Markov type and threshold embeddings (Q359547): Difference between revisions

The Markov type was introduced by \textit{K. Ball} [Geom. Funct. Anal. 2, No. 2, 137--172 (1992; Zbl 0788.46050)] as a metrical version of the notion of Rademacher type for Banach spaces, with the intention to use it for Lipschitz extension problems. In the same paper, Ball proved several results showing the applicability of this concept, and especially the notion of Markov type \(2\). However, for a long time, the only known examples of spaces with Markov type \(2\) were metric spaces admitting bilipschitz embeddings into a Hilbert space. The breakthrough in this direction was achieved in the paper by \textit{A. Naor} et al. [Duke Math. J. 134, No. 1, 165--197 (2006; Zbl 1108.46012)] where it was shown (among other results) that trees and the spaces \(L_p\) \((2\leq p<\infty)\) have Markov type \(2\). The present paper is a further significant contribution to this direction. It is shown that weighted planar graphs, doubling metric spaces (answering questions asked in [Naor et al., loc. cit.]), and spaces with finite Assouad-Nagata dimension have Markov type \(2\). The proof uses the following notion: A metric space \(X\) is said to threshold-embed into a metric space \(Y\) if there is a constant \(0<K<\infty\) and a family of maps \(\varphi_\tau:X\to Y\) for all \(\tau>0\) such that, for every \(x,y\in X\), the inequality \(d_X(x,y)\geq \tau\) implies \(d_Y(\varphi_\tau(x),\varphi_\tau(y))\geq\tau\text{Lip}(\varphi_\tau)/K\). The authors prove (Theorem 1.3) that metric spaces which threshold-embed into a Hilbert space have Markov type \(2\). After that, they apply the known facts that the mentioned classes of spaces threshold-embed into a Hilbert space. The authors also prove the following \(p\)-analogue of Theorem 1.3: Metric spaces which threshold-embeds into a \(p\)-uniformly smooth Banach space have Markov type \(p\) \((1<p\leq 2)\). The paper concludes with some interesting results which may be regarded as nonlinear analogues of \textit{S. Kwapien}'s theorem [Stud. Math. 44, 583--595 (1972; Zbl 0256.46024)] on Banach spaces having Radema\-cher type \(2\) and Radema\-cher cotype \(2\) (it is known, and is explained on p.\,1210, that the most natural nonlinear analogues of Kwapien's theorem fail to be true). Reviewer's remark: The references to [MN06] at the top of p.\,1212 should be replaced by references to [MN08].