CrPS4 Nanoflakes as Stable Direct-Band-Gap 2D Materials for Ultrafast Pulse Laser Applications

Two-dimensional (2D) materials have attracted considerable attention due to their potential for generating ultrafast pulsed lasers. Unfortunately, the poor stability of most layered 2D materials under air exposure leads to increased fabrication costs; this has limited their development for practical applications. In this paper, we describe the successful preparation of a novel, air-stable, and broadband saturable absorber (SA), the metal thiophosphate CrPS4, using a simple and cost-effective liquid exfoliation method. The van der Waals crystal structure of CrPS4 consists of chains of CrS6 units interconnected by phosphorus. In this study, we calculated the electronic band structures of CrPS4, revealing a direct band gap. The nonlinear saturable absorption properties, which were investigated using the P-scan technique at 1550 nm, revealed that CrPS4-SA had a modulation depth of 12.2% and a saturation intensity of 463 MW/cm2. Integration of the CrPS4-SA into Yb-doped fiber and Er-doped fiber laser cavities led to mode-locking for the first time, resulting in the shortest pulse durations of 298 ps and 500 fs at 1 and 1.5 µm, respectively. These results indicate that CrPS4 has great potential for broadband ultrafast photonic applications and could be developed into an excellent candidate for SA devices, providing new directions in the search for stable SA materials and for their design.


Introduction
Ultrafast laser pulses can be generated by converting a continuous laser wave into a short pulse train via a mode-locking method that has attracted considerable interest from various fields of science and technology, including material processing, the semiconductor industry, and advanced instrumentation [1][2][3][4][5][6]. A saturable absorber (SA) is a vital component of mode-locking technology. An SA generates ultrafast pulsed lasers through its nonlinear optical properties, which can periodically modulate the circulated light field in the laser cavity and thus satisfy the ever-growing demand for multiple technological applications [7][8][9][10]. As a result of the sustained efforts of scientists to discover SA materials with valuable properties, such as the semiconductor saturable-absorber mirror (SESAM) [11,12], many two-dimensional (2D) layered materials have now been considered as alternative systems [13][14][15][16]. 2D materials are characterized by their chemical diversity and structural complexity, as well as their unique optical and magnetic properties [17][18][19]. Moreover, semiconductor materials with direct band gaps have become important in a range of technologies such as solar cells and lasers. Their strong light absorption and ability to facilitate efficient light emission make them highly desirable for these applications [20][21][22][23][24]. As a result, several direct-band-gap semiconductors have been developed, including black phosphorus (BP) [25][26][27] and transition metal dichalcogenides (TMDs), for designing modelocked fiber lasers [28] such as WSe 2 and MoS 2 [29,30]. BP has gained a great deal of

Results and Discussion
Bulk CrPS 4 crystallized in the non-centrosymmetric monoclinic space group C2 (No. 5) at room temperature. The structure of layered CrPS 4 is shown in Figure 1a. The crystal structure was composed of distorted CrS 6 octahedra and PS 4 tetrahedra connected into a 2D double layer through chemical covalent bonds in a horizontal direction and vdWs forces along the c axis. The sheet thickness was 3.69 Å. Structurally, according to previous reports, the weak vdWs gap between layers was about 2.46 Å; this means that the few-layer form could be readily obtained from bulk samples by the LPE method. More importantly, the multibonded crystal structure of CrPS 4 , in contrast to the lone-pair electrons found in BP, gave it the potential for air stability. It is commonly known that layered BP possesses a honeycomb structure in which a phosphorus atom is covalently bonded to three neighboring atoms through their p-orbitals, exposing a pair of lone-pair electrons (Figure 1b). The lone pairs of phosphorus atoms can readily react with oxygen to form P x O y ; this ultimately Nanomaterials 2023, 13, 1128 3 of 14 leads to the formation of phosphoric acid and H 2 O, and thereby presents a significant obstacle to the use of BP in applications involving flexible electronics and photoelectronics. Nanomaterials 2023, 13, x FOR PEER REVIEW 3 of 15 possesses a honeycomb structure in which a phosphorus atom is covalently bonded to three neighboring atoms through their p-orbitals, exposing a pair of lone-pair electrons ( Figure 1b). The lone pairs of phosphorus atoms can readily react with oxygen to form PxOy; this ultimately leads to the formation of phosphoric acid and H2O, and thereby presents a significant obstacle to the use of BP in applications involving flexible electronics and photoelectronics. To gain deeper insights into the band-structure properties of CrPS4, the CASTEP mode in Material Studio software was utilized to perform DFT calculations of the band structure and density of states (DOS) of both bulk and monolayer forms of CrPS4 [48]. The calculated electronic band structures of bulk and monolayer CrPS4 are illustrated in Figure  2b,c; these indicate that CrPS4 is a direct-band-gap semiconductor, in agreement with previous research [49]. From Figure 2b, it can be seen that the bulk CrPS4 nanoflakes exhibited a band-gap energy of 0.97 eV. As the number of layers decreased, the energy of the band gap increased. When the CrPS4 nanoflakes were monolayered, the band-gap energy increased to about 2.15 eV. As illustrated in Figure 2d, the DOS plots reveal significant spin splitting in the d-orbitals of Cr atoms, indicating the presence of a sizable spin polarization in the bulk system. To gain deeper insights into the band-structure properties of CrPS 4 , the CASTEP mode in Material Studio software was utilized to perform DFT calculations of the band structure and density of states (DOS) of both bulk and monolayer forms of CrPS 4 [48]. The calculated electronic band structures of bulk and monolayer CrPS 4 are illustrated in Figure 2b,c; these indicate that CrPS 4 is a direct-band-gap semiconductor, in agreement with previous research [49]. From Figure 2b, it can be seen that the bulk CrPS 4 nanoflakes exhibited a band-gap energy of 0.97 eV. As the number of layers decreased, the energy of the band gap increased. When the CrPS 4 nanoflakes were monolayered, the band-gap energy increased to about 2.15 eV. As illustrated in Figure 2d, the DOS plots reveal significant spin splitting in the d-orbitals of Cr atoms, indicating the presence of a sizable spin polarization in the bulk system.
Using the LPE method, few-layer CrPS 4 nanoflakes were exfoliated simply and effectively. Subsequently, we demonstrated the powder X-ray diffraction (XRD) patterns of CrPS 4 and its samples after exfoliation treatment by XRD spectroscopy, as illustrated in Figure 3a. It can be seen that the positions of the diffraction peaks on the two patterns were in almost perfect agreement with a standard PDF card of CrPS 4 (PDF #30-0411), indicating the integrity of the structure after exfoliated CrPS 4 nanoflakes were obtained. In addition, the diffraction pattern from the CrPS 4 sample was analyzed using X'pert Highscore Plus 3.0 [50]. The characterization of the diffraction pattern occurred in the range of 2θ = 10-70 • , as shown in Figure S1. The diffraction pattern from the sample matched well with the diffraction pattern from the previous report. The reference code of CrPS 4 is 00-033-0404 [51]. The vibrational modes of the CrPS 4 were verified by the Raman spectrum and detected in the range 200-700 cm −1 (excitation wavelength: 532 nm, inVia, Renishaw, Wotton-under-Edge, UK) at room temperature. As shown in Figure 3b, about 13 vibration peaks were found in the Raman spectrum; this was in agreement with previously reported results [46], and further confirmed the rationality of the structure of the CrPS 4 . The surface morphology of the samples as exfoliated was then analyzed via scanning electron microscope (SEM, JSM-5910LV, JEOL, Tokyo, Japan). As depicted in Figure 3c, the CrPS 4 nanoflakes exhibited an obvious layered structure, indicating that the nanoflakes were successfully fabricated based on the LPE method. The morphology of the CrPS 4 nanoflakes was also tested using an atomic force microscope (AFM, MFP-3D Infinity, Asylum Research, Oxford, UK), which can observe the lateral size of nanoflakes. Figure 3d,e show CrPS 4 nanoflakes with an average Nanomaterials 2023, 13, 1128 4 of 14 thickness of~25 nm. The linear optical transmission spectrum of exfoliated CrPS 4 was also detected using a UV-vis-NIR spectrophotometer (LAMBDA, Perkin Elmer Inc., Waltham, MA, USA). We found that, at wavelengths of 1030 and 1530 nm, the transmittances were approximately 91.7% and 93.3%, respectively ( Figure 3f). BP has attracted tremendous interest because of its natural layer-dependent direct-band-gap energy, but layered BP is unstable and degrades rapidly in ambient conditions within hours. Calculated results indicated that CrPS 4 is also a direct-band-gap semiconductor, and this made us keen to discover if it was stable in ambient conditions for long time periods. Firstly, taken together, the powder XRD and Raman spectroscopy studies clearly demonstrated that the CrPS 4 structure remained stable after months in ambient conditions. As shown in Figure S1a, the intensity and peak position of CrPS 4 , both as exfoliated and after~2 months in air, remained essentially unchanged. The intensity and peak position of the Raman modes of CrPS 4 also remained essentially unchanged for~2 months, indicating the air-stability of CrPS 4 (shown in Figure 3b). The CrPS 4 nanoflakes were then analyzed via energy-dispersive X-ray spectroscopy (EDS, Oxford Instruments, Oxford, UK). As depicted in Figure S1c,d, an average Cr/P/S molar ratio of 1.0:1.0:4.0 was recorded for CrPS 4 nanoflakes as exfoliated and again after~2 months in air, further confirming the air-stability of the samples. To more intuitively demonstrate this stability after exposure to air for about 2 months, the surface morphology of the sample was analyzed via SEM. As can be seen in Figure S1c, the CrPS 4 nanoflakes continued to exhibit an obvious layered structure. Using the LPE method, few-layer CrPS4 nanoflakes were exfoliated simply and effectively. Subsequently, we demonstrated the powder X-ray diffraction (XRD) patterns of CrPS4 and its samples after exfoliation treatment by XRD spectroscopy, as illustrated in Figure 3a. It can be seen that the positions of the diffraction peaks on the two patterns were in almost perfect agreement with a standard PDF card of CrPS4 (PDF #30-0411), indicating the integrity of the structure after exfoliated CrPS4 nanoflakes were obtained. In addition, the diffraction pattern from the CrPS4 sample was analyzed using X'pert Highscore Plus 3.0 [50]. The characterization of the diffraction pattern occurred in the range of 2ϴ = 10-70°, as shown in Figure S1. The diffraction pattern from the sample matched well with the diffraction pattern from the previous report. The reference code of CrPS4 is 00-033-0404 [51]. The vibrational modes of the CrPS4 were verified by the Raman spectrum and detected in the range 200-700 cm −1 (excitation wavelength: 532 nm, inVia, Renishaw, Wotton-under-Edge, UK) at room temperature. As shown in Figure 3b, about 13 vibration peaks were found in the Raman spectrum; this was in agreement with previ- The obtained few-layer CrPS 4 nanoflakes were then dripped onto the D-shaped fiber to form an SA device. To investigate the nonlinear optical properties of the CrPS 4 -SA, a dual-channel balanced detection measurement system based on an erbium-doped fiber laser (1550 nm, 100 fs, 8.05 MHz) was employed, as shown in Figure 4 inset. Equation (1) only considers the case of single-photon absorption, and the nonlinear saturated absorption curve of CrPS 4 -SA was obtained after fitting [52,53].
CrPS4 nanoflakes were then analyzed via energy-dispersive X-ray spectroscopy (EDS ford Instruments, Oxford, UK). As depicted in Figure S1c and S1d, an average Cr/P/S lar ratio of 1.0:1.0:4.0 was recorded for CrPS4 nanoflakes as exfoliated and again aft months in air, further confirming the air-stability of the samples. To more intuit demonstrate this stability after exposure to air for about 2 months, the surface morpho of the sample was analyzed via SEM. As can be seen in Figure S1c, the CrPS4 nanof continued to exhibit an obvious layered structure. The obtained few-layer CrPS4 nanoflakes were then dripped onto the D-shaped to form an SA device. To investigate the nonlinear optical properties of the CrPS4-S dual-channel balanced detection measurement system based on an erbium-doped laser (1550 nm, 100 fs, 8.05 MHz) was employed, as shown in Figure 4 inset. Equatio only considers the case of single-photon absorption, and the nonlinear saturated abs tion curve of CrPS4-SA was obtained after fitting [52,53].  In Equation (1), T(I) is the transmission rate, ∆ is the modulation depth (MD), I is the input intensity, is the saturated intensity, and is the nonsaturable loss (NL). The fitting results are shown in Figure 4. Values of MD, , and NL were found to be ~12.2 %, ~463 MW/cm 2 , and ~25.3 %, respectively.
Next, to validate the excellent potential of the layered CrPS4 for ultrafast laser applications, we constructed 1.0 µm and 1.5 µm all-fiber laser cavities using Er-doped or Ybdoped fibers. The structural diagram of the optical fiber laser structure is shown in Figure  5.  In Equation (1), T(I) is the transmission rate, ∆T is the modulation depth (MD), I is the input intensity, I sat is the saturated intensity, and T ns is the nonsaturable loss (NL). The fitting results are shown in Figure 4. Values of MD, I sat , and NL were found to be~12.2%, 463 MW/cm 2 , and~25.3%, respectively.
Next, to validate the excellent potential of the layered CrPS 4 for ultrafast laser applications, we constructed 1.0 µm and 1.5 µm all-fiber laser cavities using Er-doped or Yb-doped fibers. The structural diagram of the optical fiber laser structure is shown in Figure 5. The laser cavity consisted of a section of Er-doped or Yb-doped fiber, a laser diode (980 nm Pump Laser, Hanyu, Shanghai, China), a wavelength division multiplexer (WDM, Mingchuang, Shenzhen, China), an optical coupler (OC, Mingchuang, Shenzhen, China), a polarization-independent optical isolator (ISO, Mingchuang, Shenzhen, China), a polarization controller (PC, General Photonics, Losa Angeles, US), and a D-shaped fiber based on CrPS 4-SA. ISO and PC were used to ensure the unidirectional propagation of light and to adjust its polarization state, respectively. The evanescent field length of the D-shaped fiber optic bare leak was 10 mm, and the distance between the surface and the core was 1 µm. The Er-doped fiber laser cavity length was 21. Importantly, we performed pre-experiments to demonstrate that, in the absence of SA, no mode-locking occurred in the laser cavity, regardless of PC modulation or pump power. By such means, we confirmed the authenticity of the experiment. After SA was applied to the laser cavity, we adjusted the PC while continuously increasing the pump power, and observed the output waveform of the oscilloscope. We achieved continuouswave mode-locking (CWML) when the pump power was higher than 170 wm. When the pump power was increased to 300 mW, the output power was 10.64 mW, and pulse energy and peak power were 1.174 nJ and 3.94 W, respectively. The mode-locking sequence diagram is presented in Figure 6a, which shows a 110.4 ns time interval between adjacent pulses, which corresponds well to the pulse repetition rate of 9.05 MHz. The inset in Figure 6a shows the amplitude intensity plot of the mode-locked pulse sequence; this indicates that the mode-locked sequence existed stably for a long time. The spectrum with a central wavelength of 1036.1 nm had a 3 dB spectral width of 0.84 nm, as shown in Figure  6b. The autocorrelation trace corresponding to the measured pulse at this time is shown Importantly, we performed pre-experiments to demonstrate that, in the absence of SA, no mode-locking occurred in the laser cavity, regardless of PC modulation or pump power. By such means, we confirmed the authenticity of the experiment. After SA was applied to the laser cavity, we adjusted the PC while continuously increasing the pump power, and observed the output waveform of the oscilloscope. We achieved continuous-wave mode-locking (CWML) when the pump power was higher than 170 wm. When the pump power was increased to 300 mW, the output power was 10.64 mW, and pulse energy and peak power were 1.174 nJ and 3.94 W, respectively. The mode-locking sequence diagram is presented in Figure 6a, which shows a 110.4 ns time interval between adjacent pulses, which corresponds well to the pulse repetition rate of 9.05 MHz. The inset in Figure 6a shows the amplitude intensity plot of the mode-locked pulse sequence; this indicates that the mode-locked sequence existed stably for a long time. The spectrum with a central wavelength of 1036.1 nm had a 3 dB spectral width of 0.84 nm, as shown in Figure 6b. The autocorrelation trace corresponding to the measured pulse at this time is shown in Figure 6c. It can be seen that the pulse width is about 298 ps, and the time-bandwidth product (TBP) is 69.9, indicating that the pulse has a serious chirp. In Figure 6d, a signal-tonoise ratio (SNR) measurement of approximately 55.3 dB can be observed with a higher signal peak at a laser cavity repetition rate of 9.05 MHz. The relationship between the mode-locked output power and pump power is shown in Figure 6e. When the pump power was 130 mW, the laser cavity output a continuous wave (CW). When the pump power was increased to 170 mW, the output was CWML, and the measured slope efficiency was about 4.9%. Subsequently, we measured the output spectrum of the laser cavity over a longer timescale, at time intervals of 1 h, for a total of 6 h. As shown in Figure 6f, the long-term spectrum was quite stable, indicating that a ytterbium-doped laser has the potential of highly stable operation. In order to verify that CrPS4 could work in a wide wavelength range, we place additionally prepared CrPS4-SA into an erbium-doped fiber laser cavity for debug Stable CWML output was achieved by adjusting the PC when the pump power was a 150 mW. The output power, pulse energy, and peak power were 6.1 mW, 0.893 nJ 1786 W, respectively, when the pump power was 270 mW. The output pulse charac tics are shown in Figure 7. The mode-locked pulse sequence is shown in Figure 7a pulse period was 146.4 ns, which corresponded to a pulse repetition rate of 6.83 MHz inset shows that the mode-locking was quite stable. Figure 7b shows the laser spec centered at 1531.6 nm with a 3 dB bandwidth of 5.6 nm. A measured pulse width o fs resulted in a TBP of 0.35, as shown in Figure 7c. It can be clearly observed that the value of the high signal at the mode-locked repetition frequency was 6.83 MHz, an signal-to-noise ratio was about 64 dB, as shown in Figure 7d. Figure 7e shows tha average output power varied linearly with increasing pump power, with a slope effic of 2.2 %. Subsequently, we also measured the spectrum for 6 h, at time intervals o and found that the erbium-doped laser mode-locking was very stable, as shown in F 7f. In order to verify that CrPS 4 could work in a wide wavelength range, we placed the additionally prepared CrPS 4 -SA into an erbium-doped fiber laser cavity for debugging. Stable CWML output was achieved by adjusting the PC when the pump power was above 150 mW. The output power, pulse energy, and peak power were 6.1 mW, 0.893 nJ, and 1786 W, respectively, when the pump power was 270 mW. The output pulse characteristics are shown in Figure 7. The mode-locked pulse sequence is shown in Figure 7a. The pulse period was 146.4 ns, which corresponded to a pulse repetition rate of 6.83 MHz. The inset shows that the mode-locking was quite stable. Figure 7b shows the laser spectrum centered at 1531.6 nm with a 3 dB bandwidth of 5.6 nm. A measured pulse width of 500 fs resulted in a TBP of 0.35, as shown in Figure 7c. It can be clearly observed that the peak value of the high signal at the mode-locked repetition frequency was 6.83 MHz, and the signal-to-noise ratio was about 64 dB, as shown in Figure 7d. Figure 7e shows that the average output power varied linearly with increasing pump power, with a slope efficiency of 2.2%. Subsequently, we also measured the spectrum for 6 h, at time intervals of 1 h, and found that the erbium-doped laser mode-locking was very stable, as shown in Figure 7f. To demonstrate the stability of CrPS4, we placed the previously prepared satu absorber back into an ytterbium-doped fiber laser, 40 days after the first experiment found that the laser could still output mode-locked pulses after adjusting the pump PC. The mode-locking output results are shown in Figure S3. The threshold of modeing was 180 mW, a slight increase (of 10 mW) compared to the first experiment. Fi S3a shows the pulse sequence with a pulse interval of 107.2 ns. The inset shows tha laser continued to work with high stability. Figure S3b shows that, in comparison the first experiment, the central wavelength of the mode-locked spectrum remained same, at 1036.1 nm, while the 3 dB bandwidth changed only slightly, from 0.84 nm t nm. The measured mode-locking pulse width was 400 ps, as shown in Figure S3c, w was wider than the first measurement. The repetition frequency of the mode-lo strong signal peak was 9.32 MHz, and the SNR was about 52.6 dB, as shown in Figure  The relationship between average output power and pump power is shown in Figure  and the slope efficiency was 3.6 %, which was slightly lower than the first time. We measured the spectrum for 6 h, and the results are shown in Figure S3f. The spec remained unchanged, indicating that the laser worked stably, and that the CrPS4 satu absorber we made exhibited high stability.
Once again, 40 days after the first experiment, we put the previously used satu absorber back into the erbium-doped fiber laser, using the adjustment method descr above. To achieve CWML, pump power needed to be increased to more than 150 mw mode-locking output characteristics are shown in Figure S4. Figure S4a shows tha pulse interval was 136.9 ns, and the inset shows that the laser cavity remained in a s state. Figure S4b shows that the central wavelength of the spectrum was 1531.4 nm, the 3dB bandwidth was 6.2 nm, which was stable compared with the first experim Figure S4c shows that the mode-locked pulse width was 594 fs, which was slightly w than the first experiment, with a TBP of 0.47. The signal-to-noise ratio of a strong s peak at a repetition frequency of 7.33 MHz was approximately 69 dB. Figure S4e sh that the slope efficiency between the average output power and the pump power 2.4%. Finally, we measured the long-term spectral changes, which indicated the exce stability of the laser cavity, as shown in Figure S4f.
We then compared the data of the two experiments, as set out in Table 1. Altho the two experiments were separated by 40 days, we found few differences in the res indicating that the CrPS4-SA we prepared had excellent stability. However, in ord more intuitively observe any variation in the experimental results, we drew a coeffi To demonstrate the stability of CrPS 4 , we placed the previously prepared saturable absorber back into an ytterbium-doped fiber laser, 40 days after the first experiment, and found that the laser could still output mode-locked pulses after adjusting the pump and PC. The mode-locking output results are shown in Figure S3. The threshold of mode-locking was 180 mW, a slight increase (of 10 mW) compared to the first experiment. Figure S3a shows the pulse sequence with a pulse interval of 107.2 ns. The inset shows that the laser continued to work with high stability. Figure S3b shows that, in comparison with the first experiment, the central wavelength of the mode-locked spectrum remained the same, at 1036.1 nm, while the 3 dB bandwidth changed only slightly, from 0.84 nm to 0.8 nm. The measured mode-locking pulse width was 400 ps, as shown in Figure S3c, which was wider than the first measurement. The repetition frequency of the mode-locked strong signal peak was 9.32 MHz, and the SNR was about 52.6 dB, as shown in Figure S3d. The relationship between average output power and pump power is shown in Figure S3e, and the slope efficiency was 3.6%, which was slightly lower than the first time. We then measured the spectrum for 6 h, and the results are shown in Figure S3f. The spectrum remained unchanged, indicating that the laser worked stably, and that the CrPS 4 saturable absorber we made exhibited high stability.
Once again, 40 days after the first experiment, we put the previously used saturable absorber back into the erbium-doped fiber laser, using the adjustment method described above. To achieve CWML, pump power needed to be increased to more than 150 mw. The mode-locking output characteristics are shown in Figure S4. Figure S4a shows that the pulse interval was 136.9 ns, and the inset shows that the laser cavity remained in a stable state. Figure S4b shows that the central wavelength of the spectrum was 1531.4 nm, and the 3dB bandwidth was 6.2 nm, which was stable compared with the first experiment. Figure S4c shows that the mode-locked pulse width was 594 fs, which was slightly wider than the first experiment, with a TBP of 0.47. The signal-to-noise ratio of a strong signal peak at a repetition frequency of 7.33 MHz was approximately 69 dB. Figure S4e shows that the slope efficiency between the average output power and the pump power was 2.4%. Finally, we measured the long-term spectral changes, which indicated the excellent stability of the laser cavity, as shown in Figure S4f.
We then compared the data of the two experiments, as set out in Table 1. Although the two experiments were separated by 40 days, we found few differences in the results, indicating that the CrPS 4 -SA we prepared had excellent stability. However, in order to more intuitively observe any variation in the experimental results, we drew a coefficient of Nanomaterials 2023, 13, 1128 9 of 14 variation diagram to represent the degree of dispersion of the pulse parameters of the two experiments, as shown in Figure 8. It can be seen that the coefficients of variation of the pulse width, output power, pulse energy, and coefficients of variation of the Yb-doped fiber laser were close to 30%, while the coefficient of variation of the peak power was as high as 59%. For Er-doped fiber lasers, the coefficients of variation of its pulse parameters were all less than 20%; among these, the coefficients of variation of pulse energy, peak power and slope efficiency were all less than 10%. In short, the output variation of the erbium-doped fiber laser cavity was lower, and this finding is related to the performance of the laser cavity itself. The ytterbium-doped fiber laser cavity worked in the total positive dispersion region, while the erbium-doped laser cavity worked in the anomalous dispersion region, and the optical solitons formed through the balance of dispersion and nonlinear effects were more stable. of variation diagram to represent the degree of dispersion of the pulse parameters of the two experiments, as shown in Figure 8. It can be seen that the coefficients of variation of the pulse width, output power, pulse energy, and coefficients of variation of the Yb-doped fiber laser were close to 30%, while the coefficient of variation of the peak power was as high as 59%. For Er-doped fiber lasers, the coefficients of variation of its pulse parameters were all less than 20%; among these, the coefficients of variation of pulse energy, peak power and slope efficiency were all less than 10%. In short, the output variation of the erbium-doped fiber laser cavity was lower, and this finding is related to the performance of the laser cavity itself. The ytterbium-doped fiber laser cavity worked in the total positive dispersion region, while the erbium-doped laser cavity worked in the anomalous dispersion region, and the optical solitons formed through the balance of dispersion and nonlinear effects were more stable.  The pulse performance of the laser has a crucial influence on the application. In Table  S1, we summarize the performance of mode-locked ytterbium-doped lasers for several representative 2D materials. It can be seen that the output of these ytterbium-doped lasers was in the order of picoseconds. In contrast, the output of our CrPS4 ytterbium-doped fiber laser was 298 ps. These results show that CrPS4 has higher generation efficiency for ultra-short pulse output, and has certain advantages in terms of pulse width. The maximum output power of our laser was 10.63 mW, which is 28.7 times that of the graphene previously reported by Zhao et al. [54], and 1.38 times that of the Mo2C previously reported by Liu et al. [55]. The single-pulse energy of most YDF lasers using 2D materials as SA is usually limited to below 1 nJ; however, our laser achieved an output of 1.174 nJ, with a peak power of 3.94 W, which is 5.5 times that of the WS2 reported by Mao et al. [56], and 2.3 times that of the NiPS3 reported by Liu et al. [56]. In a similar way, we compared The pulse performance of the laser has a crucial influence on the application. In Table S1, we summarize the performance of mode-locked ytterbium-doped lasers for several representative 2D materials. It can be seen that the output of these ytterbium-doped lasers was in the order of picoseconds. In contrast, the output of our CrPS 4 ytterbiumdoped fiber laser was 298 ps. These results show that CrPS 4 has higher generation efficiency for ultra-short pulse output, and has certain advantages in terms of pulse width. The maximum output power of our laser was 10.63 mW, which is 28.7 times that of the graphene previously reported by Zhao et al. [54], and 1.38 times that of the Mo 2 C previously reported by Liu et al. [55]. The single-pulse energy of most YDF lasers using 2D materials as SA is usually limited to below 1 nJ; however, our laser achieved an output of 1.174 nJ, with a peak power of 3.94 W, which is 5.5 times that of the WS 2 reported by Mao et al. [56], and 2.3 times that of the NiPS 3 reported by Liu et al. [56]. In a similar way, we compared erbium-doped fiber lasers based on other 2D slave materials, as shown in Table S2. As can be readily observed, because of the remarkable nonlinear optical properties of these 2D materials, their modulation depths vary in percentage terms from low single digits to many tens. In contrast, the modulation depth of our prepared CrPS 4 was 12.2%, which is higher than the 10.9% for BP reported by Chao et al. [57] and the 5.1% for Mo 2 C reported by Liu et al. [55]. At the same time, our pulse output was 500 fs, which is similar to that achieved by other lasers. In addition, our output reached a level of 6.1 mW, which is twice that of the graphene reported by Bao et al. [7], and 2.3 times that of the BP reported by Chao et al. [57]. The single pulse energy and peak power were 893 pJ and 1786 W, respectively. Not only did we measure the stability of the spectrum over 6 h, we also re-experimented with the previously fabricated saturable absorber 40 days later and found that it still achieved mode locking. These experimental results confirm that CrPS 4 is a competitive direct-band-gap material with excellent nonlinear optical modulation properties and great potential for broadband ultrafast photonics applications.

Conclusions
In conclusion, we fabricated high-quality CrPS 4 -SA by the LPE method. Theoretical calculations of the electronic band structures of CrPS 4 revealed a direct band gap. We studied the applications of few-layer CrPS 4 -SA in ultrafast photonics for the first time. The saturated intensity and modulation depth of CrPS 4 -SA were 463 MW/cm 2 and 12.2%, respectively, at 1.5 µm. Moreover, based on the excellent saturable absorption of the Dshaped CrPS 4 SA, the pulse characteristics of fiber lasers operating in conventional soliton states were measured. We successfully obtained picosecond mode-locked pulses of 298 ps and ultrashort femtosecond pulses of 500 fs in the 1 µm and 1.5 µm regions, respectively. The signal-to-noise ratio (SNR) of the mode-locked operation was as high as 55.3 dB at 9.05 MHz (YDFL), and 64 dB at 6.83 MHz (EDFL). More importantly, the few-layer CrPS 4 exhibited excellent stability during exposure to air for a period of time. Our experimental results show that CrPS 4 is an air-stable and broadband SA, with promising potential for ultrafast laser applications.

Experimental Section
Fabrication of CrPS 4 . An LPE method was used to exfoliate few-layer CrPS 4 nanoflakes, in which the vdWs forced between the layers of CrPS 4 were broken by an ultrasonic wave. Firstly, commercially available high-purity CrPS 4 (Shenzhen six carbon) powder (about 23 mg) was ground in a mortar and dispersed into N-methyl-2-pyrrolidone (NMP, 30 mL), which was exfoliated in an ultrasonic cell disruptor for 20 h at power of 400 W. In order to make the solute to form nanoscale flakes, the solvent was ultrasonicated in an ultrasound cleaner for 24 h. Then, the mixture was centrifugally treated at a speed of 5000 rpm for 20 min to separate precipitation, and the few-layer CrPS 4 containing supernatant was obtained. All experimental procedures were conducted at room temperature (16.8 • C) and a relative humidity of 58%.
Characterization. The powder X-ray diffraction (XRD) patterns of CrPS 4 and its samples after exfoliated treatment were further demonstrated by XRD spectroscopy; the vibrational modes of the CrPS 4 were verified by Raman spectrum, and detected in the range 200-700 cm −1 (excitation wavelength: 532 nm, inVia, Renishaw, Wotton-under-Edge, UK) at room temperature. The surface morphology of the samples as exfoliated was analyzed via a scanning electron microscope. The morphology of CrPS 4 nanoflakes was also tested using an atomic force microscope and the linear optical transmission spectrum of exfoliated CrPS 4 was detected by an UV-vis-NIR spectrophotometer.
DFT calculation details. Here, the Vienna Ab initio Simulation Package (VASP, University of Vienna) was utilized to optimize the crystal structures and calculate electronic structures [58][59][60]. The exchange and correlative potentials of electron−electron interactions were accounted for using the generalized gradient approximation (GGA) within the Perdew−Burke−Eruzerhof (PBE) scheme [61,62]. More specifically, an energy cut-off of 500 eV and a Monkhorst-Pack Brillouin zone sampling grid [63] with a resolution 0.02 × 2π Å −1 were applied.