1. Introduction

With the rapid development of laser technologies and optical-gain materials, fiber lasers have witnessed great progress in the past decades and have been widely applied in lots of fields owing to the advantages of high beam quality, efficient heat dissipation, and compact and simple configuration [1–8]. Working wavelength extension is one of the aims in the development of fiber lasers [9–14]. For example, the efficient three-level (Nd³⁺: ⁴F_3/2-⁴I_9/2) laser in a Nd³⁺-doped fiber operating in a wavelength range of 850–950 nm is of interest for fundamental research and practical applications [14–18]. The continuous-wave (CW) and high-energy Q-switched pulsed laser sources at this waveband enable to generate blue and deep-ultraviolet (UV) lasers (e.g., 450 nm, 226 nm) through nonlinear frequency conversion [19–21]. In particular, the compact passively mode-locked (ML) femtosecond fiber lasers at this band (e.g., 920 nm) with high fundamental repetition rates are key light sources for two-photon microscopy (TPM) [22–25], which is a powerful method in high-resolution biomedical imaging [26,27].

However, it is challenging to achieve an efficient laser oscillation of the ⁴F_3/2–⁴I_9/2 transition in Nd³⁺-doped fibers because of its relatively low emission branching ratio compared to the undesired four-level (⁴F_3/2–⁴I_11/2) transition at 1.06 µm and the reabsorption effect of the ground state (⁴I_9/2) [28]. Different strategies have been proposed to overcome these limitations. Suppression of the competitive amplified spontaneous emission (ASE) at 1.0 µm was commonly used by modifying the designed geometry of the fiber. For example, iXblue photonics has launched several commercial Nd³⁺-doped silica fibers with a “W”-type refractive index profile [29]. The majority of reports on passively ML Nd³⁺-doped fiber lasers at ∼0.9 µm are based on these fibers with a well-suppressed 1.0-µm ASE [30–33]. The only disadvantage of this type of fiber may be the low absorption at the pump wavelength, which limits the pulse repetition rate of ML lasers to tens of megahertz [30–33], as a several-meter fiber needs to be used to provide adequate pump absorption and optical gain. Regarding high-power fiber lasers, Dawson et al. demonstrated a cladding-pumped 10-W 938-nm fiber laser by cooling an Nd³⁺-doped fiber with liquid nitrogen and reducing the reabsorption of the ground state [34]. By increasing the fiber core/cladding ratio, Laroche et al. improved the output power to 20 W at 910 nm using a fiber with a core diameter of 20 µm and cladding of 80 µm [35]. Pax et al. reported a 27-W 925-nm fiber laser using an all-solid micro-structured fiber [36]. This waveguide supports a large-mode-area (LMA) operation and strong suppression at 1 µm. Corre et al. further achieved the maximum documented power of 83 W (M² ∼ 1.5) at 910 nm by using a low-numerical-aperture (NA) Nd³⁺-doped LMA (30/130) fiber [37]. In these typical studies reviewed above, silica fibers are primarily used as the doping host for Nd³⁺. Recently, Fu et al. achieved a breakthrough in Nd³⁺-doped phosphate fiber lasers. They reported a high-efficiency (42.8%) CW laser output at 880 nm using a 25-cm-long 0.25-wt% Nd³⁺-doped phosphate fiber [38]. Later, they realized 880 nm and 915 nm single-frequency fiber lasers based on 2.5-cm-long 1.0-wt% Nd³⁺-doped phosphate fibers [39,40], which confirms that phosphate glass is a more suitable host than silica for lasing at ∼0.9 µm. Importantly, Fu et al. concluded that low-level doping of Nd³⁺ yields better laser efficiency rather than high-level doping due to the concentration quenching effect [38]. Following these studies, a Nd³⁺-doped phosphate glass fiber with similar doping concentration was developed that enables passively ML at 900 nm with repetition rate up to GHz [41,42].

As well known, the main benefit of phosphate glasses in terms of Rare-earth (RE) doping is their high solubility. Yet, the studies by Fu et al. and our previous results prompt us to think that whether phosphate glasses are superior to other oxide glasses when the doping concentration of Nd³⁺ does not need to be high for an efficient three-level ∼0.9-µm lasing, such as silicate glass, which has a more constrained three-dimensional rigid network that aids in dispersing RE ions and suppressing concentration quenching [43]. In this regard, in this study, we selected a Nd³⁺-doped silicate glass, a well-developed laser glass by our laboratory (Shanghai Institute of Optics and Fine Mechanics (SIOM), to explore its three-level laser performances. A low-loss Nd³⁺-doped silicate glass single-mode fiber (NDSF) has been developed by the rod-in-tube method. An efficient all-fiber CW lasing in the range of 890 to 915 nm with a SNR higher than 49 dB is achieved by only a 4.5-cm-long NDSF. In particular, the laser slope efficiency at 910 nm reaches 31.7%, which is comparable with that of a Nd³⁺-doped phosphate fiber [38]. Owing to the superior fiber performance, a passively ML laser at 920 nm with a gigahertz fundamental repetition rate is also demonstrated from an ultrashort laser cavity. These results indicate that the Nd³⁺-doped silicate fiber is also an efficient gain medium for CW and pulse lasers at ∼0.9 µm, with promising potential in single-frequency lasers and bio-photonic systems.

2. Experimental section

The near-infrared luminescence spectra and luminescence lifetime were measured using an Edinburgh FLS 920 spectrofluorometer. An 808-nm GaAlAs semiconductor LD was used as an excitation source. Fourier-transform infrared (FTIR) spectrum was recorded via a FTIR spectrophotometer (Nicolet 6700). The microscopy images of the fiber cross section were acquired using a JXA8230 electron probe microscopy analysis system. The 2D diagram of the refractive index profile of the fiber cross section was measured using an SHR-1802 interferometric fiber analyzer developed by Shanghai University. The optical fiber was drawn by a fiber drawing tower in the temperature range of 730–740 °C. The laser spectrum was monitored by an optical spectrum analyzer (YOKOGAWA, AQ6370C). The optical power was measured by a power meter with an integrating sphere sensor (Thorlabs, S145C) and console (Thorlabs, PM100D). The temporal pulse train was detected by an InGaAs photodetector (Thorlabs, DXM30AF) with a 30 GHz bandwidth and recorded by a real-time oscilloscope (Tektronix, DPO 73304DX). A signal analyzer (Keysight, N9000A) was employed to characterize the radio-frequency (RF) spectrum. The pulse width was measured by an autocorrelator (APE, Pulsecheck USB 150).

3. Results and discussion

The bulk Nd³⁺-doped silicate glass with a doping concentration of 1.2 wt% (1.02 × 10²⁰ ions/cm³) is cut from large, high-quality laser glasses, produced by the well-developed melting process in SIOM [44]. Figure 1(a) shows the normalized luminescence spectra of the bulk glass excited by an 808-nm laser diode (LD). Three typical luminescence bands of Nd³⁺ centered at 893, 1058, and 1330 nm could be observed, which belong to the transitions from the ⁴F_3/2 energy level to ⁴I_9/2, ⁴I_11/2 and ⁴I_13/2, respectively (Fig. 1(b)); their fluorescence branching ratios are 0.32, 0.59, and 0.09, respectively. In the silicate glass, the full width at half maximum (FWHM) at 0.9 µm reaches 54 nm. The photograph of the glass sample in the inset in Fig. 1(a) shows its high transmittance and uniformity. Figure 1(c) presents the measured fluorescence decay curves, fitted by a single exponential. The lifetime of the ⁴F_3/2 level of Nd³⁺ is 568 µs, longer than that of phosphate glass [38,41]. The long lifetime could be attributed to the low hydroxyl (OH^-) content in the selected silicate glass. As shown in the inset in Fig. 1(c), the FTIR spectrum shows a high transmissivity near 3 µm that attributed to the stretching vibration of free OH^- groups. The absorption coefficient (α_OH^-) is only 0.89 cm^-1, which could effectively reduce the energy transfer from Nd³⁺ (⁴F_3/2) to OH^-. Figure 1(d) presents the calculated absorption and emission cross sections at 0.9 µm. The maximum absorption cross section is 0.33 × 10⁻²⁰ cm², lower than that of phosphate glass [38], which indicates that the reabsorption from the ground state (⁴I_9/2) in the Nd³⁺-doped silicate glass is weaker. Notably, the emission cross section (1.03 × 10⁻²⁰ cm²) is higher than that of phosphate [38]. Overall, through the comparison of these radiative properties, it can be speculated that silicate may enable a more efficient three-level lasing of Nd³⁺ than phosphate glass.

 

Fig. 1. (a) Normalized luminescence spectra of the Nd³⁺-doped silicate glass at an excitation wavelength of 808 nm. The inset shows a photograph of the glass sample (10 mm × 10 mm × 5 mm). (b) Simplified energy level diagram of Nd³⁺. (c) Fluorescence decay curves of the Nd³⁺: ⁴F_3/2 level. The inset shows a FTIR spectrum. (d) Absorption and emission cross sections at ∼0.9 µm for the Nd³⁺-doped silicate glass.

Download Full Size | PDF

To validate this speculation, an Nd³⁺-doped silicate single-mode fiber with a standard diameter of 125 µm was fabricated by the rod-in-tube method. The cladding is a commercial H-K7 silicate glass, which has a matched refractive index and softening temperature with that of the core glass. Figure 2(a) presents a microscopy image of the fabricated NDSF with a core diameter of 4 µm. A 2D diagram of the refractive index profile of the fiber cross section measured by an interferometric fiber analyzer is shown in Fig. 2(b), which shows a refractive index difference of approximately 0.008 between the core and cladding. The calculated NA and cutoff wavelength are 0.14 and 731 nm, respectively, which ensures the single-transverse mode propagation at ∼0.9 µm. The mode diameter at 900 nm is estimated to be 5.2 µm based on the normalized frequency (V) of 1.95. These parameters of the fabricated NDSF closely match those of the commercial passive silica fiber (Coring, Hi-780), allowing for a low-loss fiber connection. Measured by a cutback method, the propagation loss at 1310 nm is 0.03 dB∕cm (Fig. S1) and the absorption coefficient at 808 nm is 4.1 dB/cm (Fig. S2).

 

Fig. 2. (a) Microscopy image and (b) two-dimensional (2D) refractive index profile of the fiber cross section at 633 nm.

Download Full Size | PDF

As depicted in Fig. 3, we constructed an experiment setup of a compact all-fiber CW laser to evaluate the fiber performance. The NDSF is forward-pumped by a single-mode LD at 808 nm with a maximum power of 250 mW through wavelength division multiplexer 1 (WDM1). The signal laser and residual pump are separated by WDM2. The laser cavity consists of a 4.5-cm-long NDSF, HR-FBG, and LR-FBG, which are physically connected and fixed to each other by UV adhesive. The measured cavity loss is 0.7 dB. A cooling system regulates the temperature of the laser cavity, which is housed in a copper tube. Four pairs of FBGs inscribed in the Hi-780 fiber with center wavelengths of 890, 900, 910, and 915 nm are used to analyze the wavelength with the maximum efficiency. The reflectivity of the HR-FBGs exceeds 99%. The reflectivity of the LR-FBGs at 890, 900, 910, and 915 nm is 82.34%, 83.25%, 80.72%, and 80%, respectively, with 3-dB bandwidths of 0.03–0.04 nm. The output laser spectra measured with a resolution of 0.02 nm are shown in Fig. 4(a). To compare the laser SNRs, the optical powers at these four wavelengths coupled into an optical spectrum analyzer (OSA) are attenuated to approximate power levels. Clearly, the 890–915-nm CW lasers could oscillate. All SNRs are higher than 49 dB relative to the ASE at 1 µm. Figure 4(b) shows normalized laser spectra, where the center wavelengths are 890.26, 899.44, 910.38, and 915.01 nm. Figure 4(c) is the laser output power as a function of the absorbed pump power, the output power at 890–915 nm linearly increases with the pump power. The maximum slope efficiency of 12.0% locates at 910 nm. The slope efficiencies at 915, 900, and 890 nm are 7.5%, 3.7%, and 5.2%, respectively. The relatively low laser efficiency at 890 and 900 nm can be attributed to the increased absorption coefficient from the ground state (Nd³⁺: ⁴I_9/2-⁴F_3/2) at a short wavelength (Fig. S2).

 

Fig. 3. Experiment setup of the 890–915 nm all-fiber CW laser based on the developed Nd³⁺-doped silicate single-mode fiber; LD: laser diode, WDM: wavelength-division multiplexer, HR-FBG: high-reflection fiber Bragg grating; LR-FBG: low-reflection fiber Bragg grating; NDSF: Nd³⁺-doped silicate fiber.

Download Full Size | PDF

 

Fig. 4. (a) CW laser spectra of 850–915 nm in the wavelength range of 850–1200 nm. (b) Normalized laser spectra. (c) Laser output power at 890, 900, 910, and 915 nm versus the absorbed pump power. (d) Laser output power at 910 nm versus the launched pump power and absorbed pump power when the LR-FBGs with reflectivity of 63% and 43% are used respectively.

Download Full Size | PDF

To extract more laser power from the cavity [40], we employed LR-FBGs at 910 nm with lower reflectivity. As shown in Fig. 4(d), the output power is efficiently elevated to more than 65 mW when the LR-FBG with 63% reflectivity was used, corresponding to a slope efficiency of 31.7% and an optical-to-optical efficiency of 29.7%, which is comparable with the laser efficiency realized by Nd³⁺-doped phosphate fiber with approximate fiber length [38]. Yet, the slope efficiency and optical-to-optical efficiency are reduced to 26.0% and 22.7%, respectively, as the reflectivity of LR-FBG was further reduced to 43%. It is believed that further improvement in laser efficiency could be achieved by optimizing the doping concentration of Nd³⁺ and fiber length [38]. The laser output power at 890–915 nm is stable (Fig. S3), the fluctuation of the output power in 1.5-hours relative to the average power is within 2%.

Considering the superior CW lasing performance of this NDSF, we further analyze its potential for the generation of ∼0.9-µm ML fiber lasers with a high fundamental repetition rate. A schematic of the constructed passively ML fiber laser is presented in Fig. 5. The pump power of the 808-nm LD is coupled into the laser cavity by a WDM. Epoxy is used to fix NDSF within a ceramic ferrule. Both end facets are perpendicularly polished. A fiber-type DF is butt-coupled to one of the end facet of the NDSF. The fiber pigtail is fusion-spliced to the common port of the WDM. A commercial SESAM with a modulation depth of 14%, non-saturable loss of 8%, and saturated fluence of 60 µJ/cm² at 940 nm is connected to the other end facet of the NDSF.

 

Fig. 5. Schematic of the passively ML fiber laser at 920 nm. SESAM: semiconductor saturable absorber mirror; NDSF: Nd³⁺-doped silicate fiber; DF: dielectric film.

Download Full Size | PDF

We incorporated NDSFs of different lengths (104, 63, and 42 mm) into the laser cavity to analyze the highest fundamental repetition rate (FRR) of the ML fiber laser supported by this newly developed NDSF, as the FRR is inversely proportional with the cavity length [45]. Here, since the thickness of SESAM and DF is negligible compared to the length of the NDSF, the cavity length is almost equal to that of the NDSF. The characteristics of the ML pulses are summarized in Fig. 6. The 104-mm NDSF enables a stable CWML with a FRR of 952 MHz, as shown in Fig. 6(a) and Fig. 6(b). The temporal pulse period is 1.052 ns, which corresponds to the measured repetition rate in the RF spectrum. The wider-span pulse traces and RF spectrum in the inset of Fig. 6(a) and Fig. 6 (b), as well as the 75.2-dB SNR of the RF spectrum confirm the good pulse stability. When the NDSF length is reduced to 63 mm (Fig. 6(c)), a CWML pulse train with a period of 648 ps is obtained, which is consistent with the repetition rate of 1.55 GHz in the RF spectrum (Fig. 6(d)). Yet, the SNR of the RF spectrum is decreased to 39.3 dB, which indicates that the pulse stability is decreased compared to the output pulses with a FRR of 952 MHz. The degraded performance could be attributed to the additional heat accumulation on the SESAM caused by the higher residual pump power. Furthermore, as shown in Fig. 6(e) and Fig. 6(f), CWML pulses are not achieved when the NDSF length is further reduced to 42 mm, even under the maximum pump power of 250 mW, due to the insufficient optical gain [47].

 

Fig. 6. Oscilloscope traces (a) and RF spectra (b) when incorporating 104-mm NDSF into cavity. Inset in Fig. 6(a) is the oscilloscope traces in a 40 ns time span. Inset in Fig. 6(b) is the RF spectrum in a wider span. Oscilloscope traces (c) and RF spectra (d) when incorporating 63-mm NDSF into cavity. Inset in Fig. 6(c) is the oscilloscope traces in a 40 ns time span. Oscilloscope traces (c) and RF spectra (d) when incorporating 42-mm NDSF into cavity.

Download Full Size | PDF

The pump threshold for self-started CWML with NDSF lengths of 104 mm is 180 mW (Fig. 7(a)). As shown in Fig. 7(c), when the length of NDSF is shorten to 63 mm, it was increased to 220 mW. The increase in ML threshold is due to the decrease in net gain when the gain fiber is shortened, which could be compensated for by a higher pump intensity [45–47]. The change in the net gain in the cavity could be manifested in the laser spectra as well. The peak wavelengths are blue-shifted from 919.27 to 903.74 nm, as shown in Fig. 7(b) and Fig. 7(d), with FRR improved from 952 MHz to 1.55 GHz. As the ML operation is a result of dynamic gain and saturable absorber loss balance, the reduction in the cavity length contributes to the shift of the lasing wavelength to the short wavelength side, where the larger emission cross section enables a higher gain to compensate for the saturable losses [47]. Additionally, the steep edges in optical spectra show the ML lasers operate in the all-normal dispersion regime. Assuming a Gaussin pulses shape, the pulses durations are 4.2 ps and 4.5 ps for 955 MHz and 1.6 GHz fundamental repetition rate (Fig. S4(a) and Fig. S4(b)). These results confirm this NDSF could also be a suitable gain medium for 920 nm ultrashort pulse generation with high repetition rate.

 

Fig. 7. (a) Output power versus pump power and (b) laser spectrum with 104-mm NDSF. (c) Output power versus pump power and (d) laser spectrum with 63-mm NDSF. QSML: Q-switched mode locking. P_th: threshold pump power of CWML.

Download Full Size | PDF

4. Conclusions

In conclusion, we designed and fabricated a low-loss Nd³⁺-doped silicate glass single-mode fiber with a core diameter of 4 µm and an NA of 0.14. The established all-fiber short cavity using the 4.5-cm-long Nd³⁺-doped silicate fiber produced an efficient CW lasing in the region of 890 to 915 nm with an SNR greater than 49 dB. The highest laser slope efficiency at 910 nm reached 31.7% when the reflectivity of the LR-FBG was 63%. Furthermore, a passively ML pulsed laser output at 920 nm with a gigahertz fundamental repetition rate was achieved from an ultrashort cavity. These results confirm that the Nd³⁺-doped silicate fiber could also be an effective gain medium for CW lasing and pulse generation at ∼0.9 µm, with potential applications in narrow-linewidth single-frequency lasers as well as bio-photonic systems.