I'm having trouble understanding the following argument (which I believe to be somewhat incomplete or flawed). Let $A=C(X)$ be the set of continuous functions from the topological space $X$ to the complex plane $\mathbb{C}$. We define $m_{x} = \{f \in C(X): f(x) = 0 \}$ and $A_x$ the ring of germs at point $x$. The statement is the following $A_x \simeq A_{m_x}$.

(1) I don't see how one defines $A_{m_x}$ since the set contains global functions that might not be well-defined as we quotient by functions $f$ such that $f(x) \neq 0$, but it doesn't necessarily mean that $f \neq 0$. Though it's indeed well defined in a neighborhood of $x$.

(2) Now using the universal property of localization, we sure want to define $\phi : A_{m_x} \rightarrow A_x$ s.t we have $\phi(a/s) = \iota(a)\iota(s)^{-1}$ where $\iota$ is the inclusion map $\iota: A \rightarrow A_x$. We want $\phi$ to be an isomorphism. It is surjective; now we want it to be one-on-one. Now I don't see how this is possible as $\phi(a/s) = 0$ iff $a = 0$ in a neighborhood of $x$, which doesn't imply that $a=0$ globally.

I guess there's something I don't really fathom, or my textbook might just be flawed. Anyway, thanks for your help.


You need to assume something about the space $X$ for the claimed statement to be true. A natural assumption to make is that $X$ is completely regular; I will use this assumption in addressing (2) below.

(1) Well, $A_{m_x}$ is not (a priori) a set of functions, it's just a ring of formal "fractions" (which may or may not make sense when evaluated as pointwise fractions of functions). You can still think of it perfectly well as a ring without having to think of its elements as functions on $X$ (which, as you observe, you can't exactly do).

(2) If $a\in A$ is such that $a=0$ in a neighborhood $U$ of $x$, then in fact the image of $a$ in the localization $A_{m_x}$ vanishes: the canonical "inclusion" $A\to A_{m_x}$ is not injective! To prove this, note that by complete regularity, there is a function $f:X\to[0,1]$ such that $f(x)=1$ and $f(y)=0$ for all $y\not\in U$. We then have $f\not\in m_x$ and $fa=0$, so it follows that $a$ maps to $0$ in the localization $A_{m_x}$.

Note that you also need to use complete regularity to show that your map $\phi:A_{m_x}\to A_x$ is surjective: given a germ of a continuous function at $x$, it is not at all obvious a priori that you can write it as a quotient of two continuous functions that are defined on all of $X$. In detail, if you have a function $f:U\to\mathbb{C}$ where $U$ is an open neighborhood of $x$, let $V$ be an open neighborhood of $x$ whose closure is contained in $U$ (by regularity) and let $g:X\to[0,1]$ be a function such that $g(x)=1$ and $g(y)=0$ for all $y\not\in V$ (by complete regularity). Define $h(y)=\min(1,2g(y))$. Then $h=1$ on a neighborhood of $x$ (namely, the set where $g>1/2$), so $hf$ (which is defined on $U$) has the same germ at $x$ as $f$. But $hf$ vanishes on $U\setminus\overline{V}$, so we can continuously extend it to all of $X$ by setting it equal to $0$ outside of $U$. This continuous extension is then an element of $A$ whose germ at $x$ coincides with the germ of $f$. Thus the map $A\to A_x$ is surjective, and hence so is the map $A_{m_x}\to A_x$.

  • $\begingroup$ Ah, complete regularity being necessary to construct an inverse for a germ, in my scheme... This is an important point. $\endgroup$ – rschwieb Dec 29 '15 at 11:52
  • $\begingroup$ Er, yes, my comment was directed at the earlier version of your answer. As for the current version, it is trivial that the germ of a function not vanishing at $x$ is invertible (just restrict it to the set where it is nonzero and take its inverse there). No complete regularity is needed for that. $\endgroup$ – Eric Wofsey Dec 29 '15 at 12:07
  • $\begingroup$ I thought that was wher you comment applied, but yes, that is what I initially thought too. Time to wit for coffee to reevaluate everything... $\endgroup$ – rschwieb Dec 29 '15 at 12:20
  • $\begingroup$ Yes indeed, when thinking about it, I had to assume for instance than X was a metric space, so I could put something like $f(x) = d(x, V^{c})$ where V is the open subset where our given function doesn't get to zero...So is it true, nonetheless, that $A_{x} \simeq A_{m_x}$? (It doesn't seem like it anymore) (and yes, $A=C(X)$ ;)) $\endgroup$ – Hermès Dec 29 '15 at 12:34
  • 1
    $\begingroup$ No, it's not true for arbitrary $X$ that $A_x\simeq A_{m_x}$. For a counterexample, you can let $X=\mathbb{R}\cup\{x\}$, where a set $U\subseteq X$ is open iff it is an open subset of $\mathbb{R}$ not containing $0$ or it is of the form $V\cup\{x\}$ for an open set $V\subseteq\mathbb{R}$. Then $A_x\simeq\mathbb{C}$ (since $\{x\}$ is open) but you can show $A_{m_x}$ is not a field. $\endgroup$ – Eric Wofsey Dec 29 '15 at 18:10
  1. The collection of functions that are nonzero at $x$ are certainly a multiplicatively closed subset $S$ of the ring, so there is no problem localizing at this set.

  2. We can always map a function $f$ to its germ $[f]$ at $x$. If $f$ is nonzero at $x$, it is nonzero on a neighborhood of $x$, and the function $\frac 1f$ is defined and nonzero on the same neighborhood. Then the germ of the inverse of $f$ on the interval is the inverse of the germ of $f$. So, things in $M_x$ get mapped to units of the ring of germs. By the universal property of localization, there is now a map of the localization ontointo the ring of germs at $x$. (Addendum: I never did catch that the map is not obviously onto and I won't pretend here by duplicating the work Eric Wofsey does in his solution to show that complete regularity of $X$ is sufficient to prove the map is onto.)

    Your last comment is on the right track. But your concern that the preimage is not globally zero is irrelevant. Starting with $f/s$ in your ring of fractions, if $[\frac fs]=[0]$ is the zero germ, then by definition of the extension $[f]=[0][s]=[0]$. As you noted, $f$ is zero on a neighborhood $U$ of $x$. To close the case we need to show that there exists $t\in S$ such that $tf=0$, demonstrating that $f\sim 0$ in the localization. For this, you need complete regularity (as pointed out by Eric Wofsey). By finding a continuous function which is zero on $U^c$ and $1$ at $x$, you will have such an element $t$.


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.