The computation speed of modern computers is limited by the resistance of the submicron metal lines used to interconnect devices. One solution to the problem of inefficient interconnect is a vertical interconnect between layered devices, a key to 3-D integration. Because IC fabrication is traditionally 2-D procedure, no method has been developed to create vertical submicron diameter "wires". This electron elevator could be made by a selective metal deposition process whereby a material catalyzed its own deposition, filling a well in an insulating material. Loss of selectivity (LOS), leading to shorts between metal lines, is due to reaction with the noncatalytic insulator, gas phase reactions, nucleation and deposition on the insulator, and chemistries between reactive intermediates from the deposition reaction. Finding the process margin is a matter of understanding all of the chemical pathways and relevant transport processes of both the heterogeneous and homogeneous reactions. The LOS kinetics are quite complex for the reactive species, have very short lifetimes, and can only be seen with in situ diagnostics. The goal of the research is to answer some very specific questions which cannot be answered in commercial or UHV reactors. What is the role of thermodiffusion and recirculation in LPCVD reactors? What are the gas phase intermediates, reactions, and products during selective CVD? How do our LPCVD (1 torr) results compare with those we determine in the UHV studies funded by NSF? Over the long term, the research will lead to understanding what other reactants could be used to create selectively deposited vertical wires and what the process margins are in terms of local composition and temperature. Once the local conditions are known for selective deposition, it becomes necessary to implement mathematical models of commercial reactors so that the inlet gas composition which results in the desired near wafer composition can be determined. This global model must also include modeling of the transience within the submicron well as it fills. Current models will be extended to new chemistries.
