Fibered cusp versus \(d\)-index theory (Q2462258)

Let \(\overline{X}\) be a compact manifold-with-boundary and let \(\pi : \partial X \to Y\) be a locally trivial bundle whose base is a closed manifold. Let \(\rho\) be a defining function for \(\partial X\). Let \(T(X)^\Phi\) be the \(\Phi\)-tangent bundle over \(\overline{X}\), so that the restriction to \(\partial X\) of any smooth section \(V : \overline{X} \to T(X)^\Phi\) is vertical and \(V\) vanishes at \(\partial X\) to the second order. Let \(g_\Phi\) be a fibered cusp metric i.e. \(g_\Phi\) is the restriction to \(X = \overline{X} \setminus \partial X\) of a Riemannian metric in \(T(X)^\Phi\) which is smooth up to the boundary. One calls \(g_\Phi\) exact if there is a tensor field \(S \in C^\infty (\overline{X} , S^2 (T(X)^\Phi ))\) such that \(g_\Phi - \rho \, S\) is a product \(\Phi\)-metric, i.e., a \(\Phi\)-metric of the form \(\rho^{-4} \, d \rho^2 + \rho^{-2} \pi^* g_Y + h\) near infinity, where \(g_Y\) is a Riemannian metric on \(Y\) and \(h\) is a family of metrics on the fibres of \(\pi\). To state the main result in the paper under review let \(g_\Phi\) be an exact \(\Phi\)-metric on \(X\). Let \(E \to \overline{X}\) be a Hermitian vector bundle with connection smooth up to the boundary. Let \(X\), \(\partial X\) and the fibres of \(\pi\) be endowed with compatible spin structures. Let \(D^\Phi\) and \(D^p\) be the Dirac operators associated to \(g_\Phi\) and \(\rho^{2p} g_\Phi\) for \(p \geq 0\). Let us assume that that the twisted Dirac operators on the fibres of the portion of \(E\) over \(\partial X\), the induced metric on the fibres of \(\pi\), and the induced spin structure are invertible. Then multiplication by \(\rho^{p(n-1)/2}\) gives an isomorphism of the \(L^2\)-kernels of \(D^p\) and \(D^\Phi\). Reviewer's remark: For any compact strictly pseudoconvex CR manifold-with-boundary \(M\) the total space \(\overline{X}\) of the canonical circle bundle \(S^1 \to \overline{X} \to M\) over \(M\) is a compact manifold-with-boundary such that \(S^1 \to \partial X \to \partial M\) is a principal bundle. It is an appealing question whether these results apply to \(\overline{X}\) endowed with the Fefferman metric in the context of the spinor calculus developed by \textit{H. Baum} [Dragomir, Sorin (ed.), Selected topics in Cauchy-Riemann geometry. Rome: Aracne. Quad. Mat. 9, 39--87 (2003; Zbl 1066.53095)]