2
$\begingroup$

i was trying to solve this exercise from Royden:

Let X be a locally convex topological vector space, let $Y \subset X$ be a closed subspace and $x_0 \in X-Y$.Prove that there exists a continuous linear functional $\varphi: X \rightarrow \mathbb{R}$ such that $\varphi(y) = 0$ if $y \in Y$ and $\varphi(x_0) \ne 0$

I don't understand where to use the fact that $Y$ is closed. My attempt was to construct a positive homogeneous and subadditive functional $p_y(x) = 1 - \chi_{Y}$. It's not hard to see that it has the desired properties. Moreover it is clear that $p_y(x_0) = 1$. Thus one can take $\varphi(\lambda x_0) = \lambda$ and extend it to a linear functional $\varphi \le p_y$ on $X$. Observing that $p_y(x) \le 1$ we have that $\varphi$ is bounded on $X$ and thus, because $X$ is locally convex topological vector spaces, continuous.

Isn't it right?

$\endgroup$
  • $\begingroup$ What formulation of the hahn banach theorem are you using? I only ask because your function $p_y$ is not sublinear. $\endgroup$ – Aweygan Dec 28 '16 at 15:12
  • $\begingroup$ Why not? Let $x,y \in X$ if $x+y \in Y$ it's obvious since $p_y(x+y) =0$ otherwise, since $Y$ is a subspace of X, we must have $x \not\in Y$ or $y \not\in X$. Thus subadditivity is true also in this case. $\endgroup$ – jJjjJ Dec 28 '16 at 18:27
  • $\begingroup$ I forgot to tag @Aweygan $\endgroup$ – jJjjJ Dec 28 '16 at 20:11
  • $\begingroup$ $p_y$ is just the indicator function for $X\setminus Y$. Indicator functions are not positive homogeneous. $\endgroup$ – Aweygan Dec 28 '16 at 20:16
  • $\begingroup$ @Aweygan right. Sorry, was stupid $\endgroup$ – jJjjJ Dec 28 '16 at 20:19
0
$\begingroup$

Since both the sets $Y$ and $\{x_0 \}$ are closed and $\{x_0\}$ is compact then from separation theorem there exists a linear functional $f: X\to \mathbb{R} $ and real numbers $a< b$ such that $f(y)<a $ for $y\in Y $ and $f(x_0 ) =b .$ Now, take any $u\in Y$ then $$f(ny )<a $$ for all $n\in \mathbb{N}$ hence $f(y)\leqslant 0$ and analogously $$f(-ny )<a$$ thus $f(y) \geqslant 0$ and finally $$f(y)=0.$$

$\endgroup$
  • $\begingroup$ Yes that's ok, but I asked if my solution is correct or not, since I don't use the fact that $Y$ is closed $\endgroup$ – jJjjJ Dec 27 '16 at 17:14

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.