8

Page No.
(1-ok1
8 ))
+
mroN
H wo0d a a
8
a (B v
PHOMua
O
A

R
L
 Pb

N
 ixersi
Page No.
T988 Date...
t799-0
= KtEES L99 *o-
999 S
T98
99 '0
h-
8 8-
 Z > 5-6662-533 6D= S.5226,

2--z =3

OShtal Radius Manmvalue a

a
taknq ( a l p
all cgd Naluns:
Apectrel Radiu< 1
hem t e OuAel"
ConeR

peeta Radius
wfl Hdt ComNe

Fo acok
H- -D(Lt 0)
FH) 11
Max , ewen
Valln ef A
31e fw)1
 FM

N
11
10 NUMERICAL METHODS

10.2 ITERATIVE METHODS FOR SOLVING LINEAR SYSTEMS

AS a numerical technique, Gaussian elimination is rather unusual because it is airec
5, a solution is obtained after a single application of Gaussian climination. Once a sou
tion has been obtained, Gaussian elimination offers no method of refinement. Tne tc
retinements can be a problem because, as the previous section shows, Gaussian elimination
Is sensitive to
rounding eror.

Numerical techniques commonly involve an iterative method. For exampic, 1u

more

calculus you probably studied Newton's iterative method for

approximating the zeros of a
dilferentiable function. In this section
you will look at two iterative methods lor approxi-
mating the solution of a system of n linear
equations
in n variables.

The Jacobi Method

The first iterative
technique is called the Jacobi method, after Carl Gustav Jacob Jacobi
(1804-1851). This method makes two assumptions: (1) that the system grven by

has a unique solution and (2) that the coetficient matrix A has no zeros on its main diago-
nal. If any of the diagonal entries a . d z s a ure zero, then rows or columns must be
interchanged to obtain a cocticient marix
that has nonzero entries on the main diagonal.
To begin the Jacobi method, solve the hrst cquation lor f . the second equation for x2.
and so on, as follows,

( ai22 di tn

22

the solution,
Then make an initial appruximation of
Initial apprusimetion

side of the rewritten equations to obtain

and substitute these values of x, into the right-hand
nas been completed, one
iteration has been
Aler this procedure
the Jirst aPproximation.
 SECTION 10.2 ITERATIVE METHODS OR SOLVING LINEAR SYSTEMS
57
performed. In the way, the second approximation is formed by
same
substituting the first
approximation's r-values into the right-hand stde of the rewritten
equations. By repeated
iterations, you will form a sequence of
approximations that often converges to the actual
solution. This procedure is illustrated in Examplel.

EXAMPLE 1 Appbing the Jacobi Method

Use the Jacobi method to
approximate the solution of the
following system of linear
equalions.

Sx2x
-3x, + 9x, + x= 2

2x X-7x 3
Continue the iterations until two
successive upproximations are identical when rounded
three significant digits. top

Solution To begin, write the system in the form

-+ix-
+ -
- + ix- x.
Because you do not know the actual solution, choose
0, 0, x0 nitial approimation
as a convenient initial
approximation. So, the first approximation is
3t30) -(0)= -0.200
x, t0) -#0) = 0.222
x-+(0) %0) -0.429
Continuing this procedure, you obtain the
sequence of
apprOximations shown in Table 10.1.
TAELE 101

0.000-0.200 0.146 0.192

0.1811 0.185 0.186 0.186
0.000 0,222 0203 0.328 0.332 0.329 0.331 0.331
0.000-0.4290.517 -0416 -0.421 -0.424 0.423-0423
 NGEKIAL MEIHODS

Because the last two columns

in Table 10.1
significant digits the solution is are identical, you can conclude that to nice

X0.186, = 0.331, -0423.

For the
system of linear
cquations given in
converge. That is, repeated iterations succeed in Example I,an Jacobi method is
the said to
to three producing approximation that is
signiticant digits. As is generally true for iterative methods, correct
require more iterations. greater accuracy woul

The Gauss-Seidel Method

You will now look
modification of the Jacobi method called the Gauss-Seidel
at a
named after Carl Friedrich Gauss
(1777-1855) and Philipp L. Seidel (1821-1896). method,
modification is no more ditficult to use than
the Jacobi method, and it often requires fewer This
terations to produce the same
degree of accuracy
With the Jacobi method, the values of x, obtained in the nth
approximation remain
unchanged until the entire (n +
1)th approximation has been caleulated. With the Gauss-
Serdet method, on the other
hand, you use the new values of cach x as soon as
known. That is,
they ure
once
you have determined X, from the first cquation, its value is then uscd
in the second equation to obtain the new X2. Similarly, the new
X and x, are used in
the third cquation to obtain the new and so on.
X, This
procedurc iS demonstrated in
Example 2.

EXAMPLE 2 Appbing the Guuss-Seidel Method

Use the Gauss-Seidel iteration method to
approximate the solution to the system of
equations given in Exumple 1.

Solution The first computation is identucal to that grven in Exumple 1. That is, using (i g)
(0,0,0) as the imitial approximation. you obain the follow ing new value for
x - + 0 ) 0 ) = -0.200
Now thut you have a new vialue Tor Anowever, use t to canpute A new value tor . That

15,

+-0.200)-(0)= 0.156,
Similarly, use x, 0.200 and 0.156 to compute a new value for . That is,

x - t (-0.200)-(0.156) -0.508.

So the first approximation is x-0.200, 0.156, and a -0,508. Continued

iterations produce the sequence of upproximations shown in Table 102.
 SECTION 10.2 TERATIVE METHODS FOR SOLVING LINEAR SYSTEMS

TABLE 10 2

0.000-0.200 0.167 0.191 0.186 0.186

0.000 0.156 0.334 0.333 0.331 0.331

0.000 -0.508 -0.429 -0422-04230.423

Note that after only five iterations of the Gauss-Seidel method, you achieved the same
accuracy as was obtained with seven iterations of the Jacobi method in Example 1.

of
Nether the iterative methods presented in this section always converges. That is, it is
possible to apply the Jacobi method or the Gauss-Seidel method to a system of lincar equa-

tons
method obtain divergent sequence of approximations. In such cases, it is said that the
anddiverges. a

EXAMPLE 3 An Example of tDivergence

Apply the Jacobi method to the system

Sr,-4

using the initial upprOXIUnation ( ) ( 0 , 0), and show that the method diverges.

SolutionAs usual, begin by rewritng the given system in the form

-4+ St
6*

Then the initial approximation (0, 0) proeluces

-4
+ SI0) =

us the first approximation. Repearea tertois prouce the cquence ot

shown in Jable 10.5
approximations

TABLE 3

-4 34 174 1244 6 1 2 4 42,874 -214,374

-6 -34 244 1244 8574-42874 S00,124

 PTER 10
NUMERICAL METHODS

For of linear equations you can determine that the actual soluthon
and
this particular system
. So you can see from Table 10.3 that the approximations given by
he Jacobi method become progressively worse instead of better, and you can conciude that
the method diverges.

ne problem of divergence
method rather than the
inExample 3 is not resolved by using the
Gauss-Seidel
Jacobi mnethod. In fact. for this particular
method diverges
the system Gauss-Scdct
more rapidly, as shown in lable
104
TABLE 104

-174 -6124 -214374 -7.503,124

34 1224-42,874-1500,624-$2521,874

wih an imitul approximation of (.I,)= (0,0), nenther Jacobi

uss-Sendel method
converges to the solution of the
the
method nor the
system t lincar
cquutions
Npic at special type of cocticient matrix 4, cullcd a gnven in
now
TOu w look a

diagonaly dominant matris, for which it is

guaranteed that both methods will strictly
converge.
Definition of Stictly An n *
nmatrix A is
strictly díagonaly dominant if the absolute value
Diegonaly Dominant on the main
dingonal 1s greater than the surm of the absolute of cach entry
values of the other
in the sane row.That is, entries
Matrix
laul> lal+ lal+**+la
la2al4+ lasl +*** la

lala+ la.l+** +laa-1l

EXAMPLE 4 Strictly Diagonally Dominunt Matrices

Which of the following systems of linear cquations has a strictiy diagonally dominant
coctlicient mutrix?

(a) Jr, 4

2
2x, + Sx,
(b) Ax, + 2 , -1

3r,Sr+ 3
 SECTION 10 2 TERATIVE METHODS FOR SOLVING LINEAR SYSTEMS

Solution (a) The coefficient matrix

iS strictly diagonally dominant because | 3 1 - 1 | and |5| > 121

(6) The coeficient matrix
4 2

IS not stricily diagonally dominant because the entries in the second and third rows
do to the
ot contorm deinition. For imstance, in the second ruw 4 d third
it is not true that l4alail+ l42l. Interchanging
drows in2,the
undorigirnal the second and
systerm of linear cquations, however. produces the cocficient matrix

and this mutrix is strictly diagonuly dominant.

The following theorem, which is listed without proof, states that strict diagoal domi-
nance is suficient for the convergence of either the Jacobi method or the Ciauss-Seidel
nethod.

Theorem 10.1 then the system of linear equations given by Ax b

=

tA is
strictly diagonally dominant,
has a unique solution to which the Jacobi method and the Gauss-Seidel method will con-
Convergence o verge 1or any initial approximation.
the Jacobi and
Gauss Seidel Methods

In Example 3 you looked at a system of linear cquations for which the Jacobi and Gauss-
Seidel methods diverged. In the following example you can see that by inlerchunging the
can obtain
a coellicient mitrix that is strictly
rows ot
tne systeim gnven in
Exampie s, you
is assured.
dingonaly dominan. Aler this interchange, convergence

EXAMPLE S Interchanging Rows to Obtain Comvergence

Interchange the rows of the system

the a u s s
coetlicient matrix. Then apply
obtain one with a strictly diagonaly dominant
to sofution to four signiicant digts.
Seidel method to approximate the
ICAL METHODS

of the given system to obtain

blution Begin by interchanging the two rows

Tx X2 6

X -4.

Note that the coefficient matrix of this system is strictly diagonally dominant. Then soive
for , and 2 as follows.

Using the initial approximation (x,.x,) = (0, 0), you can obtain the sequence of approxi-
mations shown in Table 10.5.

TABLE10.5

O.0000 O.8571 0.9959 0.9999 1.000 1.000

0.0000 0.9714 0.9992 1.000 1.000 1.000

So you conclude
can
that the solution is x I and x, =1.

Do not conclude
from Theorem l10.1 that strict
diagonal dominance is a
dition for convergence of the Jacobi or Gauss-Seidel methods. For necessary con-
matrix of the system instance, the coefficient

2 3
is not a
strictly diagonallydominant
matnix,initaal
and yet
x - 1 and Iwhen you use an
boih methods converge to the solution
Exercises 21-22.)
upproximation of
(i.x) (0, 0). (See
(See
 SECTION 102 EXIPCSES

SECTHON10.2 EXERCISES
In Exercises1-4, appiy the
Jacobi metod to the gIven system of 19. Interchange the ws of the systenm of near
lincarcquations, using the inital approxation .
cqualons in
a
(0,0, 0 ) Continue performing itera
d wo successive N o obtan system with asricty diagonally
1hen appiy
dom-
appruvimatiws are identical when ruded o threc Sgnficant digits ACCtCCnthe uatri. the auss-Seidel
solution to two significant digits.
method
appr mate
20. Interchange the
4x,+2, -6 of the
rows system of linear equations
NerCEe 12lo obtaim a system with a strietly diagonally dkom-
in

nant coethicicnt matriK. Then apply the Causs- Seidel method

To
pproximate the solution to two significant digits.
, 2
In Exercises 21 and 22, the coetlicient matrix
of system of lincar the
cyualons is not strictly diagonally dominant. Show that the Jacobi
S. Apply the
6.
Gauss- Seidel miethod to Exercise
1. and Gauss-Seidel methos
converge using un nitial appronimation
Apply the Gauss-Sedel method to Exercise 2. g (0,0, 0).
7. Apply the Ciauss-Seidel method to Exercise 3. 21. -4x,
8
APpiy the
Causs- Sedel methol to Exercise 4. +Sa,= I 4r t
22 =0

In
2x3
Exereises 9-12, show that the Ciauss-Sesdel mcthod denges for
* ,5
te gvC 5ysietn using the inittal
approximution (s,, t.
10,0, 0). E In Ewrvises 25 and 24, wrnte a
computer program that applies the
auss-SiCuet metnon to sfve he
system of lincar
cquatons.
10., t 4,
S 2 2 23. 4r,
I1. 2t, 6
10r 9 s. -5
5
,
2, |
In Exercises 13-16, determine whether the matrix is stnetly diag 12
nally dminant.
22
- S, 2
A. 4
. - s
x+a A
6
. mierchange the ruws of the sysiem of lincar cquations in A
in a system with a strictly diagonalty dml
nant coeflicient matrix. Then apply the Giaun Seidel method A
t proimale the solution to wo sgnificant digits,
ot in
Lnterchange the ruws sy stem
of the
lincar equallos
TEINC 1O
lo obtuin system with stnctly dagonally dom
a a

Then upply the Gaauss- Seidel mcthod

nan
cocilicient matrx
o pproximute thhe solution to two significant digits