Japan’s EUV Failure
in 2019 asml finally shipped the first high-volume UV system completing a long and Winding Road stretching multiple decades the move completed the semiconductor lithography industry's shift away from Asia towards Europe it was a particularly big loss for the Japanese Japan discovered euv Japanese companies led the lithography industry before asml came along but they were not able to get UV out in time Japan's failed euv development journey is fascinating as it illustrates the incredible challenge of producing commercial-ready euv systems in this video we're going to look at Japan's attempt to achieve euv but first let me talk to you about the asianometry patreon Early Access members get to watch new videos and see selected references for them before their release to the public it helps support the videos since views are so volatile I appreciate every pledge thanks and on with the show as I mentioned euv was invented in Japan in 1986 Dr Hiro kinoshita then working at NTT announced that he had successfully printed a circuit pattern using euv light and a reflective optic system a great achievement but at the time euv was just one of several possible Next Generation lithography candidates others included Electron Beam direct write lithography proximity x-ray ion beam projection Electron Beam projection and 157 nanometers fluorine the Japanese lithography industry studied all of these but UV was not their main Preference they seem to have been leading towards proximity x-ray which was what most early progress reports focused on the Japanese were not alone in this assessment the department of Energy's early euv project was ending without industry follow-up and as late as 1997 sematech had evaluated euv in fourth place with x-ray again in first but the American Semiconductor leader Intel saw demonstrations in Japan and the department of energy labs and liked what they saw with the doe euv Project R D team on the verge of being disbanded they provided Bridge funding to keep the team together and a year later Intel and other American Semiconductor manufacturers founded EV LLC an industry-wide Consortium to help provide the estimated 60 million dollars a year of r d to commercialized euv lithography Intel's interest in euv and euv llc's founding caught the Japanese by surprise and unfortunately due to American restrictions on foreign participation long geopolitical history there Japan's lithography makers Nikon and Canon could not directly join euv LLC themselves however they were allowed to participate in design reviews and working groups the Japanese had to rapidly pivot and go their own path in 1998 a year after euv llc's founding Japan's Ministry of international trade and industry or midi began its own euv lithography program as originally planned the Japanese eov lithography development plan had three phases in the first phase we need to develop the basic Technologies behind UV lithography and its three system parts the exposure tool the mask and the resist this would go from 1998 to 2001. the government-funded research organization in charge during euv phase one was the aset UV laboratory aset stands for the association of super Advanced electronic Technologies and was a multi-year research Consortium fully funded by MIDI in the second phase another organization would take over to turn aset's raw technology into a working prototype Japan semiconductor industry would contribute partial financial support here this second phase would span three years from 2001 to the end of 2003. it was not exactly clear at the time however who would spearhead this part of the process finally the third year phase takes us into 2006. this is when the Japanese semiconductor industry would turn that prototype into a commercialized product presumably salute as well as the Japanese lithography makers Nikon and Canon would be in charge of this salid or semiconductor Leading Edge Technologies is a Japanese r d research Consortium their 90 million dollar annual budget is funded by industry so they focus on more Applied Technologies aset had a formidable task ahead of it I've done a few other videos about the three basic systems within the lithography machine that includes the exposure tool the mask and the resist and what they do but let's get a quick recap to get everyone up to speed the exposure tool contains an eov light source and an optic system it fires high powered light which then reflects off the mask which has the chip design printed onto it the pattern light then Falls onto the wafer which is coated with the photosensitive chemical known as the resist the resist reacts to the light so that we can use it as a stencil to later etch the wafers while the UV lithography Machine's three basic systems still shared a philosophical Heritage with its predecessors the actual engineering required brand new substantial work and resources were limited for the exposure tool aset needed to develop the basic Technologies behind a large variety of the tools subsystems this not only includes the light source and the light optic subsystems some of the euv's most charismatic Parts but also some less sexy yet nevertheless vital bits like the mechanical wafer and mask alignment technology subsystem which aligns the wafer and The Mask within tolerances of half a nanometer and the Optics Metrology subsystem which measures the mirror's performance during the exposure steps faced with hard choices aset chose to focus their limited resources on Optics manufacturing like the work of adding The Mirror's reflective multi-layers and cleaning them as well as Metrology Asad choosing to work on the Optics meant that it could not work on the exposure Tool's light source an essential piece of the puzzle this task was reassigned to a new industry organization founded in 2001 the extreme ultraviolet lithography system Development Association or Eva the organization included both Academia and Industry five Japanese equipment companies ushio Canon Nikon Komatsu and gigafoton and four semiconductor manufacturers Toshiba and EC Fujitsu and Renaissance on the industry side for the Academia side we had himeji Institute of Technology Kumamoto University Miyazaki University and Osaka University for this reason the Ministry of Education culture sports science and technology which oversees the country's universities had to get involved people proposed two possible technological approaches to producing UV light the first was a laser-produced plasma approach where you fire a laser at a material to turn it into plasma which emits UV light this was the approach that asml eventually chose the second was the discharge produced plasma approach you first create a cloud of material then within that cloud you fire off high powered Sparks between two electrodes to create the light emitting plasma UVA researched both approaches with the laser approach being researched at the hero Scott Research Center and the discharge approach being researched at the gotemba research center in this early phase aset also handled the eov photoresist the major challenge with the euvresist has to do with euv light's high absorbability for traditional resist materials the UV light can only travel about 700 angstroms deep into the resist layer this Falls literally short of the 4 000 angstrom depth that the semiconductor industry is traditionally familiar with to produce their existing patterns so the industry has to develop a new suite of Technologies to accommodate these are called thin layer Imaging or tli for this a set collaborated with the himeji Institute of Technology if you recall the mask contains the chip design aset needed brand new subsystems for producing mirror mask blanks which are masks without any patterns on them adding patterns to those blanks and cleaning those blanks surfaces critically they needed to be able to consistently produce defect-free photo masks this is immensely difficult due to the incredible tolerances involved asets intermediate but ambitious goal was just 0.1 to 1 defects for every Square centimeter but eventually they needed to get to 0.003 defects per
square centimeter furthermore the threshold for a defect on the mask blank is a deviance of just a few nanometers if you want to learn more about how asml tackled this challenge I made an older video about it another reason why this is so hard is because a duv light we previously used to detect such defects cannot penetrate deep enough into the blank's mirror layer we need to use UV light itself to do the inspection the industry phrase for this being actinic light the ministry of economy trade and Industry decided to set up a separate organization just for this part of the euv process the mask and lithography related Metrology inspection group within a research Consortium called Mirai as you might be sensing the Japanese euv effort rapidly expanded beyond the management capacity of a single research institution by 2005 a critical point in the program's history there were four research consortia focusing on their own part of the euv machine aset handled their resist contamination control and making the multi-layer mirror masks UVA handled the exposure tools light source as well as some of the Optics another consortia called leading project handled the inter-university collaboration surrounding light sources and finally Mirai handled the max defect control detection subsystems to say the least this was a complicated setup in contrast the American euv lithography project had a single entity overseeing early commercialization the aforementioned UV LLC UV LLC went six years from 1997 to 2003. after which they passed on to asml which took over from the American company Silicon Valley group lithography by the program's end the UV LLC team had completed all of their objectives and had a prototype that kind of worked the engineering test stand the American and Japanese programs shared information with one another so the Japanese benefited from being able to follow the Americans footsteps towards UV but the Americans were clearly in the lead in 2005 Japan's UV development plan was for high volume UV lithography to hit the 45 nanometer process node in 2010 five years away some parts of the project hit their schedules in 2005 Mirai had already produced an actinic blank inspection tool capable of detecting UV light scattered by blank defects so this product was essentially ready to go but what about the actual euv exposure tools and their components everything was waiting on them this Monumental task fell onto Canon and Nikon Japan's leading lithography makers the two wanted to start with producing a small field prototype kind of like a pre-alpha machine field meaning the area of projected light with the small field prototype the field was just a minuscule 200 micrometers by 600 micrometers making the device more for testing other parts of the euv system like the resist and the mask for their part the Americans of the Lawrence Livermore and Lawrence Berkeley National Laboratories working in conjunction with Carl Zeiss delivered this in mid-2001. their small field device was called the micro exposure tool or met after this you produce a more ambitious full field prototype the alpha machine this one has a field size that or near the production level field size of 26 millimeters by 33 millimeters starting off in 2005 Japan needed to move extremely fast in order to stay competitive and hit the 45 nanometer process note in 2010 five years away and euv complete system would have to be ready by 2008 or 2009. this means having
a working pre-alpha machine the aforementioned small field device by the end of the year 2005. a working Alpha tool a year later sometime in 2006 and an iteration of that Alpha tool the beta tool ready by 2007. a very optimistic schedule but all very necessary to catch up to asml which at that time was putting the finishing touches on their Alpha demo tool two such tools will be delivered in August 2006 to the Belgian Research Institute imec and the State University of New York between Canon and Nikon Canon was the Junior and lacking partner when they started on their UV efforts they had not yet mastered 193 nanometer immersion lithography the most advanced technology just short of euv in 2006 Canon and UVA completed their small field prototype UV lithography tool called the sfet and installed it at an r d Center for testing it generates euv light by firing a 248 nanometer Krypton fluorine laser had a jet of xenon gas at that time the company believed they would be able to start producing their full field 6 mirror prototype called the vs2 starting in 2009. completion was scheduled for 2011. Canon started designing this vs2 but struggled to resolve core euv problems like raising the UV power level producing defect-free masks and extending the lifetime of the optic system problems Nikon faced as well this was in the midst of the global financial crisis Canon then the world's largest digital camera maker was heavily affected with profits shrinking over 60 percent the company delayed plans for a new camera plant in Japan and in 2008 total lithography r d expense was north of 600 million dollars a year so the company probably decided to cut back the market eventually recovered enough for the camera plan to move forward the Canon never built the S2 as far as I know it would go no further in the lithography technology chain and as of this writing the company still only sells krf non-immersion or dry lithography systems Nikon was ahead though as the world's only producer of immersion lithography other than asml they were Japan's lithography leader from 2001 to 2004 Nikon worked with aset to produce three iterations of their small field prototype euv tool nickname Hina for high numerical aperture tool finally in April 2006 Nikon felt ready enough to produce their first full field exposure tool a six mirror system called the euv-1 this Alpha device was completed and installed in the r d Research Center at the Japanese semiconductor Consortium salite going live in January 2008.
this progress was about one and a half years behind asml's 2006 Alpha demo tool based on the 2005 timeline the beta euv device should have been done by now but shipping uv-1 in the first place was still a substantial achievement without ambiguity we can call uv-1 Asia's most advanced lithography machine but it was also definitely not commercially competitive uv-1's late installation pushed the UV high volume ship schedule further back from the 45 nanometer process node to the 32 nanometer node there was even talk of the 28 and 22 nanometer process nodes this presented a major problem because the more advanced the process node the more power the UV machine and its light source needed to have and the uv-1's light source power output was one of its biggest drawbacks the uv-1 used a discharge-produced plasma or DPP Contraption produced by the Japanese lighting company ushio and its subsidiary the germany-based Extreme Technologies as I briefly mentioned earlier the concept behind this DPP approach is the blast a lightning bolt through a vapor of xenon to generate UV photons the DPP approach is said to be more power efficient be physically smaller and supposedly cleaner than laser-produced plasma but it struggles to Output as much power as this technological competitor and the less power the light source has the longer the wafer needs to be exposed to that light source slowing down throughput the uv-1's light source power topped out at about 100 Watts the intermediate goal was 110 to 180 watts and the eventual goal for high volume production was 250 Watts this would require a big system change swapping out Xenon for 10. as envisioned this tin base DPP UV light source module would have two wheel shaped electrodes about 20 meters in diameter lined up together each electrode is suspended over a bath of molten tin would induce the electrode spin around such that the edges barely touched the Bath's surface picking up a layer of molten tin atoms a low power laser then fires rapid pulses at the wheel's Edge vaporizing the tin and forming a cloud of tin atoms floating around the wheel electrodes then discharge a massive current that causes the cloud to collapse onto itself like a dying Star Heating it up to 200 000 degrees Celsius the collapsed tin Cloud then emits UV photons the contraption was indeed quite Nifty but it still struggled with contamination issues as well as being able to produce enough power to become economically viable by the time the project ended they had not yet overcome this asml also tried the DPP approach themselves but eventually halted all work on it in 2013. power levels would not reach sufficient levels until asml pioneered its laser double shot method considering this it is likely that Nikon would have swapped its light source over to the laser-produced plasma approach as well the main supplier probably would have been gigafoton a specialty laser maker which has been working on such a light source since 2002. tests at salite conducted in partnership with Intel showed that the uv-1 can produce beautiful vertical lines space between 26 and 33 nanometers wide however the device struggled with contamination control UV light continually hitting the mirror's surface created carbon contamination slowly deteriorating The Mirror's reflectiveness I am strangely reminded of this scene in one of my favorite movies Godzilla versus bayalante modern day masterpiece Godzilla fires his Atomic breath at a mirror in the flying drone Super X2 causing contamination and damaging The Mirror's reflectivity who knew Japanese kaiju movies were so technologically accurate anyway the introduction of pure oxygen to the mirror surfaces as well as the use of UV dry cleaning with a low pressure mercury lamp help clean this carbon contamination however the euv team then struggled to deal with a mysterious contamination stemming from a silicon carbon compound called cyloxane going into 2011 Nikon's researchers were still struggling to overcome these widespread contamination issues after the uv-1 successful testing Nikon would proceed on to the next euv tool called the euv2 the uv-2 was scheduled to be completed in the 2011-2012 time period in time for 32 nanometers however it became clear that asml was pulling ahead with a substantial lead stretching from 18 months to as much as two years and with the global financial crisis in full force in 2009 Nikon cut their r d budgets rumors floated around that Nikon was putting their EV development entirely on hold a spokesman denied it but did clarify that the uv2 would remain an in-house unit what would come out in 2012 instead would be the euv3 2010 was another hard year with market sales for lithography machines in the chip and LCD Industries Falling by half year over year the company continued nevertheless presenting its internal Research into 2011. that year the company had a chart indicating that they were targeting the 16 to 11 nanometer process nodes so about 2016. but this would be the last significant report I was able to dig up about Nikon's progress on either euv2 or uv3 in July 2012 Intel announcer had agreed to invest up to 4 billion dollars into asml in exchange for a stake of up to 15 percent Nikon shares fell seven percent after the agreement was announced Intel had the Nikon's second biggest customer after Samsung the American Giant is known for sourcing the lithography machines from multiple vendors but the market evidently realized that this investment agreement meant that intel was moving forward with asml alone in 2011 Nikon spent a combined 28.4
billion yen or 355 million dollars on R D and capital expenditures for its lithography Technologies Nikon probably decided that without Intel as a Cornerstone customer it was not worth spending hundreds of millions of dollars a year in r d and capital expenditures to continue developing euv quietly Nikon's UV efforts tailed off uv3 never made its 2012 ship date with this asml became the last man standing in the UV lithography industry Japan's UV efforts were not totally in vain many component efforts turned into viable technologically advanced businesses for instance we still have the actinic blank inspection tool from Mirai and its collaboration with the Japanese company laser tech Japanese companies also have dominant positions in the UV photo resist space with jsr shinetsu and Tokyo Oka kogio the biggest suppliers Japanese equipment maker Tokyo electron also makes the equipment to apply those euv photo resists all that being said however I am sure that the Japanese would have preferred to have those 150 to 300 million dollar UV machines to be produced in the country as well the story of Japan's euv failure is utterly fascinating I had expected a debacle but what I found was a close race Nikon had substantial problems with the uv-1 but they were not deal breakers in the end what stopped them was economics and competition not technology my very rough back of the envelope calculation of the total cost of Japan's UV effort from 1999 to 2012 is somewhere between 5 to 10 billion dollars this is taking into account Nikon and Canon's r d expense over that time the 90 million dollar annual budgets of aset and the four consortiums and a fudge factor to account for Investments made by other semiconductor makers not mentioned here by comparison the original vlsi project that remade Japan's semiconductor industry and spawned its lithography dominance back in the late 1970s costs about 150 million dollars in total so this was an exponential jump if another government wanted to build their own independent euv machine for whatever reason that might be I reckon it would likely require at least five to ten billion dollars direct guidance and information sharing from the European and American UV makers and about 10 to 15 years something to think about all right everyone that's it for tonight thanks for watching subscribe to the Channel Sign up for the newsletter and I'll see you guys next time