Skip to main content

Several matrix trace inequalities on Hermitian and skew-Hermitian matrices

Abstract

In this paper, we present several matrix trace inequalities on Hermitian and skew-Hermitian matrices, which play an important role in designing and analyzing interior-point methods (IPMs) for semidefinite optimization (SDO).

1 Introduction

SDO is the generalization of linear optimization (LO), which is convex optimization over the intersection of an affine set and the cone of positive semidefinite matrices. Several IPMs designed for LO have been successfully extended to SDO [1, 2]. Some matrix trace inequalities are developed and applied in the analysis of IPMs for SDO (see [312]). In [3], Yang proved the arithmetic mean-geometric mean inequality for positive definite matrices, which was an open question proposed by Bellman in [4]; Neudecke used a different method in [5] to show a slightly relaxed version of Yang’s result in [3]; In [6], Coope considered alternative proofs of some simple matrix trace inequalities in [35] and further studied properties of products of Hermitian and positive (semi)definite matrices; In [7], Yang gave a new proof of the result obtained by Yang in [3] and extended it to a generalized positive semidefinite matrix. Based on the work in [35], Chang established a matrix trace inequality for products of Hermitian matrices in [8], which partly answers a conjecture proposed by Bellman in [4]. In addition, Yang gave a matrix trace inequality for products of positive semidefinite matrices in [9]; In [10], Yang et al. established a matrix trace inequality for positive semidefinite matrices, which improved the result given by Yang in [9]. Although there have been many results on matrix trace inequality, some important matrix trace inequality problems have not been fully solved. In this paper, we will provide several matrix trace inequalities on Hermitian and skew-Hermitian matrices, which play an important role in designing and analyzing IPMs for SDO.

This paper is organized as follows: In Section 2, a matrix trace inequality on 2×2 Hermitian and skew-Hermitian matrices is provided, and its simple proof is given by using an elementary method. However, it is difficult for us to use this method to deal with the high-dimensional case. Based on Lemmas 2 and 3 in Section 2, the high-dimensional case will be shown as Corollary 2 in Section 3. In Section 4, some conclusions are made.

The following notations are used throughout the paper. N, R + , C, and C n denote the set of natural numbers, the set of nonnegative real numbers, the set of complex numbers and the set of vectors with n components, respectively. C n × n is the space of all n×n matrices over C. Define

Span{ v 1 ,, v n }={ k 1 v 1 ++ k n v n | k i C,i=1,2,,n},

where v 1 , v 2 ,, v n C n . The vector inner product of α C n and β C n is defined by (α,β). || and denote modules for complex numbers and the 2-norm for vectors, respectively. For A,B C n × n , A represents the conjugate transpose of A, AB (A>B) means that AB is positive semidefinite (positive definite). For any Hermitian positive definite matrix Q, the expression Q 1 2 denotes the Hermitian square root of Q. Similarly, the power Q r can be defined for any Q>0 and rR.

2 Preliminary results

In this section, we will present a matrix trace inequality on 2×2 Hermitian and skew-Hermitian matrices.

Lemma 1 Let N>0 and M be 2×2 Hermitian and skew-Hermitian matrices, respectively. Then

tr ( ( ( N + M ) ( N + M ) ) 1 2 ) tr ( N 1 ) .

Proof From N>0, there exists a unitary matrix U such that

N=UΛ U ,

where

Λ=diag{ n 1 , n 2 }and n 1 n 2 >0.

Hence,

tr ( N 1 ) =tr ( Λ 1 )

and

tr ( ( ( N + M ) ( N + M ) ) 1 2 ) =tr ( ( ( Λ + U M U ) ( Λ + U M U ) ) 1 2 ) ,

where U MU is a skew-symmetric matrix.

Without loss of generality, let

N=diag{ n 1 , n 2 }andM= [ 0 m m ¯ 0 ] ,

where n 1 n 2 >0. Then

N+M= [ n 1 m m ¯ n 2 ] .

Thus

(N+M) ( N + M ) = [ n 1 2 + | m | 2 m ( n 2 n 1 ) m ¯ ( n 2 n 1 ) n 2 2 + | m | 2 ] .

From

det ( λ I 2 ( N + M ) ( N + M ) ) =0,

it follows that

λ 2 ( n 1 2 + n 2 2 + 2 | m | 2 ) λ+ ( n 1 n 2 + | m | 2 ) 2 =0.
(1)

Suppose λ 1 and λ 2 are two roots of (1), where λ 1 λ 2 . According to Weda’s theorem, we obtain

{ λ 1 + λ 2 = n 1 2 + n 2 2 + 2 | m | 2 , λ 1 λ 2 = n 1 2 n 2 2 + 2 n 1 2 | m | 2 + | m | 4 .
(2)

From (2), it is easy to compute that

( 1 λ 1 + 1 λ 2 ) 2 = λ 1 + λ 2 + 2 λ 1 λ 2 λ 1 λ 2 = n 1 2 + n 2 2 + 2 | m | 2 + 2 ( n 1 n 2 + | m | 2 ) ( n 1 n 2 + | m | 2 ) 2 = ( n 1 + n 2 ) 2 + 4 | m | 2 ( n 1 n 2 + | m | 2 ) 2 .
(3)

On the other hand, it is clear that

02 n 1 n 2 ( n 1 2 + n 2 2 ) | m | 2 + ( n 1 + n 2 ) 2 | m | 4 .
(4)

According to (4), it follows that

4 | m | 2 n 1 2 n 2 2 2 n 1 n 2 ( n 1 + n 2 ) 2 | m | 2 + ( n 1 + n 2 ) 2 | m | 4 .

This implies

n 1 2 n 2 2 ( n 1 + n 2 ) 2 +4 | m | 2 n 1 2 n 2 2 ( n 1 n 2 + | m | 2 ) 2 ( n 1 + n 2 ) 2 .
(5)

Thus, (5) can be written as

( n 1 + n 2 ) 2 + 4 | m | 2 ( n 1 n 2 + | m | 2 ) 2 ( n 1 + n 2 ) 2 n 1 2 n 2 2 .

This, together with (3), yields

( 1 λ 1 + 1 λ 2 ) 2 ( n 1 + n 2 ) 2 n 1 2 n 2 2 = ( 1 n 1 + 1 n 2 ) 2 ,

i.e.,

1 λ 1 + 1 λ 2 1 n 1 + 1 n 2 .

Hence,

tr ( ( ( N + M ) ( N + M ) ) 1 / 2 ) tr ( N 1 ) .

 □

Remark 1 In this lemma, Weda’s theorem is applied to prove the matrix trace inequality of 2×2 Hermitian and skew-Hermitian matrices. However, it is difficult for us to extend this result to the high-dimensional case, because it is too complex for us to compute the matrix traces of n×n Hermitian and skew-Hermitian matrices by using Weda’s theorem, when n3.

In order to prove the high-dimensional case of Lemma 1, we need to give the following two lemmas.

Lemma 2 [13]

Let A C n × n . Then x (A+ A )x2 x A A x , x C n , x=1.

Lemma 3 [13]

Let H= H C n × n with eigenvalues λ 1 (H) λ 2 (H) λ n (H) and their corresponding orthonormal eigenvectors v 1 , v 2 ,, v n , respectively. Define

V=span{ v p , v p + 1 ,, v q },1pqn.

Then

λ p (H) v Hv λ q (H)

for any vV with v=1.

3 Main results

In this section, we develop several matrix trace inequalities on Hermitian and skew-Hermitian matrices. Furthermore, Lemma 1 is extended to the high-dimensional case as Corollary 2.

Theorem 1 Let

A C n × n ,H= A + A 2 .

If σ 1 (A), σ 2 (A),, σ n (A) and λ 1 (H), λ 2 (H),, λ n (H) are singular values of A and eigenvalues of H, respectively. They are arranged in such a way that

σ 1 (A) σ 2 (A) σ n (A), λ 1 (H) λ 2 (H) λ n (H).

Then

σ k (A) λ k (H),

where k=1,2,,n.

Proof From the definition of singular values of A, it follows that

σ 1 2 (A), σ 2 2 (A),, σ n 2 (A)

are eigenvalues of A A, and their corresponding orthonormal eigenvectors of A A are denoted as u 1 , u 2 ,, u n , respectively. Suppose v 1 , v 2 ,, v n are orthonormal eigenvectors of H, and

λ 1 (H), λ 2 (H),, λ n (H)

are eigenvalues of H corresponding to these orthonormal eigenvectors, respectively. Define

W 1 k =Span{ v k , v k + 1 ,, v n }

and

W k n =Span{ u 1 , u 2 ,, u k },

where k=1,2,,n. It is clear that

dim( W 1 k )=nk+1anddim( W k n )=k.

Then

dim( W 1 k )+dim( W k n )=n+1.

According to the dimensional formula, it follows that

dim( W 1 k + W k n )+dim( W 1 k W k n )=dim( W 1 k )+dim( W k n ).

This implies that

dim( W 1 k W k n ){0}.

From Lemma 3, it follows that

σ k 2 (A) w ( A A ) wand w Hw λ k (A)

for any w W 1 k W k n with w=1. According to Lemma 2, we obtain

σ k (A) w ( A A ) w w Hw λ k (H),

where k=1,2,,n. This completes the proof. □

Remark 2 In this section, we prove Theorem 1 by using a simple and elementary method, which is slightly different from the proof method in [14]. This theorem reveals the relationship between singular values and eigenvalue of matrices. Based on it, several matrix trace inequalities on Hermitian and skew-Hermitian matrices will be obtained immediately.

Theorem 2 Let A C n × n , H= A + A 2 . The following matrix trace inequalities are satisfied.

  1. (I)

    If H0, then

    tr ( ( A A ) r ) tr ( H 2 r ) ,r R + .
  2. (II)

    If H>0, then

    tr ( ( A A ) r ) tr ( H 2 r ) ,r R + .
  3. (III)
    tr ( ( A A ) 2 k + 1 2 ) tr ( H 2 k + 1 ) ,kN.

Proof (I) From Theorem 1, it follows that

tr ( ( A A ) r ) = i = 1 n λ i ( ( A A ) r ) = i = 1 n ( λ i ( A A ) ) r = i = 1 n ( σ i ( A ) ) 2 r i = 1 n ( λ i ( H ) ) 2 r = i = 1 n λ i ( H 2 r ) = tr ( H 2 r ) , r R + .

Similar to the proof of (I), we can easily verify that (II) and (III) hold. This completes the proof. □

Let

A=M+N,H= A + A 2 ,

where N C n × n and M C n × n be Hermitian and skew-Hermitian matrices, respectively. By using Theorem 2, the following conclusions will be obtained immediately.

Corollary 1 Let N C n × n and M C n × n be Hermitian and skew-Hermitian matrices, respectively. The following matrix trace inequalities are satisfied.

  1. (I)

    If H0, then

    tr ( ( ( N + M ) ( N + M ) ) r ) tr ( N 2 r ) .
  2. (II)

    If H>0, then

    tr ( ( ( N + M ) ( N + M ) ) r ) tr ( N 2 r ) .
  3. (III)
    tr ( ( ( N + M ) ( N + M ) ) 2 k + 1 2 ) tr ( N 2 k + 1 ) .

Based on Corollary 1, the following results can be obtained when r= 1 2 . This implies that the proof of the general case of Lemma 1 is completed.

Corollary 2 Let N C n × n and M C n × n be a Hermitian and skew-Hermitian matrices, respectively. The following matrix trace inequalities are satisfied.

  1. (I)

    If N0, then

    tr ( ( ( N + M ) ( N + M ) ) 1 2 ) tr(N).
  2. (II)

    If N>0, then

    tr ( ( ( N + M ) ( N + M ) ) 1 2 ) tr ( N 1 ) .

4 Conclusions

This paper proves several matrix trace inequalities on Hermitian and skew-Hermitian matrices. These matrix trace inequalities can be applied to design and analyze interior-point methods (IPMs) for semidefinite optimization (SDO). In addition, matrix trace inequalities have many potential applications in control theory, for example, stabilization of time-delay systems (see [1518]).

References

  1. Peng J, Roos C, Terlaky T: Self-regular functions and new search directions for linear and semidefinite optimization. Math. Program. 2002,93(1):129-171. 10.1007/s101070200296

    Article  MathSciNet  MATH  Google Scholar 

  2. Zhang Y: On extending some primal-dual interior-point algorithms from linear programming to semidefinite programming. SIAM J. Control Optim. 1998,8(2):365-386. 10.1137/S1052623495296115

    Article  MATH  MathSciNet  Google Scholar 

  3. Yang YS: A matrix trace inequality. J. Math. Anal. Appl. 1988,133(2):573-574. 10.1016/0022-247X(88)90423-4

    Article  MathSciNet  MATH  Google Scholar 

  4. Bellman R: Some inequalities for positive definite matrices. 2nd Internat. Conf. on General Inequalities 1980, 88-89.

    Google Scholar 

  5. Neudecke H: A matrix trace inequality. J. Math. Anal. Appl. 1992, 166: 302-303. 10.1016/0022-247X(92)90344-D

    Article  MathSciNet  Google Scholar 

  6. Coope ID: On matrix trace inequalities and related topics for products of Hermitian matrices. J. Math. Anal. Appl. 1994,188(3):999-1001. 10.1006/jmaa.1994.1475

    Article  MathSciNet  MATH  Google Scholar 

  7. Yang XM: A generalization of a matrix trace inequality. J. Math. Anal. Appl. 1995, 189: 897-900. 10.1006/jmaa.1995.1056

    Article  MathSciNet  MATH  Google Scholar 

  8. Chang DW: A matrix trace inequality for products of Hermitian matrices. J. Math. Anal. Appl. 1999, 237: 721-725. 10.1006/jmaa.1999.6433

    Article  MathSciNet  MATH  Google Scholar 

  9. Yang XJ: A matrix trace inequality. J. Math. Anal. Appl. 2000,250(1):372-374. 10.1006/jmaa.2000.7068

    Article  MathSciNet  MATH  Google Scholar 

  10. Yang XM, Yang XQ, Teo KL: A matrix trace inequality. J. Math. Anal. Appl. 2001,263(1):327-331. 10.1006/jmaa.2001.7613

    Article  MathSciNet  MATH  Google Scholar 

  11. Yang ZP, Feng XX: A note on the trace inequality for products of Hermitian matrix power. J. Inequal. Pure Appl. Math. 2002. Article ID 78,3(5): Article ID 78

    MATH  Google Scholar 

  12. Ulukök Z, Türkmen R: On some matrix trace inequalities. J. Inequal. Appl. 2010. Article ID 201486, 2010: Article ID 201486 10.1155/2010/201486

    Google Scholar 

  13. Zhang X, Zhong GP, Gao XY: Matrix Theory of the System and Control. Heilongjiang University Press, Harbin; 2011. (in Chinese) (in Chinese)

    Book  Google Scholar 

  14. Fan K, Hoffman A: Some metric inequalities in the space of matrices. Proc. Am. Math. Soc. 1955,6(1):111-116. 10.1090/S0002-9939-1955-0067841-7

    Article  MathSciNet  MATH  Google Scholar 

  15. Zhou B, Gao HJ, Lin ZL, Duan GR: Stabilization of linear systems with distributed input delay and input saturation. Automatica 2012,48(5):712-724. 10.1016/j.automatica.2012.02.007

    Article  MathSciNet  MATH  Google Scholar 

  16. Zhou B, Lin ZL, Duan GR: Truncated predictor feedback for linear systems with long time-varying input delays. Automatica 2012,48(10):2387-2399. 10.1016/j.automatica.2012.06.032

    Article  MathSciNet  MATH  Google Scholar 

  17. Zhou B, Li ZY, Lin ZL: Observer based output feedback control of linear systems with input and output delays. Automatica 2013,49(7):2039-2052. 10.1016/j.automatica.2013.03.031

    Article  MathSciNet  Google Scholar 

  18. Zhou B: Truncated Predictor Feedback for Time-Delay Systems. Springer, Heidelberg; 2014.

    Book  MATH  Google Scholar 

Download references

Acknowledgements

The authors thank the editor and the referees for their valuable suggestions to improve the quality of this paper. This work was partially supported by the National Natural Science Foundation of China (No. 11371006), Natural Science Foundation of Heilongjiang Province (No. A201416), the Undergraduate Innovation and Entrepreneurship Program of Heilongjiang University (No. 2014SX04) and the Laboratory Project of Heilongjiang University.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Xiangyu Gao.

Additional information

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

All authors contributed equally to the writing of this paper. All authors 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

Verify currency and authenticity via CrossMark

Cite this article

Gao, X., Wang, G., Zhang, X. et al. Several matrix trace inequalities on Hermitian and skew-Hermitian matrices. J Inequal Appl 2014, 358 (2014). https://doi.org/10.1186/1029-242X-2014-358

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/1029-242X-2014-358

Keywords

  • Hermitian matrix
  • skew-Hermitian matrix
  • Hermitian positive definite matrix
  • semidefinite optimization
  • interior-point methods