632 Series
when jxj < R. Persons willing to go fishing even when no fish are caught can
try to prove the result with rudimentary equipment. To simplify writing,
let a = 0. Then, when txol < R and Ix( < R,
n
f(x) Ckxk = Jim I Ckxk,
k=0 n-- k=0
m n
f(xo) = ckxo = lim ckxo.
k=0 n-"O k=0
so
f(x') - f(xo) _
n
lim ck(xx - xo)x
n-+GO k=1
and
n
f(x) - f(xo)= lim C c1
+
ck(xk-' + xk-2xo +... + x0
X - xo n-- k=2I
If we can prove that the limits exist, we can take limits as x approaches xo to
obtain
n
f'(xo) = lim lim 1c, + ck(xk-1 + xk-2xo +. .. + xo-1).
x-.xo n-- - k=2
If we can prove that the same result is obtained by interchanging the order in
which limits are taken, we obtain
n
f'(xo) = lim lim I C1 + I Ck(xk-' + xk-2xo + + 40-1)
n-+m x--.x0 k-2
and hence
m
f'(xo) = lim Icl + kckxo-11= kckxo-1.
n-- k=2 k=1
Our fishing expedition can be ended with the remark that we ran into questions
involving iterated limits and change of order of limits that swamped us.
12.5 Taylor formulas with remainders In Section 12.4, we started
with given convergent power series and found that these series are the
Taylor series of the functions to which they converge. In this section
we start with a given function f and study the general aspects and further
applications of a method we have previously employed in special cases
to obtain power series expansions of ex, cos x, and sin x. We suppose
that a and x are confined to an interval over which f has all of the con-
tinuous derivatives we want to use. Then
(12.51) f(x) = f(a) + fax f' (t) dt.