Chinese Optics Letters, Volume. 23, Issue 7, 071402(2025)

Repetition-rate tunable ultrafast microjoule Yb-fiber lasers

Xiangming Xiao1...2,3, Wenmi Shi1,2, Chenyang Gao1,2, Yang Gui1,2,3, Meng Pang1,2,3, Gengji Zhou1,2,*, and Yuxin Leng1,23 |Show fewer author(s)
Author Affiliations
  • 1State Key Laboratory of High Field Laser Physics and CAS Center for Excellence in Ultra-intense Laser Science, Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai 201800, China
  • 2University of Chinese Academy of Sciences, Beijing 100049, China
  • 3Hangzhou Institute for Advanced Study, University of Chinese Academy of Sciences, Hangzhou 310024, China
  • show less

    We presented a repetition-rate tunable Yb-doped fiber laser system, which used a chirped fiber Bragg grating as a fiber stretcher designed to match the second- and third-order dispersion of the transmission grating compressor. The system delivered 1-µJ, 143-fs pulses at a 2 MHz repetition rate and 10-µJ, 157-fs pulses at a 200 kHz repetition rate, respectively. The pulse repetition rate can be tuned from 200 kHz to 2 MHz while the pulse duration maintains <180 fs. This compact fiber laser source was built for applications in ophthalmology, such as corneal flap cutting and tissue vaporization. Furthermore, it can be applied in micro-machining applications, such as laser marking, scribing, and drilling.

    Keywords

    1. Introduction

    Since micro-machining at the stromal region of the cornea with a microjoule (µJ) femtosecond pulse was demonstrated two decades ago, µJ-level femtosecond lasers garnered significant attention in the field of ophthalmology. Two major approaches were developed to create structural modifications in the cornea. One was laser-assisted in situ keratomileusis (LASIK)[1,2], and the other was incision lenticule extraction (SMILE)[3]. Both LASIK and SMILE benefited from precise and controllable micro-machining capabilities provided by µJ-level femtosecond lasers, which exhibited greater advantages compared with traditional mechanical blades or microkeratome. Current commercial µJ-level femtosecond lasers applied in vision correction surgery were operated at a fixed pulse repetition rate (rep-rate), such as 500 kHz or 2 MHz for the VisuMax series from Carl Zeiss. Fixed rep-rate helped to reduce the system complexity in constructing femtosecond lasers. However, it was powerless when confronting applications necessitating a femtosecond pulse with a larger pulse energy at a lower rep-rate, such as performing femtosecond laser-assisted cataract surgery (FLACS), which usually needs a 5- to 10-µJ 500-fs pulse in hundreds of kHz rep-rate[4]. Hence, a rep-rate tunable femtosecond laser with µJ-level pulse energy could provide great flexibility in dealing with applications demanding femtosecond pulses with different parameters.

    The most common means to construct a femtosecond laser with µJ-level pulse energy is to adopt the chirped-pulse amplification (CPA) technique invented by Strickland and Mourou in 1985[5]. At the early stage of CPA development, researchers utilized bulk grating pairs for both stretching and compressing the pulse, allowing the second- and third-order dispersion (GDD and TOD) to be perfectly matched and canceled out. However, this configuration typically requires a large footprint and lacks stability. Subsequently, the fiber CPA technique was introduced[68], where a normal dispersion fiber served as a pulse stretcher, while bulk grating pairs were still employed as pulse compressors. This technique benefited from the compactness and stability of the fiber stretcher but faced issues with the TOD mismatch between the fiber stretcher and grating compressor, both of which exhibit positive TOD. The accumulation of positive TOD resulted in a broad pedestal in the compressed pulse, leading to dispersed energy and reduced peak power. To address this issue, nonlinear pulse amplification (NPA) in the fiber amplifier was developed to reduce the TOD-induced pedestal, which exploited the compensation between the TOD and self-phase modulation (SPM)[913]. However, the NPA technique proved highly sensitive to the pulse peak power in the fiber amplifier, making it challenging to implement in a rep-rate tunable fiber laser system.

    Recently, a chirped fiber Bragg grating (CFBG), specially designed to match a bulk grating pair compressor, was introduced to the fiber CPA system[1417]. This method combined the merits of a bulk grating pair and a fiber stretcher, making it an effective pulse stretcher for the development of rep-rate tunable µJ-level femtosecond fiber lasers.

    In this Letter, we experimentally demonstrated a rep-rate tunable Yb-doped ultrafast fiber laser system that used a CFBG as a pulse stretcher and transmission grating as a pulse compressor. This system can deliver 1-µJ, 143-fs pulses at a rep-rate of 2 MHz and 10-µJ, 157-fs pulses at a rep-rate of 200 kHz, respectively. The pulse rep-rate can be adjusted using a fiber-type acousto-optic modulator (AOM). Pulse durations of less than 180 fs at the µJ-level energy were achieved by fine-tuning the compressor grating pair distance when operating at different rep-rates. The tuning time is within seconds.

    2. Experimental Setup

    Figure 1 depicts the schematic of the laser system including six main parts: a mode-locked fiber oscillator, a pulse stretcher, one stage of core-pumped polarization-maintaining (PM) fiber amplifier, two stages of clad-pumped PM fiber amplifier, and a grating-pair pulse compressor. Additionally, an in-line AOM was introduced to pick the pulse for rep-rate tuning.

    Schematic of the rep-rate tunable µJ-level femtosecond fiber laser system. SM-LD, single-mode laser diode; WDM, wavelength-division-multiplexer; YDF, Yb-doped fiber; COL, collimator; HWP, half-wave plate; QWP, quarter-wave plate; PBS, polarization beam splitter; ISO, isolator; HR, highly reflective mirror; PM, polarization-maintaining; CIR, circulator; BPF, bandpass filter; AOM, acousto-optic modulator; MM-LD, multi-mode laser diode; DM, dichroic mirror.

    Figure 1.Schematic of the rep-rate tunable µJ-level femtosecond fiber laser system. SM-LD, single-mode laser diode; WDM, wavelength-division-multiplexer; YDF, Yb-doped fiber; COL, collimator; HWP, half-wave plate; QWP, quarter-wave plate; PBS, polarization beam splitter; ISO, isolator; HR, highly reflective mirror; PM, polarization-maintaining; CIR, circulator; BPF, bandpass filter; AOM, acousto-optic modulator; MM-LD, multi-mode laser diode; DM, dichroic mirror.

    The mode-locked fiber oscillator was constructed in a ring configuration, with a 40-cm-long Yb-doped fiber (INO, YB401-5/125) serving as the gain medium. A wavelength-stabilized 976-nm laser diode (LD) pumped the gain fiber through a wavelength-division multiplexer (WDM). A pair of reflective gold gratings (600 lines/mm groove density) provided the necessary anomalous dispersion to the cavity. By adjusting the separation between the grating pair using a linear translation stage, we could tune the net cavity dispersion from negative to positive. This adjustment allowed us to access a wide range of mode-locking regimes, from soliton mode-locking to dissipative soliton mode-locking[1822]. The combination of two quarter-waveplates (QWPs), one half-waveplate (HWP), and a polarization beam splitter (PBS) constituted the “mode locker,” exhibiting equivalent performance as a saturable absorber. This configuration is well-known as nonlinear polarization rotation evolution (NPR). Given that the generated pulse with the broadest mode-locked spectrum manifests the lowest intensity and phase noise, we deliberately operated the oscillator in the stretched pulse mode-locked regime with a net dispersion close to zero[23,24]. We isolated the mechanical vibrations through a 35-mm-thick optical breadboard together with a vibration-absorbing mat placed underneath.

    We then sent the mode-locked pulse into a PM fiber for subsequent pulse stretching and amplification. A fiber coupler extracted 10% of the seed power to monitor the mode-locking status, while the remaining 90% of the power was fed into the stretcher. To assemble the pulse stretcher, we combined a three-port PM fiber circulator with a set of CFBG (36ps2). We used a single-stage core-pumped PM fiber amplifier to scale up the signal average power. The 60-cm-long gain fiber (INO, Yb401-PM) with a core diameter of 5 µm was forward-pumped by a 950-mW wavelength-stabilized 976-nm laser diode (LD). An in-line fiber isolator was placed after the gain fiber to prevent back reflection. Considering that a relatively weak signal was being amplified in a high-gain (600dB/m) fiber, we used a fiberized bandpass filter (BPF) to effectively suppress the amplified spontaneous emission (ASE).

    For pulse picking, we employed a fiber-coupled AOM (CETC, SGTF200), which generated a tunable rep-rate signal pulse. The AOM’s trigger signal was provided by the 1% port of the fiber coupler located right behind the BPF. Picked signal pulses were then sent into a clad-pumped fiber amplifier, which consisted of a 3-W 976-nm multimode (MM) pump, a (2+1)×1 PM pump beam combiner (PBC), a 180-cm-long Yb-doped gain fiber, and a fiber isolator. The core diameter of the fiber in use was 12 µm. The first fiber amplifier was used to compensate for the power loss from the CFBG stretcher, while the second one was designated to provide adequate seed power for the subsequent fiber power amplifier. Adequate seed power here was determined by the final output pulse shape and spectrum. More specifically, there was no observation of stimulated Raman spectra or broad pedestal shown in the compressed pulse auto-correlation (AC) trace.

    Another fiber-type BPF was employed behind to suppress the ASE and improve the signal-to-noise ratio (SNR). The last stage of the fiber power amplifier consisted of an 18-W 976-nm multi-mode pump diode, a (2+1)×1 PM PBC, a 70-cm-long Yb-doped double-cladding gain fiber with a core diameter of 20 µm and an endcap featuring a straight-cleaved coreless fiber rod. The signal pulse was injected through an input PM fiber with a 12-µm core diameter. Notably, there is a mode-field area mismatch between the input and output fibers of the PBC. This mode-area adaptation from the 12-µm core diameter fiber to the 20-µm core diameter fiber was achieved using a tapered fusion technique. In the end, the amplified µJ-level pulse was collimated using a spherical plano-convex lens (f=30mm) and then compressed with a pair of transmission gratings (Gitterwerk, 31.8mm×20.2mm, 130mm×15mm) in a double-pass configuration. A dichroic mirror (Layertec, 108980) was employed to separate the signal pulse from the residual pump light. To prevent any back-reflected light, a large-aperture optical isolator was positioned in front of the grating compressor. Additionally, several wedges and mirrors were included in the system to divert a small portion of the laser pulse for diagnosis.

    3. Results and Discussion

    Figure 2 shows the output of the mode-locked fiber ring oscillator. The mode-locked spectrum of the seed pulse is centered at 1030 nm, with a full width at half-maximum (FWHM) of 2 nm. As noticed in Fig. 2(a), an asymmetric mode-locked spectrum with a distinct bump at shorter wavelengths is observed. This is probably due to the highly nonlinear evolution accompanied with the NPR, which is common for stretched-pulse mode-locking[20]. Apart from the sharp bump, the spectrum shows a wide range extending from 1018 to 1050 nm, supporting about a 50-fs Fourier-transform limited (FTL) pulse. The central part of the spectrum is relatively flat and smooth, making it suitable as the seed spectrum for the following fiber amplifiers. Figures 2(b) and 2(c) plot the radio-frequency (RF) spectrum and pulse train waveform of the oscillator. The pulse train waveform exhibits a temporal period of 33.3 ns with even amplitude, and the RF spectrum shows a rep-rate of 29.8 MHz, with 70dB SNR at 1 kHz resolution bandwidth (RBW). Both of them confirm a stable fundamental mode-locking state. We also measured the pulse duration directly from the oscillator. Figure 2(d) depicts the autocorrelation trace. A chirped pulse of 2.45 ps was measured when applying the Gaussian line shape assumption with a deconvolution factor of 1.41.

    Characteristics of the mode-locked fiber oscillator. (a) Mode-locked spectrum, (b) RF spectrum, (c) pulse train waveform, and (d) pulse auto-correlation trace.

    Figure 2.Characteristics of the mode-locked fiber oscillator. (a) Mode-locked spectrum, (b) RF spectrum, (c) pulse train waveform, and (d) pulse auto-correlation trace.

    In order to achieve rep-rate tunable and µJ-level femtosecond pulses, a pure CPA manner should be maintained in constructing the laser system. Pure CPA here indicates that the pulse peak power-induced nonlinearities occurring in the fiber amplifiers could be negligible. As a result, the final output pulse spectra and shape are insensitive to the seed pulse rep-rate. It is commonly expected that the longer the stretched pulse, the lower the peak power, allowing linear amplification without any nonlinearities. However, it is rather complicated in the real scenario. Excessive positive chirp introduced to the seed pulse demands a large-footprint grating-pair compressor for de-chirping, which probably degrades the output beam quality or system stability. Additionally, large normal-dispersion compensation demands a longer horizontal size of the second grating, which also raises the total cost of the laser system. There exists a balance between the pulse stretcher and the compressor.

    In our laser system, the introduced positive group-delay dispersion (GDD) was determined to be 36ps2. We stretched the seed pulse through a temperature-array-controlled CFBG. As depicted in Fig. 3(b), the stretched temporal pulse is estimated to be 1ns from the measured waveform in the oscilloscope. We attributed the asymmetry of the waveform to the response curve of the InGaAs photodetector. Due to the reflectivity and 17-nm reflection bandwidth of the CFBG, together with the insertion loss from the 3-port fiber circulator, 3.7 mW out of 43.2 mW average power was left for the subsequent fiber amplifier. The CFBG reflected spectrum is plotted with the blue curve in Fig. 3(a), showing a clear sharp rise/fall edge with a structured top. The reflectivity profile of the CFBG applied in the original mode-locked spectrum resulted in the spiky structure of the stretched pulse spectrum.

    (a) CFBG reflected spectrum (blue curve) and core-pumped fiber amplifier spectrum (orange curve). (b) Stretched pulse temporal waveform. (c) Picked pulse train waveform at a rep-rate of 2 MHz. (d) Picked pulse train waveform at a rep-rate of 200 kHz.

    Figure 3.(a) CFBG reflected spectrum (blue curve) and core-pumped fiber amplifier spectrum (orange curve). (b) Stretched pulse temporal waveform. (c) Picked pulse train waveform at a rep-rate of 2 MHz. (d) Picked pulse train waveform at a rep-rate of 200 kHz.

    The first stage of the fiber amplifier was core-pumped, and a maximum of 557 mW average power was obtained. The orange curve depicted in Fig. 3(a) shows the amplified spectrum. The red part gained more amplification than the blue one, which was observed in the fiber amplifiers during the amplification of the ns-level pulse with rep-rates in the tens of MHz range[14,16]. The amplified seed pulse was picked by the fiber-type AOM. The picked pulse rep-rate can be tuned from 2 MHz to 200 kHz, as confirmed by the measured pulse train waveforms plotted in Figs. 3(c) and 3(d). Limited by the damage threshold of fiber-type AOM, the input average power of the seed pulse was set at 463 mW. Because of the insertion loss and rep-rate decrease, the picked signal power was altered from 14.9 to 1.52 mW at the rep-rates from 2 MHz to 200 kHz. Two pieces of a 1% fiber coupler, placed both before and after the AOM, were used to monitor the spectra and power of the seed pulse. The following second stage of the fiber amplifier was employed to increase the seed pulse power with a tunable rep-rate, providing adequate power for the last stage of the power amplifier. We utilized the 12-µm core size fiber amplifier and boosted the seed power to 221.6 and 25.5 mW at rep-rates of 2 MHz and 200 kHz, respectively. The pulse energy was limited to 0.1μJ, and the corresponding pulse peak power was restrained to below 100 W inside the 12-µm core size fiber. The configuration above maintained the pure CPA approach, as demonstrated by the measured spectra and the pulse AC trace. The bandwidth of the amplified spectrum was preserved and measured using an optical spectrum analyzer.

    Pulse compression was performed using a pair of transmission gratings. The resulting AC trace matched that of the compressed pulse from the first stage of amplification. With the linear amplification of the rep-rate tunable seed pulse in the second stage of the fiber amplifier, we continued to construct the following power amplifier and pulse compressor. To achieve a tunable pulse energy ranging from 1 to 10 µJ at the final output, while accounting for a final compression efficiency of approximately 78%, we boosted the average power to >2.6W across a tunable rep-rate range of 200 kHz to 2 MHz. The calculated amplifier gain ranged from 20 dB (200 kHz) to 10.6 dB (2 MHz). The launched pump power was 5.15 and 4.26 W at rep-rates of 200 kHz and 2 MHz, respectively. The calculated optical-to-optical efficiency ranged from 53.7% (200 kHz) to 67.8% (2 MHz). The final fiber power amplifier became more efficient as we increased the seed pulse rep-rate.

    Figure 4 depicts the compression results at the rep-rates of 2 MHz and 200 kHz. Figures 4(a) and 4(c) plot the output spectra, while Figs. 4(b) and 4(d) show the compressed pulse auto-correlation traces. A comparison between Figs. 4(a) and 4(c) indicates that the amplified pulse spectra shape and bandwidth are almost identical, insensitive to rep-rates changing from 2 MHz to 200 kHz. The blue curve shows the output spectra before the pulse compressor, unabsorbed pump, and broadband ASE are noticed. The SNR is above 40 dB. The orange curve plots the measured spectra after pulse compression. The residual pump and ASE spectra component are cleaned by the grating pair compressor. Spectra bandwidth of 10 nm was maintained during the power amplifier at both rep-rates of 2 MHz and 200 kHz. We numerically calculated the FTL pulse according to the orange spectrum and plotted the result with a light green curve in Figs. 4(b) and 4(d). The measured AC trace of the compressed pulse is also depicted in Figs. 4(b) and 4(d) with the orange curve. The pulse duration is estimated as 143 fs (at a rep-rate of 2 MHz) and 157 fs (at a rep-rate of 200 kHz) when applying the Lorentz line shape assumption with a deconvolution factor of 2. As shown in the figure, both of the compressed pulse AC traces hardly differ from their FTL pulse, which confirmed that the system operated in a pure CPA manner. Both of the compressed pulse AC traces showed a limited pedestal, which was attributed to uncompensated TOD. Although the CFBG is specially designed to match the GDD and TOD of the compressor gratings, the residual uncompensated TOD from the fiber, particularly from the pre- and power amplifier, still accumulates. This accumulation resulted in a limited pedestal, as observed in the AC trace. The CFBG provided a negative TOD that can be tuned within a narrow range by adjusting the temperature array, which helps to reduce the pedestal but not to eliminate it.

    Characteristics of final output pulse. (a) 2 MHz: output spectra from power amplifier (blue curve) and pulse compressor (orange curve). (b) 2 MHz: output pulse AC trace (orange curve) and FTL AC trace (light green curve). (c) 200 kHz: output spectra from power amplifier (blue curve) and pulse compressor (orange curve). (d) 200 kHz: output pulse AC trace (orange curve) and FTL AC trace (light green curve).

    Figure 4.Characteristics of final output pulse. (a) 2 MHz: output spectra from power amplifier (blue curve) and pulse compressor (orange curve). (b) 2 MHz: output pulse AC trace (orange curve) and FTL AC trace (light green curve). (c) 200 kHz: output spectra from power amplifier (blue curve) and pulse compressor (orange curve). (d) 200 kHz: output pulse AC trace (orange curve) and FTL AC trace (light green curve).

    We measured the spectra, pulse energy, and pulse duration at all eight rep-rates and recorded the results in Fig. 5(a). The pulse energy can be tuned from 1 to 10 µJ as the rep-rate decreases from 2 MHz to 200 kHz. By slightly adjusting the distance between the compressor grating pairs, a de-chirped pulse duration of less than 180 fs can be achieved for different rep-rates. To maintain pure CPA and avoid any nonlinearities, we kept the fiber power amplifier in a short length of 70 cm. Hence, the power amplifier is not as efficient as a conventional clad-pumped fiber amplifier. The measured output power as a function of the pump power is plotted with solid squares in Fig. 5(b), and the linear slope efficiency is also fitted. The power amplifier operated at 2 MHz with a slope efficiency (SE) of 68.0% is more efficient compared with operation at 200 kHz with an SE of 57.2%. We also measured the power stability of the output µJ-level femtosecond pulse. As depicted in Fig. 5(c), in a measured period of 8 h, the standard deviation is 1.3 and 3.8 mW, while the mean output power is 2 W at rep-rates of 2 MHz and 200 kHz. Figure 5(d) shows the beam quality measured with the Ophir Beam Squared analyzer. The M2 of output 10-µJ pulse at a rep-rate of 200 kHz was quantified to be 1.29 and 1.27 for horizontal and vertical directions, respectively. The inset shows the well-confined Gaussian beam profile measured by a CCD camera at the average power of 2.02 W.

    (a) Pulse energy (blue square) and pulse duration (orange triangle) corresponding to rep-rate. (b) Slope efficiency (SE) of the fiber power amplifier at 2 MHz (blue solid square) and 200 kHz (orange solid square). (c) Power stability at 2 MHz (blue line) and 200 kHz (orange line). (d) M2 measurement for both x- and y-directions of the compressed pulse beam (10 µJ). Inset: beam profile.

    Figure 5.(a) Pulse energy (blue square) and pulse duration (orange triangle) corresponding to rep-rate. (b) Slope efficiency (SE) of the fiber power amplifier at 2 MHz (blue solid square) and 200 kHz (orange solid square). (c) Power stability at 2 MHz (blue line) and 200 kHz (orange line). (d) M2 measurement for both x- and y-directions of the compressed pulse beam (10 µJ). Inset: beam profile.

    4. Conclusion

    In summary, a rep-rate tunable, ultrafast µJ-level Yb-fiber laser system was experimentally demonstrated. A pure CPA manner was maintained during all fiber amplification. The rep-rate can be tuned from 200 kHz to 2 MHz, with the output pulse energy changed from 10 to 1 µJ. A pulse duration below 180 fs was obtained via slightly adjusting the compressor grating pair distance when operating at different pulse rep-rates. The output pulse exhibited good power stability with a standard deviation of 3.8 mW at an output mean power of 2.02 W. The measured beam had a well-confined Gaussian beam profile with M2 below 1.3. This laser system can be applied in ophthalmology (e.g.,  corneal flap cutting and tissue vaporization), micro-machining (e.g.,  laser marking, scribing, and drilling), and optical brain monitoring through the iDWS method[25,26].

    Tools

    Get Citation

    Copy Citation Text

    Xiangming Xiao, Wenmi Shi, Chenyang Gao, Yang Gui, Meng Pang, Gengji Zhou, Yuxin Leng, "Repetition-rate tunable ultrafast microjoule Yb-fiber lasers," Chin. Opt. Lett. 23, 071402 (2025)

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Lasers, Optical Amplifiers, and Laser Optics

    Received: Jan. 15, 2025

    Accepted: Feb. 17, 2025

    Published Online: Jun. 5, 2025

    The Author Email: Gengji Zhou (zhougengji@siom.ac.cn)

    DOI:10.3788/COL202523.071402

    CSTR:32184.14.COL202523.071402

    Topics