Given a topological space $C$ and points $c_0, c1\in C$ we say that a topological space $(X,\tau)$ is $(C,c_0,c_1)$-connected if and only if for all $x,y\in X$ there is $f:C\to X$ continuous with $f(c_0) = x$ and $f(c_1)=y$.

(In this language, path-connectedness equals $([0,1],0,1)$-connectedness.)

What is an example of a space $C$ and points $c_0\neq c1\in C$ such that

  1. $(C,c_0,c_1)$-connectedness implies path-connectedness, and
  2. for every infinite cardinal $\kappa$ there is a topology on $\tau$ on $\kappa$ such that $(\kappa,\tau)$ is path-connected, but not $(C,c_0,c_1)$-connected


  • 5
    $\begingroup$ $(\{*\},*,*)$ ?. $\endgroup$ Mar 31 '16 at 12:01
  • 2
    $\begingroup$ More seriously, unless I'm mistaken, If $C$ is not functionally separated (more precisely if $c_0$ and $c_1$ cannot be separated by continuous functions) then $[0,1]$ is not going to be $C$-connected and conversely if $C$ is functionally separated then every path connected space will be $C$-connected. So you are looking for examples of path-connected non functionally separated spaces. Lots of non Hausdorff example, and I'm sure books like "counterexamples in topology" contains lots of Hausdorff examples. $\endgroup$ Mar 31 '16 at 12:12
  • $\begingroup$ ... and obviously it will also implies that you don't have many $C$-connected spaces: a functionally separated space will always be "$C$ totally disconected". $\endgroup$ Mar 31 '16 at 12:21
  • $\begingroup$ Right, the one-space example... I changed the question a bit to evade this example $\endgroup$ Mar 31 '16 at 14:36

Let $C$ be the space nº74 ("double origin topology") in Steen & Seebach's Counterexamples of Topology, chosen because it is the only one listed there that is T2 and path-connected but not T3:

$C$ consists of the set of points of the plane $\mathbb{R}^2$ together with an additional point $0^*$. Neighborhoods of points other than the origin $0$ and the point $0^*$ are the usual open sets of $\mathbb{R}^2\setminus\{0\}$; as a basis fo neighborhoods of $0$ and $0^*$, we take $V_n(0) = \{(x,y) : x^2+y^2 < 1/n^2 \land y>0\} \cup \{0\}$ and $V_n(0^*) = \{(x,y) : x^2+y^2 < 1/n^2 \land y<0\} \cup \{0^*\}$.

In other words we have replaced the origin in the plane by an "upper origin" $0$ and a "lower origin" $0^*$, the neighborhoods of the upper origin being sets containing an open half-disk centered at the origin, plus the upper origin itself, and similarly for the lower origin.

The space $C$ is Hausdorff, but not T2½ because $0$ and $0^*$ do not have disjoint closed neighborhoods; in particular, it is not T3 or T3½. So there is no continuous function $C \to \mathbb{R}$ taking different values on $0$ and $0^*$.

Let $c_0 = 0$ and $c_1 = 0^*$. Then there is a path connecting $c_0$ and $c_1$: indeed, there is a path connecting $c_0$ to, say, $(1,0)$, and one connecting $(1,0)$ to $c_1$ (we can even find "arcs", i.e., injective paths, if we want). If we have a continuous function $C \to X$ taking $c_0$ to $x$ and $c_1$ to $y$, then right-composing it with the path just mentioned gives a path connecting $x$ and $y$: so $(C,c_0,c_1)$-connectedness implies path-connectedness. On the other hand, $\mathbb{R}$ is not $(C,c_0,c_1)$-connected because of what was said in the previous paragraph.

I think this answers the question, with the additional constraint that $C$ is Hausdorff and $c_0 \neq c_1$. (I just noticed that Simon Henry had the same idea in the comments.)

However, it doesn't answer the question that I think should have been asked, namely to also require $C$ itself to be $(C,c_0,c_1)$-connected (in the above example, it's pretty clear that it's not).

  • $\begingroup$ With regard to your last remarks: In fact it would already be nice to have an example where $C$-connectedness is a non trivial notion with (i.e. there is as least one non trivial space that is $C$-connected) $\endgroup$ Mar 31 '16 at 14:41
  • $\begingroup$ @Gro-Tsen very good point, can you ask the question with your additional requirement? $\endgroup$ Mar 31 '16 at 14:59
  • $\begingroup$ @SimonHenry: There are still the indiscrete spaces, but you're right, it's even worse than I thought. $\endgroup$
    – Gro-Tsen
    Mar 31 '16 at 15:00
  • $\begingroup$ @DominicvanderZypen: I'll rather let you do that, but I suggest you give the matter further thought to decide what the best question should be, because it's not entirely clear. You probably want to require $C$ to be $(C,c_0,c_1)$-connected and $c_0$ and $c_1$ (hence any two points) to be connected by a path and sprinkle some magic "Hausdorff" dust in various places, but I'm not sure what else. $\endgroup$
    – Gro-Tsen
    Mar 31 '16 at 15:06
  • $\begingroup$ Thanks @Gro-Tsen - will think about it and post a new question in a week or so $\endgroup$ Apr 1 '16 at 6:47

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.