Rigidity of higher elliptic genera (Q1922785)

General equivariant elliptic genera for a closed connected manifold \(M\) of even dimension \(2m\) with a nontrivial smooth \(S^1\)-action are defined as the indices of the Dirac operator \(D_M\) twisted by a bundle depending on a positive energy representation \(E\) of the central extension of \(LG\), the loop group of a simple, simply connected compact Lie group, and an \(S^1\) equivariant principal bundle \(V\) over \(M\) with structure group \(G\). The authors prove that such elliptic genera are rigid if the first equivariant Pontryagin classes of \(M\) and \(V\) are equal up to a multiple given by the level of \(E\). The special case \(G=\text{Spin}(2m)\) has been settled by the second author [in J. Differ. Geom. 41, No. 2, 343-396 (1995; Zbl 0836.57024)]. If \(M\) is not simply connected, higher elliptic genera can be defined, and the rigidity results extend to this situation provided that the \(S^1\)-action has fixed-points. If the action is fixed-point free, the problem is more involved. It is then related to the vanishing of higher elliptic genera. In particular, the authors raise the question of analogues of the Gromov-Lawson-Rosenberg conjecture: do higher elliptic \(\widehat{A}\)-genera for closed spin manifolds of positive curvature vanish? Finally, they give an example which shows that the analogue of the Novikov conjecture is definitely wrong.