The Twistor Space of a 3-Sasakian Manifold (Q4335500)

Penrose's twistor construction was generalized by \textit{S. Salamon} [Invent. Math. 67, 143-171 (1982; Zbl 0486.53048)] to arbitrary quaternionic Kähler manifolds. If the scalar curvature is positive, the twistor space is a Fano contact manifold with a Kähler-Einstein metric. \textit{C. LeBrun} [Int. J. Math. 6, 419-437 (1995; Zbl 0835.53055)] has shown the converse: a Fano contact manifold is the twistor space of a quaternionic Kähler manifold of positive scalar curvature precisely when it admits a Kähler-Einstein metric. Recent work of \textit{C. P. Boyer, K. Galicki} and \textit{B. M. Mann} [Commun. Anal. Geom. 1, 229-279 (1993; Zbl 0856.53038); J. Reine Angew. Math. 455, 183-220 (1994)] shows that this setup is intimately related to 3-Sasakian geometry. Namely, to the twistor space, being a 2-sphere bundle over a quaternionic Kähler manifold, is associated the corresponding principal \(SU(2)\) or \(SO(3)\) bundle, which is a 3-Sasakian manifold. The twistor space can be recovered as a quotient of this 3-Sasakian manifold by the circle action generated by one of the three Killing vector fields defining the 3-Sasakian structure. The goal of this paper is to give a 3-Sasakian version of LeBrun's theorem mentioned above. The essential difficulty is that the quotient of a 3-Sasakian manifold by a circle action as above is in general not a manifold but an orbifold. This orbifold is referred to as the twistor space associated to a 3-Sasakian manifold. Using W. L. Baily's harmonic theory for orbifolds, they relate Betti numbers of compact 3-Sasakian manifolds with those of its twistor space in the same way as in the smooth case. They show, furthermore, that the twistor space of a compact 3-Sasakian manifold is a simply connected \(\mathbb Q\)-factorial Fano variety with a complex contact structure. The main theorem of this paper states that a \(\mathbb Q\)-factorial Fano contact variety, whose principal circle bundle (given by the contact structure) is a smooth compact manifold, is the twistor space of a 3-Sasakian manifold if and only if it admits a compatible Kähler-Einstein metric. This easily gives an orbifold version of LeBrun's theorem. A guiding principle for this paper is that many results that hold in the smooth case still hold in the orbifold case. However, the finiteness theorem of \textit{C. LeBrun} and \textit{S. Salamon} [Invent. Math. 118, 109-132 (1994; Zbl 0815.53078)] does not carry over to this more general setting. For example, if a (smooth) twistor space has second Betti number two, it is unique (it must be a flag manifold). In this paper, an infinite family of inequivalent (orbifold) twistor spaces with second Betti number two are constructed. A precise description of the singular locus of these examples is given in the final section.