Even though rapid advances have been made in improving Cu(InGa)Se2 thin film based solar cell efficiencies, the breakthroughs have been limited to the laboratory scale. Most commercially viable thin-film technologies are at the pre-manufacturing development stage and scale-up has proven to be much more difficult than expected. Among various thin film deposition methods, the technique of co-evaporative physical vapor deposition (PVD) has yielded the highest efficiency solar cells. This process is employed at the University of Delaware's Institute of Energy Conversion (IEC) to deposit Cu(InGa)Se2 thin-films on a flexible substrate using a roll-to-roll processing scheme. The film is deposited by thermal evaporation from a series of elemental sources located sequentially in a vacuum deposition chamber. The process throughput is currently at the pilot manufacturing stage, with substrate width of 6” and translation speeds less than 1 inch/min.

A commercially viable process is required to produce large-area high quality films at a much higher throughput. Specifically, desired film thickness (~2μm), and composition uniformity have to be achieved continuously and reproducibly on large-area substrates (12 inch-wide) at a much higher translation speeds (~1 feet/min). While achieving the desired film thickness and composition set-points is best addressed by proper control system design, the film thickness uniformity is determined by the source design and individual nozzle effusion rates. The nozzle effusion rates, in turn, depend on the melt surface temperature profile and thus on the source design as well. Proper source design, therefore, is critical in achieving film thickness uniformity. We have identified two modeling requirements for effective commercial-scale source design: (i) detailed three-dimensional thermal modeling of the evaporation source – to predict the melt surface temperature accurately and to analyze the effect of melt level on the melt temperature profile; and (ii) nozzle effusion modeling – to predict effusion rate and vapor flux distribution for a given nozzle geometry (length and diameter), melt surface temperature and evaporant. To meet the first requirement for scaling up the University of Delaware process, we have developed a first-principles three-dimensional electro-thermal model using COMSOL Multiphysics software; the Direct Simulation Monte Carlo (DSMC) method has been used for effusion modeling. We discuss the details of the two models, their experimental validation, and present a procedure by which these models are employed in general to optimize the source design with respect to film thickness uniformity and material utilization. These modeling tools are then used specifically to design evaporation sources that achieve effective scale-up of the IEC's pilot-scale PVD process. Two commercial-scale sources are proposed for a 12”-wide substrate: a three-nozzle single source, and a four-nozzle modular source. The latter is easily scalable for webs wider than 12” because of its modular structure. We also show that the proposed source designs are robust to modeling errors and such process parameters as source-to-substrate distance.