Deep ultraviolet (DUV) light sources, particularly those operating at the ArF excimer laser wavelength of 193 nm, have become indispensable for high-resolution semiconductor lithography.1^–4 For advanced interference lithography, seeding an ArF excimer amplifier with a narrow-linewidth solid-state laser becomes a considerable solution to obtaining high-power output and high spatial coherence simultaneously, promising to achieve a higher coherence and higher machining accuracy. Tanaka et al. demonstrated such a hybrid ArF excimer laser capable of generating a 193-nm laser of more than 100 W average power with high coherence.5 Compared with traditional ArF excimer lasers, the seeded beam exhibited a Gaussian intensity distribution and superior beam quality, suggesting enhanced performance in various applications. Moreover, a compact and variable 193-nm solid-state system is needed not only for seeding ArF excimer lasers due to the limitation of the space but also for the advanced mask inspection and metrology, which should be fully integrated into standard commercial ArF excimer lasers for interference lithography.5

Solid-state 193-nm lasers can be generated through harmonic generation and frequency mixing of visible or near-infrared lasers.6^–8${\mathrm{KBe}}_2{\mathrm{BO}}_3{\mathrm{F}}_2$ (KBBF) crystal is so far the only nonlinear crystal that can generate coherent light below 200 nm with more than 1 W average power output by direct second-harmonic generation (SHG). A watt-level picosecond laser at 193.5 nm by frequency quadrupling from a Ti:sapphire laser system operating at 774 nm with a ${\mathrm{LiB}}_3{\mathrm{O}}_5$ (LBO) crystal and a KBBF crystal was demonstrated by Kanai et al.9 However, scaling Ti:sapphire systems to more than 16 W average power remains challenging due to the complexity of regenerative and multipass amplifiers, as well as the need for high-power green laser pumps. In addition, the KBBF crystal is still expensive and requires a specialized prism coupling device due to its plate-like form.

Frequency mixing offers a more promising path to generate high-power 193-nm lasers. Kawai et al. demonstrated a 140-mW 193-nm laser source by frequency mixing the seventh-harmonic signal and the fundamental laser of a nanosecond pulsed laser at 1547 nm using ${\mathrm{CsLiB}}_6{\mathrm{O}}_{10}$ (CLBO) crystals.10 Benefiting from the reuse of the residual pump laser, the conversion efficiency from the fundamental laser to the 193-nm laser reached 4.5%. Xuan et al. reported a two-stage sum-frequency generation (SFG) approach in CLBO crystals using a 258-nm laser and a 1553-nm erbium-doped fiber laser as pumps.11 The configuration provides more than 1 W at 193 nm, which is the highest average power generated by SFG so far. They later developed a compact collinear cascaded SFG configuration to reduce cost, complexity, and instability.12

Although frequency mixing with DUV lasers at $258/266\mathrm{nm}$ with the power of tens of watts has been successfully demonstrated,13^–15 narrow-linewidth erbium-doped fiber amplifiers at $1.5\mu \mathrm{m}$ still face challenges in scaling to high power due to amplified spontaneous emission and nonlinear effects, such as stimulated Brillouin scattering and self-phase modulation.16^,17 The erbium concentration limitation lowers pump-to-signal conversion efficiency and requires longer gain fiber and higher pump power, which increases the complexity of the system and cost. Optical parametric amplifiers (OPAs) based on periodically poled lithium niobate (PPLN) crystals offer a more scalable alternative.18^,19 PPLN’s high nonlinear coefficient and conversion efficiency enable efficient amplification from the milliwatt to watt level using only one or two stages, simplifying system complexity. Unlike the erbium-doped fiber amplifier, the OPA process does not involve a carrier lifetime of rare ions, which makes it more suitable to scale up a pulsed laser with a low-duty cycle, getting rid of generating amplified spontaneous emission between pulses. In addition, PPLN-based OPAs offer flexibility in wavelength tuning, allowing for easy adaptation to different operating conditions by adjusting the seed laser wavelength or crystal temperature.

In this paper, we present an experimental demonstration of a 193-nm pulsed laser generation system utilizing a 1553-nm OPA-based laser and a 258-nm laser, along with the generation of vortex beams. A high-power 1030-nm pulsed laser from a Yb:YAG bulk crystal amplifier is divided into two parts: one is used to pump two stages of OPA to produce a 1553-nm laser with an average output power of 700 mW, and the other is used to generate a 1.2-W 258-nm laser by fourth-harmonic generation (FHG). The 1553 and 258 nm lasers are then injected as the pump sources into two stages of SFG using LBO crystals, successively generating 221 and 193 nm lasers. Pulsed lasers with an average power of 270 mW at 221 nm and an average power of 70 mW at 193 nm are obtained, respectively. The 193-nm laser exhibited a linewidth of less than 880 MHz, resulting in a full width at half-maximum (FWHM) of less than 0.11 pm. To explore novel applications, we introduced a spiral phase plate (SPP) into the optical path of the 1553-nm laser, transforming its Gaussian mode into a vortex beam carrying orbital angular momentum (OAM) with a topological charge of 1. This vortex beam was then used as the pump source for frequency conversion, successfully transferring the OAM to the 221 and 193 nm lasers. As a result, we obtained a vortex beam at 193 nm with a topological charge of 2. To our knowledge, this is the first compact 193-nm laser generation system using OPA and cascaded LBO crystals, as well as the first demonstration of vortex beam generation at 193 nm from a solid-state laser. This innovative configuration opens up new possibilities for applications by solid-state laser technology.

The conversion process from 1030 to 193 nm laser closely resembles that in our previous work.20 A 1030-nm laser amplifier based on a $2\mathrm{mm}\times 2\mathrm{mm}\times 30\mathrm{mm}$ Yb:YAG crystal pumped by a 100 W multimode laser diode (LD) at 969 nm provides more than 14-W 1030-nm pulsed laser with 6 kHz repetition rate and 13.1 ns pulse duration. The 1030-nm laser is used to generate a 258 nm laser through successive SHG and FHG processes, within an LBO crystal and a CLBO crystal, respectively. It is also used as the pump of two stages of OPA, to provide a high-power 1553-nm pulsed laser. Instead of fiber-based amplifiers, here, we employ an OPA-based laser source to produce a sub-watt-level pulsed laser at 1553 nm. This modification results in a more compact system and eliminates the need for electronic controllers to synchronize the pulse trains of 1553 and 258 nm in SFGs, which is now achieved using an optical delay line. Two stages of SFG, pumped by 1553 and 258 nm lasers using cascaded LBO crystals, generate 221 and 193 nm lasers, respectively. The whole experimental scheme is depicted in Fig. 1. In the following sections, we will discuss the details of the induced OPA laser and the corresponding frequency conversion process to a 193 nm laser.

The 1553-nm pulsed laser source consists of a continuous-wave (CW) single-frequency distributed feedback (DFB) LD as the seed and two stages of OPA based on PPLN crystals, as shown in Fig. 2.

The DFB LD operates at 1553 nm and emits an average power of 12 mW. First, the 1030-nm pump laser is induced, together with the seed laser, into a $1\mathrm{mm}\times 1\mathrm{mm}\times 40\mathrm{mm}$ PPLN (Covesion, poling period of $29.98\mu \mathrm{m}$ and held at 122°C by an oven) to form the first stage of OPA. The amplified signal laser, which is filtered out by dichroic mirrors from the output of OPA 1 and OPA 2, is accompanied by a residual pump laser and an idler laser at $3\mu \mathrm{m}$ during the OPA process. The signal laser power is determined by an energy meter (Ophir, PE9-ES-C) to distinguish the pulse signal components from the CW seed laser. Because of the low-duty cycle of the pump laser and the weak seed laser, the OPA pump threshold is up to nearly 600 mW. With a pump laser of $\sim 700\mathrm{mW}$ average power, over $8\mu \mathrm{J}$ pulse energy is obtained from the OPA 1, corresponding to an average power of 48 mW. Then, the amplified pulsed signal is further scaled up in OPA 2, with another $5\mathrm{mm}\times 3\mathrm{mm}\times 30\mathrm{mm}$ PPLN (HCP Corp., poling period of $30.3\mu \mathrm{m}$ and held at 47°C by an oven) and maximum pump power of 3 W. Note that the power density of pump lasers in both stages of OPA are kept at nearly $30\mathrm{MW}/{\mathrm{cm}}^2$ to avoid photorefractive damage of PPLN. The average power of signal laser versus pump power in OPA 2 is shown in Fig. 3(a). A 700-mW signal laser at 1553 nm is obtained, corresponding to an efficiency of 23.3%. The increasing efficiency indicates that the output power could be further improved with more pump power. The optical spectrum of the seed laser and signal lasers from OPA 1 and OPA 2 are shown in Fig. 3(b), respectively, which are measured by an optical spectrum analyzer (YOKOGAWA, AQ6370D) with a resolution of 0.02 nm. The central wavelength of the amplified signal laser aligns with the seed laser, but the spectrum is slightly broadened. This broadening may be attributed to phase mismatch during the OPA process, the conversion from CW to pulse mode, and the linewidth of the pump laser. Although parametric fluorescence noise may increase with higher pump power, the signal-to-noise ratio remains close to 50 dB. To accurately measure the linewidth evolution of the 1553-nm laser during OPA processes, a scanning Fabry–Perot interferometer (Thorlabs, SA30-144) with a resolution of $\sim 1\mathrm{MHz}$ and a free spectrum range of 1.5 GHz is used; the results are shown in Fig. 3(c). The initial linewidth of the CW laser is broadened from 180 to 370 and 580 MHz during OPA 1 and OPA 2, respectively, reflecting the aforementioned spectral changes. The pulse duration of the pump and signal laser are also investigated by an InGaAs photodetector, and the result is shown in Fig. 3(d). Because of the parametric conversion threshold of the OPA process, the pulse of the signal laser has a steeper leading edge than the pump laser, and the duration is shortened from 13.1 to 9 ns. Consequently, an OPA-based pulsed laser at 1553 nm with 700 mW average power and 9 ns pulse duration was obtained and used as the pump source for generating the 193-nm laser.

The remaining part of the pump laser from the Yb:YAG bulk crystal amplifier is delivered to SHG. With the pulsed laser of average power of 9 W at 1030 nm, 515 nm green light is produced within a $5\mathrm{mm}\times 5\mathrm{mm}\times 30\mathrm{mm}$ LBO crystal [(type-I phase matching, 1030 nm (o) + 1030 nm (o) → 515 nm (e)], which is held in a copper holder at 185°C to meet a noncritical phase-matching condition. The maximum generated output power and conversion efficiency for the 515 nm light reaches nearly 5.6 W and 62.2%, respectively, as shown in Fig. 4(a). The generated 515-nm green light is subsequently launched into a $5\mathrm{mm}\times 5\mathrm{mm}\times 19\mathrm{mm}$ CLBO crystal [$\theta =66.2\mathrm{deg}$, $\phi =45.0\mathrm{deg}$, type-I phase matching, 515 nm (o) + 515 nm (o) → 258 nm (e)] for FHG. The CLBO crystal is heated to more than 150°C and surrounded by noble gas (e.g., argon or nitrogen) to prevent the hydroscopic problem. As shown in Fig. 4(b), when the pump power at 515 nm reaches 5.6 W, the output highest average power of 258 nm laser is 1.2 W, corresponding to a conversion efficiency of 23%.

The SFG processes are implemented in the collinear cascaded scheme using two LBO crystals. To compensate for the deduced power of 258 nm laser compared with our previous work (about 2 W 258 nm laser) due to the pump power reduction of the 1030-nm laser, here, we choose a longer LBO-1 crystal (CASTECH, $5\mathrm{mm}\times 5\mathrm{mm}\times 30\mathrm{mm}$, $\theta =55.6\mathrm{deg}$, $\phi =90\mathrm{deg}$) to generate 221 nm laser, under type-II phase match [1553 nm (o) + 258 nm (e) → 221 nm (o)]. The 258-nm laser is collimated and combined with the 1553-nm laser from OPA with a dichroic mirror and injected into the LBO-1 crystal to generate a 221-nm laser.

Theoretically, in the SFG process, two photons with frequencies ${\omega }_1$ and ${\omega }_2$ interact within a nonlinear crystal; the two photons combine to generate a single photon with a frequency of ${\omega }_{3 }={\omega }_{1 }+{\omega }_2$, conserving energy in the process. Consequently, the ideal power ratio of the pump lasers should ideally correspond to the ratio of their respective frequencies (${\omega }_1/{\omega }_2$). In this work, the 1553-nm laser provided more power to guarantee the efficiency of SFG. Figure 5(a) shows the average power of the 221 nm laser plotted versus the pump power of the 258 nm laser. With 1553 nm laser power fixed at 700 mW, a maximum output power of 270 mW at 221 nm is obtained. The conversion efficiency stays at $\sim 23\%$ as the pump power increases. Subsequently, the 221 nm laser and the residual 1553 nm laser are used to pump the second stage of SFG. An LBO-2 crystal (CASTECH, $5\mathrm{mm}\times 5\mathrm{mm}\times 30\mathrm{mm}$, $\theta =90\mathrm{deg}$, $\phi =76.4\mathrm{deg}$) is placed closely following the LBO-1, to generate 193 nm laser under a type-I phase match [1553 nm (o) + 221 nm (o) → 193 nm (e)]. As shown in Fig. 5(b), with low pump power, the conversion efficiency from 221 to 193 nm laser remains close to 30%. As the pump power exceeds 200 mW, there is an obvious efficiency drop, from 30% to 25%, and when the pump power reaches 270 and 70 mW, a 193 nm laser is obtained. The reduction in efficiency may be attributed to the long crystal used in the first stage of SFG. The walk-off effect, which is proportional to the length of the nonlinear crystal, leads to an unavoidable reduction in the overlap of pump lasers in the LBO-2 with cascaded configuration. In addition, the linewidth broadening of the 1553 nm laser from OPA imposes further restrictions due to the acceptance bandwidth of the crystal. To some extent, these issues could be mitigated by providing a 1553 nm laser with a higher power to compensate for the photon depletion. The pulse durations of the 221 and 193 nm lasers are measured to be $\sim 4$ and 3.5 ns, respectively, using a biplanar phototube (Hamamatsu, R15590U-55), as shown in Fig. 5(c). Compared to the pump laser, the generated lasers from SFG exhibit a more symmetrical, nearly Gaussian shape, which makes it more suitable in practical applications. The linewidth of the 193-nm pulsed laser cannot be measured directly because of the lack of a suitable Fabry–Perot interferometer; it is theoretically estimated by $\mathrm{\Delta }{\nu }_{\mathrm{SF}\text{G}}=\sqrt{\mathrm{\Delta }{\nu }_{\text{Pump}1}^2+\mathrm{\Delta }{\nu }_{\text{Pump}2}^2}$ in nonlinear frequency conversion processes. The linewidths of the fundamental pump lasers at 1030 and 1553 nm lasers are measured to be 15020 and 580 MHz, respectively. Thus, it is estimated to be $<873\mathrm{MHz}$, corresponding to an FWHM of $<0.11\mathrm{pm}$ at 193 nm. As discussed in previous works,5^,7 our 193 nm laser is suitable for seeding the ArF excimer amplifier.

Optical vortex beams carrying OAM have a distinctive spiral phase and a donut-shaped spatial intensity profile. Their unique spatial structure makes them valuable in optical communication, high-precision measurement, and optical trapping.21^–23 For instance, Wang et al. introduced OAM-carrying beams in bright-field coherent Fourier scatterometry, significantly improving defect inspection performance in semiconductor manufacturing.24 It also shows a special application in laser-induced material processes.25^,26 Typically, OAM beams could be generated using spatial light modulators or SPPs, but these methods are challenging to implement in the deep ultraviolet (DUV) region, hindering the direct generation of short-wavelength OAM beams. Fortunately, numerous researchers have demonstrated that OAM can be transferred during nonlinear frequency conversion processes such as SHG and SFG.27^,28 To further expand the 193-nm laser application, we initially experimentally demonstrated a vortex beam at 1553 nm by introducing an SPP (Welloptics Co., Ltd.) after the OPA to transfer the fundamental Gaussian mode of 1553-nm pulsed laser to the Laguerre–Gaussian (LG) mode carrying OAM. The SPP with a 25-mm diameter is mounted in a lens adapter with 25.4-mm diameter. Although the SPP does not have anti-reflection coating on both ends, the transmission is $>90\%$. Subsequently, the carried OAM transfers to 221 and 193 nm lasers through SFG processes, as shown in Fig. 6.

To verify the generation of the vortex beam, a pyroelectricity camera (Ophir, Pyrocam III HR, single pixel of $80\mu \mathrm{m}$) is used to record the beam profiles of the 1553, 221, and 193 nm lasers of different modes, respectively, as shown in Fig. 7. Before inserting the SPP, the 1553, 221, and 193 nm lasers all exhibit Gaussian mode profiles. After inserting the SPP, the 1553-nm laser mode is converted and exhibits a ring-shaped intensity distribution, which is characteristic of the LG mode. To determine its topological charge, we simply introduce a cylindrical lens ($f=25\mathrm{mm}$) to obtain the diffraction pattern of the LG mode, which is known as the Hermite–Gauss (HG) mode. To minimize the influence of Gouy phase shifts on the HG modes, the 193-nm laser beam was initially focused by a 200-mm focal length ${\mathrm{CaF}}_2$ lens. Given the short focal length (25 mm) of the cylindrical lens, it was positioned near the focal point of the ${\mathrm{CaF}}_2$ lens. Because of the short focus length of the cylindrical lens (25 mm), the cylindrical lens was placed near the focus spot. The camera could be placed far from the cylindrical lens (for example, 20 to 30 cm), to ensure that the beam size was large enough and its details could be recognized by the camera, as the pyroelectricity camera we used to observe the beam profile has low resolution. The cylindrical lens changed the ring beam into two bright spots with a gap in the middle, indicating that a vortex beam with topological charge $l=1$ was generated. This result corresponds to the 2π phase shift of the SPP, as described by the manufacturer.

Due to the significant difference in intensity distribution between the vortex beam and the Gaussian mode, the beam of the 258-nm laser had to be enlarged to cover the 1553-nm laser, ensuring better OAM transfer in SFG 1 and SFG 2. However, the weak power density of the 258 nm laser resulted in a much lower conversion efficiency of SFG compared with the above all-Gaussian mode experiments, and only 30 mW of the 221 nm laser and 3 mW of the 193 nm laser are obtained. According to the OAM conservation law in nonlinear processes, the topological charge of the generated laser from SFG is equal to the sum of the pump lasers.29 Therefore, with the 1553 laser with topological charge $l=1$ and the 258 nm laser of Gaussian mode (considered to have topological charge $l=0$), the topological charge of the 221 nm laser would be 1. This could be confirmed by the similar beam profiles of the generated 221 nm laser and the pump laser of 1553 nm, as shown in the second column of Fig. 7.

For the 193 nm laser, the vortex pump sources of 1553 and 221 nm lasers are both of the order of $l=1$, which contributes to the generation of the order of $l=2$ at 193 nm. Consequently, the diffraction pattern of the 193-nm vortex beam is split into three light spots with two dark gaps, whereas the intensity distribution still maintains the donut shape. Compared with the fundamental vortex beam at 1553 nm, there are unavoidable distortions to vortex beam profiles at 221 and 193 nm lasers during SFG processes because of the phase mismatch and walk-off effect from the nonlinear crystals. The cascaded structure would also add complexity to OAM conversion and even mode degradation might occur. We believe that the quality of the OAM-carrying mode could be improved using shorter crystals or independent SFG processes.

In summary, we have developed a compact solid-state nanosecond pulsed laser system at 193 nm with narrow linewidth, combining an OPA scheme and cascaded LBO crystals. This system delivers an average power of 70 mW with a repetition rate of 6 kHz, corresponding to a pulse energy of over $10\mu \mathrm{J}$. The linewidth of the 193 nm laser is estimated to be less than 880 MHz ($\mathrm{FWHM}<0.11\mathrm{pm}$). The entire system occupies an optical table measuring $1200\mathrm{mm}\times 1800\mathrm{mm}$, of which the footprint could be further reduced to meet the requirements of industrial applications. Considering that the 1553 nm laser is pumped and scaled up by a 1030 nm laser, the whole conversion efficiency from a 1030 to a 193 nm laser is $\sim 0.55\%$. Despite the current low conversion efficiency, the power scaling of the 193 nm laser is anticipated to exceed hundreds of milliwatts by increasing the pump power at 1030 nm, potentially reaching the watt level in future developments. Furthermore, employing nonlinear crystals with higher nonlinear coefficients would significantly enhance the feasibility of achieving this goal. In addition, by inserting an SPP to transform the Gaussian mode to an LG mode, a vortex beam carrying OAM at 1553 nm is created. The OAM is then transferred to 221 and 193 nm through SFG processes, respectively, resulting in topological charges of $l=1$ and $l=2$. The order of topological charge could be altered simply by changing the phase shift of the SPP. It has been reported that the OAM-carrying beam could be amplified in the single crystal fiber and nitrogen plasma,30^,31 suggesting that the 193-nm vortex beam could also be amplified in an excimer laser. Leveraging its high-power output and unique vortex beam characteristics, the 193-nm laser could be applied in a variety of new applications.

Biographies of the authors are not available.