On totally reducible binary forms. I (Q5949621)

Let \(f(x,y)\) denote a binary form of degree \(l \geq 3\) that is a product of linear factors with integer coefficients. For such forms the author obtains a result of the type previously obtained for irreducible cubic forms in [J. Reine Angew. Math. 521, 185-240 (2000; Zbl 0989.11019)] and for certain forms of higher degree in the references quoted therein. This is to the effect that almost all the numbers represented by such a form have essentially only one representation. In fact, if \(\nu(n)\) denotes the number of positive integers not exceeding a large number \(n\) that have essentially more than one representation, then \(\nu(n) \ll n^{2/l-\eta+\varepsilon}\) for each \(\varepsilon>0\), with \(\eta = {1\over 9}\) if \(l=3\), but \(\eta=(l-2)/l^2(l-1)\) if \(l>3\). An essential tool in the proof is a theorem of \textit{E. Bombieri} and \textit{J. Pila} [Duke Math. J. 59, 337-357 (1989; Zbl 0718.11048)] on the number of integer points on curves. The meaning of ``essentially'' is that representations \(f({\mathbf u})=f({\mathbf v})\) of a number are said to be essentially the same if \({\mathbf u}=A{\mathbf v}\) where \(A\) is a rational automorphism of the form, that is to say an integer matrix \(A\) such that \(f(A{\mathbf x})=f({\mathbf x})\) for all \(\mathbf x\). A corollary of results of this type is an asymptotic formula for the number of integers not exceeding \(n\) that are representable by the form. In this instance such a corollary is deferred, because it depends on a study of the automorphisms of the form that is not needed for the result in this paper. It is also planned to extend the results to the case where the linear factors need not be homogeneous.