Issue



Moore's Law: Recognizing the major contributions of


10/01/2002







Ralph R. Dammel,
AZ Electronic Materials

The success of the semiconductor industry in printing ever-finer features is frequently seen as only a triumph of hardware engineers. While there can be no doubt that improvements in steppers and other tools have made major contributions to the development of photolithography, the improvements in the "wetware" of the lithographic process, i.e., the photoresist, have also contributed substantially to the industry's ability to keep meeting and exceeding Moore's Law. The impact of these contributions is frequently underestimated, however, and while the rates of resolution improvement by optics and photoresists are found to be roughly equal in magnitude, the funds being spent on their development are greatly dissimilar.

The comparison of the relative NA improvement rates for 365, 248, and 193nm technologies shows that, in each case, the contribution to resolution improvement from the photoresist outweighs that from the stepper NA by roughly the same factor of 1.3–1.4.

The "equivalent NA" increase is most easily determined by comparing lithographic data obtained on the same exposure tool under standardized imaging conditions. As an example, for an i-line stepper with NA = 0.54, the resolution capability of 1984 lithographic resist and process technology (0.60–0.50μm) corresponds to a value of k1 = 0.81. By 1990, resolution had improved to 0.45 and by 1996 to 0.30μm. To achieve 1996's resolution with 1984's k1 factor, one would require an exposure tool with an NA of 0.99, i.e., an increase of 0.45 NA units. This NA increase of 0.45 units is the resists' resolution improvement in terms of "equivalent NA units." Closer inspection shows that this measure is linear over time, with a rate of increase ∂NA/∂t = 0.037/yr [1, 2]. In the same period, the NA of exposure tools increased from 0.32 for early g-line steppers to 0.63 for modern i-line tools, an increase of 0.31 NA units or 0.026 units/yr. In other words, for near-UV lithography, the contribution from resist and resist processing has outweighed the exposure optics improvements by a factor of about 1.4.

For all three lithographic technologies, the equivalent NA has a common slope of approximately 0.037 NA units/yr (see figure). The historical record, therefore, supports a constant, universal rate of equivalent NA improvement for all optical lithographies [1, 2].

Comparing the contributions of resist improvements to photolithographic resolution capability with those of steppers, one might very well argue that the comparison is unfair and that exposure tool vendors had to improve other factors, such as overlay accuracy or focus repeatability, in order to meet the tightening requirements. But similar improvements beyond just resolution had to be made in the photoresists as well. Resin platforms had to be adjusted for transparency at new wavelengths, which, for the switch from the DNQ/novolak inhibition systems used at near-UV to polyhydroxystyrene at 248nm, required chemical amplification as a new imaging scheme, and which, for 193nm lithography, mandated the development of nonaromatic systems with improved dry etch resistance. Resist processing has become increasingly more complex, and new assist technologies, such as amine filtration or top and bottom antireflective coatings, were required as feature sizes continued to shrink.

The strong contribution of the chemical "wetware" is all the more impressive if the relative investments in resist and exposure tool development are taken into account. While development of a new exposure tool generation for a new wavelength will cost between US$1.5–2.5 billion (spread over 3–4 vendors), the investment in the development of a photoresist platform for a new wavelength is significantly less. For 248nm, this author's best estimate is probably less than a tenth of the exposure tool figure, and, moreover, this investment is spread over 6–7 resist vendors. For 193nm, where resist companies are only just making the investment in full-field exposure tools, the situation is even more lopsided.

Things become even worse for 157nm lithography, where the industry's exposure tool development cost probably already exceeds US$1 billion, but where the investment in resist development is, at best, only a few percent of that number. Even when staying on the conservative side of these estimates and using the numbers for 248nm lithography (where the discrepancy is the lowest, but also the least likely to be skewed by timing effects), investment in photoresist development is about 10–20¥ more cost-effective in improving photolithographic capability than investment in exposure tool development.

Another comparison that illustrates the low level at which photoresist R&D is currently funded looks at the R&D spending on coating tracks and other tools used to apply and characterize photoresist. Based on the public statements of equipment manufacturers and some admittedly subjective eyeballing, the development of these tools for working with photoresist enjoys about 3–5¥ the funding of resist R&D itself. To this author, who as a practicing photoresist chemist is admittedly biased, the situation has always smacked of putting the cart before the horse.

The size and fragmented state of the photoresist market, combined with the long leadtimes of photoresist development, make it economically impossible for the photoresist suppliers on their own to provide R&D funding levels that are optimal for the industry overall. In the past, the key developments for new resist platforms have mostly come from research groups at large semiconductor manufacturers, who began development efforts decades before the production implementation of a new technology.

Today, many of these research groups have been disbanded or refocused as the semiconductor industry has struggled to cope with changed business conditions. The long-term R&D infrastructure that photoresist vendors have been able to tie into is in danger of evaporating. Efforts such as the resist test centers and development programs of Sematech and Selete have been enormously helpful, but they are not able to fill this gap on their own. For optimum effectiveness, the industry will need to find new mechanisms to fund resist R&D more directly and beyond the pre-competitive stage.

References

  1. R.R. Dammel, Proc. SPIE, 4690, 1–10, 2002.
  2. R.R. Dammel, J. Microlith. Microfabr. Microsys., 1, in print, 2002.

Ralph R. Dammel is director of technology at AZ Electronic Materials, 70 Meister Ave., Somerville, NJ 08876; ph 908/429-3533, fax 908/429-3634, e-mail [email protected].