24-Newton’s Method, Rolle’s & Mean Value Theorem VU
Lecture # 24
Newton’s Method, Rolle’s Theorem, and the Mean Value Theorem
• Newton’s method for approximating solutions to f(x)=0
• Some difficulties with Newton’s method
• Rolle’s theorem
• Mean Value Theorem
Newton’s method for approximating solutions to f (x) =0
b
We have seen in algebra that the solution for the equation ax + b =0 is x = − .
a
Similarly we have algebraic formulas for polynomial equation up to degree 5.
But there is no algebraic solution to the equation of the kind
x − cos( x) =
0
For this equation and many like it, we settle for approximate solutions. How do we approximate the
solutions? There are many methods, and one of them is Newton’s method. Here is how Newton's
method works to find approximate solutions to equations.
What does it mean for an equation to have a solution? or that f (x) = 0 ?
It means that we are looking for those x values, for which the corresponding y value or f (x) is 0.
This means that the solutions are those points where the graph of the function crosses the x axis.
Suppose that x = r is the solution we are looking for.
Let's approximate it by an initial guess called.
We draw a line tangent to the graph of the given function at the point.
If the tangent line is not parallel to the x-axis, then it will x2 eventually intersect the x-axis at some
point x1 which will generally be closer to r than We will repeat the process, with a tangent line at x2
that meets x axis at x3, and so on…
This is Newton’s method.
We need a formula for this method.
Note that the tangent line at x1
has the equation
y − f ( x= 1) f '( x1 )( x2 − x1 )
If f '( x1 ) ≠ 0 , then the
line meets the x axis at ( x2 , 0)
Plug this coordinate into
above equation, we get
− f ( x=
1) f '( x1 )( x2 − x1 )
f ( x1 )
⇒ x2 = x1 −
f '( x1 )
Repeating this process for a third point ( x3 , 0) gives
− f ( x=
2) f '( x2 )( x3 − x2 )
f ( x2 )
⇒ x3 = x2 −
f '( x2 )
© Copyright Virtual University of Pakistan 164
24-Newton’s Method, Rolle’s & Mean Value Theorem VU
In general, then we have
f ( xn )
xn +=
1 xn −
f '( xn )
There is limiting process involved here for finding the solutions. We get as close as we like to the
solution.
Example
The equation x = cos( x) has a solution between 0 and 1. Approximate it using Newton’s method.
Rewrite as x = cos( x) so our function is f ( x)= x − cos( x) . The derivative is
f / ( x) = 1 + sin( x)
So we have
xn − cos( xn )
xn += xn −
1 + sin( xn )
1
As our approximation formula
Here is a graph of the situation. From the graph it looks like the solution is closer to 1 than 0. So we
will use x 1 = 1.So we get the following approximations.
x − cos( x1 ) 1 − cos(1)
x2 =
x1 − 1 =
1− =
0.7503
1 + sin( x1 ) 1 + sin(1)
x − cos( x2 ) 0.7503 − cos(0.7503)
x3 =
x2 − 2 =
0.7503 − =
0.7391
1 + sin( x2 ) 1 + sin(0.7503)
You may continue if you will, but we will say that the solution is approximately x ≈ 0.7391
Some difficulties with Newton’s method
• Newton’s method does not always work.
• If for some values of n f / ( xn ) then the formula for Newton’s method involves division
by 0 and we are out of business.
© Copyright Virtual University of Pakistan 165
24-Newton’s Method, Rolle’s & Mean Value Theorem VU
• Such a case will occur if the tangent line for some approximation has slope 0 or is parallel to
the x axis.
• Sometimes the approximations don’t converge to a solution.
1
Consider the equation x 3 = 0
The only solution is x = 0. Let's approximate it by Newton’s Method with initial approx x 1 = 1. We
get the following Formula
1
xn +1 =
x −
( xn ) 3 =
−2 xn
n 2
1
( xn ) 3
−
3
Plug in x1 =1 and then the following approx to see that the values do not converge
Rolle’s theorem
Rolle’s Theorem says essentially that for a certain kind of function,
if it crosses the x-axis at two point, then there is one point between
those two points where the derivative of f is 0.
Example
The function f ( x) = sin( x) is continuous and differentiable everywhere, hence continuous on [0 ,
2π] and differentiable on (0, 2π). Also, f (0) = sin (0) = 0 and
f(2π)=sin (2π) = 0
So the function satisfies the hypotheses of Rolle’s Theorem. So there exists a point c in the interval (0
, 2π) such that
= =
f / (c) cos( c) 0
f (b) − f (a )
y=
− f (a) ( x − a)
b−a
⇒
f (b) − f (a )
=y ( x − a) + f (a)
b−a
Here is a more tangible way to think of Roll’s theorem
© Copyright Virtual University of Pakistan 166
24-Newton’s Method, Rolle’s & Mean Value Theorem VU
I leave from Lahore to Islamabad.
When i start driving from Lahore, my velocity is 0, and when i reach Islamabad, my velocity is 0 as
well.
Velocity is a continuous function on the interval [0 , 376].
Also, velocity is differentiable on (0 , 376) as its derivative acceleration is defined at each point on the
velocity curve.
Hence during my drive from Lahore to ISB, there is some point on the motorway where the
acceleration of the car was 0.
One could argue: What if you keep accelerating on the motorway!
Mean Value Theorem
This says basically that under the right conditions, a function will have the same slope for the
tangent line at a point as that of a certain secant line.
Proof of MVT
From this figure we have the following:
Slope of Secant line joining A and B:
f (b) − f (a )
v( x)= f ( x) − ( x − a) + f (a)
b−a
Since f(x) is continuous [a , b] and differentiable on (a , b), so is v(x) by its formula involving f(x).
Also note that
=v(a ) 0=
and v(b) 0
So v(x) satisfies the assumptions of Roll’s theorem on the interval [a , b]. So there is a point c in (a ,
b) such that v / ( c)=0. But note that
© Copyright Virtual University of Pakistan 167
24-Newton’s Method, Rolle’s & Mean Value Theorem VU
f (b) − f (a )
v=
/
( x) f / ( x) −
b−a
f (b) − f (a )
⇒ v / (c ) = f / (c ) −
b−a
So by this last formula, at the point where v / (c) = 0 , we have
f (b) − f (a )
f / (c ) =
b−a
© Copyright Virtual University of Pakistan 168
