数学杂志  2024, Vol. 44 Issue (1): 59-72   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
陈琳
吴嘎日迪
修正的一元与二元Szasz-Mirakyan-Kantorovich算子在Orlicz空间内的加权逼近
陈琳, 吴嘎日迪    
1. 内蒙古师范大学数学科学学院, 内蒙古 呼和浩特 010022;
2. 内蒙古自治区应用数学中心, 内蒙古 呼和浩特 010022
摘要:本文主要研究了修正的一元与二元Szasz-Mirakyan-Kantorovich算子的逼近问题.为此,首先证明了在加权Orlicz空间内的Korovkin型定理,在此基础上利用Jensen不等式,Hölder不等式,Steklov平均函数并结合相关分析技巧,获得了修正的一元与二元Szasz-Mirakyan-Kantorovich算子在加权Orlicz空间内的逼近正定理和收敛定理.
关键词Szasz-Mirakyan-Kantorovich算子    Orlicz空间    Korovkin型定理    逼近    
WEIGHTED APPROXIMATION OF MODIFIED UNIVARIATE AND MULTIVARIATE SZASZ-MIRAKYAN-KANTOROVICH OPERATORS IN ORLICZ SPACES
CHEN Lin, WU Garidi    
1. College of Mathematics Science, Inner Mongolia Normal University, Hohhot 010022, China;
2. Center for Applied Mathematics, Inner Mongolia Autonomous Regin, Hohhot 010022, China
Abstract: This article mainly studies the approximation problem of modified univariate and multivariate Szasz-Mirakyan-Kantorovich operators. In order to solve this problem, the Korovkin type theorem in weighted Orlicz spaces was first proved. Based on the Jensen inequality, Hölder inequality, Steklov mean function, and related analysis techniques, we obtained the approximation positive theorems and convergence theorems of the modified univariate and multivariate Szasz-Mirakyan-Kantorovich operators in weighted Orlicz spaces.
Keywords: Szasz-Mirakyan-Kantorovich operators     Orlicz space     Korovkin type theorem     approximation    
1 引言

著名的Szasz-Mirakyan-Kantorovich算子为$ T_n(f;x)=:n\sum\limits_{k=0}^{\infty}e^{-nx}\frac{{(nx)}^k}{k!} \int_{\frac{k}{n}}^{\frac{k+1}{n}}f(t) \mathrm{d} t, $其中$ x\in R_0=[0, +\infty) $$ n\in N $. 参考文献[1]和[2]对于该算子在勒贝格可积函数空间内的逼近问题已有详细研究. 之后参考文献[3]中给出了一个修正的Szasz-Mirakyan-Kantorovich算子

$ \begin{equation*} T_n(f;a_n, b_n;x)=:b_n\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{{(a_nx)}^k}{k!}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}f(t) \mathrm{d} t, x\in R_0, n\in N, \end{equation*} $

其中$ \{a_n\}_{n=1}^{\infty} $$ \{b_n\}_{n=1}^{\infty} $是递增的无界数列,并且$ a_n\geq1 $$ b_n\geq1 $,同时$ \{\frac{a_n}{b_n}\}_{n=1}^{\infty} $是不减数列且

$ \frac{a_n}{b_n}=1+o(\frac{1}{b_n}), \; \; \lim\limits_{n\rightarrow \infty}{\frac{1}{b_n}}=0. $

同时,ZBIGNIEW WALCZAK在参考文献[3]中研究了这一算子在勒贝格可积函数空间中的相关逼近问题. 本文将函数空间放大为Orlicz空间,并在这一空间内研究修正的Szasz-Mirakyan-Kantorovich算子的加权逼近,同时给出相应的逼近正定理以及收敛定理.

结合参考文献[4],并仿照修正的一元Szasz-Mirakyan-Kantorovich算子,我们定义修正的二元Szasz-Mirakyan-Kantorovich算子如下

$ \begin{align*} &T_{n, m}(f;a_n, b_n, c_m, d_m;x, y)\\ =:&b_nd_me^{-a_nx}e^{-c_my}\sum\limits_{k=0}^{\infty}\sum\limits_{j=0}^{\infty}\frac{{(a_nx)}^k}{k!}\frac{{(c_my)}^j}{j!} \int_{\frac{j}{d_m}}^{\frac{j+1}{d_m}}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}f(u_1, u_2) \mathrm{d} u_1 \mathrm{d} u_2, \end{align*} $

其中$ u_1, \; u_2\in R_0, \; \; n\in N, \; \; \{a_n\}_{n=1}^{\infty}, \; \; \{b_n\}_{n=1}^{\infty}, \; \; \{c_m\}_{m=1}^{\infty}, \; \; \{d_m\}_{m=1}^{\infty} $是递增的无界数列,并且满足

$ \frac{a_n}{b_n}=1+o(\frac{1}{b_n}), \; \; \lim\limits_{n\rightarrow \infty}{\frac{1}{b_n}}=0;\; \; \frac{c_m}{d_m}=1+o(\frac{1}{d_m}), \; \; \lim\limits_{m\rightarrow \infty}{\frac{1}{d_m}}=0. $

之后我们证明了加权$ Orlicz $空间内的Korovkin型定理,并在此基础上研究了这个二元算子在加权Orlicz空间内的收敛问题.

我们以$ L_M^*[0, +\infty) $(简记为$ L_M^* $)表示定义在$ [0, +\infty) $上由$ N $函数$ M(u) $生成的Orlicz空间,$ N(v) $$ M(u) $的余$ N $函数,对于任意的$ f\in L_M^* $,定义Orlicz范数为

$ \begin{equation*} \|f\|_{M}=\sup\limits_{\rho(v;N)\leq1}\left|\int_0^{+\infty}f(x)v(x) \mathrm{d} x\right|, \end{equation*} $

其中$ \rho(v;N)=\int_0^{+\infty}N(v(x)) \mathrm{d} x $$ v(x) $关于$ N(v) $的模. 由参考文献[5]知,上述Orlicz范数还可以用如下方式计算

$ \begin{equation*} \|f\|_{M}=\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+\int_0^{+\infty}M(\alpha f(x)) \mathrm{d} x). \end{equation*} $

$ w_n(x_1, x_2, \cdots, x_n) $$ n $元权函数,对$ f\in L_{M, w_n}^* $($ f $$ n $元函数,且每一个自变量的取值范围均是$ [0, +\infty) $),定义其加权Orlicz范数如下

$ \begin{align*} &\|f\|_{M, w_n}\\ =&\sup\limits_{\rho(v(x_1, x_2, \cdots, x_n);N)\leq1}|\int_0^{\infty}\int_0^{\infty}\cdots\int_0^{\infty}f(x_1, x_2, \cdots;x_n)v(x_1, x_2, \cdots;x_n)\\ &\times w_n(x_1, x_2, \cdots, x_n) \mathrm{d} x_1 \mathrm{d} x_2\cdots \mathrm{d} x_n|, \end{align*} $

其中$ \rho(v(x_1, x_2, \cdots, x_n);N)= \int_0^{\infty}\int_0^{\infty}\cdots\int_0^{\infty}N(v(x_1, x_2, \cdots, x_n)) \mathrm{d} x_1 \mathrm{d} x_2\cdots \mathrm{d} x_n $表示$ v(x_1, x_2, \cdots, x_n) $关于$ N(v) $的模. 还可以用如下方式计算加权Orlicz范数

$ \begin{align*} &\|f\|_{M, w_n}\\ =&\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+\int_0^{\infty}\int_0^{\infty}\cdots\int_0^{\infty}M(\alpha f(x_1, x_2, \cdots, x_n)w_n(x_1, x_2, \cdots, x_n)) \mathrm{d} x_1 \mathrm{d} x_2\cdots \mathrm{d} x_n). \end{align*} $

定义Orlicz空间内的连续模为

$ \omega(f;t)_M=:\sup\limits_{0\leq h\leq t}\|\Delta_hf(\cdot)\|_M(t\geq0), \; $

其中$ \Delta_hf(x)=:f(x+h)-f(x). $

  本文中C在不同处表示不同的正常数.

2 相关引理

引理2.1   $ \{T_n(f;a_n, b_n;x)\} $$ L_M^* $$ L_M^* $的有界正线性算子序列,且$ \|T_n(f;a_n, b_n;x)\|_M\leq\frac{b_1}{a_1}\|f\|_M. $

  因为

$ T_n(f;a_n, b_n;x)=:b_n\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{{(a_nx)}^k}{k!}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}f(t) \mathrm{d} t, $

其中

$ \begin{align*} \sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{{(a_nx)}^k}{k!}=1, \qquad \int_{0}^{\infty}e^{-a_nx}\frac{{(a_nx)}^k}{k!} \mathrm{d} x=\frac{1}{a_n}. \end{align*} $

故利用Jensen不等式,有

$ \begin{align*} &\|T_n(f;a_n, b_n;x)\|_{M}\\ =&\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+\int_0^{\infty}M(\alpha T_n(f;a_n, b_n;x)) \mathrm{d} x)\\ =&\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+\int_0^{\infty}M(\alpha b_n\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_nx)^k}{k!}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}f(t) \mathrm{d} t) \mathrm{d} x)\\ \leq&\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+\int_{0}^{\infty}\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_nx)^k}{k!}M(\alpha b_n\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}f(t) \mathrm{d} t) \mathrm{d} x)\\ \leq&\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+\int_{0}^{\infty}\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_nx)^k}{k!}b_n \int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}M(\alpha f(t)) \mathrm{d} t \mathrm{d} x)\\ =&\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+b_n\sum\limits_{k=0}^{\infty}\int_{0}^{\infty}e^{-a_nx}\frac{(a_nx)^k}{k!} \mathrm{d} x\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}M(\alpha f(t)) \mathrm{d} t)\\ =&\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+\frac{b_n}{a_n}\sum\limits_{k=0}^{\infty}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}M(\alpha f(t)) \mathrm{d} t)\\ =&\inf\limits_{\alpha>0}\frac{1}{\alpha}(1+\frac{b_n}{a_n}\int_{{0}}^{{\infty}}M(\alpha f(t)) \mathrm{d} t)\\ \leq&\frac{b_n}{a_n}\|f\|_{M}. \end{align*} $

因为$ \{\frac{b_n}{a_n}\}_{n=1}^{\infty} $是不增数列且趋于1,从而有

$ \begin{align*} \|T_n(f;a_n, b_n;x)\|_{M} \leq\frac{b_n}{a_n}\|f\|_{M} \leq\frac{b_1}{a_1}\|f\|_{M}. \end{align*} $

引理2.2[3]   令

$ \begin{equation*} y_x(t)=: \left\{ \begin{aligned} 1&, \; \; t\leq x\; \; \; \\ 0&, \; \; t>x\; \; \; \end{aligned} \right. (x\in R_0), \end{equation*} $

再令$ w(x)=\frac{1}{1+x} $,则存在与$ a_1 $$ b_1 $有关的正常数$ M_1(a_1, b_1) $,使得

$ \begin{align*} R_n(t)=: \int_0^{\infty}\left|T_n(y_x(t);a_n, b_n;x)-y_x(t)\right|w(x) \mathrm{d} x \leq M_1(a_1, b_1)\frac{1}{\sqrt{b_n}}. \end{align*} $

引理2.3   设$ f\in L_M^* $,且$ f'\in L_M^* $,则存在与$ a_1 $$ b_1 $有关的正常数$ M_2(a_1, b_1) $,使得

$ \begin{align*} \|(T_n(f;a_n, b_n;x)-f(x))w(x)\|_M\leq M_2(a_1, b_1)\|f'\|_M\frac{1}{\sqrt{b_n}}. \end{align*} $

   因为$ f(x)-f(0)= \int_0^xf'(t) \mathrm{d} t=\int_0^{\infty}f'(t)y_x(t) \mathrm{d} t, $并且

$ \begin{align*} T_n(f;a_n, b_n;x)-f(x)=\int_0^{\infty}f'(t)\{T_n(y_x(t);a_n, b_n;x)-y_x(t)\} \mathrm{d} t. \end{align*} $

根据引理2.2有

$ \begin{align*} &\|[T_n(f;a_n, b_n;x)-f(x)]w(x)\|_{M}\\ =&\sup\limits_{\rho(v;N)\leq1}\left|\int_0^{\infty}(T_n(f;a_n, b_n;x)-f(x))w(x)v(x) \mathrm{d} x\right|\\ \leq&C\left|\int_0^{\infty}\int_0^{\infty}f'(t)\{T_n(y_x(t);a_n, b_n;x)-y_x(t)\} \mathrm{d} tw(x) \mathrm{d} x\right|\\ \leq&C\left|\int_0^{\infty}f'(t)\int_0^{\infty}\{|T_n(y_x(t);a_n, b_n;x)-y_x(t)|\}w(x) \mathrm{d} x \mathrm{d} t\right|\\ =&C\left|\int_0^{\infty}f'(t)R_n(t) \mathrm{d} t\right|\\ \leq&C\|f'\|_{M}\|R_n(t)\|_{N}\\ =&C\|f'\|_{M}\sup\limits_{\rho(u;M)\leq1}\left|\int_0^{\infty}R_n(t)u(t) \mathrm{d} t\right|\\ \leq&C\|f'\|_{M}M_1(a_1, b_1)\frac{1}{\sqrt{b_n}}\sup\limits_{\rho(u;M)\leq1}|\int_{0}^{\infty}u(t) \mathrm{d} t| \end{align*} $

根据参考文献[5]中N函数的相关性质,我们有$ \lim\limits_{u\rightarrow \infty}\frac{M(u)}{u}=\infty, \; \; \lim\limits_{u\rightarrow0}\frac{M(u)}{u}=0. $从而有$ \sup\limits_{\rho(u;M)\leq1}\left| \int_0^{\infty}u(t) \mathrm{d} t\right|\leq C. $所以

$ \begin{align*} \|[T_n(f;a_n, b_n;x)-f(x)]w(x)\|_{M} \leq C\|f'\|_{M}M_1(a_1, b_1)\frac{1}{\sqrt{b_n}} \leq M_2(a_1, b_1)\|f'\|_{M}\frac{1}{\sqrt{b_n}}. \end{align*} $

引理2.4    $ \{T_{n, m}(f;a_n, b_n, c_m, d_m;x, y)\} $$ L_M^* $$ L_M^* $的有界正线性算子序列,且

$ \|T_{n, m}(f;a_n, b_n, c_m, d_m;x, y)\|_{M}\leq\frac{b_1d_1}{a_1c_1}\|f\|_M. $

   因为

$ \begin{align*} &e^{-a_nx}\sum\limits_{k=0}^{\infty}\frac{(a_nx)^k}{k!}=1, \; \; e^{-c_my}\sum\limits_{j=0}^{\infty}\frac{(c_my)^j}{j!}=1;\\ &\int_0^{\infty}e^{-a_nx}\frac{(a_nx)^k}{k!} \mathrm{d} x=\frac{1}{a_n}, \; \; \; \; \int_0^{\infty}e^{-c_my}\frac{(c_my)^j}{j!} \mathrm{d} y=\frac{1}{c_m}. \end{align*} $

从而

$ \begin{align*} &\|T_{n, m}(f;a_n, b_n, c_m, d_m;x, y)\|_{M}\\ =&\inf\limits_{\beta>0}\frac{1}{\beta}(1+\int_0^{\infty}\int_0^{\infty}M(\beta b_nd_me^{-a_nx}e^{-c_my}\\ &\times \sum\limits_{k=0}^{\infty}\sum\limits_{j=0}^{\infty}\frac{(a_nx)^k}{k!}\frac{(c_my)^j}{j!}\int_\frac{j}{d_m}^{\frac{j+1}{d_m}}\int_\frac{k}{b_n}^{\frac{k+1}{b_n}}f(u_1, u_2) \mathrm{d} u_1 \mathrm{d} u_2) \mathrm{d} x \mathrm{d} y)\\ \leq&\inf\limits_{\beta>0}\frac{1}{\beta}(1+\int_0^{\infty}\int_0^{\infty}e^{-a_nx}e^{-c_my}\\ &\times\sum\limits_{k=0}^{\infty}\sum\limits_{j=0}^{\infty}\frac{(a_nx)^k}{k!}\frac{(c_my)^j}{j!}M(\beta b_nd_m\int_\frac{j}{d_m}^{\frac{j+1}{d_m}}\int_\frac{k}{b_n}^{\frac{k+1}{b_n}}f(u_1, u_2) \mathrm{d} u_1 \mathrm{d} u_2) \mathrm{d} x \mathrm{d} y)\\ =&\inf\limits_{\beta>0}\frac{1}{\beta}(1+\sum\limits_{k=0}^{\infty}\sum\limits_{j=0}^{\infty}\int_0^{\infty}e^{-a_nx}\frac{(a_nx)^k}{k!} \mathrm{d} x\int_0^{\infty}e^{-c_my}\frac{(c_my)^j}{j!} \mathrm{d} y\\ &\times M(\beta b_nd_m\int_\frac{j}{d_m}^{\frac{j+1}{d_m}}\int_\frac{k}{b_n}^{\frac{k+1}{b_n}}f(u_1, u_2) \mathrm{d} u_1 \mathrm{d} u_2))\\ \leq&\inf\limits_{\beta>0}\frac{1}{\beta}(1+\sum\limits_{k=0}^{\infty}\sum\limits_{j=0}^{\infty}\frac{b_nd_m}{a_nc_m}\int_\frac{j}{d_m}^{\frac{j+1}{d_m}}\int_\frac{k}{b_n}^{\frac{k+1}{b_n}}M(\beta f(u_1, u_2)) \mathrm{d} u_1 \mathrm{d} u_2)\\ \leq&\frac{b_nd_m}{a_nc_m}\inf\limits_{\beta>0}\frac{1}{\beta}(1+\int_0^{\infty}\int_0^{\infty}M(\beta f(u_1, u_2)) \mathrm{d} u_1 \mathrm{d} u_2)\\ =&\frac{b_nd_m}{a_nc_m}\|f\|_M. \end{align*} $

由于$ \frac{\{b_n\}}{\{a_n\}}, \; \frac{\{d_m\}}{\{c_m\}} $是非减数列,从而有$ \|T_{n, m}(f;a_n, b_n, c_m, d_m;x, y)\|_{M}\leq\frac{b_1d_1}{a_1c_1}\|f\|_M. $

3 主要定理

定理3.1  设$ \{L_{n}\}_{n\in N} $$ L_{M}^{*}\rightarrow L_{M}^{*} $的一致有界正线性算子序列, 对于$ w_n(x_1, x_2, \cdots, x_n)=e^{-(x_1+x_2+\cdots+x_n)} $,若

(1) $ \lim\limits _{n\rightarrow \infty}\|L_n(1;x)-1\|_{M, w_n}=0, $

(2) $ \lim\limits_{n\rightarrow \infty}\|L_n(t_{i};x)-x_{i}\|_{M, w_n}=0, \; (i=1, 2, \cdots, n) $

(3) $ \lim\limits _{n\rightarrow \infty}\|L_n(|t|^2;x)-|x|^2\|_{M, w_n}=0 $,

则对任意$ f\in L_{M}^{*} $,有

$ \lim\limits_{n\rightarrow \infty}\|L_n(f)-f\|_{M, w_n}=0. $

其中$ x=(x_1, x_2, \cdots, x_n) $$ t=(t_1, t_2, \cdots, t_n) $,且$ |t| $$ |x| $表示$ n $维欧氏空间中的欧氏距离.

  类似于参考文献[6][7][8]的证明思路,对任何一个正数$ A $,设$ \chi_1^{A}(t) $$ n $维欧氏空间中球体$ 0\leq|t|\leq A $上的特征函数.

$ \chi_2^{A}(t)=1-\chi_1^{A}(t) $,则由于$ \chi_2^{A}(t)\rightarrow0(A \rightarrow +\infty) $,从而对任意$ \varepsilon>0 $,当$ A $充分大时,有$ |\chi_2^{A}(t)|<\varepsilon $. 而

$ \begin{align*} \|f\chi_2^{A}\|_{M} =&\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n);N)\leq1)}|\int_0^\infty\int_0^\infty\cdots\int_0^\infty f(t_1, t_2, \cdots, t_n)\\ &\times\chi_2^{A}(t_1, t_2, \cdots, t_n)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n| <\varepsilon. \end{align*} $

另一方面,由于$ \{L_n\} $$ L_{M}^{*}\rightarrow L_{M}^{*} $的一致有界正线性算子,故有

$ \|L_n(f)\|_{M}\leq K\|f\|_{M}. $

因为

$ \begin{align*} &\|L_n(f)-f\|_{M, w_n}\\ =&\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n);N)\leq1}|\int_0^\infty\int_0^\infty\cdots\int_0^\infty (L_nf-f)v(t_1, t_2, \cdots, t_n)\\ &\times e^{-(t_1+t_2+\cdots+t_n)} \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n|\\ \leq&\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n);N)\leq1}\left|\int_0^\infty\int_0^\infty\cdots\int_0^\infty (L_nf-f)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n\right|\\ =&\|L_nf-f\|_M\\ =&\|L_n[(\chi_1^{A}+\chi_2^{A})f]-(\chi_1^{A}+\chi_2^{A})f\|_{M}\\ =&\|L_n(\chi_1^{A}f)+L_n(\chi_2^{A}f)-\chi_1^{A}f-\chi_2^{A}f\|_{M}\\ \leq&\|L_n(\chi_1^{A}f)-\chi_1^{A}f\|_{M}+\|L_n(\chi_2^{A}f)-\chi_2^{A}f\|_{M}\\ =:&I_{n}^{'}+I_{n}^{''}. \end{align*} $

先讨论$ I_{n}^{''} $,由线性算子$ \{L_n\} $的一致有界性,有$ \|L_n(\chi_2^{A}f)-\chi_2^{A}f\|_{M}\leq(K+1)\varepsilon. $

因为连续函数类在$ L_{M}^*[0, A] $中稠密,故对任意$ \varepsilon_{1}>0 $,存在$ [0, A] $上的连续函数$ \varphi(t) $(当$ t>A $时,令$ \varphi(t)=0) $,使得

$ \begin{equation*} \|(f-\varphi)\chi_1^{A}\|_{M}<\frac{\varepsilon_1}{(K+1)}. \end{equation*} $

从而

$ \begin{align} I_{n}^{'}=& \|L_n(\chi_1^{A}f)-\chi_1^{A}f\|_{M}\\ \leq&\|L_n[\chi_1^{A}(f-\varphi)]\|_{M}+\|L_n(\varphi\chi_1^{A})-\varphi\chi_1^{A}\|_{M}+\|(f-\varphi)\chi_1^{A}\|_{M}\\ \leq&\|L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}\|_{M}+(K+1)\|(f-\varphi)\chi_1^{A}\|_{M}\\ <&\|L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}\|_{M}+\varepsilon_{1}. \end{align} $ (3.1)

又因为对于$ A_1>A $,有$ \chi_{2}^{A_1}\chi_{1}^{A}\varphi=0 $,所以

$ \begin{align} \|L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}\|_{M} \leq&\|[L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}]\chi_1^{A_1}\|_{M}+\|\chi_2^{A_1}L_n(\varphi\chi_1^{A})\|_{M}+\|\varphi\chi_1^{A}\chi_2^{A_1}\|_{M}\\ =&\|[L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}]\chi_1^{A_1}\|_{M}+\|\chi_2^{A_1}L_n(\varphi\chi_1^{A})\|_{M}. \end{align} $ (3.2)

我们选定$ A_1 $,使

$ |\int_{0}^\infty\int_{0}^\infty\cdots\int_{0}^\infty\chi_2^{A_1}(t_1, t_2, \cdots, t_n)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n|<\frac{\varepsilon_{1}}{M_{\varphi}}, $

其中$ M_{\varphi}=\max\limits_{t\in R_0^n}|\varphi(t)|\chi_{1}^{A_1}(t) $.

从而

$ \begin{align*} &\|\chi_{2}^{A_1}L_n(\varphi\chi_{1}^{A})\|_{M}\\ =&\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n;N)\leq1)}\left|\int_{0}^{\infty}\int_{0}^{\infty}\cdots\int_{0}^{\infty}(1-\chi_1^{A_1})L_n(\varphi\chi_1^A)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n\right|\nonumber\\ \leq&\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n;N)\leq1)}\left|\int_{A_1}^{\infty}\int_{A_1}^{\infty}\cdots\int_{A_1}^{\infty}L_n(\varphi\chi_1^A)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n\right|\\ \leq&\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n;N)\leq1)}\left|\int_{0}^{\infty}\int_{0}^{\infty}\cdots\int_{0}^{\infty}L_n(\varphi\chi_1^A)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n\right|\nonumber\\ \leq&\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n;N)\leq1)}\left|\int_{0}^{\infty}\int_{0}^{\infty}\cdots\int_{0}^{\infty}M_{\varphi}L_n(1;x)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n\right|\nonumber\\ \leq&M_{\varphi}\left\{\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n;N)\leq1)}\left|\int_{0}^{\infty}\int_{0}^{\infty}\cdots\int_{0}^{\infty}(L_n(1;x)-1)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n\right|\right\}\\ &+ M_{\varphi}\left\{\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n;N)\leq1)}\left|\int_{0}^{\infty}\int_{0}^{\infty}\cdots\int_{0}^{\infty}\chi_2^{A_1}v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n\right|\right\}\\ \leq&M_{\varphi}\left\{\sup\limits_{\rho(v(t_1, t_2, \cdots, t_n;N)\leq1)}\left|\int_{0}^{\infty}\int_{0}^{\infty}\cdots\int_{0}^{\infty}(L_n(1;x)-1)v(t_1, t_2, \cdots, t_n) \mathrm{d} t_1 \mathrm{d} t_2\cdots \mathrm{d} t_n\right|\right\}\\&+\varepsilon_{1}. \end{align*} $

因此得

$ \begin{align} \|\chi_{2}^{A_1}L_n(\varphi\chi_1^{A})\|_{M}\leq M_{\varphi}\|L_n(1;x)-1\|_{M}+\varepsilon_{1}. \end{align} $ (3.3)

再将(3.3)代入(3.2)得

$ \begin{align} \|L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}\|_{M}\leq\|[L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}]\chi_1^{A_1}\|_{M} +M_{\varphi}\|L_n(1;x)-1\|_{M}+\varepsilon_{1}. \end{align} $ (3.4)

最后将(3.4)代入(3.1)得

$ \begin{equation} I_n^{'}\leq2\varepsilon_{1}+M_{\varphi}\|L_n(1;x)-1\|_{M}+\|[L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}]\chi_1^{A_1}\|_{M}. \end{equation} $ (3.5)

根据定理的条件,对于充分大的$ n $,有$ \|L_n(1;x)-1\|_{M}<\frac{\varepsilon_{1}}{M_{\varphi}}. $又因为$ \varphi\chi_1^{A} $$ [0, A] $上为连续函数,所以对于上述$ \varepsilon_{1}>0 $,存在$ \delta>0 $,当$ \; x, \; t\in R_0^n $,且$ |t-x|<\delta $时,有

$ |\varphi(t)\chi_1^{A}(t)-\varphi(x)\chi_1^{A}(x)|<\varepsilon_{1}+2 M_{\varphi}\frac{(t-x)^2}{\delta^2}. $

从而

$ \begin{align} &\|[L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}]\chi_{1}^{A_1}\|_{M}\\ \leq&\|[L_n(|\varphi(t)\chi_1^{A}(t)-\varphi(x)\chi_1^{A}(x)|);x)]\chi_1^{A_1}\|_{M}+\|\varphi(x)\chi_1^{A_1}(x)(L_n(1;x)-1)\|_{M}\\ \leq&\|[L_n|\varepsilon_{1}+2M_{\varphi}\frac{(t-x)^2}{\delta^2}|;x]\chi_1^{A_1}\|_{M}+\|\varphi(x)\chi_1^{A_1}(x)(L_n(1;x)-1)\|_{M}\\ \leq&\varepsilon_{1}\|L_n(1;x)\|_{M}+ \frac{2M_{\varphi}}{\delta^2}\|L_n(|t-x|^2;x)\|_{M}+M_{\varphi}\|L_n(1;x)-1\|_{M}\\ \leq&\varepsilon_{1}K\|1\|_{M}+ \frac{2M_{\varphi}}{\delta^2}\|L_n(|t-x|^2;x)\|_{M}+\varepsilon_{1}. \end{align} $ (3.6)

下面只对$ \|L_n(|t-x|^2;x)\|_{M} $进行研究. 根据已知条件,当$ n $充分大时,对于上述$ \varepsilon_1>0 $,有

$ \begin{align*} \text{max}\left\{\|L_n(|t|^2;x)-|x|^2\|_{M}, \sum\limits_{i=1}^{n}\|L_n(t_i;x)-x_i\|_{M}, \|L_n(1;x)-1\|_{M}\right\}<\frac{\varepsilon_1\delta^2}{2M_\varphi(1+A)^2}. \end{align*} $

$ \begin{align} &\|L_n(|t-x|^2;x)\|_{M}\\ =&\|L_n(t^2+x^2-2tx;x)\|_{M}\\ \leq&\|L_n(t^2;x)-x^2\|_{M}+x^2\|L_n(1;x)-1\|_{M}+2\sum\limits_{i=1}^{\infty}x_i\|x_i-L_n(t_i;x_i)\|_{M}\\ \leq&\|L_n(t^2;x)-x^2\|_{M}+2A\sum\limits_{i=1}^{\infty}\|x_i-L_n(t_i;x_i)\|_{M}+A^2\|L_n(1;x)-1\|_{M}\\ \leq&(1+A)^2max\left\{\|L_n(|t|^2;x)-|x|^2\|_{M}, \sum\limits_{i=1}^{n}\|L_n(t_i;x)-x_i\|_{M}, \|L_n(1;x)-1\|_{M}\right\}\\ <&\frac{\varepsilon_1\delta^2}{2M_{\varphi}}. \end{align} $

所以

$ \begin{align} \|L_n(|t-x|^2;x)\|_{M}\leq\frac{\varepsilon_{1}\delta^2}{2M_\varphi}. \end{align} $ (E3.7)

将(3.7)代入(3.6)得

$ \begin{equation} \|[L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}]\chi_{1}^{A_1}\|_{M}<\varepsilon_{1}(2+K||1||_{M}). \end{equation} $ (3.8)

最后将(3.8)代入(3.5)得

$ \begin{align*} I_n^{'}\leq&2\varepsilon_{1}+M_{\varphi}\|L_n(1;x)-1\|_{M}+\| [L_n(\varphi\chi_{1}^{A})-\varphi\chi_{1}^{A}]\chi_1^{A_1}\|_{M}\\ \leq&5\varepsilon_{1}+K\|1\|_{M}\varepsilon_{1}. \end{align*} $

从而

$ \|L_n(f)-f\|_{M, w_n}\leq I_n^{'}+I_n^{''} \leq (K+1)\varepsilon+(5+K\|1\|_{M})\varepsilon_{1}. $

所以由$ \varepsilon $$ \varepsilon_{1} $的任意性,有

$ \lim\limits_{n\rightarrow \infty}\|L_n(f)-f\|_{M, w_n}=0. $

定理3.2  设$ w(x)=\frac{1}{1+x} $,则对任意$ f\in L_M^* $,存在与$ a_1, b_1 $有关的常数$ M_3(a_1, b_1) $,使得

$ \begin{align*} \|(T_n(f;a_n, b_n;x)-f(x))w(x)\|_{M}\leq M_3(a_1, b_1)\omega(f;\frac{1}{\sqrt{b_n}})_M. \end{align*} $

  利用Steklov平均函数

$ \begin{align*} f_h(x)=\frac{1}{h}\int_0^hf(x+u) \mathrm{d} u, x\in R_0, h>0. \end{align*} $

对于$ f_h $我们有

$ \begin{align*} \|(f-f_h)w\|_M=&\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty(f-f_h)w(x)v(x) \mathrm{d} x\right|\\ =&\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty(f-\frac{1}{h}\int_0^hf(x+u) \mathrm{d} u)w(x)v(x) \mathrm{d} x\right|\\ \leq&\sup\limits_{\rho(v;N)\leq1}\frac{1}{h}\left|\int_0^\infty\int_0^h\left|f(x+u)-f(x)\right|w(x)v(x) \mathrm{d} x \mathrm{d} u\right|\\ =&\frac{1}{h}\int_0^h\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty\left|f(x+u)-f(x)\right|v(x)w(x) \mathrm{d} x\right| \mathrm{d} u\\ \leq&\frac{1}{h}\int_0^h\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty\left|f(x+u)-f(x)\right|v(x) \mathrm{d} x\right| \mathrm{d} u\\ \leq&\frac{1}{h}\int_0^h\omega(f;h)_M \mathrm{d} u\\ =&\omega(f;h)_M. \end{align*} $

又因为$ f'_h(x)=h^{-1}\Delta_hf(x) $,所以

$ \begin{align*} \|f'_h(x)\|_M=&\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty h^{-1}\Delta_hf(x)v(x) \mathrm{d} x\right|\\ =&h^{-1}\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty \Delta_hf(x)v(x) \mathrm{d} x\right|\\ \leq&h^{-1}\omega(f;h)_M. \end{align*} $

故根据引理2.3,有

$ \begin{align*} \|(T_n(f_h)-f_h)w(x)\|_M \leq M_2(a_1, b_1)\frac{1}{\sqrt{b_n}}\|f'_h\|_M \leq M_2(a_1, b_1)\frac{1}{\sqrt{b_n}}h^{-1}\omega(f;h)_M. \end{align*} $

当取$ h=\frac{1}{\sqrt{b_n}} $时,有$ \|(T_n(f_h)-f_h)w(x)\|_M\leq M_2(a_1, b_1)\omega(f;\frac{1}{\sqrt{b_n}})_M. $利用插项逼近法,有

$ \begin{align*} \|(T_n(f)-f)w(x)\|_M =&\|T_n(f-f_h)w\|_M+\|(T_n(f_h)-f_h)w\|_M+\|(f_h-f)w\|_M\\ \leq&\frac{b_1}{a_1}\|(f-f_h)w\|_M+M_2(a_1, b_1)\omega(f;\frac{1}{\sqrt{b_n}})_M+\|(f_h-f)w\|_M\\ \leq&(\frac{b_1}{a_1}+1)\omega(f;\frac{1}{\sqrt{b_n}})_M+M_2(a_1, b_1)\omega(f;\frac{1}{\sqrt{b_n}})_M\\ \leq& M_3(a_1, b_1)\omega(f;\frac{1}{\sqrt{b_n}})_M. \end{align*} $

定理3.3  设$ \{T_n(f;a_n, b_n;x)\} $$ L_M^* $$ L_M^* $一致有界正线性算子序列,$ w_1(x)=e^{-x} $,则对任意$ f\in L_M^* $,有

$ \begin{align*} \lim\limits_{n\rightarrow \infty}\|T_n(f;a_n, b_n;x)-f(x)\|_{M, w_1}=0. \end{align*} $

  因为

$ \begin{align*} T_n(1;a_n, b_n;x)=&b_n\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_n x)^k}{k!}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}1\; \; \mathrm{d} t =\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_n x)^k}{k!} =1.\; \; \; \; \; \; \; \\ T_n(t;a_n, b_n;x)=&b_n\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_n x)^k}{k!}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}t\; \; \mathrm{d} t =b_n\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_n x)^k}{k!}\frac{2k+1}{2b_n^2}\\ =&\frac{a_nx}{b_n}\sum\limits_{k=1}^{\infty}e^{-a_nx}\frac{(a_nx)^{k-1}}{(k-1)!}+\frac{1}{2b_n} =\frac{a_nx}{b_n}+\frac{1}{2b_n}. \end{align*} $
$ \begin{align*} T_n(t^2;a_n, b_n;x)=&b_n\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_n x)^k}{k!}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}t^2\; \; \mathrm{d} t\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\\ =&b_n\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_n x)^k}{k!}\frac{3k^2+3k+1}{3b_n^3}\\ =&\frac{a_nx}{b_n^2}\sum\limits_{k=1}^{\infty}e^{-a_nx}\frac{(a_nx)^{k-1}}{(k-1)!}k+\frac{a_nx}{b_n^2}+\frac{1}{3b_n^2}\\ =&\frac{a_nx}{b_n^2}\sum\limits_{k=0}^{\infty}e^{-a_nx}\frac{(a_nx)^{k}}{k!}(k+1)+\frac{a_nx}{b_n^2}+\frac{1}{3b_n^2}\\ =&\frac{(a_nx)^2}{b_n^2}\sum\limits_{k=1}^{\infty}e^{-a_nx}\frac{(a_nx)^{k-1}}{(k-1)!}+\frac{2a_nx}{b_n^2}+\frac{1}{3b_n^2}\\ =&\frac{(a_nx)^2}{b_n^2}+\frac{2a_nx}{b_n^2}+\frac{1}{3b_n^2}. \end{align*} $

所以

$ \begin{align*} \|T_n(1;a_n, b_n;x)-1\|_{M, w_1} =&\|1-1\|_{M, w_1}=0.\; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \\ \|T_n(t;a_n, b_n;x)-x\|_{M, w_1} =&\|\frac{a_nx}{b_n}+\frac{1}{2b_n}-x\|_{M, w_1}\\ \leq&(\frac{a_n}{b_n}-1)\|x\|_{M, w_1}+\|\frac{1}{2b_n}\|_{M, w_1}\\ =&(\frac{a_n}{b_n}-1)\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty xv(x)e^{-x} \mathrm{d} x\right|+\frac{1}{2b_n}\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty v(x)e^{-x} \mathrm{d} x\right|\\ \leq&C(\frac{a_n}{b_n}-1)\int_0^\infty xe^{-x} \mathrm{d} x+\frac{C}{2b_n}\int_0^\infty e^{-x} \mathrm{d} x\\ =&C(\frac{a_n}{b_n}-1)+\frac{C}{2b_n}.\; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \\ \|T_n(t^2;a_n, b_n;x)-x^2\|_{M, w_1} =&\|\frac{(a_nx)^2}{b_n^2}+\frac{2a_nx}{b_n^2}+\frac{1}{3b_n^2}-x^2\|_{M, w_1}\\ \leq&(\frac{a_n^2}{b_n^2}-1)\|x^2\|_{M, w_1}+\frac{2a_n}{b_n^2}\|x\|_{M, w_1}+\frac{1}{3b_n^2}\|1\|_{M, w_1}\\ \leq&(\frac{a_n^2}{b_n^2}-1)\sup\limits_{\rho(v;N)\leq1}\left|\int_0^\infty x^2e^{-x}v(x) \mathrm{d} x\right|+C\frac{2a_n}{b_n^2}+\frac{C}{3b_n^2}\\ =&C(\frac{a_n^2}{b_n^2}-1)+C\frac{2a_n}{b_n^2}+\frac{C}{3b_n^2}. \end{align*} $

从而有

$ \begin{align*} &\lim\limits_{n\rightarrow \infty}\|T_n(1;a_n, b_n;x)-1\|_{M, w_1}=0;\\ &\lim\limits_{n\rightarrow \infty}\|T_n(t;a_n, b_n;x)-x\|_{M, w_1}=0;\\ &\lim\limits_{n\rightarrow \infty}\|T_n(t^2;a_n, b_n;x)-x^2\|_{M, w_1}=0. \end{align*} $

故利用定理3.1得,对任意$ f\in L_M^* $,有

$ \begin{align*} \lim\limits_{n\rightarrow \infty}\|T_n(f;a_n, b_n;x)-f(x)\|_{M, w_1}=0. \end{align*} $

定理3.4   设$ \{T_{n, m}(f;a_n, b_n, c_m, d_m;x, y)\} $$ L_M^* $$ L_M^* $的一致有界正线性算子序列,$ w_2(x, y)=e^{-(x+y)} $,则对任意的$ f\in L_M^* $,有

$ \begin{align*} \lim\limits_{n, m\rightarrow \infty}\|T_{n, m}(f;a_n, b_n, c_m, d_m;x, y)-f(x, y)\|_{M, w_2}=0. \end{align*} $

  因为

$ \begin{align*} &T_{n, m}(1;a_n, b_n, c_m, d_m;x, y)\\ =&b_nd_me^{-a_nx}e^{-c_my}\sum\limits_{k=0}^\infty\sum\limits_{j=0}^\infty\frac{(a_nx)^k}{k!}\frac{(c_my)^j}{j!}\int_{\frac{j}{d_m}}^{\frac{j+1}{d_m}}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}} 1 \mathrm{d} u_1 \mathrm{d} u_2 =1.\; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \end{align*} $
$ \begin{align*} &T_{n, m}(u_1;a_n, b_n, c_m, d_m;x, y)\\ =&b_nd_me^{-a_nx}e^{-c_my}\sum\limits_{k=0}^\infty\sum\limits_{j=0}^\infty\frac{(a_nx)^k}{k!}\frac{(c_my)^j}{j!} \int_{\frac{j}{d_m}}^{\frac{j+1}{d_m}}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}}u_1 \mathrm{d} u_1 \mathrm{d} u_2\\ =&b_n(e^{-a_nx}\sum\limits_{k=0}^\infty\frac{(a_nx)^k}{k!}\frac{2k+1}{2b_n^2}\sum\limits_{j=0}^\infty e^{-c_my}\frac{(c_my)^j}{j!}) =b_n\sum\limits_{k=0}^\infty e^{-a_nx}\frac{(a_nx)^k}{k!}\frac{2k+1}{2b_n^2}\\ =&\frac{a_nx}{b_n}\sum\limits_{k=1}^{\infty} e^{-a_nx}\frac{(a_nx)^{k-1}}{(k-1)!}+\frac{1}{2b_n} =\frac{a_n}{b_n}x+\frac{1}{2b_n}. \end{align*} $

同理可求得

$ \begin{align*} T_{n, m}(u_2;a_n, b_n, c_m, d_m;x, y) =\frac{c_m}{d_m}y+\frac{1}{2d_m}. \end{align*} $

$ \begin{align*} &T_{n, m}(u_1^2+u_2^2;a_n, b_n, c_m, d_m;x, y)\\ =&b_nd_me^{-a_nx}e^{-c_my}\sum\limits_{k=0}^\infty\sum\limits_{j=0}^\infty\frac{(a_nx)^k}{k!}\frac{(c_my)^j}{j!}\int_{\frac{j}{d_m}}^{\frac{j+1}{d_m}}\int_{\frac{k}{b_n}}^{\frac{k+1}{b_n}} (u_1^2+u_2^2) \mathrm{d} u_1 \mathrm{d} u_2\\ =&b_nd_me^{-a_nx}e^{-c_my}\sum\limits_{k=0}^\infty\sum\limits_{j=0}^\infty\frac{(a_nx)^k}{k!}\frac{(c_my)^j}{j!}\frac{3k^2+3k+1}{3b_n^3 d_m}\\&+ b_nd_me^{-a_nx}e^{-c_my}\sum\limits_{k=0}^\infty\sum\limits_{j=0}^\infty\frac{(a_nx)^k}{k!}\frac{(c_my)^j}{j!}\frac{3j^2+3j+1}{3d_m^3 b_n}\\ =&\frac{(a_nx)^2}{b_n^2}+\frac{2a_nx}{b_n^2}+\frac{1}{3b_n^2}+\frac{(c_my)^2}{d_m^2}+\frac{2c_my}{d_m^2}+\frac{1}{3d_m^2}.\; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \; \end{align*} $

对于

$ \begin{align*} \|1\|_{M, w_2}=&\sup\limits_{\rho(v(x, y);N)\leq1}\left|\int_0^\infty\int_0^\infty e^{-(x+y)}v(x, y) \mathrm{d} x \mathrm{d} y\right|\\ \leq &C\left|\int_0^\infty e^{-x} \mathrm{d} x\right|\left|\int_0^\infty e^{-y} \mathrm{d} y\right| =C.\\ \|x\|_{M, w_2}=&\sup\limits_{\rho(v(x, y);N)\leq1}\left|\int_0^\infty\int_0^\infty xe^{-(x+y)}v(x, y) \mathrm{d} x \mathrm{d} y\right|\\ \leq &C\left|\int_0^\infty xe^{-x} \mathrm{d} x\right|\left|\int_0^\infty e^{-y} \mathrm{d} y\right| =C.\\ \|x^2\|_{M, w_2}=&\sup\limits_{\rho(v(x, y);N)\leq1}\left|\int_0^\infty\int_0^\infty x^2e^{-(x+y)}v(x, y) \mathrm{d} x \mathrm{d} y\right|\\ \leq&C\left|\int_0^\infty x^2e^{-x} \mathrm{d} x\right|\left|\int_0^\infty e^{-y} \mathrm{d} y\right| =C. \end{align*} $

同理我们可以求得

$ \|y\|_{M, w_2}\leq C, \; \; \; \|y^2\|_{M, w_2}\leq C. $

所以

$ \begin{align*} &\lim\limits_{n, m\rightarrow \infty}\|T_{n, m}(1;a_n, b_n, c_m, d_m;x, y)-1\|_{M, w_2}=0;\\ &\lim\limits_{n, m\rightarrow \infty}\|T_{n, m}(u_1;a_n, b_n, c_m, d_m;x, y)-x\|_{M, w_2}\\ =&\lim\limits_{n\rightarrow \infty}\|\frac{a_n}{b_n}x+\frac{1}{2b_n}-x\|_{M, w_2} \leq\lim\limits_{n\rightarrow \infty}[(\frac{a_n}{b_n}-1)\|x\|_{M, w_2}+\frac{1}{2b_n}\|1\|_{M, w_2}]\\ \leq&\lim\limits_{n\rightarrow \infty}[C(\frac{a_n}{b_n}-1)+\frac{C}{2b_n}] =0. \end{align*} $

同理可得

$ \begin{align*} &\lim\limits_{n, m\rightarrow \infty}\|T_{n, m}(u_2;a_n, b_n, c_m, d_m;x, y)-y\|_{M, w_2}=0. \end{align*} $

$ \begin{align*} &\lim\limits_{n, m\rightarrow \infty}\|T_{n, m}(u_1^2+u_2^2;a_n, b_n, c_m, d_m;x, y)-(x^2+y^2)\|_{M, w_2}\\ =&\lim\limits_{n, m\rightarrow \infty}\|(\frac{a_n^2}{b_n^2}-1)x^2+\frac{2a_nx}{b_n^2}+\frac{1}{3b_n^2}+(\frac{c_m^2}{d_m^2}-1)y^2+\frac{2c_my}{d_m^2}+\frac{1}{3d_m^2}\|_{M, w_2}\\ \leq&\lim\limits_{n, m\rightarrow \infty}\{(\frac{a_n^2}{b_n^2}-1)\|x^2\|_{M, w_2}+(\frac{c_m^2}{d_m^2}-1)\|y^2\|_{M, w_2}+ \frac{2a_n}{b_n^2}\|x\|_{M, w_2}\\&+\frac{2c_m}{d_m^2}\|y\|_{M, w_2}+(\frac{1}{3b_n^2}+\frac{1}{3d_m^2})\|1\|_{M, w_2}\}\\ \leq&\lim\limits_{n, m\rightarrow \infty}C\{(\frac{a_n^2}{b_n^2}-1)+(\frac{c_m^2}{d_m^2}-1)+ \frac{2a_n}{b_n^2}+\frac{2c_m}{d_m^2}+(\frac{1}{3b_n^2}+\frac{1}{3d_m^2})\}\\ =&0. \end{align*} $

从而根据定理3.1得,对任意的$ f\in L_M^* $,有

$ \begin{align*} \lim\limits_{n, m\rightarrow \infty}\|T_{n, m}(f;a_n, b_n, c_m, d_m;x, y)-f(x, y)\|_{M, w_2}=0. \end{align*} $
参考文献
[1] Becker M. Global approximation theorems for Szasz-Mirakjan and Baskakov operators in polynomial weight spaces[J]. Indiana University Mathematics Journal, 1978, 27(1): 127–142. DOI:10.1512/iumj.1978.27.27011
[2] Butzer P L. On the extensions of Bernstein polynomials to the infinite interval[J]. Proceedings of the American Mathematical Society, 1954, 5(4): 547–553. DOI:10.1090/S0002-9939-1954-0063483-7
[3] Walczak Z. On approximation by modified Szasz-Mirakyan operators[J]. Glasnik Matematicki, 2002, 37(57): 303–319.
[4] Ispir N, Atakut C. Approximation by modified Szasz—Mirakjan operators on weighted spaces[J]. Proceedings of the Indian Academy of Sciences-Mathematical Sciences, 2002, 112(4): 571–578.
[5] 吴从炘, 王延辅. 奥尔里奇空间及其应用[M]. 哈尔滨: 黑龙江科学技术出版社, 1983.
[6] Gadjiev A D, Aral A. Weighted Lp-approximation with positive linear operators on unbounded sets[J]. Applied Mathematics letters, 2007, 20(10): 1046–1051. DOI:10.1016/j.aml.2006.12.007
[7] 姜胜楠, 吴嘎日迪. 修正的Picard算子在Orlicz空间中的指数加权逼近[J]. 应用数学, 2023, 36(03): 631–639.
[8] 李昕昕, 吴嘎日迪. Ibragimov-Gadjiev-Durrmeyer算子在Orlicz空间内的逼近性质[J]. 数学杂志, 2022, 42(3): 237–245.