Eigenvalue estimates for the Dirac operator and harmonic 1-forms of constant length (Q1876801)

Let \((M^n,g)\) be a compact \(n\)-dimensional spin manifold. Assume the scalar curvature \(\tau\) is strictly positive and let \(\tau_0:=\inf_M\tau>0\). Let \(\lambda\) be any eigenvalue of the Dirac operator. Then the Friedrich inequality [see \textit{Th. Friedrich}, Math. Nachr. 97, 117--146 (1980; Zbl 0462.53027)] states that \[ \lambda^2\geq{{n}\over{4(n-1)}}\tau_0\,. \] Furthermore, there are manifolds where the inequality is sharp; this is equivalent to the existence of a Killing spinor on \((M^n,g)\). However, this inequality is not sharp if additional geometric hypothesis are imposed. The authors generalize a result of \textit{B. Alexandrov, G. Grantcharov} and \textit{S. Ivanov} [J. Geom. Phys. 28, No. 3--4, 263--270 (1998; Zbl 0934.58026)] and prove: Theorem. Suppose that \(n\geq3\) and that \((M^n,g)\) admits a non-trivial harmonic \(1\)-form \(\theta\) of constant length 1. Then \(\lambda^2\geq{{(n-1)}\over{4(n-2)}}\tau_0\). Equality holds if and only if \(\theta\) is parallel and if any eigenspinor corresponding to \(\lambda\) satisfies the following Killing type equation \[ \nabla_X\Psi +{\lambda\over{n-1}} \{X\cdot\Psi-(X,\theta) \theta\cdot\Psi\}=0\,. \]