compactness of closed unit ball in normed spaces

finite dimensional infinite-dimensional subspace convergent

Theorem - Let $X$ be a normed space and $\overline{B_1(0)}\subset X$ the closed ball (http://planetmath.org/Ball). Then $\overline{B_1(0)}$ is compact if and only if $X$ is finite dimensional (http://planetmath.org/Dimension2).

The above result is false, in general, if one is considering other topologies in $X$ besides the norm topology (see, for example, the Banach-Alaoglu theorem). It follows that infinite dimensional (http://planetmath.org/Dimension2) normed spaces are not locally compact.

Proof:

• $(⟸)$ This is the easy part. Since $X$ is finite dimensional it is isomorphic to some ${ℝ}^n$ (with the standard topology). The result then follows from the Heine-Borel theorem.

• $(⟹)$ Suppose that $X$ is not finite dimensional. Pick an element $x_1\in X$ such that $\parallel x_1\parallel =1$ and denote by $S_1$ the subspace (http://planetmath.org/VectorSubspace) generated by $x_1.$

  Recall that, according to the Riesz Lemma (http://planetmath.org/RiezsLemma), there exists an element $x_2\in X$ such that $\parallel x_2\parallel =1$ and $d(x_2,S_1)\ge \frac{1}{2}$, where $d(y,M):=inf\{\parallel y-z\parallel :z\in M\}$ denotes the distance between an element $y\in X$ and a subspace $M\subset X.$

  Now consider the subspace $S_2$ generated by $x_1,x_2$. Since $X$ is infinite dimensional $S_2$ is a proper subspace and we can still apply the Riesz Lemma to find an element $x_3$ such that $\parallel x_3\parallel =1$ and $d(x_3,S_2)\ge \frac{1}{2}.$

  If we proceed inductively, we will find a sequence $\{x_n\}$ of norm 1 elements and a sequence of subspaces $S_n:=⟨x_1,\mathrm{\dots },x_n⟩$ such that $d(x_{n+1},S_n)\ge \frac{1}{2}.$ Under this setting it is easily seen that the sequence $\{x_n\}$ is in $\overline{B_1(0)}$ and satisfies $\parallel x_n-x_m\parallel \ge \frac{1}{2}$ for all $m\ne n\in \mathbb{N}$. Therefore, $\{x_n\}$ is a sequence in $\overline{B_1(0)}$ that has no convergent (http://planetmath.org/ConvergentSequence) subsequence, i.e., $\overline{B_1(0)}$ is not compact. $\mathrm{\square }$

Remarks on the proof - Note that the sequence $\{x_n\}$ constructed in the proof does not have a Cauchy subsequence (http://planetmath.org/CauchySequence). Thus we have in fact proven the slightly stronger result that $X$ is finite dimensional if and only if every bounded sequence in $X$ has a Cauchy subsequence.

For Hilbert spaces the proof would be slightly simpler because one could just pick any orthonormal basis $\{e_n\},$ and it would $\parallel e_n-e_m\parallel =\sqrt{2}$ for all $m,n\in \mathbb{N}$ with $m\ne n,$ therefore having no convergent subsequence. For general normed spaces we cannot just pick orthonormal elements, since this notion does not exist. Thus, we have to use Riesz Lemma to assure the existence of elements with some .

Title	compactness of closed unit ball in normed spaces
Canonical name	CompactnessOfClosedUnitBallInNormedSpaces
Date of creation	2013-03-22 17:48:41
Last modified on	2013-03-22 17:48:41
Owner	asteroid (17536)
Last modified by	asteroid (17536)
Numerical id	13
Author	asteroid (17536)
Entry type	Theorem
Classification	msc 46B50
Synonym	closed unit ball in a normed space is compact iff the space is finite dimensional