China-based firm delivers its first chipmaking tool that stamps nanoscale chip designs onto wafers
Given Substrate's achievements so far, we are likely dealing with soft X-rays (wavelengths of 1-10 nm, lower energy) rather than hard X-rays (wavelengths of 0.1-1nm, higher energy).
Since short-wavelength light (including EUV and X-rays) is strongly absorbed by most materials, managing it requires a set of perfectly polished mirrors that reflect light at grazing angles (to avoid absorption), ultra-precise alignment, and vacuum environments. Also, X-ray lithography requires all-new resists that can handle high-energy photons without damage or blur.
To prove that its XRL method works, Substrate has shown off images of a random logic contact array with 12nm critical dimensions (CD) and 13nm tip-to-tip (T2T) spacing printed with high pattern fidelity, as well as random vias with a 30nm center-to-center pitch, possessing superb pattern quality and critical dimension uniformity. If such metrics could be achieved for mass production today, this would largely revolutionize the lithography industry, as it would enable scaling across both axis at 2nm-class nodes (and lower) without using multi-patterning.
Modern EUV scanners with 0.33 NA optics can achieve critical dimensions of 13nm–16nm in high-volume manufacturing, which is sufficient to print a 26nm minimum metal pitch (good enough for 2nm or 3nm-class process technologies) and a 25nm T2T interconnect space with a single exposure.
Such disproportions emerge because chipmakers tend to optimize resolution in the Y direction (CD) to get the tightest metal-pitch line-space pattern, but at the cost of resolution in the X direction, which means that T2T prints poorly or inconsistently, leading to bridging defects, stochastic defects, yield loss, complicated design rules, and slower scaling. To mitigate this and avoid blurred or inconsistent line ends at tip-to-tip spacing, Intel applies pattern-shaping tools in the X-direction with its 18A fabrication technology, but this complicates the overall production flow and does not fundamentally solve the issue.
Substrate's tool (assuming these are real lab results, not a simulation) can already outperform existing Low-NA EUV scanners in terms of achievable CDs with single-resolution patterning, and it leaves them behind dramatically when it comes to T2T spacing printed with high fidelity. This means that Substrate's X-ray lithography tool could possibly replace costly EUV multi-patterning used for sophisticated 3nm and 2nm-class process technologies or pattern shaping used for Intel 18A.
Our friends at SemiAnalysis have managed to get more performance claims from Substrate, which look even more impressive. The company claims it has achieved overlay accuracy of under 1.6nm, full wafer critical-dimension uniformity (CDU) of 0.25 nm, line edge roughness (LER) of under 1nm, and local critical dimension uniformity (LCDU) below 1.5 nm.
If accurate, this performance would match or surpass ASML's Twinscan NXE:3800E in uniformity, though its overlay precision is slightly worse than the 0.9nm machine-matched overlay standard in the latest EUV scanners. Also, the line-width uniformity of contacts on an image provided by Substrate is rather poor.
Assuming the results presented by Substrate are real and achieved in a lab environment, this means the company has solved three critical challenges with X-ray lithography. First, build a light source featuring an electron gun and a particle accelerator; second, create a grazing-incidence mirror system to reflect and focus X-rays at very shallow angles; and make the whole thing compact enough to fit into a lab.
However, Substrate still has a lot of work to do, turning its X-ray lithography technology from a lab success into a viable production tool. Substrate must prove that its X-ray lithography system can maintain beam stability, optical precision, resist compatibility, overlay accuracy, and commercial throughput simultaneously, something no X-ray platform has ever achieved.
Existing photoresists are incompatible with X-ray radiation, as they are optimized for EUV radiation with considerably lower photon energy. So, Substrate will have to invent a proper resist and then produce it at volume. The company will also have to develop photomasks that can sustain X-ray radiation. Grazing-incidence mirrors for X-rays are also not in mass production, and it is unknown whether they can be mass-produced cheaply and reliably by existing producers like Zeiss.
Substrate will also have to ensure that X-rays do not damage the underlying transistors or introduce stochastic defects. Achieving overlay accuracy below 1nm (to match ASML's production-level alignment precision) remains another challenge for the company. This is perhaps because the company still has to address issues such as wafer handling, stage repeatability, and other factors related to high-precision mechanics, which ASML has taken decades to solve.
Beyond that, the tool must reach commercial throughput and consistent yield, something that took years for ASML's EUV tools. In fact, ASML's EUV journey timeline is quite exemplary: it has taken the industry 12 years to evolve from an alpha demo tool (2006) to mass production (2018), and about seven years to go from the first pre-production system (2010) to a mass-production-capable scanner.
Speaking of mass-production-capable X-ray lithography tools, it is important to note that Substrate has no intention of selling them to third parties such as Intel or TSMC. Instead, Substrate plans to build its own fabs in the U.S. (a move that could give the company geopolitical importance in the eyes of the U.S. government), install additional tools, and offer foundry services, thus challenging existing chip contract manufacturers.
However, this strategy adds complexity and cost. Constructing even a single high-end semiconductor fabrication plant would require tens of billions of dollars in investment and a large ecosystem of suppliers and service infrastructure, which currently does not exist for X-ray lithography production.
Substrate would also need to integrate its XRL litho machines with hundreds of other tools in the fab, or persuade its suppliers (such as Applied Materials, KLA, Lam Research, etc.) to help it do so, which likely involves further investments from the company, making its first fab particularly expensive.
Also, running both a toolmaking activity and a chip foundry would stretch Substrate's technical and financial resources, which would make it particularly hard to achieve its promised per-wafer price of $10,000 by the end of the decade, as its investors will likely demand returns after pouring tens of billions of dollars into the company.
However, if Substrate succeeds in both roles, it could shift the balance of the semiconductor supply chain back to the U.S., as the company will likely outpace ASML's tools in terms of resolution and performance, and TSMC in terms of design cycle time and potentially volume.
Follow Tom's Hardware on Google News , or add us as a preferred source , to get our latest news, analysis, & reviews in your feeds.
Anton Shilov Social Links Navigation Contributing Writer Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/tech-industry/semiconductors/SPONSORED_LINK_URL
- https://www.tomshardware.com/tech-industry/semiconductors/american-startup-substrate-promises-2nm-class-chipmaking-with-particle-accelerators-at-a-tenth-of-the-cost-of-euv-x-ray-lithography-system-has-potential-to-surpass-asmls-euv-scanners#main
- https://www.tomshardware.com
- Taiwan increases penalty for damaging undersea cables amid increased suspected sabotage attempts — offenders face up to 7 years in prison and $325,000 in fines
- Save $40 on 1TB of Switch 2 storage at Costco — Lexar's Play Pro microSD Express card drops to $179.99
- OpenAI and Microsoft sign agreement to restructure OpenAI into a public benefit corporation with Microsoft retaining 27% stake — non-profit 'Open AI Foundation'
- Samsung unleashes high-speed P9 Express microSD Express cards for the Switch 2 at highly competitive prices — 256GB and 512GB available for $54 and $99
- NVIDIA Launches Omniverse DSX Blueprint, Enabling Global AI Infrastructure Ecosystem to Build Gigawatt-Scale AI Factories
Informational only. No financial advice. Do your own research.