Patterning the World: The Rise of Chemically Amplified Photoresists

X-ray of tBOC photoresist. Courtesy Hiroshi Ito.

By David C. Brock

In the late 20th century increasingly powerful and numerous personal computers and interconnected networks thereof were at the center of shifts in work practices, communications, and cultural production that collectively became known as the digital age. These personal computers were in no small part defined by two key types of electronic components: the microprocessor and the dynamic random access memory (DRAM). These components in turn were both species of silicon integrated circuits, owing both their existence and their growing power to new developments in the manufacturing technology used to create them. Computer chip manufacturers in the mid-1980s were pushing the limits of miniaturization using a variety of innovative manufacturing practices. The rise of the digital age depended on new materials and techniques that could both increase performance and drive down cost

For decades the semiconductor industry had used photolithography to build integrated circuits on wafers cut from large single crystals of the element silicon. In the patterning process of photolithography, a polymer film called a photoresist is deposited over a thin film of one of a variety of materials deposited atop a silicon wafer. Next, in a complex (and expensive) apparatus known as an exposure tool, light of a very specific wavelength is projected through a pattern-bearing mask onto the photoresist. Regions of the photoresist exposed to the light undergo chemical changes, making them either more or less susceptible (depending on the process) to being removed in a subsequent chemical developing process. Thus the pattern of the mask is transferred to the photoresist. The pattern from the photoresist is then transferred to the underlying thin film through a subsequent process of chemical etching. Multiple iterations of this thin-film patterning process, along with several other physical processes, produce integrated circuits. The photoresist is at the center of the photolithographic process, just as film used to be the crux of photography.

In the late 1970s photolithographic procedures used light from the near-ultraviolet (UV) and mid-UV ranges at 365 and 313 nanometers (nm), respectively. Manufacturers realized that moving to a shorter wavelength, the so-called deep UV, at 248 nm or less, would allow even smaller patterning of integrated circuits, thereby continuing the dynamics of miniaturization, exponential increases in functionality, and dramatic decreases in cost that characterize Moore’s law. Making the leap to deep UV would require dramatic materials innovations and a sea change in photoresist technology. An entirely new breed of photoresist—chemically amplified (CA) photoresists—created within IBM in the early 1980s for just this purpose would eventually come to dominate global semiconductor manufacture. More recently, a later generation of chemically amplified photoresists tuned to 193-nm light has continued to enable Moore’s law. For nearly two decades CA photoresists have stood behind the digital age, largely unrecognized and undeservedly so.

Pushing the Limits at IBM

Some commentators describe the digital computing business in the late 1970s as divided into halves, with IBM on one side and all the other companies on the other. Despite thriving competitors in the minicomputer business and the appearance of the very first personal computers, IBM dominated the computer industry with its broad offering of mainframe and mid-range computer systems, largely produced by captive suppliers within IBM. Large semiconductor fabrication operations in East Fishkill, New York, and Burlington, Vermont (among other locations), produced integrated circuits as logic and memory components. Many of the materials for these semiconductor fabrication plants, or “fabs,” came from additional operations in East Fishkill. In San Jose, California, a disk-drive manufacturing facility boasted a research laboratory. On the East Coast, Yorktown Heights, New York, was the site for the firm’s research and development headquarters.

Throughout the 1970s IBM produced its own photolithography equipment. As the decade drew to a close, however, IBM began to purchase significant numbers of sophisticated and expensive optical devices from the outside, particularly the Micralign lithography tools produced by the venerable optics house and chemical-instrumentation manufacturer PerkinElmer. IBM’s production facilities for advanced semiconductor components contained hosts of self- and PerkinElmer– produced lithography “tools.” These capital goods represented an enormous expenditure, with each tool having cost hundreds of thousands of dollars. In the same period, the fate and future utility of these existing tools were being seriously questioned within IBM.

By the time the 16K DRAM generation was launched in 1977, semiconductor memory was well on its way to displacing magnetic core memory as the dominant memory technology for digital computers. DRAMs were considered the shining examples of so-called large-scale and even very-large-scale integrated circuits in which huge numbers of components were squeezed onto tiny chips of silicon using the latest manufacturing technology, yielding expanded memory functionality at declining costs. Magnetic core memory, in contrast, hailed from the 1950s and consisted of great grid-like planes of wires with small metal rings at each intersection: think of the screen in a window, with a miniature washer around the corner of each little square. The magnetic states of these rings, or “cores,” represented the digital language of zeros and ones. First introduced in 1970, DRAMs were beating out cores on both performance and cost just six years later.

The success of DRAM depended on the semiconductor industry’s ability to push its manufacturing technology to the limits. Indeed, DRAM production became the bellwether for such technology. The semiconductor industry, led by Intel, had established a metronomic pattern in which the industry launched a new generation of DRAM with four times the capacity of the previous generation—1K, 4K, 16K—every three years. Each generation required a new level of miniaturization, thereby creating a fundamental link between DRAM generations and manufacturing technology.

In 1977 a looming question for the semiconductor industry was whether or not the existing lithography tools for the 16K DRAM generation could be used again for the upcoming 64K DRAM generation or perhaps even for the 256K DRAM generation. The ability to form smaller features depended on the wavelength of light used in the tool: the smaller the wavelength, the smaller the possible features. The existing lithography tools used 365 nm light in the near-UV region to expose patterns onto silicon wafers coated with photoresists. Could the existing lithography tools and photoresists be modified to work with smaller wavelengths of light? The economic consequence of the answer was significant. Millions of dollars could be saved if the useful life of the manufacturing equipment could be extended.