Given a qubit state $|\psi\rangle \in \mathcal{H}$, and two bipartite general mixed states $\rho$ and $\sigma$, such that, $$\langle \psi|\otimes \langle \psi|\rho - \sigma |\psi\rangle \otimes |\psi \rangle \ \leqslant \epsilon$$ Now suppose the reduced state of $\rho, \sigma$ be such that, $$ \rho_r = Tr_1(\rho) = Tr_2(\rho), \hspace{5mm} \sigma_r = Tr_1(\sigma) = Tr_2(\sigma)$$ Then can we say something about the closeness of the reduced state in terms of epsilon? In other words, $$\langle \psi| \rho_r - \sigma_r|\psi\rangle \leqslant ? $$

  • 1
    $\begingroup$ Can you explain (otherwise edit) the notation $|\psi\rangle \in SU(2)$ since it is not commonplace? $\endgroup$ Jul 13, 2020 at 9:07
  • $\begingroup$ I think you could use a distance measure to compare how close the states are considering distance measures are used to check for distinguishability. In Nielsen & Chuang, there's some discussion about how the distance measures decrease when a system is partially traced over i.e. the distinguishability further decreases or stays the same, $\endgroup$ Jul 13, 2020 at 19:50
  • $\begingroup$ Hi Purva, the closeness notion that I was looking for with respect to a fixed state $|\psi\rangle$. This is a weak notion of distinguishability. And as DaftWullie wrote in the answer, it is not possible to give non-trivial measure of the closeness of the reduced state with respect to the state $|\psi\rangle$. If, alternatively, I had the notion of distinguishability over trace 1 norm (which is the maximum overlap of $\rho - \sigma$ with $|\psi\rangle$, where the choice of state $|\psi\rangle$ is over entire bloch sphere, then I could be able to make statements about the reduced density matrix. $\endgroup$ Jul 14, 2020 at 4:38

1 Answer 1


No, there's not a lot you can say. Consider these two cases, both with $\epsilon=0$.

First, the obvious one, $\rho=\sigma=|\psi\rangle\langle\psi|\otimes |\psi\rangle\langle\psi|$. Clearly $\rho_r-\sigma_r=0$.

Second, let $|\psi^\perp\rangle$ be orthogonal to $|\psi\rangle$. You can have $$\rho=(|\psi\rangle\langle\psi|\otimes |\psi^\perp\rangle\langle\psi^\perp|+|\psi^\perp\rangle\langle\psi^\perp|\otimes |\psi\rangle\langle\psi|)/2$$ and $$\sigma=|\psi^\perp\rangle\langle\psi^\perp|\otimes |\psi^\perp\rangle\langle\psi^\perp|.$$ Now you have $$ \langle\psi|\rho_r-\sigma_r|\psi\rangle=\frac12, $$ which is more or less as far away as you can get.

  • $\begingroup$ Hi DaftWullie, in your proof, you have not considered the property that the reduced state of $\rho$ (or $\sigma$) obtained by tracing either the first subsystem or the second subsystem is the same. $\endgroup$ Jul 13, 2020 at 7:07
  • $\begingroup$ Ah, sorry. Missed that aspect. $\endgroup$
    – DaftWullie
    Jul 13, 2020 at 7:13
  • 1
    $\begingroup$ Does that fix it? $\endgroup$
    – DaftWullie
    Jul 13, 2020 at 7:17
  • $\begingroup$ Thank you DaftWullie. This indeed shows that if some more information is known about $\rho$ and $\sigma$, then one can hope for something better. $\endgroup$ Jul 13, 2020 at 7:27

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.