The conditions on the $p_j$ and $q_j$ that I believe hold do not in fact hold. See the edit for the linked question showing this.

Let $A$ be a finite dimensional $\mathrm{C}^*$-algebra.

Suppose that $p$ is a projection and $\nu\in \mathcal{S}(A)$ a state such that $\nu(p)=1$. Consider projections $p_j,\,q_j$ such that $p_j\leq p$ and $q_j\leq p$. Consider $$\sum_{j=1}^n q_j\otimes p_j,$$ so that the $p_j$ may be taken as linearly independent.

Suppose now that:

$$\sum_{j=1}^n\nu(q_j) p_j=p.$$

Does it follow that $$\sum_{j=1}^n q_j\otimes p_j=p\otimes p?$$

Edit: Perhaps there are more conditions on the $p_j$ and $q_j$ that come from the context.

The element $$T(q)=\sum_{j=1}^nq_j\otimes p_j$$ is the image of a projection $q$ under a *-homomorphism $T:A\rightarrow A\otimes A$. My understanding, therefore, is that it can be written as a sum of elementary tensors $q_j\otimes p_j$, with the $q_j$ and $p_j$ projections.

My understanding is that for such a general element of $A\otimes A$, we can choose the $p_j$ to be linearly independent projections. Is this incorrect? Furthermore, can we also take the $q_j$ to be linearly independent (or just the $p_j$)? Perhaps even then we can take the $q_j$ and $p_j$ to be orthogonal?

These questions are asked here.

A fuller context is here.


If you put no restrictions on the $q_j$, in particular if you allow $\nu(q_j)=0$, then the answer is obviously no because you can choose $p_1,\ldots,p_n=p$, $q_1=p$ with $\nu(q_1)=1$ and $\nu(q_2)=0$ and $q_3=\cdots=q_n=q_2$. Then your condition is satisfied and $$ \sum_jq_j\otimes p_j=p\otimes p+\left(\sum_{j\geq2}q_j\right)\otimes p. $$

If you require $\nu(q_j)>0$ for all $j$ your condition can still fail. Suppose that $p=p_1+p_2$, with $\nu(p_1)=1$. Let $q_1=q_2=p_1$. Then $$ \sum_j\nu(q_j)p_j=p_1+p_2=p, $$ while $$ \sum_j q_j\otimes p_j=p_1\otimes p_1+p_1\otimes p_2=p_1\otimes p $$ is a proper subprojection of $p\otimes p$.

  • $\begingroup$ Thank you for your answer. The first scenario doesn't apply because I want the $p_j$ to be linearly independent. I will in a moment slightly edit the question which may rule out the second scenario, I am not sure. $\endgroup$ Nov 25 '19 at 7:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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