An asymptotic formula for the number of smooth values of a polynomial (Q1604982): Difference between revisions

Suppose distinct irreducible polynomials \(F_1(t), F_2(t), \dots, F_k(t) \in \mathbb Z[t]\) are given and let \(F(t)=F_1(t)\dots F_k(t).\) Let \(\psi(F; x, y)\) be the number of integers \( n\leq x\) such that all prime factors of \(F(n)\) are \(\leq y;\) and let \(\psi(x;y)=\psi(t; x, y)\). It is well known that for any constant \(u>0, \lim_{\substack{ x \rightarrow\infty }} \frac{1}{x} \psi (x; x^{1/u})\) exists and equals \(\rho (u)\), the \`\` Dickman-De Bruijn function''. The author here conjectures that if each \(F_j(t)\) has degree \(d_j\) then \(\lim_{\substack{ x \rightarrow\infty }}\frac{1}{x}\psi (F; x, x^{1/u})\) exists and equals \(\rho (d_1 u)\rho (d_2 u)\dots\rho(d_k u)\); in other words that \(F_j(n)\) should be as often \`\` \(n^{1/u}\)-smooth'' as typical integers of the same size, and that the property that each given \(F_j (n)\) is \`\` \(n^{1/u}\)-smooth'' is independent from the others. Martin shows that this is true for all \(u\leq 1/(d-1/k +\epsilon)\), where \(d=\max_j d_j\) and \(k\) is the number of \(j\) with \(d_j=d\), assuming a reasonable uniform version of a well-known conjecture as to how often \(F_1(n), \dots, F_k(n)\) are simultaneously prime. The range here is only a little better than the trivial \(u\leq 1/d\) but nonetheless provides the first compelling evidence for Martin's \`\` smoothness'' conjecture.