Japan’s Semiconductor Photoresist Monopoly

Japan doesn't dominate semiconductor fabrication or lithography machines like it once did, but they still keep a mighty grip on the supply chain, particularly one very special chemical. The photoresist is often ignored, we just kind of offhandedly mention it at the end there. But without question, it is the most important chemical of the lithography process, literally indispensable. Around 90% of this market is held by Japanese companies like JSR and Tokyo Okoko-Gyo, amongst others. One of those companies is today state owned.

日本不再像过去那样在半导体制造和光刻机领域占据主导地位，但它们仍然在供应链中具有强大的影响力，特别是在一种非常重要的化学材料方面。光刻胶常常被忽视，我们通常只是随便在最后提到它。但毫无疑问，它是光刻过程中最重要的化学品，绝对不可或缺。这个市场约90%由日本公司占据，包括JSR和东京应化工业等。其中一家现在是国有企业。

In this video, we examine the many generations of photoresist used by the industry and Japan's low-key monopoly of it. If the fancy DUV lasers or UV mirrors help write the chip design, then the photoresist is the medium onto which that design is written. I argue that it is the photoresist, not the $150 million UV exposure tool that is the true star of the show throughout this whole process. At the start of the photolithography process, we apply the liquid photoresist onto what we want to pattern, silicon, germanium, silicon dioxide, metal, whatever.

在这个视频中，我们探讨了工业中使用的多代光刻胶，以及日本在这一领域低调的垄断地位。如果说华丽的DUV激光或紫外镜帮助我们写入芯片设计，那么光刻胶就是承载这些设计的媒介。我认为在整个过程中，真正的明星是光刻胶，而不是价值1.5亿美金的紫外曝光设备。在光刻过程的开始，我们会将液态光刻胶涂在我们想要刻画的材料上，无论是硅、锗、二氧化硅还是金属。

This is done using a spin coat machine that dispenses the liquid photoresist onto the wafer as it spins at a very high rate. Depending on the spin speed, the liquid itself and the dispensing rate, we get a nice even layer. After this, we need to dry this spin coated photoresist layer, and this is done with a spin in the spin dryer and then a trip to the hot plate, a dehydration bake or sometimes also called a soft bake, pre-bake or post-apply bake. Once this is done, we can expose the wafer inside the exposure tool, which transfers the chip layer design from the photo mask to the wafer using powerful UV rays.

这是使用旋涂机完成的，旋涂机在晶圆高速旋转时，将液态光刻胶均匀地分配到晶圆上。根据旋转速度、液体性质和分配速率，我们可以得到一层均匀的涂层。在此之后，我们需要对这层旋涂的光刻胶进行干燥，首先需在旋转干燥机中旋转，然后放入热板进行脱水烘烤，这个过程有时也称为“软烘烤”、“预烘烤”或“涂覆后烘烤”。完成这些步骤后，我们可以在曝光工具中处理晶圆，通过强大的紫外线将光掩膜上的芯片层设计转移到晶圆上。

Most of the time we use a positive tone photoresist. When exposed to the light, a positive photoresist undergoes a special chemical change that makes it easier to be dissolved and washed away by another chemical called the developer solution. On the other hand, we have a negative tone system, which works the opposite of the positive. The exposed areas become less likely to be dissolved and washed away by the developer solution. Negative tone resists came along first, but nowadays semiconductor fabs use and prefer positive tone resists. We'll talk about more why later.

大多数情况下，我们使用正性光刻胶。当暴露在光下时，正性光刻胶会发生一种特殊的化学变化，使其更容易被称为显影剂的其他化学物质溶解并洗去。另一方面，我们还有负性光刻胶，它的工作原理与正性光刻胶相反，曝光区域变得不容易被显影剂溶解和洗去。负性光刻胶最先出现，但现在半导体工厂使用并偏爱正性光刻胶。我们稍后会讨论更多原因。

Anyway, whether we use a positive or negative photoresist, after exposure we have this latent exposed image that we must now develop. As I just mentioned, this is done with a liquid developer solution, a hydroxide chemical that stabilizes the image enough so that it can withstand later riggers. After a final spin or bake is done to remove any trace of water left behind, we can now take the wafer to the edge bay. There, we employ an acid, wet edge, or gas plasma, dry edge, to remove whatever parts of wafer not protected by the developed resist layer.

无论我们使用正性光刻胶还是负性光刻胶，在曝光后，我们都会得到一个潜在的曝光图像，现在需要进行显影。正如我刚才提到的，这一步骤需要使用显影液，这是一种碱性化学溶液，可以稳定图像，以便其能够承受后续工序。在最后一步，通过旋干或烘烤去除残留水分后，我们可以将晶圆移动到蚀刻区。在那里，我们会采用酸性湿法蚀刻或气体等离子体干法蚀刻，去除未被显影出的光刻胶保护的晶圆部分。

Ergo, why we call it a resist? It resists the edge. The name was first coined by the French physicist Edmund Beckerle in 1840. His son Henry discovered radiation alongside the curays. In this video, I shall use the terms photoresist and resist interchangeably. Finally, we finish by stripping the remaining photoresist layer from the wafer. This is challenging. We're trying to remove very burnt gunk off a metal pan. Semi-conductor companies do this using either a wet acid bath or plasma ashing induced by dry oxygen plasma. We then suck up the ashes using a vacuum machine.

因此，为什么我们称它为“抗蚀剂”？因为它能抵抗边缘侵蚀。这个名称最初由法国物理学家埃德蒙·贝克尔在1840年创造。他的儿子亨利与居里夫妇一起发现了辐射。在这个视频中，我将交替使用"光刻胶"和"抗蚀剂"这两个术语。最后，我们需要从晶圆上剥离剩余的光刻胶层。这有点困难，就像要从金属锅上清除烧焦的残渣一样。半导体公司通常会使用湿酸浴或者干氧气等离子体诱导的等离子刻蚀来完成这项任务。接下来，我们用真空设备将灰烬吸走。

It works quite well, though the radiation used to create the plasma can disrupt the wafer. In light of that concern, there has been recent interest in the supercritical CO2-based fluid. Special CO2 made such that it acts like both a liquid and gas. The ideal semi-conductor photoresist must have several attributes. First, it must be sensitive to radiation and the wavelength you're using. The more sensitive, the better. A lithography exposure tool's cost of ownership is directly linked to its throughput, as in how many wafers per hour it can produce. If the photoresist is not that sensitive, then it will take longer to expose the wafer, which slows down throughput, which in turn raises the lithography exposure tool's total cost of ownership. That is bad.

它的效果相当好，但用于产生等离子体的辐射可能会干扰晶圆。考虑到这个问题，最近对超临界二氧化碳基流体产生了兴趣。这种特殊的二氧化碳经过处理，使其同时表现出液体和气体的特性。理想的半导体光刻胶必须具备几个属性。首先，它必须对所使用的辐射和波长敏感。越敏感越好。光刻曝光设备的拥有成本直接与其吞吐量相关，也就是每小时可以处理多少晶圆。如果光刻胶不够敏感，曝光晶圆就需要更长时间，这会降低吞吐量，从而提高光刻曝光设备的总拥有成本，这样不好。

Second, the photoresist must form an even and uniform layer on the wafer. Moreover, that layer must stick to the wafer's surface very well. Otherwise, the resist layer, after exposure, might lift up off the underlying wafer. That is bad. Third, the resist layer must faithfully keep its shape even as it endures a long and torturous series of processing steps after exposure. If the layer warps, changes or breaks, then the same goes for the transfer design pattern. That is bad. Considering all the jobs that must be done, there is no such thing as a perfect photoresist, but we keep looking.

其次，光刻胶必须在晶圆上形成均匀且一致的薄层。此外，该薄层必须牢牢粘附在晶圆的表面，否则在曝光后，这层光刻胶可能会从晶圆上剥离。这是很糟糕的。再次，光刻胶层在经过长时间且复杂的处理步骤后，必须保持其形状不变。如果这层薄膜变形、改变或断裂，那么设计的图案转移也会出现问题。这同样很糟糕。考虑到所有需要完成的任务，目前还不存在完美的光刻胶，但我们仍在不断寻找。

The resist concept itself dates back to the earliest days of photography. Like really. In the 1820s, the French inventor, Nisei for Nieps, noticed that if you exposed a layer of bitumen of Judea asphalt to light for a few hours, then the exposed areas hardened and quote-unquote resisted washing away, while the unexposed areas did not. Taking advantage of this, Nieps produced the world's oldest surviving photograph at some time between 1822 and 1827. It is William Shockley, yes, the brilliant inventor of the bipolar transistor, who most likely first came up with the idea of using photoresist to accurately reproduce patterns onto an oxide film on a wafer. This was the 1950s, and they started with what was then the photography and printing industries most commonly used resist, dichromated gelatin.

抗蚀剂的概念可以追溯到摄影术的最早期。真的，最早可以追溯到19世纪20年代。法国发明家尼埃普斯注意到，如果将犹太沥青（Bitumen of Judea）涂层暴露在光线下几个小时，曝光的部分就会变硬并“抵抗”被冲洗掉，而未曝光的部分则不会。利用这一特性，尼埃普斯在1822年至1827年之间拍摄了世界上现存最古老的照片。到1950年代时，是著名发明家威廉·肖克利——也就是那位发明了双极晶体管的人——最有可能提出利用光刻胶将图案精确地复制到晶圆上的氧化膜的方法。那时候，他们使用了在摄影和印刷行业中被广泛采用的光刻胶——二铬酸盐明胶。

The second drawback is somewhat more relevant to semiconductors. In a protein gelatin, gelatin did not resist chemicals during the etch process very well. Technicians had to add a bake step to literally burn in the image. Bell Labs wanted to use hydrofluoric acid to etch the design into the germanium or silicon. Gelatin would not have stood a chance. So Shockley called up the guys at Eastman Kodak Research Lab in Rochester, New York for some advice. Kodak turned to the work of Lewis Minks, who a few years prior, had invented a curious thing. Minks was researching a replacement for dichromatic gelatin so that Kodak can sell pre-treated lithographic plates to newspapers, like as in Ready to Go without that first drawback.

第二个缺点与半导体更加相关。在蛋白质明胶中，明胶在蚀刻过程中对化学品的抵抗能力不强。所以技术人员不得不增加一个烘烤步骤，来将图像"烧录"进去。贝尔实验室想使用氢氟酸将设计蚀刻到锗或硅上，明胶对此完全无法应对。因此，肖克利联系了位于纽约罗切斯特的伊士曼柯达研究实验室寻求建议。柯达转向了刘易斯·明克斯的研究。几年前，明克斯发明了一种奇特的东西。他正在研究一种替代二铬酸盐明胶的方法，以便柯达能向报社出售预处理好的平版印刷版材，就像是解决了第一个缺点一样，实现"开箱即用"。

Minks suspected why gelatin worked well as a photoresist was because of a chemical reaction called cross-linking. This is where adjacent chains in the protein would link together in a chain, hardening the whole thing. So he went through the literature and found the only other known substance with this same reaction, Cinnamate or Cinnamic Acid. Yes, as in cinnamon. It is derived from oil of cinnamon. Because it is based on cinnamon, Sam's aloof, the noted silicon YouTuber and now startup founder, wrote in a very cool blog post that it smells quite nice. Anyway, Minks leveraged this behavior to create a polyvinyl alcohol, cinnamon negative tone photoresist they called Kodak photoresist, or KPR. When exposed to UV light, KPR's chemical groups would cross-link to create an insoluble region that can survive the etch process. KPR first received a patent in 1950.

明克斯推测明胶之所以能很好地用作光刻胶，是因为一种叫做交联的化学反应。这种反应使得蛋白质中相邻的链彼此连接，进而使整个结构变得更加坚硬。因此，他查阅了文献，找到了唯一另一种已知能发生这种反应的物质，即桂皮酸。没错，就是和桂皮有关的。它是从肉桂油中提取的。由于它源自肉桂，著名的硅基产品YouTuber兼创业公司创始人Sam Aloof在一篇很酷的博客文章中写道，它闻起来相当不错。不管怎样，明克斯利用这种特性创造了一种聚乙烯醇和肉桂酸组成的负性光刻胶，称为柯达光刻胶，或KPR。当暴露于紫外光下时，KPR的化学基团会发生交联，形成一个不溶区域，从而在蚀刻过程中得以保存。KPR首次获得专利是在1950年。

So that was what the guys at Kodak suggested to use and sometime in 1953 or 1954, not clear when, Shockley sent Jules Andres over to the Kodak lab in New York to pick up a bottle of this stuff and learn how to use it. KPR easily survived the hydrofluoric acid, but being originally developed for newspapers it presented a new problem. It did not adhere very well to the wafers glassy oxidized metal surface. They'll complain to Kodak about this. After much effort, Kodak determined that they needed something entirely new. The problem landed at the doorstep of Kodak's head of graphic design department in England, Dr. Martin Heffer and his colleague Hans Wagner. The latter just happened to have recently read a paper about azido compounds, which are called that because they have groups of three nitrogen atoms bound to a carbon atom. When these puppies are hit by light, they decompose into reactive chemicals that are very good at sticking to stuff, brilliant as they say over there in London.

所以，这就是柯达公司的建议，大概是在1953或1954年，不确定具体时间，肖克利派朱尔斯·安德烈斯去纽约的柯达实验室取了一瓶这种材料，并学习如何使用。KPR很容易抵挡氢氟酸的侵蚀，但它最初是为报纸开发的，所以遇到了新的问题：它不能很好地粘附在硅片表面的玻璃氧化金属上。为此，他们向柯达公司抱怨。经过大量努力，柯达公司确定需要一种全新的材料。问题落到了位于英格兰的柯达平面设计部门主管马丁·赫弗博士和他的同事汉斯·瓦格纳的肩上。恰巧，汉斯·瓦格纳不久前刚读过一篇关于叠氮化合物的论文，这种化合物之所以如此命名，是因为它们具有由三个氮原子与一个碳原子相连的基团。当这些化合物受到光照时，会分解成能很好地粘附在物体上的活性化学物质，正如伦敦人所说，这真是绝妙。

So Wagner created several bizicide compounds, meaning to have two of these three nitrogen groups, and mix them with simple rubber cement as a binder for stabilization. The result was a very good negative tone photoresis called Kodak Thin Film Resist or KTFR. When exposed to UV light, the bizicide compound inside the KTFR releases a special nitrogen molecule. This intermediate molecule then reacts with the rest of the KTFR to cross-link the resin. Released in 1957, KTFR remained an industry workhorse, though a somewhat flawed one, for the next 15 years. If you are wondering when Japan is going to enter the picture, it's now. The first Japanese company to produce a domestic photoresis was Tokyo Okakokyo. Founded back in 1936 as the Tokyo Okako Research Institute, they started by producing niche chemicals for batteries and cathode ray TVs.

所以，瓦格纳创造了几种双氮化合物，这意味着它们包含三个氮基中的两个，并将它们与简单的橡胶水泥混合，作为稳定剂的粘合剂。结果形成了一种非常出色的负性光刻胶，名为柯达薄膜光刻胶（KTFR）。当暴露在紫外线下时，KTFR内部的双氮化合物会释放出一种特殊的氮分子。这种中间分子随后与KTFR的其余部分反应，使树脂交联。KTFR于1957年问世，并在接下来的15年中成为行业的中坚力量，尽管有一些缺陷。如果你在想日本何时会进入这个领域，那就是现在。第一家生产国产光刻胶的日本公司是东京冶金工业会社。该公司成立于1936年，当时名为东京冶金研究所，起初生产用于电池和阴极射线电视的专用化学品。

In 1961, they had produced some cinematic acid stuff, originally with the goal of supplying photoresis for producing the shadow mask for those old shadow mask tube cathode ray televisions. I earlier covered the shadow mask tube cathode ray TV, originally invented by RCA, in a prior video about the Sony Trinitron. Unfortunately, RCA rejected Tokyo Okako's cinematic acid, ostensibly because of fire risks. The company's employees started looking for other used cases for their new chemical. In 1961, a Tokyo Okako researcher gave a presentation about their polyvinyl cinnamon, and someone in the audience told them to try using it as a photoresis for printing circuit boards.

在1961年，他们曾生产出一些电影酸性材料，最初的目标是为那些旧式阴影罩显像管电视生产阴影罩的光刻胶。我之前在一个关于索尼特丽珑的影片中介绍过阴影罩显像管电视，这种技术最初是由RCA发明的。不幸的是，RCA出于火灾风险的考虑，拒绝了东京冈子公司的电影酸性材料。于是，公司员工开始寻找他们新化学产品的其他应用。在1961年，东京冈子的一位研究员进行了一次关于其聚乙烯肉桂酸的演讲，现场有人建议他们尝试将其用作印刷电路板的光刻胶。

They dispatched a salesperson to manufacture and work with what is now Japan's National Institute of Advanced Industrial Science and Technology to produce Tokyo Okako photoresis or TPR. TPR worked, but did not sell very well. At the start of 1963, they only sold about 80 liters each month. It improved three years later to 1000 liters thanks to changes in the way circuit boards were printed, but still nothing to write home about. In 1968, one of the workers on the TPR team suggested using it for semiconductor lithography. At this time, most Japanese companies preferred KPR from Kodak, but the industry was transitioning from polyvinyl cinnamon resist to rubber-based photoresis compounds like KTFR. Such technology transitions are always opportunities.

他们派遣了一名销售人员去与现在的日本国立先进工业科学技术研究所合作，生产东京Okako光刻胶（简称TPR）。TPR虽然有效，但销售情况并不理想。在1963年初，他们每个月仅售出约80升。三年后，随着电路板印刷方式的改变，销量提升至每月1000升，但仍然不算出色。1968年，TPR团队的一名员工建议将其用于半导体光刻。当时，大多数日本公司更青睐柯达的KPR，但行业正在从聚乙烯桂皮光刻胶转向以橡胶为基础的光刻胶配方如KTFR。这种技术转型总是充满机会。

Tokyo Okako had experience in this rubber, and in October 1968 produced Okako-Resist or OMR81, Japan's first domestically produced photoresist. The negative tone OMR81 was the first of the OMR series, with the 83 being pretty successful. KTFR hit the end of its viability with the advent of the two micron process node in 1972. Why? Because KTFR's crosslinks were not completely insoluble, the KTFR still absorbs some solvent, which causes it to swell and wrinkle, and at 2 micrometers the swelling and wrinkling was large enough to disrupt the prints overall accuracy. So the industry needed a new thing, and that turned out to be the DNQ novelac photoresist. DNQ stands for… Diosonathoquinone, and I'm not going to say it again. Novelac, I have also seen its spell with a K at the end, is a resin polymer somewhat related to bacolite.

东京大加古在橡胶领域拥有经验，并于1968年10月生产了大加古抗蚀剂（Okako-Resist，简称OMR81），这是日本首款国产光刻胶。OMR81是OMR系列的首款负胶，其中OMR83相当成功。随着1972年两微米工艺节点的出现，KTFR这种光刻胶达到了使用极限。为什么呢？因为KTFR的交联不是完全不溶的，它仍然吸收一些溶剂，导致膨胀和起皱，在2微米时，这种膨胀和起皱足以影响整体印刷精度。因此，行业需要一种新材料，这就催生了DNQ甲酚醛光刻胶的诞生。DNQ代表双氧萘醌（Diosonathoquinone），我只说这一遍。甲酚醛聚合物有时也拼作Novelac或Novolak，它是一种与电木（Bakelite）相关的树脂聚合物。

DNQ novelac is a positive tone photoresist and so works differently than its predecessors. When exposed to light, the resist becomes more soluble rather than less. Why? Who cares? Next slide please. Just kidding. Novelac films are normally quite soluble in developer solution, but when you add DNQ to the novelac, the DNQ reacts with it to make it much more insoluble. In other words, the DNQ is a dissolution inhibitor. When the UV light hits the DNQ novelac however, something mysterious happens. Now the resist, strangely enough, dissolves even faster than just the novelac alone. Why? Do you really want to know? The DNQ dissolves in the UV light to create acid. For a long time we presume that that caused the whole photoresist to dissolve faster. But later research by a ricer in co at Polytech University found that when DNQ and novelac mix together, hydrogen bonds attach to organic compounds called phenolic groups to make what is called a phenolic string, holding it together like my sanity right now and preventing it from dissolving. When the DNQ gets hit by light, it rearranges itself to release a nitrogen atom. This rearrangement releases a lot of heat energy which breaks down those hydrogen bonds which then break the whole string apart. Voila, the dissolving behavior returns. The result is that after you develop the exposed wafer, you're left with only the unexposed areas of the resist layer remaining. A positive tone. Chemistry is hard.

DNQ光刻胶是一种正型光刻胶，它的工作原理与之前的光刻胶不同。受到光照时，这种光刻胶变得更易溶解，而不是更难溶解。为什么会这样呢？有人会关心吗？下一张幻灯片请。开玩笑的。通常来说，新型树脂在显影剂中是非常易溶的，但当你加入DNQ后，DNQ会与树脂发生反应，使其变得更加不溶。换句话说，DNQ是一种溶解抑制剂。然而，当紫外线照射到DNQ光刻胶时，会发生神奇的事情。此时，这种光刻胶反而比单独使用树脂时溶解得更快。为什么会这样呢？你真的想知道吗？因为DNQ在紫外线下会分解生成酸。我们曾经认为这使整个光刻胶溶解得更快。但是，Polytech大学的研究人员发现，当DNQ和新型树脂混合时，氢键会与名为酚基的有机化合物结合，形成所谓的酚链，像保持我理智一样紧紧地将其固定，防止溶解。当DNQ受到光照时，它会发生重新排列并释放一个氮原子。这种重新排列释放大量热能，打破这些氢键，从而分解整个酚链。瞧，溶解的行为又回来了。结果是，经过显影处理后，晶圆上只剩下未曝光区域的光刻胶层。一种正型调子。化学真的很难。

DNQ novelac positive tone refoto-resists were a major development in the semiconductor industry particularly at the Mercury Land 436 nanometer and 365 nanometer wavelengths. They were first invented by the German organic chemist Oskar Zeus at the German Chemicals Company Cali AG, which is now a subsidiary of Hosh AG which is now part of Sanofi. Cali's main business in the 1940s was reproducing engineering blueprints. This required the use of a positive tone resist that worked with this special blueprint paper. So, Oskar Zeus, it means sweet in German, discovered DNQ works well and then his team discovered that novelac was a great physical binder pairing literally because the company across the street was making it. So DNQ novelac first entered the printing market in the 1940s and patented in the US as a photoresist in 1956. When the semiconductor lore has it, the DNQ novelac then made their way from Cali to Bell Labs via family. Cali's US subsidiary Azoplait is in Murray Hill, New Jersey, same as Bell Labs. The story goes that the father of an Azoplait technician worked at Bell. When the father complained to his son about KTFR's issues, the latter recommended the resist they use at Azoplait, DNQ novelac. The rest is history. Another major company was the American-based Shipley. In the mid-1960s, they introduced their own line of AZ photoresists for printing circuit boards and photo mask components. They then branched into microelectronics photoresist. Other big photoresist suppliers turned in the 1970s included DuPont, Dyna-Chem in California, GAF and Miecker Company in New York, and the PA Hunt Company in New Jersey. So until pretty recently, American companies dominated this space.

DNQ novolac 正性光刻胶是半导体行业的一项重大进展，尤其是在435纳米和365纳米波长上的应用。它们首次由德国有机化学家奥斯卡·宙斯在德国化学公司Cali AG发明，现在该公司是Hosh AG的子公司，而Hosh AG又属于Sanofi。Cali公司在20世纪40年代的主要业务是复制工程图纸，这需要使用与特殊图纸配合的正性光刻胶。因此，奥斯卡·宙斯（在德语中意为“甜”）发现DNQ效果很好，然后他的团队发现novolac是一种很好的物理结合剂，部分原因是因为街对面的公司正在生产它。因此，DNQ novolac在20世纪40年代首次进入印刷市场，并于1956年在美国被作为光刻胶申请了专利。据传，DNQ novolac通过家庭原因从Cali传到了贝尔实验室。Cali在美国的子公司Azoplait位于新泽西的默里山，正是贝尔实验室所在地。故事是这样的：Azoplait公司一名技术员的父亲在贝尔实验室工作。当父亲向儿子抱怨KTFR的使用问题时，儿子建议父亲使用他们在Azoplait公司用的光刻胶，也就是DNQ novolac。故事就此展开。另一个主要公司是美国的Shipley。他们在20世纪60年代中期推出了自己的用于印刷电路板和光掩模的AZ光刻胶产品线，并且随后进入了微电子光刻胶市场。在20世纪70年代，其他大型光刻胶供应商还包括美国的杜邦、位于加州的Dyna-Chem、纽约的GAF和Miecker公司，以及新泽西的PA Hunt公司。因此，直到最近，美国公司一直主导着这一领域。

Tokyo Oka produced their own first positive tone photoresist in 1971 with OFPR2. Eight years later, they released OFPR800, which was widely adopted by Japanese companies. Not only because of favorable pricing, but also because it did not leave as much residue after the resist stripping step. DNQ novelac photoresist dominated the semiconductor and larger electronics industries for 25 years, maintaining 80-90% market share. And Tokyo Oka's OFPR became a cornerstone of Japan's dram dominance in the late 1970s and 1980s, 80% of the Japanese market. As semiconductor manufacturing left the United States for Asia, Japanese chemical companies entered the market, replacing the American ones.

东京冈田（Tokyo Oka）在1971年生产了他们的第一款正性光刻胶OFPR2。八年后，他们推出了OFPR800，这款产品由于售价合理且在光刻胶剥离后残留物较少而被许多日本公司广泛采用。DNQ novolak光刻胶在半导体及大型电子行业中占据了25年的主导地位，市场份额保持在80-90%。在20世纪70年代末和80年代，东京冈田的OFPR成为日本在动态随机存取存储器（dram）领域占优势的基石，占据了日本市场的80%。随着半导体制造从美国转移到亚洲，日本的化学公司进入了市场，取代了美国公司。

One of the more successful entrants was JSR, and now I think it is fitting to introduce them because they're going to have a bigger role later in the story. JSR was founded in 1957 as Japan's first synthetic rubber producer. Ergo the name. JSR stands for Japan Synthetic Rubber Company. Japan's Mitty sponsored the company's founding to explore tech that can help them import less natural rubber. Their founding president was Shojiro Ishibashi, the founder of Bridgestone. Isn't that nice? Three years later, JSR finished their factory and started making synthetic rubber like nitro rubber, styrene rubber, and chloroprene to supply domestic companies, mostly making tires. Technical expertise was supplied by American companies like Goodyear and ESO, the oil company.

其中一个比较成功的参赛者是JSR，现在我认为介绍他们是合适的，因为他们将在故事后半段扮演更重要的角色。JSR成立于1957年，是日本第一家合成橡胶生产商。这正是其名字的由来：JSR代表日本合成橡胶公司。日本的通商产业省（MITI）赞助了公司的成立，旨在探索能够帮助他们减少天然橡胶进口的技术。他们的创始总裁是石桥正二郎，他也是普利司通（Bridgestone）的创始人。是不是很有意思？三年之后，JSR建成了他们的工厂，开始生产合成橡胶，如硝化橡胶、苯乙烯橡胶和氯丁橡胶，主要供应国内的轮胎制造公司。技术支持则来自于像固特异（Goodyear）和石油公司ESO这样的美国企业。

The Japanese government privatized JSR in the 1960s, and then the oil crises of the 1970s forced the company to diversify its product lineup. In 1979, they introduced their CIR photo-resist, a negative tone photo-resist using a material derived from its synthetic rubber originally made for tires. It was known for being amenable to plasma-ashing. Three years later, they opened an optical materials business, providing coating materials for optical fiber and the like. Seems random, right? Medium-sized Japanese chemical companies like JSR followed a business strategy like Tokyo Oka many years before, seeking out underserved market niches. Each niche might be worth just several tens of millions of dollars.

上世纪60年代，日本政府将JSR私有化，然后70年代的石油危机迫使公司多元化其产品阵容。1979年，他们推出了CIR光阻剂，这是一种负性光阻剂，使用源自原本用于轮胎的合成橡胶的材料制成。它以适合等离子灰化而闻名。三年后，他们开设了光学材料业务，提供用于光纤等的涂料。看似随机，对吧？像JSR这样的中型日本化工公司，早在多年前就仿效东京化成工业等的商业策略，专注于服务不足的市场小众。每个市场小众可能只有几千万美元的价值。

In the late 1970s and early 1980s, the microelectronics market was such a niche. It did not interest the big Japanese chemical companies of the time, like Mitsubishi or Sumitomo Chemical, but it does kind of move the needle for medium-sized companies like JSR. Their style would be to collaborate closely with their customers, the Japanese microelectronics manufacturers, to produce a custom solution relying on their own in-house technology and expertise. JSR built up their initial microelectronics chops this way. In 1989, they sold about 3 billion yen, or $21 million of photo-resist, 90% of which to other Japanese companies. It does not sound like much, and indeed, photo-resist were just 1.5% of their total sales, but nevertheless still make them the second largest Japanese photo-resist supplier. Then in the 1990s, it came time for another technology transition in photo-resist. The leading-edge semiconductor industry transitioned from G and I-line mercury lamps to the 248 nanometer KRF and 193 nanometer ARF wavelength light, Deep Ultraviolet Light or DUV, Eximer Lasers.

在20世纪70年代末和80年代初，微电子市场是一个非常小众的领域。像三菱或住友化学这样的大型日本化学公司对此不感兴趣，但对于像JSR这样的中型公司而言，这一领域却具有一定的吸引力。JSR的策略是与日本微电子制造商紧密合作，利用他们自身的技术和专业知识，为客户提供定制化解决方案。JSR就是通过这种方式逐步发展其微电子技术。 到1989年，JSR销售了约30亿日元（约合2100万美元）的光刻胶产品，其中90%销往日本国内的公司。虽然这似乎不是一个很大的数字，而且光刻胶只占其总销售额的1.5%，但这仍然使他们成为日本第二大光刻胶供应商。 到了1990年代，光刻胶技术迎来了新的转变。尖端半导体行业开始从G线和I线汞灯过渡到波长为248纳米的KRF和193纳米的ARF，即深紫外光（DUV）技术，使用准分子激光器。

These new photo-lithography tools shifted the industry away from the workhorse D&Q novel-act photo-resist for two technical reasons. First, the D&Q absorbed too much of these more energetic DUV photons at the surface of the resist layer, so they then react too fast, blocking the DUV beam from penetrating any deeper. So think of it as a cake's surface instantly hardening, leaving the bottom layer still uncooked and liquid. Delicious for things like lava cakes bad for resists. Second and more seriously, these new DUV Eximer Lasers need line narrowing subsystems to reduce their wavelength ranges. This narrowing also hurt the strength of their light, thusly they make fewer photons than their lamp predecessors. This means that each photon must somehow trigger more chemical reactions than it did before. It was this need for a disruptively higher jump in photon efficiency, maybe 10, 30 or even 100 times higher, that caused industry chemists to judge it necessary to move on from the venerable D&Q novel-act photo-resists.

这些新的光刻工具因两个技术原因使行业从传统的D&Q光致抗蚀剂转型。首先，D&Q在光刻胶表面吸收了太多更高能量的深紫外光（DUV）光子，使其反应过快，从而阻碍了DUV光束深入渗透。这可以想象成蛋糕表面迅速变硬，而底部仍然没有烤熟并保持液态。对于岩浆蛋糕来说这样的效果是美味的，但对于光致抗蚀剂则不理想。其次，更重要的是，这些新的DUV准分子激光器需要缩小波长范围的子系统。这种波长的收缩也削弱了其光强，因此产生的光子比旧的灯光源要少。这意味着每个光子需要比以前引发更多的化学反应。这一对光子效率极大提升的需要，可能是10倍、30倍甚至100倍，导致行业化学家认为有必要从经典的D&Q光致抗蚀剂转型。

Thus, we have the next generation of resists, the chemically amplified resists or CARs. The chemical amplification concept has been around since the one micrometer device generation. The simplified version is that the light exposure does not directly modify the resist chemical structure. Instead, it produces an intermediate chemical that then serves as a catalyst for a subsequent set of reactions within said resist. And if you recall from your high school chemistry class, a catalyst's key trait is getting stolen from your car. It is that it itself does not permanently change or get consumed when speeding up or triggering a chemical reaction, so it can keep on doing it many times over. In other words, a single photon interaction leads to many later chemical reactions. Perfect for maximally amplifying the impact of a small number of photons as long as those reactions don't diffuse too far away from where they're supposed to.

因此，我们拥有了新一代的光刻胶，即化学放大光刻胶（CARs）。化学放大的概念自一微米设备的时代就已经出现。 简单来说，光照并不直接改变光刻胶的化学结构，而是生成一种中间化学物质，该物质作为催化剂引发光刻胶中的一系列后续反应。 如果你还记得高中化学课，催化剂的一个关键特性就是它能加速或触发化学反应，但本身不会发生永久性变化或被消耗，因此可以重复使用。 换句话说，一个光子与催化剂的互动可以引发多次化学反应。这种特性非常适合在光子数量较少的情况下，最大限度地放大其影响力，只要这些反应不扩散到不该去的地方。

During in 1978, three scientists at IBM, Jean-Forsché, Grant Wilson and Hiroshi Ito worked on the mechanism that eventually produced the first practical CAR, which works using what is called a photo-acid generator. IBM introduced this first CAR, called T-Bok, in the early 1980s to help produce one megabit dram. It was quite temperamental, requiring gentle wafer stage movement and high air purification. There's a story a little later about that, but it eventually did the job. IBM initially kept their chemically amplified resists a trade secret. When trying to patent the invention, they discovered that a scientist at 3M had already patented something similar a few years before, though the company did nothing with that invention.

在1978年，IBM的三位科学家Jean-Forsché、Grant Wilson和Hiroshi Ito研究了一种机制，最终开发出第一个实用的化学放大光刻胶（简称CAR）。这种光刻胶依靠一种称为光生酸发生剂的物质来工作。IBM在20世纪80年代初推出了这一首个CAR，名为T-Bok，用于帮助生产一兆位动态随机存取存储器（DRAM）。该光刻胶相当不稳定，需要轻柔的晶圆移动和高度的空气净化。据说后来也发生了一些有趣的事情，但最终它完成了任务。起初，IBM将他们的化学放大光刻胶作为商业机密，直到他们尝试为该发明申请专利时才发现，3M的一位科学家几年前就已经申请了类似的专利，尽管该公司并未对那项发明进行任何应用。

This forced IBM to patent only this specific implementation of the CAR concept rather than the whole concept itself. The CAR concept eventually got out and that led to other commercial photo-resis companies like Tokioka, Hosh and Olin bringing out their own versions in the early 1990s. Because of this, IBM eventually decided that it did not make sense to keep holding on to what they knew. The concept was already out there and IBM was in the middle of a great transition away from products to services. It made more sense to share some of the costs with actual photo-resis companies, so they partnered with companies to do joint development.

这迫使IBM只能给CAR概念的这一特定实现申请专利，而不是整个概念本身。最终CAR概念被传播开来，导致其他商业光阻公司如Tokioka、Hosh和Olin在1990年代初推出了自己的版本。由于这个原因，IBM最终决定继续坚持保留他们所掌握的知识已无意义。该概念已经被公开，而IBM正处于从产品转向服务的大转变中。因此，与真正的光阻公司分担一些成本更有意义，于是他们与这些公司合作进行联合开发。

The IBM Shipley-DUV resist alliance seemed fruitful and another one of those partners was JSR. With Japan's semiconductor industry falling apart in the 1990s and lacking any affiliations to major k-retzoo, JSR had no choice but to go overseas. In 2000, JSR sealed a joint research cooperation pack with IBM that let them first produce the photo-resis for the 193nm wavelength, leapfrogging the rival Tokioka. The business fundamentally changed. As I mentioned in the 1980s, 90% of the company's revenues came domestically. By 2003, the opposite situation. 70% of JSR's revenues came from overseas, places like Taiwan and South Korea. This story is repeated throughout much of Japan's chemical industry. Like LCD for instance, Japanese LCD chemical companies built their proficiency by supplying Japanese LCD companies. But when the industry emigrated to other parts of Asia in the 1990s, the chemical companies internationalized and started supplying those guys to Top 10 anime betrayal.

IBM Shipley-DUV抗蚀剂联盟看起来成果丰硕，其中的合作伙伴之一是JSR。在1990年代，日本的半导体产业逐渐崩溃，且缺乏与主要行业的关联，JSR别无选择，只能转向海外发展。2000年，JSR与IBM达成了一项联合研究合作协议，让他们率先生产193nm波长的光刻胶，超过了竞争对手Tokioka。公司的业务由此发生了根本性的转变。正如我在1980年代提到的，当时公司90%的收入来自国内。但到了2003年，情况正好相反：JSR 70%的收入来自海外，像是台湾和韩国等地。这种情况在日本的化工业中屡见不鲜。以液晶显示器（LCD）为例，日本的LCD化学公司最初通过给日本的LCD公司供货来建立自己的能力。但当该行业在1990年代迁移到亚洲其他地区时，这些化学公司也开始国际化，转而向那些地区供货，如同一幕“十大动漫背叛”般的戏剧性转变。

In 2019, Japan's global photo-resis monopoly, led by JSR and Tokioka, made the news when it got embroiled in a larger geopolitical dispute. In July 2019, the Japanese government announced certain controls on exports of three semiconductor chemicals to South Korea, fluorinated polyamide, hydrogen fluoride and UV photo-resis. Korea imports nearly 90% of its photo-resis from Japan. Japanese firms now need an individual export license to send these to South Korea rather than a broad one, which was the previous situation. Why did Japan do this? Japan cited potential re-export issues to North Korea and other unfriendly nations. Korea said it was really retaliation for them pressing on forced labor issues from during the Imperial Japanese era. Let us leave it at that.

2019年，日本在全球光刻胶市场上的垄断地位，由JSR和东京冈等公司主导，引发了新闻关注。当时，日本卷入了一场更大的地缘政治争端。2019年7月，日本政府宣布对出口到韩国的三种半导体化学品（氟聚酰亚胺、氟化氢和光刻胶）实施管制。韩国近90%的光刻胶依赖从日本进口。现在，日本企业向韩国出口这些产品需要单独的出口许可证，而不是之前的统许可。日本为什么这样做？日本称这是因为担心这些材料可能会被再出口到朝鲜等不友好国家。韩国则表示，这是日本针对他们追究日本帝国时代强制劳动问题的报复。我们就先说到这里。

In response to Japan's change, South Koreans boycotted Japanese products, canceled trips to Japan and waged war with their keyboards. The actual impact of the new licensing regime, at least when it came to the photo-resists was minimal. Only UV photo-resists were affected and Samsung and SK Highnicks had not yet then ramped up on UV. And when they did, South Koreans just switched to getting their resist from Belgium, where JSR had a joint venture with the local research organization, IMac. This dispute lasted for four years, and finally ended in 2023 when the newly elected South Korean president visited Japan for the first time in 12 years. Japanese companies like JSR and Tokiooka have maintained their market grip even as photo-resists transition into the UV era.

为应对日本的变化，韩国人抵制了日本商品，取消了赴日旅行，并在网络上展开“键盘战争”。新的许可制度对光刻胶的实际影响很小，主要是因为该措施仅涉及到紫外光刻胶，而当时三星和SK海力士尚未开始大规模使用紫外技术。后来，当他们开始使用紫外技术时，韩国便转而从比利时购买光刻胶，比利时的JSR公司在当地与研究机构IMac有合资企业。这场纠纷持续了四年，最终在2023年，韩国新任总统时隔12年首次访问日本时得以结束。尽管光刻胶技术进入了紫外时代，但日本公司如JSR和Tokiooka依然保持着市场影响力。

How? The photo-resists industry, like the semiconductor industry as a whole, is dominated by big incumbent players with tight working relationships nurtured over years. Decades even. Today's resists are custom developed between user and producer, so like the rest of the chemicals industry, the photo-resist companies have to be highly specialized and very technical. PhD spend their entire careers studying just a subset of this stuff. A TSMC executive told me that during the R&D phase, the Japanese will gleefully agree to almost anything because they know that once the node goes to high volume production, you won't dare change out the resist, and that is when they have you over a barrel.

如何？光刻胶行业与整个半导体行业类似，由少数成熟的大企业主导，这些企业经过多年的合作建立了紧密的工作关系，甚至是几十年的关系。如今的光刻胶是用户和生产商之间定制开发的产品，因此，就像化学品行业的其他部分一样，光刻胶公司需要高度专业化并具备很高的技术水平。博士们往往在职业生涯中仅研究这一领域的一个子集。一位台积电的高管告诉我，在研发阶段，日本公司会非常乐意接受几乎任何建议，因为他们知道，一旦工艺节点进入大规模生产阶段，你就不敢更换光刻胶，那时他们就可以对你予取予求。

Though I should also know that this is a two-way street. The semiconductor fabs are infamous for constantly pressuring their suppliers for better pricing. Nature of horrors of vacuum, TSMC of horrors of sole supplier. As one major reason why the Japanese eventually outcompeted the Americans and Europeans in this space, they are far more willing to endure lower prices and ROIs for longer periods of time. Director customers also demand incredible unprecedented purity and quality, as it was one drop of impurity amidst the equivalent of two Olympic-sized swimming pools is more than enough to ruin everything and the fab won't take the shipment.

虽然我也应该知道这是一条双向道路。半导体工厂以不断向供应商施压以获得更好价格而闻名。自然界憎恶真空状态，台积电也憎恶独家供应商。日本最终在这个领域击败美国和欧洲的一个重要原因是，他们更愿意在更长时间内忍受较低的价格和投资回报率。主要客户也要求极高的纯净度和质量，因为即使是相当于两个奥林匹克规模游泳池中的一滴杂质都足以毁掉一切，工厂将拒绝接受这批货物。

Another example. Early in its use, IBM's T-Bok failed due to a skin forming on the layer after development. It was caused by traces of paint and liquid cleaner in the clean room, as well as ammonia on human skin and the fertilizer in the lawns outside the fab. Something that I should mention is that for all of its strategic importance, the whole photo-resists industry is a niche. In 2019, the entire photo-resists industry was worth about $1.3 billion. It's grown some since then thanks to surges in EUV photo-resists, but JSR's former chairman has joked before that Japan's ramen noodle industry is much larger. JSR made just $3 billion in revenue in 2023 and over half of that is pharmaceuticals and plastics, Tokyooka $1.1 billion.

另一个例子。IBM的T-Bok在早期使用时失败了，因为在显影后层面上形成了一层薄膜。这是由洁净室中残留的油漆和清洁液引起的，还有人类皮肤上的氨和工厂外草坪上的肥料导致的。我需要提到的是，尽管具有战略重要性，整个光刻胶行业实际上是一个小众市场。2019年，整个光刻胶行业的市场价值约为13亿美元。由于EUV光刻胶的需求增加，这个市场自那时起有了一些增长，但JSR公司的前任董事长曾开玩笑说，日本的拉面行业要大得多。JSR公司在2023年的收入仅为30亿美元，其中超过一半来自制药和塑料，Tokyooka的收入为11亿美元。

These companies are also not very profitable. JSR's net profit margin in 2023 was 3.8% and Tokyooka is about 7.8%. Being neither big or profitable opens these companies up to shareholder activism or that most googlers show financial entities private equity. The FT reported that in 2022, the German company Merck tried to acquire JSR. JSR said no and Merck was like okay, but then two private equity companies approached JSR attempting to either buy out the company or do some of that quote unquote value creation. In response, JSR approached the Japanese government and in 2023, a state controlled buyout fund called JIC bought out a majority stake of JSR for $6.4 billion, taking it private.

这些公司也不是特别赚钱。2023年，JSR的净利润率为3.8%，而Tokyooka大约为7.8%。由于既不大也不赚钱，这些公司容易成为股东激进主义的目标，或者被财务实体或私募股权所关注。《金融时报》报道，2022年，德国公司默克尝试收购JSR，但被拒绝。随后，两家私募股权公司接触JSR，试图收购该公司或进行所谓的“价值创造”。为了应对这种情况，JSR向日本政府求助。2023年，一个名为JIC的由国家控制的收购基金以64亿美元的价格收购了JSR的多数股份，使其成为私有公司。

This was weird. The Japanese government has bought entire companies before, but those companies were on the verge of collapse. JSR was far from that, but it shows that the government recognized how insane it would be to let private equity touch this company and it shows a willingness to protect one of Japan's last strongholds and semiconductors. It was a bold move, we shall see if it holds up over time.

这件事很奇怪。日本政府以前也曾收购过整个公司，但那些公司都已濒临破产。而JSR公司并没有陷入这种困境。这说明政府意识到让私募股权接触这家公司会是多么疯狂的事情，同时也表现出政府愿意保护日本最后的几个重要领域之一——半导体领域。这是一个大胆的决定，我们拭目以待它是否能长久奏效。