Approximate factorization of multivariate polynomials and absolute irreducibility testing (Q1179781)

Let \(\mathbb{C}\) be the field of complex numbers and \(F(x,y,\dots,z)\) a monic polynomia in \(\mathbb{C}\{y,\dots,z\}[x]\). \(F(x,0,\dots,0)\) is supposed to be square-free. The paper suggests a factorization algorithm for such polynomials with approximate complex coefficients. This algorithm can also be used to test the absolute irreducibility of multivariate polynomials. First, the algorithm calculates the roots of \(F(x,y_ 0,\dots,z_ 0)\) numerically, where \(y_ 0,\dots,z_ 0\) are suitable chosen numbers, then, it constructs power series \(F_ 1,\dots,F_ n\) such that \(F(x,y,\dots,z)\equiv F_ 1(x,y,\dots,z)\dots F_ n(x,y,\dots,z)\mod S^{e+2}\), where \(n=\deg_ x(F)\), \(S=(y-y_ 0,\dots,z-z_ 0)\), and \(e=\max\{\deg_ y(F),\dots,\deg_ z(F)\}\). Finally, it finds the approximate divisors of \(F\) as products of elements \(\{F_ 1,\dots,F_ n\}\). In the worst case \(2^{n-1}\) different combinations of \(F_ 1,\dots,F_ n\) have to be checked. A method how to make this algorithm more practical is proposed. However, a detailed analysis of this method for finding the relevant combinations of \(F_ 1,\dots,F_ n\) will be given in another paper.