The optimal synthesis of zeotropic efficient distillation processes is addressed in this paper. The reversible distillation sequence model (RDSM) is rigorously approximated by a sequence of adiabatic units. Alternative superstructures representations involving different energy distributions are proposed for approximating the RDSM. The efficient distillation synthesis problem is formulated as a rigorous tray-by-tray mixed-integer nonlinear programming (MINLP) problem. Discrete decisions are related to the feed and heat-exchange locations while continuous choices select the operating conditions of the process. The most suitable representation is selected by computing the entropy production of each scheme. The solution procedure includes an automatic preprocessing phase where reversible distillation theory is combined with mathematical programming to generate initial values and bounds for the variables involved in the rigorous MINLP problem. Convergence and robustness of the formulations are considerably improved by including this preliminary phase. Numerical examples are reported for the separation of zeotropic mixtures to show the performance of the proposed methodology.