A Cloud Modeling Capability for Low Temperature Plasmas for Microelectronics

Plasma processing is critical for the development of microelectronics devices. One of the problems currently faced is to obtain uniform device yield over a wafer, as there can be large variations in plasma conditions at the wafer edge. Various solutions, such as “focus rings”, have been proposed, but trying out each of these solutions can be expensive and time consuming with the need to shut down the chamber, insert a new device, restart the chamber, process a wafer, and measure the results. The process of improvement could be accelerated with computational modeling. However, there is a relative lack of CAE Computationally Aided Engineering software that can address this problem including modeling an entire chamber and computing the wafer impacting ion distributions. In addition, since these problems are computationally large, there is the problem of access to sufficient computational resources to address these problems. TechX will develop a GPU enabled computational application that will be able to address the needs of the plasma processing community through being highly performant and able to take advantage of new computational hardware. This will be cloud enabled such that an engineer will be able to carry out these computations without needing to have a local supercomputer or computer cluster. Phase I will resulted in 1 a cloud approach to computation that presented a desktop interface to the engineer, while carrying out the computation on a remote cluster, 2 an implementation of GPU capable simulations for modeling plasma processing, 3 development of flexible GPU capability that allows the needed simulation to be set up through an input file, 4 prototyping plasmawall interactions on GPUs. Phase II will build out this infrastructure, including development of GPU capable linear solvers, including an inductive solver, more and different kinds of particles, enhancing the set of collisions, including particle management such as splitting and recombining. Extensive efforts in code optimization will be undertaken. As well, the capability will be made available through public clouds, such as Amazon Web Services, and the new physics capabilities will become available through this cloud interface. The developed application will have use in the area of plasma simulation for microelectronics. Further, the capability would be made available on research clusters to benefit the plasma processing research community.