Solution.
First, assume that f=χE. Assume otherwise, and that χE=0 almost everywhere, i.e., m(E)>0. Then
∫R∫R∣x−y∣2+ε2∣f(x)∣∣f(y)∣dxdy=∫R∫R∣x∣2+ε2χE(x+y)χE(y)dxdy=∫R∣x∣2+ε21∫RχE(x+y)χE(y)dydx.(x↦x+y)
Since translation is continuous in L1(R), we see that
x→0lim∫RχE(x+y)χE(y)dy=x→0lim∫EχE(x+y)dy=∫EχE(y)dy=m(E)>0.
Thus, there exists δ>0 so that if ∣x∣<δ, then ∫RχE(x+y)χE(y)dy≥2m(E). Hence, if ε<1,
∫R∣x∣2+ε21∫RχE(x+y)χE(y)dydx≥2m(E)∫B(0,δ)∣x∣2+ε21dx=m(E)∫0δx2+ε1dx=εm(E)arctan(εδ)≥εm(E)arctan(δ)ε→0∞.
Hence, if f=χE, then it must be that m(E)=0. For the general case, suppose f=0, and so there exist a>0 and E⊆R measurable such that ∣f∣≥aχE. Then
∫R∫R∣x−y∣2+ε2∣f(x)∣∣f(y)∣dxdy≥a2∫R∫R∣x−y∣2+ε2χE(x)χE(y)dxdyε→0∞.
Thus, f=0 almost everywhere, which was what we wanted to show.