8
$\begingroup$

Let $X$ be a (smooth) manifold. It's well known that its Stone-Cech compactification $\beta X$ is homeomorphic to $\operatorname{Specm}(C(X))$, with its Zariski topology.

Is $\beta X$ also homeomorphic to $\operatorname{Specm}(C^\infty(X))$?

$\endgroup$
12
  • 3
    $\begingroup$ Related question $\endgroup$
    – Gro-Tsen
    Jan 13, 2022 at 13:18
  • 4
    $\begingroup$ It would sound more natural to ask whether some canonical map is a homeomorphism. $\endgroup$
    – YCor
    Jan 13, 2022 at 13:40
  • 1
    $\begingroup$ It would suffice to show that disjoint closed sets can be separated by smooth real-valued functions. $\endgroup$
    – KP Hart
    Jan 13, 2022 at 17:54
  • $\begingroup$ @KPHart This is just the smooth Urysohn lemma, isn't it? If so, why is this enough? $\endgroup$
    – Gabriel
    Jan 13, 2022 at 19:24
  • 1
    $\begingroup$ Because it implies disjoint closed sets in $X$ have disjoint closures in the compactification determined by $C^\infty(X)$, and that property characterizes $\beta X$. $\endgroup$
    – KP Hart
    Jan 13, 2022 at 19:26

1 Answer 1

8
$\begingroup$

Indeed, $\operatorname{Specm}(C^\infty(X))$ is homeomorphic to $\beta X$ (through an explicit homeomorphism that will be described below). Essentially all the theory described in Gillman & Jerison's classic book Rings of Continuous Functions (1960) applies: for completeness of MathOverflow, let me recall how this works.

If $f$ is a continuous real-valued function on $X$, its zero-set is the set $Z(f) := \{x\in X : f(x)=0\}$. The zero-set of a continuous function is closed, and conversely, as $X$ is a metric space, every closed set is the zero-set of a continuous function. But in fact, even more is true: every closed set of $X$ is the zero-set of a smooth function (as follows, e.g., from here). Because of this, the notions of “z-filter” and “z-ultrafilter” below do not depend on whether we are talking about zero-sets of continuous functions or of smooth functions (or, indeed, closed sets).

Now let us make the following definitions:

  • A set $\mathscr{F}$ of zero sets is said to be a z-filter iff it contains $X$, and is closed under enlargement (i.e., if $Z \subseteq Z'$ are zero-sets and $Z\in \mathscr{F}$ then $Z'\in \mathscr{F}$) and finite intersection.

  • A z-filter is said to be a z-ultrafilter iff it is proper (i.e., does not contain $\varnothing$), and maximal for inclusion among proper z-filters.

Now let $R$ be either the ring $C(X)$ of continuous real-valued functions on $X$ or the ring $C^\infty(X)$ of smooth real-valued functions on $X$. Define two maps $\mathscr{Z}\colon \{\text{ideals of $R$}\} \to \{\text{z-filters}\}$ and $\mathscr{I}\colon \{\text{z-filters}\} \to \{\text{ideals of $R$}\}$ as follows:

  • If $I$ is an ideal of $R$, then $\mathscr{Z}(I) := \{Z(f) : f\in I\}$. This is indeed a z-filter, as is easy to check (for enlargement, note that if $Z(f) \subseteq Z(f')$ with $f\in I$ then $Z(f') = Z(ff')$; and for finite intersection, note that $Z(f_1) \cap Z(f_2) = Z(f_1^2 + f_2^2)$).

  • If $\mathscr{F}$ is a z-filter, then $\mathscr{I}(\mathscr{F}) := \{f\in R : Z(f)\in \mathscr{F}\}$. This is indeed an ideal as is easy to check (note that $Z(g_1 f_1+\cdots+g_r f_r) \supseteq Z(f_1)\cap \cdots \cap Z(f_r)$ if $f_1,\ldots,f_r$ are in $\mathscr{I}(\mathscr{F})$).

These functions are not bijections, but still, they preserve inclusions and we trivially have $\mathscr{Z}(I) \subseteq \mathscr{F}$ iff $I \subseteq \mathscr{I}(\mathscr{F})$ (when $I$ is an ideal of $R$ and $\mathscr{F}$ a z-filter); from this follows $\mathscr{Z}(\mathscr{I}(\mathscr{F})) \subseteq \mathscr{F}$ for any z-filter $\mathscr{F}$, and $\mathscr{I}(\mathscr{Z}(I)) \supseteq I$ for any ideal $I$ of $R$, but in fact it is clear that the former is easily seen to be an equality: $\mathscr{Z}(\mathscr{I}(\mathscr{F})) = \mathscr{F}$ (the latter inclusion, $\mathscr{I}(\mathscr{Z}(I)) \supseteq I$ can be proper in general as evidenced by the ideal of functions vanishing to order $2$ at a point). Also note that an ideal $I$ is the unit ideal $R$ iff $\mathscr{Z}(I)$ is the improper z-filter $\mathscr{Z}(R)$ (the one consisting of every zero-set) because $Z(f)$ is empty iff $f$ is invertible in $R$.

At this point, it is easy to check that: if $M$ is a maximal ideal of $R$ then $\mathscr{Z}(M)$ is a z-ultrafilter, and if $\mathscr{U}$ is a z-ultrafilter then $\mathscr{I}(\mathscr{U})$ is a maximal ideal. So $\mathscr{Z}$ and $\mathscr{I}$ restrict to bijections between maximal ideals of $R$ and z-ultrafilters.

But because the above worked just as well for $R$ being the ring $C(X)$ of continuous real-valued or the ring $C^\infty(X)$ of smooth real-valued functions, we can conclude that their maximal ideals are in bijection by the compositions of the corresponding $\mathscr{Z}$ and $\mathscr{I}$ functions, and in fact since $\mathscr{I}_{C^\infty(X)}(\mathscr{F}) = \mathscr{I}_{C(X)}(\mathscr{F}) \cap C^\infty(X)$ the bijection from maximal ideals of $C^\infty(X)$ to those of $C(X)$ can more simply be described as $M \mapsto M \cap C^\infty(X)$. Furthermore, the Zariski topologies correspond under this bijection, because (since every z-filter is of the form $\mathscr{Z}(I)$) they both correspond to the Zariski topology on the set of z-ultrafilters whose closed sets are given by the sets of ultrafilters containing a given z-filter.

(It might be worth while examining exactly which properties are required for a ring $R$ of “functions” on $X$ to make the above reasoning ensure that $\operatorname{Specm}(R) = \beta X$.)

PS: It might also be interesting to consider what happens for bounded functions. The maximal ideals of the ring $C^*(X)$ of bounded continuous functions on $X$ are also naturally in bijection with the Stone-Čech compactification, but the bijection between ideals of $C(X)$ and $C^*(X)$ is not the naïve $M \mapsto M \cap C^*(X)$. So I'm not sure what happens to bounded smooth functions (I didn't give it thought).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.