Continuity of Linear Functionals in Initial Topology on Vector Space Generated by Linear Functionals
Theorem
Let $\GF \in \set {\R, \C}$.
Let $X$ be a vector space over $\GF$.
Let $F$ be a set of linear functionals on $X$ that form a vector space over $\GF$.
That is, for each $\lambda, \mu \in \GF$ and $f, g \in F$, we have:
- $\lambda f + \mu g \in F$
Let $\tau$ be the initial topology on $X$ generated by $F$.
Let $g : X \to \GF$ be a linear functional.
Then $g$ is $\tau$-continuous if and only if $g \in F$.
Proof
Sufficient Condition
From the definition of the initial topology, $\tau$ is the coarsest topology making all $g \in F$ continuous.
So if $g \in F$ then $g$ is $\tau$-continuous.
$\Box$
Necessary Condition
Suppose that $g$ is $\tau$-continuous.
For each $x \in \GF$ and $\epsilon > 0$, let:
- $\map {B_\epsilon} {x, \GF} = \set {y \in \GF : \cmod {y - x} < \epsilon}$
Then since $g$ is $\tau$-continuous, we have:
- $g^{-1} \sqbrk {\map {B_1} {0, \GF} } \in \tau$
and so $g^{-1} \sqbrk {\map {B_1} {0, \GF} }$ is an open neighborhood of ${\mathbf 0}_X$.
From Initial Topology on Vector Space Generated by Linear Functionals is Locally Convex, $\tau$ is generated by the seminorms:
- $\set {p_f : f \in F}$
where we define $p_f : X \to \R_{\ge 0}$ by:
- $\map {p_f} x = \cmod {\map f x}$
for each $x \in X$.
So, from Open Sets in Standard Topology of Locally Convex Space, there exists $g_1, \ldots, g_n \in F$ and $\epsilon > 0$ such that:
- $\set {x \in X : \cmod {\map {g_k} x} < \epsilon \text { for each } 1 \le k \le n} \subseteq g^{-1} \sqbrk {\map {B_1} {0, \GF} }$
We will deduce that:
- $g \in \span \set {g_1, \ldots, g_n}$
so that $g \in F$.
Let:
- $\ds x \in \bigcap_{k \mathop = 1}^n \ker g_k$
Then $\map {g_k} {\lambda x} = 0$ for each $\lambda \in \GF$ and $1 \le k \le n$.
In particular, $\cmod {\map {g_k} {\lambda x} } < \epsilon$ for each $\lambda \in \GF$ and $1 \le k \le n$.
So $\lambda x \in g^{-1} \sqbrk {\map {B_1} {0, \GF} }$ for each $\lambda \in \GF$.
Then:
- $\cmod {\map g {\lambda x} } < 1$
for each $\lambda \in \GF$.
So, by the linearity of $g$ we have:
- $\cmod \lambda \cmod {\map g x} < 1$
for each $\lambda \in \GF$.
If $\cmod {\map g x} > 0$, we could take:
- $\ds \cmod \lambda > \frac 1 {\cmod {\map g x} }$
to get:
- $\cmod \lambda \cmod {\map g x} > 1$
Hence we must have $\map g x = 0$, so that $g \in \ker f$.
So, we have:
- $\ds \bigcap_{k \mathop = 1}^n \ker g_k \subseteq \ker g$
So from Condition for Linear Dependence of Linear Functionals in terms of Kernel, we have:
- $g \in \span \set {g_1, \ldots, g_n}$
and so $g \in F$ as desired.
$\blacksquare$