Documente Academic
Documente Profesional
Documente Cultură
ek = xk − x∗ ,
where x∗ is the exact solution, i.e., f (x∗ ) = 0. To figure out the convergence order (which
will turn out to be superlinear) we have to find a relation between en+1 and en . As in the
corresponding derivation for Newton’s method we use Taylor’s theorem to write:
1
f (xn ) = f (x∗ + (xn − x∗ )) = f (x∗ + en ) = f (x∗ ) + f 0 (x∗ )en + f 00 (x∗ )e2n + O(e3n ).
2
In Eq. 1 f (xn−1 ) also appears and we write similarly
1
f (xn−1 ) = f (x∗ +(xn−1 −x∗ )) = f (x∗ +en−1 ) = f (x∗ )+f 0 (x∗ )en−1 + f 00 (x∗ )e2n−1 +O(e3n−1 ).
2
Furthermore we have
Subtracting x∗ from both sides of Eq. 1, and keeping in mind that by definition we have
f (x∗ ) = 0, gives then
which can be rewritten using (e2n + e2n−1 ) = (en − en−1 )(en + en−1 ) as
1
is of the form
en+1 = ρn en
with
f 00 (x∗ )
ρn = en−1 ,
2f 0 (x∗ )
so if the method converges, i.e., limn→∞ en = 0, we have limn→∞ ρn = 0, so we have
established superlinear convergence, provided f 0 (x∗ ) 6= 0.
To establish the order of convergence let us postulate that
en = Cern−1 ,
2
en+1 = Cern = C(Cern−1 )r = C r+1 ern−1 .
Substituting this in Eq. 4 gives
2
−r−1
lim C r ern−1
n→∞
= C 0.