Open core and small groups in dense pairs of topological structures (Q2003921)

For various classes of ``tame'' first-order \(L\)-structures \(M\), the question has been considered whether (or in which sense) they stay tame when the language \(L\) is expanded by a predicate \(P\) for an elementary sub-structure \(P(M) \prec M\). Often, one imposes some conditions on \(P(M)\); notably, if \(M\) is a topological structure, one often requires \(P(M)\) to be dense in \(M\) and calls \((M, P(M))\) a dense pair. In that situation, it makes sense to think of \(P(M)\) as being ``small'' (as a prototypical example, take \(M = \mathbb{R}\) and \(P(M)\) the set of algebraic elements in \(M\)). There are some results which describe all sets definable in the pair ``up to small sets''; other results describe all definable sets which are open. Results of the above kinds have been obtained when \(M\) is an o-minimal expansion of an ordered group, and also when \(M\) is a \(p\)-adically closed field. The present article considers dense pairs in a setting comprising both of those cases, namely when \(M\) is a ``geometric topological ring'', a notion which consists of quite a lot of assumptions, but all of which are very natural. The first main result is a new proof that \((L \cup \{P\})\)-definable sets \(X\) which are open are already \(L\)-definable. The proof introduces and uses the following useful criterion: To prove that \(X\) is \(L\)-definable, it suffices to verify that \(X \cap C\) is \(L\)-definable for every \(L\)-definable curve \(C\). The second main result considers groups definable in certain dense pairs (under more restrictive assumptions; in particular, \(M\) is required to be NIP). There are previous results describing definable groups only locally, which in particular means that they say almost nothing about small definable groups. (Formally, ``small'' is defined as: image of \(P(M)^m\) under a definable map, for some \(m\).) The result from the present article complements this by desribing small definable groups. An intersting tool introduced for that proof is a new notion of dimension suitable for small sets.