Five immersion scanners this year, twenty next year. Speculation about an imminent wave of Chinese lithography systems is causing turbulence in the stock market. In the industry itself, however, there’s little reason for alarm.
An unnamed Chinese equipment manufacturer will build five immersion scanners this year, according to reports in the US media. The stock market reacted immediately, pushing shares of ASML as well as ASM, Besi and US equipment manufacturs in the red. Investors realize that China may be about to overcome another major technological barrier. The country also appears to be making rapid progress in other strategic semiconductor technologies, such as atomic layer deposition and hybrid bonding. Will lithography be next?
My first reaction is that the news primarily served as the trigger for a correction in what had become an extremely overvalued semiconductor sector. But that’s no reason to dismiss Chinese efforts to make inroads in the semiconductor lithography arena.
The Chinese company wasn’t identified. Based on an anonymous source, Reuters reports it’s Shanghai Aishengna Electronic Technology Group, “a Chinese state-owned company” that’s leading the production effort “after incorporating teams from top Chinese lithography startups.”
“The Shanghai-based company” can also be Shanghai Micro Electronics Equipment (SMEE), a lithography and inspection equipment manufacturer heavily backed by the Chinese government. Western media already reported at the end of 2023 that SMEE had completed the development of its 28 nm immersion scanner, the SSA800-10W. It’s possible that working systems were subsequently delivered to the Chinese foundry SMIC and to research institutes. If the reports refer to five immersion machines now under construction, they’re most likely working prototypes.
Production-worthy scanners seem highly unlikely at this stage. These systems may well be capable of printing patterns while still leaving thousands of defects on every exposure field. The SSA800 doesn’t yet appear in the product overview on SMEE’s website.
Resolve
The Chinese shouldn’t be underestimated. SMEE has already developed i-line lithography systems used for mature 0.35-micron process nodes. Besides chipmakers, OSAT companies also use these steppers in back-end manufacturing, for example to create through-silicon vias (TSVs). What has stood out in recent years is the reported reliability of these machines. Well-informed and trusted industry sources say they perform at least as well as, and perhaps even better than those of Veeco (formerly Ultratech), the back-end lithography market leader.
SMEE has since spun off its i-line business under the name AMIES. By the end of last year, the company claimed to have an installed base of no fewer than 500 steppers.
Canon demonstrated during the Covid pandemic that it’s possible to gain a foothold at established ASML customers. It dealt Veldhoven a psychological blow by securing major i-line scanner orders from TSMC. ASML couldn’t supply enough tools during the pandemic, allowing Canon to re-establish itself at the Taiwanese foundry with its i-line steppers. This will have strengthened the resolve of Chinese firms.
Five and twenty
According to the reports, Shanghai will build five immersion scanners this year and twenty next year. Given the difficulties SMEE encountered with 248nm and dry 193nm systems, it seems unlikely that the company will master this extraordinarily complex technology quickly.
SMEE has a respectable track record in the development of lithography equipment. In the early 2010s – before the current technological sovereignty craze – ASML evaluated whether SMEE could become a supplier. At the time, the company was considered insufficiently mature. That’s why I find it highly unlikely that a complete unknown such as Shanghai Aishengna could suddenly bring a production-worthy immersion scanner to market.
Immersion lithography is far more complicated than simply placing a thin layer of water between the projection lens and the wafer. After the first prototype scanners had been built, it still took ASML years to eliminate the defects that prevented high-volume manufacturing. Many of the solutions developed during that period were patented, creating substantial protection around the technology.
Paradox
That creates an interesting paradox, however. Those patents protect strategic know-how in Western markets, but they’re also publicly available and effectively serve as technical textbooks for technology followers and provide valuable guidance.
One example is the problem of air bubbles. The rapid motion of the wafer stage and the need to confine the tiny puddle of water between the lens and the wafer create bubbles that interfere with imaging. The very first wafers exposed on ASML’s immersion prototype scanners at research institutes contained more than 30,000 defects.
One solution was a patented carbon dioxide gas curtain that helped transport and stabilize the water layer. Chinese engineers may well have studied this technology. Simply copying patented solutions would prevent them from selling such systems internationally, but the patents themselves may have pointed them toward alternative approaches.
If China succeeds in developing a competitive immersion scanner, TSMC and Samsung will undoubtedly follow that development with great interest. At the very least, it would strengthen their leverage in negotiations with ASML.
How about dry?
There’s another question: what about SMEE’s dry 248nm and 193nm scanners? Those technologies are considerably less complex, yet, as far as is known, not a single one has been adopted in a Western fab. Even for those systems, the learning curve is substantial – let alone for immersion lithography.
If leading chipmakers ever wanted to introduce immersion scanners from Shanghai, those machines would also have to integrate seamlessly with fleets of existing ASML systems to achieve machine-matched overlay. That’s not merely a hardware issue; it requires years of calibration, process control and software optimization. It’s one of the reasons Nikon has never managed to make a significant dent in ASML’s market share, despite offering at least as-capable yet cheaper equipment.
Throughput
One factor that Chinese sources never seem to report is throughput. They almost always focus on nanometer precision, critical dimensions or overlay performance. It’s entirely possible that SMEE or even Shanghai Aishengna can produce 28 nm patterns on prototype immersion systems or in advanced laboratory setups. They may even achieve good overlay. But producing a defect-free image is only part of the challenge. That image must also be printed at high speed.
Once you start pushing throughput and increase the speed of the wafer, the real challenges emerge. Accuracy begins to degrade, bubbles form in the immersion fluid and tiny trails of water droplets are left behind on the silicon, compromising the precise alignment on which immersion lithography and especially multiple patterning depend.
So, even if the Chinese already have a working immersion scanner, it’s likely producing only a handful of wafers per hour. A commercially viable system needs a throughput of around one hundred wafers per hour, while ASML’s most advanced immersion scanners process well over three hundred wafers in the same time.
Taking all these factors together, the gap remains enormous. Even if China succeeds in building five immersion scanners this year, they’re nowhere near matching the performance of ASML’s production systems.
ASML has also benefited for decades from an enormous learning curve. Immersion lithography wasn’t invented overnight; it required years of debugging, process optimization and customer learning before it became a mature production technology. Even if SMEE has a working immersion scanner today, reproducing that accumulated operational experience may prove far more difficult than reproducing the optics or mechanics themselves.
For Veldhoven, the message is simple: Keep a cool head, but remain relentlessly vigilant.
Top image credit: ASML
