Real reductive Cayley groups of rank 1 and 2. (Q2345559)

Consider a finite-dimensional associative \(K\)-algebra with unity, where \(K\) is a field of characteristic \(\neq 2\). If \(i\) is an involution of \(A\) over \(K\), then \(\{a\in A^*:a^ia=1\}\) is an algebraic group. If \(G\) denotes its connected component at the identity, then its Lie algebra is seen to be \[ \text{Lie}(G)=\{a\in A:a^i+a=0\}. \] As \(\text{char\,}K\neq 2\), the map \(a\mapsto (1-a)(1+a)^{-1}\) defines a \(G\)-equivariant birational map \(\lambda\) from \(G\) to \(\text{Lie}(G)\). Its inverse from \(\text{Lie}(G)\) to \(G\) is also \(b\mapsto (1-b)(1+b)^{-1}\). Therefore, we have a \(K\)-birational isomorphism between \(G\) and \(\text{Lie}(G)\); such a map is said to be a Cayley \(K\)-map. Indeed, the first example of this is the `Cayley transform' due to Cayley in 1846; this is the case \(A=M_n(K)\) and \(i\colon M\mapsto M^t\). In general, a linear algebraic group \(G\) over a field \(K\) is called a Cayley \(K\)-group if it has a Cayley \(K\)-map; viz., a \(K\)-birational \(G\)-equivariant (under the conjugation and adjoint actions) isomorphism between \(G\) and \(\text{Lie}(G)\). In an important paper [``Cayley groups'', J. Am. Math. Soc. 19, No. 4, 921-967 (2006; Zbl 1103.14026)] \textit{N. Lemire, V. L. Popov} and \textit{Z. Reichstein} studied this notion for algebraic groups over algebraically closed fields \(K\) of characteristic \(0\), raised many questions and answered some of them. In particular, they showed all connected solvable groups are Cayley. Also, over an algebraically closed field of characteristic \(0\), they proved that the groups \(\text{SL}_2\) and \(\text{SL}_3\) are Cayley (the latter was contrary to expectations) while \(\text{SL}_n\) for \(n\geq 4\) is not Cayley. A weaker notion is that of being stably Cayley. If \(G\times_K\mathbb G_{m,K}^r\) is a Cayley \(K\)-group (where \(G_{m,K}^r\) is a \(K\)-split torus), one calls \(G\) a stably Cayley \(K\)-group. Observe that Cayley groups are automatically stably Cayley but, the converse is a difficult problem which is not completely settled yet. In their paper cited above, Lemire, Popov \& Reichstein classified Cayley and stably Cayley groups which are simple over an algebraically closed field \(K\) of characteristic \(0\). By the works of Borovoi, Iskovskikh, Kunyavskii, Lemire and Reichstein, as of now, a complete classification of stably Cayley \(K\)-groups which are semisimple, is known where \(K\) is any field of characteristic \(0\). Further, it is known that all reductive \(K\)-groups of absolute rank \(\leq 2\) (where \(\text{char\,}K=0\)) are stably Cayley. In this paper, the case \(K=\mathbb R\) is considered, and the results show which -- among these (stably Cayley) reductive groups of rank \(\leq 2\) -- are Cayley and which are not. The main result is: Let \(G\) be a connected, reductive \(\mathbb R\)-group of absolute rank \(\leq 2\). If \(G\) is simple of type \(G_2\), or is isomorphic to \(\text{SL}_3\), or is \(\text{PGU}_3\), or \(\text{PGU}(2,1)\), then \(G\) is NOT Cayley. In the rest of the cases, \(G\) is Cayley. The proof is done case by case. The cases where \(G\) is Cayley (other than the case of \(\text{SU}_3\)) are proved in the main body of the paper. In appendix A, I. Dolgachev proves the Cayley-ness of \(\text{SU}_3\) and the non-Cayley-ness in the cases other than \(G_2\) which is already known from Lemire, Popov and Reichstein's work. In appendix B, some interesting remarks by a referee are given which deal with \(K\) of positive characterictic and \(G=\text{PGL}_1(A)\) where \(A\) is a central simple algebra over \(K\) whose degree is a multiple of the characteristic. In the paper, the following remarkable result refining stable Cayley-ness is proved as well: Let \(G\) be a connected, reductive \(K\)-group of absolute rank \(\leq 2\) where \(\text{char\,}K=0\). If \(G\) has rank \(1\), then it is Cayley. If \(G\) has rank \(2\), then \(G\times_K\mathbb G^2_{m,K}\) is Cayley.