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

Let $M\subset B(\mathcal H)$ be a finite von Neumann algebra of bounded operators on a Hilbert space $\mathcal H$., let $P\in M$ be a self-adjoint operator with a pure-point spectrum (for example a projection), and let $T\in M$ be another self-adjoint operator.

The question will be formulated in the case of $P$ being a projection. Let $\mathcal H_0$ and $\mathcal H_1$ be the eigenspaces of $P$.

Question: Is there a well-studied condition on $P$ and $T$ which would imply that there exists $\epsilon$ such that whenever $U\subset \mathcal H$ is a "generalized eigenspace" of $T$ (i.e. the image of a spectral projection of $T$) then there exists $u\in U$ of norm $1$ such that $u=h_1+h_2$, $h_1\in \mathcal H_1$, $h_2\in \mathcal H_2$, and the norm of $h_2$ is at least $\epsilon$?

In particular (EDIT: the previous version had a weeker ergodicity assumption, in which case the answer is negative, by the answer of Steven Deprez below)

Question: If a discrete group $\Gamma$ acts freely and in a measure preserving way on a probablity measure space $X$, each element of $\Gamma$ acts ergodically, $M=L^\infty(X) \rtimes \Gamma$, $P$ is the characteristic function of some subset of $X$ of positive measure, and $T = \sum_{\gamma\in \Gamma} f_\gamma \gamma$, where $f_\gamma$ are real-valued positive functions on $X$ such that $\sum f_\gamma=1$; do $T$ and $P$ fulfill the property from the question?

share|cite|improve this question

I don't know about your first question, but i can show that your second question fails in general.

Take any free ergodic pmp action of $\Gamma$ on $(X,\mu)$, but assume that there is a group element $g\in\Gamma$ that globally preserves some non-trivial subset $Y\subset X$. Set $P=\chi_Y$ and $T=(1-\chi_Y)+\frac12\chi_Y(u_g+u_g^\ast)$. For any subset $A$ of the spectrum of $T$, denote the corresponding spectral projection by $p_{T,A}$. Whenever $1\not\in A$, then $p_{T,A}\leq P$. Hence the $h_2$ from the first question is always $0$.

Probably the most easy example of such an action is the following: Consider the finite group $\mathbb{Z}/2\times\mathbb{Z}/2$ acting on itself by translation. For example, $(1,0)$ preserves the non-trivial subset $\{(0,0),(1,0)\}$.

The infinite tensor product gives an example where $M$ is the hyperfinite type II$_1$ factor.

share|cite|improve this answer
thanks for this example; I've added additional assumption of every element acting ergodically – Łukasz Grabowski Mar 23 '12 at 17:31

A refinement of the argument from my previous answer:

For any free, pmp, ergodic action of any non-trivial countable group, $\Gamma\curvearrowright (X,\mu)$, there is a projection $P\in L^\infty(X,\mu)$ and a $T=\sum_{\gamma\in\Gamma}f_\gamma\gamma$ such that the $f_\gamma$ are positive functions adding to $1$, and such that $T$ and $P$ do not fulfill the property from the question.

We can assume that every nontrivial element of $\Gamma$ acts ergodically on $X$. In particular, $\Gamma$ does not have any finite subgroups and sice the action is free, $X$ is non-atomic. Let $e\not=g\in \Gamma$ be any element. Take any non-trivial subset $U_0\subset X$. Since $gU_0\not=U_0$, we see that $U_1=U_0\setminus gU_0$ is a non-null set. Now we know that $U_1\cap g^{-1}U_1=\emptyset$. Choose a subset $U\subset U_1$ with measure $\mu(U)<\frac14$. This set still satisfies $U\cap g^{-1}U=\emptyset$. Denote $V=X\setminus (U\cup g^{-1}U)$ and $W=V\cap gV$. Observe that $\mu(W)\geq 1-4\mu(U)>0$.

Now, set $T=\chi_V + \chi_U u_g + \chi_{g^{-1}}u_g^\ast$ and $P=1-\chi_W$. Observe that $T\chi_W=\chi_W$. It follows that every vector $\xi\in Ran(\chi_W)$ is an eigenvector of $T$ with eigenvalue $1$. For any subset $A$ of the spectrum of $T$, denote the corresponding spectral projection by $p_{T,A}$. It follows that $p_{T,A}\leq 1-\chi_W=P$ for every subset $1\not\in A\subset \sigma(T)$ of the spectrum, not containing $1$. This contradicts the property from the question.

share|cite|improve this answer

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.