3.2 Separable Metric Spaces
Definition 3.13. A metric space \(X\) is said to be separable if there exists a countable dense subset in \(X\).
Example 3.14. Let \(X=\mathbb {R}\), then \(X\) is separable, since \(\mathbb {Q}\) is a countable and dense subset in \(\mathbb {R}\).
Example 3.15. Let \(X=[0,1]\) with discrete metric defined on it. Then \(X\) is not a separable metric space since the only dense subset of the discrete metric space is \(X\) itself.
Theorem 3.16. A metric space \((X,d)\) is separable if and only if \(X\) has a countable subset \(E\) with the following property. For every \(\varepsilon >0\) and \(x\in X\), there is a \(y\in E\) such that \(d(x,y)<\varepsilon \).
Proof. Let \(X\) be separable, then \(X\) has a countable dense subset \(E\). Let \(x\in X\) and \(\varepsilon >0\) be given. Since \(E\) is dense in \(X\), we have that \(\overline {E}=X\) and \(x\in \overline {E}\) so that the open ball \(B_{\varepsilon }(x)\) contains a point \(y\in E\). Therefore \(d(x,y)<\varepsilon \).
Conversely, if \(X\) has a countable subset \(E\) with the given property that every \(x\in X\) is a point of \(E\) or limit point of \(E\). Hence \(\overline {E}=X\).
Therefore \(X\) is separable.
A subset \(A\) of a metric space \(X\) is said to be dense in \(X\) if \(\overline {A}=X\). This is equivalent to the following:
- i
- For each \(x\in X\) and \(\varepsilon >0\), there exists an element \(a\in A\ni d(x,a)<\varepsilon \).
- ii
- For \(x\in X\), there exist \(\{a_n\}\in A\) such that \(a_n\rightarrow x\).(Every point \(x\in X\) is a limit point of \(A\)).
- iii
- Any open set in \(X\) contains a point of \(A\).
Example 3.17. Prove that the space \(C_0\) is separable where \(C_0=\{(x_n):x_n\rightarrow 0, n\rightarrow \infty \}\) and the metric on \(C_0\) defined by \(\hspace {0.2cm}\displaystyle { d(x,y)=\sup _{1\leq n\leq \infty }|x_n-y_n|}\)
Solution. Let \(E\) be the subset of \(C_0\) consisting of all sequences of rational numbers in which only finitely many terms are non-zero. This set is a countable subset of \(C_0\).
To prove that \(E\) is dense in \(C_0\), let \(x\) be any point of \(C_0\); we must find a point of \(E\) arbitrarily close to it.
That means choosing \(r=(r_1,r_2,\dots ,r_k,0,0,0,\dots )\) with each \(r_i\) rational and \(\sup _k|x_k-r_k|\) small.
Let \(\varepsilon >0\) be given, we must find an \(n_0\) such that \(|x_n|<\varepsilon \) \(\forall n\geq n_0\).
Let \(r_1,r_2,\dots ,r_{n_0}\) be rational numbers such that \(|x_n-r_n|<\varepsilon \) \(\forall n=1,2,3,\dots ,n_0\). Then \((r_1,r_2,\dots ,r_{n_0},0,0,0,\dots )\) is the required point.
Example 3.18. Show that \(l_{\infty }\), the space of bounded sequences under \(\displaystyle {d(x,y)=\sup _{1\leq i<\infty }|x_i-y_i|}\), is not separable.
Solution. Let \(A\) be the set of sequences all of whose terms are \(0\) or \(1\). Every such sequence is bounded, so \(A\subset l_{\infty }\).
\(A\) is uncountable
A member of \(A\) is exactly a choice of \(0\) or \(1\) at each of countably many places, so \(A\) is in one-one correspondence with the collection of all subsets of \(\mathbb {N}\), which Cantor’s diagonal argument shows to be uncountable.
Its points are far apart
If \(x,y\in A\) are distinct they differ in some coordinate, and there the difference is \(1\), so \[d(x,y)=\sup _i|x_i-y_i|=1 .\] Hence the balls \(B_{1/3}(x)\), for \(x\in A\), are pairwise disjoint: a point lying in two of them would put two members of \(A\) within \(\frac {2}{3}\) of each other.
No countable set can be dense
Let \(E\) be dense in \(l_{\infty }\). Each ball \(B_{1/3}(x)\) is non-empty and open, so it must contain a point of \(E\), and since the balls are disjoint these points are distinct. That assigns to each of the uncountably many \(x\in A\) its own point of \(E\), so \(E\) is uncountable.
So \(l_{\infty }\) has no countable dense subset. It is worth contrasting this with \(C_0\) above: there the approximating sequences were allowed to be eventually zero, and a \(0\)–\(1\) sequence is never approximated by one of those.
Example 3.19. Prove that \(C[a,b]\) is a separable metric space where \(\hspace {0.2cm}\displaystyle { d(f,g)=\max _{a\leq t\leq b}|f(t)-g(t)|}.\)
Proof. We prove this in three steps
Step 1
The set of all polynomials is dense in \(C[a,b]\). This is a consequence of Weierstrass approximation theorem which states that for any continuous function \(f(t)\) on \([a,b]\), given \(\varepsilon >0\) there exists a polynomial \(p(t)\) such that \(\displaystyle {\max _{a\leq t\leq b}|f(t)-p(t)|<\varepsilon }.\)
Thus for any \(f\in C[a,b]\) and \(\varepsilon >0\) we can find a polynomial \(p\in C[a,b]\) such that \(d(f,p)<\dfrac {\varepsilon }{2}\).
Step 2
Given a polynomial \(p(t)\), there is a polynomial \(r(t)\) with rational coefficients such that \(\displaystyle {d(p,r)=\max _{a\leq t\leq b}|p(t)-r(t)|<\frac {\varepsilon }{2}}.\hspace {0.2cm}\) Hence
\[d(f,r)\leq d(f,p)+d(p,r)<\frac {\varepsilon }{2}+\frac {\varepsilon }{2}=\varepsilon .\]
Thus the set of rational polynomials is dense in \(C[a,b]\).
Step 3
The set of all polynomials with rational coefficients is countable thus \(C[a,b]\) is separable. □
Questions on this section
Stuck on something here? Ask below and it stays attached to this topic.