Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Separable Metric Space: Local Compactness and Completeness

0 views
Skip to first unread message

Shozo Mori

unread,
Nov 11, 2009, 9:46:22 PM11/11/09
to
Is a locally compact separable metric space complete?
Or is a complete separable metric space locally compact?

Thank you very much.
- Shozo

José Carlos Santos

unread,
Nov 12, 2009, 2:08:28 AM11/12/09
to
On 12-11-2009 2:46, Shozo Mori wrote:

> Is a locally compact separable metric space complete?

No. Consider an open interval (a,b) in R, with its usual distance.

> Or is a complete separable metric space locally compact?

Again, no. Consider the space of all absolutely convergent series
of real numbers with the distance

d(sum_n a_n,sum_n b_n) = sum_n |a_n - b_n|.

Best regards,

Jose Carlos Santos

David C. Ullrich

unread,
Nov 12, 2009, 7:41:16 AM11/12/09
to

Good example, except you didn't say what you meant.
sum a_n is a _number_; you've defined the distance
between two reals, and in an ambiguous way since
a given real can be written as sum a_n in many different ways.

You meant he should consider the space of all absolutlely
summable _sequences_, with the metric

d((a_n), (b_n)) = sum_n |a_n - b_n|.

>Best regards,
>
>Jose Carlos Santos

David C. Ullrich

"Understanding Godel isn't about following his formal proof.
That would make a mockery of everything Godel was up to."
(John Jones, "My talk about Godel to the post-grads."
in sci.logic.)

Shozo Mori

unread,
Nov 12, 2009, 12:13:47 PM11/12/09
to
Jose and David,

Thank you very much for the responses, which I appreciate.

Another question: Is the (a ?) completion of a locally compact, separable metric space locally compact and separable?

Probably the answer is yes.

If so, I might say any locally compact, separable metric space is "almost" complete. Do you agree?

Or, the answer may be "No," and Jose's counter example would be also a counter example of the above question. Which do you agree with?

By the way, the fact that a complete separable metric space (Polish space ?) is "more general" than a locally compact separable metric space is "more or less" obvious, when we consider a vector space, i.e., a Euclidean space vs. C([0 1]).

Do you agree?
Again thank you very much.
- Shozo

José Carlos Santos

unread,
Nov 12, 2009, 6:26:58 PM11/12/09
to
On 12-11-2009 17:13, Shozo Mori wrote:

> Thank you very much for the responses, which I appreciate.
>
> Another question: Is the (a ?) completion

It is "a" completion, not "the" completion.

> of a locally compact,separable metric space locally compact

I don't know.

> and separable?

Of course, Because if A is a dense subset of B and B is a dense subset
of C, then A is a dense subset of C.

> Probably the answer is yes.
>
> If so, I might say any locally compact, separable metric space
> is "almost" complete. Do you agree?

Actually, *any* metric space is "almost complete", in the sense that
every metric space is isometric to a dense subspace of a complete metric
space.

David C. Ullrich

unread,
Nov 13, 2009, 6:54:29 AM11/13/09
to
On Thu, 12 Nov 2009 23:26:58 +0000, Jos� Carlos Santos
<jcsa...@fc.up.pt> wrote:

>On 12-11-2009 17:13, Shozo Mori wrote:
>
>> Thank you very much for the responses, which I appreciate.
>>
>> Another question: Is the (a ?) completion
>
>It is "a" completion, not "the" completion.

??? Of course that's literally true, but any two completions
of a metric space are isometric, so it's "the" completion,
up to the strongest possible relevant sort of isomorphism.

>> of a locally compact,separable metric space locally compact
>
>I don't know.
>
>> and separable?
>
>Of course, Because if A is a dense subset of B and B is a dense subset
>of C, then A is a dense subset of C.
>
>> Probably the answer is yes.
>>
>> If so, I might say any locally compact, separable metric space
>> is "almost" complete. Do you agree?
>
>Actually, *any* metric space is "almost complete", in the sense that
>every metric space is isometric to a dense subspace of a complete metric
>space.
>
>Best regards,
>
>Jose Carlos Santos

David C. Ullrich

José Carlos Santos

unread,
Nov 13, 2009, 9:13:54 AM11/13/09
to
On 13-11-2009 11:54, David C. Ullrich wrote:

>> On 12-11-2009 17:13, Shozo Mori wrote:
>>
>>> Thank you very much for the responses, which I appreciate.
>>>
>>> Another question: Is the (a ?) completion
>> It is "a" completion, not "the" completion.
>
> ??? Of course that's literally true, but any two completions
> of a metric space are isometric, so it's "the" completion,
> up to the strongest possible relevant sort of isomorphism.

Right. :-( For a moment I was thinking about another type of completion,
namely about completion for a metrizable topological space. In general
there are several.

David C. Ullrich

unread,
Nov 13, 2009, 12:45:59 PM11/13/09
to
On Thu, 12 Nov 2009 12:13:47 EST, Shozo Mori <Shoz...@AOL.com>
wrote:

>Jose and David,
>
>Thank you very much for the responses, which I appreciate.
>
>Another question: Is the (a ?) completion of a locally compact, separable metric space locally compact and separable?
>
>Probably the answer is yes.

Of course it's yes for the "separable". The following is I think a
counterexample for "locally compact". Probably stupid, there must
be much simpler examples:

Say X is the space of all sequences a = (a_1, a_2, ...) of reals such
that ||a|| = sum |a_n| < infinity; define

d(a,b) = sum |a_j - b_j| = ||a-b||.

So X is a Banach space: a complete normed vector space.

Let H_1, H_2, ... be the sequence of finite-dimensional
subspaces of X: say a is in H_n if a is in X and a_j = 0 for
all j > n.

Choose a strictly in increasing sequence of positive numbers r_n
that tends to 1.

Let E be the union of the sets

E_n = {a in H_n : r_n <= ||a|| <= r_{n+1}}

Then it seems clear to me that E is locally compact, since any point
has a neighborhood isometric to a closed subset of some euclidean
space.

But the completion of E is the same as the closure of E in X,
which is the union of E and the unit sphere S = {a in X : ||a||=1}
This is not locally compact.

0 new messages