Decoherence time maximization and partial isolation for open quantum harmonic oscillator memory networks

This paper considers a network of open quantum harmonic oscillators which interact with their neighbours through direct energy and field-mediated couplings and also with external quantum fields. The position-momentum dynamic variables of the network are governed by linear quantum stochastic differential equations associated with the nodes of a graph whose edges specify the interconnection of the component oscillators. Such systems can be employed as Heisenberg picture quantum memories with an engineered ability to approximately retain initial conditions over a bounded time interval. We use the quantum memory decoherence time defined previously in terms of a fidelity threshold on a weighted mean-square deviation for a subset (or linear combinations) of network variables from their initial values. This approach is applied to maximizing a high-fidelity asymptotic approximation of the decoherence time over the direct energy coupling parameters of the network. The resulting optimality condition is a set of linear equations for blocks of a sparse matrix associated with the edges of the direct energy coupling graph of the network. We also discuss a setting where the quantum network has a subset of dynamic variables which are affected by the external fields only indirectly, through a complementary ``shielding'' system. This holds under a rank condition on the network-field coupling matrix and can be achieved through an appropriate field-mediated coupling between the component oscillators. The partially isolated subnetwork has a longer decoherence time in the high-fidelity limit, thus providing a particularly relevant candidate for a quantum memory.