Semiconductor Lithography Economics–fuelling Moore’s law and market power

Race to profit from technology possibilities has been driving the evolution of semiconductor lithography--resulting into rise and fall of firms, and monopolisation

Semiconductor lithography has been evolving due to profit making competition from science and technology possibilities in keeping Moore's law alive--leading to monopolisation
Semiconductor lithography has been evolving due to profit making competition from science and technology possibilities in keeping Moore’s law alive–leading to monopolisation

The rise of the rudimentary microscope that an American firm GCA invented in 1961 to ASML’s $250 million complex machine is a surprise. Its uprising has fueled monopoly in the highly critical semiconductor industry and triggered the chip war. But what is the underlying force? The race to profit from the opportunity of making transistors smaller, better, and less costly has been the underlying force of the evolution of semiconductor lithography–unleashing monopoly. These simple economics also gave birth to Moore’s law.

Sustaining Moore’s law by doubling chip density every 24 months has been the underlying cause of the rise of the semiconductor industry. Technology has been playing a pivotal role in addressing the basics of business economics, as measured in cost-per-good-product out.

Not a single technology deserves full credit. Among the many contributors, semiconductor lithography stands out. It’s the crown jewel of the semiconductor industry. Though it has historical roots in the way of printing using a flat stone (in the 1820s), it has been evolving, stretching the limit of science and engineering–reaching extreme ultraviolet (EUV) lithography.

Underlying force of evolution

The underlying force of the evolution of semiconductor lithography has been economics. Due to the fundamental advantage of profit-per-die over higher-cost alternatives, the race of shrinking feature dimensions through increasing lithography resolution has been moving forward. As a result, advances have been ongoing in major subsystems such as light sources and optics. Besides, to maximize cost reduction through the integration of related modules and automation of tasks, modern steppers are enormous machines. They are far more than a microscope used in the early 1960s. Some major subsystems are the wafer loader, wafer stage, wafer alignment system, reticle loader, reticle stage, reticle alignment system, reduction lens, and illumination system.

Consequential effect

As a result, the cost and size of individual machines have been growing. For example, an ASML EUV lithography machine for a 5nm process node cost as high as $250 million. The shipment of subassemblies of this double-decker-sized machine, weighing 180 tonnes, requires 20 trucks and three fully loaded Boeing 747s. This machine has been the underpinning of the rise of TSMC as a monopoly in high-end chip making. Besides, due to the failure to master it, American icons like Intel have been falling behind. On the other hand, in the mission of attaining semiconductor independence, it’s a must-have tool for China. Hence, it has become a vital weapon in the unfolding chip war.

Birth of semiconductor lithography

The birth of lithography dates back to printing using a flat stone. Wax-coated limestones used to be scratched with a stylus. Dissolving in an acid bath would etch away the parts of the stone not protected by the wax—resulting in a printing plate. In the 1832s, Nicephore Niepce updated this plate-making method by inventing a photographic process. Instead of wax, he used Bitumen of Judea, a natural asphalt, as the first photoresist. As photolithography helps create windows on the substrate, scientists found it practical to add window-specific impurities—resulting in PN or NPN junctions on the semiconductor wafer (substrate).

In the 1950s, scientists at the Bell Laboratories started experimenting with photolithography for producing transistors. The target was to have segment (window) specific doping on the same wafer—forming NPN structure. And their success was followed by Texas Instruments’ Kilby and Fairchild Semiconductor’s Noyes. Subsequently, they invented a technique for producing multiple transistors and interconnections on the same wafer—giving birth to an integrated circuit (IC) in 1959. 

The invention of IC and its production through lithography assisted segment-specific addition of impurities led to making transistors better and cheaper. Hence, the race started to improve the lithography process so that increasingly smaller transistors could be produced for fabricating higher-density microchips. Consequentially, it gave rise to Moore’s law in 1965, doubling Transistor density yearly. Later, Dr. Moore updated it to 18 and 24 months needed for profitably doubling transistor density per mm2 of the wafer or die area.   

Evolution of lithography technology for semiconductor

In the early 1960s, semiconductor lithography technology used to be very rudimentary. It was just like photography, using photoresist-coated wafers as a film to capture the image of the designs (mask) of ICs. Hence, semiconductor companies used their own in-house lithography tools and equipment. But the scope of profiting from higher-performing lithography machines for reducing transistor size and increasing yield led to the demand for growing specialization.  

The lithography market has historically used visible or ultraviolet light to form patterns on the photoresist. But other technologies have also become targets for exploitation. For example, X-ray, E-Beam, and Ion Beams have been tried. Despite the efficacy, they became economically less attractive for the semiconductor industry, where wafers are processed at rates above 120 wafers per hour. Hence, exploiting an increasingly smaller wavelength light source sustained the evolution of the semiconductor lithography market.  

Evolution of light source:

The underlying science for reducing the feature dimension of the image is guided by the Rayleigh formula: CD=K1*λ/NA. Here, the CD is the critical dimension for exposure; K1 is the process constant; λ is the optical wavelength. Whereas NA is the optical numerical aperture of the projection objective. The lower the CD value, the higher the resolution—resulting in a smaller transistor size. Hence, profitable means of lowering the value of CD started, which could be done by reducing K1 and wavelength and increasing NA.  Among these, the scope of improving process constant and numerical aperture is quite limited. Hence, the specialization approach focused on reducing the wavelength.

The lithography process started with the visible light source in the 1960s. Subsequently, equipment makers started moving to lower wavelengths to increase the resolution. In the later 1970s, blue light with a wavelength of 436 nanometers (nm) known as the mercury g-line became a target. A lithography machine using a mercury g-line light source could print features as small as 1 micron (1,000 nm). To keep feature size smaller, invisible ultraviolet (UV) light with a wavelength of 365 nm became a target in the early 1980s.

Increasing demand for smaller feature sizes led to exploiting more minor wavelength light sources, such as krypton-fluoride (KrF) lasers produce light with a deep UV (DUV) wavelength of 248 nanometers (nm). As a result, modern KrF systems can now produce features down to 80 nm. Going further down the DUV led to using argon-fluoride (ArF) excimer lasers that produce light with a wavelength of 193 nm—enabling features sizes of 38 nm to be printed.

Ultimate optical frontier—13.5nm EUV

Exploiting smaller wavelength light sources has led to extreme EV (EUV). Molten tin droplets of around 25 microns in diameter ejected at 70 meters per second are turned into plasma with the laser to produce 13.5 nm light.

Evolution of means of getting the exposure:

In addition to the light source, progress has been made in getting exposure to chip designs on photoresist-coated wafers. In the 1960s, it began with mask aligners, which patterned the entire wafer at once. The mask would contain individual ICs patterned across the mask. This technique requires holding the photomask over the photoresist-coated silicon wafer while a bright light is shone through the mask and onto the photoresist. Improved ones like proximity aligners and projection aligners, subsequently, replaced an early version of contact aligners. Significant manual adjustments were needed to perfect the alignment. However, the continued growth of chip density kept making constructing these complex multi-chip masks very difficult. Hence, in 1975, the industry witnessed the introduction of the first step-and-scan camera by GCA, which simplified the process of making masks.

As the diameter of the wafer kept growing, the feasibility of patterning the entire wafer once started getting infeasible. Hence, in the late 1970s, the process of repeating the same pattern on the wafer emerged.  It’s called stepper, short for the step-and-repeat camera. Unlike aligners imaging the entire surface of a wafer at the same time, producing many chips in a single operation, the stepper images only one chip at a time. Hence, in the beginning, it was much slower to operate. However, as stepper images only one chip at a time, it offers a higher resolution. Hence, immense pressure on increasing chip density and advancement of stepping operation led to phasing out aligners by stepper. Consequentially, by the late 1980s, the stepper had almost entirely replaced the aligner in the high-end market. The stepper itself was replaced by the step-and-scan systems (scanner) to advance resolution further.

Immersion lithography and multiple patterning

Among other advancements, immersion lithography is notable. The replacement of the usual air gap between the final lens and the wafer surface with a liquid medium with a refractive index greater than one results in enhanced resolution.

Besides, using a double set of masks for patterning the same substrate increases the feature density. For example, using immersion and multiple patterning, in 2014, the industry succeeded in reaching 32nm node and 22nm node using 193nm light source. But multi-patterning reduces the throughput, increasing the cost.

Increasing automation and integration:

Once covering wafers with photoresist used to be a manual operation. Over the years, automation kept progressing, getting this task done without human touch. For example, a robot in the wafer loader picks up one of the wafers from the cassette and loads it onto the wafer stage. Similarly, many tasks, from mask placement to wafer handling, have been automated. Furthermore, an increasing number of modules have been added to the essential operation of steppers or scanners. Hence, the number of subassemblies kept growing, increasing the size and throughput.

Rising R&D and unit cost—fueling Economies of Scale

The advancement in light source, projection technique, and automation has also contributed to the cost of the semiconductor lithography machines—reaching $250 million for ASML’s EUV scanner. R&D costs for developing subsequent advanced versions also kept exponentially growing. For example, the R&D cost for GCA’s g-line stepper was $5 million. But the R&D cost for the i-line and DUV stepper rose to $25 million and $140 million, respectively. Increasing R&D costs kept contributing to economies of scale, adding momentum to Monopolistic market power. Hence, ASML’s R&D investment of 6 billion Euro over 17 years in developing EUV lithography machines has led to monopoly status.

Rise and fall of American lithography firms

There has been a growing concern about the loss of American inventions. Among many others, semiconductor lithography, a critical part of the chipmaking process, is a notable invention that America lost. To roll out Bell Labs’ semiconductor photolithography inventions in the market, in 1961, American GCA built the first lithography machine. In the 1970s, American Kasper Instrument and Perkin Elmer became market leaders by introducing remarkable innovations, like alignment, projection, and projection lithography technology. In the 1960s, Kulicke & Soffa introduced contact mask aligners for wafer printing. Another American company Cobilt, formed by Kasper engineers, brought significant innovations in wafer printing. The acquisition of Cobilt by Computervision in 1962 brought automation to Cobilt’s mechanical aligner. In retrospect, the decades of the 1960s and 1970s of semiconductor lithography were dominated by American firms.

But in the technology race for continued improvement, American firms started losing their edge to Japanese Canon and Nikon. Notably, due to superior optics capabilities, Japanese firms started outperforming their American counterparts in the 1980s. The growing dominance of Japanese stepper suppliers was a worry for American chipmakers. Intel’s effort to develop an alternative with a European company, Censor, failed. Subsequently, Censor was sold to Perkin-Elmer in 1984. According to Gordon Moore’s recollection, compared to the big steppers coming out of Canon and Nikon, there wasn’t a comparable piece of equipment from the US firms. Hence, American semiconductor firms like Intel got compelled to keep buying Japanese equipment because it was the best available.

Consequentially, due to the failure of attempts to release better alternatives with the support of public funds, American semiconductor lithography firms started to fold. For example, due to the loss of $60 million in government funds to restore competitiveness, American lithography pioneer GCA closed its operation in 1993.  

Rise of Japanese dominance—making Canon-Nikon duopoly

Because of the similarity of their industries, in the late 1960s, Nikon and Canon of Japan entered the semiconductor lithography market. In the early 1980s, the release of the stepper lithography, the NSR-1010G, took Nikon to a new height. Advanced Optical systems and self-developed lenses began to take away from GCA a series of large customers such as IBM, Intel, and AMD. This success enabled Nikon to equal GCA, each with a 30% market share, by 1994. The remaining 40% was left to Ultratech, Eaton, P&E, Canon, Hitachi, and a few others to divide. By this time, Canon also succeeded in rolling out advanced steppers, gaining market share at the cost of marginalizing American firms.

By the late 1980s, the American photolithography trio had fallen. As a result, Japan’s Nikon and Canon had the lion’s share of the market. And new entrant ASML managed to survive and got 10% of the market share. Such reality underscores the reality that Japan leads by being a late entrant.

By the way, the underlying cause of the failure of American firms in their inventions is not due to low-cost labor advantage or subsidies given by the Japanese government. Contrary to the belief of many experts, Americans fell behind in the lithography Innovation race due to poor performance. Perhaps, this is the underlying cause of the loss of edge in many inventions which America brought first in the market.

Uprising of late entrant ASML as monopoly and get caught in chip war

In 1984, ASML was born in a wooden hut as a joint venture between Dutch Philips and ASMI. Based on the work done by Philips in the 1970s, ASML succeeded in releasing its first system, the PAS 2000 stepper, before its first birthday. But within one year, the product became obsolete due to Japanese firms’ rising innovation power.  

To attain an innovation edge, Japanese Nikon, Tokyo Electron, and Canon had been significant applicants for lithography patents for a long time. Hence, to find a survival and growth path, ASML accelerated R&D leading to the dramatic rise in the number of lithography patents filed by ASML in the 1990s. ASML also started filling patents in Japan, Taiwan, the United States, and South Korea.

ASML’s success in accelerating and leveraging R&D resulted in releasing steppers to compete and take away market share from Japanese Canon and Nikon. Furthermore, due to the urgency of American firms in finding alternative sources and the rapid growth of American semiconductor firms to fuel the PC growth, ASML started rapidly gaining market share. Consequentially, in the late 1990s, semiconductor lithography became the era of the Nikon and ASML duopoly. By 2005, ASML took over both Canon and Nikon in lithography unit shipments.  Besides, by pursuing a EUV light source, ASML succeeded in being the sole supplier of lithography machines for sub-10 nm process nodes. And in the overall market, Dutch firm ASML attained 62% market share, leaving the rest to Canon and Nikon of Japan.

Economics of exploiting science and technology—driving the industry growth and causing rise and fall of business successes

The semiconductor industry has grown from one transistor produced in 1947. In 2022, a single Apple A16 bionic chip has 16 billion of them. The semiconductor industry is estimated to generate $600 billion in revenue in 2022. Besides, microchips have been fueling the Reinvention and Incremental innovation of a growing number of products. The underlying reason is not due to the replication of the invented transistor. Indeed, a $150 apiece transistor, offered by Fairchild in 1957, would not have succeeded in unfolding the reality we have been observing.

The underlying force of the phenomenal growth of the semiconductor industry in fueling reinvention and incremental innovation waves has been due to increasing chip density. The success of doubling chip density every two years, as Moore’s law observes, has been making transistors better and cheaper—making a trade-off irrelevant. But the cornerstone of this remarkable uprising has been the evolution of semiconductor photolithography.

The race to profit from increasing chip density kept demanding lithography machines to print wafers with decreasing feature sizes. To benefit from meeting this demand, lithography firms entered an intense race to exploit science and technology with growing precision. Due to increasing R&D investment needs, economies of scale and scope, smarter firms started gaining market power. As a result, even early entrants could not sustain themselves and left the market. In retrospect, the evolution of semiconductor lithography is a wonderful example of the exploitation of science and technology to offer higher quality at a decreasing cost, leading to gaining market power and succeeding in monopolizing. Apart from ASML’s monopoly in EUV, TSMC has emerged as a monopoly in high-end chip production by leveraging lithography innovations.  

It’s an example of consistently innovating to generate the cumulative effect of incremental innovation—resulting in a big bang.

...welcome to join us. We are on a mission to develop an enlightened community for making the world a better place. If you like the article, you may encourage us by sharing it through social media to enlighten others.
Enlighten others

Subscribe

Subscribe now to our newsletter

You May Also Like

Leave a Reply

Your email address will not be published. Required fields are marked *