The separable space is a one having the countable everywhere dense subset where the "dense" implies a topology. Consider a vector and Hilbert space $V$ of finite/nonfinite dimension. Please enlighten me on some details about relation between this formal definition of separability and availability of countable basis for this vector space. Seems the former should follow from the latter but I was not able to make sure this at the moment. Put differently, if vector space has a countable basis whether it should be separable as a topological space? Need proof just in a nutshell. Can I do without norm here? Certainly, for such a topological question I need the topology itself. But then the question above is meaningless for a `naked' vector space without additional structures like norm, metric, or scalar product. So is the question meaningless even in the case of a finite dimensional $V$-space?

  • $\begingroup$ basis = Schauder basis or Hamel basis? $\endgroup$ – Henno Brandsma Jan 19 '18 at 12:33

Two types of bases are considered for (topological) vector spaces $X$: a Hamel basis (a set of vectors $B$ such that every $x \in X$ can be written uniquely as some finite sum $x=\sum_{i=1}^n c_i b_i$ where $b_1,\ldots,b_n \in B$, and a Schauder basis where we have a sequence $(b_1, b_2, \ldots)$ of vectors such that for every $x \in X$ there is a unique sequence of scalars $(c_n)_n$ such that $\sum_{n=1}^\infty c_i b_i := \lim_{N \to \infty } \sum_{i=1}^N c_i b_i = x$, where the limit taken is w.r.t. a topology on $X$ that makes the vector space operations continuous (i.e. a topological vector space), most often taken to be some norm-topology.

If a TVS $X$ has a Schauder basis or a countable Hamel basis, it is topologically separable as well: the countable dense subset in either case can be taken to be $$D = \{\sum_{i=1}^n q_i b_i: n \in \mathbb{N}, q_1,\ldots,q_n \in \mathbb{Q}\}$$

which is countable as $\cup_{n=1}^\infty \mathbb{Q}^n$ is a countable set by standard set theory, plus the fact that we can approximate vectors like $c b_i$($ c \in \mathbb{R}$) by vectors of the form $q b_i$, where $q \in \mathbb{Q}$, as the rationals are dense in the reals and scalar multiplication is continuous. (This argument also works over other fields when these fields are separable, like $\mathbb{C}$).

The Wikipedia entry I linked to mentions a famous example by Enflo of a separable (topologically) Banach space without a Schauder basis. Most separable Banach spaces in practice do have a Schauder basis, though. No complete metric TVS can have a countable Hamel basis by Baire's theorem. So in neither definition of basis need a topologically separable TVS have a countable basis.

  • $\begingroup$ Hmmm, relation between species of bases turned out to be quite subtle, however. Thanks, anyway. But, how do I understand countable Hamel basis while you emphasised `finite'. Is it allowed to happen infinitely countable or noncountable for it? If it is always related to the word finite then what does it do in the infinite dimensional vector space cases? $\endgroup$ – Sir168 Jan 22 '18 at 12:30
  • $\begingroup$ @maximav the sums are finite sums not limits as in Schauder bases. A Hamel basis often cannot be found explicitly especially uncountable ones. Zorn’s lemma implies that all vector spaces have a Hamel basis. If a metric vector space has a countable Hamel basis, it cannot be completely metrisable. $\endgroup$ – Henno Brandsma Jan 22 '18 at 21:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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