Every metric space can be embedded isometrically into the Banach space

115 Views Asked by At

I've been trying to construct my own proof of this, but the only reference I've been able to find that I could (somewhat) understand was the following. In particular, I'm confused at the use of $f_{a}(a)$ and $f_{b}(a)$ in proving that $\|f_{a} - f_{b}\| = d(a, b)$. I realize there's an almost identical question, but it's ~4 years old and received one response which didn't help my understanding.

The proof goes as follows:

Let $X$ be a set and $B(X, \mathbb{R})$ the set of bounded functions $f: X \rightarrow \mathbb{R}$ with norm $\|f\| = \sup_{x \in X}\{|f(x)|\}$. Then every metric space $(X, d)$ can be embedded isometrically into the Banach space $E = \text{Bou}(X, \mathbb{R})$.

Proof. One can assume $X \neq \emptyset$. Fix a point $a_{0} \in X$ and for every $a \in X$ define a function $f_{a}: X \rightarrow \mathbb{R}$ by $$ f_{a}(x) = d(x, a) - d(x, a_{0}). $$ Then $|f_{a}(x)| \leq d(a, a_{0})$ for every $x \in X$ so $f_{a}$ is bounded. Set $g: X \rightarrow E$, $g(a) = f_{a}$. Then we have the mapping $g: X \rightarrow E$, and it now remains to prove that $g$ is an isometry.

Let $a, b \in X$. As $x \in X$, it can be shown that $$ |f_{a}(x) - f_{b}(x)| = |d(x, a) - d(x, b)| \leq d(a, b). $$ Therefore $\|f_{a} - f_{b}\| \leq d(a, b)$. On the other hand $$ |f_{a}(a) - f_{b}(a)| = |d(a, a) - d(a, a_{0}) - d(a, b) + d(a, a_{0})| = d(a, b). $$ Therefore $\|g(a) - g(b)\| = \|f_{a} - f_{b}\| = d(a, b)$.

End Proof.

I can understand how the proof is able to show that $|f_{a}(x)|$ is bounded for every $x$. I also understand how they derived the inequality $|f_{a}(x) - f_{b}(x)| \leq d(a, b)$. Where I'm lost is how the use of $|f_{a}(a) - f_{b}(a)| = d(a, b)$, shows that the equality holds (required for an isometric embedding). I would assume the fact that just prior, we established that $|f_{a}(x) - f_{b}(x)| \leq d(a, b)$ challenges the idea that the equality case is meaningful.

1

There are 1 best solutions below

0
On BEST ANSWER

To summarise the discussion in the comments, fix $a,b\in X$. By definition,

$$\|g(a) - g(b)\| = \sup_{x\in X}|[g(a) - g(b)](x)| = \sup_{x\in X}|(f_{a} - f_{b})(x)| = \sup_{x\in X}|f_{a}(x) - f_{b}(x)|.$$

So the result follows by showing that

$$\sup_{x\in X}|f_{a}(x) - f_{b}(x)| = d(a,b).$$

Showing that $|f_{a}(x) - f_{b}(x)| \leq d(a,b)$ for all $x\in X$ allows you to conclude that $d(a,b)$ is an upper bound for $\{|f_{a}(x) - f_{b}(x)| : x\in X \}$, from which it follows that

$$\sup_{x\in X}|f_{a}(x) - f_{b}(x)| \leq d(a,b).$$

On the other hand, showing that $|f_{a}(a) - f_{b}(a)| = d(a,b)$ allows you to conclude that

$$d(a,b) = |f_{a}(a) - f_{b}(a)| \leq \sup_{x\in X}|f_{a}(x) - f_{b}(x)|.$$

This obtains the desired equality.