Verification of example to show surjective maps of sheaves need not surject onto sections in all open sets

60 Views Asked by At

As an exercise in understanding the notion of surjectivity in the category of sheaves, I came up with this example, slightly modifying the standard ones given in my textbooks. I feel like this one is easier to visualize, since it takes place in real spaces (as opposed to the standard examples, drawn from complex analysis, which I'm less familiar with).

I would appreciate verification and any thoughts on the construction.

The idea is to take the circle $\mathbb{S}^1$ with its standard smooth manifold structure. Then, to every point $x$, we designate a distinguished open $U_x$ defined as $\mathbb{S}^1 - \{\bar{x}\}$, where $\bar{x}$ is the antipodal point of $x$, and furthermore, we assign a distinguished chart $\pi :U_x \rightarrow \mathbb{R}$ by stereographic projection from $\bar{x}$. It can be checked that $\pi$ sends $x$ to $0$.

Then, we assign to each $U_x$ the ring of real analytic functions defined on $\pi(U_x)$. (Perhaps it's also important to stipulate that they have infinite radius of convergence?) Informally, each regular function on the circle is given by its Taylor series in the "$\theta$" coordinate.

With this setup of real analytic functions on the circle on a base, we ought to be able to extend this to a sheaf on the circle. (And without needing to sheafify).

The only global sections, then, should be those analytic functions which are $2\pi$-periodic. That is, when we transport a global section $f$ to $\pi(U_x)=\mathbb{R}$ and analytically continue it on the coordinate chart, we should find that it has a period of $2\pi$. Otherwise, as we "wrap it around" the circle, it will fail to match up, and the compatibility requirement in the gluing axiom will not be satisfied.

We can define a sheaf morphism $\phi$ by stipulating that $f(x) \in \mathscr{F}_x$ gets sent to $f(2x)$. We see that it sends each ($2\pi$-periodic) global section to a $\pi$-periodic global section. And so while the germs at every point are surjected on ($\phi(\sum c_n \frac{1}{2^n} \theta^n) = \sum c_n \theta^n$), there is no global section which maps to a global section with fundamental period $2\pi$.