Neighbourhood base
Posted on 31st January 2019
Reading a paper with my office mate, we ended up having a discussion about the notion of an "open neighbourhood base"
in a topological space. For example, I might informally say that the weak topology on a Banach space
has, around a point , an open neighbourhood base is given by the sets
where are members of and .
This raises a natural question:
Suppose we have a set and for each we have specified a collection
of subsets of , such that .
When is there a topology on such that are the "basic open neighbourhoods" of ?
What precisely do we mean by this? I think this is a little unclear, so let's be precise. We seek
a topology on so that:
- (1) Each member of is in ; and
- (2) Each should be a filter-base
for the collection of all neighbourhoods of . That is, if
with then there is with . Clearly it suffices to
consider the cases when .
Consider the topology, say , generated by all the sets
that is, members of are arbitrary unions of sets of the norm for
. Condition (1) holds, and condition (2) will hold exactly when, given
each containing , there is with . In particular:
- (2a) Given there is with ;
- (2b) Given with , there is with ;
These two conditions then imply (2), for given each containing , by (2b) we
can find with , and then by (2a) there is with
.
Consider now
Then contains and , and is obviously closed under unions. Condition (2a)
implies that is closed under finite intersections, that is, that is a topology.
If and then by (2), there is with ; thus .
Hence .
Finally, given , if then there is with ,
and so is the union of members of , so . We conclude that our putative
topology is unique, and equals (or ). Further, we have shown that
exists exactly when (2a) and (2b) hold (which are conditions on the families ).
Let us return to the weak topology. (2a) is clear, but to show (2b) is a little exercise
in picking epsilons.