I trying to understand that if a TVS V is Hausdorff, that we then can separate closed sets and points by opens.
In the accepted answer of Every Topological Vector Space is Regular, I have two questions:
(1) Why can one select opens $O,O'$ such that $Y\times 0\subseteq O \times O' \subseteq f^{-1}(U)$? Somehow this uses continuity. I can see that for every $y\in Y$ we can select opens with $(y,0)\in O\times O' \subseteq f^{-1}(U)$ but why can we choose these to work for all $y\in Y$ at once?
(2) Where does this proof use that $Y$ is closed?
If someone can give an alternative proof using elementary tools (just the definitions and basic topology), that will equally answer my question. Just trying to understand why this result holds.
Thanks.
I am having some issues with the linked question and answers, so I'll explain my own approach instead.
Here's a trick I like: topological groups, vector spaces, etc. are a little weird to work at. But these types of problems actually depend only on the uniform structure of the given topological spaces.
I like metric spaces, though.
So what would the similar result be for metric spaces?
This is easy to do geometrically: Since $x\not\in Y$ and $Y$ is closed, then the distance from $x$ to $Y$ is positive. Let $d=d(x,Y)=\inf\left\{d(x,y):y\in Y\right\}>0$ be that distance. Then $$B_{d/2}(x)=\left\{z\in M:d(x,z)<d/2\right\}$$ and $$B_{d/2}(Y)=\left\{z\in M:d(z,Y)<d/2\right\}$$ have precisely the property that we need.
Ok, so what was the main point of the argument? We basically did the following steps:
Find an open set containing $x$ which does not intersect $Y$: $B_d(x)$ does that;
Take another open subset which has half of the diameter of the original open set: $B_{d/2}(x)$ does that
Translate that small open set to $Y$: $B_{d/2}(Y)$ works;
Done.
The same procedure can be done in topological vector spaces, topological groups, etc. For step 2., you should use the following facts:
If $x$ is a point in a topological vector space, then the open sets around $x$ are precisely those of the form $x+O$, where $O$ is an open set around $0$;
If $O$ is an open set around $0$, then there exists another open set $O'$ around $0$ such that $O'+O'\subseteq O$ (because addition is continuous).
Then $x+O'$ is an open set around $x$ which has "intuitive diameter half of that of $x+O$;