# Start a new discussion

## Not signed in

Want to take part in these discussions? Sign in if you have an account, or apply for one below

## Discussion Tag Cloud

Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.

• CommentRowNumber1.
• CommentAuthorUrs
• CommentTimeApr 28th 2017
• (edited Jun 16th 2021)

I have spelled out the proof at paracompact Hausdorff spaces equivalently admit subordinate partitions of unity.

This uses Urysohn’s lemma and the shrinking lemma, whose proofs are not yet spelled out on the $n$Lab.

• CommentRowNumber2.
• CommentAuthorUrs
• CommentTimeApr 30th 2017
• (edited Apr 30th 2017)

Okay, proofs of

are now all filled in, so now the proof at paracompact Hausdorff spaces equivalently admit subordinate partitions of unity is complete. I have added that it needs excluded middle and the axiom of choice (for the shrinking lemma).

• CommentRowNumber3.
• CommentAuthorTobyBartels
• CommentTimeMay 3rd 2017

If you use the axiom of choice, then you don't have to separately mention excluded middle, which follows from choice. (Excluded middle is Kuratowski-finite choice.) Of course, if you only use (say) countable choice, then excluded middle is independent of that.

• CommentRowNumber4.
• CommentAuthorTodd_Trimble
• CommentTimeJun 4th 2017

I have tweaked the statement paracompact Hausdorff spaces equivalently admit subordinate partitions of unity so that it now reads: (Assuming the axiom of choice) Let $X$ be $T_1$. Then $X$ is paracompact and Hausdorff iff every open cover admits a subordinate partition of unity. I also tweaked the “if” part in the proof so as to derive Hausdorffness.

It’s just a slight strengthening of the statement, but a useful one. :-)

• CommentRowNumber5.
• CommentAuthorUrs
• CommentTimeJun 6th 2017

Thanks. I had been thinking I should do this, but didn’t get around to. Thanks very much.

• CommentRowNumber6.
• CommentAuthorDmitri Pavlov
• CommentTimeJun 15th 2021
• (edited Jun 15th 2021)

Why exactly do we need T1 and Hausdorffness properties here?

Is it not true that paracompact topological spaces are precisely those spaces that admit a subordinate locally finite partition of unity for every open cover?

The proofs and their dependencies appear to make no use of T1 or Hausdorffness properties.

• CommentRowNumber7.
• CommentAuthorDavidRoberts
• CommentTimeJun 16th 2021
• (edited Jun 16th 2021)

Can you get away from using Urysohn’s lemma? That requires the space to be normal, but maybe you mean the version without $T_1$?

It’s a pity Bourbaki in General Topology defines “paracompact” to include the Hausdorff condition, otherwise that would be a good place to check.

• CommentRowNumber8.
• CommentAuthorDavidRoberts
• CommentTimeJun 16th 2021

I guess there’s this: https://mathoverflow.net/a/360635/, though that points back to the nLab.

• CommentRowNumber9.
• CommentAuthorDmitri Pavlov
• CommentTimeJun 16th 2021

Re #8: Curiously, it points back to the nLab article for which I wrote the cited section.

So it appears that the correct condition is that the space is a fully normal topological space.

• CommentRowNumber10.
• CommentAuthorDmitri Pavlov
• CommentTimeJun 16th 2021

Ironically, I added the relevant references to the nLab article numerable open cover and forgot about them.