The Gelfand duality says that $$X\to C(X)$$ is a contravariant equivalence between the category of compact Hausdorff spaces and continuous maps and the category of commutative unital $C^*$-algebras and continuous $*$-homomorphisms. A well known generalization of $C^*$-algebras are pro-$C^*$-algebras. Pro-$C^*$-algebras are topological $*$-algebras that are cofiltered limits of $C^*$-algebras (in the category of topological $*$-algebras). See for instance this paper by Phillips. If $X$ is a weakly Hausdorff compactly generated space, it is not hard to see that $C(X)$, the continuous functions from $X$ to $\mathbb{C}$, is a pro-$C^*$-algebra. Indeed, it is the cofiltered limits of $C(K)$, as $K$ ranges over all compact Hausdorff subspaces of $X$. My question is:

Is the functor $$X\to C(X)$$ a contravariant equivalence between the category of weakly Hausdorff compactly generated spaces and continuous maps and the category of commutative unital pro-$C^*$-algebras and continuous $*$-homomorphisms?

Few remarks:

  1. In the paper by Phillips mentioned above, Theorem 2.7, it is shown that the functor above is a contravariant equivalence between the category of completely Hausdorff quasitopological spaces and the category of commutative unital pro-$C^*$-algebras. It is also shown, in Example 2.11, that when restricted to the full subcategory of completely Hausdorff compactly generated spaces, this functor is not essentially surjective. But this still doesn't answer my question, because maybe on weakly Hausdorff compactly generated spaces this functor is essentially surjective.

  2. Apparently, the reason Phillips insists on working with completely Hausdorff (quasi)topological spaces is that the way he proves this result is by constructing an inverse equivalence which assigns to any commutative unital pro-$C^*$-algebra its spectrum. The way he defines the spectrum, it can be shown that if $A$ is a commutative unital pro-C*-algebra, the spectrum of $A$ is completely Hausdorff. But, again, this doesn't answer my question, because maybe one can define an inverse equivalence in a different way, or maybe one can just show that this functor is fully-faithful and essentially surjective.

  3. The main reason I am asking is that the category of weakly Hausdorff compactly generated spaces is very important in algebraic topology and homotopy theory, and such a result would imply that pro-$C^*$-algebras are exactly the non-commutative version of it.

Remark: I used the terminology of the paper of Phillips that I referred to. Note that the category of pro-$C^*$-algebras is not the pro-category of the category of $C^*$-algebras. In particular, its objects are topological $*$-algebras and not cofiltered diagrams of $C^*$-algebras. Some authors call pro-$C^*$-algebras locally $C^*$-algebras. It can be shown that the pro-category of the category of $C^*$-algebras contains the category of pro-$C^*$-algebras, as a full coreflective subcategory.

Edit: Due to Simon's answer the functor $X\to C(X)$ defined above is clearly not faithful. But the following question remains: Is this functor full and essentially surjective. If so then it would induce a contravariant equivalence of categories between $\overline{CGWH}$ and the category of commutative unital pro-$C^*$-algebras, where $\overline{CGWH}$ is the category of weakly Hausdorff compactly generated spaces and equivalence classes of continuous maps, where two continuous map $X\rightrightarrows Y$ are called equivalent if the induced maps $C(Y)\rightrightarrows C(X)$ are the same.

  • $\begingroup$ So the question is whether the category of CGWH spaces is the category of ind-CH spaces? I feel like if this were true someone would've told me by now. $\endgroup$ Aug 12, 2015 at 8:00
  • 1
    $\begingroup$ I'm not sure it answers your question, but I seem to remember that Chris Phillips had overlooked the work Dubuc and Porta, and that they got closer to the bottom of some of these things. Almost 20 years ago I wrote a preprint with (at that time fellow student) Kasper Andersen about fibration category structures on C^*-algebras which we never published. We reference the work of Dubuc and Porta. It can be found here: math.ku.dk/~jg/papers/fibcat.pdf $\endgroup$ Aug 12, 2015 at 9:24
  • 1
    $\begingroup$ @TimPorter I think a paper with similar results is already published: arxiv.org/abs/1011.2926 $\endgroup$ Aug 12, 2015 at 14:10
  • 2
    $\begingroup$ If one stands the question on its head and asks for a simple and natural representation of the dual category to that of the algebras described above, the answer has been known for nearly 50 years---it is the category of regular compactologies. See the article "Topologies et compactologies" by H. Buchwalter (Publ. Dept. Math. Lyon 6-2 (1969), 1-73. The definitions of these notions can be found there and the relevant result is Th. 1.4.4 on p. 14. The result is not stated in the language of category but in the spirit of functional analysis. $\endgroup$
    – priel
    Aug 19, 2015 at 16:14
  • 1
    $\begingroup$ @IlanBarnea. Afraid not. It is hard to come by and hasn't entered the secondary literature which is a pity since this is vexed question which is raised frequently here and elsewhere. If I may be allowed some self-advertisement you can find an alternative version to the generalised GN theory and Riesz representation theorem (also a frequent flyer on MO) using Saks spaces in my book "Saks spaces and Applications to Functional Analysis". Btw I tried to post a more detailed response as an answer but couldn't get it past the robot test so cut it down to a comment. $\endgroup$
    – priel
    Aug 25, 2015 at 13:24

2 Answers 2


The answer is No. Rougly, because it is not a good idea to look at continuous $\mathbb{C}$ valued function on a space which is not completely Haussdorff as completely haussdorf is exactly the hypothesis that says "your space can be understood by looing at function over it"... I really don't think you can obtain something different from Philips result by considering the exact same functor.

More precisely: Take any example of a space $X$ which weakly Hausdorff CG, but non completely Hausdorff.

then there is going to be a pair of points $x,y \in X$ such that for any continuous function $f:X \rightarrow \mathbb{C}$. $f(x) = f(y)$

Hence the two functions :$x,y: \{ *\} \rightrightarrows X$ are going to have the same image by your functor which is hence not faithfull, and hence not an equivalence of category.

Edit : This answer the edited version of the question.

If you start from a weakly Haudorf CG space $X$, then you can consider the equivalence relation on $X$ defined by $x \sim y$ if for all continuous $\mathbb{C}$ valued functions $f$ on $X$, $f(x)=f(y)$. Let $Y$ be the quotient of $X$ by this relation. By construction $Y$ is CG (it is the quotient of a CG space), any functions from $X$ to $\mathbb{C}$ is compatible with the equivalence relation hence defines a function from $Y$ to $\mathbb{C}$, hence $Y$ is completely Hausdroff and $C(Y) = C(X)$. So the image of your functor is the same as the image of the functor of Philips when restricted to CG spaces, which as you mentioned in your question is apparently not essentially surjective.

One can also probably prove it is not full by constructing an example where there won't be so many interesting maps from the quotient $Y$ to the initial space $X$ while if the functor was full then the isomorphisms map from $C(X)$ to $C(Y)$ should be represented by a map from $Y$ to $X$...

  • $\begingroup$ Thanks! Such a space indeed exists by Example 2.14 in Phillips paper. However, look at the edit I made to my question. Since I don't think it is worthwhile to start another MO question, I did not accept your answer. $\endgroup$ Aug 14, 2015 at 11:34
  • $\begingroup$ @IlanBarnea: And I've answered the edited version. $\endgroup$ Aug 14, 2015 at 11:57

The answer is yes, provided you change the category on the topological side slightly to: compactly generated functionally Hausdorff topological spaces with a distinguished family of compact sets; with continuous maps that preserve the distinguished family. See Theorem 6 of

Michael Forger, Daniel V. Paulino, Locally $C^*$ Algebras, $C^*$ Bundles and Noncommutative Spaces, arXiv:1307.4458v1

  • 1
    $\begingroup$ I turned the direct pdf link into something human-readable pointing to the abstract page. $\endgroup$
    – David Roberts
    Nov 7, 2016 at 6:48
  • $\begingroup$ The question of the definitive extension of Gelfand duality was solved by H. Buchwalter in 1969 (in the article "Topologies et Compactologies" which is easy to find online). If you want to locate it in book form, it is available in an equivalent version using Saks spaces in the monograph "Saks Spaces and Applications to Functional Analysis" (1978), also available online. $\endgroup$ Apr 2, 2021 at 13:37

Your Answer

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