Lockability and non-lockability are explained in this paper. A real valued function of a quantum state is called non-lockable if its value does not change by too much after discarding a subsystem. The max-entropy of a quantum state is defined as

$$H_{\max }(A)_{\rho}= \log \operatorname{tr}(\rho_{A}^{1 / 2})$$

For a bipartite quantum state $\rho_{AB}$, I would like to know if the max-entropy is non-lockable i.e. is there any relationship of the form

$$\text{tr}(\rho_{AB}^{1/2}) \leq \text{tr}(\rho_{A}^{1/2})\cdot|B|$$

which, after taking logs on both sides, would yield

$$H_{\max}(AB)\leq H_{\max}(A) + \log|B|$$

A couple of easy numerical examples suggests this might be true but I have not been able to prove it.


1 Answer 1


Let $D_{\alpha}(\rho\|\sigma):= \frac{1}{\alpha - 1} \log \mathrm{Tr}[\rho^\alpha \sigma^{1-\alpha}]$ be the Petz-Rényi divergence for $\alpha \in (0,1)\cup(1,\infty)$. Note that for $\alpha \in (0,1)\cup(1,2]$ this quantity satisfies the data processing inequality $$ D_{\alpha}(\rho\|\sigma) \geq D_{\alpha}(\mathcal{E}(\rho) \| \mathcal{E}(\sigma)), $$ where $\mathcal{E}$ is any CPTP map.

Now let $\rho_{AB}$ be any state and take $\mathcal{E}$ to be the partial trace over the $B$ system. Then for $\alpha = 1/2$ we have $$ \begin{aligned} -2 \log \mathrm{Tr}[\rho^{1/2}_{AB}] &= D_{1/2}(\rho_{AB}\|I_{AB}) \\ &\geq D_{1/2}(\mathrm{Tr}_{B}[\rho_{AB}]\|\mathrm{Tr}_{B}[I_{AB}]) \\ &= D_{1/2}(\rho_{A}\||B|I_{A}) \\ &= -2 \log \mathrm{Tr}[\rho^{1/2}_{A} |B|^{1/2}] \end{aligned} $$

Rearranging you get $$ \log \mathrm{Tr}[\rho^{1/2}_{AB}] \leq \log \mathrm{Tr}[\rho^{1/2}_{A}] + \log|B|^{1/2} $$ which is actually a slightly stronger inequality with the exponent of $|B|$ being only $1/2$.

  • 2
    $\begingroup$ It seems like you can also use data processing on fidelity under partial trace. $\text{Tr}(\rho_{AB}^{1/2}) = F(\rho_{AB}, I_{AB}) \leq F(\rho_A, |B|I_A) = |B|^{1/2}F(\rho_A, I_A) = |B|^{1/2}\text{Tr}(\rho_A^{1/2})$. $\endgroup$
    – rnva
    Sep 24, 2020 at 14:37

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.