The scaling of CVD rotating disk reactors to large sizes and comparison with theory

An important advantage of the rotating disk reactor (RDR) geometry for chemical vapor deposition is the one-dimensional nature of the transport process over a wide range of operating conditions. Due to geometric complexity, models for most chemical vapor depositon (CVD) reactors have been used to fit existing data and are not very useful as predictors of performance for changed dimensions, particularly larger ones. Previous RDR modeling results have shown good agreement with experiment for smaller sizes (typically single wafer) and have also resulted in a set of scaling “laws” for scaling to larger sizes. In this work, we report the use of these scaling laws to design larger RDRs and to determine optimal process conditions. We then compare the experimental results to those predicted by the model. Previously, our largest RDR utilized an 180 mm diameter disk, which holds 6 × 50 mm or 3 × 76 mm wafers. The new systems have disks of 300 and 420 mm diameter, holding 17 and 38, 50 mm wafers, respectively. In general, excellent agreement was obtained when the new systems were run at conditions predicted by the scaling laws. As the disk diameter increases, the model calls for a reduced rotational rate, a lower operating pressure and a main carrier gas flow that increases sublinearly with the disk area. Experimental growths were made with the III-V materials AlGaAs and InGaAlP, on GaAs substrates of 50, 76, and 100 mm diameters. Uniformities and electrical and optical properties similar to those obtained in smaller RDR systems were obtained under conditions predicted by the model. We also discuss the trade-offs involved where one parameter is fixed by process needs, necessitating that others must be adjusted accordingly. It should be stressed that these adjustments are all calculated from the model and are designed to maintain the ideal flow and thermal environments obtained with the smaller systems. We find that typical metalorganic chemical vapor deposition process parameters are easily accommodated for these disk sizes. We believe that this is the first report of a truly scalable CVD reactor geometry with both theoretical and experimental justification. Using the RDR technique therefore enables the grower to develop a process on a small system and transfer it to a larger system for manufacturing, without having to reinvent the process. It also enables the CVD system manufacturer to design and build new systems that will perform in a predictable manner, or to adapt new process conditions to an existing system, without requiring extensive re-engineering efforts.