Solution.
Since eitx is bounded and μ is a finite measure, we may apply Fubini's theorem in the following:
2T1∫−TT∣φ(t)∣2dt=2T1∫−TT(∫Reitxdμ(x))(∫Reitydμ(y))dt=2T1∫−TT∫R∫Reit(x−y)dμ(x)dμ(y)dt=2T1∫R∫R∫−TTeit(x−y)dtdμ(x)dμ(y)=∫R∫RT(x−y)sin(T(x−y))dμ(x)dμ(y).
Again, because the integrand is continuous and μ is finite, we may apply dominated convergence:
T→∞lim2T1∫−TT∣φ(t)∣2dt=∫R∫RT→∞limT(x−y)sin(T(x−y))dμ(x)dμ(y).
If x=y, then clearly the integrand tends to 0. If x=y, then we get 1, so this becomes
T→∞lim2T1∫−TT∣φ(t)∣2dt=∫R∫Rχ{x=y}(x,y)dμ(x)dμ(y)=∫Rμ({y})dμ(y)=0,
since singletons have μ-measurable zero, and this is what we wanted to show.