Ultrashort pulse fiber lasers have been adopted in various applications for their high brightness, low cost, maintenance-free turnkey operation, small footprint, and flexible light delivering. High repetition rate (HRR) femtosecond (fs)-pulsed fiber lasers, particularly, are of great interest for laser micromachining [1], nonlinear optical imaging [2], high-speed optical sampling [3], arbitrary waveform generation [4], and frequency metrology [5]. However, the generation of fundamental-HRR fs pulses is technically challenging [6], as cavity length (also the length of gain fiber) is inversely proportional to the repetition rate. For a GHz repetition rate, especially, a cavity length of centimeters is required, which can result in an insufficient gain for mode locking and even lasing [7]. To this end, many efforts on the highly doped optical fibers have recently been devoted to generating GHz fs pulses at 1.0 and 1.5 µm [6]. The generation of GHz fs pulses at 2.0 µm, on the other hand, is still largely unexploited [8,9], which, however, is highly demanded for eye-safe biological and mid-infrared (MIR) spectroscopic applications [10–12]. Although mode-locked ${{\rm Tm}^{3 + }}$-doped fiber (TDF) lasers with fundamental repetition of 1.25 GHz [13] and harmonic repetition of 14.5 GHz [14] have recently been demonstrated, high-power fs all-fiber lasers at 2.0 µm with even higher fundamental repetition rates are yet to be studied.

In this work, we present the first all-fiber fs-pulsed laser at 2.0 µm that operates at a fundamental repetition rate up to $\sim{2}\;{\rm GHz}$. In addition to a high repetition rate, this all-fiber laser also provides other superior performances, especially, a pulse width of 126 fs and an average power of $\sim{8}\;{\rm W}$. Such a high average power is particularly accomplished by using nonlinear chirped pulse amplification in a completely all-fiber scheme. This fs-pulsed all-fiber laser is thus anticipated to be beneficial for various applications at longer wavelengths that require a high repetition rate.

Figure 1 shows the schematic diagram of the high power GHz fs-pulsed all-fiber laser at 2.0 µm. The laser system includes two parts, i.e., the oscillator and external amplifier, both of which are all-fiber. The oscillator has a linear cavity that is mainly constructed by a semiconductor saturable absorber mirror (SESAM, Batop GmbH) and a fiber-type dielectric film (DF). The parameters of key components of the oscillator are given in Table 1. The modulation depth and recovery time of the SESAM are 12% and 10 ps, respectively. The fiber-type DF has multiple layers of ${{\rm SiO}_2}/{{\rm Ta}_2}{{\rm O}_5}$ DFs coated onto the face of an optical fiber in a ceramic ferrule. The DF has a transmittance of 58.5% at the pump wavelength (1570 nm), while a high reflectivity of 93.1% at the signal wavelength (1930 nm). In the cavity, a short piece of TDF is employed as the gain medium, which is pumped by a fiber laser at 1570 nm through a wavelength-division multiplexer (WDM). The core diameter of the TDF (Nufern SM-TSF-5/125) is about 5.5 µm, and its core absorption is about 340 dB per meter at 1560 nm. Between the SESAM and gain fiber, a piece of passive fiber (PF) is utilized to adjust the total cavity length and compensate the cavity dispersion. One end of the PF, as well as that of TDF, is inserted into ceramic ferrules and polished for better contact with the SESAM and DF, respectively. The whole resonator is placed in a compact copper package and maintained at a temperature of $\sim{12^\circ {\rm C}}$ by thermal control.

 

Fig. 1. Experimental setup of the high-power GHz fs-pulsed all-fiber laser at 2.0 µm. DF, dielectric film; ISO, isolator; DC-TDF, double-cladding ${{\rm Tm}^{3 + }}$-doped fiber; MFA, mode field adaptor; PF, passive fiber; SESAM, semiconductor saturable absorber mirror; SPC, signal and pump combiner; UHNA, ultrahigh numerical aperture fiber (Nufern UHNA4); WDM, wavelength-division multiplexer; TC, temperature controller; TDF, ${{\rm Tm}^{3 + }}$-doped fiber. Insets show the photographs of the semiconductor saturable absorber mirror and the highly reflective dielectric film.

Download Full Size | PDF

Table 1. Parameters of Key Components of the Oscillator

View Table

The oscillator is protected from optical backreflection by a fiber isolator (ISO) and is subsequently fed into the nonlinear chirped pulse amplification system. The amplification system contains three stages of optical amplifiers. The first preamplifier has a 10.5 cm gain fiber (Nufern SM-TSF-5/125) and is backward pumped by a 1550 nm fiber laser (3 W maximum power) through a WDM. The residual pump is extracted by another WDM located at the input of the first optical amplifier. After the first amplifier, the average power of the signal beam is more than 50 mW. A length of PF with opposite dispersion, i.e., an ultrahigh numerical aperture fiber (Nufern UHNA4, 5 m in length), is used to compensate the accumulated frequency chirp and thus reduce the pulse width for a higher peak power. Such a design can enable a better pulse compression via appropriate spectral broadening. Another fiber ISO is placed before the second preamplifier. The second preamplifier has a configuration similar to the first one, except that the pump wavelength is 1568 nm, and the length of the gain fiber is 12.5 cm. The output power of the second preamplifier is about 580 mW. The frequency chirp accumulated in the second preamplifier is compensated by another oppositely dispersive fiber (10 m in length). The amplified HRR pulse is then fed into the high-power amplifier to further boost up its average power. In the high-power amplifier, a double-cladding TDF (DC-TDF, Nufern SM-TDF-10P/130-M, 4 m in length) is employed, which is cladding-pumped by two high-power multimode laser diodes at 793 nm (30 W maximum power for each) through a $({2} +{1}) \times {1}$ signal and pump combiner (SPC). The residual multimode pump is extracted by using a mode field adaptor (MFA). Finally, the high-power HRR pulses are dechirped by another oppositely dispersive fiber (UHNA4, 13 cm in length) and launched into free space by a fiber optic patch cable. All the components of the laser system are mounted on an aluminum plate for cooling.

The optical spectrum of the laser system is studied by an optical spectrum analyzer (Yokogawa AQ6375). The temporal pulse train is detected by a 12.5 GHz photodiode and recorded by a 20 GHz real-time oscilloscope (Keysight DSOV204A). The pulse width is measured by an autocorrelator (APE Pulsecheck USB 50). The noise performance is investigated by a frequency signal analyzer (Rohde & Schwarz FSWP26).

First, we study the effect of the PF of the oscillator on the mode-locking threshold and output power. As shown in Fig. 2, three different single-mode PFs, which are typically used for the 2.0 µm wavelength, are investigated. During the experiments, the total length of the oscillator is fixed to 5 cm, while keeping the same gain fiber (4.2 cm). As can be observed, the UHNA7 PF with a mode area of ${A_{\rm eff}} = 8\,\, \unicode{x00B5} {{\rm m}^2}$ provides the best performance in terms of the mode-locking threshold and output power, i.e., 140 mW and $\sim{3}\;{\rm mW}$, respectively. Please note that the case using SM28e PF cannot enable mode locking even at a pump power approaching the damage threshold of the SESAM.

 

Fig. 2. Output power of the oscillator versus the pump power for different passive fibers, i.e., UHNA7 (blue circles), SM1950 (red diamonds), and SM28e (black triangles). In the experiment, the lengths of the laser cavity and TDF were fixed to 5 cm and 4.2 cm, respectively. The vertical lines indicate the mode-locking thresholds for cases of UHNA7 and SM1950, respectively.

Download Full Size | PDF

The reduction of mode-locking threshold can be mathematically understood based on the criterion achieved in a stability analysis [15,16], i.e.,

Then, we optimize the net dispersion of the oscillator to obtain a broader mode-locked optical spectrum, and thus a narrower pulse width. To this end, UHNA7 PFs with different lengths are studied, and the results of mode locking are analyzed in Fig. 3. It is clear that a 2.5 cm length of the UHNA7 PF provides a much broader optical spectrum, and thus a narrower pulse width (244 fs). The irregular distribution of small dips across the spectral profile corresponds to the featured water absorption. In this case, the corresponding net dispersion of the oscillator is calculated to be ${717}\;{{\rm fs}^2}$. Please note that other lengths of UHNA7 PFs have also been investigated (not shown in Fig. 3). The center wavelength of the mode-locked optical spectrum is notably blueshifted as the length of UHNA7 PF increases.

 

Fig. 3. (a) Optical spectra and (b) autocorrelation traces of mode-locked pulses using different lengths of UHNA7 passive fibers. In the experiment, the total lengths of the laser cavities were fixed to 5 cm.

Download Full Size | PDF

A detailed analysis is further performed for a 2.5 cm length of UHNA7, and the results are shown in Fig. 4. As shown in Fig. 4(a), the oscillator operates in the continuous wave (CW) regime at a pump threshold of 70 mW. In the pump range of 70–140 mW, on the other hand, the oscillator operates in the $Q$-switched mode-locking regime. Once the pump power is higher than 140 mW, the oscillator transits to the CW mode-locking regime. Figure 4(b) shows the pulse train of the CW mode-locking pulses. The pulse train has a temporal period of 0.507 ns, yielding a fundamental repetition rate of $\sim{2}\;{\rm GHz}$ [see Fig. 4(c)]. The parasitic sidelobes on both sides of the fundamental frequency resulted from the weak modulation, such a weak modulation caused by the birefringence effect on the mode-locked pulse train. The phase noise of the CW mode-locked pulses is also measured and shown in Fig. 3(d). As can be observed, the phase noise gradually decreases from $ - {13}\;{\rm dBc/Hz}$ to $ - {134}\;{\rm dBc/Hz}$ when the frequency changes from 10 Hz to 1 MHz. The timing jitter is high in the current free-running fiber laser system. To reduce the timing jitter, active stabilization techniques can be applied.

 

Fig. 4. Performance of the mode locking using a 2.5 cm length of UHNA7. (a) Output power as a function of pump power. QS, $Q$-switched; ML, mode locking. (b) Pulse train of the mode-locked pulses. Inset shows a longer pulse train. (c) RF spectrum of the mode-locked pulses. Inset shows a wider span RF spectrum. The resolution bandwidth of the measurement is 10 kHz. (d) Phase noise measurement of the mode-locked pulses.

Download Full Size | PDF

The high-power performance of this 2 GHz fs-pulsed laser is then studied, and results are presented in Fig. 5. Figure 5(a) shows the average power measured at the output of the high-power amplifier. The output power linearly increases with the pump power, and reaches $\sim{8}\;{\rm W}$ at the maximum pump power ($\sim{45}\;{\rm W}$ in this case). It is also noticed that no saturation effect is observed, which implies that an even higher output power can potentially be obtained by simply increasing the maximum pump power. At an average power of $\sim{8}\;{\rm W}$, the optical spectrum of the 2 GHz fs pulse is significantly broadened and redshifted [see Fig. 5(b)]. The redshifted optical spectrum can be mainly attributed to the Raman scattering effect. Such effect, however, is beneficial in this case. It is because the generated broader optical spectrum enables the reduction of pulse width from 244 fs (the oscillator) to 126 fs, as shown in Fig. 5(c). Please also note that, compared with cases of low repetition rate fs pulses (typically 10s of MHz), the pulse energy of the HRR fs pulses is 2 orders of magnitude lower. As a result, no supercontinuum spectrum is generated in such a long fiber link although a high-average power ($\sim{8}\;{\rm W}$) is delivered.

 

Fig. 5. Performance of the high-power amplification. (a) Output power of the high-power amplifier as a function of pump power. (b) Optical spectra of the oscillator and high-power amplifier. (c) Autocorrelation trace of the amplified fs pulses.

Download Full Size | PDF

 

Fig. 6. Scaling of the repetition rate. (a) Optical spectrum of the 2.7 GHz fs pulses. (b) Pulse train. (c) RF spectrum, measured with a resolution bandwidth of 10 kHz. (d) Autocorrelation trace.

Download Full Size | PDF

So far, a high power 2 GHz fs-pulsed all-fiber laser at 2.0 µm has been successfully demonstrated. Regarding the repetition rate, one may be interested in its scalability. For this concern, there are several potential solutions: (1) replacing the intracavity PF with those providing a higher opposite dispersion coefficient; (2) customizing SESAMs with appropriate dispersion for compensating the cavity dispersion; (3) performing chirped mirror coating on the fiber faced (i.e., the location of DF). As a proof of concept, we have studied the first solution, and the results are illustrated in Fig. 6. In this case, the original UHNA7 PF is replaced with a PF that has a higher dispersion coefficient, i.e., ${93}\;{{\rm fs}^2}/{\rm mm}$ at 2.0 µm (about twice higher than that of UHNA7). Thus, a shorter PF is required to compensate the cavity dispersion. Figures 6(a)–6(d) show the optical spectrum, pulse train, RF spectrum, and autocorrelation trace of the mode-locked resonator at a repetition rate of 2.75 GHz. The output power of the 2.75 GHz mode-locked pulses is 4.58 mW. The autocorrelation trace is measured after amplified to 97.2 mW due to the low sensitivity of InGaAs photoelectric detector of the autocorrelator.

In conclusion, a high-power GHz fs-pulsed all-fiber laser at 2.0 µm is demonstrated. A low mode-locking threshold, down to 140 mW, is realized by carefully choosing the intracavity PF, which also compensates the cavity dispersion and enables a broadband mode locking. The superior performance, including a compact design, high average power, ultrashort pulse width, and high repetition rate, is anticipated to be promising for a broad spectrum of applications in the fields of biophotonics, nanomaterial science, nonlinear optics, and industry.