New \(\mathrm{G}_2\)-holonomy cones and exotic nearly Kähler structures on \(S^6\) and \(S^3\times S^3\) (Q509692)

This astonishing paper solves a longstanding problem in the field of `special geometry', by exhibiting the first examples of simply connected, non-homogeneous nearly Kähler structures on the manifolds \(S^6\) and \(S^3\times S^3\), invariant by actions of \(\mathrm{SU}(2)\times\mathrm{SU}(2)\) of cohomogeneity one. Nearly Kähler (NK) manifolds \(N\) probably arose from the work of \textit{A. Gray} on holonomy, see [Math. Ann. 223, 233--248 (1976; Zbl 0345.53019)], and \textit{J. A. Wolf} and \textit{A. Gray}'s \(3\)-symmetric spaces [J. Differ. Geom. 2, 115--159 (1968; Zbl 0182.24702)]. They form the class of almost Hermitian manifolds that are closest to being Kähler in the language of the Gray-Salamon intrinsic torsion of \(G\)-structures, see [\textit{S. Salamon}, Riemannian geometry and holonomy groups. Harlow: Longman Scientific \& Technical; New York: John Wiley \& Sons (1989; Zbl 0685.53001); Milan J. Math. 71, 59--94 (2003; Zbl 1055.53039)]. They are Einstein with positive scalar curvature, and compact with finite fundamental group if complete. A deRham-like structure theorem tells us \(6\)-dimensional ones are distinguished [\textit{P.-A. Nagy}, Asian J. Math. 6, No. 3, 481--504 (2002; Zbl 1041.53021)]. As a matter of fact, it is the connection to \(G_2\) geometry that is key: the metric cone over an NK \(6\)-manifold has a holonomy \(G_2\) [\textit{C. Bär}, Commun. Math. Phys. 154, No. 3, 509--521 (1993; Zbl 0778.53037)]. This is of interest also because it ties in nicely with the general theory of Killing spinors [\textit{H. Baum} et al., Twistors and Killing spinors on Riemannian manifolds. Stuttgart etc.: B.G. Teubner Verlagsgesellschaft (1991; Zbl 0734.53003)], see also [\textit{I. Agricola} et al., J. Geom. Phys. 98, 535--555 (2015; Zbl 1333.53037)]. The scarcity of NK \(6\)-manifolds (i.e., \(G_2\)-holonomy cones) is surprising and somehow vexing, especially if compared to other special geometries: there are in fact infinitely many Calabi-Yau, hyperKähler and \(\mathrm{Spin}(7)\)-cones. Besides, it is known that \(S^6, S^3\times S^3, \mathbb{CP}^3\) and the full flag \(F^3=\operatorname{SU}(3)/(S^1\times S^1)\) are the only homogeneous NK \(6\)-manifolds [\textit{J.-B. Butruille}, Ann. Global Anal. Geom. 27, No. 3, 201--225 (2005; Zbl 1079.53044)], which opens the question of complete, non-homogeneous instances. There are two natural ways to find NK \(6\)-manifolds, namely imposing enough symmetry (cohomogeneity-one) in order to attain manageable differential equations, and resolving singular NK examples. The authors combined the two techniques: First they take the sine cone of a Sasaki-Einstein \(5\)-manifold to obtain a singular NK space with two isolated singularities modelled on a Calabi-Yau cone. Then they desingularise it as a cohomogeneity-one space by replacing said singularities with conical Calabi-Yaus. The authors take the potential models of complete cohomogeneity-one NK \(6\)-manifolds given by \textit{F. Podestà} and \textit{A. Spiro} [J. Geom. Phys. 60, No. 2, 156--164 (2010; Zbl 1184.53074); Commun. Math. Phys. 312, No. 2, 477--500 (2012; Zbl 1262.53062)], and `match', instead of glueing, the local geometries. They study carefully the evolution ODEs of one-parameter family of nearly hypo structures [\textit{D. Conti} and \textit{S. Salamon}, Trans. Am. Math. Soc. 359, No. 11, 5319--5343 (2007; Zbl 1130.53033); \textit{M. Fernández} et al., J. Lond. Math. Soc., II. Ser. 78, No. 3, 580--604 (2008; Zbl 1158.53018)] and find a way to recognise which local solutions extend to complete metrics. The conjecture posed in the paper posits that are there no more (inhomogeneous) cohomogeneity-one NK structures on simply connected six-manifolds.