Maximal function estimates and self-improvement results for Poincaré inequalities (Q1755525)

Let \(X=(X, d, \mu)\) be a metric space equipped with a metric \(d\) and a~positive complete Borel measure \(\mu\), which satisfies \(0<\mu(B) <\infty\) for any ball \(B=B(x,r)=\{y \in X: d(x,y)<r\}\), where \(x\in X\) and \(r>0\). Moreover, let \(\#X\geq 2\) and let the measure \(\mu\) be doubling with the constant \(c_\mu >1\). A nonconstant, rectifiable, continuous mapping from a compact interval of \(\mathbb R\) to \(X\) is called a curve (it is assumed that the curve is parametrized by its arc-length). A metric space \(X\) is called a geodesic space if any two points \(x, y \in X\) can be joined by a~curve whose length is equal to \(d(x, y)\). If \(A\subset X\), with \(0< \mu(A)<\infty\), then the integral average of \(u \in L^1(A)\) is denoted by \[ u_A=\fint_A u(y) \, d\mu(y)=\frac{1}{\mu(A)} \int_A u(y) \, d\mu(y). \] If \(A\subset X\), then \(u : A \rightarrow \mathbb R\) is called a \(\beta\)-Hölder function, with an exponent \(0<\beta\leq 1\) and a constant \(0\leq \kappa <\infty\), (notation \(u \in \text{Lip}_\beta(A)\)) provided that \[ |u(x)-u(y)|\leq \kappa \;d(x, y)^{\beta} \quad \text{for all}\;\;x, y \in A. \] The authors adopt the terminology of \textit{V. Gol'dshtein} and \textit{M. Troyanov} [Expo. Math. 19, No. 4, 289--336 (2001; Zbl 1006.46023)] and introduce a \(\mathcal D\)-structure in \(X\), with exponents \(1\leq p \leq \infty\) and \(0<\beta\leq 1\), and with constants \(K>0\) and \(\tau \geq 1\), as follows. Let \(1\leq p \leq \infty\) and \(0<\beta\leq 1\). Assume that to any \(u \in \text{Lip}_\beta(X)\) a~family \(\mathcal D (u)\neq \emptyset\) of measurable functions is given in such a way that the following conditions are satisfied: {\parindent=9mm \begin{itemize}\item[(D1)] There exist constants \(K>0\) and \(\tau \geq 1\) such that the \((1, p)\)-Poincaré inequality \[ \fint_B |u(x)-u_B| \, d\mu(x) \leq K^{1/p}\, \text{diam}(B)^\beta \Big( \fint_{\tau B} g(x)^p \,d\mu(x) \Big)^{1/p} \] holds if \(B\) is a ball in \(X\), \(u \in \text{Lip}_\beta(X)\) and \(g\) in \(\mathcal D (u)\) (here if \(B=B(x,r)\), then \(\tau B=B(x, \tau r)\)). \item[(D2)] If \(a \in \mathbb R\) and \(g \in \mathcal D (u)\), then \(|a|g \in \mathcal D (au)\). \item[(D3)] If \(u, v \in \text{Lip}_\beta(X)\) and \(g_u \in \mathcal D (u)\), \(g_v \in \mathcal D (v)\), then \(g_u+g_v \in \mathcal D (u+v)\). \item[(D4)] If \(v \in \text{Lip}_\beta(X)\), with a constant \(0\leq \kappa <\infty\), and \(v|_{X\setminus E}=u|_{X\setminus E}\) for a Borel set \(E\subset X\), then \(\kappa \mathbf{1}_E+g_u\mathbf{1}_{X\setminus E} \in \mathcal D (v)\) whenever \(g_u \in \mathcal D (u)\). \end{itemize}} Since the authors also need a stronger version of the condition (D1), which corresponds to a \((p,p)\)-Poincaré inequality, they prove: If \(\{\mathcal D (u): \, u \in \text{Lip}_\beta(X)\}\) is a \(\mathcal D\)-structure in a geodesic space \(X\), with exponents \(1\leq p \leq \infty\) and \(0<\beta\leq 1\), and with constants \(K>0\) and \(\tau \geq 1\), then {\parindent=9mm \begin{itemize}\item[(D1')] there exists a constant \(K_{p,p}=C(c_\mu, \beta, p, \tau)K>0\) such that the \((p,p)\)-Poincaré inequality \[ \Big(\fint_B |u(x)-u_B|^p \, d\mu(x)\Big)^{1/p} \leq K^{1/p}_{p,p}\, \text{diam}(B)^\beta \Big( \fint_B g(x)^p \,d\mu(x) \Big)^{1/p} \] holds whenever \(B\) is a ball in \(X\), \(u \in \text{Lip}_\beta(X)\) and \(g\) in \(\mathcal D (u)\). \end{itemize}} In fact, this is a consequence of a stronger assertion, which gives \((q, p)\)-Poincaré inequalities for some \(q>p\) (and where \(\mathcal D\)-structures are not needed). Namely, the authors prove the following self-improvement result: Let \(X\) be a geodesic space, \(1\leq p \leq \infty\) and \(0<\beta\leq 1\). Assume that \(u \in \text{Lip}_\beta(X)\) and \(g:X\rightarrow [0, \infty]\) is a measurable function. Suppose that there are constants \(K>0\) and \(\tau \geq 1\) such that \[ \fint_B |u(x)-u_B| \, d\mu(x) \leq K^{1/p}\, \text{diam}(B)^\beta \Big( \fint_{\tau B} g(x)^p \,d\mu(x) \Big)^{1/p} \] holds if \(B\) is a ball in \(X\). Let \(Q \geq \log_2 c_\mu >0\) satisfy \(\beta p < Q\), where \(c_\mu\) is the doubling constant of \(\mu\). If \(1\leq q<Qp/(Q-\beta p)\), then there is a constant \(C=C(c_\mu, Q, \beta, p, q, \tau)>0\) such that \[ \Big(\fint_B |u(x)-u_B|^q \, d\mu(x)\Big)^{1/q} \leq CK^{1/p}\, \text{diam}(B)^\beta \Big( \fint_B g(x)^p \,d\mu(x) \Big)^{1/p} \] holds whenever \(B\) is a ball in \(X\). Let \(B_0 \subset X\) be a fixed ball and \[ \mathcal B_0=\{B \subset X: B \;\text{is a ball such that} \;2B \subset B_0\}. \] Given \(1<p<\infty\), \(0<\beta \leq 1\) and \(u \in \text{Lip}_\beta(X)\), the authors define a fractional maximal function \[ M_{\beta, \mathcal B_0}^{\sharp,p}u(x)=\sup_{x\in B\in \mathcal B_0}\Big(\frac{1}{\text{diam}(B)^{\beta p}} \Big(\fint_B |u(y)-u_B|^p \, d\mu(y)\Big)^{1/p} , \quad x\in X. \] The main result of the paper reads as follows: {Theorem.} Assume that we are given a \(\mathcal D\)-structure in a geodesic space \(X\), with exponents \(1<p <\infty\) and \(0<\beta \leq 1\). Let \(k\in \mathbb N\), \(0\leq \epsilon <p-1\) and \(\alpha=\beta p^2/(2(s+\beta p))>0\) with \(s=\log_2 c_\mu\). If \(K_{p,p}>0\) is the constant from the \((p,p)\)-Poincaré inequality in condition (D1'), then \[ \begin{aligned} \int_{B_0}( M_{\beta, \mathcal B_0}^{\sharp,p} u)^{p-\epsilon}\,d\mu &\leq C_1\Big(2^{k(\epsilon-\alpha)}+\frac{K_{p,p}4^{k\epsilon}} {k^{p-1}}\Big)\int_{B_0}( M_{\beta, \mathcal B_0}^{\sharp,p} u)^{p-\epsilon}\,d\mu \\ &+C_1C(k,\epsilon)K_{p,p} \int_{B_0\setminus \{M_{\beta, \mathcal B_0}^{\sharp,p} u=0\}}g^p( M_{\beta, \mathcal B_0}^{\sharp,p} u)^{-\epsilon}\,d\mu \end{aligned} \] holds for any \(u \in \text{Lip}_\beta(X)\) and \(g\) in \(\mathcal D (u).\) Here the constant \(C_1>0\) depends only on the parameters \(\beta, p, c_\mu\); and \(C(k,\epsilon)=(4^{k\epsilon}-1)/\epsilon\) if \(\epsilon>0\) and \(C(k,0)=k\). Note that the first term on the right-hand side of the inequality mentioned above is finite since \(u\) is assumed to be a \(\beta\)-Hölder function. Note also that this term is absorbed by the left-hand side provided that the numbers \(k\) and \(0<\epsilon<\epsilon_0\) are chosen appropriately. The authors apply the main result to study abstract Sobolev spaces defined in terms of a \(\mathcal D\)-structure.