Linking Micro- and Macro-Scale Models of the OMVPE Process

Rajesh Venkataramani, Massachusetts Institute of Technology

The organometallic vapor phase epitaxy (OMVPE) process is widely used to produce compound semiconductors for applications ranging from solar cells to cellular phones. The OMVPE process is a chemical vapor deposition (CVD) process where thin films are synthesized from organometallic precursors. As device dimensions decrease, processing issues that affect the material properties (e.g., stringent demands on film thickness, morphology and impurity incorporation) become increasingly important. Models of OMVPE can be used in fundamental studies of the processing of thin films, as well as in predictive simulations to aid in process design and control. The difficulties in modeling the process arise from the complex nature of OMVPE; processes occur on widely different length and time scales. In this work, models on various length and time scales for the OMVPE process will be developed and methodologies for linking these models will be formulated. The “linked” models will be shown to make accurate predictions on various length scales. The complex coupling of gas-phase transport phenomena with gas-phase and surface chemical kinetics means that gaining an understanding of the entire OMVPE process requires more than one type of modeling approach. Predictions on the macroscopic length scale (growth rate, film uniformity, film composition) and on the “mesoscopic” length scale (surface morphology) are both needed to determine optimal operating conditions for OMVPE reactors. Models have been developed to make predictions on each of the relevant length scales:

  • Macroscopic reactor scale models have been developed to understand the fluid flow, heat and mass transfer combined with detailed gas-phase chemical kinetic mechanisms to predict the type and concentration of precursors arriving at the growth front. Finite Element simulations have been developed in our group to solve these problems.

  • “Mesoscopic” surface models have been developed to study the evolution of surface morphology during growth. Kinetic Monte Carlo methods have been used to investigate the importance of individual steps in surface mechanisms, as well as determine the growth mode on the surface. Parallel algorithms are also in the process of being developed in order to speed the computations of the surface problem.

A methodology is developed to “link” reactor scale and surface models in order to gain an overall understanding of the OMVPE process. Case studies of GaAs growth using a variety of precursors is used to validate the models separately, as well as to validate the “linked” model. Using the reactor scale model alone cannot give predictions of surface morphology. The surface model used alone is missing a needed input of gas-phase flux to the surface. The “linked” model is able to make extremely accurate predictions on both the macroscopic length scale (growth rate) and mesoscopic length scale (surface morphology).

Abstract Author(s): Rajesh Venkataramani