On the dependent property of solutions for general higher order periodic differential equation (Q663630)

Consider the higher order periodic differential equation \[ f^{(k)}+A_{k-2} f^{(k-2)}+\dots+ A_{0} f=0, \tag{*} \] where \(A_0, \dots, A_{k-2}\) are entire functions of period \(2\pi i\). Let \(N(r, 1/f)\) be the Nevanlinna counting function of the zeros of an analytic function \(f\), \(T(r,f)\) is its Nevanlinna characteristic. Denote \[ \sigma(f)=\limsup _{r\to+\infty} \frac{\log^+ T(r,f)}{\log r}, \quad \sigma_e(f)=\limsup _{r\to+\infty} \frac{\log^+ T(r,f)}{r}, \] \[ \lambda_e(f)=\limsup _{r\to+\infty} \frac{\log^+ N(r,1/f)}{r}. \] \textit{S. Bank} and \textit{J. Langley} [Comment. Math. Univ. St. Pauli 41, No. 1, 65--85 (1992; Zbl 0764.34024)] have proved Theorem A. Suppose \(k\geq 2\), \(A_0, \dots, A_{k-2}\) are entire functions of period \(2\pi i\), and that \(f\) is a non-trivial solution of the differential equation \((*)\). Suppose further that \(f\) satisfies \(\log^+ N(r,1/f)=o(r) \) \((r\to+\infty)\), that \(A_0\) is non-constant and rational in \(e^z\), and that if \(k\geq 3\), then \(A_1, \dots, A_{k-2}\) are constants. Then there exists an integer \(q\) with \(1\leq q\leq k\) such that \(f(z)\) and \(f(z+q2\pi i)\) are linearly dependent. The same conclusion holds if \(A_0\) is transcendental in \(e^z\), and \(f\) satisfies \(\log^+ N(r,1/f)=O(r) \), and if \(k\geq 3\), then as \(r\to+\infty\) we have \(T(r, A_j)=o(T(r,A_0))\) for \(j=1, \dots, k-2\). We note that there are a lot of results due to S. Bank, I. Laine, J. Langley, Shi-An Gao and Yik-Man Chiang which deal with the case \(k=2\). In [Acta Math. Sci., Ser. B, 27, No. 4, 743--752 (2007; Zbl 1150.30023)] \textit{Z. Chen, S. Gao} and \textit{K. Shon} proved Theorem B. Let \(A_j\) \((j=0, \dots, k-2)\) be entire functions of period \(2\pi i\), \(A_j(z)=C_j(e^{-z})+ B_j(e^z)\), where \(C_j\), \(B_j\) are entire functions of finite order of growth. Let \(B_0\) be transcendental with \(\sigma(B_0)< \frac 12\), \(\sigma(B_j)< \sigma(B_0)\) \((j=1, \dots, k-2)\) and \(\sigma(C_s)< \sigma(B_0) \) \((s=0, \dots, k-2)\) if \(\sigma(B_0)>0\); or let \(B_j\) \((j=1, \dots, k-2)\) and \(C_s \) \((s=0, \dots, k-2)\) be polynomials if \(\sigma(B_0)=0\). If \(f(z)\) is a non-trivial solution of \((*)\) and \(\log^+ N(r,1/f)=O(r) \), then \(f(z)\) and \(f(z+2\pi i)\) are linearly dependent. These theorems deal with the case when the coefficient \(A_0\) dominates. The author of the current paper considers the case when \(A_s\), \(s\in \{2, \dots, k-2\}\) is dominant. Theorem 1.6. Let \(k\geq 4\), \(s\in \{2, \dots, k-2\}\), \(A_0, \dots, A_{k-2}\) be entire functions of period \(2\pi i\) satisfying \(\max \{\sigma_e(A_j): j\neq s\}< \sigma_e(A_s)<+\infty\). If \(f(z)\) is a transcendental solution of \((*)\) satisfying \(\lambda_e(f)< \sigma_e(A_s)\), then there exists an integer \(q\) with \(1\leq q\leq k\) such that \(f(z)\) and \(f(z+q2\pi i)\) are linearly dependent. \textbf{Theorem 1.7.} Let \(k\geq 4\), \(s\in \{2, \dots, k-2\}\), \((k-s, s-1)=1\), \(A_0, \dots, A_{k-2}\) be entire functions of period \(2\pi i\) satisfying \(\max \{\sigma_e(A_j): j\neq s\}< \sigma_e(A_s)<+\infty\). Suppose further that \(A_j(z)=B_j(e^z)=B_{j1}(e^{-z})+ B_{j2} (e^z)\) \((j=0, \dots, k-2)\), where \(B_{j1}\), \(B_{j2}\) are entire functions satisfying \(\max \{\sigma(B_{j2}): j\neq s\}< \sigma(B_{s2})<+\infty\). If \(f(z)\) is a transcendental solution of \((*)\) satisfying \(\lambda_e(f)< \sigma_e(A_s)\), then there exisits an integer \(q\) with \(1\leq q\leq k\) such that \(f(z)\) and \(f(z+2\pi i)\) are linearly dependent.