The goal of this article is to prove that a compact Hausdorff ring is a profinite ring, i.e., is canonically isomorphic as a topological ring to its profinite completion , topologized as an inverse limit of finite rings with their discrete topologies.
Further commentary on this result can be found in a CafΓ© blog post of Tom Leinster, in connection with some work by his student Barry Devlin.
All topological rings considered are (see separation axiom). As is well-known from the study of uniform structures, this implies that all our rings are Hausdorff and even Tychonoff spaces. In particular, finite rings are Hausdorff and therefore discrete. If is a finite ring and is a continuous ring homomorphism, this means the kernel of must be an open ideal of (for us, βidealβ always means two-sided ideal), since is open in .
In the converse direction, if is a compact ring and is an open ideal, then the ring with the quotient topology coming from is discrete (every point is open), and compact since is compact, and so must then be finite.
Let denote the poset of open ideals of , ordered by inclusion. This poset is codirected by taking finite intersections of ideals. Each inclusion induces a surjection of finite rings , and so we get a functor whose (projective) limit is the profinite completion of ; the quotient maps of course induce a canonical map .
The map is surjective.
Since is compact and is Hausdorff, the image is closed in , so it suffices to show the image is a dense subspace.
So, we are to show that if is open and inhabited, then there exists with . Let denote a typical component map of the limit cone. A basis for the topology of consists of inhabited finite intersections of the form where each is a point of , so WLOG assume is of this form.
Let be any point belonging to (so for ), and put . Then for each the coordinate maps to under the map , precisely because of the compatibility conditions on the coordinates imposed by the limit. Then, letting be an element that maps to under , the element lies in and we are done.
A key technical result is the following (see this -Category CafΓ© comment):
Every compact Hausdorff ring is totally disconnected.
Let be the Pontryagin dual of (the additive group of) ; as a space is discrete. As and are locally compact Hausdorff, they are exponentiable as spaces, and it follows quickly that the functorial maps
are continuous homomorphisms of topological groups. Using naturality of the isomorphism , these maps are mutually inverse, and so they give an isomorphism of topological groups.
The multiplication map transforms to a continuous injection . As is compact and is Hausdorff, the injection maps homeomorphically onto its image in , i.e., is a subspace embedding. But is manifestly a subspace of a product of discrete spaces and hence totally disconnected, so is totally disconnected as well. (In more detail: if is connected, then the image of under the composite
is also connected and hence a one-point space for each , so that and therefore are one-point spaces.)
By Theorem , Pontryagin duality implies compact rings are totally disconnected, which readies us for a next wave of results.
Let be a compact Hausdorff totally disconnected space. For each and open neighborhood of , there exists a compact open set such that .
In a compact Hausdorff space , the connected component of a point equals its quasi-component; an online proof may be found here. So if is also totally disconnected, then is also the intersection of all clopens containing it. Now let be an open neighborhood of , so and therefore . As is compact, finitely many such clopens cover , and so . This is compact and open and contains , which completes the proof.
Let be a compact Hausdorff totally disconnected abelian group. For each open neighborhood of , there exists an open subgroup such that .
By Lemma , there is a compact open with . Now let be the restriction of the addition operation . Since is continuous, there is for each an open neighborhood of where we may assume is symmetric (i.e., ). Finitely many of the cover , say . Then putting , we have that , or for short. This is again symmetric: .
Hence if denotes the set of all sums of elements belonging to , then is open and by induction, whence . Since is symmetric, the union is the subgroup generated by ; it is open and contained in and therefore also in , which completes the proof.
Let be a compact Hausdorff totally disconnected ring. For each open neighborhood of , there is an open ideal such that .
By Lemma , there is an open additive subgroup . Let . It is evident that and is an ideal of . Now hold fixed. For each , continuity of multiplication implies there are open neighborhoods of , of , and of such that . By compactness of , finitely many sets of the form cover . The corresponding finite intersection is an open neighborhood of that is contained in , thus completing the proof.
Now for our final result.
For a compact ring, the canonical map is injective.
This and Proposition taken together imply that is a ring isomorphism. It is also a homeomorphism because it is a continuous bijective map from a compact space to a Hausdorff space. Therefore is an isomorphism of topological rings.
As we have an inclusion of the profinite completion as projective limit, we see the kernel of is the same as the kernel of
which is precisely . Since by Lemma every neighborhood of contains an open ideal , this intersection is contained in the intersection of all open neighborhoods of , which is since is Hausdorff. Thus the kernel of is trivial, as was to be shown.
Last revised on September 11, 2016 at 23:44:09. See the history of this page for a list of all contributions to it.