Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The problem is:

Let $X$ be a separable Banach space then there is an isometric embedding from $X$ to $\ell^{\infty}$.

My efforts:

I showed that there is an isometry from $X^*$ (topological dual) to $\ell^\infty$ in the following way:

Let $(e_{i})_{i=1}^{\infty}$ be a dense sequence in $B_{X}$ then define $\Phi:X^*\rightarrow\ell^\infty$ by $\Phi(f)=(f(e_{i}))_{i=1}^\infty$.

It is clear that $\Phi$ is an isometry.

An initial idea and a secondary question

Is there any canonical isometry from $X$ to $X^*$ since $X$ is separable (or not)?

share|cite|improve this question
up vote 5 down vote accepted

Let $(x_n)_{n=0}^\infty$ be a dense sequence in $X$. For each $n$, find $x_n^*\in X^*$ so that $\|x^*_n\|=1$ and $x^*_n(x_n)=\|x_n\|$. It is easy to check that the operator $T: X\to \ell^\infty$ defined by $T(x)=(x_n^*(x))_{n=0}^\infty$ is the desired embedding.

share|cite|improve this answer
Note that $|Tx|\leq|x|$ because $|x_{n}^*|=1$ for the other inequality let $x_{n_{j}}\rightarrow x$ then $|Tx_{n_{j}}|\geq |x_{n_{j}}|$ 'cause $x_{n_{j}}^*(x_{n_{j}})=|x_{n_{j}}|$ appears in the sequence $Tx_{n_{j}}$ taking the limit as $n_{j}$ goes to infinity we get the required inequality. – checkmath Feb 24 '12 at 1:48

This is almost the same as the other answers, with a bit of detail:

Let X be a separable Banach space. Let $(x_n)$ be a dense sequence in the unit sphere of $X$. For each $n$, use the Hahn-Banach Theorem to find a norm-one functional $f_n\in X^*$ with $f_n^*(x_n)=1$.

Define $\Phi: X\rightarrow\ell_\infty$ via $$ \Phi(x) =(f_n(x) ) $$

$\Phi$ is clearly linear.

Suppose $x\in X$ has norm one and let $1>\epsilon>0$. Choose $n_\epsilon$ so that $\Vert x_{n_\epsilon}-x\Vert<\epsilon$.

Then $$\epsilon>|f_{n_\epsilon}(x_{n_\epsilon}-x)|=|f_{n_\epsilon}(x_{n_\epsilon})-f_{n_\epsilon}(x)| = |1-f_{n_\epsilon}(x)|.$$

As $\epsilon$ was arbitrary, this implies that $\Vert \Phi(x)\Vert=\sup\limits_{n\in \Bbb N}|f_n(x ) |\ge 1$. Also, for any $n$, $$|f_n(x)|\le \Vert f_n \Vert \Vert x\Vert =1.$$ and so $\Vert \Phi(x)\Vert\le 1$.

Thus we have $\Vert \Phi(x)\Vert= 1$, whenever $\Vert x\Vert =1$.

From this it follows that for any non-zero element $x$ of $X$, we have
$$\Vert \Phi(x)\Vert= \Vert x\Vert \sup\limits_{n\in \Bbb N}|f_n(x/\Vert x\Vert) | =\Vert x\Vert;$$ and thus, $\Phi$ is an isometry.

share|cite|improve this answer
Hi @David Mitra, I was looking your answer, and I don't understand two steps, I will be grateful if you can help me. First, how can we ensure that $f_{n}(x) \in l_{\infty}$? and second, I can't understand why if the norm of $x_{n}-x$ is less than $\epsilon$ why this implie that $f_{n}(x_{n}-x)$ is less than $\epsilon$? Thank You so much – user 123456 Jun 8 '13 at 18:46
@Charles 1) For every $n$, $|f_n(x)|\le \Vert f_n\Vert\Vert x\Vert\le 1\cdot\Vert x\Vert$. 2) Same argument as in $1)$ . – David Mitra Jun 8 '13 at 19:11
thank you so much @David Mitra! – user 123456 Jun 8 '13 at 21:25

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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