文章快速检索     高级检索
  华东师范大学学报(自然科学版)  2018 Issue (4): 23-31  DOI: 10.3969/j.issn.1000-5641.2018.04.003
0

引用本文  

杨月英, 马萍. 关于, Neuman-Sándor, 平均的两个最佳不等式[J]. 华东师范大学学报(自然科学版), 2018, (4): 23-31. DOI: 10.3969/j.issn.1000-5641.2018.04.003.
YANG Yue-ying, MA Ping. Two optimal inequalities for Neuman-Sándor means[J]. Journal of East China Normal University (Natural Science), 2018, (4): 23-31. DOI: 10.3969/j.issn.1000-5641.2018.04.003.

基金项目

湖州职业技术学院教改课题,(2016xj26);浙江广播电视大学科学研究课题,(XKT-17G26)

作者简介

杨月英, 女, 副教授, 研究方向为解析不等式.E-mail:2004002@hzvtc.net.cn

文章历史

收稿日期:2017-03-27
关于, Neuman-Sándor, 平均的两个最佳不等式
杨月英, 马萍     
湖州职业技术学院 机电与汽车工程学院, 浙江 湖州 313000
摘要:运用实分析方法, 研究了Neuman-Sándor平均$M(a, b)$与第二类反调和平均$D(a, b)$和调和根平方平均$\overline{H}(a, b)$ (及调和平均$H(a, b)$)凸组合的序关系.发现了最大值$\lambda_{1}, \lambda_{2}\in(0, 1)$和最小值$\mu_{1}, \mu_{2}\in(0, 1)$使得双边不等式 $ \lambda_{1}D(a,b)+(1-\lambda_{1})\overline{H}(a,b) <M(a,b)<\mu_{1}D(a,b)+(1-\mu_{1})\overline{H}(a,b), \\ \lambda_{2}D(a,b)+(1-\lambda_{2})H(a,b)<M(a,b)<\mu_{2}D(a,b)+(1-\mu_{2})H(a,b) $ 对所有$a, b>0$$a\neq b$成立.
关键词Neuman-Sándor平均    反调和平均    根平方平均    调和平均    不等式    
Two optimal inequalities for Neuman-Sándor means
YANG Yue-ying, MA Ping    
Mechanic Electronic and Automobile Engineering College, Huzhou Vocational & Technical College, Huzhou Zhejiang 313000, China
Abstract: This paper deals with the inequalities involving Neuman-Sándor means using methods of real analysis. The convex combinations of the second contra-harmonic mean $D(a, b)$ and the harmonic root-square mean $\overline{H}(a, b)$ (or harmonic mean $H(a, b)$) for the Neuman-Sándor mean $M(a, b)$ are discussed. We find the maximum values $\lambda_{1}, \lambda_{2}\in(0, 1)$ and the minimum values $\mu_{1}, \mu_{2}\in(0, 1)$ such that the two-sided inequalities $ \lambda_{1}D(a,b)+(1-\lambda_{1})\overline{H}(a,b) <M(a,b)<\mu_{1}D(a,b)+(1-\mu_{1})\overline{H}(a,b), \\ \lambda_{2}D(a,b)+(1-\lambda_{2})H(a,b)<M(a,b)<\mu_{2}D(a,b)+(1-\mu_{2})H(a,b) $ hold for all $a, b>0$ with $a\neq b$.
Key words: Neuman-Sándor mean    contra-harmonic mean    root-square mean    harmonic mean    inequalities    
0 引言

对于$a$, $b>0$$a\neq b$, Neuman-Sándor平均[1]定义为

$ M(a, b) =\frac{a-b}{2{\rm arcsinh}\big[(a-b)/(a+b)\big]}, $ (1)

其中${\rm arcsinh}(x)=\ln(x+\sqrt{x^2+1})$是反双曲正弦函数.

$\overline{H}(a, b)=\sqrt{2}ab/\sqrt{a^2+b^2}$, $H(a, b)=2ab/(a+b)$, $G(a, b)=\sqrt{ab}$, $L(a, b)=(a-b)/(\ln a-\ln b)$, $P(a, b)=(a-b)/\big[2\arcsin\big((a-b)/(a+b)\big)\big]$, $A(a, b)=(a+b)/2$, $T(a, b)=(a-b)/\big[2\arctan\big((a-b)/(a+b)\big)\big]$, $E(a, b)=2(a^2+ab+b^2)/\big[3(a+b)\big]$, $Q(a, b)=\sqrt{(a^2+b^2)/2}$, $C(a, b)=(a^2+b^2)/(a+b)$$D(a, b)=(a^3+b^3)/(a^2+b^2)$分别是调和根平方平均、调和平均、几何平均、对数平均、第一类Seiffert平均、算术平均、第二类Seiffert平均、形心平均、二次平均、反调和平均和第二类反调和平均.我们有熟知不等式

$ \overline{H}(a, b) <H(a, b)< G(a, b)<L(a, b)<P(a, b)<A(a, b) \\ <M(a, b)<T(a, b)<E(a, b)<Q(a, b)<C(a, b)<D(a, b), $ (2)

对所有$a$, $b>0$$a\neq b$成立.

近年来, Neuman-Sándor平均和其他二元平均得到了深入的研究.特别地, 从有关Neuman-Sándor平均与其他二元平均和它们的各类组合比较中发现了许多重要不等式, 参见文献[2-6].

Neuman和Sándor[1, 7]证明了不等式

$ A(a, b) <M(a, b)<\frac{A(a, b)}{\ln(1+\sqrt{2})}, \frac{\pi}{4}T(a, b)<M(a, b)<T(a, b), \\ M(a, b)<\frac{A^{2}(a, b)}{P(a, b)}, M(a, b)<\frac{2A(a, b)+Q(a, b)}{3}, \\ \sqrt{A(a, b)T(a, b)}<M(a, b)<\sqrt{\frac{A^{2}(a, b)+T^{2}(a, b)}{2}}, $ (3)

对所有$a$, $b>0$$a\neq b$成立.

在文献[8]中, Neuman证明了不等式

$ \alpha Q(a, b)+(1-\alpha)A(a, b) <M(a, b) <\beta Q(a, b)+(1-\beta)A(a, b), \\ \lambda C(a, b)+(1-\lambda)A(a, b)<M(a, b)<\mu C(a, b)+(1-\mu)A(a, b) $ (4)

对所有$a$, $b>0$$a\neq b$成立当且仅当$\alpha \leqslant \big[1-\ln(1+\sqrt{2})\big]/\big[(\sqrt{2}-1)\ln(1+\sqrt{2})\big]=0.324~9\cdots$, $\beta \geqslant 1/3$, $\lambda \leqslant \big[1-\ln(1+\sqrt{2})\big]/\ln(1+\sqrt{2})=0.134~5\cdots$$\mu \geqslant 1/6$.

赵铁洪等[9]发现了最佳参数$\alpha_{1}$, $\alpha_{2}$, $\alpha_{3}$, $\beta_{1}$, $\beta_{2}$, $\beta_{3}\in(0, 1)$使得双边不等式

$ \alpha_{1} H(a, b)+(1-\alpha_{1})Q(a, b)<M(a, b)<\beta_{1} H(a, b)+(1-\beta_{1})Q(a, b), \\ \alpha_{2} G(a, b)+(1-\alpha_{2})Q(a, b)<M(a, b)<\beta_{2} G(a, b)+(1-\beta_{2})Q(a, b), \\ \alpha_{3} H(a, b)+(1-\alpha_{3})C(a, b)<M(a, b)<\beta_{3} H(a, b)+(1-\beta_{3})C(a, b) $ (5)

对所有$a$, $b>0$$a\neq b$成立.

钱伟茂和禇玉明[10]发现了最大值$\alpha_{1}$, $\lambda_{1}$, $\alpha_{2}$, $\lambda_{2}$和最小值$\beta_{1}$, $\mu_{1}$, $\beta_{2}$, $\mu_{2}$使得双边不等式

$ \alpha_{1} E(a, b)+(1-\alpha_{1})A(a, b)<M(a, b)<\beta_{1} E(a, b)+(1-\beta_{1})A(a, b), \\ E^{\lambda_{1}}(a, b)A^{1-\lambda_{1}}(a, b)<M(a, b)<E^{\mu_{1}}(a, b)A^{1-\mu_{1}}(a, b), \\ \alpha_{2} Q(a, b)+(1-\alpha_{2})\overline{H}(a, b)<M(a, b)<\beta_{2} Q(a, b)+(1-\beta_{2})\overline{H}(a, b), \\ \lambda_{2} C(a, b)+(1-\lambda_{2})\overline{H}(a, b)<M(a, b)<\mu_{2} C(a, b)+(1-\mu_{2})\overline{H}(a, b) $ (6)

对所有$a$, $b>0$$a\neq b$成立.

在文献[11]中, 作者证明了双边不等式

$ \lambda_{1} D(a, b)+(1-\lambda_{1})A(a, b) <T^*(a, b)<\mu_{1} D(a, b)+(1-\mu_{1})A(a, b), \\ D\big[\lambda_{2}a+(1-\lambda_{2})b, \lambda_{2}b+(1-\lambda_{2})a\big] <T^*(a, b) <D\big[\mu_{2}a+(1-\mu_{2})b, \mu_{2}b+(1-\mu_{2})a\big] $ (7)

对所有$a, b>0$$a\neq b$成立的充要条件是$\lambda_{1} \leqslant 1/8$, $\mu_{1} \geqslant 4/\pi -1=0.273~2\cdots$, $\lambda_{2} \leqslant 1/2+\sqrt{2}/8=0.676~7\cdots$$\mu_{2} \geqslant 1/2+\sqrt{(4-\pi)/(3 \pi -4)}/2=0.698~8\cdots$.其中$T^*(a, b)=\frac{2}{\pi} \int_0^{\pi/2}\sqrt{a^2 \cos^2\theta+b^2 \sin^2 \theta}{\rm{d}}\theta$是两个正数$a$$b$的Toader平均.

在文献[12-13]中作者发现了最佳参数$\alpha_{1}, \alpha_{2}, \beta_{1}, \beta_{2} \in(0, 1)$使得双边不等式

$ \alpha_{1} D(a, b)+(1-\alpha_{1})A(a, b)<T(a, b)<\beta_{1} D(a, b)+(1-\beta_{1})A(a, b), \\ \alpha_{2} D(a, b)+(1-\alpha_{2})H(a, b)<T(a, b)<\beta_{2} D(a, b)+(1-\beta_{2})H(a, b) $ (8)

对所有$a, b>0$$a\neq b$成立.

杨月英等[14]找到了最大值$\alpha_{1}, \alpha_{2}$和最小值$\beta_{1}, \beta_{2}$使得双边不等式

$ \alpha_{1} D(a, b)+(1-\alpha_{1})G(a, b)<P(a, b)<\beta_{1} D(a, b)+(1-\beta_{1})G(a, b), \\ \alpha_{2} D(a, b)+(1-\alpha_{2})G(a, b)<T(a, b)<\beta_{2} D(a, b)+(1-\beta_{2})G(a, b) $ (9)

对所有$a, b>0$$a\neq b$成立.

本文的主要结果是找到了最大值$\lambda_{1}, \lambda_{2}$和最小值$\mu_{1}, \mu_{2}$使得双边不等式

$ \lambda_{1} D(a, b)+(1-\lambda_{1})\overline{H}(a, b)<M(a, b)<\mu_{1} D(a, b)+(1-\mu_{1})\overline{H}(a, b), $ (10)
$ \lambda_{2} D(a, b)+(1-\lambda_{2})H(a, b)<M(a, b)<\mu_{2} D(a, b)+(1-\mu_{2})H(a, b) $ (11)

对所有$a, b>0$$a\neq b$成立.

1 引理

为了证明主要结论, 我们需要以下引理.

引理1.1  设$p\in(0, 1)$,

$ f(x)= (1-p)^2 x^5 + (7p^2-11p+1)x^4+6(2p^2-1)x^3 \\ -2(2p^2-11p+3)x^2+12p(1-p)x-4p^2, $ (12)

则下列结论成立.

(1) 若$p=10/21$, 则当$x\in(1, \sqrt{2})$时有$f(x) < 0$;

(2) 若$p=1/\big[2\ln(1+\sqrt{2})\big]$, 则存在$x_1 \in(1, \sqrt{2})$使得当$x\in(1, x_1)$时有$f(x)>0$, 而当$x\in(x_1, \sqrt{2})$时有$f(x) < 0$.

证明 (1)当$p=10/21, x\in(1, \sqrt{2})$时, 等式(12)变成

$ f(x)= \frac{1}{441}(x-1)(121x^4-1~048x^3-2~494x^2-920x+400) \\ < \frac{1}{441}(x-1)(121x^4-1~048-2~494-920+400) \\ = -\frac{1}{441}(x-1)(4~062-121x^4)<0, $ (13)

所以, 我们从不等式(13)容易证得(1).

(2) 当$p=1/\big[2\ln(1+\sqrt{2})\big]$时, 则由数值计算知

$ 7p^2-11p+1 =-2.987~4\cdots<0, $ (14)
$ 2p^2-1 =-0.356~3\cdots<0, $ (15)
$ 2p^2-11p+3 =-2.596~6\cdots<0, $ (16)
$ 27p^2-4p-8 =-1.579~9\cdots<0, $ (17)
$ f(1)=21p-10 =1.913~2\cdots>0, $ (18)
$ f\big(\sqrt{2}\big) =4(1+\sqrt{2})\big[4p^2+\sqrt{2}(\sqrt{2}-1)p-2\big]=-3.673~3\cdots<0 $ (19)

$ f'(x)= 5(1-p)^2 x^4 + 4(7p^2-11p+1)x^3+18(2p^2-1)x^2 \\ -4(2p^2-11p+3)x+12p(1-p). $ (20)

从不等式(14)-(17)和等式(20), 可知

$ f'(x) < 5(1-p)^2 x^4 + 4(7p^2-11p+1)x^2+18(2p^2-1)x^2\\ -4(2p^2-11p+3)x^2+12p(1-p)x^2\\ = x^2\big[5(1-p)^2 x^2+(44p^2+12p-26)\big]\\ < x^2\big[5(1-p)^2 (\sqrt{2})^2 +(44p^2+12p-26)\big]\\ = 2(27p^2-4p-8)x^2 <0 $ (21)

对所有$x\in(1, \sqrt{2})$成立.故从不等式(21)可知函数$f(x)$在区间$(1, \sqrt{2})$上严格单调递减.所以, 由不等式(18)和(19)以及函数$f(x)$的单调性知, 存在$x_1\in(1, \sqrt{2})$, 使得当$x\in(1, x_1)$时, 有$f(x)>0$; 当$x\in(x_1, \sqrt{2})$时, 有$f(x) < 0$.

引理1.2  设$p\in(0, 1)$

$ g(x)= (1-p)^2 x^7 - p(1-p)x^6-(3p+4)(1-p)x^5+(3p^2-2p-4)x^4 \\ +6px^3+12px^2+4p(1-p)x-4p^2, $ (22)

则下列结论成立.

(1) 若$p=7/18$, 则当$x\in(1, \sqrt{2})$时有$g(x) < 0$;

(2) 若$p=1/\big[2\ln(1+\sqrt{2})\big]$, 则存在$x_2 \in(1, \sqrt{2})$, 使得当$x\in(1, x_2)$时有$g(x)>0$, 而当$x\in(x_2, \sqrt{2})$时有$g(x) < 0$.

证明  (1)当$p=7/18, x\in(1, \sqrt{2})$时, 等式(22)变成

$ g(x)= \frac{1}{324}(x-1)(121x^6+44x^5-979x^4-2~380x^3-1624x^2-112x+196) \\ <\frac{1}{324}(x-1)(121x^6+44x^6-979x-2~380x-1~624x-112x+196x^6) \\ = -\frac{1}{324}x(x-1)(5~095-361x^5)<0, $ (23)

所以, 由不等式(23)容易证得(1).

(2) 当$p=1/\big[2\ln(1+\sqrt{2})\big]$时, 则由数值计算知

$ 3p^2-2p-4 =-4.169~1\cdots<0, $ (24)
$ 4p^2+\sqrt{2}p-2\sqrt{2} =-0.738~8\cdots<0, $ (25)
$ 57p^2-19p-8 =-0.434~5\cdots <0, $ (26)
$ g(1) =18p-7=3.211~3\cdots>0, $ (27)
$ g(\sqrt{2}) =4(1+\sqrt{2})(4p^2+\sqrt{2}p-2\sqrt{2})=-7.134~9\cdots<0 $ (28)

$ g'(x)= 7(1-p)^2 x^6 -6 p(1-p)x^5-5(3p+4)(1-p)x^4 \\ +4(3p^2-2p-4)x^3+18px^2+24px+4p(1-p). $ (29)

从不等式(24)、(26)和等式(29), 可知

$ g'(x)< 7(1-p)^2 x^6 -6 p(1-p)x^2-5(3p+4)(1-p)x^2 \\ +4(3p^2-2p-4)x^2+18px^2+24px^2+4p(1-p)x^2\\ = x^2\big[7(p^2-2p+1)x^4+(29p^2+37p-36)\big]\\ <x^2\big[7(p^2-2p+1)(\sqrt{2})^4+(29p^2+37p-36)\big]\\ = (57p^2-19p-8)x^2<0 $ (30)

对所有$x\in(1, \sqrt{2})$成立.故从不等式(30)可知函数$g(x)$在区间$(1, \sqrt{2})$上严格单调递减.所以, 由不等式(27)和(28)以及函数$g(x)$的单调性知, 存在$x_2\in(1, \sqrt{2})$, 使得当$x\in(1, x_2)$时有$g(x)>0$; 当$x\in(x_2, \sqrt{2})$时, 有$g(x) < 0$.

2 主要结果

定理2.1  双边不等式

$ \lambda_{1} D(a, b)+(1-\lambda_{1})\overline{H}(a, b)<M(a, b)<\mu_{1} D(a, b)+(1-\mu_{1})\overline{H}(a, b) $ (31)

对所有$a, b>0$$a\neq b$成立的充要条件是$\lambda_1\leqslant 10/21$$\mu_1\geqslant1/\big[2\ln(1+\sqrt{2})\big]=0.5672\cdots$.

证明  首先, 我们证明不等式

$ M(a, b)>\frac{10}{21}D(a, b)+\frac{11}{21}\overline{H}(a, b), $ (32)
$ M(a, b)<\frac{1}{2\ln(1+\sqrt{2})}D(a, b)+\Big[1-\frac{1}{2\ln(1+\sqrt{2})}\Big]\overline{H}(a, b) $ (33)

对所有$a, b>0$$a\neq b$成立.

根据$M(a, b)$, $D(a, b)$$\overline{H}(a, b)$是对称且一阶齐次的, 不失一般性, 我们假设$a>b>0$.设$v=(a-b)/(a+b)$, $x=\sqrt{1+v^2}$, $p\in(0, 1)$.则有$v\in(0, 1)$, $x\in(1, \sqrt{2})$

$ \frac{M(a, b)}{A(a, b)}=\frac{v}{{\rm arcsinh}(v)}, \frac{D(a, b)}{A(a, b)}=\frac{1+3v^2}{1+v^2}, \frac{\overline{H}(a, b)}{A(a, b)}=\frac{1-v^2}{\sqrt{1+v^2}}. $ (34)

我们从等式(34)得到

$ \frac{M(a, b)-\overline{H}(a, b)}{D(a, b)-\overline{H}(a, b)}=\frac{\big[v\sqrt{1+v^2}-(1-v^2){\rm arcsinh}(v)\big]\sqrt{1+v^2}} {\big[(1+3v^2)-(1-v^2)\sqrt{1+v^2}\big]{\rm arcsinh}(v)} $ (35)

$ \lim\limits_{v\rightarrow 0^{+}}\frac{\big[v\sqrt{1+v^2}-(1-v^2){\rm arcsinh}(v)\big]\sqrt{1+v^2}}{\big[(1+3v^2)-(1-v^2)\sqrt{1+v^2}\big]{\rm arcsinh}(v)}=\frac{10}{21}, $ (36)
$ \lim\limits_{v\rightarrow 1^{-}}\frac{\big[v\sqrt{1+v^2}-(1-v^2){\rm arcsinh}(v)\big]\sqrt{1+v^2}} {\big[(1+3v^2)-(1-v^2)\sqrt{1+v^2}\big]{\rm arcsinh}(v)}=\frac{1}{2\ln(1+\sqrt{2})}, $ (37)
$ M(a, b)-\big[pD(a, b)+(1-p)\overline{H}(a, b)\big] \\ =A(a, b)\bigg\{\frac{v}{{\rm arcsinh}(v)}-\Big[p\Big(\frac{1+3v^2} {1+v^2}\Big)+(1-p)\Big(\frac{1-v^2}{\sqrt{1+v^2}}\Big)\Big]\bigg\} \\ =\frac{A(a, b)\big[p(1+3v^2)+(1-p)(1-v^2)\sqrt{1+v^2} \big]}{(1+v^2){\rm arcsinh}(v)}F(x), $ (35)

其中

$ F(x)=\frac{x^2\sqrt{x^2-1}}{(p-1)x^3+3px^2+2(1-p)x-2p}-{\rm arcsinh}\big(\sqrt{x^2-1}\big). $ (39)

经简单计算得

$ F(1) =0, $ (40)
$ F(\sqrt{2}) =\frac{1}{2p}-\ln(1+\sqrt{2}), $ (41)
$ F'(x) =-\frac{(x-1)}{\sqrt{x^2-1}\big[(p-1)x^3+3px^2+2(1-p)x-2p\big]^2}f(x), $ (42)

其中函数$f(x)$由式(12)给出.

现在, 我们分两种情形证明.

情形1  当$p=10/21$时, 则从等式(42)和引理1.1(1)可看到函数$F(x)$在区间$(1, \sqrt{2})$上严格单调递增.所以, 不等式(32)容易从等式(38)-(40)及$F(x)$的单调性得到.

情形2  当$p=1/\big[2\ln(1+\sqrt{2})\big]$时, 则从等式(42)和引理1.1(2)得:存在$x_1 \in(1, \sqrt{2})$, 使得当$x\in(1, x_1)$时有$F'(x) < 0$; 当$x\in(x_1, \sqrt{2})$时有$F'(x)>0$.这样, 函数$F(x)$在区间$(1, x_1)$上严格递减, 在区间$(x_1, \sqrt{2})$上严格递增.注意到此时等式(41)变成

$ F(\sqrt{2})=0, $ (43)

所以, 不等式(33)可从等式(38)-(40)和(43)及函数$F(x)$的分段单调性得到.因此, 定理2.1容易由不等式(32)和(33)以及下面的事实得到.

•若$p>10/21$, 则由等式(35)和(36)得, 存在一个$0 < \delta_1 < 1$, 使得

$ M(a, b) <pD(a, b)+(1-p)\overline{H}(a, b) $

对所有$a$, $b>0$$(a-b)/(a+b)\in(0, \delta_1)$成立.

•若$p < 1/\big[2\ln(1+\sqrt{2})\big]$, 则由等式(35)和(37)得, 存在一个$0 < \delta_2 < 1$, 使得

$ M(a, b)>pD(a, b)+(1-p)\overline{H}(a, b) $

对所有$a, b>0$$(a-b)/(a+b)\in(1-\delta_2, 1)$成立.

定理2.2  双边不等式

$ \lambda_{2} D(a, b)+(1-\lambda_{2})H(a, b) <M(a, b) <\mu_{2} D(a, b)+(1-\mu_{2})H(a, b) $ (44)

对所有$a, b>0$$a\neq b$成立的充要条件是$\lambda_2\leqslant 7/18$$\mu_2\geqslant1/\big[2\ln(1+\sqrt{2})\big]=0.5672\cdots$.

证明  首先, 我们证明不等式

$ M(a, b)>\frac{7}{18}D(a, b)+\frac{11}{18}H(a, b), $ (45)
$ M(a, b) <\frac{1}{2\ln(1+\sqrt{2})}D(a, b)+ \Big[1-\frac{1}{2\ln(1+\sqrt{2})}\Big]H(a, b) $ (46)

对所有$a, b>0$$a\neq b$成立.

根据$M(a, b)$, $D(a, b)$$H(a, b)$是对称且一阶齐次的, 不失一般性, 我们假设$a>b>0$.设$v=(a-b)/(a+b)$, $x=\sqrt{1+v^2}$, $p\in(0, 1)$.则有$v\in(0, 1)$, $x\in(1, \sqrt{2})$

$ \frac{M(a, b)}{A(a, b)}=\frac{v}{{\rm arcsinh}(v)}, \frac{D(a, b)}{A(a, b)}=\frac{1+3v^2}{1+v^2}, \frac{H(a, b)}{A(a, b)}=1-v^2. $ (47)

从等式(47)得到

$ \frac{M(a, b)-H(a, b)}{D(a, b)-H(a, b)}=\frac{(1+v^2)\big[v-(1-v^2){\rm arcsinh}(v)\big]}{v^2(v^2+3){\rm arcsinh}(v)} $ (48)

$ \lim\limits_{v\rightarrow 0^{+}}\frac{(1+v^2)\big[v-(1-v^2){\rm arcsinh}(v)\big]}{v^2(v^2+3){\rm arcsinh}(v)}=\frac{7}{18}, $ (49)
$ \lim\limits_{v\rightarrow 1^{-}}\frac{(1+v^2)\big[v-(1-v^2){\rm arcsinh}(v)\big]}{v^2(v^2+3){\rm arcsinh}(v)}=\frac{1}{2\ln(1+\sqrt{2})}, $ (50)
$ M(a, b)-\big[pD(a, b)+(1-p)H(a, b)\big] \\ =A(a, b)\bigg\{\frac{v}{{\rm arcsinh}(v)}-\Big[p\Big(\frac{1+3v^2}{1+v^2}\Big)+(1-p)(1-v^2)\Big]\bigg\} \\ =\frac{A(a, b)\big[p(1+3v^2)+(1-p)(1-v^4)\big]}{(1+v^2){\rm arcsinh}(v)}G(x), $ (51)

其中

$ G(x)=\frac{x^2\sqrt{x^2-1}}{(p-1)x^4+(p+2)x^2-2p}-{\rm arcsinh}\big(\sqrt{x^2-1}\big). $ (52)

经简单计算得

$ G(1) =0, $ (53)
$ G(\sqrt{2}) =\frac{1}{2p}-\ln(1+\sqrt{2}), $ (54)
$ G'(x) =-\frac{(x-1)}{\sqrt{x^2-1}\big[(p-1)x^4+(p+2)x^2-2p\big]^2}g(x) $ (55)

其中函数$g(x)$由式(22)定义.

现在, 我们分两种情形证明.

情形1  当$p=7/18$时, 则从等式(55)和引理1.2(1)知函数$G(x)$在区间$(1, \sqrt{2})$上是严格单调递增的.所以, 不等式(45)容易从等式(51)-(53)及$G(x)$的单调性得到.

情形2  当$p=1/\big[2\ln(1+\sqrt{2})\big]$时, 则从等式(55)和引理1.2(2)知, 存在$x_2 \in(1, \sqrt{2})$使得当$x\in(1, x_2)$时有$G'(x) < 0$; 当$x\in(x_2, \sqrt{2})$时有$G'(x)>0$.这样, 函数$G(x)$在区间$(1, x_2)$上严格递减, 在区间$(x_2, \sqrt{2})$上严格递增.注意到此时等式(54)变成

$ G(\sqrt{2})=0, $ (56)

所以, 不等式(46)可从等式(51)-(53)和(56)以及函数$G(x)$的分段单调性得到.因此, 定理2.2容易从不等式(45)和(46)协同下面事实得到.

•若$p>7/18$, 则由等式(48)和(49)知, 存在$0 < \delta_3 < 1$, 使得

$ M(a, b) <pD(a, b)+(1-p)H(a, b) $

对所有$a, b>0$$(a-b)/(a+b)\in(0, \delta_3)$成立.

•若$p < 1/\big[2\ln(1+\sqrt{2})\big]$, 则由等式(48)和(50)知, 存在$0 < \delta_4 < 1$, 使得

$ M(a, b)>pD(a, b)+(1-p)H(a, b) $

对所有$a, b>0$$(a-b)/(a+b)\in(1-\delta_4, 1)$成立.

参考文献
[1] NEUMAN E, SÁNDOR J. On the Schwab-Borchardt mean[J]. Mathematica Pannonica, 2003, 14(2): 253-266.
[2] LI Y M, LONG B Y, CHU Y M. Sharp bounds for the Neuman-Sándor mean in terms of generalized logarithmic mean[J]. Journal of Mathematical Inequalities, 2012, 6(4): 567-577.
[3] CHU Y M, LONG B Y, GONG W M, et al. Sharp bounds for Seiffert and Neuman-Sándor means in terms of generalized logarithmic means[J]. Journal of Inequalities and Applications, 2013, 2013: 10 DOI:10.1186/1029-242X-2013-10
[4] CHU Y M, LONG B Y. Bounds of the Neuman-Sándor mean using power and identric means[J]. Abstract and Applied Analysis, 2013, 6pages.
[5] ZHAO T H, CHU Y M, JIANG Y L, et al. Best possible bounds for Neuman-Sándor mean by the identric, quadratic and contraharmonic means[J]. Abstract and Applied Analysis, 2013, 12pages.
[6] HE Z Y, QIAN W M, JIANG Y L, et al. Bounds for the combinations of Neuman-Sándor, arithmetic and second Seiffert means in terms of contra-harmonic mean[J]. Abstract and Applied Analysis, 2013, 5pages.
[7] NEUMAN E, SÁNDOR J. On the Schwab-Borchardt mean Ⅱ[J]. Mathematica Pannonica, 2006, 17(1): 49-59.
[8] NEUMAN E. A note on a certain bivariate mean[J]. Journal of Mathematical Inequalities, 2012, 6(4): 637-643.
[9] ZHAO T H, CHU Y M, LIU B Y. Optimal bounds for Neuman-Sándor mean in terms of the convex combinations of harmonic, geometric, quadratic, and contraharmonic means[J]. Abstract and Applied Analysis, 2012, 9pages.
[10] QIAN W M, CHU Y M. On certain inequalities for Neuman-Sándor mean[J]. Abstract and Applied Analysis, 2013, 6pages.
[11] QIAN W M, SONG Y Q, ZHANG X H, et al. Sharp bounds for Toader mean in terms of arithmetic and second contra-harmonic means[J]. Journal of Function Spaces, 2015, 5pages.
[12] 孟祥菊, 刘红, 高红亚. 第二类, Seiffert, 平均的最优凸组合界[J]. 宁夏大学学报(自然科学版), 2012, 33(1): 14-16.
[13] 孟祥菊, 王淑燕, 田淑环. 关于第二类, Seiffert, 平均的最佳双边不等式[J]. 数学的实践与认识, 2015, 45(18): 299-302.
[14] YANG Y Y, SHEN L C, QIAN W M. The optimal convex combination bounds of second contra-harmonic and geometric mean for the Seiffert means[J]. Pacific Journal of Applied Mathematics, 2016, 7(3): 207-217.