Mode-locked (ML) lasers around 2 µm with pulse widths in the picosecond and femtosecond orders have been used in applications, such as remote detection radar systems, eye laser minimally invasive surgery, gas detection, and material micromachining^[1–4]. Relevant literature indicates that pulse durations in the picosecond range are advantageous for precision micromachining. Additionally, they can yield high average power and beam quality, thereby qualifying micromachining technology for industrial applications^[5,6].

The 1.8–1.9 µm band intersects with the absorption peak of water^[7], impacting the stability of continuous-wave mode-locked (CWML) lasers. This interference induces dispersion in the laser spectra, resulting in the broadening of the pulse width^[8]. If progress is to be made in the direction of ultrafast lasers in this band, Tm3+-doped lasers need more in-depth exploration and new experimentation. The strong absorption band of Tm3+ ion is around 800 nm, exactly coinciding with the wavelength of the commercial AlGaAs laser diode (LD), which makes the structure of the laser system simple, efficient, and more conducive to the realization of commercialization. Currently, Tm3+ ultrafast solid-state lasers at wavelengths exceeding 2 µm have been advanced to pulse widths below 100 fs^[9–15], whereas ultrafast lasers near 1.9 µm are lagging due to the significant water absorption^[16–20], despite their importance in practical applications. Therefore, achieving stable mode-locking within the water absorption band remains a significant challenge, awaiting further development and maturation.

The existing Tm3+-doped gain media suitable for ultrafast lasers have been studied deeply in recent years. Compared to other monocrystalline hosts available, calcium fluoride (CaF2) crystals, which can be commercialized for large-size growth, have low photon energy (322cm−1), which leads to the favorable phenomenon that cluster structure can be formed even at low doping concentration levels of rare earth ions^[21]. Within the cubic lattice structure of CaF2, F− compensating ions are positioned in the corners of the cubic structure, while Ca2+ ions are positioned in the center. Clusters to a certain extent facilitate the improvement of gain medium laser characteristics. In Tm3+ clusters, energy transfer between adjacent ions results in cross-relaxation processes, where one Tm3+ ion in the ground state transitions to the excited state, thereby enhancing pump efficiency. The cross-relaxation of heavily doped crystals can completely suppress 1.5 µm fluorescence. Although the cluster effect formed by high Tm3+ doping can enhance crystal performance, higher Tm3+ concentrations do not necessarily equate to better performance. With increasing Tm3+ doping concentration, ions in the laser upper level undergo reverse cross-relaxation, leading to photon quenching. Additionally, heavily doped crystals generate a large number of defects during growth, which also contribute to laser quenching. When doped with Tm3+ and La3+ ions, these two ions and F− ions will take up the space of the lattice at random, and co-doped crystals exhibit fewer clusters and defects compared to singly doped crystals, leading to the formation of a disordered lattice field, then ground state splitting and spectral broadening^[19,22]. Observing its spectral morphology can provide insights into the material’s potential for mode-locking to some extent. Combined with the above theory, we select Tm, La:CaF2 single crystal with appropriate doping concentration as the gain medium to study its performance in mode-locking.

Zhang et al. in 2018 successfully verified the highly efficient laser output of Tm,La:CaF2 single crystal, achieving a slope efficiency of nearly twice the theoretical quantum efficiency and a tunable spectral range as wide as 192 nm^[23]. Then, the passively and actively Q-switched Tm,La:CaF2 lasers were likewise successfully obtained separately by Zu et al. in 2019^[24] and Hao et al. in 2022^[25]. However, the CWML operation of Tm,La:CaF2 single crystal has not been reported yet. The possibility of Tm,La:CaF2 single crystal for the generation of ultrashort pulses with high repetition frequency needs to be exploited.

In this paper, we report, to the best of our knowledge, the first demonstration of a highly stable CWML Tm,La:CaF2 laser. The initial investigation was conducted in the continuous-wave (CW) regime, achieving a maximum output power of 427 mW at an absorbed pump power of 2.26 W with a slope efficiency of 20.4%. A self-starting passively CWML laser was subsequently obtained at 1886.5 and 1886.7 nm, exhibiting a significantly high extinction ratio of 75 dB. The laser generated a smooth pulse train with a pulse duration of 284.8 ps and a maximum average output power of 143 mW.

We designed a tightly focusing five-mirror X-shaped cavity, as illustrated in Fig. 1, to ensure the appropriate energy density to support the stable operation of CWML. The gain medium selected for this experiment was a 5.5-mm-thick 3% (atomic fraction) Tm3+ and 2% (atomic fraction) La3+ co-doped disordered CaF2 crystal with the cross-section size of 3 mm × 3 mm, grown by the temperature gradient method. Both sides of the crystal were plated with anti-reflection (AR) films for 792 and 1780–1980 nm in order to reduce Fresnel reflection loss. The single crystal, tightly mounted in a copper block, was water-cooled to maintain 12°C to mitigate the thermal load. The room temperature absorption and emission spectra of Tm,La:CaF2 crystal were demonstrated in our previous work^[23]. There is a strong absorption band at 792 nm corresponding to the H36−H34 energy level transition of Tm3+ ion. The pumping source was a fiber-coupled LD with a wavelength of 792 nm, a numerical aperture of 0.22, and a core diameter of 105 µm. The pumping laser was focused into the crystal through a collimation system with a focusing ratio of 1:1, ensuring the attainment of the appropriate mode volume. The absorption of the incident pump power by the Tm,La:CaF2 sample was about 30%.

When the oscillator worked in the CW state, the input mirror (M1) (R=−100mm) and other mirrors (M2−M4) (M2:R=−200mm, M3:R=−100mm, and M4:R=∞) were all AR coated at 780–810 nm and high-reflection (HR) coated at 1900–2000 nm. The output mirror (M5) was a plane mirror with the transmission of 2% for the oscillating laser. The CW laser output performance of the Tm,La:CaF2 crystal was shown in Fig. 2. The laser threshold was observed to be low when the absorbed pump power reached 160 mW and the slope efficiency was about 20.4%. When the absorbed pump power reached 2.26 W, the maximum output power was 427 mW, with optical-to-optical efficiency of 18.9%, at which point laser saturation was approached. The CW laser spectrum, centered at 1886.2 and 1886.5 nm, was scanned using an optical spectrum analyzer (MS3504i), as depicted in Fig. 3(a). The spectrum showed a narrow and double-peak structure mainly due to the inhomogeneous spectral splitting in the disordered crystals, the influence of water molecules, and the imperfect stability of the coating process.

To investigate the performance of the CWML, the semiconductor saturable absorber mirror (SESAM) (BATOP GmbH, SAM-1920-2-30 ps-4.0-25.4s-e) served as an excellent replacement for M4, positioned at the second beam waist generated by M3. The resonator length and mode of the laser will affect pulse characteristics. The length of the gain medium, one of the influencing factors, can impact the optical path length of the resonator, thereby affecting the laser’s mode and pulse frequency. By computing and simulating the relationship between the crystal length and cavity modes, the optimal cavity length for achieving the desired pulse state is determined. The total physical length of the cavity was 1.56 m, and the mode radius at the single crystal and the SESAM were calculated to be ∼56 and ∼34.5μm. Simultaneously, the small laser pattern facilitates low-threshold laser operation.

When the absorption of pump power exceeded 0.9 W, the unstable Q-switched ML state transitioned into a CWML state, with a minimum output power of 52 mW. Once established, the CWML state could restart itself each time without the need for adjustment and remained continuous for hours in the absence of external interference. As Fig. 2 illustrates, the maximum average output power of 143 mW and pulse energy of 1.5 nJ at 96.2 MHz pulse repetition frequency were obtained under an absorbed pump power of 1.73 W, corresponding to the maximum average fluence (Imax) of 3.94mJ/cm2 on the SESAM. Imax can be estimated by the following equation:Imax=Pmaxf·1+ROC1−ROC·1Aeff,A,(1)where Pmax, f, Roc, and Aeff,A are the maximum output power, repetition frequency, reflectivity of the output mirror, and effective laser mode area on the SESAM. The damage threshold of the SESAM is 4mJ/cm2, which is very close to Imax. Keeping on increasing the incident pump power, the CWML state will be destabilized and revert to the Q-switched ML state. At this point, a hasty increase of power is most likely to lead to damage to the SESAM. In subsequent research, to surpass the current maximum output power, options such as adjusting the spot size at the SESAM to increase its size moderately or exploring ML elements with higher damage thresholds may be considered.

As can be seen from Fig. 3, the CWML optical spectrum and autocorrelation trace were detected. It was observed that the output spectrum under the CWML state still owned double central peaks, with a slight red shift to 1886.5 and 1886.7 nm. Although a spectral broadening phenomenon was observed, it became less prominent after data processing. The essential parameter, pulse width, was measured by a high dynamic range autocorrelator (A.P.E. GmbH, pulseCheck SM). By assuming a Gaussian-fit envelope profile, the pulse duration was found to be 284.8 ps. The corresponding time-bandwidth product was calculated to be 9.6. The wide pulse width obtained is mainly limited by several aspects. First, dispersion compensation is not applied to the cavity. Second, the SESAM has a small saturable fluence (45μJ/cm2) and a slow recovery time (30 ps). The shorter the recovery time is, the narrower the CWML pulse width that can be obtained^[26]. In addition, the output wavelength is centered at ∼1886nm, which will be directly impacted by the humidity of the ultra-clean room, especially during the summer rainy season. Meantime, the band is at the edge of the SESAM’s high-reflection band (1880–1980 nm), creating a huge obstacle in realizing wide-band ultrafast mode-locking. Hence, it is anticipated that the pulse width of the Tm,La:CaF2 single crystal could be significantly reduced through the preparation of more parameterized SESAM devices, combined with a more proficient technical approach by experimentalists.

Using the 90/10 knife-edge method, we analyzed the diffracted beam quality in the two orthogonal directions (Mx2=1.42, My2=1.43), shown in Fig. 4. The ML beam profile and 3D intensity distribution were measured by a mid-infrared broadband phasor-type beam quality analyzer (WinCamD-IR-BB, 2–16 µm, USA), as shown in the inset of Fig. 4. Under the combination of the astigmatism of multiple cavity mirrors and the presence of measurement errors that cannot be ruled out, M2 factors showed only slight differences in x and y directions.

Thereafter, to evaluate the stability of the CWML laser, the temporal waveforms of real-time pulse trains were recorded and monitored by a high-speed photodetector (EOT, ER-5000, rise/fall time of 28 ps, >110GHz) and a digital oscilloscope (Tektronix DPO4104, 1 GHz), as displayed in Fig. 5(a). The good pulse-to-pulse stability of the CWML laser was evidenced by the clean individual pulse at nanosecond time scale as well as the smooth and neat periodic pulse sequences over 20 ms/div wide time scale. To further verify the high stability of the CWML laser, a spectrum analyzer (ROHDE & SCHWARZ-FSC) was selected to measure the radio frequency spectrum (RFS) and signal-to-noise ratio (SNR). As the lab results showed, the clean fundamental peak at 96.2 MHz without side peaks was in perfect agreement with the round-trip time of the cavity, demonstrating an excellent CWML state without the accompanying Q-switched instability. The SNR was up to 75 dB, indicating a pure mode-locking with a minimal CW component present. In addition, the single-frequency ML state had also been confirmed by the comb-like spectrum at a broad scan range (1 GHz), as depicted in Fig. 5(c). All of the aforementioned observations indicate that we have achieved a CWML laser characterized by extremely high stability in pulse energy fluctuation at the nearly 1.9 µm wavelength band. This is crucial for applications requiring ultrashort lasers within the 1.8–1.9 µm spectral window.

In conclusion, the first CWML Tm,La:CaF2 laser has been demonstrated, to the best of our knowledge. For the CW mode, the maximum output power was 427 mW with 20.4% slope efficiency. By replacing the end mirror with a SESAM, we obtained a highly stable CWML laser, delivering a laser pulse with 284.8 ps width at wavelengths of 1886.5 and 1886.7 nm. The SNR is up to 75 dB, which can represent a prominent level available in LD-pumped crystalline single-doped Tm3+ CWML lasers under the fundamental frequency of 96.2 MHz. We have succeeded in making Tm,La:CaF2 lasers capable of realizing high-frequency mode-locking in the picosecond scale in the band of 1.9 µm.