**1. Introduction**

Optical lithography is a core technique used in the industrial mass production of semiconductor memory chips. To increase the memory size per chip, shorter wavelength light is required for the light source. ArF excimer laser light (193 nm) is used at present and extreme ultraviolet (EUV) light (13.5 nm) is proposed in next-generation optical lithography. There is currently worldwide research and development for lithography using EUV light (Bakshi, 2005). EUV lithography (EUVL) was first demonstrated by Kinoshita et al. in 1984 at NTT, Japan (Kinoshita et al., 1989). He joined our laboratory in 1995 and has since been actively developing EUVL technology using our synchrotron facility NewSUBARU. Today, EUVL is one of the major themes studied at our laboratory.

To use EUVL in industry, however, a small and strong light source instead of a synchrotron is required. Our group began developing laser-produced plasma (LPP) sources for EUVL in the mid-1990s (Amano et al., 1997). LPP radiation from high-density, high-temperature plasma, which is achieved by illuminating a target with high-peak-power laser irradiation, constitutes an attractive, high-brightness point source for producing radiation from EUV light to x-rays.

Light at a wavelength of 13.5 nm with 2% bandwidth is required for the EUV light source, which is limited by the reflectivity of Mo/Si mirrors in a projection lithography system. Xe and Sn are known well as plasma targets with strong emission around 13.5 nm. Xe was mainly studied initially because of the *debris problem*, in which debris emitted from plasma with EUV light damages mirrors near the plasma, quickly degrading their reflectivity. This problem was of particular concern in the case of a metal target such as Sn because the metal would deposit and remain on the mirrors. On the other hand, Xe is an inert gas and does not deposit on mirrors, and thus has been studied as a deposition-free target. Because of this advantage, researchers initially studied Xe. To provide a continuous supply of Xe at the laser focal point, several possible approaches have been investigated: employing a Xe gas puff target (Fiedrowicz et al., 1999), Xe cluster jet (Kubiak et al., 1996), Xe liquid jet (Anderson et al., 2004; Hansson et al., 2004), Xe capillary jet (Inoue et al., 2007), stream of liquid Xe droplets (Soumagne et al., 2005), and solid Xe pellets (Kubiak et al., 1995). Here, there are solid and liquid states, and their cryogenic Xe targets were expected to provide higher laser-to-EUV power conversion efficiency (CE) owing to their higher density compared with the gas state. In addition, a smaller gas load to be evacuated by the exhaust pump system was expected.

Laser-Plasma Extreme Ultraviolet Source Incorporating a Cryogenic Xe Target 355

Fig. 1. Illustration of (a) the top view of the rotating cryogenic drum, (b) the side view, and

First, we formed a solid Xe layer with thickness of 300–500 m on the drum surface and measured the size of the laser crater, which depends on the laser pulse energy. The crater diameter was measured directly from a microscope image, and its depth was roughly estimated from the number of shots needed to burn through the known thickness of the layer. A Q-switched 1064 nm Nd:YAG laser was focused on the Xe target surface with a spot

for a laser energy range of 0.04–0.7 J. From the results in Fig. 2, a thickness of more than 200 m was found to be sufficient for a laser shot of 1 J not to damage the drum surface. We

Two wipers are mounted on the container wall as shown in Fig.1 (a) to adjust the thickness of the solid Xe layer to 500 m. As shown in Fig. 1 (c), the V-figure wipers also collect the Xe target powder on the craters produced by laser irradiation, thereby increasing the recovery

*<sup>c</sup>* are plotted in Fig. 2

diameter of 90 m. Measured crater diameters *Dc* and crater depths

then decided the target thickness to be 500 m.

(c) the wiper.

We have also studied a cryogenic Xe solid target. In that study, we measured the EUV emission spectrum in detail, and we found and first reported that the emission peak of Xe was at 10.8 nm, not 13 nm (Shimoura et al., 1998). This meant we could only use the tail of the Xe plasma emission spectrum, not its peak, as the radiation at 13.5 nm wavelength with 2% bandwidth. From this, improvements in the CE at 13.5 nm with 2% bandwidth became a most critical issue for the Xe plasma source; such improvements were necessary to reduce the pumped laser power and cost of the whole EUV light source. On the other hand, the emission peak of a Sn target is at 13.5 nm; therefore, Sn intrinsically has a high CE at 13.5 nm with 2% bandwidth. The CE for Sn is thus higher than that for Xe at present, in spite of our efforts to improve the CE for Xe. This resulted in a trend of using Sn rather than Xe in spite of the debris problem. Today, Cymer (Brandt et al., 2010) and Gigaphoton (Mizoguti et al., 2010), the world's leading manufacturers of LPP-EUV sources, are developing sources using Sn targets pumped with CO2 lasers while making efforts to mitigate the effects of debris. In the historical background mentioned above, we developed an LPP-EUV source composed of 1) a fast-rotating cryogenic drum system that can continuously supply a solid Xe target and 2) a high-repetition-rate pulse Nd:YAG slab laser. We have developed the source in terms of its engineering and investigated potential improvements in the CE at 13.5 nm with 2% bandwidth. The CE depends on spatial and temporal Xe plasma conditions (e.g., density, temperature, and size). To achieve a high CE, we controlled the condition parameters and attempted to optimize them by changing the pumping laser conditions. We initially focused on parameters at the wavelength of 13.5 nm with 2% bandwidth required for an EUV lithography source, but the original emission from the Xe plasma has a broad spectrum at 5– 17 nm. We noted that this broad source would be highly efficient and very useful for many other applications, if not limiting for EUVL. Therefore, we estimated our source in the wavelength of 5–17 nm. Though Xe is a deposition-free target, there may be sputtering due to the plasma debris. We therefore investigated the plasma debris emitted from our LPP source, which consists of fast ions, fast neutrals, and ice fragments. To mitigate the sputtering, we are investigating the use of Ar buffer gas. In this chapter, we report on the status of our LPP-EUV source and discuss its possibilities.

### **2. Target system – Rotating cryogenic drum**

We considered using a cryogenic solid state Xe target and developed a rotating drum system to supply it continuously, as shown in Fig. 1 (Fukugaki et al., 2006). A cylindrical drum is filled with liquid nitrogen, and the copper surface is thereby cooled to the temperature of liquid nitrogen. Xe gas blown onto the surface condenses to form a solid Xe layer. The drum coated with a solid Xe layer rotates around the vertical z-axis and moves up and down along the z-axis during rotation, moving spirally so that a fresh target surface is supplied continuously for every laser shot. A container wall surrounds the drum surface, except for an area around the laser focus point. This maintains a relatively high-density Xe gas in the gap between the container wall and the drum surface so as to achieve a high growth rate of the layer and fast recovery of the laser craters during rotation. The container wall also suppresses Xe gas leakage to the vacuum chamber to less than 5%, and the vacuum pressure inside the chamber is kept at less than 0.5 Pa. The diameter of the drum is 10 cm. Its mechanical rotation and up–down speed are tunable at 0–1200 rpm and 0–10 mm/s in a range of 3 cm respectively.

We have also studied a cryogenic Xe solid target. In that study, we measured the EUV emission spectrum in detail, and we found and first reported that the emission peak of Xe was at 10.8 nm, not 13 nm (Shimoura et al., 1998). This meant we could only use the tail of the Xe plasma emission spectrum, not its peak, as the radiation at 13.5 nm wavelength with 2% bandwidth. From this, improvements in the CE at 13.5 nm with 2% bandwidth became a most critical issue for the Xe plasma source; such improvements were necessary to reduce the pumped laser power and cost of the whole EUV light source. On the other hand, the emission peak of a Sn target is at 13.5 nm; therefore, Sn intrinsically has a high CE at 13.5 nm with 2% bandwidth. The CE for Sn is thus higher than that for Xe at present, in spite of our efforts to improve the CE for Xe. This resulted in a trend of using Sn rather than Xe in spite of the debris problem. Today, Cymer (Brandt et al., 2010) and Gigaphoton (Mizoguti et al., 2010), the world's leading manufacturers of LPP-EUV sources, are developing sources using Sn targets pumped with CO2 lasers while making efforts to mitigate the effects of debris. In the historical background mentioned above, we developed an LPP-EUV source composed of 1) a fast-rotating cryogenic drum system that can continuously supply a solid Xe target and 2) a high-repetition-rate pulse Nd:YAG slab laser. We have developed the source in terms of its engineering and investigated potential improvements in the CE at 13.5 nm with 2% bandwidth. The CE depends on spatial and temporal Xe plasma conditions (e.g., density, temperature, and size). To achieve a high CE, we controlled the condition parameters and attempted to optimize them by changing the pumping laser conditions. We initially focused on parameters at the wavelength of 13.5 nm with 2% bandwidth required for an EUV lithography source, but the original emission from the Xe plasma has a broad spectrum at 5– 17 nm. We noted that this broad source would be highly efficient and very useful for many other applications, if not limiting for EUVL. Therefore, we estimated our source in the wavelength of 5–17 nm. Though Xe is a deposition-free target, there may be sputtering due to the plasma debris. We therefore investigated the plasma debris emitted from our LPP source, which consists of fast ions, fast neutrals, and ice fragments. To mitigate the sputtering, we are investigating the use of Ar buffer gas. In this chapter, we report on the

We considered using a cryogenic solid state Xe target and developed a rotating drum system to supply it continuously, as shown in Fig. 1 (Fukugaki et al., 2006). A cylindrical drum is filled with liquid nitrogen, and the copper surface is thereby cooled to the temperature of liquid nitrogen. Xe gas blown onto the surface condenses to form a solid Xe layer. The drum coated with a solid Xe layer rotates around the vertical z-axis and moves up and down along the z-axis during rotation, moving spirally so that a fresh target surface is supplied continuously for every laser shot. A container wall surrounds the drum surface, except for an area around the laser focus point. This maintains a relatively high-density Xe gas in the gap between the container wall and the drum surface so as to achieve a high growth rate of the layer and fast recovery of the laser craters during rotation. The container wall also suppresses Xe gas leakage to the vacuum chamber to less than 5%, and the vacuum pressure inside the chamber is kept at less than 0.5 Pa. The diameter of the drum is 10 cm. Its mechanical rotation and up–down speed are tunable at 0–1200 rpm and 0–10

status of our LPP-EUV source and discuss its possibilities.

**2. Target system – Rotating cryogenic drum** 

mm/s in a range of 3 cm respectively.

Fig. 1. Illustration of (a) the top view of the rotating cryogenic drum, (b) the side view, and (c) the wiper.

First, we formed a solid Xe layer with thickness of 300–500 m on the drum surface and measured the size of the laser crater, which depends on the laser pulse energy. The crater diameter was measured directly from a microscope image, and its depth was roughly estimated from the number of shots needed to burn through the known thickness of the layer. A Q-switched 1064 nm Nd:YAG laser was focused on the Xe target surface with a spot diameter of 90 m. Measured crater diameters *Dc* and crater depths *<sup>c</sup>* are plotted in Fig. 2 for a laser energy range of 0.04–0.7 J. From the results in Fig. 2, a thickness of more than 200 m was found to be sufficient for a laser shot of 1 J not to damage the drum surface. We then decided the target thickness to be 500 m.

Two wipers are mounted on the container wall as shown in Fig.1 (a) to adjust the thickness of the solid Xe layer to 500 m. As shown in Fig. 1 (c), the V-figure wipers also collect the Xe target powder on the craters produced by laser irradiation, thereby increasing the recovery

Laser-Plasma Extreme Ultraviolet Source Incorporating a Cryogenic Xe Target 357

Although flaking of the target layer due to superimposition of shock and/or thermal waves produced by continuous laser pulses was a concern for high-repetition pulse operation, model experiments and calculations show that there is no problem up to 1 J per pulse and 10

From the above results, we conclude that the rotating drum system we developed can supply the target continuously, achieving the required laser irradiation of 10 kHz and 1 J,

High peak power and high focusability (i.e., high beam quality) are required for a driving laser to produce plasma. In addition, high average power is required for high throughput in industrial use such as EUVL. We express such a laser as a *high average and high peak brightness laser*, for which the average brightness and peak brightness are defined as average power/(M2)2 and peak power/(M2)2, respectively; we began studying such lasers in the

We attempted to realize a *high average and high peak brightness laser* using a solid-state Nd:YAG laser (Amano et al., 2001). The thermal-lens effect and thermally induced birefringence in an active medium are serious for such a laser; thus, thermal management of the amplifier head is more critical, and the design of the amplifier system must more efficiently extract energy and more accurately correct the remaining thermally induced wavefront aberrations in the pumping head. To meet these requirements, we developed a phase-conjugated master-oscillator-power-amplifier (PC-MOPA) Nd:YAG laser system consisting of a diode-pumped master oscillator and flash-lamp-pumped angularmultiplexing slab power-amplifier geometry incorporating a stimulated-Brillouin-scattering phase-conjugate mirror (SBS-PCM) and image relays (IR). The system design and a photograph are shown in Fig. 3. This laser demonstrated simultaneous maximum average power of 235 W and maximum peak power of 30 MW with M2 = 1.5. The maximum pulse energy was 0.73 J with pulse duration of 24 ns at a pulse repetition rate of 320 pps. We therefore obtained, simultaneously, both high average brightness of 7 × 109 W/cm2sr and

This peak brightness is enough to produce plasma but the average brightness needs to be higher for EUVL applications. The maximum average power is mainly limited by the thermal load caused by flash-lamp-pumping in amplifiers. The system design rules that we confirmed predicted that average output power at the kilowatt level can be achieved by replacing lamp pumping in the amplifier with laser-diode pumping. Since our work, it seems that there has been no major progress in laser engineering for such *high average and high peak brightness lasers.* Average power of more than 10 kW has been achieved in continuous-wave solid-state lasers using configurations of fibers (ex. IPG Photonics Corp.) or thin discs (ex. TRUMPF GmbH). On the other hand, for the short-pulse lasers mentioned above, the maximum average power remains around 1 kW (Soumagne et al., 2005), which is more than an order of magnitude less than the ~30 kW required for an industrial EUVL source. This is one of the reasons why CO2 lasers have been preferred over Nd:YAG lasers as the driving laser. To further the industrial use of solid-state lasers, there needs to be a

and thus realizing a high-average-power EUV light source.

**3. Drive laser – Nd:YAG slab laser** 

high peak brightness of 1 × 1015 W/cm2sr.

breakthrough to increase the average power.

1990s (Amano et al, 1997,1999).

kHz (Inoue et al., 2006).

speed. The wipers demonstrated a recovery speed of 150 m/s up to a rotation speed of 1000 rpm, at a Xe flow rate of 400 mL/min.

Fig. 2. Measured diameter and depth of a crater as a function of the irradiating laser energy.

Next, operational parameters of the drum are discussed to achieve high-repetition-rate laser pulse irradiation. In Fig. 1(b), *R* is the rotation speed, *r* is the radius of the drum, and *L* is the range of motion (scanning width of the target) along the rotational axis (z-axis). When the laser pulses are irradiated with frequency *f*, craters form on the target with separation length *d* between adjacent craters. The recovery time of a crater is *T*. Under the condition that craters do not overlap, *f* and *T* can be written as

$$f = \frac{2\pi r \cdot R}{d} \tag{1}$$

$$T = \frac{2\pi rL}{f \cdot d^2} \tag{2}$$

For example, if we assume laser energy of *EL* = 1 J, a formed crater has a diameter of *Dc* = 300 m and a depth of *<sup>c</sup>* = 160 m, and *d* must be at least 300 m for the craters not to overlap. At *r* = 5 cm and *R* = 1000 rpm, we obtain *f* = 17 kHz from Eq. (1). When *f* = 10 kHz and *L* = 3 cm, *T* is calculated to be 10 s using Eq. (2), and we know that a recovery speed of the crater (*Vc* = *<sup>c</sup>*/*T*) of 16 m/s is required. Here, we have already obtained *Vc* = 150 m/s via the wiper effect and the required speed has been achieved.

Although flaking of the target layer due to superimposition of shock and/or thermal waves produced by continuous laser pulses was a concern for high-repetition pulse operation, model experiments and calculations show that there is no problem up to 1 J per pulse and 10 kHz (Inoue et al., 2006).

From the above results, we conclude that the rotating drum system we developed can supply the target continuously, achieving the required laser irradiation of 10 kHz and 1 J, and thus realizing a high-average-power EUV light source.
