The National Ignition Facility (NIF) can output megajoule-level laser energy, which provides the conditions for research on high-energy-density physics and thermonuclear fusion.1 In recent decades, inertial confinement fusion (ICF) research has attracted much attention, especially after several ignition experiments at the NIF.2 In ICF experiments, the interaction between laser and target involves complex laser–plasma physical processes, and laser–plasma instability (LPI) will have significant effects on the fusion process.3,4 Precise laser control technology, advanced target manufacturing capabilities, and higher laser output energy are needed to realize successful ignition at the NIF.2 In addition, recent research results indicate that the use of broadband lasers can effectively suppress LPI, and it has been estimated that 10% bandwidth can completely suppress LPI and that multibeam broadband lasers can greatly increase the threshold for LPI.5–7 However, the bandwidth achievable with existing large-scale laser devices is less than 0.1%.8

With the aim of obtaining broadband laser output for experiments, much work has been done in developing more advanced laser techniques, especially those based on stimulated Raman scattering (SRS). Frequency conversion techniques based on SRS were investigated as early as the 1980s,9–11 the coupling process of multi-order SRS was analyzed theoretically, and it was demonstrated that this technique could be used to provide broadband lasers. Following on from this, a large number of studies on broadband laser generation using stimulated rotational Raman scattering (SRRS) in various gaseous media have been performed.12,13 To improve the efficiency of spectral broadening, a variety of experimental schemes have been proposed, such as two-color vibrational excitation,14,15 high-gain gas mixing,16 and hollow fibers filled with gas,17 and a wide range of spectral broadening from ultraviolet to infrared has been achieved. Thus, spectral broadening based on SRRS in gas has great promise for application in large-scale laser devices.

From another perspective, there have been a number of studies of second-frequency broadband laser techniques,18,19 which have led to energy outputs at the kilojoule level with bandwidths exceeding 10 nm. However, it should be noted that second-frequency lasers are inherently inferior to third-frequency lasers in terms of their physical effects.20

For large laser devices like the NIF, the broadening of a triple-frequency laser is limited by the frequency-tripling bandwidth of the triple-frequency crystal used, and this broadening must be performed after the frequency-tripling has been completed. It is expected that broadband ultraviolet lasers will ultimately be achieved using SRRS in gas. However, there are still many technical and engineering problems to be overcome in broadening the spectrum of large-aperture and high-flux lasers for ICF, in which a long transmission distance21 and the use of a high-gain gas medium are the two key factors.

To solve these problems, Stokes light, which is gradually excited during the long-distance transmission of the laser, can be injected as signal light, greatly reducing the propagation distance required for SRRS excitation and allowing spectral broadening to be achieved over a short distance. Such a scheme is compatible with the overall configuration of existing devices and is able to realize spectral broadening while maintaining energy output. At the same time, the intensity of the signal light can be adjusted to control the transient spectral broadening of the SRRS. In addition, with the injection of high-quality signal light instead of Stokes light (which is gradually excited by noise) to stimulate SRRS, it is possible to maintain laser quality after spectral broadening.

In this paper, we propose a scheme of spectral broadening based on SRRS on nitrogen excited with signal laser injection, and we are able to experimentally realize kJ-level ultraviolet laser output with 1% bandwidth while retaining good beam quality.

During the propagation of a high power laser in air, the nonlinear SRRS effect is easily excited, and Stokes components are produced by rotational scattering of the pump pulse on nitrogen molecules. Tasking account of the effects of diffraction during the long-range propagation of finite-aperture beams, the SRRS process can be described in the slowly varying amplitude approximation as follows:22,23∇⊥2+2ikL∂∂z′EL=−2κ2kLQEs,(1)∇⊥2+2ikS∂∂z′ES=−2κ2kLQ*EL,(2)∂Q*∂t′=−ΓQ*+iκ1ESEL*+F*,(3)where E_L is the pump field, E_S is the scattering field (Stokes field), Q is the atomic polarization field, Γ is the Raman bandwidth, and F is the molecular collision force. In the evolution Eq. (3) for the atomic polarization field, the first term on the right-hand side represents the relaxation effect of the field, and the second term indicates that the pump and scattering fields E_L and E_S are the driving sources of the growth in the atomic polarization field. The initial atomic polarization field and molecular collision force are excited by random noise. It can be seen that when there is no initial scattered light field, the molecular collision force excited by noise gradually increases to excite the atomic polarization field, and this is the process by which SRRS is excited during the long-range propagation of the laser.

On this basis, we propose an innovative scheme for fast excitation of the SRRS process with signal laser injection serving as first-order Stokes light.24 We inject the pump laser and signal laser simultaneously, thus realizing rapid amplification of the Stokes light field, as shown in Fig. 1(a). SRRS can be excited quickly by two-color injection of pump laser and signal light to realize transfer of pump laser energy to signal light and higher-order Stokes light. Because the excitation of the atomic polarization field depends on both the pump laser and the signal light, it becomes important in a short distance with increasing signal light intensity at a constant pump laser injection intensity. A homemade code that mainly considers the forward SRRS process along the laser beam axis is used to compute the laser evolution. Figure 1(b) shows the calculated energy depletion rate at different signal laser injection intensities, with the intensity of the pump laser set at I = 2.2 GW/cm², the beam aperture set at L = 36 cm, and the propagation distance set at 10 m in air. The pump and signal laser both start with a super-Gaussian flat-top profile, with linear polarization, which is consistent with practical conditions. Here, the energy depletion rate can be taken to roughly represent the final degree of spectral broadening, with 80% pump laser energy depletion corresponding to 2% spectral broadening according to experimental results.11 The simulation results show that with increasing initial signal laser intensity, the energy depletion of the pump laser is greatly accelerated, which demonstrates that the addition of initial signal light greatly shortens the distance requirement for the initial noise seed. This is the theoretical foundation for the concept of short-distance spectral broadening with signal laser injection. With increasing signal light power density, the depletion rate tends to saturate after propagating a distance of 10 m. Figure 1(c) shows the energy depletion rate after 10 m propagation for different signal light intensities. At the same propagation distance, the two-color energy depletion rate is greatly improved, corresponding to the degree of spectral broadening. When the signal light intensity reaches I_s = 2 × 10⁴ W/cm², the energy transfer rate is close to 80%, which corresponds to 2% spectral broadening.

Spectral broadening with signal laser injection also has a beneficial effect on suppression of near-field modulation. This is because the initial Stokes laser, which is susceptible to random noise, has been replaced by an artificially injected signal laser with a certain spatial distribution and temporal waveform. This makes the subsequent excitation process less random and uncontrollable, and thus suppresses high-frequency modulation in the near-field distribution. The near-field differences between single-laser and two-color injection excitation at the same energy conversion efficiency of 50% are illustrated in Fig. 2(a). The results show that a large amount of high-frequency modulation occurs in the near field under single-laser excitation, which is due to excitation of random noise in the Stokes light. With two-color laser injection, the initial Stokes light is replaced by the signal light, which avoids the accumulation of noise during long-distance propagation and the generation of near-field high-frequency modulation.

Because the pump laser energy is gradually transferred to the Stokes light during the SRRS process and there is almost no loss of energy in the excitation of the atomic polarization field, the total time waveform changes little, as shown in Fig. 2(b). This time-evolution behavior also shows that the broadened spectrum exhibits the characteristics of first narrowing and then widening with time.

The experiment is performed on a high-throughput laser platform25 using the setup shown in Fig. 3. The platform contains four major subsystems: front-end, preamplifier, main amplifier, and final optical assembly. The front-end system can deliver a 100 nJ-level pulsed laser source, which is then amplified about 10⁷ times by the preamplifier system into a J-level laser beam. It is then injected into the main amplifier to be amplified about 10⁵ times, giving an energy output approaching 20 kJ.26 With this platform, synchronous injection of signal light is realized in the front-end through a two-color fiber bundle combiner, and the time waveform and relative delay of the two-color light can be adjusted independently, as shown in Fig. 3(a). Considering the amplification ability of different wavelengths and to prevent the signal laser from obtaining gain energy, we choose the fundamental wavelengths of the pump and signal lasers as 1051.650 and 1055.015 nm, respectively. This wavelength difference takes into account the difference the third frequency after frequency conversion, which corresponds to the rotational energy level wavenumber difference of nitrogen. In the experiment, by adjusting the birefringent filter crystal, the preamplifier system maintains good control of amplitude–frequency modulation under dual-wavelength operation. At the output position of the preamplifier, the energy of the two-color light is measured independently through an optical fiber filter, and the energy ratio is adjusted by the front-end injection energy. In the main amplification section, the energy after amplification is calculated and analyzed by the two-color light amplifier module, and spectral measurement is performed on the basis of the original energy and time waveform. The output fundamental light (1ω) has a beam aperture of 375 mm. A potassium dihydrogen phosphate (KDP) crystal converts approximately two-thirds of the 1ω laser energy to the second harmonic (2ω). Then, the 2ω laser mixes with the residual 1ω to produce the third harmonic (3ω) in a potassium dideuterium phosphate (DKDP) crystal. The generated 3ω laser is focused by a wedge focusing lens (WFL) and passes through a vacuum window (VW), a main debris shield (MDS), and a disposable debris shield (DDS) to the target. The VW separates the laboratory atmosphere from the vacuum chamber region.27 The frequency conversion is set according to the best matching angle of the pump wavelength (1051.650 nm) to achieve the highest triple-frequency energy output. After the frequency conversion process, a triple-frequency pump laser at the several kJ level and signal light at the tens of J level are generated, and the spectrum is then broadened by the SRRS process in air. As shown in Fig. 3(b), the SRRS process takes place in the atmospheric environment between the third-frequency crystal and the vacuum window, over a distance of about 5.5 m. The parameters of the broadened triple-frequency laser are sampled and measured by a beam sample grating (BSG) on the VW.

Figure 4(a) shows the final spectrum after the SRRS process. It exhibits discrete peaks with equal intervals, and the spectral interval is determined by the wavelength difference between the pump light and signal light. Stokes light components of about ten orders are produced, and the spectral spacing wavenumber difference is 30 cm⁻¹, which is nearest to the S(2) transition. This is because the two-color fundamental-frequency light produces four kinds of triple-frequency spectral components through the frequency conversion process, including frequency-doubling and frequency combination processes. The conversion efficiency of 2ω₁ + ω₂ is much higher than that of 3ω₂ in third-harmonic generation, and an atomic polarization field corresponding to the wavenumber difference under stronger signal light injection (3ω₁ → 2ω₁ + ω₂) is preferentially established.

Figure 5 shows the whole energy transformation process, consisting of the harmonic conversion and SRRS processes. The total energy of the fundamental light that we measured in the experiment is 4100 J, and the pump laser energy (ω₁) according to the calculation and the spectral distribution is 2900 J, which means that the energy of the signal laser (ω₂) is 1200 J. Using the harmonic converter program under the experimental conditions, the energy distributions for 350.55, 350.92, 351.30, and 351.67 nm are 947, 296, 89, and 0.15 J, respectively, which are represented by the histogram in Fig. 5(b). Here, we utilize the results obtained from the program calculations rather than experimental measurements because of the difficult in sampling and measuring the spectrum after the third harmonic crystal prior to propagation in air. The program for the two-frequency harmonic converter is an extension of the original single-frequency program, whose accuracy has been experimentally verified. The final spectral distributions calculated by the SRRS program and measured in the experiment are shown in Fig. 5(c), where the experimental results is derived by integration of the spectral distribution in Fig. 4(a). Comparison of the results reveals good consistency between them in terms of the energy distribution, taking account of the conversion process from spectral distribution to energy and the spatial selection in the spectral sampling process. The significant intensity of the anti-Stokes part of the spectrum in Fig. 4(a) is derived from SRRS at the fundamental frequency, serving as another shorter-wavelength signal light. Also, because the short-wavelength component has a higher gain in the amplification system, its initial injection intensity is higher, resulting in a significant intensity after the THG and SRRS processes. We contend that by THG, the two-color fundamental-frequency light produces pump light consisting of one main energy and three weaker signal lights, and then, during propagation through several meters, SRRS on nitrogen results in energy redistribution, ultimately leading to the laser spectral broadening observed in the experiment.

Figure 4(b) shows the near-field distribution after spectral broadening, which does not exhibit any obvious differences from that in the monochrome case. It is found that no additional high-frequency modulation component is introduced, which is in accord with the theoretical results. Figure 4(c) compares the temporal waveform of the fundamental-frequency light and the broadened triple-frequency light. It is found that the square-wave signal, which has undergone harmonic conversion and SRRS broadening, is not subject to any additional obvious distortion except for a small amount of waveform distortion caused by self-modulation in the harmonic conversion process. What is important is that trailing-edge collapse of the temporal waveform is not introduced. This characteristic of the temporal waveform is also consistent with our simulation results. The broadened Stokes light and the original triple-frequency light are both used as driving lasers, and this will not cause the trailing edge of the temporal waveform to collapse.

These experimental results show that the ultraviolet spectral broadening realized by SRRS with signal laser injection can achieve last-stage broadening in an atmospheric environment, and the spectral width can reach 1%, with good preservation of temporal waveform and spatial distribution quality. The experimental results are in good agreement with theoretical results, demonstrating the engineering feasibility of achieving final SRRS broadening by coaxial injection of a signal laser.

Compared with the use of long-range propagation to stimulate SRRS by a KrF laser,21 this scheme can achieve high-gain spectral broadening at short distances by signal light injection and transient spectral control by signal light modulation. Meanwhile, the quality of the broadened beam is guaranteed owing to the high quality of the signal light. In contrast to table-top laser beam combination,14,15 through the use of a front end with two-color beam combination, we can achieve pump light and signal light co-injection with little modification of the original laser device, successfully realizing SRRS spectral broadening in the third-harmonic stage. At the same time, this scheme can guarantee a central wavelength in the ultraviolet suitable for driving ICF and with little impact on the output capability.18,19

Based on SRRS, we have proposed a technique for fast spectral broadening by two-color injection in a short distance. The results of theoretical calculation indicate that 1% spectral broadening can be achieved within several meters in a high-power laser facility. The near-field distribution is maintained well, and the total temporal waveform is hardly distorted. Finally, we have experimentally demonstrated a kJ-level 3ω laser with 1% spectral broadening, confirming the accuracy of our calculation model. The approach proposed here is expected to be applied to generate high-bandwidth ultraviolet driven lasers.

Acknowledgment. We thank Sun Zhihong, Shen Zhichao, and Men ting for their help in the experiments, and we thank Xie Yu for useful discussions of laser amplification and frequency conversion calculations. This work is supported by the Presidential Foundation of CAEP (Grant No. YZJJZL2023116), the National Nature Science Foundation of China (Grant No. 12275249), and the Youth Talent Fund of the Laser Fusion Research Center, CAEP (Grant Nos. RCFCZ7-2024-2 and RCFPD4-2020-4).