Inside today's most advanced semiconductor chips, electronic circuits are etched with linewidths of just a few nanometers. EUV (extreme ultraviolet) lithography systems, which transfer these ultra-fine patterns onto silicon wafers, are a core technology underpinning the modern information society. Yet operating a single EUV lithography machine requires more than 1 MW of power. Reports indicate that TSMC, the world's largest contract chipmaker, now consumes an amount of electricity approaching roughly 10% of Taiwan's total power supply, making reductions in power consumption an urgent priority from both a manufacturing-cost and environmental standpoint.
An international research team led by Team Director Eiji Takahashi of the RIKEN Center for Advanced Photonics and Professor Takeshi Higashiguchi of Utsunomiya University has successfully demonstrated a multi-laser irradiation method—simultaneously firing multiple laser beams—in an experiment extracting EUV light from tin using a 2µm-band solid-state laser. The lead author is Naoki Nagahama, a master's student at Utsunomiya University's Graduate School. The paper was accepted on July 19, 2026, and published in the peer-reviewed optics journal Optics Letters (DOI: 10.1364/OL.607856). RIKEN officially announced the results on August 25, 2026.
The Power Wall: Over 1 MW Per EUV Lithography Machine
Shrinking semiconductor circuit patterns further requires exposure using EUV light with an extremely short wavelength of 13.5 nm. In today's mass-production EUV lithography systems, a high-power laser irradiates tiny droplets of tin supplied inside a vacuum chamber, generating high-temperature, high-density tin plasma that emits light at the 13.5 nm wavelength.
The driving light source widely used to excite this plasma is a carbon dioxide () laser operating at a wavelength of 10.6 µm. While CO2 lasers can readily achieve high output, their efficiency in converting input electrical energy into laser light is only a few percent. This drive laser for the EUV light source is a major contributor to power consumption, and it is one reason the entire lithography system draws more than 1 MW per unit.
As next-generation High-NA (numerical aperture) and Hyper-NA lithography systems demand even greater source power, there has been strong pressure to reduce laser power consumption. This has driven worldwide research into replacing CO2 lasers with 2µm-band solid-state lasers, which offer superior electrical-to-optical conversion efficiency.
However, 2µm solid-state lasers have not yet reached the output levels required by EUV lithography systems, and practical implementation remains uncertain. As the paper's abstract notes, scaling EUV source power for next-generation lithography requires reducing per-pulse energy while simultaneously increasing conversion efficiency. A new physical approach was needed to achieve high optical conversion efficiency while keeping the laser energy per pulse low.
From Single-Beam to Simultaneous Dual-Beam Irradiation
Rather than relying on a single ultra-high-power laser, Team Director Takahashi and colleagues attempted an approach that combines multiple relatively lower-power lasers fired together.
In the experiment, a holmium-doped yttrium aluminum garnet (Ho:YAG) solid-state laser—operating at a wavelength of 2,090 nm, a pulse width of 20 ns, and a repetition rate of 1 kHz—was focused onto a flat tin metal target placed inside a vacuum chamber.
First, under single-beam irradiation conditions, the tin plate was heated at a normal incidence angle of 0 degrees, a total energy of 40 mJ, and a focal spot diameter of 40 µm. Varying the laser intensity, the team found that EUV conversion efficiency peaked at 2.6% under a focusing condition of approximately . At this point, spectral purity—the fraction of emission concentrated in the band around the 13.5 nm wavelength—reached a maximum of 8%, confirming that unwanted out-of-band emission was strongly suppressed.
Next, keeping the total laser energy fixed at 40 mJ, the researchers split the beam into two (20 mJ + 20 mJ) and irradiated the tin flat-plate target simultaneously from incidence angles of ±30 degrees. By adjusting each beam's focal spot diameter to 30 µm, they maintained the same surface laser intensity of as in the single-beam case.
As a result, EUV conversion efficiency rose from 2.6% with a single beam to 3.6% with two beams—an approximately 40% improvement in efficiency. Notably, total energy and irradiation intensity were held constant; only the geometric configuration—splitting the beam and irradiating from multiple directions simultaneously—was changed, and efficiency still improved. This 3.6% figure represents the highest reported value to date for tin flat-plate target experiments using a 2µm driver. Radiation-hydrodynamic simulations also calculated a conversion efficiency of approximately 3.2%, supporting the experimental trend. In these calculations, the plasma's electron temperature was estimated at about 30 eV, with an electron density of roughly .
| Irradiation Condition | Laser Configuration | Energy Per Beam | Spot Diameter | Laser Intensity | EUV Conversion Efficiency |
|---|---|---|---|---|---|
| Single beam | 1 beam (0° incidence) | 40 mJ | 40 µm | 2.6% | |
| Dual beam simultaneous | 2 beams (±30° incidence) | 20 mJ | 30 µm | 3.6% |
Suppressing Plasma Cooling While Preserving Emission Profile
Why would simply splitting the irradiating beam improve conversion efficiency? The likely explanation lies in suppressing rapid plasma expansion and cooling.
When a powerful pulsed laser strikes a metal target, the surface material instantly ionizes into plasma. For tin plasma to strongly emit EUV light (13.5 nm), the plasma's electron temperature must be maintained within an appropriate range of roughly 30 eV, and electron density around .
When a single laser beam is focused tightly on a narrow spot, the resulting plasma expands rapidly and three-dimensionally into the vacuum, losing heat and cooling within a short time. In contrast, when two beams strike from different angles, the spatial region of plasma heated by the laser expands. The paper explains that this suppresses excessive cooling from three-dimensional plasma expansion, allowing the electron temperature favorable for EUV emission (around 30 eV) to be sustained for longer.
Measurements of the time waveform recorded by an energy meter showed that peak amplitude itself was comparable between single-beam and dual-beam conditions. However, the full width at half maximum (FWHM) of the signal extended from approximately 150 ns for a single pulse to approximately 200 ns under dual-beam irradiation, increasing the time-integrated value. Since the energy meter's response reflects the integrated signal, this lengthening of the waveform does not by itself directly indicate an extended EUV emission duration—but it does clearly demonstrate an increase in total EUV energy. The paper notes that the efficiency gain is driven not only by an increase in emission area but substantially by this temporal effect as well.
Meanwhile, the energy spectrum of tin ions observed using a Faraday cup showed matching profiles and peak energies between the single-beam and dual-beam cases, confirming that comparable plasma conditions were reproducibly generated.
In the optical design of a lithography system, the size of the emitting light source is a critical parameter—if the source is too large, the collector mirrors inside the lithography system cannot efficiently gather the light. Observations using an EUV pinhole camera found that the emission region size remained at approximately 60–70 µm for both single- and dual-beam irradiation, showing almost no change. This satisfies the source size requirement of 100 µm or less demanded by future High-NA and Hyper-NA lithography systems.
Droplet-Target Verification Remains Before Real-World Application
This research provides important insight toward power reduction in future EUV lithography systems. The research group had previously proposed the multi-laser irradiation concept in 2024 using a 1 µm Nd:YAG laser (pulse energy 500 mJ), but because the wavelength and energy conditions differed, a direct efficiency comparison is not possible. This time, the team used the leading candidate technology—a 2µm-band solid-state laser—and demonstrated the efficiency improvement under lower pulse energy conditions (40 mJ).
Because this method requires no complex active control equipment and can be built using only passive optics, it is inherently compatible with existing laser optical system architectures.
However, it should be noted that this result comes from an experiment using a flat tin plate target in a laboratory setting. Actual EUV lithography systems use extremely tiny liquid tin droplets ejected into vacuum as the target. While solid flat-plate target experiments are positioned as the most accessible and reproducible benchmark, verifying this method's effectiveness and implementability with the droplet-type tin targets used in actual mass-production systems remains a task for future work.
Additionally, 2µm solid-state lasers themselves have not yet reached the output levels required by EUV lithography systems, and a path to practical implementation is not yet established. Further optimization of the number of irradiating beams, demonstration of effectiveness with droplet-type tin targets, and continued advances in 2µm high-power laser technology itself must all come together to bring a power-efficient next-generation semiconductor lithography process closer to reality.