A survey on fixed divisors (Q2074030)

This article provides an extensive and exhaustive survey on the topic of fixed divisor of polynomials, from its early beginnings in the late 19th century to recent years. Given an integral domain \(R\), a polynomial \(f\in R[X]\) and a subset \(S\) of \(R\), the fixed divisor of \(f\) over \(S\) is the ideal of \(R\) generated by the values taken by \(f(X)\) over \(S\), namely, \(d(S,f)=(f(a)\mid a\in S)\). This definition is easily generalized to polynomials in several variables. The fixed divisor of a polynomial has a tight connection with integer-valued polynomials, which are polynomials over \(K\) which map \(S\) into \(R\); for example, problems related to the study of irreducibility of such polynomials are established by looking at the fixed divisor of the polynomial. In section 2 the authors provide an thorough excursus on the basilar known facts about fixed divisors of univariate polynomials, starting from the work of \textit{K. Hensel} [J. Reine Angew. Math. 116, 350--356 (1896; JFM 27.0322.03)] and culminating in the important results of \textit{M. Bhargava} [J. Reine Angew. Math. 490, 101--127 (1997; Zbl 0899.13022)], when he introduced the notion of \(P\)-orderings, which among several different applications allowed him to compute the fixed divisor of a polynomial over an arbitrary subset of a Dedekind domain. It is also worth mentioning the pioneering work of \textit{G. Pólya} [J. Reine Angew. Math. 149, 97--116 (1919; JFM 47.0163.04)], who provided a complete solution to the problem of establishing which ideals of a Dedekind domain \(R\) with finite residue fields occur as fixed divisor of some primitive polynomial \(f\in R[X]\). That result is now a special case of a theorem of Bhargava. The last part of this section addresses some results related to the case of multivariate polynomials. In section 3, other notions related to fixed divisors are presented, such as simultaneous \(P\)-orderings, fixed divisor sequences, Newton and Schinzel sequences, universal sets and regular basis. The authors discuss briefly about the interplay between these different concepts and give several references about them. In section 4, the authors recall some analytic lower and upper bounds for the fixed divisor expressed in terms of some data of the given polynomial. In particular, the results by Turk about the probability that a polynomial in \(\mathbb Z[X]\) has fixed divisor equal to some integer \(d\) are also mentioned. In section 5, some applications of fixed divisors in the study of irreducibility of integer-valued polynomials are presented. It is worth recalling that the classic ring of integer-valued polynomials, \(\text{Int}(\mathbb Z)=\{f\in\mathbb{Q}[X] \mid f(\mathbb{Z})\subseteq\mathbb{Z}\}\), which lies in between the two UFDs \(\mathbb Z[X]\) and \(\mathbb Q[X]\), is far from being a UFD and the same integer-valued polynomial may have factorizations into irreducible which have different lenghts (i.e., the number of different irreducibles in the factorization). In section 6, the authors shows some applications of fixed divisor in number fields, focusing on the notion introduced by \textit{H. Gunji} and \textit{D. L. McQuillan} [J. Number Theory 2, 207--222 (1970; Zbl 0199.37402)]: given two number fields \(K\subset L\) with ring of integers \(O_K,O_L\), respectively, \(\mathfrak{I}(L\mid K)\) is defined as the ideal of \(O_K\) given by the l.c.m. of the fixed divisors \(d(f_a,O_K)\) of all the minimal polynomials of primitive elements \(a\in O_L\) such that \(L=K(a)\). \textit{C. R. MacCluer} [J. Number Theory 3, 33--34 (1971; Zbl 0215.36002)] established when \(\mathfrak{I}(L\mid K)\) is non-trivial. Recent works about this concept have been given by \textit{M. Ayad} and \textit{O. Kihel} [Int. J. Number Theory 7, No. 5, 1173--1194 (2011; Zbl 1223.11128)], who connected \(\mathfrak{I}(L\mid K)\) with the classical index of \(K\), i.e., the g.c.d. of \([O_K:\mathbb Z[a]]\), as \(a\) ranges in the set of primitive elements of \(K\) over \(\mathbb Q\). Some number-theoretic results related to the fixed divisors of polynomials conclude the section. In the last section 7 of this paper, the authors suggest a definition of fixed divors of polynomials with matrix coefficients, i.e., \(f\in M_m(R)[X]\), when they are evaluated at a given subset \(S\) of \(M_m(R)\): \(d(S,f)\) is the ideal of \(R\) generated by the entries of \(f(A)\), for \(A\in S\). This ideal is easily seen to be realted to a class of ideals of \(M_m(R)[X]\) studied by \textit{N. J. Werner} [Commun. Algebra 40, No. 12, 4717--4726 (2012; Zbl 1272.16028)] in connection with rings of integer-valued polynomials over matrices. The authors conclude the paper by raising some open questions in this setting.