Characterization of von Neumann-Boundedness in Hausdorff Locally Convex Space
Theorem
Let $\GF \in \set {\R, \C}$.
Let $\struct {X, \PP}$ be a Hausdorff locally convex space over $\GF$.
Let $U \subseteq X$.
Then $U$ is von Neumann-bounded if and only if:
- for each $p \in \PP$ there exists $C_p > 0$ such that:
- $\map p x < C_p$
- for each $x \in U$.
Proof
For each $p \in \PP$, let:
- $B_p = \set {y \in X : \map p y < 1}$
Note that by the definition of the standard topology, $B_p$ is an open neighborhood of $\mathbf 0_X$.
Let $r > 0$.
From Seminorm Axiom $\text N 2$: Positive Homogeneity, for $y \in X$ we have:
- $\map p y < 1$
- $\map p {r y} < r$
So, we have:
- $r B_p = \set {y \in X : \map p y < r}$
Necessary Condition
Suppose that $U$ is von Neumann-bounded.
Then there exists $s > 0$ such that:
- $U \subseteq t B_p = \set {y \in X : \map p y < t}$
for $t > s$
In particular:
- $U \subseteq \set {y \in X : \map p y < s + 1}$
That is:
- $\map p x < s + 1$
for all $x \in U$.
$\Box$
Sufficient Condition
Suppose that for each $p \in \PP$ there exists $C_p > 0$ such that:
- $\map p x < C_p$
for each $x \in U$.
Let $V$ be an open neighborhood of $\mathbf 0_X$.
From Open Sets in Standard Topology of Locally Convex Space, there exists $p_1, \ldots, p_n \in \PP$ and $\epsilon > 0$ such that:
- $\ds \bigcap_{k \mathop = 1}^n \epsilon B_{p_k} = \set {y \in X : \map {p_k} y < \epsilon \text { for each } 1 \le k \le n} \subseteq V$
Let:
- $\ds r = \frac 1 \epsilon \max_{1 \le k \le n} C_{p_k}$
Then:
- $\ds r \bigcap_{k \mathop = 1}^n \epsilon B_{p_k} = \bigcap_{k \mathop = 1}^n \paren {\max_{1 \le k \le n} C_{p_k} } B_{p_k} \subseteq r V$
and generally if $s > r$:
- $\ds \bigcap_{k \mathop = 1}^n \paren {\max_{1 \le k \le n} C_{p_k} } B_{p_k} \subseteq s \bigcap_{k \mathop = 1}^n \epsilon B_{p_k} \subseteq s V$
Note that for $x \in U$, $1 \le k \le n$ and $s > r$ we then have:
- $\ds \map {p_k} x < C_{p_k} \le \max_{1 \le k \le n} C_{p_k} < s \epsilon$
for $x \in U$, and hence:
- $U \subseteq s V$
for $s > r$.
So $U$ is von Neumann-bounded.
$\blacksquare$