1
$\begingroup$

Given:

$X$ - any Banach space
$F : X \to X$ (linear bounded and non-invertible)
$P_n$, which is projector that strongly converges to the identity operator $I$ as $n \to\infty$

Can you help me come up with a linear bounded and non-invertible operator $F$ such that the conditions are satisfied:

$P_nFP_n$ is invertible on $\operatorname{Im}P_n$
$\|(P_nFP_n)^{-1} \| < C$ (some constant)
$P_n \to I$ strongly

If you can help at least with idea it also will be great. Thank you very much! If you think that such example doesn't exist: please can you prove it or help with idea of proving it.

I've tried a lot of things. I'm already desperate, I really hope for your help, thanks.

$\endgroup$

1 Answer 1

5
$\begingroup$

(In the following I assume that the word "invertible" in the question means "bijective".)

Your assumptions do not imply that $F$ is bijective (however, they imply that $F$ is injective and has closed range, see the partial result below).

Counterexample. Let $X = \ell^1$, let $P_n$ be the projection onto the first $n$ components and let $F: \ell^1 \to \ell^1$ be given by $$ F(x_1,x_2,x_3,\dots) = (\sum_{k=1}^\infty x_k, x_1, x_2, \dots) . $$ An explicit computation shows that all your assumptions are satisfied. However, the range of $F$ has co-dimension $1$, so $F$ is not surjective.

Partial results

(1) Your assumptions imply that $F$ is bounded below (and thus is injective and has closed range).

(2) If, in addition to your assumptions, the dual projections $P_n^*: X^* \to X^*$ also converge strongly to the identity, then $F$ is bijective.

Proof. (1) For each $x \in X$ and each $n$ one has $$ \|P_n x\| = \|(P_nFP_n)^{-1}(P_nFP_n)x \| \le C \|P_n F P_n x\|. $$ By letting $n$ tend to $\infty$ we get $\|x\| \le C \|Fx\|$.

(2) By (1) it suffices to show that $F$ has dense range. To this end it suffices to show that the dual operator $F^*$ is injective.

Define the operators $G_n := (P_n F P_n)^{-1} P_n: X \to X$. Since $P_n F G_n = P_n$, it follows that $G_n^* F^* P_n^* = P_n^*$, which converges strongly to the identity on $X^*$.

Now if $x^* \in X^*$ satisfies $F^* x^* = 0$, then $F^* P_n^* x^*$ converges strongly to $0$ and since the sequence $(G_n^*)$ is bounded, it follows that $G_n^* F^* P_n^* x^*$ converges strongly to $0$, too. But from the previous paragraph we know that the same sequence is convergent to $x^*$, so $x^* = 0$. $\square$

$\endgroup$
20
  • $\begingroup$ But $F$ is invertible. For example: because $Ker(F) = \{0\}$. Or I am wrong? $\endgroup$ Commented Nov 28, 2023 at 6:47
  • $\begingroup$ @TorteDeline: If you don't use "invertible" as a synonym for "bijective", could you please specify what definition of the notion you use? $\endgroup$ Commented Nov 28, 2023 at 7:10
  • $\begingroup$ This link. But I understand what you mean: "The inverse of $f$ exists if and only if $f$ is bijective". Yeah, hope so. But it the same moment: if $Ker(F)=\{0\}$ it means that $F$ is invertible. I don't know how to link this two points in this case $\endgroup$ Commented Nov 28, 2023 at 7:21
  • $\begingroup$ Or maybe Im wrong and this $F$ has another kernel $\endgroup$ Commented Nov 28, 2023 at 7:21
  • $\begingroup$ Addition: the theorem I’m talking about only works when $F$ is linear, but we are considering such a case, so it still works $\endgroup$ Commented Nov 28, 2023 at 7:26

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .