Processing math: 4%

数学物理学报, 2021, 41(4): 921-935 doi:

论文

一类广义的十次Freud-型权函数

王丹,1, 朱孟坤,2, YangChen,1, 王晓丽,2

A Generalised Decic Freud-Type Weight

Wang Dan,1, Zhu Mengkun,2, Yang Chen,1, Wang Xiaoli,2

通讯作者: 朱孟坤, E-mail: zmk@qlu.edu.cn

收稿日期: 2020-08-4  

基金资助: 澳门科学技术发展基金.  FDCT023/2017/A1
澳门大学基金.  MYRG2018-00125FST
国家自然科学基金.  11801292
广东省自然科学基金.  2021A1515010361

Received: 2020-08-4  

Fund supported: the Macau Science and Technology Development Fund.  FDCT023/2017/A1
the Universidade de Macau.  MYRG2018-00125FST
the NSFC.  11801292
the NSF of Guangdong Province.  2021A1515010361

作者简介 About authors

王丹,E-mail:bohewan@126.com , E-mail:bohewan@126.com

YangChen,E-mail:yayangchen@um.edu.mo , E-mail:yayangchen@um.edu.mo

王晓丽,E-mail:wxlspu@qlu.edu.cn , E-mail:wxlspu@qlu.edu.cn

Abstract

In this paper, the authors focus on a generalised decic Freud-type weight function

w(x;t,α)=|x|αex10+tx2,t,xR,α>1,
and study the properties of the orthogonal polynomials with respect to this weight. The difference-differential equations of their associated recurrence coefficients are derived; meanwhile, the authors also find the asymptotic behavior of recurrence coefficients via above mentioned equations. What's more, the authors discuss the Hankel determinant in regard to this weight as n, and calculate the smallest eigenvalues of large Hankel matrices generated by this weight when α=t=0.

Keywords: Orthogonal Polynomials ; Recurrence coefficients ; Hankel determinant ; Smallest eigenvalue

PDF (559KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

王丹, 朱孟坤, YangChen, 王晓丽. 一类广义的十次Freud-型权函数. 数学物理学报[J], 2021, 41(4): 921-935 doi:

Wang Dan, Zhu Mengkun, Yang Chen, Wang Xiaoli. A Generalised Decic Freud-Type Weight. Acta Mathematica Scientia[J], 2021, 41(4): 921-935 doi:

1 引言

正交多项式是在19世纪末由Chebyshev对连分式的研究发展起来的. 随后有越来越多的数学家, Szegö, Bernstein, Akhiezer, Ismail等等, 对它进行研究. 正交多项式有很多有用的性质, 并且广泛应用于数学及数学物理中, 例如近似理论, 随机过程, 随机矩阵. 自20世纪60年代, Géza Freud开始研究正交于实数域R上的指数型权函数e|x|m, 其中m属于自然数集N, 的正交多项式Pn(x)以及相关的循环系数βn的渐近行为. 此类最高次项系数为1的(亦称首一的)正交多项式符合递推关系

xPn(x)=Pn+1(x)+βnPn1(x).

上式中的循环系数βn和Painlevé方程的解有关, 这也是研究正交多项式的一个重要方面. 在应用数学, 计算数学以及物理中, 经常遇到的问题是将上述递推关系中的循环系数βn与权函数的性质联系起来. 这些关系可以转化为特殊的循环系数性质. 相当多的理论研究[1-3]已经涉及了正交性质和循环系数的关系.

Freud[4]给出了关于权函数|x|ρe|x|m, ρ>1, m>0的循环系数βn的一个猜想

lim
(1.1)

并证明了当 m = 2, \; 4, \; 6 时猜想成立, 于是大胆猜测对于任意的正偶数 m , (1.1) 式都成立. 这个猜想在1985年被Magnus[5]证明. 随后, Lubinsky[6]又给出了一个更广义的证明.

关于Freud - 型权函数 {w}(x;t) = {\rm e}^{-x^4+tx^2}, \; \; t, x\in{{\mathbb R}} , Magnus[7]指出循环系数 \beta_n 所满足的非线性离散方程 \beta_n(\beta_{n-1}+\beta_n+\beta_{n+1})+2t\beta_n = n, 正是一个离散的二阶Painlevé方程. 后来, Clarkson[8-9]又研究了一个更为广义的Freud - 型权函数

{w}(x;t, \lambda) = |x|^{2\lambda+1}{\rm e}^{-x^4+tx^2}, \; \; t, \; x\in{{\mathbb R}} , \; \; \lambda>-1,

它所对应的循环系数 \beta_n 还可以用含有抛物柱面函数的Wronskian行列式表示. 最近, 王丹等[19]又讨论了一类六次的Freud - 型权函数

{w}(x;t, \lambda) = |x|^{\alpha}{\rm e}^{-x^6+tx^2}, \; \; x, \; t\in{{\mathbb R}} , \; \; \alpha>-1.

朱孟坤等在文献[27]中构造了一类介于高斯权函数和四次Freud - 型权函数之间的权函数

{w}(x;s, \alpha, N) = |x|^{\alpha}{\rm e}^{-N\left[x^2+s\left(x^4-x^2\right)\right]}, \; x\in{{\mathbb R}} , \; \; \alpha>-1, \; \; s\in[0, 1], \; \; N>0,

得到相关的循环系数满足第一类离散的Painlevé方程, 并讨论了该方程的解. 另外, 这篇文章中建立了正交于该权函数的多项式与一类双合流的Heun方程的关系. 这对正交多项式与Heun方程的交叉研究提供了一个思路. 更多关于Freud - 型权函数的研究, 可参考文献[10-13, 15].

\{P_n(x)\} 为最高次项系数和次数分别为1 (亦称首一的)和 n 的多项式, 并且正交于一个十次的Freud - 型权函数

\begin{equation} {w}(x;t, \alpha) = |x|^{\alpha}{\rm e}^{-x^{10}+tx^2}, \; \; t, \; x\in{{\mathbb R}} , \; \alpha>-1. \end{equation}
(1.2)

根据正交条件, 可知

\begin{equation} \int_{{{\mathbb R}} }P_m(x)P_n(x){w}(x;t, \alpha){\rm d}x = h_n(t, \alpha)\delta_{(m, n)}, \end{equation}
(1.3)

这里, h_n(t, \alpha)>0 , \delta_{(m, n)} 是指克罗内克 \delta 函数, 即

\delta_{(m, n)} = \left\{ \begin{array}{ll} 0&, \; \; m\neq n, \\ 1&, \; \; m = n. \end{array} \right.

对正交于Freud - 型权函数(1.2)的多项式而言, 它们的三项循环递推关系式是

\begin{equation} xP_n(x) = P_{n+1}(x)+\beta_nP_{n-1}(x), \; \; n = 0, \; 1, \; 2, \cdots, \end{equation}
(1.4)

其中

\begin{equation} \beta_n(t, \alpha) = \frac{1}{h_{n-1}(t, \alpha)}\int_{{{\mathbb R}} }xP_{n-1}(x)P_n(x){w}(x;t, \alpha){\rm d}x; \end{equation}
(1.5)

h_n 表示 P_n L^2 -范数的平方. 当 n = 0 1 时, 规定

\beta_0(t, \alpha)P_{-1}(x) = 0, {\quad} P_0(x) = 1.

另外, P_n(x) 的展开式[10] P_{2k}(z) = z^{2k}+{\cal P}(2k; t)z^{2k-2}+\dots+P_{2k}(0)

\begin{eqnarray*} P_{2k+1}(z)& = &z^{2k+1}+{\cal P}(2k+1;t)z^{2k-1}+\dots+Cz\\ & = &z\Big(z^{2k}+{\cal P}(2k+1;t)z^{2k-2}+\dots+C\Big). \end{eqnarray*}

将(1.4) 式带入(1.5) 式, 可以得到 \beta_n(t, \alpha) 的另外一种表述

\begin{equation} \beta_n(t, \alpha) = \frac{h_n(t, \alpha)}{h_{n-1}(t, \alpha)}. \end{equation}
(1.6)

关于十次Freud - 型权函数(1.2) 的矩定义为

\begin{equation} \mu_k(t, \alpha) = \int_{{{\mathbb R}} }x^k|x|^{\alpha}{\rm e}^{-x^{10}+tx^2}{\rm d}x, \; \; k = 0, \; 1, 2, \cdots, \end{equation}
(1.7)

它所决定的Hankel行列式记为 D_n(t, \alpha) . 根据正交多项式的广义定理[14, 16-18], D_n (t, \alpha) 还有两种表现形式

\begin{equation} D_n(t, \alpha) = \det\Big(\mu_{j+k}(t, \alpha)\Big)_{j, k = 0}^{n-1} = \prod\limits_{j = 0}^{n-1}h_j(t, \alpha), \; \; n\geq 1, \end{equation}
(1.8)

这里 D_0 = 1 , D_{-1} = 0 . 此外, \beta_n(t, \alpha) 也可以用 D_n(t, \alpha) 表示[12]

\begin{equation} \beta_n(t, \alpha) = \frac{D_{n+1}(t, \alpha)D_{n-1}(t, \alpha)}{D_n^2(t, \alpha)}. \end{equation}
(1.9)

近年来, 较大病态Hankel矩阵的最小特征值的渐近行为成为一个研究热点, 具体内容可参考文献[28-38], 当 \alpha = t = 0 时, (1.7) 式所决定的Hankel矩阵的最小特征值在本文中也有讨论.

Hankel矩阵是随机矩阵中最基本的研究对象之一. 其行列式可以代表一个特定随机矩阵的配分函数, 或者可以代表一个随机变量与系综相关的母函数, 亦或者与最大特征值的分布有关. 在给定一个权函数的前提下, 最大或最小特征值的研究又可以提供有关Hankel矩阵非常有用的信息.

在本文第2、3节, 我们将分别给出权函数(1.2)所对应的循环系数 \beta_n 的行列式表示, 和它所符合的一个非线性离散方程. 基于第3节结果, 我们在第四部分给出了 \beta_n 的一个近似值. 作为第2–4节的应用, 我们在第5、6节分别得到了其所对应的Hankel行列式满足的一个方程, 和当 t = \alpha = 0 时, 得到对应Hankel矩阵的最小特征值的渐近行为.

2 循环系数的行列式表达

定义两个新的Hankel行列式

\begin{eqnarray} \widehat{D}_n(t, \alpha) = \det\Big(\mu_{2j+2k}(t, \alpha)\Big)_{j, k = 0}^{n-1} = \left|\begin{array}{ccccc} \mu_0(t, \alpha) &\mu_2(t, \alpha)& \cdots&\mu_{2n-2}(t, \alpha) \\ \mu_2(t, \alpha) & \mu_4(t, \alpha) & \cdots &\mu_{2n}(t, \alpha)\\ \vdots & \vdots &\ddots&\vdots\\ \mu_{2n-2}(t, \alpha)&\mu_{2n}(t, \alpha)&\cdots &\mu_{4n-4}(t, \alpha) \end{array}\right| \end{eqnarray}
(2.1)

\begin{eqnarray} \widetilde{D}_n(t, \alpha) = \det\Big(\mu_{2j+2k+2}(t, \alpha)\Big)_{j, k = 0}^{n-1} = \left|\begin{array}{ccccc} \mu_2(t, \alpha) &\mu_4(t, \alpha)& \cdots&\mu_{2n}(t, \alpha) \\ \mu_4(t, \alpha) & \mu_6(t, \alpha) & \cdots &\mu_{2n+2}(t, \alpha)\\ \vdots & \vdots &\ddots&\vdots\\ \mu_{2n}(t, \alpha)&\mu_{2n+2}(t, \alpha)&\cdots &\mu_{4n-2}(t, \alpha) \end{array}\right|, \end{eqnarray}
(2.2)

它们和(1.8)式可以建立起这样的联系

\begin{equation} D_{2n}(t, \alpha) = \widehat{D}_n(t, \alpha)\widetilde{D}_n(t, \alpha), \; \; \; D_{2n+1}(t, \alpha) = \widehat{D}_{n+1}(t, \alpha)\widetilde{D}_n(t, \alpha). \end{equation}
(2.3)

与此同时, Freud -型权函数(1.2)作为定义在实数范围内恒正的对称函数, 所以 \widehat{D}_n(t, \alpha) \widetilde{D}_n(t, \alpha) 还可用Wronskian行列式表示.

引理2.1  设 W(\psi_1, \psi_2, \cdots, \psi_n) 是Wronskian行列式, 对于Freud - 型权函数(1.2),

\begin{equation} \widehat{D}_n(t, \alpha) = W\bigg(\mu_0(t, \alpha), \frac{{\rm d}\mu_0(t, \alpha)}{{\rm d}t}, \cdots, \frac{{\rm d}^{n-1}\mu_0(t, \alpha)}{{\rm d}t^{n-1}}\bigg) \end{equation}
(2.4)

\begin{equation} \widetilde{D}_n(t, \alpha) = W\bigg(\frac{{\rm d}\mu_0(t, \alpha)}{{\rm d}t}, \frac{{\rm d}^2 \mu_0(t, \alpha)}{{\rm d}t^2}, \cdots, \frac{{\rm d}^{n}\mu_0(t, \alpha)}{{\rm d}t^{n}}\bigg) \end{equation}
(2.5)

成立.

  对于任一个非负整数 k , 有

\frac{{\rm d}^k\mu_0(t, \alpha)}{{\rm d}t^k} = \frac{{\rm d}^k}{{\rm d}t^k}\int_{{{\mathbb R}} } {w}(x;t, \alpha){\rm d}x = \int_{{{\mathbb R}} }x^{2k}|x|^{\alpha}{\rm e}^{-x^{10}+tx^2}{\rm d}x = \mu_{2k}(t, \alpha),

再结合 \widehat{D}_n(t, \alpha) \widetilde{D}_n(t, \alpha) 的定义(2.1) 和(2.2), 即可得证(2.4) 和(2.5)式.

注2.1

\mu_{2k+1}(t, \alpha) = 0.

借助Mathematica软件, 我们可以用超几何函数 {}_1F_4(a_1; b_1, b_2, b_3, b_4; z) 表示0 -阶矩

\begin{eqnarray} \mu_0(t, \alpha)& = &\int_{{{\mathbb R}} }|x|^{\alpha}{\rm e}^{-x^{10}+tx^2}{\rm d}x = 2\int_0^{\infty}|x|^{\alpha}{\rm e}^{-x^{10}+tx^2}{\rm d}x\\ & = &\frac{1}{5}\Gamma\Big(\frac{1+\alpha}{10}\Big)\; _1\textbf{F}_4\Big(\frac{1+\alpha}{10};\frac{1}{5}, \frac{2}{5}, \frac{3}{5}, \frac{4}{5};\frac{t^5}{5^5}\Big) \\ && +\frac{t}{5}\Gamma\Big(\frac{3+\alpha}{10}\Big)\; _1\textbf{F}_4\Big(\frac{3+\alpha}{10};\frac{2}{5}, \frac{3}{5}, \frac{4}{5}, \frac{6}{5};\frac{t^5}{5^5}\Big)\\ &&+\frac{t^2}{10}\Gamma\Big(\frac{5+\alpha}{10}\Big)\; _1\textbf{F}_4\Big(\frac{5+\alpha}{10};\frac{3}{5}, \frac{4}{5}, \frac{6}{5}, \frac{7}{5};\frac{t^5}{5^5}\Big) \\ && +\frac{t^3}{30}\Gamma\Big(\frac{7+\alpha}{10}\Big)\; _1\textbf{F}_4\Big(\frac{7+\alpha}{10};\frac{4}{5}, \frac{6}{5}, \frac{7}{5}, \frac{8}{5};\frac{t^5}{5^5}\Big)\\ & &+\frac{t^4}{120}\Gamma\Big(\frac{9+\alpha}{10}\Big)\; _1\textbf{F}_4\Big(\frac{9+\alpha}{10};\frac{6}{5}, \frac{7}{5}, \frac{8}{5}, \frac{9}{5};\frac{t^5}{5^5}\Big). \end{eqnarray}
(2.6)

图 1

图 1   左图是当 \alpha = -\frac{1}{3}, 0, \frac{3}{2} 3 时, \mu_0(t, \alpha) 图像; 右图是当 t = -4, 0, \frac{1}{2} 1 时, \mu_0(t, \alpha) 图像


引理2.2  关于Freud - 型权函数(1.2)的Hankel行列式 \widetilde{D}_n(t, \alpha) \widehat{D}_n(t, \alpha) 具有如下关系

\begin{equation} \widetilde{D}_n(t, \alpha) = \widehat{D}_n(t, \alpha+2). \end{equation}
(2.7)

   \forall m\in [0, 2n-2] , 有

\begin{eqnarray*} \mu_{2m+2}(t, \alpha)& = &\int_{{{\mathbb R}} }x^{2m+2}|x|^{\alpha}{\rm e}^{-x^{10}+tx^2}{\rm d}x = 2\int_0^{\infty}x^{2m+2+\alpha}{\rm e}^{-x^{10}+tx^2}{\rm d}x\\ & = &2\int_0^{\infty}x^{2m}|x|^{\alpha+2}{\rm e}^{-x^{10}+tx^2}{\rm d}x = \int_{{{\mathbb R}} }x^{2m}|x|^{\alpha+2}{\rm e}^{-x^{10}+tx^2}{\rm d}x\\ & = &\mu_{2m}(t, \alpha+2), \end{eqnarray*}

对照(2.1) 和(2.2) 式, 即可得证(2.7)式.

定理2.1   正交于Freud - 型权函数(1.2)的首一的正交多项式 P_n(x) 满足

\begin{equation} P_{n+1}(x) = xP_n(x)-\beta_n(t, \alpha)P_{n-1}(x), \end{equation}
(2.8)

这里

\begin{equation} \beta_{2n}(t, \alpha) = \frac{\widehat{D}_{n+1}(t, \alpha)\widehat{D}_{n-1}(t, \alpha+2)}{\widehat{D}_n(t, \alpha)\widehat{D}_n(t, \alpha+2)} = \frac{\rm d}{{\rm d}t}\ln\frac{\widehat{D}_n(t, \alpha+2)}{\widehat{D}_n(t, \alpha)}, \end{equation}
(2.9)

\begin{equation} \beta_{2n+1}(t, \alpha) = \frac{\widehat{D}_{n}(t, \alpha)\widehat{D}_{n+1}(t, \alpha+2)}{\widehat{D}_{n+1}(t, \alpha)\widehat{D}_n(t, \alpha+2)} = \frac{\rm d}{{\rm d}t}\ln\frac{\widehat{D}_{n+1}(t, \alpha)}{\widehat{D}_n(t, \alpha+2)}, \end{equation}
(2.10)

并且 P_0(x) = 1 , \beta_0(t, \alpha)P_{-1}(x) = 0 .

  根据(1.9)和(2.3)式, 有

\beta_{2n}(t, \alpha) = \frac{\widehat{D}_{n+1}(t, \alpha)\widetilde{D}_{n-1}(t, \alpha)}{\widetilde{D}_{n}(t, \alpha)\widehat{D}_{n}(t, \alpha)}{, }

用引理2.2, 代换上式的 \widetilde{D}_{n-1}(t, \alpha) \widetilde{D}_n(t, \alpha) , 可得

\beta_{2n}(t, \alpha) = \frac{\widehat{D}_{n+1}(t, \alpha)\widehat{D}_{n-1}(t, \alpha+2)}{\widehat{D}_{n}(t, \alpha+2)\widehat{D}_{n}(t, \alpha)}{.}

另一方面, Vein和Dale[20]已经证明

\widehat{D}_{n+1}(t, \alpha)\widetilde{D}_{n-1}(t, \alpha) = \widehat{D}_{n}(t, \alpha)\frac{{\rm d}\widetilde{D}_{n+1}(t, \alpha)}{{\rm d}t}-\widetilde{D}_{n}(t, \alpha)\frac{{\rm d}\widehat{D}_{n}(t, \alpha)}{{\rm d}t},

那么 \beta_{2n}(t, \alpha) 也可写成

\beta_{2n}(t, \alpha) = \frac{1}{\widetilde{D}_{n}(t, \alpha)}\frac{{\rm d}\widetilde{D}_{n} (t, \alpha)}{{\rm d}t}-\frac{1}{\widehat{D}_{n}(t, \alpha)}\frac{{\rm d}\widehat{D}_{n}(t, \alpha)}{{\rm d}t} = \frac{\rm d}{{\rm d}t}\ln\frac{\widetilde{D}_{n}(t, \alpha)}{\widehat{D}_{n}(t, \alpha)},

再结合引理2.2, 可得

\beta_{2n}(t, \alpha) = \frac{\rm d}{{\rm d}t}\ln\frac{\widehat{D}_{n}(t, \alpha+2)}{\widehat{D}_{n}(t, \alpha)}.

至此, 得证(2.9)式. 同理可得(2.10)式.

图 2

图 2   \alpha\equiv 1, t = -4, -\frac{1}{2}, 0 2 , n = 2, 3, 4 5 时, P_n(x;t, \alpha) 图像


图 3

图 3   \alpha\equiv 1, t = -4, -\frac{1}{2}, 0 2 时, P_n(x;t, \alpha) 图像, n = 2, 3, 4 5


注2.2   \beta_0(t, \alpha) = 0 , \beta_1(t, \alpha) = \frac{\rm d}{{\rm d}t}\ln\mu_0(t, \alpha) .

3 循环系数满足的差分方程

定理3.1   对于(1.4) 式中的循环系数 \beta_n(t, \alpha) 符合如下的非线性离散方程

\begin{eqnarray} & &10\beta_n\Big(\beta_{n-4}\beta_{n-3}\beta_{n-2}\beta_{n-1}+\beta_{n-3}^2\beta_{n-2}\beta_{n-1}+2\beta_{n-3}\beta_{n-2}^2\beta_{n-1}+\beta_{n-2}^3\beta_{n-1} \\ &&+2\beta_{n-3}\beta_{n-2}\beta_{n-1}^2+3\beta_{n-2}^2\beta_{n-1}^2+3\beta_{n-2}\beta_{n-1}^3+\beta_{n-1}^4 +2\beta_{n-3}\beta_{n-2}\beta_{n-1}\beta_n\\ &&+2\beta_{n-2}^2\beta_{n-1}\beta_{n}+6\beta_{n-2}\beta_{n-1}^2\beta_{n} +4\beta_{n-1}^3\beta_{n}+3\beta_{n-2}\beta_{n-1}\beta_{n}^2+6\beta_{n-1}^2\beta_{n}^2 \\ &&+4\beta_{n-1}\beta_n^3+\beta_n^4+\beta_{n-3}\beta_{n-2}\beta_{n-1}\beta_{n+1} +\beta_{n-2}^2\beta_{n-1}\beta_{n+1}+2\beta_{n-2}\beta_{n-1}^2\beta_{n+1}\\ & &+\beta_{n-1}^3\beta_{n+1}+4\beta_{n-2}\beta_{n-1}\beta_{n}\beta_{n+1} +6\beta_{n-1}^2\beta_{n}\beta_{n+1}+9\beta_{n-1}\beta_{n}^2\beta_{n+1}+4\beta_{n}^3\beta_{n+1}\\ &&+\beta_{n-2}\beta_{n-1}\beta_{n+1}^2+\beta_{n-1}^2\beta_{n+1}^2+6\beta_{n-1} \beta_{n}\beta_{n+1}^2+6\beta_{n}^2\beta_{n+1}^2+\beta_{n-1}\beta_{n+1}^3 +4\beta_{n}\beta_{n+1}^3\\ & &+\beta_{n+1}^4+\beta_{n-2}\beta_{n-1}\beta_{n+1}\beta_{n+2}+\beta_{n-1}^2\beta_{n+1}\beta_{n+2} +3\beta_{n-1}\beta_n\beta_{n+1}\beta_{n+2}+2\beta_{n}^2\beta_{n+1}\beta_{n+2} \\ & &+\beta_{n-1}\beta_{n+1}^2\beta_{n+2}+4\beta_{n}\beta_{n+1}^2\beta_{n+2}+2\beta_{n+1}^3\beta_{n+2} +\beta_{n+1}^2\beta_{n+2}^2+\beta_{n-1}\beta_{n}\beta_{n+1}\beta_{n+3} \\ & & +\beta_{n}^2\beta_{n+1}\beta_{n+3}+\beta_{n-1}\beta_{n+1}^2\beta_{n+3} +2\beta_{n}\beta_{n+1}^2\beta_{n+3}+\beta_{n+1}^3\beta_{n+3} +\beta_{n-1}\beta_{n+1}\beta_{n+2}\beta_{n+3}\\ & &+3\beta_{n}\beta_{n+1}\beta_{n+2}\beta_{n+3} +3\beta_{n+1}^2\beta_{n+2}\beta_{n+3}+2\beta_{n+1}\beta_{n+2}^2\beta_{n+3}+\beta_{n-1}\beta_{n+1}\beta_{n+3}^2 \\ && +\beta_{n}\beta_{n+1}\beta_{n+3}^2+\beta_{n+1}^2\beta_{n+3}^2+2\beta_{n+1}\beta_{n+2}\beta_{n+3}^2 +\beta_{n+1}\beta_{n+2}\beta_{n+3}\beta_{n+4}\Big)-2t\beta_{n}\\ & = &n+\frac{\alpha}{2}-\frac{(-1)^n\alpha}{2}. \end{eqnarray}
(3.1)

  为了方便起见, 记

{w}(x) = |x|^{\alpha}{\rm e}^{-x^{10}+tx^2} = |x|^{\alpha}{w}_0(x) = |x|^{\alpha}{\rm e}^{-v(x)}, \; \; x, t\in{{\mathbb R}} , \; \; \alpha>-1,

这里 {w}_0(x) = {\rm e}^{-x^{10}+tx^2}, \; v(x) = x^{10}-tx^2. 反复应用(1.4)式, 可以得到

\begin{eqnarray} x^5P_n(x)& = &P_{n+5}(x)+\beta_{n+4}P_{n+3}(x)+(P_{n+3}(x)+\beta_{n+2}P_{n+1}(x))(\beta_n+\beta_{n+1}+\beta_{n+2}+\beta_{n+3})\\ & &+(P_{n+1}(x)+\beta_nP_{n-1}(x))(\beta_n\beta_{n-1}+\beta_n^2+2\beta_n\beta_{n+1}+\beta_{n+1}^2+\beta_{n+1}\beta_{n+2})\\ &&+(P_{n-1}(x)+\beta_{n-2}P_{n-3}(x))\beta_{n-1}\beta_n(\beta_{n-2}+\beta_{n-1}+\beta_n+\beta_{n+1})\\ & &+(P_{n-3}(x)+\beta_{n-4}P_{n-5}(x))\beta_{n-3}\beta_{n-2}\beta_{n-1}\beta_n, \end{eqnarray}
(3.2)

\begin{eqnarray} x^4P_{n-1}(x) & = &P_{n+3}(x)+\beta_{n+2}P_{n+1}(x)+(P_{n+1}(x)+\beta_{n}P_{n-1}(x))(\beta_n+\beta_{n+1}+\beta_{n-1})\\ &&+(P_{n-1}(x)+\beta_{n-2}P_{n-3}(x))\beta_{n-1}(\beta_{n-2}+\beta_{n-1}+\beta_{n})\\ &&+(P_{n-3}(x)+\beta_{n-4}P_{n-5}(x))\beta_{n-3}\beta_{n-2}\beta_{n-1}. \end{eqnarray}
(3.3)

将(3.2) 和(3.3) 式带入到如下的积分计算中, 并结合(1.3) 式, 我们得到

\begin{eqnarray} &&\int_{{{\mathbb R}} }\Big(P_n(x)P_{n-1}(x)|x|^{\alpha}\Big)^{\prime}{w}_0(x){\rm d}x \\ & = &P_n(x)P_{n-1}(x){w}(x)\mid_{-\infty}^{+\infty} -\int_{{{\mathbb R}} }P_n(x)P_{n-1}(x)|x|^{\alpha}{\rm d}{w}_0(x)\\ & = &\int_{{{\mathbb R}} }P_{n-1}(x)P_n(x){w}(x)(10x^9-2tx){\rm d}x\\ & = &10\int_{{{\mathbb R}} }(x^5P_n(x))(x^4P_{n-1}(x)){w}(x){\rm d}x-2t\int_{{{\mathbb R}} }xP_{n-1}(x)P_n(x){w}(x){\rm d}x\\ & = &10\beta_n\Big[\beta_{n-4}\beta_{n-3}\beta_{n-2}\beta_{n-1}+\beta_{n-3}^2\beta_{n-2}\beta_{n-1} +2\beta_{n-3}\beta_{n-2}^2\beta_{n-1}+\beta_{n-2}^3\beta_{n-1}\\ & &+2\beta_{n-3}\beta_{n-2}\beta_{n-1}^2+3\beta_{n-2}^2\beta_{n-1}^2+3\beta_{n-2}\beta_{n-1}^3 +\beta_{n-1}^4+2\beta_{n-3}\beta_{n-2}\beta_{n-1}\beta_n\\ &&+2\beta_{n-2}^2\beta_{n-1}\beta_{n}+6\beta_{n-2}\beta_{n-1}^2\beta_{n}+4\beta_{n-1}^3\beta_{n} +3\beta_{n-2}\beta_{n-1}\beta_{n}^2+6\beta_{n-1}^2\beta_{n}^2+4\beta_{n-1}\beta_n^3\\ &&+\beta_n^4+\beta_{n-3}\beta_{n-2}\beta_{n-1}\beta_{n+1}+\beta_{n-2}^2\beta_{n-1}\beta_{n+1}+2\beta_{n-2} \beta_{n-1}^2\beta_{n+1}+\beta_{n-1}^3\beta_{n+1}\\ &&+4\beta_{n-2}\beta_{n-1}\beta_{n}\beta_{n+1}+6\beta_{n-1}^2\beta_{n}\beta_{n+1}+9\beta_{n-1} \beta_{n}^2\beta_{n+1}+4\beta_{n}^3\beta_{n+1}+\beta_{n-2}\beta_{n-1}\beta_{n+1}^2\\ &&+\beta_{n-1}^2\beta_{n+1}^2+6\beta_{n-1}\beta_{n}\beta_{n+1}^2+6\beta_{n}^2\beta_{n+1}^2 +\beta_{n-1}\beta_{n+1}^3+4\beta_{n}\beta_{n+1}^3+\beta_{n+1}^4\\ & &+\beta_{n-2}\beta_{n-1}\beta_{n+1}\beta_{n+2}+\beta_{n-1}^2\beta_{n+1}\beta_{n+2}+3 \beta_{n-1}\beta_n\beta_{n+1}\beta_{n+2}+2\beta_{n}^2\beta_{n+1}\beta_{n+2}\\ &&+\beta_{n-1}\beta_{n+1}^2\beta_{n+2}+4\beta_{n}\beta_{n+1}^2\beta_{n+2}+2\beta_{n+1}^3 \beta_{n+2}+\beta_{n+1}^2\beta_{n+2}^2+\beta_{n-1}\beta_{n}\beta_{n+1}\beta_{n+3}\\ &&+\beta_{n}^2\beta_{n+1}\beta_{n+3}+\beta_{n-1}\beta_{n+1}^2\beta_{n+3}+2\beta_{n}\beta_{n+1}^2 \beta_{n+3}+\beta_{n+1}^3\beta_{n+3}+\beta_{n-1}\beta_{n+1}\beta_{n+2}\beta_{n+3}\\ &&+3\beta_{n}\beta_{n+1}\beta_{n+2}\beta_{n+3}+3\beta_{n+1}^2\beta_{n+2}\beta_{n+3} +2\beta_{n+1}\beta_{n+2}^2\beta_{n+3}+\beta_{n-1}\beta_{n+1}\beta_{n+3}^2\\ &&+\beta_{n}\beta_{n+1}\beta_{n+3}^2+\beta_{n+1}^2\beta_{n+3}^2+2\beta_{n+1}\beta_{n+2} \beta_{n+3}^2+\beta_{n+1}\beta_{n+2}\beta_{n+3}\beta_{n+4}\Big]h_{n-1}\\ &&-2t\beta_{n}h_{n-1}. \end{eqnarray}
(3.4)

另一方面

\begin{eqnarray} & &\int_{{{\mathbb R}} }\Big(P_n(x)P_{n-1}(x)|x|^{\alpha}\Big)^{\prime}{w}_0(x){\rm d}x\\ & = &\int_{{{\mathbb R}} }P_n^{\prime}(x)P_{n-1}(x){w}(x){\rm d}x+\int_{{{\mathbb R}} }P_n(x)P_{n-1}^{\prime}(x){w}(x){\rm d}x+\alpha\int_{{{\mathbb R}} }\frac{P_n(x)P_{n-1}(x)}{x}{w}(x){\rm d}x\\ & = &n h_{n-1}+\alpha h_{n-1}\Delta_n, \end{eqnarray}
(3.5)

这里 \Delta_n = \frac{1-(-1)^n}{2} , 理由是: 当 n 是奇数时, \frac{P_n(x)}{x} n-1 次多项式; 当 n 是偶数时, \frac{P_{n-1}(x)}{x} n-2 次多项式[21]. 合并(3.4)和(3.5)式, 同时约去 h_{n-1} , 结果得证.

注3.1  方程(3.1)是一个八阶差分方程, 属于离散的Painlevé I方程族, 可参考文献[22]. 类似地, 我们可称该方程为 d_8{ P}_{I} .

4 循环系数的近似值

利用Freud[4]中的结论, 关于权函数 {w}(x;0, \alpha) = |x|^{\alpha}{\rm e}^{-x^{10}} 的循环系数 \beta_n 的渐近行为可表示为

\begin{equation} \lim\limits_{n\rightarrow \infty}\frac{\beta_n}{\sqrt[5]{n}} = \bigg(\frac{\Gamma(5)\Gamma(6)}{\Gamma(11)}\bigg)^{\frac{1}{5}} = \frac{1}{\sqrt[5]{1260}}, \end{equation}
(4.1)

于是, 可以得到关于权函数(1.2)的循环系数 \beta_n(t, \alpha) 的近似值.

图 4

图 4   \alpha\equiv 1, t = -4, -\frac{1}{4}, \frac{1}{2} 5 , \beta_n(t, \alpha) 图像


定理4.1   对于(1.4) 式中的循环系数 \beta_n(t, \alpha) , 有

\bullet n 是偶数时

\begin{eqnarray} \beta_n(t, \alpha)& = &\frac{n^{\frac{1}{5}}}{\phi}+\frac{\phi^3t}{3150n^{\frac{3}{5}}}-\frac{152+37\phi^5\alpha}{29925\phi n^{\frac{4}{5}}}-\frac{\phi^7 t^2}{9922500n^{\frac{7}{5}}}+\frac{(2888-4773\phi^5\alpha)\phi^3 t}{895505625n^{\frac{8}{5}}}\\ & &+\frac{1056355312+21866464\phi^5\alpha+11317\phi^{10}\alpha^2}{14328090000\phi n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.2)

\bullet n 是奇数时

\begin{eqnarray} \beta_n(t, \alpha)& = &\frac{n^{\frac{1}{5}}}{\phi}+\frac{\phi^3t}{3150n^{\frac{3}{5}}}-\frac{608-167\phi^5\alpha}{119700\phi n^{\frac{4}{5}}}-\frac{\phi^7 t^2}{9922500n^{\frac{7}{5}}}+\frac{(23104+37101\phi^5\alpha)\phi^3 t}{7164045000n^{\frac{8}{5}}}\\ & &+\frac{1056355312-21820256\phi^5\alpha-653\phi^{10}\alpha^2}{14328090000\phi n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.3)

这里, \phi = \sqrt[5]{1260} .

  通过观察(3.1) 式, (-1)^n 取决于 n 是奇数或偶数. 于是, 不妨令

\begin{equation} \beta_n = \left\{ \begin{array}{ll} u_n&, \; \; {{ n }是偶数{, }} \\ v_n&, \; \; {{ n }是奇数{, }} \end{array} \right. \end{equation}
(4.4)

这里通过(4.1)式, 可知

\lim\limits_{n\rightarrow \infty}\frac{u_n}{\sqrt[5]{n}} = \frac{1}{\sqrt[5]{1260}}, \; \; \lim\limits_{n\rightarrow \infty}\frac{v_n}{\sqrt[5]{n}} = \frac{1}{\sqrt[5]{1260}}.

这就意味着, u_n v_n 也满足(3.1) 式, 分别是

\begin{eqnarray} & &10u_n[u_{n-4}v_{n-3}u_{n-2}v_{n-1}+v_{n-3}^2u_{n-2}v_{n-1}+2v_{n-3}u_{n-2}^2v_{n-1}+u_{n-2}^3v_{n-1} +2v_{n-3}u_{n-2}v_{n-1}^2\\ & &+3u_{n-2}^2v_{n-1}^2+3u_{n-2}v_{n-1}^3+v_{n-1}^4+2v_{n-3}u_{n-2}v_{n-1}u_n+2u_{n-2}^2v_{n-1}u_{n}+6u_{n-2}v_{n-1}^2u_{n}\\ & &+4v_{n-1}^3u_{n}+3u_{n-2}v_{n-1}u_{n}^2+6v_{n-1}^2u_{n}^2+4v_{n-1}u_n^3+u_n^4+v_{n-3}u_{n-2}v_{n-1}v_{n+1}\\ & &+u_{n-2}^2v_{n-1}v_{n+1}+2u_{n-2}v_{n-1}^2v_{n+1}+v_{n-1}^3v_{n+1}+4u_{n-2}v_{n-1}u_{n}v_{n+1}+6v_{n-1}^2u_{n}v_{n+1}\\ &&+9v_{n-1}u_{n}^2v_{n+1}+4u_{n}^3v_{n+1}+u_{n-2}v_{n-1}v_{n+1}^2+v_{n-1}^2v_{n+1}^2+6v_{n-1}u_{n}v_{n+1}^2+6u_{n}^2v_{n+1}^2\\ &&+v_{n-1}v_{n+1}^3+4u_{n}v_{n+1}^3+v_{n+1}^4+u_{n-2}v_{n-1}v_{n+1}u_{n+2}+v_{n-1}^2v_{n+1}u_{n+2}+3v_{n-1}u_nv_{n+1}u_{n+2}\\ &&+2u_{n}^2v_{n+1}u_{n+2}+v_{n-1}v_{n+1}^2u_{n+2}+4u_{n}v_{n+1}^2u_{n+2}+2v_{n+1}^3u_{n+2}+v_{n+1}^2u_{n+2}^2\\ &&+v_{n-1}u_{n}v_{n+1}v_{n+3}+u_{n}^2v_{n+1}v_{n+3}+v_{n-1}v_{n+1}^2v_{n+3}+2u_{n}v_{n+1}^2v_{n+3}+v_{n+1}^3v_{n+3}\\ &&+v_{n-1}v_{n+1}u_{n+2}v_{n+3}+3u_{n}v_{n+1}u_{n+2}v_{n+3}+3v_{n+1}^2u_{n+2}v_{n+3}+2v_{n+1}u_{n+2}^2v_{n+3}\\ &&+v_{n-1}v_{n+1}v_{n+3}^2+u_{n}v_{n+1}v_{n+3}^2+v_{n+1}^2v_{n+3}^2+2v_{n+1}u_{n+2}v_{n+3}^2+v_{n+1}u_{n+2}v_{n+3}u_{n+4}]\\ &&-2tu_{n} = n \end{eqnarray}
(4.5)

\begin{eqnarray} & &10v_n[v_{n-4}u_{n-3}v_{n-2}u_{n-1}+u_{n-3}^2v_{n-2}u_{n-1}+2u_{n-3}v_{n-2}^2u_{n-1}+v_{n-2}^3u_{n-1} +2u_{n-3}v_{n-2}u_{n-1}^2\\ &&+3v_{n-2}^2u_{n-1}^2+3v_{n-2}u_{n-1}^3+u_{n-1}^4+2u_{n-3}v_{n-2}u_{n-1}v_n+2v_{n-2}^2u_{n-1}v_{n}+6v_{n-2}u_{n-1}^2v_{n}\\ &&+4u_{n-1}^3v_{n}+3v_{n-2}u_{n-1}v_{n}^2+6u_{n-1}^2v_{n}^2+4u_{n-1}v_n^3+v_n^4+u_{n-3}v_{n-2}u_{n-1}u_{n+1}\\ & &+v_{n-2}^2u_{n-1}u_{n+1}+2v_{n-2}u_{n-1}^2u_{n+1}+u_{n-1}^3u_{n+1}+4v_{n-2}u_{n-1}v_{n}u_{n+1}+6u_{n-1}^2v_{n}u_{n+1}\\ & &+9u_{n-1}v_{n}^2u_{n+1}+4v_{n}^3u_{n+1}+v_{n-2}u_{n-1}u_{n+1}^2+u_{n-1}^2u_{n+1}^2+6u_{n-1}v_{n}u_{n+1}^2+6v_{n}^2u_{n+1}^2\\ & &+u_{n-1}u_{n+1}^3+4v_{n}u_{n+1}^3+u_{n+1}^4+v_{n-2}u_{n-1}u_{n+1}v_{n+2}+u_{n-1}^2u_{n+1}v_{n+2}+3u_{n-1}v_nu_{n+1}v_{n+2}\\ & &+2v_{n}^2u_{n+1}v_{n+2}+u_{n-1}u_{n+1}^2v_{n+2}+4v_{n}u_{n+1}^2v_{n+2}+2u_{n+1}^3v_{n+2}+u_{n+1}^2v_{n+2}^2\\ & &+u_{n-1}v_{n}u_{n+1}u_{n+3}+v_{n}^2u_{n+1}u_{n+3}+u_{n-1}u_{n+1}^2u_{n+3}+2v_{n}u_{n+1}^2u_{n+3}+u_{n+1}^3u_{n+3}\\ & &+u_{n-1}u_{n+1}v_{n+2}u_{n+3}+3v_{n}u_{n+1}v_{n+2}u_{n+3}+3u_{n+1}^2v_{n+2}u_{n+3}+2u_{n+1}v_{n+2}^2u_{n+3}\\ & &+u_{n-1}u_{n+1}u_{n+3}^2+v_{n}u_{n+1}u_{n+3}^2+u_{n+1}^2u_{n+3}^2+2u_{n+1}v_{n+2}u_{n+3}^2+u_{n+1}v_{n+2}u_{n+3}v_{n+4}]\\ & &-2tv_{n} = n+\alpha. \end{eqnarray}
(4.6)

设存在常数 a_j b_j , j = 0, 1, 2, \cdots, 9 , 使得[8-9, 27]

\begin{equation} u_n = \frac{\sqrt[5]{n}}{\sqrt[5]{1260}}+\sum\limits_{j = 0}^{9}\frac{a_j}{n^{\frac{j}{5}}}+{\cal O}(n^{-2}), \; \; \; v_n = \frac{\sqrt[5]{n}}{\sqrt[5]{1260}} +\sum\limits_{j = 0}^{9}\frac{b_j}{n^{\frac{j}{5}}}+{\cal O}(n^{-2}). \end{equation}
(4.7)

进而

\begin{eqnarray} u_{n\pm 4}& = &\frac{\sqrt[5]{n}}{\phi}+a_0+\frac{a_1}{\sqrt[5]{n}}+\frac{a_2}{n^{\frac{2}{5}}}+\frac{a_3}{n^{\frac{3}{5}}}+\frac{a_4\pm\frac{4}{5\phi}}{n^{\frac{4}{5}}}+\frac{a_5}{n} +\frac{a_6\mp\frac{4a_1}{5}}{n^{\frac{6}{5}}}+\frac{a_7\mp\frac{8a_2}{5}}{n^{\frac{7}{5}}}\\ & &+\frac{a_8\mp\frac{12a_3}{5}}{n^{\frac{8}{5}}}+\frac{a_9\mp\frac{16a_4}{5}-\frac{32}{25\phi}}{n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.8)

\begin{eqnarray} v_{n\pm 4}& = &\frac{\sqrt[5]{n}}{\phi}+b_0+\frac{b_1}{\sqrt[5]{n}}+\frac{b_2}{n^{\frac{2}{5}}}+\frac{b_3}{n^{\frac{3}{5}}}+\frac{b_4\pm\frac{4}{5\phi}}{n^{\frac{4}{5}}}+\frac{b_5}{n} +\frac{b_6\mp\frac{4b_1}{5}}{n^{\frac{6}{5}}}+\frac{b_7\mp\frac{8b_2}{5}}{n^{\frac{7}{5}}}\\ & &+\frac{b_8\mp\frac{12b_3}{5}}{n^{\frac{8}{5}}}+\frac{b_9\mp\frac{16b_4}{5}-\frac{32}{25\phi}}{n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.9)

\begin{eqnarray} u_{n\pm 3}& = &\frac{\sqrt[5]{n}}{\phi}+a_0+\frac{a_1}{\sqrt[5]{n}}+\frac{a_2}{n^{\frac{2}{5}}}+\frac{a_3}{n^{\frac{3}{5}}}+\frac{a_4\pm\frac{3}{5\phi}}{n^{\frac{4}{5}}}+\frac{a_5}{n} +\frac{a_6\mp\frac{3a_1}{5}}{n^{\frac{6}{5}}}+\frac{a_7\mp\frac{6a_2}{5}}{n^{\frac{7}{5}}}\\ & &+\frac{a_8\mp\frac{9a_3}{5}}{n^{\frac{8}{5}}}+\frac{a_9\mp\frac{12a_4}{5}-\frac{8}{25\phi}}{n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.10)

\begin{eqnarray} v_{n\pm 3}& = &\frac{\sqrt[5]{n}}{\phi}+b_0+\frac{b_1}{\sqrt[5]{n}}+\frac{b_2}{n^{\frac{2}{5}}}+\frac{b_3}{n^{\frac{3}{5}}}+\frac{b_4\pm\frac{3}{5\phi}}{n^{\frac{4}{5}}}+\frac{b_5}{n} +\frac{b_6\mp\frac{3b_1}{5}}{n^{\frac{6}{5}}}+\frac{b_7\mp\frac{6b_2}{5}}{n^{\frac{7}{5}}}\\ & &+\frac{b_8\mp\frac{9b_3}{5}}{n^{\frac{8}{5}}}+\frac{b_9\mp\frac{12b_4}{5}-\frac{8}{25\phi}}{n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.11)

\begin{eqnarray} u_{n\pm 2}& = &\frac{\sqrt[5]{n}}{\phi}+a_0+\frac{a_1}{\sqrt[5]{n}}+\frac{a_2}{n^{\frac{2}{5}}}+\frac{a_3}{n^{\frac{3}{5}}}+\frac{a_4\pm\frac{2}{5\phi}}{n^{\frac{4}{5}}}+\frac{a_5}{n} +\frac{a_6\mp\frac{2a_1}{5}}{n^{\frac{6}{5}}}+\frac{a_7\mp\frac{4a_2}{5}}{n^{\frac{7}{5}}}\\ & &+\frac{a_8\mp\frac{6a_3}{5}}{n^{\frac{8}{5}}}+\frac{a_9\mp\frac{8a_4}{5}-\frac{8}{25\phi}}{n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.12)

\begin{eqnarray} v_{n\pm 2}& = &\frac{\sqrt[5]{n}}{\phi}+b_0+\frac{b_1}{\sqrt[5]{n}}+\frac{b_2}{n^{\frac{2}{5}}}+\frac{b_3}{n^{\frac{3}{5}}}+\frac{b_4\pm\frac{2}{5\phi}}{n^{\frac{4}{5}}}+\frac{b_5}{n} +\frac{b_6\mp\frac{2b_1}{5}}{n^{\frac{6}{5}}}+\frac{b_7\mp\frac{4b_2}{5}}{n^{\frac{7}{5}}}\\ & &+\frac{b_8\mp\frac{6b_3}{5}}{n^{\frac{8}{5}}}+\frac{b_9\mp\frac{8b_4}{5}-\frac{8}{25\phi}}{n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.13)

\begin{eqnarray} u_{n\pm 1}& = &\frac{\sqrt[5]{n}}{\phi}+a_0+\frac{a_1}{\sqrt[5]{n}}+\frac{a_2}{n^{\frac{2}{5}}}+\frac{a_3}{n^{\frac{3}{5}}}+\frac{a_4\pm\frac{1}{5\phi}}{n^{\frac{4}{5}}}+\frac{a_5}{n} +\frac{a_6\mp\frac{a_1}{5}}{n^{\frac{6}{5}}}+\frac{a_7\mp\frac{2a_2}{5}}{n^{\frac{7}{5}}}\\ & &+\frac{a_8\mp\frac{3a_3}{5}}{n^{\frac{8}{5}}}+\frac{a_9\mp\frac{4a_4}{5}-\frac{2}{25\phi}}{n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.14)

\begin{eqnarray} v_{n\pm 1}& = &\frac{\sqrt[5]{n}}{\phi}+b_0+\frac{b_1}{\sqrt[5]{n}}+\frac{b_2}{n^{\frac{2}{5}}}+\frac{b_3}{n^{\frac{3}{5}}}+\frac{b_4\pm\frac{1}{5\phi}}{n^{\frac{4}{5}}}+\frac{b_5}{n} +\frac{b_6\mp\frac{b_1}{5}}{n^{\frac{6}{5}}}+\frac{b_7\mp\frac{2b_2}{5}}{n^{\frac{7}{5}}}\\ & &+\frac{b_8\mp\frac{3b_3}{5}}{n^{\frac{8}{5}}}+\frac{b_9\mp\frac{4b_4}{5}-\frac{2}{25\phi}}{n^{\frac{9}{5}}}+{\cal O}(n^{-2}), \end{eqnarray}
(4.15)

这里 \phi = \sqrt[5]{1260} . 将(4.7)–(4.15) 式代入(4.5) 式和(4.6)式, 可以计算出

a_0 = b_0 = 0, \; \; a_1 = b_1 = 0, \; \; a_2 = b_2 = 0, \; \; a_3 = b_3 = \frac{\phi^3 t}{3150}, \; \; a_4 = \frac{-37\phi^5\alpha-152}{29925\phi},

b_4 = \frac{167\phi^5\alpha-608}{119700\phi}, \; \; \; a_5 = b_5 = a_6 = b_6 = 0, \; \; \; a_7 = b_7 = -\frac{\phi^7 t^2}{9922500},

a_8 = \frac{(2888-4773\phi^5\alpha)\phi^3 t}{895505625}, \; \; \; b_8 = \frac{(23104+37101\phi^5\alpha)\phi^3 t}{7164045000},

a_9 = \frac{11317\phi^{10}\alpha^2+21866464\phi^5\alpha+1056355312}{14328090000\phi},

b_9 = \frac{-653\phi^{10}\alpha^2-21820256\phi^5\alpha+1056355312}{14328090000\phi}.

定理4.1得证.

图 5

图 5   t\equiv 1, \alpha = -\frac{1}{2}, 0, \frac{1}{2} 5 时, \beta_n(t, \alpha) 图像


5 Hankel行列式满足的方程

定理5.1   关于Freud - 型权函数(1.2), 当 n\rightarrow \infty 时, 对应的Hankel行列式满足关系式

\begin{equation} D_{2n+1} = {\rm e}^{\Theta_n{+C}}D_{2n}, \end{equation}
(5.1)

这里, C 为任意常数,

\begin{eqnarray*} \Theta_n& = &\frac{2^{\frac{4}{5}}tn^{\frac{1}{5}}}{3^{\frac{2}{5}}\times 35^{\frac{1}{5}}}+\frac{t^2}{5\times 3^{\frac{4}{5}}\times 70^{\frac{2}{5}}n^{\frac{3}{5}}}+\frac{(299+315\alpha)t}{3150\times 3^{\frac{2}{5}}\times 70^{\frac{1}{5}}n^{\frac{4}{5}}}-\frac{2^{\frac{2}{5}}t^3}{225\times 3^{\frac{1}{5}}\times {35}^{\frac{3}{5}}n^{\frac{7}{5}}}\nonumber\\ & &-\frac{(945\alpha+913)t^2}{31500\times 3^{\frac{4}{5}}\times 70^{\frac{2}{5}}n^{\frac{8}{5}}} +\frac{(264533850\alpha^2-313931205\alpha+16010711)t}{895505625\times 3^{\frac{2}{5}}\times 70^{\frac{1}{5}}n^{\frac{9}{5}}} +{\cal O}\left(n^{-2}\right). \end{eqnarray*}

  在(2.9) 和(2.10) 式的两边同时进行积分, 可得

\begin{equation} \int\beta_{2n}(t, \alpha){\rm d}t = \ln\widehat{D}_n(t, \alpha+2)-\ln\widehat{D}(t, \alpha){+C_1} \end{equation}
(5.2)

\begin{equation} \int\beta_{2n+1}(t, \alpha){\rm d}t = \ln\widehat{D}_{n+1}(t, \alpha+2)-\ln\widehat{D}(t, \alpha+2){+C_2}. \end{equation}
(5.3)

联立(5.2) 和(5.3) 式, 并且在消去 \ln\widehat{D}_n(t, \alpha+2) 的同时, 应用(2.3) 式, 可得

\begin{equation} \int(\beta_{2n}(t, \alpha)+\beta_{2n+1}(t, \alpha)){\rm d}t = \ln\widehat{D}_{n+1}(t, \alpha)-\ln\widehat{D}_n(t, \alpha){+C_3} = \ln\frac{D_{2n+1}(t, \alpha)}{D_{2n}(t, \alpha)}{+C_3}, \end{equation}
(5.4)

其中 C_1-C_3 为任意常数. 又根据(4.2) 和(4.3) 式, 有

\begin{eqnarray} \beta_{2n}(t, \alpha)& = &\frac{n^{\frac{1}{5}}}{3^{\frac{2}{5}}\times 70^{\frac{1}{5}}}+\frac{t}{5\times 3^{\frac{4}{5}}\times 70^{\frac{2}{5}}n^{\frac{3}{5}}}-\frac{t^2}{75\times 3^{\frac{1}{5}}\times 70^{\frac{3}{5}}n^{\frac{7}{5}}}-\frac{(38+11655\alpha)2^{\frac{4}{5}}}{29925\times 3^{\frac{2}{5}}\times 35^{\frac{1}{5}}n^{\frac{4}{5}}}\\ & &-\frac{2^{\frac{8}{5}}(1503495\alpha-722)\times 3^{\frac{1}{5}}\times 35^{\frac{3}{5}}t}{298501875n^{\frac{8}{5}}}+\frac{1122929325\alpha^2+1721984040\alpha+66022207}{3582022500\times 3^{\frac{2}{5}}\times 70^{\frac{1}{5}}n^{\frac{9}{5}}} \\ && +{\cal O}\left((2n)^{-2}\right), \end{eqnarray}
(5.5)

\begin{eqnarray} \beta_{2n+1}(t, \alpha)& = &\frac{(2n+1)^{\frac{1}{5}}}{6^{\frac{2}{5}}\times 35^{\frac{1}{5}}}+\frac{2^{\frac{1}{5}}t}{5\times 3^{\frac{4}{5}}\times 35^{\frac{2}{5}}(2n+1)^{\frac{3}{5}}}-\frac{2^{\frac{4}{5}}t^2}{75\times 3^{\frac{1}{5}}\times 35^{\frac{3}{5}}(2n+1)^{\frac{7}{5}}}\\ & &+\frac{52605\alpha-152}{29925\times 6^{\frac{2}{5}}\times 35^{\frac{1}{5}}(2n+1)^{\frac{4}{5}}} +\frac{2^{\frac{1}{5}}(11686815\alpha+5776)t}{2842875\times 3^{\frac{4}{5}}\times 35^{\frac{2}{5}}(2n+1)^{\frac{8}{5}}}\\ & &-\frac{64793925\alpha^2+1718345160\alpha-66022207}{895505625\times 6^{\frac{2}{5}}\times 35^{\frac{1}{5}}(2n+1)^{\frac{9}{5}}}+{\cal O}\left((2n+1)^{-2}\right), \end{eqnarray}
(5.6)

将(5.5) 和(5.6) 式代入(5.4) 式, 并且进行积分运算, 可得

\begin{eqnarray*} \ln\frac{D_{2n+1}(t, \alpha)}{D_{2n}(t, \alpha)} & = &\frac{2^{\frac{4}{5}}tn^{\frac{1}{5}}}{3^{\frac{2}{5}}\times 35^{\frac{1}{5}}}+\frac{t^2}{5\times 3^{\frac{4}{5}}\times 70^{\frac{2}{5}}n^{\frac{3}{5}}}+\frac{(299+315\alpha)t}{3150\times 3^{\frac{2}{5}}\times 70^{\frac{1}{5}}n^{\frac{4}{5}}}-\frac{2^{\frac{2}{5}}t^3}{225\times 3^{\frac{1}{5}}\times {35}^{\frac{3}{5}}n^{\frac{7}{5}}}\nonumber\\ &&-\frac{(945\alpha+913)t^2}{31500\times 3^{\frac{4}{5}}\times 70^{\frac{2}{5}}n^{\frac{8}{5}}} +\frac{(264533850\alpha^2-313931205\alpha+16010711)t}{895505625\times 3^{\frac{2}{5}}\times 70^{\frac{1}{5}}n^{\frac{9}{5}}}\nonumber\\ & &+{\cal O}\left(n^{-2}\right){+C}, \end{eqnarray*}

C 为常数, 定理5.1得证.

6 Hankel矩阵的最小特征值

在这一节, 我们讨论由权函数

\begin{equation} {w}_{E}(x) = {\rm e}^{-x^{10}}, \; \; \; \; x\in{{\mathbb R}} \end{equation}
(6.1)

生成的Hankel矩阵的最小特征值在 n\rightarrow \infty 时的渐近行为. 显然, 这个权函数是(1.2) 式当 t = \alpha = 0 时的特例. 该节要用到的方法, 可参见文献[23, 31]. 它可被用于求任意一个指数型权函数 {\rm e}^{-2Q(x)} 生成的Hankel矩阵的最小特征值在高维时的渐近行为, 这里 Q(x) 是偶函数并且所符合的条件请参阅文献[31]. 对于函数(6.1)而言, Q(x) = \frac{1}{2}x^{10} . 根据平衡测度以及Mhaskar-Rakhmanor-Saff数字 a_n (参见文献[23, 31]), 循环系数 \beta_n 的近似值可以表出, 这里 a_n 是方程 n = \frac{2}{\pi}\int_0^1\frac{a_ntQ^{\prime}(a_nt)}{\sqrt{1-t^2}}{\rm d}t 唯一的正根. 除此之外, 当 n\rightarrow \infty 时, \beta_n a_n 的关系式[23]

\begin{equation} a_n^2 = 4\beta_n. \end{equation}
(6.2)

接下来, 我们给出本文关于最小特征值的结论.

定理6.1  关于权函数(6.1), 它的最大Hankel矩阵所对应的最小特征值可以表示成

\begin{equation} \lambda_n\simeq\frac{n^{\frac{9}{20}}}{{\cal K}^{\frac{1}{2}}}\exp\left(-\frac{525n^{\frac{1}{10}}}{864{\cal K}^9}+\frac{25n^{\frac{3}{10}}}{84{\cal K}^7}-\frac{3n^{\frac{1}{2}}}{10{\cal K}^5} +\frac{10n^{\frac{7}{10}}}{21{\cal K}^3}-\frac{20n^{\frac{9}{10}}}{9{\cal K}}\right), \end{equation}
(6.3)

这里 {\cal K} = \left(\frac{256}{315}\right)^{\frac{1}{10}} .

  不妨令 v_E(x) = -\ln{w}_E(x) = x^{10} , 则 v_E^{\prime}(x) = 10x^9 . 对于函数(6.1), 根据库仑流体方法[24-26]及其补充条件 \int_a^b\frac{xv^{\prime}(x)}{\sqrt{(b-x)(x-a)}}{\rm d}x = 2\pi n 可知

2\pi n = \int_{-b_E}^{b_E}\frac{10x^{10}}{\sqrt{b_E^2-x^2}}{\rm d}x = 20\int_0^{b_E}\frac{x^{10}}{\sqrt{b_E^2-x^2}}{\rm d}x = 20b_E^{10}\int_0^{\frac{\pi}{2}}\sin^{10}\theta {\rm d}\theta = \frac{315\pi}{128}b_E^{10},

于是

b_E^2\simeq \left(\frac{256n}{315}\right)^{\frac{1}{5}}.

从而[24], 有

\begin{eqnarray} \beta_n\simeq\left(\frac{b_E-a_E}{4}\right)^2 = \frac{b_E^2}{4} = \left(\frac{n}{1260}\right)^{\frac{1}{5}}, \; \; \; n\rightarrow \infty, \end{eqnarray}
(6.4)

这个结果也对应了(4.1)式. 将(6.4) 式应用于(6.2)式, 有

a_n = \left(\frac{256n}{315}\right)^{\frac{1}{10}} = {\cal K}n^{\frac{1}{10}},

这里 {\cal K} = \left(\frac{256}{315}\right)^{\frac{1}{10}} .

基于Chen和Lubinsky的理论[31], 同时利用 \ln(x+\sqrt{1+x^2}) x = 0 处的麦克劳林展开式

\ln(x+\sqrt{1+x^2}) = \sum\limits_{k = 0}^{\infty}(-1)^k\frac{(2k)!}{2^{2k}(k!)^2(2k+1)}x^{2k+1}, \; \; \; |x|\leq 1,

可以计算出最小特征值 \lambda_n 如下

\begin{eqnarray*} \lambda_n&\simeq& \sqrt{\frac{n}{a_n}}\exp\Big(-2\int_0^n\ln\Big(\frac{1}{a_s}+\sqrt{1+\frac{1}{a_s^2}}\Big){\rm d}s\Big)\nonumber\\ & = &\sqrt{\frac{n}{a_n}}\exp\Big(-2\int_0^n\sum\limits_{k = 0}^{\infty}(-1)^k\frac{(2k)!}{2^{2k}(k!)^2(2k+1)}a_s^{-2k-1}{\rm d}s\Big)\nonumber\\ & = &\frac{n^{\frac{9}{20}}}{{\cal K}^{\frac{1}{2}}}\exp\Big(-2\sum\limits_{k = 0}^{\infty}(-1)^k\frac{(2k)!}{2^{2k}(k!)^2(2k+1)}\int_0^n{\cal K}^{-2k-1}s^{\frac{-2k-1}{10}}{\rm d}s\Big)\nonumber\\ & = &\frac{n^{\frac{9}{20}}}{{\cal K}^{\frac{1}{2}}}\exp\Big(-2n\sum\limits_{k = 0}^{\infty}(-1)^k\frac{(2k)!}{2^{2k}(k!)^2(2k+1)}\frac{10a_n^{-2k-1}}{9-2k}\Big)\nonumber\\ &\simeq &\frac{n^{\frac{9}{20}}}{{\cal K}^{\frac{1}{2}}}\exp\Big(-2n\Big(\frac{10}{9{\cal K}n^{\frac{1}{10}}}-\frac{5}{21{\cal K}^3n^{\frac{3}{10}}} +\frac{3}{20{\cal K}^5n^{\frac{1}{2}}}-\frac{25}{168{\cal K}^7n^{\frac{7}{10}}}+\frac{175}{576{\cal K}^9n^{\frac{9}{10}}}+{\cal O}(n^{-1})\Big)\Big)\nonumber\\ &\simeq &\frac{n^{\frac{9}{20}}}{{\cal K}^{\frac{1}{2}}}\exp\Big(-\frac{175n^{\frac{1}{10}}}{288{\cal K}^9}+\frac{25n^{\frac{3}{10}}}{84{\cal K}^7}-\frac{3n^{\frac{1}{2}}}{10{\cal K}^5} +\frac{10n^{\frac{7}{10}}}{21{\cal K}^3}-\frac{20n^{\frac{9}{10}}}{9{\cal K}}\Big). \end{eqnarray*}

定理6.1得证.

参考文献

Chihara T S . An Introduction to Orthogonal Polynomials. New York: Gordon and Breach, 1978

[本文引用: 1]

Freud G . Othogonal Polynomials. Budapest, Hungary: Pergamon, 1971

Van Assche W . Asymptotics for Orthogonal Polynomials. Berlin: Springer, 1987

[本文引用: 1]

Freud G .

On the coefficients in the recursion formulae of orthogonal polynomials

Proc R Irish Acad, 1976, 76, 1- 6

URL     [本文引用: 2]

Magnus A P. A Proof of Freud's Conjecture About the Orthogonal Polynomials Related to |x|^{\rho}\exp(-x^{2m}), for Integer m//Brezinski C, Draux A, Magnus A P, et al. Orthogoanl Polynomials and Applications. Berlin: Springer-Verlag, 1985: 362-372

[本文引用: 1]

Lubinsky D , Mhaskar H , Saff E .

A proof of Freud's conjecture for exponential weights

Constr Approx, 1988, 4, 65- 83

DOI:10.1007/BF02075448      [本文引用: 1]

Magnus A P .

On Freud's equations for the exponential weights

J Approx Theory, 1986, 46, 65- 99

DOI:10.1016/0021-9045(86)90088-2      [本文引用: 1]

Clarkson P A , Jordaan K , Kelil A .

A generalized Freud weight

Stud Appl Math, 2016, 136, 288- 320

DOI:10.1111/sapm.12105      [本文引用: 2]

Clarkson P A , Jordaan K .

Properties of generalized Freud polynomials

J Approx Theory, 2018, 25, 148- 175

[本文引用: 2]

Bleher P , Its A R .

Semiclassical asymptotics of orthogonal polynomials: Riemann-Hilbert problem, and universality in the matrix model

Ann of Math, 1999, 150, 185- 266

DOI:10.2307/121101      [本文引用: 2]

Bleher P , Its A R .

Double scaling limit in the random matrix model: the Riemann-Hilbert approach

Comm Pure Appl Math, 2003, 56, 433- 516

DOI:10.1002/cpa.10065     

Clarkson P A , Jordaan K .

The relationship between semiclassical Laguerre polynomials and the fourth Painlevé equation

Constr Approx, 2014, 39, 223- 254

DOI:10.1007/s00365-013-9220-4      [本文引用: 1]

Levin E , Lubinsky D . Orthogonal Polynomials with Exponential Weights. New York: Springer-Verlag, 2001

[本文引用: 1]

Lyu S L , Chen Y .

The largest eigenvalue distribution of the Laguerre unitary ensemble

Acta Math Scientia, 2017, 37 (2): 439- 462

DOI:10.1016/S0252-9602(17)30013-9      [本文引用: 1]

Wong R , Zhang L .

Global asymptotics of orthogonal polynomials associated with |x|^{2\alpha}{\rm e}^{-Q(x)}

J Approx Theory, 2010, 162, 723- 765

DOI:10.1016/j.jat.2009.09.007      [本文引用: 1]

Bonan S , Clark D S .

Estimates of the orthogonal polynomials with weight \exp(-x^m), m an even positive integer

J Approx Theory, 1986, 46, 408- 410

DOI:10.1016/0021-9045(86)90074-2      [本文引用: 1]

Bonan S , Nevai P .

Orthogonal polynomials and their derivatives I

J Approx Theory, 1984, 40, 134- 147

DOI:10.1016/0021-9045(84)90023-6     

Chen Y , Pruessner G .

Orthogonal polynomials with discontinuous weights

J Phys A: Math Gen, 2005, 38, 191- 198

DOI:10.1088/0305-4470/38/12/L01      [本文引用: 1]

Wang D , Zhu M K , Chen Y .

On semi-classical orthogonal polynomials associated with a Freud - type weight

Math Meth Appl Sci, 2020, 43 (8): 5295- 5313

DOI:10.1002/mma.6270      [本文引用: 1]

Vein P R , Dale P . Determinants and Their Applications in Mathematical Physics. New York: Springer-Verlag, 1999

[本文引用: 1]

Van Assche W. Discrete Painlevé Equations for Recurrence Coefficients of Orthogonal Polynomials//Elaydi S, Cushing J, Lasser R, et al. Difference Equations, Special Functions and Orthogonal Polynomials. Hackensack: World Scientific, 2007: 687-725

[本文引用: 1]

Cresswell C , Joshi N .

The discrete first, second and thirty-fourth Painlevé hierarchies

J Phys A: Math Gen, 1999, 32, 655- 669

DOI:10.1088/0305-4470/32/4/009      [本文引用: 1]

Damelin S B .

Asymptotics of recurrence coefficients for orthonormal polynomials on the line-Magnus method revisited

Math Comp, 2004, 73, 191- 209

URL     [本文引用: 3]

Chen Y , Ismail M E H .

Thermodynamic relations of the Hermitian matrix ensembles

J Phys A: Math Gen, 1997, 30, 6633- 6654

DOI:10.1088/0305-4470/30/19/006      [本文引用: 2]

Chen Y , Lawrence N .

On the linear statistics of Hermitian random matrices

J Phys A: Math Gen, 1998, 31, 1141- 1152

DOI:10.1088/0305-4470/31/4/005     

Chen Y , Mckay M R .

Coulomb fluid, Painlevé transcendents and the information theory of MIMO systems

IEEE Trans Inform Theory, 2012, 58, 4594- 4634

DOI:10.1109/TIT.2012.2195154      [本文引用: 1]

Zhu M K , Chen Y .

On properties of a deformed Freud weight

Random Matrices Theory Appl, 2019, 8, 1950004

DOI:10.1142/S2010326319500047      [本文引用: 2]

Berg C , Chen Y , Ismail M E H .

Small eigenvalues of large Hankel matrices: The indeterminate case

Math Scand, 2002, 91, 67- 81

DOI:10.7146/math.scand.a-14379      [本文引用: 1]

Berg C , Szwarc R .

The smallest eigenvalue of Hankel matrices

Const Approx, 2011, 34, 107- 133

DOI:10.1007/s00365-010-9109-4     

Chen Y , Lawrence N D .

Small eigenvalues of large Hankel matrices

J Phys A: Math Gen, 1999, 32, 7305- 7315

DOI:10.1088/0305-4470/32/42/306     

Chen Y , Lubinsky D S .

Smallest eigenvalues of Hankel matrices for exponential weights

J Math Anal Appl, 2004, 293, 476- 495

DOI:10.1016/j.jmaa.2004.01.032      [本文引用: 4]

Emmart N , Chen Y , Weems C .

Computing the smallest eigenvalue of large ill-conditioned Hankel matrices

Commun Comput Phys, 2015, 18, 104- 124

DOI:10.4208/cicp.260514.231214a     

Szegö G .

On some Hermitian forms associated with two given curves of the complex plane

Trans Amer Math Soc, 1936, 40, 450- 461

DOI:10.1090/S0002-9947-1936-1501884-1     

Widom H , Wilf H S .

Small eigenvalues of large Hankel matrices

Proc Amer Math Soc, 1966, 17, 338- 344

DOI:10.1090/S0002-9939-1966-0189237-7     

Zhu M K , Chen Y , Emmart N , Weems C .

The smallest eigenvalue of large Hankel matrices

Appl Math Comput, 2018, 334, 375- 387

Zhu M K , Emmart N , Chen Y , Weems C .

The smallest eigenvalue of large Hankel matrices generated by a deformed Laguerre weight

Math Meth Appl Sci, 2019, 42, 3272- 3288

DOI:10.1002/mma.5583     

Chen Y , Sikorowski J , Zhu M K .

Smallest eigenvalue of large Hankel matrices at critical point: Comparing conjecture with parallelised computation

Appl Math Comput, 2019, 363, 124628- 124646

URL    

Zhu M K , Chen Y , Li C Z .

The smallest eigenvalue of large Hankel matrices generated by a singularly perturbed Laguerre weight

J Math Phys, 2020, 61, 073502

DOI:10.1063/1.5140079      [本文引用: 1]

/