Skip to main content

Inequalities for eigenvalues of matrices

Abstract

The purpose of the paper is to present some inequalities for eigenvalues of positive semidefinite matrices.

MSC:15A18, 15A60.

1 Introduction

Throughout this paper, M n denotes the space of n×n complex matrices and H n denotes the set of all Hermitian matrices in M n . Let A,B H n ; the order relation AB means, as usual, that AB is positive semidefinite. We always denote the singular values of A by s 1 (A) s n (A). If A has real eigenvalues, we label them as λ 1 (A) λ n (A). Let denote any unitarily invariant norm on M n . We denote by |A| the absolute value operator of A, that is, |A|= ( A A ) 1 2 , where A is the adjoint operator of A.

For positive real number a, b, the arithmetic-geometric mean inequality says that

a b a + b 2 .

It is equivalent to

( a b ) m ( a + b 2 ) 2 m ,m=1,2,.
(1.1)

Let A,B M n be positive semidefinite. Bhatia and Kittaneh [1] proved that for all m=1,2, ,

λ j ( ( A B ) m ) λ j ( A + B 2 ) 2 m .
(1.2)

This is a matrix version of (1.1). For more information on matrix versions of the arithmetic-geometric mean inequality, the reader is referred to [111] and the references therein.

It is easy to see that the arithmetic-geometric mean inequality is also equivalent to

( a 3 / 4 b 3 / 4 ) 2 / 3 a + b 2 .
(1.3)

As pointed out in [10], p.198], although the arithmetic-geometric mean inequalities can be written in different ways and each of them may be obtained from the other, the matrix versions suggested by them are different.

In this note, we obtain a refinement of (1.2) and a log-majorization inequality for eigenvalues. As an application of our result, we give a matrix version of (1.3).

2 Main results

We begin this section with the following lemma, which is a question posed by Bhatia and Kittaneh [1] (see also [8, 10]) and settled in the affirmative by Drury in [2].

Lemma 2.1 Let A,B M n be positive semidefinite. Then

s j (AB) s j ( A + B 2 ) 2 .

As a consequence of Lemma  2.1, we have

| A B | 1 / 2 1 2 A+B.
(2.1)

It is a matrix version of the arithmetic-geometric mean inequality. By properties of the matrix square function, we know that this last inequality is stronger than the assertion

AB ( A + B 2 ) 2 ,

which is due to Bhatia and Kittaneh [1]and is also a matrix version of (1.1).

Theorem 2.1 Let A,B M n be positive semidefinite. Then for all m=1,2, ,

λ j ( ( A B ) m ) λ j ( A + B + A 1 / 2 B 1 / 2 + B 1 / 2 A 1 / 2 4 ) 2 m .
(2.2)

Proof By Lemma 2.1, we have

λ j ( ( A 2 B 2 ) m ) = ( λ j ( A 2 B 2 ) ) m = ( λ j ( A B 2 A ) ) m = ( s j ( A B ) ) 2 m s j ( A + B 2 ) 4 m = λ j ( A + B 2 ) 4 m .
(2.3)

Replacing A, B by A 1 / 2 , B 1 / 2 in (2.3), we have

λ j ( ( A B ) m ) λ j ( A + B + A 1 / 2 B 1 / 2 + B 1 / 2 A 1 / 2 4 ) 2 m .

This completes the proof. □

Remark 2.1 Let A,B M n be positive semidefinite. Note that

0 ( A 1 / 2 B 1 / 2 ) 2 2 = A + B 2 A + B + A 1 / 2 B 1 / 2 + B 1 / 2 A 1 / 2 4 .

Therefore, the inequality (2.2) is a refinement of the inequality (1.2).

Remark 2.2 For m=1, by (1.2), we have

λ j (AB) λ j ( A + B 2 ) 2 .
(2.4)

For m=1, by (2.2), we have

λ j ( A 2 B 2 ) λ j ( A + B 2 ) 4 .
(2.5)

In view of the inequalities (2.4) and (2.5), one may ask whether it is true that

λ j ( A m B m ) λ j ( A + B 2 ) 2 m
(2.6)

for all m=1,2, . The answer is no. For m=3, the inequality (2.6) is refuted by the following example:

A= [ 5 1 1 9 ] ,B= [ 6 4 4 5 ] .

Theorem 2.2 Let A,B M n be positive semidefinite. Then

j = 1 k | λ j ( A ( A v B 1 v + A 1 v B v 2 ) B ) | j = 1 k λ j ( A + B 2 ) 3 .

Proof By Weyl’s inequality, Horn’s inequality and Lemma 2.1, we have

j = 1 k | λ j ( A X B ) | = j = 1 k | λ j ( X A B ) | j = 1 k s j ( X A B ) j = 1 k s j ( X ) s j ( A B ) j = 1 k s j ( X ) j = 1 k s j ( A + B 2 ) 2 .
(2.7)

Putting

X= A v B 1 v + A 1 v B v 2 ,0v1,

in (2.7) gives

j = 1 k | λ j ( A ( A v B 1 v + A 1 v B v 2 ) B ) | j = 1 k s j ( A v B 1 v + A 1 v B v 2 ) j = 1 k s j ( A + B 2 ) 2 .
(2.8)

In response to a conjecture by Zhan [11], Audenaert [3] proved that if 0v1, then

s j ( A v B 1 v + A 1 v B v 2 ) s j ( A + B 2 ) .
(2.9)

The special case where v= 1 2 was obtained earlier in [6, 12] and the special case where v= 1 4 was obtained earlier in [13]. It follows from (2.8) and (2.9) that

j = 1 k | λ j ( A ( A v B 1 v + A 1 v B v 2 ) B ) | j = 1 k λ j ( A + B 2 ) 3 .

This completes the proof. □

Remark 2.3 As an application of Theorem 2.2, we now present a matrix version of (1.3). Taking v= 1 2 in this last inequality, we have

j = 1 k | λ j ( A 3 / 2 B 3 / 2 ) | j = 1 k s j ( A + B 2 ) 3

and so

j = 1 k s j ( A 3 / 4 B 3 / 4 ) j = 1 k s j ( A + B 2 ) 3 / 2 ,

which is equivalent to

j = 1 k s j ( | A 3 / 4 B 3 / 4 | 2 / 3 ) j = 1 k s j ( A + B 2 ) .

Since weak log-majorization is stronger than weak majorization, we have

j = 1 k s j ( | A 3 / 4 B 3 / 4 | 2 / 3 ) j = 1 k s j ( A + B 2 ) .

By Fan’s dominance theorem [4], p.93], we get

| A 3 / 4 B 3 / 4 | 2 / 3 1 2 A+B.
(2.10)

This is a matrix version of (1.3).

Next, we give another proof of the inequality (2.10). Araki [14] (also see [15]) obtained the following log-majorization inequality:

j = 1 k s j ( ( A p / 2 B p A p / 2 ) q / p ) j = 1 k s j ( A q / 2 B q A q / 2 ) ,0<pq.
(2.11)

Putting

p= 3 2 ,q=2

in (2.11) gives

j = 1 k s j ( ( A 3 / 4 B 3 / 2 A 3 / 4 ) 1 / 3 ) j = 1 k s j ( A B 2 A ) 1 / 4 ,

and so

j = 1 k s j ( | A 3 / 4 B 3 / 4 | 2 / 3 ) j = 1 k s j ( | A B | 1 / 2 ) .

By Fan’s dominance theorem [4], p.93], we get

| A 3 / 4 B 3 / 4 | 2 / 3 | A B | 1 / 2 .
(2.12)

It follows from (2.1) and (2.12) that

| A 3 / 4 B 3 / 4 | 2 / 3 1 2 A+B.

References

  1. Bhatia R, Kittaneh F: Notes on matrix arithmetic-geometric mean inequalities. Linear Algebra Appl. 2000, 308: 203–211. 10.1016/S0024-3795(00)00048-3

    Article  MATH  MathSciNet  Google Scholar 

  2. Drury SW: On a question of Bhatia and Kittaneh. Linear Algebra Appl. 2012, 437: 1955–1960. 10.1016/j.laa.2012.04.040

    Article  MATH  MathSciNet  Google Scholar 

  3. Audenaert KMR: A singular value inequality for Heinz means. Linear Algebra Appl. 2007, 422: 279–283. 10.1016/j.laa.2006.10.006

    Article  MATH  MathSciNet  Google Scholar 

  4. Bhatia R: Matrix Analysis. Springer, New York; 1997.

    Book  Google Scholar 

  5. Bhatia R, Davis C: More matrix forms of the arithmetic-geometric mean inequality. SIAM J. Matrix Anal. Appl. 1993, 14: 132–136. 10.1137/0614012

    Article  MATH  MathSciNet  Google Scholar 

  6. Bhatia R, Kittaneh F: On the singular values of a product of operators. SIAM J. Matrix Anal. Appl. 1990, 11: 272–277. 10.1137/0611018

    Article  MATH  MathSciNet  Google Scholar 

  7. Bhatia R: Interpolating the arithmetic-geometric mean inequality and its operator version. Linear Algebra Appl. 2006, 413: 355–363. 10.1016/j.laa.2005.03.005

    Article  MATH  MathSciNet  Google Scholar 

  8. Bhatia R, Kittaneh F: The matrix arithmetic-geometric mean inequality revisited. Linear Algebra Appl. 2008, 428: 2177–2191. 10.1016/j.laa.2007.11.030

    Article  MATH  MathSciNet  Google Scholar 

  9. Kosaki H: Arithmetic-geometric mean and related inequalities for operators. J. Funct. Anal. 1998, 156: 429–451. 10.1006/jfan.1998.3258

    Article  MATH  MathSciNet  Google Scholar 

  10. Bhatia R: Positive Definite Matrices. Princeton University Press, Princeton; 2007.

    Google Scholar 

  11. Zhan X Lecture Notes in Mathematics 1790. In Matrix Inequalities. Springer, Berlin; 2002.

    Chapter  Google Scholar 

  12. Ando T: Matrix Young inequalities. Oper. Theory Adv. Appl. 75. Operator Theory in Function Spaces and Banach Lattices 1995, 33–38.

    Chapter  Google Scholar 

  13. Tao Y: More results on singular value inequalities of matrices. Linear Algebra Appl. 2006, 416: 724–729. 10.1016/j.laa.2005.12.017

    Article  MATH  MathSciNet  Google Scholar 

  14. Araki H: On an inequality of Lieb and Thirring. Lett. Math. Phys. 1990, 19: 167–170. 10.1007/BF01045887

    Article  MATH  MathSciNet  Google Scholar 

  15. Hiai F: Matrix analysis: matrix monotone functions, matrix means, and majorization. Interdiscip. Inf. Sci. 2010, 16: 139–248.

    MATH  MathSciNet  Google Scholar 

Download references

Acknowledgements

The authors wish to express their heartfelt thanks to the referees for their detailed and helpful suggestions for revising the manuscript.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Xiaozeng Xu.

Additional information

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

All authors contributed equally to the manuscript and read and approved the final manuscript.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License (https://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and Permissions

About this article

Cite this article

Xu, X., He, C. Inequalities for eigenvalues of matrices. J Inequal Appl 2013, 6 (2013). https://doi.org/10.1186/1029-242X-2013-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/1029-242X-2013-6

Keywords

  • singular values
  • eigenvalues
  • unitarily invariant norm