Monotonicity of the zeros of orthogonal polynomials through related measures (Q557022): Difference between revisions

In this quite interesting paper, the authors use known facts about orthogonal polynomials where the associated measures on the real line are related by a linear polynomial factor (in numerator/denominator, of constant sign ), specifically the three term recurrence relation containing one `old' polynomial and two `new' ones, to derive results on the monotonicity of the zeros. Let \(d\phi\) be a determinate measure on \([a,b]\;(-\infty<a<b\leq \infty)\) (for \(b<\infty\) the measure is always determinate) and let \(N,\kappa\) be real numbers with \(0\leq N<\infty,\;\kappa<a\) (or exceptionally \(\kappa\leq a\)). Apart from the monic orthogonal polynomials \(P_n^{(\phi)}\) associated with \(d\phi\), consider the monic orthogonal polynomials \(P_n^{(\phi,\kappa,N)}\) with associated measure given by \[ \int p(x)d\phi^{\kappa,N)}(x)={1\over N+1}\,\left\{Np(\kappa)+{{\mathcal M}^{(\phi,\kappa)}[p] \over {\mathcal M}^{(\phi,\kappa)}[1]}\right\}, \] where \(p\) is an arbitrary polynomial and \[ {\mathcal M}^{(\phi,\kappa)}[p]=\int_a^b\,p(x)(x-\kappa)^{-1}d\phi(x). \] The value \(\kappa=a\) is allowed, only if \({\mathcal M}^{(\phi,\kappa)}[1]\) is convergent. The main result of the paper is then a theorem on the monotonicity behavior of the zeros of the monic orthogonal polynomials: Let \(x_{n,r}^{(\phi,\kappa,N)}\) and \(x_{n,r}^{(\phi)},\;1\leq r\leq n\), be the respective zeros of \(P_n^{(\phi,\kappa,N)}\) and \(P_n^{(\phi)}\), arranged in \textit{decreasing order}. Then \textbf{1.} \(x_{n,r}^{(\phi)}<x_{n,r}^{(\phi,\kappa,N)},\;1\leq r\leq n\) and \(\kappa < x_{n,n}^{(\phi,\kappa,N)}<x_{n,n}^{(\phi)}\). \textbf{2.} \(x_{n,r}^{(\phi,\kappa,N)}\) is a decreasing function of \(N\). \textbf{3.} If \(N=0\), then \(x_{n,r}^{(\phi,\kappa,N)}\) is a decreasing function of \(\kappa\). \textbf{4.} If \(N>0\), then \(x_{n,r}^{(\phi,\kappa,N)}\) is an increasing function of \(\kappa\), provided that \(\kappa\) varies in the range \((-\infty,\check{\kappa}^{(\phi,N)}]\), where \[ \check{\kappa}^{(\phi,N)}=a-{\sqrt{[N(\beta_2^{(\phi)}-a)-(\beta_1^{(\phi)}-a)]^2+4Nv_1^{(\phi)}(a)} - [N(\beta_2^{(\phi)}-a)-(\beta_1^{(\phi)}-a)]\over 2N}. \] Here \(v_1^{(\phi)}=(\beta_1^{(\phi)}-a)(\beta_2^{(\phi)}-a)-\alpha_2^{(\phi)}>0\) and \(\alpha_n^{(\phi)},\,\beta_n^{(\phi)}\) are the coefficients from the three term recurrence relation for the \(P_n^{(\phi)}\). This theorem is then applied to the cases of the Jacobi, Laguerre and Charlier polynomials. The functionals \({\mathcal M}\) are respectively given by \[ {\mathcal M}_J^{(\alpha,\beta,\kappa)}[p]=\int_{-1}^1\,p(x)\,{(1-x)^{\alpha}(1+x)^{\beta}\over x-\kappa}\,dx, \] \[ {\mathcal M}_L^{(\alpha,\kappa)}[p]=\int_0^{\infty}\,p(x)\,{x^{\alpha}e^{-x}\over x+\kappa}\,dx, \] \[ {\mathcal M}_C^{(\alpha,\kappa)}[p]=\int_0^{\infty}\,p(x)\,d\phi_C^{(\alpha)}(x) = \sum_{r=0}^{\infty}\,p(r)\,{1\over r+\kappa}\,{e^{-\alpha}\alpha^r\over r!}. \]