MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Related to Jon's question, I have two questions. Let $\mathcal{A}$ be a concrete $C^*$-algebra on a Hilbert space $\mathcal{H}$. For any selfadjoint subset $S$ of $\mathbb{B}(\mathcal{H})$, let $S^m$ denote the set of elements of $\mathbb{B}(\mathcal{H})_{sa}$ that can be obtained as the strong limits of monotone increasing nets from $S$.

Question 1. Is $((\mathcal{A}_{sa})^m)^m=(\mathcal{A}_{sa})^m$? (Maybe this is very basic.)
Question 2. Does the $C^*$-algebra $C^*((\mathcal{A}_{sa})^m)$ generated by $(\mathcal{A}_{sa})^m$ in $\mathbb{B}(\mathcal{H})$ coincide with the strong closure of $\mathcal{A}$ in $\mathbb{B}(\mathcal{H})$?

For Question 2, I have been thinking that Pedersen's up-down-up theorem [Theorem 2 in American Journal of Mathematics 94 (1972), 955-962] might be useful, but I couldn't figure out.

share|cite|improve this question
up vote 8 down vote accepted

Edited. As Masayoshi points out, my reading of Hamana's paper was incorrect. I'm quite sure question 1 is false in general but I don't have a reference. (Masayoshi, did you look in Pedersen's book? I feel the answer may be there but I don't have access to it right now.)

I guess I'd better be more explicit about question 2. For example, take $A = C[0,1]$ acting by multiplication on $l^2[0,1]$. Then the strong closure of $A$ equals $l^\infty[0,1]$, but $A_{sa}^m$ is contained in e.g. the set of bounded Borel measurable functions, so the C*-algebra it generates is also contained in that class.

share|cite|improve this answer
Thank you Nik for your answer. Please allow me to ask further questions following your answer. For Question 1: Yes, I know that paper by Hamana. In fact, I previously highlighted the sentence $B_{sa}$ is itself a unique real linear subspace $S$ of $B_{sa}$ such that $A_{sa}\subset S$ and $S^m=S$'' in Introduction. I think that you are referring this part. In this statement, $S$ is assumed to be a real linear subspace of $B_{sa}$. However, in my question, $(A_{sa})^m$ is just a selfadjoint subset'' of $B(H)_{sa}$ and not a linear subspace in general. Could you please clarify this point? – Masayoshi Kaneda Aug 4 '12 at 10:45
For Question 2: If you don't mind, could you please provide me with your counterexample in the abelian case? Thank you. P.S. My Question 1 seems exactly what Jon is asking. – Masayoshi Kaneda Aug 4 '12 at 11:14
Nik, thank you again for your response. For Question 2, I see, the point is to consider $l^{\infty}[0,1]$ instead of $L^{\infty}[0,1]$. Thank you for a nice example. For Question 1, of course, I looked over Pedersen’s book, but I couldn’t find the answer. This is why I am placing the question on this board. – Masayoshi Kaneda Aug 5 '12 at 13:05
(Continuation) In Pedersen’s book (Theorem 2.4.4), Kadison’s other (but related) result “A $C^*$-subalgebra $M$ of $B(H)$ is a von Neumann algebra if and only if $(M_{sa})^m= M_{sa}$” (Lemma 1 of [Kadison, Operator algebras with a faithful weakly-closed representation, Ann. of Math. (2) 64 (1956), 175-181]) is presented. But the proof in Pedersen’s book is not Kadison’s original one, but one using his own lemma which was used to prove his “up-down” theorem. – Masayoshi Kaneda Aug 5 '12 at 13:07
(Continuation) Kadison’s result cited in Hamana’s paper can be found on pp. 316-317 of [Kadison, Unitary invariants for representations of operator algebras, Ann. of Math. (2) 66 (1957), 304-379], though it is not presented as a theorem or lemma. (Note that in this paper, $A_1^m$ is used in a different meaning from the usual one. It is the union of the monotone closures of $A_1$ taken “transfinitely recursively”. Also the monotone closures are taken in both directions, i.e., both increasing and decreasing.) – Masayoshi Kaneda Aug 5 '12 at 13:11

I can give you a partial answer to question 1, which you may already be aware of if you're familiar with Pedersen's "C*-algebras and their automorphism groups". Specifically, if $A\subseteq A^{**}$, i.e. if we are considering the universal representation of $A$, and if $A$ is either unital or separable then you do indeed have $(A_\mathrm{sa}^m)^m=A_\mathrm{sa}^m=\overline{A_\mathrm{sa}^m}(=$ norm closure of $A_\mathrm{sa}^m$). For if $A$ is unital then, by Pedersen 3.11.7, $A_\mathrm{sa}^m=\overline{A_\mathrm{sa}^m}$, while if $A$ is separable then we again have $A_\mathrm{sa}^m=\overline{A_\mathrm{sa}^m}$, by Corollary 3.25(a) of Lawrence G. Brown's "Semicontinuity and Multipliers of C*-algebras". By Pedersen 3.11.5, $\overline{A_\mathrm{sa}^m}$ consists precisely of those $x\in A^{**}_\mathrm{sa}$ that are lower semicontinuous on $Q=A^{*1}_+$, which are immediately seen to be closed under taking supremums of bounded increasing nets.

If we consider $A$ with its atomic representation instead, then the above remarks still apply, as there is then a normal morphism $\pi$ from $A^{**}$ onto $A''$ which is faithful on $A_\mathrm{sa}^m$, by Pedersen 4.3.13 and 4.3.15. Arbitrary representations can be faithful on $A$ but not on $A^m_\mathrm{sa}$, although in general you might perhaps try to use the predual $A''_*$ of $A''$, in the topology induced by $A$, as a replacement for $A^*$ with the weak* topology. The only problem is that $A''^1_{*+}$ may not be compact in this topology so you would have to somehow adjust the proof of Pedersen 3.11.2, on which 3.11.5 relies.

Incidentally, it is a problem of Akemann and Pedersen from 1973 (still open as of 2014 according to Brown) whether $A_\mathrm{sa}^m=\overline{A_\mathrm{sa}^m}$ for arbitrary $A$ in its universal representation, so if you or Nik have a counterexample it would be quite important.

share|cite|improve this answer
Nice answer. I didn't know that problem was open. – Nik Weaver Jan 3 '15 at 16:23
Thanks. I was a bit surprised that problem was still open too but that's what Brown says in a recent article posted to the ArXiv - see – Tristan Bice Jan 4 '15 at 2:00

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.