Momentum state engineering and control in Bose-Einstein condensates

We demonstrate theoretically the use of genetic learning algorithms to coherently control the dynamics of a Bose-Einstein condensate. We consider specifically the situation of a condensate in an optical lattice formed by two counterpropagating laser beams. The frequency detuning between the lasers acts as a control parameter that can be used to precisely manipulate the condensate even in the presence of a significant mean-field energy. We illustrate this procedure in the coherent acceleration of a condensate and in the preparation of a superposition of prescribed relative phase.