In classical surgery theory, there is a map $$L_{n+1}(\pi_1M)\to S(M^n)$$

Element in $L_{n+1}(\pi_1M)$ is realized as surgery obstruction of a surgery problem to $M\times I$ with one boundary piece the identity map and the other a homotopy equivalence (Wall's realization). The map is defined by sending the element in $L$ to the boundary piece which is a homotopy equivalence (a structure on $M$).It is NOT clear to me if the domain manifold of the structure is homeomorphic to $M$.

In Homology spheres and fundamental group

(when $n+1$ is even) Danny Ruberman commented that "The effect of the action is to change the multisignature, and hence it changes the homeomorphism type of M"

What's the correct reference if i want to understand some details about the effect of Wall's realization on multisignature?

  • $\begingroup$ I would look at Wall's book (search for "multisignature" in the index) and also Wall's paper "On the classification of hermitian forms VI. Group rings". $\endgroup$ – Igor Belegradek Jan 23 at 12:13

There is perhaps some confusion over the terminology. Wall (chapter 13A) uses the term multisignature to denote a collection of invariants of certain Hermitian forms over group rings, giving rise to a function from $L_{2k}(\pi) \to \mathbb{Z}^n$. In that chapter, he interprets the multisignature in terms of equivariant signatures. Such signatures occur in many places in geometric topology, for instance as the Tristram-Levine signatures and Casson-Gordon invariants of knots.

One main use of the multisignature comes in the study of odd-dimensional manifolds. Given such a manifold $Y^{2k-1}$ and a finite regular covering with covering group $G$, classified by a homomorphism $\phi:\pi_(Y) \to G$, then one shows that some multiple (disjoint union) $N\cdot(Y,\phi)$ extends to a $2k$-manifold $(X,\Phi)$. Then (suitably normalized) the equivariant signature of $(X,\Phi)$ is an invariant $\rho(Y,\phi)$.

This can be packaged in various ways, see for instance chapter 14E in Wall, where for a manifold with finite fundamental group $G$, one obtains an invariant $\rho$. Technically one has to choose an identification of $\pi_1(Y)$ with G, and $\rho$ depends on this identification. A priori, the fact that one gets an invariant depends on the Atiyah-Singer index theorem, and so is really an invariant up to diffeomorphism. But a bordism argument (Wall 14B) shows that it is actually a homeomorphism invariant.

Putting these notions together, the way that the invariant $\rho$ is defined tells you that if one acts on (the identity map from $Y$ to itself) by an element $A \in L_{2k}(\pi_1(Y))$ then the invariant $\rho$ of the manifold $Y'$ at the other end of the resulting normal cobordism satisfies $\rho(Y') - \rho(Y) = $ multisignature of $A$. If this multisignature is non-trivial, then there is no homeomorphism of $Y$ and $Y'$ that respects the identification of their fundamental groups given by the normal cobordism. If you want to argue that they are simply not homeomorphic you need to look a little more closely.

  • $\begingroup$ Is there an explicit example in the literature s.t. the action of $L(\pi_1 M)$ on $S(M)$ is nontrivial, but the homeomorphism type of the domain manifold is not changed? $\endgroup$ – student Jan 25 at 12:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.