Contents

1. Introduction The exponential map is a fundamental instrument of Lie theory that yields local linearization of various problems involving Lie groups and their actions, see [Bou1 ]. Let L be a real Lie group with Lie algebra l. As the differential at 0 of the exponential exp : l → L is bijective, exp yields a diffeomorphism of an open neighborhood of 0 in l onto an open neighborhood U of the identity element e in L. The inverse diffeomorphism λ (logarithm) is equivariant with respect to the action of L on l via the adjoint representation AdL : L → Aut l and on L by conjugation, i.e., λ(gug −1 ) = AdL g (λ(u)) if g ∈ L, u ∈ U and gug −1 ∈ U . This shows that the conjugating action of L on its underlying manifold is linearizable in a neighborhood of e. In this paper we study what happens if L is replaced with a connected linear algebraic group G over an algebraically closed field k : what is a natural algebraic counterpart of λ for such G and for which G does it exist? In the sequel we assume that char k = 0 (in fact in many places this assumption is either redundant or can be bypassed by modifying the relevant proof). 1.1. The classical Cayley map. Let g be the Lie algebra of G. One way to look at the problem is to replace the Hausdorff topology in the Lie group setting by the ´ etale topology, i.e., to define the algebraic counterpart of λ as a G-equivariant morphism G → g ´ etale at e. Then, at least for reductive groups, there is no existence problem: such morphisms always exist, see Corollary of Lemma 10.3 below. Properties of some of them has been studied by Kostant and Michor in [KM], see Example 10.4 below. In the general case, there always exists a rational dominant G-equivariant map G g, see Proposition 10.5 below. In the present paper we look at the problem differently. Our point of view stems from a discovery made by Cayley in 1846, [Ca], cf. [Weyl], [Pos]. It suggests that the most direct approach, i.e., replacing the Hausdorff topology by the Zariski one, leads to something really interesting. Namely, let G be the special orthogonal group, G = SOn := {X ∈ Matn×n | X T X = In }, where In is the identity n × n-matrix. Then g = on := {Y ∈ Matn×n | Y T = −Y }, and the adjoint representation AdG : G → Aut g is given by (1.2) AdG g (Y ) = gYg −1 , λ: G
CAYLEY GROUPS
NICOLE LEMIRE, VLADIMIR L. POPOV, AND ZINOVY REICHSTEIN Abstract. The classical Cayley map, X → (In − X )(In + X )−1 , is a birational isomorphism between the special orthogonal group SOn and its Lie algebra son , which is SOn -equivariant with respect to the conjugating and adjoint actions respectively. We ask whether or not maps with these properties can be constructed for other algebraic groups. We show that the answer is usually “no”, with a few exceptions. In particular, we show that a Cayley map for the group SLn exists if and only if n 3, answering an old question of Luna.
≃
g ∈ G, Y ∈
Cayley discovered that there exists a birational isomorphism (1.3) g
equivariant with respect to the conjugating and adjoint actions of G on the underlying varieties of G and g respectively, i.e., such that (1.4) λ(gXg −1 ) = AdG g (λ(X ))
Date : September 1, 2004. 2000 Mathematics Subject Classification. 14L35, 14L40, 14L30, 17B45, 20C10. Key words and phrases. Algebraic group, Lie algebra, reductive group, algebraic torus, Weyl group, root system, birational isomorphism, Cayley map, rationality, cohomology, permutation lattice. N. Lemire and Z. Reichstein were supported in part by NSERC research grants. V. L. Popov was supported in part by ETH, Z¨ urich, Switzerland and Russian grant NX–123.2003.01.
CAYLEY GROUPS
3
if g and X ∈ G and both sides of (1.4) are defined. His proof is given by the explicit formula defining such λ: (1.5) λ : X → (In − X )(In + X )−1
1
2
NICOLE LEMIRE, VLADIMIR L. POPOV, AND ZINOVY REICHSTEIN 9. Which stably Cayley groups are Cayley? 9.1. The group SL3 9.9. The group G2 10. Generalization 10.1. Generalized Cayley maps 10.9. The Cayley degree Appendix. Alternative proof of Proposition 9.2: an outline
Contents 1. Introduction 1.1. The classical Cayley map 1.6. Basic definitions, main problem, and examples 1.25. Notational conventions. 1.30. Main results 1.40. Application to Cremona groups 2. Preliminaries 2.1. Γ-fields and Γ-varieties. 2.2. Γ-lattices 2.8. Stable equivalence and flasque resolutions 2.15. Speiser’s Lemma 2.17. Homogeneous fiber spaces 3. Cayley maps, generic tori, and lattices 3.1 Restricting Cayley maps to Cartan subgroups 3.8. Generic tori 3.13. Proof of Theorem 1.31 4. Reduction theorems 4.1. Unipotent normal subgroups 4.3. The Levi decomposition 4.7. From reductive to semisimple 4.10. From semisimple to simple 5. Proof of Theorem 1.32: an overview 6. The groups SLn /µd and their character lattices 6.1. Lattices Qn (d) 6.9. Properties of Qn (d). 7. Stably Cayley groups of type An 7.1. Restricting Qn (d) to some subgroups 7.3. Lattices Λp and Λ2p 7.10. Completion of the proof of Proposition 5.1 8 Stably Cayley groups of type Dn 8.1. Root systems of type Dn 8.2. Lattices Y2m and Z2m