Researchers at Argonne National Laboratory have linked artificial intelligence to a system for applying thin films of atomic thickness to silicon during the manufacture of microcircuits. Such films can carry many functional loads: from insulating elements to giving special characteristics to transistors on a chip. AI optimizes the film application process and saves time and money.
The thin-film coating process is called Atomic Layer Deposition (ALD). Most often, the ALD process uses two precursors alternately, which are pumped as a gas into a silicon wafer chemical reactor. Each of them needs to be pumped out over time, and so many times. The quality and properties of the film applied in this way depend on the duration of each cycle. During the optimization process, researchers must retrieve the sample multiple times and evaluate the coverage.
American scientists were able to connect AI to the process with feedback from the reactor. The algorithm calculates the expected ideal cycle of atomic film deposition and almost immediately receives data on the chemical reaction performed. There is no longer any need to remove a sample, take measurements and place it for further processing. Automation immediately adjusts the installation parameters to improve the film build-up result. Thanks to AI, the process is much faster.
The introduction of such systems in production will allow chip manufacturers to significantly speed up the development of new technical processes and even improve existing ones. Today, when the traditional CMOS process has come to its limit, it is more than relevant.