3

The Heine-Borel-Property says:

A subset M is compact iff it is closed and bounded.

It is well known that the euclidean space $\mathbb{R}^n$ has this property. In a narrower sense I found the following.

A normed space $X$ is finite-dimensional iff it obeys the Heine-Borel-Property.

I was wondering if the property of the space $X$ to have a norm or a metric is necessary. In the proofs I found they are all based on balls, convergence and other notions where norms or metrics are involved. But in the statement of the H-B-Property, all concepts could be defined without a reference to a norm or a metric. For example, let $K$ be a complete, ordered field (like $\mathbb{R}$), then we could define the sets $(a,b) := \{ x : a < x < b \}$ and consider the topology $\tau$ generated by the collection of all such sets. So compactness could be defined as usual by requiring a compact set to have a finite open-cover, boundedness of a set $M$ could be defined as existence of a set $(a,b)$ with $M \subseteq (a,b)$ and closedness as $M$ is closed iff $K \setminus M$ is open, i.e. $K \setminus M \in \tau$. So now could the Heine-Borel-Property be proved for $K$? I don't find any proofs that doesn't use concepts like norm or metric (even the concept of limit could be defined without these concepts, like $x$ is limit point of set $M$ iff for every set $U$ with $x \in U$ if holds that $M \cap U \ne \emptyset$). Do you know any proofs that just use these basic notions?

rschwieb
  • 153,510
StefanH
  • 18,086

1 Answers1

9

The Schwartz space $\mathcal D(\mathbb R^n)$ of test functions has the Heine-Borel property and is not metrizable.

Mariano Suárez-Álvarez
  • 135,076
  • How do you define bounded without a metric? – JSchlather Jan 15 '13 at 18:32
  • 2
    Here bounded should be taken to have its usual meaning in the context of topological vector spaces, of course; there is a Wikipedia page on that at http://en.wikipedia.org/wiki/Bounded_set_%28topological_vector_space%29 – Mariano Suárez-Álvarez Jan 15 '13 at 18:32
  • Yeah. It is silly that spaces that have property X be called Y... :-/ – Mariano Suárez-Álvarez Jan 15 '13 at 18:44
  • I suppose your last comment was a response to my deleted comment pointing out that (barreled spaces) with this "Heine-Borel property" are called Montel spaces. I think it makes sense to have the distinction because boundedness in the topological vector space sense cannot be detected by a metric -- otherwise this would force the space to be locally compact, hence finite-dimensional, which the space of distributions manifestly isn't. – Martin Jan 15 '13 at 19:44
  • s/distributions/test functions – Martin Jan 15 '13 at 19:53