1. Introduction

Extreme ultraviolet (EUV) lithography by use of a 13.5-nm sources [1] is focused on high-volume production of semiconductor devices. In fact, the fabrication tools for integrated circuits with 0.33NA (numerical aperture), to attain a half-pitch (HP) of 22 nm, have already been developed and are capable of reaching a HP of 16 nm with off-axis illumination techniques [2]. In order to increase the resolution of the HP to less than 16 nm, the NA in the exposure tools needs to be improved by advances in design and fabrication of reflective optics, such as the multilayer mirrors (MLMs) and masks. However, it is difficult to improve these given the current status of EUV technology. In order to achieve very high resolution HP, a reduction of the radiation wavelength, which couples to highly reflective MLMs with a theoretical reflective coefficient of ∼ 70%, is a more attractive proposition and explains the present trend in lithography source development [3, 4]. Wavelengths around 6.x nm are especially useful for the final stage beyond the 13.5-nm EUV source [5]. In fact, the resist sensitivity has been evaluated for the 6.x-nm radiation for future lithography application [6].

One efficient EUV source at 6.x nm is provided by laser-produced plasmas (LPPs) of the rare-earth elements gadolinium (Gd) and terbium (Tb). These elements produce strong narrowband emission, which is attributed to a n = 4 − n = 4 (Δn = 0) unresolved transition array (UTA), at 6.x nm. The spectral behavior of Gd and Tb plasmas is expected to be similar to that of tin (Sn) plasmas, because their emission is dominated by 4d open-shell ions [7, 8]. EUV emission produced from hot dense Gd plasmas may in future be used with La/B₄C multilayer mirrors to produce a viable source [3, 4]. The precise value of x is yet to be determined but will be decided by the source and reflectivity combination that provides the brightest in-band EUV yield and conversion efficiency. Previous works have proposed a wavelength of 6.76 nm as the optimum choice based on the fact that the strongest lines, originating from Ag- and Pd-like ions, are observed in this spectral region [9]. To achieve an efficient beyond EUV source, it is very important to produce an optically thin plasma either by increasing the laser wavelength [10] or decreasing the Gd concentration [11].

As the ion stages involved in the 6.x-nm radiation from Gd plasmas are higher than those of the 13.5 nm emitting ions in Sn plasmas, higher electron temperatures are needed and thus higher laser intensities. The ion stage distribution and density play a crucial role in the transport of radiation through the plasma due to opacity effects. Their distribution strongly depends on experimental conditions, such as laser wavelength, laser pulse duration, focal spot size, and target geometry. It is essential to produce an optically thin hot Gd plasma to achieve high energy conversion efficiency (CE), defined as the in-band EUV energy at 6.x nm to the incident laser energy ratio within a bandwidth (BW) of 0.6% in a solid angle of 2π sr. In the case of a laser wavelength of 1 μm, the maximum CE has been observed to be 0.54% using dual laser pulse irradiation, together with a low-density Gd target [11]. The production of low-density plasma by use of mid-IR (for example, CO₂ (carbon dioxide)) LPPs has been proposed, because the critical electron density n_ec depends on the laser wavelength, λ_L, i.e. $n_{\text{ec}}\propto {\lambda }_L^{-2}$ [12]. The critical density at a laser wavelength of λ_L = 10.6 μm for a CO₂ laser is two orders of magnitude smaller than at λ_L = 1.06 μm for the solid-state laser. As a result, a reduction of self-absorption and satellite emission in the wavelength region longer than 6.x nm is expected in CO₂ LPPs due to the lower plasma density. By analogy with efficient mid-IR CO₂ laser-produced Sn plasma EUV sources at 13.5 nm, the CE and spectral efficiency, which is important when considering out-of-band spectral suppression, should be increased in an optically thin plasma. In addition, previous work on rare-earth plasma EUV sources, has focused on plasma EUV source production by solid-state lasers operating typically in the near infrared [10, 11, 13]. In order to ascertain the applicability of a mid-IR CO₂ LPP EUV source at 6.x nm, its behavior needs to be clarified in a manner similar to the work performed on CO₂ LPP EUV sources at 13.5 nm [12, 14].

In this paper, we characterize a mid-IR CO₂ laser-produced Gd plasma EUV source with a highest peak output at 6.76 nm, and compare its performance to near-IR solid-state Nd:YAG (Nd:yttrium-aluminum-garnet) LPPs. The CE and spectral efficiency in the CO₂ LPP were higher than those of the Nd:YAG LPP, resulting in a CE of about 0.7% at a constant peak wavelength at 6.76 nm. The maximum CE at a CO₂ laser intensity of 1.3 × 10¹¹ W/cm² was evaluated to correspond to an electron temperature by the hydrodynamic simulation code.

2. Experimental setup

Transversely Excited Atmospheric (TEA) CO₂ (Lambda Physik AG) and Q-switched Nd:YAG lasers (Continuum Inc.) were used to initiate the plasmas. The TEA CO₂ laser operating at a wavelength of 10.6 μm produced an output pulse energy of 1.15 J per pulse with a pulse duration of 70 ns in a full width at half-maximum (FWHM) followed by a low-intensity tail lasting shorter than 1 μs in an unstable cavity configuration when the ratio of N₂, He, and CO₂ gases was optimized. The laser beam was focused onto a planar Gd target with a thickness of 1 mm placed in a vacuum chamber by use of an anti-reflection (AR) coated plano-convex ZnSe lens with a focal length of 6 cm. The maximum focused intensity was 6.6 × 10¹⁰ W/cm² with an expected focal spot size of about 100 μm (FWHM). The output energy of the Nd:YAG laser at the wavelength of 1.06 μm was about 1.3 J per pulse with a pulse duration of 10 ns (FWHM), and the laser beam was focused by use of an AR plano-convex BK7 lens with a focal length of 10 cm. The range of the Nd:YAG laser intensity was varied from I_L ≈ 10¹¹ to 10¹² W/cm² by changing the laser pulse energy, keeping the focal spot diameter constant. Both lasers were operated in single shot mode.

The absolute EUV energy was measured by use of a calibrated EUV calorimeter equipped with a calibrated Mo/B₄C MLM and a Zr filter. A flat-field grazing incidence spectrometer with an unequally ruled 2400 grooves/mm grating was used to record spectra. The measurement instruments were positioned at 30° with respect to the incident laser axis. Time-integrated spectra were recorded by a thermoelectrically cooled back-illuminated x-ray charge coupled device (CCD) camera (Andor Technology). The typical spectral resolution was better than 0.005 nm.

3. Results and discussion

Figure 1 shows time-integrated EUV emission spectra from the Nd:YAG LPPs at different laser intensities ranging from 9.7 × 10¹¹ to 6.6 × 10¹² W/cm². The peak wavelength shifts from 6.7 to 6.8 nm, and is mainly due to n = 4 − n = 4 (Δn = 0) transitions in ions with an open 4f or 4d outermost subshell. The sharp peak at 6.65 nm and the dip structure below 6.59 nm first appear at a laser intensity of 2.4×10¹¹ W/cm². Emission at wavelengths less than 6 nm, increases with increasing laser intensity and according to numerical evaluation, lines in the λ = 2.5 – 6 nm (hν = 207 – 496 eV) spectral region originate from Gd ionic charge states between Gd¹⁹⁺ and Gd²⁷⁺, and arise from n = 4 − n = 5 (Δn = 1) transitions.

 

Fig. 1 Time-integrated EUV emission spectra from the Nd:YAG LPPs at different laser intensities of 9.7 × 10¹¹ (a), 2.2 × 10¹² (b), and 6.6 × 10¹² W/cm² (c), respectively. The peak wavelength shifts from 6.7 to 6.8 nm with increasing the laser intensity.

Download Full Size | PDF

In the case of CO₂ LPPs, on the other hand, the main spectral features near 6.7 nm are narrower than for Nd:YAG laser irradiation, as shown in Fig. 2. The CO₂ laser intensity was varied from 5.5 × 10¹⁰ to 1.2 × 10¹¹ W/cm². The spectral structure is dramatically different to that from the Nd:YAG LPPs. The peak wavelength of 6.76 nm remains constant with increasing laser intensity. Moreover the intensity of the peak at 6.76 nm increases more rapidly with laser intensity than emission in the ranges λ = 3 – 6.6 nm and λ = 6.8 – 12 nm. Under the optically thin plasma conditions imposed by the CO₂ LPP, this peak, which is mainly due to the 4d^{10 1}S₀ − 4d⁹4f¹P₁ transition of Pd-like Gd¹⁸⁺ overlapped with ²F − ²D lines of Ag-like Gd¹⁷⁺, known to lie near 6.76 nm [9, 15] shows that these ions are indeed present in the plasma. Somewhat similar structure has been also observed in a discharge-produced plasma [9], which like the CO₂ LPP has low density and is optically thin.

 

Fig. 2 Time-integrated EUV emission spectra from the CO₂ LPPs at different laser intensities of 5.5×10¹⁰ (a), 8×10¹⁰ (b), 9.8×10¹⁰ (c), and 1.3×10¹¹ W/cm² (d), respectively. The peak wavelength of 6.76 nm remains constant with increasing the laser intensity.

Download Full Size | PDF

From Figs. 1 and 2 it is seen that the peak wavelength shifts in the case of the Nd:YAG laser irradiation. It is difficult to maintain efficient coupling of a tunable peak to the reflective coefficient, R(λ), of the MLM. In the case of CO₂ LPPs the peak wavelength, on the other hand, remains constant at λ = 6.76 nm, so that wavelength matching could be readily maintained and may be obtained by turning the thickness of the bi-layer of the MLM for 6.76 nm. It should be noted that the value of x in 6.x nm should be tuned the value of x = 0.76 to optimally couple to the spectral structure in the CO₂ LPP. Longer wavelength emission in the wavelength range λ = 6.8 – 12 nm, which may be attributed to satellite emission from the dense region in the Nd:YAG LPP, is weaker in the CO₂ LPP case. As a result, the out-of-band emission is reduced in CO₂ LPPs due to their low critical electron density and consequently lower ion density.

The spatial and temporal evolution of the Gd plasma were studied using the modified 1-dimensional Lagrangian, laser-plasma hydrodynamic simulation code MED103 [16] for a cylindrical solid Gd target. Simulations were performed for a Nd:YAG LPP for laser parameters: λ_L = 1064 nm, pulse width, τ_L = 8.5 ns (FWHM) and peak intensity 2 × 10¹² W/cm² incident on a target with a radius of 10 μm, and for a plasma produced by a CO₂ laser, λ_L = 10.6 μm, τ_L = 70 ns (FWHM), peak intensity 2 × 10¹¹ W/cm² incident on a target with a radius of 100 μm, respectively. The electron temperature and density obtained as a function of time and space are presented in Fig. 3. A higher temperature plasma, T_e up to 200 eV, was obtained with the CO₂ laser, as shown in Fig. 3(c), at a distance of 103 μm and a time of around 100 ns. The electron temperature in a plasma rises with increasing laser intensity as $T_e\propto {\left(I_L{\lambda }_L^2\right)}^{0.4}$, where I_L and λ_L are the laser intensity and wavelength, respectively. The higher electron temperature for longer wavelength of λ_L = 10.6 μm at lower laser intensity of the order of 10¹¹ W/cm² is expected. Based on a collisional-radiative (CR) simulation [17], ionic populations of Rh- to Ag-like are maximized at an electron temperature of 80 – 130 eV [15]. The plasma will thus have strong emission at 6.76 nm when T_e is close to 80 eV [15]. When the plasma emission is integrated over its lifetime it then produces a sharp peak at 6.76 nm as shown in Fig. 2. To further illustrate this point, Cowan’s suite of atomic structure codes were used to calculate synthetic spectra for each relevant ion stage. The resulting spectra were then weighted by ion populations calculated within the CR model and summed to give theoretical spectra for given temperatures [17, 18]. The evolution of the spectra around 6.x nm at electron temperatures of T_e = 62, 68, and 74 eV is shown in Fig. 4(a). The theoretical evaluation was performed for an ideal optically thin plasma. The peak intensity at 6.76 nm is expected to increase rapidly due to increased populations of the Gd ionic charged states of Gd¹⁷⁺ and Gd¹⁸⁺. Figure 4(b) shows the theoretical evaluations that best fit the peak and falling edges of the most intense spectrum shown in Fig. 2. An electron temperature of 66 eV was found to give best agreement. The maximum electron temperature was calculated to be T_e ≈ 200 eV. Since the plasma cools by expansion and the bulk of the emission from hydrodynamic modeling has been shown in the past to come from the corona of the plasma, the value of T_e ≈ 66 eV deduced here from fitting to the spectral profile appears reasonable. As the present results were therefore essentially obtained at an electron temperature lower than optimum, a significantly higher CE could be achieved experimentally by improvement of the CO₂ laser system, such as by amplifying the pulse energy or using a shorter pulse duration. It is seen that electron density of Nd:YAG LPP is almost two orders of magnitude higher than that of the CO₂ LPP. The rate of three-body recombination is proportional to $n_e^2$, where n_e is the electron density. Thus a higher degree of ionization is obtained in the CO₂ laser plasma than from Nd:YAG laser plasma. Once the plasma reaches critical density, the remaining part of radiation is mainly absorbed by the coronal plasma. Higher ionic charge states are obtained in the coronal region, which results in less absorption.

 

Fig. 3 Electron temperature, T_e (a,c) and electron density, log n_e (b,d) distributions as a function of time and space using Nd:YAG (left column (a,b)) and CO₂ (right column (c,d)) lasers simulated by the modified hydrodynamic code: MED103.

Download Full Size | PDF

 

Fig. 4 Numerical evaluation of the Gd plasma spectra at the electron temperatures of 62, 68, and 74 eV (a). Theoretical spectra (red, solid line) that best fit the experimental time-integrated spectrum (d) (blue, solid line) shown in Fig. 2 in CO₂ LPP at the laser intensity of 1.3 × 10¹¹ W/cm² with the CE of 0.7% (b).

Download Full Size | PDF

In order to improve the energy CE, it is important to increase the spectral efficiency, i.e., the spectral purity by decreasing the out-of-band intensity. Figure 5(a) shows the laser intensity dependence of the spectral efficiency. Here the spectral efficiency is defined the ratio of the in-band energy at 6.76 nm within the 0.6%BW to the spectral range from 3 to 12 nm. For Nd:YAG LPPs the spectral efficiency decreased from 1.2% to 0.7% with increasing laser intensity because of the increase of emission energy at short wavelengths from 3 to 6.5 nm. The spectral efficiency increased from 1.5% to 2.1% with increasing laser intensity for CO₂ LPPs, and yielded a 2–4 times improvement of the spectral efficiency for CO₂ laser pulse irradiation.

 

Fig. 5 Laser intensity dependence on the spectral (a) and conversion efficiencies (b) in the CO₂ (circles) and Nd:YAG LPPs (rectangles), respectively.

Download Full Size | PDF

The energy CE as a function of the laser intensity at the different laser wavelengths is shown in Fig. 5(b), where the data points and the error bars correspond, respectively, to the average values and the standard deviations with more than 10 shots. The CE grew with increasing CO₂ laser intensity and reached its maximum value of 0.7% at a laser intensity of around 1.2 × 10¹¹ W/cm². In the case of Nd:YAG laser pulse irradiation, however, the CE was lower than in the case of CO₂ laser pulse irradiation, again due to the optical thickness of the plasma produced at a laser wavelength of 1 μm, resulting in low spectral efficiency. The maximum CE was 0.4% at this intensity, which corresponds closely to that obtained at the optimum value of I_L ≈ 4 × 10¹² W/cm² determined in previous work [13]. It is noted that the CE in the CO₂ LPPs was slightly increased by production in a cavity target [14], as previous observed with Sn cavity targets and is due to the presence of a longer density gradient, resulting in an observed maximum CE of about 1% after 5 – 10 shot irradiation at the same position. The cavity structure on the planar target surface was formed by laser ablation. Although this irradiation test was not pursued in detail, the same effect was found in pre-formed plasmas such as with expanded low-density targets which produce longer density gradients for efficient emission [14, 19].

4. Summary and outlook

In summary, we have demonstrated high-efficiency for a beyond EUV source at 6.76 nm, and have proposed a value of x = 0.76 in 6.x nm, because the peak at this wavelength was constant with high spectral and energy conversion efficiencies in optically thin mid-IR CO₂ laser-produced Gd plasmas. The maximum CE was observed to be 0.7%.

To produce not only a Gd plasma with a high electron temperature of the order of 100 eV but also one that has low density and is optically thin, it is necessary to use a mid-IR CO₂ laser due to the low critical density with a short pulse duration less than 1 ns, while at the same time maintaining a laser intensity higher than 2 × 10¹² W/cm². In the near future, a short pulse mid-IR CO₂ laser system will be developed utilizing optical parametric oscillation. After that the seed 10.6-μm pulse with low energy of 100 μJ/pulse will be amplified by a regenerative amplifier system by use of a modified TEA CO₂ laser system, resulting in an expected output energy higher than 100 mJ/pulse with sub-ns pulse duration. It is noted that the pulse duration is limited by the spectral bandwidth in the CO₂ amplifier.

The experimental results also provide a guideline for development of mid-IR laser-induced EUV sources for short wavelength applications, such as EUV lithography and in vivo biological imaging [20, 21].