2
$\begingroup$

I'm trying to solve problem 11.3 in Nielsen Chuang:

(3) Prove the conditional version of the triangle inequality: $$ S(A,B|C)\geq S(A|C)-S(B|C) $$

But the inequality seems incorrect. For example, let $$ |\psi\rangle = \frac{|0\rangle|0\rangle|00\rangle+|0\rangle|1\rangle|01\rangle+|1\rangle|0\rangle|10\rangle+|1\rangle|1\rangle|11\rangle}{2},\\ \rho^{ABC} = |\psi\rangle\langle\psi|. $$ Then: $$ \rho^{C}=\frac{I}{4},\\ \rho^{AC}=\rho^{BC}=\frac{I}{2},\\ S(A,B|C)=S(A,B,C)-S(C)=0-2,\\ S(A|C)=S(B|C)=1-2=-1,\\ S(A,B|C)=-2<0=S(A|C)-S(B|C). $$

Am I doing something wrong, or did the problem mean to say $S(A,B)\geq S(A|C)-S(B|C)$ or something?

$\endgroup$
2
$\begingroup$

You've computed $\rho^{AC},\rho^{BC}$ incorrectly $-$ they are matrices of size $8\times 8$, also they are not equal to $I/8$. Nevertheless their entropy is equal, i.e. $S(A,C) = S(B,C)$. So $|\psi \rangle$ is indeed a counterexample to the statement.

The inequality $S(A,B)\geq S(A|C)-S(B|C)$, which is equivalent to $$ S(A,B) + S(B,C) \geq S(A,C) $$ is probably what they had in mind.

$\endgroup$

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.