4
$\begingroup$

Suppose that $M$ is a von Neumann algebra with no minimal projections. Let $p$ be a nonzero projection in $M$ and $\rho$ be a normal state on $M$.

For any $\epsilon>0$, can we find a projection $e$ in $M$ such that $0\leq e\leq p$ and $\rho(e)=\epsilon \rho(p)$?

$\endgroup$
2
  • 4
    $\begingroup$ Yes, if $\epsilon\le1$, by the intermediate value theorem. The space of projections in $pMp$ is connected in the SOT (if $M$ is diffuse). $\endgroup$ Commented Sep 22, 2021 at 23:02
  • $\begingroup$ Would you like to give me some hint on how to prove that $pMp$ is connected in SOT when $M$ is diffuse? $\endgroup$
    – Yanyu
    Commented Apr 12, 2023 at 15:38

1 Answer 1

1
$\begingroup$

I think the proof goes something like this. We need to assume that $\epsilon\le 1$ as pointed out.

Let $M$ be a von Neumann algebra with no minimal projections. Let $\rho$ be a normal state on $M$.

Claim-1: Given $\epsilon>0$, there exists a non-zero projection $p\in \text{Proj}(M)$ such that $\rho(p) <\epsilon$.

Proof. Towards a contradiction, let us assume otherwise. Then, $\inf\{\rho(p) : p \in \text{Proj}(M), p\ne 0\}>0$. Let us denote the infimum by $l$. Given $\epsilon> 0$, we can find a non-zero projection $p\in M$ such that $l\le \rho(p) < l +\epsilon$. Since $M$ contains no minimal projections, we can find a non-zero projection $q<p$. Let us now observe that $l \le \rho(q)$ and $0\le\rho(p-q)<l+\epsilon-l=\epsilon.$ Now, since $p-q$ is a projection, choosing $\epsilon<l$ gives us a contradiction. Note that since $q\le p$, $pq=qp=q$.

Claim-2: The map $\rho: \text{Proj}(M)\to [0,1]$ is onto.

Proof: Let $t\in [0,1]$. We want to show that there exists $q\in \text{Proj}(M)$ such that $\rho(q)=t$. We let $T=\{p\in \text{Proj}(M): \rho(p)\le t\}$. We give it the natural order, i.e., $p\le p'\iff p'-p\ge 0$. We just need to show that there is a maximal element in the set. Let $\{p_i\}$ be an increasing net of projections (bounded above in norm by 1). Then $p_i\to p$ in the strong operator topology. Moreover, $p$ is the supremum of $\{p_i\}$ under the natural order. Since $\rho$ is a normal state, $\rho$ is strong operator continuous on the unit ball of $M$. Therefore, $\rho(p)=\lim_i\rho(p_i)\le t$. By Zorn's lemma, there exists a maximal element which we call $q$. We want to show that $\rho(q)=t$. Suppose otherwise. Let $\epsilon=t-\rho(q)$. Using claim-1, we can find $p\in \text{Proj}(M)$ such that $\rho(p)<t-\rho(q)$. This shows that $\rho(p+q)<t$. However, $p+q\ge q$, contradicting the maximality of $q$.

Now, to answer your question, for any non-zero projection $p$ and $\epsilon\le 1$, we see that $\epsilon\rho(p)\in [0,1]$ and hence, the existence of $e$ is guaranteed by Claim-1 and 2. Clearly $e\le p$. Otherwise, $\rho(e)>\rho(p)\ge \epsilon\rho(p)$ which would lead to a contradiction.

$\endgroup$
3
  • $\begingroup$ I don't see why towards the end of your proof, $p+q$ is a projection. In fact, I think you used nowhere that $\rho$ is normalized. $\endgroup$ Commented Sep 26, 2023 at 13:32
  • 1
    $\begingroup$ After some thought, I think what you would like to do is to apply claim-1 to the von Neumann algebra $(1-q)M(1-q)$, which is not zero provided that $t\leq1$. This algebra is clearly diffuse as well, hence contains a projection $p$ such that $\rho(p) < t-\rho(q)$. Then $p+q$ is also a projection and we can continue the proof as above. In particular, we need to have $t\leq1$ in order for the proof to work. $\endgroup$ Commented Sep 26, 2023 at 20:41
  • $\begingroup$ Thank you for pointing that out. $\endgroup$ Commented Sep 27, 2023 at 11:09

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .