Parametrizing the solutions of an analytic differential equation (Q1909535)

Recall that a holomorphic function \(f\) (defined in some non-empty open connected subset of the complex plane) is said to be differentially algebraic if it is solution of a differential equation \((*)\) \(P(z, f, \dots, f^{(m)})= 0\), where \(P\) is some nonzero polynomial. The aim of this paper is to prove that, given a sequence \((z_n)\) in \(\mathbb{C}^{\mathbb{N}}\), the set of sequences \((w_n)\) of \(\mathbb{C}^N\) for which there exists an entire function \(f\) such that \(f(z_n)= w_n\) for all \(n\) is meagre and hence is strictly contained in \(\mathbb{C}^{\mathbb{N}}\). The proof is based on a result of Malgrange. A direct proof of an adapted version of this latter is given in the first part of the paper. More precisely, it is proved that solutions in the disk \(D(R)= \{z; |z|< R\}\) of the differential equation \((*)\) can be parametrized by a countable set of holomorhic functions \(F_n: U_n\times D(r)\to \mathbb{C}\) with \(U_n\) open in \(\mathbb{C}^{k(n)}\), \(0\leq n\leq m\) (namely there exists \(n\) and \(a\) in \(U_n\) such that \(f(z)= F(a, z)\) for \(|z|< r\)). After linearization of the equation \((*)\) this is a consequence of the index theorem for linear differential equations: any differential operator \(L\) of \(\mathbb{C}[z, {d\over dz}]\) has a finite-dimensional cokernel as operator on the vector space of functions holomorphic in \(D(R)\).