Power residue symbols and the central sections of \(\text{SL}(2,A)\) (Q1273634)

The problem of classifying the normal subgroups of a classical linear (algebraic) group over a commutative ring \(R\) has attracted considerable attention over many years. For the important special case \(G=\text{SL}_n\) it has been known for some time that, when the rank is ``sufficiently high'', such subgroups can be classified in terms of the \(R\)-ideals. Following a long series of special cases it was finally shown (by Vaserstein) that, when \(n\geq 3\), a subgroup \(N\) of \(\text{SL}_n(R)\) is normalized by \(E_n(R)\), the subgroup generated by the elementary matrices, if and only if \[ C_n(R,{\mathfrak q})\leq N\leq H_n(R,{\mathfrak q}), \] where \(C_n(R,{\mathfrak q})=[E_n(R),H_n(R,{\mathfrak q})]\) and \(H_n(R,{\mathfrak q})\) is the set of matrices which are scalar mod \({\mathfrak q}\). (The \(R\)-ideal \(\mathfrak q\) is uniquely determined by \(N\).) The normal subgroup structure of \(\text{SL}_2(\mathbb{Z})\) is known to be so complicated that a classification similar to the above for every \(\text{SL}_2(R)\), in terms of some kind of substructure of \(R\), seems very unlikely to exist. For anything like the above ladder relation to hold for \(n=2\) it appears that \(R\) has to contain ``many'' units. From now on \(A\) denotes either an \(SR_2\)-ring or an arithmetic Dedekind domain with infinitely many units. It is known that the above classification holds for \(\text{SL}_n(A)\) when \(n=2\), provided, for example, 6 is a unit in \(A\). In general however the \(A\)-ideals are not sufficient to classify the normal subgroups of \(\text{SL}_2(A)\). The extra complications arise when 2 is not a unit. To extend the above one seeks some kind of ``generic subgroup'' \(G_\alpha\) of \(\text{SL}_2(A)\), defined in terms of a special type of additive subgroup of \(A\), with the following property. A subgroup \(N\) of \(\text{SL}_2(A)\) is normal if and only if there exists a (unique) subgroup \(G_{\alpha'}\) of ``generic'' type such that \[ C_{\alpha'}\leq N\leq G_{\alpha'},\quad\text{where }C_{\alpha'}=[\text{SL}_2(A),G_{\alpha'}]. \] Menal and Vaserstein have proved that such a classification holds for an \(SR_2\)-ring \(A\), provided \(A\) has no quotient of order 2. Their classification involves the concept of a Jordan ideal. (An additive subgroup \(\Gamma\) of \(A\) is called a \(\mathcal J\)-ideal if and only if, for all \(x\in\Gamma\) and \(t\in A\), we have \(t^2x,tx^2\in\Gamma\).) In a previous paper [Ann. Math., II. Ser. 134, 159-188 (1991; Zbl 0743.20048)] the authors remove this restriction on the quotients of \(A\) by introducing the notion of a radix. (An additive subgroup \(\mathcal R\) of \(A\) is called a radix if and only if, for all \(x\in{\mathcal R}\) and \(t\in A\), we have \(tx^3,(t^3-t)x^2+t^2x\in{\mathcal R}\).) They use radices to extend the above classification theorem to \textit{all} \(A\). (The differences between radices, \(\mathcal J\)-ideals and \(A\)-ideals are largely determined by ``2''. For example, if \(A\) has no quotient of order 2, then every radix is a \(\mathcal J\)-ideal. In addition, if 2 is a unit, then every radix (and consequently every \(\mathcal J\)-ideal) is an ideal.) In a celebrated paper \textit{H. Bass, J. W. Milnor} and \textit{J.-P. Serre} [Publ. Math., Inst. Hautes Étud. Sci. 33, 59-137 (1967; Zbl 0174.05203)] have classified the normal subgroups of \(\text{SL}_n(B)\), for \(n\geq 3\), where \(B\) is \textit{any} arithmetic Dedekind domain. This paper is very much concerned with the so-called Congruence Subgroup Problem and central to this problem is the quotient group \(\text{SL}_n(B,{\mathfrak q})/C_n(B,{\mathfrak q})\), where \(\text{SL}_n(B,{\mathfrak q})\) is the set of all matrices congruent to the identity \((\text{mod }{\mathfrak q})\) and \(C_n(B,{\mathfrak q})=[\text{SL}_n(B),\text{SL}_n(B,{\mathfrak q})]\). Most of their paper is devoted to the structure of this quotient group and involves the concept of a Mennicke Symbol, together with the theory of power residue and Hilbert symbols from algebraic number theory. This quotient is shown to be a finite, cyclic group which is trivial when \(A\) is not totally imaginary. (A totally imaginary arithmetic Dedekind domain is the ring of integers of an algebraic number field not embeddable in \(\mathbb{R}\).) In this paper the authors determine the structure of the ``radix'' equivalent of the quotient group of Bass, Milnor and Serre for \(\text{SL}_2(A)\), denoted by \(\text{SL}_2(A,{\mathcal R})/C({\mathcal R})\), where \(C({\mathcal R})=[\text{SL}_2(A),\text{SL}_2(A,{\mathcal R})]\). This proves to be a very complicated problem whose solution is based on an extension of the theory of Mennicke symbols from (ordinary) ideals to \(\mathcal J\)-ideals. The most difficult case (not surprisingly perhaps, given the Bass, Milnor, Serre paper) is where \(A\) is a totally imaginary Dedekind domain. This involves a great deal of the theory of power residue and Hilbert symbols.