Consider derivative of the convolution of a given function $f(\cdot)$ with a fixed $C^\infty$ function $s(\cdot)$, evaluated say at $1/2$. Is there a measure which generates the functional so defined?
1 Answer
For your linear functional (say $L$) and all $f\in C[0,1]$ we have $$L(f)=(f*s)'(1/2)=\int_0^1 f(y)s'(1/2-y)\,dy,$$ so that $L$ is generated by the signed measure $\mu(dy)=s'(1/2-y)\,dy$.
The question in your title, "Are ALL linear functionals on C[0,1] generated by measures?", is different from the one in the body of your post, and the answer to the "title" question is negative, assuming the axiom of choice. Indeed, then $C[0,1]$ has a Hamel basis $(e_i)_{i\in I}$, which is infinite (and even uncountable). Let $J=\{i_1,i_2,\dots\}$, with pairwise distinct $i_1,i_2,\dots$, be a countable subset of $I$. Define the linear functional $K$ on $C[0,1]$ by the formulas $$K(e_{i_k}):=k\|e_{i_k}\|$$ for natural $k$, with $K(e_i):=0$ for $i\in I\setminus J$. Then the linear functional $K$ is not bounded and hence not continuous. So, $K$ is not generated by any signed measure.
-
$\begingroup$ Is there a simple way to prove that C[0,1]* is isomorphic to the space of signed measures? $\endgroup$ Commented Nov 29 at 18:44
-
6$\begingroup$ @HTomaszGrzybowski : I think the original proof by Riesz is quite simple. This answers the third question you have asked on this page. According to MathOverflow guidelines, users should refrain from answering posts that "request answers to multiple questions". Also, these other guidelines may be relevant here. $\endgroup$ Commented Nov 29 at 18:59
-
$\begingroup$ @HTomaszGrzybowski : Do you have a further response to my answer and comment? $\endgroup$ Commented Dec 1 at 15:14