0
$\begingroup$

Let $A_n$ and $A$ be bounded self-adjoint operators in a Hilbert space, such that $A_n\to A$ strongly. Then it is well known that $(z-A_n)^{-1}\to(z-A)^{-1}$ strongly for each $z\in\mathbb{C}\setminus\mathbb{R}$, and even strong convergence of $f(A_n)$ to $f(A)$ for nice enough functions $f$.

My question is a bit similar to this, but involves series, instead of sequences. Let $T_n$ and $T$ be bounded self-adjoint operators, such that $$ T_1+\ldots+T_n\to T \qquad \textrm{strongly}. $$ Is it true that $$ \frac{T_1}{z-T_1} + \ldots + \frac{T_n}{z-T_n} , $$ converges strongly to some self-adjoint operator? Do we have convergence of $$ f(T_1)T_1+\ldots+f(T_n)T_n , $$ for nice enough functions? A particular case of interest is $f$ being the indicator function of the positive half line. If it helps, $T$ can be assumed to be invertible.

Regarding the aforementioned expectations:

  • Are they hopelessly false?
  • If it is salvageable, what kind of extra conditions does one need?
  • What is the correct keyword to look up?
$\endgroup$
0

1 Answer 1

2
$\begingroup$

Hopelessly false. Consider the one-dimensional case (so the $T_j$ are just numbers). Any function $f$ for which $\sum_j T_j$ convergent implies $\sum_j f(T_j)$ convergent is linear in a neighbourhood of $0$.

This was proved by G. Waldenberg, American Mathematical Monthly 95 (1988) 542-544. Y. Benjamini's solution to Problem E3404, American Mathematical Monthly 99 (1992) 466-467 contains an extension:

Let $f: X \to Y$ be a mapping of normed spaces such that $\sum_{n=1}^\infty f(a_n)$ converges whenever $\sum_{n=1}^\infty a_n$ converges (both in the norm topology). Then there is a neighbourhood of 0 on which $f$ is equal to a bounded linear operator.

I have a slight additional extension (unpublished):

Let $f: X \to Y$ be a mapping of Banach spaces such that $\sum_{n=1}^\infty f(a_n)$ converges weakly whenever $\sum_{n=1}^\infty a_n$ converges (strongly). Then there is a neighbourhood of 0 on which $f$ is equal to a bounded linear operator.

$\endgroup$
2
  • $\begingroup$ Thank you for the answer! So $f(x)=\max\{0,x\}$ seems to be also OK. My main interest is in this function. I am still hoping.. $\endgroup$
    – timur
    Aug 20, 2014 at 3:48
  • 1
    $\begingroup$ No, it is not OK. A neighbourhood of $0$ includes both sides. $\endgroup$ Aug 20, 2014 at 3:53

Your Answer

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

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