On a Number Theory Problem:

Chasing a Red Herring

Titu Andreescu & Marian Tetiva

1. Introduction: stating the problem. We all know that any two consecutive integers n
and n + 1 are relatively prime, which can also be expressed as gcd(n, n + 1) = 1 for any integer
n. (Here and further throughout this note by gcd(x, y) we denote the greatest common divisor
of the integers x and y. We have that x and y are relatively prime if and only if gcd(x, y) = 1.)
This is because any common divisor of n and n + 1 must also divide (n + 1) − n = 1. Similarly,
any two consecutive odd numbers are relatively prime, that is, gcd(2n − 1, 2n + 1) = 1 for any
integer n, since any common divisor of 2n − 1 and 2n + 1 must divide their difference, which is 2.
But both 2n − 1 and 2n + 1 are odd, hence the conclusion follows. We invite the reader to show
similarly that gcd(2n + 1, 4n + 1) = 1, or gcd(30n + 3, 24n + 2) = 1 for all n. On the other hand,
we evidently do not have gcd(2n + 3, 3n + 2) = 1 for every integer n, as long as this does not hold
for (at least) n = 1. So, naturally, we asked ourselves about the following
Problem 1. Let a, b, c, and d be integers. What necessary and sufficient conditions must they
satisfy in order to have gcd(an + b, cn + d) = 1 for all integers n?
The very simple (but, as we will see, also very useful to solving our problem) identity

a(cn + d) − c(an + b) = ad − bc

immediately shows that gcd(an + b, cn + d) = 1 holds for all n whenever ad − bc is either 1, or

−1. Nevertheless, it is naive to believe that this can be a necessary and sufficient condition as
long as we have a very simple example such as gcd(2n + 1, 4n + 1) = 1 (where a = 2, b = 1, c = 4,
d = 1, therefore ad − bc = −2). (Although many examples belong to this particular situation.)
The above identity also shows that gcd(an + b, cn + d) = 1 for all n whenever ad − bc is nonzero
and divides both a and c, since we then can rewrite it in the form
a c
(cn + d) − (an + b) = 1,
ad − bc ad − bc
with integer coefficients for an + b and cn + d. Although many particular examples can be framed
here, we see that gcd(30n + 3, 24n + 2) = 1, or gcd(2n + 17, 4n + 66) = 1 do not belong to this
case. So, until now, we found nothing.
2. We did not find Problem 1 in the literature, although we are pretty sure that it has been
studied and solved, possibly in much more general forms, so we tried to find a solution. (We mention
that writing this note is not at all based on any ambition of originality. We rather intended to show
how one finds a path to solving a problem through the maze of already known results, sometimes
wondering and getting lost on undesired and nowhere leading trails.) We actually started from the
following contest item.
Problem 2. Find all integers k for which gcd(4n + 1, kn + 1) = 1 for all integers n.
Solution. If d1 = gcd(4n + 1, kn + 1), we have (of course) that d1 | 4n + 1, and also

d1 | k − 4 = k(4n + 1) − 4(kn + 1),

therefore
d1 | d2 = gcd(4n + 1, k − 4).
But d2 | 4n + 1, too, and
d2 | kn + 1 = n(k − 4) + 4n + 1

Mathematical Reflections 5 (2023) 1

hence d2 | d1 . It follows that d1 = d2 , implying

gcd(4n + 1, kn + 1) = 1 ⇔ gcd(4n + 1, k − 4) = 1

for every integer n. Thus the condition from the statement of the problem is equivalent to gcd(4n+
1, k − 4) = 1 for all integers n. This is true if k − 4 = ±2s for some nonnegative integer s and some
choice of the signs plus/minus, because 4n + 1 is odd and has no common factors (other than 1
and −1) with ±2s . On the other hand, if k − 4 has an odd factor greater than 1, that factor will
be a common factor for k − 4 and 4n + 1 for some n (this is clear if the odd factor is of the form
4t + 1; when it is of the form 4t − 1, it will be also a factor of (4t − 1)2 = 4t(t − 1) + 1). Since,
under this assumption, k − 4 and 4n + 1 cannot be relatively prime for all n, it follows that an
odd factor greater than 1 is not allowed for k − 4, and we conclude that the numbers required by
the problem are those of the form 4 ± 2s , s being a nonnegative integer.
This still doesn’t suggest any general necessary and sufficient condition as required by Problem
1, but it makes a connection between gcd(an + b, cn + d) and gcd(cn + d, ad − bc) which, at first
glance, seemed to us to be true in general (but is not). Namely, because

a(cn + d) − c(an + b) = ad − bc

it follows that
gcd(an + b, cn + d) | gcd(cn + d, ad − bc)
for all n. On the other hand, we also have the equality

n(ad − bc) + b(cn + d) = d(an + b)

showing that the greatest common divisor of cn + d and ad − bc also divides d(an + b) – so, if we
had d = 1 (as in the previous example), then

gcd(cn + d, ad − bc) | gcd(an + b, cn + d)

and, hence,
gcd(cn + d, ad − bc) = gcd(an + b, cn + d)
would follow.
(Similarly, when b = 1, gcd(an + b, ad − bc) = gcd(an + b, cn + d) holds.) Thus we considered
the case d = 1, and got the next result.
Problem 3. Let a, b, and c be integers. Then we have

gcd(an + b, cn + 1) = 1

for every integer n if and only if any prime divisor of a − bc is also a factor of c.
Solution. As we just seen, the equality

a(cn + 1) − c(an + b) = a − bc

implies
gcd(an + b, cn + 1) | gcd(cn + 1, a − bc),
while
n(a − bc) + b(cn + 1) = an + b
implies
gcd(cn + 1, a − bc) | gcd(an + b, cn + 1)

Mathematical Reflections 5 (2023) 2

so we actually get
gcd(cn + 1, a − bc) = gcd(an + b, cn + 1)
for all n. Thus we have

gcd(an + b, cn + 1) = 1, ∀ n ∈ Z
⇔ gcd(cn + 1, a − bc) = 1, ∀ n ∈ Z.

Then it is very easy to see that the condition ”any prime divisor of a − bc is also a factor of c” is
sufficient to have gcd(an + b, cn + 1) = 1, or, equivalently, gcd(cn + 1, a − bc) = 1 for all n. Indeed,
if there exists some integer n for which gcd(cn + 1, a − bc) > 1, then a common prime divisor p
exists for both cn + 1 and a − bc. Since we assumed that p | a − bc ⇒ p | c, this p would divide
both c and cn + 1, which is impossible, so no n exists with gcd(an + b, cn + 1) > 1.
The condition ”any prime divisor of a − bc is also a factor of c” is also necessary to have
gcd(cn + 1, a − bc) = 1 for all n. If not, we would have gcd(cn + 1, a − bc) = 1 for all n, while a
prime q would exist such that q | cn + 1, and q does not divide c. But, this being the case, we can
find an n such that cn + 1 ≡ 0 mod q (the congruence cx + 1 ≡ 0 mod q is solvable). Since q
also divides a − bc, we get the contradiction q | gcd(cn + 1, a − bc), thus finishing the proof.
Well, this was the red herring that troubled our way towards the demonstration for the general
case: the misleading idea that we could use a connection between gcd(an + b, cn + d) and gcd(cn +
d, ad − bc) (or gcd(an + b, ad − bc)), as we did in the previous Problems 2 and 3. Nevertheless,
Problem 3 (and its particular case, Problem 2) finally led us to the general necessary and sufficient
conditions for which Problem 1 asks (but only when we decided to give up chasing chimeras).
Observing that ”any prime divisor of a − bc is also a factor of c” implies ”any prime divisor of
a − bc is also a factor of a”, too (and, anyway, some symmetry about a and c is inevitable) we
finally realized what we were looking for.
3. The solution. We now solve Problem 1, after we reformulate it as
Problem 4. For integers a, b, c, d the following statements are equivalent.
(i) The numbers an + b and cn + d are relatively prime for any integer n.
(ii) We have that b and d are relatively prime, and any prime divisor of ad − bc is also a factor
of both a and c.
Solution. The condition gcd(b, d) = 1 is obviously necessary in order to have gcd(an + b, cn +
d) = 1 for any integer n (take n = 0) – and we assume further that this is the case. Then note
that the equality
a(cn + d) − c(an + b) = ad − bc
holds for any n, and assume that a prime p divides ad − bc, but it does not divide a. Since a is
relatively prime to p, the congruence ax + b ≡ 0 mod p is solvable, hence we can find an integer
n satisfying it, that is, such that
an + b ≡ 0 mod p.
Multiplying this by d, and using the divisibility of ad − bc by p, we get

bcn + bd ≡ adn + bd ≡ 0 mod p,

or
b(cn + d) ≡ 0 mod p.
Now, if p divides b, since it also divides ad − bc, it follows that p divides ad. But p does not divide
a, hence we get p | d, and the assumption that b and d are relatively prime is contradicted. So p
does not divide b, hence b(cn + d) ≡ 0 mod p implies cn + d ≡ 0 mod p. We summarize: when
gcd(b, d) = 1, if a prime p exists such that p divides ad − bc, but p does not divide a, then we can

Mathematical Reflections 5 (2023) 3

find an integer n such that gcd(an + b, cn + d) > 1 (p divides both an + b and cn + d). Similarly,
the existence of a prime that divides ad − bc, but it does not divide c leads to the same conclusion.
Thus, for gcd(an + b, cn + d) = 1 to hold for all integers n it is necessary to have gcd(b, d) = 1,
and, also, to have that any prime factor of ad − bc is a prime factor of both a and c.
Now we show that these two conditions are also sufficient in order to have gcd(an+b, cn+d) = 1
for all n. That is, we assume that gcd(b, d) = 1, and that any prime dividing ad − bc also divides
both a and c, and we show that gcd(an + b, cn + d) = 1 for all n.
Indeed, let q be a common prime factor of an + b and cn + d, for some integer n. By using
again the identity
a(cn + d) − c(an + b) = ad − bc
we see that q divides ad − bc. But then, by hypothesis, q divides a, and q divides c, hence q divides
b = (an + b) − an, and q divides d = (cn + d) − cn. This comes in contradiction with the hypothesis
cd(b, d) = 1, hence the assumption that a prime common factor exists for an + b and cn + d (for
some n) is false, and the desired conclusion gcd(an + b, cn + d) = 1 for any integer n follows.
4. Final remarks. Note that, by contraposition, we get the next rewording of our main result
(Problem 4):
Problem 40 . For integers a, b, c, d the following statements are equivalent.
(i) There exists an integer n such that an + b and cn + d are not relatively prime.
(ii) We either have gcd(b, d) > 1, or there exists a prime divisor of ad − bc that does not divide
either a, or c.
Also, note some particular cases of the main result.

• In Problem 2 we have a = 4, b = 1, c = k, and d = 1, therefore the condition gcd(b, d) = 1

is fulfilled. Since ad − bc = 4 − k, by the result of Problem 4, for gcd(4n + 1, kn + 1) = 1 to
hold it is necessary and sufficient that any prime factor of 4 − k is also a factor of 4 and of
k. This means 4 − k = ±2s , hence k = 4 ± 2s for some nonnegative integer s, and if this is
the case, 2 (the only prime factor of ad − bc = 4k) is, indeed, a factor of not only 4, but of
k too. The result (as proved above) follows.
• When ad − bc = 1, or ad − bc = −1 the conditions gcd(b, d) = 1 and p | ad − bc ⇒ p | a
and p | c (for a prime p) are automatically satisfied (the second because no prime divisor
of ad − bc exists), hence gcd(an + b, cn + d) = 1 follows for any integer n. Slightly more
generally, if gcd(b, d) = 1, ad − bc | a, and ad − bc | c, then gcd(an + b, cn + d) = 1 for any n.
(Again, the equality a(cn+d)−c(an+b) = ad−bc immediately implies these results.) Most of
the usual examples one meets in elementary arithmetics textbooks, such as gcd(n, n + 1) = 1,
gcd(n + 1, 2n + 1) = 1, or gcd(2n + 1, 4n + 1) = 1 belong to one of these two particular cases
– which we also discussed in the Introduction. (Nevertheless, there also exist situations that
do not fit into these cases: for – one more – example, we have gcd(6n + 5, 12n + 6) = 1 for
all n.)
• Also, observe that when ad − bc = 0, we have gcd(an + b, cn + d) = 1 for any n if and only
if a = c = 0 and gcd(b, d) = 1.
• If we have a = c = 1 the condition ”any prime dividing ad − bc also divides both a and c”
can only be fulfilled for ad − bc = 1, or ad − bc = −1, meaning that d − b ∈ {1, −1}. This
closes a circle, since it leads us to the very first (and simplest, and most known) example we
gave: any two consecutive integers are relatively prime.

Finally, we invite the reader to see that 2n + 1 and 4n − 17 are not relatively prime for any n,
and to find such an n that 2n + 1 and 4n − 17 have a common divisor greater than 1.

Mathematical Reflections 5 (2023) 4