Zhu and Rabitz [J. Chem. Phys. 109, 385 (1998)] presented a rapidly converg
ent iterative algorithm for optimal control of the expectation value of a p
ositive-definite observable in a pure-state quantum system. In this paper w
e generalize this algorithm to a quantum-statistical mechanics setting and
show that it is both efficient in the mixed-state case and effective in ach
ieving the control objective of maximizing the ensemble average of arbitrar
y observables in the cases studied.