Metrical characterization of super-reflexivity and linear type of Banach spaces (Q2474121)

A mapping \(f:X\to Y\) between two metric spaces is called a \textit{Lipschitz embedding} if there are constants \(0<c<C<\infty\) such that \[ \forall u,v\in X~~ cd_X(u,v)\leq d_Y(f(u),f(v))\leq Cd_X(u,v). \] The \textit{distortion} of a Lipschitz embedding is the ratio \(\frac{C}c\) between the optimal constants for which the inequality above is true. \textit{J. Bourgain} [Isr. J. Math. 56, 222--230 (1986; Zbl 0643.46013)] proved that a Banach space \(Y\) is not super-reflexive if and only if complete binary trees (with their graph distances) of all finite depths are Lipschitz embeddable into \(Y\) with uniformly bounded distortions. One of the main purposes of this paper is to strengthen the ``only if'' part of this result to: An infinite complete binary tree with its graph distance is Lipschitz embeddable into an arbitrary non-super-reflexive Banach space. The other purpose is to prove a similar (infinite) characterization of Banach spaces whose type does not exceed a given \(p\), \(1\leq p<2\). With the insight provided by later work, see \textit{F. Baudier} and \textit{G. Lancien} [Proc. Am. Math. Soc. 136, No. 3, 1029--1033 (2008; Zbl 1142.46037)] and \textit{M.I.~Ostrovskii} [Coarse embeddability into Banach spaces, Topology Proceedings, to appear] we can describe the proofs in the following way. A metric space is called \textit{locally finite} if any ball of finite radius in it has finite cardinality. Both results can be obtained combining well-known results with the following lemma. Lemma. Let \(X\) be a locally finite space and \(O\in X\). Suppose that a Banach space \(Y\) contains a subspace with finite dimensional decomposition \(\{F_n\}_{n=1}^\infty\) such that the sets \(B_n:=\{x\in X: d_X(O,x)\leq n\}\), \(n\in{\mathbb N}\), considered as metric spaces are Lipschitz embeddable into the respective \(F_n\) with uniformly bounded distortions. Then \(X\) is Lipschitz embeddable into \(Y\). Reviewer's remark. The first of the results of this paper shows an important difference between the tree properties introduced by [\textit{R.~C.~James}, Some self-dual properties of normed linear spaces. Sympos. infinite dim. Topology, Baton Rouge 1967, Ann. Math. Stud. 69, 159--175 (1972; Zbl 0233.46025)] and the tree properties introduced by J.~Bourgain [op. cit.]. For James' tree properties infinite tree property does not follow from finite tree property, see R.~C.~James [op. cit.].