Uncommon Goals

I had the opportunity to attend the Common Platform event recently. This is a technology and business showcase sponsored by Global Foundries, IBM and Samsung with major support from ARM, Cadence, Synopsys and Mentor. Wow, that’s some serious sponsorship. The event was well run and provided a good balance of technology details and business outlook. The wine at the evening reception was decent and there was even entertainment throughout the day.

Enough of the pleasantries—what did I learn by being there? FinFETs are coming. Almost every presentation touched on this technology. These device structures clearly will lay down more road for the semiconductor technology roadmap. Other tidbits: Gate first works. 3D stacks are coming. For the 20th year in a row, IBM has the most patents. These are all interesting stories, but the big news is that EUV lithography is really hard, but it has to work.

Extreme ultraviolet lithography is incredibly hard to perfect. A lot of smart people have been at it for a long time. Some folks are projecting 7nm as the node that will see this technology finally flourish—maybe. I won’t go into the long list of chemical, mechanical, physical and optical challenges to get this technology off the ground. You can read plenty about that from many sources. But I do want to explore why this technology is so critical.

There are options. New materials can change the game (e.g., carbon nanotubes). You could do away with optics altogether (e.g., direct ebeam write). There is even research in chemical processes that will form devices at a molecular level. While many of these approaches show promise, they all share one common obstacle: They require significant change in the manufacturing supply chain to adopt. And that could be the fatal flaw for all of them.

In a past life, I did some work in design for manufacturability. I got an up-close look at the mask data prep and lithography workflow. To say the folks who run these processes are conservative is an understatement. Change happens very, very slowly here. When there are billions of dollars of equipment (and its depreciation) on the line, everyone obsesses about uptime, stability and reliability. Small perturbations in the workflow need to be very, very stable to be integrated. EUV is in itself a change, but far less traumatic than some of the other options.

So, in my opinion at least, we really need EUV lithography to float. Otherwise, the semiconductor technology roadmap might indeed run out of road.