数学杂志  2024, Vol. 44 Issue (2): 165-168   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
胡兴凯
弋苑
刘武双
两个酉不变范数不等式的推广
胡兴凯, 弋苑, 刘武双    
昆明理工大学理学院, 云南 昆明 650500
摘要:本文研究了矩阵酉不变范数不等式的问题. 在v∈[0, 1]时, 利用函数φ(v)=||AvXB1-v+A1-vXBv||的凸性, 推广了两个酉不变范数不等式.
关键词半正定矩阵    凸函数    酉不变范数    
GENERALIZATIONS OF TWO UNITARILY INVARIANT NORM INEQUALITIES
HU Xing-kai, YI Yuan, LIU Wu-shuang    
Faculty of Science, Kunming University of Science and Technology, Kunming 650500, China
Abstract: In this paper, unitarily invariant norm inequalities for matrices are studied. Using the convexity of φ(v)=||AvXB1-v+A1-vXBv|| on the interval [0, 1], two unitarily invariant norm inequalities for matrices are generalized.
Keywords: positive semidefinite matrices     convex function     unitarily invariant norms    
1 引言

$ M_n $是复数域上$ n\times n $阶矩阵的集合. 设$ \|\cdot\| $$ M_n $上的范数, 若对所有$ n $阶矩阵$ A $和酉矩阵$ U, V\in M_n $, 都有$ \|UAV\|=\|A\| $成立, 则称$ \|\cdot\| $为酉不变范数. 矩阵$ A\in M_n $的奇异值定义为$ A^ *A $的特征值的非负平方根, 用$ s_1(A)\geq s_2(A)\geq \cdots \geq s_n(A) $表示$ A\in M_n $的奇异值. 其中, 有两类重要的酉不变范数, 第一类是Ky Fan $ k $-范数$ \|A\|_{(k)} $, 即

$ \|A\|_{(k)} = \sum\limits_{j=1}^ks_j(A), k=1, \cdots, n, $

第二类是Schatten $ p $-范数$ \|A\|_{p} $, 即

$ \|A\|_{p}= \left(\sum\limits_{j=1}^ns_j^p(A)\right)^{\frac{1}{p}}= ({\rm{tr}}|A|^p)^{\frac{1}{p}}, 1\leq p < \infty. $

$ A, B, X\in M_n $$ A, B $是半正定矩阵, Bhatia和Davis在文献[1]证明了:若$ 0 \leq v \leq 1 $, 则

$ \begin{align} ||A^\frac{1}{2}XB^\frac{1}{2}||\leq ||\frac{A^vXB^{1-v}+ A^{1-v}XB^v}{2}||\leq ||\frac{AX + XB}{2}||, \end{align} $ (1.1)

$ \varphi(v)=||A^vXB^{1-v}+ A^{1-v}XB^v|| $, 则不等式(1.1)可简写为:

$ \varphi(\frac{1}{2})\leq \varphi(v)\leq \varphi(0). $

$ A, B, X\in M_n $$ A, B $是半正定矩阵, 函数$ \varphi(v) $在区间$ [0, 1] $上是凸函数, 当$ v=\frac{1}{2} $时取最小值(见文献[2]).

2012年, Zou和He在文献[3]中改进了不等式(1.1)得到

$ \begin{align} \varphi\left(\frac{1}{2}\right)+2\left(\int_{0}^{1}\varphi(v)dv-\varphi\left(\frac{1}{2}\right)\right)\leq \varphi(0), 0\leq v\leq 1. \end{align} $ (1.2)

$ A, B\in M_n $为半正定矩阵, Bhatia和Kittaneh在文献[4]中证明了:

$ \begin{align} ||AB||\leq \frac{1}{4}||(A+B)^{2}||. \end{align} $ (1.3)

Zou和He在文献[3]中改进了不等式(1.3)得到

$ \begin{align} ||AB||+\left(\int_{0}^{1}\psi(v)dv-2||AB||\right)\leq \frac{1}{4}||(A+B)^{2}||, \end{align} $ (1.4)

其中

$ \psi(v)=||A^{\frac{1}{2}+v}B^{\frac{3}{2}-v}+ A^{\frac{3}{2}-v}B^{\frac{1}{2}+v}||. $

近年来, 有很多文献[5-9]对酉不变范数不等式进行了研究. 本文将继续研究矩阵的酉不变范数不等式, 利用函数的凸性, 对不等式(1.2)和(1.4)进行推广.

2 主要结果

下面我们首先给出两个引理, 它们将在本文定理证明过程中起到重要作用.

引理2.1 [10]   设$ A, B, X\in M_n $$ A, B $是半正定矩阵, 则

$ \varphi(v)\leq (1-r_0)\varphi(0)+r_0\varphi(\mu), $

其中$ 0\leq v\leq 1 $, $ 0< \mu< 1 $, $r_0 = \left\{ {\begin{array}{*{20}c} {\frac{v}{\mu }, 0 \le v \le \mu , } \\ {\frac{{1 - v}}{{1 - \mu }}, \mu < v \le 1.} \\ \end{array}} \right.\\$

引理2.2 [4]   设$ A, B\in M_n $$ A, B $是半正定矩阵, 则

$ ||A^{\frac{1}{2}}(A+B)B^{\frac{1}{2}}||\leq \frac{1}{2}||(A+B)^2||. $

定理2.1    设$ A, B, X \in M_n $$ A, B $是半正定矩阵, 则

$ \begin{align} \varphi\left(\mu\right)+2\left(\int_{0}^{1}\varphi(v)dv-\varphi(\mu)\right)\leq \varphi(0), 0\leq v\leq 1, 0<\mu<1 , \end{align} $ (2.1)

其中$ \varphi(v)=||A^{v}XB^{1-v}+A^{1-v}XB^{v}||. $

   当$ 0\leq v\leq \mu $时, 由引理2.1可得

$ \varphi(v)\leq (1-\frac{v}{\mu})\varphi(0)+\frac{v}{\mu}\varphi(\mu). $

于是

$ \int_{0}^{\mu}\varphi(v)dv\leq \varphi(0)\int_{0}^{\mu}(1-\frac{v}{\mu})dv+\varphi(\mu)\int_{0}^{\mu}\frac{v}{\mu}dv. $

则有

$ \begin{align} \int_{0}^{\mu}\varphi(v)dv\leq \frac{\mu}{2}\left(\varphi(0)+\varphi(\mu)\right). \end{align} $ (2.2)

$ \mu < v\leq 1 $时, 由引理2.1可得

$ \varphi(v)\leq (1-\frac{1-v}{1-\mu})\varphi(0)+\frac{1-v}{1-\mu}\varphi(\mu). $

于是

$ \int_{\mu}^{1}\varphi(v)dv\leq \varphi(0)\int_{\mu}^{1}(1-\frac{1-v}{1-\mu})dv+\varphi(\mu)\int_{\mu}^{1}\frac{1-v}{1-\mu}dv, $

则有

$ \begin{align} \int_{\mu}^{1}\varphi(v)dv\leq \frac{1-\mu}{2}\left(\varphi(0)+\varphi(\mu)\right). \end{align} $ (2.3)

由不等式(2.2)和(2.3), 有

$ \begin{align} \int_{0}^{1}\varphi(v)dv=\int_{0}^{\mu}\varphi(v)dv+\int_{\mu}^{1}\varphi(v)dv\leq \frac{1}{2}(\varphi(0)+\varphi(\mu)). \end{align} $ (2.4)

不等式(2.4)可写成

$ 2\int_{0}^{1}\varphi(v)dv\leq \varphi(0)+\varphi(\mu), $

$ \varphi\left(\mu\right)+2\left(\int_{0}^{1}\varphi(v)dv-\varphi(\mu)\right)\leq \varphi(0). $

注2.1    在定理2.1中, 当$ \mu=\frac{1}{2} $时, 可得不等式(1.2).

定理2.2    设$ A, B\in M_n $$ A, B $是半正定矩阵, 则

$ \begin{align} \frac{1}{2}||A^{\frac{1}{2}+\mu}B^{\frac{3}{2}-\mu}+A^{\frac{3}{2}-\mu}B^{\frac{1}{2}+\mu}||+&\left(\int_{0}^{1}\psi(v)dv-||A^{\frac{1}{2}+\mu}B^{\frac{3}{2}-\mu}+ A^{\frac{3}{2}-\mu}B^{\frac{1}{2}+\mu}||\right) \\ &\leq \frac{1}{4}||(A+B)^2||, \end{align} $ (2.5)

其中$ 0\leq v\leq 1, 0<\mu<1 , \psi(v)=||A^{\frac{1}{2}+v}B^{\frac{3}{2}-v}+ A^{\frac{3}{2}-v}B^{\frac{1}{2}+v}|| $.

   在不等式$ (2.1) $, 令$ X=A^{\frac{1}{2}}B^{\frac{1}{2}} $, 可得

$ \begin{align} ||A^{\frac{1}{2}+\mu}B^{\frac{3}{2}-\mu}+ A^{\frac{3}{2}-\mu}B^{\frac{1}{2}+\mu}||+&2\left(\int_{0}^{1}\psi(v)dv-||A^{\frac{1}{2}+\mu}B^{\frac{3}{2}-\mu}+ A^{\frac{3}{2}-\mu}B^{\frac{1}{2}+\mu}||\right) \\ &\leq ||A^{\frac{1}{2}}(A+B)B^{\frac{1}{2}}||. \end{align} $ (2.6)

由引理2.2, 不等式(2.6)可写成

$ ||A^{\frac{1}{2}+\mu}B^{\frac{3}{2}-\mu}+ A^{\frac{3}{2}-\mu}B^{\frac{1}{2}+\mu}||+2\left(\int_{0}^{1}\psi(v)dv-||A^{\frac{1}{2}+\mu}B^{\frac{3}{2}-\mu}+ A^{\frac{3}{2}-\mu}B^{\frac{1}{2}+\mu}||\right)\leq \frac{1}{2}||(A+B)^2||. $

注2.2    在定理2.2中, 当$ \mu=\frac{1}{2} $时, 可得不等式(1.4).

参考文献
[1] Bhatia R, Davis C. More matrix forms of the arithmetic-geometric mean inequality[J]. SIAM J. Matrix Anal. Appl., 1993, 14(1): 132–136. DOI:10.1137/0614012
[2] Bhatia R. Matrix analysis[M]. New York: Springer-Verlag, 1997.
[3] Zou Limim, He Chuanjiang. On some inequalities for unitarily invariant norms and singular values[J]. Linear Algebra Appl., 2012, 436: 3354–3361. DOI:10.1016/j.laa.2011.11.030
[4] Bhatia R, Kittaneh F. Notes on matrix arithmetic-geometric mean inequalities[J]. Linear Algebra Appl., 2000, 308: 203–211. DOI:10.1016/S0024-3795(00)00048-3
[5] Xue Jianming, Hu Xingkai. A note on some inequalities for unitarily invariant norms[J]. J. Math. Inequal., 2015, 9(3): 841–846.
[6] 邹黎敏, 吴艳秋. 矩阵酉不变范数Hölder不等式及其应用[J]. 数学物理学报, 2017, 37A(3): 450–456. DOI:10.3969/j.issn.1003-3998.2017.03.004
[7] Al-Natoor A, Benzamia S, Kittaneh F. Unitarily invariant norm inequalities for positive semidefinite matrices[J]. Linear Algebra Appl., 2022, 633: 303–315. DOI:10.1016/j.laa.2021.10.012
[8] Cao Haisong, Wu Junliang. Unitarily invariant norm inequalities involving Heron and Heinz means[J]. J. Inequal. Appl., 2014, 2014: 288. DOI:10.1186/1029-242X-2014-288
[9] Cao Haisong. A new Hermite-Hadamard type inequality for coordinate convex function[J]. J. Inequal. Appl., 2020, 2020(1): 162. DOI:10.1186/s13660-020-02428-3
[10] Liu Wushang, Hu Xingkai, Shi Jianping. On some inequalities related to Heinz means for unitarily invariant norms[J]. J. Math. Inequal., 2022, 16(3): 1123–1128.