2

I'd like to understand what connected subsets and connected components of a GO-space look like.

Some definitions:

A linearly ordered topological space (LOTS) is a totally ordered set $(X,<)$ together with the corresponding order topology generated by all the open rays $(\leftarrow,x)$ and $(x,\rightarrow)$ for $x\in X$. For a point $x$ that is not the minimum or the maximum of $X$, the open intervals $(a,b)$ with $a<x<b$ form a local base of open nbhds at $x$. Let's write $\lambda(<)$ for this order topology.

A GO-space (generalized ordered space) is a totally ordered set with a more general topology, not determined by the order. Specifically and formally, it's a triple $(X,<,\tau)$ where $<$ is a total order on $X$ and $\tau$ is a topology on $X$ containing $\lambda(<)$ with a base of open sets that are order-convex. (A set $A\subseteq X$ is order-convex if for any two points $x<y$ in $A$ the whole interval $[x,y]$ is contained in $A$.) So, a subbase for $\tau$ is given by the collection of sets consisting of:

  • (1) all sets of the form $(\leftarrow,x)$ and $(x,\to)$ for $x\in X$;

  • (2) all sets of the form $[x,\to)$ such that $[x,\to)$ is $\tau$-open, $x$ is not the minimum element of $X$ and $x$ has no immediate predecessor in $(X,<)$;

  • (3) all sets of the form $(\leftarrow,x]$ such that $(\leftarrow,x]$ is $\tau$-open, $x$ is not the maximum element of $X$ and $x$ has no immediate successor in $(X,<)$.

See Equivalent definitions for GO-spaces (generalized ordered spaces) for some equivalent characterizations.

As an example, the Sorgenfrey line is a GO-space with $(\mathbb R,<)$ as underlying ordered set. The topology contains the usual open rays, as well as all closed rays $[x,\rightarrow)$ (which are not open in the Euclidean topology). Another example is the Michael line.


First the situation with LOTS (then I'll get to the question for GO-spaces). Suppose $X$ is a LOTS with order $<$ and $A$ is a nonempty subset of $X$.

Let $<_A$ denote the restriction of the order $<$ to the set $A$. The order $<_A$ is a total order on $A$ with associated order topology $\lambda(<_A)$. In general, the topology $\lambda(<_A)$ on $A$ can be strictly coarser than the subspace topology induced by the topology $\lambda(<)$ on $X$. (Example: $A=\{0\}\cup(1,2)\subseteq\mathbb R$) But if $A$ is order-convex in $X$, the two topologies coincide (easy to check).

On the other hand, being order-convex is a necessary condition for $A$ to be connected. Indeed, if $a<c<b$ with $a,b\in A$ and $c\notin A$, then the disjoint open sets $(\leftarrow,c)$ and $(c,\rightarrow)$ cover $A$ and give a non-trivial separation of $A$.

Therefore, the subset $A\subseteq X$ is connected iff $A$ is order-convex in $X$ and the ordered set $(A,<_A)$ itself is connected in its order topology. (As is well known, that condition can be expressed entirely by properties of the order $<_A$, namely the order is Dedekind-complete and order-dense.)

Aso, every locally connected LOTS is locally compact. See for example here.


What of these results still hold for GO-spaces?

Even though the order does not determine the topology for a GO-space, is there a way to detect that a subset is connected just by looking at the order?

PatrickR
  • 7,165
  • Probably you are aware of the following, aren't you? Connected GO-spaces are LOTS, hence they are locally compact, locally connected, ... – Ulli Sep 08 '24 at 06:58
  • It may not be as simple as that. How does one know that the orders and topologies of the full space and the subspace are all compatible? (They may not be for arbitrary subsets of a GO-space.) I.e., it would be helpful to see this in detail. – PatrickR Sep 08 '24 at 17:35
  • I find it much easier to look at GO-spaces just as subspaces of LOTS. And it's also easy to show then that a connected GO is LOTS (in fact, its order topology coincides with the subspace topology). Would you like to get a proof for that? – Ulli Sep 08 '24 at 18:54
  • I already know that if the full space is a connected GO-space, it must be a LOTS (with the same order). That's in the Lutzer paper mentioned in https://topology.pi-base.org/theorems/T000131/references. What I am interested in here is the case of a subset of a GO-space, not the full space itself. – PatrickR Sep 08 '24 at 19:00
  • This was also my first assumption. Since you were asking, what results hold for GO-spaces, it was just the intent of my comment, to answer that "if a subspace A of a LOTS X is connected, then it is locally compact" also holds for GO-spaces. – Ulli Sep 08 '24 at 19:07
  • I meant something slightly more general: a locally connected LOTS is locally compact, Changing this in the question. – PatrickR Sep 09 '24 at 00:43
  • 2
    But this still holds for GO-spaces as well: A locally connected GO is locally compact. – Ulli Sep 09 '24 at 05:16

1 Answers1

2

It turns out things are a lot simpler than I was making it to be. We can use this other characterization of GO-spaces:

$(X,<,\tau)$ is a GO-space iff (a) $\lambda(<)\subseteq\tau$ and (b) there is a base for the topology $\tau$ consisting of order-convex sets.

where $\lambda(<)$ is the order topology induced by the order $<$.

Also a basic fact for LOTS: Suppose $(X,<)$ is a totally ordered set with order topology $\lambda(<)$ and let $A$ be an arbitrary subset of $X$. Let $<_A$ be the order $<$ restricted to $A$. There are two topologies on $A$: the order topology $\lambda(<_A)$ induced by $<_A$ and the subspace topology $\lambda(<)_A$ induced by the topology $\lambda(<)$ on $X$. In general, the two topologies are not equal. But it is easy to check that $\lambda(<_A)\subseteq\lambda(<)_A$. And if $A$ is order-convex, the two topologies are equal: $\lambda(<_A)=\lambda(<)_A$.


Now suppose $(X,<,\tau)$ is a GO-space and let $A\subseteq X$ be a connected set for the topology $\tau$. Let $\tau_A$ be the subspace topology on $A$ induced by $\tau$. I'll show that $(A,<_A,\tau_A)$ is also a GO-space.

The set $A$ has to be order-convex in $(X,<)$. (Just as in the LOTS case: if $a<c<b$ with $a,b\in A$ and $c\notin A$, then $A=(A\cap(\leftarrow,c))\cup(A\cap(c,\to)$ is a partition of $A$ into two disjoint nonempty $\tau_A$-open sets.)

We have $\lambda(<)\subseteq\tau$ as topologies on $X$. Therefore restricting the topologies to $A$ gives $\lambda(<)_A\subseteq\tau_A$. And since $A$ is order-convex, $\lambda(<_A)=\lambda(<)_A$. Thus $\lambda(<_A)\subseteq\tau_A$.

And on the other hand, there is a base for the topology $\tau$ on $X$ consisting of order-convex sets in $(X,<)$. Intersecting each of these basic sets with $A$ gives a base for the topology $\tau_A$. And since $A$ is order-convex in $X$, the intersections will also be order-convex in $(X,<)$, and in $(A,<_A)$.

By the characterization of GO-spaces at the top, $(A,<_A,\tau_A)$ is a GO-space. We can now use the known result that a connected GO-space is actually a LOTS with the same order. That is, $\tau_A=\lambda(<_A)$.


Summary: The connected sets in a GO-space $(X,<,\tau)$ can be determined just by looking at the order $<$. They are exactly the order-convex subsets $A$ of $X$ that satisfy the usual condition for being connected as a LOTS when restricting attention to the order $<_A$ on $A$. Namely, $(A,<_A)$ is Dedekind-complete (= least upper bound condition) and order-dense.

And from that, some local results for LOTS imply local results for GO-spaces. In particular, any interval $[x,y]$ in a connected LOTS is compact (since it is Dedekind complete and has a minimum element and a maximum element). Hence if a GO-space is locally connected, each point $x$ has a connected nbhd $V$. That nbhd $V$ with the restriction of the order from $X$ is a LOTS and the point $x$ has a compact nbhd within $V$. That shows:

A locally connected GO-space is locally compact.

PatrickR
  • 7,165