Deformation theory of \(\mathrm{G}_2\) conifolds (Q2208314)

Let \((M, \varphi, g)\) be a connected \(G_2\) manifold, i.e., a 7-dimensional manifold equipped with a 3-form \(\varphi\) satisfying an appropriate non-degeneracy condition, that implies the existence of a canonically associated Riemannian metric \(g = g_\varphi\) with \(G_2\) holonomy. This manifold is called \textit{asymptotically conical} (AC) of rate \(\nu > 0\) if (a) it is complete and (b) there exists a diffeomorphism \[ f: M \setminus L \to C \simeq (0, + \infty) \times \Sigma \] between the complement \( M \setminus L\) of a compact set \(L \subset M\) and a \(G_2\)-cone \((C, \varphi_C, g_C)\), with dilatation parameter \(r\), satisfying the following condition: for each \(j \geq 0\), the norm \(|\nabla^j(f^{-1}{}^*(g) - g_C)|\) of the \(j\)-th covariant derivative of \(f^{-1}{}^*(g) - g_C\) is an \( O(r^{\nu -j})\). Similarly, \((M, \varphi, g)\) is called \textit{conical singular} (CS) with \(n\) conical singularities of rates \(\nu_1, \dots, \nu_n\) if \((\alpha)\) it is the complement \(M = \overline M \setminus \{x_1, \dots, x_n\}\) in a compact topological space \(\overline M\) of a set of \(n\) points and \((\beta)\) there exists a diffeomorphism \[ f: M \setminus L \to C_1 \times \dots \times C_n \] between the complement \( M \setminus L\) (with \(n\) connected components) of a compact subset \(L \subset M\) and the disjoint union of \(n\) distinct \(G_2\)-cones \(C_\ell\), \(1 \leq \ell \leq n\), with dilatation parameters \(r_{(1)}, \dots, r_{(n)}\), respectively, satisfying the following condition: For each cone \(C_\ell\) and each \(j \geq 0\) the norm \(|\nabla^j(f^{-1}{}^*(g) - g_{C_\ell})|\) is a function of \(C_\ell\) which is an \(O(r^{\nu_\ell}_{(\ell)} -j)\). A manifold belonging to either one of these two classes is called a \textit{\(G_2\) conifold}. The vertices of the connected components of the complement \(M \setminus L\) are the \textit{ends} of the conifold. In this long but very nice and very clearly written paper, the authors develop the deformation theory of \(G_2\) conifolds. More precisely, after three preliminary sections in which they thoroughly discuss the infinitesimal deformations of \(G_2\) structures (with a particular attention to the deformations of the structures of the \(G_2\) cones) and establish useful Hodge theoretic results on the spaces of \(k\)-forms of \(G_2\) conifolds, they study the moduli spaces \(\mathcal M_\nu\) of the \(G_2\) structures of rate \(\nu\) on a fixed \(G_2\) conifold \(M\). These moduli spaces are defined as the spaces of \(G_2\)-structures modulo the actions of appropriate groups of diffeomorphisms preserving the prescribed rates. More precisely, given a rate \(\nu\) and denoting by \(\mathcal T_\nu\) the set of all \(G_2\) conifold structures on \(M\), which converge at the rate \(\nu\) at all of its ends and by \( \mathcal D_{\nu +1}\) the group of diffeomorphisms generated by the vector fields which decay (or grow) of rate \(\nu + 1\) near each end, the corresponding moduli space is defined as the quotient \(\mathcal M_\nu = \mathcal T_\nu/\mathcal D_{\nu +1}\). The main result of the paper consists in the proof the following properties: (1) In the AC case, if \(\nu\) is in the interval \((-4, 0)\) and does not belong to an appropriate (fully determined) finite set of critical values, the moduli space \(\mathcal M_\nu\) is a smooth manifold, whose dimension is given by a formula involving topological and analytic contributions (which is explicitly given in the paper). (2) In the AC case with \(\nu < -4\) and in the CS case with \(\nu > 0\), for all rates not belonging to an appropriate (fully determined) finite set of critical values, the space \(\mathcal M_\nu\) is just a topological space and the deformation theory may be obstructed. However, in all such cases, the virtual dimension of the space \(\mathcal M_\nu\) is explicitly determined and it is computable using a formula involving topological and analytic contributions. This result, combined with some refinements of the arguments used in its proof, yields diverse applications and corollaries. For instance, it is proved that the Bryant and Salamon AC \(G_2\) manifold is locally rigid. It is also shown that, for any CS \(G_2\) manifold \(M\) with singularities modeled on \(G_2\) cones over \(\mathbb C P^3\) or \(S^3 \times S^3\), the deformation theory is not obstructed and the moduli spaces \(\mathcal M_\nu\) mentioned in the above main result are not just topological spaces, but actually smooth manifolds.