A mathematical model of atomic layer deposition of alumina is developed using a TMA + H2O ALD operation as a physical base. The model coherently bridges the process dynamics of ALD reactor with the detailed surface chemical kinetics, where the time-dependent variations of surface species are modeled during each precursor pulse. The model is validated by the experimental data. Upon the validation, continuous ALD cycling operation is simulated under different operating conditions. Results show that temperature rising can reduce the surface reaction time in two half cycles, and meanwhile intensify the desorption rate of precursors and backward surface reactions. Suitably increasing the system pressure can improve the completion of surface reactions and film formation in the cycling. The sensitivity analysis shows that surface species variations of vertical bar-OH, vertical bar-Al(CH3)(2) and vertical bar-Al(CH3)(2) have different degree of sensitivities to the surface reaction steps, and reveals the rate-limiting steps in two precursor treatment cycles. (C) 2014 Elsevier B.V. All rights reserved.