You are on page 1of 6

Chapter 2

2.1 Solution of equations of one variables

The problem
To find an approximate value 𝒑𝒊 to a root 𝝀 of
𝒇(𝒙) = 𝟎
to within a given accuracy 𝜺.

Example 1
Consider the equation 𝒙𝟐 = 𝟐 with the positive solution
𝝀 = √𝟐 = 𝟏. 𝟒𝟏𝟒𝟐𝟏 …
𝒑𝟐 = 𝟏. 𝟒𝟑𝟓𝟒~𝝀 with absolute error < 10−1 .
𝒑𝟓 = 𝟏. 𝟒𝟏𝟒𝟔~𝝀 with absolute error < 10−3 .

Existence of a solution
Let 𝒇 be continuous function on [𝒂, 𝒃] such that
𝒇(𝒂)𝒇(𝒃) < 0. Then ∃𝝀 ∈ [𝒂, 𝒃] such that 𝒇(𝝀) = 𝟎
Example 2
Locate the roots of the following equations on the given
interval:
a) 𝒙𝟑 + 𝟐𝒙𝟐 − 𝟑𝒙 = 𝟏 on [−𝟐, 𝟐]

b) (𝒙 − 𝟐)𝟐 − 𝒍𝒏𝒙 = 𝟎 on (𝟎, 𝟑)

1
2.3 Newton’s Method
Given a good approximation 𝒑𝟎 to a root 𝝀 of 𝒇(𝒙) = 𝟎.
𝒇(𝒑𝒏 )
Then 𝒑𝒏+𝟏 = 𝒑𝒏 − , 𝒏 = 𝟎, 𝟏, 𝟐, … .
𝒇′ (𝒑𝒏 )

Example 1
Use the Newton’s method to find an approximate solution
to within 𝟏𝟎−𝟓 accuracy for the equation 𝒙𝟐 = 𝟐 on the
interval [𝟎, 𝟐] , using 𝒑𝒐 = 𝟏 .

𝒑𝒏 𝒇(𝒑𝒏 ) 𝒇′(𝒑𝒏 ) 𝒇(𝒑𝒏 )/𝒇′(𝒑𝒏 )

2
Example 2
Use Newton’s methods to find approximate solution to
within 𝟏𝟎−𝟏𝟎 accuracy for the first positive root of the
equation 𝒕𝒂𝒏𝒙 = 𝒙.

a) Choosing 𝒑𝒐 = 𝟒. 𝟕.
n 𝒑𝒏 𝒇(𝒑𝒏 ) 𝒇′(𝒑𝒏 ) 𝒇(𝒑𝒏 )/𝒇′(𝒑𝒏 )

b) Maple 𝒇𝒔𝒐𝒍𝒗𝒆 command.

3
2.4 Rate of Convergence of Numerical Methods
Let {𝒑𝒏 } be a sequence of real numbers generated by a
numerical method which converges to a solution 𝝀.
The sequence converges linearly if, for large value of 𝒏,
|𝒑𝒏+𝟏 − 𝝀| < 𝑀|𝒑𝒏 − 𝝀| .
𝒆𝒏+𝟏
i.e. ~𝑴
𝒆𝒏
Note that the bisection method generates a linear
convergence sequence of approximations

The sequence converges quadratically if, for large value of


𝒏,

|𝒑𝒏+𝟏 − 𝝀| < 𝑀|𝒑𝒏 − 𝝀|𝟐 .


𝒆𝒏+𝟏
i.e. ~𝑴
𝒆𝒏 𝟐

Example 1
Let 𝑀 = 0.5, 𝑒0 = 0.2. Estimate the errors in the 3rd
iteration using:
a) A linear convergence method

b) A quadratic convergence method

4
Note that in general, Newton’s method generates a
quadratic convergence sequence of approximations.

Example 2
Show that Newton’s method generate a quadratic
convergence sequence of approximate solutions for the
positive root of the equation 𝒙𝟐 − 𝟑 = 𝟎.

Example 3
Show that Newton’s method generate a linear
convergence sequence of approximate solutions for the
root 𝝀 = 0 of the equation 𝑒 𝑥 − 𝒙 − 𝟏 = 𝟎.

5
Note
Newton’s method generates a quadratic convergence
sequence of approximations if the root is simple.

Theorem
Let {𝒑𝒏 } be a sequence generated by Newton’s method to
approximate a simple root 𝝀 of 𝒇(𝒙) = 𝟎. Then
𝑴
|𝒑𝒏+𝟏 − 𝝀| < |𝒑𝒏 − 𝝀|𝟐
𝟐|𝒇 (𝝀)|

You might also like