Algebraic osculation and application to factorization of sparse polynomials (Q434417)

Is quite of interest to find good algorithms for factorizations of a form \(f \in K[t_1,t_2]\), where \(K\subset {\mathbb C}\) is a number field. The main existing algorithms present problems when \(f\) is sparse (i.e. when there are many zeros as coefficients in a monomial expression of \(f\)), in fact they are not able to use the amount of information encoded in the Newton polytope \(N_f\) associated to \(f\). In this paper a new algorithm is found which allows to factorize sparse polynomials using the geometry of their Newton polytope. The way to achieve this result is to consider an embedding of the complex affine curve given by \(\{ f=0\}\subset {\mathbb C}^2\) into a compactification \(X\) of \({\mathbb C}^2\). If \(X\) is well-chosen, \(N_f\) can be recovered from the Picard class of a compactification \(C \subset X\) of \(\{ f=0\}\). Let \(\partial X\) be \(X \backslash {\mathbb C}^2\), then Pic(\(\partial X\)) = Pic(\(X\)), so the idea is to use \(C|_D\), where \(D\) is a divisor with support in \(|\partial X|\). The main result of the paper is a theorem which detects the irreducible components of \(C\) via necessary and sufficient conditions for a Cartier divisor on \(D\) to extend to \(C\). This ``osculation criterion'' is expressed via residues.