If we have the following upper bound on the sum of trace distances:

$$ \frac{1}{N} \sum_{a, b}||p_1(a | b) \rho_{ab} - p_2(a | b) \sigma_{ab}|| \le \epsilon, $$ where $p_1$ and $p_2$ are two probability distributions and $\rho_{ab}$ and $\sigma_{ab}$ are quantum states depending on $a, b$. Then, what is the best that we could say about the following in terms of $\epsilon$:

$$ ||\rho_{ab} - \sigma_{ab}|| \le ? $$


1 Answer 1


In general I would imagine not much. For simplicity consider the case where $p_1 = p_2$. Then the worst case would be that for all $(i,j) \neq (a,b)$ we have $\rho_{ij} = \sigma_{ij}$. Thus the entire difference in the first equation is coming only from the difference between the pair $\rho_{ab}$ and $\sigma_{ab}$. In such a case you'd have $$ \|\rho_{ab} - \sigma_{ab}\| \leq \frac{N \epsilon}{p(a|b)}. $$ which could even be bigger than the trivial bound of $1$ unless $\epsilon < 1/N$.

Without more context to the problem I don't see how you'd get around this.

  • $\begingroup$ I see your point, Rammus. Thanks a lot for the clarification. $\endgroup$ Commented Mar 10, 2023 at 3:06

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.