# The estimates of diagonally dominant degree and eigenvalues distributions for the Schur complements of matrices

## Abstract

By applying the properties of Schur complement and some inequality techniques, some new estimates of diagonally dominant degree on the Schur complement of matrices are obtained, which improve the main results of Liu (SIAM J. Matrix Anal. Appl. 27:665-674, 2005) and Liu (Linear Algebra Appl. 432:1090-1104, 2010). As an application, we present some new distribution theorems for eigenvalues of the Schur complement. Finally, we give a numerical example to illustrate the theory results.

MSC:15A45, 15A48.

## 1 Introduction

Let ${\mathbb{C}}^{n×n}$ denote the set of all $n×n$ complex matrices, $N=\left\{1,2,\dots ,n\right\}$ and $A=\left({a}_{ij}\right)\in {\mathbb{C}}^{n×n}$ ($n\ge 2$). Denote

$\begin{array}{c}{R}_{i}\left(A\right)=\sum _{j\ne i}|{a}_{ij}|,\phantom{\rule{2em}{0ex}}{C}_{i}\left(A\right)=\sum _{j\ne i}|{a}_{ji}|,\phantom{\rule{1em}{0ex}}i\in N,\hfill \\ {N}_{r}\left(A\right)=\left\{i:|{a}_{ii}|>{R}_{i}\left(A\right),i\in N\right\},\phantom{\rule{2em}{0ex}}{N}_{c}\left(A\right)=\left\{i:|{a}_{ii}|>{C}_{i}\left(A\right),i\in N\right\}.\hfill \end{array}$

We call A a strictly diagonally dominant matrix (abbreviated $S{D}_{n}$) if

$|{a}_{ii}|>{R}_{i}\left(A\right),\phantom{\rule{1em}{0ex}}\mathrm{\forall }i\in N.$

A is called an Ostrowski matrix (abbreviated $O{S}_{n}$) (see ) if

$|{a}_{ii}||{a}_{jj}|>{R}_{i}\left(A\right){R}_{j}\left(A\right),\phantom{\rule{1em}{0ex}}\mathrm{\forall }i,j\in N,i\ne j.$

As in , for all $i\in N$ and $\alpha \in \left[0,1\right]$, we call $|{a}_{ii}|-{R}_{i}\left(A\right)$, $|{a}_{ii}|-\alpha {R}_{i}\left(A\right)-\left(1-\alpha \right){C}_{i}\left(A\right)$ and $|{a}_{ii}|-{\left[{R}_{i}\left(A\right)\right]}^{\alpha }{\left[{C}_{i}\left(A\right)\right]}^{1-\alpha }$ the i th dominant degree, α-dominant degree and product α-dominant degree of A, respectively.

For $\beta \subseteq N$, denote by $|\beta |$ the cardinality of β and $\overline{\beta }=N/\beta$. If $\beta ,\gamma \subseteq N$, then $A\left(\beta ,\gamma \right)$ is the submatrix of A lying in the rows indexed by β and the columns indicated by γ. In particular, $A\left(\beta ,\beta \right)$ is abbreviated to $A\left(\beta \right)$. If $A\left(\beta \right)$ is nonsingular, then

$A/\beta =A/A\left(\beta \right)=A\left(\overline{\beta }\right)-A\left(\overline{\beta },\beta \right){\left[A\left(\beta \right)\right]}^{-1}A\left(\beta ,\overline{\beta }\right)$

is called the Schur complement of A with respect to $A\left(\beta \right)$.

The comparison matrix of A, $\mu \left(A\right)=\left({\alpha }_{ij}\right)$ is defined by

A matrix A is called an M-matrix if there exist a nonnegative matrix B and a real number $s>\rho \left(B\right)$ such that $A=sI-B$, where $\rho \left(B\right)$ is the spectral radius of B. It is well known that A is an H-matrix if and only if $\mu \left(A\right)$ is an M-matrix, and if A is an M-matrix, then the Schur complement of A is also an M-matrix and $detA>0$ (see ). We denote by ${H}_{n}$ and ${M}_{n}$ the sets of H-matrices and M-matrices, respectively.

The Schur complement has been proved to be a useful tool in many fields such as control theory, statistics and computational mathematics, and many works have been done on it (see ). Meanwhile, studying the locations of eigenvalues of the Schur complement of matrices is of great significance as shown in [2, 3, 914]. In this paper, we present some new estimates of diagonally dominant degree on the Schur complement of matrices and use them to study the distributions for the eigenvalues of the Schur complement of matrices.

The paper is organized as follows. In Section 2, we give several new estimates of the diagonally dominant degree, the α-diagonally dominant degree and product α-diagonally dominant degree on the Schur complement of matrices. In Section 3, several new distribution theorems for eigenvalues of the Schur complements are obtained. In Section 4, we present a numerical example to illustrate the theory results.

## 2 Estimates of diagonally dominant degree for the Schur complement

In this section, we present several new estimates for the diagonally dominant degree, the α-diagonally dominant degree and product α-diagonally dominant degree of the Schur complement of matrices.

Lemma 1 

If $A\in {H}_{n}$, then ${\left[\mu \left(A\right)\right]}^{-1}\ge |{A}^{-1}|$.

Lemma 2 

If A is an $S{D}_{n}$ or A is an $O{S}_{n}$, then $A\in {H}_{n}$, i.e., $\mu \left(A\right)\in {M}_{n}$.

Lemma 3 

If A is an $S{D}_{n}$ or A is an $O{S}_{n}$ and $\beta \subseteq N$, then the Schur complement of A is an $S{D}_{|\overline{\beta }|}$ or an $O{S}_{|\overline{\beta }|}$, where $\overline{\beta }=N-\beta$ is the complement of β in N, and $|\overline{\beta }|$ is the cardinality of $\overline{\beta }$.

Lemma 4 

Let $a>b$, $c>b$, $b>0$ and $0\le \alpha \le 1$. Then

${a}^{\alpha }{c}^{1-\alpha }\ge {\left(a-b\right)}^{\alpha }{\left(c-b\right)}^{1-\alpha }+b.$

Theorem 1 Let $A=\left({a}_{ij}\right)\in {\mathbb{C}}^{n×n}$, $\beta =\left\{{i}_{1},{i}_{2},\dots ,{i}_{k}\right\}\subseteq {N}_{r}\left(A\right)\ne \varphi$, $\overline{\beta }=\left\{{j}_{1},{j}_{2},\dots ,{j}_{l}\right\}$, $k, and let $A/\beta =\left({a}_{ts}^{\prime }\right)$. Then for all $1\le t\le l$,

$|{a}_{tt}^{\prime }|-{R}_{t}\left(A/\beta \right)\ge |{a}_{{j}_{t}{j}_{t}}|-{R}_{{j}_{t}}\left(A\right)+{\delta }_{{j}_{t}}\ge |{a}_{{j}_{t}{j}_{t}}|-{R}_{{j}_{t}}\left(A\right),$
(1)

and

$|{a}_{tt}^{\prime }|+{R}_{t}\left(A/\beta \right)\le |{a}_{{j}_{t}{j}_{t}}|+{R}_{{j}_{t}}\left(A\right)-{\delta }_{{j}_{t}}\le |{a}_{{j}_{t}{j}_{t}}|+{R}_{{j}_{t}}\left(A\right),$
(2)

where

$\begin{array}{c}{\delta }_{{j}_{t}}=\underset{1\le u\le k}{min}\frac{|{a}_{{i}_{u}{i}_{u}}|-{P}_{{i}_{u}}\left(A\right)}{|{a}_{{i}_{u}{i}_{u}}|}\sum _{v=1}^{k}|{a}_{{j}_{t}{i}_{v}}|,\phantom{\rule{2em}{0ex}}r=\underset{1\le u\le k}{max}\frac{{\sum }_{v=1}^{l}|{a}_{{i}_{u}{j}_{v}}|}{|{a}_{{i}_{u}{i}_{u}}|-{\sum }_{v\ne u}^{k}|{a}_{{i}_{u}{i}_{v}}|},\hfill \\ {P}_{{i}_{u}}\left(A\right)=r\sum _{v\ne u}^{k}|{a}_{{i}_{u}{i}_{v}}|+\sum _{v=1}^{l}|{a}_{{i}_{u}{j}_{v}}|,\phantom{\rule{1em}{0ex}}{i}_{u}\in \beta ,1\le u\le k.\hfill \end{array}$

Proof Since $\beta \subseteq {N}_{r}\left(A\right)\ne \varphi$, then $A\left(\beta \right)\in {H}_{k}$, $\mu \left(A\left(\beta \right)\right)\in {M}_{k}$. From Lemma 1 and Lemma 2, we have

${\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\ge |{\left[A\left(\beta \right)\right]}^{-1}|.$

Thus, for $1\le t\le l$,

$\begin{array}{r}|{a}_{tt}^{\prime }|-{R}_{t}\left(A/\beta \right)\\ \phantom{\rule{1em}{0ex}}=|{a}_{tt}^{\prime }|-\sum _{s\ne t}^{l}|{a}_{ts}^{\prime }|\\ \phantom{\rule{1em}{0ex}}\ge |{a}_{{j}_{t}{j}_{t}}|-\sum _{s\ne t}^{l}|{a}_{{j}_{t}{j}_{s}}|-\sum _{s=1}^{l}\left(|{a}_{{j}_{t}{i}_{1}}|,\dots ,|{a}_{{j}_{t}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{s}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{s}}|\end{array}\right)\\ \phantom{\rule{1em}{0ex}}=|{a}_{{j}_{t}{j}_{t}}|-{R}_{{j}_{t}}\left(A\right)+\sum _{v=1}^{k}|{a}_{{j}_{t}{i}_{v}}|+{\delta }_{{j}_{t}}-{\delta }_{{j}_{t}}-\sum _{s=1}^{l}\left(|{a}_{{j}_{t}{i}_{1}}|,\dots ,|{a}_{{j}_{t}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{s}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{s}}|\end{array}\right)\\ \phantom{\rule{1em}{0ex}}=|{a}_{{j}_{t}{j}_{t}}|-{R}_{{j}_{t}}\left(A\right)+{\delta }_{{j}_{t}}\\ \phantom{\rule{2em}{0ex}}+\frac{1}{det\left[\mu \left(A\left(\beta \right)\right)\right]}det\left(\begin{array}{cccc}{\sum }_{v=1}^{k}|{a}_{{j}_{t}{i}_{v}}|-{\delta }_{{j}_{t}}& -|{a}_{{j}_{t}{i}_{1}}|& \cdots & -|{a}_{{j}_{t}{i}_{k}}|\\ -{\sum }_{s=1}^{l}|{a}_{{i}_{1}{j}_{s}}|\\ ⋮& \mu \left(A\left(\beta \right)\right)\\ -{\sum }_{s=1}^{l}|{a}_{{i}_{k}{j}_{s}}|\end{array}\right)\\ \phantom{\rule{1em}{0ex}}\stackrel{def.}{=}|{a}_{{j}_{t}{j}_{t}}|-{R}_{{j}_{t}}\left(A\right)+{\delta }_{{j}_{t}}+\frac{1}{det\left[\mu \left(A\left(\beta \right)\right)\right]}detB.\end{array}$

Similarly to the proof of Lemma 4 in , we can prove that $detB\ge 0$. Thus, we obtain Inequation (1). Similarly, we can prove Inequation (2). □

Remark 1 Note that

$\frac{{P}_{{i}_{u}}\left(A\right)}{|{a}_{{i}_{u}{i}_{u}}|}\le r\le \frac{{R}_{{i}_{u}}\left(A\right)}{|{a}_{{i}_{u}{i}_{u}}|},\phantom{\rule{1em}{0ex}}1\le u\le k.$

This shows that Theorem 1 improves Theorem 1 of .

Theorem 2 Let $A=\left({a}_{ij}\right)\in {\mathbb{C}}^{n×n}$, $\beta =\left\{{i}_{1},{i}_{2},\dots ,{i}_{k}\right\}\subseteq {N}_{r}\left(A\right)\cap {N}_{c}\left(A\right)\ne \varphi$, $\overline{\beta }=\left\{{j}_{1},{j}_{2},\dots ,{j}_{l}\right\}$, $k, and let $A/\beta =\left({a}_{ts}^{\prime }\right)$. Then for all $1\le t\le l$, $0\le \alpha \le 1$,

$|{a}_{tt}^{\prime }|-{\left({R}_{t}\left(A/\beta \right)\right)}^{\alpha }{\left({C}_{t}\left(A/\beta \right)\right)}^{1-\alpha }\ge |{a}_{{j}_{t}{j}_{t}}|-{\left({R}_{{j}_{t}}\left(A\right)-{\delta }_{t}\right)}^{\alpha }{\left({C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}\right)}^{1-\alpha },$
(3)

and

$|{a}_{tt}^{\prime }|+{\left({R}_{t}\left(A/\beta \right)\right)}^{\alpha }{\left({C}_{t}\left(A/\beta \right)\right)}^{1-\alpha }\le |{a}_{{j}_{t}{j}_{t}}|+{\left({R}_{{j}_{t}}\left(A\right)-{\delta }_{t}\right)}^{\alpha }{\left({C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}\right)}^{1-\alpha },$
(4)

where

$\begin{array}{c}{\delta }_{t}=\underset{1\le \omega \le k}{min}\frac{|{a}_{{i}_{\omega }{i}_{\omega }}|-{P}_{{i}_{\omega }}\left(A\right)}{|{a}_{{i}_{\omega }{i}_{\omega }}|}\sum _{v=1}^{k}|{a}_{{j}_{t}{i}_{v}}|,\phantom{\rule{2em}{0ex}}\eta =\underset{1\le \omega \le k}{max}\frac{{\sum }_{v=1}^{l}|{a}_{{i}_{\omega }{j}_{v}}|}{|{a}_{{i}_{\omega }{i}_{\omega }}|-{\sum }_{v\ne \omega }^{k}|{a}_{{i}_{\omega }{i}_{v}}|},\hfill \\ {\delta }_{t}^{T}=\underset{1\le \omega \le k}{min}\frac{|{a}_{{i}_{\omega }{i}_{\omega }}|-{Q}_{{i}_{\omega }}\left(A\right)}{|{a}_{{i}_{\omega }{i}_{\omega }}|}\sum _{v=1}^{k}|{a}_{{i}_{v}{j}_{t}}|,\phantom{\rule{2em}{0ex}}\xi =\underset{1\le \omega \le k}{max}\frac{{\sum }_{v=1}^{l}|{a}_{{j}_{v}{i}_{\omega }}|}{|{a}_{{i}_{\omega }{i}_{\omega }}|-{\sum }_{v\ne \omega }^{k}|{a}_{{i}_{v}{i}_{\omega }}|},\hfill \\ {P}_{{i}_{\omega }}\left(A\right)=\eta \sum _{v\ne \omega }^{k}|{a}_{{i}_{\omega }{i}_{v}}|+\sum _{v=1}^{l}|{a}_{{i}_{\omega }{j}_{v}}|,\phantom{\rule{2em}{0ex}}{Q}_{{i}_{\omega }}\left(A\right)=\xi \sum _{v\ne \omega }^{k}|{a}_{{i}_{v}{i}_{\omega }}|+\sum _{v=1}^{l}|{a}_{{j}_{v}{i}_{\omega }}|.\hfill \end{array}$

Proof Since $\beta \subseteq {N}_{r}\left(A\right)\ne \varphi$, then $A\left(\beta \right)\in {H}_{k}$, $\mu \left(A\left(\beta \right)\right)\in {M}_{k}$. From Lemma 1 and Lemma 2, we have

${\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\ge |{\left[A\left(\beta \right)\right]}^{-1}|.$

Thus, for all $1\le t\le l$, $0\le \alpha \le 1$,

$\begin{array}{r}|{a}_{tt}^{\prime }|-{\left({R}_{t}\left(A/\beta \right)\right)}^{\alpha }{\left({C}_{t}\left(A/\beta \right)\right)}^{1-\alpha }\\ \phantom{\rule{1em}{0ex}}\ge |{a}_{{j}_{t}{j}_{t}}|-\left(|{a}_{{j}_{t}{i}_{1}}|,\dots ,|{a}_{{j}_{t}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{t}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{t}}|\end{array}\right)\\ \phantom{\rule{2em}{0ex}}{-\left(\sum _{s\ne t}^{l}\left[|{a}_{{j}_{t}{j}_{s}}|+\left(|{a}_{{j}_{t}{i}_{1}}|,\dots ,|{a}_{{j}_{t}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{s}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{s}}|\end{array}\right)\right]\right)}^{\alpha }\\ \phantom{\rule{2em}{0ex}}{×\left(\sum _{s\ne t}^{l}\left[|{a}_{{j}_{s}{j}_{t}}|+\left(|{a}_{{j}_{s}{i}_{1}}|,\dots ,|{a}_{{j}_{s}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{t}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{t}}|\end{array}\right)\right]\right)}^{1-\alpha }.\end{array}$

Let

$\zeta =\left(|{a}_{{j}_{t}{i}_{1}}|,\dots ,|{a}_{{j}_{t}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{t}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{t}}|\end{array}\right).$

By the proof of Theorem 1, we have

$\sum _{s\ne t}^{l}\left[|{a}_{{j}_{t}{j}_{s}}|+\left(|{a}_{{j}_{t}{i}_{1}}|,\dots ,|{a}_{{j}_{t}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{s}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{s}}|\end{array}\right)\right]\le {R}_{{j}_{t}}\left(A\right)-{\delta }_{t}-\zeta .$

Similarly,

$\sum _{s\ne t}^{l}\left[|{a}_{{j}_{s}{j}_{t}}|+\left(|{a}_{{j}_{s}{i}_{1}}|,\dots ,|{a}_{{j}_{s}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{t}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{t}}|\end{array}\right)\right]\le {C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}-\zeta .$

By Lemma 4, we have

$\begin{array}{r}|{a}_{tt}^{\prime }|-{\left({R}_{t}\left(A/\beta \right)\right)}^{\alpha }{\left({C}_{t}\left(A/\beta \right)\right)}^{1-\alpha }\\ \phantom{\rule{1em}{0ex}}\ge |{a}_{{j}_{t}{j}_{t}}|-\zeta -{\left({R}_{{j}_{t}}\left(A\right)-{\delta }_{t}-\zeta \right)}^{\alpha }{\left({C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}-\zeta \right)}^{1-\alpha }\\ \phantom{\rule{1em}{0ex}}\ge |{a}_{{j}_{t}{j}_{t}}|-\zeta -\left[{\left({R}_{{j}_{t}}\left(A\right)-{\delta }_{t}\right)}^{\alpha }{\left({C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}\right)}^{1-\alpha }-\zeta \right]\\ \phantom{\rule{1em}{0ex}}=|{a}_{{j}_{t}{j}_{t}}|-{\left({R}_{{j}_{t}}\left(A\right)-{\delta }_{t}\right)}^{\alpha }{\left({C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}\right)}^{1-\alpha }.\end{array}$

Thus, we obtain Inequation (3). Similarly, we can prove Inequation (4). □

Remark 2 Note that

$\begin{array}{c}{\delta }_{t}=\underset{1\le \omega \le k}{min}\frac{|{a}_{{i}_{\omega }{i}_{\omega }}|-{P}_{{i}_{\omega }}\left(A\right)}{|{a}_{{i}_{\omega }{i}_{\omega }}|}\sum _{v=1}^{k}|{a}_{{j}_{t}{i}_{v}}|\ge \underset{1\le \omega \le k}{min}\frac{|{a}_{{i}_{\omega }{i}_{\omega }}|-{R}_{{i}_{\omega }}\left(A\right)}{|{a}_{{i}_{\omega }{i}_{\omega }}|}\sum _{v=1}^{k}|{a}_{{j}_{t}{i}_{v}}|,\hfill \\ {\delta }_{t}^{T}=\underset{1\le \omega \le k}{min}\frac{|{a}_{{i}_{\omega }{i}_{\omega }}|-{Q}_{{i}_{\omega }}\left(A\right)}{|{a}_{{i}_{\omega }{i}_{\omega }}|}\sum _{v=1}^{k}|{a}_{{i}_{v}{j}_{t}}|\ge \underset{1\le \omega \le k}{min}\frac{|{a}_{{i}_{\omega }{i}_{\omega }}|-{C}_{{i}_{\omega }}\left(A\right)}{|{a}_{{i}_{\omega }{i}_{\omega }}|}\sum _{v=1}^{k}|{a}_{{i}_{v}{j}_{t}}|.\hfill \end{array}$

This shows that Theorem 2 improves Theorem 2 of .

Similarly to the proof of Theorem 2, we can prove the following theorem.

Theorem 3 Let $A=\left({a}_{ij}\right)\in {\mathbb{C}}^{n×n}$, $\beta =\left\{{i}_{1},{i}_{2},\dots ,{i}_{k}\right\}\subseteq {N}_{r}\left(A\right)\cap {N}_{c}\left(A\right)\ne \varphi$, $\overline{\beta }=\left\{{j}_{1},{j}_{2},\dots ,{j}_{l}\right\}$, $k and $A/\beta =\left({a}_{ts}^{\prime }\right)$. Then for all $1\le t\le l$, $0\le \alpha \le 1$,

$\begin{array}{r}|{a}_{tt}^{\prime }|-\alpha {R}_{t}\left(A/\beta \right)-\left(1-\alpha \right){C}_{t}\left(A/\beta \right)\\ \phantom{\rule{1em}{0ex}}\ge |{a}_{{j}_{t}{j}_{t}}|-\alpha {R}_{{j}_{t}}\left(A\right)-\left(1-\alpha \right){C}_{{j}_{t}}\left(A\right)+\alpha {\delta }_{t}+\left(1-\alpha \right){\delta }_{t}^{T}\\ \phantom{\rule{1em}{0ex}}\ge |{a}_{{j}_{t}{j}_{t}}|-\alpha {R}_{{j}_{t}}\left(A\right)-\left(1-\alpha \right){C}_{{j}_{t}}\left(A\right),\end{array}$

and

$\begin{array}{r}|{a}_{tt}^{\prime }|+\alpha {R}_{t}\left(A/\beta \right)+\left(1-\alpha \right){C}_{t}\left(A/\beta \right)\\ \phantom{\rule{1em}{0ex}}\le |{a}_{{j}_{t}{j}_{t}}|+\alpha {R}_{{j}_{t}}\left(A\right)+\left(1-\alpha \right){C}_{{j}_{t}}\left(A\right)-\alpha {\delta }_{t}-\left(1-\alpha \right){\delta }_{t}^{T}\\ \phantom{\rule{1em}{0ex}}\le |{a}_{{j}_{t}{j}_{t}}|+\alpha {R}_{{j}_{t}}\left(A\right)+\left(1-\alpha \right){C}_{{j}_{t}}\left(A\right).\end{array}$

## 3 Distribution for eigenvalues of the Schur complement

In this section, we present two new distribution theorems for eigenvalues of the Schur complement.

Lemma 5 

Let $A=\left({a}_{ij}\right)\in {\mathbb{C}}^{n×n}$ and $0\le \alpha \le 1$. Then for any eigenvalue λ of A, there exists $1\le t\le n$ such that

$|\lambda -{a}_{tt}|\le {\left({R}_{t}\left(A\right)\right)}^{\alpha }{\left({C}_{t}\left(A\right)\right)}^{1-\alpha }.$

Theorem 4 Let $A=\left({a}_{ij}\right)\in {\mathbb{C}}^{n×n}$, $\beta =\left\{{i}_{1},{i}_{2},\dots ,{i}_{k}\right\}\subseteq {N}_{r}\left(A\right)\ne \varphi$, $\overline{\beta }=\left\{{j}_{1},{j}_{2},\dots ,{j}_{l}\right\}$, and let $A/\beta =\left({a}_{ts}^{\prime }\right)$. Then for any eigenvalue λ of $A/\beta$, there exists $1\le t\le l$ such that

$|\lambda -{a}_{{j}_{t}{j}_{t}}|\le {R}_{{j}_{t}}\left(A\right)-{\delta }_{{j}_{t}}\le {R}_{{j}_{t}}\left(A\right).$
(5)

Proof Let λ be an eigenvalue of $A/\beta$. From the famous Gerschgorin circle theorem, we know that there exists $1\le t\le l$ such that $|\lambda -{a}_{tt}^{\prime }|\le {R}_{t}\left(A/\beta \right)$. Hence

$\begin{array}{rcl}0& \ge & |\lambda -{a}_{tt}^{\prime }|-{R}_{t}\left(A/\beta \right)\\ =& |\lambda -{a}_{{j}_{t}{j}_{t}}+\left({a}_{{j}_{t}{i}_{1}},\dots ,{a}_{{j}_{t}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{t}}\\ ⋮\\ {a}_{{i}_{k}{j}_{t}}\end{array}\right)|\\ -\sum _{s=1,\ne t}^{l}|{a}_{{j}_{t}{j}_{s}}-\left({a}_{{j}_{t}{i}_{1}},\dots ,{a}_{{j}_{t}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{s}}\\ ⋮\\ {a}_{{i}_{k}{j}_{s}}\end{array}\right)|\\ \ge & |\lambda -{a}_{{j}_{t}{j}_{t}}|-\sum _{s=1,\ne t}^{l}|{a}_{{j}_{t}{j}_{s}}|-\sum _{s=1}^{l}\left(|{a}_{{j}_{t}{i}_{1}}|,\dots ,|{a}_{{j}_{t}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{s}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{s}}|\end{array}\right)\\ =& |\lambda -{a}_{{j}_{t}{j}_{t}}|-{R}_{{j}_{t}}\left(A\right)+\sum _{v=1}^{k}|{a}_{{j}_{t}{i}_{v}}|+{\delta }_{{j}_{t}}-{\delta }_{{j}_{t}}\\ -\sum _{s=1}^{l}\left(|{a}_{{j}_{t}{i}_{1}}|,\dots ,|{a}_{{j}_{t}{i}_{k}}|\right){\left[\mu \left(A\left(\beta \right)\right)\right]}^{-1}\left(\begin{array}{c}|{a}_{{i}_{1}{j}_{s}}|\\ ⋮\\ |{a}_{{i}_{k}{j}_{s}}|\end{array}\right)\\ \ge & |\lambda -{a}_{{j}_{t}{j}_{t}}|-{R}_{{j}_{t}}\left(A\right)+{\delta }_{{j}_{t}},\end{array}$

that is

$|\lambda -{a}_{{j}_{t}{j}_{t}}|\le {R}_{{j}_{t}}\left(A\right)-{\delta }_{{j}_{t}}\le {R}_{{j}_{t}}\left(A\right).$

Thus, Inequation (5) holds. □

Theorem 5 Let $A=\left({a}_{ij}\right)\in {\mathbb{C}}^{n×n}$, $\beta =\left\{{i}_{1},{i}_{2},\dots ,{i}_{k}\right\}\subseteq {N}_{r}\left(A\right)\cap {N}_{c}\left(A\right)\ne \varphi$, $\overline{\beta }=\left\{{j}_{1},{j}_{2},\dots ,{j}_{l}\right\}$, $k, and let $A/\beta =\left({a}_{ts}^{\prime }\right)$. Then for any $0\le \alpha \le 1$ and every eigenvalue λ of $A/\beta$, there exists $1\le t\le l$ such that

$|\lambda -{a}_{{j}_{t}{j}_{t}}|\le {\left({R}_{{j}_{t}}\left(A\right)-{\delta }_{t}\right)}^{\alpha }{\left({C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}\right)}^{1-\alpha }\le {\left({R}_{{j}_{t}}\left(A\right)\right)}^{\alpha }{\left({C}_{{j}_{t}}\left(A\right)\right)}^{1-\alpha }.$
(6)

Proof Let λ be an eigenvalue of $A/\beta$. By Lemma 5, we know that there exists $1\le t\le l$ such that

$|\lambda -{a}_{tt}^{\prime }|\le {\left({R}_{t}\left(A/\beta \right)\right)}^{\alpha }{\left({C}_{t}\left(A/\beta \right)\right)}^{1-\alpha },\phantom{\rule{1em}{0ex}}0\le \alpha \le 1.$

Thus,

$\begin{array}{rcl}0& \ge & |\lambda -{a}_{tt}^{\prime }|-{\left({R}_{t}\left(A/\beta \right)\right)}^{\alpha }{\left({C}_{t}\left(A/\beta \right)\right)}^{1-\alpha }\\ =& |\lambda -{a}_{{j}_{t}{j}_{t}}-\left({a}_{{j}_{t}{i}_{1}},\dots ,{a}_{{j}_{t}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{t}}\\ ⋮\\ {a}_{{i}_{k}{j}_{t}}\end{array}\right)|\\ {-\left[\sum _{s=1,\ne t}^{l}|{a}_{{j}_{t}{j}_{s}}+\left({a}_{{j}_{t}{i}_{1}},\dots ,{a}_{{j}_{t}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{s}}\\ ⋮\\ {a}_{{i}_{k}{j}_{s}}\end{array}\right)|\right]}^{\alpha }\\ {×\left[\sum _{s=1,\ne t}^{l}|{a}_{{j}_{s}{j}_{t}}+\left({a}_{{j}_{s}{i}_{1}},\dots ,{a}_{{j}_{s}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{t}}\\ ⋮\\ {a}_{{i}_{k}{j}_{t}}\end{array}\right)|\right]}^{1-\alpha }\\ \ge & |\lambda -{a}_{{j}_{t}{j}_{t}}|-|\left({a}_{{j}_{t}{i}_{1}},\dots ,{a}_{{j}_{t}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{t}}\\ ⋮\\ {a}_{{i}_{k}{j}_{t}}\end{array}\right)|\\ {-\left(\sum _{s=1,\ne t}^{l}\left[|{a}_{{j}_{t}{j}_{s}}|+|\left({a}_{{j}_{t}{i}_{1}},\dots ,{a}_{{j}_{t}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{s}}\\ ⋮\\ {a}_{{i}_{k}{j}_{s}}\end{array}\right)|\right]\right)}^{\alpha }\\ {×\left(\sum _{s=1,\ne t}^{l}\left[|{a}_{{j}_{s}{j}_{t}}|+|\left({a}_{{j}_{s}{i}_{1}},\dots ,{a}_{{j}_{s}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{t}}\\ ⋮\\ {a}_{{i}_{k}{j}_{t}}\end{array}\right)|\right]\right)}^{1-\alpha }.\end{array}$

Similarly to the proof of Theorem 2, we can prove

$\begin{array}{r}|\left({a}_{{j}_{t}{i}_{1}},\dots ,{a}_{{j}_{t}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{t}}\\ ⋮\\ {a}_{{i}_{k}{j}_{t}}\end{array}\right)|\\ \phantom{\rule{2em}{0ex}}{+\left(\sum _{s=1,\ne t}^{l}\left[|{a}_{{j}_{t}{j}_{s}}|+|\left({a}_{{j}_{t}{i}_{1}},\dots ,{a}_{{j}_{t}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{s}}\\ ⋮\\ {a}_{{i}_{k}{j}_{s}}\end{array}\right)|\right]\right)}^{\alpha }\\ \phantom{\rule{2em}{0ex}}{×\left(\sum _{s=1,\ne t}^{l}\left[|{a}_{{j}_{s}{j}_{t}}|+|\left({a}_{{j}_{s}{i}_{1}},\dots ,{a}_{{j}_{s}{i}_{k}}\right){\left[A\left(\beta \right)\right]}^{-1}\left(\begin{array}{c}{a}_{{i}_{1}{j}_{t}}\\ ⋮\\ {a}_{{i}_{k}{j}_{t}}\end{array}\right)|\right]\right)}^{1-\alpha }\\ \phantom{\rule{1em}{0ex}}\le {\left({R}_{{j}_{t}}\left(A\right)-{\delta }_{t}\right)}^{\alpha }{\left({C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}\right)}^{1-\alpha }.\end{array}$

Therefore, we have

$0\ge |\lambda -{a}_{tt}^{\prime }|-{\left({R}_{t}\left(A/\beta \right)\right)}^{\alpha }{\left({C}_{t}\left(A/\beta \right)\right)}^{1-\alpha }\ge |\lambda -{a}_{{j}_{t}{j}_{t}}|-\left({R}_{{j}_{t}}\left(A\right)-{\delta }_{t}\right){\left({C}_{{j}_{t}}\left(A\right)-{\delta }_{t}^{T}\right)}^{1-\alpha }.$

That is, Inequation (6) holds. □

## 4 A numerical example

In this section, we present a numerical example to illustrate the theory results.

Example 1 Let

$A=\left(\begin{array}{ccccc}16& 1& 5& 2& 2\\ 3& 15& 2& 4& 3\\ 2& 2& 18& 1& 4\\ 5& 3& 5& 8& 2\\ 5& 2& 2& 3& 9\end{array}\right),\phantom{\rule{2em}{0ex}}\beta =\left\{1,3\right\}.$

By calculation with Matlab 7.1, we have that

$\begin{array}{c}{R}_{1}\left(A\right)=10;\phantom{\rule{2em}{0ex}}{R}_{2}\left(A\right)=12;\phantom{\rule{2em}{0ex}}{R}_{3}\left(A\right)=9;\phantom{\rule{2em}{0ex}}{R}_{4}\left(A\right)=15;\phantom{\rule{2em}{0ex}}{R}_{5}\left(A\right)=12;\hfill \\ {C}_{1}\left(A\right)=15;\phantom{\rule{2em}{0ex}}{C}_{2}\left(A\right)=8;\phantom{\rule{2em}{0ex}}{C}_{3}\left(A\right)=14;\phantom{\rule{2em}{0ex}}{C}_{4}\left(A\right)=10;\phantom{\rule{2em}{0ex}}{C}_{5}\left(A\right)=11;\hfill \\ {\delta }_{2}=2.7273;\phantom{\rule{2em}{0ex}}{\delta }_{4}=5.4545;\phantom{\rule{2em}{0ex}}{\delta }_{5}=3.8182;\hfill \\ {\delta }_{2}^{T}=0.2143;\phantom{\rule{2em}{0ex}}{\delta }_{4}^{T}=0.2143;\phantom{\rule{2em}{0ex}}{\delta }_{5}^{T}=0.4286.\hfill \end{array}$

Since $\beta \in {N}_{r}\left(A\right)$, by Theorem 4, any eigenvalue λ of $A/\beta$ satisfies

$\lambda \in \left\{\lambda :|\lambda -15|\le 9.2727\right\}\cup \left\{\lambda :|\lambda -8|\le 9.5455\right\}\cup \left\{\lambda :|\lambda -9|\le 8.1818\right\}={\mathrm{\Gamma }}_{1}.$

From Theorem 3 of , any eigenvalue λ of $A/\beta$ satisfies

$\lambda \in \left\{\lambda :|\lambda -15|\le 10.1250\right\}\cup \left\{\lambda :|\lambda -8|\le 11.2500\right\}\cup \left\{\lambda :|\lambda -9|\le 9.3750\right\}={\mathrm{\Gamma }}_{1}^{\prime }.$

Evidently, ${\mathrm{\Gamma }}_{1}\subset {\mathrm{\Gamma }}_{1}^{\prime }$, we use Figure 1 to show the fact.

Meanwhile, since $\beta \in {N}_{r}\left(A\right)\cap {N}_{c}\left(A\right)$, by taking $\alpha =0.5$ in Theorem 5, any eigenvalue λ of $A/\beta$ satisfies

$\lambda \in \left\{\lambda :|\lambda -15|\le 8.4968\right\}\cup \left\{\lambda :|\lambda -8|\le 9.6648\right\}\cup \left\{\lambda :|\lambda -9|\le 9.3002\right\}={\mathrm{\Gamma }}_{2}.$

From Theorem 4 of , any eigenvalue λ of $A/\beta$ satisfies

$\lambda \in \left\{\lambda :|\lambda -15|\le 8.8939\right\}\cup \left\{\lambda :|\lambda -8|\le 10.5067\right\}\cup \left\{\lambda :|\lambda -9|\le 9.9804\right\}={\mathrm{\Gamma }}_{2}^{\prime }.$

Evidently, ${\mathrm{\Gamma }}_{2}\subset {\mathrm{\Gamma }}_{2}^{\prime }$, we use Figure 2 to show the fact.

## References

1. Cvetković LJ: A new subclass of H -matrices. Appl. Math. Comput. 2009, 208: 206–210. 10.1016/j.amc.2008.11.037

2. Liu JZ, Huang ZJ: The Schur complements of γ -diagonally and product γ -diagonally dominant matrix and their disc separation. Linear Algebra Appl. 2010, 432: 1090–1104. 10.1016/j.laa.2009.10.021

3. Liu JZ, Li JC, Huang ZH, Kong X: Some properties on Schur complement and diagonal Schur complement of some diagonally dominant matrices. Linear Algebra Appl. 2008, 428: 1009–1030. 10.1016/j.laa.2007.09.008

4. Carlson D, Markham T: Schur complements on diagonally dominant matrices. Czechoslov. Math. J. 1979, 29: 246–251.

5. Horn RA, Johnson CR: Topics in Matrix Analysis. Cambridge University Press, New York; 1991.

6. Ikramov KD: Invariance of the Brauer diagonal dominance in Gaussian elimination. Vestn. Mosk. Univ., Ser. 15 Vyčisl. Mat. Kibern. 1989, 2: 91–94.

7. Smith R: Some interlacing properties of the Schur complement of a Hermitian matrix. Linear Algebra Appl. 1992, 177: 137–144.

8. Zhang FZ: The Schur Complement and Its Applications. Springer, New York; 2005.

9. Li B, Tsatsomeros M: Doubly diagonally dominant matrices. Linear Algebra Appl. 1997, 261: 221–235. 10.1016/S0024-3795(96)00406-5

10. Liu JZ, Huang YQ: The Schur complements of generalized doubly diagonally dominant matrices. Linear Algebra Appl. 2004, 378: 231–244.

11. Liu JZ, Huang YQ: Some properties on Schur complements of H -matrices and diagonally dominant matrices. Linear Algebra Appl. 2004, 389: 365–380.

12. Liu JZ, Huang ZJ: The dominant degree and disc theorem for the Schur complement. Appl. Math. Comput. 2010, 215: 4055–4066. 10.1016/j.amc.2009.12.063

13. Liu JZ, Zhang FZ: Disc separation of the Schur complements of diagonally dominant matrices and determinantal bounds. SIAM J. Matrix Anal. Appl. 2005, 27: 665–674. 10.1137/040620369

14. Liu JZ: Some inequalities for singular values and eigenvalues of generalized Schur complements of products of matrices. Linear Algebra Appl. 1999, 286: 209–221. 10.1016/S0024-3795(98)10156-8

## Acknowledgements

The authors are very indebted to the referees for their valuable comments and corrections, which improved the original manuscript of this paper. This work was supported by the National Natural Science Foundation of China (71161020, 11361074) and IRTSTYN.

## Author information

Authors

### Corresponding author

Correspondence to Yao-tang Li.

### Competing interests

The authors declare that they have no competing interests.

### Authors’ contributions

All authors read and approved the final manuscript.

## Authors’ original submitted files for images

Below are the links to the authors’ original submitted files for images.

## Rights and permissions

Reprints and Permissions

Li, Yt., Wang, F. The estimates of diagonally dominant degree and eigenvalues distributions for the Schur complements of matrices. J Inequal Appl 2013, 431 (2013). https://doi.org/10.1186/1029-242X-2013-431

• Accepted:

• Published:

• DOI: https://doi.org/10.1186/1029-242X-2013-431

### Keywords

• matrix
• Schur complement
• diagonally dominant degree
• eigenvalue distribution 