1. Introduction

Passive mode-locked fiber lasers (PML-FLs) have gained considerable attention owing to their intriguing output characteristics in the field of physics and various industrial and biomedical applications, such as micromachining, tissue diagnosis, and disease treatment. Because of the broad gain bandwidth, a relatively high slope efficiency and various intrinsic advantages, such as robustness, compactness, reliability and portability, PML Yb-doped fiber laser (PML-YDFLs) have become one the most popular light sources for operating at approximately 1 $\mu$m. Through dispersion compensation inside the cavity, conventional solitons (CSs) with subpicosecond pulsewidth but lower pulse energy have been demonstrated in the net-anomalous cavity regime of the PML-YDFLs [1,2]. To boost the pulse energy of PML-YDFLs, dissipative solitons (DSs) have been studied in a net-normal or an all-normal dispersion (ANDi) cavity with an extra insertion length of single-mode fibers (SMFs) inside the cavity [3–5]. Owing to the extraordinarily complicated interaction among gain, nonlinearity, and dispersion inside the cavity, peculiar dynamics, such as bound soliton [6], multiple pulses [7], rain soliton [8], Q-switched and mode-locked pulses [9], and noise like pulses (NLPs) [10], have been widely investigated from PML-YDFLs.

NLPs exhibit relatively long (subnanosecond) wave packets that consist of a fine inner structure of subpicosecond pulses with randomly varying amplitude and duration. Since the first report by Horowitz et al. [10] about PML Er-doped fiber lasers (PML-EDFLs), various studies have been conducted to investigate this intriguing phenomenon of NLPs. NLPs have been demonstrated in numerous PML-FLs based on artificial saturable absorbers, such as nonlinear polarization rotation (NPR) [10–14], nonlinear loop mirrors [15], and real saturable absorber, such as graphene [16] and tungsten disulfide (WS2) [17]. One of the distinctive features of NLPs is the double-scale intensity autocorrelation (IAC) trace with a subpicosecond coherent spike riding on a broadband pedestal. Through the pulse compression based on grating pair (GP) outside the laser cavity [13], the 14.5-fs coherent spike has been experimentally generated from a Yb-doped fiber amplifier [14]. Benefiting from superior properties of a broad emission spectrum and low temporal coherence, NLPs possess relatively high potential to be applied in optical metrology, generation of low speckle image [18], and optical coherent tomography.In addition, a near-infrared (NIR) short-pulse laser can be adopted to obtain the multiphoton image of biological tissue and novel optical materials [19]. For the noninvasive measurement of the human eye and skin with ultrahigh resolution, an ultrafast laser of approximately 1.3 $\mu$m is a suitable light source for second-harmonic and third-harmonic microscopy [20,21].

In comparison with the CSs and DSs, a Raman scattering effect can assist NLPs to extend the output spectrum of PML-YDFL efficiently toward a longer wavelength and produce an even broader emission bandwidth [18,22] or supercontinuum generation (SCG) [23]. Based on the cascaded Raman scattering (CRS), the second-order Stokes wave (2$^{nd}$-SW) is excited to extend the flat emission spectrum of PML-YDFL to more than 1.1 $\mu$m after insertion of a certain length of SMFs inside the cavity [18]. In addition, the SCG can also be generated using the amplified NLPs with a relatively high pulse energy as seeded pulses through a certain length of SMFs [24,25]. Zaytsev et al. [24] reported NIR SCG toward 1050-1250 nm by launching a relatively high pulse energy NLPs into an amplifier state and following a certain length of SMFs. In their work, a 10 W pump power was adopted for a PML-YDFL using double cladding YDF as a gain medium to generate 45 nJ NLPs. Following, the output power of the NLPs was boosted up $\sim$ 3W by the amplifier stage and then passed through 100 m long SMFs for the NIR SCG. In this work, the 5.7 nJ NLPs was generated in PML-YDFL using single mode YDF as a gain medium. The broad and flat spectrum from 1148 to 1278 nm can be easily achieved by means of the nonlinear fiber amplifier with 532 mW pump power. In addition, the short-wavelength component of the NLP was filtered, and the first-order Stokes wave ($1^{st}$-SW) was adopted in order to reduce the gain competition between the Yb-doped fiber (YDF) and stimulated Raman scattering (SRS) for the extension of the emission spectrum toward 1.3 $\mu$m.

2. Experimental setup

The schematic setup of the ring cavity configuration of the YDFL is shown in Fig. 1(a). The gain medium was a 40-cm-long YDF (absorption of 1200 dB/m at 980 nm), which was pumped by a 976-nm laser diode (LD) through a 980/1040-nm wavelength division multiplexer (WDM). To generate stabilized mode-locked pulses, the NPR mechanism was adopted in combination with one half-wave-plate ($\lambda /2$, insertion loss $\sim$ 0.05 dB), two quarter-wave-plates ($\lambda /4$, insertion loss $\sim$ 0.05 dB), and a polarization beam splitter (PBS) in free space. A GP comprising two separated gratings (1000-lines/mm groove density) approximately about 3.5 cm, was used for dispersion compensation. The isolator (ISO, insertion loss $\sim$ 0.8 dB at 1040 nm) was used to ensure uni-directional propagation of pulsed light inside the laser cavity. The output of the YDFL was obtained from the rejection port of the PBS. To shift the emission spectrum of the NLP toward the longer wavelength and broaden the emission bandwidth around NIR, two YDFAs, i.e., preamplifier and a nonlinear amplifier, were adopted in this work as shown in Figs. 1(b) and (c). In this work, we did not control the polarization state of pump pulses after the oscillator and pre-amplifier. For the pre-amplifier in Fig. 1(b), the gain medium was a 5-m-long double-cladding Yb-doped fiber (absorption of cladding $\sim$ 0.715 dB/m, and estimated absorption of core $\sim$ 446 dB/m at 915 nm). In Fig. 1(c), a 20-m-long germanium-zirconia-silica Yb$^{3+}$-doped fiber (GZY fiber), manufactured by the modified chemical vapor deposition process coupled with the solution doping technique [26], was used as a gain medium of the nonlinear amplifier. The V number and mode field diameter of the GYZ fiber were 2.4 and 3.57$\mu$m, respectively, at 1064 nm. To induce third-order nonlinearity and SRS, the core of the silica-glass-based optical fibers was doped with high levels of GeO$_2$ and ZrO$_2$. The nonlinaer coefficient $\gamma$ of GYZ fiber is estimated to be about $\sim$ 19 W$^{-1}$Km$^{-1}$[26]. The time traces of NLP were detected by a high-speed photodetector (EOT Inc.) and monitored using a 2-GHz high-speed oscilloscope (200-GHz sampling scope, WaveRunner 620Zi, LeCroy Inc.). The optical spectra and pulsewidth of the NLPs were recorded by an optical spectrum analyzer (OSA, AQ-6370, Ando Inc.) and an intensity autocorrelator (FR-103XL, Femtochrome Research Inc.).

 

Fig. 1. (a) Ring cavity configuration of YDFL, LD: laser diode, WDM: wave division multiplexing, YDF: Yb-doped fiber, SMF: single mode fiber, Col: collimator, $\lambda /4$: quarter-wave plate, ${\lambda }/2$: half-wave plate, PBS: polarization beam splitter, GP: grating pair, ISO: isolator, (b) pre-amplifier, Yb-doped DCF: Yb-doped double-cladding fiber, and (c) nonlinear amplifier, GZY fiber: germanium-zirconia-silica Yb-doped fiber.

Download Full Size | PDF

3. Results and discussion

In our previous report [18], a number of SMFs with different lengths, from 25 m to 150 m, have been inserted inside cavity of PML-YDFL to induced the 2$^{nd}$-SW. Typically, the threshold of the SRS was reduced as the length of SMFs increased that was consistent with the formula:

 

Fig. 2. (a) Optical spectrum of NLP without (blue curve) and with (red curve) filter, and corresponding IAC trace of NLP (b) without and (c)with filter (Inset: coherent spike on the top of pedestal).

Download Full Size | PDF

3.1 Spectrum extension by means of the nonlinear fiber amplifier

The spectrum extension was investigated only by means of the nonlinear YDFA in Fig. 1(c). The NLP-I was adopted as seeded pulses with a power of approximately 10.3 mW, corresponding to 5.8-nJ pulse energy. Figures 3(a)-(f) show the evolution of the optical spectrum as the pump power of the amplifier increases. When the pump power of the nonlinear YDFA did not turn on ($P_{na}$= 0), the emission peak of the output pulses after the nonlinear YDFA revealed almost the same wavelength as the injected NLPs around 1075 nm in Fig. 3(a). Owing to the gain peak for the typical Yb-doped fiber of approximately 1030 nm, the emission peak at a shorter wavelength of approximately 1042.3 nm became stronger than that at a longer wavelength of about 1078 nm, with $P_{na}$ = 74 mW, as shown in Fig. 3(b). As $P_{na}$ increased to 195 mW, another emission peak emerged at a short wavelength of approximately 1027.3 nm, termed the pump wave, emerged, as in Fig. 3(c). In the meantime, the largest emission peak (S1) shifted to the 1080.7 nm, ($1^{st}$-SW). Relative to the S1, the other emission peak (S2) at 1136 nm revealed a frequency shift of approximately 13.5 THz, the $2^{nd}$-SW. As $P_{na}$ increased to 312 and 429 mW in Figs. 3(d) and (e), the third-order Stokes wave ($3^{rd}$-SW, $S_3$) and the fourth-order Stokes wave ($4^{th}$-SW, $S_4$) emerged at 1190 and 1250 nm, respectively. In combination with a number of nonlinear optical effect, such as self-phase modulation, four-wave mixing, and cross phase modulation, a relatively broad spectrum was generated in Fig. 3(f), as $P_{pn}$ at 532 mW. The spectrum reveals flat-top emission, ranging from 1148.3 to 1277.8 nm and 3-dB bandwidth of approximately 130 nm. The output power of NLP was approximately 98.4 mW which corresponds to a pulse energy of about 55.4 nJ.

 

Fig. 3. Spectrum evolution of NLP through nonlinear YDFA with (a) $P_{na}$ = 0, (b) $P_{na}$ = 74, (c) $P_{na}$ = 195, (d) $P_{na}$ = 312, (e) $P_{na}$ = 429, and (f) $P_{na}$ = 532 mW (see Visualization 1).

Download Full Size | PDF

3.2 Spectrum extension through YDFA system

To further shift the peak wavelength of NLP further toward the longer wavelength, the emission peak of NLP around 1075 nm was first amplified through a pre-amplifier, as in Fig. 1(b), and then passed through the nonlinear YDFA, as in Fig. 1(c). Two different kinds of NLP, NLP-I (blue solid curve) and NLP-II (red solid curve) in Fig. 2(a), were used as seeded pulses. Here, the power of the two seeded NLPs was fixed at approximately 1 mW, corresponding to a pulse energy of approximately 0.55 nJ. The amplified emission spectra of NLP-I and NLP-II after passing through the pre-amplifier with a pump power ($P_{pa}$) of about 1.3 W are illustrated by the blue (w/o filter) and red (w/ filter) curves in Fig. 4, respectively, with corresponding output powers of approximately 93.1 and 107 mW. When the NLP-I was used as a seeded pulse, the peak and shoulder wavelengths at 1076.2 and 1054.0 nm were amplified simultaneously by the pre-amplifier. Owing to the SRS, the longer emission peak at 1128.0 nm ($2^{nd}$-SW) was generated, with a 12.8-THz frequency shift relative to the peak wavelength of 1076.2 nm. By launching the NLP-II as a seeded pulse into the pre-amplifier, the short wavelength component of approximately 1054 nm was not apparent, and the peak wavelength revealed a slightly red shift to 1080 nm in the amplified spectrum (red curve) in contrast to the NLP-I (blue curve) in Fig. 4(a). In addition to the enhancement of peak intensity of $2^{nd}$-SW at 1130 nm in Fig. 4(a), the $3^{rd}$-SW with a peak wavelength at 1186.6 nm started to emerge in the logarithm scale of the spectrum, as in the inset of Fig. 4(a). It is attributed to the better Raman conversion efficiency of NLPs inside YDFA after filtering the short-wavelength component. By means of the pre-amplifier, the corresponding IAC traces of the amplified NLP-I and NLP-II show that the pedestal width $\tau _{p}$ of approximately 150 and 140 ps, Fig. 4(b) and the width of the coherent spikes $\tau _{s}$ of approximately 169 and 175 fs, as in the insets of Figs. 4(b) and (c) are relatively close.

 

Fig. 4. (a) Optical spectrum of NLP-I (blue curve, without filter) and NLP-II (red curve, with filter) through the pre-amplifier and corresponding IAC trace of (b) NLP-I and (c) NLP-II (inset: enlargement of coherent spike from IAC trace)

Download Full Size | PDF

In the following, an attempt is made to shift the amplified spectrum of NLP-I and NLP-II in Fig. 4 further toward longer wavelengths by means of the nonlinear YDFA, simultaneously. Figures 5(a)-(h) illustrate the spectra evolution of NLP-I as a function of the pump power ($P_{na}$) of the nonlinear amplifier. Here, the seed power of NLP-I before the nonlinear-amplifier is 93.1 mW, corresponding to a pulse energy of approximately 51.7 nJ. As the pump power of the nonlinear YDFA does not turn on ($P_{na}$ = 0 mW), the wavelengths of the three emission peaks, i.e., $S_1$ at 1056.0 nm, $I_1$ at 1076.1 nm, and $I_2$ at 1126.9 nm, in Fig. 5(a), were almost the same as the injected NLP-I blue curve in Fig. 4(a). Here, the largest emission peak, located at 1126.9 nm ($I_2$), and the Stokes wave $I_3$ emerged at 1184.8 nm. If the pump power of the nonlinear YDFA was futher increased, the intensity of emission peak $I_2$ declined obviously and converted its energy to the higher-order Stokes waves, such as $I_3$ at 1184.8 nm and $I_4$ at 1245.3 nm, owing to the CRS, as shown in Figs. 5(b)-(f). At same time, the 2$^{nd}$-SW ($S_2$ at 1098.9 nm) and 3$^{rd}$-SW ($S_3$ at 1164.0 nm), relative to the pump wave ($S_1$), are also shown in Figs. 5(c) and (e), with $P_{na}$ at 74 and 253 mW, respectively. As the pump power increased, the emission peaks, i.e., $S_1$, $S_2$, and $S_3$ shifted obviously toward the shorter wavelength, as in Figs. 5(c)-(h). This is shown (brown arrow) in the contour plot in Fig. 7(a). Owing to the largest gain of YDFA of approximately 1034.7 nm, $S_1$ revealed a blue shift as pump energy increased and then caused the following shift of the high-order Stokes waves, such as $S_2$ and $S_3$, toward a short wavelength. Because the gain competition between YDF and stimulated Raman scattering, the fourth order stoke wave $S_4$ at 1212 nm in Fig. 5(g) obtained higher gain than the other fourth-order stoke wave $I_4$ at 1245.3 nm and became the largest emission peak with $P_{na}$= 532 mW in Fig. 5(h). Thus, the emission peak of the fourth Stoke wave from NLP-I in combination of pre-amplifier and nonlinear amplifier reveals slightly shorter wavelength than that by only using nonlinear YDFA in Fig. 3(f). Here, the 3-dB bandwidth ($\Delta \lambda$) of $S_4$ is approximately 106.1 nm and the output power of NLP is approximately 122 mW, corresponding to a pulse energy of about 67.8 nJ.

 

Fig. 5. Spectrum evolution of amplified NLP-I through pre-amplifier and nonlinear amplifier as (a) $P_{na}$ = 0, (b) $P_{na}$ = 20, (c) $P_{na}$ = 74, (d) $P_{na}$ = 195, (e) $P_{na}$ = 253, (f) $P_{na}$ = 312, (g) $P_{na}$ = 429, and (h) $P_{na}$ = 532 mW.

Download Full Size | PDF

In the following, the spectrum evolution was investigated using the filtered NPs as seeded pulses of the amplifier system. After pre-amplifier, the power and corresponding pulse energy of NLP-II were 107 mW and 59.4 nJ, respectively. Figures 6(a)-(f) show the spectrum evolution as a function of the pump power of the nonlinear amplifier. Because $P_{na}$ = 0 mW, the wavelengths of the three emission peaks, i.e., $I_1$ at 1083.2 nm, $I_2$ at 1130.0 nm, and $I_3$ at 1186.6 nm, in Fig. 6(a) are also similar to the spectrum of amplified NLP-II, as in the red curve in Fig. 4(a). Nevertheless, the additional emission peak $S_1$ emerged at 1071.5 nm, resulting from the gain of the nonlinear YDFA. Unlike the previous case in Fig. 5, the short-wavelength components of NLP-II less than 1050 nm were filtered, so that the gain competition between nonlinear YDFA and SRS can be avoided. Thus, the Raman conversion efficiency is better than the previous one. In Fig. 6(b), the Stokes wave $I_4$ was obvious at 1246 nm as $P_{na}$ increased to 74 mW. As $P_{na}$ =195 mW in Fig. 6(c), the emission peak $I_1$ revealed a slightly blue shift toward the higher gain of YDF and then merged into the emission peak $S_1$. This caused a slight blue shift of the higher-order Stokes wave such as $I_2$ and $I_3$. The fifth-order Stokes wave (5$^{th}$-SW, $I_5$) was induced at 1303.5 nm, as in Fig. 6(e), as $P_{na}$ = 429 mW and became the highest emission peak at 1304.2 nm in Fig. 6(f) as $P_{na}$ = 532 mW. The output power of NLP after the nonlinear fiber amplifier system was approximately 117 mW, which corresponds to the pulse energy of approximately 65.0 nJ. The emission peak of NLP-II can extend to a longer wavelength around 1304 nm with a 3-dB bandwidth ($\Delta \lambda$) around 108.5 nm. The evolution of the amplified spectrum of the NLP-II as a function of pump power is also revealed in the contour plot shown in Fig. 7(b). In comparison with Fig. 7(a), with obvious gain competition between two groups of modes, the red shift of Raman soliton (red area) in Fig. 7(b) is more continuous and efficient.

 

Fig. 6. Spectrum evolution of amplified NLP-II through the Yb-doped fiber amplifier system as (a) $P_{na}$ = 0, (b) $P_{na}$ = 74, (c) $P_{na}$ = 195, (d) $P_{na}$ = 312, (e) $P_{na}$ = 429, and (f) $P_{na}$ = 532 mW (see Visualization 2.)

Download Full Size | PDF

 

Fig. 7. Contour plot of spectrum evolution from (a) NLP-I and (b) NLP-II as a function of pump power $P_{na}$

Download Full Size | PDF

3.3 Application of NIR NLPs for the photoluminescence measurement

The optical spectra of amplified NLP-I (blue curve) and NLP-II (red curve) through the YDFA system with $P_{na}$ = 532 mW are shown in Fig. 8(a). The emission peaks are located at 1212.1 and 1304.2 nm with a 3-dB bandwidth ($\Delta \lambda$) of approximately 106.1 and 108.5 nm. In comparison with NLP-I, the red shift of the emission peaks of the amplified spectrum of approximately 92.2 nm for the NLP-II is obvious. In addition, the intensity contrast ($I_m$/$I_s$) of the main emission peak ($I_m$) relative to the side emission peak ($I_s$) is higher for the NLP-II by 3.13 relative to the NLP-I of about 1.79. As in the previous description, the conversion efficiency of Raman scattering is higher using the filter NLP-II as a seeded pulse, because the gain competition with YDFA was avoided. The corresponding IAC traces of NLP-I and NLP-II after the YDFA system are shown in Figs. 8(b) and (c). The pedestal width of both NLPs broadens over 90 ps, so that the $\tau _p$ cannot be obtained by the IAC trace. The inset of Figs. 8(b) and (c) reveals that the width of coherent spikes ($\tau _s$) are approximately 174 and 167 fs, respectively.

 

Fig. 8. (a) Amplified spectrum of NLP-I (blue curve) and NLP-II (red curve) through the YDFA system (inset: corresponding spectrum in logarithm scale), corresponding IAC trace from (b) NLP-I and (c) NLP-II (inset: coherent spike on the top of pedestal), (d) experimental setup of nonlinear microscopy (inset: picture of perovskite crystal), and(e) transmittance spectrum (blue curve) and PL spectrum resulting from three-PA (brown solid and dash curve) and one-PA (red solid curve). Inset shows the logarithm peak intensity (Log($I_{TPL}$)) from three photon absorption of perovskite as a function of the logarithm excitation pump power (Log($P_{exc}$)) of NIR NLPs.

Download Full Size | PDF

The generated NIR pulsed light (approximately 1.3 $\mu$m) can be an excellent light source to obtain the emission spectrum from three-photon absorption (three-PA) of CH$_3$NH$_3$PbBr$_3$ perovskite by a nonlinear microscopy system in Fig. 8(d). After reflection by a dichroic mirror (DM), the pulse light was focused onto the perovskite crystal by an objective lens (40x, N.A. = 1.55). Then, the emission spectrum of the sample passed through the DM and was then measured by the spectrometer. Here, the CH$_3$NH$_3$PbBr$_3$ perovskite was synthesized by a modified solution inverse temperature crystallization method (dimension $\sim$5$\times$5$\times$1 mm), with the picture shown in lower left of Fig. 8(d). The transmittance spectrum (blue solid curve) and the excited photoluminescence (PL) spectrum (red solid curve) by the ultraviolet light are also shown in Fig. 8(e) (red line). In comparison with the PL from one-photon absorption (one-PA) with an emission peak at 545 nm (2.275 eV), the PL spectrum by the three-PA appears obviously red-shifted (brown solid line).

Unlike the PL excited near the surface, the NIR source has a deeper penetration depth to excite the emission light at the interior of pervoskite. As the generated photon inside interior of sample diffused to the surface of the sample, a reabsorption effect occurred to cause a redshift of the sample [19]. It is obviously to see that the the emission peak of the perovskite reveals slightly red shift from 2.245 eV (552 nm, dash brown line) to the 2.222 eV (558 nm, solid brown line) while the peak wavelength of NIR NLP shift from 1212 nm to the 1300 nm. For three-photon absorption processes, the photoluminescence intensity $I_{TPL}$ is connected with the excitation power $P_{exc}$ by $I_{TPL}$ = a $P^n_{exc}$. Here, $a$ is a constant and $n$=3 represents the number of absorbed photons. Inset of Fig. 8(e) represent the relation of logarithm peak intensity (Log($I_{TPL}$)) from three photon absorption of perovskite as a function of the logarithm excitation pump power (Log($P_{exc}$)) of NIR NLPs. By the fitting, the slope around 3 can be further used to demonstrate three photon absorption of perovskite by the excitation of NIR NLPs.

4. Conclusion

NIR noise like pulses (NLPs) were demonstrated through a YDFA system. After insertion of a 100-m-long SMF inside the cavity of the YDFL, 1.8-MHz low-repetition-rate NLPs were generated through the nonlinear polarization rotation technique. By means of the nonlinear YDFA using a GZY fiber as a gain medium, the emission spectrum of NLPs was extended to approximately 1.2 $\mu$m with a flat 3dB bandwidth of approximately 130 nm based on the CRS to excite the fourth-order Stokes wave. The high-order Stokes wave generation was also compared using NLPs with a broad spectrum bandwidth and filtered spectrum (with emission peak of approximately 1075 nm) as seeded pulses. Owing to the better conversion efficiency of SRS without gain competition of Yb-doped fiber, using filtered NPLs as seed pulses the fifth-order Stokes wave could be excited to shift the peak wavelength of NLP further at around 1.3 $\mu$m. The amplified NIR NLPs after the YDFA system with a 167-fs coherent spike can be an excellent light source for the PL emission from the three-PA of perovskite. The emission peak reveals a slight red shift relative to the one-PA owing to the deep penetration depth inside the material to produce the reabsorption effect.