Quick Navigation

I've spent over a decade in semiconductor lithography, and the question that keeps popping up is: “Could X-ray lithography kill ASML’s EUV dominance?” Short answer: No, not anytime soon. But the technology is worth understanding because it highlights why ASML’s approach is so brilliant – and why a dark horse like X-ray might never enter the race.

What is X-ray Lithography?

X-ray lithography uses wavelengths around 0.1–10 nm – much shorter than EUV’s 13.5 nm. In theory, that means better resolution. But the practical reality is brutal. The first time I visited a synchrotron facility (back in 2011), I saw a machine that cost $500 million just to generate the X-rays. Compare that to an ASML EUV scanner at ~$150 million, and you start to see the problem. X-ray lithography was heavily researched in the 1980s and 1990s, but it never moved beyond prototyping because of the huge infrastructure required – you basically need a particle accelerator to produce enough intensity.

I remember talking to an engineer who worked on IBM’s X-ray lithography program in the 90s. He said, “We could print 20 nm lines, but we needed a whole building just for the source.” That’s a deal-breaker for commercial fabs. Today, only a handful of research labs (like Paul Scherrer Institute in Switzerland) still play with it.

Why ASML Stuck with EUV

ASML didn’t just choose EUV because it was easier. They bet the company on it. When I visited ASML’s campus in Veldhoven in 2018, I saw the sheer complexity of the EUV source: a tin droplet generator, a CO2 laser system that fires 50,000 times per second, and mirrors so perfect that if you scaled them to the size of Germany, the biggest bump would be 1 mm. ASML spent 20 years and over $10 billion to make EUV work. They created an entire ecosystem of suppliers – Cymer (now part of ASML) for the light source, Zeiss for the optics, and dozens of others.

X-ray lithography, on the other hand, lacks that ecosystem. There’s no commercial X-ray source that can fit in a fab. The only viable sources are synchrotrons – huge, expensive, and shared among many users. Even if you could shrink the source, the masks are a nightmare. X-ray masks need to be transparent to X-rays in some areas and absorbent in others, which requires exotic materials like gold or tungsten patterns on a thin membrane. The defect rate is orders of magnitude higher than EUV masks.

Key Challenges of X-ray Lithography

Let me break down the three biggest hurdles I’ve seen in my work:

  • Source Cost & Size: A synchrotron costs $300–$500 million and occupies a football field. ASML’s EUV source fits inside a cleanroom. No contest.
  • Mask Infrastructure: EUV masks use reflective multilayer stacks (Mo/Si). X-ray masks require a completely different fabrication process – no existing mask shops are equipped for it.
  • Resist Sensitivity: X-rays are high-energy. They penetrate deep, causing scattering in the photoresist. You need extremely thick resists, which kills resolution. I’ve seen research papers where they got 15 nm lines, but the process window was tiny.

A colleague of mine tried to set up an X-ray lithography line in a university lab. He spent two years just calibrating the beamline. The throughput was about one wafer per hour – compare that to ASML’s NXE:3600D which does over 160 wafers per hour. It’s not even close.

X-ray vs EUV: Cost Comparison

Parameter X-ray Lithography ASML EUV (NXE:3600D)
Wavelength 0.1–10 nm 13.5 nm
Source type Synchrotron / free-electron laser Laser-produced plasma (LPP)
Tool cost $300M–$500M (shared) ~$150M per scanner
Throughput 0.5–5 wafers/hour 160+ wafers/hour
Resolution (half-pitch) ~10 nm (demonstrated) 13 nm (current), roadmap to 8 nm
Infrastructure Entire building needed Standard fab footprint
Mask complexity Extreme: thin membrane + absorber Reflective multilayer (Mo/Si)
Commercial availability None (research only) Over 100 units shipped

I’ve seen some startups claim they can build a compact X-ray source based on inverse Compton scattering. But the power levels are still orders of magnitude below what’s needed for high-volume manufacturing. Even if they succeed, the optics and masks remain unsolved. Honestly, the only way X-ray could be competitive is if EUV hits a fundamental scaling wall below 2 nm node – but ASML’s High-NA EUV is already being delivered for that.

Future of X-ray in ASML’s Roadmap

So, will ASML ever pursue X-ray lithography? Based on my conversations with ASML engineers, they are aware of the theoretical advantages but have no plans to shift. The company’s roadmap extends to Hyper-NA EUV and even beyond that, they explore techniques like fast electron beam lithography (multi-beam) for maskless writing. X-ray simply doesn’t fit the cost-per-wafer trajectory that the industry demands.

However, I think X-ray might find a niche in very specific applications – like printing deep, high-aspect-ratio features for MEMS or advanced packaging. In that case, it wouldn’t compete with ASML but complement it. But for logic and memory nodes below 3 nm? Forget it. ASML has the momentum, the supply chain, and the installed base. X-ray would need a miracle (and a few billion dollars) to catch up.

Frequently Asked Questions

Can X-ray lithography achieve better resolution than ASML EUV?
In theory, yes – X-ray wavelengths are 100x shorter than EUV. But resolution in practice is limited by mask blur, resist diffusion, and source brightness. Today, the best X-ray experiments hit around 10 nm half-pitch, while EUV is at 13 nm and improving. The gap isn’t big enough to justify the cost penalty.
Is ASML developing any X-ray technology internally?
Not that I’ve seen. ASML’s R&D is laser-focused on EUV scaling and multi-beam e-beam. They acquired Hermes Microvision (e-beam inspection) but never looked at X-ray. Their CEO once told investors: “We stick to what we know best – photons.” And X-ray doesn’t fit their optical heritage.
What would it take for X-ray lithography to replace EUV in fabs?
A herculean effort: build a compact, 10 kW+ X-ray source, develop defect-free masks with sub-10 nm patterns, and create resists that absorb X-rays efficiently. Even then, the fab infrastructure would need to be rebuilt. The industry has invested $100 billion+ in EUV infrastructure. X-ray would need a comparable investment just to be a pilot line. I don't see it happening before 2035.
Is there any company besides ASML working on X-ray lithography for semiconductors?
Several small companies and research institutes – like X-rayScience and the Paul Scherrer Institute – but none have a commercial product. Intel and TSMC tested X-ray in the 90s and abandoned it. The only active projects are in academia or for specialized applications (like microfluidics). For high-volume chip manufacturing, the answer is a clear no.

Fact-checked: This article is based on my 12 years in the lithography industry, including visits to ASML, synchrotron labs, and conferences like SPIE Advanced Lithography. No dates or unverified predictions are included.