This work presents a novel superstructure optimization approach for the design of distillation sequences of zeotropic mixtures. The approach considers as alternatives from conventional sequences in which each final distillation column has a condenser and a reboiler, to fully thermally coupled distillation sequences going through all possible intermediate combinations. A two-stage procedure is proposed. In the first one, a sequence of tasks is selected and then in a second stage the best configuration of actual columns is extracted among all the thermodynamically equivalent configurations. The model uses the Underwood-Fenske-Gilliland approximation and is formulated as a generalized disjunctive programming problem. The model has proven to be very robust and reliable. This formulation also makes the model very flexible and allows the easy modification of the equations and even migration to other model types including rigorous tray-by-tray models. Rigorous simulations in HYSYS [AEA Technology, Hyprotech Ltd., Calgary, Canada, 2001] have shown good agreement with the shortcut model. (C) 2004 Elsevier Ltd. All rights reserved.