A proof of the axial anomaly (Q1079861)

This article is written mostly in physical dialect and fluency on the part of the reader is assumed. ''The axial anomaly is the variation of the determinant of the Dirac operator under axial changes of the gauge field.'' The author wishes ''to show that in fact the exact form of the anomaly is given from topological arguments'' but according to the abstract, this is done via a local formula. Let \(A^{\pm}\) be connections on a principal G-bundle \(P\to M\) over an \(n(=even)\)- dimensional Riemannian spin manifold, let \(V\to M\) be an associated (vector?) bundle and let \(S=S^+\otimes S^-\) be the bundle of spinors of M. Define \({\mathbb{D}}_{A^+,A^-}: \Gamma (S\otimes V)\to \Gamma (S\otimes V)\) as the appropriate combination of the spinor covariant derivative determined by \(A^{\pm}\). Let \(a_ n\in End(V\otimes S)\) be the nth term in the asymptotic expansion of \(\exp (- T{\mathbb{D}}_{A^+,A^-})\). If projection from \(\Gamma\) (S) to \(\Gamma (S^+)\) is written as \(1/2(1+\epsilon)\), then \(A=trace \epsilon (a_ n(A^+,A^-)+a_ n(A^-,A^+))\) is a differential form on M given by the polynomial in \(A^{\pm}\), \(g,g^{-1}\) and their derivatives. The implication of the rest of the article is that the symmetrization (necessary to ensure Bose symmetry) of \(\underset \tilde{} A\) is the axial anomaly. Now let \({\mathcal Q}\) denote the space of connections on P and \({\mathfrak G}\), the group of gauge transformations (vertical automorphisms) of P. A vertical 1-form \(\omega\) on \({\mathcal Q}\times {\mathcal Q}\to {\mathcal Q}\times {\mathcal Q}/{\mathfrak G}\times {\mathfrak G}\) is given by \(\int_{M}Tr(X- Y)\underset \tilde{} A\) and restricts to a closed form on \({\mathfrak G}\times {\mathfrak G}\). Let \(Z_ A\) be the index bundle for \(\phi_ A\) over \(D^ 2\) with a specific connection \(\phi\). Lemma 3. \[ \int_{D^ 2}c_ 1(\phi)(Z_{A^+})-\int_{D^ 2}(\phi)(Z_{A^-})=\frac{1}{i\pi}\int_{\partial D^ 2}\omega (A^+,A^-). \] (Apparently the author means \(c_ 1(\phi A^{\pm})\) where \(\phi A^{\pm}\) is the corresponding connection on \(Z_{A^{\pm}}.)\) Trace \(\phi\) given a U(1) connection on a line bundle over \({\mathcal Q}/{\mathfrak G}\) associated to Index \({\mathbb{D}}_ A\) and hence determines a class \(c_ 1(Tr\phi)\in H^ 2({\mathcal Q}/{\mathfrak G})\approx H^ 1({\mathfrak G})\). The families index theorem gives \[ c_ 1(\phi)=ch \eta \hat A(g) \] where \(\eta\) is an \({\mathfrak G}\)-equivariant connection on \({\mathcal Q}\times P\to {\mathcal Q}\times M\). [Apparently Tr is understood now and \(\hat A(g)\) is what is usually written \(\hat A(M)\).] The author gives a complicated formula to show that the suspension of \(c_ 1(\phi)\) into \(H^ 1({\mathfrak G})\) is given by Chern's transgression formula. Lemma 4 shows that a vertical 1-form \({\hat \omega}\) on \({\mathcal Q}\times {\mathcal Q}\) of the type \(\int_{M}Tr(X-Y)\) \(\sigma\) considered above determines a trivial cohomology class in \(H^ 1({\mathfrak G}\times {\mathfrak G})\) only if the symmetrization of \(\sigma\) is zero. The details are sufficiently complicated that the converse is not transparent; can the anomaly be identified with \(\omega\) ?