15
$\begingroup$

This is an exercise of Ravi Vakil that I solved by a very trivial argument without using the hint. For this reason I'm worried that I might have missed something.

If $f$ is a function on a locally ringed space $X$, show that the subset of $X$ where $f$ doesn't vanish is open. (Hint: show that if $f$ is a function on a ringed space $X$, show that (sic) the subset of $X$ where the germ of $f$ is invertible is open)

Solution: Assuming that the first sentence means "$f$ is an element of $\mathcal O_X$", and that $\mathcal O_X$ is a ring of continuous functions on $X$, we just pull back the closed set $\{0\}$ to $X$ via $f^{-1}$. Under a continuous function the pull back of a closed set is closed, and therefore the complement of the locus where $f$ doesn't vanish is closed.

Questions:

1) is my solution correct?

2) I know it is true for the special example of schemes, but for a ringed space in general why is the locus of where a function $f$ doesn't vanish the same of where the germ of $f$ is invertible? The most I can say is that this is true for locally ringed spaces (and not for just ringed spaces as the hint suggests), because if the germ of a function is non-zero at a point $p$ in a locally ringed space $X$, $f_p$ can't belong to the maximal ideal $\mathfrak m_p$. For if it did, then under the homomorphism $ev_p: \mathcal O_{X,p} \rightarrow k$ the maximal ideal of $k$, which is $(0)$, would not pull back to $\mathfrak m_p$ (it would miss $f_p$).

$\endgroup$

2 Answers 2

21
$\begingroup$

There is a subtle and interesting point in this context that one really has to understand:
let me try to be as clear as possible rather than concise.

a) Let $(X,\mathscr O_X)$ be a ringed space that is not necessarily locally ringed.
Then given $f\in \mathscr O_X(X)$, we can ask whether the germ $ f_x\in \mathscr O_{X,x}$ at $x\in X$ is invertible i.e. if $f_x\in \mathscr O_{X,x}^\star$.
Note carefully that the ring $\mathscr O_{X,x}$ is absolutely arbitrary and should not be assumed local.

In this set-up the set of points $x\in X$ with $f_x\in \mathscr O_{X,y}^\star$ is open:
Indeed if $f_x\in \mathscr O_{X,x}^\star$ there exists a germ $g_x$ with $f_xg_x=1\in \mathscr O_{X,x}$ and thus on a suitable open neighbourhood $U$ of $x$ there exists a representative $g\in \mathscr O_X(U)$ of $g_x$ satisfying $f|U\cdot g=1\in \mathscr O_X(U)$.
It is then clear that for all $y\in U$ we also have $f_y\in \mathscr O_{X,y}^\star$ since $f_y\cdot g_y=1\in \mathscr O_{X,y}$ : the promised openness has been proved .

As a digression, let me add that the reasoning above easily implies (by covering $X$ by suitable open $U$'s and using the unicity of inverses in a ring) that a global section $f\in \mathscr O_X(X)$ is invertible if and only if all its germs are invertible: $$f\in \mathscr O_X(X)^\star \iff (\forall x\in X) \; f_x\in \mathscr O_{X,x}^\star \quad (\bigstar)$$

$\\$

b) Suppose now that $(X,\mathscr O_X)$ is a locally ringed space.
Each point $x\in X$ then acquires a residue field $\kappa(x)= \mathscr O_{X,x}/\mathfrak m_x$ and a function $f\in \mathscr O_X(X)$ now receives a value $f(x)=\text {class} (f_x) \in \mathscr O_{X,x}/\mathfrak m_x=\kappa(x)$.
This value is close to the classical interpretation of the value of a function but it is crucial to understand the enormous difference between $f_x$ and $f(x)$.
The charm of local rings is that invertibility of $f_x$ is now exactly equivalent to the rather concrete requirement of having non-zero value: $$f_x\in \mathscr O_{X,x}^\star \iff f(x)\neq 0\in \kappa(x) \quad \quad (\bigstar \bigstar) $$ In particular, taking $(\bigstar)$ into account, we get: $f\in \mathscr O_X(X)^\star \iff (\forall x\in X) \; f(x)\neq 0\in \kappa(x)$

$\\$

c) So the set of points where $f(x)\neq 0$ is open thanks to the above equivalence $(\bigstar \bigstar)$ and the openness result in a).
As an aside, I find it a little intriguing that the trivial-looking assertion of openness of $f(x)\neq 0$ requires such a roundabout proof.

$\endgroup$
3
  • $\begingroup$ Thank you Georges! It's always great to receive a detailed explanation :-) $\endgroup$
    – Rodrigo
    Jan 21, 2014 at 21:33
  • $\begingroup$ Dear Rodrigo, you are welcome. Actually this is a subject that took me some time to really understand and organize. $\endgroup$ Mar 16, 2014 at 7:39
  • $\begingroup$ @GeorgesElencwajg How do we show that a global section $f\in \mathscr O_X(X)$ is invertible if and only if all of its germs are invertible? $\endgroup$ Jan 13, 2021 at 3:08
6
$\begingroup$

When he says function, he refers to any $f \in \mathcal{O}_X(U)$

1) No, $\mathcal O_X$ is just a sheaf and $f$ need not be a function at all.

2) To vanish at $p$ means that the germ $[f]_x$ is in the maximal ideal of the local ring $\mathcal O_{X,p}$

3) To be invertible at $p$ means that the stalk $[f]_x$ is in $\mathcal O_{X,p}^*$, a unit.

Now:

If $f$ is invertible at $p$ means that $[f]_p [g]_p=[1]_p$. By the definition this holds on a open nhood $f|_V g|_V = 1|_V$ so that the invertible locus is open.

Now if $\mathcal O_{X,p}$is a local ring and $[f]_x$ does not vanish, so $[f]_x \notin \mathfrak{m}_x$ is invertible

$\endgroup$
4
  • $\begingroup$ Your (2) is incorrect: $f$ vanishes at $p$ if it is zero in the residue field of $\mathscr{O}_{X, p}$. $\endgroup$
    – Zhen Lin
    Jan 21, 2014 at 14:09
  • $\begingroup$ Thank you Blah, but tapping in to ZhenLin's remark it should also be "Now if your stalk is a local ring $[f]_x \neq 0 \Rightarrow [f]_x \notin \mathfrak m \Rightarrow [f]_x$ is a unit. Also, at 2) and 3) I gather you meant to say germ. $\endgroup$
    – Rodrigo
    Jan 21, 2014 at 14:13
  • $\begingroup$ @Zhen, Rodrigo: sorry, made the necessary edits. $\endgroup$
    – Blah
    Jan 21, 2014 at 17:42
  • $\begingroup$ When he says function on $X$, he refers to any $f \in \mathcal O_X(X)$. $\endgroup$ Jan 13, 2021 at 3:07

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.