Let $M$ be a von Neumann algebra. If it has a semifinite faithful normal trace $\tau$, then we have $\tau(p\vee q)\le \tau(p)+\tau(q)$. However, for the weight (even a faithful normal state) $\omega$ case, can we compare $\omega(p\vee q)$ and $\omega(p+q ) $? Say, if $p_n\downarrow 0$ and $q_n\downarrow 0$, then can we find some $p_m$ and $q_l$ sucht that $\omega(p_m \vee q_l)$ is very small?
1 Answer
Let $M = M_2$ be the $2\times 2$ complex matrices, acting on $\mathbb{C}^2$. Let $p$ be the orthogonal projection onto ${\rm span}(e_1)$ and for each $n \in \mathbb{N}$ let $q_n$ be the orthogonal projection onto ${\rm span}(e_1 + \frac{1}{n}e_2)$. Both $p$ and $q_n$ are rank one projections, but their join is the identity operator on $\mathbb{C}^2$. Now define $\phi_n(A) = \frac{1}{2^n}\langle Ae_1,e_1\rangle + \frac{2^n-1}{2^n}\langle Ae_2, e_2\rangle$; this is a faithful state with $\phi_n(p) = \frac{1}{2^n}$, $\phi(p \vee q_n) = \phi(I) = 1$ for all $n$, and $\phi(q_n) \to 0$.
For a counterexample to the second question, Let $P_n = 0 \oplus \cdots \oplus 0 \oplus p \oplus p \oplus \cdots \in M_2 \oplus M_2 \oplus \cdots$, with $n$ initial $0$'s, acting on $\mathbb{C}^2 \oplus \mathbb{C}^2 \oplus \cdots$. Let $Q_n = 0 \oplus \cdots \oplus 0 \oplus q_{n+1} \oplus q_{n+2} \oplus \cdots$ and let $\omega$ be the weight $\omega = \phi_1 \oplus \phi_2 \oplus \cdots$. Here $\omega(P_n),\omega(Q_n) \to 0$, and $\omega(P \vee Q_n) = \infty$ for all $n$.
-
1$\begingroup$ Thank you very much for your answer. However, the state $\phi$ is not faithful. What about the case when it is required to be faithful? $\endgroup$ Commented Nov 4, 2023 at 2:56
-
$\begingroup$ Oh, I missed that. Let me think about it ... $\endgroup$ Commented Nov 4, 2023 at 3:29
-
-
$\begingroup$ Thank you. I think it is OK now. I was thinking about how to controll the SIZE of projections in type III factor but it seems to be difficut. $\endgroup$ Commented Nov 4, 2023 at 3:54
-
$\begingroup$ Hi again. You didn't define \phi above. We only have \phi_n(q_n)\to 0, right? $\endgroup$ Commented Nov 5, 2023 at 14:18