Example 6.  Consider the function  [Graphics:Images/NewtonAccelerateMod_gr_334.gif].  
6 (a).  Use Newton's method to find the multiple root  [Graphics:Images/NewtonAccelerateMod_gr_335.gif].  

Solution 6 (a).

[Graphics:../Images/NewtonAccelerateMod_gr_338.gif]


[Graphics:../Images/NewtonAccelerateMod_gr_339.gif]

Graph the function.

[Graphics:../Images/NewtonAccelerateMod_gr_340.gif]


[Graphics:../Images/NewtonAccelerateMod_gr_341.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_342.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_343.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_344.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_345.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_346.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_347.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_348.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_349.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_350.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_351.gif]


It appears that round off error in the calculation of f[x] is introducing quite a bit of round-off error or noise.
Let's see what happens!

The Newton-Raphson iteration formula  g[x]  is

[Graphics:../Images/NewtonAccelerateMod_gr_352.gif]


[Graphics:../Images/NewtonAccelerateMod_gr_353.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_354.gif]

Investigate linear convergence at the double root  [Graphics:../Images/NewtonAccelerateMod_gr_355.gif],  using the starting value  [Graphics:../Images/NewtonAccelerateMod_gr_356.gif]

First, do the iteration one step at a time.  
Type each of the following commands in a separate cell and execute them one at a time.

[Graphics:../Images/NewtonAccelerateMod_gr_357.gif]
[Graphics:../Images/NewtonAccelerateMod_gr_358.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_359.gif]
[Graphics:../Images/NewtonAccelerateMod_gr_360.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_361.gif]
[Graphics:../Images/NewtonAccelerateMod_gr_362.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_363.gif]
[Graphics:../Images/NewtonAccelerateMod_gr_364.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_365.gif]
[Graphics:../Images/NewtonAccelerateMod_gr_366.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_367.gif]
[Graphics:../Images/NewtonAccelerateMod_gr_368.gif]

[Graphics:../Images/NewtonAccelerateMod_gr_369.gif]
[Graphics:../Images/NewtonAccelerateMod_gr_370.gif]

Notice that convergence is slow, but the sequence is converging to the multiple root  [Graphics:../Images/NewtonAccelerateMod_gr_371.gif].  

[Graphics:../Images/NewtonAccelerateMod_gr_372.gif]



[Graphics:../Images/NewtonAccelerateMod_gr_373.gif]
[Graphics:../Images/NewtonAccelerateMod_gr_374.gif]

Observe.  The function value is very close to zero, and the approximation [Graphics:../Images/NewtonAccelerateMod_gr_375.gif]  is a long way from the root !

At the multiple root  [Graphics:../Images/NewtonAccelerateMod_gr_376.gif]  we can explore the relationship  [Graphics:../Images/NewtonAccelerateMod_gr_377.gif]  for  k  sufficiently large.

This will be done by investigating the ratio  [Graphics:../Images/NewtonAccelerateMod_gr_378.gif]  for  k  sufficiently large.

[Graphics:../Images/NewtonAccelerateMod_gr_379.gif]


[Graphics:../Images/NewtonAccelerateMod_gr_380.gif]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(c) John H. Mathews 2004