Sufficient Condition for stable Convergence

functional-analysisoperator-theory

Let $X$ be some Banach space, $\mathcal{L}(X)$ be the space of bounded linear operators $T:X\to X$ with $\mathrm{dom}(T)=X$.

Let $T\in \mathcal{L}(X)$ and $(T_n)$ be a sequence in $\mathcal{L}(X)$.

We define stable convergence of $T_n$ to $T$ ($T_n\overset{s}{\to}T)$ by

  1. $\forall x\in X:\ T_nx\to Tx$,
  2. $\exists M>0,N\in \mathbb{N}:\forall n>N: T_n^{-1}\in \mathcal{L}(X)$ and $\Vert T_n^{-1}\Vert \leq M$.

I am struggling now with following argument: Let $S\in \mathcal{L}(X)$ and $z\in\rho(S)=\{z\in \mathbb{C}:(S-z)^{-1}\in \mathcal{L}(X)\}$, consider a sequence $S_n\in \mathcal{L}(X)$ with $S_nx\to S x\ \forall x\in X$. The proof I am trying to understand now uses

Stability of $S_n-z$ can be written as $\Vert (S_n-z)x_n \Vert\geq M \Vert x_n\Vert$ for $x_n\in X, n\geq N$. (*)

I See that (*) is necessary for the stable convergence, but why is this condition sufficient for $S_n-z$ having an inverse in $\mathcal{L}(X)$?

For example: If I take $T:\mathcal{l}^2\to\mathcal{l}^2,\ (x_1,x_2,x_3,\dots,)\mapsto (0,x_1,x_2,\dots)$, I clearly have $\Vert Tx\Vert=\Vert x \Vert$ but $T$ is not surjective.

What am I missing?

Best Answer

It is obviously not sufficient. Consider the operators $S_n$ in $\ell^2$ defined by $$ x=(x_1,x_2,\dots)\mapsto S_nx=(x_1,\dots,x_n,0,x_{n+1},\dots) $$ Then $S_nx\to x$ for all $x$ and you have the lower bound (with $z=0$) but not invertibility for every $S_n$.

Most likely, the words "can be written as" should be read as "implies", not as "is equivalent to" or, as noted in the comments, the book uses some non-standard terminology that hasn't been thought through thoroughly enough not to lead to occasional contradictions.

Related Question