I am not an expert of this area but I need help to answer this question: For a compact metric space $X$, let $C(X,X)$ be the space of all continuous functions $X\to X$, equipped with the uniform metric $$d(f,g)=\sup_x d(f(x),g(x)).$$ Can we say the space $C(X,X)$ is compact?
Asked
Active
Viewed 157 times
2
-
3For $X=[0,1]$ it will be non-compact. – erz Mar 10 '20 at 06:10
-
@erz Thank you so much. – Sh.M1972 Mar 10 '20 at 06:16
-
See the Arzelà–Ascoli theorem. – abx Mar 10 '20 at 06:40
-
2Note that a compact metric space is always complete and separable. – YCor Mar 10 '20 at 07:27
-
2YCor yes thank you. – Sh.M1972 Mar 10 '20 at 08:19
-
2@M.Shahryari it's quite trivial to find $X$ with $C(X,X)$ not compact, so maybe you should reformulate the question taking this into account (e.g., asking about a characterization of $X$ such that $C(X,X)$ is compact). Note that (for $X$ compact) this is equivalent to asking when $C(X,X)$ is closed in $X^X$, and hence a related question is this one. – YCor Mar 10 '20 at 08:31
-
@YCor, Indeed I only need the case when $X=A^G$ the shift space over finite alphabet $A$ and a f.g group $G$. The set of all cellular automata is a closed subset in this case but I want to see if it is compact or not. – Sh.M1972 Mar 10 '20 at 09:52
1 Answers
2
Indeed this space is (almost, see the comment by YCor) never compact except for trivial cases (i.e. the compact set $X$ having only finitely many points). For more information on the topology (which is colloquially known as the compact-open topology) of the space you are asking about, see e.g. the book Engelking: General topology. (Chapters 2.6 and 3.4 will have most of the information you could possibly need).
Thanks for the clarification by @YCor, the never compact of the original answer was a bit hasty and comes from my use of these spaces where $X$ is always locally euclidean (i.e. the $X$ should be a manifold). Which is of course not necessary.
Alexander Schmeding
- 2,063
-
3This seems to be in contradiction with this answer, which says that there exists an infinite connected compact metrizable space, in which every continuous self-map is identity or constant. – YCor Mar 10 '20 at 07:37
-
2Still for $X$ compact metrizable $C(X,X)$ is non-compact in many cases, notably (a) when $X$ has infinitely many connected components (b) when $X$ contains a non-trivial arc. Also, for $X$ compact metrizable $C(X,[0,1])$ is compact iff $X$ is finite. – YCor Mar 10 '20 at 08:28
-
@YCor so $C(A^G, A^G)$ is not compact, because $A^G$ is totally disconnected ($G$ is f.g). – Sh.M1972 Mar 10 '20 at 09:54
-
@YCor: Thanks for the correction, I am always thinking nice metrisable spaces, i.e. manifolds... – Alexander Schmeding Mar 10 '20 at 12:08