If You Had to Pick Just One…

It is too early to pick any one. Meeting the timing for the technology and the production ramp for the customer products is the primary goal of the litho technology.  We may be use more than one. They can be complimentary to each other or not.  One can look at each technology as to the timing and to the market that it is intended to satisfy.  Over time this will change. 

Thanks, Paul. So, does it not matter at this point to GlobalFoundries which technology comes first? Does EUV, e-beam direct write, nanoimprint or other novel approach hold a leadership position for you at all?

Aaron, EDA's highest objective is focused on IC design productivity.  One element is to shield chip designers from the esoteric details of semiconductor manufacturing as much as possible.   The usual evolution is to first develop a verification methodology, then a reasonable repair flow.  The end-game is to prevent all issues in the first place with sufficient automation in the physical synthesis tools.  

The multiplicity of alternative lithography schemes we are seeing now increases the challenge.  One approach may be to conform to a common set of constraints so that layouts are more or less portable among alternatives.  On the other hand, If there isn't an optimal intersection of constraints, then EDA tools must support generating layouts optimized for specific lithography methods.

Aaron, the answer is "everything's difficult in its own unique way."  If we define mask extensibility as how far we can build masks to the demanded specifications, each technology has its benefits and challenges.  For instance, while double patterning provides a relaxed pitch requirement (benefit), it has very tight placement and CD accuracy specs (challenge).  And while EUV may have the most relaxed CD specs (benefit), defect management will be a challenge.  The other technologies have their puts and takes as well.  And don't forget that there is always the possibility (probability?) of combining more than one of the approaches in a multi-patterning scenario.

Thank you, Mike and Franklin. You both come from a supplier's perspective, and you will of course in all likelihood be providing the full range of technology solutions. I think it will be interesting to see how each of those challenges (aka "opportunities") are approached and how they will change over time.

You're right, Aaron!   I'd pick EbDW if I have to pick one. But I want to first say that EUV and NIL both have their places. No matter what way it goes, Ebeam either writes masks or wafers.  So investment in Ebeam technologies is crucial to the health of the semiconductor industry.

I'd pick EbDW if I'm forced to pick only one because of Eb and because of DW.  DW first.  Maskless is important for the "long tail" (in semiconductor, it's more like the long body, since only the head is above water).  To enable innovation at the leading edge nodes, we need DW (direct write).  Eb next.  Ebeam is scalable on the accuracy side.  Parallelization (multiple ways to do this) enables scalability in the write time dimension.  And Ebeam is much cheaper to operate than EUV.  NIL has that going for it too, but it isn't DW.  So at the end of the day DW is more important than Eb.

Thanks, Aki. It's an unrealistic exercise, I admit, but I think it's interesting to consider where you'd want to put the most emphasis anyway. It seems our lives are filled with the idea that we don't have enough time, money or other resources to do everything, and yet we seem to keep forging ahead with "everything" nonetheless.

I especially appreciate your argument that the DW is more crucial than the EB. It's a very interesting point.