Outer-space plasmas and arc-jet plasmas are well known to many researchers at NASA. Efforts to understand the structures of these plasmas have consumed years of research at various NASA laboratories. Computational modeling of such plasmas involves analyses of multicomponent, multitemperature flows, and many computer codes developed by NASA are available for this purpose. Now, researchers at Ames Research Center have applied their expertise to understand a different kind of plasma — the kind used in manufacturing integrated circuits.
Typically, in such manufacturing, a weakly ionized plasma is used to etch fine patterns (smallest dimensions of the order of 10 –7 m) in a silicon wafer. The discharge gas (the gas used to generate the plasma) could be chlorine or some type of fluorocarbon, depending on the material to be etched or the required functionality. The objective is to obtain a highly directional (even vertical) etched profile at a rapid rate, with minimum use of chemicals. If one were to use the chemicals in their liquid— form in what is known as "wet etching" — this objective would be compromised. Because chemical etches are "directionally blind," wet etching results in isotropic features. In contrast, dry etching with the help of a plasma achieves desired directionality through the combined action of reactive atoms or radicals and vertical bombardment of the surface with ions to clear off chemical-reaction products.
It is of interest to the microelectronics community to understand this dry process in order to optimize the process and design the right kind of reactor. That is where computational modeling is valuable. A computer code called Semiconductor Equipment Modeling Software (SEMS) has been developed to enable such modeling. This program solves equations of multicomponent, multitemperature, chemically reacting flows. The set of equations includes the Navier-Stokes equations of subsonic flow of gases, a gas-energy equation, an electron-energy equation, multicomponent conservation-of-chemical-species equations, and Maxwell's equations for determining the power coupled to the plasma from an inductive coil or other external power source.
The coupling of several such equations with disparate time scales results in a stiff (in the mathematical sense) problem. Time scaling of the equations has been explored in an effort to find ways to solve the problem rapidly. SEMS has been used to model commercial reactors and processes common in the fabrication of integrated circuits. Results have been obtained for nitrogen, chlorine, carbon tetrafluoride, and other plasmas in inductively coupled plasma reactors for processing 300-mm wafers in preparation for the next-generation circuits.