You are on page 1of 9

Newton’s Method

Suppose we want to solve x 4 + x − cos x = 0

Question: How can we solve for x without graphing ?

Cannot solve algebraically (no closed form solution), must use numerical
techniques:
I Bisection Method
I Secant Method
I False Position Method
I Newton’s Method "

Section 3.8 MATH 173


Newton’s Method

Newton’s Method for Approximating the Zeros(Roots) of a Function

Let c be a root of a function, ie f (c) = 0.

1. Initial estimate (guess) x1 , that is close to c.

2. Determine new approximation

f (xn )
xn+1 = xn −
f 0 (xn )

3. If |xn − xn+1 | < ε (within the desired accuracy), let xn+1 serve as the final
approximation. Otherwise, return to step 2 and calculate a new approximation.

Section 3.8 MATH 173


Example 1: Using Newton’s Method

Calculate three iterations of Newton’s Method to Approximate a zero of


f (x) = x 2 − 2. Use x1 = 1 as the initial guess.

f 0 (x) = 2x

xn2 − 2
Iteration Formula: xn+1 = xn −
2xn

f (xn ) f (xn )
n xn f (xn ) f 0 (xn ) xn − |xn − xn+1 |
f 0 (xn ) f 0 (xn )
1 1.000000 -1.000000 2.000000 -0.500000 1.500000 0.5
2 1.500000 0.2500000 3.000000 0.083333 1.416667 0.083333
3 1.416667 0.006945 2.833334 0.002451 1.414216 0.002451

Section 3.8 MATH 173


Example 2: Using Newton’s Method

Use Newton’s Method to approximate the zeros of f (x) = e x + x. Continue the


iteration until two successive approximations differ by less than 0.0001.

f 0 (x) = e x + 1

e xn + xn
Iteration Formula: xn+1 = xn −
e xn + 1

f (xn ) f (xn )
n xn f (xn ) f 0 (xn ) xn − |xn − xn+1 |
f 0 (xn ) f 0 (xn )
1 -0.60000 -0.05119 1.54881 -0.03305 -0.56695 0.03305
2 -0.56695 0.00030 1.56725 0.00019 -0.56714 0.00019
3 -0.56714 0.00000 1.56714 0.00000 -0.56714 0

Section 3.8 MATH 173


Conditions Where Newton’s Method Fails

From the previous examples, the approximations approach a limit


x1 , x2 , x3 , . . . , xn , . . . ⇒ c
 
f (xn )
lim xn − = c (Sequence Converges)
n→∞ f 0 (xn )

If the limit is c, then c is a zero of f .


However, Newton’s Method does not always guarantee convergence.
I Newton’s Method involves division by f 0 (xn ).
I Possible division by zero
I Choose a different initial guess, x1 .
I Another way Newton’s Method can fail is shown in the next example.
I The sequence x1 , x2 , x3 , . . . does not converge.
I Converges to the wrong value (some functions have more than one zero).
I Choose your initial value close to the zero you sought.

Section 3.8 MATH 173


Example 3: An Example in Which Newton’s Method Fails

The function f (x) = x 1/3 is not differentiable at x = 0. Show that Newton’s


Method fails to converge using x1 = 0.1.

f 0 (x) = 31 x −2/3

1/3
xn
Iteration Formula: xn+1 = xn − 1 −2/3
= −2xn
x
3 n

f (xn ) f (xn )
n xn f (xn ) f 0 (xn ) xn − |xn − xn+1 |
f 0 (xn ) f 0 (xn )
1 0.10000 0.46416 1.54720 0.30000 -0.20000 0.30000
2 -0.20000 -0.58480 0.97467 -0.60000 0.40000 0.60000
3 0.40000 0.73681 0.61401 1.20000 -0.80000 1.20000
4 -0.80000 -0.92832 0.38680 -2.40000 1.60000 2.40000

Section 3.8 MATH 173


Example 3: An Example in Which Newton’s Method Fails

Section 3.8 MATH 173


Condition for Convergence

Condition (sufficient) for convergence of Newton’s Method to a zero of f


f (x)f 00 (x)

[f 0 (x)]2 < 1 (1)

on an open interval containing zero.

From Example 1: f (x) = x 2 − 2, f 0 (x) = 2x, f 00 (x) = 2

f (x)f 00 (x) (x 2 − 2)(2) 1



= − 1 < 1.

=
[f 0 (x)]2 4x 2 2 x2

√ √
6 6
∴ Any initial guess chosen from (−∞, − 3
) ∪( 3
, +∞) guarantees
convergence.

Section 3.8 MATH 173


Condition for Convergence

1 −2/3 00 2
From Example 3: f (x) = x 1/3 , f 0 (x) = x , f (x) = − x −5/3
3 9

f (x)f 00 (x) x 1/3 (−2/9)(x −5/3 )



[f 0 (x)]2 (1/9)(x −4/3 ) = 2.
=

∴ Cannot conclude that Newton’s Method will converge.

Section 3.8 MATH 173

You might also like