There are different models of odd $K$-theory. In one case, one takes the group $U=\lim\limits_{\longrightarrow}U(n)$ as classifying space. Similarly, if $\mathcal U$ denotes the unitary group of a separable Hilbert space, one can consider the subspace $$\mathcal U_0=\bigl\{\,g\in\mathcal U\bigm|g+\mathrm{id}\text{ is Fredholm}\,\bigr\}\;.$$

In another model, one considers instead the space $\mathcal F$ of selfadjoint Fredholm operators on a Hilbert space that have an infinite number of positive as well as of negative eigenvalues. Both spaces are related by a Cayley transformation $$\mathcal F\owns D\mapsto \frac{D-i}{D+i}\in\mathcal U_0\subset\mathcal U_0\;.$$

Characteristic classes are cohomology classes on these classifying spaces. It is well-known that the cohomology ring is the exterior algebra over $\mathbb Z$ with one generator $\gamma_k$ in each odd degree $2k-1$. There is a de Rham type description for maps $u\colon M\to U$ of these generators of the form $$\gamma_k(u)=c_k\,\mathrm{tr}\bigl((u^{-1}\,du)^{2k-1}\bigr)$$ with appropriate constants $c_k\in\mathbb C$.

On the other hand, for a map $D\colon X\to\mathcal F$, the characteristic classes of $[D]\in K^1(X)$ should be determined entirely by the way that $\mathrm{spec}(D)\cap(-\varepsilon,\varepsilon)$ (and maybe the corresponding eigenspaces) behave for some small $\varepsilon>0$.

My question is, how are these descriptions related? In there some book/article that makes characteristic classes of selfadjoint Fredholm operators more explicit and relates them to those of maps to $U$?

  • $\begingroup$ Is it obvious that $\mathcal U _{0}$ is a subgroup? $\endgroup$ – Ali Taghavi Mar 9 '16 at 18:41
  • 1
    $\begingroup$ @AliTaghavi I am not entirely sure. But maybe it does not have to be a subgroup to be a valid classifying space. I think you get a subgroup if you demand that $1$ is the only essential eigenvalue. $\endgroup$ – Sebastian Goette Mar 9 '16 at 19:00

I know something about it when $X$ is compact. Since $U$ is the classifying space of $K^1$, for $x\in K^1(X)$, there exists $u\in C(X, U)$, such that $[u]=x$. In fact, we could choose $N$ large enough, such that $u\in C(X, U(N))$.

For $(b,t,v)\in X\times [0,1]\times \mathbb{C}^N$, the relation $(b,0,v)\sim (b,1, u(b)v)$ makes a vector bundle $W$ over $X\times S^1$. Let $U=X\times S^1\times \mathbb{C}^N$ be the trivial bundle. Then $[W]-[U]\in K^0(X\times S^1)$ corresponds to $[u]\in K^1(X)$.

Let $\nabla^u=d+t\cdot u^{-1}du$. Then $\nabla^u$ is a connection on $W$ over $X\times S^1$ under the paste above. $$\mathrm{ch}([u]):=\int_{S^1}\mathrm{ch}(W)=\left[\int_{S^1}\mathrm{Tr}\left[\exp\left(\frac{(\nabla^{u})^2}{2\pi i}\right)\right]\right].$$ (There is a sign notation here for the integral on odd dimensional manifold.) Calculate it carefully, $$\mathrm{ch}([u])= \mathrm{Tr}\left[\sum_{n≥0}\frac{(−1)^n} {(2πi)^{n+1}}\frac{n!}{ (2n + 1)!}(u^{−1}du)^{2n+1}\right].$$ This is the familiar form in the paper of Getzler: The odd chern character in cyclic homology and spectral flow

For Fredholm type interpretation, we could only consider the Dirac type case. Let $\pi:Z\rightarrow X$ be a submersion with odd-dimensional compact Spin fiber $Y$. Let $E$ be a complex vector bundle over $Z$. We have the fiberwise Dirac operator $D_Y^E$ and $\mathrm{ind}(D_Y^E)\in K^1(X)$.

For any $x\in K^1(X)$, there exists such submersion with appropriate geometric structure such that $\mathrm{ind}(D_Y^E)=x$. See Section 4.2.3 in the book Index theory, eta forms and Deligne cohomology. In fact, choose submersion $X\times S^1\rightarrow X$ above, we have $\mathrm{ind}(D_{S^1}^W)=[u]\in K^1(X)$.

For the Caylay transform, maybe you could find something in Section 5 of the paper of Melrose-Piazza: An index theorem for families of Dirac operators on odd-dimensional manifolds with boundary.

  • $\begingroup$ Thanks for the nice list of references. The equation $\operatorname{ind}(D^W_{S^1})=[u]\in K^1(X)$ is already pretty close to what I am looking for. But is there a nice relation to the Cayley transform? The transformation in Melrose-Piazza's article seems to be something different. $\endgroup$ – Sebastian Goette May 19 '16 at 20:43
  • 1
    $\begingroup$ @sebastian-goette Sorry that I made a mistake. The transformation in Melrose-Piazza's paper is the relation to the definition of $K^1$ group in Atiyah's paper: Index theory for skew-adjoint Fredholm operators, not the Cayley transform. I don't know the relation to the Cayley transform. Maybe there is something in Quillen's article: Superconnection character forms and the Cayley transform. But I haven't read this. $\endgroup$ – Bo Liu May 20 '16 at 9:53

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.