This blog post is about a short perspective article that I recently wrote for the ILTPC newsletter. ILTPC stands for the International Low Temperature Plasma Community. It is one of the two main communities where I feel at home. The other one is the atomic scale processing community. This one is basically the atomic layer deposition (ALD) community but nowadays extended with related and overlapping areas such as atomic layer etching (ALE) and area-selective deposition (ASD). Furthermore, there is also considerable overlap between the low temperature plasma community and the atomic scale processing community as the methods such as ALD, ALE and ASD are also often plasma-enhanced. My research interests lie especially (but not only) there were these two fields come together. Also note that both the communities are strongly rooted in the AVS.
To organize the low temperature plasma community to a higher degree, the ILTPC was established in April 2020 with Mark Kushner (University of Michigan) and Peter Bruggeman (University of Minnesota) as main driving forces. The goal was to be a forum for communication, collaboration, and conversation for the low-temperature plasma community. The ILTPC newsletter is doing an excellent job in that respect, I very much like it. So please check it out and sign up if you like it too.
Obviously, I immediately agreed when Mark Kushner approached me to write a short perspective article for the ILTPC newsletter. I decided to devote it to artificial intelligence (AI) as this topic started to draw my attention big time in recent months. And I’m certainly not the only one! My interest was particularly triggered when playing around with – and being amazed by – ChatGTP. See also post that I wrote about ChatGTP and ALD earlier this year: Let’s chat GPC – Putting ChatGPT to the test and a first (?) article about ALD by AI.
Below you find the text of the perspective article in the April issue of the ILTPC newsletter reproduced. Please also make sure you check out the ILTPC newsletter itself. Also note that what is being written about the plasma community, basically also holds for the atomic scale processing community. This community is also enabling AI and next it will certainly start benefitting from it too. Finally, I also want to thank Aaf from the De Digitale Vormgever for the great images. She is responsible for many of the images on this blog site.
After Enabling AI It Is Time for the Plasma Community to Benefit from AI
That artificial intelligence (AI) draws a great deal of attention lately probably hasn’t gone unnoticed by the readers of this article. Particularly the availability of bots such as ChatGPT has demonstrated many of us the remarkable capabilities of AI: after decades of just being a promise, we suddenly have the opportunity to experience the power of AI ourselves. We start to realize that it will soon transform the way we live, work, and interact with the world around us.
Our low temperature plasma community has obviously played a key role in enabling the breakthrough of AI. By developing highly advanced etch and deposition methods for high-fidelity processing steps at the nanoscale, the low temperature plasma community has contributed to the development of all kinds of powerful hardware technologies important for AI. These include central processing units (CPUs), graphics processing units (GPUs), field-programmable gate arrays (FPGAs), and application-specific integrated circuits (ASICs). Although these don’t always require the real cutting-edge process nodes, they do involve specialized circuits that have enabled the rapid progress of AI in recent years, allowing developers to train larger, more complex models and deploy AI applications more efficiently. Furthermore, AI will bring into play new computing paradigms, for example those based on neuromorphic chips involving a range of novel devices such as memristors that impose also other requirements on deposition and etching of materials.
Now the time has come that plasma processes do not only enable AI but also benefit from AI. AI can improve semiconductor manufacturing processes, for example by increasing efficiency, reducing defects, and improving product quality. By leveraging machine learning algorithms to analyze large amounts of data, semiconductor manufacturers can gain insights into the manufacturing process that would be difficult or impossible to obtain manually. This will enable them to make more informed decisions and optimize their operations. Note that the notion that AI tools can be helpful for monitoring and controlling plasma processes is nothing new, it was already realized more than 30 years ago. For example, an IBM team involving the highly valued member of our community, Prof. Gottlieb Oehrlein, published an article about the use of optical emission spectroscopy and mass spectrometry data for reactive ion etch process parameter estimation using neural networks in 1992. However, now – due to the advancement of technology to which our plasma community has contributed – we have entered the next stage. This is underlined by an article that was published last month in Nature: Keren Kanarik and co-workers in the team of Richard Gottscho at Lam Research Corporation report how semiconductor process development can be improved by humans partnering with computers and AI algorithms now that large amounts of data are becoming available and can be processed easily. They set out a strategy to be followed and it becomes absolutely clear that AI will soon revolutionize the way our plasma community works in the semiconductor industry.
(written with the help of ChatGPT)