3
$\begingroup$

It is a well-known result in functional analysis that the sum $M+N$ of two subspaces of a Banach space with $M\cap N=0$ is closed if and only if the inclination $$\widehat{(M,N)} := \inf_{x\in M, \|x\|=1} d(x,N)$$ is positive, i.e. $$ M+N \text{closed} \Leftrightarrow \widehat{(M,N)}>0.$$ Typically this is quoted from

Gurariĭ, V. I.: Openings and inclinations of subspaces of a Banach space. Teor. Funkciĭ Funkcional. Anal. i Priložen. Vyp. 1 1965 194–204. 

where this is stated as a Theorem but without proof.

I am looking for a reference containing a proof of this result.

$\endgroup$
1
  • $\begingroup$ You should assume something else. Otherwise $M=N$ where $M$ is a closed subspace of $X$ is an obvious counterexample. $\endgroup$
    – fedja
    Nov 21, 2018 at 14:24

3 Answers 3

4
$\begingroup$

You can find this result in the book of T. Kato. Perturbation theory for linear operators. Springer 1980, 1995. In Theorem IV.4.2, page 219.

$\endgroup$
1
  • $\begingroup$ The result in Kato's book is a bit more general. It does not need $M\cap N=\{0\}$. $\endgroup$ Nov 23, 2018 at 8:42
2
$\begingroup$

References are boring if the proof is simple:

The open mapping theorem tells you that $L+M$ is closed if and only the canonical map $S:L\times M \to L+M$ $(x,y)\mapsto x-y$ is an isomorphism (where the product is endowed with the norm $\|(x,y)\|=\|x\|+\|y\|$). The inclination of $L\times\{0\}$ and $\{0\}\times M$ in this product is $1$ which gives you the necessity of $\widehat{(M,N)}>0$. For the other implication, positive inclination $c$ means $\|x-y\|\ge c\|x\|$ for all $(x,y)\in L\times M$ which gives you the continuity of the first component of $S^{-1}$, and this also implies the continuity of the second.

$\endgroup$
1
$\begingroup$

I think that in the meantime, I found the paper where the Theorem originally appeared. It seems to be

Grinblyum, M. M.: On the representation of a space of type B in the form of a direct sum of subspaces. Doklady Akad. Nauk SSSR 70, (1950). 749–752.

In this article the mentioned theorem is Theorem 1 on p. 749.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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