθ2
  
1
Solve this to get ≥ (2 + 2Re(ei(θ1 −θ2 ) )
1 + ei(θ1 +θ2 ) 1 + ei(θ1 +θ2 ) 4

38
 eiθ1 − eiθ2 eiθ1 − eiθ2
   
For second term , observe =
1 − ei(θ1 +θ2 ) 1 − ei(θ1 +θ2 )
So, it is real .
eiθ1 − eiθ2 eiθ1 − eiθ2
  
1
Thus using similar arguments we get ≥ (2 − 2Re(ei(θ1 −θ2 ) )
1 − ei(θ1 +θ2 ) 1 − ei(θ1 +θ2 ) 4

Adding both the equations, we get

2 2 !
eiθ1 + eiθ2 eiθ1 − eiθ2
 
+ ≥1
1 + ei(θ1 +θ2 ) 1 − ei(θ1 +θ2 )

2  iθ1 2 !
eiθ1 + eiθ2 e − eiθ2

1 1
+ ≥ 2
r2 1+e i(θ1 +θ2 ) 1−e i(θ1 +θ2 ) r

z1 + z 2 2 z 1 − z2 2
   
1
+ ≥ 2
r 2 + z 1 z2 r 2 − z1 z2 r

39
9 Geometry
1. Observe that AF H and HEA are right-angled triangles and L is the mid-point of AH. Hence LF = LA = LE.
Similarly, considering the right triangles BF C and BEC, we get N F = N E. Since M is the mid-point of F E it
follows that ∠LM F = ∠N M F = 90◦ and L, M, N are collinear. Since LY and N X are perpendiculars to XY ,
we conclude that Y F M and F XN M are cyclic quadrilaterals. Thus

∠F LM = ∠F Y M, and ∠F XM = ∠F N M.

We also observe that CF B is a right triangle and N is the mid-point of BC. Hence N F = N C. We get

∠N F C = ∠N CF = 90◦ − ∠B.

Similarly, LF = LA gives
∠LF A = ∠LAF = 90◦ − ∠B.

We obtain

∠LF N = ∠LF C + ∠N F C = ∠LF C + 90◦ − ∠B = ∠LF C + ∠LF A = ∠AF C = 90◦ .

In triangles Y M X and LF N , we have

∠XY M = ∠F Y M = ∠F LM = ∠F LN,

and
∠Y XM = ∠F XM = ∠F N M = ∠F N L.
It follows that ∠Y M X = ∠LF N = 90◦ . Therefore Y M ⊥ M X.
2. (a) We first show that BIF X is a cyclic quadrilateral. Since ∠BIC = 90◦ + (A/2), we see that ∠BIX =
90◦ − (A/2). On the other hand, F AE is an isosceles triangle so that ∠AF E = 90◦ − (A/2). But ∠AF E = ∠BF X
as they are vertically opposite angles. Therefore ∠BF X = 90◦ − (A/2) = ∠BIX. It follows that BIF X is a cyclic
quadrilateral. Therefore ∠BXI = ∠BF I. But ∠BF I = 90◦ since IF ⊥ AB. We obtain ∠BXC = ∠BXI =
90◦ .
A similar consideration shows that ∠BY C = 90◦ . Therefore ∠BXC = ∠BY C which implies that BCY X is a
cyclic quadrilateral.
(b) We also observe that BDIX is a cyclic quadrilateral as ∠BXI = 90◦ = ∠BDI and therefore ∠BXI +∠BDI =
180◦ . This gives ∠DXI = ∠DBI = B/2. Now the concyclicity of B, I, F, X shows that ∠IXF = ∠IBF = B/2.
Hence ∠DXI = ∠IXF . Hence XI bisects ∠DXY . Similarly, we can show that Y I bisects ∠DY X. It follows
that I is the incenter of △DY X as well.
3. We may assume that B lies between C and D. Let AB = c, BC = a and CA = b. Then b > c. Let BD = x
and AD = y. Observe that ∠DAB = ∠DCA. Hence △DAB ∼ △DCA. We get
x c y
= = .
y b x+a

abc
Therefore xb = yc and by = c(x + a). Eliminating x, we get y = b2 −c2
.
Suppose gcd(b, c) = 1. Then gcd(b, b2 − c2 ) = 1 = gcd(c, b2 − c2 ). Since y is an integer, b2 − c2 divides a. Therefore
b + c divides a. Hence
a ≥ b + c.

This contradicts the triangle inequality. We conclude that gcd(b, c) > 1.

40
4. Draw a line parallel to OA through P . Let it intersect OB in M . Using compasses, draw an arc of a circle
with center M and radius M O to cut OB in L, L ̸= O. Again with L as center and with the same radius OM ,
draw one more arc of a circle to cut OB in D, D ̸= M . Join DP and extend it to meet OA in C. Then CD is the
required line segment such that CP : P D = 1 : 2. This follows from similar triangles OCD and M P D.
5. Draw the tangent P Q at D such that D is between P and Q. Join D to A, B and C. Let L = DA ∩ Γ1 and
M = DB ∩ Γ1 . Join C to L and M . Observe that

∠ADP = ∠LM D = ∠ABD. (1)

Therefore LM is parallel to AB and hence ∠LM C = ∠M CB (alternate angles). Again observe that

∠ADC = ∠LDC = ∠LM C = ∠M CB = ∠M DC = ∠BDC. (2)

Thus CD bisects ∠ADB. Hence X is the midpoint of the arc AB not containing D. Similarly Y is the midpoint
of the arc AB not containing F . Thus the arc XY is half of the sum of two arcs that together constitute the
circumference of Ω and hence it is a diameter.
6. Let D be the midpoint of AB. Observe that DX · DB = DG · DA = DY · DC. But DB = DC. Hence
DX = DY . This means that D is the midpoint of XY as well. Hence AD is also a median of △AXY . Now we
know that AG : GD = 2 : 1. If G′ is the median of △AXY , then G′ must lie on AD and AG′ : G′ D = 2 : 1. We
conclude that G = G′ .
7. First observe that ADBI is a cyclic quadrilateral since ∠AID = ∠ABD = 90◦ . Hence

∠ADI = ∠ABI = 45◦ .

Hence ∠DAI = 45◦ . But we also have

∠ADB = ∠ADI + ∠IDB = 45◦ + ∠IAB = ∠DAI + ∠IAC = ∠DAC.

Therefore CDA is an isosceles triangle with CD = CA. Since CI bisects ∠C it follows that CI ⊥ AD.
This shows that DB = CA − CB = b − a. Therefore

AD2 = c2 + (b − a)2 = c2 + b2 + a2 − 2ba = 2b(b − a).

But then 2ID2 = AD2 = 2b(b − a) and this gives

p
ID = b(b − a).

8. Extend EI to meet CB extended in D. First observe that ADBI is a cyclic quadrilateral since ∠AID = ∠ABD.
Hence
∠ADI = ∠ABI = 45◦ .
Hence ∠DAI = 45◦ . Therefore IA = ID.
Consider the triangles AIE and DIF . Both are right triangles. Moreover ∠IAE = ∠IAB = ∠IDB. Since
IA = ID, the triangles are congruent. This means IE = IF .
9. Let P be the center of the circumcircle Γ of △ABC. Let the tangent at D to Γ intersect AC in E. Then
P D ⊥ DE. Since DE bisects ∠ADC, this implies that DP bisects ∠ADB. Hence the circumcenter and the
incenter of △ABD lies on the same line DP . This implies that DA = DB. Thus DA = DB = DC and hence D
is the circumcenter of △ABC. This gives ∠A = 90◦ .
10. Extend JI to meet CB extended at L. Then ALBI is a cyclic quadrilateral. Therefore

∠BLI = ∠BAI = ∠IAC.

41
Therefore ∠LAD = ∠IBD = 45◦ . Since ∠AIL = 90◦ , it follows that ∠ALI = 45◦ . Therefore AI = IL. This
shows that the triangles AIJ and LID are congruent giving IJ = ID. Similarly, IK = IE. Since IJ ⊥ ID and
IK ⊥ IE and since DJ = EK, we see that IJ = ID = IK = IE. Thus E, D, J, K are concyclic.
This implies together with DJ = EK that EDJK is an isosceles trapezium. Therefore DE ∥ JK. Hence

∠EDA = ∠DAC = ∠A/2

and
∠DEC = ∠ECA = ∠C/2.
Since IE = ID, it follows that ∠A/2 = ∠C/2. This means BA = BC.
11. Draw ID ⊥ AC. Then ID = r, the inradius of △ABC. Observe EF ∥ BC and hence ∠AEF = ∠ABC = 90◦ .
Hence ∠AIF = 90◦ . Therefore ID2 = F D · DA. If a > c, then F A > DA and we have
b
DA = s − a, and F D = F A − DA = − (s − a).
2

Thus we obtain
(b + c − a)(a − c)(a − c)
r2 = .
4

But r = (c + a − b)/2. Thus we obtain

(c + a − b)2 = (b + c − a)(a − c).

Simplification gives 3b = 3a + c. Squaring both sides and using b2 = c2 + a2 , we obtain

BC a 4
= = .
BA c 3

(If a ≤ c, then I lies outside the circumcircle of △AEF .)

12. If O is the centre of Σ, then we have
1 1
∠AEB = ∠AOB = (180◦ − ∠ACB)
2 2
1 1 1
= ∠EDB = (180◦ − ∠EAB) = 90◦ − ∠EAB.
2 2 2
But we know that ∠AEB + ∠EAB + ∠EBA = 180◦ .
Therefore
1
∠EBA = 180◦ − ∠AEB − ∠EAB = 180◦ − 90◦ + ∠EAB − ∠EAB.
2
This shows that ∠AEB = ∠EBA and hence AE = AB.
13. Given that X is the incentre of triangle ABY , we have ∠BAX = ∠XAY . Therefore,

∠BDC = ∠BAC = ∠BAX = ∠XAY = ∠XDY = ∠BDY.

This shows that C, D, Y are collinear. Therefore,

∠CY X + ∠XY D = 180◦ .

But the left-hand side equals (180◦ − ∠CBD) + (180◦ − ∠CAD). Since ∠CBD = ∠CAD, we obtain

180◦ = 360◦ − 2∠CAD.

42
This shows that ∠CAD = 90◦ .
14 We first show that AA′ is perpendicular to B ′ C ′ . Observe ∠C ′ A′ A = ∠C ′ CA = ∠C/2; ∠A′ C ′ C = ∠A′ AC =
∠A/2; and ∠CC ′ B ′ = ∠CBB ′ = ∠B/2. Hence

∠C ∠A ∠B
∠C ′ AP + ∠AC ′ P = ∠C ′ AB + ∠BAP + ∠AC ′ P = + + = 90◦ .
2 2 2

It follows that ∠AP C ′ = ∠A′ P C ′ = 90◦ . Thus ∠IP Q = 90◦ . Since P IRQ is a kite, we observe that ∠IP R =
∠IRP and ∠QP R = ∠QRP . This implies that ∠IRQ = 90◦ . Hence the kite IRQP is also a cyclic quadrilateral.
Since ∠IRQ = 90◦ , we see that BB ′ ⊥ AC. Since BB ′ is the bisector of ∠B, we conclude that ∠A = ∠C.
We also observe that the triangles IRC and IP B ′ are congruent triangles: they are similar, since ∠IRC =
∠IP B ′ = 90◦ and ∠ICR = ∠IB ′ P (= ∠BCC ′ ); besides IR = IP . Therefore IC = IB ′ . But B ′ I = B ′ C. Thus
IB ′ C is an equilateral triangle. This means ∠B ′ IC = 60◦ and hence ∠ICR = 30◦ . Therefore ∠C/2 = 30◦ . Hence
∠A = ∠C = 60◦ . It follows that ABC is equilateral.
15. Let C be the reflection of O′ with respect to O. Then in triangle O′ AC, the midpoints of the segments O′ A
and O′ C are M and O, respectively. This implies AC is parallel to OM , and hence B lies on AC. Let the line
AC intersect Γ again at N . Since O′ C is a diameter of Γ, it follows that ∠O′ N C = 90◦ . Since O′ A = O′ B, we
can now conclude that N is the midpoint of the segment AB.

43
10 Coordinate geometry
10.1 Exercise 1
1. The distance between two points (x1 , y1 ) and (x2 , y2 ) is given by:
p
d = (x2 − x1 )2 + (y2 − y1 )2 .

Let the vertices be A(4, −4), B(−2, 4), and C(6, 10).

p p √
AB = (−2 − 4)2 + (4 − (−4))2 = (−6)2 + 82 = 36 + 64 = 10.

p p √
BC = (6 − (−2))2 + (10 − 4)2 = (8)2 + 62 = 64 + 36 = 10.

p p √ √
AC = (6 − 4)2 + (10 − (−4))2 = (2)2 + (14)2 = 4 + 196 = 200.

Since AB = BC = 10, the triangle is isosceles.

2. Three points are collinear if the slope of AP is equal to the slope of P B. The slope of a line between two points
(x1 , y1 ) and (x2 , y2 ) is given by:
y2 − y1
m= .
x2 − x1

- Slope of AP :
2−2 0
mAP = = = 0.
6 − (−2) 8

- Slope of P B:
5−2 3
mP B = = = 0.5.
12 − 6 6

Since both slopes are equal, the points A, P , and B are collinear.
Let P divide AB in the ratio k : 1. Using the section formula, the coordinates of P are:
 
kx2 + x1 ky2 + y1
P = , ,
k+1 k+1

where A(x1 , y1 ) and B(x2 , y2 ).

Substitute P (6, 2), A(−2, 2), and B(12, 5):

12k − 2 5k + 2
6= , 2= .
k+1 k+1

Solve for k using the x-coordinate:

4
6(k + 1) = 12k − 2 =⇒ 6k + 6 = 12k − 2 =⇒ 6k = 8 =⇒ k = .
3

Verify using the y-coordinate:

4
2(k + 1) = 5k + 2 =⇒ 2k + 2 = 5k + 2 =⇒ 3k = 0 =⇒ k = .
3

Hence, P divides AB in the ratio 4 : 3.

44
4. Show that (2, −1) is the center of the circumcircle of ∆ABC, where A = (−3, −1), B = (−1, 3), and
C = (6, 2). Find the circumradius.
The circumcenter of a triangle is the intersection of the perpendicular bisectors of its sides.
- Midpoint of AB:  
−3 + (−1) −1 + 3
M1 = , = (−2, 1) .
2 2
- Midpoint of BC:    
−1 + 6 3 + 2 5 5
M2 = , = , .
2 2 2 2

- Slope of AB:
3 − (−1) 4
mAB = = = 2.
−1 − (−3) 2
Perpendicular slope = − 12 . - Slope of BC:

2−3 −1
mBC = = .
6 − (−1) 7

Perpendicular slope = 7.
- Perpendicular bisector of AB: Using point-slope form y − y1 = m(x − x1 ), the equation is:
1
y − 1 = − (x + 2) =⇒ 2y − 2 = −x − 2 =⇒ x + 2y = 0.
2
- Perpendicular bisector of BC: Using the midpoint 25 , 52 and slope 7:

 
5 5
y− =7 x− =⇒ y = 7x − 30.
2 2

Solve the system of equations:

x + 2y = 0, y = 7x − 30.
Substituting y = 7x − 30 into x + 2y = 0:

x + 2(7x − 30) = 0 =⇒ x + 14x − 60 = 0 =⇒ 15x = 60 =⇒ x = 4.

Substituting x = 4 into y = 7x − 30:

y = 7(4) − 30 = 28 − 30 = −2.

The circumcenter is (2, −1).

The circumradius is the distance from the circumcenter to any vertex, e.g., A(−3, −1):
p p
r = (2 − (−3))2 + (−1 − (−1))2 = (5)2 + (0)2 = 5.

Thus, the circumcenter is (2, −1), and the circumradius is 5.

5. Let the vertices of the quadrilateral be A(x1 , y1 ), B(x2 , y2 ), C(x3 , y3 ), and D(x4 , y4 ).
- Midpoint of AB:  
x1 + x2 y1 + y2
M1 = , .
2 2
- Midpoint of CD:  
x 3 + x 4 y3 + y4
M2 = , .
2 2

45
- Midpoint of AD:  
x1 + x4 y1 + y4
N1 = , .
2 2
- Midpoint of BC:  
x2 + x3 y2 + y3
N2 = , .
2 2

- Line joining M1 and M2 : The midpoint of this line is:

! 
y1 +y2
x1 +x2
+ x3 +x + y3 +y 4
4

2 2 2 2 x1 + x2 + x3 + x4 y1 + y2 + y3 + y4
P = , = , .
2 2 4 4

- Line joining N1 and N2 : The midpoint of this line is:

! 
x1 +x4 x2 +x3 y1 +y4 y2 +y3 
2 + 2 2 + 2 x1 + x2 + x3 + x4 y1 + y2 + y3 + y4
Q= , = , .
2 2 4 4

The midpoints of both lines are identical:

 
x1 + x2 + x3 + x4 y1 + y2 + y3 + y4
P =Q= , .
4 4

Thus, the lines joining the midpoints of opposite sides of a quadrilateral bisect one another.
6. To determine the ratio in which (7, 5) divides AB and CD, we use the section formula.
If a point P (x, y) divides a line segment joining A(x1 , y1 ) and B(x2 , y2 ) in the ratio k : 1, then:
 
kx2 + x1 ky2 + y1
P (x, y) = , .
k+1 k+1

Let (7, 5) divide AB in the ratio k : 1. Using the section formula:

5k + 1 4k + 2
7= , 5= .
k+1 k+1

Solve for k using the x-coordinate:

7(k + 1) = 5k + 1 =⇒ 7k + 7 = 5k + 1 =⇒ 2k = −6 =⇒ k = −3.

Verify using the y-coordinate:

5(k + 1) = 4k + 2 =⇒ 5k + 5 = 4k + 2 =⇒ k = −3.

Thus, (7, 5) divides AB in the ratio −3 : 1.

Let (7, 5) divide CD in the ratio k : 1. Using the section formula:

3k − 5 3k − 1
7= , 5= .
k+1 k+1

Solve for k using the x-coordinate:

7(k + 1) = 3k − 5 =⇒ 7k + 7 = 3k − 5 =⇒ 4k = −12 =⇒ k = −3.

46
Verify using the y-coordinate:

5(k + 1) = 3k − 1 =⇒ 5k + 5 = 3k − 1 =⇒ 2k = −6 =⇒ k = −3.

Thus, (7, 5) divides CD in the ratio −3 : 1.

7. Let the vertices of ∆ABC be A(x1 , y1 ), B(x2 , y2 ), and C(x3 , y3 ), and the centroid G is given by:
 
x1 + x2 + x3 y1 + y2 + y3
G , .
3 3

- AB 2 = (x2 − x1 )2 + (y2 − y1 )2 , - BC 2 = (x3 − x2 )2 + (y3 − y2 )2 , - CA2 = (x1 − x3 )2 + (y1 − y3 )2 .

The squared distance between G and any vertex (e.g., A) is:
 2  2
2 x1 + x2 + x3 y1 + y2 + y3
GA = − x1 + − y1 .
3 3

Simplify:
 2  2
2 −2x1 + x2 + x3 −2y1 + y2 + y3
GA = + .
3 3

Similarly, find GB 2 and GC 2 .

By using coordinate transformations and algebraic simplifications, it can be shown that:

AB 2 + BC 2 + CA2 = 3(GA2 + GB 2 + GC 2 ).

- OA2 = (x1 − xO )2 + (y1 − yO )2 , - OB 2 = (x2 − xO )2 + (y2 − yO )2 , - OC 2 = (x3 − xO )2 + (y3 − yO )2 .

2
2
Using the centroid’s coordinates and the squared distances GO2 = xO − x1 +x32 +x3 + yO − y1 +y32 +y3 , it can
be shown that:
OA2 + OB 2 + OC 2 = GA2 + GB 2 + GC 2 + 3GO2 .

8. The incenter of a triangle is the point where the angle bisectors of the triangle intersect. It is also the center
of the inscribed circle. The incenter can be calculated using the formula:
 
ax1 + bx2 + cx3 ay1 + by2 + cy3
I= , ,
a+b+c a+b+c

where: - a, b, c are the lengths of the sides opposite to vertices A(x1 , y1 ), B(x2 , y2 ), and C(x3 , y3 ), respectively.
Let the vertices be A(0, 0), B(20, 15), and C(36, 15).
1. Length of side a (opposite to vertex A):
p √
a = BC = (36 − 20)2 + (15 − 15)2 = 162 = 16.

2. Length of side b (opposite to vertex B):

p p √ √
b = AC = (36 − 0)2 + (15 − 0)2 = 362 + 152 = 1296 + 225 = 1521 = 39.

3. Length of side c (opposite to vertex C):

p p √ √
c = AB = (20 − 0)2 + (15 − 0)2 = 202 + 152 = 400 + 225 = 625 = 25.

47
Substitute the values of a, b, c, and the coordinates of the vertices A(0, 0), B(20, 15), and C(36, 15) into the formula
for the incenter:
ax1 + bx2 + cx3 ay1 + by2 + cy3
Ix = , Iy = .
a+b+c a+b+c

1. Calculate Ix :
16(0) + 39(20) + 25(36) 0 + 780 + 900 1680
Ix = = = = 21.
16 + 39 + 25 80 80

2. Calculate Iy :
16(0) + 39(15) + 25(15) 0 + 585 + 375 960
Iy = = = = 12.
16 + 39 + 25 80 80

Thus, the incenter is:

I = (21, 12).

9. We use the midpoint formula and the distance formula to prove the result.
- A(x1 , y1 ), - B(x2 , y2 ), - C(x3 , y3 ), - D, the midpoint of BC, is:
 
x2 + x3 y2 + y3
D= , .
2 2

1. Square of AB:
AB 2 = (x2 − x1 )2 + (y2 − y1 )2 .

2. Square of AC:
AC 2 = (x3 − x1 )2 + (y3 − y1 )2 .

3. Square of AD: Using the coordinates of D:

 2  2
2 x2 + x3 y2 + y3
AD = − x1 + − y1 .
2 2

4. Square of DC: Using the coordinates of D:

2
y2 + y3 2
  
2 x2 + x3
DC = x3 − + y3 − .
2 2

Simplify:
 2  2
2 x3 − x2 y3 − y2 1
(x3 − x2 )2 + (y3 − y2 )2 .


DC = + =
2 2 4

1. AB 2 + AC 2 = (x2 − x1 )2 + (y2 − y1 )2 + (x3 − x1 )2 + (y3 − y1 )2 .

2. Expand 2AD2 + 2DC 2 : - Expand AD2 :
 2  2
2 x2 + x3 y2 + y3
AD = − x1 + − y1 .
2 2

Expand each term:

(x2 + x3 − 2x1 )2 (y2 + y3 − 2y1 )2
AD2 = + .
4 4

- Expand 2DC 2 :
1
2DC 2 = (x3 − x2 )2 + (y3 − y2 )2 .


2

48
3. Add 2AD2 + 2DC 2 : After simplifying and grouping terms, it can be shown that:

AB 2 + AC 2 = 2AD2 + 2DC 2 .

Thus, the result is proved.

⃗ and OB
10. The dot product of two vectors OA ⃗ is given by:

⃗ · OB
OA ⃗ = |OA||
⃗ OB|
⃗ cos ∠AOB,

where: q q
⃗ =
|OA| x21 + y12 , ⃗ =
|OB| x22 + y22 .

The dot product can also be written as:

⃗ · OB
OA ⃗ = x1 x2 + y1 y2 ,

where x1 , y1 are the coordinates of point A, and x2 , y2 are the coordinates of point B.
From the definition of the dot product:
⃗ · OB
OA ⃗ = |OA||
⃗ OB|
⃗ cos ∠AOB.

⃗ and |OB|:
Substitute the magnitude of |OA| ⃗
q q
⃗ =
|OA| x21 + y12 , ⃗ =
|OB| x22 + y22 .

Thus: q q
x1 x2 + y1 y2 = x21 + y12 · x22 + y22 · cos ∠AOB.

Rearranging:
x1 x2 + y1 y2
cos ∠AOB = .
⃗ OB|
|OA|| ⃗

Finally:
OA · OB cos ∠AOB = x1 x2 + y1 y2 .

11.
Let P (x, y) be the moving point. The condition states:

Distance from P to (−1, 0) = 3 · Distance from P to (0, 2).

Using the distance formula: p p

(x + 1)2 + (y − 0)2 = 3 · (x − 0)2 + (y − 2)2 .

(x + 1)2 + y 2 = 9 x2 + (y − 2)2 .

Expand the left-hand side:

(x + 1)2 + y 2 = x2 + 2x + 1 + y 2 .

Expand the right-hand side:

9 x2 + (y − 2)2 = 9x2 + 9(y 2 − 4y + 4) = 9x2 + 9y 2 − 36y + 36.

49
Thus:
x2 + 2x + 1 + y 2 = 9x2 + 9y 2 − 36y + 36.

Group terms and simplify:

x2 + y 2 − 9x2 − 9y 2 + 2x + 36y + 1 − 36 = 0.

Combine like terms:

−8x2 − 8y 2 + 2x + 36y − 35 = 0.

Divide through by −1:

8x2 + 8y 2 − 2x − 36y + 35 = 0.

Factorize coefficients of x2 and y 2 :

x 9y
8(x2 − ) + 8(y 2 − ) = −35.
4 2

Complete the square for x and y: 1. For x, x2 − x4 :

1 2
 
2 x 1
x − = x− − .
4 8 64

9y
2. For y, y 2 − 2 :  2
2 9y 9 81
y − = y− − .
2 4 16

Substitute these back into the equation:

" # " #
1 2 9 2 81
 
1
8 x− − +8 y− − = −35.
8 64 4 16

Simplify:
1 2 9 2
   
8 x− +8 y− = 5.5.
8 4

Divide through by 5.5 to normalize:

1 2 9 2



x− 8 y− 4
5.5 + 5.5 = 1.
8 8

Thus, the locus of P is an ellipse.

12. (a) P A2 − P B 2 = 2k 2 = constant.
Using the distance formula:
P A2 = (x − a)2 + y 2 , P B 2 = (x + a)2 + y 2 .

Substitute into the given condition:

P A2 − P B 2 = (x − a)2 + y 2 − (x + a)2 + y 2 .
   

Simplify:
(x − a)2 − (x + a)2 = 2k 2 .

Expand:
x2 − 2ax + a2 − (x2 + 2ax + a2 ) = 2k 2 .

50
Simplify:
k2
−4ax = 2k 2 =⇒ x = − .
2a

Thus, the locus of P is a vertical line:

k2
x=− .
2a

(b) P A + P B = c = constant.
Using the distance formula: p p
PA = (x − a)2 + y 2 , PB = (x + a)2 + y 2 .

The condition becomes: p p

(x − a)2 + y 2 + (x + a)2 + y 2 = c.

This is the standard equation for an ellipse with foci at A(a, 0) and B(−a, 0), and the major axis c satisfying
c > 2a.
Thus, the locus of P is an ellipse:
P A + P B = c.

(c) P B 2 + P C 2 = 2P A2 , where C = (c, 0).

Using the distance formula:

P A2 = (x − a)2 + y 2 , P B 2 = (x + a)2 + y 2 , P C 2 = (x − c)2 + y 2 .

Substitute into the given condition:

(x + a)2 + y 2 + (x − c)2 + y 2 = 2 (x − a)2 + y 2 .

 

Expand all terms:

(x + a)2 + (x − c)2 + 2y 2 = 2 (x − a)2 + y 2 .
 

Expand and simplify:

x2 + 2ax + a2 + x2 − 2cx + c2 + 2y 2 = 2x2 − 4ax + 2a2 + 2y 2 .

Simplify further:
2x2 + 2ax − 2cx + a2 + c2 + 2y 2 = 2x2 − 4ax + 2a2 + 2y 2 .

Cancel terms:
2ax − 2cx + a2 + c2 = −4ax + 2a2 .

Combine like terms:

6ax − 2cx = c2 − a2 .

Simplify:
c2 − a2
x= .
6a − 2c

Thus, the locus of P is a vertical line:

c2 − a2
x= .
6a − 2c

51
10.2
1. Let the equation of the straight line be:
x y
+ = 1,
a b
where a and b are the x- and y-intercepts of the line, respectively. The endpoints of the segment intercepted by
the line on the coordinate axes are:
A(a, 0) and B(0, b).
The midpoint (x1 , y1 ) of the segment is given by:

a+0 a 0+b b
x1 = = , y1 = = .
2 2 2 2
Expressing a and b in terms of x1 and y1 , we get

a = 2x1 , b = 2y1 .

Substituting a and b into the equation of the line:

x y x y
+ = 1 =⇒ + = 1.
a b 2x1 2y1
Thus, the equation of the line is:
x y
+ = 1.
2x1 2y1

11 Sequences
1. The relations
1 1
am + am = (a2m + a0 ) and a2m + a0 = (a2m + a2m )
2 2
a2 +a4
imply a2m = 4am , as well as a0 = 0. We compute a2 = 4, a4 = 16. Also, a1 + a3 = 2 = 10, so a3 = 9. At this
point, we guess that ak = k 2 for all k ≥ 1.
We prove our guess by induction on k. Suppose that aj = j 2 for all j < k. The given equation with m = k − 1
and n = 1 gives
1
an = (a2n−2 + a2 ) − an−2 = 2an−1 + 2a1 − an−2 .
2

Substituting values:
an = 2(n2 − 2n + 1) + 2 − (n2 − 4n + 4).

Simplify:
an = 2(n2 − 2n + 1) + 2 − (n2 − 4n + 4) = n2 .

2. Examining the sequence, we see that the mth term of the sequence is equal to n exactly for those m that
satisfy
n2 − n n2 + n
+1≤m≤ .
2 2
So the sequence grows about as fast as the square root of twice the index. Let us rewrite the inequality as

n2 − n + 2 ≤ 2m ≤ n2 + n,

then try to solve for n. We can almost take the square root. And because m and n are integers, the inequality is
equivalent to
1 1
n2 − n + < 2m < n2 + n + .
4 4

52
Here it was important that n2 − n is even. And now we can take the square root. We obtain
1 √ 1
n− < 2m < n + ,
2 2
or
√ 1
n< 2m + < n + 1.
2
√
Now this happens if and only if n = ⌊ 2m + 21 ⌋, which then gives the formula for the general term of the
sequence:
√
 
1
am = 2m + , m ≥ 1.
2
√
3. The sequence xn = n n − 1 is clearly positive, so we only need to bound it from above by a sequence converging
to 0. For that we employ the binomial expansion:
     
n n n 2 n
n = (1 + xn ) = 1 + xn + x + ··· + xn−1 + xnn .
1 2 n n−1 n

Forgetting all terms but one, we can write:  

n 2
n> x ,
2 n
which translates to: r
2
xn < , for n ≥ 2.
n−1
q
2
The sequence n−1 , n ≥ 2, converges to 0, and hence by the squeezing principle, (xn ) itself converges to 0, as
desired. □
4. We consider the sequence (mod 4):
1, 1, 2, 3, 3, 2, 3, 3, . . .
It has period 2, 3, 3 and does not contain a zero.
5. For a strictly increasing function an , we have

a2n = an + a2 ≥ a2 + (n − 1).

This is impossible for any finite value a2 .

6. We have
n n n
Y k3 − 1 Y k − 1 Y k2 + k + 1
= .
k3 + 1 k+1 k2 − k + 1
k=2 k=2 k=2

The first product is

2
.
n(n + 1)
To find the second product, we observe that if bk = k 2 + k + 1, ck = k 2 − k + 1, then ck = bk−1 . Hence, the second
product is
n2 + n + 1
.
3
Finally,
2 n2 + n + 1 2
lim 2
= .
n→∞ 3 n + n 3

53
7. The table for an suggests an+2 = 4an − an−2 , (n = 3, 4, 5, . . .). We prove this by induction. Suppose that the
formula is valid for n − 1. That is,

an−1 an+2 = 1 + an+1 an = 1 + (4an−1 − an−3 )an = 4an−1 an − an−1 an−2 ,

an+2 = 4an − an−2 .

This proves that an is an integer for all n

8 Observe that we have
1 1 1 1
xk+1 = x2k + xk =⇒ = = − .
xk+1 xk (1 + xk ) xk 1 + xk

We get
1 1 1 1 1 1 1 1 1
+ + ··· + = − + ··· + − = − ,
x1 + 1 x2 + 1 x101 + 1 x1 x2 x100 x101 x1 x101
and this is
1
2− .
x101
The integer part is 1 since x101 > 1.

11.1
1. An1 ≤ An1 + · · · + Ank ≤ kAn1 , so we have

A1 = lim (An1 )1/n ≤ lim (An1 + · · · + Ank )1/n ≤ lim (kAn1 )1/n = A1 .
n→∞ n→∞ n→∞

showing that the limit equals A1 .

2. Obviously, xn ≥ 1 for all n; so, if the limit exists, it is ≥ 1, and we can pass to the limit in the recurrence
relation to get
3 + 2x∞
x∞ = ;
3 + x∞
in other√words, x2∞ + x∞ − 3 = 0. So x∞ is the positive solution of this quadratic equation, that is, x∞ =
1
2 (−1 + 13).

To prove that the limit exists, we use the recurrence relation to get

3 + 2xn 3 + 2xn−1 3(xn − xn−1 )

xn+1 − xn = − = .
3 + xn 3 + xn−1 (3 + xn )(3 + xn+1 )

Hence, |xn+1 − xn | ≤ 13 |xn − xn−1 |. Iteration gives

1
|xn+1 − xn | ≤ 3−n |x1 − x0 | = .
3n ·4

The seriesP ∞
P 1 P∞ −n
n=1 (xn+1 −xn ), of positive terms, is dominated by the convergent series 4 n=1 3 and so converges.
We have ∞ n=1 (xn+1 − xn ) = limn→∞ xn − x1 and we are done.

3. By the given relation xn − xn−1 = (α − 1)(xn − xn−1 ). Therefore, by the Induction Principle, we have
xn − xn−1 = (α − 1)n−1 (x1 − x0 ). Hence,
n
X n
X
xn − x0 = (xk − xk−1 ) = (x1 − x0 ) (α − 1)k−1 .
k=1 k=1

54
Taking limits, we get
(1 − α)x0 + x1
lim xn = .
n→∞ 2−α

4. It can be proved by induction on n that

3 3
P (n + 1) = P (n) and A(n + 1) = A(n)
2 4
holds for every n ≥ 1. Hence we get
 n−1  n−1
3 3
P (n) = P (1) and A(n) = A(1)
2 4
3

n
for every n ≥ 1. Since 0 < 4 < 1, we know that 43 → 0 as n → ∞, which implies that

lim A(n) = 0.
n→∞

3 3 n


On the other hand, since 2 > 1, we know that 2 diverges to +∞ as n → ∞. Therefore, we can write

lim P (n) = ∞.
n→∞

5. Using the inequalities

 k  k
1 1
1+ <e< 1+ (k ≥ 2),
k k−1
we get
2n 2n 2n 2n 2n
!     !
Y k X 1 k X 1 X 1 k+1 Y k+1
log 2 = log = log > > log = log
k−1 k k−1 k k k k
k=n+1 k=n+1 k=n+1 k=n+1 k=n+1
 
2n + 1
= log ;
n+1
therefore, we have
2n
X 1
log 2 ≥ lim ≥ log 2
n→∞ k
k=n+1

and the result follows.

an+1
6. Observe that using the recursive relation given we can say that an +1 = . Plugging this into the expression
n
of P , we can see that its a telescopic product and we get that
an + 1
Pn =
n!
Splitting the terms and using the recursion, we get that
n
X 1
Pn =
k!
k=1

Thus the limit of this required product is e

7. Since this is nothing but an AGP(Arithmetico Geometric Progression) hence we can directly find the sum, we
get that
lim un = 1
n→∞

55
12 Differentiation
1. Let f : [1, ∞) → R, f (x) = x1/x . We seek distinct integers a, b such that f (a) = f (b). Observe:
• f (x) has a local maximum at x = e; logarithmic differentiation.
• In fact, f ′ (x) < 0 for x > e and f ′ (x) > 0 for 1 < x < e.
• limx→∞ f (x) = 1; logarithms and L’Hôpital’s Rule.
Now f (2) = f (4), and this is the only solution! Indeed, by the Intermediate Value Theorem for n ≥ 5, f (n) = f (α),
for some α ∈ (1, 2). But there are no integers between 1 and 2.
2. Consider the function x1/x . Differentiating gives x1/x x12 (1 − ln x), so the function attains its global maximum

at x = e.
Thus e1/e ≥ π 1/π , and it is clear that the inequality is strict, so eπ > π e .
3. If x = y then x = y = 0 which is not possible as x and y are both positive. Assume that there are x ̸= y > 0
such that
x2y + y2−x = x + y
and let y = x1 − x2 ,
4. We prove that no such functions exists. Assume the contrary and let k be an integer. From the mean value
theorem we obtain
φ(k + 1) − φ(k) = φ′ (ζ) , ζ ∈ (k, k + 1)
Since φ(k) and φ(k + 1) are integers hence φ(k + 1) − φ(k) is also an integer and so is φ′ (ζ). On the other ζ is
not an integer and hence φ′ (ζ) is not an integer. Thus we get a contradiction.
k
5. Let us define xk = a + (b − a), 0 ≤ k ≤ n. Observe that
n
b−a
xk+1 − xk =
n
The mean value theorem tells us that
b−a ′
f (x1 ) − f (x0 ) = f (θ1 ) , θ1 ∈ (x0 , x1 )
n
b−a ′
f (x2 ) − f (x1 ) = f (θ2 ) , θ2 ∈ (x1 , x2 )
n
..
.
b−a ′
f (xn ) − f (xn−1 ) = f (θn ) θn ∈ (xn−1 , xn )
n
Adding these we get that
n
b−aX ′
f (xn ) − f (x0 ) = f (θi )
n
i=1
n
f (b) − f (a) 1X ′
= f (θi )
b−a n
i=1

6. Let F : R → R, F (x) = f (x) − g(x) and note that that F is a differentiable function. Since F (x1 ) = F (x2 ) = 0,
from Rolle’s Theorem there is a c ∈ (x1 , x2 ) such that F ′ (c) = 0. On the other hand, F ′ (x) = f ′ (x) − g ′ (x) =
f (x) + g(x) and therefore,
f (c) + g(c) = 0

7. We have
x
x
x

lim (φ(x) + φ +φ + ··· + φ
x→0 2 3 n

56
    
x x
2 − φ(0) n − φ(0)
!
φ(x) − φ(0) 1 φ 1 φ
= lim + x + ··· + x
x→0 x−0 2 2 −0 n n −0
 1 1
= φ′ (0) 1 + + · · · +
2 n

Since φ(0) = 0 and φ is differentiable at the origin

ln ax
8. Consider the function f : (0, ∞) → R, f (x) = . We have
x
1 − ln ax
f ′ (x) =
x2
e e
Thus we can say that f ′ (x) = 0 if and only if x = . It follows that that µ = is the point of maxima of the
a a
function so µ is the only point such that f (x) ≤ f (µ) for all positive real numbers x. Hence

µx
≥ aµ−x
xµ

9. Consider the function F : [a, b] → R

F (x) = f (x)e−λf (x)
This function F is differentiable, since f and f ′ are differentiable and F (a) = F (b). By Rolle’s theorem it follows
that there is a c ∈ (a, b) such that F ′ (c) = 0. On the other hand

F ′ (x) = e−λf (x) (f ′′ (x) − λ(f ′ (x))2 )

Thus we can say that f ′′ (c) − λ(f ′ (c))2 = 0

57
13 Riemann Integration
1. Note that for 0 ≤ x ≤ 1, we have
Z 1 Z x Z 1
x3 x 1
f (x) = |t − x|t dt = (x − t)t dt + (t − x)t dt = − + .
0 0 x 3 2 3

For x < 0,
1
1 1
t3 t2
Z Z
1 x
f (x) = |t − x|t dt = (t − x)t dt = −x = − .
0 0 3 2 3 2
0

Finally, for x > 1,

1
1 1
t2 t3
Z Z
x 1
f (x) = |t − x|t dt = (x − t)t dt = x − = − .
0 0 2 3 2 3
0

It is quite easy to draw the graph of this function f (x), because for x ∈ [0, 1] it is just a cubic polynomial and for
both x ≤ 0 and x ≥ 1 it is a straight line.
2. By symmetry, it is enough to find the number of lattice points (points having integer coordinates) in the first
quadrant. This can be calculated by first fixing the x-coordinate to be k and then summing up for k = 1, 2, . . . , n.
Note that jp k
# (x, y) : x = k, y ∈ Z, y ≥ 0 and x2 + y 2 ≤ n2 =

n2 − k 2 .

Hence,
n jp
X k
C(n) = 4 n2 − k 2 + 1
k=1
2
where the last +1 is for the origin (0, 0). Next, inj√order tokcalculate the limit of C(n)/n as n → ∞, observe that
n
it is enough to calculate the limit of n−2 k=1 n2 − k 2 and we can handle the floor function using sandwich
P

principle. The inequality x − 1 ≤ ⌊x⌋ ≤ x produces the following bounds

n n n
!
1 X p 2 2
 1 X jp 2 k
2 ≤
1 Xp 2
n − k − n ≤ n − k n − k2 .
n2 n2 n2
k=1 k=1 k=1

Observe that
n p n
! Z 1p
1 X  1 Xp 2
lim n2 − k 2 − n = lim n − k 2 = 1 − x2 dx.
n→∞ n2 n→∞ n2 0
k=1 k=1

This integral calculates the area of one quarter of the unit circle, hence equals π/4. (Alternatively, you can use
integration by parts.) Finally, applying the Sandwich theorem we conclude that
n
C(n) 4 X jp 2 2
k π
lim = lim 2 n − k = 4 × = π.
n→∞ n2 n→∞ n 4
k=1

This is intuitive, because the area of the circle x2 +y 2 ≤ n2 being πn2 (square units), it should include approximately
πn2 many unit squares. The above limit makes this idea precise.
3. Since h(x) = f (x) − g(x) is a polynomial of degree less than or equal to n, and h satisfies
Z 1
xk h(x)dx = 0 for each k = 0, 1, . . . , n,
0

58
we can easily deduce that Z 1
h(x)2 dx = 0.
0

But h(x)2 is a non-negative and continuous function, so the above equation can hold if and only if h is identically
zero on [0, 1]. Therefore, f (x) = g(x) for every x ∈ [0, 1]. Since f and g are polynomials, this is enough to conclude
that f = g.
R1 R1 R1 2
4. To start with, note that y0 = 0 f = 0 g, and y1 = 0 fg . How to show y0 ≤ y1 ? Well, the Cauchy-Schwarz
inequality gives
Z 1 2  Z 1  Z 1 2
f
g ≥ f =⇒ y1 y0 ≥ y02 =⇒ y1 ≥ y0 .
0 g 0 0

Let’s proceed by strong induction. Suppose that yk ≤ yk+1 holds for all k ≤ n − 1. How can we show that
yn ≤ yn+1 ? Cauchy-Schwarz inequality gives
1 1 1 2
f n+2 fn f n+1
Z  Z  Z
≥
0 g n+1 0 g n−1 0 gn

which tells us that yn+1 yn−1 ≥ yn2 . Hence yn+1 /yn ≥ yn /yn−1 and yn /yn−1 ≥ 1 holds by induction hypothesis.
This completes the induction and hence the proof.
Rx
6. Define F (x) = a f ′ (t) dt for t ∈ [a, b]. Then by the previous exercise, we can say that F is differentiable on
[a, b], with F ′ (t) = f ′ (t) for every t ∈ [a, b]. In other words, the function g = F − f will be a differentiable function
having derivative equal to 0 on the entire [a, b], which implies that g must be a constant function (this may be
justified using MVT). Thus, F (x) − f (x) = c for every x ∈ [a, b]. Putting x = a, and using F (a) = 0 (from its
definition), we get that c = −f (a). Therefore,
Z b
f ′ (t) dt = F (b) = f (b) + c = f (b) − f (a),
a

which completes the proof.

7. Just apply Rolle’s theorem or the Mean Value Theorem on the function
Z x
F (x) = f (t)dt, x ∈ [a, b]
a

which is differentiable here since f is continuous.

8. Define Z x
g(x) = f (t) dt − x3 , x ∈ [0, 1].
0
Note that g(1) = g(0) = 0, and invoking the Fundamental Theorem of Calculus (FTC), we can say that g is
continuous on [0, 1] and differentiable on (0, 1). Hence we can apply Rolle’s theorem on g, which gives the desired
conclusion.
9. Using the fact that f (x) = sin x is an increasing function on [0, π/2], for π/6 < x < π/2 we have 1/2 < sin x < 1.
Hence Z π/2 Z π/2 Z π/2
x x x
dx ≤ dx ≤ dx.
π/6 1 π/6 sin x π/6 1/2

Observing that
π/2
π2
Z
x dx = ,
π/6 9

59
1. (a) This is really straightforward. On one hand we have
Z 1  Z 1 
n 1
lim fn (x) dx = lim x dx = lim = 0,
n→∞ 0 n→∞ 0 n→∞ n + 1

while on the other hand we have

(
0 if 0 ≤ x < 1,
lim fn (x) = lim xn =
n→∞ n→∞ 1 if x = 1.

Hence Z 1 
lim fn (x) dx = 0.
0 n→∞

(b) First note that Z 1  Z 1 

n n
lim fn (x) dx = lim nx dx = lim = 1.
n→∞ 0 n→∞ 0 n→∞ n+1
Next, we show that
lim fn (x) = lim nxn = 0, for every 0 ≤ x ≤ 1.
n→∞ n→∞

This is trivial for x = 0 and x = 1 (since fn (1) = 0 for all n ≥ 1). For 0 < x < 1, we take the help of r = 1/x > 1,
and see that
n n 2
0 < nxn = n
≤ n
= 2
→ 0 as n → ∞.
(1 + r − 1) 2 (r − 1)
2 (r − 1) (n − 1)
Therefore in this problem we have
Z 1  Z 1 
lim fn (x) dx = 0 ̸= 1 = lim fn (x) dx .
0 n→∞ n→∞ 0

2. For any x > x0 , we apply MVT on F to say that there exists cx ∈ (x0 , x) such that

F (x) − F (x0 )
= F ′ (cx ) = f (cx ). (2)
x − x0

Now letting x → x+ ′
0 , the above LHS converges to F (x0 ) = f (x0 ). What about the RHS? Since x0 < cx <
+
x =⇒ cx → x0 , so the RHS of (2) converges to limx→x+ f (cx ) = limcx →x+ f (cx ) = a. Hence we get the desired
0 0
conclusion.
3. One way toRattack this problem is by means of contradiction. But a simpler way is to use FTC, as follows.
x
Define F (x) = a f (t) dt for x ∈ [a, b]. Since f is continuous, it holds by FTC that F is differentiable and F ′ = f
on (a, b). But it is given that F is a constant function, implying that f (x) = F ′ (x) = 0 for any x ∈ (a, b). Finally,
continuity ensures that f must also vanish at the endpoints of [a, b].
Rx
4. Define F (x) = a f (t) dt for x ∈ R. Since f is continuous, it holds by FTC that F is differentiable and F ′ = f
on R. Now, Z a+T
def
g(a) = f (t) dt = F (a + T ) − F (a)
a
is given to be a constant function (since g(a) = g(0) for every a ∈ R). Hence for every a ∈ R, we must have
g ′ (a) = 0. But
g ′ (a) = F ′ (a + T ) − F ′ (a) = f (a + T ) − f (a).
So we get the desired conclusion that f (a + T ) = f (a) must hold for every a ∈ R.
5. Since x2 + 1 ≥ 2x > 0, we have
1 1
xn xn
Z Z
1
In = √ dx ≤ √ dx = √ .
0 x2 + 1 0 2x 2(n + 1/2)

60
 √
6. First recall that 1 < sin x + cos x < 2 for every x ∈ (0, π/2). Also recall that, when x ∈ (0, 1), a > b actually
implies xa < xb (not its opposite). Therefore,
√
2
∀x ∈ (0, 1], x < xsin x+cos x < x1 .

(At x = 0, these are all equal.) Upon integration, the above inequalities produce the following:
Z 1 √ Z 1 Z 1
x 2 dx < xsin x+cos x dx < x1 dx.
0 0 0

(Do you see why strict inequality holds here?) Finally note that
Z 1 √ √
x 2 dx = 2 − 1 > 0.4.
0

Thus we get the desired inequality

Z 1 √ Z 1 Z 1
2 sin x+cos x
0.4 < x dx < x dx < x1 dx = 0.5.
0 0 0

7. We start by defining Z a+1

1
I(a) = dx.
a−1 1 + x8
Applying the Leibniz rule, we get
1 1
I ′ (a) = 8
− .
1 + (a + 1) 1 + (a − 1)8
Note that I ′ (a) = 0 ⇐⇒ (a + 1)8 = (a − 1)8 ⇐⇒ a = 0, 21 . Now I ′ (a) is of the form

I ′ (a) = c(a) · (a − 1)2 − (a + 1)2 = c(a) · (−4a),

where c(a) is positive for any a ∈ R. Therefore, I ′ (a) changes its sign only when a = 0.
Furthermore, note that I ′ (a) > 0 for a < 0 and I ′ (a) < 0 for a > 0. Hence, I(a) is maximised at a = 0.
8. If f had an anti-derivative, the problem would have ′
′
R b been much easier. How? Suppose that F = f . Since
F = f is non-decreasing, F must be convex. Also, a f (u)du = F (b) − F (a) for any a, b ≥ 0. Hence, the given
inequality simplifies as follows: Z z Z z
(z − x) f (u)du ≥ (z − y) f (u)du
y x

⇐⇒ (z − x)(F (z) − F (y)) ≥ (z − y)(F (z) − F (x))

F (z) − F (y) F (z) − F (x)
⇐⇒ ≥
z−y z−x
z−y y−x
⇐⇒ F (y) ≤ F (x) + F (z) (*)
z−x z−x
Now observe that y = λx + (1 − λ)z ⇐⇒ λ = (z − y)/(z − x). So (*) is the same as saying

F (λx + (1 − λ)z) ≤ λF (x) + (1 − λ)F (z)

which follows from the convexity of F . Another way to finish the above solution is as follows. Observe that
F (z) − F (y) F (z) − F (x) F (z) − F (x) F (y) − F (x)
≥ ⇐⇒ ≥ .
z−y z−x z−x y−x
Now by MVT, there exists c1 ∈ (x, y) and c2 ∈ (y, z) such that
F (y) − F (x) F (z) − F (y)
= F ′ (c1 ) = f (c1 ), and = F ′ (c2 ) = f (c2 ).
y−x z−y

61
Since c1 < y < c2 and f is non-decreasing, we get f (c1 ) ≤ f (c2 ), which completes the proof.

Next, let us discuss a proof that does not rely on the assumption that f has an anti-derivative:
Z z Z z
(z − x) f (u)du ≥ (z − y) f (u)du
y x
Z z Z y Z z
⇐⇒ (z − x) f (u)du ≥ (z − y) f (u)du + (z − y) f (u)du
y x y
Z z Z y
⇐⇒ (y − x) f (u)du ≥ (z − y) f (u)du
y x

Since f is non-decreasing, for any u ∈ (y, z) we have f (u) ≥ f (y), and for any u ∈ (x, y) we have f (u) ≤ f (y).
Therefore, Z z
(y − x) f (u)du ≥ (y − x)(z − y)f (y)
y
Z y Z y
= f (y)du ≥ (z − y) f (u)du.
x x
This completes the proof.
9. Let [x] = m. The LHS divided by 2 can be written as

x m−1
X k(k + 1) Z k+1 m(m + 1) x
Z Z
[u]([u] + 1)
f (u)du = f (u)du + f (u)du. (3)
1 2 2 k 2 m
k=1

On the other hand, the RHS divided by 2 can be simplified as

[x] Z x Z x Z x Z x
X
n f (u)du = f (u)du + 2 f (u)du + · · · + m f (u)du. (4)
n=1 n 1 2 m

Now we can break each integral on the RHS of (4) as the sum of ‘consecutive’ integrals, e.g.,
Z x Z 2 Z 3 Z m Z x
f (u)du = f (u)du + f (u)du + · · · + f (u)du + f (u)du.
1 1 2 m−1 m
R k+1
In this manner, note that for any 1 ≤ k ≤ m − 1, the integral k f (u)du appears on the RHS of (4) exactly
Rx
(1 + 2 + · · · + k) = k(k+1)
2 times, while the integral m f (u)du appears m(m + 1)/2 times. Therefore, the RHS of
(4) is the same as the RHS of (3), which completes the proof. □

62
1.
R1 1
(a) It equals 0 xk dx = k+1 . (Ans)
(b) Taking log, we get
n   Z 1
1X k x=1
lim log 1 + = log(1 + x)dx = x log x − x = 2 log 2 − 1.
n→∞ n n 0 x=2
k=1

So the desired limit equals exp(2 log 2 − 1) = 4/e. (Ans)

R1 1
(c) It equals 0 1+x3 dx. Evaluating this is usually carried out using a partial fraction decomposition: by assuming
that
1 A Bx + C
2
= +
(1 + x)(1 − x + x ) x + 1 1 − x + x2
is an identity we solve for A, B, C, and then use standard integrals. Another way is to do some algebra and
cleverly write it as
1 1 1 1 1 2x − 1 1 1
Z Z Z
1
dx − dx + dx.
6 0 x+1 6 0 x2 − x + 1 2 0 x2 − x + 1
Anyway, these are some very standard methods that I hope you already are (or, going to be) familiar with
π
them. The final answer is 13 log 2 + 3√ 3
. (Ans)

Qn n+k
(d) Since 2n
n = k=1 k , the given limit equals
Z 1   Z 2 Z 1
1
log 1 + dx = log xdx − log xdx = (2 log 2 − 1) − (−1) = log 4.
0 x 1 0

(Ans)
2. The first one can be calculated as follows.
Z e Z 1 Z e
| log x| dx = | log x| dx + | log x| dx
1/e 1/e 1
Z 1 Z e
= − log x dx + log x dx
1/e 1

= x − x log x|x=1 x=e

x=1/e + x log x − x|x=1 = 2(1 − 1/e). (Ans)

3. First we write
Z π Z π Z π
I= xf (sin x)dx = (π − x)f (sin(π − x))dx = (π − x)f (sin x)dx
0 0 0

and then adding up these two alternate expressions for the same integral, we get
Z π Z π/2
2I = π f (sin x)dx = 2π f (sin x)dx
0 0
R 2a Ra
where in the last step we used 0 f (x)dx = 0 (f (x) + f (2a − x))dx. □
Using the above formula/idea, we get
Z π π/2
x sin2n x sin2n x
Z
dx = π dx
0 sin2n x + cos2n x 0 sin2n x + cos2n x
Ra Ra
Now using 0 f (x)dx = 0 f (a − x)dx,
π/2 π/2 π/2
sin2n x cos2n x
Z Z Z
1 π
I= 2n dx = 2n dx = dx = .
0 sin x + cos2n x 0 sin x + cos2n x 2 0 4

63
Therefore, the desired integral equals π 2 /4. (Ans)
4. First we substitute y = nx to write
Z π Z nπ Z nπ
sin nx sin y dy sin y
dx = = dy.
0 x 0 y/n n 0 y
R π R 2π
Now break the integral as the sum of integrals 0 , π , etc. as follows.

Z nπ n Z kπ
sin y X sin y
dy = dy
0 y (k−1)π y
k=1
n Z kπ
X |sin y|
≥ dy (since (k − 1)π < y < kπ ⇒ 1/y > 1/kπ)
(k−1)π kπ
k=1
n Z π  
X 1 2 1 1
= |sin y| dy = 1 + + ··· +
kπ 0 π 2 n
k=1
as required.
5. Let me do the first two, and leave the rest for you. For n ≥ 1, define
Z π/2 Z π/2
n
In = sin x dx = cosn x dx.
0 0

For instance, I0 = π/2, and I1 = 1. How to calculate In for a general n? The idea is to get a recursion for In and
then solve that recursion. For n > 1, we integrate by parts to get
Z π/2
In = (sin x)n−1 · sin x dx
0
 Z π/2 Z π/2 Z 
n−1 d
= (sin x) sin x dx − (sin x)n−1 sin x dx dx
0 0 dx
π/2
Z π/2
= −(sin x)n−1 cos x 0 + (n − 1)(sin x)n−2 cos2 x dx

0
Z π/2
=0+ (n − 1)(sin x)n−2 (1 − sin2 x) dx = (n − 1)(In−2 − In ).
0
Thus, In = (n − 1)(In−2 − In ), which can also be written as
n−1
In = In−2 , n ≥ 2.
n

Now, for an even n, say n = 2k where k ≥ 1, we have

2k − 1 2k − 1 2k − 3 1 × 3 × · · · × (2k − 1)
I2k = I2k−2 = · I2k−4 = · · · = I0 .
2k 2k 2k − 2 2 × 4 × · · · × 2k
Similarly, for odd n, say n = 2k + 1 where k > 1, we have
2k 2k 2k − 2 2 × 4 · · · × 2k
I2k+1 = I2k−1 = · I2k−3 = · · · = I1 .
2k + 1 2k + 1 2k − 1 3 × 5 · · · × (2k + 1)

We can also write 

 (2k−1)!! π = (2k
k)
π
if n = 2k ≥ 0,
(2k)!! 2 22k+1
In = (3)
 (2k)!! = 2k 2k
−1 if n = 2k + 1 ≥ 1.
(2k+1)!! 2k+1 k

64
These integrals (In ) are commonly known as Wallis’ integrals.
6. Using the Binomial theorem,
n  
2 n
X n
(1 − x ) = (−x2 )k .
k
k=0
Integrating both sides, and noting that the RHS being a finite summation we can pass the integral sign inside the
summation, we get
Z 1 n  Z 1 n  
2 n
X n 2 k
X n (−1)k
(1 − x ) dx = (−x ) dx = .
0 k 0 k 2k + 1
k=0 k=0

Now, we can calculate the integral on the LHS directly (using by parts or by substitution) and hence get an
expression for the sum on the RHS.
Z 1 Z π/2 Z π/2
2 n 2 n 2 × 4 × · · · × 2k
(1 − x ) dx = (1 − sin θ) cos θ dθ = (cos θ)2n+1 dθ = .
0 0 0 1 × 3 × · · · × (2n + 1)
where the last integral was evaluated using (3). Therefore,
       
1 n 1 n 1 n n 1 n (2n)!!
− + − · · · + (−1) = . (Ans)
1 0 3 1 5 2 2n + 1 n (2n + 1)!!

7. For any x > 1, we calculate the following integral by substituting u = 1/t

Z 1/x Z 1 Z x
log t log(1/u) (−1) du log u
dt = 2
= du.
1 1+t x 1 + 1/u u 1 1+u

Therefore, Z x Z 1 Z x
log t log t log t 1 x 1
f (x) + f (1/x) = dt + dt = dt = (log t)2 = (log x)2 .
1 1+t x 1+t 1 t 2 1 2

So, f (x) + f (1/x) = 2 ⇐⇒ (log x)2 = 4 ⇐⇒ log x = ±2 ⇐⇒ x = e2 or e−2 .

Ra Ra
8. Using the formula −a f (x)dx = 0 (f (x) + f (−x))dx, we get
Z a
g(x)dx = 0
0

for all a ∈ R where g(x) = f (x) + f (−x). In a previous exercise we saw that this implies g ≡ 0, which here forces
f to be an odd function.
√ f ′ (x)
9. The derivative of x is 12 x1/2−1 = 2√1 x . So, dx
d
p
f (x) = √ . Now we can proceed in many ways. One way
2 f (x)
is to say that the function
p 1
g(x) = f (x) − x
2
has derivative
f ′ (x) 1
g ′ (x) = p − ≥ 0,
2 f (x) 2
hence g is increasing and therefore for any x ≥ 1, we have g(x) ≥ g(1), which gives the desired inequality.
10. First note that f ′ (x) > 0 so f is increasing. Hence for x ≥ 1, we can say that f (x) ≥ f (1) = 1. Therefore,
1 1
f ′ (x) = ≤ 2 for all x ≥ 1. (4)
x2 + f (x)2 x +1

Now Z x Z x
1 π π π
f (x) − f (1) = f ′ (t)dt ≤ dt = tan−1 x − tan−1 1 < − = .
1 1 1 + t2 2 4 4

65
Since f is increasing and bounded above, we can say that limx→∞ f (x) exists, and from the above inequalities, it
is immediate that the limit should be less than or equal to π/4.
But how to claim that the limit is strictly less than π/4? Showing that is quite tricky, because even if you have
f (x) < g(x) for all x, taking limit as x → ∞ (or x → a) would change the < sign to a ≤ sign. Here we adopt the
following approach.
If f never crosses c where 1 < c < 1 + π/4 then it is trivial that limx→∞ f (x) ≤ c < 1 + π/4.
Else, f (x0 ) > c for some x0 > 1, then f (x) ≥ f (x0 ) > c for all x > x0 , and hence
1 1
f ′ (t) = ≤ 2 , for t ≥ x0 .
t2 + f (t)2 t + c2
Integrating this inequality from x0 to x and integrating (4) from 1 to x0 , we obtain
Z x0 Z x
1 1
f (x) − f (1) ≤ 2+1
dt + 2 + c2
dt
1 t x0 t
for every x > x0 . Letting x → ∞ here, we get
Z x0 Z ∞ Z ∞
1 1 1 π
lim f (x) ≤ 1 + 2
dt + 2 2
dt < 1 + 2
dt = 1 + .
x→∞ 1 1+t x0 t + c 1 t +1 4

11. First observe that f (x) + f (1 − x) = 1 for every x ∈ R. Then note that
f (f (1 − x)) = f (1 − f (x)) = 1 − f (f (x)).
In fact, you can do induction on n to show that if g be f composed with itself n times, then g also satisfies
g(x) + g(1 − x) = 1. Hence, for any n ≥ 1, we can write
Z 1 Z 1 Z 1 
I= f [n] (x) dx = f [n] (1 − x) dx = 1 − f [n] (x) dx
0 0 0

and then add up these two alternate expressions for I to show that I = 1/2. (Ans)
12. We observe that
Z 1 n−1 Z n−1 Z n−1
1 n 1 X k+1 1X 1
Z
1X
f (nx)dx = f (y)dy = f (y)dy = f (u + k)du = ak .
0 n 0 n k n 0 n
k=0 k=0 k=0

Now you have to use the following fact: if (an )n≥0 be a sequence that converges to a, then the sequence (bn )n≥1
defined by
n−1
1X
bn = ak
n
k=0
also converges to a. This tells us that the desired limit also equals a. (Ans)
13. Applying integration by parts, we get
Z b  Z b Z b  Z 
f (x) sin(nx)dx = f (x) sin(nx)dx − f ′ (x) sin(nx)dx dx
a a a
b
f (a) cos na − f (b) cos nb 1
Z
= − f ′ (x) cos(nx)dx. (†)
n n a

Now, since f is continuously differentiable on [a, b], we can say that f ′ is bounded on [a, b]. In other words, we can
find an M > 0 such that |f ′ (x)| < M holds for every x ∈ [a, b]. So, 0 ≤ |f ′ (x) cos nx| ≤ M also holds for x ∈ [a, b]
and therefore we obtain from (†) that
Z b
f (a) cos na − f (b) cos nb 1 b ′
Z
0≤ f (x) sin(nx)dx ≤ + f (x) cos(nx)dx
a n n a

66
 b
|f (a) cos na| + |f (b) cos nb| 1
Z
≤ + |f ′ (x) cos(nx)|dx
n n a
|f (a)| + |f (b)| M (b − a)
≤ + → 0, as n → ∞.
n n
This proves that the desired limit is 0.

14 Combinatorics
1. Let abcdef denote the 6 vertical stripes in order from the left. Then a can be of any one of the 4 colours; then
b can be of any one of the other 3 colours; then c can be again of any one of 3 colours since the colour used for a
is now available. Similarly, there are 3 possible colours for each of d, e, f . Hence, by MP, there are 4 × 35 ways of
designing the flag.
2.
(i) A 5-digit number is of the form abcde where a, . . . , e are the digits in it and a is the leading digit. Now a
cannot be 0 and so can be chosen in 9 ways (any one of 1, . . . , 9). Each of the remaining digits can be chosen
in 10 ways since repetition is allowed. So, in all 9 × 104 ways.
(ii) The number abcde is even if and only if the units digit e is even; so there are 5 choices for e: 0, 2, 4, 6, 8. So
in all 9 × 103 × 5 ways.
(iii) Two cases: either only a = 3 or exactly one of b, c, d, e = 3. There are 94 ways in the first case and 8 × 93 × 4
in the second case (factor 4 for the 4 subcases b = 3 or c = 3 etc.); so in all 94 + 8 × 93 × 4 ways.
(iv) By the given condition, the values of a, b, c fix the values of d, e : e = a, d = b. Hence, the number of ways is
9 × 102 .
3. We are considering the integers t such that 1 ≤ t ≤ 3333. Clearly, the largest number t having 0 in the units
place is 3330. So there are 333 numbers t having 0 in the units place: they are 10, 20, . . . , 3330. We can describe
these numbers as t = x0 where x is any one of 1, 2, . . . , 333.
Similarly, numbers t = x0y i.e. numbers having 0 in the tens place are in all 33 × 10 because x can be any one of
1, 2, . . . , 33 and y can be any one of 0, 1, 2, . . . , 9. Thus there are 33 × 10 = 330 numbers like x0y. In the same
way, there are 3 × 102 = 300 numbers with 0 in the hundreds place (i.e. x0yz where 1 ≤ x ≤ 3, 0 ≤ y, z ≤ 10).
Hence the total number of times 0 is written is 333 + 330 + 300 = 963.
4. Since we are counting the occurrences of digit 5, consider integers t such that 1 ≤ t ≤ 105 . Clearly, the largest
number t having 5 in the units place is 99995. So there are 1 + 9999 = 104 numbers t having 5 in the units
place: they are 5, 15, 25, . . . , 99995. We can describe these numbers as t = x5 where x is any one of 0, 1, 2, . . . ,
9999.
Similarly, numbers t = x5y, i.e., numbers having 5 in the tens place are in all (1 + 999) × 10 = 104 because x can
be any one of 0, 1, 2, . . . , 999 and y can be any one of 0, 1, 2, . . . , 9. In the same way, there are 104 numbers
in each of the following cases: numbers with 5 in the hundreds place or thousands place or ten thousands place.
Hence the total number of times 5 is written is 104 × 5.
5. A collection is determined once we know the number of times A occurs in it and the number of times B occurs
in it. Let these numbers be i and j respectively. Then i takes values from 0 to 3 and j from 0 to 5. Hence to
obtain a collection we have 4 choices for i and 6 choices for j so that there are 4 × 6 = 24 collections (for example,
i = 1, j = 2 gives the collection {A, B, B}). The empty collection corresponds to letting i = j = 0; so that omitting
this case there are 24 − 1 = 23 non-empty collections.
6. To make a collection we have to select a certain number of objects of each kind. Now for each value of r,
1 ≤ r ≤ k, from nr like objects we can choose 0, 1, . . . , or nr objects; i.e. there are nr + 1 choices. Hence,
by MP, there are in all (n1 + 1)(n2 + 1) . . . (nk + 1) collections and so the number of non-empty collections is
(n1 + 1)(n2 + 1) . . . (nk + 1) − 1.

67
7. Let S = {a1 , a2 , . . . , an }. Note that we form a subset T of S in n stages as follows: we have 2 choices for a1 :
either a1 is included in T or a1 is not included in T . Similarly, we have 2 choices for a2 : either a2 is included in T
or a2 is not included in T , etc. Finally, we have 2 choices for an : either an is included in T or an is not included
in T . (For example, if n = 4, then the subset {a2 , a4 } corresponds to the sequence of choices no, yes, no, yes.)
Hence, by MP, the total number of subsets is 2 × 2 × · · · × 2 (n factors) i.e. 2n .
8. Here m = 30030 = 2 × 3 × 5 × 7 × 11 × 13, and each prime factor occurs only once. Hence every factor of m
corresponds to the product of elements of a particular subset of the 6-set S = {2, 3, 5, 7, 11, 13}. Note that the
empty subset corresponds to the factor 1. Hence the number of factors of m is equal to the number of subsets of
S, namely 26 , by the last problem.
9. A function f from A to B corresponds to the ordered set given by (f (1), . . . , f (m)) of m elements where f (i)
is the value of f at i. Now for each i in A, f (i) can be chosen in n ways from B. Hence, by the multiplication
principle, there are n × n × n × · · · × n (m factors) = nm functions from A to B.
The number of one-one functions will be n(n − 1) . . . (n − m + 1). Thus, if n < m then the number of one-one
functions is zero and if n ≥ m then the number of one-one functions is
n!
.
(n − m)!

10. The function is to be onto B and so it must take on both the values b1 and b2 . So there are 2 choices for the
value of the function at each of the n elements of A, namely b1 or b2 , excepting the case when all values are equal
to b1 and the case when all values are equal to b2 . Hence the number of required functions is 2n − 2.

14.1 .2
1. There are 32 black squares and of these 12 can be chosen to put the 12 white pawns in 32


12 ways. Then out
20


of the remaining 20 black squares 12 can be chosen to put the 12 black pawns in 12 ways. So the answer is
32 20



12 × 12 .

2. Since the 2 rooks are identical, the order in which they are placed in a row (or column) is not important. First
a row can be chosen in 8 ways. In any row, 2 of the 8 squares can be chosen in 8 C2 ways. Hence the 2 rooks can
be placed in a row in 8 × 8 C2 ways. Similarly, 2 rooks can be placed in a column in 8 × 8 C2 ways. So the total
number is 2 × 8 × 8 C2 . For an n × m board the answer is n × m C2 + m × n C2 .
3. A row can be chosen in 8 ways. In a row, there are in all 8 C2 pairs of squares of which 7 are pairs of adjacent
squares. So there are [8 C2 − 7] pairs of places in any of which the two identical kings can be placed. Hence the
kings can be placed in a row in 8 × [8 C2 − 7] ways and in the same number of ways in a column. So the total
number of ways is 2 × 8 × [8 C2 − 7]. For an n × m board, the number is

n [m C2 − (m − 1)] + m [n C2 − (n − 1)] .

4. The different necklaces are determined by the number of red beads between the two blue beads, taking both
arcs into consideration. So there are exactly 3 distinct necklaces:

bbrrrrr, brbrrrr, brrbrrr.

5. This is very straightforward. The answers are

1. 9! × 3!
2. 8! × 9 P3
6. There are q + 1 places between the q 0’s (namely, q − 1 places between the q successive 0’s and 2 places at the
ends). The required arrangements are obtained by putting the p 1’s in p of these q + 1 places; and this can be
done in q+1


p ways.

68
7. An r-subset T of S corresponds, in a one-to-one way, to an arrangement a1 , a2 , . . . , an of r 1’s and n − r 0’s in
a row as follows: ai = 1 if i ∈ T and ai = 0 if i ∈
/ T . Hence the r-subsets of S = {1, 2, . . . , n} that do not contain
a pair of consecutive integers exactly correspond to the arrangements of r 1’s and n − r 0’s in
a line such that no
two 1’s are adjacent. Hence, by the last problem, the number of required r-subsets is n−r+1 r .

15 Pigeon Hole Principle

1. A person goes into box #i if they have made i handshakes. We have n persons and n boxes numbered
0, 1, . . . , n − 1. But the boxes with the numbers 0 and n − 1 cannot both be occupied. Thus, there is at least one
box with more than one person.
2. Same as problem 1 with handshakes replaced by contests.
3. Denote the 20 integers a1 to a20 . Then 0 < a1 < · · · < a20 < 70. We want to prove that there is a k, so that
aj − ai = k has at least four solutions. Now

0 < (a2 − a1 ) + (a3 − a2 ) + · · · + (a20 − a19 ) = a20 − a1 ≤ 68.

We will prove that, among the differences ai+1 − ai , i = 1, . . . , 19, there will be four equal ones. Suppose there are
at most three differences equal. Then

3 · 1 + 3 · 2 + 3 · 3 + 3 · 4 + 3 · 5 + 3 · 6 + 7 ≤ 68,

that is, 70 ≤ 68. Contradiction!

4. Consider the three coordinates mod 2. There are 23 = 8 possible binary 3-words. Since there are nine words
altogether, at least two sequences must be identical. Thus there are two points (a, b, c) and (r, s, t) with integral
midpoint M = ((a + r)/2, (b + s)/2, (c + t)/2).
5. Subdivide the unit square
√ into 25 small squares of side 1/5. There will be√three insects in one of these squares
of side 1/5 and diagonal 2/5. A circumcircle of this square has radius 2/10 < 1/7. If we circumscribe a
concentric circle with radius 1/7, it will cover this square completely.
6. Consider the n integers 1, 11, . . . , 11 · · · 1 mod n. There are n possible remainders 0, 1, . . . , n − 1. If 0
occurs, we are finished. If not, two of the numbers have the same remainder mod n. Their difference 11 · · · 100 · · · 0
is divisible by n. Since n is not divisible by 2 or 5, we can strike the zeros at the end and get the number consisting
of ones and divisible by n.
7. We use the same motive. Consider the sums

a1 , a1 + a2 , a1 + a2 + a3 , . . . , a1 + a2 + · · · + an .

If any of the n sums is divisible by n, then we are done. Otherwise, two of the sums a1 + · · · + ai and a1 + · · · + aj
have the same remainder upon division by n. Suppose j > i. Then the difference ai+1 + · · · + aj is divisible by
n.
8. Among n + 1 integers from 1, . . . , 2n, there are two successive integers. They are coprime.
9. A set S of 10 numbers with two digits, each one ≤ 99 has 210 = 1024 subsets. The sum of the numbers in any
subset of S is ≤ 10 · 99 = 990. So there are fewer possible sums than subsets. Thus there are at least two different
subsets S1 and S2 having the same sum. If S1 ∩ S2 = ∅, then we are finished. If not, we remove all common
elements and get two nonintersecting subsets with the same sum of their elements.
10. Use induction from n to 2n, which corresponds to induction from k to k + 1.
(1) For n = 1, the statement is correct.
(2) Suppose that, from 2n − 1 integers, we can always select n with sum divisible by n. Of the 2(2n) − 1 positive
integers, we can select n numbers three times, which are divisible by n. After the first selection, there will remain

69
3n − 1 numbers, after the second selection, 2n − 1 numbers. Let the sum of the first choice be a · n, the sum of
the second choice be b · n, and the last choice be c · n. At least two of the numbers a, b, c have the same parity,
e.g., a and b. Then an + bn = (a + b)n is divisible by 2n, since a + b is even.
11. Consider all subsets {i1 , . . . , ik } of the set {1, . . . , n}. Let S(i1 , . . . , ik ) = ai1 + · · · + aik . The number of such
sums is 2n − 1. Since 2n − 1 > n2 for n ≥ 5, two of these sums will have the same remainder upon division by
n2 . Their difference will be divisible by n2 . This difference has the form ±as1 ± as2 ± · · · ± ast for some t ≥ 1 and
some selection of indices s1 , . . . , st .
12. Considering the fractional parts of these numbers, we get n − 1 reals in the interval [0, 1]. Subdivide this
unit interval into n equal parts, each of length 1/n. If one of the n points falls into the first interval, then we are
finished. Otherwise, two points, say {ia} and {ka}, fall into the same interval. Then the point {(k − i)a} is a
distance ≤ 1/n from 0.
√
13. Two of six points placed into a 3 × 4 rectangle will have distance ≤ 5.
14. This problem contains all necessary hints for a solution. It is a problem for the box principle, since all existence
problems about finite sets somehow√rely on the box principle. Furthermore, it contains the hint to the addition
theorem for tan, and 0 = tan 0, 1/ 3 = tan(π/6) give the missing hints for the boxes. So we set yi = tan xi ,
yj = tan xj and get
π
tan 0 ≤ tan(xi − xj ) ≤ tan .
6
Because tan is monotonically increasing everywhere, we get
π
0 ≤ xi − xj ≤ .
6

The yi can lie anywhere in the infinite interval −∞ < yi < ∞. But the xi are confined to the interval −π/2 < xi <
π/2. For at least two of the seven xi we have 0 ≤ xi − xj ≤ π/6. The original inequality follows from this.

16 Invariance Principle
1. In one move the number of integers always decreases by one. After (4n − 2) steps, just one integer will be left.
Initially, there are 2n even integers, which is an even number. If two odd integers are replaced, the number of
odd integers decreases by 2. If one of them is odd or both are even, then the number of odd numbers remains the
same. Thus, the number of odd integers remains even after each move. Since it is initially even, it will remain
even to the end. Hence, one even number will remain.
2.
3. Consider the remainder mod 9. It is an invariant. Since 106 = 1 mod 9 the number of ones is by one more
than the number of twos.
4. Here, I(x1 , x2 , . . . , x6 ) = 2x1 + 4x2 + 6x3 + 8x4 + 10x5 + 12x6 mod 10 is the invariant. Starting with
I(1, 0, 1, 0, 1, 0) = 8, the goal I(0, 1, 0, 1, 0, 1) = 4 cannot be reached.
5. We proceed by contradiction. Suppose all the remainders 0, 1, . . . , 2n − 1 occur. The sum of all integers and
their place numbers is
S1 = 2 (1 + 2 + · · · + 2n) = 2n (2n + 1) ≡ 0 (mod 2n).
The sum of all remainders is

S2 = 0 + 1 + · · · + 2n − 1 = n (2n − 1) ≡ n (mod 2n).

Contradiction!
6. We partition the participants into the set E of even persons and the set O of odd persons. We observe that,
during the hand shaking ceremony, the set O cannot change its parity. Indeed, if two odd persons shake hands,

70
O increases by 2. If two even persons shake hands, O decreases by 2, and, if an even and an odd person shake
hands, |O| does not change. Since, initially, |O| = 0, the parity of the set is preserved.
7. The number of heads is invariant mod 3. Initially, it is 1 and it remains so.
8. Observe that the initial sum is odd. At some step if we remove the numbers a, b and replace it with their
difference then we see that the sum changes to

S → S − (a + b) + (a − b) = S − 2b

Hence the parity of the sum remains the same. Hence all the numbers can not be 0
9. For any collection of n numbers on the blackboard we consider the following quantity X: the sum of all the
numbers decreased by n. Assume that we have transformed the collection as described in the statement. How
would the quantity X change? If the sum of all the numbers except a and b equals S, then before the transformation
X = S + a + b − n, and after the transformation X = S + (a + b − 1) − (n − 1) = S + a + b − n. So the value of X
is the same: it is invariant. Initially (for the collection in the statement) we have

X = (1 + 2 + · · · + 19 + 20) − 20 = 190.

Therefore, after 19 operations, when there will be only one number on the blackboard, X will be equal to 190.
This means that the last number, which is X + 1, is 191.
10. Its type is B. Consider the parities of the differences N (A) − N (B), N (B) − N (C), and N (C) − N (A), where
N (X) is the number of type X amoebae. These parities do not change in the course of the merging process. This
means, in particular, that in the end (when there is only one amoeba in the tube) the numbers of A-amoebae and
C-amoebae have the same parity, which is possible only if the only amoeba left belongs to type B.

71