Equivalent Characterizations of Finite Dimensionality for Banach Spaces
This post lists out various conditions on Banach spaces equivalent to finite dimensionality. Finite dimensionality in Banach-space theory is exactly the point at which boundedness behaves like compactness. Once the unit ball ceases to be compact, we enter the genuinely infinite-dimensional world. Nearly every equivalent characterization is some reformulation of that fact.
Let \(X\) be a real Banach space. The following conditions are equivalent to \(\dim X < \infty\).
Basic Norm-Topological Characterizations
\(X\) is finite dimensional.
The closed unit ball \(B_X = \{x \in X : \|x\| \le 1\}\) is compact in norm.
Some closed ball \(\overline B(x,r)\) with \(r>0\) is compact in norm.
Every closed and bounded subset of \(X\) is compact.
Every bounded subset of \(X\) is totally bounded.
Every bounded sequence in \(X\) has a norm-convergent subsequence.
Every bounded infinite subset of \(X\) has a norm-accumulation point.
The unit sphere \(S_X = \{x \in X : \|x\| = 1\}\) is compact.
These are all versions of the Heine–Borel property.
Local Compactness Characterizations
\(X\) is locally compact.
\(0\) has a compact neighborhood.
Some nonempty open set has compact closure.
For normed linear spaces, local compactness forces finite dimensionality; that is a standard consequence of Riesz’s lemma.
Operator-Theoretic Characterizations
The identity operator \(I_X:X \to X\) is compact.
Every bounded linear operator \(T:X \to Y\) into every Banach space \(Y\) is compact.
Every bounded linear operator \(T:X \to X\) is compact.
Indeed, \(I_X\) is bounded, so any statement implying \(I_X\) is compact gives finite dimensionality, and conversely if \(X\) is finite dimensional then every bounded operator out of \(X\) is compact.
Weak Versus Norm Compactness
The closed unit ball is weakly compact and the weak topology agrees with the norm topology on it.
Every weakly convergent sequence converges in norm.
Every weakly Cauchy bounded sequence has a norm-convergent subsequence.
These are not standard as the first line of attack, but they are equivalent because on a finite-dimensional space weak and norm topologies coincide, while on an infinite-dimensional Banach space the unit ball cannot be norm compact.
Condition 16 is especially strong: if every weakly convergent sequence converges in norm, then in particular on the weakly compact unit ball every sequence has a norm-convergent subsequence, so the unit ball is norm compact and \(X\) is finite dimensional.
Coordinate/Topological Vector Space Characterizations
The norm topology on \(X\) is equivalent to the Euclidean topology on \(\mathbb R^n\) for some \(n\).
\(X\) is linearly homeomorphic to \(\mathbb R^n\) for some \(n\).
Every linear subspace of \(X\) is closed.
This last condition is classical. In infinite dimensions one can always construct a proper dense subspace, so if every subspace is closed then \(X\) must be finite dimensional.
Compactness Of The Unit Ball In Other Standard Senses
For the closed unit ball, the following are also equivalent to finite dimensionality:
\(B_X\) is sequentially compact in norm.
\(B_X\) is countably compact in norm.
\(B_X\) is limit-point compact in norm.
\(B_X\) is complete and totally bounded.
In metric spaces these are just reformulations of compactness.
Equivalent Reformulations Via Riesz’s Lemma
Every bounded sequence has a Cauchy subsequence.
The unit ball cannot contain an infinite sequence with pairwise distance bounded below by a positive constant.
For every \(\varepsilon>0\), the unit ball can be covered by finitely many balls of radius \(\varepsilon\).
Condition 27 is just total boundedness. In infinite dimensions Riesz’s lemma gives, for example, a sequence \((x_n)\) in the unit ball such that \[ \|x_n-x_m\| \ge \frac12 \qquad n \ne m, \] which rules out total boundedness and hence compactness.
Dual-Space And Functional Characterizations
The weak-* topology on \(X^*\) coincides with the weak topology on bounded subsets of \(X^*\).
Every linear functional on \(X^*\) is weak-* continuous.
The canonical embedding \(J:X \to X^{**}\) identifies \(X^{**}\) with a space of the same algebraic dimension as \(X\).
These are more abstract, and in practice one usually proves finite dimensionality first and then deduces them. But they are equivalent. In finite dimensions there is no distinction among norm, weak, and weak-* topologies up to the obvious identifications on bounded sets.
Compactness Of Closed Bounded Sets Under Continuous Images
Every continuous map from a closed bounded subset of \(X\) into \(\mathbb R\) attains its maximum and minimum.
Every closed bounded subset of \(X\) is sequentially compact.
Every bounded closed convex subset of \(X\) is compact.
Condition 33 is enough: apply it to the unit ball.
A Useful Minimal Core List
The most important equivalences:
- \(\dim X < \infty\);
- the closed unit ball is norm compact;
- every bounded set is totally bounded;
- every bounded sequence has a norm-convergent subsequence;
- \(X\) is locally compact;
- the identity operator is compact.
Those six already generate most of the rest.
Remarks On Scope
A few warnings help keep the list honest.
Reflexivity Is Not Enough
Weak compactness of the unit ball is equivalent to reflexivity, not finite dimensionality. So \[ B_X \text{ weakly compact} \] is not on the finite-dimensional list by itself.
Separability Is Not Enough
Separability, second countability, and metrizability are not equivalent to finite dimensionality; many infinite-dimensional Banach spaces have them.
Attainment Statements Usually Give Reflexivity, Not Finite Dimensionality
For example, “every continuous linear functional attains its norm on the unit ball” is related to James-type theorems and reflexivity, not finite dimensionality.
Standard Proof Pattern
Most of the norm-compactness equivalences boil down to one theorem.
Riesz’s Lemma
If \(Y\) is a proper closed subspace of a normed space \(X\) and \(0<\alpha<1\), then there exists \(x \in X\) with \(\|x\|=1\) such that \[ \|x-y\| \ge \alpha \qquad \text{for all } y \in Y. \]
From this we inductively construct a sequence \((x_n)\) in the unit ball with \[ \|x_n-x_m\| \ge \frac12 \qquad n \ne m. \] Hence the unit ball is not totally bounded in any infinite-dimensional normed space. Therefore it cannot be compact. That proves \[ \text{compact unit ball} \Longrightarrow \dim X < \infty. \]
The converse is the familiar Heine–Borel theorem in \(\mathbb R^n\) together with equivalence of norms on finite-dimensional spaces.