Why is the product of weakly measurable operator valued functions also weakly measurable?

66 Views Asked by At

I've been working through parts of Schrödinger Operators by Cycon, Froese, Kirsch, and Simon, and I'm a bit stuck on one of the results they use without proving. Let $(\Omega,\mathcal{M})$ be a measurable space, and $\mathcal{H}$ some Hilbert space. We say an operator valued map $A:\Omega \rightarrow \mathcal{B}(\mathcal{H})$ is weakly measurable if for any $\varphi, \psi \in \mathcal{H}$, the function $\omega \mapsto \langle \varphi, A(\omega)\psi\rangle$ is measurable in the usual sense, i.e. pre-images of open sets in $\mathbb{C}$ are in $\mathcal{M}$.

The text says it's clear that the product of weakly measurable functions is measurable, i.e. if $A,B: \Omega \rightarrow \mathcal{B}(\mathcal{H})$ are both weakly measurable, then operator valued function $\omega \mapsto A(\omega)B(\omega)$ should also be weakly measurable. The only thing I can think of is to consider pre-images, working with open balls for simplicity, so my current plan of attack is to show: $$\forall \psi,\varphi \in \mathcal{H}, z \in \mathbb{C}, r >0 \qquad\{\omega \in \Omega\,:\, \lvert \langle \varphi, A(\omega)B(\omega)\psi \rangle - z \rvert < r\} \in \mathcal{M}$$

While this is more or less exactly what it would mean for the product to be measurable, I'm not sure such a direct approach will be fruitful, and the way the book puts it suggests to me that there's some obvious trick I'm missing that makes this a short proof.

(Edited to make notation consistent.)

1

There are 1 best solutions below

0
On BEST ANSWER

Most likely these authors only consider separable Hilbert spaces $\mathcal{H}$. In this case pick an orthonormal basis $(e_n)_{n=0}^{\infty}$ and write $$ \langle\varphi,A(\omega)B(\omega) \psi\rangle= \lim_{N\rightarrow \infty}\sum_{n=0}^{N}\langle \varphi,A(\omega)e_n\rangle\ \langle e_n, B(\omega)\psi\rangle\ . $$ Sums and products of measurable functions are measurable so the expression is measurable for finite $N$. Finally, since pointwise limits of sequences of measurable functions are measurable, this takes care of the $N\rightarrow\infty$ limit.