${\mathrm{Nd}}^{3+}$-doped silica fiber (NDF) has good thermo-mechanical properties of silica glass and a perfect energy level structure of ${\mathrm{Nd}}^{3+}$ ions, which is widely used in communication, biomedical, military, material processing, high-power laser, and other fields^[1,2]. The ${{}^4\mathrm{F}}_{3/2}\to {{}^4\mathrm{I}}_{11/2}$ transition of ${\mathrm{Nd}}^{3+}$ can generate a $\sim 1060\mathrm{nm}$ laser with extremely low laser threshold and long fluorescence lifetime, which benefits the laser output and energy storage. ${\mathrm{Nd}}^{3+}$-doped glass, crystal, and ceramics are well developed for this band^[3–5]. Moreover, the application of ${\mathrm{Nd}}^{3+}$ in several other bands has also been widely considered by researchers. The ${{}^4\mathrm{F}}_{3/2}\to {{}^4\mathrm{I}}_{9/2}$ transition of NDF can generate a $\sim 900\mathrm{nm}$ laser, which can be directly used in atmospheric detection, lidar, and ${\mathrm{Yb}}^{3+}$ ion pumping^[6]. The deep blue lasers produced by its frequency doubling can be used in underwater communication, precision optics, atom cooling, and other fields^[7,8]. The ${{}^4\mathrm{F}}_{3/2}\to {{}^4\mathrm{I}}_{13/2}$ transition of NDF can generate lasers in the E-band of wavelengths (1350–1450 nm), which plays an important role in the fields of remote sensing, surgery, optical fiber communication, information storage, and so on^[9,10]. These new applications require NDFs with a larger core-to-clad ratio and higher doping concentrations.

However, the ${\mathrm{Nd}}^{3+}$ ion has high cation field strength and requires coordination with more oxygen to reduce the energy of the system. But, the rigid network of silica glass makes it difficult for oxygen ions to gather around the ${\mathrm{Nd}}^{3+}$ ion in sufficient numbers. The ${\mathrm{Nd}}^{3+}$ ion can only consume less oxygen ions by agglomeration. Consequently, the ${\mathrm{Nd}}^{3+}$ ion is very easy to agglomerate and even crystallize in silica glass when the ${\mathrm{Nd}}^{3+}$ ions doping concentration is more than 1000 ppm (parts per million), resulting in deterioration of optical quality. Co-doping ${\mathrm{Al}}^{3+}$ is one of the good methods to eliminate agglomeration to some extent^[11,12]. In the 1990s, the application of NDF to $\sim 1060\mathrm{nm}$ lasers was widely pursued using a modified chemical vapor deposition (MCVD) method^[13,14]. IPHT (Institut für Physikalische Hochtechnologie, Jena, Germany) fabricated ${\mathrm{Al}}^{3+}/{\mathrm{P}}^{3+}/{\mathrm{Ge}}^{4+}$ co-doped NDF with ${\mathrm{Nd}}^{3+}$ contents of 1300 ppm by the MCVD method. Using laser diode (LD) pumping, researchers have obtained several tens of watts of the $\sim 1060\mathrm{nm}$ laser output in this NDF with a slope efficiency of 46%^[15]. Such a high slope efficiency is enabled by the ultralow optical loss of fibers prepared with the MCVD method. Nevertheless, the MCVD process combined with solution doping is difficult to fabricate large-diameter fiber core glass with heavy ${\mathrm{RE}}^{3+}$ contents while maintaining high homogeneity^[16]. As a result, it is difficult to meet the needs of the development of a new type of optical fiber towards a large core-to-clad ratio and large mode area (LMA).

Non-chemical vapor deposition (Non-CVD) methods have a significant advantage: they can be used to prepare large-size silica glass; hence, these methods have been adopted to prepare ${\mathrm{RE}}^{3+}$-doped silica glass and fiber^[17–19]. Fujimoto et al. used the zeolite method to prepare ${\mathrm{Nd}}^{3+}$-doped silica glass and fibers. The prepared fiber could provide mode-locked laser output with a high repetition frequency through short-cavity amplification^[20]. The maximum laser power they reported was 27.2 mW, with a slope efficiency of 19.6%. Wu et al.^[21,22] also used sol-gel methods to fabricate the NDF. However, the laser output power in the aforementioned studies was limited to the order of milliwatts.

The modified sol-gel method combined with high-temperature melting and molding technology has been demonstrated as an innovative technique to prepare large-size ${\mathrm{RE}}^{3+}$-doped silica glass with heavy doping and high homogeneity. A series of achievements have been achieved in LMA photonic crystal fibers (PCFs) prepared by this method^[23–25]. In this Letter, large-size ${\mathrm{Al}}^{3+}/{\mathrm{Nd}}^{3+}$ co-doped silica glass was prepared using this method. To test the laser performance, a 20/125 NDF was drawn at 1800°C–2000°C using this core glass, preliminarily. A maximum output power of 14.6 W was obtained with a slope efficiency of 39.6% at $\sim 1060\mathrm{nm}$ from the NDF under an 808 nm LD pump. An amplifier with a maximum power of 16.6 W corresponding to a slope efficiency of 30.5% at 1061 nm was also achieved. The properties of silica glass, fiber fabrication process, and its optical properties will be demonstrated.

In the sol-gel preparation process, tetraethoxysilane (TEOS), ${\mathrm{C}}_2{\mathrm{H}}_5\mathrm{OH}$, ${\mathrm{AlCl}}_3·6{\mathrm{H}}_2\mathrm{O}$, and ${\mathrm{NdCl}}_3·6{\mathrm{H}}_2\mathrm{O}$ were used as precursors. Deionized water was added to sustain the hydrolysis reaction. The preparation process of ${\mathrm{Nd}}^{3+}$-doped silica glass by the sol-gel method is the same as that of ${\mathrm{Yb}}^{3+}$-doped silica glass, which was described in details in Refs. [23,26]. The centimeter-scale Nd-doped monolithic glass can be further ground and polished directly into a preform core of the NDF. The mean doping concentrations of ${\mathrm{Nd}}^{3+}$ and ${\mathrm{Al}}^{3+}$ were 5000 and 50,000 ppm, respectively. The results of inductively coupled plasma-optical emission spectrometry analysis show that the ${\mathrm{Nd}}^{3+}$ and ${\mathrm{Al}}^{3+}$ contents in the silica glasses were close to the theoretical values.

The glass rod was cut and polished into 2-mm-thick sheets to study their physical and spectral properties. To investigate the doping homogeneity, the distributions of ${\mathrm{Al}}^{3+}$ and ${\mathrm{Nd}}^{3+}$ were characterized using an electron probe micro-analyzer (EPMA, Shimadzu, 1720H). A spectrophotometer (Lambda 900 UV-VIS-NIR, Perkin-Elmer) was used to record the absorption spectra of the glass sheets in the range of 200–1000 nm. The fluorescence lifetime and fluorescence spectrum of the glass were measured using a time-resolved spectrometer (Edinburgh Instruments, FLS920) with an excitation of an 808 nm LD. Fourier transform infrared (FT-IR) spectra were measured using a Nexus FT-IR spectrometer (Thermo Nicolet).

The rod-in-tube method was used to prepare the preform of the NDF. First, the ${\mathrm{Al}}^{3+}/{\mathrm{Nd}}^{3+}$ co-doped bulk silica glass was processed into a round rod with a size of $\varphi 2.8\times 30\mathrm{mm}$. It was then inserted into a matched pure silica tube to prepare a fiber preform. Finally, the fiber preform was drawn to 20/125 NDF at 2000°C in a drawing tower. The refractive index profile of the 20/125 NDF was measured using an interferometer fiber analyzer (IFA-100, Interfiber Analysis, LLC). The core transmission loss of the NDF was measured using a broadband light source and the cutback method.

Figure 1 shows a schematic of the experimental laser setup. A multimode LD with an output power of 40 W at 808 nm was used as the pump source. The 2 + 1 combiner had double-clad input/output fibers with core/clad diameters of 20/125 µm and an numerical aperture (NA) of 0.08/0.46, which matched well with the 20/125 NDF. The two fiber end-faces were cleaved at 0° to provide a Fresnel reflection (FR) of $\sim 4\%$, thereby forming the laser cavity. To separate the signal power from the residual pump power, a 1000 nm long-pass filter (LP1000) was used. A power meter (PM, Thorlabs S442C) and an optical spectrum analyzer (Yokogawa AQ6370D) were utilized to record the laser output power and laser spectrum, respectively. In the experiment, the NDF was bent into 15 cm diameter rings and fixed on a metal plate for cooling.

Figure 2(a) shows a photograph of a polished large-size core-glass block and rod. The original size of the glass block before cutting out the glass rod was $\varphi 30\times 12\mathrm{mm}$. There were no bubbles or stripes in either the glass block or the rod. The uniform distributions of ${\mathrm{Nd}}^{3+}$ and ${\mathrm{Al}}^{3+}$ in the glass rod are displayed in Figs. 2(b) and 2(c), respectively. Thus, large ${\mathrm{Al}}^{3+}/{\mathrm{Nd}}^{3+}$ co-doped silica glass rods with high optical quality and high ${\mathrm{Nd}}^{3+}$ and ${\mathrm{Al}}^{3+}$ doping levels can be prepared by the sol-gel method. It can be used to fabricate a new type of NDFs like LMA fibers and fibers with large diameter ratio of core/cladding.

Figure 3(a) labels the absorption peaks of ${\mathrm{Nd}}^{3+}$ from the ground state to the excited state in the wavelength range of 200–1000 nm. The peak absorption cross section at the ${{}^2\mathrm{H}}_{9/2}\to {{}^4\mathrm{F}}_{5/2}$ level was $1.09{\mathrm{pm}}^2$, located at 804 nm. The main emission peak of the fluorescence spectrum [see Fig. 3(b)] was located at 1059 nm, while the secondary emission peak was located at 1090 nm. The effective linewidth of the emission peak was 44.5 nm, which was obtained by dividing the integral area of the peak by the maximum peak. Emission cross sections (${\sigma }_{\mathrm{emi}}$) were calculated based on the Judd–Ofelt theory^[27–29]. The maximum value of ${\sigma }_{\mathrm{emi}}$ at 1059 nm was $1.77{\mathrm{pm}}^2$, which was higher than that of the ${\sigma }_{\mathrm{emi}}$ of ${\mathrm{Nd}}^{3+}$-doped silica glass prepared using other methods with similar compositions ($1.4–1.5{\mathrm{pm}}^2$)^[30]. As shown in Fig. 3(c), the fluorescence decay spectrum exhibited a typical exponential decay. The fluorescence lifetime (460 µs) was also longer than that reported in other similar studies ($<400\mathrm{\mu s}$)^[11,30]. A high ${\sigma }_{\mathrm{emi}}$ and long fluorescence lifetime improve the laser behavior of ${\mathrm{Nd}}^{3+}$-doped glass. The longer fluorescent lifetime of ${\mathrm{Nd}}^{3+}$ ions indicates that the agglomeration of ${\mathrm{Nd}}^{3+}$ in silica glass was considerably suppressed by co-doping with ${\mathrm{Al}}^{3+}$. According to Fig. 3(d), the hydroxyl content in silica glass was calculated as 1.8 ppm. The calculation method is described elsewhere^[31]. Low hydroxyl content is also a contributor to long fluorescence lifetime.

Figure 4(a) shows the refractive index profile of the 20/125 NDF. The refractive index difference ($\mathrm{\Delta }n$) between the core and silica cladding was $\sim 2.83\times {10}^{-3}$, and the calculated core NA was 0.09. The inset in Fig. 4(a) shows the fiber cross section. It has a round cladding. Figure 4(b) depicts the core transmission loss of the NDF. The typical 0.75 dB/m transmission loss of the fiber core at 1200 nm was mainly caused by impurities introduced during the preparation process. The cladding pump absorption coefficient of the NDF at 808 nm was $\sim 2\mathrm{dB}/\mathrm{m}$.

Figure 5(a) depicts the output power of three NDFs with fiber lengths ($L$) of 3, 4, and 5 m. The output power was the sum of the forward and reverse output powers. In the case of $L=5\mathrm{m}$, a maximum laser output power of 14.6 W was achieved when the 808 nm pump power was 38.3 W, and the corresponding slope efficiency was 39.6%. Under the same pump power, for $L=4$ and 3 m, laser powers of 14 and 13 W, respectively, were obtained, with slope efficiencies of 38.2% and 35.1%. A smaller fiber length corresponds to a smaller absorbed pump power, which reduced the output power or slope efficiency in our experiment. However, a fiber length exceeding 5 m did not significantly improve the slope efficiency because of the relatively high transmission loss of the NDF [0.7 dB/m, see Fig. 4(a)].

The absorbed 808 nm pump power was further calculated according to the residual pump power. For $L=5$, 4, and 3 m, the absorbed pump powers were 29.7, 27.8, and 24.7 W, respectively; thus, the corresponding slope efficiencies were 51.2%, 52.5%, and 54.9%, respectively. Therefore, through optimized fiber structure tailoring (such as increasing the ratio of core/cladding, adopting octagonal cladding) and/or by reducing the fiber transmission loss, the laser efficiency of the NDF can be expected to be improved further.

Figure 5(b) depicts the laser spectrum for $L=4\mathrm{m}$ at the forward output port when the output power reached 14 W. The spectrum consisted of two broad laser peaks at 1060–1070 nm and 1090–1100 nm. These peaks corresponded to the fluorescence peaks of ${\mathrm{Nd}}^{3+}$ [Fig. 3(b)]. Therefore, for an NDF-based amplifier, achieving a gain bandwidth $>40\mathrm{nm}$ is possible. The laser spectra for $L=5$ and 3 m exhibited no significant difference from the spectrum for $L=4\mathrm{m}$.

Table 1 lists the laser performance at $\sim 1060\mathrm{nm}$ of different ${\mathrm{Nd}}^{3+}$-doped fibers in Refs. [15,20,21] for comparison. It can be found that the NDF prepared by the MCVD method has higher output power and slope efficiency due to the very low optical loss. Compared with other non-CVD methods, the NDF prepared in this work shows higher output power. It should be pointed out that the NDF prepared in this work was preliminary and has a high transmission loss without the purification technology. But, the loss can be significantly suppressed by the purification technology that we have developed and have applied in ${\mathrm{Yb}}^{3+}$-doped silica fiber (YDF). The purification technology can reduce the optical loss of YDF prepared by the modified sol-gel method to 50 dB/km, and it is as low as the losses of fibers prepared by MCVD^[26]. Moreover, the modified sol-gel method is also beneficial to adjust the refractive index through co-doping ${\mathrm{F}}^{-}$ or ${\mathrm{P}}^{5+}$ to reduce the NA of the core and thus to improve the beam quality of the LMA fiber^[19].

Further, a monolithic all-fiber master oscillator power amplifier (MOPA) was built based on the setup in Fig. 1 to investigate the laser performance of the 20/125 NDF. The seed laser used in the MOPA is a fiber laser consisting of a 5 m commercial PM-NDF-5/125 fiber and two fiber Bragg gratings with a wavelength at 1061 nm, and it has an output power of 1 W at 1061 nm. The seed laser was connected at the left side of the 2 + 1 combiner. To reduce the insertion loss due to mode field mismatch, a 5/125 to 20/125 mode-field adapter was spliced between them. For high-power amplifiers, short fiber lengths are desirable. So, a 3.3-m-long 20/125 NDF was used in MOPA. In addition, the NDF was cleaved with an angle of $\sim 8°$ as the output port of the MOPA.

The output power of the MOPA is plotted versus the launched pump power in Fig. 6(a). A maximum output of 16.6 W is achieved when the pump power is increased to 54 W, and the corresponding slope efficiency is 30.3%. It should be noted that an 808 nm LD with an output power of 60 W was used here as the pumping source for the MOPA. Figure 6(b) shows the laser spectra of MOPA at the maximum output power and seed power (depicted in the inset). The peak-to-peak contrast between the emission at 1061 nm and the amplified spontaneous emission around 1060 nm reaches $\sim 50\mathrm{dB}$, so the output power can be further improved. Comparing the spectra of the amplifier and seed laser, it can be found that there is a spectral depression near the 1061 nm laser peak. However, the reason for this remains unclear.

In this study, large-size ${\mathrm{Al}}^{3+}/{\mathrm{Nd}}^{3+}$ co-doped silica glass with high ${\mathrm{Nd}}^{3+}$ and ${\mathrm{Al}}^{3+}$ doping concentrations and high doping homogeneity was prepared using a modified sol-gel method combined with high-temperature melting and molding technology. The large emission cross section ($1.77{\mathrm{pm}}^2$) and long fluorescence lifetime (460 µs) indicated superior spectral properties and excellent ${\mathrm{Nd}}^{3+}$ ions dispersity. Based on this core glass, an NDF with core and cladding diameters of 20 and 125 µm, respectively, was drawn. A maximum output power of 14.6 W was achieved with 38.3 W pumping at 808 nm in the laser oscillation experiment. To the best of our knowledge, this is the highest laser output power reported for NDF operated at $\sim 1060\mathrm{nm}$ among the non-CVD methods. Besides, a maximum power of 16.6 W corresponding to a slope efficiency of 30.5% at 1061 nm was demonstrated in the MOPA experiment. By further optimizing fiber structure (such as increasing the ratio of core/cladding, adjusting the shape of cladding and core) and/or reducing the fiber loss, it is expected to increase the laser power and efficiency of NDF.