6
$\begingroup$

Let $\rho,\sigma\in\text{L}(\mathcal{H}_{XAB})$ be given by $$ \rho = \sum_x |x\rangle\langle x|\otimes p_x\rho_x, \quad \sigma = \sum_x |x\rangle\langle x|\otimes q_x\sigma_x, $$ and consider operators $M$ be given by $$ M = \sum_x |x\rangle\langle x|\otimes M_x, \quad\quad M_x\geq 0, \quad \sum_x M_x = id, $$ that is $\{M_x\}$ is a POVM measurement. Now suppose $$ \lVert M(\rho-\sigma)\rVert_1 = \sum_x \left|\operatorname{Tr}M_x(p_x\rho_x - q_x\sigma_x)\right| = 0 $$ for all measurement operators $M$ given as above, where $\{M_x\}$ is implementable by LOCC. Do we necessarily have $\rho = \sigma$?

Using $M_x = id/|X|$ yields $p_x = q_x$, and I have shown it is true for $\rho,\sigma$ pure using Schmidt decomposition, so I do believe it should be possible to prove $\rho_x = \sigma_x$ for all $x$, but I have not managed to do so.

Any help is appreciated!

$\endgroup$
1
  • $\begingroup$ isn't Bennett et al. (1999) (10.1103/PhysRevA.59.1070) a counterexample of this (referring to the question in the title)? $\endgroup$
    – glS
    Jul 11, 2021 at 10:41

1 Answer 1

1
$\begingroup$

We can take, for example, $ M = |0 \rangle \langle 0| \otimes I $, right? But then:

$$ \Big|\Big| M(\rho - \sigma) \Big|\Big|_1 = \Big|\Big| |0 \rangle \langle 0| \otimes (p_0 \rho_0 - q_0 \sigma_0) \Big|\Big|_1 = \Big|\Big| p_0 \rho_0 - q_0 \sigma_0 \Big|\Big|_1 = 0 \implies p_0 \rho_0 = q_0 \sigma_0 $$

Similary, taking $ M = |x \rangle \langle x| \otimes I $, we get $ p_x \rho_x = q_x \sigma_x, \forall x $. We conclude that $ \rho = \sigma $.

$\endgroup$
5
  • $\begingroup$ Yeah, your are right. However, I just changed my question back to the original, which is slightly different. Sorry about the inconvenience. $\endgroup$
    – user114158
    Oct 13, 2020 at 14:18
  • $\begingroup$ Sorry for that, but could you explain? Which part of the question has changed and the answer is not valid? Also i think you can even take $ M_x = id / |X| $ and conclude since $ ||M(\rho - \sigma)||_1 = \sum_x \text{Tr}\big[ |M_x(p_x \rho_x - q_x \sigma_x)| \big] $, meaning $ | . | $ should be inside the trace. $\endgroup$
    – tsgeorgios
    Oct 13, 2020 at 14:45
  • $\begingroup$ In the sum, I have absolute value of the trace, not trace of the absolute value. I guess, I have used a misleading notation for the norm, but take the assumption as the sum over absolute values of traces being equal to zero. $\endgroup$
    – user114158
    Oct 13, 2020 at 14:50
  • $\begingroup$ Oh, i see now. So just to be clear you are assuming basically that $ \text{Tr} [M_x (\rho_x - \sigma_x) ] = 0 $ for every $ M_x \geq 0 $ and $ \sum_x M_x = I $. right? $\endgroup$
    – tsgeorgios
    Oct 13, 2020 at 15:15
  • $\begingroup$ Yes, that is indeed what I mean! $\endgroup$
    – user114158
    Oct 13, 2020 at 15:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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