A note on the Gurov-Reshetnyak condition. (Q1812503): Difference between revisions

Let \(\mu\) be a positive measure on \(\mathbb R^n\). Let \(f_{Q,\mu}=\frac{1}{\mu(Q)}\int_Q \, d\mu\) denote the average of \(f\) on \(Q\) and \(\Omega_\mu (f,Q)=| f-f_{Q,\mu}| _{Q,\mu}\) denote the mean oscillation of \(f\) over \(Q\). Given \(0<\epsilon<2\) one says that \(f\) satisfies the Gurov-Reshetnyak condition of order \(\epsilon\) (denoted \(f\in \text{GR}_\mu(\epsilon)\)) if, for any cube, one has \(\Omega_\mu (f,Q)\leq \epsilon f_{Q,\mu}\). The condition was introduced by Gurov and Reshetnyak in the 1970s as a type of analogue of BMO, with applications similar to those for BMO in mind. Here the authors prove that \(A_\infty(\mu)=\cup_{0<\epsilon<2} GR_\mu(\epsilon)\). Here \(A_\infty(\mu)\) denotes the Muckenhoupt class with respect to the measure \(\mu\). The point is to start with the appropriate definition of \(A_\infty(\mu)\), namely, \(w\in A_\infty\) if and only if there exist \(0<\alpha,\beta<1\) such that, for any cube \(Q\), \(\mu(\{x\in Q: w(x)>\beta w_{Q,\mu}\})>\alpha \mu(Q)\). The authors relate the parameters \(\alpha\) and \(\beta\) with \(\epsilon\) in the definition of \(\text{GR}(\epsilon)\). The authors go on to estimate the average rearrangement \(\frac{1}{t}\int_0^t f^\ast_\mu (\tau)\, d\tau\) in terms of \(f^\ast_\mu(t)\) and these parameters.