A (0:02)

Social media is passing around an announcement that Huawei is testing a China domestically developed EUV machine. The machine uses an EUV light source known as Laser Induced discharge plasma or LDP. This is in contrast to ASML's method which is called the laser produced Plasma or LPP method. It is claimed that LDP is much more efficient than lpp smaller, simpler and better energy efficiency. Has ASML just been deep seeked? I've been asked to speak on this via email and Twitter and I guess I have to do it. There is so little out there about how this machine works, so I'm not going to speculate, but people have tried LDP before and we can talk about that. Feel free to extrapolate from there in today's COPE video. Let's take a look at the Laser Induced Discharge plasma EUV light source. When we are talking about the EUV light source, we're talking about producing 13.5nm light. ASML's LPP based light source is famous. In it, a carbon dioxide laser hits a droplet of tin twice. The first laser pulse flattens the tin droplet into a disk shape and the second laser pulse is more powerful. When it hits the droplet, it creates a tin plasma. The plasma emits the 13.5 nanometer light we want. From there, this light must make its way through a complicated optical system consisting of anywhere from 6 to 10 mirror reflections and unfortunately, each mirror reflects a little less than 70% of the EUV light that hits it. Since so few photons ever make it to the wafer, an EUV light source has to generate a lot of photons. This is measured in units of watts or the number of photons generated per second. The initial goal was something like 250 watts. Go much lower than that and it's like trying to warm a pizza with a dim heating lamp. You wait forever or never get there. ASML didn't automatically choose LPP at the start. In the early 2000s, people knew that LPP had serious problems, like the part that depends on an unstable CO2 laser hitting an unstable near microscopic sized tin target from several meters away. At the time, the community considered this difficult, so vendors explored a range of light generation options. One of those alternate options was dpp. DPP stands for Discharge Produced Plasma method. The DPP method produces EUV light by discharging a strong, very brief electrical current between two electrodes, essentially a lightning bolt. Between the two electrodes there is a vaporized fuel, pre ionized to ensure a more conductive path for the Coming discharge. And to achieve a nice uniform effect, the discharge comes from a powerful capacitor bank. When fired, the current flows in a line along the Z axis between the two electrodes through the fuel vapor, ideally a straight line, but not always. This creates a powerful magnetic field that forms circles around the current like fingers closing around a wire. The field constricts and confines the fuel's ionized particles, creating a plasma inside. The field squeezes it inwards, heating it to higher densities and temperatures of up to 200,000 degrees Celsius. When done right, it creates a plasma hot and dense enough to emit sufficient EUV radiation that we can collect. Because the electric current discharge moves along that z axis, we call it the Z pinch. Does this name sound somewhat familiar? Perhaps you heard of it from when Z pinch was investigated back in the 1950s as one of the first approaches to nuclear fusion energy? Yes. That is how crazy EUV is. We need nuclear fusion energy tech, in a manner of speaking. In the early days of EUV, like the late 1990s and early 2000s, DPP was investigated as a possible light source by two Extreme Ultraviolet Lithography System Development association, or EUVA, over in Japan. This was a large joint organization with participation from various Japanese companies and academics. And the second company was Philips Extreme uv, a joint venture between Philips and the Fraunhofer Institute for Laser Technology. One of the major differences between the two setups in the early 2000s appears to be the choice of vaporized fuels. There are three fuel xenon, tin, and lithium. UVA studied DPP using xenon, While Philips Extreme UV's variant of DPP championed tin. Nobody seriously tried to commercialize lithium, apparently, because lithium atoms and ions are very small. If used as a fuel, lithium atoms risk diffusing into the mirrors and all other solid parts of the tool poisoning them. Xenon fueled DPP sources were used in early EUV prototypes like the Micro exposure tool in 2004. A few others were installed in the R and D labs of intel and Sematech. But in the end, tin emerged as the best candidate. More of its ions produced light bands within 13.5nm, and that light was less likely to be absorbed by its own ions. So in all, a tin setup was expected to convert more of its wall plug power into EUV power, 2% rather than the 0.5 to 0.9% for xenon. DPP is a simple method, and it let Philips Extreme ship a light source by 2003. The main issue with the DPP methodology has to do with repetition. If we want to raise the EUV light source's power level, then we must have many discharges. By some calculations, the electrodes must discharge thousands of times per second to achieve the minimum economic goals of 120 wafers per hour. The pulsed power technology struggled to achieve that kind of speed. Such speeds presented serious thermal damage concerns for the electrodes. You know, from having to deal with thousands of lightning strikes and fusion class plasma temperatures. One more issue. Tin vapor condenses on cold surfaces, necessitating extensive engineering to heat the tool surfaces to keep tin films from forming. For these reasons, in the Mid to early 2000s, Philips Extreme UV evolved their DPP method into LDP. The early LDP setups had two rotating tin supply disks. These disks, connected to large energy storage capacitors, will also serve as the electrodes. The disks rotate through baths of carefully temperature controlled liquid tin, coating their surfaces with a tin film. Then either a pulsed carbon dioxide or neodymium doped YAG laser fires at the tin film on one of the disk surface, creating a tin cloud called a pre plasma between the electrodes. From there, the same thing happens. We fire a big electric discharge to generate the Z pinch and and from there, the hot and dense plasma. For EUV light, this method offers some compelling upsides. LDP's wheel setup sidesteps some of the technology challenges of LPP, namely trying to hit the tin droplet in LPP. Every time the laser pulse misses the droplet, we suffer an efficiency loss. But with ltp, since the wheels are always turning and replenishing their tin films, the the laser is basically guaranteed to hit some tin somewhere on the wheel. So indeed, there are efficiency benefits to be had here. It was also argued that any heat transferred to the wheels by the Z pinch, and there's probably a whole lot of it, can be mitigated by said wheels dipping into the liquid tin bath. This new design also offers some debris mitigation benefits. A foil trap is included to capture tin debris flying off the surface and at the cost of some EUV light loss. In 2005, Japan's Ushio Group bought 50% of the Philips Extreme UV company. A few years later, in 2008, they purchased the whole thing, renaming the company to Extreme Technologies and started doing joint research on EUV light sources. The resulting LDP light source was revealed in 2010 as the Tin DPP Source Collector Module, or SoCoMo. Its electrodes fire powerful discharges of up to 20,000amps, basically a small lightning strike. Lasting for just a few hundred nanoseconds. During operations, we discharge anywhere from 10,000 to 20,000 to up to an insane 100,000 times per second. By adjusting either the discharged pulse repetition pace or the energy of each pulse, we get stronger EUV output power. A higher pulse energy at a lower frequency can produce the same output as a lower pulse energy at a higher frequency. They pitch LDP as having the best of both LPP and DPP scalability and stability LDP also seemed physically smaller and simpler in concept, probably why ASML used an LDP light source in its first alpha demo tool. And as reported in an October 2010 IEEE Spectrum article on EUV, they kept close tabs on both LDP through Extreme and LPP through the startups Cymer and Giga Photon. In the end though, ASML switched away from the LDP source, adopting LPP without talking to someone who worked there, and no one responded to my emails. It is hard for me to find out exactly why, but I have a few theories. The simplest theory is that LDP could not scale output power as fast as LPP can. In 2008, Extreme announced that it was capable of producing 500 watts of EUV power output at plasma at Plasma refers to the measurement of the EUV power emitted by the plasma itself right at the light source stage. Sounds quite high, but remember that the light still must go through all the reflections towards the wafer. The EUV power actually reaching the wafer is measured using a metric known as power at intermediate focus. Sokomo, whilst using one set of EUV light collectors, had a measured power at intermediate focus of 14 watts, corresponding to about seven to eight wafers per hour. If equipped with an improved EUV light collector armed with additional reflective surfaces, then that theoretically goes up to 34 watts better, but still short of Cymer's 250W and Giga Photon's 104 watts at intermediate focus. As reported by IEEE spectrum. To achieve 500W at intermediate focus, Extreme said that LDP would theoretically need to hit 4,000 watts at plasma. Based on their accompanying graph, this can be done if they discharged 50,000 plasma pulses per second, with each pulse discharge releasing about 80 millijoules pulse energy. Perhaps this was too much to engineer with the heat generated by the z pinches too difficult to dissipate. The paper's authors did seem to imply that there was some kind of physical limitation at the aforementioned 80 millijoules per pulse metric. Not clear based on the reading. Maybe it was something else. Earlier studies with XENON plasma found that as we pump more power into the Z pinch, the generated plasma becomes larger rather than brighter. This larger plasma emits UV light because beyond what the collector mirrors can physically collect, essentially wasting energy. Maybe something similar happened here. One of the later announcements I dug UP was in October 2011 saying that USHIO had achieved 30W power and intermediate focus. This was up from the previous 14 watts, but by then I think ASML had seen enough and moved forward with lpp, though the community did keep debating over the choice. As late as 2014, Ushio kept on using the LTP method to generate EUV light for special mask blank inspection tools. A 2022 paper about the tools showed output of about 250W at Plasma. I'd probably muse that that is your local maximum for this technology, and it works fine for checking mass blanks but not for litho. LPP, on the other hand, has scaled quite well for ASML. Late last year it was mentioned that ASML's research team in San Diego hit 740W EUV power. Recently, I presume at intermediate focus. They expect to productize that shortly. China continues working on a technique to produce EUV light for their own domestic use. They're chasing all the known EUV light source methods, including LPP and free electron lasers, so it should not be a big surprise that they are looking at laser assisted discharge plasma too. In 2023 there was a publication from a research team in the School of Aeronautics and Astronautics at the Harbin Institute of Technology. There seems to be a small cluster of researchers working on a DPP on LDP Ever since the early 2010s. There are a few papers out there from other Chinese organizations investigating LDP plasma dynamics and simulations. A casual reading of these papers didn't find anything I considered groundbreaking. If anything, it seems rather unimpressive with lower conversion efficiencies than setups in 2005 and 2006. But I suspect they are hiding their strengths. There will certainly be some people out there who just want me to say something like oh, China can't catch up with asml or China will certainly achieve EUV or China will destroy America's blockade. Oh, and how that all affects ASML stock? I don't do stock analysis. I do want to say that I think China achieves EUV eventually. After all, EUV is a technology made by man. Nothing is impossible. I'm also willing to say that it happened sooner than you think. They're spending so much money and putting so many smart people on it and there is already so much published research out there on the Golden Path. The Chinese have the benefit of knowing that it can work and knowing conceptually how it's done. That's like 60% of the job, kinda. Side note, I recently saw an interesting paper posted by Fred Chen on LinkedIn discussing an exploratory effort by Russian academics in 11.2nm EUV light and ruthenium beryllium mirrors. I don't see that particular effort scaling, but do respect that people are exploring alt EUV setups. I'd never presume people back then had all the answers. Anyway, I doubt that any China EUV machine will be economically competitive with an ASML EUV machine. At the start it might pattern far few wafers per hour, less than 100 wafers even, or pattern with worse resolution so on. People might ridicule it as a boondoggle or a toy, but I do believe that the machine will be clever with an interesting twist. For example, if the LDP method is not bright enough, then perhaps it has a nifty two mirror optics system to deal with that shortcoming. ASML should study whatever does come out whenever it does, and consider possibly integrating it. And finally, I believe one of China's superpowers, an Asian one in general, really is being able to tolerate a business making no money for years on end. ASML's reported gross margins are about 51%. I don't know what's in that service tools, components, etc. But I reckon those need to go down when the time comes to beat back the Chinese Challenge. Alright, that's all for tonight. Sorry about my voice, it's not sounding that great. Subscribe to the channel, sign up for the Patreon and I'll see you guys next time.