In recent years, mid-infrared (mid-IR) laser sources operating between 2 and 5 µm have been in great demand for various applications, such as molecular spectroscopy, free-space communication, mid-IR supercontinuum generation, material processing, and medical applications [1–3]. Due to multiple molecules undergoing strong vibrational absorption, the mid-IR range is often referred to as the molecular fingerprint regime. Therefore, the availability of ultrafast mid-IR coherent sources with broad spectral spans is an important prerequisite in molecular spectroscopy compared to other research fields. In the mid-IR wavelength range above 2 µm, solid-state lasers based on Cr²⁺-doped ZnS or ZnSe, named Ti:sapphire of the mid-IR, operate between 2 and 3.5 µm covering a spectrally critical portion of the fingerprint section [1, 3]. Compared with Cr²⁺:ZnSe (Cr:ZnSe), Cr²⁺:ZnS (Cr:ZnS) crystal is distinguished by a higher damage threshold, higher thermal conductivity, better chemical and mechanical stability, and lower thermal lensing parameter dn/dT [3]. Consequently, various crystal characteristics of Cr:ZnS crystal make it attractive for developing high-power femtosecond mid-IR solid-state lasers.

Kerr-lens mode-locked polycrystalline Cr:ZnS laser has successfully generated three optical cycle pulses of <29 fs [4] and watt-level output power [1]. However, this promising technique requires precise optical alignment and intracavity dispersion control to produce ultrashort pulses. One critical consideration in achieving shorter pulse duration is optimizing the intracavity negative group delay dispersion (GDD) for a flat spectral profile across a broad range. However, from the perspective of mid-IR optical components, fabricating thick multilayer mirrors with homogeneous dispersion properties is a nontrivial task [3]. Furthermore, the mode-locked operation with a polycrystalline gain medium may be another challenge by its asymmetric grain distribution, which causes cavity misalignment and power ramp-up [5]. Thus, achieving a stable Kerr-lens mode-locked operation in mid-IR polycrystalline lasers requires significant effort [1]. As a different approach, femtosecond Cr:ZnS lasers were relatively easily demonstrated with passive mode-locking (ML) techniques based on saturable absorbers (SAs), such as semiconductor saturable absorber mirror (SESAM) [6, 7], single-walled carbon nanotube saturable absorber (SWCNT-SA) [8, 9], and graphene-SA [10, 11]. However, manufacturing SESAM and SWCNT-SA for mid-IR solid-state lasers presents a significant technical hurdle owing to the requirement of sophisticated epitaxial growth techniques and the lack of commercially available SWCNT materials with a large diameter for nonlinear absorption around 2.3 µm, respectively.

On the other hand, the graphene-SAs have various strengths, such as a cost-effective manufacturing process, ultrabroad bandwidth, and spectrally even saturable absorption [10]. To date, various femtosecond Cr:ZnS lasers have been demonstrated utilizing different types of graphene-SA [2]. Based on reflecting graphene-SAs, passively mode-locked Cr:ZnS laser produced ultrashort pulses as short as 41 fs with a pulse energy of 2.3 nJ [10]. And the highest pulse energies of up to 15.5 nJ and pulse duration of 870 fs were also obtained from the chirped-pulse Cr:ZnS oscillator, which requires an extra-cavity pulse compressor for reducing the pulse width down to 187 fs [12]. With transmitting graphene-SAs, an ultrafast Cr:ZnS laser capable of continuous tuning over a wide range of wavelengths (~300 nm) was demonstrated. And the laser produced pulses as short as 220 fs with a pulse energy of 7.8 nJ [11].

Despite the requirement for a second focusing regime in the oscillator to ensure adequate energy fluence for the SA, transmission-type graphene-SAs offer two clear benefits for laser operation compared to reflection-type SAs. Firstly, the transmitting graphene-SA typically comprises a graphene layer with universal linear optical absorption properties and a transparent substrate supporting the layer. It means that the graphene-SA has the potential to operate over an extremely wide spectral range, encompassing all currently available ultrashort pulse laser technologies [11, 13–19]. Secondly, the energy fluence on the graphene-SA can be controlled by only varying its position along the optical path in the focusing regime, simplifying the optimization of the laser’s performance.

For developing transmissive graphene-SAs, quartz has been commonly used as the base plate in the near-IR regime. However, the material exhibits a transparency problem beyond 2 µm [15]. Therefore, transparent substrates such as calcium fluoride (CaF₂), magnesium fluoride (MgF₂), and sapphire can be utilized as base substrates, especially in the mid-IR spectral range. At the 2.3 μm spectral regime, these materials exhibit a small second-order nonlinear refractive index (n²) of around 1–2 × 10⁻²⁰ m²/W, comparable to quartz’s nonlinearity at the same wavelength [20].

In this paper, to achieve a shorter pulse duration through an additional self-phase modulation effect, we utilized ZnSe with high nonlinearity as a new base plate for developing transmission-type SAs in the mid-IR range. The ZnSe material exhibits a nonlinear refractive index (n₂,_ZnSe = 113 ± 12 × 10⁻²⁰ m²/W) approximately two orders of magnitude higher than typical CaF₂ substrates (n_2,CaF2 = 1.09 ± 0.11 × 10⁻²⁰ m²/W) used in previous works [11, 15], and even higher than the host material of a Cr:ZnS laser crystal at 2.3 µm (n_{2,ZnS,Polycrystalline} = 47.8 ± 5 × 10⁻²⁰ m²/W) [20]. By only replacing the substrate material, we can expect an enhanced nonlinear effect that will broaden the spectrum of the generated pulses. To compare the laser performance using different SAs, we developed two transmitting graphene-SAs based on CaF₂ and ZnSe with the same thickness. With these SAs, we successfully demonstrated passively mode-locked Cr:ZnS laser at the same laser configuration. With graphene-SA based on CaF₂ substrate, femtosecond Cr:ZnS laser produced relatively long pulses with 540 fs pulse duration at 2,330 nm by uneven round-trip GDD compensation. By simply replacing the SA with graphene on a ZnSe substrate, the Cr:ZnS laser generated transform-limited ~130 fs pulses at 2,340 nm in a stable single-pulse regime. And the laser produced an average power of up to 635 mW at a repetition rate of 234 MHz. To the best of our knowledge, this is the first study to use a substrate with high nonlinearity to fabricate transmission-type graphene-SAs for achieving shorter pulse widths from a polycrystalline Cr:ZnS laser.

2.1. Monolayer Graphene Saturable Absorber and Cr:ZnS Laser Configuration

To fabricate a high-quality monolayer graphene-based SA, the fabrication method reported in [13] was employed. First, monolayer graphene synthesized by chemical vapor deposition on a copper (Cu) foil was rapidly cooled down. Then, a supporting layer of 5 wt% poly(methyl-methacrylate) (PMMA) was spin-coated onto the graphene. Afterward, the Cu layer was wet-etched in ferric chloride (FeCl₃), allowing the layer consisting of graphene and PMMA to be transferred onto CaF₂ and zinc selenide (ZnSe) substrates with a diameter of 1 inch and thickness of 2 mm each. Finally, the PMMA layer was removed using acetone, leaving behind the high-quality monolayer graphene-SAs.

We investigated the linear optical characteristics of each transmitting graphene-SAs, as shown in Fig. 1. From the linear transmission curve depicted in Fig. 1(a) and 1(b), the monolayer graphene-SA deposited on different substrates exhibit approximately 2% absorption near 2,340 nm, which is comparable with the theoretical value of πα = 2.3%. The insets of Fig. 1(a) and 1(b) display photo images of the monolayer graphene-Sas used in the present work, with four dots on each CaF₂ and ZnSe substrate indicating the boundary edges of the transferred graphene layer. For passive ML of a polycrystalline Cr:ZnS laser employing graphene-Sas, we developed an experimental setup for the astigmatically compensated Cr:ZnS laser as shown in Fig. 2. As pumping sources, we used a continuous wave (CW) Er-doped fiber laser (λ_center = 1,550 nm; IPG Photonics, MA, USA), delivering power up to 20 W. The pump beam was focused on a polycrystalline Cr:ZnS crystal by the 100-mm focusing lens (L). The 5.05-mm long Cr:ZnS crystal was supplied by IPG Photonics, and the Cr²⁺ concentration in ZnS host material is about 8.56 × 10¹⁸ cm⁻³, regarding small signal absorption at 1,550 nm is around 94.2%.

Figure 1. Optical transmission of high-quality monolayer graphene saturable absorber (G-SA) and photos of the G-SAs developed on CaF₂ and ZnSe substrate (inset). (a) Graphene transferred onto CaF₂ substrate. (b) Graphene transferred onto ZnSe substrate.

Figure 2. Experimental setup for graphene mode-locked Cr:ZnS laser. Pump laser, CW Er-doped fiber laser; L, pump focusing lens (f = 100 mm); M1 and M2, high reflecting concave mirrors (ROC = −100 mm); M3 and M4, high reflecting concave mirrors (ROC = −75 mm); OC, output coupler (T = 15%); Graphene-ZnSe, monolayer graphene saturable absorber with ZnSe substrate. CW, continuous wave; ROC, radius of curvature.

We mounted the crystal on a water-cooled Cu block for efficient heat removal from the gain medium, which stabilized at 15 ℃. The laser crystal was oriented at Brewster angle to minimize Fresnel reflection losses and positioned between two concave mirrors with a radius of curvature (ROC) of −100 mm, M1 and M2. The focused pump beam waist size was calculated to be ~57 µm, and the resonator beam waist size was estimated to be 65–75 µm, depending on distances between M1 and M2. These two beam sizes are comparable and allow efficient energy transfer from the pump to the resonator beam. With two additional concave mirrors (M3 and M4) with an ROC = −75 mm, the second focusing regime was developed at the long arm to deliver sufficient energy fluence enough for bleaching the absorption of the SAs. The minimum beam waist size in the second focusing regime was calculated to be 80–90 µm, depending on the position of M4. Developed graphene-SAs were placed at the Brewster angle to minimize the insertion losses near the focus. Substrate materials of two graphene-SAs are CaF₂ and ZnSe, regarding Brewster angles of each substrate are 54.9° (G-CaF₂) and 67.7° (G-ZnSe) at 2,340 nm, respectively. After exchanging the SAs with each other, we optimized the position of M4 along the horizontal direction of the optical axis, considering the parallel shift of the beam caused by the differences in Brewster angles of the substrate materials. In the short arm of the laser cavity, we installed a flat-wedged output coupler (OC) with a 15% transmission at the lasing wavelength regime. To isolate the laser from environmental effects such as airflow and humidity, the Cr:ZnS laser was enclosed with acrylic blocks; we did not purge it with nitrogen gases to reduce humidity.

All laser mirrors in the cavity are manufactured by ion beam sputtering (IBS) coating process and have the trait of GDD-reflection optimized (Layertec GmbH, Mellingen, Germany). However, the mirrors are not optimized for higher-order dispersion. The GDD of 15% OC is nearly 0 fs² at lasing spectral ranges, but the focusing mirrors (M1–M4) provide negative GDD between 2,100 nm and 2,600 nm.

Therefore, with four focusing mirrors, we sufficiently compensated the opposite sign GDD from the Cr:ZnS laser crystal and ZnSe substrate of the SA. The mirrors, however, were not designed like the typical chirped mirror with homogeneous GDD at specific spectral ranges; The mirrors introduced inhomogeneous and relatively strong negative GDD near 2,430 nm (GDD = −450 fs² per bounce @ 2,340 nm, central wavelength). Figure 3(a) displays the wavelength dependence of GDD for different optical components within the laser cavity, including the laser crystal, different substrates of the SAs, OC, and a GDD-optimized mirror. The solid vertical green lines in Fig. 3 indicate the central wavelength of the passively mode-locked Cr:ZnS laser with graphene-SAs. As shown in Fig. 3(b), the net round-trip GDD of the laser cavity was calculated to be −2,100 fs² and −1,070 fs² at 2,340 nm for the CaF₂ and ZnSe substrates of graphene-SAs, respectively. Since the net GDD values are similar to the GDD of −1,500 fs² in [1], which is adequate to chirp <100 fs pulses, no additional optical components were utilized to minimize the difference in GDD values. In addition, the GDD data shown for the mirror in Fig. 3(a) is simulation data at normal incidence (0 degrees). Therefore, by three mirrors (M1–M3) with a folding angle of 15°, the actual cavity round-trip GDD may differ from the estimated GDD values shown in Fig. 3(b). In this configuration, we achieved soliton ML in the negative dispersion regime by shaking the output coupler [21].

Figure 3. Intracavity dispersion compensation. (a) Theoretical GDD spectra of 5.05 mm thick Cr:ZnS (Cr:ZnS), output coupler (OC), concave mirrors (mirror), 2 mm thick ZnSe plate (ZnSe), and 2 mm thick CaF₂ plate (CaF₂); (b) Round-trip GDD spectra of the Cr:ZnS resonator with graphene saturable absorbers based on CaF₂ (G-CaF₂) and ZnSe (G-ZnSe). The solid vertical green lines denote the central wavelength (λ_center) regime of the passively mode-locked Cr:ZnS laser.

2.2. Femtosecond Cr:ZnS Laser with Graphene on CaF₂ Substrate

Firstly, with graphene on a 2-mm-thick CaF₂ plate (G-CaF₂), a passively mode-locked Cr:ZnS laser was obtained with a small perturbation achieved by moving an output coupler mounted on a linear stage. In the ML operation, we studied the average output power versus the incident pump power in the femtosecond mode-locked regime, as shown in Fig. 4(a). A solid vertical red line shows the ML threshold, 200 mW output at 1.75 W incident pump power. The maximum output power of 760 mW was achieved for a pump level of 4.4 W. The mode-locked operation was sustained for several hours in a whole power range above the ML threshold. In addition, there were no symptoms of degradation in the SA due to damage to the graphene layer. Higher average power was readily obtained with increasing pump power, but the laser became unstable, and the emerging multiple-pulsing tendency and strong CW components could not be suppressed easily.

Figure 4. Passively mode-locked Cr:ZnS laser utilizing graphene on CaF₂ substrate: (a) Laser output powers as a function of incident pump power, solid vertical red line denotes the mode-locking threshold. (b) Radio-frequency (RF) spectrum of the fundamental beat note at 234.8 MHz and 1 GHz wide-span RF spectrum (inset). (c) Optical spectrum. (d) Intensity autocorrelation trace with sech² fit.

The radio-frequency (RF) spectrum of the femtosecond Cr:ZnS laser was measured to verify stable mode-locked operation, as shown in Fig. 4(b). The fundamental beat note at 234.8 MHz exhibited an extinction ratio of >57.7 dB above carrier (dBc), measured at a resolution bandwidth of 3 Hz within 400 kHz span. The inset of Fig. 4(b) depicts a 1 GHz wide-span RF measurement. Two RF spectra measured at different spans apparently exhibit a stable and clean CW mode-locked operation without unwanted laser operation, such as Q-switching instability and multiple-pulsing.

Figures 4(c) and 4(d) show the laser spectrum and the intensity autocorrelation trace, recorded at an output power of around 630 mW. The output spectrum and the pulse width were measured simultaneously with the WaveScan USB MIR (A.P.E GmbH, Berlin, Germany) and the PulseCheck USB MIR (A.P.E GmbH), respectively. The spectrum was centered at 2,330 nm and exhibited a spectral bandwidth of 11.4 nm. The autocorrelation trace was fitted well by assuming a sech²-shaped pulse, and its pulse width (full width at half maximum, FWHM) was 540 fs. The corresponding time-bandwidth product is 0.34, close to the theoretical value for a transform-limited sech² pulse, 0.315.

Comparing other laser cavities with similar round-trip GDD values [1, 7], the Cr:ZnS laser produced relatively long pulses despite sufficient negative net GDD of −2,000 fs² at the central wavelength regime. In our opinion, the proper explanation for generating long femtosecond pulses is inhomogeneous GDD compensation over the lasing spectral range rather than insufficient dispersion compensation of the laser oscillator and the non-optimized parameters of the gain element. Therefore, we expect to achieve an even shorter pulse width by additional optimization to make the dispersion more flat in the negative GDD regime.

2.3. Femtosecond Cr:ZnS Laser with Graphene on ZnSe Substrate

For comparison in this laser cavity, we replaced the 2-mm-thick G-CaF₂ with the graphene on the ZnSe base plate (G-ZnSe) with the same thickness. After optimizing the horizontal position of the concave mirror (M4), passive ML in the Cr:ZnS laser was also achieved. Figure 5(a) displays the output properties for the Cr:ZnS laser with the same 15% output couplers, and a red vertical line indicates the ML threshold, 185 mW output at 2.19 W pump power. The laser produced as high as 635 mW of ML output power at the incident pump power of 4.95 W.

Figure 5. Passively mode-locked Cr:ZnS laser utilizing graphene on ZnSe substrate: (a) Laser output powers as a function of incident pump power, solid vertical red line denotes the mode-locking threshold. (b) Radio-frequency (RF) spectrum of the fundamental beat note at 234.01 MHz and 1 GHz wide-span RF spectrum (inset). (c) Optical spectrum. (d) Intensity autocorrelation trace with sech² fit.

In this case, the relatively higher ML thresholds were attributed to increased cavity losses caused by the ZnSe substrate, which has relatively lower transmittance than the CaF₂ substrate. The mode-locked operation was stable for several hours in the whole ML power range, and there were no visible damages on the SA. At higher ML power levels, unwanted CW components and multiple pulsing interrupted stable mode-locked operation.

To verify ML stability, we also recorded RF spectra in different spans [Fig. 5(b)]. The first beat note at 234.01 MHz displayed a pedestal peak separation of 60 dBc, recorded with a resolution bandwidth of 3 Hz within 400 kHz span. This high signal-to-noise ratio and the 1 GHz span measurement [inset of Fig. 5(b)] evidently show a stable single-pulse laser operation without Q-switching instability. The frequency of the fundamental beat note in this figure is comparable with the frequency value in Fig. 4(b), which indicates two resonator lengths are almost identical. Figures 5(c) and 5(d) show the optical spectrum and intensity autocorrelation trace of the generated pulses, recorded near the maximum output power level of 605 mW. The spectral bandwidth (full width at half maximum, FWHM) of the pulses was measured to be 46 nm centered at 2,340 nm. By assuming a sech²-shaped pulse, the autocorrelation trace reveals a pulse width of 128 fs (FWHM) with a corresponding time-bandwidth product of 0.322. This is close to the theoretically expected value of 0.315 for the transform-limited sech² pulse. The slight asymmetrical spectrum shown in Fig. 5(c) was observed because of intracavity atmospheric absorption above 2,400 nm, which can be moderated by purging the cavity with nitrogen gas.

In the condition of inhomogeneous round-trip GDD compensation, we achieved shorter pulses by only replacing the substrate of graphene-SAs from CaF₂ to ZnSe. The high nonlinearity of the ZnSe substrate led to an expansion in the spectral bandwidth of the oscillating pulse as a result of the enhanced self-phase modulation effect in the substrate of the graphene-SA. Despite negative net cavity GDD of −1,000 fs² at the central wavelength range, the Cr:ZnS laser did not generate sub-100-fs pulses, compared with other laser cavities with similar round-trip GDD values around –1,500 fs² [1, 7]. Based on these results, we estimate that inhomogeneous GDD compensation may be the dominant factor in achieving >100 fs pulse duration. This result leads us to expect that optimizing the intracavity net GDD for a flat profile over a broad spectral range will shorten the pulse duration further.

In conclusion, we demonstrated passively mode-locked 2.3 µm Cr:ZnS laser with monolayer graphene coated on different substrates, CaF₂ and ZnSe. With graphene on CaF₂, the mode-locked Cr:ZnS laser produced relatively long pulses as short as 540 fs near 2,330 nm. In the stable single-pulse operation, output powers up to 760 mW at 234.8 MHz repetition rates were achieved. Uneven round-trip GDD compensation would be the dominant cause for long pulse generation in the negative dispersion cavity. For direct comparison within the same laser cavity, we replaced the G-CaF₂ with the G-ZnSe, and made adjustments only to the position of the end mirror along the horizontal direction of the optical axis. By the high nonlinearity of the ZnSe plate of the SA, the femtosecond Cr:ZnS laser generates shorter transform-limited pulses with a duration of ~130 fs at 2,340 nm in a stable single-pulse regime. Average powers up to 635 mW at a repetition rate of 234 MHz were obtained. By further optimization of the net cavity GDD for a flat profile, we expect to achieve an even shorter pulse width. To the best of our knowledge, this represents the first endeavor to attain shorter pulse durations from a polycrystalline Cr:ZnS laser by utilizing graphene deposited on a high nonlinearity ZnSe substrate. Without precise controls of intracavity net GDD, we quickly achieved ultrashort pulses with ~130 fs pulse duration in the polycrystalline Cr: ZnS lasers.

The National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning, Korea (RS-2023-00208484); the Institute of Information & communications Technology Planning & Evaluation (IITP) grant funded by the Korea government (MSIT) (No.2021-0-00019, Research on Optical Learning Technology for AI).

The authors declare no conflicts of interest.

Data underlying the results presented in this paper are not publicly available at the time of publication, which may be obtained from the authors upon reasonable request. No data were generated or analyzed in the current study.