Semiconductor manufacturers now print transistor features using light with a wavelength of roughly 13.5 nanometers, a dimension so small that a single coronavirus particle, measuring 80 to 100 nanometers across, dwarfs the beam doing the work. That gap between biology and engineering captures a quiet revolution in how chips are made: the shift from older 193-nanometer lithography to extreme ultraviolet (EUV) systems forced the industry to abandon glass lenses entirely, replace them with multilayer mirrors, and generate light by vaporizing tin with high-powered lasers. The result is visible in the devices people carry every day, where fin and metal pitches have dropped to about 30 nanometers, packing more transistors into less space while trimming power consumption.
Why 13.5-nanometer light rewrites the rules of chipmaking
For decades, chipmakers shrank transistors by using progressively shorter wavelengths of deep-ultraviolet light. The workhorse was 193-nanometer lithography, but by the mid-2010s that wavelength could no longer reliably define features at the densities chip designers demanded. The industry’s answer was EUV, operating at roughly 13.5 nanometers, a jump of more than 14 times shorter than the prior generation. NIST’s metrology program documents this transition as a direct response to feature-size scaling limits that 193-nanometer tools could not overcome, no matter how many patterning passes engineers stacked on top of one another.
The hypothesis that EUV would deliver energy-per-switching gains beyond what simple wavelength scaling predicts rests on a practical reality: fewer patterning steps mean fewer deposited and etched layers, which in turn means thinner interconnects and tighter pitches. Peer-reviewed analysis in Nature Electronics shows that EUV-enabled processes have achieved fin pitch and minimum metal pitch of roughly 30 nanometers. Shorter wires and smaller capacitances reduce the energy each transistor switch consumes, an effect that compounds across billions of transistors on a single die. The gains are real, but isolating how much comes from the wavelength change versus simultaneous advances in transistor architecture and materials remains an open measurement problem, because no published study has controlled for all variables at matched densities.
Tin plasma, multilayer mirrors, and the physics of 13.5-nanometer light
Producing usable EUV light is an engineering challenge with no precedent in earlier lithography generations. ASML, the sole manufacturer of production EUV scanners, described the process in its 2023 annual report filed with the U.S. Securities and Exchange Commission: a high-powered carbon-dioxide laser strikes tiny tin droplets tens of thousands of times per second, creating a plasma that emits light at 13.5 nanometers. Independent peer-reviewed research has confirmed the same laser-produced tin plasma mechanism as the basis for high-volume manufacturing EUV sources.
At 13.5 nanometers, conventional glass optics are useless. The photons are absorbed by virtually all solid materials, so NIST researchers have documented that EUV projection systems rely on specialized mirrors rather than lenses. Each mirror consists of dozens of alternating layers of molybdenum and silicon, each only a few nanometers thick, tuned to reflect the target wavelength through constructive interference. Even so, each mirror surface absorbs a significant fraction of the incoming light, and the optical column contains multiple mirrors in series. The cumulative loss means only a small share of the original photons reach the wafer, placing extreme demands on source brightness and mirror cleanliness.
To put the scale in perspective, virology measurements of coronavirus particles with spikes found diameters of roughly 80 to 100 nanometers. The light used to print modern chip features is about six to seven times narrower than a single virus particle. That ratio helps explain why contamination control inside EUV scanners is so demanding: particles that would be invisible in older lithography systems can block or scatter enough photons to ruin a pattern at this scale. The same kinds of sub-micrometer aerosols that matter for infection control in public-health settings, documented in resources from the U.S. Centers for Disease Control, become a yield risk when they drift through an EUV tool’s optical path.
Gaps in the public record on EUV yield and cost
Despite the technical achievements, several questions lack public answers. No primary-source data on fab-level yield or throughput tied directly to 13.5-nanometer production lots has been published in regulatory filings or peer-reviewed journals. ASML’s public reports discuss the technology but do not disclose scanner uptime percentages, defect densities, or per-wafer maintenance costs in a way that outside analysts can benchmark. Tool users occasionally hint at learning curves on earnings calls, yet those comments are qualitative and often bundled with broader process-node ramp discussions, making it impossible to attribute outcomes solely to EUV exposure.
Metrology facilities provide another example of partial transparency. NIST’s Synchrotron Ultraviolet Radiation Facility (SURF) supports EUV mirror and resist characterization, but only a subset of its results reach the open literature, often in the form of summary plots or derived metrics rather than full raw datasets. Without wafer-scale statistics on mirror reflectivity drift, contamination rates, and cleaning cycles, independent researchers cannot construct a bottom-up cost model that connects optical degradation to downtime and scrap. The result is a patchwork understanding in which the physics of EUV are well described, while the economics remain largely anecdotal.
There is also little quantitative information about how EUV interacts with the rest of the process stack. Public sources rarely specify how many EUV layers a given node actually uses, how often multi-patterning is still required even at 13.5 nanometers, or how defectivity varies between EUV and legacy deep-UV layers on the same product. Without these details, it is difficult to test claims that EUV simplifies process integration enough to offset its higher tool costs and power consumption.
What better data could unlock
More granular disclosure would not just satisfy academic curiosity; it could materially influence how the industry allocates capital. If independent analyses could show, for example, that EUV layers deliver a specific percentage yield improvement at a given design rule, or that certain classes of defects dominate at particular pitches, chipmakers could target their investments in resists, pellicles, or in-line inspection more effectively. Likewise, a clearer picture of uptime and maintenance burdens could inform whether it is more economical to deploy additional scanners or to push existing ones closer to their theoretical throughput limits.
Regulators and policymakers would also benefit from a firmer empirical foundation. EUV tools consume substantial electrical power and require intricate supply chains for optics, lasers, and vacuum components. Quantitative data on tool efficiency and material utilization could shape incentives for more sustainable fab designs, from waste-heat recovery to more efficient gas handling. In an era when semiconductor capacity has become a strategic priority for many governments, understanding how much performance-per-watt and performance-per-dollar EUV truly delivers is essential for crafting effective industrial policy.
Finally, richer public datasets could catalyze innovation beyond the current dominant players. Universities and smaller firms, which lack direct access to leading-edge fabs, could use open yield and cost models to explore alternative architectures, new resist chemistries, or novel optical schemes that complement or eventually succeed EUV. The history of lithography shows that each generational shift, from mercury lamps to excimer lasers to today’s tin plasma sources, opened space for fresh ideas once the underlying parameters were widely understood. EUV has transformed how chips are made; making its economic and reliability contours equally visible would help determine what comes next.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
