Mathematics Olympiad

Number Theory

Narasimhan R. Chari

om
References:

l.c
1) AoPS: Active Website for solutions to international math olympiad problems, https://artofproblemsolving.com,
(since 1993)
2) Alexander Bogomolny, www.cut-the-knot.org, (1996-onwards)

ai
3) Arthur Engel, Problem-Solving Srategies, Springer-Verlag, 1998
4) H. Eves, A Survey of Geometry, Allyn and Bacon, Inc., 1972
gm
5) H. S. M. Coxeter and S. L. Greitzer, Geometry Revisited, MAA, (Mathematical Association of
America, 1967
5) R. Graham, D. Knuth, O. Patashnik, Concrete Mathematics, Addison-Wesley, 1994
6) Roger A. Johnson, Advanced Euclidean Geometry, Dover, 1929
ri@

7) S. L. Loney, Plane Trigonometry, Cambridge, 1915)

8) Mactutor History of Mathematics, www.history.mcs.st-and.ac.uk, Univ of St. Andrews, Scotland
9) I. F. Sharygin, Problems in Plane Geometry, (1982), (free download available online):
ha

Main references for number theory:

1) G. H. Hardy and E. M. Wright, An Introduction to the Theory of Numbers, 1938, (Sixth Edition,
D. R. Heath-Brown, J. H. Silverman, eds, OUP, 2008)
sic

2) Ivan Niven, Herbert S. Zuckerman and Hugh L. Montgomery, An Introduction to the Theory of
Numbers, Wiley, 1960, (Fifth Edition, 1991)
ra

Euclid wrote for mature persons preparing for the study of philosophy. Geometry was the best intro-
duction to deductive reasoning.
na

H. S. M. Coxeter
 2

1 a) Division algorithm: Given integers a and b, we can divide b by a and get the
quotient q and remainder r ;
i.e., there exist unique integers q and r such that b = aq + r, with 0 ≤ r ≤ |a| − 1;
b and r leave the same remainder when divided by a; this is written as
i.e., b = r (mod a) or b ≡ r (mod a)
e.g., a = 7, b = 15 gives 15 = 2 × 7 + 1 ⇒ 15 ≡ 1 ( mod 7)

om
a = 7, b = −15 gives −15 = −3 × 7 + 6 ⇒ −15 ≡ 6 ( mod 7)
Euclidean algorithm: Given non-zero integers a and b, we can find their GCD d = (a, b) by
repeated division; there exist integers x and y such that d = xa + yb; d is the least positive in-

l.c
teger which can be expressed as an integer linear combination of a and b; such a representation
(Bézout’s identity, French, 1779) is not unique.
An integer can be expressed as a linear combination of two integers if and only if it is a multiple

ai
of their GCD.
Algorithm: Start with the first row, R1 : a = 1 × a + 0 × b and
gm
the second row, R2 : b = 0 × a + 1 × b; let a = bq + r as in the division algorithm;
use the row transformation R1 − q R2 to get R3 : r = a − qb = 1 × a − q × b;
Now repeat the process with (R1 , R2 ) replaced by (R2 , R3 ), and so on...till r = 0;
ri@

the last non-zero remainder is the gcd d of a and b and this row gives the linear combination
d = xa + yb.
Example: Find integers x and y such that 67x + 29y = 1
R1 : a = 67 = 1(a) + 0(b)
ha

R2 : b = 29 = 0(a) + 1(b
R3 = R1 − 2R2 : 67 − 2(29) = 9 = 1(a) − 2(b)
R4 = R2 − 3R3 : 29 − 3(9) = 2 = −3(a) + 7(b)
sic

R5 = R3 − 4R4 : 9 − 4(2) = 1 = 13(a) − 30(b) ⇒ x = 13, y = −30

R6 = R4 − 2R5 : 2 − 2(1) = 0 = −29(a) + 67(b)
ra
na

[email protected]
 3

1 b) (Inverses mod p): If p is a prime and i an integer such that 1 ≤ i ≤ p − 1, there exists a
unique j, 1 ≤ j ≤ p − 1, such that i j ≡ 1 (mod p); further, i = j only when i = 1 or i = p − 1.
Ans: Since i and p are coprime, i.e., gcd(i, p) = 1 the Euclidean algorithm gives xi + yp = 1
for some integers x and y. Going modulo p we get xi = 1 (mod p); take j = x (mod p).
If i is its own multiplicative inverse mod p, i.e., i = j , we get i j = i2 = 1 ⇒ i2 − 1 = 0 (mod p)
⇒ p | i2 − 1 = (i + 1)(i − 1) ⇒ p |(i + 1) or p |(i − 1) and 1 ≤ i ≤ p − 1 ⇒ i = 1 or i = p − 1

om
1 c) Addition, subtraction, multiplication and division (mod n)
Fix a positive integer n. For all integers x, y, a, b we have addition, subtraction and multiplica-
tion of residues modulo n , i.e.,

l.c
If x = a (mod n) and y = b (mod n) we have integers k, m such that x − a = kn, y − b = mn
⇒ x − a + y − b is a multiple of n ⇒ x + y = a + b (mod n); similarly x − y = a − b (mod n)
Again, x = a (mod n) and y = b (mod n) we have integers k, m such that x = a + kn ,

ai
y = b + mn ⇒ xy = (a + kn)(b + mn) = ab + amn + bkn + kmn2 = ab + ( a multiple of n)
⇒ xy = yx = ab (mod n) gm
To define division of integers modulo n we repeat the method we used to find inverses of non-
zero integers mod a prime p.

Suppose gcd (b, n) = 1. The Euclidean algorithm gives bx + ny = 1, for some integers x, y
1
ri@

Hence bx −1 is a multiple of n; i.e., bx = 1 (mod n); bx = 1 ⇒ b−1 = x; this defines (mod n)

b
If gcd (b, n) = d > 1 then b−1 does not exist, because bx = 1 (mod n) ⇒ bx − 1 = kn
⇒ bx − kn = 1; now d divides b and d also divides n; hence d divides bx − kn = 1; but d > 1,
contradiction.
 
a 1
ha

Conclusion: = a = ab−1 (mod n) is defined if and only if gcd(b, n) = 1

b b
sic
ra
na

[email protected]
 4

1 d) Divisibility: Give algorithms for testing the divisibility of a given natural number by any
given prime.
Ans: Every positive integer n has the usual decimal representation, i.e., it can be written in base
10 as
n = a0 10k + a1 10k−1 + . . . + ak



(1)
All the terms on the right side of this equation, except the last term ak , are multiples of 10.

om
Hence a positive integer n is divisible by 2 or 5 if and only if its last (units) digit is divisible
by 2 or 5 respectively.
The same argument with eq (1) proves that a number is divisible by 4 or 25 if and only only if

l.c
its last two digits give a number divisible by 4 or 25 respectively.
Similarly, a number is divisible by 2m or 5m if and only only if its last m digits form a number
which is divisible by 2m or 5m respectively.

ai
Going modulo 3 in eq (1), we get 10 = 1 (mod 3) ⇒ 10m = 1m = 1 (mod 3), hence
n = a0 + a1 + . . . + ak (mod 3)
gm
Conclusion: n is a multiple of 3 if and only if the sum of the digits of n is divisible by 3.
Since 10 = 1 (mod 9) the same argument can be repeated with 3 replaced by 9. Hence
n is a multiple of 9 if and only if the sum of the digits of n is divisible by 9.
ri@

Divisibility by 7: Let n = a0 10k + a1 10k−1 + . . . + ak be a positive integer. Then 7 divides

n if and only if 7 divides n − 21ak , i.e.,

7 divides a0 10k + a1 10k−1 + . . . − 20ak = 10 a0 10k−1 + a1 10k−2 + . . . − 2ak






⇔ 7 | a0 10k−1 + a1 10k−2 + . . . − 2ak .




ha

This has reduced the problem to one digit less than that of n. Hence we can repeat the process;
the process terminates after at most k − 1 steps.
sic
ra
na

[email protected]
 5

Example: Test whether n = 951496 is divisible by 7.

Ans: Method: 7 × 3 = 21 = 2 × 10 + 1; hence, multiply the last digit by 2 and subtract from the
rest of the number, omitting the last digit.
951496
− 12
95137; repeat the process;

om
− 14
9499
− 18
931
−2

l.c
91
This is a multiple of 7; hence n is also a multiple of 7.

ai
Divisibility by 13: Let n = a0 10k + a1 10k−1 + . . . + ak be a positive integer. Then 13

divides n if and only if 13 divides n + 39ak , i.e.,

gm
13 divides a0 10k + a1 10k−1 + . . . + 40ak = 10 a0 10k−1 + a1 10k−2 + . . . + 4ak






⇔ n | a0 10k−1 + a1 10k−2 + . . . + 4ak .





Again, this has reduced the problem to one digit less than that of n. Hence we can repeat the
process. Hence the process terminates after at most k − 1 steps.
ri@

Example: Show that n = 89572769 is divisible by 13.

Ans: Method: 13 × 3 = 39 = 4 × 10 − 1 . Hence, multiply the last digit by 4 and add it to the
rest of the number, omitting the last digit.
89572769
ha

+ 36
8957312; repeat the process;
+8
sic

895739
+ 36
89609
+ 36
ra

8996
+ 24
na

923
+ 12
104
+ 16
26
This is a multiple of 13; hence n is also a multiple of 13.

[email protected]
 6

Divisibility by any prime greater than 5: Let p be any prime number greater than 5. The
procedure used to test divisibility by 7 and 13 can be extended to test divisibility by p as fol-
lows:
The last digit of p must be one of 1, 3, 7, 9. Hence p or 3p must have last digit 9 or 1, i.e.,
p = 10m + 1 or p = 10m − 1 or 3p = 10m − 1 or 10m + 1.
Case 1: If p = 10m − 1 or 3p = 10m − 1, then multiply the last digit of the given k-digit posi-

om
tive integer n by m and add it to the rest of the number, omitting the last digit. Then repeat the
process. After at most k − 1 steps we get a 2-digit number; this is a multiple of p if and only if
the given number n is a multiple of p.
Case 2: If p = 10m + 1 or 3p = 10m + 1, then multiply the last digit of the given k-digit

l.c
positive integer n by m and subtract it from the rest of the number, omitting the last digit. Then
repeat the process. Again, after at most k − 1 steps we get a 2-digit number; this is a multiple
of p if and only if the given number n is a multiple of p.

ai
gm
ri@
ha
sic
ra
na

[email protected]
 7

1 1
2 Show that Hn = 1 + + . . . + is never an integer. (‘Harmonic numbers’).
2 n
[Hint: Take the lcm; there is a unique highest power of 2 which is less than or equal to n; the
corresponding numerator term is odd and all the others are even; hence the numerator is odd and
the denominator is even].

om
3 Show that 7744 is the only 4-digit number which is a perfect square having equal first two digits
and equal last two digits.

4 a) Show that an integer of the form 4n + 3 is never a sum of two squares.

l.c
Hint: Any integer leaves remainder 0 or 1 or 2 or 3, when divided by 4. Hence x = 0 or x = 1
or x = 2 or x = 3 mod 4. Hence x2 = 02 = 0 or 12 = 1 or 22 = 4 = 0 or 32 = 9 = 1 mod 4;
i.e., x2 = 0 or 1 mod 4.

ai
Similarly, y2 = 0 or 1 mod 4. Hence x2 +y2 = 0 +0 = 0 or 0 +1 = 1 or 1 +0 = 1 or 1 +1 = 2
mod 4. In all cases, x2 + y2 6= 3 mod 4. gm
4 b) Show that an integer of the form 8n + 7 cannot be written as the sum of three squares.
Generalise.
Hint: For any integer x we have x = 0 or ±1 or ±2 or ±3 or 4 mod 8. Hence
ri@

x2 = 0, (±1)2 = 1, (±2)2 = 4, (±3)2 = 9 = 1, (4)2 = 16 = 0 mod 8; x2 = 0 or 1 or 4 mod 8;

similarly for y2 and z2 ;
x2 + y2 + z2 = 0 + 0 + 0 = 0, 0 + 0 + 1 = 1, 0 + 1 + 1 = 2, 1 + 1 + 1 = 3,
0 + 2 + 2 = 4, 1 + 1 + 4 = 6, 0 + 1 + 4 = 5; in all cases x2 + y2 + z2 6= 7 mod 8.
ha

Following a) and b) the student may be tempted to make a general conjecture about integers of a
certain form which cannot be represented as a sum of four or more squares.
sic

But the famous four squares theorem of Lagrange (1770), says that every positive integer can
be written as a sum of four squares, (some of which may be zero).
We do not give the proof here.
ra

4 c) Show that the equation x2 + y2 = 3 z2 has no non-trivial integer solutions.

na

Equivalently, the equation x2 + y2 = 3 has no rational solutions.

In this problem, 3 can be replaced by any prime which is 3 mod 4.
√
Ans: Method 1: (Infinite descent mod 4). First note that 3 is irrational, hence none of the
variables can be zero. Suppose if possible there is a non-trivial integer solution, i.e., at least one
of (x, y, z) is non-zero. Changing the signs of the variables if necessary, we may assume x, y, z
are all positive. Start with a smallest positive solution, (in the sense |x| + |y| + |z| is the least
possible positive integer).
We know that, x2 = 0 or 1 mod 4; similarly for y2 and z2 . RHS is equal to 0 or 3; LHS is 0
or 1 or 2.
Hence the two sides can be equal only when they are both 0; i.e., when x = y = z = 0 (mod 4).

[email protected]
 8

Hence the re-scaling, (i.e., expansion or contraction, equivalently, mulltiplying a vector by a

positive constant),
x y z
, , gives a smaller positive solution of the given homogeneous equation, contradiction.
4 4 4
Method 2: (Infinite descent mod 3): To prove that the equation x2 + y2 = 3 z2 has no non-trivial
integer solutions.

om
Again, start with a smallest positive solution, (in the sense |x| + |y| + |z| is the least possible).
We know that x2 = 0 or 1 mod 3; similarly for y2 and z2 .
RHS is equal to 0 ; LHS is 0 or 1 or 2. Hence the two sides can be equal only when they are
both 0; this is possible only when x = y = 0 ( mod 3).

l.c
x = 3x1 , y = 3y1 ; this gives 9 x12 + y21 = 3z2 ⇒ z = 0 (mod 3);


x y z
hence x, y, z are all divisible by 3. Hence , , gives a smaller solution of the homoge-

ai
3 3 3
neous equation, contradiction.
gm
4 d) Show that the equation x2 + y2 = 3 (z2 + w2 ) has no non-trivial integer solutions.
Ans: (Infinite descent mod 3). We know that x2 is either 0 or 1 (mod 3); similarly y2 is either 0
or 1 (mod 3); hence, x2 +y2 = 0 (mod 3) only when x = y = 0 mod 3; substitute x = 3x1 ; y = 3y1
in the given problem and repeat the same argument.
ri@

9(x12 + y21 ) = 3(z2 + w2 ) ⇒ z = w = 0 (mod 3); hence z = 3z1 ; w = 3w1

x y z w
Hence , , , gives a smaller non-trivial solution of the homogeneous equation (1).
3 3 3 3
Any variable which was non-zero to begin with remains non-zero in the scaling procedure.
ha

Hence this process gives an infinite descending chain of positive integers, contradiction.

4 e) Show that the equation x2 + y2 + z2 = 7w2 has no non-trivial integer solutions.

sic

Equivalently, 7 cannot be written as the sum of three rational squares.

Ans: Infinite descent mod 4. We know that x2 is either 0 or 1 (mod 4); similarly y2 , z2 , w2 are
either 0 or 1 (mod 4); hence, x2 + y2 + z2 = 7w2 ⇒ x2 + y2 + z2 = 3w2 (mod 4); this is possible
only when x2 , y2 , z2 , w2 are all either 0 or 1 (mod 4); i.e., either all are odd or all are even.
ra

Case 1: If x, y, z, w are all odd, we get

na

x2 = (2x1 + 1)2 = 4x12 + 4x1 + 1 = 4x1 (x1 + 1) + 1 = 1 (mod 8), [because, either x1 is even or
x1 + 1 is even; hence x1 (x1 + 1) is even; so 4x1 (x1 + 1) is a multiple of 8]; similarly y2 , z2 , w2
are all equal to 1 (mod 8)
Hence the LHS is equal to 3 (mod 8), but the RHS is equal to 7 (mod 8). This is a contradiction.
Case 2: If x, y, z, w are all positive even integers, we use descent to get a contradiction,
i.e., substitute x = 2x1 , y = 2y1 , z = 2z1 , w = 2w1 in the given problem to get a smaller positive
solution.

[email protected]
 9

5 Show that the sum of three/four/five/six/seven/eight/nine/ten consecutive squares is never a square.

Ans: a) All the proofs are consequences of (mod 4) considerations.

a) For instance, consider (a − 1)2 + a2 + (a + 1)2 = 3a2 + 2 = b2
If a is even, LHS is 2 (mod 4), but RHS is 0 or 1 (mod 4); hence there are no solutions.
If a is odd, then b is odd; a2 = (2n + 1)2 = 4n(n + 1) + 1 = 1 (mod 8); LHS is 5 (mod 8), but

om
RHS is 1 (mod 8); again there are no solutions.

b) Consider a2 + (a + 1)2 + (a + 2)2 + (a + 3)2 = 4a2 + 12a + 14 = b2 ; LHS is 2 (mod 4); RHS
is either 0 or 1 (mod 4); no solutions.

l.c
c) Consider (a − 2)2 + (a − 1)2 + a2 + (a + 1)2 + (a + 2)2 = 5a2 + 10 (mod 4);
We know that a2 is either 0 or 1 (mod 4); hence 5a2 + 10 = 2 or 3 (mod 4);

ai
this number cannot be a perfect square because a square is either 0 or 1 (mod 4).
gm
d) Consider a sum of seven consecutive squares:
(a − 3)2 + (a − 2)2 + (a − 1)2 + a2 + (a + 1)2 + (a + 2)2 + (a + 3)2 = b2
7a2 + 28 = b2
ri@

If a is odd, LHS = 3 (mod 4) and RHS = 1 (mod 4), contradiction.

If a is even, a = 2a1 , b = 2b1 ⇒ 7a21 + 7 = b21 ;
b1 = 7b2 ⇒ a21 + 1 = 7b22 ⇒ 7 divides a21 + 1;
we will see later, (Problem 10), that the Euler-Fermat little theorem implies that the only primes
ha

which divide n2 + 1 are p = 2 or p = 1 (mod 4). Hence there are no solutions.

6 Show that 7744 is the only four digit number which is a perfect square having equal first two
sic

digits and equal last two digits.

Hint: Let N = aabb = k2 ; 1000a + 100a + 10b + b = k2 ⇒ 11(100a + b) = k2 ⇒ 100a + b

ra

100a + b
is a multiple of 11 and = m2 ; 1 ≤ a ≤ 9 and 0 ≤ b ≤ 9 (1)
11
na

Going mod 11, we get 100a + b = a + b (mod 11); hence a + b = 0 (mod 11); but

1 ≤ a + b ≤ 18 ⇒ a + b must be equal to 11.

Trying 1 ≤ a ≤ 9, we get a = 7, b = 4 as the only solution to eq (1); 7744 = 112 × 64 = 882

[email protected]
 10

7 a) Find a sum of 11 consecutive squares which is a square.

b) ‘Square pyramid problem’, Edouard Lucas, (French, 1875), (De Gang Ma, Chinese, 1985),
Anglin, (Amer Math Monthly, 1990), Baker and Davenport, (British, 1966),
‘Simultaneous solutions of 3x2 − 2 = y2 , 8x2 − 7 = z2 ’
Find n such that 12 + 22 + . . . + n2 is a square.

om
Ans: a)
(a − 5)2 + (a − 4)2 + (a − 3)2 + (a − 2)2 + (a − 1)2 + a2 + (a + 1)2 + (a + 2)2 + (a + 3)2
+ (a + 4)2 + (a + 5)2 = b2
11a2 + 110 = b2 ⇒ 11(a2 + 10) = b2 ⇒ b = 11b1 ⇒ a2 − 11b21 = −10 (1)

l.c
To solve such equations in integers we require some theory.

General Pell’s equation: x2 − dy2 = N.

ai
Given: x, y, d, N are integers and d is positive and square-free.
√
Let u = a + b d, a > 0, b > 0, be a solution of Pell’s equation a2 − db2 = 1.
gm
Then every integer solution of x2 − dy2 = N is given by the Pell multiples
√  √  √ k
xn + yn d = ±a ± b d x + y d , k ∈ Z
r
|N| u
ri@

2 2
p
where (x, y) is an integer solution of x − dy = N, with |x| ≤ |n| u ; |y| ≤
d
We do not give the proof here. (The theorem is not completely proved in Niven and Zuckerman
either). Refer to the book, ‘Introduction to Number Theory’, J E Shockley, (Holt, Reinhart and
Winston Publishers), New York, 1967 and online lecture notes, ‘Pell’s equation II’, Keith Con-
rad.
ha

For x2 − 11y2 =
√ −10 we have N = −10, |N| = 10, d = 11 and an obvious solution is (1, 1);
hence u = 1 + 11
The Pell equation x2 − 11y2 = 1 has an obvious solution 100 − 99 = 1 ⇒ (x, y) = (10, 3)
sic

s  √ 
10 10 + 3 11
r
|N| u
To apply the theorem consider the condition |y| ≤ = ⇒ |y| ≤ 4
d 11
√
ra

y = ±1 gives x = ±1, x2 − 11y2 = −10 ⇒ u = ±1 ± 11

There are no solutions for other values of y. Hence the general solution of x2 − 11y2 = −10 is
√  √ k
na


given by ±1 ± 11 10 + 3 11
We get all the (infinitely many) solutions of equation (1), i.e., sums of 11 consecutive squares
which are perfect squares, e.g.,
 √  √  √
−1 + 11 10 + 3 11 = 23 + 7 11 ⇒ (a, b1 ) = (23, 7); (a, b) = (23, 77);
182 + 192 + . . . + 282 = 772 ;
 √  √  √
1 + 11 10 + 3 11 = 43 + 13 11 ⇒ (a, b1 ) = (43, 13); (a, b) = (43, 143)
382 + 392 + . . . + 482 = 1432 ;
 √  √ 2 √
−1 + 11 10 + 3 11 = 461 + 139 11 ⇒ (a, b1 ) = (461, 139); (a, b) = (461, 1529)
4562 + 4572 + . . . + 4662 = 15292

[email protected]
 11

n(n + 1)(2n + 1)
b) Consider 12 + 22 + . . . + n2 = = b2
6
By trial and error, we get n = 24; this gives the smallest positive integral solution.
(24)(25)(49)
12 + 22 + . . . + 242 = = 702 .
6
In fact this is the only non-trivial solution in integers.

om
But the proof is not obvious. In this problem we have a smooth cubic Diophantine equation,
(much deeper than Pell’s equation above). (See the references at the beginning of the problem).

l.c
ai
gm
ri@
ha
sic
ra
na

[email protected]
 12

8 a) (Pythagorean triples) Show that every integer solution of the equation x2 + y2 = z2 is given
by x = k(u2 − v2 ), y = k(2uv), z = k(u2 + v2 ), where k, u, v are integers, u and v are of opposite
parity and gcd(u, v) = 1
x y
Hint: De-homogenise the equation, i.e., put X = and Y = to get the standard circle
z z
X 2 + Y 2 = 1. Every non-degenerate conic is rational, i.e., any conic with rational coefficients

om
which is not a product of two lines with conjugate irrational slopes can be parametrised by ratio-
nal functions, (i.e., a quotient of two polynomials in one variable).
To get a rational parametrisation of the circle, consider one rational point, say, P ≡ (−1, 0).

l.c
Consider all lines through P with rational slopes.
A general line intersects the circle in two points. For a quadratic equation with rational coeffi-
cients, if one root is rational then the sum of the roots formula says that the other root is also

ai
rational.
If the equation ax2 + bx + c = 0 has roots α and β then
gm
ax2 + bx + c = a(x − α)(x − β ), α + β = −
b
and α β =
c
(Vieta’s formulas).
a a
Hence, if one point is rational and the slope of the line is rational then the other point will also
be rational.
ri@

Conversely, every rational point Q on the circle will give a rational slope for the line PQ; hence
Q can be obtained as the intersection of the line PQ with the circle. This gives the parametrisa-
tion.
ha

Consider a general line with rational slope m passing through P, given by the equation

Y = m(X + 1). Substitute in the equation of the circle X 2 +Y 2 = 1. We get

sic

X 2 − 1 + m2 (X + 1)2 = 0; cancel X + 1 which corresponds to P; we get

1 − m2 2m
X − 1 + m2 (X + 1) = 0 ⇒ X = 2 . Then Y = m(X + 1) ⇒ Y = 2 .
m +1 m +1
1 − m2 2m
 
ra

(X, Y ) = , gives a rational parametrisation of the circle,

1 + m2 1 + m2
i.e., it gives all rational points on the circle as m varies over Q.
na

We have repeated the trigonometric formulas:

1 − t2
 
2 2 θ 2t
cos θ + sin θ = 1; put tan = t; cos θ = 2
; sin θ =
2 1+t 1 + t2
(Note that P is obtained as a limiting case when m → ∞).
x y u
Let X = , Y = , m = , where x, y, z, u, v are integers, gcd(u, v) = 1 Then
z z v
x v2 − u2 y 2uv
= 2 2
; = 2
z u +v z u + v2
Hence the general solution is given by x = k(v2 − u2 ), y = k(2uv), z = k(u2 + v2 ), for some inte-
gers k, u, v.
[email protected]
 13
8 b) Brahmagupta’s quadrilateral, Ujjain, 628 AD: Consider two Pythagorean triples

a2 + b2 = c2 and x2 + y2 = z2 . We construct a cyclic quadrilateral ABCD

with integer sides and diagonals such that the diagonals AC and BD are perpendicular to each
other.
Let P be the point of intersection of the diagonals.

Let DP = ax, PA = bx, AD = cx; CP = ay, PB = by, BC = cy,

i.e., ∆ APD and ∆ BPC on opposite sides of P

om
are obtained by magnifying the first Pythagorean triple
{a, b, c}, x times and y times respectively.
Let CD = az, AB = bz, i.e., ∆ DPC and ∆ APB

l.c
are obtained by magnifying the second Pythagorean triple
{x, y, z}, a times and b times respectively.
ay by
∠ PDC = θ ⇒ sin θ = = ⇒ ∠ BDC = ∠ BAC ; hence ABCD is a cyclic quadrilateral.

ai
az bz
We verify Ptolemy’s theorem for the Brahmagupta cyclic quadrilateral:
gm
AC. BD = (ax + by)(bx + ay)
= abx2 + aby2 + a2 xy + b2 xy
= ab(x2 + y2 ) + xy(a2 + b2 )
= abz2 + xyc2
ri@

= (az)(bz) + (cx)(cy)
= AB.CD + AD. BC
Example: Take {a, b, c} = {3, 4, 5} and {x, y, z} = {5, 12, 13}. We get
AB = 52, BC = 60, CD = 39, DA = 25, AP = 20, BP = 48, CP = 36, DP = 75
ha

c) Similar parametrisations can be used to find all triangles with integer sides having one angle
equal to 60◦ or 120◦ .
sic

If ∠ A = 60◦ , the cosine rule gives a2 = b2 + c2 − 2bc cos 60◦ = b2 − bc + c2

If ∠ A = 120◦ , a2 = b2 + c2 − 2bc cos 120◦ = b2 + bc + c2
ra

The last two parts of this problem are related because

x2 − xy + y2 = x(x − y) + y2 = (x − y)(x − y + y) + y2 = (x − y)2 + y(x − y) + y2

na

Repeating the method of part a) above, we find rational parametrisations for the equations
 2 √ !2
Y 3Y
X 2 ± XY +Y 2 = 1, which are ellipses X ± + = 1.
2 2
Start with the rational point P(−1, 0), consider a general line through P and find the other point
of intersection with the ellipse. We get the integer points

a2 = b2 − bc + c2 ⇒ a = u2 − uv + v2 , b = u2 − 2uv, c = u2 − v2 , u > 2v > 0

a2 = b2 + bc + c2 ⇒ a = u2 − uv + v2 , b = u2 − 2uv, c = 2uv − v2 , u > 2v > 0

The geometric relation between these formulas can be seen by combining these two parametri-
sations to construct a cyclic quadrilateral (similar to Brahmagupta’s construction).

[email protected]
 14

Draw an isosceles trapezium ABCD with AD = BC and a point E on the segment AB such that
AE = u2 − 2uv, EB = 2uv − v2 , AB = AE + EB = u2 − v2 , BC = u2 − 2uv, CD = 2uv − v2 ,
DA = u2 − 2uv

∆ AED is an equilateral triangle; ∠ A = ∠ B = 60◦ ; ∠C = ∠ D = 120◦ ; the diagonals are integers

given by AC = BD = u2 − uv + v2 . We can verify Ptolemy’s theorem:

2
AB.CD + AD. BC = (u2 − v2 ) 2uv − v2 + u2 − 2uv

om


2
= v4 − v2 u2 + 2uv + 2u3 v + u2 − uv − uv




2
= v4 − v2 u2 + 2uv + 2u3 v + u2 − uv − 2uv u2 − uv + u2 v2




2
2

l.c
= v4 + u2 − uv + 2u2 v2 − 2uv3 = v4 + u2 − uv + 2v2 u2 − uv



2
= v2 + u2 − uv
= AC. BD

ai
Example: u = 3, v = 1 gives ABCD with E on AB and AE = ED = AD = 3
EB = 5, BC = 3, CD = 5, AC = BD = 7
gm
8 d) Find all Pythagorean triangles whose legs are consecutive positive integers.
Ans: x2 + (x + 1)2 = y2 ⇒ 2x2 + 2x + 1 = y2 . (1)
ri@

This is a hyperbola (2x + 1)2 − 2y2 = −1. Intersect with a general line passing through the ra-
tional point (0, 1), i.e., y = mx + 1 (2)
2 − 2m 2m − m2 − 2
We get 2x2 + 2x + 1 = m2 x2 + 2mx + 1 ⇒ x = ⇒ y =
m2 − 2 m2 − 2
ha

u 2v2 − 2uv 2uv − u2 − 2v2

Put m = , to get x = 2 , y =
v u − 2v2 u2 − 2v2
 √ k √
The solutions of the Pell equation u2 − 2v2 = 1 are given by 1 + 2 = u + v 2, where k
sic

is an even natural number.

The corresponding Pythagorean triangle with consecutive integer legs is given by

ra

(x, x + 1, y) = 2uv − 2v2 − 1, 2uv − 2v2 , u2 − 2uv + 2v2

 √ k √
Similarly, the solutions of the Pell equation u2 − 2v2 = −1 are given by 1 + 2 = u + v 2,
na

where k is odd.
The corresponding Pythagorean triangle with consecutive integer sides is given by

(x, x + 1, y) = 2uv − 2v2 , 2uv − 2v2 + 1, u2 − 2uv + 2v2

Examples: k = 2, u = 3, v = 2 gives (3, 4, 5);

k = 3, u = 7, v = 5 gives (20, 21, 29); k = 4, u = 17, v = 12 gives (119, 120, 169)

If x = m2 − n2 , y = 2mn, z = m2 + n2 ; m1 = 2, n1 = 1 gives (3, 4, 5); the recurrence relation

mi+1 = 2mi + ni , ni+1 = mi generates all Pythagorean triangles with consecutive integer legs.

[email protected]
 15

8 e) Find all Pythagorean triangles having one side and hypotenuse as consecutive integers.
Ans: Let x = m2 − n2 , y = 2mn, z = m2 + n2 , and gcd(m, n) = 1, be the parametrisation of a
Pythagorean triangle and suppose a side and the hypotenuse are consecutive integers.
If z = x + 1 then m2 + n2 = m2 − n2 + 1 ⇒ 2n2 = 1 ⇒ n is irrational.
Hence we must have m2 + n2 = 2mn + 1 ⇒ (m − n)2 = 1 ⇒ m = n ± 1

om
Hence (x, y, z) = 2n + 1, 2n(n + 1), 2n2 + 2n + 1 , n ∈ N, is the general solution.

Examples:
n = 1 gives (x, y, z) = (3, 4, 5); n = 2 gives (x, y, z) = (5, 12, 13)

l.c
n = 3 gives (x, y, z) = (7, 24, 25); n = 4 gives (x, y, z) = (9, 40, 41)

ai
gm
ri@
ha
sic
ra
na

[email protected]
 16
9 a) Fermat’s little theorem: If p is a prime number and n is a natural number which is not a
multiple of p, we have n p−1 − 1 is a multiple of p.
If p is a prime number and n is a natural number, then n p − n is a multiple of p.
Ans: By induction on n. Assume n p − n is a multiple of p. Use the binomial theorem,
n      
n n−r r n n n−1 n n−2 2
n
(a + b) = ∑ a b =a + a b+ a b + . . . + bn
r=0 r 1 2
     

om
p p p p−1 p p−2 p
Consider (n + 1) − (n + 1) = n + n + n +...+ n+1−n−1
1 2 p−1
     
p p p−1 p p−2 p
= n −n+ n + n +...+ n
1 2 p−1

l.c
= n p − n+ (some multiple of p). Hence this expression is also a multiple of p.
 
p
Note: We have used the fact that is a multiple of p, for 1 ≤ r ≤ p − 1. This follows from
r

ai
two observations:
 
n
r  
gm
is the number of ways of selecting r objects out of n given objects);
n
Hence is an integer for all positive integers n and all r, 0 ≤ r ≤ n,
r
 
p
ri@

Hence is an integer. Further if p is prime and 1 ≤ r ≤ p − 1,

r
 
p p(p − 1)(p − 2) . . . (p − r + 1)
= ⇒ r! must divide p(p − 1)(p − 2) . . . (p − r + 1);
r r!
but r! has no common factor with p; hence r! must divide (p − 1)(p − 2) . . . (p − r + 1); hence
ha

 
p
is a multiple of p.
r
sic

b) Converse of Fermat’s little theorem: If n > 1 and an−1 − 1 is divisible by n, for all a which
are not divisible by n, then n is prime.
Proof: Suppose n is not prime. Take a, 1 < a < n, such that a divides n.
ra

By hypothesis, an−1 − 1 = kn . Then a divides both an−1 and n implies

a| an−1 − kn = 1 ⇒ a = 1, contradiction. Hence n is a prime number.


na

But there exist composite numbers n and positive integers a such that gcd (a, n) = 1 and
an − a is a multiple of n.
For example, 341 = 11 × 31 is not prime; 210 − 1 = 1023 = 3 × 341;
also an − bn is divisible by a − b, for all distinct integers a and b and all positive integers n.
Hence 2340 − 1 is a multiple of 210 − 1, which is a multiple of 341;
hence, 2341 − 2 = 2 2340 − 1 is also a multiple of 341.

There also exist numbers n (called Carmichael numbers), (Robert Carmichael, USA, 1910),
which are composite and satisfy the condition, an = a (mod n) for all integers a.

[email protected]
 17

Equivalently, an−1 − 1 is a multiple of n whenever a is an integer such that gcd(a, n) = 1.

Example: 561 = 3×11×17 and 1105 = 5×13×17 and the Hardy-Ramanujan taxi-cab number

1729 = 7 × 13 × 19 are Carmichael numbers, using Korselt’s criterion (1899):

n is a Carmichael number if and only if n is square-free and (p − 1) divides (n − 1), for all
primes p dividing n.

om
10 Euler-Fermat theorem: (Euler, 1736): Given a positive integer n, define
φ (n) = the number of integers in {1, 2, . . . , n} which are co-prime to n, i.e., integers which

l.c
do not have any common factor (greater than 1) with n. This function is called the Euler-phi
function.
Statement: If gcd(a, n) = 1 then aφ (n) ≡ 1 (mod n), where a ≡ b (mod n), or a = b (mod n)

ai
means a − b is divisible by n.
If n = p is prime, we get φ (p) = p − 1 ⇒ a p−1 = 1 (mod p) ⇒ a p = a (mod p). This is Fermat’s
little theorem.
gm
Proof: The possible remainders obtained when any integer is divided by n are given by
S = {0, 1, . . . , n − 1}. This, like {1, 2, . . . , n}, is called a ‘complete residue system’ mod n.
ri@

Let R = {x1 , x2 , . . . , xk } be the numbers in S which are co-prime to n.

By definition, k = φ (n); R is called a ‘reduced residue system’ mod n.
By unique factorisation, if a and b have no common factors with n, ab also has no common
factors with n. Hence, if x ∈ R and y ∈ R we get xy ∈ R, i.e., R is closed under multiplication
ha

mod n.
Let a ∈ R; consider aR = {ax1 , ax2 , . . . , axk }, all numbers being reduced modulo n.
sic

If axi ≡ ax j (mod n) then n| a(xi − x j ); but a and n have no common factors; hence n must
divide xi − x j , i.e., xi ≡ x j (mod n).
This says that multiplication by a gives a one-one mapping from the finite set R to itself; hence
ra

this map is also onto, i.e., R = aR; the elements of R and aR are permutations of each other.
Hence ax1 . ax2 . . . axk = x1 . x2 . . . xk (mod n); as before, we can cancel the product of all the
na

factors x1 . . . . . xk , (since they are all co-prime to n);

we get ak = 1 (mod n), i.e., aφ (n) ≡ 1 (mod n).

Example 1) For any natural number a, show that the number a2 + 1 has no factors of the form
4n + 3.
Ans: We have (4a + 1)(4b + 1) = 16ab + 4a + 4b + 1 = 4(4ab + a + b) + 1. Hence the product
of any two numbers of the form 4n + 1 is again of the form 4n + 1. Hence a number of the form
4n + 3 must have a prime factor of the form 4k + 3.
Suppose p is a prime factor of a2 + 1.
To prove that: either p = 2 or p = 4k + 1

[email protected]
 18

Suppose p > 2, i.e., p is an odd prime and p| a2 + 1 ⇒ a2 = −1 ( mod p);

Obviously p does not divide a; hence gcd(a, p) = 1; hence by Fermat’s little theorem
p−1 p−1
p−1 2


we have a = 1 (mod p) ⇒ a 2 = 1(mod p) ⇒ (−1) 2 = 1 (mod p)

p−1 p−1
⇒ is even; hence = 2k ⇒ p = 4k + 1.

om
2 2

Example 2) Show that the equation y2 = x3 + 7 has no integer solutions.

Ans: We have x ≥ 1, y ≥ 2

l.c
Case 1: Suppose x is even. x3 + 7 is odd; y is odd;

x = 2m ⇒ x3 + 7 = 8m3 + 7 = 7 ( mod 8) (1)

ai
y = 2n + 1 ⇒ y2 = (2n + 1)2 = 4n2 + 4n + 1 = 4n(n + 1) + 1 = 1 ( mod 8) (2)
Eq (1) and (2) contradict LHS=RHS. Hence this case is not possible.
gm
Case 2: Suppose x is odd. x3 + 7 is even; y is even.

y2 = x3 + 7 ⇒ y2 + 1 = x3 + 8 = (x + 2)(x2 − 2x + 4) ⇒ x2 − 2x + 4 divides y2 + 1
ri@

x2 − 2x + 4 = x2 − 2x + 1 + 3 = (x − 1)2 + 3; but x is odd; x − 1 is even; hence (x − 1)2 + 3 is

equal to 3 (mod 4)
Hence y2 + 1 has a factor of the form 4k + 3. This is impossible, by the previous problem.

The general equation y2 = x3 + k is called Mordell’s equation. Mordell (British, 1920), proved
ha

that this equation has only finitely many integral solutions, for each integer k 6= 0.
The study of integer and rational points on this smooth cubic curve involves deeper concepts in
sic

number theory and algebraic geometry.

Ref: Keith Conrad, ‘Examples of Mordell’s equation’, www.math.uconn.edu.mordell, 2008
ra
na

[email protected]
 19

11 Wilson’s theorem: If p is a prime, then (p − 1)! + 1 is divisible by p, (and conversely).

Proof: If n > 1 is composite, then n = ab, 1 ≤ a, b ≤ n − 1 ⇒ ab|(n − 1)! ⇒ n divides (n − 1)!,
hence n cannot divide (n − 1)! + 1.
If p is prime, consider S = {1, 2, . . . , p − 1}. Every integer a in S is co-prime to p; hence
gcd(a, p) = 1; hence, by the Euclidean algorithm, there exist integers x and y such that
ax + py = 1 ⇒ ax = 1 (mod p); ie., a has an inverse x (mod p).

om
In S, a = a−1 (mod p) if and only if a2 = 1 (mod p) if and only if

p|a2 − 1, i.e, [p divides a + 1 or a − 1] and a ∈ S

l.c
if and only if [ a = 1 or a = p − 1].
Hence every element ai in S has a distinct inverse a−1
i 6= ai except 1 and p − 1, which are
their own inverses.

ai
(p − 1)! = 1 × (p − 1) × ai × a−1
i = p − 1 = −1 (mod p).
gm
This proves Wilson’s theorem.
ri@
ha
sic
ra
na

[email protected]
 20

12 Chinese Remainder Theorem (Linear Diophantine equations) Given pairwise co-prime posi-
tive integers {n1 , . . . , nk } and positive integers {a1 , . . . , ak },
there exist an integer solution to the system of
simultaneous congruences x ≡ ai (mod ni ), for 1 ≤ i ≤ k.
Proof: For k = 2, we have the Euclidean algorithm: given co-prime integers n1 and n2 there
exist integers m1 and m2 such that m1 n1 + m2 n2 = 1. (1)

om
m1 and m2 are obtained from the penultimate convergent to the continued fraction expansion of
n1
n2
60 4 1 1 1 1
e.g., = 8+ = 8+ = 8+ = 8+ = 8+ ; omitting the last fraction

l.c
7 7 7 3 1 1
1+ 1+ 1+
4 4 4 1
1+
3 3
1 1 1 17

ai
gives the penultimate convergent 8 + = 8+ = .
3 1 2 2
1+
1 gm
Hence (60)(2) − (17)(7) = 1
Continuing from eq (1), given a1 and a2 , take x = a1 (m2 n2 ) + a2 (m1 n1 ) (2)
Then m1 n1 = 1 − m2 n2 ⇒ x = a1 (m2 n2 ) + a2 (1 − m2 n2 ) ⇒ x ≡ a2 (mod n2 )
ri@

Similarly, m2 n2 = 1 − m1 n1 ⇒ x = a1 (1 − m1 n1 ) + a2 (m1 n1 ) ⇒ x ≡ a1 (mod n1 )

This proves the Chinese remainder theorem for k = 2

By induction, assume we have found y such that y ≡ ai (mod ni ), for 1 ≤ i ≤ k − 1.
ha

Use the k = 2 case to solve simultaneously the congruences

x ≡ y (mod n1 n2 · · · nk−1 ) and x ≡ ak (mod nk ).

sic

This proves the general case.

N
We can also do it constructively as follows. Let N = n1 n2 · · · nk ; Ni = .
ni
ra

Given ni are pairwise co-prime; hence gcd (ni , Ni ) = 1.

The Euclidean algorithm gives integers mi and Mi such that
na

mi ni + Mi Ni = 1. Take
k
x = ∑ ai mi Ni
i=1

Then going mod ni , we get N j ≡ 0 (mod ni ), for j 6= i.

x ≡ ai Mi Ni (mod ni );
x ≡ ai (1 − mi ni ) (mod ni );
x ≡ ai (mod ni )
Hence x is a solution of the given system of congruences.

[email protected]
 21

Example (CRT): a) (Sun Zi Suanjing, Chinese, 237 AD, completed by Qin Jiushao, 1247 AD)
b) Find integers x and y such that 60x + 7y = 1
60 4 1 1 1 1 1
= 8+ = 8+ = 8+ = 8+ = 8+ ; omitting the last fraction
7 7 7 3 1 1 3
1+ 1+ 1+
4 4 4 1
1+
3 3

om
1 1 17
gives the penultimate convergent 8 + = 8+ = .
1 2 2
1+
1
We find (60)(2) + (7)(−17) = 1; hence x = 2 and y = −17.

l.c
In general, x = 2 + 7m and y = −17 − 60m, (where m is any integer), give all the integer solu-
tions of 60x + 7y = 1

ai
(CRT): c) Brahmagupta (628 AD, Ujjain, Kuttaka method of repeated division): A woman
had some eggs in her basket. When she counted them two at a time she had one egg left. The
same thing happened when she picked up the eggs 3 or 4 or 5 or 6 at a time. When she took out
gm
7 eggs at a time there were none left. Find the number of eggs in her basket.
Ans: x ≡ 1 (mod 2); x ≡ 1 (mod 3); x ≡ 1 (mod 4); x ≡ 1 (mod 5); x ≡ 1 (mod 6).
Take lcm (2, 3, 4, 5, 6) = 60; to solve x ≡ 1 (mod 60); x ≡ 0 (mod 7):
ri@

60 4 1 1 1 1
We find the continued fraction = 8+ = 8+ = 8+ = 8+ = 8+
7 7 7 3 1 1
1+ 1+ 1+
4 4 4 1
1+
60 1 1 1 3 3
i.e., = 8+ ; the penultimate (i.e., the second last) convergent is
7 1+ 1+ 3
ha

1 1 17
8+ =
1+ 1 2
sic

Hence we get 60 × 2 + 7 × −17 = 1; then the Chinese remainder theorem gives

x = (1)(−17)(7) + (0)(2)(60) = −119;

ra

x ≡ −119 (mod lcm (60, 7) = 420); x = −119 + 420 = 301; this is the least positive integer so-
lution.
na

There is a bigger solution given by Wilson’s theorem. For any prime p, we have p divides
(p − 1)! + 1; hence (7 − 1)! + 1 = 6! + 1 = 721 is divisible by 7; also 6! + 1 obviously leaves
remainder 1 when divided by 2 or 3 or 4 or 5 or 6. Note that 721 = −119 + 2(420) is the next
step in the previous method.

[email protected]
 22

13 Pell’s equation, (Examples of Brahmagupta, Fermat); (Proofs given by Brouncker, Lagrange,

Galois):
Consider positive integer solutions of the Diophantine equation x2 − dy2 = 1, where d is a pos-
itive square-free integer.
equation x2 − dy2 = 1 where d is a non-square,
General method: In general, to solve Pell’s √
consider the continued fraction expansion of d.

om
We have the following procedure (no proofs).
The continued fraction expansion is periodic (although the period is not an obvious function of
d ).

l.c
√ √
Let d = [a0 ; a1 , a2 , . . . a p ], with a0 = b dc, a p = 2a0 . Then a1 , a2 , . . . , a p−1 is a palindromic
sequence, i.e., it reads the same in the reverse order.

ai
If p is even, all the solutions
√ are precisely given by the n(p − 1) convergents in the continued
fraction expansion of d, for n ≥ 1 gm √
If p is odd, all the solutions are given by the 2n(p − 1) convergents to d, for n ≥ 1; in this
case suppose the first (smallest) integer solution, (corresponding to n = 1 ), is (x1 , y1 ). Then all
 √ n
further solutions are given by (xn , yn ) = x1 + y1 d .
ri@

Example 1: x2 − 2y2 = 1 has a minimal (fundamental) easy-to-see positive solution

√ √ √
(x1 , y1 ) = (3, 2), since 32 − 2(22 ) = 1. x2 + y2 2 = (3 + 2 2)2 = 17 + 12 2 gives (17, 12) as
the next larger solution, and so on.
√  √ n
All solutions are given by xn + yn 2 = x1 + y1 2 .
ha

√
The continued fraction expansion of 2 is periodic with period p = 2√and is given by taking the
1 2+1
integer part 1 , then the reciprocal of the fractional part √ = and then repeating
2−1
sic

2−1
the process, i.e.,
√ 1 1 1
2 = 1+ = 1+ . . . = [1; 2̄]
1 2+ 2+
2+
ra

1
2+
2+...
na

The period is p = 1. The alternating successive convergents give solutions of x2 − 2y2 = ±1.
1 3
1+ = ⇒ 32 − 2(22 ) = 1
2 2
1 7
1+ = ⇒ 72 − 2(52 ) = −1
1 5
2+
2
1 17
1+ = ⇒ 172 − 2(122 ) = 1 and so on...
1 12
2+
1
2+
2

[email protected]
 23
Example 2 (Pell’s equation): x2 − 7y2 = 1
√
7 has integer part 2; subtract the integer part and take the reciprocal.
√
1 7+2
√ = , which has integer part 1; subtract the integer part and take the reciprocal. (1)
7−2 3
√
7+1
We get which has integer part 1; subtract the integer part and take the reciprocal.
2
√
7+1

om
We get which has integer part 1; subtract the integer part and take the reciprocal.
3
√
We get 7 + 2 which has integer part 4; subtract the integer part and take the reciprocal.
1 √ 1 1 1 1
We get √ ; we are back to eq. (1). Hence 7 = 2+ . . . = [2; 1, 1, 1, 4]; p = 4
1+ 1+ 1+ 4+

l.c
7−2
p4n−1
The successive convergents which give solutions of x2 − 7y2 = 1 are
q4n−1
8

ai
[2, 1, 1, 1] = ⇒ 82 − 7(32 ) = 1;
3
127 √ √
[2, 1, 1, 1, 4, 1, 1, 1] = ⇒ 1272 − 7(482 ) = 1; verify that (8 + 3 7)2 = 127 + 48 7, etc.
48
gm
Example 3: History (Brahmagupta (628) and also Fermat (1637)). x2 − 61y2 = 1; d = 61
√ p10 29718
We have 61 = 7; 1, 4, 3, 1, 2, 2, 1, 3, 4, 1, 14 ; p = 11; the 10th convergent is
 
=
q10 3805
ri@

√  √ 2
The minimal positive integer solution is given by x + y d = p10 + q10 d

⇒ x = p210 + d q210 ; y = 2 p10 q10

⇒ x = (29718)2 + 61(3805)2 = 1766319049; y = 2(29718)(3805) = 226153980

ha
sic
ra
na

[email protected]
 24

14 History: a) (Euler, 1737) Show that e is irrational.

x2 −x x2
Hint: ex= 1 + x + + . . .; changing x to −x on both sides we get e = 1 − x + − . . ..
2! 2!
1 1 1 1
Suppose if possible e is rational; then 1/e is also rational. = 1 − 1 + − + − . . .
e 2! 3! 4!
1 p
Let = , q ≥ 1, gcd(p, q) = 1. Let Sn denote the sequence of partial sums of the alternating
e q

om
1 3 1 1 3−1 2 p 2 p 3
series. Then S2 = = ; S3 = − = = ; hence S3 < < S2 , i.e., < < ;
2! 3! 2! 3! 3! 3! q 3! q 3!
hence, q does not divide 3!

l.c
Similarly, 3!(S2 − S3 ) = 1; 4!(S3 − S4 ) = −1; n!(Sn−1 − Sn ) = (−1)n−1 ; this follows from
1
the definition of Sn as Sn = Sn−1 + (−1)n ;
n!

ai
hence, for every positive integer n, there exists a positive integer k such that
k p k+1
< <
n! q n!
gm
; hence q does not divide n!, for all n ≥ 3. This is a contradiction.

1
Hence is irrational; hence e is also irrational.
e
ri@

History: b) Show that π is irrational. [J. H. Lambert, German, 1761; Hermite, French, 1873;
Ivan Niven, Bulletin A.M.S., 1947; Miklós Laczkovich, Amer Math Monthly, 1997]

Ans: Method 1 (Niven’s proof): Let b be a non-negative integer. Consider the integral
Zπ
xn (π − x)n
ha

n
An (x) = b sin x dx
n!
0

Then An (0) = 0; An (b) > 0, for all positive integers b and n.......................................(1)
sic

π
The maximum value of x (π − x) in [0, π] is attained at x = ;
2
π  π  π2
x (π − x) ≤ π− = ;
2 2 4
ra

Zπ  2 n
n π 1
An (b) ≤ b sin x dx
na

4 n!
0
2n n
2 π2 b

n π
= 2b 2n =
2 n! n! 4

⇒ ∀ b ≥ 1 there exists n0 ∈ N such that An (b) < 1, ∀ n ≥ n0 ..................................(2)

a
Suppose if possible π is rational; let π = ; a = b π, where a, b are positive integers.
b
xn (a − bx)n xn (b π − bx)n bn xn (π − x)n
f (x) = = =
n! n! n!
f (x) is a polynomial of degree 2n with rational coefficients;
hence all derivatives of order 2n + 1 or higher are zero.

[email protected]
 25
Zπ
An (b) = f (x) sin x dx, use repeated integration by parts
0
h iπ
= f (x) (− cos x) − f 0 (x) (− sin x) + f 00 (x) (cos x) − . . . ± f (2n) (x) cos x
0
h iπ
= 0 + 0 − f 00 (π) − f 00 (0) + . . . ± f (2n) (x) cos x ..........................................(3)
0

Using the binomial theorem we get

om
xn (a − bx)n
f (x) =
n!

n n
∑k=0 k ak (−b)n−k x2n−k
=

l.c
n!
2n k
dk x
=∑ , where dk ∈ Z, dk = 0, for k < n
k=0 n!

ai

0 , if k < n
(k)
⇒ f (0) = k! dk
, if n ≤ k ≤ 2n

n!
gm
Hence f (k) (0) ∈ Z, for 0 ≤ k ≤ 2n............................................................................(4)
bn xn (π − x)n
Also, f (x) = ⇒ f (π − x) = f (x)
ri@

n!
Differentiating k times, we get (−1)k . f (k) (π − x) = f (k) (x), ∀ k ≥ 0
Put x = π; (−1)k . f (k) (0) = f (k) (π), ∀ k ≥ 0
Equation (4) gives f (k) (π) ∈ Z, for k ≥ 0
Equation (3) gives An (b) ∈ Z, for b, n ∈ N
ha

This contradicts (1) and (2). Hence π is irrational.

Method 2: (Lambert) To prove that π is irrational.

sic

x2 x4 x6 x3 x5
We know that cos x = 1 − + − + . . . and sin x = x − + − . . ..
2! 4! 6! 3! 5!
ra

Consider the infinite (hypergeometric) series

x x2
F(n, x) = 1 + + + . . ., where n is a non-negative integer and
na

1!(a + n) 2!(a + n)(a + n + 1)

a is a real number which is not 0 or a negative integer. This series is convergent for all x ∈ R.
For a = 1/2, we get F 0, −x2 /4 = cos x and x F 1, −x2 /4 = sin x

The xm term in F(n + 1, x) − F(n, x) is equal to

xm xm
−
m!(a + n + 1)(a + n + 2) · · · (a + n + m) m!(a + n)(a + n + 1) · · · (a + n + m − 1)
−xm
= .
(m − 1)!(a + n)(a + n + 1) · · · (a + n + m)

[email protected]
 26

x2
 
−x x
F(n + 1, x) − F(n, x) = 1+ + +...
(a + n)(a + n + 1) 1!(a + n + 2) 2!(a + n + 2)(a + n + 3)
−x
= F(n + 2, x)
(a + n)(a + n + 1)
F(n + 1, x) −x F(n + 2, x) F(n + 1, x)
−1 =
F(n, x) (a + n)(a + n + 1) F(n + 1, x) F(n, x)
F(n + 1, x)

om
Let G(n, x) = ; then
F(n, x)
−x
G(n, x) − 1 = G(n + 1, x) G(n, x)
(a + n)(a + n + 1)
1
⇒ G(n, x) = x

l.c
1+ G(n + 1, x)
(a + n)(a + n + 1)
Starting from n = 0 and repeating the process gives the continued fraction expansion

ai
1 x/{a(a + 1)} x/{(a + 1)(a + 2)}
G(0, x) = ···
1+ 1+ 1+
F(1, x) a x x
⇒ =
F(0, x) a+ a + 1+ a + 2+
···
gm
−x2 −x2
   
2 sin x
But for a = 1/2 and x = −x /4; we get F 0, = cos x; F 1, =
4 4 x
ri@

The previous equation becomes

x x2 x2 x2 x2 x x2 x2 x2 x2
tan x = · · · and tanh x = ···
1− 3− 5− 7− 9− 1+ 3+ 5+ 7+ 9+

Now, let us prove a condition for the irrationality of a continued fraction.

ha

b1 b2 b3
Theorem: Suppose x = . . . is an infinite continued fraction with positive integers
a1 − a2 − a3 −
an and bn ;
sic

suppose bn+1 ≤ an , for all n ≥ 1 and bn+1 < an , for infinitely many values of n. Then x is
irrational.
x1
Proof: Suppose x = , where x0 and x1 are positive integers. Then x < 1; x0 > x1 ;
ra

x0
x1 b1 a1 x1 − b1 x0
x= := ; hence y1 = < 1;
x0 a1 − y1 x1
na

now repeat the process with y1 , etc. We get a strictly decreasing sequence of positive integers
x0 > x1 > . . . This is a contradiction. Hence x is irrational.
π a
Finally, suppose if possible, π is rational; let = .
4 b
π
Put x = in the continued fraction expansion of tan x;
4
a/b a2 /b2 a2 /b2 a a2 a2
then 1 = ··· = · · · ; for large n, we have
1− 3− 5− b− 3b− 5b−
nb > a2 + 1; by the previous theorem, this continued fraction on the right side is irrational,
contradiction.

[email protected]
 27

15 If a, b, c are natural numbers and a|b3 , b|c3 and c|a3 , show that abc|(a + b + c)13 .
Hint: Let b3 = ak, c3 = bl and a3 = cm, where k, l, m are natural numbers.
(13)! p q r
Each monomial in the multinomial expansion of (a + b + c)13 is of the form a b c,
p!q!r!
where p, q, r are non-negative integers with p + q + r = 13.
If all of p, q, r are positive, we are done.

om
Assume p = 0. Then q+r = 13; 13 = 0+13 = 1+12 = 2+11 = 3+10 = 4+9 = 5+8 = 6+7.
Consider c13 = cc12 = cc3 c9 = c(bl)(b3 l 3 ) = c(bl)(akl 3 ), which is divisible by abc. Similarly,
bc12 = bc9 c3 = bb3 l 3 c3 = bakl 3 c3 , divisible by abc;

l.c
b2 c11 = b2 c2 c9 = b2 c2 (b3 l 3 ) = b2 c2 akl 3 , divisible by abc;
b3 c10 = bc9 c3 = bb3 l 3 c3 = bakl 3 c3 , divisible by abc;
b4 c9 = bb3 c9 = bakc9 , divisible by abc;

ai
b5 c8 = b2 b3 c8 = b2 akc8 , divisible by abc;
b6 c7 = b3 c3 c7 = akb3 c7 , divisible by abc. gm
This verifies all the possible cases.
Note that 13 is the smallest power of a + b + c which is divisible by abc, as seen from the fact
that c13 is divisible by abc in the previous calculation but not by any higher power of a or b
or c; in general c12 is not divisible by abc.
ri@

Example: a = 29 , b = 23 , c = 2; 29+3+1 = 213 does not divide 52212 .

16 a) Solve the equation 22x 23{x} = 11. 25x + 5. 22[x] , for rational values of x, where [x] or bxc
means the integer part of x, i.e., the greatest integer which is less than or equal to x and {x} is
ha

the fractional part of x, i.e., x = [x] + {x} = bxc + {x}, e.g., 3.1 = 3 + 0.1 ⇒ [3.1] = b3.1c = 3
and {3.1} = 0.1; −3.1 = −4 + 0.9 ⇒ [−3.1] = b−3.1c = −4 and {−3.1} = 0.9
Hint: let x = n + f , where n = [x] = bxc and f = {x}, 0 ≤ f < 1. If x is rational then f is also
sic

rational.
5.22n
Given 22n+3 f .23 f = 11.25 f + 5.22n ; hence 25 f = . The LHS is non-negative, hence
22n − 11
ra

5m
22n − 11 ≥ 0 ⇒ n ≥ 2; 0 ≤ f < 1 ⇒ 1 ≤ 25 f < 32; let 22n = m; 1 ≤ < 32. This gives
m − 11
352
m> ⇒ m > 13, i.e., n ≥ 2 and f is rational only when n = 2; hence f = 0.8
na

27

b) (Pre-RMO, 2012) Find all a ∈ R such that 3 < a < 4 and a (a − 3{a}) is an integer.
Hint: Let a = 3 + f , 0 < f < 1; bac = 3; {a} = f = a − bac; given, a (a − 3{a}) ∈ Z.
Hence (3 + f )(3 + f − 3 f ) ∈ Z; 2 f 2 + 3 f ∈ Z;
but 0 < f < 1 ⇒ 2 f 2 + 3 f < 5; hence

2 f 2 + 3 f is equal to 1 or 2 or 3 or 4; also 0 < f < 1; solving each quadratic separately, we get,

√ √ √
1 −3 + 17 −3 + 33 −3 + 41
f= , f= , f= ; f= ; a = 3+ f
2 4 4 4
this gives four solutions.

[email protected]
 jxk jxk 28
c) Find the number of non-negative integral solutions of the equation =
5 7
jxk jxk
Hint: If k = = , we get x ∈ {5k, 5k + 1, 5k + 2, 5k + 3, 5k + 4}
5 7
Also x ∈ {7k, 7k + 1, 7k + 2, 7k + 3, 7k + 4, 7k + 5, 7k + 6}
If k ≥ 3 we get 7k > 5k + 4; hence there is no solution. Hence k = 0 or 1 or 2.

x = 0, 1, 2, 3, 4, 7, 8, 9, 14

om
d) (Pre-RMO 2012) Find the sum of the squares of the real roots of the equation x2 − 7bxc + 5 =
0
Hint: If bxc ≤ 0, then each term on the left side is positive; hence there is no solution.

l.c
If bxc ≥ 8, then x ≥ bxc ≥ 8 ⇒ the left side is positive; again there is no solution.

bxc = 1, 2, 3, 4,√5, 6, 7. Of these the√

only possible solutions are

ai
bxc = 1 ⇒ x = √2; bxc = 4 ⇒ x = √ 23;
bxc = 5 ⇒ x = 30; bxc = 6 ⇒ x = 37. gm
The sum of the squares is 92.

17 If n2 + 3n + 51 is divisible by 13 for some natural number n,

ri@

show that 21n2 + 89n + 44 is divisible by 169.

Hint: Put n = 0, ±1, . . . , ±6 in the given expression and go mod 13. We get
n2 + 3n + 51 6= 0 ( mod 13), unless n = 5.
Then put n = 5 in the second expression;
ha

21n2 + 77n + 12n + 44 = (7n + 4)(3n + 11); 7n + 5 = 39 and 3n + 11 = 26;

both these terms are divisible by 13, hence the expression is divisible by 169.
sic

m
n
22 + 1 22 + 1 ,


18 (Bulg. Math. Olym. 2016) If m and n are positive integers and mn divides
show that
ra

(m, n) = (1, 1), (1, 5) or (5, 1)

Hint: First let us recall the fact that any two distinct Fermat numbers are coprime.
na

n+1
h n+1 i n
n n
Proof: Fn+1 = 22 + 1 = 22 − 1 + 2 = 22 + 1 22 − 1 + 2 = Fn 22 − 1 + 2;

now repeat this process; we get

Fn+1 = Fn Fn−1 · · · F1 F0 + 2. (1)
Suppose if possible, Fm , Fn are not coprime, where m < n. Let p > 1 be a common prime factor.
Now Fn is always odd; hence both Fm and Fn are odd; hence p must be odd; but eq (1) implies
Fm is one of the factors in the product Fn Fn−1 · · · F1 F0 ; hence p divides 2, contradiction.
Hence any two distinct Fermat numbers are coprime.
Now, let p − 1 = 2k with k < n.
(p − 1) divides 2k which divides 2n+1 ; but p divides 2 p−1 − 1 (by Fermat’s little theorem) and
k
2 p−1 − 1 divides 22 − 1
[email protected]
 29

k n−k n
which divides (22 )2 − 1 = 22 − 1, since x − 1 divides xm − 1
m n m n n n
Let p|mn, and mn|(22 + 1)(22 + 1) ⇒ p|(22 + 1)(22 + 1); if p|22 + 1, and also p|22 − 1,
which gives
m n n
p = 1 or p = 2. But 22 +1 and 22 +1 are both odd; hence p = 1; m = 1 or n = 1, 22 +1 = 5

19 Give an example of positive integers a and b such that a 6= b3 and b 6= a3 and ab + 1 divides

om
b4 + 1
Hint: a = 30, b = 8; 241 divides 84 + 1 = 4097 = 17 × 241

l.c
ai
gm
ri@
ha
sic
ra
na

[email protected]
 30

20 (Hermite’s identity, 1873) (Greatest integer, or integer part function, Legendre, 1798; square
bracket notation, Gauss, 1808; ‘floor function’, Kenneth Iverson, 1962). Let x be a real number
and let n be a positive integer. Then
   
1 n−1
bnxc = bxc + x + +...+ x+
n n
Hint: The integer part or greatest integer function is a non-decreasing function,

om
bxc ≤ x + 1n ≤ x + 2n ≤ . . .; we know that x = bxc + {x}; 0 ≤ {x} < 1;
   

if x is not an integer, −1 < {x} − n1 < 1; {x} − 1n is equal to −1 or zero. Hence

 
   
i n−1
x+ ≤ x+ , for 1 ≤ i ≤ n − 1
n n

l.c
 
1
= bxc + {x} + 1 −
n
 
1

ai
= bxc + 1 + {x} −
n
≤ bxc + 1
gm
Hence each term on the RHS is equal to bxc or bxc + 1

Case 1: If x is an integer then both the sides are equal to nx.

Case 2: If x is not an integer, 0 < {x} < 1. All the terms on the RHS form a non-decreasing
ri@

sequence; we also know that each of them is equal to either bxc or bxc + 1
1 i−1
Hence there exists i, 1 ≤ i ≤ n such that bxc = bxc + = . . . = bxc + and
n n
i i+1 n−1
bxc + = bxc + = . . . = bxc + = bxc + 1.
n n n
i−1 i
ha

Hence, {x} + < 1 and {x} + ≥ 1

n n
n{x} < n − i + 1 and n − i ≤ n{x}; add n bxc on both sides,
sic

n − i + n bxc ≤ nx < n bxc + n − i + 1

Hence LHS = bnxc = n bxc + n − i

ra

RHS = bxc + bxc + . . . (i times) + (bxc + 1) + (bxc + 1) + . . . (n − i) times

Hence RHS = n bxc + n − i; hence LHS = RHS

na

[email protected]
 31

21 (AIME, 1991) Let r be a real number. If

     
19 20 91
r+ + r+ +...+ r + = 546, find b100rc
100 100 100

Hint: From Hermite’s identity, we know that the terms on the left side form a non-decreasing
sequence and each term has only two possible values, brc or brc + 1;

om
546 35
there are 91 − 19 + 1 = 73 terms; also = 7+ ;
73 73
suppose the first i terms are equal to 7 and the remaining (73 − i) terms are equal to 8. Then
7i + 8(73 − i) = 546 ⇒ i = 38; counting 38 terms after 19 we get to 56; using Hermite’s iden-

l.c
tity, we get,
       
1 56 57 99
b100rc = brc + r + +...+ r + + r+ +...+ r +
100 100 100 100

ai
= 7(57) + 8(43) = 743.
gm
22 If x, y are real numbers, prove that b2xc + b2yc ≥ bx + yc + bxc + byc
   
1 1
Hint: By Hermite’s identity, LHS = bxc + x + + byc + y + ; to prove that
    2 2
1 1
ri@

bx + yc ≤ x + + y + . For this, x = m + f1 , y = n + f2 , where 0 ≤ fi < 1, for i = 1, 2;

2 2  
1 1 1
then x + y = m + n + f1 + f2 ; if f1 + f2 < 1, then either f1 < or f2 < ; hence x + =m
  2 2 2
1
or y + = n; hence both the sides are equal to m + n.
2
ha

In other cases (the left side) is equal to (the right side −1).

√
23 If f (x) = round( x), find
sic

1 1 1
+ +...+
f (1) f (2) f (2018)
ra

Hint: The round function gives the integer closest to a real number x; for example,
round(3.14) = 3; round(2.71) = 4; round(2.5) is usually taken as 3; the square root of a positive
na

integer can never be a half-integer, so there is no ambiguity in this problem.

√ 
f 2 = round(1.41...) = 1. Note that (1.5)2 = 2.25; (2.5)2 = 6.25, . . . , (4.5)2 = 20.25,
(44 × 45) = 1980, 2018 = 1980 + 38, etc. Hence f (1) = f (2) = 1;
f (3) = f (4) = f (5) = f (6) = 2; f (7) = f (8) = f (9) = f (10) = f (11) = f (12) = 3, and so on...
In general
1 √ 1
f (i) = n ⇐⇒ n − < i < n + ⇐⇒ n2 − n + 1 ≤ i ≤ n2 + n
2 2
Hence f (i) = n is constant on each such interval of length 2n. The required sum is equal to
       
1 1 1 1 38 38
2 +4 +6 + . . . + 88 + = 88
1 2 3 44 45 45

[email protected]
 32

 
1
24 a) Show that round (x) = x +
2
b) Show that round (x) + bxc = b2xc
a) round (x) = bxc, if {x} < 0.5, and
round (x) = bxc + 1, if {x} ≥ 0.5
x + 12 = bbxc + {x} + 0.5c
 
We know that x = bxc + {x}; hence

om
Case 1: If {x} < 0.5, then {x} + 0.5 < 1 ⇒ bbxc + {x} + 0.5c = bxc
Case 2: If {x} ≥ 0.5, then {x} + 0.5 ≥ 1 ⇒ bbxc + {x} + 0.5c = bxc + 1
 
1
b) By Hermite’s identity, b2xc = bxc + x + = bxc + round(x)

l.c
2
25 If a function f : N → N satisfies f (1) = 1, f (2) = 3, f (n + 2) = (n + 3) f (n + 1) − (n + 2) f (n),
prove that all the numbers of the sequence {3 f (n) , n ≥ 7}, leave the same remainder when di-

ai
vided by 2018.
Hint: The first seven values of the function are {1, 3, 9, 33, 153, 873, 5913}; 2018 = 2 × 1009
gm
is the prime factorisation of 2018; φ (2018) = 2018(1 − 1/2)(1 − 1/1009) = 1008; by Euler’s
theorem, 31008 ≡ 1 (mod 2018).
f (n) is a strictly increasing sequence, as can be seen by induction: assuming
f (n + 1) > f (n), we get f (n + 2) = (n + 2)[ f (n + 1) − f (n)] + f (n + 1) > f (n + 1).
ri@

To prove that 3 f (n+1)− f (n) ≡ 1(mod 2018), for n ≥ 7, by induction,

f (n + 2) − f (n + 1) = (n + 2)[ f (n + 1) − f (n)] ⇒ 3 f (n+2)− f (n+1) = [3 f (n+1)− f (n) ]n+2
3 f (n+2)− f (n+1) = (1)n+2 = 1 (mod 2018).
ha

Consider f (7) − f (6) = 5040 = 24 × 32 × 5 × 7 × 13, which is a multiple of 1008 = 24 × 32 × 7;

hence f (n + 2) − f (n + 1) is also a multiple of 1008, for n ≥ 6, (but not for n ≤ 5). This com-
pletes the induction.
sic

Note that in forward difference notation, we have ∆ f (n + 1) = (n + 2)∆ f (n), hence

∆ f (n + 1) = (n + 2)(n + 1) · · · = (n + 2)! ⇒ ∆ f (n) = (n + 1)! ⇒ f (n) = 1! + 2! + . . . + n!
ra
na

[email protected]
 33

a b c
26 a) Find positive integer solutions to the equation + + =2
b+c c+a a+b
a b c
b) Can you find positive integer solutions to the equation + + =4
b+c c+a a+b
a b c 3
Hint: Note that Nesbitt’s inequality for positive real numbers gives + + ≥
b+c c+a a+b 2
Proof: By AM-HM inequality, for positive real numbers x, y, z, we have

om
 
1 1 1
(x + y + z) + + ≥9
x y z  
1 1 1
Hence (a + b + b + c + c + a) + + ≥9
a+b b+c c+a

l.c
 
1 1 1 9
⇒ (a + b + c) + + ≥
a+b b+c c+a 2
           
a+b c a b+c b c+a 9

ai
⇒ + + + + + ≥
a+b a+b b+c b+c c+a c+a 2
a b c 9 3
⇒ + + ≥ −3 =
b+c c+a a+b 2 2
gm
a) Try a = b; we get c = 3a; hence (a, b, c) = (a, a, a), where a is any positive integer.
b) There are positive integer solutions but these points cannot be found by hand; the smallest
ri@

such solutions have around 80 digits. (Andrew Bremner, http://ami.ektf.hu, 2014)

If 4 is replaced by 178, the smallest positive integer solutions have ≈ 4 × 108 digits.

27 (Infinite descent, Vieta jumping) If a, b, c are positive integers and 0 < a2 + b2 − abc ≤ c, show
ha

that a2 + b2 − abc is a perfect square.

Hint: Assume, by symmetry, a ≥ b ≥ 1; a + b is minimal and 0 < a2 + b2 − abc = q ≤ c and q is
not a perfect square; hence a 6= bc; the monic quadratic equation a2 − abc + (b2 − q) = 0, with
sic

b2 − q 6= 0 and bc 6= a, has one integer root a; hence the other root β is also an integer. Given
−q ≥ −c,
bc + b2 − q ≥ b2 + bc − c = b2 + c(b − 1) ≥ 0; 0 = a2 − abc + b2 − q;
bc + b2 − q ≥ a2 − abc + b2 − q; hence bc(1 + a) ≥ a2 ; hence bc > a; otherwise bc ≤ a − 1 gives
ra

bc(a + 1) ≤ a2 − 1 < a2 ; the second root β = bc − a > 0 by Vieta’s formula for the sum and prod-
uct of the roots; aβ > 0; 0 < aβ = b2 − q < b2 ; hence a ≥ b and aβ < b2 gives β < b; but (β , b)
is a solution in positive integers of the same quadratic and by minimality we get β + b ≥ a + b;
na

hence β ≥ a ≥ b; this a contradiction.

[email protected]
 34

28 (Crux mathematicorum, R. K. Guy, J. Nowakowski, Prob. 1746) Find infinitely many pairs of
integers (a, b), 1 < a < b such that ab divides a2 + b2 − 1
Hint: (1, 2) is a solution; b = a + 1 gives a2 + (a + 1)2 − 1 = 2a(a + 1)
Let a2 + b2 − 1
=k (1)
ab
⇒ a2 − kab + (b2 − 1) = 0 (2)

om
Eq. 1 has (a, b) = (1, k) as a solution; consider eq. 2 as a quadratic in a; let α = a and β be the
roots. Then by Vieta’s formulas, α + β = kb and αβ = b2 − 1; β = kb − α; hence (β , b) also
satisfies eq. 2.

l.c
If (a, b) = (1, k) then (β , b) = (kb − a, b) = (k2 − 1, k) is also an integer solution. Hence k can
take all possible integer values. (x0 , y0 ) = (1, k) or (k, 1) and (xn+1 , yn+1 ) = (yn , kyn − xn ) gives
infinitely many integer solutions for each fixed value of k.

ai
For k = 1, this equation is x2 − xy + y2 = 1 which represents an ellipse; hence it will have only
finitely many integer points; in this case the integer solutions are
(x, y) = (1, 0), (0, 1), (−1, 0), (0, −1)
gm
For k = 2, we have x2 − 2xy + y2 = 1 which is a pair of lines x − y = 1 and x − y = −1; hence
the integer points are (n, n + 1) and (n + 1, n)
For k ≥ 3, the equation x2 − kxy + y2 = 1 is a hyperbola, i.e., the equation is Pell’s equation of
the form x2 − dy2 = 1. It is symmetric around the 45◦ line through the origin, y = x, (since the
ri@

equation is symmetric with respect to x and y).

√
3± 5
For k = 3, the asymptotes are y = mx where m = , both positive real numbers. The point
2
(1, 3) is on the top branch of the hyperbola x2 − 3xy + y2 = 1 above the line y = x; the next point
(3, 8) is on the bottom branch below the line y = x; the next point (8, 21) is on the top branch,
ha

etc. In general, xn+1 = yn ,

yn+1 = kyn − xn ; hence xn < yn , k > 1 ⇒ (k − 1)yn > xn ; xn+1 > yn+1 . Hence this method is
sic

called ‘Vieta jumping’. Tracing the Vieta jumping backwards we get all the integer solutions
starting from (1, k) or (k, 1).

29 (1992, Westinghouse Science Talent Search, Kiran Kedlaya, Math. Magazine, Vol. 71, No. 1,
ra

(Feb. 1998), pp 61-73)

x, y, z are positive integers. Show that the product (xy + 1)(yz + 1)(zx + 1) is a perfect square if
na

and only if each of the factors xy + 1, yz + 1 and zx + 1 is a perfect square.

Ans: Let t be a positive integer. A set of positive integers is called a Pt −set if the product of
any two elements of the set is t less than a perfect square. For example, (t + 1)2 = t 2 + 2t + 1
and (t + 2)2 = t 2 + 4t + 4 gives the Pt 2 −set {1, 2t + 1, 4t + 4}.
{1, 3, 8, 120} is a P1 −set; {1, 33, 68, 105} is a P256 -set, i.e., 1 × 33 + 256 = 172 ,
1 × 68 + 256 = 182 ; 1 × 105 + 256 = 862 , 33 × 68 + 256 = 502 , 33 × 105 + 256 = 612 ,

68 × 105 + 256 = 862 .

[email protected]
 35

If {p, q, r} is a P1 −set, let

p
s = p + q + r + 2pqr ± 2 (pq + 1)(qr + 1)(rp + 1)

Assume s > 0 . Let pq + 1 = u2 , pr + 1 = v2 , qr + 1 = w2 . Then

ps + 1 = p2 + pq + pr + 2p2 qr + 2puvw + 1 = p2 + p2 qr + pq + pr + p2 qr + 1 + 2puvw

om
= p2 (qr + 1) + (pq + 1)(pr + 1) + 2puvw = p2 w2 + u2 v2 + 2puvw = (pw + uv)2

Hence {p, q, r, s} is also a P1 −set.

The definition of s is equation is equivalent to

l.c
4p2 + q2 + r2 + s2 − 2(pq + pr + ps + qr + qs + rs) − 4pqrs − 4 = 0 , or (p + q − r − s)2 =
4(pq + 1)(rs + 1) or (p + r − q − s)2 = 4(pr + 1)(qs + 1) or

ai
(p + s − q − r)2 = 4(ps + 1)(qr + 1)
For proving the main result, assume without loss of generality, p ≤ q ≤ r, atleast one of the
gm
factors pq + 1, pr + 1, qr + 1 is not a perfect square but (pq + 1)(qr + 1)(rp + 1) is a perfect
square. Take a solution (p, q, r) such that p + q + r is minimum.
p
Let s = p +q +r +2pqr −2 (pq + 1)(qr + 1)(rp + 1). To prove that 0 < s < r and the product
(pq + 1)(qr + 1)(rp + 1) is a perfect square but atleast one of the factors is not a perfect square.
ri@

Then p + q + s < p + q + r and (p, q, s) is also a solution. This will contradict the minimality
of (p, q, r).
Multiply two of the equivalent forms of s; we get
16(pq + 1)2 (pr + 1)(qr + 1)(ps + 1)(qs + 1) = (pq + 1)2 (p + r − q − s)2 (p + s − q − r)2 , which
is a perfect square; hence the LHS must be a square; also given, (pq + 1)(qr + 1)(rp + 1) is a
ha

square; hence the remaining part (pq + 1)(ps + 1)(qs + 1) must be a square. But the equivalent
forms of s imply that ps + 1 is a square if and only if qr + 1 is a square and qs + 1 is a square
if and only if pr + 1 is a square; hence not all of pq + 1, pr + 1 , qr + 1 are squares. (Otherwise
sic

each of them would be a square). Also

(p + q − r − s)2 −1
= rs + 1 ≥ 0 ⇒ s ≥
4(pq + 1) r
ra

If r = 1 and p ≤ q ≤ r , then p = q = r = 1 ⇒ (pq + 1)(qr + 1)(rp + 1) = 8 , which is not a

square, contradiction. Hence r > 1 and s ≥ 0 , (since s ≥ −1 r > −1 and s is an integer). If
na

s = 0 we get (p + q − r) = 4(pq + 1), (q + r − p) = 4(qr + 1), (r + p − q)2 = 4(pr + 1) , hence

2 2

pq + 1, qr + 1, rp + 1 are all squares, contradiction. Hence s > 0. Let s0 be the other root of the
quadratic equation, i.e., s0 is the conjugate of s , i.e.,
s0 = p + q + r + 2pqr + 2 (pq + 1)(qr + 1)(rp + 1); the product of the roots is given by
p

ss0 = p2 + q2 + r2 − 2pq − 2qr − 2pr − 4 < r2 − p(2r − p) − q(2r − q)

But p ≤ q ≤ r ⇒ p < 2r, q < 2r → 2r − p > 0, 2r − q > 0;
hence s < s0 , s2 < ss0 < r2 ⇒ s < r ⇒ p + q + s < p + q + r, contradicting the minimality of
p + q + r.
The proof can be repeated for t = 4, P4 − sets, with the equivalent form
4(pq + t)(rs + t) = t(p + q − r − s)2

[email protected]
 36

30 a) Show that a2 + b2 = 3ab + 3 has no integer solutions.

9b2 9b2
 
2 2 2
Hint: a − 3ab + b = 3 ⇒ a − 3ab + + b −2 =3
4 4
⇒ 4a2 − 12ab + 9b2 − 5b2 = 12 ⇒ (2a − 3b)2 = 5b2 + 12
5b2 is a multiple of 5 ; hence its last digit is either 0 or 5; hence the last digit of the right side is
either 2 or 7. But the left side is a perfect square; hence its last digit can be neither 2 nor 7.

om
a2 + b2
b) (IMO, 1988) If a, b are positive integers and is a positive integer k, show that k
ab + 1
must be a perfect square.
Hint: (Stan Dolan) Fix k; assume (A, B) is a solution of the equation A2 + B2 = k(AB + 1) such

l.c
that min{A, B} is the least posssible.
Without loss of generality, we may assume that B ≤ A.

ai
The quadratic equation A2 − kAB + (B2 − k) = 0 has one root A. Suppose the other root is C.
By Viete’s formulas for the sum and product of the roots, A +C = kB and AC = B2 − k.
gm
B2 k B2
Hence C = − < ≤ B; hence min{C, B} = C < B = min{A, B} and
A A A
(C, B) is a solution of the same equation.
Reason: Put A = C (since C is a root of the sme equation), in A2 − kAB + (B2 − k) = 0 ⇒
ri@

C2 − kCB + B2 − k = 0 ⇒ C2 + B2 = k(CB + 1)

C2 + B2
⇒ =k
CB + 1
Further, C = kB − A is an integer and
ha

(A+1)(C +1) = AC +A+C +1 = B2 −k +kB+1 = B2 +k(B−1)+1 > 0 ⇒ C +1 > 0 ⇒ C ≥ 0.

This contradicts the minimality of the solution (A, B) , unless C = 0; hence k = B2 .
sic

For example, k = 4 , we have (a − 2b)2 − 3b2 = 4 and Vieta jumping gives the integer points
(a, b) = (0, 2), (8, 2), (30, 8), (112, 30), etc.
x √
All these points (x, y) are on one branch of the hyperbola whose asymptote is lim = 2+ 3
ra

31 If k is a positive integer, show that 3k can be written as x2 + 2y2 , with gcd(x, y) not a multiple
na

of 3.
Hint: Start with 3 = 12 + 2(1)2 . Then use Brahmagupta’s identity, (650 AD)
(a2 + nb2 )(c2 + nd 2 ) = (ac − nbd)2 + n(ad + bc)2 = (ac + nbd)2 + n(ad − bc)2
By induction on n, 3n = a2 + 2b2 and 3 = c2 + 2d 2 implies 3n+1 = x2 + 2y2

[email protected]
 37

32 a) Prove that n3 ≡ 0, ±1 ( mod 9), for all positive integers.

Hint: Modulo 9, we have (±1)3 = ±1; (±2)3 = ±8 = ∓1; (±3)3 = ±27 = 0;
(±4)3 = ±64 = ±1

b) Deduce that, if k ≡ ±4 ( mod 9), there are no integer solutions of the equation k = x3 + y3 + z3

c) Verify the identities:

om

3
3
3
i) 9t 4 + 3t − 9t 4 + 1 − 9t 3 = 1

3
3
3
ii) −6t 2 + 1 + 6t 3 + 1 − 6t 3 = 2
When t varies over all integers, these formulas give some, but not all, representations of 1 and

l.c
2 as a sum/difference of three integer cubes. For example, as discovered by computer search,
2 = 12149283 + 34802053 − 35288753 is not given by this paramatrisation.

d) The equation 74 = a3 + b3 − c3 , has integer solutions, where a, b, c are 14 or 15 digit num-

ai
bers. (Found by computer search in 2016)
gm
e) (Choudhry, 1998): All integral solutions of the equation x3 + y3 + z3 = w3 are given by

2
x = c c3 − a3 − b3 , y =c3 (a + b) − a2 − ab + b2 , z = c3 (2a − b) + a2 − ab + b2 ,


2

w = c a3 + c3 + (a − b)3 , where a > b and c3 > a3 + b3

For example, put b = 1, a = 2, c = 3 to get 543 + 723 + 903 = 1083 ; then cancel 183 on both
ri@

sides to get 33 + 43 + 53 = 63 .

f) Ramanujan’s taxicab number given in 1911, (already stated by Frenicle de Bessy, 1657)
1729 = 93 + 103 = 13 + 123 is a particular case of the identity c, i) above: put t = −1.
The general parametrisation is given by Euler’s formula (1753), simplified by Binet (1841):
ha

Every rational solution of the equation x3 + y3 = z3 + w3 is proportional to

x = 1 − (a − 3b) a2 + 3b2 , y = −1 + (a + 3b) a2 + 3b2 ,



sic

2
2
z = (a + 3b) − a2 + 3b2 , w = −(a − 3b) + a2 + 3b2 , where a, b are rational.
For instance, a = 3b, t = −2b gives the formula in part c, i) above.
ra

g) Euler gave the example 594 + 1584 = 1334 + 1344

Euler’s sum of powers conjecture that a k−th power cannot be written as a sum of less than k
na

k−th powers is false.

(Lander, Parkin, Selfridge, 1967): 275 + 845 + 1105 + 1335 = 1445
(Noam Elkies, 1986): 26824404 + 153656394 + 187967604 = 206156734
The first example of a fourth power which is a sum of four fourth powers is
304 + 1204 + 2724 + 3154 = 3534 , (R Norrie, 1911)
EEC (Euler’s extended conjecture) states: The diophantine equation
x1k + x2k + . . . + xm
k
= yk1 + yk2 + . . . + ykn has no integral solutions for k > m + n

except for m = n, xi = yi , for all i

[email protected]
 38

The particular case m + n = 3 states that xk + yk = zk , for k > 3 has no solutions in positive
integers for k > 3; this follows from Fermat’s last theorem, (FLT), that xn + yn = zn has no
solutions in positive integers for n ≥ 3.
Fermat (1637) himself proved FLT for n = 4; Euler proved the case n = 3; the case n = 5 was
solved by Sophie Germain, Dirichlet and Legendre in 1825. The complete solution was given
by Andrew Wiles in 1994.

om
h) (Open problem, 2018, youtube, ’Numberphile’): Can 33 be expressed as a sum/difference
of three cubes?

l.c
ai
gm
ri@
ha
sic
ra
na

[email protected]
 39

33 History: a) Euler, ‘The Basel problem’ (1735), made rigorous by Cauchy (1821):
∞
1 π2
∑ =
n=1 n2 6

Ans: Assume the Maclaurin series formula (Scotland, 1742), or Taylor series (England, 1715):
f 0 (0) f 00 (0) 2 f (n) (0)
f (x) = f (0) + x+ x + + Rn . In particular,
1! 2! n!

om
x2 x3
ex = 1 + x + + + . . .
2! 3!
x 3 x5 x2 x4 x6
sin x = x − + − . . . ; cos x = 1 − + − + . . ., (Madhava, Kerala, India, around 1400)

l.c
3! 5! 2! 4! 6!
Change x to ix, where i2 = −1; using i3 = −i, i4 = 1, i5 = i, etc., we get

i2 x2 i3 x3

ai
eix = 1 + ix + + +...
2! 3!
x2 x4 x6 x3 x5
   
= 1− + − +... +i x− + −...
2! 4! 6! 3! 5!
gm
Hence eix = cos x + i sin x (Euler’s formula)

eix cos x + i sin x

ri@

Hence, = = cot x + i; raise both sides to the power n; hence

sin x sin x
einx n inx = cos nx + i sin nx
n = (cot x + i) ; again use Euler’s formula, e
sin x
cos nx + i sin nx
= cotn x + n1 cotn−1 x i + n2 cotn−2 x i2 + . . .





ha

n
sin x
Equating imaginary parts on both sides we get,
sic

sin nx n
n−1 x − n cotn−3 x + n cotn−5 x + . . .

= 1 cot
sinn x 3 5

rπ
Put n = 2m + 1; x = xr = , 1 ≤ r ≤ m;
ra

2m + 1

sin(rπ) 2m+1

cot2m x − 2m+1

2m−2
= cot x+...
na

sinn x 1 3

2m+1

cot2m x − 2m+1

2m−2
0= 1 3 cot x+...

π
Let {xr } be the distinct roots of this equation, 0 < xr < ; then cot2 xr are also distinct.
2
2m+1

− 2m(2m − 1)
The sum of the roots is ∑αi = ∑ cot2 xr = − 3
2m+1

=
6
.
1
sin x π 1
Also, 0 < < 1, for 0 < x < , implies cot2 x < 2 < csc2 x;
x 2 x
1 rπ
∑ cot2 xr < ∑ xr2 < ∑ csc2 xr ; use xr = 2m + 1
[email protected]
 40

2m(2m − 1) (2m + 1)2 2

< < ∑ 1 + cot xr ;
6 r2 π 2

2m(2m − 1) (2m + 1)2 2m(2m − 1) 2m2 + 2m

 
1
< ∑ r2 < m + = ;
6 π2 6 3
m
2m(2m − 1)π 2 1 2m(2m + 2)π 2
< ∑ < .

om
6(2m + 1)2 r=1 r2 6(2m + 1)2
∞
1 π2
Taking the limit as m → ∞ , we get ∑ 2 6 =
r=1 r

l.c
ai
gm
ri@
ha
sic
ra
na

[email protected]
 41
33 b) History (Euler, 1737) Let k ≥ 1 be a positive integer. Then
∞
1 k+1 (2π)2k
ζ (2k) := ∑ = (−1) B2k
n=1 n2k 2(2k)!

Ans: The notation ζ (s) stands for the Riemann zeta function, (German, 1859).
The Bernoulli numbers Bn (Jakob Bernoulli, brother of Euler’s mentor Johann Bernoulli, died
in 1705, paper published in 1713), are defined by the equation

om
∞
z Bn n
z
:= ∑ z
e −1 n=0 n!

Step 1: We start with the Taylor series for x cot x obtained using the Fourier series expansion as

l.c
follows:
Let f (x) = cos(ax), −π ≤ x ≤ π, where a is not an integer.
We assume the Fourier series expansion formula of a periodic function, f (x + 2π) = f (x).

ai
a0 ∞  nπx  ∞  nπx 
f (x) = + ∑ an cos + ∑ bn sin
2 n=1 l n=1 l
l=
b−a
= π; a0 =
1 Rπ
f (x) dx =
1 Rπ
gm
cos(ax) dx =
2 sin(aπ)
2 l −π π −π aπ
1 Rπ 1 Rπ 2(−1)n a sin(aπ)
an = f (x) cos(nx) dx = cos(ax) cos(nx) dx =
l −π π −π π (a2 − n2 )
ri@

1 Rπ 1 Rπ
bn = f (x) sin(nx) dx = cos(ax) sin(nx) dx = 0, (since f (x) is an even function).
l −π π −π
Hence, we get the Fourier series
" #
ha

sin(aπ) 1 ∞
2(−1)n a sin(aπ)
cos (ax) = +∑ cos (nx)
π a n=1 π (a2 − n2 )
∞
1
Putting x = π, we get, aπ cot(aπ) = 1 + 2a2
sic

∑ a2 − n2
n=1
 
Let |a| ≤ r < 1. Then a2 − n2  ≥ n2 − r2 > 0, ∀ n ∈ N

The series is convergent for a ∈ B(0, r), a ∈ C \ Z

ra

i
a2
∞ ∞ ∞ 
1 1 −1 ∞
Consider ∑ =∑  2 =∑ 2 ∑
a2 − n2 n=1 a n=1 n i=0 n2
na

n=1 2
n 2
−1
n
 i
2
∞
1 ∞ a2
⇒ aπ cot(aπ) = 1 − 2a ∑ 2 ∑
n=1 n i=0 n2
∞  2
a4

a
= 1−2 ∑ 2 + 4 +...
n=1 n n

= 1 − 2 a2 ζ (2) + a4 ζ (4) + . . .
 

Let a = z ∈ C \ Z. Then this identity becomes

∞
π z cot πz = 1 − 2 ∑ z2k ζ (2k) (1)
k=1

[email protected]
 42
Step 2: This can be matched with an expansion for the analytic function z cot z in B(0, r), using
the Bernoulli numbers as follows.
z ∞ Bn −1
z
= ∑ zn ; B1 = , Bn = 0, for n odd, n > 1 and B2n = 0
e − 1 n=0 n! 2
∞
z z Bn n z
⇒ z + = ∑ z +
e −1 2 n=0 n! 2
z(ez + 1) ∞
Bn n z
⇒ = z +

om
∑
2(ez − 1) n=0 n! 2
 
z ez/2 + e−z/2 ∞
B z
⇒
= ∑ n zn +
2 ez/2 − e−z/2 n=0 n! 2

l.c
∞
z z Bn n z
⇒ coth = ∑ z +
2 2 n=0 n! 2
∞
Bn
(2z)n + z

ai
⇒ z coth z = ∑
n=0 n!
∞
Bn
⇒ iz coth (iz) = (2iz)n + iz
n=0
∑ n!
gm
∞
Bn
⇒ z cot z = ∑ (2iz)n + iz
n=0 n!
B1  B2
ri@

⇒ z cot z = 1 + 
iz + (2iz) + (2iz)2 + . . .
1 2!
∞
2 2k
⇒ z cot z = 1 + ∑ (−1)k B2k z2k
k=1 (2k)!
∞
22k
⇒ πz cot πz = 1 + ∑ (−1)k B2k (πz)2k (2)
ha

k=1 (2k)!

Comparing eq. 1 and eq. 2, we get

sic

∞
1 k+1 (2π)2k
ζ (2k) = ∑ = (−1) B2k , k≥1
n=1 n2k 2(2k)!

There is no such explicit formula for ζ (2k + 1)

ra

Use the formal, easy-to-use, recurrence relation (B + 1)n+1 − Bn+1 = 0,

n  
n+1 n
na

i.e., ∑ B = 0; replace Bn by Bn ; we get

k=0 k
n  
n+1
∑ k Bn = 0; use the initial value B0 = 1. We get
k=0
1
n = 1 ⇒ 20 B0 + 21 B1 = 0 ⇒ B1 = −



2
1
n = 2 ⇒ 30 B0 + 31 B1 + 32 B2 = 0 ⇒ B2 =




6
−1 1
Similarly, B4 = ; B6 = ; hence we get the first few standard formulas,
30 42
∞
1 π 2 ∞
1 π4 ∞ 1 π6
= ;
∑ 2 6 ∑ n4 90 ∑ n6 945 = ; =
n=1 n n=1 n=1

[email protected]
 43
33 c) History: Wallis’ infinite product formula for π, (Oxford, 1656)
 2 2
(2n − 2)2 1

2 4
π = lim 2 . 2 . . . . ; equivalently,
n→∞ 1 3 (2n − 3)2 n

π ∞
4n2 2.2.4.4.6.6 . . .
=∏ 2
=
2 n=1 4n − 1 1.3.3.5.5.7 . . .

Proof: Let a > 0; 0 ≤ x ≤ a

om
y=∞ 1 −1 y y=∞
 
R dy π
2 2
= tan =
y=0 x +y x x y=0 2x
Differentiating under the integral sign (Leibnitz, 1675), we get,

l.c
y=∞
R −2x dy −π y=∞
R dy π
2
= ⇒ 2
=
y=0 (x2 + y2 ) 2x2 y=0 (x2 + y2 ) 4x3

ai
Repeating DUIS n − 1 times we get
y=∞
R dy 1 3 2n − 3 π 1
= . ... . .
√
y=0 (x2 + y2 )
n
gm
2 4 2n − 2 2 x2n−1
Put x = n on both sides
y=∞
R dy 1 3 2n − 3 π 1
n = . ... . . √ 2n−1
ri@

2 2 4 2n − 2 2 ( n)
y=0 (n + y )
y=∞
√
R dy 1 3 2n − 3 π n
⇒ n = . . . . . .
y2 2 4 2n − 2 2 nn

y=0 n
n 1+
n
a bn
ha


Take the limit as n → ∞; using lim 1 + = eab , we get
n→∞ n
y=∞ 1 3 2n − 3 π √
2
e−y dy = lim . . . .
R
. . n
sic

y=0 n→∞ 2 4 2n − 2 2
The Euler gamma function (1729), (i.e., double integrals in polar coordinates), gives
y=∞
  √
R −y2 1 1 π
e dy = Γ = . Hence
ra

y=0 2 2 2
√
π 1 3 2n − 3 π √
= lim . . . . . . n; squaring, we get Wallis’ product formula
na

2 n→∞ 2 4 2n − 2 2
22 42 (2n − 2)2 1
π = lim . . . . . ; change n → n + 1
n→∞ 12 32 (2n − 3)2 n

22 42 (2n)2 1
π = lim 2
. 2
. . . 2
.
n→∞ 1 3 (2n − 1) n + 1

22 .42 .62 . . . .(2n)2 2n + 1

π = lim .
n→∞ (1.3)(3.5) . . . (2n − 1)(2n + 1) n + 1

22 .42 .62 . . . .(2n)2 4n2

  ∞
π
π = lim . 2 ⇒ = ∏ 4n2 − 1
n→∞ (22 − 1)(42 − 1) . . . ((2n)2 − 1) 2 n=1

[email protected]
 44
33 d) History: See ‘The life of Pi’, Jonathan M. Borwein, (2011)
22
Show that π <
7
x4 (1 − x)4
Hint: The function f (x) = is non-negative for 0 ≤ x ≤ 1;
1 + x2
hence the area under the curve y = f (x), 0 ≤ x ≤ 1, is positive.
x4 (1 − x)4 = x4 (1 − 4x + 6x2 − 4x3 + x4 ) = x8 − 4x7 + 6x6 − 4x5 + x4 ; long division

om
x4 (1 − x)4 6 − 4x5 + 5x4 − 4x2 + 4 − 4
= x
x2 + 1 x2 + 1
4 4
R1 x (1 − x) dx
Area = >0
x2 + 1

l.c
x=0
 7 1
x x6 x5 x3
⇒ − 4 + 5 − 4 + 4x − 4 arctan x > 0
7 6 5 3 0

ai
1 2 4 π 22
⇒ − +1− +4−4 > 0 ⇒ >π
7 3 3 4 7 gm 1 1 1
34 a) Show that the n−th partial sum of the harmonic series, Hn = 1 + + + . . . + is not an
2 3 n
integer, for n ≥ 2.
1 1 1
b) Show that if n > m ≥ 1, + + . . . + is not an integer.
ri@

m m+1 n

2 4 2016
c) Show that + +...+ is not an integer.
3 5 2017

1 1 1
d) Show that + + ...+ is not an integer, where 2 = p1 , 3 = p2 , . . . , pn are the first n
ha

p1 p2 pn
primes.
a
sic

e) If H p − 1 = , where p is prime, p > 2 and gcd (a, b) = 1, show that p divides a.

b
f) In the previous problem, if p > 3, we have p2 | a.
1 1 (−1) p − 2
g) Let p ≥ 3 be an odd prime. Show that 1 − + − . . . + ≡ 0 (mod p)
ra

22 32 (p − 1)2

Ans: a) This has been done earlier, (in problem 2). Let m = 2t , t ≥ 1 be the highest power of 2
na

which is less than or equal to n, i.e., m = 2t ≤ n < 2t+1 = 2m

1 1 1
Suppose k = 1 + + . . . + t + . . . + is an integer, for n ≥ 2; multiply both sides by 2t−1
2 2 n
2t−1 2t−1 2t−1
2t−1 k = 2t−1 + +...+ t +...+ .
2 2 n
1 t−1 t−1 2t−1 2t−1
= 2 k−2 − −...−
2 2 n
All the fractions on the right side have odd denominators, hence we get
1 a
= , with b odd, contradiction.
2 b

[email protected]
 45
1 1
Method 2: Suppose k = 1 + + . . . + is an integer, for n ≥ 2.
2 n
Let p be the largest prime such that p ≤ n.
We know that, by Bertrand’s postulate, for any positive integer n > 1, there always exists a prime
strictly between n and 2n .
Hence n < 2p . (Otherwise, if 2p ≤ n there would exist a prime between p and 2p ≤ n, con-
tradicting the choice of p).

om
Hence the only primes which occur in the prime factorisation of all the numbers (other than p)
between 1 and n are all strictly less than p.
1 1 1
Suppose k = 1 + + . . . + + . . . + . Multiplying both sides by n! we get

l.c
2 p n
n! n! n!
kn! = n! + + . . . + + . . . + ; each term is an integer.
2 p n

ai
Going mod p, we get
n!
LHS is divisible by p ; is the only term on the RHS which is not divisible by p; all the other
p
gm
terms on the RHS are divisible by p. This is a contradiction.
1 1 1
32 b) To prove that, if n > m ≥ 1, + + . . . + is not an integer.
m m+1 n
ri@

We can repeat the first method. Let 2t , t ≥ 1 be the highest power of 2 such that 2t |i, for
m ≤ i ≤ n. Then there exists a unique j such that m ≤ j ≤ n and j is an odd multiple of 2t ,
(since, between two odd multiples there would be an even multiple of 2t dividing the integers
between m and n, which contradicts the definition of t as the maximal power of 2 with this
ha

property). Suppose, if possible,

1 1 1 1
k= + + + . . . + is an integer.
m m+1 j n
sic

Multiply both sides by 2t−1 to get

1 a 2am + b
2t−1 k = + , with both m and b odd ⇒ k = t
2m b 2 bm
ra

The left side is an integer; the numerator of the fraction on the right side is odd and the denomi-
nator is even; this is a contradiction.
na

2 4 2016 1 1 1 1
c) + +...+ = 1− +1− +1− +...+1− .
3 5 2017 3 5 7 2017
1 1 1
If this is an integer then + + . . . + is also an integer.
3 5 2n + 1
Repeat the second method used in part a) to get a contradiction.

a
d) If H p − 1 = , where p is prime, p > 2 and gcd (a, b) = 1, show that p divides a.
b
Hint: Combine the first term with the last term, the second term with the second last term, etc.

[email protected]
 46

e) In part d), if p > 3, we have p2 | a.

(p − 1)/2  
1 1
[This is Wolstenholme’s theorem, (1862): H p − 1 = ∑ + ;
k=1 k p−k
(p − 1)/2 (p − 1)/2  (p − 1)/2
p2 (p2 − 1)

1 −1 2 2
= p. ∑ = p. ∑ 2
(mod p Z p ) = p. ∑ k = ∈ p2 Z p
k=1 k(p − k) k=1 k k=1 24

om
1 1 (−1) p − 2
f) Let p ≥ 3 be an odd prime. Show that 1 − + − . . . + ≡ 0 (mod p).
22 32 (p − 1)2

l.c
g) (Putnam Exam, 1996): Let p > 3 and k = [2p/3]. Show that
     
p p p
+ +...+ ≡ 0 (mod p2 )
1 2 k

ai
p(p − 1) . . . (p − i + 1) (−1)i−1 1 1 −1
Hint: = p. (mod p); + ≡ , etc..
i! i 2i p−i 2i
gm
1 1 1 p
h) (IMO, 1979) If 1 − + − · · · + = where p and q are coprime positive integers,
2 3 1319 q
show that p is divisible by 1979.
ri@

   
1 1 1 1 1 1
Ans: 1 + + + · · · + − + +···+ ; add and subtract the even terms;
3 5 1319 2 4 1318
   
1 1 1 1 1 1
= 1+ + +···+ −2 + +···+
2 3 1319 2 4 1318
ha

   
1 1 1 1 1 1 1 1
= 1+ + +···+ + +...+ − 1+ + +···+
2 3 659 660 1319 2 3 659
1 1 1
= + +···+ ; combine the first and last, the second and second-last terms, etc.
sic

660 661 1319

1979 1979 1979 p m
= + +···+ ⇒ = (1979) ,
(660)(1319) (661)(1318) (989)(990) q n
ra

where n is a product of positive integers less than 1979. But 1979 is prime;
hence p must be a multiple of 1979.
na

[email protected]
 47

35 Find the maximum possible value of k for which 2018 can be written as a sum of k consecutive
integers.
Ans: Let the sum of n − m = k consecutive numbers be

(m + 1) + (m + 2) + · · · + n = 2013 ⇒ (1 + 2 + · · · + n) − (1 + 2 + · · · + m) = 2018
n(n + 1) m(m + 1)
− = 2018 ⇒ n2 − m2 + n − m = 4036

om
2 2
(n + m)(n − m) + (n − m) = 4036 ⇒ (n − m)(n + m + 1) = 4036 = 4 × 1009 = 1 × 4036
The possible values are n − m = k = 1, 4; the maximum value is k = 4.

l.c
36 Find the sum (in base 10) of all natural numbers less than 64 which have exactly three ones in
their binary representation.

ai
Ans: The base 10 representation is the usual decimal expansion, e.g.,

(314)10 = 4 + 1(10) + 3(10)2 ; a number which is less than 103 has at most 3 digits in its decimal
expansion.
gm
The base 2 (binary) expansion is similar, e.g., (1011)2 = 1 + 1(2) + 0(2)2 + 1(2)3 = 11;

(100)2 = 0 + 0(2) + 1(2)2 = 4; (1000)2 = 8; (1000000)2 = 64

ri@

A number which is less than 26 = 64 has at most 6 digits in its binary expansion.
We must select any 3 places out of 6 places for placing the ones. This can be done in
6
6×5×4
3 = = 20 ways. The remaining 3 places will contain zeroes.
ha

×2 × 3
Every number in binary has a unique complement obtained by changing the ones to zeroes and
zeroes to ones. For example, (101110) and (010001) are complements of each other.
sic

The sum of any binary 6-digit number and its complement is given by

(111111)2 = 1 + 1(2) + 1(2)2 + 1(2)3 + 1(2)4 + 1(2)5 = 1 + 2 + 4 + 8 + 16 + 32 = 63

ra

Hence the twenty 6-digit numbers (having three ones) can be split into two parts of ten elements
each; and the total sum will be 10 × 63 = 630.
na

3 4 2018 1 1
37 Find a + b if + +...+ = −
1! + 2! + 3! 2! + 3! + 4! 2016! + 2017! + 2018! a! b!
n+2 n+2 1 n+1 (n + 2) − 1
Ans: = 2
= = =
n! + (n + 1)! + (n + 2)! n!(n + 2) n!(n + 2) (n + 2)! (n + 2)!
1 1
= − ;
(n + 1)! (n + 2)!
1 1
hence we get a telescoping series which gives − ⇒ a + b = 2020
2! (2018)!

[email protected]
 48

38 Find the remainder when N = 111 . . . 1 with 124 digits is divided by 271.
Ans: We must calculate 11, 111, 1111, etc. mod 271. The 5-digit number 11111 = 41 × 271.
Hence 105 − 1 = 99999 is divisible by 271.

10125 − 1 is divisible by 271; 111 . . . 1 with 125 digits is divisible by 271.

10(111 . . . 1) (124 digits) = −1 = 270 mod 271 ⇒ N = 27 mod 271

om
39 If F1 = 1, F2 = 1, Fn+1 = Fn + Fn−1 , for n ≥ 2 evaluate
∞
1
∑ Fn−1 Fn+1
n=2

l.c
 
1 1 Fn+1 − Fn−1 1 1
Ans: ∑ =∑ . =∑ − ;
n≥2 Fn−1 Fn+1 n≥2 Fn Fn−1 Fn+1 n≥2 Fn Fn−1 Fn Fn+1

ai
1 1 1 1 1
this is a telescoping series which sums to − + − +... = = 1.
F1 F2 F2 F3 F2 F3 F3 F4
gm F1 F2

40 If x0 = 2018 and n xn = (n − 2) xn−1 + 1 for n ≥ 1, find x2018

1
Ans: n = 2 gives 2 x2 = 1 ⇒ x2 = ; n(xn − xn−1 ) = −2 xn−1 + 1 ⇒ x3 − x2 = 0 ⇒ x3 = x2 ; by
2
1
induction we get, xn = , for all n ≥ 2.
ri@

2
ha
sic
ra
na

[email protected]
 49
n
1 1
41 History: Show that ∑ 2
< 2−
r=1 r n
1
Ans: Draw the graph of the decreasing convex function y = for x > 0
x2
Consider the Riemann sum, i.e., the sum of the areas of the lower rectangles with bases
2 3 n
[1, 2], [2, 3], . . . , [n − 1, n] and heights
2
, 2 , . . . , 2 respectively.
n n n
This sum is less than the area under the curve, i.e., the definite integral.

om
n Z n
1 1
Hence, ∑ 2 < 1 + 2
dx
r=1 r 1 x
−1 n

1
LHS < 1 + = 2−

l.c
x 1  n 
1 1 1 1 1 1
Second method: 1 + 2 + 2 + . . . + 2 < 1 + + +...+
2 3 n 2. 3 3. 4 n(n + 1)
1 1 1 1 1 1

ai
< 1+ − + − +...+ −
2 3 3 4 n n+1
1 1 3 1 3 1 1 1
< 1+ − = −
2 n+1 2 n+1 2 2 n
gm
< + − = 2 − , for n ≥ 2
n
ri@
ha
sic
ra
na

[email protected]
 50

1 1 1
42 History: Infinite series The harmonic series ∑ n = 1 + 2 + . . . + n + . . . is divergent.
     
1 1 1 1 1 1 1 1 1
Proof: 1 + + + + + + + + +...+ +...
2 3 4 5 6 7 8 9 16
   
1 1 1 1 1 1 1
> 1+ + + + + + + +...
2 4 4 8 8 8 8

om
   
1 1 1
> 1+ +2 +4 +...
2 4 8  
n 1 1
Hence the partial sum upto N = 2 terms gives SN > 1 + + + . . . , (N times); hence
2 2

l.c
N
SN > 1 + ⇒ SN → ∞ as N → ∞, i.e., the series is divergent.
2
∞  
1 1 1 1 1 1 1
Obviously, the series of even terms, ∑ = + + +... = 1 + + + . . . is diver-
n=1 2n 2 4 6 2 2 3

ai
∞
1 1 1 1 1 1
gent; the series of odd terms, ∑ = 1 + + + . . . > + + + . . . is also divergent;
n=1 2n − 1 3 5 2 4 6
But the corresponding alternating series ∑
∞
gm
(−1) n−1 1 1
= 1 − + − . . . is convergent.
n=1 n 2 3
This a particular case of Abel’s test, (Norwegian, 1822) for conditional convergence:
ri@

A series of alternating terms (of real numbers) is convergent if the terms are numerically strictly
(monotonically) decreasing and tending to zero.
The value (i.e., limit) of this series is given by the general Maclaurin (Scottish) / Taylor (English,
1715) series expansion:
f 0 (0) x f 00 (0) x2
ha

f (x) = f (0) + + +...

1! 2!
x2 x3
In particular, loge (1 + x) = x − + − . . .
sic

2 3
1 1 1
⇒ ln(2) = log 2 = 1 − + − + . . .
2 3 4
Other classical alternating series were first discovered by Indian mathematicians (the Kerala
ra

school) in the calculation of the circumference of a circle:

Madhava (Kerala, 1340-1425, recorded in Nilkantha’s ‘Tantra Sangraha’ (1501)), found a for-
na

mula (i.e., series expansion) for π :

π 1 1 1
= 1− + − +...
4 3 5 7
This series is called Gregory’s series, (Scotland, 1668). It also follows from the Maclaurin series
for tan−1 x :
x3 x5 x7
tan−1 x = x − + − − . . .; the convergence can be accelerated as follows:
3 5 7
       
π 1 1 1 1 1 1 1 1 1 1
= 1− + + − − + − + + − −...
4 4 4 8 3 8 12 5 12 16 7
1 1 2 2 π 3 1 1
but + − = 3 ⇒ = + 3 − 3 +...
n−1 n+1 n n −n 4 4 3 −3 5 −5

[email protected]
 51
43 Divergent series: A divergent series should not be re-arranged in evaluating the sum.
In fact, a conditionally convergent alternating series can be re-arranged to sum to any pre-
assigned number.
1 1
Example: The series 1 − + − . . . is conditionally convergent.
2 3
Given any real number k, take n1 terms of the divergent series of odd positive terms

om
1 1
S1 = 1 + + ... + such that this sum is just greater than k ; then take n2 terms of the
3 2n1 − 1
1 1 1
divergent series of even positive terms S2 = + + . . . + such that the sum S1 − S2 is just
2 4 2n2
less than k .

l.c
Repeat this process to get the sum arbitrarily close to k.
Example: It is easy to get contradictions and pure nonsense by re-arranging divergent series.

ai
S = 1−1+1−1+1−...

S= 1−1+1−1+1−...
gm 1
Adding, we get, 2S = 1 + (−1 + 1) + (−1 + 1) + . . . = 1 ⇒ S = (1)
2
Similarly, T = 1 − 2 + 3 − 4 + 5 − . . .
ri@

T= 1−2+3−4+5−...
1 1
Adding, we get, 2T = 1 − (2 − 1) + (3 − 2) + . . . = 1 − 1 + 1 − 1 + . . . = S = ⇒T = (2)
2 4
X = 1+2+3+4+...
ha

T = 1−2+3−4+5−...
1
Subtracting, we get, X − T = 4 + 8 + 12 + . . . = 4X ⇒ −3X = T =
4
sic

1
⇒ X = 1+2+3+... = − ; but a sum of positive numbers cannot be negative.
12
1 1 1
Similarly, 1 + + + + . . . is divergent. The terms cannot be arranged and added or sub-
2 3 4
ra

tracted arbitrarily.
   
1 1 1 1 1 1 1
1+ + +... = 1+ + +... + + + + . . . ; add and subtract the even terms
na

2 3 3 5 2 4 6
        
1 1 1 1 1 1 1 1 1 1
⇒ 1+ + +... = 1− + − + − +... +2 + + +...
2 3 2 3 4 5 6 2 4 6
   
1 1 1 1 1 1 1 1
⇒ 1+ + +... = + + +... +2 + + +...
2 3 1. 2 3. 4 5. 6 2 4 6
   
1 1 1 1 1 1 1
⇒ 1+ + +... = + + +... + 1+ + +...
2 3 1. 2 3. 4 5. 6 2 3
1 1 1
⇒ + + + . . . = 0 ; but a sum of positive real numbers cannot be zero.
1. 2 3. 4 5. 6
Conclusion: Never manipulate a divergent or conditionally convergent series.

[email protected]
 52
1
44 Region of convergence: The infinite geometric series 1 − x + x2 − x3 + . . . = ,
1+x
is convergent for −1 < x < 1, (or in the open unit disc |x| < 1 for complex numbers).
We get wrong conclusions by using this series at points outside this region of convergence. For
example,
1
x = 1 ⇒ 1 − 1 + 1 − 1 + . . . = , which is nonsense.
2
Such contradictions and erroneous conclusions emphasise the fact that algebraic operations with

om
infinite series should be undertaken only at points inside the established region of convergence.
Standard series expansions:
1
We know the geometric series = 1 − x + x2 − x3 + . . ., for |x| < 1

l.c
1+x
Integrating term-by-term, we get
x2 x3

ai
log(1 + x) = x − + − . . ., for |x| < 1
2 3
Changing x to x2 on both sides of the geometric series, we get
1
gm
= 1 − x2 + x4 − x6 + . . ., for |x| < 1; integrating term-by-term, we get
1 + x2
−1 x3 x5
tan x = x − + − . . ., for −1 < x ≤ 1
3 5
ri@

This series is convergent at x = 1, by the alternating series test. This gives

1 1 π
1− + −... = (Madhava-Gregory-Leibniz formula)
3 5 4
ha

Madhava, (1400), Gregory, (1668) and Leibniz, (German, 1671)

The Kerala school also discovered the standard expansions of sin x and cos x:
sic

x3 x5 x7 x2 x4 x6
sin x = x − + − + . . . ; cos x = 1 − + − + . . .
3! 5! 7! 2! 4! 6!
(The Taylor (1715) series for sin x, cos x and ex were used for real and complex numbers by
ra

Euler and were later proved to be convergent by Cauchy, (French, 1821), for all real as well as
complex numbers).
x2 x3
na

ex = 1 + x + + +... ;
2! 3!
Changing x to ix, we get Euler’s identity, (Swiss, 1748): eix = cos x + i sin x ; eiπ + 1 = 0
The binomial theorem, (Newton, 1665, for rational exponents), can be proved by induction for
positive integral powers with integer binomial coefficients nk :

 
n n k n(n − 1) 2 n(n − 1)(n − 2) 3
(1 + x) = ∑ x = 1+nx+ x + x + . . . , for |x| < 1
k≥0 k 2! 3!

The Pascal triangle, (French, 1653), (al-Karaji, Arabic, 1000 CE) satisfies the recurrence relation
     
n n n+1
+ =
r r−1 r
[email protected]
 45 The base e of the natural logarithm: (Jakob Bernoulli, Swiss, 1683), (Euler, 1727)
1 n
 
Consider the binomial expansion of 1 + , for any natural number n ≥ 1 : 53
n
1 n
 
xn = 1 +
n
     n
1 n(n − 1) 1 1
= 1+n + + . . . +
n 2! n2 n
         
1 n−1 1 n−1 n−2 1 n−1 n−2 1
= 1+1+ + +...+ ...
2! n 3! n n n! n n n
         
1 1 1 1 2 1 1 2 1

om
= 2+ 1− + 1− 1− +...+ 1− 1− ... (1)
2! n 3! n n n! n n n
Change n to n + 1. This gives
    
1 1 1 1 2
xn+1 = 2 + 1− + 1− 1− +...

l.c
2! n+1 3! n+1 n+1
    
1 1 2 1
+ 1− 1− ... (2)
n! n+1 n+1 n+1

ai
k 1 1 1 1 1 1
For 1 ≤ k ≤ n − 1, we have < 1; > ⇒− <− ⇒ 1− < 1−
n n n+1 gm n n+1 n n+1
hence each bracket in (1) is positive and is less than the corresponding bracket in (2); also (1)
contains n + 1 positive terms and (2) contains n + 2 positive terms.
Hence xn < xn+1 , for all n ≥ 1, i.e., the sequence {xn } is monotone (strictly) increasing.
k
ri@

Further, each bracket in (1) lies between 0 and 1, since 0 < < 1 for 1 ≤ k ≤ n − 1
n
Equation (1) gives xn > 2, for n ≥ 2
Also, from equation (1), we get
         
1 1 1 1 2 1 1 2 1
xn = 2 + 1− + 1− 1− +...+ 1− 1− ...
ha

2! n 3! n n n! n n n
1 1 1
< 2 + (1) + (1) + . . . + (1), (since each bracket is less than 1)
2! 3! n!
sic

1 1 1
⇒ xn < 2 + + + . . . +
2! 3! n!
But 3! = 1.2.3 > 1.2.2 = 22 ; 4! = 1.2.3.4 > 1.2.2.2 = 23 , etc., n! > 2n−1 for n ≥ 3
ra

1 1 1
⇒ xn < 2 + + 2 + . . . + n−1 ; this gives a geometric progression;
2 2 2
"  n−1 #
na

1
⇒ xn < 2 + 1 − < 2 + 1 = 3; ⇒ 2 < xn < 3 , for n ≥ 2;
2
hence {xn } is monotone increasing and bounded above; hence the sequence is convergent.
1 n
 
1 1 1
The limit is denoted by e, i.e., lim 1 + = e = 1+ + + +...
n→∞ n 1! 2! 3!

[email protected]
 54
46 a) Show that there are infinitely many primes of the form 4n + 3 .
Ans: The proof is similar to Euclid’s proof (350 BC) that there are infinitely many prime num-
bers.
Suppose if possible there are only finitely many primes, {p1 , p2 , . . . pk } of the form 4n + 1.
Let N = 4 p1 p2 . . . pk + 3. Then N leaves remainder 3 when divided by each pi ; hence N is
not divisible any of the primes {p1 , p2 , . . . , pk }.

om
Then the odd number N must have at least one prime factor of the form 4n + 3 , since any prod-
uct of numbers of the form 4n + 1 is again of the form 4n + 1.
This prime factor cannot be any of the primes {p1 , p2 , . . . , pk }. This contradicts our assumption
that these are all the only primes of the form 4n + 3.

l.c
b) Show that there are infinitely many primes of the form 4n + 1 .

ai
Ans: First note that the method of part a) is not strong enough to prove this theorem.
Method 1: Quadratic Residues. Proof by contradiction. Suppose, if possible, there are only
gm
finitely many prime numbers, {p1 , p2 , . . . , pk }, of the form 4n + 1.
Let N = 4p21 p22 . . . p2k + 1 = (2p1 p2 . . . pk )2 + 1 (1)

N is odd; so it must have only odd primes factors; these are of the form 4n + 1 or 4n + 3. But
ri@

equation (1) shows that N leaves the remainder 1 when divided by any prime of the form 4n + 1;
hence it must have a prime factor p of the form 4n + 3.

⇒ (2p1 p2 . . . pk )2 + 1 ≡ 0 (mod p) ⇒ x = 2p1 p2 . . . pk , x2 ≡ −1 (mod p)

This says that −1 is a quadratic residue mod p; hence p must be of the form 4n + 1 , contra-
ha

diction.
Method 2: Given a natural number n > 1, let p be a prime factor of N := (n!)2 + 1. Then
p > n; hence p must be odd and n! cannot be a multiple of p.
sic

(n!)2 + 1 ≡ 0 (mod p) ⇒ (n!)2 ≡ −1 (mod p)

ip−1 p−1
ra

h
2 2 ≡ (−1) 2 (mod p)
⇒ (n!)
p−1
na

p−1
⇒ (n!) ≡ (−1) 2 (mod p)

But, by Fermat’s little theorem, a p−1 ≡ 1(mod p), for all a coprime to p.
We have n! coprime to p, (since p > n); hence (n!) p−1 ≡ 1(mod p)
p−1
p−1 p−1
⇒ (−1) 2 ≡ 1(mod p) and p is odd; hence is even ⇒ = 2m ⇒ p = 4m + 1
2 2
This proves that, for every n > 1, there exists a prime p, p > n and p ≡ 1 (mod 4)
When n → ∞, we get p → ∞. Hence there exist infinitely many primes of the form 4n + 1.

[email protected]
 55

47 Generating functions. Show that

Fn 1 2 3 5 8
∑ n
= 1+ + 2 + 3 + 4 + 5 +... = 4
n≥0 2 2 2 2 2 2
where Fn is the Fibonacci sequence {1, 1, 2, 3, 5, 8, . . .} defined by

F0 = 1, F1 = 1, Fn+1 = Fn + Fn−1 , for n ≥ 1

Ans: The generating function of Fn is given by F(x) = ∑ Fn xn

om
F(x) = F0 + F1 x + F2 x2 + . . . = 1 + x + 2x2 + 3x3 + 5x4 + . . .

Fn+1 = Fn + Fn−1 ⇒ ∑ Fn+1 xn = ∑ Fn xn + ∑ Fn−1 xn , ( F−1 = 0)

l.c
Consider x F(x) = x + x2 + 2x3 + 3x4 + . . . = F0 x + F1 x2 + F2 x3 + . . . = ∑n≥0 Fn−1 xn ,
F(x) − 1
= 1 + 2x + 3x2 + 5x3 = ∑n≥0 Fn+1 xn = F(x) + x F(x)
x

ai
Hence the recurrence relation gives the identity for the generating function,

F(x) − 1
x
gm
= F(x) + x F(x); F(x) − 1 = x F(x) + x2 F(x) ⇒ F(x) =
1
1 − x − x2
1 1
1 + x + 2x2 + 3x3 + 5x4 + . . . = ; put x = on both sides;
1 − x − x2 2
we get
ri@

1 2 3 5 8
1+ + 2 + 3 + 4 + 5 +... = 4
2 2 2 2 2

Further, the generating function gives the formula for Fn , as follows.

ha

√ √
−1 −1 −1 + 5 −1 − 5
F(x) = 2 = ; α= ; β= ; αβ = −1
x + x − 1 (x − α)(x − β ) 2 2
sic

1
 
  1
1 1 1 1  α β 
By partial fractions, F(x) = − √ − = −√  − .
5 x−α x−β 5  x − 1 x − 1
α β
ra

But α and β are negative reciprocals of each other. Hence

na

  " #
∞ ∞
1 β α 1
F(x) = − √ −
5 1 + β x 1 + αx
= √ −β
5
∑ (−β x)n + α ∑ (−αx)n
n=0 n=0
 √ !n+1 √ !n+1 
∞ ∞
1 5 + 1 5−1
∑ Fn xn = F(x) = √5 ∑  2 + (−1)n
2
 xn
n=0 n=0

Equating coefficients on both sides, we get,

 √ !n+1 √ !n+1 
1  5+1 5−1
Fn = √ + (−1)n 
5 2 2

[email protected]
 56

48 If the sides of a right-angled triangle are integers show that the in-radius, the three ex-radii and
the area are also integers; the area is a multiple of 6; the circumradius of a primitive Pythagorean
triangle is a non-integer rational.
Ans: We have Euclid’s formula for Pythagorean triangles, a2 + b2 = c2 with a, b, c integers
if and only if a = k m2 − n2 , b = 2kmn, c = k m2 + n2 , for some positive integer k and co-

prime integers m and n with m > n ≥ 1. If k = 1, we get primitive Pythagorean triples.

a+b−c

om
Drop perpendiculars from the in-centre to the sides to get r = = k n (m − n), which is
2
a+b+c
an integer. s = = k m(m + n), which is an integer. Hence the area is given by
2
∆ = rs = k2 mn m2 − n2 , which is an integer.

l.c
∆ ∆
The ex-radii are given by ra = = km(m − n); rb = = kn(m + n);
s−a s−b
∆ abc

ai
rc = = km(m + n). The circumradius R is given by the general formula ∆ = . For a
s−c 4R
m2 + n2
right triangle, R is half the hypotenuse; hence we get R = . Now m and n are of op-
gm 2
posite parity for a primitive Pythagorean triple; hence m2 + n2 is odd; hence R is not an integer.

49 a) Let {ai , bi , ci } be positive real numbers which form an arithmetic progression, for 1 ≤ i ≤ n.
ri@

Let xi , 1 ≤ i ≤ n, be positive real numbers. Prove that

!
n xibi xibi
∑ xbi + xai + xbi + xci = n
1 i i i i
99
5100
ha

b) Evaluate ∑
n=1 5100 + 52n
sic

b b
Ans: a) If a, b, c are in GP, we get b2 = ac ; this is equivalent to the equation + =1
b+a b+c
i.e., {b + a, b + b = 2b, b + c} form a harmonic progression..
5100 5100 5100
ra

b) Given + + . . . +
5100 + 25 5100 + (25)2 5100 + (25)99
To use the idea of part a), combine the first term with the last term, the second term with the
na

second-last term, and so on. We get 49 terms, and the central stand-alone term corresponding
to n = 50

2550 2550 2550 2550

   
+ + + +...
25 + 2550 2599 + 2550 252 + 2550 2598 + 2550
2550 2550 2550
 
+ + +
2549 + 2550 2551 + 2550 2550 + 2550
1 99
= 49(1) + =
2 2

[email protected]
 57

50 (Dutch contest, 2018): Let a, b, c be distinct positive integers such that a+2b+3c < 12. Which
of the following conditions must be satisfied?
a) 3a + 2b + c < 17 b) a + b + c < 7 c) a − b + c < 4 d) b + c − a < 3 e) 3b + 3c − a < 6.
Ans: a + 2b + 3c < 12 gives c ≤ 2 , since c ≥ 3 ⇒ a + 2b ≤ 3 , which is impossible for distinct
positive integers.

c = 2 gives a 6= 2, b 6= 2 and a + 2b < 6 ⇒ a + 2b ≤ 5 ⇒ a = 3, b = 1 ⇒ (a, b, c) = (3, 1, 2)

om
c = 1 gives a 6= 1, b 6= 1 and a+2b < 9 ⇒ a+2b ≤ 8 ⇒ (a, b, c) = (3, 2, 1), (2, 3, 1), (4, 2, 1)
Of all the given conditions, only d) is satisfied.

l.c
∞  
1 1
51 Find the sum of the infinite series ∑ −
n=1 5n − 1 5n + 1

ai
Ans: Method 1) The standard procedure to evaluate such infinite series is to use Fourier series.
Consider the Fourier series expansion of the periodic function
gm
f (x) = cos(ax), −π ≤ x ≤ π, f (x + 2π) = f (x), where a is not an integer.
a0 ∞  nπx  ∞  nπx 
f (x) = + ∑ an cos + ∑ bn sin
2 n=1 l n=1 l
ri@

b−a 1 Rπ 2 sin(aπ)
l= = π; a0 = cos(ax) dx =
2 l −π aπ
1 R π n
2(−1) a sin(aπ)
an = cos(ax) cos(nx) dx =
l −π π (a2 − n2 )
1 Rπ
ha

bn = cos(ax) sin(nx) dx = 0, (since f (x) is an even function).

l −π

Hence, we get the Fourier series expansion

sic

" #
sin(aπ) 1 ∞
2(−1)n a sin(aπ)
cos (ax) = +∑ cos (nx)
π a n=1 π (a2 − n2 )
∞
1
ra

Putting x = π, we get, aπ cot(aπ) = 1 + 2a2 ∑

n=1 a2 − n2
 2 2
 2 2
Let |a| ≤ r < 1. Then a − n ≥ n − r > 0, ∀ n ∈ N
na

The series is convergent for a ∈ B(0, r), a ∈ C \ Z

1
Using this formula with a = , we get
5
π π 2 ∞ 1
cot = 1 + ∑ 1
5 5 25 n=1
− n2
25
∞
π π 1
cot = 1 − 2 ∑ 2
5 5 n=1 25n − 1
Hence we get
∞   ∞
√
1 1 2 π π 5+1
∑ 5n − 1 − 5n + 1 = ∑ 2
= 1 − cot ; we can use cos 36◦ =
n=1 n=1 25n − 1 5 5 4
[email protected]
 58
∞  
1 1
Method 2: To find the sum of the infinite series ∑ −
n=1 5n − 1 5n + 1
∞  
1 1
Ans: Let S = ∑ − (1)
n=1 5n − 1 5n + 1
x5n−1 x5n+1
∞  
Let f (x) = ∑ − (1)
n=1 5n − 1 5n + 1
x4 x6 x9 x11 x14 x16

om
⇒ f (x) = − + − + − +... (2)
4 6 9 11 14 16

⇒ f 0 (x) = x3 − x5 + x8 − x10 + x13 − x15 + . . .

f 0 (x) = x3 1 + x5 + x10 + . . . − x5 1 + x5 + x10 + . . .

l.c
= x3 − x5 1 + x5 + x10 + . . .

x3 − x5

ai
= , for |x| < 1
1 − x5
Z 1 3
x − x5
⇒ S = f (1) =
0 1 − x5
dx
gm
Z 1 3
x − 1 + 1 − x5
S= dx
0 1 − x5
ri@

Z 1 3
x −1
= 1− dx
0 x5 − 1
x2 + x + 1
Z 1
= 1− dx (3)
0 x4 + x3 + x2 + x + 1
ha

 x 2 x2
To factorise the denominator: use reciprocal equations or x2 + + 1 = x4 + 2x2 + + x3 +
2 4
x+1
sic

2 5x2  2 √ !2
 x x 5x
⇒ x4 + x3 + x2 + x + 1 = x2 + + 1 − = x2 + + 1 −
2 4 2 2
√ √
5+1 1− 5
ra

= f g = x2 + αx + 1 x2 + β x + 1 , with α =



and β =
2 2
√
Hence f + g = 2x2 + x + 2; f − g = 5 x; α β = −1; hence
na

1 1 2x2 + 2x + 2
Z
S = 1− dx
2 0 fg
1 1 (2x2 + x + 2) + x
Z
S = 1− dx
2 0 fg
1
Z 1 ( f + g) + √ ( f − g)
1 5
S = 1− dx
2 0 fg
   
1 1
1+ √ f + 1− √ g
1 1
Z
5 5
S = 1− dx
2 0 fg

[email protected]
 59

1 √
  Z 1 dx √  Z 1 dx 
S = 1− √ 5+1 + 5−1
2 5 0 g 0 f
1 √
 Z1 dx √ Z1 dx

S = 1− √ 5+1 2
, dx + 5−1 2
2 5 0 x +βx+1 0 x + αx + 1
√ √
π 5+1 1+ 5 π
We know that cos = ⇒α = = 2 cos ;
5 4 2 5

om
√
1− 5 2π 3π
β= = −2 cos = 2 cos ;
2 5 5
"
1 √ Z1 dx

l.c
S = 1− √ 5+1
2π 2
 
2 5 0 2π
x − cos + sin2
5 5
#
√

ai
Z1 dx
+ 5−1
0
 π 2 π
x + cos + sin2
5 5
gm
2π
 
x − cos
"
1 √  1
5 
= 1− √ 5+1 tan−1 

2 5 2π 2 π 
sin sin
ri@

5 5  
π #1
√  1 x + cos
+ 5−1 tan−1  5
π  π 
sin sin 0
5 5
ha

π π
Verify that this is equivalent to S = 1 − cot
5 5
sic

52 Let fk (x) : N ∪ 0 → N ∪ 0, k ∈ N be defined as

y = fk (x) = xk (mod 10)

A) y = fk (x) is a periodic function with period 3
ra

B) f2019 ( f2017 (x) − x) = f2019 ( f2013 (x) − x) has infinitely many solutions.
na

C) f2019 ( f2018 ( f2017 (. . . f1 (x)))) = x is true only when the units place of x is 0, 1, 5
D) y = fk (x) is a non-periodic function.
Ans: Option B) is the only correct answer.
Details: The units digit of any power of an integer depends only on the units digit of the integer,
because, by the binomial theorem,

(x + 10)k = xk + 1k (10) xk−1 + 2k (10)2 xk−2 + . . . + 10k

⇒ (x + 10)k − xk is divisible by 10, i.e., (x + 10)k ≡ xk (mod 10), for all integers.

fk (x + 10) = fk (x), for all k ≥ 1

Hence fk (x) is periodic with period at most 10; options (A) and (D) are incorrect. (1)
[email protected]
 60
The units digit of xk is 5 ⇔ the units digit of x is also 5, i.e., fk (5) = 5, for all k ≥ 1
If the units digit of x is 0 or 1 or 6 we get fk (x) = x. Hence option (C) is wrong. (2)

fk (2) = {21 , 22 , . . .} = {2, 4, 8, 6, 2, 4, 8, 6, . . .}

fk (3) = {31 , 32 , . . .} = {3, 9, 7, 1, 3, 9, 7, 1, . . .}
fk (7) = {71 , 72 , . . .} = {7, 9, 3, 1, 7, 9, 3, 1, . . .}
fk (8) = {81 , 82 , . . .} = {8, 4, 2, 6, 8, 4, 2, 6, . . .}
If the units digit of x is 2, 3, 7, 8 we have f4k+1 (x) = x. (2)

om
fk (4) = {41 , 42 , . . .} = {4, 6, 4, 6, . . .}
fk (9) = {91 , 92 , . . .} = {9, 1, 9, 1, . . .}
If the units digit of x is 4, 9 we have f2k+1 (x) = x. (2)

l.c
The values (modulo 10) of the function fk (x) = xk are tabulated:
k x

ai
1 2 3 4 5 6 7 8 9 0
1 1 2 3 4 5 6 7 8 9 0
2 1 4 9 6 5 6 9 4 1 0
gm
3 1 8 7 4 5 6 3 2 9 0
4 1 6 1 6 5 6 1 6 1 0
5 1 2 3 4 5 6 7 8 9 0
ri@

6 1 4 9 6 5 6 9 4 1 0
7 1 8 7 4 5 6 3 2 9 0
8 1 6 1 6 5 6 1 6 1 0
9 1 2 3 4 5 6 7 8 9 0
ha

Recall the fact: x5 and x have the same last (units) digit, for all positive integers x.
Proof: x5 − x = x(x4 − 1) = x(x2 + 1)(x2 − 1) = x(x2 + 1)(x + 1)(x − 1)
sic

Either x is even or x + 1 is even; hence x5 − x is always a multiple of 2

Going mod 5, we have x = 0 or ±1 or ±2 (mod 5).
ra

If x = ±1 we get x2 − 1 = 0

If x = ±2 we get x2 + 1 = 4 + 1 = 5 = 0 (mod 5); hence x5 − x = 0 (mod 5)

na

Hence x5 − x is always a multiple of 10.

Option B), f2019 ( f2017 (x) − x) = f2019 ( f2013 (x) − x), holds for all integers.
Proof: 2013 and 2017 are congruent to 1 (mod 4); hence we have

x = 2, 3, 7, 8 ⇒ f4k+1 (x) = x ⇒ f2019 ( f2017 (x) − x) = f2019 ( f2013 (x) − x) = f2019 (0) = 0

x = 4, 9 ⇒ f2k+1 (x) = x ⇒ f2019 ( f2017 (x) − x) = f2019 ( f2013 (x) − x) = f2019 (0) = 0

x = 0, 1, 5, 6 ⇒ fk (x) = x ⇒ f2019 ( f2017 (x) − x) = f2019 ( f2013 (x) − x) = f2019 (0) = 0

[email protected]
 61

53) (INMO 2019) Let m, n be distinct positive integers. Prove that

gcd(m, n) + gcd(m + 1, n + 1) + gcd(m + 2, n + 2) ≤ 2 |m − n| + 1

Also find the values of m, n such that equality holds.
Ans: We know that gcd(a, b) = gcd(a, a − b).
[Proof: Suppose d divides both a and b. Then d divides a − b. Conversely, if d divides both

om
a and a − b. Then d divides a − (a − b) = b. Hence any common factor of a and b is also a
common factor of a and a − b, and vice versa.]
Both the left and right sides of the given inequality remain unchanged when we interchange a
and b. Hence, by symmetry, we may assume m > n.

l.c
Let m − n = k ≥ 1.

gcd(m, n) = gcd(m, m − n) = gcd(m, k) ⇒ gcd(m, n) divides k ⇒ k = a gcd(m, n), for some

ai
natural number a. (1)
Similarly, gcd(m + 1, n + 1) = gcd(m + 1, m − n) = gcd(m + 1, k) ⇒ gcd(m + 1, n + 1) divides
gm
k ⇒ k = b gcd(m, n), for some natural number b (2)
Similarly, gcd(m + 2, n + 2) = gcd(m + 2, m − n) = gcd(m + 2 k) ⇒ gcd(m + 2, n + 2) divides
k ⇒ k = c gcd(m + 2, n + 2), for some natural number c (3)
ri@

From eqs (1) and (2), k divides am and k divides b(m + 1).
Hence k divides a(b(m + 1)) − b(am) = ab ⇒ k divides ab. Hence k ≤ ab. (4)
From eqs (2) and (3), k divides b(m + 1) and k divides c(m + 2).
Hence k divides b(c(m + 2)) − c(b(m + 1)) = bc ⇒ k divides bc. Hence k ≤ bc. (5)
ha

Consider the left side of the given problem:

LHS = gcd(m, n) + gcd(m + 1, n + 1) + gcd(m + 2, n + 2)
sic

k k k
= + +
a b c
k k
Eq (4) gives ≤ b; eq (5) gives ≤ b
ra

a c
k k
Hence LHS ≤ b + + b = 2b +
b b
na

k
To prove that 2b + ≤ 2k + 1
b
⇔ 2b2 − 2kb − b + k ≤ 0

⇔ 2b(b − k) − 1(b − k) ≤ 0

⇔ (b − k)(2b − 1) ≤ 0
But this is clear, since, b ≥ 1 and k = b gcd(m + 1, n + 1) ⇒ k ≥ b
Equality holds in the last line only when k = b ⇒ gcd(m + 1, k) = 1.

[email protected]
 62

Also, equality in the given problem implies

gcd(m, k) + gcd(m + 1, k) + gcd(m + 2, k) = 2k + 1

⇔ gcd(m, k) + 1 + gcd(m + 2, k) = 2k + 1

⇔ gcd(m, k) + gcd(m + 2, k) = 2k; but each term on the left side is ≤ k

⇔ gcd(m, k) = k = gcd(m + 2, k)

om
⇔ k|m and k|(m + 2) ⇔ k = 1 or k = 2

⇔ (m, n) = (n + 1, n) or (n + 2, n ) or (n, n + 1) or (n, n + 2)

l.c
54) Let P(x) = x3 + ax2 + b; Q(x) = x3 + bx + a , where a and b are real numbers. The roots of
P(x) = 0 are reciprocals of the roots of Q(x) = 0. Find

ai
gcd (P(2013! + 1), Q(2013! + 1))
1
Hint: Change x to ; we get
 
1 1 a
x
gm ax 1
P = 3 + 2 + b = 0 ⇒ bx3 + ax + 1 = 0 ⇒ x3 + + = 0
x x  x b b
1
But x3 P and Q(x) are both monic polynomials having the same roots; hence they must
ri@

x
be identical.
ax 1 a 1
Hence x3 + + ≡ x3 + bx + a ⇒ = b and = a
b b b b
Hence b2 = a; b3 = 1; also a, b are real; hence b = 1; a = 1
ha

P(x) = x3 + x2 + 1; Q(x) = x3 + x + 1
Suppose d divides P(k) = k3 + k2 + 1 and d also divides Q(k) = k3 + k + 1. Then d divides
sic

P(k) − Q(k) = k2 − k = k(k − 1)

d|k(k − 1) ⇒ d|k or d|k − 1. Let k = (2013)! + 1

ra

If d|k and d|P(k) = k3 + k2 + 1 ⇒ [d|k3 , d|k2 ] ⇒ d = 1

If d|(k − 1) and d|P(k) = k3 + k2 + 1 = (k3 − 1) + (k2 − 1) + 3 ⇒ [d|(k3 − 1), d|(k2 − 1)] ⇒ d|3
na

If d|(k − 1) and d|Q(k) = k3 + k + 1 = (k3 − 1) + (k − 1) + 3 ⇒ d|3

Hence gcd (P(k), Q(k)) = 3

55) The Sylvester sequence is defined as s0 = 1, s1 = 2, sn+1 = s2n − sn + 1, for n ≥ 1

Show that a) sn+1 = s0 s1 · · · sn + 1, for n ≥ 1

b) Show that gcd (sm , sn ) = 1, for all distinct positive integers, m 6= n

c) Identify the sequence given by s0 = 1, s1 = 3, sn+1 = s0 s1 · · · sn + 2

[email protected]
 63

56) (Singapore Math Olympiad, 2009, Junior section): m and n are two positive integers with digits
in reverse order whose product is 1446921630. m + n = ab4ba. Find ab.
Hint: One number must be 5 ∗ ∗ ∗ 2 and the other must be 2 ∗ ∗ ∗ 5.
Perform long multiplication. The second-last digit of the final answer is 3. The second-last digit
of the first line of the answer must be either 1 or 6; but any multiple of 2 has an even last digit;
hence 1 + 2 = 3 is the only possibility for the second-last digit.

om
5∗∗∗2
×2 ∗ ∗ ∗ 5
————–

l.c
∗ ∗ ∗1 0
......2 .
.........

ai
.........
∗ ∗ ∗.. gm
————–
1446921630
————–
The second-last digit of the first number must be even, i.e., 2, 4, 6 or 8 . The second-last digit
ri@

of the second number must be 1 or 6.

The third-last digit of the first three lines of the multiplication may be 1, 2, 0 or 4, 8, 4 to get 6
as the third-last digit of the answer. By trial and error, we get

51282 × 28215 = 1446921630

ha

Second method: Factorise mn = 1446921630 = 2.35 .7.112 .19.37

The first digit of the answer is 1; hence we may assume the last digit of m is 5 and the last digit
sic

of n is 2, and not 4, 6 or 8.
The product is divisible by 35 ; hence 9 divides at least one of m or n. But m and n have the
same digits in reverse order. Hence the sum of their digits is the same. Hence both m and n
ra

must be multiples of 9.
Again, the product mn is divisble by 112 ; hence 11 divides at least one of m or n. But by the
na

same argument, m and n have the same digits in reverse order. Hence if 11 divides m then 11
divides n. Hence both m and n must be multiples of 11.

n is even; hence n = 2 × 9 × 11 × k = 198k (1)

The last digit of n is 2; hence the last digit of k is either 4 or 9. But 4 does not divide n;
hence the last digit of k must be 9.

k is 7 × 37 or 3 × 7 × 19 or 3 × 19 × 37 (2)
The first digit of n is 5; hence 50000 < n = 198k < 60000 ; trying (2), we get k = 7 × 37;
n = 51282; m = 28215.

[email protected]
 64

57) (math stack exchange, May 29, 2019)

Find positive integer solutions of the equation 3u2 + 3v2 + 3w2 = 10(uv + vw + wu)
Ans: Note that, if (u, v, w) is a solution, then (−u, −v, −w) and λ (u, v, w) are also solutions
of the given equation.
u v
De-homogenise by putting x = ; y = ; 3x2 − 10xy − 10x + 3y2 − 10y + 3 = 0.
w w

om
Completing the square in the x−variable, we get (3x − 5y − 5)2 − 16y2 − 80y − 16 = 0
Completing the square in the y−variable, we get (3x − 5y − 5)2 − 4(2y + 5)2 = −84
The first term must be even. Put 3x − 5y − 5 = 2X; X 2 − (2y + 5)2 = −21 (1)

l.c
This is Pell’s equation. P(X, y) ≡ (2, 0) is one rational point on this hyperbola. To find all ra-
tional points, intersect the conic with a general straight line through P having rational slope, m.

ai
Let y = m(X − 2); eq (1) gives X 2 − (2mX − 4m + 5)2 = −21

X 2 − (2mX − 4m)2 − 10(2mX − 4m) − 25 + 21 = 0

gm
(X 2 − 4) − 4m2 (X − 2)2 − 20m(X − 2) = 0
8m2 − 20m + 2
(X + 2) − 4m2 (X − 2) − 20m = 0 ⇒ X(1 − 4m2 ) + 8m2 − 20m + 2 = 0 ⇒ X =
4m2 − 1
ri@

8m2 − 20m + 2
 
−4m(5m − 1)
y = m(X − 2) ⇒ y = m 2
−2 =
4m − 1 4m2 − 1
64m2 + 20m + 1
Eq (1) gives 3x = 5y + 5 + 2X ⇒ x =
3(1 − 4m2 )
64p2 + 20pq + q2 4p(5p − q)
 
p
ha

m ∈ Q ⇒ m = ⇒ (x, y) = , 2
q 3(q2 − 4p2 ) q − 4p2
u v
(x, y) = , ⇒ (u, v, w) = (64p2 + 20pq + q2 , 12p(5p − q), 3(q + 2p)(q − 2p))
w w
sic

1 p 1
We can take p and q to be positive integers with < <
5 q 2
There are infinitely many integer solutions. Examples:
ra

p = 1, q = 2 ⇒ (x, y, z) = (108, 36, 0) ≡ (3, 1, 0)

p = 1, q = 3 ⇒ (x, y, z) = (133, 24, 15)

na

p = 1, q = 4 ⇒ (x, y, z) = (160, 120, 36) ≡ (40, 3, 9)

p = 2, q = 5 ⇒ (x, y, z) = (481, 120, 27)

p = 2, q = 7 ⇒ (x, y, z) = (585, 72, 99) ≡ (65, 8, 11)

1 1 1 p
58) If 1 − + − · · · + = in reduced form, show that 83 divides p
2 3 55 q
   
p 1 1 1 1 1 1
Ans: = 1 + + + · · · + −2 + +···+
q 2 3 55 2 4 54
   
1 1 1 1 1 1
= 1+ + +···+ − 1+ + +···+
2 3 55 2 3 27
[email protected]
 65

p 1 1 1
= + +···+
q 28 29 55
As usual, reverse the order and add the two equations
p 1 1 1
= + +···+
q 55 54 28
2p 83 83 83 83m

om
= + +···+ =
q (28)(55) (29)(54) (28)(55) n
But gcd(83, n) = 1; hence 83 must divide p.

l.c
59) A student was asked to add all the numbers from 1 to a certain integer n. While doing this, he
double-counted one number k by mistake and got the answer m.

ai
a) If m = 2020 find n and k.
b) If m is any given positive integer, can we always determine n and k uniquely?
gm
Ans: b) Suppose if possible, (1 + 2 + · · · + n) + k = (1 + 2 + · · · + m) + l with (1)

1 ≤ k ≤ n < m and 1 ≤ l ≤ m.

(1 + 2 + · · · + n) + k < (1 + 2 + · · · + n) + (n + 1) ≤ (1 + 2 + · · · + m), contradicting eq (1)

ri@

Hence, given 1 ≤ k ≤ n, the sum (1 + 2 + · · · + n) + k uniquely determines n and k.

a) Given k ≤ n and (1 + 2 + · · · + n) + k = 2020

n(n + 1)
+ k = 2020
ha

2
4n2 + 4n + 1 + 2k = 16161; (2n + 1)2 + 8k = 16161; k ≤ n ⇒ 16161 < 4n2 + 4n + 1 + 8n <
(2n + 3)2
sic

The odd square just below 16161 is 1272 = 16129; 2n + 1 = 127 ⇒ n = 63;

8k = 16161 − 16129 = 32; k = 4

ra
na

[email protected]