$\DeclareMathOperator\spin{spin}\DeclareMathOperator\ch{ch}\DeclareMathOperator\ind{ind}$In the paper Čadek, Crabb, and Vanžura - Obstruction theory on 8-manifolds, the authors discussed the "$\spin^c$-index" for a $\spin^c$ manifold $M$ (display (3.1) of the paper): $$y\in K^0(M)\mapsto \ind(y)=(e^{c/2}\hat{A}(\tau M)\ch(y))[M]\in\mathbb{Z}, $$ where $c$ is the $\spin^c$ class, $\tau M$ is the tangent bundle of $M$, and $\hat{A}$ is the Hirzebruch signature: $$\hat{A}(\tau M)=1-p_1(\tau M)+\dotsb. $$ I was wondering if there is a more coherent context in which the invariant $\ind(y)$ is discussed.
-
3$\begingroup$ If if you think of $y$ as defined by a complex vector bundle, then $\mathrm{ind}(y)$ is the index of the $spin^c$ Dirac operator corresponding to the $spin^c$ structure $c$ and twisted by the vector bundle $y$ . This idea goes back to work of Atiyah and Hirzebruch $\endgroup$– Liviu NicolaescuCommented Jul 29, 2020 at 18:01
-
$\begingroup$ @Liviu Nicolaescu Thank you! $\endgroup$– Xing GuCommented Jul 30, 2020 at 8:58
1 Answer
I suggest consulting section 26 of Hirzebruch's Topological Methods in Algebraic Geometry and the references therein. In particular, it contains the following statement:
Theorem 26.1.1. Let $d$ be an element of $H^2(X, \mathbb{Z})$ whose reduction mod 2 is the Whitney class $w_2(X)$, and $\eta$ a continuous $GL(q, \mathbb{C})$-bundle over $X$. Then
$$\hat{A}(X, \tfrac{1}{2}d, \eta) = \chi^m\left[e^{\frac{1}{2}d}\cdot\operatorname{ch}\eta\cdot\sum_{j=0}^{\infty}\hat{A}_j(p_1, \dots, p_j)\right]$$
is an integer.
References to proofs are given in subsections 26.3-26.5.