You are on page 1of 18

Chapter 2

Iterative Methods
2.1 Introduction
In this section, we will consider three dierent iterative methods for solving a sets of equations.
First, we consider a series of examples to illustrate iterative methods.
To construct an iterative method, we try and re-
arrange the system of equations such that we gen-
erate a sequence.
2.1.1 Simple Iteration Example
Example 2.1.1: Let us consider the equation
f(x) = x + e
x
2 = 0 . (2.1)
When solving an equation such as (2.1) for y=2x
y=e
x

2
1
2
where f() = 0, 0 < < 2, we can generate a sequence {x
(k)
}

k=0
from some initial value (guess)
x
(0)
by re-writing the equation as
x = 2 e
x
,
i.e. by computing x
(k+1)
= 2 e
x
(k)
from some x
(0)
. If the series converges, it will converge to the
solution. For example, let us consider x
(0)
= 1 and x
(0)
= 1:
21
k x
(k)
x
(k)
0 1.0 -1.0
1 1.63212 -0.71828
2 1.80449 -0.05091
3 1.83544 0.947776
4 1.84046 1.61240
5 1.84126 1.80059
6 1.84138 1.83480
7 1.84140 1.84124
8 1.84141 1.84138
9 . . . . . .
In this example, both sequences appear to converge to a value close to the root = 1.84141 where
0 < < 2. Hence, we have constructed a simple algorithm for solving an equation and it appears
to be a robust iterative method.
However, (2.1) has two solutions: a positive root at 1.84141 and a negative root at -1.14619. Why
do we only nd one root?
If f (x) = 0 has a solution x = then x
(k+1)
= g(x
(k)
) will converge to , provided |g

()| < 1
and x
(0)
is suitably chosen.
The condition |g

()| < 1 is a necessary condition.


In the above example,
g(x) = 2 e
x
and g

(x) = e
x
,
and
|g

(x)| < 1 if x > 0 .


So this method can be used to nd the positive root of (2.1). However, it will never converge to the
negative root. Hence, this kind of approach will not always converge to a solution.
2.1.2 Linear Systems
Let us adopt the same approach for a linear system.
Example 2.1.2:
22
Consider the following set of linear equations:
10x
1
+ x
2
= 12
x
1
+ 10x
2
= 21
Let us re-write these equations as
x
1
= (12 x
2
)/10
x
2
= (21 x
1
)/10 .
Thus, we can use the following:
x
(k+1)
1
= 1.2 x
(k)
2
/10
x
(k+1)
2
= 2.1 x
(k)
1
/10 ,
to generate a sequence of vectors x
(k)
= (x
(k)
1
, x
(k)
2
)
T
from some starting vector, x
(0)
.
If
x
(0)
=
_
0
0
_
,
then
x
(0)
=
_
0
0
_
, x
(1)
=
_
1.2
2.1
_
, x
(2)
=
_
0.99
1.98
_
, x
(3)
=
_
1.002
2.001
_
, . . .
where
x
(k)

_
1
2
_
as k ,
which is indeed the correct answer. So we have generated a convergent sequence.
Let us consider the above set of linear equations again. Possibly the more obvious rearrangement
was
x
1
= 21 10x
2
x
2
= 12 10x
1
.
Thus, we can generate a sequence using:
x
(k+1)
1
= 21 10x
(k)
2
x
(k+1)
2
= 12 10x
(k)
1
,
If, we again use
x
(0)
=
_
0
0
_
,
then
x
(0)
=
_
0
0
_
, x
(1)
=
_
21
12
_
, x
(2)
=
_
99
198
_
, x
(3)
=
_
1011
1992
_
, . . .
23
Clearly, this sequence is not converging! Why?
Example 2.1.3:
Let us consider the above example (2.1.2) again. Can we nd a method that allows the system to
converge more quickly?
Let us look at the computation more carefully. In the rst step x
(1)
1
is computed from x
(0)
2
and in
the second step we compute x
(1)
2
from x
(0)
1
.
It seems more natural, from a computational point of view, to use x
(1)
1
rather then x
(0)
1
in the second
step. i.e. to use the latest available value. In eect, we want to compute the following:
x
(k+1)
1
= 1.2 x
(k)
2
/10
x
(k+1)
2
= 2.1 x
(k+1)
1
/10 ,
which gives,
x
(0)
=
_
0
0
_
, x
(1)
=
_
1.2
1.98
_
, x
(2)
=
_
1.002
1.9998
_

_
1
2
_
,
which converges to
_
1
2
_
much more rapidly!
In the following sections, we will consider, in general terms, iterative methods for solving a system
Ax = b. First, though we introduce some important results about a sequence of vectors
2.2 Sequences of Vectors
2.2.1 The Limit of a Sequence
Let
_
x
(k)
_

k=0
be a sequence in a Vector Space V . How do we know if this sequence has a limit?
First observe that x = y x = y. i.e. two distinct objects in a Vector Space can have the
same size. However, from rule 1 for norms (1.1) we know that if x y = 0, then x y.
So if
lim
k
x
(k)
x = 0
then
lim
k
x
(k)
= x
The vector x is the limit of the sequence.
24
2.2.2 Convergence of a Sequence
Suppose the sequence {x
(k)
}

k=0
converges to x, where
x
(k+1)
= Bx
(k)
+c .
If x
(k)
x for k , then x satises the equation:
x = Bx +c ,
and so we have
x
(k+1)
x = B(x
(k)
x) ,
and thus, taking norms,
x
(k+1)
x B x
(k)
x .
If B < 1 then
x
(k+1)
x < x
(k)
x ,
i.e. we have a monotonically decreasing sequence, or, in other words, the error in the approximations
decreases.
Say we start from an initial guess x
(1)
x = B(x
(0)
x). Then
x
(2)
x = B(x
(1)
x)
= B
_
B(x
(0)
x)
_
= B
2
(x
(0)
x) ,
and so on, to give
x
(k)
x = B
k
(x
(0)
x) .
Taking norms, and using rule 5 (1.9) for sub-ordinate matrix norms
x
(k)
x B
k
(x
(0)
x) B
k1
B(x
(0)
x) B
k2
B
2
(x
(0)
x) B
k
(x
(0)
x) .
If B < 1, then B
k
0 as k and hence, x
(k)
x as k .
Recall that (B) B (1.5) so a necessary condition for convergence is (B) < 1. Furthermore,
it is possible to show that if
(B) < 1 , then B < 1 .
and if
(B) > 1 , then B > 1 ,
although we do not prove these results in this course.
Hence, (B) < 1 is not only a necessary condition, but it is also sucient condition.
25
2.2.3 Spectral radius and rate of convergence
In numerical analysis, to compare dierent methods for solving systems of equations we are interested
in determining the rate of convergence of the method. As we will see below the spectral radius is a
measure of the rate of convergence.
Consider the situation where B
NN
has N linearly independent eigenvectors. As before we have
x
(k+1)
x = B(x
(k)
x) ,
or substituting in for v
(k)
= x
(k)
x, we have
v
(k+1)
= Bv
(k)
.
Now write v
(0)
=

N
i=1

i
e
i
where e
i
are the eigenvectors (with associated eigenvalues
i
) of B,
then
v
(1)
= B
_
N

i=1

i
e
i
_
=
N

i=1

i
Be
i
=
N

i=1

i
e
i
,
v
(2)
= B
_
N

i=1

i
e
i
_
=
N

i=1

i
Be
i
=
N

i=1

2
i
e
i
,
continuing this sequence gives
v
(k)
=
N

i=1

k
i
e
i
.
Now suppose |
1
| > |
i
| (i = 2, . . . , N), then
v
(k)
=
1

k
1
e
1
+
N

i=2

k
i
e
i
=
k
1
_

1
e
1
+
N

i=2

i
_

1
_
k
e
i
_
.
Given that
i
/
1
< 1, for large k,
v
(k)

1

k
1
e
1
.
Hence, the error associated with x
(k)
, the kth vector in the sequence, is given by v
(k)
which varies
as the kth power of the largest eigenvalue. In other words, it varies as the kth power of the spectral
radius (B) (= |
1
|). So the spectral radius is a good indication of the rate of convergence.
2.2.4 Gerschgorins Theorem
The above result means that if we know the magnitude of the largest vector of the iteration matrix we
can estimate the rate of convergence of a system of equations for a particular method. However, this
26
requires the magnitudes of all eigenvalues to be known, which would probably have to be determined
numerically.
The Gerschgorin Theorem is a surprisingly simple result concerning eigenvalues that allows us to put
bounds on the size of the eigenvalues of a matrix without actually nding the eigenvalues themselves.
The equation Ae = e, where (, e) are an eigenvalue, eigenvector pair of the matrix A, can be
written in component notation as
N

j=1
a
ij
e
j
= a
ii
e
i
+
N

j=1
j=i
a
ij
e
j
= e
i
.
Rearranging implies
e
i
(a
ii
) =
N

j=1
j=i
a
ij
e
j
,
and thus,
|e
i
| |a
ii
|
N

j=1
j=i
|a
ij
| |e
j
| .
Suppose the component of eigenvector e with the largest absolute value is |e
l
|, such that |e
l
| |e
j
|
for j (note e
j
= 0 for all j). Then from above
|e
l
| |a
ll
|
N

j=1
j=l
|a
lj
| |e
j
|
N

j=1
j=l
|a
lj
| |e
l
|
so,dividing by |e
l
| gives
|a
ll
|
N

j=1
j=l
|a
lj
| .
Each eigenvalue lies inside a circle with centre a
ll
and radius

N
j=1
|a
lj
| with j = l.
However, we dont know l without nding and e.
But we can say that the union of all such circles must contain all the eigenvalues. This is
Gerschgorins Theorem.
Example 2.5.1: Determine the bounds on the eigenvalues for the matrix
A =
_

_
2 1 0 0
1 2 1 0
0 1 2 1
0 0 1 2
_

_
.
27
Gerschgorins Theorem implies that the union of all circles
|a
ll
|
N

j=1
j=l
|a
lj
| .
must contain all eigenvalues.
For l = 1 and 4 we get the relation |2 | 1.
For l = 2 and 3 we get |2 | 2.
The matrix is symmetric - the eigenvalues are real so Gerschgorins Theorem implies
0 4 .
The eigenvalues of A are

1
= 3.618,
2
= 2.618,
3
= 1.382, and
4
= 0.382.
hence, the largest eigenvalue is indeed less than 4.
0 1 2 3 4
2.3 The Jacobi Iterative Method
The Jacobi Iterative Method follows the iterative method shown in Example 2.1.2.
Consider the linear system
Ax = b, A
NN
= [a
ij
] , x
N
= [x
i
] , b
N
= [b
i
] .
Let us try to isolate x
i
. The ith equation looks like
N

j=1
a
ij
x
j
= b
i
.
Assuming a
ii
= 0 for all i, we can re-write this as
a
ii
x
i
= b
i

j=1
j=i
a
ij
x
j
,
so,
x
i
=
1
a
ii
_
_
_
_
b
i

j=1
j=i
a
ij
x
j
_
_
_
_
giving the recurrence relation
x
(k+1)
i
=
1
a
ii
_
_
_
_
b
i

j=1
j=i
a
ij
x
(k)
j
_
_
_
_
, (2.2)
28
for each x
i
(i = 1, . . . , N). This is known as the Jacobi Iterative Method.
In matrix form, we have
A = DL U, (2.3)
where D is a diagonal matrix with elements a
ii
,
L is a strictly lower triangular matrix, L = [l
ij
] such that
l
ij
=
_

_
a
ij
, i > j
0, i j ,
and U is a strictly upper triangular matrix, U = [u
ij
] such that
u
ij
=
_

_
a
ij
, i < j
0, i j .
The system becomes
(DL U)x = b,
or,
Dx = (L +U)x +b.
Dividing each equation by a
ii
is equivalent to writing
x = D
1
(L +U)x +D
1
b
where the elements of D
1
are 1/a
ii
, so we have pre-multiplied by the inverse of D. Hence, the
matrix form of the iterative method (2.2), known as the Jacobi Iteration Method is
x
(k+1)
= D
1
(L +U)x
(k)
+D
1
b. (2.4)
The matrix B
J
= D
1
(L +U) is called the iteration matrix for the Jacobi Iteration method.
2.3.1 Convergence of the Jacobi Iteration Method
From 2.2.2 recall that an iterative method of the form x
(k+1)
= Bx
(k)
+ c will converge provided
B < 1 and that a necessary and sucient condition for this is to be true is (B) < 1.
Thus, for the Jacobi method, we require B
J
= D
1
(L + U) < 1 for convergence and, hence,
(B
J
) < 1.
29
Example 2.3.1: Let us return once more to Example 2.1.2 and recast it in the form of the Jacobi
iterative method. The linear system we wish to solve is
Ax =
_
_
10 1
1 10
_
_
_
_
x
1
x
2
_
_
=
_
_
12
21
_
_
= b .
The rst thing we need to do is nd D and L +U where A = DL U:
A =
_
_
10 1
1 10
_
_
D =
_
_
10 0
0 10
_
_
and L +U =
_
_
0 1
1 0
_
_
hence, D
1
(L +U) = B
J
=
_
_
0 1/10
1/10 0
_
_
.
Now choosing the matrix norm sub-ordinate to the innity norm we nd
B
J

=
1
10
< 1 .
Alternatively we can consider the spectral radius of B
J
. The eigenvalues of B
J
are given by

2
1/100 = 0
and so
(B
J
) =
1
10
,
which in this case is equal to B
J
.
So if x is the limit to our sequence then
x
(k+1)
x


1
10
x
(k)
x

.
In Example 2.1.2. we had
x
(0)
=
_
0
0
_
and x =
_
1
2
_
,
so x
(0)
x

= 2 and
x
(1)
x


1
10
2 = 0.2 .
Remember,
x
(1)
=
_
1.2
2.1
_
,
so,
x
(1)
x =
_
0.2
0.1
_
.
and indeed,
x
(1)
x

0.2 .
Since the size of (B
J
) is an indication of the rate of convergence we see here that this system
converges at a rate of (B
J
) = 0.1. The smaller the spectral radius the more rapid the convergence.
So is it possible to modify this method to make it faster?
30
2.4 The Gauss-Seidel Iterative Method
To produce a faster iterative method we amend the Jacobi Method to make use of the new values
as they become available (e.g. as in Example 2.2.2.).
Expanding out the Jacobi Method (2.4) we have
x
(k+1)
= D
1
(L +U)x
(k)
+D
1
b
= D
1
Lx
(k)
+D
1
Ux
(k)
+D
1
b.
Here D
1
L is a lower triangular matrix so the ith row of D
1
Lx
(k)
contains the values
x
(k)
1
, x
(k)
2
, x
(k)
3
, . . . , x
(k)
i1
.
(components up to, but not including the diagonal).
Likewise, D
1
U is an upper triangular matrix so the ith row contains
x
(k)
i+1
, x
(k)
i+2
, . . . , x
(k)
N
.
If we compute the x
(k+1)
i
s in the order of increasing i (i.e. from the top of the vector to the bottom)
then when computing
x
(k+1)
i
,
we have available
x
(k+1)
1
, x
(k+1)
2
, . . . , x
(k+1)
i1
.
Hence, a more ecient version of the Jacobi Method is to compute (in the order of increasing i)
x
(k+1)
= D
1
Lx
(k+1)
+D
1
Ux
(k)
+D
1
b.
This is equivalent to nding x
(k+1)
from
(I D
1
L)x
(k+1)
= D
1
Ux
(k)
+D
1
b,
or,
x
(k+1)
= (I D
1
L)
1
D
1
Ux
(k)
+ (I D
1
L)
1
D
1
b.
This is known as the Gauss-Seidel Iterative Method.
The iteration matrix becomes
B
GS
= (I D
1
L)
1
D
1
U
= [D(I D
1
L)]
1
U
= (DL)
1
U.
31
The way of deriving the Gauss-Seidel method formally is as follows:
A = DL U,
so Ax = b becomes
(DL)x = Ux +b,
and hence,
x = (DL)
1
Ux + (DL)
1
b,
generating the recurrence relation
x
(k+1)
= (DL)
1
Ux
(k)
+ (DL)
1
b. (2.5)
The iteration matrix for the Gauss-Seidel method is given by B
GS
= (D L)
1
U. Thus, for
convergence (from 2.2.2) we require that
B
GS
= (DL)
1
U < 1 .
Example 2.4.1: Again we reconsider the linear system used in Examples (2.1.2, 2.1.3 & 2.3.1) and
recast it in the form of the Gauss-Seidel Method:
A =
_
_
10 1
1 10
_
_
,
and since A = DL U, we have
DL =
_
_
10 0
1 10
_
_
and U =
_
_
0 1
0 0
_
_
.
Then
(DL)
1
=
_
_
1/10 0
1/100 1/10
_
_
, so (DL)
1
U =
_
_
1/10 0
1/100 1/10
_
_
_
_
0 1
0 0
_
_
,
and thus the Gauss-Seidel iteration matrix is
B
GS
= (DL)
1
U =
_
_
0 1/10
0 1/100
_
_
.
Clearly, the norm of the iteration matrix is
B
GS
= (DL)
1
U

=
1
10
< 1 ,
and hence, the method will converge for this example.
Let us look at the eigenvalues to get a feel for the rate of convergence. The eigenvalues are given by
det
_
_
1/10
0 1/100
_
_
= 0 ,
32
or,
_

1
100
_
= 0 ,
so we have
= 0 or =
1
100
,
and hence,
(B
GS
) =
_
(DL)
1
U

=
1
100
.
Observe that in this example even though B
GS
= B
J
we have (B
GS
) = [(B
J
)]
2
(cf Example
2.3.1), implying that Gauss-Seidel converges twice as fast as Jacobi.
2.5 The Successive Over Relaxation Iterative Method
The third iterative method we will consider is a method which accelerates the Gauss-Seidel method.
Consider the system Ax = b, with A = D L U as before. When trying to solve Ax = b, we
obtain an approximate solution x
(k)
of the true solution x. The quantity r
(k)
= b Ax
(k)
is called
a residual and it is a measure of the accuracy of x
(k)
. Clearly, we would like to make the residual
r
(k)
to be as small as possible for each approximate solution x
(k)
.
Now remember, when calculating x
(k)
i
, the components x
(k+1)
1
, . . . , x
(k+1)
i1
are already known. So in
the Gauss-Seidel iterative method for the most recent approximation, the residual vector is given by
r
(k)
= b Dx
(k)
+Lx
(k+1)
+Ux
(k)
.
Ultimately, we wish to make x x
(k)
as small as possible. However, as we dont know x yet, we
instead consider x
(k+1)
x
(k)
as a measure for x x
(k)
. We now wish to calculate x
(k+1)
such that
D(x
(k+1)
x
(k)
) = (b Dx
(k)
+Lx
(k+1)
+Ux
(k)
) ,
where is called the relaxation parameter. Re-arranging, we get
(DL)x
(k+1)
= ((1 )D+ U)x
(k)
+ b,
and hence, the recurrence relation is given by
x
(k+1)
= (DL)
1
((1 )D+ U)x
(k)
+ (DL)
1
b. (2.6)
The process of reducing residuals at each stage is called Successive Relaxation. If 0 < < 1,
the iterative method is known as a Successive Under Relaxation and they can be used to
obtain convergence when the Gauss-Seidel scheme is not convergent. For choices of > 1 the scheme
33
is a Successive Over-Relaxation and is used to accelerate convergent Gauss-Seidel iterations.
Note, = 1 is simply the Gauss-Seidel Iterative Method.
The iteration matrix for the S.O.R. method - Successive Over-Relaxation with > 1 is given
by
B
SOR
= (DL)
1
[(1 )D+ U] .
The iteration matrix B
SOR
can be derived by splitting A in the following way:
A = DL U = D
_
1
1

_
+
1

DL U, > 0 .
Thus Ax = b can be written as
_
1

DL
_
x =
_

_
1
1

_
D+U
_
x +b
(DL)x = ((1 )D+ U) x + b,
so,
B
SOR
= (DL)
1
[(1 )D+ U]
The aim is to choose such that the rate of convergence is maximised, that is the spectral radius,
(B
SOR
()), is minimised. How do we nd the value of that does this? There is no complete
answer for general N N systems, but it is known that if, for each, 1 i N, a
ii
= 0 then
(B
SOR
) |1 | .
This means that for convergence we must have 0 < < 2.
Example 2.6.1: We return once more to the linear system considered throughout this chapter
in Examples (2.1.1, 2.1.2, 2.3.1 & 2.4.1) and recast it here in terms of the SOR iterative method.
Recall,
A =
_
_
10 1
1 10
_
_
,
and A = DL U such that
(1 )D+ U = (1 )
_
_
10 0
0 10
_
_
+
_
_
0 1
0 0
_
_
=
_
_
10(1 )
0 10(1 )
_
_
,
and
(DL) =
_
_
10 0
0 10
_
_

_
_
0 0
1 0
_
_
=
_
_
10 0
10
_
_
Now
(DL)
1
=
_
_
1/10 0
/100 1/10
_
_
,
34
thus the iteration matrix
B
SOR
= (DL)
1
[(1 )D+ U] =
_
_
1 /10

(1)
10

2
100
+ 1
_
_
.
The eigenvalues of this matrix are given by
[(1 ) ]
_

2
100
+ 1
_


2
(1 )
100
= 0 ,

_
(1 ) +

2
100
+ 1
_
+ (1 )
_

2
100
+ 1
_


2
(1 )
100
= 0 ,

_
2(1 ) +

2
100
_
+ (1 )
2
= 0 .
Solving this quadratic for gives
=
1
2
_
2(1 ) +

2
100

_
4(1 )
2
+ 4(1 )

2
100
+

4
10
4
4(1 )
2
_
1/2
_
= (1 ) +

2
200

1
2
_
4(1 )

2
100
+

4
10
4
_
1/2
= (1 ) +

2
200


20
_
4(1 ) +

2
100
_
1/2
.
When = 1 (the Gauss-Seidel Method), one root is 0 and the other is
1
100
. Changing changes
these roots. Suppose we select such that
4(1 ) +

2
100
= 0 ,
so there are equal roots to the equation. Then this implies,

2
200
= 2(1 ) and = 1 .
The smallest value of ( > 1) producing equal roots is
= 1.002512579 ,
which is not very dierent ( 1) to Gauss-Seidel!
However, the spectral radius of the SOR iteration matrix is just
(B
SOR
) = 0.002512579
compared with (B
GS
)=0.01.
(B) is very sensitive to . If you can hit the right value, the improvement in speed of convergence
of the iteration method is signicant.
Although this example is only a 2 2 matrix, the comments apply in general. For a larger set of
equations, convergence of Gauss-Seidel can be slow and SOR with an optimum value of (if it can
be found) can be a major improvement.
35
2.6 Convergence of the SOR Method for Consistently Or-
dered Matrices
In general, it is not easy to nd an appropriate for the SOR method and so an is usually chosen
which lies in the range 1 < < 2 and leads to a spectral radius, (B
SOR
) which is as small as
reasonably possible. However, there are a set of matrices for which it is relatively easy to nd the
optimum .
Consider the linear system Ax = b and let A = DL U. If the eigenvalues of
_
D
1
L +
1

D
1
U
_
, = 0 ,
are independent of , then the matrix is said to be Consistently Ordered, and the optimum
for the SOR iterative method is
w =
2
1 +
_
1
2
(B
J
)
.
Explanation
First, we note that for such a matrix, consistently ordered (eigenvalues are the same for all ) implies
that the eigenvalues of
D
1
L +
1

D
1
U
are the same as those for D
1
L +D
1
U = B
J
, the Jacobi iterative matrix (i.e. put = 1).
Now consider the eigenvalues of B
SOR
. They satisfy the polynomial
det(B
SOR
I) = 0
or
det
_
(DL)
1
((1 )D+ U) I

= 0 ,
and hence,
det(DL)
1
. .
det [(1 )D+ U(DL)] = 0 ,
= 0
so satisfy
det [(1 )D+ U+ L] = 0 .
Since = 0, the non-zero eigenvalues satisfy
det
__
(1 )

D+
1

U+

L
_

_
= 0 ,
36
and thus,
det
_

D
1
L +
1

D
1
U
( + 1)

I
_
= 0 .
When consistently ordered, the eigenvalues of

D
1
L +
1

D
1
U
are the same as those of D
1
(L +U) = B
J
.
Let the eigenvalues of B
J
be , then the non-zero eigenvalues of B
SOR
satisfy
=
+ 1

.
If we put = 1 (i.e. recover Gauss-Seidel), then =

, or =
2
. (Recall Example 2.4.1 where
this result was also found).
For = 0, we have

2
=
2
+ 2( 1) + ( 1)
2
,
or,

2
+ (2 2
2

2
) + ( 1)
2
= 0 .
The eigenvalues of B
SOR
are then given by
= ( 1) +

2

2
2

1
2
_
4( 1)
2
4( 1)
2

2
4( 1)
2
+
4

4
= 1 +

2

2
2

_
(1 )
2

2
+

4

4
4
= 1 +

2

2
2

_
(1 ) +

2

2
4
.
For each
2
there are 2 values of , these may be real or complex. If complex (note > 1), then
= ( 1)
2
or || = 1 .
Hence,
(B
SOR
) = 1 .
For the fastest convergence we require (B
SOR
) to be as small as possible. It can be shown that
the best outcome is to make the roots of B
SOR
equal when = (B
J
), i.e. when is largest. This
implies

2
4
+ 1 = 0 .
Solving for yields
=
1
_
1
2

2
/2
,
=
2

2
_
1 (1
2
)
1
_
1
2
_
,
=
2
1
_
1
2
.
37
We are looking for the smallest value of and so we take the positive root of the above equation.
Hence, with = (B
J
), the best possible choice for is
=
2
1 +
_
1 ((B
J
))
2
.
Example 2.6.1: We again return to Example (2.2.1, 2.2.2, 2.3.1 & 2.4.1) and show that it is a
consistently ordered matrix and determine the minimum , and hence, the fastest rate of convergence
for the SOR method.
As before we have
A =
_
_
10 1
1 10
_
_
,
then
D
1
L +
1

D
1
U =
_
_
0
1
10

10
0
_
_
,
and the eigenvalues are given by

1
10
_
_

10
_
= 0 so
2
=
1
100
,
and hence, the matrix is consistently ordered.
Then by applying the above formulae and recalling that the eigenvalues of B
J
are = 1/10 (Example
2.3.1) we have
=
2
1 +
_
1 ((B
J
))
2
=
2
1 +
_
1 (1/100)
,
= 1.0025126 .
This is essentially the same value as we found in Example 2.5.1.
Thus the fastest rate of convergence for this particular system is
(B
SOR
) = 0.0025126 ,
as shown in Example 2.5.1
38

You might also like