The light sources needed for the next generation of chipmaking lithography remain significantly short of their power target.
The shift to extreme ultraviolet (EUV), which will use 13.5nm wavelength light to expose patterns on wafers, is seen as crucial to move beyond 25nm-scale production processes. The alternative, extending the use of 193nm deep ultraviolet sources based on argon fluoride (ArF) lasers, threatens to make the move to sub-30nm processes prohibitively expensive.
But since the start of work on EUV lithography more than ten years ago, it has proved difficult to scale the power needed to illuminate the pattern on a chip mask much beyond a few watts.
Although source manufacturers such as Cymer have demonstrated output power of more than 100W for short periods, Cymer’s source, which is based on a 25kW carbon dioxide laser, was only able to exceed 10W for extended periods in 2010.
Although likely to increase energy consumption to more than 70kW, a boost in output power to 100W should make it possible to expose 60 wafers per hour instead of the five possible today. The improved throughput would still be 40 wafers short of the number considered practical for economic chipmaking.
However, production deadlines for sources that can be operated for long periods have slipped. In 2010, Cymer claimed it would have a 100W source ready for use by customers of lithography tool supplier ASML by the beginning of 2011. The deadline then moved to the middle of this year. Ahead of the Semicon West show, the expected date had moved to the end of this year.
Skip Miller, vice president of sales at ASML, conceded: “The leap to 100W for 60 wafers per hour is still a challenge.”
ASML is hedging its bets by working with three competing suppliers: Cymer; Gigaphoton; and Xtreme Technologies.
Marc Corthout, president of Xtreme, argued: “Increasing the power is a technological challenge but certainly not a physics challenge anymore.”
In the absence of a guarantee that production tools will be ready in time, chipmakers such as Globalfoundries are choosing to extend the use of 193nm scanners to the 20nm node, invoking complex and relatively expensive techniques such as double patterning .
Yayi Wei, senior member of technical staff at Globalfoundries said: “On our schedule we need test chips for 20nm in 2011. At this point, the solution for us is to use double patterning for 20nm. There is a chance in the future that we might transfer to EUV, but we will have to see.”
Nikon Precision, a competitor to ASML, has decided to focus on double patterning on 193nm equipment for the medium term. Stefan Renwick, said he expects EUV to “finally come along in the 2015 timeframe”.