Topological calculation of the phase of the determinant of a non self-adjoint elliptic operator (Q2575394): Difference between revisions

Let \(D\) be an elliptic partial differential operator of positive order on the space of smooth sections to a vector bundle \(E\) over a smooth compact manifold \(M\). Let \(\theta\) be an Agmon angle for \(D\). This means that the symbolic spectrum of \(D\) avoids the complex cone \(L_\theta:=\{z:\arg(z)\in[\theta-\varepsilon,\theta+\varepsilon]\}\) for some \(\varepsilon>0\). Using the logarithm defined on the complement of this cone, one can define the associated zeta function \(\zeta(s,D)\) and the regularized functional determinant \(\det_\theta(D):= \exp\{-\partial_s\zeta_\theta(s,D) | _{s=0}\}\). By replacing \(\varepsilon\) by a smaller value if need be, one has that there are only a finite number of eigenvalues of \(D\) in the complex cone \(L_\theta\). The authors show \(\det_{\tilde\theta}(D)=\det_\theta(D)\) if \(| \theta-\tilde\theta| <\varepsilon\); thus the determinant is independent of small perturbations of the defining Agmon angle. Suppose that the spectrum of \(D\) is symmetric with respect to the imaginary axis. The authors show that the determinant is real. For many geometrical operators, the phase of the determinant is a topological invariant -- the authors calculate the phase of such determinants in many examples in topological terms.