Consider the functor $C$ from the category of compact Hausdorff spaces to that of unital abelian C*-algebras given by $C: X \mapsto C(X, \mathbb{C})$ (the continuous maps from $X$ to $\mathbb{C}$) and also, for continuous $f: X \to Y$ ($Y$ being another compact Hausdorff space), $Cf: C(Y) \to C(X), h \mapsto h \circ f$. If $Cf$ is injective, it is then the case that $f$ is surjective -- my question is how/why? Several resources that I have found brush this off by merely invoking Tietze - and then moving on. However, I don't see how this works, exactly. Clearly I'm overlooking something elementary. Here is as far as I get with my reasoning.

First off, $X$ is a compact Hausdorff space and is hence normal. Moreover, $f(X)$ is a closed (and compact) subset of $Y$. If $Cf$ is injective, then we have that for $h_1,h_2 \in C(Y)$, if $h_1|_{f(X)} = h_2|_{f(X)}$ then it must be that $h_1 = h_2$.

I suppose a slight variant of Tietze's theorem is being considered, where in the real and imaginary parts of the function are being extended, however, I see no value in extending $h_1|_{f(X)}$ and $h_2|_{f(X)}$ to all of $Y$, for they would have to be identically $h_1$ and $h_2$ respectively, which are equal!

What should be done?


I see no value in extending $h_1|_{f(X)}$ and $h_2|_{f(X)}$ to all of $Y$, for they would have to be identically $h_1$ and $h_2$ respectively, which are equal!

The point is that if $f(X)\neq Y$, then it is possible to extend $h_1|_{f(X)}$ to a function that is not equal to $h_1$. That is, Tietze's theorem leads to a proof by contraposition. In particular, there would be a nonzero element of $C(Y)$ that vanishes on $f(X)$.

  • $\begingroup$ Tietze guarantees the existence of $\textit{an}$ extension. Please justify the statement that $h_1|_{f(X)}$ can be extended something not equal to $h_1$. $\endgroup$ – georgC Dec 9 '11 at 5:03
  • $\begingroup$ @georgC: There exists $y\not\in f(X)$, and you can map $y$ wherever you want, then extend from there. E.g., for the vanishing on $f(X)$ case, there exists $h\in C(Y)$ such that $h(y)=1$ and $h|_{f(X)}=0$. (Edit: I see that Sam's answer made this explicit as well.) $\endgroup$ – Jonas Meyer Dec 9 '11 at 5:08

If $f$ is not surjective, then - choosing a point $x_0\notin f(X)$ - there is a continuous function $h$ for which $h(x_0) = 1$ and $h|_{f(X)} \equiv 0$.

This follows from Tietze extension and the fact that $\{x_0\}$ and $f(X)$ are closed subsets of $Y$. (The latter because $f(X)$ is compact and $Y$ is Hausdorff).

For this $h$ we have $Cf(h) = Cf(0)$, although $h\ne 0$. So $Cf$ is not injective.

  • $\begingroup$ Ah. Finally, thank you very much. However, what you have invoked here is the Urysohn lemma, not the Tietze extension theorem. I am aware that these two propositions are equivalent, but, again, what you gave was not a straight-forward application of Tietze's theorem. $\endgroup$ – georgC Dec 9 '11 at 5:13
  • 1
    $\begingroup$ @georgC: Please note that $\{x_0\}\cup f(X)$ is closed, and the function that is $1$ at $x_0$ and vanishes on $f(X)$ is continuous on $\{x_0\}\cup f(X)$, then extend it. $\endgroup$ – Jonas Meyer Dec 9 '11 at 5:16
  • $\begingroup$ @Jonas: Yes, you are correct. When you put it like this, it is most definitely an application of Tietze. Thank you. $\endgroup$ – georgC Dec 9 '11 at 5:20

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.