The Callias index formula revisited (Q267846)

In the later 1970's, \textit{C. Callias} [Commun. Math. Phys. 62, 213--234 (1978; Zbl 0416.58024)] found a formula for the Fredholm index of a particular first-order differential operator related to a supersymmetric Dirac-type operator additively perturbed by a potential. The approach to calculating the Fredholm indices initiated by Callias had a profound influence on theoretical physics. It is remarkable for its simplicity and it is consistent with the more complicated Hörmander-Fedosov formula. These lecture notes aim at providing an accessible proof of the Callias index formula. For a given spatial dimension \(n\in\mathbb N\), let the elements of the Euclidean Dirac algebra be denoted by \(\gamma_{j,n}\), \(j\in \{1,\dots,n\}\). For \(\hat{n}\in\mathbb N\) with the elements \(\gamma_{j,n}\) in place, one can introduce the constant coefficient first-order differential operator \(Q\) in \(L^2(\mathbb R^n)^{2^{\hat{n}}}\) by \(Q:=\sum_{j=1}^n\gamma_{j,n}\partial_j\) with the domain \(H^1(\mathbb R^n)^{2^{\hat{n}}}\), where \(H^1(\mathbb R^n)\) is the standard Sobolev space. The lecture notes under review are devoted to the proof of the following main result. Theorem 1.1. Let \(n\in\mathbb N\) be odd and \(n\geq 3\). For \(d\in\mathbb N\), assume that \(\Phi\) is a matrix function in \(C_b^2(\mathbb R^n;\mathbb C^{d\times d})\) such that \(\Phi(x)=\Phi(x)^\ast\) for \(x\in\mathbb R^n\), there exist \(c>0\) and \(R\geq 0\) such that \(|\Phi(x)|\geq cI_d\) for \(x\in\mathbb R^n\) satisfying \(|x|\geq R\), and there exists \(\varepsilon>1/2\) such that, for all \(\alpha\in\mathbb N_0^n\) with \(|\alpha|<3\), there is \(\kappa>0\) such that \[ \|(\partial^\alpha\Phi)(x)\|\leq\begin{cases} \kappa(1+|x|)^{-1}, & |\alpha|=1, \\ \kappa(1+|x|)^{-1-\varepsilon}, & |\alpha|=2,\quad x\in{\mathbb R}^n.\end{cases} \] Then the closed operator \(L:=Q\otimes I_d+\Phi\) in \(L^2({\mathbb R}^n)^{2^{\hat{n}}d}\) is Fredholm with index given by the formula \[ \text{ind}(L)= \lim_{\Lambda\to\infty}\frac{C_n}{\Lambda} \sum_{i_1,\dots,i_n=1}^n\varepsilon_{i_1\dots i_n} \int\limits_{\Lambda {\mathbb S}^{n-1}}\text{tr} (U(x)(\partial_{i_1}U)(x)\dots(\partial_{i_{n-1}}U)(x))x_{i_n}\text{d}^{n-1}\sigma(x), \] where \[ C_n:=\frac{1}{2} \left(\frac{i}{8\pi}\right)^{(n-1)/2}\frac{1}{[(n-1)/2]!} \] and \(U(x):=|\Phi(x)|^{-1}\Phi(x)=\text{sgn}(\Phi(x))\) for \(x\in{\mathbb R}^n\). Here, \(\varepsilon_{i_1\dots i_n}\) denotes the totally anti-symmetric symbol in \(n\) coordinates, \(\text{tr}(\cdot)\) represents the matrix trace in \({\mathbb C}^{d\times d}\), and \(\text{d}^{n-1}\sigma(\cdot)\) is the surface measure on the unit sphere \({\mathbb S}^{n-1}\) of \({\mathbb R}^n\). Under these hypotheses on the potential \(\Phi\), the above result is new, all previous derivations of the Callias index formula were made under additional hypotheses on \(\Phi\), in particular, asymptotically \(\Phi\) had to be homogeneous of order zero. A generalization of the index formula to certain classes of non-Fredholm operators \(L\) is also presented. For such a non-Fredholm operator, the above index formula gives its generalized Witten index based on a resolvent regularization scheme. The authors at first follow the original strategy of Callias, however the bulk of the arguments differ from those in [loc. cit.] as some of the claims of that paper can be disproved. The given proof requires a number of additional steps not present in the paper by Callias.