Hardy-Littlewood maximal inequality
Let f∈Lloc1(Rn) and let
Mf(x)=r>0supm(B(r,x))1∫B(r,x)∣f(y)∣dy
be the Hardy-Littlewood maximal function.
-
Show that
m({x∣Mf(x)>s})≤sCn∫{∣f(x)∣>s/2}∣f(x)∣dx,s>0,
where the constant Cn depends on n only. The Hardy-Littlewood maximal theorem may be used.
-
Prove that if φ∈C1(R), φ(0)=0, and φ′>0, then
∫φ(Mf(x))dx≤Cn∫∣f(x)∣(∫{0<t<2∣f(x)∣}tφ′(t)dt)dx.
Solution.
-
Let g=fχ{∣f(x)∣>s/2} and h=fχ{∣f(x)∣≤s/2} so that f=g+h. Observe that for all x∈Rn and r>0, we have
m(B(x,r))1∫B(x,r)∣h(y)∣dy≤2s⟹Mh(x)≤2s.
Thus, {x∣Mh(x)>2s}=∅, and so {x∣Mf(x)>s}⊆{x∣Mg(x)>2s}. Hence,
m({x∣Mf(x)>s})≤m({x∣Mg(x)>2s})≤s/2Cn′∥g∥L1=sCn∫{∣f(x)∣>s/2}∣f(x)∣dx.(Hardy-Littlewood maximal inequality)
-
By the fundamental theorem of calculus and the fact that φ(0)=0,
∫φ(Mf(x))dx=∫Rn∫0Mf(x)φ′(t)dtdx=∫Rn∫0∞χ{0<t<Mf(x)}φ′(t)dtdx=∫0∞∫Rnχ{0<t<Mf(x)}φ′(t)dxdt=∫0∞m({0<t<Mf(x)})φ′(t)dt≤∫0∞φ′(t)tCn∫{∣f(x)∣>t/2}∣f(x)∣dxdt=Cn∫tφ′(t)∫χ{0<t<2∣f(x)∣}∣f(x)∣dxdt=Cn∫∣f(x)∣(∫{0<t<2∣f(x)∣}tφ′(t)dt)dx,(Fubini-Tonelli)(1)(Fubini-Tonelli)
which completes the proof.