5
$\begingroup$

$\DeclareMathOperator{\res}{res}$Let $X$ be a topological space, $p \in X$ a point, $U \subset X$ an open subset covered by $\bigcup_{i \in I}U_i$, and $S$ a set (or an abelian group). I'm trying to show that the skyscraper sheaf $i_pS$ given by $$ i_pS(U) = \begin{cases} S & \text{if } \, p \in U \\ \{ e \} & \text{else} \, \end{cases} $$ is indeed a sheaf. Here's what I've tried so far.

First I want to show the gluing axiom, so I take sections $a_i \in i_pS(U_i)$ such that $\res^{U_i}_{U_i \cap U_j} a_i = \res^{U_j}_{U_i \cap U_j} a_j$. We need for there to exist a section $a \in i_pS(U)$ such that $\res^U_{U_i}a = a_i$ for all $i \in I$. But here's my problem - since the sections are elements of the set (or abelian group) $S$, I'm not sure what the restriction of one of these sections actually is (as opposed to the clear nature of a restriction when we talk about, say, the sheaf of differentiable functions).

I tried to do it case-wise, so say $p \in U_i$ and $p \notin U_j$ (so $p \notin U_i \cap U_j$) for example. Then $a_j \in \{e\}$ so $a_j =e$, and by our assumption, $\res^{U_i}_{U_i \cap U_j} a_i = e$ which means that $a_i$ is $e$ on the part of $U_i$ which overlaps with $U_j$. Does this mean that $a_i$ must be $e$ on the whole of $U_i$? I feel not, because then by this logic I think we would end up with $i_pS(U)= \{e \}$. But then if the $a_i$ take different values on different parts of $U$, how does one glue them together to construct a valid $a$, and what does this $a$ even look like?

I haven't yet attempted the other axiom for sheaves, but I think I'd run into similar problems with the above reasoning. Thank you for any help.

$\endgroup$
3
$\begingroup$

For simplicity we look at sheaves of abelian groups.

You can describe $i_pS$ as the sheaf of functions $f : U \to S$ with the property that $f(u) = e$ for all $u \neq p$. The restriction is the usual restriction of functions.

Now let $f_i : U_i \to S$ be a family of sections of $i_pS$, which agree on intersections. Then they glue uniquely to a function $f : X \to S$ and obviously $f \in i_pS(X)$.

$\endgroup$
3
  • $\begingroup$ Thanks for your answer. So I suppose this way of thinking about it works because these functions "detect" whether $U$ contains $p$ or not. Say $p \in U$. Then $f(p) \neq e$ and $f(p) \in S$. But $f(p)$ only represents one point of $S$; is this not a problem? The definition of the skyscraper sheaf I gave in the question assigns the whole of $S$ rather than just a point of it when $p \in U$. $\endgroup$
    – mathphys
    Jul 23 '18 at 23:40
  • $\begingroup$ Ah, so I suppose the fact that we can take many possible functions $f$ (all these functions are presumably what makes up the whole sheaf $i_pS(U)$), in particular, for each element of $S$, a function $f$ mapping $p$ to that element of $S$, is what makes this work out? $\endgroup$
    – mathphys
    Jul 23 '18 at 23:47
  • 1
    $\begingroup$ It is of course possible that $f(p) = e$ since $e \in S$ and $f(u) = e$ for all $u \in U$ is in $i_pS(U)$. Yes, there are exactly $|S|$ sections if $p \in U$, and a single section else. $\endgroup$ Jul 24 '18 at 5:04

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.