5

Let $H$ be an infinite dimensional Hilbert space and $B(H)$ the bounded linear operators on $H$. Then thre is no ultra weakly continous non-zero positve trace $tr:B(H)\rightarrow \mathbb{C}$.

I talked to my professor about it and he showed me how he would do it. But he used without further clarifcation that $tr$ is SOT-contionous, which I don't see right now.

Zolf69
  • 763

1 Answers1

4

No continuity is needed, nor positivity (which would imply continuity, but that's beside the point). Only linearity and traciality. The argument below uses $H$ separable, but only for notational simplicity.

Let $\phi:B(H)\to\mathbb C$ be a linear and tracial functional.

We can write $I=P+Q$ for two infinite projections, both equivalent to $I$. Explicitly, for matrix units $E_{kj} $ we can take $$ P=\sum E_{2n-1,2n-1},\ \ \ Q=\sum E_{2n,2n},\ \ \ V=\sum E_{2n,2n-1}, \ \ \ W=\sum E_{2n-1,n}. $$ Then $V^*V=P$, $VV^*=Q$, $W^*W=I$, $WW^*=P$. So $$ \phi (P)=\phi (I)=\phi (P)+\phi (Q)=2\phi (P). $$ It follows that $\phi (P)=0$, and so $\phi (I)=\phi (P)=0$.

For a finite projection $P$, it is easy to construct a sequence of projections $P_n$, each equivalent to $P$ (i.e., there exist $W_n$ such that $W_n^*W_n=P$, $W_nW_n^*=P_n$) and with $\sum P_n=I$. It is also easy to show that $\sum_{n>s}P_n$ and $\sum_{n>t}P_n$ are equivalent for all $s,t$. Then, applying $\phi$ to $$ \sum_1^mP_n+\sum_{m+1}^\infty P_n=I, $$ we obtain $$ m\phi (P)+\phi (\sum_{m+1}^\infty P_n)=0. $$ As the second term is the same for all $m$, this forces $\phi (P)=0$.

So $\phi$ is zero on all projections. As every operator in $B(H)$ is a linear combination of projections, $\phi (T)=0$ for all $T\in B(H)$.


As mentioned by Ali in the comments, another argument goes by using Halmos' result that every $T\in B(H)$ is a sum of two commutators. Corollary in Page 198 of Halmos, P. R. American Journal of Mathematics Vol. 76, No. 1 (Jan., 1954), pp. 191-198 (8 pages)

Martin Argerami
  • 217,281
  • Thank you again =). Could you elaborate on why there are P,Q with I=P+Q? – Zolf69 Jul 09 '14 at 07:53
  • Please see the edit. – Martin Argerami Jul 09 '14 at 08:15
  • Thank you very much. I hope that this was my last questions concering my bachelor thesis. You have helped me a great deal, so I'd like to mention you in the preface, if that is okay. If it is, could you tell me where you work exactly, the link in your profile is not working. – Zolf69 Jul 09 '14 at 08:21
  • Yes, of course. I work at the University of Regina. And thanks because I had not noticed that the link was not working; I have replaced it with the good one. – Martin Argerami Jul 09 '14 at 12:56
  • @MartinArgerami I think there is no even an algebraic trace(without any other assumption) since the span of commutators is whole B(H). – Ali Taghavi Mar 01 '16 at 13:10
  • ...By a Paper of Halmos every operator is a sum of two commutator. – Ali Taghavi Mar 01 '16 at 13:28
  • @Ali: good point! – Martin Argerami Mar 02 '16 at 03:13
  • @MartinArgerami As a related subject: I search for a C* algebra which is either stable or properly infinite but admit an algebraic trace(Without any extra assumption on the trace). Is there an example? I explained the motivation here: http://math.stackexchange.com/questions/1678621/a-nilpotent-element-of-an-algebra-which-does-not-lie-in-the-span-of-commutator-e – Ali Taghavi Mar 02 '16 at 08:09
  • By "algebraic" you mean not even linear? – Martin Argerami Mar 02 '16 at 12:21
  • @MartinArgerami By Algebraic I mean a linear functional which is zero on commutators, but I do not assume positivity or boundedness – Ali Taghavi Mar 02 '16 at 15:51
  • 1
    I have edited the question to avoid using continuity; still, your argument is more direct. Regarding your example,it is not obvious to me, but I'm not big on examples. – Martin Argerami Mar 02 '16 at 16:59
  • 2
    As a matter of fact, there exist Banach spaces $X$ for which $B(X)$ admits a trace (or even a character). For example this is the case for the James space. – Tomasz Kania Mar 02 '16 at 16:59
  • How to check that $P,Q$ are two infinite projections? – math112358 May 22 '19 at 23:54
  • $P$ is equivalent to the identity, and it's not the identity. $Q$ is equivalent to $P$. – Martin Argerami May 23 '19 at 02:37