Microdroplet-tin plasma sources of EUV radiation driven by solid-state-lasers (Topical Review)

Plasma produced from molten-tin microdroplets generates extreme ultraviolet light for state-of-the-art nanolithography. Currently, CO2 lasers are used to drive the plasma. In the future, solid-state mid-infrared lasers may instead be used to efficiently pump the plasma. Such laser systems have promise to be more compact, better scalable, and have higher wall-plug efficiency. In this Topical Review, we present recent findings made at the Advanced Research Center for Nanolithography (ARCNL) on using 1 and 2 µm wavelength solid-state lasers for tin target preparation and for driving hot and dense plasma. The ARCNL research ranges from advanced laser development, studies of fluid dynamic response of droplets to impact, radiation-hydrodynamics calculations of, e.g. ion ‘debris’, (EUV) spectroscopic studies of tin laser-produced-plasma as well as high-conversion efficiency operation of 2 µm wavelength driven plasma.


Introduction
Multiply excited states in multiply charged tin ions, bred in laser-produced transient plasma, are responsible for emitting narrow-band extreme ultraviolet (EUV) light at 13.5 nm wavelength [1] for nanolithography. EUV lithography * Author to whom any correspondence should be addressed.
Original Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
(EUVL) has successfully entered high-volume manufacturing. EUVL enables the continued shrinking of electronic devices as captured by Moore's law. This law drives the semiconductor industry [2,3] by postulating, or rather by demanding, that the number of transistors on an affordable CPU doubles every two years [4,5]. The production of any semiconductor device is a repetitive opto-chemical process. Photolithography is a key step in this device manufacturing [6,7]. It is a photochemical process in which a thin layer of material, a photoresist, is exposed to light. This light images onto the photoresist a so-called mask that is imprinted with the desired shapes, or 'features'. Photolithography is the defining step in setting the minimum obtainable feature size on a device. Following Abbe's law of limiting resolution, the shorter the wavelength of the light used, the better the resolution. The better the resolution, the smaller the features that can be produced. EUVL, at 13.5 nm wavelength, can thus be seen as the logical, if ambitious, next step from the 193 nm light that is still used for many lithography applications. The production of the required EUV light at sufficient power, within a narrow 'in-band' 2% wavelength bandwidth centered at 13.5 nm that can be reflected by multilayer optics [8,9], continues to present both industry and science with physics challenges, in particular in the optics and photonics fields.
Following our 2019 review [11], there are several crucial requirements for EUV light sources. Such sources should have high conversion efficiency (CE). Here, CE is defined as the ratio of in-band EUV light emitted into a half-sphere backwards towards the laser (this half-sphere being covered by the collector mirror, see figure 1) over the laser energy used to obtain it. From the perspective of sustainability, the total efficiency of converting wall-plug electricity to useful EUV light is another relevant performance indicator. Next, the amount of optics-lifetime-limiting-debris that the plasma produces (typically consisting of fragments or high-energy particles) should remain manageable. Historically, there were three promising candidate elements: Li, Sn and Xe, all of which have ions with strong electronic resonance transitions within the required bandwidth. However, for various reasons, the reported CE of Li-and Xe-based plasma sources are much lower than that of Sn-based plasmas [12,13]. Current state-of-the-art sources of EUV light are based on the irradiation of so-called mass-limited, micrometer-sized droplets of molten tin by high-energy CO 2 -laser pulses of 10.6 µm wavelength at several 10-kHz repetition rates, which creates an EUV-emitting laser-produced plasma (LPP) in several steps, see figure 1.
First, a low-energy pre-pulse deforms a spherical tin droplet into a target shape optimized for interacting with a high-energy main pulse. This deformation process involves a plethora of physical processes, ranging from laser-matter interaction, to plasma radiation-hydrodynamics, to pure fluid mechanicswith a clear separation of the relevant time scales. The ordering of these time scales determines to a large degree the overall target morphology [10,14]. On one extreme end, short fsps laser pulses create bubble-like targets (see figure 1 bottom panel) as described by, e.g. Kurilovich et al [15] (also see [16][17][18][19][20]) and more recently by De Faria Pinto et al [21] addressing also laser polarization dependencies. On the other extreme end, long (ns) laser pulses create strongly propelled, flattened targets (see figure 1 upper panel). The industry has opted for using relatively long (∼10 ns) laser pulses, creating a quasi-stationary, strongly radiating plasma ablation front that leads to the propulsion [22][23][24] and the required deformation of the droplet [22,23,25,26] into a 'pizza' target shape (see section 2), i.e. a relatively flat thin disk bounded by a thick rim [27,28]. Subsequent pulses of even lower energy may be used to rarefy a tin target in order to, e.g. make a preheated plasma [29] that is later to be laser-reheated [13,30,31]. Vaporization of the several 10 nm thick targets [27] can also be employed to obtain further information about the thickness profile itself. Schematic illustration of droplet-based EUV source operation using a pre-pulse (PP) plus main-pulse (MP) irradiation scheme including actual shadowgraph images. The top panel shows the typical deformation occurring after impact of a ns-duration pre-pulse and the lower panel the expansion following impact of a ps-duration pre-pulse. In both cases a typical desired target size is attained after several µs, at which point the target is irradiated by a more energetic main pulse, generating plasma that emits 13.5 nm EUV radiation. The EUV light is collected by the multilayer 'collector' mirror and focused at the intermediate focus (IF), the entry point of the illuminator part of an industrial wafer stepper machine. Reproduced with permission from [10].
Next, a high-energy laser main pulse drives the dense plasma from the target as prepared by the preceding prepulse(s). In industrial sources of EUV light the drive laser light is obtained from high-power CO 2 -gas lasers operating at a wavelength of 10.6 µm [32]. Solid-state laser systems, typically operating at a shorter near-infrared wavelength could potentially provide an attractive alternative [33,34] to the gas laser technology for driving the plasma (see section 3) with the promise of higher wall-plug efficiency and reduced complexity. Understanding the origins of the plasma-emitted EUV light, and thus the complex atomic structures of multiply charged tin ions (section 3.1) is crucial to be able to optimize future sources through predictive modeling [35]. Given the size of the outstanding challenges and the impact of new developments on society, it should not come as a surprise that modeling efforts have been stepped up worldwide. The Code Comparison session in the yearly EUV Source Workshop, organized jointly by EUV Litho, Inc. [36] and Advanced Research Center for Nanolithography (ARCNL) since 2019, has the ambition to bring together plasma modelers to improve simulation capabilities of EUV sources [35]. Besides the in-band EUV, there is interest in out-of-band (OOB) radiation in the EUV and also in the deep-ultraviolet (DUV) range. This OOB radiation may be detrimental for imaging contrast [37,38], but can also provide a unique window into the radiative heart of the plasma motivating detailed spectroscopy efforts [39][40][41]. Predictive modeling should also take the expansion of the hot and dense plasma into account: ions may reach typical kinetic energies of several keV (section 3.2). The impingement of such energetic particles on nearby optical elements may be detrimental for their performance. As such, plasma expansion and the generation (and stopping) of energetic particles is a subject of particular interest [42][43][44][45][46][47]. Producing the currently required 250 W of EUV light at the so-called intermediate focus (IF), where the source is connected to the scanner, has been a daunting, and continuing challenge. The semiconductor roadmap now asks for a stable 1000 Watt (at IF) source, quadrupling the present source performance. This calls for game-changing innovation driven by research at the fundamental level.
This topical review builds on a review from 2019 on the topic by Versolato et al [11], itself building on the earlier work of Banine et al [43] of ASML and indeed on the work of many others, e.g. of the University College Dublin (UCD) Spectroscopy Group, Lawrence Livermore National Lab (LLNL), Los Alamos National Laboratory (LANL), the Center for Materials Under eXtreme Environments (CMUXE) at Purdue University, the Institute of Spectroscopy of the Russian Academy of Sciences (ISAN), the Keldysh Institute of Applied Mathematics (KIAM), the Laboratory for Energy Conversion at ETH Zürich, the EUV Photonics Laboratory of the University of Central Florida, the Institute of Laser Engineering of Osaka University, Tokyo Metropolitan University and Gigaphoton (and still many others). This review will combine an overview of recent, relevant literature with discussions of key processes that govern the dynamics in each step in the process of generating EUV light, while focusing on recent results from the Source Department of the ARCNL in Amsterdam. ARCNL is tasked to focus on the fundamental physics and chemistry behind current and future technology for nanolithography, especially for application in the semiconductor industry.

Laser-droplet interaction for tin target preparation
Current high-power industrial EUV sources are based on the pulsed irradiation of a high-frequency stream of micrometersized droplets of molten tin. The reason for using individual, well-separated droplets is that each droplet serves as an isolated reservoir of tin with limited mass. Such mass-limited targets enable careful optimization of CE while using a minimum amount of tin and, moreover, generate a minimal amount of debris. The careful shaping of a spherical droplet into a shape optimally suited for interaction with the energetic main pulse is key for such careful optimization and control.
The transformation of a spherical tin droplet into a suitable target using laser 'pre-pulses' involves a broad range of physical processes. Of key importance for the final morphology of the target is the ordering of the time scales associated with the various physical processes. Among the shortest timescales is the electron-ion relaxation time τ e−i ∼ 10 ps, which is related to the exchange of energy between electron and ion subsystems, with the electron subsystem directly heated by the laser pulse through the process of inverse brehmsstrahlung. The next longest time scale is the plasma-hydrodynamic time scale, which is set by the ratio of the plasma flow length scale distance to the velocity of this flow. This flow length scale, i.e. the distance between droplet and critical surface (where the plasma electron density equals the critical density for the incoming laser beam) is of order ∼10 µm and the typical flow velocity is given by the speed of sound in the laser produced plasma corona ∼10 5 m s −1 [48]. Together they yield a hydrodynamic time scale τ h ∼ 100 ps [48]. Next is the acoustic time scale of the tin liquid τ a on which pressure waves, launched by the plasma produced by the laser pulse, acoustically travel (with sound speed c ∼ 2500 m s −1 ) through the droplet (with radius R 0 ∼ 25 µm), τ a ∼ R 0 /c ∼ 10 ns. This time scale is followed by the inertial time scale τ i = R 0 /U ∼ 100-1000 ns on which the droplet relevantly deforms. Following the practice well-established in the fluid dynamics literature, the typical deformation velocity is here taken to be the propulsion, or impact velocity U ∼ 100 m s −1 . This impact velocity should be compared to the more relevant velocitẏ R 0 , the initial radial expansion rate. TypicallyṘ 0 ∼ U, however the precise relation depends on, e.g. how tightly the laser is focused onto the droplet. More tightly focused beams lead to largerṘ 0 (andṘ 0 > U). Hernandez-Rueda et al [49] have explored a large parameter space to investigate the influence of the tin droplet diameter as well as the laser beam diameter and energy on theṘ 0 /U ratio. Next, they compared the experimental U andṘ 0 values to those obtained with detailed radiation-hydrodynamic simulations using the RALEF-2D code which in turn enabled validating analytical fluid-dynamics modeling [26].
The near-equivalenceṘ 0 ∼ U may break down if the laser pulse is of particularly short duration (Ṙ 0 ≫ U with U ≈ 0, cf figure 1).
Last in line is the capillary time scale τ c = √ ρR 3 0 /σ ∼ 10 µs, with liquid density ρ and surface tension σ [26]. This is the time scale on which retraction of the sheet due to surface tension occurs. Theoretical work by Reijers et al [14], recently experimentally validated by Meijer et al [10], made clear that the time scale of significant changes in laser intensity (typically of the order of the laser pulse duration τ p ) over the time scale ordering τ e−i < τ h < τ a < τ i < τ c is a crucial factor determining the fluid-dynamic deformation of the droplet and final target morphology.
From the whole palette of pre-pulse settings, the EUVL industry hamicros opted for using relatively long (τ p ∼ 10 ns) laser pulses which, given that τ p ≫ τ h [48], create a quasistationary plasma that leads to the propulsion and the required deformation of the droplet into a relatively flat, thin disk bounded by a thick rim. This 'pizza' shape was uncovered and experimentally mapped in detail by Liu et al [28], who studied the morphology of a radially expanding sheet of liquid tin formed by ns-pulse Nd:YAG laser impact on a spherical microdroplet. Specifically, the sheet thickness profile and its time evolution were captured over a range of laser-pulse energies for two droplet sizes. Two complementary methods were employed to determine the thickness profile. In the first method, the finite transmissivity of the several 10 nm thick stretched liquid-metal sheet was recorded and, with the known optical constants of tin [50], the transmissivity could be converted to a local thickness h(r) as a function of the radial coordinate r (see figures 2(a1.2) and (a2.2)). In the second 3)) Corresponding side-view images (laser impacts from the left, propelling the droplet to the right). In ((a1.1), (a1.3)), the cylindrical coordinate system (r, θ, z) with its origin at the center of sheet is depicted. At later times, holes appear cf ((a3), (b1), (b2)). The overlay of a hole is highlighted by a white box in both ((b1), (b2)). The arrows in inset (b3) indicate the receding edge of the hole the velocity of which gives the local thickness (see main text). (c) Sheet thickness as a function of the radial coordinate. Comparison of results obtained using a backlight-transmission method (cf panel (a2.2)) and from a complementary method using hole-opening velocities (cf panels (b1) and (b2)). Reproduced from [28]. CC BY 4.0. Volume ratio of the sheet to that of the initial droplet V sheet /V 0 as a function of non-dimensional time t/τc for various droplet sizes and Weber numbers (see [28]). Brown data (circles) are obtained from [28] using the method illustrated in figure 2. Dark gray data (squares) are obtained from a laser vaporization method [27]. The analytical prediction for V sheet /V 0 is presented as a black solid line. The inferred volume ratios V rim /V 0 and V fragment /V 0 are presented as blue and red lines, respectively, following [28]. The vertical line marks t/τc ≈ 0.55 from where the model validity is unclear. Figure data taken from [27,28]. method, the speed of opening of spontaneously formed holes (cf figures 2(b1)-(b3)) in the stretching sheet is experimentally measured, which again allows one to obtain h(r) following Culick [51]. Results from the two methods, indicating the presence of a thin tin sheet just several 10 nm in thickness, were shown to be in excellent agreement cf figure 2(c). Moreover, all obtained thickness profiles were shown to collapse onto a single self-similar curve enabling the prediction of the thickness profile under a wide range of experimental conditions. Spatial integration of the thickness profiles enables one to determine the volume of the sheet as a function of time after pre-pulse laser impact, see figure 3. Remarkably, less than half of the initial amount of tin remains in the sheet under these conditions. Further analysis shows that the largest fraction of the mass lost from the sheet during its expansion ends up as fine fragments (see figure 3). Liu et al [28] proposed that such mass loss can be minimized by producing the sheet targets on the shortest possible timescales. A followup study validated this proposal, where Liu et al [27] irradiated thin tin sheets with a low-intensity laser pulse. This auxiliary pulse, used as a probe with an intensity below plasma threshold (following Meijer et al [52]), induced vaporization which enabled investigation of the thickness profile of the sheet and its mass also at earlier times (see figure 3). The results demonstrated that increasing the energy of the Nd:YAG laser pulse, which enabled reaching the predetermined target radius more quickly, resulted in a larger mass fraction remaining in the sheet. As a corollary, less tin ended up in other channels of the mass distribution such as fragments surrounding the sheet, thus leaving more mass in the target sheet available for interaction with the more energetic main laser pulse to produce EUV light.

Main pulse: using solid-state lasers to drive EUV emission
Once a suitable target shape is reached, the main laser pulse transforms the liquid target into a hot and dense EUV emitting plasma [11,73]. In the industry, the laser providing these pulses currently is an infrared 10.6 µm wavelength CO 2 gas laser. Tin plasma driven at this laser wavelength has particularly high CE (see figure 4). Solid-state lasers, typically operating at shorter, near-infrared wavelengths may soon be able to provide a viable alternative to these CO 2 gas lasers [74]. Such solid-state laser systems may be more compact and have higher wall-plug efficiencies. Reducing the drive laser wavelength from 10.6 µm will, however, also increase the typical tin ion densities in the region where EUV light is generated in the plasma. This effect can be qualitatively understood [11] from the increase in critical electron density n c with decreasing drive laser wavelength λ through n c ∼ λ −2 . Beyond n c , where the electron plasma frequency is equal to the laser light frequency, laser light cannot propagate. Relevant plasma densities increase at a slower pace ∼λ −1 [72] (see below), but nevertheless the plasma gets more dense with decreasing drive laser wavelength. Increased plasma density is associated with larger optical depths (the product of atomic opacity, mass density, and path length [72,75]), which causes opacity broadening of the 13.5 nm emission feature beyond the 2% acceptance bandwidth, thus reducing the spectral purity (SP, defined as the ratio of spectral energy in a 2% bandwidth around 13.5 nm to the total EUV energy) and with it the maximally obtainable CE [55,61,63,65,71,[75][76][77].
Nishihara et al [12] provided simulation results for optimal drive laser intensities, with predictions for obtainable CE for idealized '0D' plasmas. In figure 4(a), CE values from the simulations are indeed shown to increase with increasing drive laser wavelength. The optimum laser intensity decreases with increasing laser wavelength (and, thus, with decreasing plasma density). Several recent simulation efforts (see, e.g. [33]) have drawn the attention of the EUV source community to the use of a 2 µm drive laser, further supported by the recent introduction of novel concepts for high-power solid-state laser systems operating at this wavelength. Such promising simulations are particularly challenging, not least because of the complex atomic physics involved, and require experimental benchmarks.
Schupp et al [72,78] and Behnke et al [34] recently reported the first experimental study of 2 µm laser-driven tin plasmas. These studies, discussed in section 3.3, confirmed the simulation results with regards to the particular promise of the 2 µm driver. Additional experiments are required in unison with detailed predictive modeling efforts. The modeling efforts need as input accurate atomic physics data to enable understanding the origins of the plasma-emitted EUV light. These origins, discussed in section 3.1, were recently shown to be much more complex than previously thought. Predictive modeling should also take into account the expansion of the hot and dense laser-driven plasma, where ions reach kinetic energies on the order of several keV. This topic will be discussed further in section 3.2. The development of advanced  [53][54][55][56][57][58][59][60][61][62][63][64][65][66][67][68][69][70][71][72]). Overlapping data points are shifted vertically for visibility. Simulation results [12] for plasma under optimal conditions are shown (open black triangles) connected by straight lines. The vertical lines, moving from right to left, indicate the wavelengths of CO 2 laser light, Nd:YAG laser light (fundamental), as well as the second and fourth harmonics of Nd:YAG laser light. The results of the work of Behnke et al [34] are shown as open red squares (see main text). Adapted with permission from [34]. © The optical Society. solid-state laser systems at ARCNL will enable dedicated studies to find the true optimum conditions to drive tomorrow's plasma sources of EUV light, as we conclude in section 4.

Strong contributions of multiply excited states to EUV emission
The complex electronic structures of multiply charged tin (Sn) ions are the root cause of their particular attractiveness for use in next-generation nanolithography [3,11,32,43,70,79,80]. They are employed as emitters of in-band 13.5 nm EUV photons. The suitability of Sn ions for this application stems from their open-4d-subshell structures [39,[81][82][83][84][85][86][87][88][89][90]. Within these structures, ∆n = 0 one-electron-excited configurations are well-known to decay to the ground state manifold via a multitude of transitions clustered together in unresolved transition arrays (UTAs) [91] centered around 13.5 nm. An exceptional feature of Sn ions is the fact that the average excitation energies of these configurations are similar across the isonuclear sequence Sn 11+ -Sn 14+ , making these charge states excellent radiators of 13.5 nm light. Recently, however, a team of researchers from ARCNL and Los Alamos National Laboratory (LANL) found that the characteristics of Sn ions are even more special and that the EUV light generated in these plasmas originates predominantly from transitions from multiply excited states [1]. Contrary to the prevailing view, contributions from one-electron-excited states are not the prime sources of EUV light in the in-band spectral region. This serendipitous alignment of transitions in singly, doubly, and triply excited systems occurs over a range of charge states Sn 11+ -Sn 14+ . The work [1] revealed the doubly magic behavior of tin and the origins of the EUV light.
The importance of the multiply excited states for the EUV emission is demonstrated in figure 5, where a comparison between an experimental emission spectrum (black curve, produced from plasma generated by impinging a Nd:YAG laser pulse onto a tin microdroplet) and the spectral flux obtained from one-dimensional radiation-transport modeling (red curve, which uses as input the ATOMIC opacity calculations) is shown [1]. The spectral flux calculated using a singledensity, single-temperature approach clearly reproduces the measured emission strikingly well. Without the contributions from the multiply excited states, it would clearly not be possible to explain the experimental spectrum to any degree of satisfaction. To further highlight the importance of these transitions, as well as the accuracy of the work of Torretti et al [1], the results are compared with calculations from previous works [76,89].
In follow-up work by Sheil et al [92], the Los Alamos ATOMIC code was used to investigate the spectral contribution from transitions from multiply excited states in CO 2 laserdriven (λ = 10.6 µm) tin plasma conditions. Here, in comparison the Nd:YAG drive laser case, much lower plasma densities are obtained where local-thermodynamic equilibrium (LTE) conditions are not met. Busquet's [93] ionisation temperature method was employed to match the average charge state of a non-local-thermodynamic equilibrium (non-LTE) plasma with an LTE one, establishing a so-called ionization temperature T Z . This approach is found to generate LTE-computed configuration populations in excellent agreement with the non-LTE populations. A corollary of this observation is that the non-LTE populations are well-described by Boltzmann-like exponential distributions characterized by the effective temperature T Z . Subsequently, extensive level-resolved LTE opacity calculations were performed at T Z . Sheil et al [92] also found that the leading contributions to the opacity near 13.5 nm arise from transitions from multiply excited states. These results reinforce the need to include multiply excited states in atomic models from which the radiative properties of laserdriven tin plasmas are generated. The work of Sheil et al [92] paves the way for the generation of detailed LTE opacity tables at non-LTE plasma conditions, which can be incorporated in radiation-hydrodynamic simulations of laser-driven tin plasmas. Such simulations will enable reliable predictions of EUV emission from LPPs and will play a key role in guiding experimental efforts in the characterization and optimization of laserdriven plasma sources of EUV light.
Aside from the in-band EUV radiation, spectroscopy in other wavelength ranges can provide further valuable insight into the properties of the laser-driven tin plasma. Indeed, detailed information about the contribution of the various charge states to the main 13.5 nm emission feature can be obtained from OOB transitions. Short-wavelength emission, located between 7 and 12 nm, enables the assessment of the charge state distribution [39,40] and served to diagnose industrial CO 2 -laser-driven plasmas as demonstrated by Torretti et al [41]. Using the method developed by Scheers et al [94] for charge-state-resolved spectroscopy, Bouza et al [95] compared laser-driven plasma emission with electron beam ion trap spectroscopies at longer EUV wavelengths. This plasma spectroscopy work was later extended to include DUV and UV wavelengths, where an intensity-calibrated spectrum from 5 all the way up to 265 nm wavelength [96] was recorded using a novel transmission grating spectrometer (TGS) developed by the MESA+ XUV Optics group at Twente University, in tandem with a smart choice of filters. The TGS will soon be upgraded with one-dimensional imaging capabilities, enabling space-resolved characterization of plasma ionicity (see Byers et al [97]). Scheers et al investigated the UV and optical emission from the lower charge states in the plasma [98]. This work was followed by a temporally and spatially resolved study of the optical emission from these lower charge states, where the evolution of electron density and temperature in the 'afterglow' of the LPP was quantified using a fiber-coupled imaging spectrometer [99]. Further spectroscopic studies will undoubtedly drive predictive modeling efforts and the development of future EUV light sources.

Plasma expansion-'fast ionic' debris
Besides the sought-after EUV light, a laser-produced plasma produces debris in the form of fast ions, the impact of which Figure 6. Schematic view of the crossed-beam setup that will be used to measure charge exchange and energy loss in the forward-scattering direction by means of a combined electrostatic Time-of-Flight analyzer system, while target fragments from the interactions of Sn q+ ions colliding with H 2 are extracted sideways into a Time-of-Flight spectrometer. The full setup can be operated on high voltage to decelerate the incoming ion beams allowing for a coverage of the full kinetic energy range of plasma ions from close to 0 to tens of keV. may limit the lifetime of the light collecting multilayer optics. Rai et al [100] studied single-collision scattering of keVenergy ions off surfaces to elucidate the absence of a singlescattering peak in the here relevant Sn-Ru collisions [101] and to test the predictive power of standard Monte Carlo binary collision codes such as SRIM (Stopping and Range of Ions in Matter) [102]. These codes, which were primarily developed for modeling swift particle interactions with solid state targets, are also used to simulate the stopping and mitigation of plasma ions in hydrogen buffer gas surrounding the plasma. Corresponding experimental data on slow (E ⩽ 100 eV amu −1 ) heavy ions is by and large lacking. To investigate the interactions between Sn q+ ions with H 2 we have commissioned an advanced crossed-beam type setup based on experience from previous crossed-beam experiments [103,104], see figure 6.
Fast plasma ions are generated during the expansion phase of the plasma [105]. Understanding such plasma expansion at the fundamental level is key for modeling the laser-produced plasma. Hemminga et al [106] presented the results of a joint experimental and theoretical study of plasma expansion driven by Nd:YAG laser ablation (laser wavelength λ = 1.064 µm) of tin microdroplets under conditions relevant for nanolithography. The experimental measurements indicate a nearplateau in the ion energy distribution for kinetic energies in the 0.03-1 keV range, a peak near 2 keV followed by a sharp fall-off in the distribution for energies above 2 keV (red curve in figure 7). Charge-state resolved measurements, performed with a cross-calibrated electrostatic analyzer in time-of-flight mode (ESA-ToF) [107], attributed this peaked feature at 2 keV to the existence of peaks (centered near 2 keV) in the Sn 3+ − Sn 8+ ion energy distributions. To understand the physical origin of this peaked feature, Hemminga et al [106] performed two-dimensional simulations of the plasma initiation, growth and subsequent expansion using the radiation hydrodynamic code RALEF-2D. As is evident from figure 7, excellent agreement was found between the simulated ion energy distribution and the measurements both in terms of the shape of the distribution and the absolute number of detected ions. The peak in the ion energy distribution near 2 keV was attributed to a The distribution of the number of ions over ion kinetic energy. The experimental ion energy distribution is shown in red (solid curve) and the RALEF-2D ion energy distribution is shown in black (solid curve). Also illustrated in green are the predictions of analytic models of plasma expansion (see discussion in [106]). Reproduced from [106]. © The Author(s). Published by IOP Publishing Ltd. CC BY 4.0.
quasi-spherical expanding shell formed at early times in the expansion.
These results demonstrate that the single-fluid singletemperature approach implemented in RALEF-2D can not only reproduce the general shape of the experimental ion energy distribution, but it can also provide a reliable prediction for the absolute number of ions, thus paving the way for future predictive modeling of plasma EUV light sources.

Comparing plasmas driven by 1 and 2 µm wavelength lasers
Recent simulation efforts at Lawrence Livermore National Laboratory (LLNL, and see, e.g. [33]) have indicated that use of a drive laser at ∼2 µm wavelength may be particularly beneficial for driving the tin plasma. Concurrently, LLNL introduced the concept of high-energy, high-power Big Aperture Thulium (BAT) laser systems operating at 1.9 µm wavelength.
At ARCNL, experiments were started to investigate the potential of such novel lasers and novel wavelengths to drive plasma. Behnke et al [34] were the first to present an experimental study of a 2 µm laser-driven tin plasma in which a planar tin target was irradiated with laser pulses of 5 ns duration. The 2 µm laser light source (cf figure 8(a) inset) used comprised a potassium titanyl phosphate (KTP) based master oscillator power amplifier (MOPA) operated in type-2 phase matching. The MOPA is pumped at a 10 Hz repetition rate by a seeded, Q-switched Nd:YAG laser providing pulses of 10 ns duration. First, a 2170 nm idler seed beam is created in a singly-resonant optical parametric oscillator (OPO). To create the seed beam, a fraction of the pump light is demagnified to a beam diameter of 1.5 mm and is coupled into the OPO, operated in a collinear alignment. About 20% of the pump radiation is converted into a 2090 nm wavelength signal beam and a 2170 nm idler beam. A dichroic mirror separates the signal and idler beams from the partially depleted pump. The idler beam is subsequently expanded to 11 mm in diameter to seed the optical parametric amplifier (OPA), while the signal beam is removed through polarization optics. The OPO and OPA are pumped by the same laser. For pumping the OPA, 1.3 J of the Nd:YAG laser light is delayed by 1.3 ns and is reduced to a beam size of 10 mm in diameter.
Seed and pump beams are overlapped on a dichroic mirror after which they pass several 18 mm long KTP crystals. The crystal orientation is alternated to compensate for walk-off. A total (sum + idler) energy of several 100 mJ can routinely be achieved. Pump and signal beams are separated from the idler using a dichroic mirror and polarization optics, respectively. To adjust the energy of the idler beam, a waveplate/ polarizer combination is employed before focusing the beam onto the target. The size of the focal spot was approximately 100 × 100 µm 2 . Spectroscopic investigations were performed for plasmas driven by this 2 µm wavelength pulsed laser light and comparisons were made with plasmas driven by the 1 µm pump laser light at several laser intensities. Very similar EUV spectra, and thus underlying plasma ionicities, were obtained when the intensity ratio was kept fixed at I 1 µm /I 2 µm = 2.4 (7). Crucially, the CE was found to be a factor of two larger (at the given 60 degree observation angle) for the 2 µm laserdriven plasma compared to the case of the denser, 1 µm driven plasma.
Following soon after, Schupp et al [72] presented experimental spectroscopic studies of EUV light emitted from plasma produced by the irradiation of tin microdroplets with 5 ns pulsed, 2 µm wavelength laser light (cf figure 8) using the same laser system as in [34]. Emission spectra were compared to those obtained from plasma driven by 1 µm wavelength Nd:YAG laser light over a range of laser intensities. Over the studied range of drive laser intensities, it was found that similar spectra (and thus underlying plasma charge state distributions) were obtained when the ratio of the 1-2 µm laser intensities was fixed at a constant value, in good agreement with the findings of Behnke et al [34]. Schupp et al [72] also performed laser-plasma simulations using the radiationhydrodynamic code RALEF-2D and found good agreement regarding the intensity ratio. Their experimental findings, supported by the simulations, indicate an approximately inversely proportional scaling ∼ λ −1 of the relevant plasma electron density and the aforementioned required drive laser intensities with drive laser wavelength λ in line with the predictions of Nishihara et al [12] (cf figure 4) for a plasma system of much reduced complexity. The ∼λ −1 scaling was also found to extend to the optical depth as captured in the observed changes in spectra over a range of droplet diameters. The decrease of optical depth with increasing drive laser wavelength is illustrated in figure 9, where experimental spectra from 1, 2, and 10 µm drivers are seen to decrease in width with increasing λ.
In a detailed follow-up work, Schupp et al [78] reported on the EUV emission properties of tin plasmas produced by the irradiation of pre-pulse-preformed liquid tin targets by 2 µm wavelength laser pulses (cf figure 8). In a two-pulse scheme, much like in the current industrial setting, a pre-pulse laser is first used to deform tin microdroplets into thin, extended disks (cf figure 8(c)) before the main (2 µm) pulse creates the EUV-emitting plasma. The effects of a change in 2 µm drive laser intensity and laser pulse duration (3.7-7.4 ns; for the longer pulses, both signal and idler were used) were studied. It was found that the angular dependence of the emission of light within a 2% bandwidth around 13.5 nm and within Figure 9. EUV spectrum produced with a 2 µm laser-driven tin plasma (red line) compared to that obtained from a 1 µm laser-driven plasma (blue line) at intensities of 1.1 and 2.4 × 10 11 W cm −2 , respectively. Also shown is a spectrum obtained using a 10 µm CO 2 laser that here represents the case of small optical depth (reproduced from [108] in [72]). Reproduced from [72]. CC BY 4.0.  [34]. All SP values provided are calculated with respect to the measured spectral range of 5.5-25.5 nm. Reproduced from [78]. © The Author(s). Published by IOP Publishing Ltd. CC BY 4.0. the backward 2π hemisphere around the incoming laser beam is almost independent of intensity and duration of the 2 µm 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 backward 2π hemisphere around the incoming laser beam. For direct comparison, a similar set of experiments was performed with a 1 µm wavelength drive laser. EUV emission spectra were found to exhibit prominent self-absorption of light around 13.5 nm in the 1 µm case, while in the 2 µm case only a modest opacity-related broadening of the spectral feature at 13.5 nm was observed. The work demonstrated the enhanced capabilities and performance of 2 µm driven plasmas produced from disk targets when compared to 1 µm driven plasmas both in terms of CE (cf figure 10(a)) and SP (cf figure 10(b)). The significant improvement of the spectral performance of the 2 µm vs 1 µm driven plasma provides strong motivation for the development of high-power, high-energy nearinfrared lasers to enable the development of more efficient and powerful sources of EUV light. Such solid-state lasers may soon become a viable alternative given the fact that solid state lasers are more compact, are more flexible, and are expected to be more stable than presently used CO 2 -gas lasers. Moreover, solid-state lasers are expected to have a significantly higher efficiency in converting electrical power to laser light, thus enabling obtaining a higher overall efficiency converting wallplug electrical power to useful in-band EUV radiation. The optimum laser wavelength of such a solid-state drive laser, operating under realistic conditions, is however still unclear. At ARCNL, we are developing advanced high-energy, lowrepetition rate laser systems [109], flexible in both wavelength (1-4 µm) and spatiotemporal beam profile. These laser systems will be employed to drive microdroplet tin plasma to produce EUV light under industrially relevant conditions, facilitating the design of the ultimate laser-driven plasma source of EUV light to drive tomorrow's nanolithography.

Conclusions
In this topical review in the JOPT special issue on Advances in Optics in The Netherlands, key physics aspects of laserdriven tin plasmas are discussed. Such plasmas are the source of extreme ultraviolet light at 13.5 nm wavelength for stateof-the-art nanolithography. Generating ever-more EUV light at ever-increasing efficiencies and machine up-times provides a challenge to both science and industry. This is especially true in the context of a 1000 W EUV source, which now lies on the horizon of the semiconductor roadmap (such a source would quadruple the present source performance). In this review, we combine an overview of selected literature, focusing on recent results from the Advanced Research Center for Nanolithography (ARCNL) in Amsterdam. ARCNL is tasked to focus on the fundamental physics and chemistry underpinning current and future technologies for nanolithography. Progress towards, and beyond, a 1000 W EUV source will continue to be supported by achievements in the fields of optics and photonics field by combining fundamental research with industrial innovation.

Data availability statement
No new data were created or analysed in this study.