Spin (7)-manifolds with parallel torsion form (Q1048111)

Let \(M^8\) be a \(Spin(7)\)-manifold, that is an \(8\)-dimensional Riemannian manifold together with a reduction of the principal bundle of oriented orthonormal frames to the compact simply connected Lie group \(Spin(7)\subset SO(8)\). Any such manifold admits a unique metric connection \(\nabla^c\) (the \textit{characteristic connection}) with totally skew-symmetric torsion \(T^c\) (the \textit{characteristic torsion}) preserving the underlying \(Spin(7)\)-structure, and making a non-trivial spinor field parallel [see \textit{S. Ivanov}, Math. Res. Lett. 11, No.~2--3, 171--186 (2004; Zbl 1073.53065)]. The author assumes that \(T^c\neq 0\) and \(\nabla^c(T^c)=0\), and explains why such structures are of particular interest [see also \textit{I. Agricola, T. Friedrich, P.-A. Nagy} and \textit{C. Puhle}, Classical Quantum Gravity 22, No.~13, 2569--2577 (2005; Zbl 1074.81056); \textit{T. Friedrich} and \textit{S. Ivanov}, Asian J. Math. 6, No.~2, 303--335 (2002; Zbl 1127.53304); \textit{C. Puhle}, J. Geom. Phys. 58, No.~10, 1355--1375 (2008; Zbl 1215.53042)]. This implies that the holonomy algebra \(\mathfrak{hol}(\nabla^c)\) of the connection \(\nabla^c\) is a subalgebra of the isotropy algebra \(\mathfrak{iso}(T^c)\) of the torsion \(T^c\), these two being Lie subalgebras of \(\mathfrak{spin}(7)\). Studying the non-abelian subalgebras of \(\mathfrak{spin}(7)\), the author provides the classification of complete, simply connected \(Spin(7)\)-manifolds with \(\nabla^c\)-parallel \(T^c\neq 0\) and non-abelian isotropy algebra \(\mathfrak{iso}(T^c)\leq \mathfrak{spin}(7)\). More precisely, he proves that there exist eight admissible isotropy algebras, and describes the admissible torsion forms \(T^c\) and Ricci tensors associated to \(\nabla^c\). The main result of this paper is that these manifolds are isometric to either a Riemannian product or a homogeneous naturally reductive space, some of them being uniquely determined. Moreover, every such structure admits at least two \(\nabla^c\)-parallel spinor fields. There are examples exhibiting 16 \(\nabla^c\)-parallel spinor fields. An important tool in the author's considerations is the splitting theorem of de Rham generalized to geometric structures with totally skew-symmetric torsion [see \textit{R. Cleyton} and \textit{A. Moroianu}, ``Connections with parallel torsion in Riemannian geometry'', to appear]. Also, results previously established by several authors [see \textit{B. Alexandrov, T. Friedrich} and \textit{N. Schoemann}, J. Geom. Phys. 53, No.~1, 1--30 (2005; Zbl 1075.53036); \textit{T. Friedrich}, Differ. Geom. Appl. 25, No.~6, 632--648 (2007; Zbl 1141.53019); \textit{T. Friedrich} and \textit{S. Ivanov}, Asian J. Math. 6, No.~2, 303--335 (2002; Zbl 1127.53304); \textit{T. Friedrich} and \textit{S. Ivanov}, J. Geom. Phys. 48, No.~1, 1--11 (2003; Zbl 1029.81037)]; \textit{T. Friedrich, I. Kath, A. Moroianu} and \textit{U. Semmelmann}, J. Geom. Phys. 23, No.~3--4, 259--286 (1997; Zbl 0898.53038)] are used in the proofs concerning the geometries related to the algebraic cases.