Bank-Laine functions, the Liouville transformation and the Eremenko-Lyubich class (Q1996442)

The Bank-Laine conjecture states that if \(A\) is a transcendental entire function of finite order \(\rho(A)\), \(f_1\), \(f_2\) are lineary independent solutions of \[ f'' + A(z)f=0 ,\tag{1}\] and the exponent of convergence \(\lambda(f_1f_2)\) is finite, then \(\rho(A)\in \mathbb{N}\). This conjecture was recently disproved in the papers of \textit{W. Bergweiler} and \textit{A. Eremenko} [J. Eur. Math. Soc. (JEMS) 19, No. 6, 1899--1909 (2017; Zbl 1381.34111); J. Anal. Math. 137, No. 2, 751--812 (2019; Zbl 1419.34235)] using quasiconformal techniques. Let \(T(r,f)\) denote the Nevanlinna characteristic of \(f\), and \(n(r, 1/f)\) the number of zeros of \(f\) in \(\{z: |z|\le r\} \). The first main result of the paper shows that the conjecture is true if the coefficient \(A\) is from the Eremenko-Lubich class \(\mathcal{B}\) of functions with a finite set of critical and asymptotic values. Theorem 1.3. Suppose that \(A\) is a transcendental entire function in \(\mathcal{B}\), and let \(E=f_1f_2\), where \(f_1\), \(f_2\) are linearly independent solutions of (1). Then exactly one of the following holds. (A) The functions \(A\) and \(E\) satisfy \(\rho(A)=\rho(E)=1\) and \[ T(r,A)+T(r,E)=O(r), \quad r\to +\infty,\] (B) There exists \(d>0\) such that the zeros of \(E\) satisfy \[ n(r, 1/E )>e^{dr^\frac 12}, \quad r\to +\infty, \] and in particular \(\rho(E)=\lambda(E)=+\infty\). An example shows that the exponent \(1/2\) in statement B is sharp. The second main result concerns the zeros of Bank-Laine functions, that is, entire functions \(E\) such that \(E(z)=0\) implies \(E'(z)=\pm 1\). The author obtains a sharp lower bound for the exponent of convergence \(\lambda(E)\) for Bank-Laine functions with positive zeros, improving a result from [\textit{D. Drasin} and the author, Lond. Math. Soc. Lect. Note Ser. 348, 165--178 (2008; Zbl 1158.30019)]. Theorem 1.5. Let \(E\) be a real Bank-Laine function of finite order, with infinitely many zeros, all real and positive. Then the exponent of convergence \(\lambda(E)\) of \(E\) is at least \(3/2\). Moreover, if \(\lambda(E)=\frac 32\) then \(E\) and the assocoated coefficient function \(A\) have order \(\rho(E)=\rho(A)=\frac 32\). An example shows that the exponent \(3/2\) in Theorem 1.5 is attained. To prove Theorems 1.3 and 1.5 a refinement of Hille's method and quasicnformal surgery are used, respectively.