7

I have sometimes heard that the Borel functional calculus maps bounded pointwise convergent sequences of Borel functions to strongly convergent sequences of operators. I gather "sequence" is important here, due to the measure theory aspect, we can't use nets. But why must it converge strongly? I am only familiar with the result of weak convergence.

Martin
  • 8,942
Jeff
  • 7,557
  • 1
    I wonder how do we get the weak convergent at first? how do we know that $\int f_n d\mu_{x,y}\rightarrow \int_f d\mu_{x,y}$ if we just have pointwise convergence? I guess we should use the dominant converge theorem, but do we know that $f_n$ is uniformly bounded if they only converges to $f$ pointwise? – Yanyu Apr 22 '22 at 09:19

1 Answers1

9

$\def\norm#1{\left\|#1\right\|}\def\skp#1{\left<#1\right>}$Suppose $T \in L(X)$ is a normal operator on the Hilbert space $X$ and $f_n$, $f \colon \mathbb C \to \mathbb C$ are bounded and measurable such that $\sup_n \|f_n\|_\infty<\infty$ and $f_n \to f$ pointwise. As you remark, we have $f_n(T)x \rightharpoonup f(T)x$ weakly for each $x \in H$. We moreover have \begin{align*} \norm{f_n(T)x}^2 &= \skp{f_n(T)x, f_n(T)x}\\ &= \skp{f_n(T)^*f_n(T)x, x}\\ &= \skp{(\bar f_n f_n)(T)x, x}\\ &\to \skp{(\bar f f)(T)x,x}\\ &= \norm{f(T)x}^2 \end{align*} Hence, as for Hilbert spaces $x_n \rightharpoonup x$ weakly plus $\|x_n\| \to \norm x$ implies $x_n \to x$, we have $f_n(T) \to f(T)$ strongly, as wished.

Teun
  • 489
martini
  • 86,011
  • 1
    I wonder how do we get the limitation in the fourth line? I guess we should use the dominant converge theorem, but do we know that $f_n$ is uniformly bounded if they only converges to $f$ pointwise? – Yanyu Apr 22 '22 at 09:19
  • I think Yanyu is right, there is a missing assumption: f_n should also be uniformly bounded. A counterexample can be found using Tx(t)=tx(t) for x in the Hilbert space L^2([0,1]). Then for the constant function 1 in L^2([0,1]) we have f_n(T)1=f_n. One takes f_n pointwise converging to 0 but with diverging L^2-norm. – Teun Jun 15 '23 at 00:48
  • Edited. By the way, an alternative reference for the strong convergence statement is Reed and Simon I (Functional analysis): Theorem VIII.5(d) on page 262. – Teun Jul 03 '23 at 03:21