Fix any $1<p\leq 2$. Let us recall that $E:=(\oplus\ell_2^n)_{\ell_1}$ is just the space of sequences $(x_n)_{n=1}^\infty$, $x_n\in\ell_2^n$, such that $(\|x_n\|_{\ell_2^n})_{n=1}^\infty\in \ell_1$, endowed with the norm $\|(x_n)_{n=1}^\infty\|_E=\sum_{n=1}^\infty\|x_n\|_{\ell_2^n}$. Also, $\ell_1\oplus_\infty\ell_p$ is just the space of ordered pairs $(x,y)$, $x\in\ell_1$, $y\in\ell_p$, endowed with the norm $\|x\oplus y\|_\infty=\max\{\|x\|_{\ell_1},\|y\|_{\ell_p}\}$.

Question. Is the space $E=(\oplus\ell_2^n)_{\ell_1}$ complemented in $\ell_1\oplus_\infty\ell_p$? In other words, does there exists a continuous linear projection $P\in\mathcal{L}(\ell_1\oplus_\infty\ell_p)$ onto a subspace isomorphic to $E$?

Recall that $\ell_2^n$'s are uniformly complemented in $\ell_p^{k_n}$ for some increasing sequence of positive integers $k_1<k_2<k_3<\cdots$, and uniformly embedded (but not complementably) into $\ell_1^{k_n}$. (See Theorem 4.10 here for the first fact. The second fact is just an application of Dvoretsky's Theorem.) Perhaps these two facts might help to construct a projection.


  • $\begingroup$ I think this should be on Math Overflow. $\endgroup$ – Kugelblitz Nov 9 '15 at 14:17
  • $\begingroup$ Is it clear that $E$ is isomorphic to a subspace of $\ell_1\oplus_{\infty}\ell_p$? $\endgroup$ – uniquesolution Nov 9 '15 at 14:27
  • $\begingroup$ If the second fact I mentioned is true, then yes. In other words, if $\ell_2^n$ embeds uniformly into $\ell_p^{k_n}$ then $E$ is indeed a subspace of $\ell_1\oplus\ell_p$. But as I said, this fact should be checked. (I will try to do so now...) EDIT: Yes, it's true. $\endgroup$ – Ben W Nov 9 '15 at 14:32
  • $\begingroup$ $l_2^n$ is just $\mathbb{C}^n$ space with standart norm? $\endgroup$ – kp9r4d Nov 9 '15 at 15:14
  • $\begingroup$ Yes. More generally, $\ell_p^n=L_p(n)=(\mathbb{K}^n,\|\cdot\|_p)$, where $\mathbb{K}=\mathbb{R}$ or $\mathbb{C}$. $\endgroup$ – Ben W Nov 9 '15 at 15:28

The answer is no. Since $\ell_1$ and $\ell_p$ ($p\in (1,\infty)$) are totally incomparable (every operator from one space to the other is strictly singular), by a result of Edelstein and Wojtaszczyk (Theorem 3.5 here) infinite-dimensional, complemented suspaces of $\ell_1\oplus \ell_p$ are the obvious ones, that is, they are isomorphic either to $\ell_1, \ell_p$ or $\ell_1\oplus \ell_p$. Neither of those spaces is isomorphic to $(\oplus_n \ell_2^n )_{\ell_1}$.

The moral reason of why $(\oplus_n \ell_2^n )_{\ell_1}$ is not isomorphic to a complemented suspace of $\ell_1\oplus \ell_p$ is that $\ell_2^n$ are not uniformly complemented in $\ell_1$ but they are uniformly complemented in $\ell_p$, so essentially this subspace would have to live in $\ell_p$ which is impossible. This heuristics can be made precise.

  • $\begingroup$ Thanks! Now that I think about it, it should just be an elementary basic sequence argument. $\endgroup$ – Ben W Nov 9 '15 at 22:25

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.