Patterning the World: The Rise of Chemically Amplified Photoresists

Willson and Ito turned to another polymer that Fréchet had worked on earlier at IBM San Jose during his sabbatical there in 1979: poly(p-hydroxystyrene), or PHOST. PHOST is a styrene-based polymer, chemically similar to the Novolac resins used in conventional photoresists. Willson suggested modifying the polymer to include a new side chain: tertiary butoxycarbonyl, or tBOC. The resulting polymer was poly(p-t-butyloxycarbonyloxystyrene), or PBOCST. Willson, who had worked mainly in biochemistry before joining IBM, was aware that tBOC—a mainstay in peptide work—was susceptible to cleavage from the basic polymer through the action of both heat and acid. Willson and Fréchet recall early, inconclusive attempts by Willson and several coworkers to make a PBOCST resist based on acid-catalyzed cleavage of the tBOC groups using photosensitive orthonitrobenzyl esters to produce the acid. From Ito’s perspective the PBOCST work was dormant when he reached the lab. However, Ito also began investigations of photoacid-catalyzed cleavage of a different tBOCprotected polymer as a potential basis for a chemically amplified resist. Looking at Ito’s results, Willson and Ito decided to pursue a hybrid course: mixing PBOCST with the onium-salt PAG. The result of this mixture—the brew resulting from the experiences and interests of Fréchet, Willson, and Ito—stopped the researchers in their tracks.

The tBOC resist displayed dramatic chemical amplification. After exposing the tBOC resist to 248-nm deep-UV light, the resist-coated silicon wafer was heated in a post-exposure bake. The acid generated by the onium salt catalyzed the cleavage of the tBOC groups. The resulting fragments then generated additional acid, catalyzing further tBOC cleavages in a cascade of de-protection. The reaction was both extremely fast and extraordinarily sensitive to the deep-UV light. At the beginning of his search for a CA resist Willson knew that he needed a 30-fold improvement in sensitivity over conventional resists. With the tBOC resist, Willson, Fréchet, and Ito had generated a 100- to 200-fold improvement.

By 1983 Willson was confident enough in the new tBOC resist to promote it within IBM. At East Fishkill he presented it to a collection of researchers and engineers from a variety of IBM sites, including representatives from East Fishkill’s own photoresist operation and staff from the cutting-edge fab in Burlington. John Maltabes, a lithography engineer from the Burlington plant, had been helping develop a manufacturing process for a 1M DRAM using deep-UV radiation to meet a “1 micron design rule.” Deep-UV lithography would be used to produce features as small as 1 micron on the new powerful memory chip. Maltabes had been evaluating the possibility of replacing the mercury lamps within the PerkinElmer lithography tools in Burlington with excimer lasers. But Willson’s tBOC presentation persuaded Maltabes that using the new photoresist with the existing mercury lamps was the better strategy: when he returned to Burlington, Maltabes tried to convince his supervisors to kill his project. Three months later they did just that. Maltabes’s new job would be to help implement the tBOC resist for manufacturing the 1M DRAM.

Something in the Air?

IBM had staked the future of its cutting-edge products on CA photoresists. The advantages were tremendous: the tBOC resist could save IBM millions of dollars in modification and replacement of its existing lithography tools. The downside was the uncertainty that the new resists would work in an active manufacturing environment.

Production trials at Burlington, however, revealed new, unanticipated problems with the CA resist. For one, its sensitivity varied widely. After eliminating the lithography tools as the source of this unpredictability by installing new, exacting filters, the blame rested squarely on the tBOC resist. Eventually, the production engineers in Vermont resorted to the kind of highly empirical “black magic” practices that characterized much of semiconductor manufacturing in its early years. They did not know why certain things worked, only that they did. The engineers found, for instance, that letting silicon wafers that had been coated with the tBOC resist sit for several hours in the factory before exposing them stabilized the sensitivity, but at a lower level.

More troubling was the occasional formation of “skins” in the uppermost layer of the tBOC resist. These skins were regions of the photoresist in which sensitivity had catastrophically collapsed. Exposed regions of the resist near the surface would not develop properly and thus formed a skin that could not be removed by the solvent. Puzzlingly, these skins were all at the surface of the resist. Regions of the resist directly below these skins developed perfectly. The issue was serious: these skins would result in fatally defective DRAMs.

The groups at San Jose, Burlington, and East Fishkill were troubled by the new resist’s difficulties. Maltabes recalls a lunch conversation in San Jose about these issues in which a researcher who had experience manufacturing disk-drive systems suggested that these troubles stemmed from “something in the air.” This researcher and his colleagues had attributed certain failures of disk-drive systems to airborne contaminants and had used air-filtration systems with activated charcoal and HEPA filters to get around the problem. Surplus filtration units sat in a warehouse, and he offered them to the tBOC team.

Maltabes and Scott McDonald from Willson’s team returned to Burlington with the surplus units. With a series of experiments the pair determined that in filtered-air environments, and indeed environments of air pumped in from outside the fab, the skins disappeared and the resist sensitivity was both high and consistent. The atmosphere of the fab itself harbored contaminants that were responsible for the problems with the tBOC resist. With pressure mounting to get the 1M DRAM into full production, Burlington decided to filter the air rather than hunt down the unknown contaminant or contaminants. Once wafers were coated with the tBOC resist, they remained in a filtered-air environment until they entered the lithography tool.

By 1986 1M DRAM production was in full swing. IBM manufactured several million of these DRAMs, all dependent on the CA tBOC resist. Reflecting the criticality of tBOC resists to the success of this project in moving IBM to the first deep-UV manufacturing technology, the firm kept the tBOC resist as a proprietary material and the use of filtered air as a closely held trade secret into the early 1990s. Several million working DRAMs within IBM’s flagship computer products offered powerful testimony: the era of CA photoresists had arrived.

Coda

For IBM, possession of the first CA photoresist conferred significant competitive advantage. By the mid-1990s, however, a combination of accidental and systematic factors broke IBM’s exclusive hold on this class of material. Willson, Fréchet, and Ito had patented the tBOC resist in 1982, but the patent was limited to just the tBOC material, not the very idea of a CA photoresist. This limited scope was the product of multiple factors: the large role played by the researchers rather than attorneys in writing the patent; the vagaries of process patenting in comparison with patents on particular materials; and the discovery of “prior art” in the patenting process. One of the developers of onium-salt photoacid generators at 3M, George Smith, had previously patented a photoresist involving a very similar mechanism to the tBOC resist. These accidental factors allowed commercial photoresist producers—inspired by IBM’s success— to bring their own versions of CA deep-UV resists to the market by the early 1990s.

More systematically, CA photoresists escaped IBM as the computer giant participated in the growing trend among semiconductor manufacturers to obtain manufacturing equipment and materials from specialized external suppliers. As IBM came to rely more heavily on lithography tools produced by outsiders, the close coupling of tool with resist meant not only that the tool makers would need access to the best CA resists but that the tool makers’ other customers would also require access. Moreover, specialized photoresist houses had greater resources and incentives for pushing CA photoresists forward. In the mid-1990s IBM actively transferred the second- and third-generation CA photoresists developed by Ito and others to the outside world. In doing so, IBM accelerated future developments in CA resists, empowering the continued evolution of the digital age.