Rising lithography costs: Where’s the crisis?
12/01/2007
As the semiconductor industry has advanced to the beat of Moore’s Law, executives have repeatedly expressed concern about the rising cost of lithography. There is some concern that next-generation extreme ultraviolet (EUV) scanners could be cost-prohibitive, creating an economic crisis for the semiconductor industry. But how serious is this economic threat?
Moore lithography
The primary driver of Moore’s Law has been the scaling enabled by advances in optical lithography with shorter wavelengths, higher lens numerical apertures (NAs), and more effective lithography processing (k1).
Over the past 20 years, lithography systems have made dramatic contributions to scaling. Enabling this progress has been an increase in optical complexity, high-performance servo-mechanics, and environmental control. These advances have increased lithography system prices by a factor >12. Yet concurrent increases in productivity, together with increased resolution, have enabled the effective patterning cost/hour of the minimum feature to be scaled by as much as 2×10-4, in-line with the 35% annual cost reduction often inferred from Moore’s Law.
Strong demand for the latest immersion lithography systems for sub-45nm processes offers the best evidence that this simple economic relationship continues to work, despite the rising price of state-of-the-art lithography systems.
The prospects for continuing this economic relationship appear to be challenged, however, as optical lithography hits the limits of physics. NAs already have reached the limit possible with the refractive indices of current materials. Higher-index fluids and optical materials are being developed, but these are years from achieving the performance required for advanced lithography, and even then will offer less than a one-node shrink. Process k1 factors are also at their limit. The only way to extend them is by using double-patterning processes, which significantly increases cost due to the doubling of process steps, halving of throughput, and requirement for two masks. EUV, with its significantly shorter wavelength of 13.5nm, offers the best prospect for continued scaling. However, it is challenged, especially in source power, to reach competitive throughputs and also requires development of a new mask infrastructure.
Beyond Moore?
So is this the end of Moore’s law? First, it is important to remember that only a few lithography process steps typically require the most advanced technology. Most steps can be handled by more mature lithography, which continues to deliver cost reductions through improved productivity. Second, it is necessary to evaluate the integral processing costs of the lithography operation, which have risen significantly with the move to lower k1 and will increase further with the use of double patterning. The growing number of processing steps increases the risk of additional yield loss. EUV, however, should result in simpler lithography processing, offering cost reduction and higher yield.
Design and mask costs have also increased as a consequence of the challenges of low-k1 lithography, which requires sophisticated design optimization and optical proximity correction. Although current lithography systems use a reduction factor of 4×, the amplification of mask errors with low-k1 processing is such that the errors transfer 1:1 from the mask, requiring very precise mask manufacturing and degrading both mask and die yield. Higher-k1 EUV will provide some relief from these challenges.
Finally, the efficiency of scaling itself is determined not just by the lithographic resolution, but also by the critical dimension uniformity (CDU) tolerances required for patterning, and overlay tolerances from layer to layer. Here again, simpler EUV process requirements should offer more effective shrinks compared to double-patterning schemes, and should enable the integral wafer processing costs to be leveraged over more die per wafer.
Therefore, as long as scaling remains so technically and economically attractive, EUV lithography will remain on customers’ roadmaps.
 Peter Jenkins, VP, ASML |
Contact Peter Jenkins at ASML, De Run 6501 5502 DR Veldhoven, The Netherlands; ph 31/40-268-3000, e-mail [email protected].