Example 3. Use
Newton's method to construct a linearly convergent
sequence
which
converges slowly to the multiple root
of
.
Then use the Aitken
process to construct
which
converges faster to the root
.
Solution 3.
Graph the function.
![[Graphics:../Images/AitkenSteffensenMod_gr_121.gif]](../Images/AitkenSteffensenMod_gr_121.gif)
Starting with
,
use the Newton-Raphson method to find a numerical approximation to
the root.
![[Graphics:../Images/AitkenSteffensenMod_gr_125.gif]](../Images/AitkenSteffensenMod_gr_125.gif)
Since we know the root is
, we
can determine the error for each iteration.
![[Graphics:../Images/AitkenSteffensenMod_gr_128.gif]](../Images/AitkenSteffensenMod_gr_128.gif)
Newton's method is converging linearly (or slowly), the error at
each step is being reduced by approximately one-half. Let
us apply Aitken's acceleration process to a
sequence
of
iterations generated by Newton's method.
![[Graphics:../Images/AitkenSteffensenMod_gr_131.gif]](../Images/AitkenSteffensenMod_gr_131.gif)
Again, we can determine the error for each term.
![[Graphics:../Images/AitkenSteffensenMod_gr_133.gif]](../Images/AitkenSteffensenMod_gr_133.gif)
The sequence
is
converging to p faster than the
sequence
converges
to p.
![[Graphics:../Images/AitkenSteffensenMod_gr_137.gif]](../Images/AitkenSteffensenMod_gr_137.gif)
(c) John H. Mathews 2004