1. Introduction

Laser produced plasmas (LPPs) play an important role in the development of sources for extreme ultraviolet lithography (EUVL) in semiconductor processing. In particular, LPPs of tin have been demonstrated to be the most promising candidate for EUV light sources for high-volume manufacturing (HVM) EUVL since the EUV spectrum of tin is dominated by intense emission at 13.5 nm which can be reflected efficiently by Mo-Si multilayer mirrors (MLMs) with a reflectivity of ~67.5% [1]. EUV emission at 11.4 nm from Xe LPP [2], was also proposed as a source for EUVL due to a higher reflectivity of near 70% obtainable using Be-based MLMs [3]. Longer wavelength EUVL operation was also proposed by Maloney & Smith [4] since the scaling of optical scatter and flare [5] can decrease as 1/λ². Since the light source has to fulfil a common set of specifications [6], which are agreed by the manufactures of exposure tools for EUVL, it is essential and critical to perform EUV metrology during the development of EUVL tools. Unlike the sources for HVM EUVL which need high-flux EUV emission in a narrow band [7], metrology EUV sources require high average brightness emission with a much broader bandwidth [8] since the testing of EUV optics needs to be performed over an extended wavelength range. Broad-band EUV sources are also of great interest for EUV spectrophotometry which requires broad-band EUV emission with low power and small source size in the 8-40 nm range [9]. Besides, broad-band EUV sources are also required for applications such as surface patterning [10], structural and chemical surface analysis [11,12], reflectometry [13], etc. Highly-ionized Bi plasmas were observed to have a strong and broad emission in the 8-17 nm spectral range in our spectra, which is more continuum-like than the spectra of either tin or alloys containing tin, e.g. galinstan [14]. LPPs of Au and tin-gold alloy have also been used as metrology sources due to the broad-band nature of their spectral emission in the 10-18 nm wavelength range [15,16]. Since the EUV spectra at shorter wavelength (2-8 nm) of Au and Bi are quite similar allowing for the shift of the main features towards shorter wavelength due to the Z-scaling along isoelectronic sequences [17,18]. LPPs of Bi are also expected to provide an efficient broad-band source in the 12-18 nm region. This is further corroborated by spectra reported from a previous survey of LPP emission from a range of medium and high Z targets [19]. In addition Bi and its alloys provide a lower cost alternative to Au, while the fact that some Bi alloys are liquid close to room temperature should render them easy to use in liquid droplet targets. The technology for this source is well understood because of its widespread use in EUVL source schemes.

The unresolved transition array (UTA) emission of ∆n = 0, n = 4-4 transitions from highly-charged Bi ions lies in the “water window” range of 2.3-4.4 nm which already makes Bi a potential material for “water window” imaging [20]. A detailed analysis of Bi spectra in the 1-7 nm wavelength region was reported recently [18]. The spectra of BiI [21,22], BiII and BiIII [23,24], BiIV [25] and BiV [26] were measured and analyzed in the 1930s and 1940s. The spectra of Bi XV and Bi XVI were studied by Churilov et al in 2001 and 2002 [27,28]. Strong resonance lines from highly-ionized BiXXII and BiXXIII were identified in 2014 [29]. In 1998, the Bi spectrum in the 12.4-17 nm from a KrF LPP measured by a spectrometer housing a grating with 600 lines/mm was reported by Shevelko et al as part of a survey to identify EUV sources based on LPPs of heavy elements [19]. However, to date, changes in Bi LPP emission in the 8-17 nm range due to changes in laser power density have not been studied in detail while transitions likely to contribute in this wavelength range remain largely unidentified.

In this paper, the EUV spectral emission in the 8-17 nm region produced by an 8 ns pulse duration Nd:YAG laser were recorded and compared for a range of laser power densities obtained ether by changing the laser energy or adjusting the focal lens positions. The spectra at high laser power densities are expected to be dominated by emissions from highly-charged Bi ions (Bi²³⁺-Bi³⁵⁺), while the spectra at low laser power densities exhibit spectral features mainly due to emission from lower ionized Bi ions (Bi⁶⁺-Bi²⁰⁺). A number of new identifications of spectral features were made by comparing the experimental spectra with calculated results from the Hartree-Fock with configuration interaction (HFCI) code of Cowan [30].

2. Experimental setup

The experiments were performed using a Q-switched Nd:YAG laser (Continuum Surelite III) with an 8 ns full-width at half-maximum (FWHM) pulse duration at a wavelength of 1064 nm. The maximum laser energy reaching the targets is 810 mJ ( ± 20 mJ). The laser beam was guided into a vacuum chamber and focused by a plano-convex lens with a 75 mm focal length. A Bi foil target with a thickness of 1 mm and an area of 2 cm × 3 cm was placed on a remotely adjustable stage with motion along the x-, y- and z-axes to provide a fresh surface for each shot. The beam focus radius was ~20.9 ± 0.4 µm which is the calculated second moment beam waist radius of a heavily attenuated focal spot recorded using a high-quality charge coupled device (CCD) camera. The calculated beam quality factor of the laser was M² = 3. This value was obtained by fitting the beam radius at different positions to the laser propagation equation [31].

The Bi spectra were recorded at various laser power densities, typically two laser shots were required to achieve sufficient signal. Power densities between 7.3 × 10¹² W/cm² and 3.5 × 10¹² W/cm² were obtained by changing the laser pulse energy from 810 mJ to 393 mJ, using polarizing optics, and power densities between 6.8 × 10¹² W/cm² and 5.5 × 10¹⁰ W/cm² were obtained by adjusting the position of the focusing lens whilst maintaining an average laser energy of 750 mJ ( ± 6 mJ). The plasma was viewed at 45ᵒ with respect to the direction of the incident laser beam which was incident normally onto the target using a flat field, grazing incidence spectrometer housing a 1200 grooves mm⁻¹, variable line space grating with a spectral range of 2-18 nm. The spectra were recorded by a thermoelectrically cooled, back-illuminated, X-Ray CCD camera (Andor DX436-BN, 2048 × 2048 pixels). The instrumental spectral resolution was evaluated to be close to 0.01 nm. Wavelength calibration of measured spectra was achieved by using known emission lines from ions of silicon and nitrogen from an LPP formed on a silicon nitride target which gave an uncertainty of ± 0.002 nm. The spectra display a dip at 12.3 nm, a wavelength corresponding to the L-edge in silicon. While the spectra were adjusted to account for this feature in the camera response, the influence of absorption in silicon is still evident due to imperfect detector calibration of the gain discontinuity at 12.3 nm. Evidence from other spectra of bismuth [19] supports the view that the strong dip at 12.3 nm is indeed due to absorption in silicon and that the spectrum would be much flatter in this region with this effect not present. The spectra are presented in this paper with the dip present, but the analysis does not attribute this dip to an absence of emission from the bismuth plasma in this region.

3. Experimental results

The measured EUV spectra of ns-laser produced Bi plasmas in the 8-17 nm spectral range at six different laser power densities from 7.3 × 10¹² W/cm² to 3.5 × 10¹² W/cm² are shown in Fig. 1(a). The laser power density was changed by adjusting the laser energy while keeping the laser beam focus radius at approximately 21 µm. A second series of Bi EUV spectra at a larger range of laser power densities varying from 6.8 × 10¹² W/cm² and 5.5 × 10¹⁰ W/cm² is shown in Fig. 1(b). The observed spectra are dominated by a broad intense feature peaking near 11 nm which contains a plethora of individual peaks from both discrete lines and unresolved transition arrays (UTAs) [32] from various ion stages. The spectra in Fig. 1 are labeled as I (8-10.5 nm), II (10.5-12.4 nm), III (12.4-13.5 nm) and IV (13.5-17 nm). The distinction between regions II and III is somewhat artificial as in the absence of the Si K-edge feature they would merge to form a single UTA. When examining the spectra in Fig. 1(a) it is possible to identify two distinct trends, firstly, the relative intensity in region I increases with increasing power density, while in regions III and IV, the relative intensity drops with increasing power density. This is reflective of the role that higher ion stages play in emission as the plasma temperature increases with increasing power density. All of the spectra in Fig. 1(a) share a similar general shape with a relatively stronger component of the broad-band UTA providing emission in regions I and II and a weaker but flatter component of the UTA band covering the 13.5 nm range in region III, which makes the spectrum of Bi very feasible for metrology application in EUVL (both at 11.4 nm and 13.5 nm) where the broad-band spectrum is needed and also source size is an issue [9]. In Fig. 1(a), it is noteworthy that the spectral profiles at 12.5-14.5 nm remain unchanged and relatively flat within a significant range of laser power densities thus exhibiting both the stability and reliability required for EUV mask spectral reflectometry applications [9].

 

Fig. 1 Experimental spectra of a bismuth laser produced plasma with a range of power densities created by (a) varying the laser energy and (b) varying the focused spot size.

Download Full Size | PDF

The estimated electron temperature of the Bi LPP at the lowest laser power density of 3.5 × 10¹² W/cm² in Fig. 1(a) is 459 eV using the collisional-radiative (CR) model [33]. This plasma temperature is high enough to produce ions of Bi²³⁺-Bi³⁵⁺, since at plasma temperature of 459 eV the ions predicted to dominate in the plasma are Bi³⁰⁺-Bi³⁹⁺ using the CR model. At the highest power density of 7.3 × 10¹² W/cm², the ions predicted to dominate in the plasma are Bi³⁴⁺-Bi⁴⁵⁺. Since the spectrum is integrated over time and therefore a range of plasma temperatures stages lower than these are expected to contribute also with their emission coming from increasing larger plasma volumes since the plasma temperature decreases away from the interaction region. Since the ion stages present are not too dissimilar, all of the spectra in Fig. 1(a) possess the similar profiles except that the intensity decreases with decreasing laser power density. Since the diameter of the laser beam focus was held constant when recording these spectra in Fig. 1(a), any increase in laser pulse energy helps to ablate more material and produces a slightly larger but hotter plasma which contains more highly-charged ions that emit in this region. An increase in power will thus increase the intensity of the feature and cause the peak of the lower wavelength feature (region II) move slightly to shorter wavelength as shown in Fig. 1(a).

Apart from the reported spectra of Bi²¹⁺ and Bi²²⁺ [29], very little information is available on the spectral analysis of moderately or highly-charged Bi ions. The calculated theoretical results indicate that resonance emission stretches from 8 to 17 nm depending on the ion stages and corresponding transitions involved. A significant fraction of the complex structure from 8.5 to 17 nm in Fig. 1(a) is primarily the result of excited to excited state transitions of the type 5s-5p, 5p-5d, 5d-5f and 5f-5g outside open 4f subshells in ions from Bi²³⁺ to Bi³⁵⁺ where different transition arrays contribute at different wavelengths. The wavelength ranges contributed by the above four sets of transitions from Bi²³⁺-Bi³⁵⁺ are shown in Table 1. No contribution is expected from ion stages higher than Bi³⁵⁺ since their resonance lines are predicted to lie at significantly shorter wavelengths [18].

Table 1. Wavelength ranges (∆λ) of emission from transitions of 4d¹⁰4f^m-15p¹ – 4d¹⁰4f^m-15d¹, 4d¹⁰4f^m-15d¹ – 4d¹⁰4f^m-15f¹, 4d¹⁰4f^m-15s¹ – 4d¹⁰4f^m-15p¹ and 4d¹⁰4f^m-15f¹ – 4d¹⁰4f^m-15g¹, written in short as 5p-5d, 5d-5f, 5s-5p and 5f-5g respectively in the table, in ions of Bi²³⁺-Bi³⁵⁺.

View Table | View all tables in this article

To more clearly show the different spectral features at the larger range of laser power densities as shown in Fig. 1(b), the same series of Bi spectra, normalized to the maximum intensity measured in the 8-17 nm region are shown in Fig. 2. Here, the power density was changed by adjusting the position of the focusing lens with using a motorized linear actuator. The laser energy was kept constant while the lens was moved to various positions including the focal position. In these experiments the focus of the laser was moved into the target, so that the laser was not focused in the plasma. The decrease in power density is the result of defocusing and the associated increase in spot size produces a larger, cooler plasma that will exhibit greater opacity effects.

 

Fig. 2 Bi spectra shown in Fig. 1(b) normalized to the maximum intensity in the 8-17 nm range.

Download Full Size | PDF

Again, the increased relative intensity in region I obtained with increasing power density is evident in Fig. 2. As the laser power density decreases, the intensity of the main UTA in region II increases until the laser power density reaches 1.1 × 10¹² W/cm² as shown in Fig. 1(b). Moreover, a dramatic narrowing of region II with decreasing intensity is seen as the laser power decreases, along with a more dramatic increase in the relative intensity of the emission in regions III & IV in Fig. 1(b) and Fig. 2, which is due to decreasing population of highly ionized ions and the increased presence of lower ionized ions along with a greater influence from opacity in the larger size plasmas. The maximum intensity was obtained at a laser power density of 1.1 × 10¹² W/cm² as shown in Fig. 1(b) and Fig. 2 in black, which still exhibits broad-band UTA emission in the studied range and thus makes it good candidate for applications where flux is a major issue regardless of the large source size.

The ionic populations of Bi ions at different power densities corresponding to those in Fig. 1(b) and Fig. 2 were estimated using the CR model and are also in good agreement with the calculated results from Sasaki et al [34]. The results are shown in Fig. 3. The measured spectra arise from the time and space integrated emission recorded over the entire plasma lifetime, therefore transitions due to the lower ion stages also contribute but the calculated populations would be expected to dominate when the plasma is hottest, which is when the bulk of the EUV emission is generated. When the laser power density decreases from 6.8 × 10¹² W/cm² to 1.1 × 10¹² W/cm², more open 4f subshell ions (Bi²³⁺-Bi³⁵⁺) are produced as shown in Fig. 3, which is the reason why the spectral intensity increases with the decreasing power density as shown in Fig. 1(b). Moreover, since the transitions involved are between excited states, opacity effects are negligible. However, when the power density decreases to 1.8 × 10¹¹ W/cm² or even lower, no open 4f subshell ions are expected, but open 5p (Bi¹⁵⁺-Bi²⁰⁺) and 5d (Bi⁵⁺-Bi¹⁴⁺) subshell ions are produced as shown in Fig. 3, which results in the decrease of spectral intensity over the whole range, the shift to longer wavelength and narrowing of region II and new features that emerge in region I and IV with the decrease in laser power density as shown in Fig. 1(b) and Fig. 2. At the lower laser power densities, opacity effects due to resonance absorption from ions with outermost 5d and 5p subshells in the plasma periphery are expected to become more important also and significantly affect the profile of the main UTA.

 

Fig. 3 The ionic populations of Bi ions at six different power densities corresponding to those that generated the spectra seen in Fig. 1(b) and Fig. 2.

Download Full Size | PDF

4. Theoretical calculations and analysis

To interpret the influence of emission from the contributing ions, computations of transitions were performed for Bi⁵⁺ to Bi³⁵⁺ using the Hartree-Fock with configuration interaction (HFCI) code of Cowan [30]. The Slater-Condon parameters were scaled to 70% of their ab initio values which gives best agreement with experimental results while the spin-orbit integrals were unchanged. No additional wavelength shift is given to the calculated transitions since the calculated results are in good agreement with the measured spectra and the experimental arrays overlap significantly. In each figure, the theoretical oscillator strength (gf) is shown in the same scale in the form of binned stick plots with widths of 0.01 nm in line with the instrumental resolution of the spectrometer. The calculated results in each figure share a common scale, where the relative intensities are the calculated oscillator strength (gf) for each specific ion stage. Since the corresponding transition probabilities (gA) can be obtained from the relationship between gA and gf, where gA = 0.66702σ²gf(s⁻¹) [30] and σ is the wavenumber in Kaysers, the relative intensities are proportional to the calculated gf values within a given ion stage. The calculated theoretical spectra in each figure represent solely the gf value distributions.

4.1. Bi theoretical spectra at higher laser power densities

The experimental spectrum recorded at the maximum laser power density, along with calculated positions of ∆n = 0, 5p-5d, 5d-5f and 5s-5p type transition spectra of Bi²³⁺ – Bi³⁵⁺, is shown in Fig. 4, where the green, red, blue and purple lines correspond to the calculated spectra, which is the calculated gf values, while the black curve corresponds to the experimental spectrum of Bi. The appearance of a few discernable strong peaks from 8.5 to 9.5 nm is mainly due to the overlapping of broad spectral features from the 4d¹⁰4f^m-15p¹ – 4d¹⁰4f^m-15d¹ transitions in adjacent ion stages from Bi²⁹⁺ to Bi³⁴⁺ at high power densities as shown in Fig. 4(a).

 

Fig. 4 Comparisons of experimental (black) and theoretical spectra of Bi from ns LPPs at the maximum laser power density for the transition type ∆n = 0, n = 5, 4d¹⁰4f^m-15p¹ – 4d¹⁰4f^m-15d¹, where m = 14-2 (Bi²³⁺–Bi³⁵⁺) (green); ∆n = 0, n = 5, 4d¹⁰4f^m-15d¹ – 4d¹⁰4f^m-15f¹, where m = 14-2 (Bi²³⁺–Bi³⁵⁺) (red); ∆n = 0, n = 5, 4d¹⁰4f^m-15s¹ – 4d¹⁰4f^m-15p¹, where m = 14-2 (Bi²³⁺ – Bi³⁵⁺) (blue) and ∆n = 0, n = 5, 4d¹⁰4f^m-15f¹ – 4d¹⁰4f^m-15g¹, where m = 14-2 (Bi²³⁺–Bi³⁵⁺) (purple).

Download Full Size | PDF

The peaks labeled as “a” through “l” from 10.6 to 11.6 nm in Fig. 4(a) become weaker and then disappear with the decreasing laser power density as shown in Fig. 1(b) and Fig. 2, since they are predominantly the result of the overlapping of spectral lines from the 4d¹⁰4f^m-15d¹ – 4d¹⁰4f^m-15f¹ transitions in adjacent ions from Bi²⁵⁺ to Bi³⁴⁺ and 4d¹⁰4f^m-15s¹ –4d¹⁰4f^m-15p¹ in Bi³¹⁺-Bi³⁵⁺ as shown in Fig. 4(a). The assignment of these discrete peaks to specific ionization stages was performed by comparing the structured features in the experimental spectrum with calculated UTA peaks given in Table 2. In this table, the positions of the lower and upper wavelengths of the observed arrays corresponding to the labels in Fig. 4(a) and their origins are presented. From the results in Fig. 4(a), it should be noted that the 5d-5f transitions are completely separated into two distinct bands due to the large 5f spin-orbit splitting.

Table 2. Maximum and minimum wavelength (nm) assignment of features from 10.6 to 11.6 nm.

View Table | View all tables in this article

A number of discrete spectral features appear in the region extending from 12.4 to 13.3 nm and 13.7–15 nm as shown in Fig. 4(b), making it possible to associate individual peaks with emission from specific ions. The features labeled as “A” through “H” in the 12.4-13.3 nm region in Fig. 4(b) are contributed by the overlapping of spectral lines from transitions 4d¹⁰4f^m-15p¹ - 4d¹⁰4f^m-15d¹ in Bi²⁸⁺- Bi³³⁺ and 4d¹⁰4f^m-15s¹ - 4d¹⁰4f^m-15p¹ in Bi²⁶⁺- Bi²⁸⁺ as shown in Figs. 4(a) and 4(b) respectively. As shown in Fig. 4(b), due to the 5f spin-orbit splitting, the 5f-5g transitions are separated into two distinct bands and the overlapping of spectral lines from adjacent ion stages from Bi²³⁺-Bi³⁰⁺ is responsible for the jagged appearance of the peaks in the 13.7-15 nm region at high power densities while transitions 4d¹⁰4f^m-15p¹ - 4d¹⁰4f^m-15d¹ in Bi²³⁺-Bi²⁷⁺ and 4d¹⁰4f^m-15s¹ - 4d¹⁰4f^m-15p¹ in Bi²³⁺-Bi²⁴⁺ also make a small contribution to this region as shown in Figs. 4(a) and 4(b) respectively. The tentative assignment of discrete peaks in the 12.5-15 nm region to specific ion stages is given in Table 3 where the positions of the lower and upper wavelengths of the observed arrays correspond to the alphanumerical labels in Fig. 4(b) and their origins are presented.

Table 3. Maximum and minimum wavelength (nm) assignment of features from 12.5 to 15.7 nm.

View Table | View all tables in this article

4.2. Bi theoretical spectra at lower laser power densities

The experimental spectrum at the minimum laser power density, along with calculated spectra of the type ∆n = 0, 5p-5d + 5d-5f, ∆n = 1, 5d-6f and ∆n = 0, 5s-5p + 5p-5d from lower ionized Bi ions is shown in Fig. 5. The black curve corresponds to the experimental spectrum of Bi, while the red, green, purple, blue and yellow lines correspond to the calculated gf values.

 

Fig. 5 Comparisons of experimental (black) and theoretical spectra of Bi from ns LPPs at the minimum laser power density for the transition types: ∆n = 1, n = 5 - n = 6, 5d^m – 5d^m-16f¹ (Bi⁵⁺ - Bi¹⁴⁺) (red); ∆n = 0, n = 5, 5p^m –5p^m-15d¹ (Bi¹⁶⁺–Bi²⁰⁺) (green); ∆n = 0, n = 5, 5s^m – 5s^m-15p¹ (Bi²¹⁺–Bi²²⁺) (purple) and ∆n = 0, n = 5, 5p⁶5d^m – 5p⁵5d^{m + 1} (blue) + 5p⁶5d^m-15f¹ (yellow) (Bi⁷⁺–Bi¹⁴⁺).

Download Full Size | PDF

The discrete peaks in Bi spectrum shown in Fig. 4(a) in the 8.5-9.5 nm region become very weak at the minimum laser power density due to it being too low to produce ions of Bi²⁹⁺ to Bi³⁴⁺ which contribute in this region at high power densities. The weak peaks apparent in this range are due to emissions from transitions of 5d^m - 5d^m-16f¹ in Bi¹¹⁺ -Bi¹⁴⁺ at low power densities. The peaks labeled as “a1” through “a3” in Fig. 5 show up when the laser power density drops to 1.8 × 10¹¹ W/cm² as shown in Fig. 1(b). These are mainly due to transitions of the type 5d^m - 5d^m-16f¹ in Bi¹⁰⁺ and Bi¹¹⁺ as shown in Fig. 5.

The peaks labeled as “a” through “l” in Fig. 4(a) completely disappear at low power densities, along with the narrowing of the broad peak in region II as shown in Fig. 2, which is due to the decrease in the populations of the ions Bi²⁵⁺ to Bi³⁴⁺ and a corresponding increase of the populations of Bi⁸⁺-Bi¹⁴⁺ in the plasma. The sharper peak is attributed to transitions 5p⁶5d¹ - 5p⁵5d² in Bi¹⁴⁺, 5d^m - 5d^m-16f¹ in Bi⁹⁺ and 5p^m - 5p^m-15d¹ in Bi¹⁶⁺-Bi¹⁷⁺ as shown in Fig. 5. Note that the 5p⁶5d^m - 5p⁵5d^{m + 1} will mainly influence absorption, due to the low probability of excitation from subvalence subshells until configuration with m = 1 is reached [18].With decreasing laser power density, the peaks labelled as “A” through “H” in Fig. 4(b) in the 12.4-13.3 nm region become weaker and finally disappear and some new features emerge as shown in region III in Fig. 2, which are due to transitions of the type 5p⁶5d^m - 5p⁶5d^m-15f¹ in Bi⁶⁺-Bi¹²⁺ and 5d^m - 5d^m-16f¹ in Bi⁸⁺ that predominantly contribute in this region as shown in Fig. 5. Similarly, with decreasing laser power density, the peaks labeled as “I” through “M” shown in Fig. 4(b) also disappear, along with new features and a new broad peak centered at around 14 nm appears as shown in region IV in Fig. 2, which is due to transitions of the type 5p⁶5d^m - 5p⁶5d^m-15f¹ in Bi⁶⁺-Bi¹⁴⁺, 5d^m - 5d^m-16f¹ in Bi⁵⁺-Bi⁷⁺ and 5p^m - 5p^m-15d¹ in Bi¹⁶⁺-Bi²⁰⁺ as shown in Fig. 5.

For transitions of the type 5p⁶5d^m - 5p⁵5d^{m + 1} + 5p⁶5d^m-15f¹, configuration interaction (CI) effects are not negligible, as in the case of 4p⁶4d^m - 4p⁵4d^{m + 1} + 4d^m-14f¹ transitions in the open 4d subshell ions [35,36], due to the proximity of excitation energies of the upper configurations. CI effects are taken into account in the calculated spectra in Fig. 5. As a specific example, the effect of CI on Bi¹⁰⁺ n = 5-5 resonance transitions is shown in Fig. 6 which indicates that the configuration interaction results in a shift of the long wavelength subarray beyond 15 nm to shorter wavelengths and the redistribution of oscillator strength towards the shorter wavelength subarray, causing a narrowing of the spectral width of the band that is repeated for each of the ions Bi⁶⁺-Bi¹⁴⁺. Note that in practice, the 5p⁶5d^m - 5p⁶5d^m-15f¹ transitions are expected to contribute to both absorption and emission whilst the 5p⁶5d^m - 5p⁵5d^{m + 1} will mainly influence absorption, as stated earlier.

 

Fig. 6 The 5p⁶5d⁵- (5p⁵5d⁶ + 5d⁴5f¹) configuration interaction in the Bi¹⁰⁺ spectrum, the configurations are not mixed in the upper plot but are mixed in the lower plot. The coordinate ranges are the same for both plots.

Download Full Size | PDF

5. Conclusion

EUV emission spectra of Bi were recorded in the 8-17 nm spectral region using the LPP technique at various laser power densities obtained by adjusting either the laser energy or the lens position, using a Nd:YAG laser with a pulse duration of 8 ns, where the maximum focused power density was around 7.3 × 10¹² W/cm². With decreasing laser energy, the observed Bi spectral profiles between 8 and 17 nm remain almost the same except the intensity becomes lower, rendering such plasmas very suitable for metrology applications such as EUV lithography or EUV spectrophotometry. This is due to the broad-band UTA feature covering both 11.4 nm and 13.5 nm and the small EUV source size. In addition, with increasing laser spot size incident on the target, the intensity of Bi spectra from 8 to 17 nm increases first and is strongest at a power density of around 1.1 × 10¹² W/cm², and then decreases when the power density goes down to 1.8 × 10¹¹ W/cm². With a further reduction in power density, new features are seen to appear. The Bi LPP with the maximum intensity at 8-17 nm occurs at the laser power density of 1.1 × 10¹² W/cm² and is a good candidate for applications where EUV flux is critical regardless of the source size such as EUV reflectometry.

The main features of the experimental EUV spectra were identified by comparisons with the results calculated using the Cowan suite of atomic structure codes. Quasicontinuous radiation in the 8-12.5 nm and 12.5-17 nm regions dominates the spectra resulting from n = 5 - n = 5 UTAs in Bi²³⁺–Bi³⁵⁺ at high power densities and Bi⁵⁺–Bi²²⁺ at low power densities.