Centralizers of Elements in Locally Finite Simple Groups

From MaRDI portal





The aim of this paper is to obtain information about centralizers of elements in locally finite simple groups, on the basis of the classification of finite simple groups (CFSG). This is likely to be useful in that centralizers provide a useful source of proper subgroups in simple groups, and some applications of this work in that way have already been made. The first result (Theorem A2) applies to all infinite simple locally finite groups and states that the centralizer of every element in such a group is infinite. The main result on which it depends is a general theorem of Brauer-Fowler type. Theorem A1. Given n and k, there are only finitely many finite simple groups containing an element of order n with centralizer of order k. As is to be expected, this depends heavily on CFSG. Subsequent results distinguish between linear and non-linear simple locally finite groups. The former (when infinite) are known to be Chevalley groups (including twisted types) over infinite locally finite fields (fields algebraic over finite fields), and we show (Theorem C) that the centralizers of their semisimple elements have a structure very much as in the finite case. More precisely, we prove the following, which applies to the finite as well as the infinite case. Let \(G={\mathcal G}(K)\), where K is a locally finite field of characteristic p and \({\mathcal G}=A_ l,B_ l,...,^ 2G_ 2\). Our convention is that \({\mathcal G}(K)\) denotes a group generated by unipotent elements, so that it is almost always simple. Let x be a semisimple element of order n in G. Let \(C=C_ G(x)\), \(\bar K\) be the algebraic closure of K, and F be the fundamental group of the algebraic group over \(\bar K\) corresponding to G in the obvious sense. Let \(P=O_{p'}(C)\). Then (i) C contains a normal subgroup \(C_ 0\) containing P, such that \(C/C_ 0\) is isomorphic to a subgroup of F (and so is cyclic or elementary abelian of order 4), and \(C_ 0/P\) is an abelian \(p'\)-group, (ii) \(Z(C_ 0)\) is a \(p'\)-group, and \(| C_ 0:PZ(C_ 0)|\) is finite and bounded in terms of the type \({\mathcal G}\) but independently of K, (iii) \(Z(P)=P\cap Z(C_ 0)\) and \(P/Z(P)=Y_ 1\times...\times Y_ r\) is the direct product of a number of groups of types \({\mathcal G}^*(K^*)\), where the \({\mathcal G}^*\) are possibly twisted parts of the extended Dynkin diagram associated to \({\mathcal G}\) and the \(K^*\) are extensions of K, \[ (iv)\quad 0\leq r\leq r(n)=n+(\frac{4}{n}) \] if G is of classical type, (v) \(\sum l(Y_ i)\to \infty\) as l(G)\(\to \infty\), where l is the Lie rank, (vi) C is infinite if G is infinite. The proof involves putting together the contributions from the various finite subfields, which are analyzed by following an approach of \textit{D. Deriziotis} [Commun. Algebra 9, 1997-2014 (1981; Zbl 0473.20036)]. The proof is given in some detail in order to bring out the connection between the number of \(Y_ i\) and the order of x. Also we are able to give a little extra information at the algebraic group level in the ``bad characteristic case. This is as follows. We introduce Construction 4.1 for a simply connected simple algebraic group over an algebraically closed field K of characteristic p. It proceeds as follows. (i) Draw the extended Dynkin diagram of G, obtained from the usual Dynkin diagram by adding a node \(-r_ 0\) connected by appropriate bonds to the other nodes, and weighting the nodes 1,2,...,l with the coefficients \(m_ 1,m_ 2,...,m_ l\) of the highest root \(r_ 0\). (ii) label the nodes 1,2,...,l with integers \(b_ 1,b_ 2,...,b_ l\geq 0\) in such a way that \(\gcd (b_ 1,b_ 2,...,b_ l,n)=1\) and \(\sum^{l}_{i=1}b_ im_ i\leq n\). (iii) Strike out the nodes for which \(b_ i>0\), and in addition, strike out \(r_ 0\) if and only if \(\sum b_ im_ i<n\). Theorem 4.2. With G as above, let n be an integer \(>0\) such that \(p\nmid n\) and let s be a semisimple element of G whose order mod Z(G) is n. Then \(C_ G(s)\) is a connected reductive algebraic group. The Dynkin diagram of its semisimple part is obtained by Construction 4.1. Conversely all Dynkin diagrams obtaind by Construction 4.1 arise in this way. The last statement appears to be new in ``bad characteristic. The following (Corollary C1) follows fairly easily from Theorem C. Let G be an infinite locally finite simple linear group over a field of characteristic p, and let \(g\in G\). If g is a \(p'\)-element, then g belongs to an abelian subgroup containing elements of infinitely many prime orders. Otherwise, g centralizes an infinite elementary abelian p- subgroup of G. Theorem D augments Theorem C by giving conditions under which \(C_ G(x)\) involves an infinite simple group. These conditions are not complete as the field plays a part which we have not fully analyzed. Theorem D. Let \(G={\mathcal G}(K)\), where K is an infinite locally finite field of characteristic p, and let n be a positive integer not divisible by p. Let \(\Pi =\{r_ 1,...,r_ l\}\) be a set of fundamental roots of the Lie algebra from which G was constructed, and let \(r=m_ 1r_ 1+...m_ lr_ l\) be the highest root. (i) If \(n\leq \sum^{l}_{i=1}m_ i\), then the centralizer in G of every element of order n (if these exist) involves an infinite simple group. (ii) Suppose that \(n>\sum^{l}_{i=1}m_ i\) and that K contains a primitive n-th root of 1 if n is odd, or a primitive 2n-th root of 1 if n is even. Then G contains an element of order n whose centralizer is an extension of an abelian \(p'\)-group by a subgroup of the fundamental group. Finally, the non-linear case is discussed in Theorem B. We conjecture that here, the centralizer of every element involves a non-linear simple group, but we can prove this only for direct limits of finite groups that are not too far from simple. Theorem B. Suppose that every finite set of elements of G lies in a finite simple subgroup, and that G is not linear. Then there exists a prime p with the following property. Let n be any natural number not divisible by p, let g be any element of order n in G, and let \(r(n)=n+(\frac{4}{n})\). Then \(C_ G(g)\) has a finite series of length at most \(2r(n)+1\) in which each factor is either non-abelian simple or soluble. The number of non-abelian factors is at most r(n), and at least one of them is nonlinear. The derived length of each soluble factor is at most 6, and there are at most \(r(n)+1\) of them. A slightly more general result is given as Theorem \(B'\).




Cited in
(36)








This page was built for publication: Centralizers of Elements in Locally Finite Simple Groups

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4731361)