Pairs of non-homogeneous linear differential polynomials (Q5894325)

Let \(L:=d^{k}/dz^{k}+\sum_{j=0}^{k-1}a_{j}d^{j}/dz^{j}\) and \(M:=d^{n}/dz^{n}+\sum_{j=0}^{n-1}b_{j}d^{j}/dz^{j}\) be linear differential operators with rational functions as their coefficients such that the differential equations \(L(\omega )=0\), \(M(\omega )=0\) have no common non-trivial local solutions. Then there exist linear differential operators \(P,Q\) such that \(P(L)+Q(M)\) is the identity operator. Moreover, define non-vanishing linear differential polynomials \(F:=L(f)+a\) and \(G:=M(f)+b\), where \(f\) is meromorphic in the complex plane and \(a,b\) are rational functions. Further, define a rational function \(c:=P(a)+Q(b)\) and set \(g:=f+c=P(F)+Q(G)\). Let then \(\Omega\) be a non-empty simply-connected domain such that \(a,b\) and all of \(a_{j},b_{j}\) are analytic on \(\Omega\) and let \(u_{1},\ldots ,u_{k}\) be linearly independent solutions of \(L(\omega )=0\), \(v_{1},\ldots ,v_{n}\) linearly independent solutions of \(M(\omega)=0\), and let \(u,v\) be solution of \(L(\omega )=0\) and \(M(\omega )=0\), respectively. \textit{J. K. Langley} proved in [``Pairs of non-homogeneous linear differential polynomials'', Proc. R. Soc. Edinb., Sect. A, Math. 136, No. 4, 785--794 (2006; Zbl 1112.30025)] that whenever \(f\) is transcendental and meromorphic in the plane and \(N(r,1/F)+N(r,1/G)=S(r,f)\), then either \(F=L(g)\) and \(G=M(g)\), or \(f=R(u_{1},\ldots ,u_{k},v_{1},\ldots ,v_{n},u,v)\) for a rational function \(R\) in \(k+n+2\) variables. In the present paper, the author relaxes the preceding assumption on the counting functions to \(N(r,1/F)\leq\gamma_{1}T(r,f)+S(r,f)\), resp. \(N(r,1/G)\leq\gamma_{2}T(r,f)+S(r,f)\). Then a third alternative appears, namely that for \(\gamma_{0}:=\max\{ \gamma_{1},\gamma_{2}\}\) and \(\gamma_{3}:=\gamma_{1}+\gamma_{2}\), \[ 1\leq\frac{2\gamma_{3}(k+n+1)+1}{k+n+1}+\frac{2\gamma_{3}+1}{k+n}+\gamma_{0}. \] If \(\gamma_{1}=\gamma_{2}=0\), then the third alternative becomes impossible, and the situation reduces back to the preceding result due to Langley. The author also proves that if \(k=n\), then \(\overline{N}(r,1/F)+\overline{N}(r,1/G)=S(r,f)\) suffices to imply the original two-alternative case. Similarly in the special situation when \(b=0\), \(a-L(c)\neq 0\) and \(n>k+2\). The proofs require a refinement of the original arguments due to Langley and non-trivial applications of Nevanlinna theory.