Sie sind auf Seite 1von 29

Roots of equations

Sea y  f ( x ) . The values of x that make y = 0 are called roots of the equation. The
fundamental theorem of algebra states that every polynomial of degree n has n roots.
In the case of real estate, it must correspond to x values that make the feature cut the
x-axis:
y

f(x)=cos(x)

x
x1 x2 x3 x4

The roots of a polynomial can be real or complex. If a polynomial has coefficients


a0 , a1 , a2 ,...an 1 , an real, then all the complex roots always occur in complex conjugate
pairs. For example, a cubic polynomial has the following general form:
f ( x)  a0 x 3  a1 x 2  a2 x  a3
The fundamental theorem of algebra states that a polynomial of degree n has n roots.
In the case of cubic polynomial can be the following:
 Three distinct real roots.
 A real root with multiplicity 3.
 A simple real root and one real root with multiplicity 2.
 One real root and a complex conjugate pair.
Example. The roots of these polynomials are summarized below.
1. Three distinct real roots:
Raíces reales distintas
50

f1 ( x)  x 3  3x 2  x  3
 ( x  3)( x  1)( x  1) 0

x1  3, x2  1, x3  1
-50

f1(x)
-100

-150

-200
-5 -4 -3 -2 -1 0 1 2 3 4 5

2. A real root with multiplicity 3:


Raíces reales triples
50

f 2 ( x)  x 3  6 x 2  12 x  8 0

 ( x  2)3
-50

x1  2, x2  2, x3  2 -100
f2(x)

-150

-200

-250

-300

-350
-5 -4 -3 -2 -1 0 1 2 3 4 5

3. A simple real root and one real root with multiplicity two:
Raíz sencilla, raíz doble
100

f 3 ( x )  x 3  12 x  16
 ( x  4)( x  2) 2
x1  4, x2  2, x3  2 50
f3(x)

-50
-5 -4 -3 -2 -1 0 1 2 3 4 5

4. One real root and a complex conjugate pair:


Raíz sencilla, raíces complejas
100

f 4 ( x)  x 3  2 x 2  3 x  10
 ( x  2)( x  (2  i ))( x  (2  i )) 50

x1  2, x2  2  i, x3  2  i
0

f4(x)
-50

-100

-150
-5 -4 -3 -2 -1 0 1 2 3 4 5

For their study, the functions can be classified into algebraic and transcendental.

Algebraic functions
Let g = f (x) the function expressed
f n y n  f n 1 y n 1    f1 y  f 0  0
Where f i is a polynomial of order i in x. Polynomials are a simple case of algebraic
functions that are usually represented as
f n ( x)  a0  a1 x  a2 x 2    an x n
Where n is the order of the polynomial.

Example.

f 2 (x) = 1-2.37 x + 7.5 x 2

f 6 (x) = 5 x 2-x 3 + 7 x 6

Transcendental functions
Are those that are not algebraic. They include the trigonometric, exponential,
logarithmic, among others.

Example.
f ( x)  ln x 2  1

f ( x )  e 0.2 sen(3 x  5)
The methods described in this unit require that the function is differentiable in the
range where they apply. If the methods used in non-differentiable or discontinuous
functions at some points, to reach the result will depend, at random, that during the
implementation of the method do not touch those points.
On the other hand, the roots of the equations can be real or complex. The standard
numerical methods for finding roots can be classified into two categories:

1. Determining the real roots of algebraic and transcendental equations. The


techniques used in these cases were designed to find the value of a simple root
in accordance with a prior knowledge of its approximate position.
2. The determination of all real and complex roots of a polynomial, for which
numerical methods are designed specifically for polynomials. Systematically
determining all roots of polynomial rather than only one, given the
approximate position.

Real roots of algebraic and transcendental equations


In general, the methods for finding the real roots of algebraic equations and
transcendental methods are divided into intervals and open methods.
The interval methods exploit the fact that typically a function changes sign in the
vicinity of a root. They get this name because it needs two initial values to be
"encapsulated" to the root. Through such methods will gradually reduce the size of
the interval so that the repeated application of the methods always produce
increasingly close approximations to the actual value of the root, so methods are said
to be convergent.
The open methods , in contrast, are based on formulas that require a single initial
value x (Initial approach to the root). Sometimes these methods away from the real
value of the root grows the number of iterations, ie diverge.
This unit will study a method known as interval bisection method and open two
methods: the fixed point method and the Newton-Raphson method.

2.1.1 Bisection method


This method, also known as range partitioning method, part of an algebraic or
transcendental equation f (x) and an interval [x 1, x 2] such that f (x 1) and f (x 2) have
opposite signs , ie such that there is at least one root in that interval.
Once the interval [x 1, x 2] and secured the continuity of function within this range, it is
evaluated at the midpoint x m the interval. If f (x m) and f (x 1) have opposite signs, will
reduce the range of x 1 to x m, and that within these values is the desired root. If f (x m)
and f (x 1) have the same sign, will reduce the range of x m to x 2. By repeating this
process to make the difference between the two values of x m is a good approximation
of the root.
The algorithm of the method is as follows:
1. Choosing values x 1 and x 2 of the interval.
2. Checking the existence of a root in the interval [x 1, x 2] making sure
f ( x1 ) f ( x2 )  0 . Otherwise, it will be necessary to choose other values for x
1
and x 2.
 x2  x1 
xm  x1 
3. Take 2 and calculate f ( xm ) .

4. If f ( xm ) = 0 is found the root of the function (end of method). Otherwise, go


to step 5.
 x2  x1 
5. Let T the desired tolerance (the margin of error). If 2 <T was found
closer to the root with a margin of error of less than T (end method).
Otherwise, go to step 6.

6. If f ( x1 ) f ( xm )  0 , Then do x2  xm and repeat from 3, otherwise, make


x1  xm and repeat from 3.

Example. Determine the positive real root of the following function considering a
tolerance of 0.001 using the bisection method.

f ( x)  x3  x  1

Solution.
In the graph of the function can be seen that the positive real root of the function is in
the interval [1,2]:
x3-x-1
40

30

20

10

-10

-20

-30

-40
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2
x

For this reason, we choose x 1 = 1 and x 2 = 2. When calculating f ( x1 )  1 and


f ( x2 )  5 we have f ( x1 ) f ( x2 )  5  0 That is, in fact there is at least one root in
the interval.
x3-x-1

20

15

y
10

x1 x2
0 [ ]

0 0.5 1 1.5 2 2.5 3


x

Now, we calculate the midpoint of the interval


 x2  x1  2 1
xm  x1   1  1.5
2 2
and evaluate the function at that point
f ( xm )  (1.5)3  1.5  1  0.875
x3-x-1

20

15
y

10

x1 xm x2
0 [ | ]

0 0.5 1 1.5 2 2.5 3


x

 x2  x1 
 0.5
Like f ( xm )  0 , Ie has not been found and also the root 2 is greater
than the desired tolerance, we proceed to calculate the product f ( x1 ) f ( xm ) :
f ( x1 ) f ( xm )  0.875

That is, f ( x1 ) f ( xm )  0 and then we x2  xm


x2  0.5

We repeat the procedure based on a calculation of xm .


x3-x-1

20

15

y
10

x1 x2
0 [ ]

0 0.5 1 1.5 2 2.5 3


x

As shown, the interval in which the root is reduced. As the method applies interval is
reduced until it is so small that is very close to the actual value of the root.
Below is a table summarizing the results of the method. We see that in the tenth
 x2  x1 
iteration 2 <T, so we can conclude that an approximation to the root is
1.32519532 with an error margin of 0.1%.
x2  x1
i x1 x2 xm f ( xm ) f ( x1 ) f ( xm ) 2
1 1 2 1.5 0875 -0875 0.5
2 1 1.5 1.25 -0.296875 0.296875 0.25
3 1.25 1.5 1375 0.22460938 -0.06668091 0125

4 1.25 1375 1.3125 -0.05151367 0.01529312 0.0625


5 1.3125 1375 1.34375 0.08261108 -0.0042556 0.03125
6 1.3125 1.34375 1.328125 0.01457596 -0.00075086 0.015625

7 1.3125 1.328125 1.3203125 -0.01871061 0.00096385 0.0078125

8 1.3203125 1.328125 1.32421875 -0.00212795 3.9815E-05 0.00390625

9 1.32421875 1.328125 1.32617188 0.00620883 -1.3212E-05 0.00195313


1
0 1.32421875 1.32617188 1.32519532 0.00203666 -4.3339E-06 0.00097656

While convergence of the method is assured as in the range must be the root, it is
usually very slow and it takes a large number of iterations to find a good
approximation to the root.

2.1.2 Fixed Point Method


Sea f ( x)  0 algebraic or transcendental equation. If we add x on both sides have
f ( x)  x  x ... (2.1)
Sea g ( x)  f ( x)  x (Another feature). Substituting (2.1)
x  g ( x) ... (2.2)
That is, the root of the equation is at the intersection g (x) = x and g (x) = f (x) + x.
If x = a is a root of x, then
f (a) = 0
and substituting in equation (2.2)
a = g (a) ... (2.3)

The fixed point method of an initial value x0 near the root


x0  g ( x0 )
And then take as the next approximation
x1  g ( x0 )
By proceeding in this way consistently, it leads to the n-th approximation is
xn  g ( xn 1 ); n  1, 2,3,
To analyze the convergence of the method, we return to the expression (2.3)
a = g (a)

Subtracting both members g ( xn 1 ) and since g ( xn 1 )  xn


a  g ( xn 1 )  g (a)  g ( xn 1 )
a  xn  g (a)  g ( xn 1 )
a  xn 1
1
Multiplying both sides by a  xn 1

g (a )  g ( xn 1 )
a  xn   a  xn1 
a  xn 1 ... (2.4)
On the Mean Value Theorem, we know that
g (a )  g ( xn 1 )
g ( )  ; xn 1    a
a  xn 1
So we replace (2.4)
a  xn  g ( )  a  xn 1  ; xn 1    a
Y
a  xn
g ( )  ; xn 1    a
a  xn 1
In absolute terms:
a  xn
g ( )  ; xn 1    a
a  xn 1

If g ( )  1 , Then a  xn  a  xn1 and obviously the absolute value of the
difference between the root and the last approximation xn is less than the difference
between the root and the penultimate approximation.
Then, if the n-th iteration converges:
a  xn
g ( )   1; xn1    a
a  xn 1
g ( )  1; xn 1    a
In contrast, the method is divergent if
a  xn  a  xn 1
That is, if
g ( )  1; xn 1    a

Since the approximation xn moves more to the root to xn1 .

Finally, it says there is a deadlock if a  xn  a  xn 1 , For the penultimate


approximation is equal to the last (the method does not advance, but neither moves
away).
 
It should be noted that the criteria of the derivatives g ( )  1 and g ( )  1 to
analyze the convergence of the method are valid for one iteration, and for practical
purposes it is not possible to apply in each of them. For this reason, just that: there is
xn 1  xn  xn  xn 1
convergence if There is disagreement whether
xn 1  xn  xn  xn 1 and there is a deadlock if xn 1  xn  xn  xn 1 . That is, the
method converges if the absolute difference between the values provided in two
successive iterations is becoming smaller as n increases.
In summary, the geometric interpretation of the method is as follows: starting from an
initial value x0 directed vertically to the curve y = g (x) thereof, horizontally to the line
y = x, and again the curve vertically, horizontally to the line, etc.
The algorithm of the method is as follows:

1. Choose an initial approximation x0 .

2. Calculate g ( x0 ) and make x0  g ( x0 ) .


Let n = 1.

3. Calculate xn  g ( xn 1 ) .

4. Compare xn 1  xn with xn  xn 1 :

a) If xn 1  xn  xn  xn 1 The method converges. Go to step 5.


b) If xn 1  xn  xn  xn 1 , The method diverges. He stops the method and
choosing a new approach x0 .

c) If xn 1  xn  xn  xn 1 The method has stalled. He stops the method and


choosing a new approach x0 .
Note that the first iteration is not yet possible to apply this criterion, so you
omit this step and continues in May.

5. If xn  xn 1 = 0, found the root of the function (so the method). Otherwise,


go to step 6.

6. Let T the desired tolerance (the margin of error). If xn  xn 1 <T was found
closer to the root with a margin of error of less than T (end method).
Otherwise, go to step 3 by n = n +1.

Example. Determines the root of the equation e  x  0


x
the fixed point method
considering a tolerance of 0.001

Solution. We
f ( x)  e x  x
Then
g ( x)  f ( x)  x
 e x
g(x)=exp(x), g(x)=x

1
g(x)=exp(-x)
0.9
g(x)=x
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


x

The graph of the equation shows that the value of the root is close to 0.6, so we
choose the initial approximation x0  0.4

Estimate g ( x0 )
g ( x0 )  e 0.4  0.67
g(x)=exp(x), g(x)=x

1
g(x)=exp(-x)
0.9
g(x)=x
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


x

And we have x1  g ( x0 )  


g(x)=exp(x), g(x)=x

1
g(x)=exp(-x)
0.9
g(x)=x
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


x

Like x1  x0  0 , Ie has not been found and also the root x1  x0   is greater
than the desired tolerance, we make a new iteration.
The following table summarizes the results by applying the method. By comparing
differences x2  x1 and x1  x0 shows that x2  x1  x1  x0 Therefore
concluded that the method converges. The method is stopped at iteration 11 because
x11  x10  T , So it can be concluded that 0.56748681 is an approximation to the
value of the root with a margin of error of 0.1%.
n xn g ( xn ) xn  xn 1
0 0.4 0.67032005 -

1 0.67032005 0.51154483 0.27032005

2 0.51154483 0.59956863 0.15877521

3 0.59956863 0.54904843 0.0880238

4 0.54904843 0.57749908 0.0505202

5 0.57749908 0.56130038 0.02845065

6 0.56130038 0.57046676 0.0161987

7 0.57046676 0.56526154 0.00916638

8 0.56526154 0.56821152 0.00520522

9 0.56821152 0.56653778 0.00294998

10 0.56653778 0.56748681 0.00167374

11 0.56748681 0.5669485 0.00094903


2.1.3 Newton-Raphson Method

This method is part of a first approximation xn and by applying a recursive formula is


close to the root of the equation, so that new approach xn1 is located at the
intersection of the tangent to the curve of the function at the point xn and the x-axis.

y=f(x)

f(xn)

a xn+1 xn

We know that
tan( )  f ( xn )
And that by definition
f ( xn )
tan( ) 
xn  xn 1
Namely
f ( xn )
f ( xn ) 
xn  xn 1
And then
f ( xn )
xn 1  xn  ; n  0,1, 2,
f ( xn )
To determine where this method converges, we will use the same approach as in the
fixed point:
f ( x)
h( x )  x 
f ( x)

And if a  in the interval xn 1    a it holds that


h( )  1
then the method converges in the nth iteration.
For practical purposes we will apply the same criterion of convergence in the fixed
point method.
Although the method almost always converges to the solution in a small number of
iterations, are listed below some of the most common cases of divergence:

1. Vicious circle. When using xn is obtained by the tangent value xn1 that by
replacing the recursive formula returns the same value xn .
2. Indeterminacy. When we evaluate the formula at a point where the
function has a maximum or minimum.

3. Apparent discrepancy. When the initial approximation x0 is far from the


real value of the root, it is possible that in the first iteration the method and
provide apparently divergent values, however, the method would after a
few iterations to the solution.
The geometric interpretation of Newton-Raphson method is very similar to the fixed
point: from an initial approximation x0 , Addresses a vertical line to the curve y = f (x),
draw a tangent and is taken as x1 entering the intersection the tangent and the x-
axis, again, vertically to the curve, etc.
The algorithm of the method is:

1. Calculate f ( x) .

2. Choose an initial approximation x0 .


Let n = 0.
f ( xn )
xn 1  xn 
3. Assess f ( xn ) .

4. Compare xn 1  xn with xn  xn 1 :

a) If xn 1  xn  xn  xn 1 The method converges. Go to step 5.

b) If xn 1  xn  xn  xn 1 The method differs, apparently. He stops


the method and choosing a new approach x0 .

c) If xn 1  xn  xn  xn 1 The method has stalled. He stops the


method and choosing a new approach x0 .
Note that the first iteration is not yet possible to apply this criterion, so you
omit this step and continues in May.

5. If xn  xn 1 = 0, found the root of the function (end method). Otherwise,


go to step 6.
6. Let T the desired tolerance (the margin of error). If xn  xn 1 <T was
found closer to the root with a margin of error of less than T (end method).
Otherwise, go to step 3 by n = n +1.

Example. Get a positive root of the equation using the Newton-Raphson method and
considering a tolerance of 0.001.
2cos ( x)  e x  0

Solution. We must

f ( x)  2 sen( x)  e x
So our recursive formula is given by
2cos( xn )  e xn
xn 1  xn  ; n  0,1, 2,
2 sen( xn )  e xn
2 cos(x)-exp(x)

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
x

The chart can be seen that the value of the root is in the interval [0.2, 0.8] and we
choose 0.2, which represents no objection, since at that point there is not a maximum
or a minimum of the function.

Now evaluate the formula for finding the value of x1

It should be noted that the calculations of the trigonometric functions are given in
radians.
2cos(x)-exp(x)

f(x)

x
x0=0.2 x1=0.66

Like x1  x0  0 , Ie has not been found and also the root x1  x0   is greater
than the desired tolerance, we make a new iteration.
In the next chart you will find the results of each iteration. By comparing differences
x2  x1 and x1  x0 shows that x2  x1  x1  x0 Therefore concluded that the
method converges. The method found an approximation to the value of the root in
just four iterations 0.53978516, with a margin of error of 0.1%.

n xn f ( x) f ( x ) xn 1 xn 1  xn
0 0.2 0.7387304 -1.61874142 0.65636097 -
1 0.65636097 -0.34332806 -3.14824033 0.54730701 0.45636097
2 0.54730701 -0.02073363 -2.76937066 0.53982025 0.10905395
3 0.53982025 -9.6261E-05 -2.74366205 0.53978516 0.00748677
4 0.53978516 -2.1119E-09 -2.74354166 0.53978516 3.5085E-05

Real and complex roots of a polynomial


The methods seen so far allow to obtain the real roots of algebraic and
transcendental equations. However, none of them allow the calculation of the
complex roots of the same. This section is devoted to the study of two methods to
obtain the roots, both real and complex, of a polynomial.

Lin Method
This method is not in itself the roots of the polynomial, but an expression which can
be derived roots. The advantage of this method is that it can be obtained through all
the roots of the polynomial, whether real or complex.
Lin's method is to factor an equation of degree n in a quadratic polynomial by a
polynomial of degree n-2, so that the roots are obtained by pairs of quadratic factor,
and the procedure is repeated as necessary.
Sea P( x)  0 an algebraic equation of the form
P( x)  a0 x n  a1 x n 1  a2 x n  2    an 1 x  an ... (2.5)

Obtain a quadratic of the form factor


x 2  px  q
And again we express (2.5)
P( x)  ( x 2  px  q)(b0 x n 2  b1 x n 3  b2 x n  4    bn 3 x  bn 2 )  bn 1 x  bn ... (2.6)

where bn 1 x and bn are the residuals of the polynomial.

To determine the coefficients bi ; i  0,1, 2, , n  2 we make small polynomial


multiplication in (2.6)
P( x)  b0 x n  pb0 x n 1  qb0 x n  2  b1 x n 1  pb1 x n  2  qb1 x n 3 
b2 x n  2  pb2 x n 3  qb2 x n  4    bn 3 x 3  pbn3 x 2  qbn3 x 
bn  2 x 2  pbn 2 x  qbn 2  bn 1 x  bn ... (2.7)
Now, we equate the coefficients of these powers in (2.5) and (2.7)
a0  b0
a1  b1  pb0
a2  b2  pb1  qb0

an 1  bn 1  pbn  2  qbn 3
an  bn  qbn  2
And solving the polynomial coefficients reduced
b0  a0
b1  a1  pb0
b2  a2  pb1  qb0

bn 1  an 1  pbn  2  qbn 3
bn  an  qbn  2
So the reduced polynomial coefficients are given by
bk  ak  pbk 1  qbk  2 ; k  0,1, 2, , n  2; b1  b2  0
And the waste
bn1  an 1  pbn  2  qbn 3
bn  an  qbn  2

For x  px  q
2
is a factor of the polynomial P (x) is necessary bn 1 and bn are
equal to zero
0  an 1  pbn  2  qbn 3
0  an  qbn  2 ... (2.8)
Solving for p and q (2.8)
an 1  qbn 3
p
bn  2
a
q n
bn  2 ... (2.9)
If you know the values of p and q we can calculate the coefficients
bi ; i  0,1, 2, , n  2 reduced polynomial.

From initial values for p and q and through an iterative process these values are
determined with the precision required. To this end, we define the increments p
and q :
p  p *  p; q  q *  q ... (2.10)

Where p * and q * are the new approaches p and q, respectively, and are given by
(2.9)
an 1  qbn3
p* 
bn  2
a
q*  n
bn  2 ... (2.11)
Substituting (2.11) in (2.10)
an 1  qbn 3
p  p
bn  2
a
q  n  q
bn  2
So
an 1  pbn  2  qbn 3 bn 1
p  
bn  2 bn  2
a  qbn  2 b
q  n  n
bn  2 bn  2
Namely
bn 1
p*  p 
bn  2
b
q*  q  n
bn  2

The method converges when bn 1 , bn , p and q tend to zero, and for any of them
can fix the error tolerance.
Note that if bn  2  0 is not possible to apply the method.
The algorithm of the method consists of the following steps:
1. Make p = q = 0.
2. Calculate the polynomial coefficients reduced
bk  ak  pbk 1  qbk  2 ; k  0,1, 2, , n  2; b1  b2  0
And waste
bn1  an 1  pbn  2  qbn 3
bn  an  qbn  2

3. Verify bn  2  0 and calculate the new p and q approaches


bn 1
p*  p 
bn  2
b
q*  q  n
bn  2

If bn  2  0 concluded that it is not possible to apply the method to solve


polynomial in question.

4. Let T be the desired tolerance (the maximum permissible error). If p *  p <T


and q * q <T found approximations to the values of p and q with a margin of
error of less than T (end method). Otherwise it is necessary to make a new
iteration starting at step 2 by p = p * and q = q *.

Example. Get an approximation to the roots of the next polynomial using the method
of Lin considering a tolerance of 0.01 in the values of p and q and rounding to three
significant figures:

x 4  x3  6 x 2  3 x  4  0

Solution. We must
a0  1, a1  1, a2  6, a3  3, a4  4; n  4
Let p = q = 0 baseline.
The reduced polynomial coefficients are given by
b0  a0  1
b1  a1  pb0  (1)  (0)(1)  1
b2  a2  pb1  qb0  (6)  (0)(1)  (0)(1)  6
waste, by
b3  a3  pb2  qb1  (3)  (0)(6)  (0)(1)  3
b4  a4  qb2  (4)  (0)(6)  4
Like bn  2  b2  0 , The method can be applied and new approaches are
b3
p*  p   0.5
b2
b
q*  q  4  0.667
b2

We then p *  p = 0.5 and q * q = 0667 and it is necessary that these values are
lower than 0.01, so it is necessary to make a new iteration.
The results of each iteration are summarized in the following table:

i b0 b1 b2 b3 b4 p* q* p* p q * q

0 1 -1 6 -3 4 -0.5 0667 0.5 0667

1 1 -0.5 5.8 -0127 0612 -0525 0787 0025 0.12

2 1 -0475 4.96 -0.0222 0.0965 -0529 0806 0004 0019

3 1 -0471 4.94 -0.00711 0.0184 -0530 0810 0001 0.004

The polynomial P ( x)  x  x  6 x  3 x  4  0 can then be expressed as


4 3 2

P( x)  ( x 2  0.530  0.810)( x 2  0.471  4.94)  0.00711x  0.0184  0


As it reached the desired tolerance, we assume that the residue can be neglected, and
we have
P ( x)  ( x 2  0.530  0.810)( x 2  0.471  4.94)  0
Where it can be concluded that the roots of the polynomial are
x1  0.265  i 0.86, x2  0.265  i 0.86, x3  0.236  i 2.21, x4  0.236  i 2.21

Bairstow Method
This method relies on dividing by a factor quadratic polynomial. Let P (x) = 0, the
general polynomial of degree n of the form
P( x )  a0 x n  a1 x n 1  a2 x n  2    an 1 x  an
We know that to get the quadratic factor
x 2  px  q
we have
P( x)  ( x 2  px  q)(b0 x n 2  b1 x n 3  b2 x n  4    bn 3 x  bn 2 )  bn 1 x  bn

As in the method of Lin, we conclude that


b0  a0
b1  a1  pb0
b2  a2  pb1  qb0

bn 1  an 1  pbn  2  qbn 3
bn  an  qbn  2
and reduced the polynomial coefficients are given by
bk  ak  pbk 1  qbk  2 ; k  0,1, 2, , n  2; b1  b2  0
and waste
bn1  an 1  pbn  2  qbn 3
bn  an  qbn  2

Bairstow studied the possibility of finding approximations of waste bn 1 and bn


through a Taylor series for the independent variables p*  p  p and q*  q  q :
 bn 1 b
bn 1 ( p*, q*)  bn 1  p  n 1 q
p q
b b
bn ( p*, q*)  bn  n p  n q
p q
Equating to zero we
 bn 1 b
p  n 1 q  bn 1
p q

 bn b
p  n q  bn
p q

In this way, one can calculate the values of p and q to solve the system of linear
equations and, consequently, get the value of new approaches p*  p  p and
q*  q  q .

The algorithm of the method consists of the following steps:


1. Make p = q = 0.
2. Calculate the polynomial coefficients reduced
bk  ak  pbk 1  qbk  2 ; k  0,1, 2, , n  2; b1  b2  0
And waste
bn1  an 1  pbn  2  qbn 3
bn  an  qbn  2

3. Calculate the partial derivatives of waste bn 1 and bn :


 bn 1  bn 1  bn  bn
, , ,
 p q  p q
4. Solve the system
 bn 1 b
p  n 1 q  bn 1
p q

 bn b
p  n q  bn
p q

5. Get the values of the new approaches p*  p  p and q*  q  q .

6. Let T be the desired tolerance (the maximum permissible error). If p *  p <T


and q * q <T found approximations to the values of p and q with a margin of
error of less than T (end method). Otherwise it is necessary to make a new
iteration starting at step 2 by p = p * and q = q *.

Example. Get an approximation to the roots of the next polynomial using Bairstow's
method considering a tolerance of 0.01 in the values of p and q and rounding to three
significant figures:

x 4  x3  6 x 2  3 x  4  0

Solution. We must
a0  1, a1  1, a2  6, a3  3, a4  4; n  4
Let p = q = 0 baseline.
The reduced polynomial coefficients are given by
b0  a0  1
b1  a1  pb0  (1)  (0)(1)  1
b2  a2  pb1  qb0  (6)  (0)(1)  (0)(1)  6
waste, by
b3  a3  pb2  qb1  (3)  (0)(6)  (0)(1)  3
b4  a4  qb2  (4)  (0)(6)  4
and the partial derivatives by
 b3  b3
 b2  6.00;  b1  1.00
p q
 b4  b4
 0.00;  b2  6.00
p q
Solving the system
6.00p  1.00q  3.00
0.00p  6.00q  4.00

we have p  0.389 and q  0.667 . Then p*  p  p  0.389 and


q*  q  q  0.667

It is noted that p *  p = 0.389 and q * q = 0667 and it is necessary that these


values are lower than 0.01, so it is necessary to make a new iteration.
The results of each iteration are summarized in the following table:

i b0 b1 b2 b3 b4 p* q* p* p q * q

0 1 -1 6 -3 4 -0389 0667 0389 0667

1 1 -0611 5.10 -0609 0598 -0368 0784 0021 0117

2 1 -0632 4.98 -0672 0.0957 -0501 0765 0133 0019

3 1 -0499 4.99 -0118 0183 -0521 0802 0.02 0037

4 1 -0479 4.95 -0.0369 0.0301 -0528 0808 0007 0006

The polynomial P ( x)  x  x  6 x  3 x  4  0 can then be expressed as


4 3 2

P ( x)  ( x 2  0.528  0.808)( x 2  0.479  4.95)  0.0369 x  0.0301  0


As it reached the desired tolerance, we assume that the residue can be neglected, and
we have
P( x)  ( x 2  0.528  0.808)( x 2  0.479  4.95)  0
Where it can be concluded that the roots of the polynomial are
x1  0.264  i 0.86, x2  0.264  i0.86, x3  0.240  i 2.21, x4  0.240  i 2.21

Bibliography
 Terrence J. Akai, numerical methods applied to engineering. Mexico, Limusa
Wiley, 2004.
 Richard L. Burden & J. Douglas Faires, Numerical Analysis. 2nd. Ed., Mexico,
Grupo Editorial Iberoamérica, 1993.
 Steven C. Chapra & Raymond P. Canale, Numerical methods for engineers. 4th.
ed., Mexico, McGraw-Hill, 2003.
 R. Iriarte & V. Balderrama, numerical methods. Mexico, UNAM Faculty of
Engineering., Trillas, 1990.

Das könnte Ihnen auch gefallen