An \(abc\) theorem on the disk (Q616728)

Given a polynomial \(p\) (in one complex variable), write \(\text{deg\,}p\) for the degree of \(p\) and \(\widetilde N(p) =\widetilde N_\mathbb C(p)\) for the number of its distinct zeros in \(\mathbb C\). The so-called \textit{abc} theorem, often referred to as Mason's theorem (but essentially due to Stothers), reads as follows: {Theorem A.} Suppose \(a\), \(b\) and \(c\) are polynomials, not all constant, having no common zeros and satisfying \(a + b = c\). Then \[ \max\big\{\text{deg}\,a, \text{deg}\,b, \text{deg}\,c\big\} \leqslant\widetilde N(abc) - 1.\tag{1} \] One impressive -- and immediate -- application is a simple proof of Fermat's last theorem for polynomials, saying that there are no nontrivial polynomial solutions to the equation \(P^n + Q^n = R^n\) when \(n\geqslant 3\). Besides, it was Theorem A that led (via the classical analogy between polynomials and integers) to the famous \(abc\) conjecture in number theory. In this note, the author presents an \textit{abc} type theorem that applies to a much more general situation. Namely, replace the polynomial equation \(a + b = c\) by \[ f_0+\cdots+f_n = f_{n+1},\tag{2} \] where the \(f_j\)'s are analytic functions on the (closed) disk \(\mathbb D\cup\mathbb T\), where \(\mathbb D=\{z\in\mathbb C: |z|< 1\}\) and \(\mathbb T=\partial\mathbb D\). The functions are thus assumed to be analytic in some open neighborhood of \(\mathbb D\cup\mathbb T\). We associate with each \(f_j\) the (finite) Blaschke product \(B_j\) built from the zeros of the function. This means that \(B_j\) is given by \[ z\mapsto\prod_{k=1}^s\left(\frac{z-a_k}{1-{\overline a}_kz}\right)^{m_k},\tag{3} \] where \(a_k^{} = a_k^{(j)}\) (\( 1\leqslant k\leqslant s =s_j\)) are the distinct zeros of \(f_j\) in \(\mathbb D\), and \(m^{}_k = m_k^{(j)}\) are their multiplicities. Further, let \(\mathbf B\) denote the least common multiple of the Blaschke products \(B_0,\dots, B_{n+1}\) (defined in the natural way), and put \[ \mathcal B:=\text{rad\,}(B_0B_1\dots B_{n+1}). \] Here, we use the notation \(\text{rad\,}(B)\) for the radical of a Blaschke product \(B\). Write \(W = W(f_1,\dots,f_n)\) for the Wronskian of the (analytic) functions \(f_1,\dots,f_n\). Introduce the quantities \[ \kappa=\kappa(W) := \| W'\|_{L^1(\mathbb T)}\| 1/W\|_{L^\infty(\mathbb T)},\quad \lambda =\lambda(W) := \| W'\|_{L^2(\mathbb D)}\| 1/ W\|_{L^\infty(\mathbb T)} \] and \[ \mu = \mu(W):= \big\| W\big\|_{L^\infty(\mathbb T)}\big\| 1/W\big\|_{L^\infty(\mathbb T)}. \] The main result of the paper is the following theorem. {Theorem B.} Suppose \(f_j\) \((j = 0,1,\dots,n + 1)\) are analytic functions on \(\mathbb D\cup\mathbb T\), related by (2) and such that the Wronskian \(W\) vanishes nowhere on \(\mathbb T\). Then \[ N_{\mathbb D}({\mathbf B})\leqslant\kappa+n\mu N_{\mathbb D}(\mathcal B)\quad\text{and}\quad N_{\mathbb D}({\mathbf B})\leqslant\lambda^2+n\mu ^2N_{\mathbb D}(\mathcal B), \] where \(N_{\mathbb D}({\cdot})\) denotes the number of zeros of the function in \(\mathbb D\), counting multiplicities.