As always, the opening of the symposium began with some awards. Our community’s biggest and most prestigious is the Frits Zernike Award for Microlithography, and it was wonderful to see Harry Levinson receive this year’s honor. (Full disclosure – I’m on the award selection committee.) Additionally, since last year’s award ceremony was virtual, Bruce Smith was giving his 2021 Frits Zernike Award for Microlithography as well. Congratulations to them both!
Four presentations of SPIE fellow were made next: Nelson Felix of IBM, Kevin Lucas of Synopsys, Uzodinma Okoroanyanwu of the University of Massachusetts, and Tatyana Sizyuk of Argonne National Laboratory. It’s a shame that Kevin and Uzo could not be here this week.
We next heard two of the three plenary talks (Eric Hosler’s talk on quantum computers will be given on Tuesday.) The first talk was by Luc Van den hove, President and CEO of imec.
A quick digression. We have had over the years a number of plenary talks given by various industry executives covering topics of interest to our community such as compute scaling, artificial intelligence, the automotive industry, progress in GPUs, etc. My biggest fear for these kinds of speakers is getting what I call the “kid on a skateboard” talk. The executive, giving the same talk they might give at an investor conference, says things like “Technology A is very important” while showing a kid on a skateboard, “Our company is ahead on technology B” then shows a family playing with a dog, etc. Very slick, and devoid of useful content. So when a CEO is asked to give a talk on a Wall Street-friendly topic such as “The endless progress of Moore’s Law”, I usually get worried.
But Luc Van den hove is not your typical CEO. He is a lithographer deep in his bones. He published his first SPIE paper in 1990, and was chairman of the Optical Microlithography conference in 1998 and 1999. He knows what he is talking about, and cares deeply about this community. So when the first few slides in his talk were of the “kid on a skateboard” variety, I was not worried. He soon got into the technical meat of the topic, and we were all rewarded for our patience. Taking the broad view of what Moore’s Law means that is typical of today, he described four general areas that will keep progress in semiconductors moving for quite some time (though not the hyperbolic “endless”): Shrinking the transistor, improving the transistor, moving into the third dimension, and shifting compute paradigms. I suspect that he is correct on all counts.
The second plenary by H.S. Philip Wong of Stanford went into considerably more detail on two of Luc’s topics, system-level optimization and 3D integration. Dr. Wong is an expert on these topics and I learned quite a bit. He would have been better off, however, if he had not tried to force lithography relevance into his talk through his provocative title and subsequent discussion of EUV lithography throughput (Tony Yen – you were a bad influence!).
For the rest of the day I alternated between the metrology conference and the two keynote talks at the Optical and EUV Nanolithography conference. Nelson Felix gave a nice review of metrology needs for nanosheet transistors, though I was very surprised when he showed that 1/3 of all the process steps in IBM’s latest generation process were metrology steps, and that this hasn’t changed since the 45 nm node. There is no doubt that IBM does more metrology than your typical fab. Mark Phillips of Intel gave a very optimistic view of when high-NA EUV lithography could be inserted into manufacturing, beating by a year the roadmap shown by Luc Van den hove (which, coming from imec, could also be assumed to be optimistic). It sounds to me that Intel is tired of being behind in EUV and is hoping that high-NA EUV will give them a chance to leapfrog ahead.
I ended the day with a Fractilia hospitality event at a favorite San Jose brew pub. Thanks to all who joined us!
The rumors of the death of Moore’s Law were greatly exaggerated! (What happened to the Lotus?)
Are the SPIE talks recorded and published anywhere?
If the author gives SPIE permission, then that presentation will be recorded and will show up on the SPIE Digital Library in about one week.