We begin by stating the dominated convergence theorem. This may be found in any standard book that has measure theory - e.g. Chung's A Course in Probability Theory, or Royden's Real Analysis.
Now we state and prove the theorem of interest.
Proof. For convenience, let =
.
Suppose
, then by the mean
value theorem (Theorem 5.10, p. 108 of Rudin),
if
and
,
then
for some , and in particular for all
,
Now let be any sequence in
converging to
0. Then by linearity of the integral,
Thus, we have for each fixed
, the sequence of functions
converges -a.e. to
, and by (
),
,
-a.e., and G is
-integrable by
assumption.
Hence, by the dominated convergence theorem (Theorem
),
,
i.e.
Since the sequence
0 was arbitrary,
it follows that
(See e.g. Theorem 4.2, p. 84 of Rudin for this result which
extends a limit from an arbitrary sequence to a continuous limit.)
This last display states that
is differentiable and the derivative is
the r.h.s. This proves the Theorem.