Optical materials with nonlinear properties have always been important for the field of nonlinear optics (NLO) since they can be used for practical applications such as optical switching, mode-locking, and optical limiting ^[1–9]. Optical nonlinearity means that the physical parameters of a material, such as absorptivity and refractive index, are electric field intensity dependent. In the past several decades, a number of materials have been found to possess significant nonlinear optical properties ^[10–12]. In particular, nanostructured materials have recently been a focus of research since they provide a flexible platform for the implementation of various functional electronic or photonic devices that are difficult to realize with traditional bulk-structured material platforms. Nanostructured nonlinear optical materials include carbon nanotubes (CNTs) ^[13,14], graphene ^[15–17], graphene oxide (GO) ^[18,19], topological insulators (TIs) ^[20–25], transition-metal dichalcogenides (TMDCs) ^{[6,8,26–45]}, black phosphorus (BPs) ^[46–48], gold nanoparticles ^[49,50], skutterudites ^[51,52], MXenes ^[53,54], and transition-metal monochalcogenides (TMMCs) ^[55]. It should be noted that nanostructured materials do not imply two-dimensional (2D) materials even if 2D materials are included in the material structure.

One family of recently discovered nanostructured materials consists of TMDCs ^[56]. TMDCs are materials with a formula of MX2, where M is a transition metal element (e.g., Mo, W, and Nb) and X is a chalcogen (e.g., S, Se, or Te). TMDCs have a layered X-M-X structure, wherein a layer of the transition metal atom is sandwiched between two layers of chalcogen atoms. The adjacent layers are weakly bonded together by van der Waals forces, and therefore thin sheets can be easily exfoliated from the bulk material. A number of investigations of the nonlinear optical properties of TMDCs have been conducted thus far, and their utilization in optical devices has been extensively demonstrated. For example, Zhang et al. reported the saturable absorption property of molybdenum disulfide (MoS2) for the first time ^[6], to the best of our knowledge. Two-photon absorption (TPA) of tungsten disulfide (WS2) and MoS2 was investigated by Zhang et al. ^[57]. The second and third-order optical susceptibilities of MoS2 were investigated by a number of research groups ^[58–60].

Rhenium diselenide (ReSe2) is a rhenium-based TMDC. Compared to commonly known molybdenum- or tungsten-based TMDCs, rhenium-based TMDCs such as ReS2 and ReSe2 have attracted less technical attention until recently, even though they possess electronic or optical properties comparable to the aforementioned common TMDCs ^[61–63]. Very recently, quite a few investigations of the nonlinear optical properties of ReS2 and ReSe2 have been conducted ^[34–42]. In particular, most of the investigations of ReSe2 have focused on saturable absorption. For example, Du et al. reported saturable absorption properties of ReSe2 for the first time, to the best of our knowledge, and demonstrated a ReSe2-based saturable absorber (SA) for a Q-switched fiber laser ^[39]. Li et al. systematically investigated nonlinear saturable absorption properties of ReSe2 at 1030 and 515 nm using an open-aperture (OA) Z-scan technique and demonstrated its use for picosecond mode-locking of a 1064 nm waveguide laser ^[40]. Investigations regarding ReSe2-based SAs have also been conducted by Li et al. ^[41,42].

In this paper, we report a detailed investigation of the nonlinear optical properties of ReSe2 at a wavelength of 1560 nm. This investigation was conducted with three-dimensional (3D) structured ReSe2, as opposed to 2D structured. First, we show experimental measurements of the nonlinear absorption coefficient and the nonlinear refractive index of a thin ReSe2 film with an OA and closed-aperture (CA) Z-scan technique. Note that unlike our measurements, the Z-scan measurement of ReSe2 by Li et al. was based on an OA technique to obtain limited information about saturable absorption properties at 1030 and 515 nm ^[40]. Our measurements are conducted with a femtosecond fiber laser at a wavelength of 1560 nm and show that ReSe2 has both saturable absorption and self-defocusing properties. Next, a fiberized SA based on ReSe2 is fabricated with a side-polished fiber platform. Finally, the fabricated SA is tested as an ultrafast mode-locker and was found to be capable of producing femtosecond pulses operating at a wavelength of 1560 nm. To the best of the author’s knowledge, there have been no previous reports of ReSe2-based SAs for use in the femtosecond mode-locking of a laser.

To fabricate a ReSe2 thin film, ReSe2 bulk flakes (99.99%, HQ graphene) were used as a starting material. Nanometer- or micrometer-sized ReSe2 particles were first obtained using bath-type ultrasonification in 10 mL of distilled water for 16 h. Note that no additional centrifugation step for elimination of large agglomerates was conducted in our case, since our target particles were not nano-sized particles. Figure 1(a) shows scanning electron microscopy (SEM) image of the obtained ReSe2 particles. Their sizes are observed to vary from tens of nanometers to around 2 μm. An energy-dispersive spectroscopy (EDS) measurement was also conducted, as shown in Fig. 1(b). For the EDS measurement, a small amount of solution containing ReSe2 particles was dropped and dried on top of a glass slide. The two strong peaks observed at 1.85 keV and 1.4 keV are associated with Re (Mα) and Se (Lα) ^[64], and the atomic ratio between Re and Se was approximately 1:2.

To form a thin film, polyvinyl alcohol (PVA) was mixed with the prepared ReSe2 solution. The concentration of ReSe2 in the ReSe2/PVA composite was ∼1mg·mL−1. An atomic force microscope (AFM) image of our prepared ReSe2 thin film has been measured, as shown in Fig. 2. The thickness of the prepared ReSe2 particles within the target range was measured to be ∼450nm. However, note that the thickness of prepared ReSe2 particles was random since no centrifugation process was used after ultrasonication. It was thus impossible to obtain a reasonable average thickness and layer number from a limited number of AFM-measured values.

To analyze the stoichiometry of the ReSe2/PVA thin film, X-ray photoelectron spectroscopy (XPS) measurements were performed. High-resolution spectra of the Re 4f and Se 3d region are shown in Figs. 3(a) and 3(b), respectively. The two peaks of Re 4f7/2 and Re 45/2 at ∼41.6 and ∼44eV are consistent with previously reported results ^[64,65]. Two peaks at 54.6 and ∼55.6eV in the Se-3d region of Fig. 3(b) are consistent with the peaks of Se 3d5/2 and Se 3d3/2 ^[64,65]. The measured high-resolution spectrum for the C 1s region confirms the existence of CH2 and CH–OH groups in the PVA ^[66]. The peaks centered at the binding energies of ∼284eV, ∼286eV, ∼288.7eV in the C 1s region can be attributed to CH2, CH–OH, and C═O groups, respectively ^[66,67].

Next, we measured the Raman spectrum of the ReSe2/PVA thin film at 532 nm (InVia Raman Microscope, Renishaw) [Fig. 4(a)]. Many small peaks were observed due to the complex lattice vibration and low symmetry of ReSe2 used ^{[41,42,64,65,68–70]}. Three main Raman peaks at ∼124, ∼159, and ∼173cm−1 indicate in-plane (Eg) and out-of-plane (Ag) vibrational modes ^{[41,42,64,65,68–70]}. Figure 4(b) shows the measured linear optical absorption of the ReSe2/PVA composite film over a spectral range of 1000–1800 nm, which was measured using a spectrophotometer (UV-3600PLUS, Shimadzu). For the measurement, the ReSe2/PVA composite solution was deposited onto a glass slide. A broad absorption band of the prepared ReSe2/PVA composite was clearly observed, which included absorption in the 1.5 μm wavelength band.

It was reported that the energy bandgap of bulk ReSe2 is around 0.995 eV (1246 nm), while that of a ReSe2 monolayer is 1.24 eV (1000 nm) ^[70,71]. From the perspective of the energy bandgap, it should thus be impossible to observe photon absorption from the ReSe2 particles at wavelengths above 1246 nm. However, as shown in Fig. 4(b), the optical absorption of the prepared ReSe2/PVA film was clearly observed in the 1.5 μm wavelength region. Such absorption behavior could be attributed to Re and Se atomic defects ^[40]. As mentioned in Ref. ^[40], the atomic defects could reduce the energy bandgap of ReSe2, allowing for broadband photon absorption irrespective of the energy bandgap of the pristine material. In fact, sub-bandgap absorption has been previously reported in other TMDC materials. It is well-known that the broken symmetry and unsatisfied bonds between M and X atoms at edges can modify the electronic structures of TMDCs ^[72].

The measurements for the nonlinear optical response of the ReSe2 particles were carried out by using the OA and CA Z-scan techniques ^[73]. The experimental setup is shown in Fig. 5(a). A homemade mode-locked fiber laser with a repetition rate of 22.26 MHz and a pulse width of 300 fs at a wavelength of 1560 nm was used as an input source. To obtain the nonlinear absorption coefficient, an OA Z-scan measurement was first performed with the aperture fully opened (S=1). The beam was focused through a plano-convex lens to obtain a peak intensity of about 50MW/cm2. The position of the sample was changed from z=−15mm to z=15mm with a step size of 0.5 mm. The spot size was estimated from the obtained curve, which was approximately 22 μm at the focal point. The average thickness of the prepared ReSe2/PVA film was approximately 9 μm. The formed ReSe2/PVA thin film had a circular shape with a diameter of ∼10mm, which is much larger than the beam waist of ∼22μm. The Z-scan setup was calibrated considering the effect of the substrate.

As shown in Fig. 5(b), the change of the transmittance of the sample was measured with a photodiode sensor; the normalized values of the measured data were plotted against the z position and fitted with the following formula ^[73,74]: where T(z) is the normalized transmittance, β is the nonlinear absorption coefficient, I0 is the peak on-axis intensity at the focus, Leff is the effective length, z is the position of the sample, and z0 is the Rayleigh length.

To determine the sign and magnitude of the nonlinear refractive index of the ReSe2 sample, a CA Z-scan measurement was subsequently conducted, while the aperture at the far field was kept closed (S∼0.7). In the CA Z-scan measurement, the transmitted beam is sensitive to both nonlinear absorption and nonlinear refraction. By dividing the normalized results of a CA Z scan by those of an OA Z scan, we can distinguish the nonlinear refraction from the nonlinear absorption ^[73]. The normalized values of the measured data were plotted against the z position and fitted with the following formula ^[75]: where ΔΦ=2πn2I0Leff/λ is the phase change, n2 is the nonlinear refractive index, Leff=(1−e−α0L)/α0 is the effective thickness of the sample, α0 is the linear absorption coefficient, L is the thickness of the sample, z0=πω02/λ is the Rayleigh length, ω0 is the beam waist, λ is the wavelength, and I0 is the intensity at the focusing point. The measured nonlinear absorption coefficient of the ReSe2/PVA film was −(5.67±0.35)×103cm/GW, and its nonlinear refractive index was −(2.81±0.13)×10−2cm2/GW.

The curves in Fig. 5(b) show the nonlinear saturable absorption and self-defocusing behaviors of the ReSe2 film at a wavelength of 1560 nm. Considering the bandgap of the pristine ReSe2 ^[70,71], the ReSe2 was supposed to exhibit TPA at 1560 nm; however, no TPA was observed in our ReSe2 sample ^[76,77]. On the other hand, our measurement results revealed that ReSe2 possesses a saturable absorption property, even for incident photons with energies lower than its bulk bandgap energy. The nonlinear saturable absorption of ReSe2 could be attributed to a defect-driven bandgap decrease or edge state absorption ^[72]. Note that TMDC materials with bandgap energies larger than the incident photon energy are known to exhibit such a saturable absorption behavior in the 1550 nm wavelength region ^[76,77]. Table 1 compares the Z-scan-measured nonlinear optical parameters of a variety of TMDCs with those of ReSe2. Note that the previous measurements for nonlinear absorption coefficients and the nonlinear refractive indices of TMDCs were conducted at wavelengths of 515, 800, 1030, or 1064 nm. Since our measurements were conducted at 1560 nm, it is impossible to directly compare our measured nonlinear optical parameters of ReSe2 to those of the other MDCs. It is believed that ReSe2 has nonlinear optical properties comparable to various TMDCs.

In this experiment, we have not observed TPA in our prepared ReSe2/PVA film. This could be attributed to the fact that our used pulse laser for the Z-scan measurements might not have a peak power high enough to induce TPA. Regarding the effect of a large nonlinear absorption coefficient β on the mode-locking performance, we believe that a large nonlinear absorption coefficient β of a nonlinear optical material is associated with a large modulation depth of the material-based SA. Note that the use of an SA with a larger modulation depth is known to increase the mode-locking stability as well as to reduce the mode-locked pulse width ^[80,81]. The relaxation time measurement of the ReSe2 sample was not conducted since no pump–probe spectroscopy setup was available in our laboratory.

Next, we fabricated a fiberized SA that can be used for ultrafast mode-locking based on ReSe2. A side-polished fiber was employed as a platform. The side-polished fiber was prepared through the polishing of one side of an SMF-28 fiber, which was fixed onto the V groove of a glass slide. The cross-sectional structure of the prepared SA is shown in Fig. 6(a). The insertion loss and polarization-dependent loss (PDL) of the prepared side-polished fiber were ∼0.8dB and ∼0.05dB at 1560 nm, respectively. The physical distance of the flat side to the edge of the fiber core was of ∼7μm, and the polishing length was ∼2.5mm. The ReSe2/PVA solution was deposited onto the flat side of the side-polished fiber using the solution-drop method and then dried at room temperature for 24 h. The insertion loss and PDL of the ReSe2/PVA-deposited side-polished fiber were ∼2.9dB and ∼1.6dB, respectively.

Nonlinear transmission curves of the ReSe2/PVA-based SA were measured as a function of the incident peak power of the input pulses for transverse electric (TE) and transverse magnetic (TM) modes, respectively. For this measurement, we used a mode-locked, 1.56 μm fiber laser with a temporal width of ∼680fs at a repetition rate of ∼22.26MHz, and the measurement setup is shown in Fig. 6(b). Since our SA had a non-negligible PDL of ∼1.6dB, a polarization controller (PC) was incorporated into the measurement setup. It was straightforward to distinguish the TE and TM modes by measuring the output power while the PC was adjusted. A variable optical attenuator (VOA) was used to adjust the optical power of the mode-locked pulses. A 50:50 coupler was used to split the mode-locked pulses into two ports. One of the two ports was connected to the prepared ReSe2/PVA-based SA, while the other was directly connected to a power meter to monitor the input beam power on the prepared SA. Another power meter was used to monitor the output power from the ReSe2/PVA-based SA for comparison with the input power. Figure 7 shows the nonlinear transmission curves for the TE and TM modes fitted to the following equation ^[82]: where T is the transmission, ΔT is the modulation depth, I is the incident pulse energy, Isat is the saturation energy, and Tns is the nonsaturable loss. The estimated modulation depth and saturation power for the TE mode are ∼3.9% and ∼42W, respectively, while they are ∼2.4% and ∼53W, respectively, for the TM mode. Due to the asymmetric structure of our prepared ReSe2/PVA-deposited side-polished fiber, the TE and TM modes have different interaction strengths between the ReSe2/PVA film and oscillating beam. This is the reason why the TE and TM modes exhibit different saturable absorption properties. Regarding the main mode for the actual mode-locking, we believe that the TE mode would play a key role in our laser setup since the TE mode experiences less insertion loss. It is believed that the modulation depth level is high enough to induce soliton mode-locking in the fiberized laser cavity with proper anomalous dispersion ^[81].

In order to measure the damage threshold of the proposed ReSe2/PVA-based SA, we have launched a 1550 nm continuous-wave amplified laser beam of 1 W power into it. However, we have not observed any substantial damage of the device within the optical power level. Therefore, we believe that the damage threshold of the prepared ReSe2/PVA-based SA must be larger than 1 W. However, it was impossible to measure the precise value due to the limited availability of a high-power laser in our laboratory.

The experimental schematic of our passively mode-locked fiber laser is shown in Fig. 8. The gain medium is a 2.3-m-long erbium-doped fiber (EDF) with peak absorption of ∼20dB/m at a wavelength of 1530 nm. A 980 nm pump laser diode was used as the pump source that was inserted into the gain medium through a 980/1550 nm wavelength division multiplexer (WDM). The polarization-independent isolator was used to force unidirectional light propagation. The mode-locked laser output was extracted from the ring cavity using the 10% port of a 90:10 coupler. A PC was used to optimize the polarization state of the oscillating beam within the laser cavity. The ReSe2/PVA-deposited side-polished fiber was placed after the PC.

Stable mode-locked pulses were readily obtained when the pump power was set to ∼19.6mW, with the PC properly adjusted. The average output power was measured to be ∼0.5mW. The main factor limiting the output power is the unoptimized cavity configuration. To optimize the laser cavity to produce higher average output power, we tried to reduce the insertion loss of the prepared SA as well as to change the output coupler. To control the insertion loss of the prepared SA, we need to optimize the side-polished fiber in terms of its polishing depth, interaction length, and the concentration of the solution. Such work to enhance the laser performance is reserved for future work. Figure 9(a) shows the measured optical spectrum of the mode-locked pulses. Kelly sidebands were clearly observed, indicating that the fiber laser was operating in the soliton regime ^[83]. The center wavelength and 3 dB bandwidth of the pulses were measured to be ∼1561.2nm and ∼3.4nm, respectively. Figure 9(b) shows the measured oscilloscope trace of the output pulses with a 16 GHz real time oscilloscope combined with an 11 GHz photodetector. The inset of Fig. 9(b) shows a magnified view of a single output pulse. The temporal period of the output pulse is ∼66.8ns, which corresponds to the fundamental repetition rate of ∼14.97MHz.

Next, an autocorrelation measurement was performed using a TPA-based autocorrelator. Figure 10(a) shows the measured autocorrelation trace of the output pulses with a hyperbolic secant-squared fitting curve. The estimated temporal width of the output pulses is ∼862fs. Considering the 3 dB bandwidth of ∼3.4nm, the estimated time-bandwidth product is ∼0.361, indicating that the output pulses are slightly chirped. The electrical spectrum was measured to confirm the stability of the output pulses, as shown in Fig. 10(b). A strong signal peak with an electrical signal-to-noise ratio (SNR) of ∼65dB was obtained at a fundamental repetition rate of ∼14.97MHz. The inset of Fig. 10(b) shows the electrical spectrum with a 1 GHz frequency span. Strong harmonic signals were clearly observed, strongly indicating that the output pulses are stable mode-locked pulses.

The solution dropping is a simple method to form a thin film on a substrate by dropping and drying a small amount of solution. However, it is difficult to control the uniformity of the film thickness using this method. The saturable absorption performance of our prepared SA was observed to change in some degree sample by sample. However, the output performance of our mode-locked laser was not observed to significantly vary depending on the SA sample change. It should be noticed that the mode-locking performance of a fiber laser with net anomalous cavity dispersion is not so dependent on the SA performance if the minimum parameter conditions of the SA used in the laser cavity such as modulation depth and saturation energy are satisfied ^[81]. We have also checked the long-term stability of the mode-locked fiber laser by observing the variation of the output pulses for several hours and did not find any substantial variations of the output pulses.

Finally, the output performance of our mode-locked fiber laser was compared to those of mode-locked fiber lasers using SAs based on other types of nonlinear optical materials. The results are summarized in Table 2. Our fabricated ReSe2/PVA-based SA exhibited a modulation depth (∼3.9%) that is similar to systems based on GO, Bi2Se3, MoSe2, SnS2, gold nanorods, and CoSb3. In terms of the pulse width of the output pulses, our fiber laser exhibits a reasonable level of the output pulse width (862 fs) for femtosecond laser applications. In terms of saturable absorption properties, ReSe2 does not exhibit noticeable advantages over other nanomaterials. However, it should be noticed that the ReSe2 particles used in this work are bulk-structured but not nanostructured. Their sizes were observed to vary from tens of nanometers to around 2 μm. Since bulk-structured materials do not require a complicated fabrication process, we believe that they have significant advantages over nanostructured ones in terms of ease of sample preparation.

We have systematically investigated the nonlinear optical properties of ReSe2. The nonlinear absorption coefficient and the nonlinear refractive index of ReSe2 were measured to be −(5.67±0.35)×103cm/GW and −(2.81±0.13)×10−2cm2/GW, respectively, using the OA and CA Z-scan technique at a wavelength of 1560 nm. The prepared ReSe2 sample was shown to possess nonlinear saturable absorption and self-defocusing properties. The values of the nonlinear absorption coefficient and the nonlinear refractive index of ReSe2 are comparable to those of other TMDC materials, as reported in previous studies. A fiberized SA was also fabricated using a side-polished-fiber platform by depositing a ReSe2/PVA composite. Using the fabricated ReSe2/PVA-based SA, stable mode-locked pulses with a temporal width of ∼862fs could be readily obtained from an EDF laser.

Further investigations for the improvement of the uniformity of the prepared ReSe2 film are required to enhance the repeatability of the ReSe2-based SA fabrication process. We believe that the results of this investigation confirm that ReSe2 is a good candidate for the development of nonlinear optic or photonic devices.