A stochastic optimization framework is proposed for optimizing semi-batch reactive crystallizations. Semi-batch reactive crystallization models normally consist of highly non-linear reaction and crystallization kinetics that deter the use of conventional deterministic optimization methodologies. Through this framework, numerous degrees of freedom in the reactive crystallization systems can be optimized simultaneously. The initial amount of the reactants, feed addition profiles, addition time, number of continuously added feed and interaction between two feed flow rates are optimized comprehensively, subject to various process and system constraints. Two objective functions are considered: (i) maximizing crystal size subject to a specified maximum coefficient of variation and (ii) minimizing coefficient of variation subject to a specified minimum crystal size. Results indicate that the optimized operation produced significant improvement in the overall crystal quality compared with the conventional semi-batch operation of constant feeding rate. (C) 2004 Elsevier Ltd. All rights reserved.