1. Introduction

Low noise continuous-wave (CW) lasers with high beam quality, high power and stable single-longitudinal-mode (SLM) operation are invaluable for various applications, including optical communication, quantum technology and coherent detection [1–4]. The optical power fluctuations of a laser are common to be specified as relative intensity noise (RIN). In the realm of inversion lasers, the RIN has been widely studied both theoretically [5,6] and experimentally [7]. Numerous studies have demonstrated the interplay between RIN and factors such as the pump scheme [8,9] and the mode structure [10–12], which have revealed that stable SLM operation is essential for achieving low-level RIN. Additionally, in inversion lasers, the intensity noise is often dominated by resonant relaxation oscillation (RRO) due to the longer population inversion lifetime than the cavity photon lifetime [7]. This RRO manifests within the Fourier frequency range of kilohertz to megahertz and forms a distinctive peak on the noise spectrum. To mitigate this RRO peak, various approaches have been adopted, including the use of electronic negative feedback controllers [13,14], the implementation of additional nonlinear loss [15,16], and the application of an injection-locking system [17]. In 2018, researchers from Shanxi University utilized an lithium triborate (LBO) crystal as a buffer reservoir to reduce the RRO peak, successfully developing an all-solid-state CW SLM 1064 nm laser, which reached an output power of 50.3 W and exhibited low intensity noise of 30 dB over the quantum noise limit, in the frequency range from 100 Hz to 10 MHz [10].

However, the inherent limitation in the energy-level structures of gain medium prevents inversion lasers from achieving wavelength further expansion. In addition, the spatial hole burning effect which is defined as a distortion of the intracavity gain field, complicates the design of laser cavity for generating SLM output [18,19]. In light of these challenges, stimulated Raman scattering (SRS) laser technology has emerged as a favor due to their flexible wavelength tunability and the gain nature of free spatial hole burning effect [20–22]. However, crystalline Raman lasers have typically been plagued by poor heat dissipation within the Raman gain medium. These thermal issues will degrade beam quality and confine the CW operation with high output power. Therefore, the characteristics of power fluctuations and RIN properties in solid-state Raman lasers are seldom investigated [23,24].

Synthetic single-crystal diamond has been found to be an excellent Raman gain medium to produce lasers with versatile wavelength and high power due to its extraordinary optical and thermal properties, including a high Raman gain coefficient (10 cm/GW @ 1 µm), a broad transmission window (ranging from 0.23 µm to far-infrared regions), a largest Raman shift (1332.5 cm^-1), an exceptional thermal conductivity (2000 W/m·K) and an extremely low thermal expansion coefficient (1.1 × 10⁻⁶K^-1) [25]. Leveraging these impressive characteristics, together with the free spatial hole burning effect and pump-to-Stokes energy funneling effect, diamond Raman lasers show an innate advantage for SLM operating even with simple laser cavity design [26–28]. Therefore, DRL provides a more suitable route to achieve high power, low RIN solid-state Raman laser output as well as to investigate the Stokes RIN properties including pump RIN transfer and RIN reduction. In this work, the RIN characteristics of a DRL are investigated for the first time. We find that the Raman RIN from 200 Hz to 1 MHz is sensitive to the parasitic stimulate Brillion scattering and the number of Raman oscillating longitudinal modes. The Raman RIN can be significantly decreased by suppressing the SBS and narrowing the Raman spectra, which is achieved by inserting spatial aperture and intracavity nonlinear loss.

2. Experiment setup

The experimental setup is depicted in Fig. 1. The pump source is composed of a single-frequency seed laser at 1064 nm and a linearly polarized narrow-band Yb-fiber amplifier, and delivers up to 70 W CW output power. A half-wave plate (HWP) is incorporated to optimize the Raman gain in diamond by rotating the pump polarization parallel to the diamond <111 > axis. A plane-convex focus lens (L1, f = 100 mm) is utilized to adjust the pump beam size and realize spatial mode-matching between the pump and Stokes beams at the center of diamond crystal. The diamond Raman resonator is designed as a V-shaped standing-wave structure to alleviate pump resonance as well as to improve the Raman resonance stability [29]. The cavity mirrors used in the experiment are listed in Table 1. The Raman medium is a 7 × 2 × 2 mm³ CVD-grown single-crystal diamond (Type IIa, Element Six Ltd.) with anti-reflective coatings at 1064 nm and 1240 nm on both sides with the cutting for propagation along the <110 > crystallographic direction. The diamond crystal is affixed to a water-cooled copper holder to maintain a constant temperature at 20 °C. In the experiment, three distinct configurations of diamond Raman cavity are used and presented in Fig. 1(a), (b) and (c), respectively. After traversing a long-pass filter with a cutoff at 1200 nm (LP1200), the Stokes light is divided using a wedge and two beam splitters (BS). The output Stokes light is monitored by the measurement system, including a power meter (PM: Ophir Nova II), a scanning F-P interferometer (FPI: THORLABS, SA210-8B), an optical spectrum analyzer (AQ6370D, Yokogawa, Inc.) and a photodetector (PD) with a signal analyzer (Keysight, N9010B).

 

Fig. 1. Experimental configuration for the diamond Raman laser. (a) Setup of free-running equipment; (b) setup of utilizing a spatial aperture; (c) setup of introducing intracavity nonlinear loss using an LBO crystal. HWP: half-wave plate; L1: plane-convex focus lens (f = 100 mm); L2: plane-convex focus lens (f = 150 mm); LP1200: long-pass filter with a cutoff at 1200 nm; BS: beam splitter; PM: power meter; FPI: F-P interferometer; PD: photodetector.

Download Full Size | PDF

Table 1. Properties of the diamond Raman cavity mirrors.

View Table

3. Experiment result and discussion

The RIN spectrum of lasers can be characterized by performing spectral analysis on the electrical signal obtained from a photodetector [30]. Generally, the PD collects both the direct current (DC) and alternating current (AC) signals. The former one represents the laser average output power and the latter one encompasses the intensity jitter noise. The DC component can be directly read using an oscilloscope, and the AC component can be analyzed using a signal analyzer. Subsequently, the power spectral density (PSD) reveals the RIN characteristics of the laser. An InGaAs detector (PDA10CF, THORLABS) with a bandwidth of 150 MHz is used for signal detection and a signal analyzer with a wide frequency range of 10 Hz to 44 GHz with up to 40 MHz analysis bandwidth is employed for signal processing. To improve measurement accuracy across the whole frequency range from 10 Hz to 10 MHz, we divide the frequency range into six segments, including 10 Hz to 100 Hz, 100 Hz to 1 kHz, 1 kHz to 10 kHz, 10 kHz to 100 kHz, 100 kHz to 1 MHz and 1 MHz to 10 MHz, and the resolution bandwidths (RBW) of each segment are 3 Hz, 3 Hz, 3 Hz, 30 Hz, 100 Hz and 300 Hz, respectively, correlating to the video bandwidths (VBW) of 3 Hz, 3 Hz, 3 Hz, 3 Hz, 10 Hz and 30 Hz.

To initiate the RIN investigation, the free-running DRL configuration (Fig. 1(a)) is firstly operated with a threshold of 19.5 W and maximum Raman output power of 23.8 W at the pump power of 67.3 W. The total cavity length is 281.2 mm (131.2 mm from M1 to M2 and 150 mm from M1 to M3) and this cavity configuration provides the Stokes and the pump beam waist of 31 µm and 17 µm, respectively. Due to lack of approaches to manipulating the resonance modes, the Raman output tends to operate in multiple-longitudinal modes and is accompanied by parasitic SBS with the increase of pump power, which is similar to the results in Ref. [29,31,32]. The RIN of the free-running diamond Raman laser is analyzed and the results are plotted in Fig. 2(a). The RIN curves of the 1064 nm Yb-fiber amplified pump laser present a typical low-pass and high-cutoff filtering property [33–35], and the RIN of the low-pass frequency continuously decreases with the increasing of the output power. The RIN characteristics of the 1240 nm Stokes laser with output power of 1 W and 20 W are illustrated in Fig. 2(a), from which the results show that the RIN of the Raman laser is significantly higher than that of the pump laser in the frequency range below 500 kHz. With the increasing of Raman power from 1 W to 20 W, the RIN curves demonstrate a prominent rise in the range of 100 Hz to 4 MHz, especially observe a noticeable broadening between 25 kHz and 4 MHz. In addition, the Raman RIN curves exhibit pronounced peaks at 150 Hz and its harmonics, and those noise peaks which are speculated due to environmental perturbations such as mechanical vibrations and air flow maintain stably across the whole experiment.

 

Fig. 2. (a) RIN spectra as a function of varying Stokes output power along with the corresponding pump powers. Black curve: background of PD; Light-blue curve: RIN spectrum of pump with a power of 24 W; Blue curve: RIN spectrum of pump with a power of 57 W; Red curve: RIN spectrum of Stokes with a power of 1 W at the pump power of 24 W; Purple curve: RIN spectrum of Stokes with a power of 20 W at the pump power of 57 W. (b) and (c) were the spectrum at the Stokes output power of 1 W and 20 W, respectively.

Download Full Size | PDF

To investigate the Stokes RIN deterioration, the Stokes output spectra are recorded by using a spectrum analyzer with the resolution of 0.03 nm (6 GHz at 1240 nm), presented in Fig. 2(b) and (c). The central wavelength of the first-order diamond Raman laser is at 1239.8 nm and the 20-dB spectral bandwidths of the SRS are 0.26 nm and 0.39 nm, respectively at the Stokes output powers of 1 W and 20 W. When the Stokes output power is 1 W, the first-order diamond Brillouin frequency at a separation of roughly 70 GHz from the longer wavelength side of the SRS is observed, as shown in Fig. 2(b). This parasitic SBS has very low spectral intensity ratio and is found to be unstable because it is near the SBS threshold. At higher Stokes output power of 20 W, the first SBS peak becomes stable and even the second order SBS appears at a spectral distance of 144 GHz from the SRS central frequency, as shown in Fig. 2(c). The proportion of the SBS spectral intensity to the total spectral intensity is increased from 3.5% to 7.6%, when the Stokes output power increases from 1 W to 20 W. In the previous report [36], the parasitic SBS was identified as a limitation to achieving intrinsic SLM operation for the SRS, as it could deplete the energy of the SRS main mode and disrupt the gain saturation of the SRS neighboring axial modes and then lead to multi-longitudinal modes operation of the SRS. In our case, the Stokes RIN deterioration with the increase of output power is mainly due to the depletion of SRS wave by the Stochastic SBS wave, as well as the enhancement of SRS modes beating caused by spectral broadening.

To further demonstrate the impact of parasitic SBS on the Stokes output performance, we insert an aperture into the cavity close to the M3 mirror, as shown in Fig. 1(b), to suppress the parasitic SBS which is tended to oscillate with higher-order spatial modes in a V-shaped cavity [32]. In the experiment, a square aperture (Daheng Optics, GCM-5711 M) rather than a circular one is used to suppress higher-order SBS spatial modes without obstructing the Raman fundamental spatial mode whose T and S planes are split due to the astigmatism from the turning mirror of M1. The results as depicted in Fig. 3 and Fig. 4 show that eliminating the SBS is capable of significant improvements on the output power, spectral purification and RIN decrease. The Stokes output powers before and after suppressing the SBS are illustrated in Fig. 3(a). The lasing threshold is the same in both cases, but with the increase of pump power, there is an obvious climb in Stokes output power after eliminating the SBS, such as increases of 0.2 W, 1.3 W, 3.9 W and 4.1 W at the pump powers of 24 W, 29 W, 38 W, 57 W, respectively. It indicates that the SBS is generated near the Raman threshold, and the stronger the SBS intensity, the lower the pump to Stokes conversion. Figure 3(b) depicts the Stokes spectrum at the output power of 24 W after suppressing the SBS, and only the first-order Raman peak centered at 1239.8 nm is observed with a 3-dB spectral linewidth of 6 GHz, which is the measurement limit of the spectrometer. Additionally, without the parasitic SBS, the Raman spectrum displays a typical single-frequency profile and the fine spectral characteristic recorded using a F-P scanning interferometer exhibits only one longitudinal mode during the scanning period, which is similar to the results in Ref. [36].

 

Fig. 3. (a) The Stokes output power as a function of the pump power before (orange dots) and after (purple triangle) using an aperture. (b) The Stokes spectrum at the output power of 24 W after suppressing the parasitic SBS by using an aperture; The inset is the measured Stokes longitudinal mode structure by using the F-P interferometer.

Download Full Size | PDF

 

Fig. 4. The Stokes RIN curves before (orange and dark blue curve) and after (yellow and cyan curve) the suppression of SBS at the output power of about 24 W (a) and 1 W (b). (c) The RIN comparison of the SLM Stokes with output powers of 1 W and 24 W. (d) The Stokes output spectra recorded using “Maxhold” function of the spectrum analyzer.

Download Full Size | PDF

The RIN curves of the Stokes at the pump power of 57 W are illustrated in Fig. 4(a). The significant Stokes RIN reduction at the frequency range from 100 Hz to 4 MHz is achieved after the parasitic SBS is suppressed and the broad multiple-longitudinal modes SRS is narrowed to SLM. By contrast, when the pump power is fixed at 24 W, yielding the Stokes output power of about 1 W, the spectrum envelope of the Stokes RIN decreases only in the frequency range of 30 kHz to 500 kHz after preventing the SBS oscillation, as shown in Fig. 4(b). Figure 4(c) depicts the comparison of the Raman RIN curves at the output powers of 1 W and 24 W after using aperture. The Raman RIN at low frequency from 30 Hz to 30 kHz slightly decreases but significantly increases from 30 kHz to 580 kHz when the output power is scaled from 1 W to 24 W. Note that in the experiment, the measurement duration for each RIN curve using the frequency segmentation method is extended up to about 10 minutes. In order to record the Stokes spectral variation simultaneously during the RIN measurement, an optical spectrum analyzer with “Max-hold” function is used to obtain the long-term spectral peak values of the output Stokes. The Stokes spectra for each condition are illustrated in Fig. 4(d). The spectral results indicate that the SBS can steadily parasitize oscillation in a free-running diamond Raman V-shaped cavity even at very low output power, and the highest spectral intensity of the SBS can be comparable to or exceed that of the SRS. The signal-to-noise ratio of the SRS spectrum is greatly improved when the SBS is eliminated by the aperture. In addition, after long-term peak values acquisition, the spectral shape of the SRS does not maintain the SLM shape (Fig. 3(b)), which is due to the SLM SRS frequency drift and mode-hopping induced by cavity length fluctuations. Notably, at the maximum output power of 24 W, some spurious peaks emerge at the first-order SBS frequency with the intensity of 14 dB lower than that of the SRS. This slight SBS component possibly results to a RIN spectrum increase of approximately 10 dBc/Hz in the frequency range from 30 kHz to 580 kHz (Fig. 4(c)).

An alternative approach to effectively suppress the parasitic SBS and reduce the adjacent axial modes of the SRS is to insert an χ⁽²⁾ crystal into the resonator to introduce additional nonlinear loss by axial-modes sum frequency generation [37,38]. To deeply investigate the influence of parasitic SBS spectral intensity and SRS spectral bandwidth on the output RIN characteristic, an LBO crystal is incorporated into the resonator to modulate the Stokes output spectrum by controlling the intracavity nonlinear loss, as shown in Fig. 1(c). The LBO crystal with dimensions of 10 × 4 × 4 mm³ and cutting angle of θ = 85.8°, Φ = 0° is mounted in a temperature-controlled copper holder with the adjustment accuracy of 0.1 °C and positioned at the beam waist between M1 and M4. Note that the planar mirror M3 is replaced with M4 with a radius of curvature of 150 mm to provide a focusing point (beam waist of 170 µm) in the center of LBO crystal to achieve sufficient nonlinear efficiency. Consequently, the beam waists of the Stokes and the pump are adjusted to approximately 54.5 µm and 16.5 µm, respectively. The level of χ⁽²⁾ phase-matching is the key to adjust the nonlinear loss induced by LBO. The nonlinear loss coefficient $\eta $ is defined as the ratio of the power of SHG laser to the total output laser, referring to [32].

We continuously tune the phase-matching temperature of the LBO crystal to change the intracavity nonlinear loss coefficient and observe the variations of the Stokes RIN and spectrum. All the measured results are illustrated in Fig. 5(a) and 5(b). At an optimal phase-matching temperature of 46 °C, corresponding to the nonlinear loss coefficient of 56.3%, the Stokes spectrum exhibits that no parasitic SBS appears and the SRS maintains SLM due to the high nonlinear loss to the adjacent axial modes within the LBO acceptance bandwidth. As expected, Stokes RIN reaches its minimum at this temperature (orange curve in Fig. 5(a)). Nevertheless, the cavity mirror coatings are not tailored for efficient frequency doubling output at 620 nm, and thus the output contains both the fundamental light at 1240 nm and the frequency-doubled light at 620 nm with powers of 4.5 W and 5.8 W, respectively, at the pump power of 57.3 W. The 620 nm SHG power is measured after M4 mirror and the total SHG power is calculated using the coating reflections of M1, M2 and M4 at 620 nm, similar to the Ref. [32]. When the intracavity nonlinear loss coefficient is varied to 43.6%, the criticality of the SRS operation from SLM to multiple modes, parasitic SBS is observed occasionally, as shown in Fig. 5(b). The slight SBS intensity causes an increase in the RIN spectrum within the range of 550 Hz to 100 kHz, as indicated by the yellow line in Fig. 5(a). When the temperature of LBO varies from the optimal 46 °C to 38 °C and 54 °C, corresponding to the nonlinear loss coefficient of 10.8% and 9.7% respectively, the 3-dB linewidth of the SRS spectrum is broadened to 7.8 GHz and 13.7 GHz, and the parasitic SBS is consistently oscillating with spectral intensity ratios of 1.3% and 1.56%, respectively. The reduction of the Stokes RIN almost in the entire frequency range from 10 Hz to 1 MHz is achieved with a maximum disparity about 40 dBc/Hz at 100 kHz, when the nonlinear loss coefficient increases from 9.7% to 56.3%.

 

Fig. 5. RIN (a) and spectrum (b) under the nonlinear loss coefficient of 56.3%, 43.6%, 10.8% and 9.7%. (c) Comparison of RIN for two methods of suppressing SBS (the blue and the red curves using an aperture, and the purple one using a χ⁽²⁾ crystal).

Download Full Size | PDF

Figure 5(c) illustrated the Stokes RIN spectra with output power of 1 W using aperture (blue), output power of 24 W using aperture (red), and output power of 4 W using LBO crystal (purple). The results indicate that using LBO crystal to introduce additional loss to the axial modes is able to achieve much lower Stokes RIN level than that using spatial aperture to suppress higher-order SBS transverse modes at the frequency range from 200 Hz to 100 kHz, approximately 25 dBc/Hz reduction at 10 kHz. However, the RIN level at the frequency below 200 Hz seems immune to the SBS suppression. We believe this is due to the ambient vibration noise. Intracavity nonlinear loss not only can completely suppress the parasitic SBS and achieve SLM SRS output but also can filter out the high intensity Stokes fluctuations that undergo high SHG conversion, as well as can constrict the Raman spontaneous emission present in the adjacent non-lasing modes. Although the SRS spectral stability and RIN level are significantly improved by using χ⁽²⁾ nonlinear loss within the cavity, the cavity structure of the diamond Raman laser becomes more complicated compared to that by employing a spatial aperture. Furthermore, suppressing parasitic SBS using an aperture provides higher light-to-light conversion efficiency due to no extra intracavity loss introduced.

4. Conclusion

In conclusion, we investigate the RIN characteristics of a CW DRL for the first time. In the free-running standing wave DRL, the Stokes RIN deteriorates with the increase of output power due to the depletion of SLM SRS by the Stochastic parasitic SBS. Moreover, SBS is found to contribute to initiate SRS multimode operation which causes the increase of SRS modes beating and elevates of the RIN level. Two strategies including inserting an aperture and introducing extra nonlinear loss through an LBO crystal are used to counteract the parasitic SBS and reduce the Stokes RIN. The results indicate that the suppression of SBS is able to enhance output power, optimize spectral purification, and reduce RIN. Compared to the insertion of spatial aperture, the introduction of LBO crystal enables a lower Stokes RIN because the intracavity nonlinear loss results in more SHG conversion to the Raman fluctuations with higher intensity.