MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

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

Recall that (for $1\le p<\infty$), $\ell^p = \{\{a_n\}_{n=1}^\infty:\sum\limits_{i=1}^\infty|a_i|\lt\infty\}$, with norm $||\{a_n\}||=(\sum\limits_{i=1}^\infty|a_i|^p)^{\frac{1}{p}}$.

It is well known that $(\ell^p)^*\cong\ell^q$ where $\frac{1}{q}+\frac{1}{p}=1$, and so $$(\ell^p)^{**}=(\ell^q)^*=\ell^p=(\ell^p)\oplus(0).$$

Note that for $\ell^2$ we have $(\ell^2)^*\cong{\ell^2}$, since $\ell^2$ is a Hilbert space.

For $\ell^\infty=\{\{a_n\}: \sup |a_n| \lt \infty\}$, we have $$ (\ell^\infty)^* \cong \ell^1\oplus {\rm Null}(C_0),$$ and $$ (\ell^\infty)^{**}=({\ell}^1)^*\oplus\operatorname{Null}(C_0)^*,$$ but $(\ell^1)^\ast=\ell^\infty$, hence $$ (\ell^\infty)^{**} = \ell^\infty \oplus{\rm Null} (C_0)^*. $$

Seeing this pattern, is it true that the double dual of any space $X$ can be written in the form of $X\oplus Y$ for some other space $Y$?

share|cite|improve this question

closed as too localized by Andrés E. Caicedo, Yemon Choi, Gerald Edgar, Bill Johnson, Alain Valette Sep 9 '12 at 20:49

This question is unlikely to help any future visitors; it is only relevant to a small geographic area, a specific moment in time, or an extraordinarily narrow situation that is not generally applicable to the worldwide audience of the internet. For help making this question more broadly applicable, visit the help center.If this question can be reworded to fit the rules in the help center, please edit the question.

What category does $X$ live in? Is it a Banach space, a topological vector space, or any vector space? – Alex Becker Sep 8 '12 at 23:54
If you want a topological decomposition in your final question, the answer is "no" - the double dual of $c_0$ is $\ell^\infty$, but you cannot write $\ell^\infty$ as $c_0\oplus E$ for any closed subspace $E$. – Yemon Choi Sep 9 '12 at 0:17
That said, I think this question would have been better suited to – Yemon Choi Sep 9 '12 at 0:17
@Yemon I now see why I should refresh my window more often... – Alex Becker Sep 9 '12 at 0:47
Since the question has an answer with upvotes, it will likely not be deleted, so I did one final pass at editing it to fix the most painful typos. – Andrés E. Caicedo Sep 9 '12 at 23:24
up vote 2 down vote accepted

Note that there exists such a $Y$ iff the sequence $$0\to X\overset{\varphi}{\to} (X^\*)^\*\overset{\eta}{\to} \mathrm{coker}\:\varphi\to 0$$ splits, where $\varphi$ is the canonical injection and $\eta$ the canonical projection. This happens iff there exists an injection $j:\mathrm{coker}\:\varphi\to (X^\*)^\*$ such that $\varphi(X)\oplus j(\mathrm{coker}\:\varphi)=(X^\*)^\*$, so can only happen if $\varphi(X)$ is split in $(X^\*)^\*$.

For a counterexample, consider the space $c_0$ of sequences in $\mathbb C$ which converge to $0$. The double dual of $c_0$ is $\ell^\infty$, yet $c_0$ does not split as a subspace of $\ell^\infty$. See for reference page 46 of this PDF.

share|cite|improve this answer
Thanks, would you please explain more why there exists such a Y iff the sequence 0$\to$ X$\overset{\varphi}{\to}$ $(X^*)^*$$\overset{\eta}{\to}$ $\mathrm{coker}\:\varphi\to$ 0 splits? – student Nov 12 '12 at 9:30
@student If it splits, then we have $(X^*)^*\cong X\times \mathrm{coker} \phi$. If we have $(X^*)^*\cong X\times E$ then the canonical injection must take $X$ to $X\times \{1\}$ and so the sequence splits. – Alex Becker Nov 23 '12 at 20:50
Oops, I meant $X\times \{0\}$. – Alex Becker Nov 23 '12 at 20:51

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