TSMC: e-beam winning on cost over EUV for lithography

TSMC’s latest progress report on sub-10nm lithography options favors multiple e-beam lithography over the more heavily R&D funded extreme ultraviolet alternative.

TSMC has said that it is willing to use “viable” EUV and is an important backer of research projects looking to bring the technology forward. And Dr Burn Lin, its VP for R&D, held to that position in a presentation to this month’s Semicon Taiwan: “We are trying very hard to make EUV work,” he said, “even at 10nm if we can.”

However, during his Semicon speech Dr Lin showed some internal TSMC research data for the relative costs of ArF immersion, EUV (both 125W and 250W power sources) and multiple e-beam direct write lithography at the 7nm node. The benchmark index was provided by three use cases on ArF immersion lithography at 10nm, the technology TSMC expects to use at that node alongside multi-patterning. The results are summarized in this table:

Relative costs of 7nm lithography options against 10nm immersion benchmark (TSMC)

That none of the options in TSMC’s table meets the 10nm benchmark today is not surprising. This is all about work in progress. The disturbing issue for EUV’s more ardent proponents is the cost advantage TSMC still assigns to multiple e-beam.

That gap will raise further questions both over TSMC’s enthusiasm for EUV (long the subject of some debate in the business) and just how much of a return current R&D investments in the technology are delivering.

Lithography as ideology?

EUV has received the bulk of the industry’s R&D support for lithography while e-beam’s development has largely been a solo and parallel effort on TSMC’s part, running since 2005. This does mean that TSMC is talking its own book, though it was equally reasonable for Dr Lin to observe that “over those years we have accumulated a lot of knowledge [in e-beam technology].”

“We have the means on paper to get there,” he added. “And there are other potential [e-beam] users that could joint the effort.”

Similarly, while EUV is the industry’s first choice in public, several executives are worried in private that – as one non-TSMC manger put it – “the technology is becoming more ideological than technological”.

For his company’s part, Dr Lin cited a number of major manufacturing concerns that remain about EUV. TSMC’s research has highlighted major vulnerabilities to mask blank and contamination-based defects, a continued need for double patterning at 7nm, and difficulties inherent in the need to mount a pellicle on the mask. These go away with a shift to a direct write technology.