Polynomial Division in Number Theory

James Rickards
Canadian Winter Camp 2017

1 Introduction
One technique which can be very useful in solving number theory problems is the use of polynomial division.
With some careful algebra and a will to carry on, you can solve some difficult problems without any tricky
ideas. The general setup is when you know that a quotient of two expressions is an integer. The method you
use is to find an approximation to what the quotient should be, ideally involving “nice” terms and a “nasty”
term, where the nasty term is small. We then impose the conditions of the quotient being in Z to deduce
what the nasty term should be, and then we have nice algebraic equations to work with.
A classical example looks like the following: Find all integers x such that

3x3 − 5x + 1
= Q ∈ Z.
2x − 1
For this problem, it’s best to multiply by 8 to get

24x3 − 40x + 8 12x2 − 40x + 8 −34x + 8 −9

8Q = = 12x2 + = 12x2 + 6x + = 12x2 + 6x − 17 +
2x − 1 2x − 1 2x − 1 2x − 1
−9
Since Q, x are integers, we see that 2x−1 is an integer, which reduces us to finitely many cases: 2x − 1 =
−9, −3, −1, 1, 3, 9, so x = −4, −1, 0, 1, 2, 5. In each of these cases we have shown that 8Q ∈ Z, but since
the denominator of Q in lowest terms is a divisor of 2x − 1 which is odd, this implies that Q ∈ Z. Thus
x = −4, −1, 0, 1, 2, 5 is the set of solutions.

2 An Example
Let’s go on to a more difficult example: the famous IMO 1988 problem 6: Let a, b be two positive integers
such that ab + 1 | a2 + b2 . Prove that
a2 + b2
ab + 1
is a perfect square. The normal technique used to solve this problem is Vieta jumping. While this is a very
standard trick now, back in the day it wasn’t well known, and as such this problem was very difficult. The
solution we give here will be somewhat similar, but not requiring any clever ideas.

To start off, suppose WLOG that b ≥ a. Thus the b term is the “dominant term”, and to get a small term
in the numerator we wish to eliminate that.

a2 + b2 b − b + a2 b a3 − b
Q= = + a = + 2 .
ab + 1 a ab + 1 a a b+a
Since a ≤ b, we see that either 0 ≤ a3 − b < a3 ≤ a2 b < a2 b + a or 0 ≤ b − a3 < b < a2 b + a. In any case, we
3
−b
have a “nice” term of ab and a “nasty” term aa2 b+a which satisfies

a3 − b
< 1.
a2 b + a

1
In particular, Q ∼ ab ; it is the ceiling or floor of ab if a3 − b ≥ 0 or a3 − b ≤ 0 respectively. Thus it is natural
to write b = an + r, where n ∈ Z+ and r ∈ Z, 0 ≤ r < a. We must get Q = n or Q = n + 1, and plugging in
the expression for b yields

r a3 − an − r (nra2 + r2 a + r) + (a3 − an − r) a2 + nra + r2 − n

Q=n+ + 2
=n+ 2
=n+ .
a a(na + ra + 1) a(na + ra + 1) na2 + ra + 1
a2 +nra+r 2 −n
Therefore na2 +ra+1 = 0, 1.

If it is 0, then note our expression is linear in n and we solve to get

a2 + r2
n= .
1 − ar
But n > 0 whence 1 ≥ 1 − ar > 0, so ar = 0 and thus r = 0. This gives n = a2 and Q = n = a2 is a perfect
square (this is the solution (a, b) = (a, a3 )).

a2 +nra+r 2 −n
The other case is na2 +ra+1 = 1. When we multiply out, it is again linear in n, and we solve to get

a2 + r2 − ra − 1 2a2 + r2 − 2ar (a − r)2 + a2

Q=n+1=1+ 2
= 2
=
a − ra + 1 a − ra + 1 (a − r)a + 1

where a, a − r ∈ Z+ . In particular, starting with the pair (a, b) = (a, an + r) we get (a − r, a) giving the
same quotient. However this decreases the sum a + b (except if a = b, where we can check (1, 1) is they only
such possibility; this gives a quotient of 1), whence we can only do this finitely many times, whereupon we
must be in the first case or a = b = 1, and the quotient is a square in both cases. Note that our solution also
describes how to find all possible pairs (a, b) where ab + 1 | a2 + b2 + 1.

3 Problems
1. a) Find infinitely many pairs of integers a, b with 1 < a < b such that ab | a2 + b2 − 1
b) For a, b as in part a), find the possible values of

a2 + b2 − 1
ab
which occur for infinitely many pairs (a, b).

2. Find all pairs of positive integers (a, b) such that

a2 + b b2 + a
, ∈Z
b2 − a a2 − b
3. Consecutive positive integers m, m + 1, m + 2, m + 3 are divisible by consecutive odd positive integers
n, n + 2, n + 4, n + 6 respectively. Find the smallest possible m in terms of n.

4. Let x, y be integers such that xy + 1 | x2 + y 2 . Prove that if

x2 + y 2
N := < 0,
xy + 1
then N = −5.

2
 5. Find all a ∈ Z such that
x2 + axy + y 2 = 1
has infinitely many distinct integer solutions (x, y).

6. Find all pairs of positive integers (a, b) such that

a2 b + b + 7 | ab2 + a + b

7. Find all positive integers which can be represented uniquely as

x2 + y
,
xy + 1
for x, y positive integers.

8. Let x, y be positive integers such that xy | x2 + y 2 + 1. Prove that

x2 + y 2 + 1
=3
xy

9. Find all pairs of positive integers (a, b) such that

ab | a2 + b2 + 3.

NOTE: you can leave your answer in terms of a set of recursive sequences.

10. Find all pairs of positive integers (m, n) such that

n3 + 1
∈ Z.
mn − 1
11. Find all pairs of integers (a, b) such that

a2
∈ Z+ .
2ab2 − b3 + 1
12. Determine all triples (a, b, c) of positive integers such that each of the numbers

ab − c, bc − a, ca − b

is a power of 2.

12’. Prove that there are no quadruples (p, a, b, c) where p is an odd prime and a, b, c ∈ Z+ such that each
of the numbers
ab − c, bc − a, ca − b
is a power of p.

12”. Let a ≤ b ≤ c be positive integers and p a prime such that each of the numbers

−(ab − c), bc − a, ca − b

is a power of p. Prove that either (a, b, c) = (pu , pu , p2u + 1) for some non-negative integer u, or a = 1.