Let $D=i\frac{d}{dx}$ be the Dirac operator on $\mathbb{R}$, acting on the spinor bundle $\mathbb{R}\times\mathbb{C}$. The bounded operator $F=\frac{D}{\sqrt{D^2+1}}$ has a coarse index

$$\text{Ind}(F)\in K_1(C^*(|\mathbb{R}|)),$$

where $C^*(|\mathbb{R}|)$ is the Roe algebra of the metric space (as opposed to the group) $\mathbb{R}$.

It has been stated in various places (for example, Higson & Roe's book "Analytic $K$-Homology" chapter 12) that this coarse index in fact generates the group $K_1(C^*(|\mathbb{R}|))\cong\mathbb{Z}$.

However, I haven't seen this explicitly computed anywhere, and it would be nice and instructive to see a detailed calculation showing. For example, the analogous fact that the Dirac operator on $\mathbb{R}^n$ has non-trivial coarse index can be used to show that there exists no Riemannian metric of positive scalar curvature on $\mathbb{T}^n$.

To be specific, my question is just about the one-dimensional case, from which I would assume higher-dimensional computations follow:

Question: Show that $\text{Ind}(F)$ generates $K_1(C^*(|\mathbb{R}|))\cong\mathbb{Z}$.

So if there are experts in the area reading this who wouldn't mind sharing their thoughts, that would be very helpful.


There are a number of ways to do this calculation, but at risk of shamelessly plugging my own work there is a nice way to see it using a Mayer-Vietoris principle.

Decompose $\mathbb{R}$ as the union of the rays $\mathbb{R}^+ = [0, \infty)$ and $\mathbb{R}^- = (-\infty, 0]$, intersecting at a point. This decomposition admits Mayer-Vietoris sequences in both K-homology and and coarse K-theory, and there is a coarse index map from the former sequence to the latter. This gives a commuting square:

$$ \require{AMScd} \begin{CD} K_1(\mathbb{R}) @>Index>> K_1(C^*(\mathbb{R})) \\ @VVM{V}V @VVM{V}V \\ K_0(\text{point}) @>Index>> K_0(C^*(\text{point})) \end{CD} $$

Now, the Mayer-Vietoris boundary maps are isomorphisms because the rays $\mathbb{R}^\pm$ have trivial K-homology and trivial coarse K-theory - this follows from homotopy invariance of K-homology and an Eilenberg swindle argument for coarse K-theory. The group $K_1(\mathbb{R})$ is generated by the K-homology class of the Dirac operator, and a quick calculation shows that the Mayer-Vietoris boundary map in this case is just the ordinary boundary map induced by the inclusion of a point into $\mathbb{R}$. A difficult but standard K-homology calculation shows that this boundary map sends the Dirac class to a generator of $K_0(\text{point})$, i.e. a Fredholm operator of index 1.

That completes the calculation, up to the construction of the Mayer-Vietoris sequences and and the index maps between them. I worked all that out in my PhD thesis, and you can see a preprint here. By inductively chopping $\mathbb{R}^n$ into half-spaces you get the Gromov-Lawson theorem about the torus for free, and the argument has a nice and geometric flavor to it. That said, it's hard to turn the argument into explicit formulas - calculating the boundary of the Dirac class is hard, and I doubt it's any cleaner to go the other way around the diagram.

  • $\begingroup$ Thanks - do you have a reference for the difficult but standard calculation that you refer to, showing that the $K$-homology class of the Dirac operator on $\mathbb{R}$ maps to an operator of index $1$ under $MV$? $\endgroup$ – ougoah Jun 15 '19 at 15:31
  • $\begingroup$ The assertion that the ordinary K-homology boundary map sends the Dirac class to $1$ is basically proposition 9.6.6 in Higson and Roe's Analytic K-homology, though the proof with all details uses most of the machinery built in chapters 8 and 9. Passing from this to a calculation with Mayer-Vietoris boundary maps is sort of folk wisdom in operator K-theory; I honestly don't know a place where it's all written out other than my PhD thesis, available here: pwsiegel.github.io/resources/paul-siegel-phd-thesis.pdf. This is not a claim of priority; someone else might know another reference. $\endgroup$ – Paul Siegel Jun 15 '19 at 17:29

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.