SummaryBy choosing the optimal steering history of a spacecraft, it is possible to maximize the mass delivered from a park orbit to a mission orbit. A low‒thrust orbit transfer that models coasts when passing through the Earth's shadow can be formulated as a large‒scale optimal control problem with many distinct phases. This paper presents a technique that constructs an initial guess using a receding horizon algorithm. A series of large‒scale multiphase optimal control problems are then solved to refine the phase structure of the problem. The final optimal solution incorporates high fidelity physics and mesh refinement techniques within a large sparse nonlinear programming approach.