Metric embeddings. Bilipschitz and coarse embeddings into Banach spaces (Q1945323)

It is a classical fact that every metric space embeds isometrically into some Banach space. Recently, the problem of embedding metric spaces like graphs with the graph distance or groups with the word metric into ``good'' Banach spaces (like Hilbert spaces or just uniformly convex spaces) by means of Lipschitz or more general maps has attracted a lot of attention. The present monograph is dedicated to this circle of ideas. One says that a metric space \((X,d)\) bilipschitz embeds into a Banach space \(E\) if there exists an injective Lipschitz map \(f:X\to E\) such that \(f^{-1}:f(X)\to X\) is Lipschitz as well, and one says that \(X\) coarsely embeds into \(E\) if there are a map \(f:X\to E\) and increasing functions \(\rho_1, \rho_2: [0,\infty) \to [0,\infty)\) such that \(\lim_{t\to\infty} \rho_1(t)=\infty\) and \[ \rho_1(d(x,y)) \leq \|f(x)-f(y)\| \leq \rho_2(d(x,y)) \] for all \(x,y\in X\). The meaning of the latter is that such an \(f\) respects large distances; however, a coarse ``embedding'' neither needs to be injective nor continuous. It was M.~Gromov who advocated the investigation of coarse embeddings into Hilbert spaces in the context of geometric group theory, an approach that was successfully worked out in papers by G.~Kasparov, V.~Lafforgue, G.~Yu and others; however, this is not the focus of the book under review. Bilipschitz embeddings are even relevant for finite metric spaces when one has quantitative estimates for Lip\((f)\cdot\)Lip\((f^{-1})\); the infimum of these products over all bilipschitz embeddings is called the distortion \(c_E(X)\) -- this is similar to the point of view of local Banach space theory. In the first chapter, the author introduces bilipschitz and coarse embeddings and gives some examples. The second chapter is devoted to the proof of the fact -- in this generality due to the author -- that bilipschitz resp.\ coarse embeddability are finitely determined for locally finite metric spaces. For this, several notions from Banach space theory such that ultraproducts, finite representability and type and cotype are introduced. The third chapter presents a class of embeddings that eventually leads to the estimate \(c_{\ell_2}(X) \leq C \log n\) for \(n\)-point metric spaces~\(X\). Originally, this distortion estimate was obtained by Bourgain using different methods. The theme of the fourth chapter is Poincaré inequalities, which can be used to estimate \(c_E(X)\) from below. Also, some families of graphs, viz.\ expander graphs, are defined; and with the help of Poincaré inequalities it can relatively easily be shown that expanders are only badly (uniformly) coarsely embeddable into \(L_1\) or \(L_2\). In the fifth chapter, the existence of expanders is proved, which is anything but obvious. Various constructions are provided; e.g., based on Kazhdan's property~(T), or probabilistic constructions. Also, another family of graphs, those with large girth, is introduced. (Having a large girth is another obstacle for ``nice'' embeddings.) Chapter~6 is more functional analytic in flavour than the previous two and discusses the question of which Banach spaces permit only bad embeddings of expanders, like \(L_1\) or \(L_2\). Stable Banach spaces are introduced, and a construction of nonreflexive spaces of type \({>1}\) is given (the one due to Pisier and Xu, not James's original construction). Chapter~7 exhibits expander-like constructions that prevent coarse embeddability into Hilbert spaces. In Chapter~8, the author considers Markov chains in metric spaces, the idea being that Markov chains might behave very differently in different metric spaces. A nonlinear analogue of type of a Banach space, namely Markov type, is introduced and used to study embeddability into Hilbert spaces. This approach is, in particular, applicable to graphs with large girth. On the Banach space side, this chapter deals with superreflexivity, \(p\)-smoothness, \(q\)-convexity and \(q\)-convex renormings of uniformly convex spaces. Chapter~9 takes a glance at the so-called Ribe program, that is, the classification of Banach spaces purely by properties of the underlying metric space, forgetting the linear structure. The main result of this chapter is Bourgain's discretisation theorem from which Ribe's theorem on (crude) finite representability of uniformly homeomorphic Banach spaces is derived. Finally, Chapter~10 looks at the Lipschitz free space (also known as Arens-Eells space) of a metric space \(X\), i.e., the canonical predual of the Banach space of Lipschitz functions on \(X\), and Chapter~11 gathers open problems. Each chapter comes with exercises (some of them with hints) and a very useful notes and remarks section with pointers to the literature and further results. There are also detailed author and subject indices. The book lies at the interface of Banach space theory and metric graph theory. The author has succeeded in making profound material from both the functional analytic and the discrete side accessible to his readers. All the arguments are exposed in great detail, and the line of reasoning is laid out clearly. The book can be highly recommended to researchers who wish to enter the rapidly developing area of metric embeddings.