Let $\mathcal M$ be a semi finite von Neumann algebra with a normal faithful semi finite trace $\tau$. Let $(e_i)_{I\in I}$ be a net of projections in the von Neumann algebra which converges to an element $u$ in $w^*$ topology. Is it true that for any $x\in L_1(\mathcal m),$ $\|xe_i-xu\|_1\to 0$? If you have any reference please point that to me.
2 Answers
No, this is not true. Even when $M$ is abelian and $\tau$ is a normal faithful tracial state, this need not be true. An easy counterexample can be constructed as follows. Denote by $\mu_0$ the probability measure on $\{0,1\}$ defined by $\mu_0(0) = \mu_0(1) = 1/2$. Then define $(X,\mu) = (X_0,\mu_0)^\mathbb{N}$. Define the projections $e_n \in L^\infty(X,\mu)$ by $e_n(x) = x_n$. Then, $e_n \to 1/2$ in the weak$^*$ topology, but $\|e_n - 1/2\|_1 = 1/2$ for all $n$.
$L^1(\mathcal M)$ is the predual of $\mathcal M$. So the weak*-topology on your von Neumann algebra $\mathcal M$ is the same as the weak topology induced by the pairing $L^1(\mathcal M)\times \mathcal M\to \mathbb C$. So your result holds true. No assumption are needed on the von Neumann algebra (e.g. doesn't need to have a trace), and no assumptions are needed on the elements $e_i$ (they don't need to be projections).
Edit: my answer is wrong. I had misread the question -- see Yemon Choi's comment.
-
$\begingroup$ Note that by duality $\tau(xe_i)\to \tau(xu)$ for all $x\in L_1(\mathcal M).$ But what I need is different from that! $\endgroup$ Commented May 10 at 13:04
-
3$\begingroup$ Andre, I think that your final claim can't be right, because if I take S to be the forward shift on l^2(N) then I believe S^n \to 0 ultraweakly, but since S is an isometry, the trace norm of S^nT equals the trace norm of T for every trace-class operator T. $\endgroup$ Commented May 10 at 13:08