Discriminants of classical quasi-orthogonal polynomials with application to Diophantine equations (Q2330990)

The authors give explicit formulae for the \textit{discriminant\/} of classical \textit{quasi-orthogonal polynomials\/}, generalizing the result given by \textit{K. Dilcher} and \textit{K. B. Stolarsky} for the quasi-orthogonal Chebyshev polynomials of the second kind [Trans. Am. Math. Soc. 357, No. 3, 965--981 (2005; Zbl 1067.12001)]. The main results are: Theorem 3.1. (quasi-Jacobi polynomials) Let \(c\) be a constant and let \(P_{n,c}^{(\alpha,\beta)}(x) = P_{n}^{(\alpha,\beta)}(x)+cP_{n-1}^{(\alpha,\beta)}(x)\). Then \[ \begin{split} \hbox{disc}(P_{n,c}^{(\alpha,\beta)})=&\frac{(2n+\alpha+\beta)^{2n+1}}{2^{n(n-1)}} \prod_{k=1}^n k^{k-2n+3} \cdot\prod_{k=1}^{n-1}(k+\alpha)^{k-1}(k+\beta)^{k-1}(n+k+\alpha+\beta)^{n-k-1}\\ &\cdot\frac{(-c)^nP_{n,c}^{(\alpha,\beta)}(-(2n(n+\alpha+\beta)c^2 + (\alpha^2-\beta^2)c+2(n+\alpha)(n+\beta))/(2n+\alpha+\beta))^2c)}{(n+\alpha+cn)(n+\beta-cn)}.\end{split} \] Furthermore, \(\hbox{disc}(P_{n,c}^{(\alpha,\beta)}\) is a polynomial in \(c\) of degree \(2(n-1)\). Theorem 3.6 (quasi-Laguerre polynomials) Let \(c\) be a constant and let \(L_{n,c}^{\alpha)}(x)= L_{n}^{\alpha)}(x)+cL_{n-1}^{\alpha)}(x)\). Then \[ \hbox{disc}(L_{n,c}^{\alpha)}= \frac{1}{n+\alpha+cn}\prod_{k=1}^n (k+\alpha)^{k-1}\\ \cdot (-1)^nL_{n,c}^{\alpha)}\left(\frac{nc^2+(2n+\alpha)c+n+\alpha)}{c}\right). \] Theorem 3.8 (quasi-Hermite polynomials) Let \(c\) be a constant and let \(H_{n,c}(x)= H_{n}(x)+cH_{n-1}(x)\). Then \[\hbox{disc}(H_{n,c})=2^{n(3n-5)/2}\prod_{k=1}^{n-1} k^k \cdot H_{n,c}\left(-\frac{c^2+2n}{2c}\right).\] Here the \textit{discriminant\/} of a polynomial \(f(x)=a_0x^n+\cdots+a_n\) of degree \(n\), is defined as \[\hbox{disc}(f)=\frac{(-1)^{n(n-1)/2}}{a_0}\hbox{Res}(f,f'),\] where the \textit{resultant\/} of two polynomials \(f(x)=a_0x^n+\cdots+a_n\) resp. \(g(x)=b_0x^m+\cdots+b_m\) of fegree \(n\) resp. \(m\) follows from the order \((n+m)\times (n+m)\) determinant where the rows, each with \(n+m\) entries, are given by the coefficients of the polynomials \[\left|\begin{matrix} a_0 & a_1 & \ldots & a_n & & \cr & \ddots & & & \ddots & \cr & & a_0 & a_1 &\ldots a_n \cr b_0 & b_1 & \ldots & b_m & & \cr & \ddots & & & \ddots & \cr & & b_0 & b_1 & \ldots & b_m \cr \end{matrix}\right|.\] After giving background and an historical overview, the proofs follow. Moreover, the authors give a section of \(8\) pages with the title \textbf{Applications}: -- a generalization of the Hausdorff equations \[\sum_{i=1}^{2r+1}x_iy_i^j=\frac{1}{\sqrt{\pi}}\int_{-\infty}^{\infty}\,t^j e^{-t^2}dt,\ j=0,1,\ldots,2r+1)\] has rational solutions \(x_i,\,y_i\ (1\leq i\leq 2r+1)\), -- Hausdorff type equations for rational solutions, -- results of the solutions of these afore mentioned solutions to be the zeros of quasi-orthogonal polynomials or not, -- existence of rational points on certain curves generated by the discriminant. An interesting paper, covering quite some ground.