二、三阶方阵的等迹分解
吴马威1,2, 陈益智2, 骆莉芳2     
1. 广西大学 数学与信息科学学院, 广西 南宁 530004;
2. 惠州学院 数学系, 广东 惠州 516007
摘要:利用构造法探究二、三阶方阵的等迹分解.首先证明任意二阶方阵都存在等迹分解;然后证明对角线元素均不为零且对角线元素互不相等的三阶方阵可以分解为迹为1的两个方阵相乘,如果该方阵还满足迹不为零的条件,则该方阵存在等迹分解,并证明了一类特殊三阶块对角阵也存在等迹分解。最后通过数值算例给出了主要结果的一个应用。
关键词二阶方阵     三阶方阵     等迹分解    
The equal trace decompositions of second order and third order square matrices
WU Mawei1,2, CHEN Yizhi2, LUO Lifang2     
1. Department of Mathematics and Information Sciences, Guangxi University, Nanning 530004, China;
2. Department of Mathematics, Huizhou University, Huizhou 516007, Guangdong, China
Abstract: The equal trace decompositions of second and third order square matrices are mainly studied by using construction methods. Firstly, it is proved that the equal trace decompositions exists if the matrix is second order square matrix; and then some third order square matrices satisfying that any diagonal entry is different from each other and nonzero, can be decomposed into two matrices with trace-1; Moreover, it is showed that the equal trace decompositions exists if the traces of these matrices is nonzero, and the equal trace decompositions of some kind of third-order block diagonal matrices also exist. Finally, one application of the main results is given by an example.
Key words: second order square matrix     third order square matrix     equal trace decomposition    
1 引言及预备知识

矩阵是线性代数的基本内容之一, 在物理、工程技术、经济等领域中也有着广泛应用.同时矩阵的迹是矩阵的一个重要的数值特征, 在矩阵理论中有着广泛应用.目前, 关于矩阵分解和矩阵的迹的研究已经有许多结果, 如满秩分解[1-3]、LU分解和QR分解等[4-9].关于矩阵的迹的研究主要包括对其性质的探讨, 以及对一些特殊矩阵的迹的研究, 比如实对称矩阵的迹的性质、埃尔米特矩阵的迹的性质等[10-17].但这些研究没有将矩阵的分解和矩阵的迹相结合, 而且对矩阵的迹的研究也主要集中于对其不等式方面的研究.满秩分解可以将一个秩为r的矩阵分解为两个秩为r的矩阵的乘积, 本文则考虑将一个迹为t的方阵分解为两个迹均为t的方阵的乘积, 主要对二、三阶方阵进行了探究, 证明了任意二阶方阵都存在等迹分解; 对角线元素均不为零且对角线元素互不相等的三阶方阵可以分解为迹为1的两个方阵相乘, 进一步地, 如果该方阵还满足迹不为零的条件, 则该方阵存在等迹分解; 一类特殊三阶块对角阵也存在等迹分解.

为方便起见, 首先引入以下预备知识.

定义1  设A是一个n阶方阵, 如果方阵BC满足A=BC, 且trA=trB=trC.则称A=BC是方阵A的一个等迹分解.

引理1  若A, B均为n阶方阵, 且B=kA, (kR), 则trB=k(trA).

证明  设A=(aij)n×n, B=(bij)n×n, 则$\text{tr}\boldsymbol{B}\text{=}\sum\limits_{i=1}^{n}{{{b}_{ii}}}=\sum\limits_{i=1}^{n}{k{{a}_{ii}}}=k\sum\limits_{i=1}^{n}{{{a}_{ii}}}=k\left( \text{tr}\boldsymbol{A} \right)$.

2 二阶方阵的等迹分解

定理1  任意二阶方阵都存在等迹分解.

证明  可用构造法来证明.设A=(aij)2×2为任意的二阶方阵, 往证存在二阶方阵B=(bij)2×2C=(cij)2×2, 使得A=BC, 且trA=trB=trC.

下面分trA=0和trA≠0两种情形来讨论B, C的存在性.

情形一:当trA=0时.

① 设$\boldsymbol{A=}\left( \begin{matrix} 0&{{a}_{12}} \\ {{a}_{21}}&0 \\ \end{matrix} \right)$, 此时可取$\boldsymbol{B=}\left( \begin{matrix} 0 &-{{a}_{12}} \\ {{a}_{21}}&0 \\ \end{matrix} \right), \boldsymbol{C=}\left( \begin{matrix} 1&0 \\ 0 &-1 \\ \end{matrix} \right)$.

② 设$\boldsymbol{A=}\left( \begin{matrix} a&{{a}_{12}} \\ {{a}_{21}} &-a \\ \end{matrix} \right), \left( a\ne 0 \right)$, 此时可取$\boldsymbol{B=}\left( \begin{matrix} b&1 \\ 1 &-b \\ \end{matrix} \right), \boldsymbol{C=}\left( \begin{matrix} c&{{c}_{12}} \\ {{c}_{21}} &-c \\ \end{matrix} \right)$, 其中

$ b = \frac{{{a_{12}} + {a_{21}}}}{{ - 2a}},{c_{21}} = \frac{{b{a_{12}} - a}}{{1 + {b^2}}},c = b{c_{12}} - {a_{12}},{c_{21}} = a - bc. $

情形二:当trA≠0时.

① 设$\boldsymbol{A=}\left( \begin{matrix} {{a}_{11}}&{{a}_{12}} \\ {{a}_{21}}&{{a}_{22}} \\ \end{matrix} \right), \left( {{a}_{11}}+{{a}_{22}}\ne 0, {{a}_{12}}\ne 0 \right)$, 此时可取

$ \mathit{\boldsymbol{B}} = \left( {\begin{array}{*{20}{c}} {\frac{{{\rm{tr}}\mathit{\boldsymbol{A}}}}{2}}&0\\ {{b_{21}}}&{\frac{{{\rm{tr}}\mathit{\boldsymbol{A}}}}{2}} \end{array}} \right),\mathit{\boldsymbol{C = }}\left( {\begin{array}{*{20}{c}} {{c_{11}}}&{{c_{12}}}\\ {{c_{21}}}&{{c_{22}}} \end{array}} \right). $

其中${{c}_{11}}=\frac{2{{a}_{11}}}{\text{tr}\boldsymbol{A}}, {{c}_{22}}=\text{tr}\boldsymbol{A}-\frac{2{{a}_{11}}}{\text{tr}\boldsymbol{A}}, {{c}_{12}}=\frac{2{{a}_{12}}}{\text{tr}\boldsymbol{A}}$, ${{b}_{21}}=\frac{{{a}_{22}}-\frac{\text{tr}\boldsymbol{A}}{2}{{c}_{22}}}{{{c}_{12}}}, {{c}_{21}}=\frac{2\left( {{a}_{21}}-{{b}_{21}}{{c}_{11}} \right)}{\text{tr}\boldsymbol{A}}$.

② 设$\boldsymbol{A=}\left( \begin{matrix} {{a}_{11}}&0 \\ {{a}_{21}}&{{a}_{22}} \\ \end{matrix} \right), \left( {{a}_{11}}+{{a}_{22}}\ne 0 \right)$, 再分四种情况讨论.

(ⅰ) 设$\boldsymbol{A=}\left( \begin{matrix} 0&0 \\ 0&{{a}_{22}} \\ \end{matrix} \right), \left( {{a}_{22}}\ne 0 \right)$.此时可取$\boldsymbol{B=}\left( \begin{matrix} {{a}_{22}} &-1 \\ {{a}_{22}}&0 \\ \end{matrix} \right), \boldsymbol{C}=\left( \begin{matrix} 0&1 \\ 0&{{a}_{22}} \\ \end{matrix} \right)$.

(ⅱ) 设$\boldsymbol{A=}\left( \begin{matrix} 0&0 \\ {{a}_{21}}&{{a}_{22}} \\ \end{matrix} \right), \left( {{a}_{21}}{{a}_{22}}\ne 0 \right)$.此时可取$\boldsymbol{B=}\left( \begin{matrix} 0&1 \\ \frac{{{a}_{21}}}{{{a}_{22}}}&{{a}_{22}} \\ \end{matrix} \right), \boldsymbol{C=}\left( \begin{matrix} {{a}_{22}}&\frac{{{\left( {{a}_{22}} \right)}^{2}}}{{{a}_{21}}} \\ 0&0 \\ \end{matrix} \right)$.

(ⅲ) 设$A=\left( \begin{matrix} {{a}_{11}}&0 \\ {{a}_{21}}&{{a}_{22}} \\ \end{matrix} \right)$, $\left( {{a}_{11}}{{a}_{22}}\ne 0, {{a}_{11}}+{{a}_{22}}\ne 0, {{a}_{21}}\ne {{a}_{11}}+{{a}_{22}} \right)$.此时可取$\boldsymbol{B=}\left( \begin{matrix} 0&{{a}_{11}} \\ \frac{{{a}_{21}}}{\text{tr}\boldsymbol{A}}-1&\text{tr}\boldsymbol{A} \\ \end{matrix} \right)$, $\boldsymbol{C=}\left( \begin{matrix} \text{tr}\boldsymbol{A}&\frac{{{a}_{22}}}{\frac{{{a}_{21}}}{\text{tr}\boldsymbol{A}}-1} \\ 1&0 \\ \end{matrix} \right)$.

(ⅳ) 设$\boldsymbol{A=}\left( \begin{matrix} {{a}_{11}}&0 \\ {{a}_{11}}+{{a}_{22}}&{{a}_{22}} \\ \end{matrix} \right)$, (a11a22≠0, a11+a22≠0).此时可取$\boldsymbol{B=}\left( \begin{matrix} 0&2{{a}_{11}} \\ \frac{1}{2}&\text{tr}\boldsymbol{A} \\ \end{matrix} \right), \boldsymbol{C=}\left( \begin{matrix} \text{tr}\boldsymbol{A}&2{{a}_{22}} \\ \frac{1}{2}&0 \\ \end{matrix} \right)$.

3 三阶方阵的等迹分解

定理2  设A是一个对角线元素均不为0且对角线元素彼此不等的三阶方阵, 则存在三阶方阵B, C,使得A=BC, 且满足trB=trC=1.

证明  设A=(aij)3×3, a11, a22, a33均不为0且彼此不等, 往证存在三阶方阵B, C,使得A=BC, 且满足trB=trC=1.

显然存在aii≠1, 不妨设a11≠1.此时可取B=diag{b11, b22, b33}, $\boldsymbol{C=}{{\left( \frac{{{a}_{ij}}}{{{b}_{ii}}} \right)}_{3\times 3}}$.其中, b11=a11, b22=$\frac{\left( {{a}_{11}}-1 \right){{a}_{22}}}{{{a}_{33}}-{{a}_{22}}}, {{b}_{33}}=\frac{\left( 1-{{a}_{11}} \right){{a}_{33}}}{{{a}_{33}}-{{a}_{22}}}$.易验证方阵B, C即为所求.

定理3  设A是一个迹和对角线元素均不为0且对角线元素彼此不等的三阶方阵, 则存在三阶方阵B, C,使得A=BC, 且满足trA=trB=trC.

证明  记t=trA≠0,$\boldsymbol{D = }\frac{1}{{{t^2}}}\boldsymbol{A}$, A=(aij)3×3, D=(dij)3×3.因为${d_{ii}} = \frac{1}{{{t^2}}}{a_{ii}}$, aii≠0且彼此不等, 由引理1知dii≠0, 且彼此不等.

由定理2可知存在方阵E, F, 使得D=EF且trE=trF=1.令B=tE, C=tF.又由引理1可知trB=t(trE)=t, trC=t(trF)=t.又t2D=(tE)(tF), 即A=BC.且由上可知trA=trB=trC=t, 故A=BC是方阵A的一个等迹分解.

注1  特别地, 当A是对角线上的元素为互不相等的正(负)数的三阶方阵时, 也可以利用定理3中的方法构造出方阵A的一个等迹分解.

定理4  设三阶方阵${{\boldsymbol{A}}_{3}}=\left( \begin{matrix} {{\boldsymbol{A}}_{2}}&0 \\ 0&1 \\ \end{matrix} \right)$, 其中, A2=(aij)2×2.则存在三阶方阵${{\boldsymbol{B}}_{3}}=\left( \begin{matrix} {{\boldsymbol{B}}_{2}}&0 \\ 0&1 \\ \end{matrix} \right)$${{\boldsymbol{C}}_{3}}=\left( \begin{matrix} {{\boldsymbol{C}}_{2}}&0 \\ 0&1 \\ \end{matrix} \right)$, 使得A3=B3C3, A2=B2C2且trA3=trB3=trC3, trA2=trB2=trC2均成立.

证明  由定理1可知, 对于二阶方阵A2, 存在二阶方阵B2, C2, 使得A2=B2C2且满足trA2=trB2=trC2.取${{\boldsymbol{B}}_{3}}=\left( \begin{matrix} {{\boldsymbol{B}}_{2}}&0 \\ 0&1 \\ \end{matrix} \right), {{\boldsymbol{C}}_{3}}=\left( \begin{matrix} {{\boldsymbol{C}}_{2}}&0 \\ 0&1 \\ \end{matrix} \right)$, 此时易验证A3=B3C3且trA3=trB3=trC3成立.

注2  定理4可以推广到n维的情形, 即设n阶方阵${{\boldsymbol{A}}_{n}}={{\left( \begin{matrix} {{\boldsymbol{A}}_{2}}&0&\cdots &0 \\ 0&1&\cdots &0 \\ \vdots &\vdots &\ddots &\vdots \\ 0&0&\cdots &1 \\ \end{matrix} \right)}_{n\times n}}$, 其中, A2=(Aij)2×2.则存在n阶方阵${{\boldsymbol{B}}_{n}}=\left( \begin{matrix} {{\boldsymbol{B}}_{2}}&0&\cdots &0 \\ 0&1&\cdots &0 \\ \vdots &\vdots &\ddots &\vdots \\ 0&0&\cdots &1 \\ \end{matrix} \right)_{n \times n}$${{\boldsymbol{C}}_{n}}=\left( \begin{matrix} {{\boldsymbol{C}}_{2}}&0&\cdots &0 \\ 0&1&\cdots &0 \\ \vdots &\vdots &\ddots &\vdots \\ 0&0&\cdots &1 \\ \end{matrix} \right)_{n \times n}$, 使得An=BnCn, A2=B2C2且trAn=trBn=trCn, trA2=trB2=trC2均成立.

4 数值算例

定理3的构造性证明实际上给出了一种求对角线元素和迹均不为0且对角线元素彼此不等的这一类三阶方阵的等迹分解的方法.步骤如下:

Step 1  根据$\boldsymbol{D=}\frac{1}{{{\left( \text{tr}\boldsymbol{A} \right)}^{2}}}\boldsymbol{A}$D;

Step 2  根据定理2求E, F, (D=EF, trE=trF=1);

Step 3  根据(B=(trA)E, C=(trA)F)求B, C.

例1  求方阵A的一个等迹分解.其中

$ \mathit{\boldsymbol{A = }}\left( {\begin{array}{*{20}{c}} { - \frac{2}{{25}}}&{\frac{1}{{25}}}&{\frac{2}{{25}}}\\ {\frac{7}{{25}}}&{\frac{3}{{25}}}&{\frac{5}{{25}}}\\ {\frac{1}{{25}}}&{\frac{8}{{25}}}&{\frac{4}{{25}}} \end{array}} \right). $

  显然方阵A的对角线元素均不为0且彼此不等以及满足$\text{tr}\boldsymbol{A=}\frac{1}{5}\ne 0$.故

$ \mathit{\boldsymbol{D = }}\frac{1}{{{{\left( {{\rm{tr}}\mathit{\boldsymbol{A}}} \right)}^2}}}\mathit{\boldsymbol{A = }}\left( {\begin{array}{*{20}{c}} { - 2}&1&2\\ 7&3&5\\ 1&8&4 \end{array}} \right). $

根据定理2,可得

$ \mathit{\boldsymbol{E = }}{\rm{diag}}\left\{ { - 2, - 9,12} \right\}, $
$ \mathit{\boldsymbol{F = }}\left( {\begin{array}{*{20}{c}} 1&{ - \frac{1}{2}}&{ - 1}\\ { - \frac{7}{9}}&{ - \frac{1}{3}}&{ - \frac{5}{9}}\\ {\frac{1}{{12}}}&{\frac{2}{3}}&{\frac{1}{3}} \end{array}} \right). $

由以上可解出

$ \mathit{\boldsymbol{B = }}\left( {{\rm{tr}}\mathit{\boldsymbol{A}}} \right)\mathit{\boldsymbol{E = }}{\rm{diag}}\left\{ { - \frac{2}{5}, - \frac{9}{5},\frac{{12}}{5}} \right\}, $
$ \mathit{\boldsymbol{C = }}\left( {{\rm{tr}}\mathit{\boldsymbol{A}}} \right)\mathit{\boldsymbol{F = }}\left( {\begin{array}{*{20}{c}} {\frac{1}{5}}&{ - \frac{1}{{10}}}&{ - \frac{1}{5}}\\ { - \frac{7}{{45}}}&{ - \frac{1}{{15}}}&{ - \frac{1}{9}}\\ {\frac{1}{{60}}}&{\frac{2}{{15}}}&{\frac{1}{{15}}} \end{array}} \right). $
参考文献
[1] 房月华, 陈萍. 矩阵的满秩分解及其方法[J]. 衡水学院学报, 2011, 13(4): 16-18
FANG Yuehua, CHEN Ping. The methods of the full rank decomposition of the matrix[J]. Journal of Hengshui University, 2011, 13(4): 16-18
[2] 李先崇. 矩阵的满秩分解和强满秩矩阵的三角分解的初等变换法[J]. 贵州师范大学学报(自然科学版), 2004, 22(4): 57-59
LI Xianchong. Elementary transformation method of full rank decomposition of matrix and triangular decomposition of strongly full rank matrix[J]. Journal of Guizhou University(Natural Sciences Edition), 2004, 22(4): 57-59
[3] 王金林. 矩阵满秩分解的二种方法[J]. 南昌航空工业学院学报(自然科学版), 2004, 18(3): 37-40
WANG Jinlin. Two kinds of methods of full rank decomposition on matrix[J]. Journal of Nanchang Institute of Aeronautical Technology(Natural Sciences Edition), 2004, 18(3): 37-40
[4] 王群英. 矩阵分解方法的探究[J]. 长春工业大学学报(自然科学版), 2011, 32(1): 95-101
WANG Qunying. Study on the matrix decomposition methods[J]. Journal of Changchun University of Technology(Natural Sciences Edition), 2011, 32(1): 95-101
[5] 冯天祥, 李世宏. 矩阵的QR分解[J]. 西南民族学院学报(自然科学版), 2001, 27(4): 418-421
FENG Tianxiang, LI Shihong. The QR decomposition of matrix[J]. Journal of Southwest University for Nationalities(Natural Sciences Edition), 2001, 27(4): 418-421
[6] 李建东. 矩阵QR分解的三种方法[J]. 吕梁高等专科学校学报, 2009, 25(1): 16-19
LI Jiandong. Three methods of matrix QR decomposition[J]. Journal of Lvliang Higher College, 2009, 25(1): 16-19
[7] 喻方园. 矩阵QR分解的一个简单应用[J]. 工科数学, 1997, 13(1): 139-140
YU Fangyuan. A simple application of matrix QR decomposition[J]. Journal of Mathematics for Technology, 1997, 13(1): 139-140
[8] 刘秀梅. 矩阵QR分解途径的研究[J]. 内江师范学院学报, 2007, 22(4): 18-20
LIU Xiumei. The way of matrix QR decomposition[J]. Journal of Neijiang Teachers College, 2007, 22(4): 18-20
[9] 邓勇. 矩阵QR分解初等变换法的改进[J]. 高师理科学刊, 2008, 28(3): 19-20
DENG Yong. Improvemeton the elementary transformation method of QR decomposition of matrix[J]. Journal of Science of Teachers College and University, 2008, 28(3): 19-20
[10] 吴马威, 陈益智. 秩为1方阵的几个性质及应用[J]. 纺织高校基础科学学报, 2015, 28(1): 4-8
WU Mawei, CHEN Yizhi. Some properties and applications for rank-1 square matrix[J]. Basic Sciences Journal of Textile Universities, 2015, 28(1): 4-8
[11] 钟镇权. 对矩阵的迹的性质研究[J]. 柳州师专学报, 1997, 12(3): 54-60
ZHONG Zhenquan. Study on the nature of matrix trace[J]. Journal of Liuzhou Teachers College, 1997, 12(3): 54-60
[12] 胡永谟. 实对称矩阵的迹的几点性质[J]. 工科数学, 1997, 13(2): 137-139
HU Yongmo. On several properties of trace of real symmetric matrices[J]. Journal of Mathematics for Technology, 1997, 13(2): 137-139
[13] 邱润之, 王曼英. 矩阵迹的若干性质[J]. 南京邮电学院学报, 1994, 14(1): 83-88
QIU Runzhi, WANG Manying. On some properties of traces of matrices[J]. Journal of Nanjing Institute of Posts and Telecommunications, 1994, 14(1): 83-88
[14] 邱双月. 矩阵的迹[J]. 邯郸学院学报, 2005, 15(3): 18-19
QIU Shuangyue. The trace of the matrices[J]. Journal of Handan College, 2005, 15(3): 18-19
[15] 唐鹏程. 矩阵的迹及其应用[J]. 孝感学院学报(自然科学版), 2000, 20(4): 11-13
TANG Pengcheng. Matrix trace and its application[J]. Journal of Xiaogan University(Natural Sciences Edition), 2000, 20(4): 11-13
[16] 马菊侠. 关于矩阵迹的运算[J]. 陕西师范大学学报(自然科学版), 2003, 31(S1): 39-41
MA Juxia. Operations about matrix trace[J]. Journal of Shaanxi Normal University(Natural Sciences Edition), 2003, 31(S1): 39-41
[17] 侯识忠, 袁明华. 等迹矩阵[J]. 邵阳高专学报, 1994, 7(1): 1-5
HOU Shizhong, YUAN Minghua. Equal trace matrix[J]. Journal of Shaoyang College, 1994, 7(1): 1-5
西安工程大学、中国纺织服装教育学会主办
0

文章信息

吴马威, 陈益智, 骆莉芳.
WU Mawei, CHEN Yizhi, LUO Lifang.
二、三阶方阵的等迹分解
The equal trace decompositions of second order and third order square matrices
纺织高校基础科学学报, 2017, 30(1): 1-5
Basic Sciences Journal of Textile Universities, 2017, 30(1): 1-5.

文章历史

收稿日期: 2016-08-11

相关文章

工作空间