Reflexive representability and stable metrics (Q627468): Difference between revisions

The main purpose of this paper is to establish a connection, within the field of metrisable groups, between stable metrics and reflexive representability, where a topological group is called reflexive if it can be represented as a group of isometries of a reflexive Banach space. In particular, the authors focus on groups equipped with a left-invariant metric. The first main result, i.e., Proposition 3.2, is actually a reformulation of previous general theorems, correctly quoted by the authors (here, in particular, a theorem due to Grothendieck), with the only original contribution given by Lemma 3.1, quite technical, in which the authors show that any stable metric \(d\) may be replaced by the ``equivalent'' (but bounded) \[ \delta(x,y):=\frac{d(x,y)}{1+d(x,y)}. \] The other main result is Theorem 3.3 (the converse of Proposition 3.2) that states that a necessary condition for a group \(G\) equipped with a left-invariant metric to be reflexive is that \(G\) admits an ``equivalent'' stable metric. This result again rests on an important theorem, essentially due to Shtern, which states that a topological group \(G\) is reflexive iff the topology of \(G\) is induced by weakly almost periodic (WAP) functions. In Theorem 3.3, the authors use such WAP functions in order to build a pre-metric \(h\) on \(G\), then, based upon general results that already appeared in the literature, they show, in the first part of the paper, that starting from a pre-metric \(h\), it is possible to generate a metric \(d\) uniformly equivalent to \(h\) (the only original contribute of this part, contained in Section 2, seems to be Lemma 2.7, quite technical, where the authors show how it is possible to explicitly construct such distance \(d\), under some precise assumptions). Finally, the applications of Theorem 3.3 are in general what one can expect from a necessary condition, i.e., to negate that some particular groups are reflexive, because they cannot admit a stable metric; in several cases, as the authors correctly specify, these examples have already been shown in different ways, but their method turns out to be easier.