International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Volume 7, Issue 8, 2019, PP 3-7
ISSN No. (Print) 2347-307X & ISSN No. (Online) 2347-3142 DOI: http://dx.doi.org/10.20431/2347-3142.0708002 www.arcjournals.org
The Inequalities of Positive Semi-Definite Block Matrix with Partial Order Relations Hui Quan* Department of mathematics, Xiangtan University, Xiangtan411105, China
*Corresponding Author: Hui Quan, Department of mathematics, Xiangtan University, Xiangtan 411105, China
Abstract: In 2019,Zübeyde Ulukök obtained an important theorem in reference [1]:When ACis a H * CB A positive semi-definite matrix, then rr1 ,where A , B are n-order square matrices. In HAB3[11 ( ) ( )] B this paper, we firstly do the same thing for a 33 positive semi-definite block matrix, and give a generalization of the above theorem. Next, we further generalize the case of kk positive semi-definite block matrix, and discuss the partial ordering relationship between the sum of matrices on quasi-diagonal lines and block matrices at other locations. Thus, we gave a new eigenvalue inequality. Keywords: Positive Semi-Definite Matrix Hermitian Matrix Partial Order Relation
1. INTRODUCTION AND PRELIMINARIES Inequalities of positive semi-definite block matrices have been widely used in matrix theory. In recent years, inequalities about block matrices have become a hot topic of research. At the same time, some very good results have been obtained, such as references [1, 3, 4, 5]. In this paper, we mainly discussed some positive semi-definite block matrices and obtained some matrix inequalities. As we all know, the positive semi-definite block matrices have very good properties. Their eigenvalues are real numbers, so they can always be arranged in ascending order and we recorded then as
n ()()()AAA 21. In this paper, we use symbol 1()A to represent the largest eigenvalue of a positive semi-definite matrix A . And use symbol AB to represent BA be a positive semi-definite matrix, obviously, "" is a partial order relation. In particular, A 0denotes that matrix A is positive semi-definite. In addition, we call U a unitary matrix if it satisfiesUUI ,we call A a Hermitian matrix if it satisfies AA . Last, the AB denotes the direct sum of A and B , the block diagonal matrix A 0 ; and 0 represents a zero block matrix. 0 B
2. MAIN RESULTS Let's start with the following lemmas
Lemma1.1 Let AM mn, with mn , then (AA ) ( A A 0) with0Mmn .
Lemma1.2 Let be positive semi-definite matrix. Then , where denotes AM n n ()()AIAAI1 I
International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page | 3 The Inequalities of Positive Semi-Definite Block Matrix with Partial Order Relations
the identity matrix in . M n
AB Lemma1.3 If is a positive semi-definite matrix, then AB . BA
1 ABI Proof Just consider the equation ABII 0 and we can get the conclusion. 2 BAI
AB Lemma1.4 If * is a positive semi-definite matrix, AC, are square matrices of the same BC order, then we have A C B B * and ACBB ()* .
AB Proof Becase* is a positive semi-definite matrix, So, for any unitary matrix U ,we have BC
* AB ABAB * UU* 0 , thus there is **UU0 . In particular, take BC BCBC
0 I ACBB* U ,then we have .From lemma 1.3 we can get that * 0 I 0 BBAC
. And we can draw another conclusion by replacing the original B with B .
Lemma1.5 If HM n is a positive semi-definite matrix, then for any nk unitary matrixU satisfying
* * UUI k and for eachik1,..., , we have i m k()()()H i U HU i H .
ADE Theorem 1 * is a positive semi-definite matrix, where A , B ,C are all n-order square HDBF ** EFC matrices,then A rr1 for r 1. HABCB3[ ( )( )( )] 111 C ProofBecause H is a positive semi-definite matrix, so there exists an invertible matrix P such that
* HPP . We divide P into blocks: PXYZ[,,], where XYZM,, 3nn , and we can know that
XXA , YYB , ZZC . From Lemma 1.1, we can get the following results:
* 111()(0)(XXX XA), 111()(0)(YYY YB ),
1(ZZ ) 1 ( Z Z 0) 1 ( C ).
Noting that the following equation holds: Hr()() P* P r P * PP * r 1 P ,and we denote
International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page | 4 The Inequalities of Positive Semi-Definite Block Matrix with Partial Order Relations
T P P()*1r . So, for H r , there is the following decompositions:
* X*** TXX TYX TZXTTTX r *** HP TPY TXY TYY TZYTTTY *** Z TXZ TYZ TZZTTTZ
* 11 TTT TT 220000 We notice that TTT , and . 11 TTTT 3() TTTTT 220000 TTT TTT 11 22 TT 0000 So, from lemma 1.2. We can get that: * XXAA HTYYTBPPBrr3().3()3() *1 111 ZZCC
A A ***1 r 3() r1***XXYYZZB 3[()()()]XXYYZZB 1 111 C C
A . So, the theorem is proved. 3[()()()]ABCB r1 111 C On the basis of this conclusion, we can easily get the results of Zübeyde Ulukök in reference[1]:
AC Corollary 1 is a positive semi-definite matrix, where A , B are n-order square H * CB matrices,then
rr1 A HAB3[()()]11 for r 1. B Next, we will discuss the case of kk positive semi-definite block matrices and explore the relationship between quasi-diagonal matrices and other block matrices at other locations.
MMM11121 k Theorem 2 MMM* is a positive semi-definite matrix, and M , …, M are H 12222 k 11 kk ** MMM12kkkk all n-order square matrices, if it satisfied M ij are all Hermitian matrices, for
(1i j n ), then
k 2 MMii ij i11k 1 i j n
Proof First of all, notice the following facts: ifU1 ,…, U k are all unitary matrices, A is a positive
** semi-definite matrix, thenU11 AU... Ukk AU is still a positive semi-definite matrix.
International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page | 5 The Inequalities of Positive Semi-Definite Block Matrix with Partial Order Relations
kk1 MM iiil : After calculation, we can get that 0 . From lemma 1.4 we can get that ili11MMilll k11 k k k , (MMMMMii ll ) ( il il ) 2 ij i1 l i 1 i 1 l i 1 1 i j n k 2 So, there is MMiiij . iijn11k 1
AB Corollary 2If * is a positive semi-definite matrix, AC, are square matrices of the same BC order, and B is a Hermitian matrix, then trBtrAtrB max, .
MMM11121 k * Theorem3 MMM12222 k is a positive semi-definite matrix, and M , …, M are all H 11 kk ** MMM12kkkk n-order square matrices, if it satisfied M ij are all Hermitian matrices, for(1 i j n ),then for these
1in, we have 2 iiji()() MH k 1ijn
I * Proof Take a unitary matrix 1 I , thenU satisfiedUUI n , noticed that M ij are all Hermitian U k I matrices, for ( ),so we have
k * 12 122 U HUMM iiij+ , MMijij+ (from Theorem 2) kkiij11 n k kk111ij nij n
2 1 2 MMij( 1) ij . k11i j n k1 k i j n
2 Then for 1in, from lemma 1.5 we have iijii()()( MU ) HUH , thus the k 1i j n
2 conclusion i()() MH ij i is proved. k 1i j n
REFERENCES [1] Zübeyde Ulukök, More inequalities for positive semidefinite 2 ×2block matrices and their blocks [J]. Linear Algebra and its Applications 572 (2019) 51–67 [2] X. Zhan, Matrix Inequalities, Lecture Notes in Math., vol. 1790, Springer-Verlag, Berlin, 2002.
International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page | 6 The Inequalities of Positive Semi-Definite Block Matrix with Partial Order Relations
[3] Y. Zhang, Eigenvalue majorization inequalities for positive semidefinite block matrices and their blocks, Linear Algebra Appl. 446 (2014) 216–223. [4] F. Zhang, Matrix inequalities by means of block matrices, Mathematical Inequalities and Applications, Vol. 4, No. 4, (2001), 481-490. [5] R. Türkmen and Z. Ulukök, Inequalities for singular values of positive semidefinite block matrices[J]. International Mathematical Forum, Vol. 6, 2011, no. 31, 1535 – 1545
Citation: Hui Quan (2019).The Inequalities of Positive Semi-Definite Block Matrix with Partial Order Relations. International Journal of Scientific and Innovative Mathematical Research (IJSIMR), 7(8), pp. 3-7. http://dx.doi.org/ 10.20431/2347 -3142.0708002 Copyright: © 2019 Authors, this is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page | 7