Symmetry factors for differential equations with applications to orthogonal polynomials (Q810214)

In an earlier paper the author [[1] Ann. Mat. Pura Appl., IV. Ser. 138, 35-53 (1984; Zbl 0571.34003)] developed a method to construct an orthogonalizing measure for a sequence of orthogonal polynomials \(\{\phi_ m\}^{\infty}_{m=0}\) which satisfy a differential equation of the form \[ (1)\quad \sum^{2n}_{j=0}a_ j(x)y^{(j)}(x)=\lambda_ my(x)\quad (m=0,1,...). \] For example, the classical differential equations of Jacobi, Laguerre and Hermite (for which, of course, the measures are known) satisfy equations of the form (1) with \(n=1\). There are numerous equations of the form (1) with \(n>1\) which have sequences of orthogonal polynomial solutions (see [1] for some examples); this method allows for an effective construction of the corresponding orthogonalizing measures in each of these cases. Basically, to construct the measure associated with the polynomial solutions to (1) by the author's constructive technique, one must simultaneously solve a system of n distributional differential equations. These equations, called the symmetry equations associated with the expression \[ L[y]:=\sum^{2n}_{j=0}a_ j(x)y^{(j)}(x), \] are of odd orders 1,3,...,2n-1. The simultaneous classical solution of these symmetry equations produces an integrating (symmetry) factor which allows L[\(\cdot]\) to be put into formally symmetric form. One drawback that this method has is that these differential equations involve the Bernoulli numbers and they are somewhat cumbersome to use. In the present paper, the author replaces the n symmetry equations associated with L[\(\cdot]\) by an equivalent system of symmetry equations of odd orders 1,3,...,2n-1. These equations do not involve the Bernoulli numbers and are easier to use in practice. A complete proof of the equivalence of these two sets of symmetry equations can be found in a recent paper by the author and \textit{D. Race} [Proc. Lond. Math. Soc., III. Ser. 60, 344-364 (1990; Zbl 0685.33011)]. Recently, in a very important paper [Bull. Lond. (to appear)] \textit{K. H. Kwon}, \textit{S. S. Kim} and \textit{S. S. Han} used the author's constructive method to construct a real orthogonalizing measure of bounded variation for the Bessel polynomials. This concludes a fifty year search for such a measure.