A standard result in real analysis says that if $f : (\mathbb{R}, +) \rightarrow (\mathbb{C}^*, *)$ is a Lebesgue measurable (group) homomorphism with $|f| = 1$, then $$(\exists ~\xi \in \mathbb{R})~ f(x) = e^{\large 2\pi{}i\langle \xi, x\rangle},~ \forall x \in \mathbb{R}.$$ The only proof I know uses the Lebesgue Differentiation Theorem to establish that there exists an $a > 0$ such that $$\int_0^a f(x)dx \neq 0.$$ I'd like to see an alternative proof of this statement, that such an $a$ exists, using a more elementary argument.
For instance, if $f$ is continuous, then this is very easy to prove. Thus, following one of Littlewood's three principles of analysis (namely, that every measurable function is almost continuous), one should be able to extend this to measurable functions, say, via Lusin's Theorem, but so far I haven't been able to do so.
Any suggestions?