Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I'm having trouble constructing the Hausdorff completion of a uniform space $(X,U)$ using pseudometrics. I know that every uniformity on a space $X$ is made by pseudometrics. Here is my idea:

Let $(X,U)$ be a uniformity generated by pseudometrics $(d_i: X \rightarrow\mathbb{R}_{\geq 0})_{i \in I}$. Define 'Generalized' Cauchy Sequences (GCS) as followed: $(x_n)_{n \in \mathbb{N}}$ is a GCS iff $$\forall \varepsilon > 0 \forall K \subset I \text{ finite }\forall i \in K \exists n_0 \forall p,q \geq n_0 : d_i(x_p,x_q)<\varepsilon$$ Let $\mathcal{C}$ be the set of all GCS and define a equivalence relation as followed: $$ (x_n)_{n \in \mathbb{N}}R(y_n)_{n \in \mathbb{N}} \Leftrightarrow \forall \varepsilon > 0 \forall K \subset I \text{ finite }\forall i \in K \exists n_0 \forall n \geq n_0 : d_i(x_n,y_q)<\varepsilon$$

Let $X'$ be the set $\mathcal{C}/R$. Define pseudometrics $d'_i((x_n)_{n \in \mathbb{N}},(y_n)_{n \in \mathbb{N}})$ as $\lim_{n \rightarrow \infty} d_i(x_n,y_n)$. Let $k: X \rightarrow X'$ be the canonical function.

Now, I can prove that $X'$ is Hausdorff (thanks to the equivalence relation), $k(X)$ is dense is $X'$ (every point in $X'$ is the limit of images of points in $X$), but I'm having trouble proving that $X'$ is complete.

Basically, I have 2 questions: is what I'm doing correct (and if so, how do I proceed) or else, what am I doing wrong here (and how to correct it).

As always, any help would be appreciated.

share|improve this question

1 Answer 1

up vote 2 down vote accepted

Notice first that you’ve not gained anything by looking at finite $K\subseteq I$: your definition of GCS is equivalent to saying that the sequence is $d_i$-Cauchy for every $i\in I$. The real problem, though, is that in general sequences don’t suffice in uniform spaces: you have to look at nets or filters. Thus, you can’t expect that adding points for sequences that ought to converge will be enough. Let me suggest a different approach (which is the standard one).

For $i\in I$ let $\langle X_i,d_i\rangle$ be the pseudometric space with underlying set $X$ and pseudometric $d_i$. For $x,y\in X$ define $x\stackrel{i}\sim y$ iff $d_i(x,y)=0$, let $Y_i$ be the quotient $X_i/\stackrel{i}\sim$, let $q_i$ be the quotient map, and let $\rho_i$ be the metric on $Y_i$ induced by $d_i$: $$\rho_i\big(q_i(x),q_i(y)\big)=d_i(x,y)$$ for any $q_i(x),q_i(y)\in Y_i$. Let $Y=\prod_{i\in I}Y_i$. Then the map

$$e:X\to Y:x\mapsto\langle q_i(x):i\in I\rangle$$

is a uniform embedding. (This takes a bit of argument, but it’s fairly straightforward.)

Now let $\hat Y_i$ be the metric completion of $Y_i$ for each $i\in I$. Show that $\prod_{i\in I}\hat Y_i$ is a complete uniform space, and conclude that the closure of $e[X]$ in $\prod_{i\in I}\hat Y_i$ is a Hausdorff completion of $X$.

share|improve this answer
1  
@MinimusHeximus: A uniformly continuous embedding. –  Brian M. Scott Jun 3 '13 at 6:35

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.