Regular Functions Satisfying Irregular Ordinary Differential Equations

From MaRDI portal





Consider the linear differential equation \[ u^{(n)} + r_{n-1}u^{(n-1)} +\cdots+r_0u = 0 \tag{1} \] in the complex domain, where the \(r_j\) are rational functions. If all \(r_j\) are polynomials, it is well-known that all solutions to (1) are entire functions. On the other hand, it may happen that (1) admits nontrivial entire solutions although some of the \(r_j\) have poles. For example, the Bessel function of order \(m\) (where \(m\) is a nonnegative integer) is a solution to the equation \[ u + \frac{1}{z}\,u' + \left(1-\frac{m^2}{z^2}\right)u = 0. \] A further example is the equation \[ u - \left(1+\frac{1}{z}\right)u' + \frac{1}{z}\,u = 0. \] Here, all solutions are linear combinations of \(e^z\) and \(z+1\), and thus entire. The authors consider the question whether a given nontrivial entire function \(u\) can satisfy a differential equation of the form (1) with polynomial coefficients \(r_j\). More precisely, consider the second-order equation \[ u = a_0u'+b_0u\,, \tag{2} \] where \(a_0\), \(b_0\) are rational functions, and suppose that \(z=0\) is a so-called regular singular point of (2) which means that \(P_0(z):=za_0(z)\) and \(Q_0(z):=z^2b_0(z)\) are holomorphic in a neighbourhood of \(0\) and that \(p_0:=P_0(0) \neq 0\) and \(q_0:=Q_0(0) \neq 0\). With these notations the main result reads as the following theorem: Let \(u\) be a nontrivial entire function having infinitely many zeros and satisfying the differential equation (2). Furthermore, assume that the indicial equation \[ r(r-1)-p_0r-q_0=0 \tag{3} \] has a root \(r\) which is not a positive integer. Then \(u\) cannot satisfy a differential equation of the form (1) with polynomial coefficients \(r_j\). The authors provide two different proofs of this theorem, one based on the indicial equation and the other using the Frobenius notion of irreducibility. They also give examples that the hypothesis on the zeros of \(u\) and the roots of (3) cannot be omitted. The result applies to the Bessel function of order \(m\) for every positive integer \(m\). Although it is not applicable to the Bessel function of order \(0\), a supplementary argument shows that the conclusion of the theorem still holds for this function.











This page was built for publication: Regular Functions Satisfying Irregular Ordinary Differential Equations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4433238)