= Hardy-Littlewood maximal function
{c}
{title2=$Mf$}
{wiki=Hardy–Littlewood_maximal_function}
For a <locally integrable function> $f$ on $\mathbb R^d$, the centred <Hardy-Littlewood maximal function> is
$$
Mf(x)=\sup_{r>0}\frac1{|B_r(x)|}\int_{B_r(x)}|f(y)|\,dy.
$$
On the real line the balls are intervals, giving $Mf(x)=\sup_{r>0}(2r)^{-1}\int_{x-r}^{x+r}|f(t)|dt$. It records the largest absolute average visible at any scale around $x$; the absolute value prevents cancellation from hiding large values.
For $f=\mathbf1_{[0,1]}$ on the line, $Mf(x)\leq1$, and it equals1 at every interior point. Far outside that interval it decays only on the order of $1/|x|$, because a large centred interval can still capture the unit mass. Thus an integrable input need not have an integrable maximal function. The correct endpoint bound is the <Hardy-Littlewood maximal inequality> of weak type$(1,1)$, while <Strong Lp bound for the Hardy-Littlewood maximal function> holds for $p>1$. These bounds control the exceptional sets in the <Lebesgue differentiation theorem>.
Back to article page