10 July 2012
Will Intel's investment break the euv log jam?
Lithography is probably the biggest challenge facing those developing next generation semiconductor manufacturing equipment. The reason is that the current approach – 193nm immersion lithography – has just about run out of steam.
The problem is the features being created are much smaller than the wavelength of the light being used. Innovative ways of getting around this fundamental problem have been developed – including double patterning – but they have only provided breathing room, rather than the way forward. Double, triple or even quad patterning can produce the feature sizes required by advanced nodes, but at the expense of many more process steps – and therefore cost.
For a decade or so, extreme ultraviolet (euv) lithography systems have been under development and seen as the way forward. EUV generates energy – it's a bit of a stretch to call it light – with a much smaller wavelength. But there are problems. It needs to operate in a vacuum and uses a complex system of mirrors which absorb a lot of the energy generated. It's also slow – currently an order of magnitude behind 193nm – and consumes huge amounts of power. Unusually for the industry, it appears the challenges are not being solved.
In essence, more energy needs to be generated for longer in order to speed throughput. A US start up called Zplasma claims its euv solution – which creates a xenon plasma using a variant of fusion energy – does just that.
However, an indication of the industry's frustration – even desperation –with euv progress comes with Intel's massive investment into lithography specialist ASML. It's ploughing $1billion into R&D over the next five years, including addressing some challenges related to 450mm wafers. It's thought this investment could bring euv development forward by two years.
Intel has moved subtly over the recent past from saying 'we'll use euv as and when' to saying 'we'll use self aligned double patterning and if euv comes along, we'll use it'. But does the investment mean Intel has a more pressing need than it appears?
Investing $1bn in ASML's R&D may not unblock the logjam unless some of the money flows to those companies developing euv light sources. The 'state of the art' for a light source is currently around 100W. Improving throughput by a factor of 10 implies the need for a 1kW light source – and that's not on the cards at the moment. Without that, euv goes nowhere.
Recently, a TI executive claimed euv was 'tomorrow's technology – and always will be'. Is it possible this perennial throwaway line might actually be true in this case?
This material is protected by Findlay Media copyright
One-off usage is permitted but bulk copying is not.
For multiple copies contact the