WOOD
one can define the Fourier transform F of T by p(x) = s-“m eiZt dK(t), where
K(t) is the function of bounded variation associated with T. This definition
generalizes the Fourier transform of an L1 function, for if f E L1, the function
w = sLof(4 d u is of bounded variation and f(x) = l-“m eiat dF(t).
In the paper, we investigate some generalizations of these ideas. We show,
for instance, that any closed operator defined on a dense subspace of L1 and
commuting with convolution (in a sense to be made precise later) is necessarily
linear and has a Fourier transform (which includes the Fourier transforms
mentioned above). With the aid of a theorem of Bochner, new necessary and
sufficient conditions are derived in order that a continuous function on the
real line be the Fourier transform of a function of bounded variation. The
ideas considered here are also closely related to “factor theory” for Fourier
transforms. This connection, as well as the above mentioned results, are
contained in Section 2. The brief second section contains an application to
semigroup theory.