Linear groups definable in o-minimal structures (Q5957502)

The paper concerns subgroups \(G<GL(n,R)\) definable in an o-minimal structure \(M=(R,<,+,\cdot,\ldots)\), where \((R,<,+,\cdot)\) is a real closed field. The authors give a number of conditions which guarantee that \(G\) is semialgebraic (that is, definable in \((R,+,\cdot)\)), or at least definably isomorphic to a semialgebraic group. They also give examples of metabelian \(G\) which are not even abstractly isomorphic to a semialgebraic group. The paper begins with an overview of the Lie algebra techniques developed by the same authors [``Definably simple groups in o-minimal structures'', Trans. Am. Math. Soc. 352, 4397-4419 (2000; Zbl 0952.03046)]. This, with some preliminaries on commutative definable groups, yield that if \(G\) is definable connected and nilpotent then \(G\) is definably isomorphic to a linear semialgebraic group. The authors also use Lie methods to show that for definable \(G\) (in their usual setting), there are semialgebraic \(G_1,G_2 <GL(n,R)\) with \(G_2<G<G_1\), \(G_2\) normal in \(G_1\), and \(G_1/G_2\) abelian. This yields easily that either \(G\) is itself semialgebraic, or \(G\) is definable in \((R,+,\cdot,f_1,\ldots,f_k)\) where the \(f_i\) are definable automorphisms of \((R_{>0},\cdot)\), or there is a definable isomorphism \(e:(R,+) \rightarrow (R_{>0},\cdot)\) and \(G\) is definable in \((R,+,\cdot,e)\). It is also shown that if \(G\) is semisimple (that is, has no infinite definable normal abelian subgroup) then \(G\) is semialgebraic; also, if \(G\) is bounded and closed in \(R^{n^2}\), then \(G\) is semialgebraic. The paper includes a number of results about abstract/semialgebraic simplicity. For example, if \(G\) is an infinite group, then \((G,\cdot)\) is elementarily equivalent to some simple Lie group if and only if \(G\) is definable in some o-minimal structure and there is no proper non-trivial normal subgroup of \(G\) definable in \((G,\cdot)\). It is also shown that if \(G\) is a semialgebraic non-abelian group over some real closed field, and \(G\) has no proper non-trivial semialgebraic normal subgroup, and \((G,\cdot)\) is not elementarily equivalent to any compact simple linear real Lie group, then \(G\) is abstractly simple.