NUMERICAL SoLUTIONS OF

ALGEBRAIC EQUATIONS
6
A1.Introduction.
mathematics and engineering we frequently face
Inapplied finding one or more roots of the equation
Lpnoblem of
f(*) = 0 (1)

f(x) is, in general, a nonlinear function of the real

chere difficult to have explicit
coriable t. But in most cases, it is very proceed to look
solutions of the equation (1) and, therefore, we
specified degree of
for a root of (1) numercially with any finding these roots are
AcCuracy. The numerical methods. of
caled iterativemethods.
following forms:
The function f(z)may have any one of the
degree n,
(ifE) is an algebraic or pclynomial function of
say, so that
+a,x"
f(x) = a, + a,*+ayt t..tan-1
real or complex, and
where a,(i=0,1,2,-.,n) are constants, +°-4x+3 etc. are
40. For example, r-7:+1
algebraic functions. In sych cases, the equation f(*) =0is
called algebraic equation
function
tt7 f(x) is a transcendental function, i.e. f(*) is a
of the form
2
f(x)= a, + aj* + ayt +.to o,
all
where a,(i =00,1,2,), are constants, real or complex, and
9 0. For example, sinx+2x-1, e* +logx +5 ete. are
Here the equation f(*) =0 is called
Lranscendental functions,
transcendental equation.)
Every value a of xfor which the function f(*) is zero, i.e.,
fla) =0 is called a root or zero of the equation (1). In this
Chapterwe shall discuss different numerical methods to compute the
transcendental
pproximate =0
tquationf(x)
real roots of an algebraic or
 NUMERICALANALYSIS.
174 -Theoretical. &&
that
Practical
To develop the methods we assume
and
WY The functionf(x) is continuoustimes,
differentiablefora sufficient number of
mf() has no multiple root, i.e., if a is a real ros.
continuously
f(x) =0 then
f(a) =0, f'(a) +0.)
Determiation of approximate (real) root of (1)by numerical
methods to be discussed here, consists, in general, of tha
following twosteps.
(i) Isolating the roots, i.e., finding the smallest possible
interval (a, b]containing one and only one root of (1).
(ii) Improving the values of the approximate roots, i.e
refinding them to the desired degree of accuracy.
To implement the first step, we use the following theorem
of a copnuous function :
Zheoren 1.If real valued function f(x) is continuous in
, b]and f(a).f(b) are of opposite igns, then there is at least
one real root of f(x)= 0 in (a, b)
6.2. Iteration Processes.
Let the sequence }of iteratesof a root a of the equation
f(x) = 0
is produced by a given method. Then theerror en involved
at the n th iteration is given by
(2)
and
’0 as n ’ oo, we say that the iteration converges
the sequencex,, converges to a . Otherwise, the iteration is
our
divergent and the method of computation fails. Thusofthe
primary task is to find
iteration processes. "The the
error condition
En+l can beof expressed
convergence
interns
of E, E- En-2;**** which we call error
equation. h,isan
Jf we define h, by h, = then iteration
,
X-, =E, -Entl,
approximation of E, if nl approximates a. If the terns
converges, then we can find an upper bound for en+in
of h,.This is called estimution
oferror.
VMERICALSOLUTONS OF ALGEBRAIC EQUATIONS 175

case an
iterative method converges, we can find two
In
Onstants p21 and q>0such that

lim=q (3)

Here called the order of convergence and gis known

a^ymptotic.error COnstant. The iterative method with p>1
aS
renerally converges rapidly.
The convergence of the iterative method also depends on
he initi£l approximation x, of the root a. . If this initial
approximation is not satisfactory,the iterative method does not
converge and then we look for the new computation.
The iterative method is self correct, ie. if there is an
sccidental error in the calculations of iteration, the erroneous
iterate acts as a new initial approximation leading to a correct
result, provided that the error is not large enough for which
the method fails.
6.3, Bisection Method.
A. Basic principle and formula
The method of bisection is the most simplest iterative
method. It is also known as half-interval or Bolzanomethod.
Ihis method is based on Theorem 1 on the change of sign.
In this method, we first find out asufficiently small interval
Ayoo containing the required root a of the equation (1).
en f(a)f(b,) <0 and f(x)has the same sign in
IPio and so f(x) isstrictly monotonic in ao3boJ
generate the sequence x,}of iterates, we put xo = 0
1
and *j =a0
2 +bo) and find f(*1). If flao) and
flz)are of opposite signs, then set a, =ao,6, =*, So that
-lagi]. On the other hand, if f(xi) and f(bo) are of
'pposite si bo].
signs then put aja, = x,,b, =bo, i.e.[a, b,]=[x,
See that a,b.lcontains the root ain either case.
176 NUMERICAL ANALYSIS-TheoreticnlL&Practinl,

Next set t, =aj toi) and repeat the above prOcess em

obtain

with desired accuracy with x, ’ a as n ’ o,

B. Convergence of bisection method

Suppose the interval a, ,b, contains the root a
fla, )f(b,) <0.
1,
Let *nal =an
2 tbn).
If f(a, f(n41)<0, then set an+i =ansOn+1 =
n+1:
On theother hand, if f(inif() <0, then
an1 =n41,bn+1 =b.
Thus in any case

and
aelan1d,1} f(an+)f(bn)<0
1
bo-a0
If En+ be the error in
approximating a by n+1) then
En: |.-sb, -a, <00 as n ’ .
2
Thus the iteration
converges.
Since so the methodis
linear.
2
convergence in bisection

C. Advantage and disadvantage of bisection method

Stageof
Advantage. This method is very
iteration the approximate value ofsimple, as at any ofthe
root
the desired
 es D1.NIRÂNur pasdaimal, ing

Sazin Let i=rmr-1-X.

-jl=-1-S<4i=0-27408>0.

Tate =0, ,=l so thst =+1)= S. Sìne

ro: Bs beiwen 0.5
f25 =-10849<0d fu>0, the

tabe:
Prenzinthis wag, we obtain the flowing

faj<o|fè)>0
1 0.3 -1.006S49

0.5 1 0.75 0.381303

0.75 1 0.875 -0.232236
2
3 0.875 1 0.9375-0.003058
0.9375 1 0.9637500.129S19
0.9375 0.96375 0.933125 0.061675
6
9375 0.961675 0.9453120.023s9S
93750.945312| 0.941406 0.012819
93750.941406| 0.9394530.004856
9375 0.939453| 0.938477 0.000893
93750.938477| 0.937988-0.001084
 ERICAIL SOLUTIONS OF ALGEBRAIC EQUATIONS
177

quationf(x) = 0 does not depend on the values f(z, ) but on

theirsignsonly. Also the method is unconditionally and surely
onvergent.

Disadvantage. The method is very slow and requires large

numberoffiteration to obtain moderately accurate results and
laborious,
henceitis
ExampletFind the root of the equation xtan x=128, that
lesintheinterval(0,1), correct. to four places of decimal, using
bisection method. (W.B.U.T., CS-312,2005]
Solution. Let f«)= xtanx -1-28.
f(0) =-1-28 < 0, f(1)= 0-277408 > 0.
So a root lies between 0 and 1.
Take a, =0, bo =1 8o that *1=0+1) = 0.5. Since
f(05) =-1006849 < 0and f(1) > 0, the root lies between 0.5
1
and 1. Thus we have t, = (05 +1) =0.75
Proceeding in this way, we obtain the following table :
No. ofiteration b, f(*n)
2
(n) fe)<o fB,) >0
0 1 0.5 -1.006849
1
0.5 1 0.75 -0.581303
2
0.75 1 0.875 -0.232256
3
0.875 0.9375 -0.003058
4
5
0.9375 1 0.9687500.129819
6
0.9375 0.96875 0.953125 0.061675
7 0.93750.961675| 0.945312 0.028898
0.9375 0.945312| 0.941406 0.012819
9 0.9375 0.941406| 0.939453 0.004856
10 0.93750.939453| 0.938477 0.000893
0.93750.938477 0.937988 -0.001084
178
NUMERICAL ANALYSIS-Theoretion! &.Proy
11 0.9379880.938477 0.938232 -0.00009
12 0.938232 0.938477 |0.938354
13
14
0.938232 0.938354 |0.938298
0.938232 |0.938293 0.938263 00..00 003195%1
Thus a real root of the given equation is
four decimal places.
0.
0.9383
00002
Corred)
6.4 Regula-Falsi Method.
A. Basic principle and formula [W.B.U.T., CS-312,20n
The regula-falsi method or false position method is
sometimes referred to as the method of linear internolat
and it is the oldest method for computing real roots of
equation f(r) = 0.
To findareal root a of f(x) =0,we first choose asufickent
small interval [xo,X,]in which the root a lies. Then f(M)
f(x) must be of opposite signs so that f()f(x) <0and
graph of f(x) must cross the X-axis between x= Xyand
Since the interval [Xo,x]is sufficiently small, the portion

the curve between A[x.f(xo)] and B[X1.f(4)andcaso

approximated by a secant line (straight line)
U-axis gives
intersection of the secant AB with the
approximate value x,,say, of the root.

(*0.f(o)

(2.f(a)
wERICALSOLUTIONS OF ALGEBRAIC EQUATIONS
179

equation of the secant line AB is

The
y-f(x)=
We derive
f(xo)-J(312(x-)
Putting y=0,X = X2,
f()
(5)

If,)= 0, then *, isa root of f) =0; otherwise, f(x,)<0

He.)>0 It f(*.) and f(*g) are of opposite signs the root
Eas between xo and t9 and in this case we set x =x.

and &, =I1. On the other hand iff(x; )and f(x) are of
opposite signs, the root lies between x, and x, and thus, in
either case,
f(*)f(*;) <0
Hence the next approximation of the root, say xa lies
between x and x and get
f(*)
f(*z)-fa)l2-1) (6)

The general formula based on the above process is

fa,) -(I, -x-), n=1,2,... (72)
fa,)-f(-)
This isregula falsi iterationformula. The process is repeated
until the root is obtained to
required degree of accuracy.
Convergence of regula falsi method.
Let o. be simple root of the equationf(*) =0. Then
a
puting x, in (7), we get
(en -En-1)f(a +e,)
en+1 = En
f(a +e,)-f(a+en-)
180 NUMERICAL ANALYSIS.Theoretical

1 1

(Expanding fin Taylor's series and notingf(a) =

E,f(a) +¿"ia)
2 +.
f(a)+n2 +ea-i"a)+.
1, fa) f(a)
f(a)
1f"(a)
2 f(a) en-1ên + e~en1tE,e). (8)
Sothat
Ep-1 =C En-1en (9)

where C 17() and we have neglect higher

2 f(e) power oi e,
The relation (9) is called the error equation.
To find the order of convergence, we set [n-1 =An and
where theconstants A and m are to be determined
Then the equation (9)gives
AE = C"A "EE, = Ca mE, (10)

m=1+
m

m 5)
Neglecting the minus sign, we find that the order of the
convergence ofnt is m=l·618, Also fron (10) we get
A-Cmm+l
 NUMERICAL SOLUTIONS OF ALGEBRAIC EQUATIONS 181

c. Advantage and disadvantage of regula falsi

method, [W.B. U.T., CS-312,2003]
Advantage. The method is very simple and does no required
n calculate the derivative of f(x) which is difficult for some
reover, the method is evidently convergent.
problems. Moreover,
Disadvantage. Sometimes the method is very slow and not
suitable for practical computation. Also the initial interval in
which the root lies is to be chosen very small.
Example.2. Find a root of the equation 3 - 2x-5 =0 by
Regula-Falsi method correct upto 4 places of decimal
[W.B.U.T. CS-312, 2004]
Solultion. Let f()=-2x-5.
:f(0) =-5, f(1) = -6,f(2) =-1, f(3) =16. So a real root
lies between 2 and 3. We choose x=2,*1 =3 giving
f(xo) =-l and fx) =16. Then the iteration (7) gives

16
=3 -(3 2) = 2-05882 and f(x2)=-0-39082
16 -(-1)
Proceeding in this way, the iteration (7) gives the following
table :
No, of
iteration f(n-1) fon) Xn+l f(Nn+l)
(n)

1 2 3 -1 16 2.05882 -0.39084

2 2.05882 -0.39084 16 2.08126-0.147244

3 2.08126 -0.147244| 16 2.08964-0.054667|
4 2.08964 -0.054667 16 2.09274-0.020198
5 2.09274 3 -0.020198| 16 2.09388-0.007491
6 2.09388 3 -0.007491| 16 2.09430-0.002806
7
2.09430 -0.002806|16 2.09445-0.001133
2.09445 3 -0.001133| 16 2.094451