$η$ invariant of massive Wilson Dirac operator and the index
Shoto Aoki, Hidenori Fukaya, Mikio Furuta, Shinichiroh Matsuo, Tetsuya Onogi, Satoshi Yamaguchi
TL;DR
This work reframes the lattice Dirac index problem in terms of $K$-theory, showing that the index defined by the overlap operator coincides with the $η$-invariant of the massive Wilson Dirac operator due to the suspension isomorphism between $K^0(\text{point})$ and $K^{1}(I,\partial I)$. It proves that for sufficiently small lattice spacing, the $η$-invariant of the Wilson operator matches the continuum Dirac index, thereby making the Ginsparg-Wilson relation and exact lattice chiral symmetry unnecessary for encoding gauge-field topology. The authors establish a concrete main theorem bounding the lattice spacing below which the lattice and continuum $K^{1}$-classes coincide, and discuss extensions to domains with boundaries, real Dirac operators, and curved backgrounds via KO-theory and APS-type indices. The results broaden the toolkit for lattice gauge theory topology beyond the overlap construction, highlighting the robustness of the $K$-theoretic viewpoint and spectral-flow centrality in nonperturbative topology.
Abstract
We revisit the lattice index theorem in the perspective of $K$-theory. The standard definition given by the overlap Dirac operator equals to the $η$ invariant of the Wilson Dirac operator with a negative mass. This equality is not coincidental but reflects a mathematically profound significance known as the suspension isomorphism of $K$-groups. Specifically, we identify the Wilson Dirac operator as an element of the $K^1$ group, which is characterized by the $η$-invariant. Furthermore, we prove that, at sufficiently small but finite lattice spacings, this $η$-invariant equals to the index of the continuum Dirac operator. Our results indicate that the Ginsparg-Wilson relation and the associated exact chiral symmetry are not essential for understanding gauge field topology in lattice gauge theory.
