Fiber lasers with outstanding beam characteristics, excellent energy conversion efficiencies, and exceptional integration capabilities have gained significant attention within the field of laser technologies at present^[1–5]. Particularly, due to the merits of narrow linewidth, low noise, and high stability, single-frequency fiber lasers (SFFLs) have experienced huge progress in the past decades and opened wide applications in many fields such as high-sensitivity sensing, high-precision measurement, and coherent optical communication and lidar. Based on different high-gain rare-Earth (RE)-doped fibers, SFFLs at 0.9, 1.0, 1.5, 1.9, and 2.9 µm have been widely investigated and well-developed^[6–12]. In addition, these wavelengths located at the gain edge of RE ions, such as 1120 nm of Nd-doped fiber^[13], 1150 nm of Yb-doped fiber^[14], 1627 nm of Er-doped fiber^[15], and 1720 nm of Tm-doped fiber^[16], have also seen exploration in cavity structure design or fiber core material optimization, taking full advantages of RE ion gain bandwidth. However, certain spectral bands remain inaccessible to RE-doped fibers for the generation of single-frequency lasers. These bands hold significant potential for various applications, especially at 1440 nm, which is not only applicable in laser-assisted liposuction and skin tightening treatments due to the high absorption rate of fat and water at 1440 nm^[17] but is also suitable for E-band optical communication. In this case, bismuth-doped fibers (BDFs) show great potential due to their broadband luminescence (1100–1800 nm) covering O, E, S, C, L, and U bands^[18,19].

At present, researchers have not only achieved broadband optical amplifiers^[20–25] but also realized highly efficient laser output^[26–29], broadband tunable lasers^[30–33], and multi-wavelength laser output^[34–36] in the 1300 nm band, 1400 nm band, and 1700 nm band based on BDFs. Nevertheless, there are few reports about SFFLs based on BDFs. In 2010, Kelleher et al.^[37] reported a narrow-linewidth BDFL operating at 1177 nm with a 30 m bismuth-doped aluminum silica fiber (BASF) as gain fiber. An output power of 10 mW in a 4 GHz linewidth was achieved under 7 W pump power. In 2015, Lobach et al.^[38] reported the realization of quasi-continuous self-sweeping near a central wavelength of $\sim 1460\mathrm{nm}$ in the single-frequency regime based on a 60 m polarization-maintaining (PM) bismuth-doped germanium silica fiber (BGSF) without information about the linewidth of the single-frequency (SF) laser. In 2023, Kharakhordin et al.^[39] reported a narrow-linewidth BDFL with a random cavity and operational wavelength at 1.67 µm. The maximum output power was $\sim 20\mathrm{mW}$, and the width of the laser emission line was narrower than 0.02 nm. There is still a significant gap between these results and the SFFLs achieved with RE-doped fibers. The primary reason preventing the realization of SFFLs based on BDFs lies in the low gain coefficient of BDFs, which generally requires tens or even hundreds of meters of BDFs to provide adequate optical gain for laser oscillation.

In recent years, thanks to the effective improvement of the absorption coefficient of BGSF, high gain can be obtained by utilizing shorter fiber length. For example, Zhai et al.^[40] reported a bismuth-doped fiber amplifier (BDFA) with a high gain of 40 dB at 1440 nm by utilizing a 35 m BGSF under the double-pass amplification configuration, and the gain per unit length reached 1.14 dB. Wang et al.^[41] presented a BDFA with a gain of 28.8 dB at 1420 nm based on 30 m BGSF, and the gain coefficient reached 0.96 dB/m. Liu et al.^[42] demonstrated a BGSF with a high absorption coefficient and obtained a high gain of 47.9 dB at 1450 nm based on 16 m BGSF under double-pass amplification, and the gain coefficient reached 4.06 dB/m. Effectively enhancing the gain coefficient of BDF provides the possibility for achieving SF output using BDF.

To propel the application of BDFs toward SFFLs, in this Letter, we develop a BGSF with a high gain coefficient, and an SFFL at 1440 nm is successfully demonstrated by a ring cavity structure with a short BGSF and two cascaded sub-ring cavities. The length of the BGSF in the cavity is only 10 m, which helps shorten the cavity length to a large extent. A maximum single-longitudinal-mode (SLM) output power of about 6 mW with an optical signal-to-noise ratio (OSNR) of over 75 dB and a narrow linewidth of less than 745 Hz is achieved. The output power fluctuation of $\pm 0.1\mathrm{dB}$ is exhibited in half an hour. These results confirm the feasibility and great potential of BDF in SFFL applications.

The BGSF is fabricated in-house through the modified chemical vapor deposition (MCVD) method combined with solution doping technology. The core and cladding diameters of the BGSF are 8 and 125 µm, respectively. The contents of Bi and Ge in the fiber are 0.0405% (mass fraction) and 9.3% (mass fraction), respectively. The refractive index difference between the core and cladding is $\sim 0.01$, as shown in Fig. 1(a), suggesting that the core numerical aperture (NA) is $\sim 0.17$. The cross-sectional image of the BGSF is displayed in the inset of Fig. 1(a). The absorption spectrum, measured by the standard cut-off method, is presented in Fig. 1(b). The BGSF exhibits an absorption peak at 1400 nm, which belongs to the bismuth-active center related to Si (BAC-Si), and the absorption coefficient at 1320 nm is 1.67 dB/m.

The BGSF has a wide emission spectrum in the E-band, and the gain coefficient at different wavelengths is different. In general, a higher gain coefficient means that the laser can operate with a higher efficiency, leading to increased output power. To explore the possibility of developing SFFL based on the homemade BGSF, we first characterize its gain performance. The experimental setup of BDFA is shown in Fig. 2(a). A 25 m long BGSF is used as the gain medium. Two laser diodes (LDs) operating at 1320 nm are utilized as the pump sources. A tunable laser source (TLS) is employed as the signal source, and the input signal power is controlled at the level of $-30\mathrm{dBm}$. Figure 2(b) shows the measured gain coefficient in the range of 1400–1480 nm under a total pump power of 785 mW. The gain spectrum of the BGSF exhibits a peak at a central wavelength of approximately 1430 nm. A remarkable gain coefficient of 1.05 dB/m at 1440 nm is obtained, which is higher than that of 0.66 and 0.875 dB/m reported in Refs. [43,44], respectively. It indicates that the homemade BGSF is beneficial for laser operation at 1440 nm.

Then, we adopt a simple ring cavity to explore the dependence of laser performance on the BGSF length. As depicted in Fig. 3(a), the ring cavity consists of a wavelength division multiplexer (WDM), an optical circulator (CIR), a fiber Bragg grating with a central wavelength of 1440 nm, and a 50:50 optical coupler (OC). An LD operating at 1320 nm serves as the pump source. Figure 3(b) illustrates the relationship between output power and pump power for different fiber lengths. It is observed that the laser efficiency exhibits a modest increase from 15.1% at a 25 m fiber length to 16.1% at a 20 m fiber length. However, with the fiber length further reduced to 10 m, the efficiency drops to 10.2%. Despite this reduction, an output power of 28 mW is still obtained with a 10 m BGSF under a pump power of 350 mW at 1320 nm.

Ring cavity configuration is a widely adopted solution for SF oscillation with low gain active fiber, as this cavity structure enables a much longer active fiber length compared to the typical centimeter-scale distributed Bragg reflector (DBR) or distributed feedback (DFB) single-frequency laser cavity^[45,46]. Thus, the single-frequency bismuth-doped fiber laser (SF-BDFL) is possible in theory when the length of the BDF could be shortened to tens or several meters.

According to the above results, in order to shorten the cavity length as much as possible while preserving output power, the 10 m BGSF is selected as the gain fiber for SF-BDFL. Figure 4(a) illustrates the experimental setup of the SF-BDFL. A 1320 nm LD with a maximum pump power of 500 mW is used as the pump source and protected by an isolator (ISO). The pump light is coupled into the BGSF through a 1320/1440 nm WDM. A CIR and an ISO are used to ensure the unidirectional transmission of the intracavity laser in the ring cavity. The high-reflectivity fiber Bragg grating (HR-FBG) with a reflectivity of 97.8% is written in a single-mode fiber by femtosecond laser technology and employed as a coarse wavelength selector determining the resonant wavelength. The transmission spectrum of HR-FBG is shown in Fig. 4(b).

The central wavelength and 3 dB bandwidth of the HR-FBG are 1440.1 and $\sim 0.18\mathrm{nm}$ (corresponding to 26 GHz), respectively. The laser is output through the 20% port of OC4, and 80% of the laser is returned to the main cavity. A polarization controller (PC) is equipped to control the laser polarization state in the laser cavity. The total length of the main cavity is about 15 m, corresponding to a free spectral range (FSR) of $\sim 13.8\mathrm{MHz}$. Obviously, the current longitudinal mode interval is too small to achieve SLM operation.

To expand the longitudinal mode spacing and further filtering, a compound sub-cavity consisting of two types of sub-rings is introduced. The first type of sub-ring (R1) is composed of a $2\times 2$ OC (OC1) with a coupling ratio of 50:50, and the cavity length is 0.9 m, giving an FSR of $\sim 230\mathrm{MHz}$. The second type of sub-ring (R2) is formed by two of the same $2\times 2$ OCs (OC2, OC3). The length of R2 is designed to make the effective FSR of the compound sub-cavity larger than half of the 3 dB bandwidth of the HR-FBG^[47]. According to the Vernier effect^[48], the effective FSR of the compound sub-cavity is the least common multiple of the FSR of all sub-rings. Based on this, the length of R2 is chosen to be 1.2 m, giving an FSR of 172 MHz, and the effective FSR of the compound sub-cavity is about 19.8 GHz, which is 0.7 times the 3 dB bandwidth of the HR-FBG. Besides, to achieve stable SLM operation, half of the effective 3 dB transmission bandwidth of the compound sub-cavity should be narrower than the FSR of the main cavity^[49,50]. The 3 dB transmission bandwidth of the sub-ring can be calculated as follows^[51]: $$\mathrm{\Delta }\nu =\frac{c\delta }{2\pi nL},$$(1)where $\delta$ is the one-way loss of the sub-ring cavity, $c$ is the speed of light in vacuum, $n$ is the effective refractive index, and $L$ is the sub-ring cavity length. Different coupling ratios of the OC will cause different losses and affect the effective bandwidth of the sub-cavity^[49]. The 3 dB bandwidth of R1 is calculated at about 29 MHz, half of which is larger than the FSR of the main cavity, and thus does not meet the requirement for SLM operation. Through careful calculations, we determine that the effective 3 dB bandwidth of R2 is $\sim 22\mathrm{MHz}$, half of which is smaller than the FSR of the main cavity when the coupling ratio of OC2 and OC3 is 0.7 and the OC’s insertion loss of 0.2 dB is taken into account. A large FSR and effective narrow bandwidth of the sub-cavity can guarantee SLM operation in theory.

The experimental setup is built on a common optical platform at room temperature without other vibration isolation and temperature control techniques. The output power of the BDFL is measured by a power meter (Thorlabs S145C). Figure 5 shows the relationship between laser output power and pump power. The laser threshold is about 115 mW, and after the laser threshold is exceeded, the laser output power and the pump power show a linear relationship with a slope coefficient of 3.4%. The maximum output power is about 11 mW under the pump power of 400 mW. The low efficiency is mainly caused by the low output ratio and the insertion loss of sub-cavity. It should be noted that mode-hopping is observed about every 30 s as the pump power increases to over 260 mW. This phenomenon may be explained as follows: more modes are excited within the cavity as the pump power increases, the gain competition between modes becomes more intense, and the effect of environmental disturbances becomes significant^[48]. The highest stable SLM output power is about 6 mW under the pump power of 258 mW. The pump power is fixed at 258 mW in the following measurement.

The longitudinal mode characteristic is confirmed first by the delay self-heterodyne detection system composed of a 50.4 km delay fiber and a 200 MHz frequency shift acousto-optic modulator (AOM). The heterodyne signal is measured by a radio frequency (RF) electrical spectrum analyzer (ESA) (Keysight, N9020B). The detected result is shown in Fig. 6(a). Only the zero-frequency signal and beat frequency signal at 200 MHz are in the whole frequency range, indicating that the laser is in an SLM state. At the same time, the SLM state is verified by a scanning Fabry–Perot interferometer (FPI) (Thorlabs, SA200-12B) with an oscilloscope (Agilent DSO5052A). The resolution and FSR of the FPI are 7.5 MHz and 1.5 GHz, respectively. As shown in the inset of Fig. 6(a), there are no additional parasitic peaks in the spectrum except between the main formants of the FPI, which is consistent with the results of the RF spectrum. Furthermore, to ensure the stability of the SLM operation over a 30 min period, the heterodyne signal of the laser is recorded at 5 min intervals, and the result is shown in Fig. 6(b). In addition to the zero-frequency signal, only one beat frequency signal at 200 MHz is presented, and no mode-hopping is exhibited during the entire measurement period.

The output spectrum of the single-frequency laser is recorded by an optical spectrum analyzer (OSA, YOKOGAWA AQ6370D) and shown in Fig. 7(a). The central wavelength is 1440.12 nm with an excellent OSNR exceeding 75 dB, and the detailed laser spectrum measured with the spectrometer resolution of 0.02 nm is shown in the inset of Fig. 7(a). The 3 dB bandwidth is measured to be 0.0233 nm, which is close to the minimum resolution of the OSA. To investigate the stability of the laser output, the output power and spectra are measured during a 30 min period. As shown in Fig. 7(b), the fluctuation of output power is less than 0.1 dB relative to the average power, and the laser central wavelength almost does not change. The small instability that exists is primarily caused by fluctuations in the ambient temperature and unavoidable environmental vibrations.

Moreover, linewidth is an important parameter of the SFFL, which is investigated by the delayed self-heterodyne method^[52]. Figure 8 shows an RF beat frequency signal near 200 MHz, with a scan interval of 100 kHz and a bandwidth resolution of 62 Hz. It should be noted that there are interference fringes in the measured RF power spectrum, which implies that the coherence length of the laser is considerably longer than the 50.4 km delay fiber^[52], but they are ignored during spectral shape fitting. From the Lorentz line fitting results of the heterodyne signal, the 20 dB bandwidth is estimated to be 14.9 kHz, corresponding to a 3 dB bandwidth of 745 Hz. It should be noted that having an insufficient length of delay fiber and ignoring the impact of the beat signal during fitting can cause the measured linewidth to be wider. Therefore, we conservatively estimate that the true linewidth of the SF-BDFL does not exceed 745 Hz.

In conclusion, we successfully demonstrate an SF-BDFL operating at 1440 nm based on the homemade BGSF. The short ring cavity laser configuration with two cascaded sub-ring cavities ensures the SLM operation. A maximum SLM laser output power of about 6 mW is obtained with an OSNR of over 75 dB under a pump power of 258 mW. Additionally, the measured linewidth is less than 745 Hz using the delayed self-heterodyne method. To the best of our knowledge, this is the first SF-BDFL at 1440 nm reported thus far, and it preliminarily shows the potential of BDF in developing a new band SFFL.