Академический Документы
Профессиональный Документы
Культура Документы
K.E.Srivatsav
PSBB Numgambakkam
Anshul Ravichandar
Sri Sankara Senior Secondary
May 2012
ii
ISBN . . .
. . . Publications
Dedicated to the one who introduced us to real
math, our beloved teacher Mr.Sadagopan Rajesh,
and presented to him on his birthday (the 21st of
June 2012)
iv
Contents
Preface 1
Foreword 1 3
2 Problems in Algebra 11
3 Problems in NT 15
5 Solutions to NT problems 41
Appendix -1 62
References 78
vi CONTENTS
Preface
tion!!).
With regard to the appendix, we sincerely thank Ms. Uma
Girish and Ms. Shruthi.S (IMOTC Attendees of 2012) for
their invaluable assistance. In the appendix, we have included
some very useful geometric tools supplemented by some re-
ally tough problems (mostly from TSTs of countries) solved
trivially using the ideas.
Being our first book, we understand that some of our proofs
may be primitive and may contain some typos or errors. We
sincerely encourage the readers to contact us regarding editing
work or feedback. We hope you enjoy reading this book.
K.E.Srivatsav
(srivatsavke@gmail.com)
Anshul Ravichandar
(r.anshul@gmail.com)
Foreword by Dr.Balaji
Sampath
4 CONTENTS
Chapter 1
Strong Induction
Wilson’s Theorem
Number Of Divisors
k
Y k
Y
⌧( p↵i i ) = (↵i + 1)
i=1 i=1
Note that another important comes to play in this theorem.
It is that every natural number has a unique canonical factor-
ization of prime powers.
Bernoulli’s Inequality
Bezout’s Identity
Uniqueness Theorem I
Uniqueness Theorem II
Cauchy-Schwarz Inequality
If a > b > 0 are two coprime integers, then for any natural
number n > 1, there exists a prime number p such that p|(an
bn ) and does not divide ak bk for all positive integers k < n,
with the following exceptions-
• a = 2, b = 1 and n = 6; or
• a + b is a power of 2 and n = 2.
ap 1
⌘ 1(mod p) for all primes p and gcd(a, p) = 1
Legendre formula
Roots Of Unity
Dirichlet’s Theorem
Lagrange’s Theorem
Quadratic Reciprocity
✓ ◆
a2
Corollary 1.0.4 = 1, gcd(a, p) = 1
p
✓ ◆ ✓ ◆
a b
Corollary 1.0.5 a ⌘ b(mod p) ) =
p p
Problems in Algebra
mm+1 + nn+1
Problem 2.2 (USAMO) Let a = m + nn
where m and
m
n are positive integers. Prove that am + an mm + nn
✓ ◆2 ✓ ◆2
1 1 25
Problem 2.3 Prove that a + + b+ if a, b >
a b 2
0, a + b = 1
21n + 4
Problem 2.6 (IMO) Prove that the fraction is irre-
14n + 3
ducible (already in lowest form) for every natural number n.
12 Problems in Algebra
Problem 2.15 (RMO India 2011) Solve for reals x,y such
2 2
that, 16x +y + 16y +x = 1.
13
Problem 2.21 Prove that P (x) = x(n 1)t + x(n 2)t ... + 1 is
divisible by Q(x) = xn 1 + ... + 1 whenever gcd(n, t) = 1
5125 1
Problem 2.22 (IMO Shortlist)Prove that is a com-
525 1
posite number.
1 1
Problem 2.23 (IMO Shortlist) Prove that 1 + 3 + ... + 3 <
2 n
5
.
4
Problem 2.27 (India 2004) Given that x,y,z are positive re-
als, and that x + y + z = 1, let
Problems in NT
Problem 3.1 Find all pairs of natural numbers (n, k) for which
(n + 1)k = n! + 1
Problem 3.12 Find all the positive integers m such that the
number obtained by erasing the last digit is a divisor of m.
Problem 3.29 Prove that for all positive integers n the fol-
lowing is not an integer.
1 1 1 1
+ + + ... +
2 3 4 n
Problem 3.34 Find all positive a and b such that (a+b2 )(b+
a2 ) = 2m for some integer m.
gcd(m, n) n
Problem 3.37 (Putnam) Prove that the expression m
n
is an integer for all integers n m 1.
2p 1
1
Problem 3.39 Find all primes p for which is a per-
p
fect square.
Solutions to Algebra
problems
mm+1 + nn+1
Problem 4.2 (USAMO) Let a = where m and
mm +n nn m
m
n are positive integers. Prove that a + a m + nn
22 Solutions to Algebra problems
✓ ◆m
a m
Solution : If we apply Bernoulli’s inequality to 1 +
✓ ◆n ✓ ◆m ✓ m ◆n
a n m a m n a n
and 1 + , we get m 1+ +n 1 +
n m n
mm + nn + a(mn + nn ) (mm+1 + nn+1 ) = mm + nn
4x2 + 4z 2 = 4 (4.1)
2y 2 + 4y(x + z) = 12 (4.2)
lemma : p
acos✓ + bsin✓ a2 + b 2
26 Solutions to Algebra problems
Solution : Assuming that both the roots are real and lie
between 3 and 4, let them be 3 + a, 3 + b where 0 < a, b < 1.
b c
From Vieta’s relations, = 6+a+b and = 9+3(a+b)+ab.
a a
1
Using these two relations gives (24a+7b+2c) = 2ab (a+b)
p a
2ab 2 ab < 0. From this it is evident that a and 24a+7b+2c
are of di↵erent signs. Hence the assumption is wrong and both
roots cannot lie between 3 and 4.
|a1 1| = n1
28 Solutions to Algebra problems
|a2 2| = n2
|a3 3| = n3
|a4 4| = n4
..
.
2011
X 2010
X
[a2i +i2
2(iai )] = n2i (4.7)
i=1 i=0
P P2011 2
Note that 2011i=1 a 2
i is always going to be i=1 i in some dif-
ferent order, and similar is the sum of squares of ni . Here is the
place where we are using the invariance principle. Simplifying
the equation (7) we get,
2011
X 2011
X
20112 + i2 = 2(iai ) (4.8)
i=1 i=1
Clearly from (8), the RHS is even and the LHS is odd. Hence
we have arrived at a contradiction starting from a contrapos-
itive statement. Therefore, there will exist two such integers
such that |aj j| = |ak k|.
Problem 4.15 (RMO India 2011) Solve for reals x,y such
2 2
that, 16x +y + 16y +x = 1.
29
16x
2 +y
+ 16y
2 +x
p
16x2 +y .16y2 +x
2
1 2 +4x+4y 2 +4y 1
) (24x )2 (4.9)
2
(2x+1)2 +(2y+12 )
1
)1 2.(2 2 ) (4.10)
(2x+1)2 +(2y+12 )
)1 (2 2 ) (4.11)
Now, we have a power of two whose exponent is a positive real
(why?) and the power of two is 1. But this can happen
only at equality because, 2k where is a non negative real is an
increasing function of k whose minimal value is 1 when k = 0.
Which means,
) 2x + 1 = 2y + 1 = 0
Hence, x = y = 12 . is the only real solution satisfying the
above diophantine equation.
✓ ◆✓ ◆✓ ◆
1 1 1
than 3. 2a 2b 2c > 1 and 2a + 2b + 2c
✓ ◆ b c a
1 1 1
+ + > 3. Since abc = 1, the first inequality reduces
a b c ✓ ◆
1 1 1
to 2(a+b+c) + + < 3. But this is a contradiction.
a b c
Hence our assumption is false.
Lemma
Proof
The proof is very simple. Since P (x) has no real roots (why?)
Q(x) and R(x) too have no real roots. Therefore, if Q(ak ) is
31
x 4, y 5, z 6
x2 + y 2 + z 2 = 77 + (t1 + t2 )2 + (3 t1 )2 + (2 t2 )2 (4.15)
lemma
Proof
S2 = (t1 + t2 )2 + (3 t1 )2 + (2 t2 ) 2 (4.17)
Note that S1 is the case where t2 = 0. Using these two equa-
tions, if we prove that, for every possible value of t2 , S1 > S2 ,
we have proved that S is a decreasing function with respect to
t2 and hence attains maximality when t2 = 0.
Consider (17) (18). If we prove that the resulting value is
positive no matter what value t2 takes, we are done.
4 > 2(t2 + t2 )
Problem 4.21 Prove that P (x) = x(n 1)t + x(n 2)t ... + 1 is
divisible by Q(x) = xn 1 + ... + 1 whenever gcd(n, t) = 1
5125 1
Problem 4.22 (IMO Shortlist)Prove that is a com-
525 1
posite number.
1 1
Problem 4.23 (IMO Shortlist) Prove that 1 + 3 + ... + 3 <
2 n
5
.
4
35
1
Solution : Observe that n3 > (n 1)n(n+1) and =
✓ ◆ (n 1)(n)(n + 1)
1 1 2 1
+
2 n 1 n n+1
1 1 1
These two yield 1 + 3 + · · · + 3 < 1 + + ··· +
2 n 1⇥2⇥3
1
(n 1)(n)(n + 1)
✓ ◆ ✓ ◆
2 1 1 2 1 1
= 1 + + + + ··· +
2 3 2 3 4 (n 1)(n)(n + 1)
✓ ◆
1 1 1 1 1 5
=1+ + <1+ =
2 2 n n+1 4 4
Solution : Note that we must use the fact that y > x. Hence
the only values we can substitute for x and y individually will
be 0 and 1 respectively. In each case we get,
⇣y ⌘
f = af (y) (4.18)
2
✓ ◆
x+1
f = (1 a)f (x) + a (4.19)
2
Note that from 21, we can say that for some positive m less
than 12 , we can say that, f (m) = af (2m). Taking m = 17 in
the above equation, we get, f ( 17 ) = af ( 27 ) and further more,
f ( 27 ) = af ( 47 ), and hence we can say,
✓ ◆
4 f ( 17 )
f = 2 (4.20)
7 a
36 Solutions to Algebra problems
Problem 4.27 (India 2004) Given that x,y,z are positive re-
als, and that x + y + z = 1, let
Solution : At first sight this looks kind of easy and one might
start o↵ by saying why not start o↵ with algebraic manipula-
tion. Even I started like that but I found that it was taking me
nowhere. But it is actually an easy problem. If we reframe the
equation as c2 b2 = a2 3, we can make an observation. If a is
even, we know that the RHS is odd, and it can be represented
as di↵erence of two perfect squares. (n + 1)2 n2 = 2n + 1.
Here, from a2 3 = 2n + 1, we can find the value of n for
each a. Then we can substitute this in b and c. Hence we
get infinitely many solutions in this process. Very subtle and
simple.
Note that the number primes less than every number in the
set {(t + 2)! + 2, ..., (t + 2)! + t + 2} is the same because they
are t + 1 consecutive composite numbers. If we rewrite the
given condition as P (x) Q(x)⇡(x) = 0, then, obviously we
must have t + 1 distinct roots. Note that the degree of this
new polynomial P (x) Q(x)⇡(x) is deg P . If we set deg P = t,
then we get a violation of Fundamental Theorem of Algebra
- we must have at most t roots, but here we get t + 1 roots.
Hence there do not exist two such polynomials.
Solutions to NT problems
Problem 5.1 Find all pairs of natural numbers (n, k) for which
(n + 1)k = n! + 1
(x2 y 2 )2 1 = 16y
) (x2 y 2 + 1)(x2 y2 1) = 16y
(x2 y 2 + 1)(x2 y2 1)
) = 16 (5.1)
y
Now, we see that the numerator is a product of two consecutive
integers. Now we use the basic number theory property that if
a|bc and GCD(b, c) = 1 then, a|b or a|c. Here, we know the y
divides the numerator because the quotient is an integer. But
we are making an assumption that y 3. Now we get two
2 2 2 2
cases, y|(x y + 1) or y|(x y 1).
Note that, 16 = 4 ⇤ 4(or)8 ⇤ 2(or)16 ⇤ 1 and also the negatives
of each pairing. So we have six cases to check for. Also, y can
divide bothe the multiples in the numerator. Hence there are
totally 12 cases to check for.
After trials we arrive at the solutions (1,0) and (4,5). But
we need positive integers. Hence, x = 4, y = 5 are the only
solution to this diophantine.
Lemma
Proof
Using this let us see what remainder 229 leaves with respect
to division by 9. 23 ⌘ 1(mod9), (23 )9 ⌘ 1(mod9) and hence,
229 ⌘ 4(mod9).
Since it is given that 229 has nine distinct digits, the digit that
is absent will be the remainder when divided by 9. (why?)
Hence, 4 is the left out digit.
76 ⌘ 1(mod43)
Note that the residue class becomes cyclic with a period of 6.
Hence we can generalize the residue left by 7n .
61 ⌘ 6(mod43)
62 ⌘ 7(mod43)
63 ⌘ 1(mod43)
64 ⌘ 6(mod43)
65 ⌘ 7(mod43)
66 ⌘ 1(mod43)
Note that we get only a period of 3 in powers of 6. Since p is
a prime, 7p ⌘ 7, 1, 6(mod43) and 6p ⌘ 6, 1, 7(mod43).
Note that 7p 6p ⌘ 1(mod43) only when p ⌘ 1, 5(mod6).
But since p > 3, the middle possibility of p being 3mod(6) is
eliminated. Hence we are done.
7.101 ⌘ m1 (modn)
47
7.102 ⌘ m2 (modn)
7.103 ⌘ m3 (modn)
7.104 ⌘ m4 (modn)
..
.
Note that if we extend the above statements infinitely large
powers of 10, we will get a particular mk that is repetitive for
infinitely many LHSs’. This is true because there exists only
a finite possible distinct residues left with n.
Hence, we take that remainder mk and write A as the sum
of n such LHS forms that leave remainder mk . Hence we get
the collective remainder to be 0. The number A will look like
...770700700077.... Hence, we are done.
Problem 5.12 Find all the positive integers m such that the
number obtained by erasing the last digit is a divisor of m.
x2 (x2 + 1) = 7z y 2
(2n)!
Solution : 2n Cn = . From Chebyshev’s theorem, we
n!n!
know that there exists a prime p between 2n and n. Its next
multiple is obviously greater than 2n. Since the highest power
of p in the numerator is 1 and it does not occur in the de-
nominator, 2n Cn can never be a perfect power greater than
1.
(p 1)! ⌘ 1 (mod p)
p 1
) 1 2 (q1 q2 q3 ...q p 1 )2 ⌘ 1 (mod p)
2
We have hence proved our claim. Now let us see some exten-
sions to this statement. Note that it is interesting to see that,
it is possible to determine the exact remainder that the prod-
uct of the quadratic residues leave depending on the nature of
the prime. Let us see that.
Extensions : It is possible to make use of the Wilson’s theo-
rem to find the remainder left by the product. For any prime
p, we have,
(p 1)! ⌘ 1 (mod p)
Q Q
Now, ((p 1)!)2 = ( qi )2 ⌘ 1 (mod p), Hence, qi ⌘ ±1
(mod p). Infact, it is more interesting
Q Q to identify that, since
(p 1)! ⌘ 1 (mod p), Q we have qi nqi ⌘ 1 (mod p).
It simply follows that nqi ⌘ ⌥1 (mod p). Hence we have
proved the theorem. Another interesting investigation is to
find out the nature of the prime when the product of the
quadratic residues is 1 or -1. We leave it as a small challenge
to the readers.
Problem 5.29 Prove that for all positive integers n the fol-
lowing is not an integer.
1 1 1 1
+ + + ... +
2 3 4 n
Solution : Let us start with the canonical factorization of
n. Let n = pa11 pa22 pa33 .... Having this in mind, let us see the
expression at hand.
n!
1 1 1 1 + n!3 + n!4 + ... + n!
+ + + ... + = 2 n
(5.9)
2 3 4 n n!
Now, if the above expression is an integer, then the denomi-
nator must divide the numerator or rather, all divisors of the
denominator must divide the numerator. Let the largest power
of two less than n be 2k , and let the largest power of two di-
viding n! be 2m . Knowing that 2k < n, consider divisibility by
2m k+1 in the numerator.
Note that all the terms in the numerator except 2n!k are divisible
by 2m k+1 . Hence, all factors of the denominator do not divide
the numerator. Hence, we are done.
57
) k2 ⌘ 112 (modp)
Hence, ✓ ◆
112
=1
p
But using the product lemma, we can say that,
✓ 2◆✓ ◆
11 1
=1
p p
⇣ 2⌘
But by definition, 11p = 1. Hence,
✓ ◆
1
=1 (5.10)
p
a1 ⌘ a1 a2 (mod n) (5.11)
..
.
ak 1 ⌘ ak 1 ak (mod n) (5.12)
Note that from the above set of equations, we can conclude
that,
a1 ⌘ a1 a2 a3 ...ak (mod n) (5.13)
Let us assume that n divides ak (a1 1). Then, we have,
a1 ak ⌘ k (mod n)
) a1 a2 ...ak 1 ak ak ⌘ ak (mod n)
From 5.10 and 5.9, we can say that,
) a1 a2 ...ak 1 ak ⌘ ak (mod n)
) a1 ⌘ ak (mod n)
But from the given conditions, since, all ai are less than n
and are distinct, such a claim is not possible and is clearly a
contradiction to our assumption. Hence proved.
Problem 5.34 Find all positive a and b such that (a+b2 )(b+
a2 ) = 2m for some integer m.
gcd(m, n) n
Problem 5.37 (Putnam) Prove that the expression m
n
is an integer for all integers n m 1.
b
Solution : Let an 1 = kbn , k 2 N ) ab = (1 + kbn ) n .
kbn+1
Applying Bernoulli’s inequality gives ab 1+ . Since
n
kbn+1 3n
b 3, 1 + .
n n
2p 1
1
Problem 5.39 Find all primes p for which is a per-
p
fect square.
Geometric intuition
Geometric intuition is often the key to solving many Olympiad
problems. It involves making good guesses. Using appropriate
properties, these intuitive insights can be supplemented. Let
us see an Indian Longlist (2000) problem involving geometric
intuition. *
Figure 5.1:
So, we have covered the condition of equal sums using the in-
variant property of the circle, the radius. Now, let us see if we
can prove that AP, BQ, CR are concurrent. This part is quite
trivial using the converse of Ceva’s Theorem.
ar(4OP B)
BP
PC
= ar(4OP C)
= ar(4XP B)
ar(4XP C
= ar(4OBX)
ar(4OCX
[Componendo-Dividendo]
1 2
R .sin\XOB
= 21 R2 .sin\XOC = sin\XOB
sin\XOC
Similarly, CQ
QA
= sin\Y OC
sin\Y OA
AR
and RB =
2
sin\AOZ
sin\ZOB
\CAX \Y BC
But by simple angle chasing, \XOC = 2
= 2
=
\Y OC So, \XOC = \Y OC
Interesting observation:
See Figure1.2. P, Q, R also happen to be the points for which
the sums OP + P H, OQ + QH, OR + RH are minimum ! This
comes from a well known idea of reflection.
So in fact, if an ellipse is drawn with foci at O, H, that ellipse
is tangential to the sides.
Figure 5.2:
Power of point
Consider a circle at centre O radius 0 r0 and a point P and a
line through P cutting the circle at A and B. Let P D be a
tangent. Power of P with respect to this circle is given by
P A.P B. Using similar triangles it is possible to prove that
this quantity P A.P B is constant for a given point and circle
and does not depend on how the line is drawn.
P A.P B = P C.P D = P E 2
Constructions
Many complicated problems involve a simple construction. This
construction may not be immediately arrived at. However,
once the construction is made, the problem is half solved. Let
us take the example of an IMO (1995) problem.
Figure 5.3:
Solution :
How do we use the 120 condition? We need to make use of
the following property.
Figure 5.4:
Figure 5.5:
Consider 4P CE and 4P F B
P E = P B ; CE = F B ; \P EC = \P BF Therefore, 4P CE
and 4P F B are congruent. So, \CP E = \F P B = \EP
2
B
=
\CAB
2
72 Solutions to NT problems
Angle Chasing
However insignificant and trivial this may sound, the following
problem will leave you stunned. It is indeed, one of its kind! It
requires a brilliant construction and careful investigation of the
angles involved. The elegance of this solution is unsurpassed!
Figure 5.6:
74 Solutions to NT problems
Transformations
Homothety
AK DL AK 0
= =
BK LC BK 0
AK AK0
) BK = BK 0
Since K, K 0 lie on AB, they both must be the same point. So,
E, K, L are collinear. Now, we need to prove that CHGB
is cyclic. ) \GBC = \LHC , \DHL = \GBA Let
\GBA = ↵. Note that the sum of the green and red angles is
180
Reflection
A tranformation that preserves distances is reflection about a
line, which behaves like a mirror. The reflected image of a
point A about the mirror is the point A0 on a line through the
perpendicular from A to the mirror such that AA0 is bisected
by the mirror.
Note that this transformation preserves distances and angles.
This idea has also been used in the first section.