3
$\begingroup$

Let $X$ be a reflexive Banach space and $T:X\to X$ is a linear operator. Is it true that $$ X = \mathrm{ker}(I-T) \oplus \overline{\mathrm{range}(I-T^\ast)}, $$ where $\oplus$ is the direct sum?

I know this is true when $X$ is a Hilbert space (due to the fact that $Tx=x$ implies $T^\ast x =x$), and I would suspect something like that is true when $T$ is a compact operator. How about the general case?

$\endgroup$
  • $\begingroup$ This isn't even true in finite dimensional spaces unless you take some further assumptions (such as $T = T^*$) $\endgroup$ – Omnomnomnom Dec 20 '18 at 0:01
  • $\begingroup$ @Omnomnomnom What if range(I-T) is changed to range(I-T*)? $\endgroup$ – user58955 Dec 20 '18 at 0:05
  • $\begingroup$ Then that's at least correct in the finite dimensional case $\endgroup$ – Omnomnomnom Dec 20 '18 at 1:28
  • $\begingroup$ What does it mean to add a subspace of $X$ and a subspace of $X^{*}$? $\endgroup$ – Kavi Rama Murthy Dec 20 '18 at 5:58
0
$\begingroup$

This fails even in finite dimension. Let $X=\mathbb C^2$, and $$ T=\begin{bmatrix} 1&1\\0&1\end{bmatrix}. $$ Then $(I-T)^2=0$, so $$\overline{\operatorname{ran}(I-T)}\subset\ker(I-T).$$

When $X$ is a Hilbert space, it is true that $$\tag1 X=\ker T\oplus \overline{\operatorname{ran} T^*} $$ for any bounded linear operator $T$, since $\overline{\operatorname{ran} T^*}=(\ker T)^\perp$.

For a general Banach space, the equality $(1)$ makes no sense, as $\ker T\subset X$ and $\operatorname{ran} T^*\subset X^*$.

More dramatically, it is known that any Banach space that is not isomorphic to a Hilbert space has a non-complemented subspace $M$. And if $X$ is separable, it is known that there exists a bounded linear $T:X\to X$ with $\ker T=M$. This impedes $(1)$ in general, and there is nothing you can put on the second summand that will make it work.

$\endgroup$
  • $\begingroup$ For a general Banach space, will it make sense in the sense of isomorphism? $\endgroup$ – user58955 Dec 20 '18 at 2:00
  • $\begingroup$ No. I have edited the answer. $\endgroup$ – Martin Argerami Dec 20 '18 at 2:32

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.