Sharp estimates of the modified Hardy-Littlewood maximal operator on the nonhomogeneous space via covering lemmas (Q2574441)

Let \((X, d)\) be a separable metric space with a Radon measure \(\mu\) such that \(0<\mu(B(x,r))<\infty\) for any ball \(B(x,r)\) \((r>0)\). Relating to the recent works on non-homogeneous spaces (doubling condition: \(\mu(B(x,2r))\leq C\mu(B(x,2r))\) is not assumed) by Nazarov, Treil and Volberg, and X. Tolsa, the author considers the weak \((1,1)\)-property of the modified Hardy-Littlewood maximal operator \[ M_{\mu,k}f(x)=\sup_{r>0}\mu(B(x, kr))^{-1}\int_{B(x,r)}| f(y)| d\mu(y). \] His main result is: \(M_{\mu,2}\) is weak \((1,1)\)-bounded. And \(2\) is optimal. To prove this, he refines Vitali's covering lemma: Let \(\delta>0\). If \(B(x_j, r_j)\) \((j=1,\ldots,n)\) are balls, then one can select \(\{B(x_j,r_j)\}_{j\in A}\) such that (1) \(\{B(x_j,r_j)\}_{j\in A}\) is disjoint, (2) \(\cup_{j=1}^n B(x_j, \delta r_j)\subset \cup_{j\in A}B(x_j, (2+\delta)r_j)\). (The case \(\delta=1\) is the usual Vitali's covering lemma.) In the Euclidean case, as is easily checked, \(M_{\mu,1}\) is weak \((1,1)\) bounded (using Besicovitch covering lemma), but if we consider the uncentered maximal operator \[ \tilde M_{\mu,k}f(x) =\sup_{x\in B(y,r)}\mu(B(y, kr))^{-1}\int_{B(x,r)}| f(y)| d\mu(y), \] as is known, \(\tilde M_{\mu,1}\) is not weak \((1,1)\) bounded. He gives weak \((1,1)\)-boundedness of \(\tilde M_{\mu,k}\) for \(k>1\). To prove this, he gives a new covering lemma: For any \(k>1\) there exists an integer \(N\) such that if \(\{B(x_\lambda, r_\lambda)\}_{\lambda\in L}\) is a family of balls in the Euclidean space such that \(\sup_{\lambda\in L}r_\lambda<\infty \), then one can take disjoint subfamilies \(\{B(x_\rho, r_\rho)\}_{\rho\in L_j}\) \((j=1,\ldots, N)\) such that \(\cup_{\lambda\in L}^n B(x_\lambda, r_\lambda)\subset \cup_{j=1,\ldots, N}\cup_{\rho\in L_j}B(x_\rho, kr_\rho)\). He also discusses the Fefferman-Stein type vector-valued inequality for \(M_{\mu,k}\).