Characterization of angularly resolved EUV emission from 2-$\mu$m-wavelength laser-driven Sn plasmas using preformed liquid disk targets

The emission properties of tin plasmas, produced by the irradiation of preformed liquid tin targets by several-ns-long 2-$\mu$m-wavelength laser pulses, are studied in the extreme ultraviolet (EUV) regime. In a two-pulse scheme, a pre-pulse laser is first used to deform tin microdroplets into thin, extended disks before the main (2$\mu$m) pulse creates the EUV-emitting plasma. Irradiating 30- to 300-$\mu$m-diameter targets with 2-$\mu$m laser pulses, we find that the efficiency in creating EUV light around 13.5nm follows the fraction of laser light that overlaps with the target. Next, the effects of a change in 2-$\mu$m drive laser intensity (0.6-1.8$\times 10^{11}$W/cm$^2$) and pulse duration (3.7-7.4ns) are studied. It is found that the angular dependence of the emission of light within a 2\% bandwidth around 13.5nm and within the backward 2$\pi$ hemisphere around the incoming laser beam is almost independent of intensity and duration of the 2-$\mu$m drive laser. With increasing target diameter, the emission in this 2\% bandwidth becomes increasingly anisotropic, with a greater fraction of light being emitted into the hemisphere of the incoming laser beam. For direct comparison, a similar set of experiments is performed with a 1-$\mu$m-wavelength drive laser. Emission spectra, recorded in a 5.5-25.5nm wavelength range, show significant self-absorption of light around 13.5nm in the 1-$\mu$m case, while in the 2-$\mu$m case only an opacity-related broadening of the spectral feature at 13.5nm is observed. This work demonstrates the enhanced capabilities and performance of 2-$\mu$m-driven plasmas produced from disk targets when compared to 1-$\mu$m-driven plasmas, providing strong motivation for the use of 2-$\mu$m lasers as drive lasers in future high-power sources of EUV light.


I. INTRODUCTION
Laser-produced plasmas containing highly charged tin ions are the light source of choice for state-of-the-art extreme ultraviolet (EUV) lithography [1][2][3][4][5][6][7][8][9][10][11]. Tin is used because no less than five of its charge states (Sn 10+ -Sn 14+ ) strongly emit in a narrow band around 13.5 nm [12][13][14][15][16][17][18][19][20][21][22] that matches the peak reflectivity of available multilayer optics [23,24]. The tin ions are bred in a hot (∼30-60 eV) and dense (10 19 -10 21 e − /cm 3 ) plasma. Starting from mass-limited tin-microdroplets, a low intensity pre-pulse (PP) deforms the droplets into a shape better suited for interaction with a second high-intensity main pulse (MP), used to create the EUV-emitting plasma. This two-step process is crucial for reaching source efficiencies and power levels that allow for the industrial utilization of EUV lithography [25][26][27][28]. Currently, 10.6-µm-wavelength CO 2 gas lasers are used to drive the plasma. Recent simulation studies however have drawn significant attention to the use of a 2-µm main pulse, a wavelength at which high-power solid-state lasers may soon become available [29]. Typical advantages of near-to mid-infrared solid state over gas lasers are their more compact build and higher wall-plug efficiency. Shorter drive laser wavelengths have the additional advantage of a higher coupling efficiency of laser light with the tin plasma. Given a higher critical electron density (n c ∝ λ −2 ), * o.versolato@arcnl.nl the shorter wavelength laser light is absorbed in regions with higher emitter and absorber density. This may benefit the obtainable source brightness but an associated increase in optical depth [30,31] leads to increased broadening of spectral features outside of the for EUV lithography relevant inband region, defined as a 2% bandwidth centered at 13.5 nm. This broadening may limit the obtainable conversion efficiency (CE) of laser energy into in-band radiation into the backward hemisphere towards the laser origin. Limitations on the spectral performance of EUV sources imposed by optical depth have been studied in detail for plasmas driven by 1-µm Nd:YAG lasers [30,32,33]. A 2-µm wavelength, between the widely investigated 1 and 10 µm, is an interesting candidate providing intermediate plasma densities and hence optical depth. In addition to 2-µm systems based on difference frequency generation [31,34,35], high-power Big Aperture Thulium (BAT) laser systems, operating at 1.9-µm wavelength, are currently under development [36,37].
Despite its importance for EUV lithography, literature on the production of in-band radiation from mass-limited, prepulse deformed targets remains rather scarce, with the majority of studies to-date focusing on λ = 10 µm main pulses (see, e.g., [28,[38][39][40]). A broad body of literature has investigated experimentally the emission from undeformed liquid tin droplet targets [9,32,41,42], coated spheres [43] or solidplanar targets [44][45][46][47] using 1-µm solid-state lasers. First results from experiments using 2-µm laser light driving plasma from solid-planar tin targets [35] as well as from undeformed liquid tin droplet targets [31] have recently been presented. Both works demonstrate the potential of the 2-µm drive wavelength, showing a doubling of the obtainable CE over the 1µm driver, to a 3% level for the solid-planar tin target. At this 3% level the overall conversion efficiency of wall plug power to in-band EUV may be at par with the current CO 2 -driven industrial solutions. Literature on industrially-relevant masslimited deformed targets interacting with either 1-or 2-µm main pulse beams at high CE is however not yet available.
In this publication, EUV light production from masslimited tin targets, suitably shaped by a 1-µm laser pre-pulse, is investigated using 1-and 2-µm main pulse laser systems to drive the plasma. First, plasma produced with a 2-µm laser beam is investigated for three different laser intensities and three laser pulse durations. For each laser parameter set a wide range of target diameters (30-300 µm) is investigated and the resulting angular distribution of in-band emission, the efficiency of converting laser light into in-band EUV radiation and overall EUV radiation, as well as the spectral performance of the plasma, are characterized. Secondly, plasmas generated by 1-and 2-µm drive laser light are characterized and compared.

II. EXPERIMENT
In the experiments, micrometer-sized liquid tin droplets are first irradiated with a relatively low intensity (∼10 9 W/cm 2 ), 1-µm wavelength PP (see Fig. 1(a)) from an Nd:YAG laser. The PP propels the droplets and deforms them into extended, disk-like targets of diameter d T [48][49][50][51] with typical radial expansion speeds of ∼90 m/s. The target diameter is precisely controlled via the time delay between pre-and main pulse, which ranges 0-3 µs. In all measurements concerning 2-µm main pulses, the spatial beam profile of this PP laser is Gaussian with a size of 120 µm at full width at half maximum (FWHM) and a FWHM duration of 25.9 ns (see Fig. 1(d)). A constant PP energy of 8.4 mJ is used throughout the experiments. We use circular polarization, to ensure that the produced targets are radially symmetric [50,52]. The produced tin targets are observed under angles of 90°(side view) and 150°(front view) with respect to the laser beam using combinations of CCD cameras and long-distance microscopes (see Fig. 1(b)). Temporal resolution is achieved by back-lighting the tin targets with spatially and temporally incoherent 560nm-wavelength pulses of 6 ns duration. Examples of typical targets are shown in Fig. 1(c).
After a set time delay the targets are irradiated with highintensity 2-µm-wavelength laser pulses. The pulses are produced in a master oscillator power amplifier (MOPA) [31,35] that was built following the work of Arisholm et al. [34]. Signal and idler pulses having energies up to 180 mJ each are produced at wavelengths of 1.9 and 2.1 µm, respectively. Depending on the precise experimental conditions, the pulse duration of the MOPA system can vary between 3.7 to 4.6 ns, and the exact, measured pulse durations are stated for each measurement in the following.
For  via polarization optics and only the idler beam of the MOPA is focused onto the tin targets while scanning laser intensity and target size. The focal spot of the idler beam is Gaussian and has a FWHM of 106 µm (see Fig. 1(e)).
In a second set of experiments, the pulse duration of the 2µm beam is varied between 3.7 ns (idler only) and 7.4 ns (idler and signal) at equal laser intensity. To achieve the longer pulse duration, the signal beam, which is separated from the idler via a thin film polarizer (TFP), is sent into an optical delay line of 1.2 m length. The focusing conditions of the delayed signal beam are matched to those of the idler by adjustment of beam size and collimation via a telescope within the delay line, resulting in a Gaussian focal spot size of 106 µm FWHM for both beams. After collimation adjustment the beams are combined again with a second TFP. The signal energy is adjusted via a λ /2-waveplate before this second TFP.
To enable a direct comparison of plasmas driven by 1-and 2-µm-wavelength laser pulses, a separate and final set of ex-periments is conducted where the 1-µm beam of a third laser system is used as MP (lower part of Fig. 1(a)). This 1-µm MP laser is a seeded Nd:YAG laser with arbitrary temporal pulse-shaping capabilities [53] that are used to reproduce the temporal profile of the 4.3 ns and the 7.4 ns 2-µm beam. The pump laser of the MOPA is used for pre-pulsing in this third set of experiments. The PP parameters are tuned to obtain similar radial expansion speeds as in experiments 1 and 2 and no significant change in EUV emission is observed when using either PP. Using telescopes, 1-µm PP and MP are adjusted in size and collimation. For the MP a symmetric, 90-µm-sized focal spot is achieved close to the dimensions of the 2-µm MP. The PP has a spot size of 100 µm and a pulse duration of 10.4 ns. PP and MP are combined via a polarizing beam cube at which their pulse energy is adjusted via a preceding λ /2waveplate. Consecutively, the combined beams are steered onto the droplet, passing a λ /4-waveplate just before the final focusing element. The latter changes the polarization of the beams from linear to circular, ensuring a symmetric deformation of the tin droplets by the PP laser as before. Laser intensity is calculated with respect to its peak in space and time according to I L = (2 2 ln 2/2π) 3 E L /abt p , where E L is the laser energy, a and b are the major and minor axis of the slightly elliptical beam given as FWHM, and t p is the pulse duration.
The spectral emission from the plasma is recorded using a transmission grating spectrometer [54]. The spectrometer is mounted at an angle of 60°with respect to the laser axis (see Fig. 1(b)) and is operated with a 25 µm slit and a 10 000 lines/mm grating. Subsequent to recording the spectra with a CCD camera, a dark exposure is subtracted from the images in order to account for thermal and readout counts on the CCD. Next, spectra are corrected for second order diffraction from the grating, the grating's first order diffraction efficiency and the quantum efficiency of the CCD.
The absolute amount of in-band radiation is measured with four photodiode assemblies installed under angles of θ = 30, 42, 64 and 115°with respect to the target's surface normal. The photodiode assemblies use a multilayer mirror near normal incidence to reflect the in-band light onto the photodiode detectors. Remaining optical radiation is filtered by an EUVtransmissive Si/Zr coating on the photodiodes. Next, following the approach of Ref. [32] (and analogously to [44] and [55]) the measured in-band energy values are corrected for the respective solid angle and fitted with the monotonous smooth function with amplitude α, offset β and power γ. We note that this function differs from the fit function ∼ cos (θ ) γ used in Schupp et al. [32]. The adaption is needed to capture the emission under angles θ > 90 • also in cases γ < 1. The resulting fitted curves, using either fit function, are found to be indistinguishable over the backward hemisphere where θ < 90 • . The integral amount of in-band radiation over the hemisphere around the incoming laser beam is then calculated by integra-tion of Eq. (1) Using this value, CE is defined as E IB,2π /E L . We further define an anisotropy factor of the in-band emission to gauge which fraction of in-band light is emitted into the backward hemisphere via E IB,2π /E IB,4π , where E IB,4π is obtained by changing the upper integration boundary in Eq. 2 to π, hence integrating over all angles. A large anisotropy factor is favorable for industrial application as light-collecting optics typically cover the backward hemisphere. A second measure for source performance is the spectral purity (SP), defined as the ratio of in-band energy to the total EUV energy. All SP values provided in the following are calculated with respect to the spectral range of 5.5-25.5 nm measured under a 60 degree angle. A third measure, the radiative efficiency η rad of the plasma is defined as the ratio of CE over SP following Ref. [32] and yields the total amount of EUV light emitted in the 5.5-25.5 nm wavelength band per incoming laser energy. In the current work, and in contrast to Ref. [32], η rad is defined using the values measured at 60 • , i.e., using the in-band EUV energy emitted per steradian under an angle of 60 • (this value is subsequently multiplied with 2π), meaning (E IB,60 • /E L )/SP, instead of using CE that is defined over the entire laser-facing hemisphere. This is undertaken because the spectra and hence the SP values are expected to show strong angular dependence for expanded targets [47,56], a dependence that is further expected to be influenced by the target diameter.

III. EUV GENERATION USING 2-µm LIGHT ON PREFORMED TARGETS
In the following, the influence of target diameter, laser intensity and laser pulse duration on CE, SP, η rad , E IB,2π and E IB,2π /E IB,4π are investigated. Further, the spectral emission and angular distribution of the in-band emission are discussed.

A. Target diameter
To investigate the influence of target size, tin targets of various sizes as produced from pre-pulse-impacted 27-µmdiameter droplets are irradiated with laser pulses having an intensity of 0.6 × 10 11 W/cm 2 and a pulse duration of 4.5 ns (downward triangles in Fig. 2). Data recorded using higher laser intensities and different pulse lengths are also shown but will be discussed in Sections III B and III C, respectively. Starting from an undeformed droplet target, CE has a low value of approximately 0.15% that almost linearly increases with target diameter (see Fig. 2(a)). At diameters of 170 µm and above, CE starts to plateau at a value of approximately 2.6%. This plateau sets in where the target is roughly a factor of 1.5-2 larger than the FWHM size of the beam. A similar result was observed for 1-µm-driven plasmas from tin coated glass spheres by Yuspeh et al. [57]. An increase in CE with target diameter is expected as an enlarged target diameter increases the geometrical overlap of laser beam and target. This overlap fraction is referred to as energy-on-target (EoT, following the definition in Refs. [48] and [49]). We note that there is a near complete absorption of laser light by the efficient inverse bremsstrahlung mechanism for this wavelength laser light at the here-relevant intensities [49], where laser light geometrically overlaps with the target. For a circular spot, EoT is given by the function which is an excellent approximation for the current focal spot shapes (cf. Fig. 1(e)), where d L = (a + b)/2 and a and b are similar to within a few percent. The free fit factor A, the amplitude of the EoT curve, is obtained from a global fit to the asymptotic CE value of all data presented in Fig. 2(a) and 2(e). The observed dependence of CE on target size is seen to be well approximated by the geometric overlap function (dashed line in Fig. 2(a)).
All CE values are calculated from angular-resolved measurements of the plasma's in-band emission and all data points represent the mean over 300 individual laser shots. The angular dependence of the measured in-band energies (see Fig. 2(k)) is first fitted with Eq. (1) and the CE value is subsequently calculated using the integration result from Eq. (2). For better visibility, all curves in Fig. 2(k) are normalized at 0°. Eq. (1) is seen to accurately describe the angular dependence of the in-band EUV emission. The increase of in-band light observed at a target diameter of 94 µm and under an angle of 42°is however not fully captured by the fit function and this data point may in fact be an outlier.
The reduced emission in the propagation direction of the laser at 180°with increasing target size may be caused by the one-sided heating of the expanded tin target by the laser beam. Plasma emission in this direction is shielded by the target (which is still thought to be present in liquid form during the laser pulse) and only plasma formed on the edge of the target presumably contributes to in-band emission under angles larger than 90°. The dashed line in Fig. 2(k) shows Lambert's cosine law for the angular emission from a planar surface. With increasing target diameter the in-band emission starts to converge to this cos(θ ) dependence, however without fully reaching it. This might be expected because of the 3-dimensional extent of the plasma, causing departure of the emission characteristics from a Lambert-type distribution. Furthermore, emission will always exist at angles larger than 90°from the edge of the plasma unless d T d L .
Next, the angular in-band distribution is quantified via an anisotropy factor, defined as the ratio of E IB,2π /E IB,4π , meaning the fraction of all in-band energy that is emitted into the backward hemisphere of the incoming laser beam, relevant for EUV lithography applications. The anisotropy factor starts at a value of 0.57 for the droplet target, where 0.5 indicates an equal split between both hemispheres (see Fig. 2(d)). The anisotropy factor then linearly increases with target diameter up to a value of 0.79, meaning that an ever-larger fraction of the in-band EUV is emitted into the backward hemisphere.
The in-band energy emitted in the 2π hemisphere towards the laser (obtained from multiplying the CE with the laser pulse energy) is shown in Fig. 2(i). Naturally, the same trends are observed as in the case of CE. A maximum of ∼1 mJ of inband EUV energy is produced (per laser shot) for this lowest intensity case. Fig. 2(j) depicts spectra from the lowest intensity case for a succession of four target diameters. For visibility and ease of comparison the spectra are normalized to their respective maximum values. Similar to previously published work, where a 2-µm laser was used to produce plasma from spherical, undisturbed droplet targets [31], all spectra show a strong emission feature at 13.5 nm originating from transitions in Sn 8+ -Sn 14+ ions [6]. Transitions in these charge states further lead to the radiation observed in the 6-12 nm wavelength region [58]. With increasing target diameter, the feature at 13.5 nm widens significantly from 0.8 to 1.5 nm (FWHM). These are values between those of 1-µm-driven low-density plasmas from planar SnO 2 targets of 0.5-1.5 nm [45,59] and plasmas from solid-planar tin targets of typically 2-3 nm [56]. In Ref. [32] a FWHM of 0.9 nm was reached in the case of a 1µm-driven plasma on a tin microdroplet target. 10-µm laserdriven tin plasmas typically have a narrower 13.5-nm feature, reaching 0.6 nm (FWHM) [60]. The 2-µm-driven plasma is thus seen to produce spectra with spectral widths typically in between those produced with 1-and 10-µm laser-driven pure tin plasmas.
In contrast to the scaling of CE, SP has its maximum at the smallest target diameter. The highest value shown in Fig. 2(b) is 11.5% at a 90 µm diameter down from a maximum value of ∼14% for the case of an undeformed droplet. SP for the undisturbed droplet target for this lowest intensity case is omitted from Fig. 2(b) due to the low signal-to-noise ratio in the recorded spectrum. With increasing target diameter SP is observed to decrease monotonously, an effect previously observed for 1-µm beams on droplet targets [32]. Above 200 µm diameter the decrease in SP plateaus towards a value of 7.5% at a 260 µm diameter. These SP values are consistent with Ref. [35], where the SP for planar-solid tin target plasmas was measured at 7.4% (dashed line in panel (b)).
Using these SP values as input, a monotonic increase of η rad with target diameter is observed up to a value of ∼0.3 at a 260 µm target diameter (cf. Fig. 2(c)). Also indicated is the dependence of EoT on target size, now with an amplitude fit value of 0.29 (dashed curve). The here measured maximum radiative efficiency at 60°is slightly lower than a value of approximately 0.4 measured for an extended solid-planar target [35]. This difference originates from the slightly higher inband emission at 60°measured on the planar target and may be explained by the smaller beam-spot size in the planar target case that allows for the in-band emission to escape more freely. Over the entire range of target diameters, the radiative efficiency qualitatively follows the EoT trend.
At this point, it is worthwhile to consider the fact that both overall CE and η rad qualitatively follow the EoT curve, describing the increasing geometrical overlap between an enlarging target and laser beam spot. However, we also note that the SP, which serves as input for calculating η rad , monotonically decreases with increasing target size. This apparent contradiction is resolved when considering the angular dependence of the plasma emission, with the increasing anisotropy factor serving to offset the decrease in SP with increasing target size (cf. Fig. 2(b) and 2(d)).

B. Laser intensity
Next, target diameter scans are performed for two higher laser intensities of 1.0 and 1.3 × 10 11 W/cm 2 . The CE curves for all three laser intensities show a trend very similar to the 0.6 × 10 11 W/cm 2 intensity case presented in Sec. III A (see Fig. 2(a)). For all target diameters, CE values are found to be slightly higher at higher laser intensities. The maximum CE value at large target diameters increases modestly from 2.6 to 2.9%. The angular distribution of in-band emission is observed to be independent of laser intensity within the scanned range (cf. Fig. 2(m)). Given the constant maximum CE values, the amount of in-band radiation increases linearly with increasing laser intensity, with up to 2.3 mJ of in-band light obtained for the highest intensity case (see Fig. 2(i)).
In Fig. 2(l) normalized spectra for a target diameter of 250 µm are shown. For all spectra, the 8-25 nm region looks remarkably alike. The most prominent difference between the spectra is seen in the 5-8 nm region where the amount of radiation is observed to increase with increasing laser intensity. This emission could stem from charge states Sn 14+ (and above) as well as from an increased fraction of light being emitted from electronic states having higher excitation energies. The corresponding SP values are shown in Fig. 2(b). All laser intensity cases follow the same trend. The maximum SP value for the two higher intensities is 14.5% and is reached at the smallest target diameter, i.e., the undeformed droplet target. SP steadily decreases with increasing target size, and levels off at a value of up to 8% depending on intensity. Even though the differences in SP are small, we do note that optimum SP values are observed for an intermediate intensity of 1.0 × 10 11 W/cm 2 .
Given the similar scaling of CE and SP with target diameter for the three laser intensities, the radiative efficiency in Fig. 2(c) closely follows the 0.6 × 10 11 W/cm 2 results, with a slightly higher value of 0.35 obtained for the highest intensity case.
No significant changes are thus observed in the scaling of emission characteristics with target size when changing laser intensity and in-band EUV emission is seen to exhibit a linear dependence on laser energy.

C. Pulse duration
Lastly, the laser pulse duration is varied from 3.7 to 7.4 ns (blue markers in Fig. 2) by combining signal and idler beams while delaying the signal with respect to the idler pulse. To obtain identical focal spot conditions for both beams, the beam size of the signal beam is carefully matched to that of the idler beam by means of a telescope. The energy of both beams is individually controlled via combinations of λ /2plates and TFPs. Signal and idler beams are set to have equal energies. The resulting temporal pulse profile is shown in Fig. 1(d). Measurements are taken with 35-µm-diameter droplets, slightly larger than those used during the laser intensity scan. We note that no dependence of CE on the initial droplet diameter is observed for target size scans performed for a range of droplet diameters from 19-45 µm diameter as long as the target thickness is sufficient to supply tin throughout the laser pulse. As an additional data set for the laser pulse duration we add the high-intensity (1.3 × 10 11 W/cm 2 ) scan from the previous Sec. III B that had slightly longer idler pulse duration of 4.6 ns due to slightly different settings of the MOPA system.
Again, CE is found to follow the same trend with increasing target diameter as described in Sec. III A. A slightly higher maximum CE value of 2.9% is observed for the short and intermediate pulse durations, compared with a value of 2.6% obtained for the longest pulse duration. We note that the small observed changes in CE may lie within the systematic uncertainties of the experiment. The rather small changes in CE with pulse duration, as well as laser intensity, imply a nearlinear scaling of in-band energy with laser energy. The inband energy (Fig. 2(i)) indeed almost doubles when doubling the laser pulse duration, and up to 3.8 mJ of in-band energy are measured per pulse. The angular distributions of in-band EUV emission are all similar up to an angle of 64°from which point the 7.4-ns case shows increased emission. The anisotropy factor in this 7.4-ns case (see Fig. 2(h)) roughly follows the one of the short and medium pulse duration cases up to a target diameter of 180 µm, after which it remains roughly constant at values around 0.6. This observed difference may be attributed to an increased EUV emission volume (which increases with pulse duration), partially extending beyond the liquid disk target, which enables radiating into the forward hemisphere. Another factor contributing to this difference may be a slight tilt of the disk target due to finite drift of the alignment of the laser to the droplet [61].
The recorded spectra, shown in Fig. 2(n), are remarkably similar when comparing the different laser pulse lengths. The only minor visible difference is the slightly wider main emission feature in the 7.4 ns case. SP also remains virtually independent of laser pulse duration and follows the same trend as discussed as in Sec. III A. For the longest pulse duration case, we note that the decrease in SP with increasing target size is no longer fully compensated by increases in the anisotropy factor and this is reflected in slightly lower overall CE values.
No major changes were thus observed in the scaling of emission characteristics with target size when changing laser pulse length, besides a levelling off of the anisotropy factor for the longest pulse length. In-band EUV emission is also here seen to follow linearly the input laser energy.

IV. EUV GENERATION USING 1-AND 2-µm LASER LIGHT
In a separate, third set of experiments, plasma is produced using the 1-µm laser beam at an intensity of 1.9(4) × 10 11 W/cm 2 . This choice of laser intensity yields identical emission features in the 5-12 nm region compared to the 1.0(2) × 10 11 W/cm 2 intensity at 2 µm. Equivalent spectral features ensure a similar charge state distribution of the plasma as each feature is charge-state specific. The factor of 1.9 difference in relative intensity is in agreement with the intensity ratio of 2.1(6) observed for droplet targets [31] and 2.4(7) for planar-solid targets [35].
As shown in Fig. 3(a), CE is found to increase monotonically with increasing target size in the 2-µm case until the laser beam reaches full overlap with the target, as described in Sec. III. In line with the findings in Ref. [62] CE is virtually identical for both laser wavelength cases for small target sizes. With increasing target diameter, the 2-µm CE values exceed the ones of the 1-µm case however significantly. The 1-µm CE is found to increase in a near-linear fashion until it plateaus at a value of close to 1.8% above ∼150 µm target diameter. In the case of planar-solid targets (effectively infinitely extended disk targets), very similar CE values of 1.7% were inferred from measurements performed at a representative angle of 60°for a comparable 4.8 ns pulse duration and a slightly smaller circular 66-µm-diameter (FWHM) spot [35].
In the 2-µm case, CE values measured on planar-solid targets reach up to 3.1%, only slightly higher than the here-observed values of up to 2.8%.
The angular dependence of the in-band EUV emission for the droplet target (see Fig. 3(g)) is very similar for both drive lasers and is consistent with previous studies where a 1µm laser was used to drive the LPP from a droplet target [32,41,55]. The similarities in the angular distributions is in line with a previous study by Chen et al. [63] where only minor changes in emission anisotropy were found between 1-and 10-µm-driven droplet plasmas. Similarly to Sec. III, the in-band emission decreases more strongly with increasing angle, in particular for the angles 90°, for the 250 µm target diameter than for the undeformed droplet target case (see Fig. 3(i)).
For the droplet target the anisotropy factor attains a value of 0.57, identical for both laser wavelength cases (see Fig. 3(d)). In both laser wavelength cases, the anisotropy factor increases linearly with increasing target diameter as discussed above in Sec. III B. Uncertainties related to the slightly worse fit of the anisotropy function to the 1-µm disk-target data, which unexpectedly converges to 0 at 180 • , may lead to a minor systematic overestimation of the anisotropy factor.
The in-band energy values displayed in Fig. 3(e) trivially follow the scaling of CE. In the 1-µm case, values of approximately 1.4 mJ per pulse are achieved while reaching approximately 1.7 mJ in the 2-µm case. The relative difference in the in-band energy values is slightly smaller than for the CE values because of the lower laser energy used in the 2-µm case (viz. 62 vs 80 mJ pulse energy). Despite this smaller laser energy more in-band energy is produced by the 2-µm pulses at target diameters above 160 µm demonstrating here the significant advantage of the 2x longer drive laser wavelength.
For undeformed droplet targets, the spectra for the 2-µm case show a much narrower main emission feature and, relative to the 13.5-nm peak, less out-of-band radiation is emitted in the 2-µm case (see Fig 3(f)). These observations are consistent with earlier results [31,32]. With increasing target diameter the differences in the spectral emission between the two drive laser wavelengths become even more pronounced. In the 2-µm case, the main emission feature simply broadens with target diameter (see Fig. 2(h)), in line with expectations from an associated increase in optical depth [30]. In the 1-µm case, the shape of the main emission feature changes drastically due to strong self-absorption, causing the emission maximum to shift to shorter wavelength. These absorption effects become even stronger for increased laser pulse duration, redistributing even more light into other energy channels (see Appendix A). This 'missing' emission in the main feature in the 1-µm case is qualitatively well explained by absorption of in-band radiation from a hotter plasma zone by a colder less emissive plasma zone [35,64].
SP is found to decrease monotonously with increasing target diameter for both wavelength cases (see Fig. 3(b)). In the 1-µm case, the SP for small target diameters is on the order of 10%, and is observed to decrease more rapidly than its 2-µm counterpart down to 4.5% at a 250 µm diameter. This value is consistent with the marginally lower SP value of 4% obtained from measurements on planar-solid tin targets which may be taken as the asymptotic value of SP towards infinite target size [35]. SP in the 2-µm case levels off at a higher value of about 8%, in part explaining the higher CE values observed using this drive laser wavelength.
Radiative efficiency, measured at an angle of 60°, follows a very similar trend in both wavelength cases (see panel (c)) with a larger amplitude in the 1-µm case due to the significantly lower SP values in this laser wavelength case.
No major differences between the anisotropy factors are observed. Comparing the two drive laser wavelength cases, the differences in CE are largely attributable to the decrease in SP. This link between CE and SP further supports the findings of Behnke et al. [35], where experiments on solid tin targets demonstrated that plasmas driven by 1-µm laser light exhibit strong EUV self-absorption which is absent in the 2-µm spectra. This makes 2-µm-driven plasmas the more efficient emitter of in-band EUV radiation. We note that conversion efficiencies of up to 3% can in fact be achieved for the 1-µm drive laser case for a homogeneous heating of undeformed droplet targets with several 10-ns-long, spatially flattop-shaped laser pulses [32]. Thus, on one hand, the current limitations to CE for the 1-µm drive laser case for extended disk targets may be eased using alternate target and illumination designs. On the other hand, even larger CE values for the 2-µm drive laser case may be obtained from optimally shaped targets homogeneously heated by a long, flattop laser pulse.

V. CONCLUSIONS
We have studied plasmas produced from laser pre-pulse preformed liquid tin disk targets with diameters ranging 30-300 µm using 1-and 2-µm drive laser systems. For the 2µm driver, the conversion efficiency of laser energy to EUV radiation closely follows the fraction of the laser energy absorbed by the tin target and CE values of up to 3% are obtained for the largest targets. Conversion efficiency (CE), spectral purity (SP), radiative efficiency (η rad ), and spectral emission are found to be nearly independent of laser intensity and laser pulse duration in the here-studied parameter range. Consequently, a linear increase of in-band radiation towards the backward hemisphere with laser energy is observed when increasing either parameter and further scaling of in-band output per tin target with laser intensity and pulse duration may be possible at little to no cost regarding CE.
Direct comparison of the emission characteristics of 1-and 2-µm-driven plasmas reveals significantly lower CE values for the 1-µm driver under the current experimental conditions when using extended disk targets. The lower 1-µm CE is explained by the particularly strong self-absorption of the emitted EUV radiation in the 1-µm-driven plasma. Further improvements in terms of CE may be obtainable by homogeneous heating of suitably shaped tin targets with longer laser pulses leading to full evaporation of the then truly 'masslimited' tin target. Such future studies should include research on the influence of laser pulse duration, intensity and laser wavelength on any possible fast ionic or liquid debris that may harm nearby optics elements in possible future industrial EUV light sources based on 2-µm-laser-driven tin plasmas. In addition to the comparison between 1-and 2-µm-driven plasmas for the short-pulse case in Sec. IV, more data is obtained using longer 1-µm pulses that mimic the temporal profile of the 7.4 ns case in Sec. III C. The parameters of CE, SP, η rad , E IB,2π /E IB,4π and E IB,2π show similar trends compared to the short-pulse data in Sec. III C.
The only significant change in comparison with the shortpulse data is observed in the spectral emission of the two drive-laser wavelength cases for disk-targets in Fig. 4(h). Here, the spectral emission of plasmas produced with the 1µm laser beam show even stronger self-absorption of emission at and around 13.5 nm. This stronger absorption manifests itself in slightly lower SP values. The slight deviation in the trend of η rad above 160 µm target diameter in the 2-µm case was already discussed in the main text.
In summary, good spectral performance of the plasma and high CE values are continued to be observed when increasing laser pulse duration of the 2-µm driver. In contrast, increasing the pulse duration of the 1-µm-wavelength driver, selfabsorption effects increase significantly and lead to an even larger unwanted redistributing of in-band radiation into other energy channels than in the short-pulse case.  [35]. The spectral emission of plasmas for a target diameter of (f) 35 µm (droplet target) and (h) 250 µm (disk target) are shown normalized to their respective peak intensities. The 1-µm spectrum shown in (h) is normalized to laser energy, instead of its peak emission. The at 0°normalized angular dependencies of the in-band EUV emission, corresponding to the spectra shown in (f) and (h), are shown in (g) and (i), respectively. Solid lines depict fits of Eq. (1) to the data.