by Aung Kyaw Myo and Prarthana Srikanth
On May 30, 2013, Dutch microelectronics firm ASML completed its $3.7 billion cash-and-stock acquisition of San Diego-based Cymer. The deal – financed in part by industry leaders Intel, Taiwan Semiconductor Manufacturing Co., and Samsung – could prove pivotal in the development of the next generation of microprocessor fabrication technology. At stake is industry leadership in the face of upstart technologies and the continued exponential growth in processing power dictated by Moore’s Law.
During the 21st century, historic advancements in telecommunications have spurred change on the macro level. According to a recent study by IDC,1 49.4 percent of the U.S. population has a smartphone, and about 67.8 percent of the population will use smartphones in 2017. There are more than 400 million devices connected to the Internet in U.S. homes, averaging more than one device per person, according to the NPD Group.2 One-third of the world’s population was online in 2011, and 87 percent of the global population has a mobile phone, according to a report from the International Telecommunication Union3. These giant booms in connectivity are made possible by advancements in semiconductor manufacturing processes that take place on the nano level.
Since the first commercially available microprocessor, Intel 4004, was released in 1971,4 the fabrication processes for silicon chips have changed dramatically. Gordon Moore, Intel’s co-founder, famously said, “The number of transistors incorporated in a chip will approximately double every 24 months.” His prediction has become known as Moore’s Law, and it has been realized along the way via technical breakthroughs by the semiconductor industry. According to Intel, “The original transistor built by Bell Labs in 1947 was large enough that it was pieced together by hand. By contrast, more than 100 million 22-nanometer tri-gate transistors could fit onto the head of a pin.”5
Twenty-two nanometers is the current standard in the semiconductor manufacturing process. A nanometer is one millionth of a millimeter, and those tiny transistors are designed using light in a process called photolithography. There are several types of lithography technologies. Currently, 193-nanometer immersion lithography, called deep ultraviolet (DUV), is used to manufacture 22-nanometer chips. The distance of 193 nanometers is the wavelength of the light source, and because its wavelength is long when compared with the size of a transistor, expensive improvements such as double patterning or triple patterning are required. Semiconductor manufacturers are begging for a better lithography technology to make Moore’s prediction true. Extreme ultraviolet lithography, or EUV, at a wavelength of 13.5 nanometers, was supposed to be that technology – ready for commercialization by 2009. After four years, in 2013, it is still not ready for high-volume manufacturing. To accelerate the research and development process of EUV, ASML Holding from the Netherlands on May 30 acquired San Diego’s Cymer for $3.7 billion in cash and stock.
Founded in the Netherlands in 1984, ASML is one of the largest suppliers of photolithography systems in the world by revenue.6 At the end of 2012, ASML employed more than 8,000 employees and had customers in 16 countries. Wafer scanners are used to print chips onto silicon wafers. ASML’s scanners use light sources from both Cymer and Gigaphoton, and its major competitors include Canon and Nikon. It has a wide range of customers such as Intel, Taiwan Semiconductor Manufacturing Co. (TSMC), and Samsung.
ASML’s revenue reached its highest point at 5.65 million euros in 2011. It has been selling scanners with DUV technology and has been working on EUV technology since 2006. In 2010, ASML reached a key milestone by shipping six NXE:3100 first preproduction systems using EUV to manufacturers and labs for testing. Five NXE:3100 machines use Cymer’s light source for EUV. The remaining machine uses a light source from Xtreme Technologies, which is owned by Ushio, a Japanese company.
San Diego-based Cymer Inc. is the maker, supplier, and marketer of superiorquality UV light sources that are used in microfabrication to design performanceenabling patterns on semiconductor chips. The company supplies deep ultraviolet (DUV) light sources and is working to develop extreme ultraviolet (EUV) light sources for chip manufacturing. Its lithography tool manufacturing customers include Canon, Nikon, and ASML, which has been Cymer’s top customer.
The company was founded in 1986 by Dr. Robert Akins and Dr. Richard Sandstrom, both graduates of UC San Diego. Cymer has more than 1,200 employees worldwide and has field service offices around the world. It also offers an OnPulse service for installed base products so that light sources can maintain production goals for customers. Cymer also had silicon crystallization process tool products, which were used to manufacture LCD and OLED displays, but it discontinued new product development in this business in January.
Gigaphoton, based in Japan and owned by Komatsu, competes with Cymer in the DUV light source segment. Gigaphoton and Xtreme Technologies are also competing with Cymer in developing EUV technology.
With rapid changes in the semiconductor industry, ASML needs to keep up with technological innovation. Its systems are very expensive (with an average price of 22.4 million euros per unit sold in 2012), and it only sells a relatively small number of machines annually. Facing intense competition from alternative technologies in the race to continue Moore’s Law, ASML needs to advance development of EUV as soon as possible so it can maintain its position in the industry. It has been working intensely with Cymer for about two years; to speed up the integration of light sources with its EUV system, ASML acquired Cymer.
Between July 9, 2012, and Aug. 27, 2012, ASML entered into an agreement with Intel, TSMC, and Samsung for a “Customer Co-Investment Program” to accelerate development of EUV lithography.7 ASML sold 23 percent of its shares to those customers for 3.85 billion euros in cash. That transaction essentially provided the company the necessary cash to acquire Cymer. This acquisition could be pivotal for the advancement of technologies in semiconductor manufacturing; ASML made a very important strategic decision. However, there are a lot of challenges ahead of the firm, and if ASML fails to get EUV technology ready before competing technologies mature, it might have wasted $3.7 billion.
Competition to release more efficient chips remains heated. Intel, an industry leader, has already released a 22-nanometer chip for its latest generation of processors. Intel plans to release a 14-nanometer chip in late 2013 or early 2014, and the industry expects to manufacture a 10-nanometer chip by 2015 or 2016. Research on 10-nanometer chips is going steady. Creating a 10-nanometer size could involve a number of experimental technologies, potentially based around photonics, graphene, and EUV lithography.8
Photolithography, as defined by Cymer,9 is the process by which semiconductor circuitry is patterned on silicon wafers. There is a light-sensitive material called photoresist on the wafer, and the deep ultraviolet (DUV) light emitted from the lithography light source exposes the photoresist material through the masks. The wavelength of the light limits the resolution of the lithography systems and hence the size of the transistors. Deep ultraviolet for lithography is invisible to human eyes and is generated by excimer light sources. A class of molecules called “excited dimer” exists only in an excited state, which lasts for a very short period of time; it does not exist in a stable, non-excited state. That is the origin of the term “excimer,” which, in turn, is the origin of the name Cymer. Two types of excimer lasers are currently used for DUV: a krypton fluoride laser (KrF laser) at a wavelength of 248 nanometers and an argon fluoride laser (ArF laser) at a wavelength of 193 nanometers. Lasers with shorter wavelengths do not yet have enough power for production.
Immersion lithography enhances the resolution of photolithography. Current immersion lithography tools use highly purified water and achieve 45-nanometer nodes using DUV.
To achieve a smaller transistor size to fulfill Moore’s Law, double patterning is used. As the name suggests, the wafers are patterned twice to achieve the necessary size of the pattern. Because the wafer needs to be patterned twice, double patterning has less productivity than single patterning. In order to create smaller transistors before EUV arrives, some firms are working on multiple patterning. Double patterning and all other types of multiple patterning are considered expensive work-arounds that push the limit of the resolution the light source can achieve.
Instead of emitting photons as in photolithography, electronbeam lithography (e-beam) emits a beam of electrons in a pattern on the resist. It can be used to manufacture chips or create nanotechnology architectures. However, e-beam takes too much time to expose the entire wafer and therefore throughput is low. As the next generation of photolithography, EUV lithography uses extreme ultraviolet light with a wavelength of 13.5 nanometers. Because of its extremely short wavelength, EUV is absorbed by traditional lenses; therefore reflective (opaque) multilayer mirrors are used to concentrate the light. The light is also absorbed by any gas in its path; therefore the light has to travel through a vacuum chamber. It also has stricter requirements on masks and photoresist, both of which are essential components of chip-making.
The development of EUV lithography has missed a number of milestones and opportunities so far. It was supposed to be ready for 32-nanometer processes in 2009. Not only was that milestone missed, but EUV also was not ready for 22-nanometer processes and might also miss the chance for 14-nanometer processes, which are expected to be ready in 2014. EUV was supposed to reduce the cost of manufacturing, but because it has missed all those milestones, it may have lost its cost-effectiveness for production of smaller transistors. There is a consensus that EUV is necessary for the future and therefore it must be developed for smaller-than-10- nanometer-node production.
TSMC, Samsung, and Intel have invested in ASML for the development of EUV technology. TSMC has already installed an EUV scanner, but it is also supporting the development of e-beam as a backup. Both EUV and e-beam have faced delays and multiple challenges and may miss the 10-nanometer mark. Some industry experts say that EUV may not be used until 7-nanometer processes or later.10 Currently, chip makers have no choice but to use 193-nanometer immersion lithography with multiple patterning. The industry forecasts that if such delays continue, chip makers would have to use 193-nanometer immersion with multiple patterning at 7 nanometers.
While Intel has invested $4.1 billion in ASML for R&D funding and equity for development of technologies including EUV, it is already planning to move ahead with multiple patterning using DUV probably down to 11 nanometers.11 Even though multiple patterning is expensive, the smaller transistor results in a cheaper cost per transistor and, therefore, more transistors per chip for the same or cheaper price. From what it seems, the industry is preparing to continue with Moore’s Law with or without EUV.
A first-generation EUV system was delivered by ASML and Cymer for research and development in 2009, and six second-generation EUV systems, called NXE:3100, were shipped in 2010 and 2011. However, none of the systems have reached the throughput required for high-volume manufacturing. ASML received 11 orders for third-generation systems, called NXE:3300B, and plans to ship them in 2013. These are still not expected to be ready for high-volume manufacturing. According to a 2009 presentation,12 ASML had planned to deliver an EUV system with a throughput of 150 wafers per hour by 2013, but it is still trying to demonstrate that it can reach 100 wafers per hour. These throughput issues are attributed to the light source that is developed by Cymer. With low throughput, EUV systems are more expensive than the DUV systems with multiple patterning, and that reality could prevent manufacturers from adopting EUV. The price of the NXE:3100 is already twice that of DUV scanners, and EUV systems that would be ready for high-volume manufacturing are expected to cost more than $100 million, compared with $20 million for DUV systems.
Cymer has been struggling with the development of EUV. By the end of 2012, the company had hoped to ship a 100-watt source, but so far it has only generated 50 watts of sustained EUV power in the lab. According to semimd.com,13 a 55-watt EUV source translates to a throughput of 43 wafers an hour. Cymer’s efforts in developing EUV lithography will now be bolstered by ASML’s acquisition. This is good for Cymer as it reduces the risk-difficulty burden. An EUV system needs to reach 250 watts for high-volume manufacturing. That is a lot of improvement to achieve. Aside from power and throughput, defect control is also very important.
Although the acquisition was largely funded by Intel and other major players in the industry, they are looking at options such as multipatterning, a combination of multipatterning, directed selfassembly, limited EUV, and e-beam lithography.14If it becomes difficult to shrink features, improve performance, and lower power per transistor, the industry would then look at other options such as energy scavenging, die stacking, better customization, and differentiation.
There are two different types of light source for EUV, according to ASML’s website: a laser-produced plasma source (LPP) and a laserassisted discharge plasma (LDP).15 With the former, “a high-energy laser fires on a microscopic droplet of molten tin and turns it into plasma, emitting EUV light, which then is focused into a beam.” The latter consists of using “a strong electrical current through a tin vapor to generate EUV photons.”
Cymer has been working on an LPP source, and Gigaphoton (under the Komatsu Group) is also working on an LPP source. Xtreme Technologies is working on LDP sources. ASML has used an LPP source from Cymer in five of the six NXE:3100 systems and used an LDP source from Xtreme in the remaining one. Nikon is working on EUV1 system,16and aside from an EUV system named SFET, Canon is also eyeing nanoimprint lithography and maskless lithography.17
Aside from the competition for EUV light sources and systems, ASML also faces challenges from e-beam lithography from startups IMS Nanofabrication, Mapper Lithography, and Multibeam. E-beam throughput is not yet high enough to be a serious threat to EUV. Another form of competition is nanoimprint lithography. These all still need development time to replace EUV, but if EUV’s development does not mature soon enough, the situation can be quite dangerous for EUV and ASML. EUV has been very expensive to develop. ASML has spent the same amount on the development of EUV as it has on the creation of the previous two generations of lithography technologies. It has been spending close to 600 million euros per year on research and development, and its main priority has been EUV. ASML is also expecting to spend 50 percent more on research and development for the next five years. If it cannot produce an EUV scanner earlier than competitors or before competing technologies get a stronghold, ASML’s acquisition of Cymer could be considered a failure.
One piece of good news is that EUV has more industry support than other next-generation lithography candidates. E-beam and maskless technologies have seen more delays and challenges than EUV. EUV also faces another dilemma – the potential end of Moore’s Law or “Moorepocalypse,” especially for complementary metal oxide semiconductor (CMOS) devices. Many believe the beginning of the end is here. It is common to see headlines such as “Moore’s law savior EUV faces uncertain future.”18 In another sign, chip maker AMD is taking longer than expected to get from a 28-nanometer chip to a 20-nanometer one. This could lead to the end of Moore’s Law and ASML could have additional time to deliver EUV. If no alternative can be found, the industry could wait for ASML and its EUV scanners.
On a larger scale, advancements in semiconductor manufacturing have brought astounding improvements to everyday life such as cars that park themselves, prosthetic hands that function like real ones, laptops that last more than a day with a single charge to the batteries, embedded computing devices in glasses and watches, and flying robots that are only the size of a penny. If Moore’s Law holds, there could be a slew of daring new creations that challenge the imagination. With EUV as the frontrunner to keep Moore’s Law alive, ASML has a burden on its shoulders to meet key milestones. The success of EUV could mean not only continued historical progress for the semiconductor industry but also a sustained innovation trajectory for the global economy.
Aung Kyaw Myo (Rady MBA, 2013) has worked in product development in software and consumer electronics. His post-MBA focus is technology, operations, and entrepreneurship, and his goal is to build businesses with a positive social impact.
Prarthana Srikanth (Rady MBA, 2013) is a life sciences and health care professional. She developed an interest in the high-technology industry after coming to the Rady School of Management. She is an avid follower of tech trends that will shape the life sciences industry