Normal families and value distribution in connection with composite functions (Q2573443): Difference between revisions

Using a theorem of \(\cos \pi \rho\) type, Zalcman's lemma and some elegant arguments, the author shows that if \(f\) is a transcendental entire function of order \(< 1/k\) then the analytic functions \(g\) on any domain \(\Omega\) such that \((f \circ g)^{(k)}\) has no zeros in \(\Omega\) form a normal family on \(\Omega\). More generally \((f \circ g)^{(k)}\) can be replaced by \((f \circ g)^{(k)}(z) + a(z) + \sum_{j=0}^{k-1} a_j(z) (f \circ g)^{(j)}(z)\) for \(a, a_0, \ldots, a_{k-1}\) analytic on \(\Omega\). A consequence is that, if \(k \geq 2\), \(f\) is a transcendental entire function of order \(< 1/k\) and \(g\) is nonconstant entire, than \((f \circ g)^{(k)}\) takes every value infinitely often. If \(f\) is transcendental entire of order \(< 1\) and \(g\) is transcendental entire, then \((f \circ g)'\) takes every value infinitely often. The main technical result is that if \(k \in {\mathbb N}\), \(f\) is transcendental entire of order \(< 1/2\), \(P\) a nonconstant polynomial and \(Q\) a polynomial, then \((f \circ P)^{(k)}(z) - Q(z)\) has infinitely many zeros.