2,129
Views
0
CrossRef citations to date
0
Altmetric
Notes

Polynomials Consisting of Quadratic Factors with Roots Modulo Any Positive Integer

ORCID Icon
Pages 178-182 | Received 17 Nov 2020, Accepted 24 Feb 2021, Published online: 10 Jan 2022

Abstract

We give an infinite family of polynomials that have roots modulo every positive integer but fail to have rational roots. Each polynomial in this family is made up of monic quadratic factors that do not have linear terms. The proofs in this note are accessible to anyone with a basic knowledge of undergraduate elementary number theory.

1 Introduction

We concern ourselves with polynomials f with integer coefficients such that f(x)0(modm) is solvable for every positive integer m. If a polynomial f has an integer root, then f clearly has roots modulo every positive integer. However, there exist polynomials that have roots modulo every positive integer but do not have any rational root. Such polynomials provide counterexamples to the local-global principle in number theory (see [Citation3, pp. 99–108] for more details on the local-global principle).

Given a cube-free integer n1, Hyde, Lee, and Spearman proved in [Citation4] thatg(x)=(x3n)(x2+3)has no rational roots but has roots modulo every positive integer if and only if n1(mod9) and all prime factors of n are congruent to 1 modulo 3.

It is well known that if p, q are distinct odd primes such that pq1 (mod 4) and p is a square modulo q, then the polynomialh(x)=(x2p)(x2q)(x2pq)has no rational roots but has roots modulo every positive integer (see [Citation6, pp. 139–140]). Let c and d be square-free integers not equal to 1, let c1=cgcd(c,d) and d1=dgcd(c,d). Hyde and Spearman obtained necessary and sufficient conditions for the polynomialp(x)=(x2c)(x2d)(x2c1d1) to have a root modulo every integer but have no rational root [Citation5]. We obtain necessary and sufficient conditions for polynomials, made up of any number of similar quadratic factors, to have roots modulo every integer but have no rational root.

Note that the congruence f(x)0(modm) is solvable for every positive integer m if and only if the congruence f(x)0(modpb) is solvable for each prime number p and each positive integer b. This is a consequence of the Chinese remainder theorem.

Given a prime p and an integer n, we denote the Legendre symbol of n with respect to p by (np). When pn,(np)=+1 if n is a square modulo p and (np)=1 otherwise. When p|n,(np)=0. Given a prime p and a nonzero integer l, pa||l will denote that pa(a1) is the highest power of p dividing l. Our main result is the following theorem.

Theorem 1.

Let a1,a2,,an be n (3) distinct nonzero square-free integers, none of which is 1. Then the polynomial f(x)=(x2a1)(x2a2)(x2an) has roots modulo every positive integer if and only if the following conditions are satisfied:

  1. There exists T{1,2,,n} of odd cardinality such that:

    1. the product jTaj is a perfect square;

    2. for every jT and for every odd prime p dividing aj, there exists i{1,,n},ij, such that (aip)=+1.

  2. One of the ai is of the form 8m+1 for some mZ and m0.

We note that the polynomial f(x)=i=1n(x2ai) in Theorem 1 cannot have a rational root, since none of the ai is a perfect square.

2 Results on Quadratic Residues

In this section, we collect some results on quadratic residues that will be used to prove Theorem 1.

Proposition 1.

Let p be an odd prime, e1, and aZ such that pa. Then a is a square modulo pe if and only if (ap)=+1.

Since pa, the proof of this statement is an immediate consequence of Hensel’s lemma applied to the polynomial q(x)=x2a. (See [Citation6, p. 135] for a proof.)

The next proposition characterizes square-free integers, not equal to 1, that are square modulo every power of 2.

Proposition 2.

  1. Let aZ be an odd number. Then a is a square modulo 2i for every i1 if and only if a=8m+1 for some mZ. (See [Citation6, p. 136].)

  2. Let a be a square-free integer not equal to 1. Then a is a square modulo 2i for every i1 if and only a=8m+1 for some mZ and m0. (See [Citation5, Lemma 2.5].)

Proof.

If a is odd, then the result follows from 2(a). On the other hand, if a is even, then a2(mod 4) because it is square-free. Since 2 is not a square modulo 22, a cannot be a square modulo 22.

3 Some Useful Lemmas

We will also use the following lemma proved by Fried (see [Citation1] and [Citation2]).

Lemma 1.

Let a1,a2,,an be finitely many nonzero integers. Then the following conditions are equivalent:

  1. For each prime p that does not divide i=1nai, at least one of the ai is a square modulo p.

  2. There exists T{1,2,,n} of odd cardinality such that jTaj is a perfect square.

Lemma 2.

Let p be an odd prime and aZ a square-free integer. If (ap)+1, then a cannot be a square modulo p2.

Proof.

If pa, then (ap)+1; along with Proposition 1, this gives that a is not a square modulo p2.

If p|a, then p||a because a is square-free. Assume that a is a square modulo p2, i.e., x2a(modp2) for some xZ. Then we have p2|(x2a), i.e., p|(x2a).

Since p|a and p|(x2a), we have that p|x2 implying p2|x2. However, p2|(x2a) and p2|x2 gives that p2|a, contradicting that a is square-free. Therefore, a cannot be a square modulo p2.

Lemma 3.

Let p be a prime number, k,nN, and f(x)=i=1n(x2ai)Z[x]. If x2ai0(modpk) is not solvable for any 1in, then f(x)0(modpkn) is not solvable.

Proof.

For the sake of contradiction, assume that f(x)0(modpkn) is solvable for some xZ, i.e., pkn|(x2a1)(x2an). Then we must have pk|(x2aj), for some j{1,2,,n}.

However, pk|x2aj implies that x2aj0(modpk) is solvable, a contradiction to the fact that x2ai0(modpk) is not solvable for any i. Therefore, we have the result.

4 Proof of Theorem 1

Proof of necessity

The necessity of the condition 1(a) immediately follows from Lemma 1. Now we will show that if condition 1(b) or condition 2 fails then f fails to have a root modulo some integer.

Assume that the condition 1(b) in Theorem 1 fails: if T is a subset of {1,2,,n} of odd cardinality for which jTaj is a perfect square, then there exists jT and an odd prime p|aj such that (aip)+1 for any ij. Since each ai is square-free, Lemma 2 implies that for any ij,x2ai0(modp2) is not solvable. Similarly, x2aj0(modp2) is not solvable because (ajp)=0+1.

Since x2ai0(modp2) is not solvable for any 1in, Lemma 3 with k = 2 implies that the congruence f(x)0(modp2n) is not solvable.

If condition 2 in the statement of Theorem 1 fails, then none of the ai is equal to 8m+1 for any nonzero mZ. Since each of the ai is square-free, part (b) of Proposition 2 implies for any i{1,2,,n},x2ai0(mod8) is not solvable. Then applying Lemma 3 with p = 2 and k = 3 gives that the congruence f(x)0(mod8n) is not solvable.

Proof of sufficiency

As a consequence of the Chinese remainder theorem, it suffices to show that f(x)0(modpb) is solvable for every prime p and every integer b1.

If p is an odd prime that does not divide jTaj, an application of Lemma 1 for {aj}jT implies that for some j0T,aj0 is a square modulo p. Since pjTaj, Proposition 1 implies that aj0 is a square modulo every positive power of p. Therefore, f(x)0(modpb) is solvable for every integer b1.

If p is an odd prime dividing jTaj, then p|aj for some jT. Condition 1(b) of Theorem 1 ensures that there exists ij such that (aip)=+1. Since (aip)=+1, we have that pai. Therefore, ai is a square modulo all powers of p, using Proposition 1. Hence, f(x)0(modpb) is solvable for every integer b1.

If p = 2, then condition 2 of Theorem 1 ensures that for some k{1,2,,n} and for some nonzero integer m, ak=8m+1. Then x2ak0(mod2b) is solvable for every b1, by part (b) of Proposition 2. Therefore, f(x)0(mod2b) is solvable for every integer b1.

5 Corollaries and Examples

The first corollary gives a necessary and sufficient condition for the polynomial (x2p)(x2q)(x2pq) to have roots modulo every positive integer. Here p and q are distinct odd primes.

Corollary 1.

Let p and q be distinct odd primes. The polynomialf(x)=(x2p)(x2q)(x2pq)has roots modulo every positive integer if and only if (pq)=(qp)=+1.

Proof.

Note that p×q×pq=(pq)2 and that one of p or q or pq has to be of the form 8m+1 for 0mZ. So the result follows by applying Theorem 1 to a1=p,a2=q, and a3=a1a2=pq.

When p and q are distinct primes such that pq1(mod4),(pq)=+1 if and only if (qp)=+1. Hence, the polynomial h(x)=(x2p)(x2q)(x2pq) has roots modulo every integer when pq1(mod4) and (pq)=+1, as mentioned in the Introduction.

Corollary 2.

Let c and d be two square-free integers not equal to 1, c1=cgcd(c,d), and d1=dgcd(c,d). Then f(x)=(x2c)(x2d)(x2c1d1) has roots modulo every integer if and only if the following conditions are satisfied:

  1. For each odd prime p dividing c, (dp)=+1 or (c1d1p)=+1 and for each odd prime p dividing d, (cp)=+1 or (c1d1p)=+1.

  2. At least one of c, d and c1d1 is of the form 8m+1 for some nonzero integer m.

Necessary and sufficient conditions for the polynomial in the above corollary to have a root modulo every positive integer were obtained by Hyde and Spearman in [Citation5]. In our Corollary 2, there are fewer conditions to check compared to the result obtained in [Citation5].

Proof.

Apply Theorem 1 for three square-free integers a1=c,a2=d, and a3=c1d1 and note that a1×a2×a3=c×d×c1d1=(rc1d1)2, where r= gcd (c, d).

It is instructive to see some examples of application of Theorem 1. The polynomials listed below are obtained from different values of n and square-free integers a1,a2,,an in Theorem 1.

  1. (x213)(x217)(x2221) has roots modulo every integer because13×17×221=(221)2,(1317)=+1,(1713)=+1,and17=8(2)+1.

  2. (x27)(x211)(x219)(x231)(x2209) has roots modulo every integer because11×19×209=(209)2,(3111)=+1,(719)=+1, and 209=8(26)+1.

  3. (x27)(x211)(x219)(x231)(x245353) has roots modulo every integer because7×11×19×31×45353=(45353)2,(117)=+1,(3111)=+1,(719)=+1,(1931)=+1,and 45353=8(5669)+1.

6 Discussion

Readers may notice that if f has a root modulo every positive integer, then for any integer c, g(x):=f(x)(x2c) also has a root modulo every positive integer. This raises the question whether there exists a polynomial

f(x)=(x2a1)(x2an) such that

  1. f has roots modulo every positive integer, but

  2. every proper divisor of f fails to have a root modulo some positive integer.

This question is answered in the affirmative in [Citation7]. Moreover, for every n4, there exists infinitely many polynomials of the form f(x)=(x2a1)(x2an) that satisfy Properties 1 and 2 (see [Citation7]).

Acknowledgments

I am grateful to the editor and the anonymous referees for their valuable suggestions and comments. This work was partly supported by the NSF, under grant DMS-1812028.

References