"Japan's Glue, China's Soil, EUV for Reverse Engineering"
These past few days, there has been considerable discussion about China's Manhattan Project—the reverse engineering of ASML lithography machines. Everyone is excited, as it represents a major breakthrough.
However, from an engineering perspective, setting aside ideology, a closer look reveals a significant gap between reality and ideal.
Because if we truly break down the issue at the engineering level, we find that what determines the long-term sustainability of a technological path is often not the hottest media headlines, but rather seemingly insignificant details.
This article will not discuss the details of the progress of reverse lithography machines, but will only discuss another recent, less prominent news item: Japan's (limited) supply of photoresist to China.
In advanced processes, photoresist determines the width of the process window, whether random defects can be suppressed, and whether the yield tail will get out of control.
In other words, even with EUV, without photoresist, the yield of advanced processes cannot be improved. Without yield, costs cannot be reduced. Without reduced costs, there are no orders, at least no international orders.
This is why photoresist is often more concealed than the equipment itself, but it's more fatal to engineering.
Few amateur scientists understand the difficulty of replicating Japanese photoresist, habitually assuming it's just a consumable in the photolithography process. But the reality is far more complex.
High-end Japanese photoresist is extremely difficult to replicate because it involves not only certain chemical structures, but more importantly, the entire production and control process—a linear, non-iterative process.
From ultra-high purity raw material control, polymerization reaction path selection, and molecular weight distribution management, to impurity statistics, batch consistency, and long-term aging behavior, it's a highly engineered, long-evolving system.
It's been built up slowly through decades of failed samples.
This has resulted in a large number of patents, but more importantly, many key judgments cannot be written into academic papers or fully encapsulated in patents.
They reside in engineers' intuition about whether a batch of material can be used on the production line, in the production line's experiential judgment of anomalies, in the company's decades of process parameters and failure data, and in the accumulated improvements to processes and controls.
This is the true meaning of "Japanese photoresist."
This isn't just about a single product; it's about an entire, long-term, operational materials industry capability.
China has a very interesting, but often overlooked, reference point in this regard: its heavy rare earth refining and processing capabilities.
What's truly difficult to replicate isn't the resource itself, but the process system for separating, purifying, and stabilizing complex minerals to an engineering-ready state.
This is a process of countless failures and repeated trial and error, consuming vast amounts of resources and generating significant pollution.
While Europe and America aren't lacking in rare earth resources, the real challenge lies in transforming "earth" into scalable, controllable, and long-term available industrial materials.
That, too, requires a highly engineered and long-accumulated capability.
This is why China's "rare earth" resources can be used to control Europe and America.
Even more interestingly, Shin-Etsu Chemical, one of Japan's main suppliers of photoresist, is also one of the few producers in Europe, America, and Japan with heavy rare earth refining capabilities. (We'll discuss in detail why Shin-Etsu Chemical can produce both photoresist and rare earths in a future article.)
Similarly, Japan's photoresist technology can also hold China hostage.
The disruption of Japanese photoresist supply affects stability and yield.
Japan holds an absolute monopoly on EUV lithography machine mass production. This means that even with EUV, without Japanese photoresist, it's impossible to produce chips at 5nm and below.
Even in China's current 7nm production process, although it doesn't use EUV but rather 193nm ArF DUV with multiple exposures, yield is still limited by high-end Japanese photoresist: For non-critical layers, domestically produced photoresist can be used stably; for secondary critical layers, a mix of domestic and imported photoresist can be used; however, the truly critical layers still heavily rely on high-end Japanese ArF photoresist.
This is because multiple exposures amplify even the smallest instabilities.
Once the supply of high-end photoresist is cut off, the already low yield of 7nm will further decrease, and costs will further increase.
Why is EUV so dependent on Japanese photoresist? Because only mature photoresist can suppress photon statistical noise and random defects.
Achieving truly controllable domestic substitution for the 7nm key layer may still require multiple R&D cycles, not to mention the EUV photoresist needed for 5nm.
One such R&D cycle typically takes 3-5 years because it must include at least five time-consuming stages that are almost impossible to parallelize: Basic formulation exploration; Laboratory → Pilot-scale amplification; Equipment integration (Scanner + Track); Production line verification (Wafer level); Long-term stability verification. The components of the equipment may be explicit, but the materials and processes are implicit; machines can be disassembled and replicated, but the time required for materials and industrial processes cannot be compressed by reverse engineering.
Japanese photoresist, like China's heavy rare earth elements, is a capability whose true value is only realized on the production line, at the tail end of the yield curve, and through years of stable operation without incidents.
If reverse EUV might "open the door," then photoresist determines whether one can persevere on this path.
And the cruelest aspect of the semiconductor industry is that—a single success is meaningless; only years of problem-free operation count as success. And time, even in the age of AI, is the only thing that cannot be reverse engineered.
