Chapter 19
Linear Algebraic Solvers and Eigenvalue Analysis
Henk A. van der Vorst
Utrecht University, Utrecht, The Netherlands
1 Introduction 551
2 Mathematical Preliminaries 551
3 Direct Methods for Linear Systems 553
4 Preconditioning 560
5 Incomplete LU Factorizations 562
6 Methods for the Complete Eigenproblem 567
7 Iterative Methods for the Eigenproblem om
Notes 374
References 315
1 INTRODUCTION
In this chapter, an overview of the most widely used
numerical methods for the solution of Linear systems of
‘equations and for eigenproblems is presented.
For linear systems Ax = b, with A a real square non-
singular n x n matrix, direct solution methods and iterative
methods are discussed, The direct methods are variations
‘on Gaussian elimination, The iterative methods are the s0-
called Krylov projection~type methods, and they include
popular methods such as Conjugate Gradients, MINRES,
Bi-Conjugate Gradients, QMR, BE-CGSTAB, and GMRES,
Iterative methods are often used in combination with
the so-called preconditioning operators (easily invertible
approximations for the operator ofthe system to be solved),
We will give a brief overview of the various preconditioners
that exist,
For the eigenproblems of the type Ax = Xx, the QR
method, which is often considered to be a direct method
Encyclopedia of Computational Mechanics, Edited by Erwin
Stein, René de Borst and Thomas J.R. Hughes. Volume I: Funda-
‘mentals. € 2004 John Wiley & Sons, Ltd. ISBN: 0-470-84609-0,
because of its very fast convergence, is discussed, Strictly,
speaking, there are no direct methods for the eigenprob-
Jem; all methods are necessarily iterative. The QR method
is expensive for larger values of n, and for these larger
values, a number of iterative methods, including the Lanc-
zo8 method, Arnoldi’s method, and the Jacobi~Davidson
method are presented,
For a general background on linear algebra for numerical
applications, see Golub and Van Loan (1996) and Stewart
(1998). Modem iterative methods for linear systems are
discussed in van der Vorst (2003). A basic introduction
with simple software 1s presented In Barett er al, (1994),
‘A complete overview of algorithms for elgenproblems,
including pointers to software, is given in Bai et al. (2000),
Implementation aspects for high-performance computers
are discussed in detail in Dongarra et al, (1998).
Some useful state-of-the-art papers have appeared; we
‘mention papers on the history of iterative methods by Golub,
and van der Vorst (2000) and Saad and van der Vorst
(2000), An overview on parallelizable aspects of sparse
‘matrix techniques is presented in Duff and van der Vorst
(1999), A state-of-the-art overview for preconditioners is
‘presented in Benzi (2002).
‘The purpose of this chapter isto make the reader familiar
with the ideas and the usage of iterative methods, We expect
that guided with sufficient knowledge about the background
of iterative methods, one will be able to make a proper
choice for a particular class of problems. It will also provide
guidance on how to tine these methods, in perticular, for
the selection or construction of effective preconditioners,
2. MATHEMATICAL PRELIMINARIES
In this section, some basic notions and notations on linear
systems and eigenproblems have been collected.552 _Linear Algebraic Solvers and Eigenvalue Analysis,
2.1 Matrices and vectors
‘We will be concerned with linear systems Ax = b, where
Ais usually ann x m matrix:
AeR™
‘The elements of A will be denoted as a, ). The vectors
= (Xp, -+4,%,)7 and b belong to the linear space R*.
Sometimes we will admit complex matrices A € C™** and
vectors x, b € C%, but that will be explicitly mentioned
Over the space R", we will use the Euclidean inner
product between two vectors x and y:
and for v,w €C%, we use the standard complex inner
product
2
‘These inner products lead to the 2-norm or Euclidean length
of a vector
Wl, = Vee forx eR
lvl, = VOR for vec
With these norms, we can associate a 2-norm for matrices:
for A € R"™*", its associated 2-norm [|All is defined as
yl
vento Nylh
Ala =
and in the complex case, similarly, using the complex
inner product. This matrix norm gives the maximal length
‘multiplication effect of A on a vector (where the length is
defined by the given norm).
‘The associated matrix norms are convenient because they
can be used to bound products. For Ae R™#, BERE™,
wwe have that
NABI; $ IAI BU,
in particular,
Daz < lAllallzlh,
‘The inverse of a nonsingular matrix A is denoted as A~!.
Particularly useful is the condition number of a square
nonsingular matrix A defined as
(A) = [AlAs
‘The condition number is used to characterize the sensitivity
of the solution x of Ax = b with respect to perturbations
in b and A. For perturbed systems, we have the following
theorem,
‘Theorem 1 (Golub and Van Loan, 1996; Th. 2.7.2)
Suppose
Ar=b, AERP", 0¢bER"
(A+ DAy= b+ Ab, AAER™, AbeR™
with Ally = ¢|IAy and [LAbIy < elbla.
Ifee(A) = 1 <1, then A+ A is nonsingular and
iy =a
lh
With the superscript 7, we denote the transpose of a
matrix (or vector): for Ae R™*, the matix B= AT €
RP" js defined by
If E <0", then the superscript # is used to denote its
complex conjugate F = EY, defined as
fi
Sometimes, the superscript 7 is used for complex matrices
in order to denote the transpose of a complex matrix.
‘The matrix A is symmetric if A = AT, and BEC is
Hermitian if B = BS, Hermitian matrices have the attrac-
live property that their spectrum is real. In particular,
Hermitian (or symmettic real) matrices that are positive-
definite are attractive because they can be solved rather
easily by proper iterative methods (the CG method).
A Hermitian matrix Ae C™" is positive-definite if
attr > 0 for all 0 2x. C*. A positive-definite Hermi-
tian matrix has only positive real eigenvalues,
‘We will encounter some special matrix forms, in parti
ular tridiagonal matrices and (upper) Hessenberg matzices.
The matrix T = (¢,;) €R"™" will be called ridiagonal,
if all elements for which |i ~ j| > 1 are zero. It is called
upper Hessenberg if all elements for which i > j-+1 are
zero. In the context of Krylov subspaces, these matrices are
often (k-+ 1) x & and they will then be denoted a8 Ty.
2.2. Higenvalues and eigenvectors
For purposes of analysis, it is often helpfl or instructive
to transform a given matrix to an easier form, for instance,
diagonal oF upper triangular form.‘The easiest situation is the symmetric case: for a real
symmetric matrix, there exists an orthogonal mattix Q <
R™*, so that QTAQ = D, where D ¢ BY" is a diagonal
‘matrix. The diagonal elements of D are the eigenvalues of
‘A, and the columns of Q are the corresponding eigenvectors
of A. Note that the eigenvalues and eigenvectors of A are
all real.
IFA € C™*" is Hermitian (A = AY), then there exist Q €
C™" and a diagonal matrix D €R™, so that Q#Q = 1
and Q#AQ = D. This means that the eigenvalues of a
Hermitian matrix are all real, but its eigenvectors may be
‘complex.
‘Unsymmetric matrices do not, in general, have an ortho-
normal set of eigenvectors and may not have a complete
set of eigenvectors, but they can be transformed unitarily
to Schur form:
QAQ=R
in which R is upper triangular.
If the matrix A is complex, then the mattices Q and R
may be complex as well. However, they may be complex
even when A is real unsymmetric. It may then be advan-
tageous to work in real arithmetic. This can be realized
because of the existence of the real Schur decomposition. If
A €R™*, then it can be transformed with an orthonormal
QeR™ as
oAg=
with
Rs
> R,
R= erm
0 0 Ruy
Bach R,, is either 1x 1 or a 2 x 2 (real) matrix having
complex conjugate eigenvalues. For a proof of this, see
Golub and Van Loan (1996, Chapter 7.4.1). This form of
referred to as an upper quasi-triangular matrix.
If all eigenvalues are distinct, then there exists a nonsin-
gular matrix X Gin general not orthogonal) that transforms
A to diagonal form:
XTAX=D
‘A general matrix can be transformed to Jordan form with
a nonsingular X:
XTMAX = diag, doy. JD)
Linear Algebraic Solvers and Eigenvalue Analysis _553
where
ed
Oo. 0%
Te there isa J, with dimension greater than 1, then the
matrix A is defective. In this case, A does not have a
complete set of independent eigenvectors. In numerical
computations, one may argue that small perturbations lead
to different eigenvalues, and hence that it will be unlikely
that A has a true Jordan form in actual computation.
However, if A is close to a mateix with a nontrivial Jordan
block, then this is reflected by a (severely) ill-conditioned
eigenvector matrix X.
We will also encounter eigenvalues that are called Ritz
values. For simplicity, we will introduce them here for
the real case. The subspace methods that are collected in
this chapter are based on the approach to identify good
solutions from certain low-dimensional subspaces V* CR",
where k