Question: Let $X$ be a Banach space, and let $E \subset X$ be a closed subspace such that $E$ is not complemented in $X$. Does it follow that $E \oplus \{0\}$ is not complemented in $X \oplus Y$, where $Y$ is some other fixed arbitrary Banach space?
I ask this question because I want to understand the answer to this question. I would just ask the answerer for clarification, but it appears that they no longer use this site.
Discussion: Suppose one wants to prove the answer to the question is "yes". A natural approach would seem to be to prove the contrapositive. So, suppose that $F \subset X \oplus Y$ is a closed subspace complementary to $E \oplus \{0\}$. From here we want to argue that $E$ is complemented in $X$. A natural candidate for a complementary subspace to $E$ would be $\pi_1(F)$, where $\pi_1$ is the factor projection $X \oplus Y \to X$. It's simple to see that $\pi_1(F) + E = X$, that is no problem. However:
- I don't see any reason why $\pi_1(F)$ should be closed in $X$. For example, if $F \subset X \times Y$ is the graph of a closed operator $X \to Y$, then $\pi_1(F)$ is the domain of the operator which needn't be closed.
- I don't see why $\pi_1(F) \cap E = \{0\}$ should hold.
Thanks for reading.
The answer to my question is "yes". I was just looking at it from the wrong angle somehow. Bit stunned of me, in hindsight.
In fact something slightly more general is true (and basically trivial to prove).
Added: I had may as well point out that there is statement can be made about quotients.
Briefly, if $E$ is a complement for $Y$ in $Z$, then $E/X$ is a complement for $Y/X$ in $Z/X$.
Added: Note you can also slightly extend the applicability of Proposition 1 as follows:Along similar lines, if $E$ is a complement for $X$ in $Z$, then $E \cap Y$ is a complement for $X \cap Y$ in $Y$.