Let $n\in \mathbb N$ and let $D^n$ be the closed $1$-ball in $\left(\mathbb R^n, \|\,\cdot\,\|_1\right)$. It is not too hard to show that $[0,1]^n \cong D^n$ in this case.

This observation leads to the question whether we also have $[0,1]^\omega \cong D^\omega$, where

$$D^\omega = \left\{(x_n)_{n\in \mathbb N} \in \ell^1 \; \Bigg| \; \sum_{n=1}^\infty |x_n|\le 1\right\}$$

is the closed unit ball in $\ell^1$.

Now if we view $D^\omega$ as a subspace of $\ell^1$, then we can prove that it is not homeomorphic to $[0,1]^\omega$ by arguing that the latter is compact, while the former is not.

But what happens if we view $D^\omega$ as a subspace of $\mathbb R^\omega$?

Is $D^\omega$ homeomorphic to $[0,1]^\omega$, when both sets are endowed with the subspace topologies induced by $\mathbb R^\omega$ (in the product topology)?

In this case both spaces are compact ($D^\omega$ is the intersection over all $n\in \mathbb N$ of the compact sets $\{x\mid \sum_{i=1}^n |x_i| \le 1\} \cap [-1,1]^\omega$) and all topological properties I can think of are preserved when we go from $[-1,1]^\omega \cong [0,1]^\omega$ to the subspace $D^\omega$. I have also tried to explicitly construct a homeomorphism, but to no avail.

I hope some of the topologically savvy guys on this site could help me out here! Thanks =)

  • $\begingroup$ Sorry for going off-topic here: in the question on $A \subset f(B)$ you deleted your answer, but can't you massage it into a full solution using paracompactness and a partition of unity subordinate to a locally finite subcovering of $\{B \smallsetminus A\} \cup \{B_{r(a)}(a)\,:\,a \in A\}$ of $B$? $\endgroup$ – t.b. May 9 '12 at 13:44

The answer is yes, they are homeomorphic.

Edit: Here's a better way to put the answer below:

Theorem (Klee, 1955). Let $K$ be a compact, convex and metrizable set in a locally convex space $E$. If $K$ is not contained in a finite-dimensional subspace of $E$ then $K$ is homeomorphic to the Hilbert cube $[0,1]^\omega$.

This applies directly to the compact convex set $K = D^\omega$ in the metrizable space $E = \mathbb{R}^\omega$. This result is not explicitly stated this way in Klee's work [1], but it follows immediately from Theorem (1.2) of that paper. The proof idea is outlined a bit further down.

Original Answer:

Klee proved in 1955, based on a theorem due to Keller, the following remarkable result (Theorem (1.2) of reference [1] below):

Let $X$ be a separable normed space and let $E$ be either $X$ with the norm topology, $X$ with the weak topology or $X^\ast$ with the weak$^\ast$ topology. If $K$ is an infinite-dimensional compact convex subset of $E$ then $K$ is homeomorphic to the Hilbert cube $[0,1]^\omega$.

The idea is that in each case one can find a countable family of continuous linear functionals $(f_n)$ separating the points of $K$ and, normalizing them appropriately, the map $x \mapsto (f_1(x),f_2(x),\ldots)$ gives a linear homeomorphism from $K$ onto an infinite-dimensional norm-compact convex subset of $\ell^2$. Keller had previously shown that all infinite-dimensional compact convex sets in $\ell^2$ are homeomorphic to $[0,1]^\omega$.

Since $D^\omega$ is a compact convex subset of $\ell^1 = (c_0)^\ast$, Klee's theorem applies. To finish up, note that the topology on $D^\omega$ viewed as a subset of $(c_0)^\ast$ and the topology induced from $\mathbb{R}^\omega$ coincide by the standard proof of Alaoglu's theorem.

Here are the relevant papers:

  1. V. L. Klee, Some topological properties of convex sets, Trans. Amer. Math. Soc. 78 (1955), 30–45. MR0069388

  2. Ott-Heinrich Keller, Die Homoiomorphie der kompakten konvexen Mengen im Hilbertschen Raum, Math. Ann. vol. 105 (1931) pp. 748–758. MR1512740

  • $\begingroup$ Of course, it would be desirable to have a more explicit argument! $\endgroup$ – t.b. Apr 18 '12 at 14:21
  • $\begingroup$ Wow! Thank you very much. I'm going to have a look at these papers, then. $\endgroup$ – Sam Apr 18 '12 at 14:32
  • 1
    $\begingroup$ @Sam: I slightly expanded my answer. $\endgroup$ – t.b. Apr 19 '12 at 19:59

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.