The development of nonlinear control strategies for high purity distillation columns requires the availability of suitable nonlinear dynamic models. Relatively simple models are needed for the design of nonlinear model predictive controllers (NMPC) which are implemented on-line via repeated solution of a constrained nonlinear optimization problem. We utilize the compartmental modeling approach to construct control relevant dynamic models of a very high purity nitrogen purification column. Reduced-order nonlinear models are derived from a stage-by-stage balance model by partitioning the column into distinct compartments and applying time-scale arguments to convert most of the ordinary differential equations (ODE) describing each compartment into simpler algebraic equations. The compartmental models are dynamically simulated using the differential-algebraic equation (DAE) solver available in MATLAB. The stage-by-stage and compartmental models are compared to a rigorous dynamic simulator constructed in Aspen Dynamics to assess the merits of the proposed modeling strategy with regard to prediction accuracy and computational efficiency. (C) 2004 Elsevier B.V. All rights reserved.