Lorentzian cones in real Lie algebras (Q1061876)

A cone K in a finite dimensional real vector space L is called a Lorentzian cone iff there is a linear form \(\omega\) : \(L\to {\mathbb{R}}\) and a quadratic form \(q: L\times L\to {\mathbb{R}}\) such that \(K=\{x\in L:\) \(\omega\) (x)\(\leq 0\) and q(x,x)\(\leq 0\}\). Then q is a Lorentzian form, determined uniquely up to scalar multiples and called a defining Lorentzian form for W. If L is a Lie algebra and \(x*y=x+y+[x,y]+..\). is the Campbell-Hausdorff-multiplication defined for all sufficiently small x and y, then we say that a closed cone K is a Lie semialgebra in L iff x*y\(\in K\) for all sufficiently small x,y\(\in K.\) This paper determines all Lorentzian Lie semialgebras in real Lie algebras. If L is the Lie algebra of the group of translations of \({\mathbb{R}}^ n\) extended by the group of scalar non-zero multiplications, then every Lorentzian cone is a Lie semialgebra. In all other cases, the defining Lorentzian form is invariant, i.e. satisfies q([x,y], z)\(=q(x, [y,z])\) (Theorem of Guts and Levichev), and L decomposes into a direct orthogonal sum (with respect to q) of two ideals one of which is compact with q inducing a positive definite quadratic form, while the other is of one of three types which are precisely described; it is either solvable and one of a countable series of solvable algebras of length 3; or else is sl(2,\({\mathbb{R}})\) with the Cartan-Killing form as Lorentzian form, or else is compact with a non-trivial center and a precisely described Lorentzian form. [Cf. \textit{A. K. Guts} and \textit{A. Levichev}, Dokl. Akad. Nauk SSSR 277, 1299-1302 (1984) and \textit{A. P. Medina}, Sémin. géométrie différentielle 1981-1982, Montpellier 1982, Exp. No.6 (1982; Zbl 0532.22018)].