An inequality of Frank, Steinmetz and Weissenborn (Q655531)

The inequality referred to in the title of this paper is given by \[ N(r, F) \leq N(r, 1/F) + (2 + \varepsilon)N(r, f) + S(r, f), \quad \varepsilon>0, \] which was used by \textit{G. Frank} [Math. Res. 53, 39--48 (1989; Zbl 0697.30032)], \textit{G. Frank} and \textit{G. Weissenborn} [Bull. Lond. Math. Soc. 18, 29--33 (1986; Zbl 0586.30025)], and \textit{N. Steinmetz} [Bull. Lond. Math. Soc. 20, No. 5, 525--531 (1988; Zbl 0661.30025)] to find the direct relation of the frequency of poles of a function \(f\) meromorphic in the plane to the frequency of zeros of a linear differential polynomial \(F:=L(f)\) in that function \(f\) with small coefficients \(a_j\), where \[ L(z) = \left(\frac{d}{dz}\right)^p + a_{p-1}\left(\frac{d}{dz}\right)^{p-1} + \cdots + a_0(z). \] Here \(S(r, f)\) denotes any quantity which is \(o\bigl(T(r, f)\bigr)\) as \(r\) tends to infinity outside a set of finite measure and independent of \(\varepsilon\). Each function \(a_j\) is said to be small with respect to \(f\) when it is meromorphic in the plane and when its Nevanlinna characteristic \(T(r, a_j)\) is \(S(r, f)\). One also assumes \(L(f)\not\equiv 0\) for the estimate. The method of Frank, Weissenborn and Steinmetz is deeply related to the proof of \textit{N. Steinmetz} of the second fundamental theorem of Nevanlinna for small functions [J. Reine Angew. Math. 368, 134--141 (1986; Zbl 0598.30045)]. The paper under review also applies this approach to generalize their estimate cited above from the following two points of view: (i) the coefficients \(a_j\) do not need to be small with respect to \(f\) anymore, and (ii) the frequency of the zeros of \(F=L(f)\) can ignore counting multiplicities, that is, \(N(r,1/F)\) can be replaced by \(\overline{N}(r,1/F)\), which estimates the frequency of the distinct zeros of \(F\). The author notes that ``the constants which arise are not so easy to control, and it is necessary to keep track of the orders of the differential operators which appear in the proof'' for this purpose. Then he gives his generalized inequality \[ N(r, F) \leq C\overline{N}(r, 1/F) + (1 + A)N(r, f) + S^{\ast}(r, F) \] with two positive constants \(A\) and \(C\) given arbitrarily and controlled by \(A\), respectively. Here, \(S^{\ast}(r, F)\) denotes any quantity which is \(O\bigl(\log^{+} T(r, F) +\log r + \sum_{j=0}^{p-1}T(r,a_j)\bigl)\) as \(r\) tends to infinity outside a set of finite measure and independent of the constant \(A\). Corollary 1.1 is an interesting consequence of the generalization for (ii): Let \(p \geq 2\) and let the function \(f\) be transcendental and meromorphic in the plane. Assume that \(\overline{N}\bigl(r, 1/f^{(p)}\bigr) = S(r, f)\) and that all but finitely many poles of \(f\) have multiplicity at most \(p-1\). Then \(N(r, f) = S(r, f)\). Besides my recommendation for this paper, two things could be emphasized here. (1) A key role in the method of Frank and Steinmetz is played by the combination of a Wronskian identity and a result by \textit{R. Spigler} [J. Math. Anal. Appl. 98, 130--147 (1984; Zbl 0476.34008)] about linear differential equations whose solutions are the products of solutions of two given differential equations. An essential reasoning of the present paper seems to be made for the estimate of the ratio of dimensions of some vector spaces over the complex number field generated by all elements of these products of fundamental solutions. The constant \(A\) acts here as the constant \(1\) does in the proof of Steinmetz for the small \(a_j\). (2) As a concluding remark, it is shown that when \(\varepsilon = A-1\) is small and positive, a better bound for the constant \(C\) is obtained so that we can replace \(N(r,1/F)\) appearing in the inequality of of Frank, Steinmetz and Weissenborn by the term \(2{\varepsilon}^{-p}\overline{N}(r,1/F)\).