Volume 2011, Issue 1 713795
Research Article
Open Access

Convergence of GAOR Iterative Method with Strictly α Diagonally Dominant Matrices

Guangbin Wang

Corresponding Author

Guangbin Wang

Department of Mathematics, Qingdao University of Science and Technology, Qingdao 266061, China qust.edu.cn

Search for more papers by this author
Hao Wen

Hao Wen

Department of Mathematics, Qingdao University of Science and Technology, Qingdao 266061, China qust.edu.cn

Search for more papers by this author
Ting Wang

Ting Wang

Department of Mathematics, Qingdao University of Science and Technology, Qingdao 266061, China qust.edu.cn

Search for more papers by this author
First published: 15 November 2011
Citations: 1
Academic Editor: Yongkun Li

Abstract

We discuss the convergence of GAOR method for linear systems with strictly α diagonally dominant matrices. Moreover, we show that our results are better than ones of Darvishi and Hessari (2006), Tian et al. (2008) by using three numerical examples.

1. Introduction

Sometimes we have to solve the following linear system:
()
where
()
is invertible. Yuan proposed a generalized SOR(GSOR) method to solve linear system (1.1) in [1]; afterwards, Yuan and Jin [2] established a generalized AOR(GAOR) method to solve linear system (1.1) as follows:
()
where
()

In [24], some people studied the convergence of GAOR method for solving linear systems Hy = f. In [2], Yuan and Jin studied the convergence of GAOR method and show that the GAOR method is better than the GSOR method under certain conditions. In [3], Darvishi and Hessari studied the convergence of GAOR method for diagonally dominant coefficient matrices and gave the regions of convergence. In [4], Tian et al. studied the convergence of GAOR method for strictly diagonally dominant coefficient matrices and gave the regions of convergence.

Sometimes, the coefficient matrices of linear systems  Hy = f  are not strictly diagonally dominant. In this paper, we will discuss the convergence of GAOR method for linear systems Hy = f with strictly α diagonally dominant matrices which need not to be strictly diagonally dominant.

Throughout this paper, we denote the set of all complex matrices by Cn×n, the spectral radius of iterative matrix Lω,r by ρ(Lω,r). And
()

Definition 1.1 (see [5].)Let  A = (aij) ∈ Cn×n. If there exists α ∈ [0,1],

()
holds, then we call A as α  diagonally dominant and denote it as  AD0(α). If all the inequalities are strict, then we denote it as  AD(α).

Obviously, if  A  is a strictly diagonally dominant matrix (ASD), then AD(α). But not vice versa.

In [3], Darvishi and Hessari obtained the following result.

Theorem 1.2. Let HSD  and assume that ωr ≥ 0. Then the sufficient condition for convergence of the GAOR method is

()
where Ji and  Ki are the i-row sums of the modulus of the entries of J and K, respectively.

In [4], Tian et al. obtained the following result.

Theorem 1.3. If HSD, then the sufficient conditions for the convergence of the GAOR method are either

  • (i)

    0 ≤ r ≤ 1 and 0 < ω ≤ 1 or

  • (ii)

    |r | ≤ min i{(1 − Ji)/Ki} and 0 < ω < 2/(1 + max i(J + rK) i).

This work is organized as follows. In Section 2, we obtain bound for the spectral radius of the iterative matrix Lω,r of GAOR iterative method. In Section 3, we present two convergence theorems of GAOR method. In Section 4, we present three numerical examples to show that our results are better than ones of Theorems 1.2 and 1.3.

2. Upper Bound of the Spectral Radius of Lω,r

In this section, we obtain upper bound of the spectral radius of iterative matrix Lω,r.

Theorem 2.1. Let HD(α), then  ρ(Lω,r)  satisfies the following inequality:

()

Proof. Let λ  be an arbitrary eigenvalue of iterative matrix Lω,r, then

()
Equation (2.2) holds if and only if
()
If (λ + ω − 1)IωJωrKD(α), that is,
()
where  (ωJ+ωrK)ii  denotes the diagonal element of matrix  ωJ + ωrK. From [5], we know that
()
then  λ is not an eigenvalue of iterative matrix  Lω,r, and when
()
λ  is not an eigenvalue of iterative matrix  Lω,r.

When λ is an eigenvalue of iterative matrix Lω,r, then there exists at least one i  (iN), such that

()
that is,
()
Hence,
()

3. Convergence of GAOR Method

In this section, we investigate the convergence of GAOR method to solve linear system (1.1). We assume that H  is a strictly  α  diagonally dominant coefficient matrix and get some sufficient conditions for the convergence of GAOR method.

Theorem 3.1. Let HD(α), then GAOR method is convergent if ω, r satisfy either

  • (I)

    0 < ω ≤ 1 and

    ()

or
  • (II)

    and

    ()

Proof. Since HD(α), then GAOR method is convergent if we have  ρ(Lω,r) < 1, that is,

()

Firstly, when 0 < ω ≤ 1, then

()
That is,
()
which leads to
()
So
()

Secondly, when 1 < ω, then

()
That is,
()
which leads to
()
So
()
then
()
so
()
Thus we complete the proof.

When r = 0, we can get the following theorem.

Theorem 3.2. Let HD(α), then ρ(Lω,0) < 1 if

()

Proof. Because HD(α), from Theorem 1.2, when r = 0, we have

()
Firstly, when 0 < ω ≤ 1,
()
That is
()
Secondly, when 1 < ω < 2,
()
That is
()
which leads to
()
So  ω  must satisfy
()
Thus we complete the proof of Theorem 3.2.

4. Examples

In the following examples, we give the regions of convergence of GAOR method to show that our results are better than ones of Theorems 1.2 and 1.3.

Example 4.1. Let

()

Obviously, H is not a strictly diagonally dominant matrix, so we cannot use the results of Theorems 1.2 and 1.3. But HD(1/2), and

()
By Theorem 3.1, we obtain the following regions of convergence:
  • (I)

    0 < ω ≤ 1 and |r| < 1/10

or
  • (II)

    1 < ω < 24/23 and |r| < (2 − (23/12)ω)(6/5ω).

Example 4.2 (Example 1 of paper [4]). Let

()
It is easy to test that HD(1/4), and
()
By Theorem 3.1, we obtain the following regions of convergence:
  • (I)

    0 < ω ≤ 1 and |r| < 3/4

or
  • (II)

    1 < ω < 1.2 and |r| < ((18 − 5ω)/4ω).

In addition, H is a strictly diagonally dominant matrix.

By Theorem 1.3, we obtain the following regions of convergence:

  • (I)

    0 ≤ r ≤ 1 and 0 < ω ≤ 1,

  • (II)

    0 ≤ r < 0.3  and 0 < ω < 1.2,

or
  • (III)

    −0.3 < r < 0 and 0 < ω < (18/(15 − 10r)).

By Theorem 1.2, we obtain the following region of convergence:

0 ≤ rω and  0 < ω < (18/(15 + 10r)).

And we get Figure 1, where the regions of convergence got by Theorems 3.1 and 1.3, Theorem 1.2 are bounded by blue lines, red lines, green lines, respectively.

Description unavailable

This figure shows that the regions of convergence got by Theorem 3.1 are larger than ones got by Theorems 1.2 and 1.3.

Example 4.3. Consider

()
where
()

Obviously, H is not a strictly diagonally dominant matrix, so we cannot use the results of Theorems 1.2 and 1.3. But HD(1/2), and

()
By Theorem 3.1, we obtain the following regions of convergence:
  • (I)

    0 < ω ≤ 1 and |r| < (2208/3481)

or
  • (II)

    1 < ω < (192/169) and |r| < ((18432 − 16224ω)/3481ω).

Acknowledgments

The authors would like to thank the referees for their valuable comments and suggestions, which greatly improved the original version of this paper. This work was supported by the National Natural Science Foundation of China (Grant no. 11001144) and the Science and Technology Program of Shandong Universities of China (J10LA06).

      The full text of this article hosted at iucr.org is unavailable due to technical difficulties.