Chemical Vapor Deposition (CVD) is one of the efficient ways of implementing vapor phase deposition in a manufacturing environment. Among its many applications are the synthesis of silicon in the wafer production industry, for implanting dopants in wafers, for depositing metallic interconnects between devices in integrated circuits, for creating compound semiconductor lasers and for synthesizing the high thermal conductivity diamond needed for high circuit density multichip modules. It is also the key technology for producing flat panel displays, creating efficient photovoltaic devises, synthesizing high temperature superconductors or ferroelectric materials. Still other applications include the fiber coating, metal matrices for titanium matrix composites, thermal barrier oxide coatings for aircraft engine components, infrared detectors based on Vanadium Oxides and perovskites.
While much progressing in improving the efficiency and reliability of CVD technology has been achieved, it is still a rather costly process, especially when one needs to obtain the high quality required in many of the above applications. Among the serious problems encountered is the minimization of defects in the production process; point defects, for instance, can seriously degrade optoelectronic performance and they are responsible for undesirable increases in the resistivity in interconnects. It is known that point defects, point defect clusters and voids develop under some processing conditions (e.g. too high deposition rate or low adatom energies). Dislocations and stacking faults can cause immature failure in strained layer solid state laser structures. Planar grain boundaries my be responsible for the failure of metal interconnects by electromigration (which may cause nucleation of voids).
Unlike physical deposition, CVD involves chemical reactions on the surface. Such reactions can be modeled at several levels, including rate equations at the PDE level and probability distributions at the MD level, but it is possible to do much better than that (i.e. by defining appropriate adiabatic potential surfaces and employing quasiclassical theories for surface induced molecular breakup/binding). Some chemical processes (such as dopant deposition, e.g. in using MBE) can be simulated in such a way. Other features include various modifications to classical hydrodynamics, such as the Soret effects. Numerous other open questions include the chemical significance of defects, imperfect surface coverage, surface corrugation and other surface structures.
We have made progress on these problems in two general ways. First, by the development of a general geometry computer code that includes the hydrodynamical description together with the novel surface dynamics outlined above. This code has been run on prototypical cases of interest to applications and will be emphasized in the studies to be made in the next two years of this MURI. A research paper on the new numerical techniques involved is in preparation.
Second, we have completed studies of two and three-dimensional CVD deposition in vias of a very low pressure gas in which kinetics is emphasized. The results of this study show the effect of flux and energy on the shape of the deposited surface and the utility of these methods for coating of vias. A publication on these results is in preparation.