3
$\begingroup$

Atiyah and Singer proved that the nontrivial component of the set of skew-adjoint Fredholm operators $ \hat{\mathcal{F}_{*}}(\mathscr{H})$ is homotopic to the loop space of Fredholm operators $\Omega\mathcal{F}(\mathscr{H})$.

But in their paper and in other sources what they prove is homotopy to a so-called relative loop space $\Omega(\mathcal{L},\mathcal{-C})$ where $\mathcal{L}$ is a retraction of the group of units of $\mathcal{B}(\mathscr{H})$, the set of bounded operators of a complex separable Hilbert space $\mathscr{H}$ and $\mathcal{-C}$ are operators of the form $I - T$, where $T$ is compact.

1) Does anyone know how to pass from the relative loop space $\Omega(\mathcal{L},\mathcal{-C})$ to $\Omega\mathcal{F}(\mathscr{H})$?

A related question is the following: By Atiyah-Singer $\mathbb{Z}\approx\mathcal{K}^{-1}(point) \approx [point,\hat{\mathcal{F}_{*}}(\mathscr{H})]$

2) Since $\forall A(t) \in \hat{\mathcal{F}_{*}}(\mathscr{H}), (0\leq t\leq 1)$, $Ind\, A(t) = 0 \,\,\forall t$ and $Ker\,A(t) \in \mathbb{Z}^{+} \,\,\forall t$, what property of $A(t)$ determines the equivalence class it belongs to in $\mathbb{Z}$?

Thank you for reading

$\endgroup$
0

1 Answer 1

0
$\begingroup$

On a complex Hilbert space $H$ we can easily identify the space of skew-adjoint operators with the space of self-adjoint operators. (Multiplication by $\sqrt{-1}$ will do the trick. The space $\newcommand{\FS}{\mathscr{FS}}$ $\FS$of Fredholm selfadjoint operators on $H$ has three components $\FS_{\pm}$, $\FS_*$. The first two are homotopically trivial, the third $\FS_*$ classifies $K^1$. In particular, $\newcommand{\bZ}{\mathbb{Z}}$ $\pi_1(\FS_*)\cong \bZ$ and there is a canonical isomorphism $\pi_1(\FS_*)\to\bZ$ called the spectral spectral flow. You can visualize this isomorphism as follows.

The space $\FS_*$ is homotopy equivalent with the infinite dimensional Banach manifold $\newcommand{\L}{\mathscr{L}}$ $\L(H)$ consisting of the subspaces $L\subset H\oplus H$ satisfying the conditions

  • The pair $(L, H\oplus 0)$ is a Fredholm pair, i.e. $\dim L\cap (H\oplus 0)<\infty$ and $L+(H\oplus 0)$ is closed in $H\oplus H$.
  • The subspaces $L$ is Lagrangian, i.e., $L^\perp=JL$, where $J:H\oplus H\to H\oplus H$ is defined by $J(x,y)=(-y,x)$.

There is a natural map $\FS(H)\to \L(H)$ that associated to the operator $T$ its graph $\Gamma_T\subset H\oplus H$. This map is a homotopy equivalence.

As I mentioned above, $\L(H)$ is a Banach manifold and, moreover, it is equipped with a Schubert-like stratification by smooth strata with finite codimension. This stratification contains a single codimension-1 stratum $\newcommand{\M}{\mathscr{M}}$ $\M$ that is naturally co-oriented. (This stratum is sometime called the Maslov divisor.) Given smooth map $\alpha:S^1\to \L(H)$ we denote by $\mu(\alpha)$ its intersection number with $\M$. The resulting map $\mu:\pi_1(\L(H))$\to\bZ$ is an isomorphism. There are similar descriptions for the isomorphisms

$$\pi_{2k-1}(\L(H))\to\bZ,\;\;k=1,2,\dotsc. $$

For details, I refer to Daniel Cibotaru's dissertation.

$\endgroup$
1
  • $\begingroup$ You're right. I'am always thinking of reduced $K$-theory. $\endgroup$ May 15, 2015 at 18:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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