We present a dynamic, coupled macroscopic model (DOGMA) for the process of reactive sputtering suited for description of real in-line processing. In our model we divide the complex volume of a processing chamber into simple cells, e.g. parallelepipeds. For each cell gettering kinetics of reactive gas at metallized surfaces as well as sputtering kinetics of the target are calculated using a Runge-Kutta time step method. The inert and reactive gas flow is obtained via flow conductances defined for each connection between adjacent cells. For deposition of sputtered particles on chamber and substrate surfaces, a pressure dependent distribution matrix is used, which is obtained from single-particle Monte Carlo calculations. The flow conductances for different pressure conditions are calculated using the well known 'Direct Simulation Monte Carlo' method implemented on a Linux cluster, where for each volume cell an individual calculation task can be spawned. Tuned with flow conductances and particle distribution factors from Monte Carlo calculations our coupled model is capable to describe real in-line sputtering processes on a single-processor PC at almost real time speed. All numerical models are implemented within a simulation system RIG-VM developed at Fraunhofer IST, which enables easy communication between separate modules, e.g. during modeling of a controlling cycle and ensures interchangeability and reusability by using XML datasheets and web browser based user interfaces. In this paper we apply our model to reactive sputtering of SiO2 with X-probe based oxygen partial pressure stabilization. The model is adapted to experiment via material parameter variation. The resulting simulation system can be applied for optimization of process parameters with respect to process stability and film homogeneity. (C) 2003 Elsevier Science B.V. All rights reserved.