Efficient extreme ultraviolet emission by multiple laser pulses

Advanced semiconductor circuits, such as central processing unit (CPU), graphics processing unit (GPU), and DRAM, are produced by extreme ultraviolet (EUV) lithography.¹ According to the recent International Roadmap for Devices and Systems (IRDS^TM), the next technologies are higher numerical-aperture (NA) of 0.55 or greater, shorter wavelength EUV, so-called beyond EUV (B-EUV) at a wavelength of $6. x$ nm, and EUV free electron laser (FEL).² A higher NA optical system and higher power EUV source are needed for higher resolution in the exposure tool. For efficient optical coupling in the exposure tool, the EUV source size should be the order of 100 μm. For a higher NA system, the EUV source size should be smaller, and EUV power should be increased. Current sources are based on plasmas produced by laser irradiation of tin (Sn) droplets. The present maximum EUV source power is $∼ 350$ W at a wavelength of 13.5 nm, which can be coupled to Mo/Si mirrors in the EUV exposure tool. The drive laser for a high-power EUV source with a higher EUV CE of $5 % − 6 %$ in 2% bandwidth (BW) is a high-power CO₂ laser with an operating power higher than 20 kW at a typical repetition rate of 50 kHz.³ However, the wall-plug efficiency of the pulsed CO₂ laser amplifier is low. Recently, the theoretical maximum EUV CE for CO₂ laser irradiation has been reported to be 10.3%.⁴ It is challenging to reduce the electric power and to develop high power CO₂ amplifiers due to the limitation of the damage on the optics and electrodes in the CO₂ laser amplifiers. Therefore, improving the EUV conversion efficiency (CE) for a higher power EUV source is an important challenge.

Mid-infrared solid-state lasers operating at a wavelength of around 2 μm provide potential candidates to overcome the wall-plug efficiency problem. Tm-doped yttrium-lithium-fluoride (YLF) laser or Ho-doped yttrium-aluminum-garnet (YAG) laser is suitable due to their higher wall-plug efficiency. High pulse energy of 1 J class at a low repetition rate has been achieved.^5,6 In addition, the EUV CE has been reported to be 5%, a level comparable to that of a CO₂ LPP EUV source, experimentally^7–10 and theoretically.^11,12 However, the solid-state laser media suffer damage even at lower average power, as compared to a gas-discharge pulsed CO₂ laser, and the pulse energy is expected to be low at high repetition rates of several 10 kHz using recent laser technologies incorporating an appropriate cooling system.^13,14 The solid-state laser pulse energy is thus lower than that of the gas-discharge CO₂ laser under high repetition rate operation at 50 kHz. It would be extremely challenging to realize a solid-state laser amplifier with a laser output power of beyond 10 kW using current laser technologies. In any case, we again have the limitation imposed by high-power amplifier development for high-power EUV sources.

To overcome these problems, we propose and demonstrate the use of quasi-simultaneous spatially separated multiple solid-state-laser pulse irradiation. By irradiating with multiple $10 − 100$ mJ/pulse class laser pulses from independent synchronized laser systems, the energy loss due to plasma expansion, which is associated with plasma cooling, can be reduced,¹⁵ thereby increasing the energy fraction available for radiation. Higher EUV CE is expected to be achieved under irradiation by multiple lasers because plasma expansion losses can be reduced in plasmas from targets irradiated by solid-state laser pulses.^16,17 This paper aims to demonstrate that the EUV CE can be increased using multi-beam and multiple-laser systems. In other words, it indicates that while obtaining the required output power with a single laser system is relatively hard, using multiple systems and beams it becomes possible. This result will reduce the laser output power requirement for mid-infrared solid-state and CO₂ laser systems and the associated burden on laser amplifiers.

In this Letter, we demonstrate enhanced efficiency for an EUV source using spatially separated multiple solid-state-laser pulse irradiation. The EUV CE is expected to be higher than that for single pulse irradiation for the reasons stated earlier. The fast ion spectra also show indirect evidence for production of plasmas with essentially similar properties under multiple laser pulse irradiation when the same total laser energy is maintained.

A schematic diagram of the experimental apparatus is shown in Fig. 1. The experiment was performed using two Q-switched Nd:yttrium-aluminum-garnet (Nd:YAG) laser systems at a wavelength of 1064 nm (Spectra Physics, Quanta-Ray, Pro-230–10). Each laser pulse had the same duration of 16 ns [full width at half-maximum (FWHM)]. In the present experiments, we set the total laser output energy at 500 mJ, and we changed the number of laser beams, keeping the intensity of each fixed at $2 × 10 11$ W/cm², corresponding to the optimum laser intensity for maximizing the EUV emission from a plasma and thus the EUV CE.¹⁸ The EUV emission at 13.5 nm originates from highly charged ions ranging from Sn⁷⁺ to Sn¹⁴⁺.^19,20 The optimum electron temperature required is 30 eV,^21,22 which is consistent with the predictions of collisional radiative (CR) models.^23,24 At a laser wavelength of 1 μm, the optimum laser intensity required to achieve the electron temperature of about 30 eV and the maximum EUV CE, is expected to be $( 1 − 2 ) × 10 11$ W/cm². Therefore, each laser beam needs to be tuned to the optimum laser intensity by changing the focusing lens positions and focal spot diameters. The control system used was set so that the output pulses from the two independent laser systems were produced at the same instant by synchronizing the Pockels cell switching with a digital delay pulse generator.

The EUV source images were captured using an EUV pinhole camera with a thermoelectrically cooled back-illuminated x-ray charge-coupled device (CCD) camera (Andor Technology). The EUV pinhole camera system was positioned at $90 °$ with respect to the incident laser axis. The EUV pinhole camera consisted of a 500-nm Zr filter to cut the infrared (IR), visible, and ultraviolet (UV) emission in front of the x-ray CCD camera by which the emission in the spectral region of $6 − 19$ nm was observed. Therefore, the 13.5-nm in-band EUV emission size was expected to be smaller than the $6 − 19$ nm pinhole image. An EUV energy meter, which was positioned at $45 °$ with respect to the incident laser axis, was used to evaluate the in-band (2% bandwidth) EUV CE at 13.5 nm. Absolute EUV energy was measured using a calibrated EUV energy meter equipped with a calibrated Mo/Si multilayer mirror and a Zr filter. In addition, a Faraday cup, placed at $30 °$⁠, was used to measure the fast ion time-of-flight (TOF) signal.

Before evaluating the EUV CE, we observed the EUV source images using the EUV pinhole camera. Figure 2 shows the EUV source images recorded for different numbers of incident laser beams at a total irradiated laser energy of 500 mJ. Note that the EUV source size should be of the order of 100 μm to satisfy the etendue requirement of the exposure tool. Figure 2(a) shows a typical EUV image recorded for single beam irradiation at 500 mJ/pulse and a focal spot diameter of 140 μm. The resulting source size was 105 μm (FWHM) in the radial direction (horizontal) and 115 μm (FWHM) in the laser axis direction (vertical). Under these conditions, we expect a three-dimensional plasma expansion with significant expansion loss, mainly the radiation loss,⁸ resulting in a measured EUV CE of 1.7%, which was reproduced by the two-dimensional radiation hydrodynamic simulation Star-2D (not shown).^4,25 Note that the expansion loss can be reduced using a set of focal spot diameters larger than 300 μm to produce a quasi-one-dimensional expansion.¹⁵ On the other hand, the focal spot diameter should be set a value close to 100 μm for optical coupling in a high-NA exposure tool. The EUV source sizes and the EUV CEs are summarized in Fig. 3 and Table I.

To suppress the plasma expansion loss, we irradiated the target with two laser beams each with a pulse energy of 250 mJ/pulse giving a total laser energy of 500 mJ at incident angles of $± 30 °$ as shown in Fig. 2(b) and $± 60 °$ as shown in Fig. 2(c). The laser intensity in each case was tuned to be $2 × 10 11$ W/cm² by adjusting the lens position. The laser beams irradiated the planar Sn target quasi-simultaneously. The source sizes were 90 μm (horizontal) and 80 μm (vertical) at $± 30 °$ irradiation in Fig. 2(b) and 200 μm (horizontal) and 100 μm (vertical) at $± 60 °$ irradiation in Fig. 2(c). The measured EUV CEs were 3.8% at $± 30 °$ injection and 4.7% at $± 60 °$ injection. The EUV CE of 4.7%, which was one of the highest values ever reported, in the case of a 1-μm solid-state laser-produced planar Sn target plasma by multiple laser pulse irradiation. Note that the CE standard deviation was measured to be less than $± 0.1 %$⁠.

The EUV emission originates primarily from the plasma surface. The effective EUV emissivity is maximized at an electron density of 10¹⁹ cm⁻³ due to the balance between the emissivity and absorption. Sn laser produced plasmas possess a large optical depth for in-band EUV radiation, and thus, radiation emitted in the plasma core is reabsorbed before it can reach the plasma surface. Only radiation emitted very close to the surface can escape.²⁶ As a result, the EUV emission increases with plasma surface area and EUV source size. In detail, as the oblique laser pulse heated a larger volume low density region, the EUV emission originated from a larger plasma surface. The EUV CE was maximized at the laser incident angle of $± 60 °$⁠. To understand more fully the detailed source size, we separately irradiated the target with a single beam at a pulse energy of 250 mJ/pulse and an intensity of $2 × 10 11$ W/cm². Therefore, we achieved an effective method of multiple laser pulse irradiation. Figure 4 shows the EUV images and source sizes at laser incidence angles of $± 30 °$ and $± 60 °$⁠, respectively, for a pulse energy of 250 mJ/pulse and a laser intensity of $2 × 10 11$ W/cm². The EUV CEs were observed to be 1.8% and 1.7% for the plasmas shown in Figs. 4(a) and 4(b) recorded for an incidence angle of $± 30 °$ and 2.5% and 1.9% for the plasma shown in Figs. 4(c) and 4(d) produced at an incident angle of $± 60 °$⁠, in the present experiments. The EUV CEs for dual beam irradiation at incident angles of $± 30 °$ in Fig. 2(b) and $± 60 °$ in Fig. 2(c) are essentially equal to the sum of each EUV CE in Figs. 4(a)/4(b) and 4(c)/4(d), and Table II, respectively. Note that the CEs were defined by the ratio between the in-band (2% bandwidth) EUV emission energy at 13.5 nm in a solid angle of $2 π$ sr and the incident laser pulse energy. In Fig. 2, we used the CE defined as the EUV emission energy divided by the total laser pulse energy of 500 mJ. On the other hand, the CEs in Fig. 4 were defined by the ratio between the EUV emission and the single laser pulse energy of 250 mJ.

The EUV CEs were 3% for three beams (about 165 mJ/pulse) at incident angles of $0 °$ and $± 30 °$ in Fig. 2(d), 2.6% for three beams (about 165 mJ/pulse) at incident angles of $0 °$ and $± 60 °$ in Fig. 2(e), 2.9% for four beams (125 mJ/pulse) at incident angles of $± 30 °$ and $± 60 °$ in Fig. 2(f), and 2.4% for five beams (100 mJ/pulse) at incident angles of $0 ° , ± 30 °$⁠, and $± 60 °$ in Fig. 2(g); the intensity of each beam was kept at $2 × 10 11$ W/cm² by changing the positions and focal spot diameters of the lens. The EUV CEs for multiple laser beams were approximately equal to the sum of each individual EUV CE for each individual laser beam. This means that the EUV CE for multiple laser beams can increase roughly by twice the value of the EUV CE for each single laser beam. This result implies that it is not necessary to confine operation to a single laser to obtain high EUV output power. Multiple beam irradiation has a potential for higher output power and efficiency. In other words, a laser amplifier for a single pulse is unnecessary to obtain the required output power. Although multiple beam irradiation can achieve higher EUV CE at a lower pulse energy, the collector (C₁) mirror needs to have some small holes through which the laser pulses can pass. Under multiple pulse irradiation, the diameter of each laser beam is expected to be small. Thus, multiple laser beam irradiation has advantages, not only due to the laser system separation but also to the increase in the EUV CE due to the suppression of the plasma expansion loss. We cannot do the simulation for oblique laser incidence because we need to improve the simulation code to allow for with the ray trace of the laser pulse in the plasma. Although we cannot fully understand the apparent reason for the increase in the CE, we discuss the emission size and loss processes.

We evaluated the spatial distributions of the electron temperature and the electron density along the laser axis at the laser peak time for a laser pulse energy of 500 mJ and a laser intensity of $2 × 10 11$ W/cm². The 13.5-nm EUV emission was attributed to the ion charge states of $Sn 7 + − Sn 14 +$⁠, originating from the region at an electron temperature of $15 − 40$ eV and an electron density of $5 × 10 19 − 1 × 10 21$ cm⁻³. The width of these regions was evaluated to be 100 μm, which was in good agreement with the observed EUV source size in Fig. 2(a).

The plasma expansion along the laser axis and the radial directions were separated. Along the laser axis direction, the plasma was heated by laser absorption, and the optimum charge state ions required to emit the EUV radiation were maintained throughout the duration of laser pulse. On the other hand, the plasma expanded toward the vacuum along the radial direction, and consequently, the plasma cooled down, and the optimum charge state ions could not be maintained. In the case of one beam with a pulse energy of 500 mJ at a laser intensity of $2 × 10 11$ W/cm², the EUV CE reached 1.7% with the plasma expansion loss and the cooling along the radial direction. On the other hand, we expect a suppression of the radial plasma expansion cooling for multiple beam irradiation, since the optimum plasma conditions can be maintained longer than that for single laser pulse irradiation.

We calculated the conversion efficiency for normally incident laser pulses at the pulse energies of 500, 250, 167, 125, and 100 mJ at the same laser intensity of $2 × 10 11$ W/cm². The conversion efficiency for each pulse energy was predicted to be $1.8 % − 2.1 %$⁠. The CEs for the laser pulse energy of 250 mJ were also reproduced to be about 2% in Fig. 4. In the multiple beam irradiation, the EUV emission is incoherent, so the EUV CE was the sum of the CEs of the respective beams. This indicates that the energy of the EUV emission strongly depends on the surface area of the plasma. This is because multi-beam irradiation suppresses the effect of plasma cooling in the radial direction and maintains plasma conditions suitable for EUV emission.

We also explored the EUV emission spectra at two incident angles of the laser pulse at a pulse energy of 250 mJ at an intensity of $2 × 10 11$ W/cm² for a single beam. The spectral purity (spectral efficiency) at the incident angle of 60° was higher than that at the incident angle of 0°. The spectral difference is expected to be one of the reasons for high CEs for the oblique laser incidence. The spectral purity and conversion efficiency for oblique laser incidence are both improved.

To find evidence of similar plasma conditions under total laser energy irradiation at 500 mJ, we observed the fast ion spectra from the Faraday cup measurements. The fast ion energy depends on the electron temperature, initial electron density, and laser pulse duration.^27–29 The fast ion spectra are expected to be the almost same for similar electron temperatures for multiple laser beam irradiation. We compared the fast ion spectra for different numbers of laser pulses, as shown in Fig. 5. The fast ion spectra have essentially a similar profile and energy, resulting in similar plasma production at different numbers of laser beams. This result suggests that we can use separate beams from the same laser or multiple laser systems provided each beam has the optimum laser intensity since the EUV emission originates from highly charged Sn ions at the plasma surface. To keep the reflection coefficient of the Mo/Si multilayer collector mirror (C₁ mirror), we should reduce the energy to less than 3 keV with the decrease in the fast ion flux in the exposure tool. We reproduced the numerical maximum energy of 2.1 keV with the experimental maximum energy of 2.3 keV at the laser intensity of $2 × 10 11$ W/cm² for a single beam.

In summary, we have demonstrated an efficient EUV source using quasi-simultaneous spatial multi-solid-state-laser pulse irradiation. The maximum conversion efficiency achieved was 3.8% for $± 30 °$ oblique laser pulse injection with a source size of about 100 μm for two separated laser pulses at a total laser energy of 500 mJ and a laser intensity of $2 × 10 11$ W/cm². This EUV CE of 3.8% with an EUV source size of 100 μm was about twice as high as that for single laser pulse irradiation of 1.7%. In addition, we achieved the EUV CE of 4.7% for $± 60 °$ oblique laser pulse injection, which was one of the highest values ever reported, in the case of a 1-μm solid-state laser-produced planar Sn target plasma by multiple laser pulse irradiation. The EUV CE for spatially separated multi-laser pulse irradiation was higher than that for single pulse irradiation. The fast ion spectra were also similar under multi-laser-pulse irradiation. This means that the plasma parameters were expected to be essentially similar. This result implies that it is not necessary to use a single laser beam to maximize EUV output power. Multiple laser systems and beams can achieve higher EUV output power and higher EUV CE, a significant result that provides freedom in laser development. We propose the use of solid-state ten lasers with a 1-kW solid-state laser to achieve the 10-kW laser power injection. In addition, we can extend the present extension to shorter wavelength B-EUV sources^30–32 and to at wavelength metrology sources.

The authors are indebted to Masaki Kume, Takeru Niinuma, and Tatsuya Soramoto (Utsunomiya University) for useful technical support and discussion. T.H. acknowledges the support from the Japan Society for the Promotion of Science (JSPS) (Nos. S20063 and JP 23H01416) and the Sumitomo Foundation (No. 2200578). E.J.T. acknowledges the support from the RIKEN TRIP initiative (Leading-edge semiconductor technology).