Quantum mutual information redistribution by Number Partitioning algorithm

Quantum information distribution in a tripartite state plays a fundamental role in quantum information processes. Here we investigate how a bipartite unitary transformation $U_{AB}$ redistributes the quantum mutual information with the third party $C$ in a tripartite pure state $|\psi\rangle_{ABC}$ in a $d_A\times d_B\times d_C$ dimensional Hilbert space. In particular, we focus on finding out the optimal unitary transformation $U_{AB}^{\ast}$ that maximizes the quantum mutual entropy between party $A$ and party $C$, $I(A:C)=S(\rho_A)-S(\rho_B)+S(\rho_C)$. We show that the mutual entropy $I(A:C)$ is upper bounded by $2S(\rho_C)$ derived from the Araki-Lieb inequality. This upper bound can be realized via an optimal unitary transformation for any pure state with the rank $r_{C}$ of $\rho_C$ satisfying $r_C\le d_A$. For a generic pure state with $r_C> d_A$, the upper bound can not be realized by any bipartite unitary transformation. To maximize the mutual entropy in the latter case, we propose a fast numerical algorithm to produce an approximate optimal unitary transformation, where our optimization is transformed into a modified number partition problem. The validness of our algorithm is confirmed by its comparison with the results from the Adam algorithm for parameterized unitary transformations. Our approximate algorithm thus provides a practical protocol to implement redistribution of quantum mutual information for a tripartite quantum state with high dimensions.