**5. Conversion efficiency for EUVL**

In this section, we report our studies carried out to improve the CE at 13.5 nm with 2% bandwidth required for the EUVL source (Amano et al., 2008, 2010a). To achieve the highest CE, we attempted to control the plasma parameter by changing the driving laser conditions. We investigated dependences of the CE on the drum rotation speed, laser energy, and laser wavelength. We also carried out double-pulse irradiation experiments to improve the CE.

To obtain data of EUV emission, a conventional Q-switched Nd:YAG rod laser (Spectra-Physics, PRO-230) was used in single-shot operation. By changing the position of the focusing lens to change the laser spot, the laser intensity on the target was adjusted to find the optimum intensity. We note that the lens position (LP) is zero at best focus, negative for in-focus (the laser spot in the target before the focus) and positive for out-of-focus (beyond the focus).

Laser-Plasma Extreme Ultraviolet Source Incorporating a Cryogenic Xe Target 361

increases. Therefore, the vaporized Xe gas from the target surface was considered as the gas on the target. Although additional Xe gas was added from outside the vacuum chamber, the EUV intensity did not increase and in fact decreased owing to gas absorption. Therefore, it is supposed that Xe gas with adequate pressure localizes only near the target surface. From these results, we conclude that Xe gas on the target surface in the rotating drum produces optically thick plasma that has optimized density and temperature for emitting EUV radiation, and satellite lines of the plasma contribute effectively to increasing the EUV

Fig. 6. Images of visible emissions from the plasma on the resting (a) and rotating (b) targets. Next, the dependence of the laser pulse energy was investigated. We measured the CE as a function of laser energy at different LPs in the rotating drum. For laser energies exceeding 0.3 J, a CE of nearly 0.9% was achieved by tuning the LP with the laser intensity optimized as ~1010 W/cm2. In the energy range, the maximum CE did not depend on the laser energy. At the LP in this experiment, the spot size on the target was larger than 500 m and plasma energy loss at the edges could be ignored for this large spot. Therefore, the same CE was achieved at the same laser intensity. However, in the lower energy region, the spot size must be small to achieve optimal laser intensity, and edge loss due to three-dimensional expansion in plasma cannot then be ignored and a decrease in the CE was observed.

Therefore, it is concluded that laser energy must exceed 0.3 J to achieve a high CE.

the plasma was too great in these experiments and the best condition was not achieved.

In conclusion, the maximum CE was found to be 0.9% at 13.5 nm with 2% bandwidth for the

The dependence of the laser wavelength was also investigated. Additionally, we carried out 1 double-pulse irradiation experiments in which a pre-pulse produces plasma with optimal density and temperature, and after a time delay, a main laser pulse effectively injects emission energy into the expanded plasma to increase the CE. Under the rest condition, there were increases in CE for the shorter laser or the double pulse irradiation (Miyamoto et al., 2005, 2006). In both cases, the long-scale plasmas and their emission spectra were observed to be similar to those under the rotation condition for 1 single-pulse irradiation. Therefore, we supposed that in the both cases, the CE was increased by the same mechanism described above. However, when the shorter pulses or the double pulses were emitted under the rotating condition, the CE did not increase but decreased. It is considered that the opacity of

intensity (Sasaki et al., 2004).

optimal condition.

Figure 5(a) shows the CE per solid angle as a function of LP (laser intensity), which was measured by an EUV energy detector calibrated absolutely—*Flying Circus* (SCIENTEC Engineering)—located 45 degrees from the laser incident axis. The laser pulse energy was 0.8 J. We see that the CE was higher under the rotating-drum condition than under the rest condition. Here, the rest condition is as follows. Xe gas flow is stopped (0 mL/min) after the target layer has formed, and the drum rests (0 rpm) during a laser shot and stepwise rotates after every shot so that a fresh target is supplied to the point irradiated by the laser. The rotation condition is as follows. Laser pulses irradiate quasi-continuously the target on the rotating drum (>3 rpm), supplying Xe gas (>40 mL/min) and forming the target layer. The EUV intensity increased immediately with slow rotation (>3 rpm) and appeared to be almost independent of the rotation speed. In Fig. 5(a), we see that the maximum CE per solid angle was for an optimized laser intensity of 1 × 1010 W/cm2 (LP = –10 mm) during rotation. The EUV angular distribution could be expressed by a fitting curve of (cos)0.38, and taking into account this distribution, we obtained the maximum spatially integrated CE of 0.9% at 13.5 nm with 2% bandwidth. EUV spectra at laser intensity of 1 × 1010 W/cm2 are shown in Fig. 5(b). It is obvious that the emission of the 13.5 nm band was greater in the case of rotation than it was in the case of rest.

Fig. 5. (a) CE at the wavelength of 13.5 nm with 2% bandwidth as a function of LP under the rotation (130 rpm) and at-rest (0 rpm) conditions. The laser energy was 0.8 J. Insets show the laser beam focusing on the target. (b) Spectra of EUV radiation from the cryogenic Xe drum targets under the rotation (bold line) and at-rest (narrow line) conditions with laser intensity of 1 × 1010 W/cm2 for LP of –10 mm.

We considered the mechanism for the increase in EUV intensity with rotation of the target. Figure 6 shows photographs of the visible emission from the Xe target observed from a transverse direction. It shows an obvious expansion of the emitting area with longer (optically thicker) plasma in the rotating case compared with the at-rest case. These images indicate the existence of any gas on the target surface. Under the rotation condition, Xe gas is supplied continuously to grow the target layer and the wipers form the layer. However, the wipers are not chilled especially, and the temperature of the target surface might increase owing to contact with the wipers in the rotating case so that the vapor pressure

Figure 5(a) shows the CE per solid angle as a function of LP (laser intensity), which was measured by an EUV energy detector calibrated absolutely—*Flying Circus* (SCIENTEC Engineering)—located 45 degrees from the laser incident axis. The laser pulse energy was 0.8 J. We see that the CE was higher under the rotating-drum condition than under the rest condition. Here, the rest condition is as follows. Xe gas flow is stopped (0 mL/min) after the target layer has formed, and the drum rests (0 rpm) during a laser shot and stepwise rotates after every shot so that a fresh target is supplied to the point irradiated by the laser. The rotation condition is as follows. Laser pulses irradiate quasi-continuously the target on the rotating drum (>3 rpm), supplying Xe gas (>40 mL/min) and forming the target layer. The EUV intensity increased immediately with slow rotation (>3 rpm) and appeared to be almost independent of the rotation speed. In Fig. 5(a), we see that the maximum CE per solid angle was for an optimized laser intensity of 1 × 1010 W/cm2 (LP = –10 mm) during rotation. The EUV angular distribution could be expressed by a fitting curve of (cos)0.38, and taking into account this distribution, we obtained the maximum spatially integrated CE of 0.9% at 13.5 nm with 2% bandwidth. EUV spectra at laser intensity of 1 × 1010 W/cm2 are shown in Fig. 5(b). It is obvious that the emission of the 13.5 nm band was greater in the case

Fig. 5. (a) CE at the wavelength of 13.5 nm with 2% bandwidth as a function of LP under the rotation (130 rpm) and at-rest (0 rpm) conditions. The laser energy was 0.8 J. Insets show the laser beam focusing on the target. (b) Spectra of EUV radiation from the cryogenic Xe drum targets under the rotation (bold line) and at-rest (narrow line) conditions with laser intensity

We considered the mechanism for the increase in EUV intensity with rotation of the target. Figure 6 shows photographs of the visible emission from the Xe target observed from a transverse direction. It shows an obvious expansion of the emitting area with longer (optically thicker) plasma in the rotating case compared with the at-rest case. These images indicate the existence of any gas on the target surface. Under the rotation condition, Xe gas is supplied continuously to grow the target layer and the wipers form the layer. However, the wipers are not chilled especially, and the temperature of the target surface might increase owing to contact with the wipers in the rotating case so that the vapor pressure

of rotation than it was in the case of rest.

of 1 × 1010 W/cm2 for LP of –10 mm.

increases. Therefore, the vaporized Xe gas from the target surface was considered as the gas on the target. Although additional Xe gas was added from outside the vacuum chamber, the EUV intensity did not increase and in fact decreased owing to gas absorption. Therefore, it is supposed that Xe gas with adequate pressure localizes only near the target surface. From these results, we conclude that Xe gas on the target surface in the rotating drum produces optically thick plasma that has optimized density and temperature for emitting EUV radiation, and satellite lines of the plasma contribute effectively to increasing the EUV intensity (Sasaki et al., 2004).

Fig. 6. Images of visible emissions from the plasma on the resting (a) and rotating (b) targets.

Next, the dependence of the laser pulse energy was investigated. We measured the CE as a function of laser energy at different LPs in the rotating drum. For laser energies exceeding 0.3 J, a CE of nearly 0.9% was achieved by tuning the LP with the laser intensity optimized as ~1010 W/cm2. In the energy range, the maximum CE did not depend on the laser energy. At the LP in this experiment, the spot size on the target was larger than 500 m and plasma energy loss at the edges could be ignored for this large spot. Therefore, the same CE was achieved at the same laser intensity. However, in the lower energy region, the spot size must be small to achieve optimal laser intensity, and edge loss due to three-dimensional expansion in plasma cannot then be ignored and a decrease in the CE was observed. Therefore, it is concluded that laser energy must exceed 0.3 J to achieve a high CE.

The dependence of the laser wavelength was also investigated. Additionally, we carried out 1 double-pulse irradiation experiments in which a pre-pulse produces plasma with optimal density and temperature, and after a time delay, a main laser pulse effectively injects emission energy into the expanded plasma to increase the CE. Under the rest condition, there were increases in CE for the shorter laser or the double pulse irradiation (Miyamoto et al., 2005, 2006). In both cases, the long-scale plasmas and their emission spectra were observed to be similar to those under the rotation condition for 1 single-pulse irradiation. Therefore, we supposed that in the both cases, the CE was increased by the same mechanism described above. However, when the shorter pulses or the double pulses were emitted under the rotating condition, the CE did not increase but decreased. It is considered that the opacity of the plasma was too great in these experiments and the best condition was not achieved. In conclusion, the maximum CE was found to be 0.9% at 13.5 nm with 2% bandwidth for the

optimal condition.

Laser-Plasma Extreme Ultraviolet Source Incorporating a Cryogenic Xe Target 363

indicated (Kubiak et al., 1995). Since these reports, liquid Xe targets have been preferred over solid Xe targets, with the exception of our group. It is therefore necessary to clarify characteristics of fragment debris from a solid Xe target on a rotating cryogenic drum. After exposing a Si sample to the Xe plasmas pumped by 100 laser pulses, we observed fragment impact damage on its surface using a scanning electron microscope. We observed damage spots on the samples at laser energy of 0.8 J irrespective of whether the drum rotates. Conversely, we did not observe spots at laser energy of 0.3 J. To explain these results, we consider that the fragment speed (kinetic energy) might drop below a damage threshold upon reducing the laser pulse energy because the fragment speed is a function of incident laser energy (Mochizuki et al., 2001). Observing the damage spots, we know that the fragment size was larger than a few microns, and the gas curtain might not be effective for such large fragments. This would explain why the fragment impact damage was independent of the state of drum rotation. From these results, we conclude that fragment impact damage, which occurs especially for the solid Xe target, can be avoided simply by

The laser pulse energy was set to 0.3 J to avoid fragment impact damage and the laser repetition rate was 320 pps, giving an average power of 100 W. Next, we investigated damage to a Mo/Si mirror, which was the result of total plasma debris (mainly fast ions) from the laser multi-shots experiments. After 10 min plasma exposure, the sputtered depth was measured to be 50 nm on the surface of a Mo/Si mirror placed 100 mm from the plasma at a 22.5-degree angle to the incident laser beam. Because a typical Mo/Si mirror has 40 layer pairs and the thickness of one pair is approximately 6.6 nm, all layers will be removed within an hour by the sputtering. Although Xe is a deposition-free target, sputtering by debris needs to be mitigated. However, the major plasma debris component is ions, and we believe their mitigation to be simple compared with the case of a metal target such as Sn, using magnetic/electric fields and/or gas. We are now studying debris mitigation by Ar buffer gas. Ar gas was chosen because of its higher stopping power for Xe ions and lower absorption of EUV light, and its easy handling and low cost. After the vacuum chamber was filled with Ar gas, total erosion rates were measured using a gold-coated quartz crystal microbalance sensor placed 77 mm from the plasma at a 45-degree angle, and simultaneously, EUV losses were monitored by an EUV detector placed 200 mm from the plasma at a 22.5-degree angle. Figure 8 shows the erosion rates as a function of Ar gas pressure. The rates were normalized by the erosion *N0* at a pressure of 0 Pa. When the Ar pressure was 8 Pa, we found the erosion rate was 1/18 of that without the gas, but the absorption loss for EUV light was only 8%. The erosion rates (N/N0) in Fig. 8 can be fitted to

exp <sup>0</sup>

where *PAr* is the Ar pressure, *k* is the Boltzmann constant, *T* is the gas temperature,

cross section and *l* is the debris flight length. From this fitting, we obtain

*Ar NP N <sup>l</sup> Ar kT*

The Ar buffer gas successfully mitigated the effect of plasma debris with little EUV attenuation. Increasing the Ar pressure, mirror erosion decreases but EUV attenuation increases. Compromising the erosion and EUV attenuation, an optimized pressure is achieved. We should localize the higher density Ar gas to only the debris path so that EUV attenuation is as small as possible. We can design the optimized pressure condition using

*P*

(3)

is the

= 2.0 × 10–20 m2.

reducing the incident laser pulse energy to less than 0.3 J.

an exponential curve:
