A generalized MIT bag operator on spin manifolds in the non-relativistic limit (Q2149431)

This paper extends to arbitrary spin manifolds the results of \textit{A. Moroianu} et al. [Commun. Math. Phys. 374, No. 3, 1963--2013 (2020; Zbl 1445.35273)], valid on flat Euclidean spaces, concerning the eigenvalues of a Dirac operator modified with a mass term converging to infinity. Let \((\mathcal{N},g)\) be a complete spin Riemannian manifold and \(\mathcal{K}\subset\mathcal{N}\) a compact domain with smooth boundary. The author looks at the Dirac operator \(D\) on the product Riemannian cylinder \((\mathcal{C},g_\mathcal{C})=(\mathcal{N}\times \mathbb{R},g+dt^2)\) and restricts it to spinors oscillating at speed \(m\in\mathbb{R}\) in the \(t\)-direction, in the sense that \(\nabla_\frac{\partial}{\partial t} \phi= im\phi\). This operator can be described as a Dirac operator on spinors from \(\mathcal{C}\) restricted to \(\mathcal{N}\) as follows: \[ \mathcal{D}_m=\nu(D+im) \] where \(\nu\) is the Clifford multiplication by the \(1\)-form \(\mathrm{d}t\), and \(D\) is the spin Dirac operator on \(\mathcal{N}\) in even dimension, respectively two copies of this operator in odd dimensions. The MIT bag boundary condition at \(\partial \mathcal{K}\) reads \[ i\nu \mathbf{n} \psi=\psi, \] where \(\mathbf{n}\) denotes Clifford multiplication by the exterior unit normal to \(\partial\mathcal{K}\). The Dirac operator \(\mathcal{D}_m\) endowed with this boundary condition is self-adjoint. The author proves that as \(m\to-\infty\), the eigenvalues of \(\mathcal{D}_m^2\) on the compact manifold \(\mathcal{K}\) converge to the eigenvalues of the square of the induced Dirac operator on \(\partial\mathcal{K}\). The author proves also a similar result for a truncated operator on the complete manifold \(\mathcal{N}\). Let \(M\in\mathbb{R}\) and define \[ \mathcal{D}_{m,M}= \nu(D+iM +i(m-M)\chi_{\mathcal{K}}) \] where \(\chi_{\mathcal{K}}\) is the characteristic function of the set \({\mathcal{K}}\). This operator is essentially self-adjoint. When \(M\to\infty\) and \(m\to-\infty\), the eigenvalues of \(\mathcal{D}_{m,M}^2\) converge again to the eigenvalues of \(D_{\partial\mathcal{K}}^2\).