Due to the low reliance on polarization states of saturable absorber (SA), passively mode-locked fiber laser based
on SA have recently attract great attention
[1]–
[3]. In order to search for high-performance SA, different kinds of nonlinear
optical material including semiconductor saturable absorber mirrors (SESAMs), single-wall carbon nanotubes (SWCNTs),
graphene, and graphene oxide (GO) have been widely exploited
[4]–
[6]. However, SESAMs are usually considered as expensive and
complicated-fabrication devices; moreover, its operation bandwidth is limited at tens of nanometers
[7]. Concerning SWCNTs, it needs additional band-gap engineering technologies
by controlling the diameter and chirality in order to fit for special wavebands
[8]. On the other hand, graphene SA, owing to its Dirac electron property,
shows some advantages, such as wavelength-independent saturable absorbing characters, low saturable absorbing
threshold, and large modulation depth
[9]. Recently, topological insulators (TIs), as a rising material, have
attracted extensive interest in the field of photonics. It was experimentally found that
$\hbox{Bi}_{2}\hbox{Te}_{3}$,
$\hbox{Bi}_{2}\hbox{Se}_{3}$, and
$\hbox{Sb}_{2}\hbox{Te}_{3}
$ are also characterized by graphene-like electronic-band structure and exhibits Dirac-like
linear band dispersion
[10],
[11]. Bernard
et al. found that TIs exhibited saturable absorption behavior around at the wavelength of 1.55
$\mu\hbox{m}$
[12]. On this basis, several different research teams had achieved ultrafast
fiber laser mode-locked by inserting the TIs based SA into the laser cavity
[13],
[14]. However, in these reports, the TI based SA was fabricated with a quartz
plate or onto the fiber end facet, which rendering the interaction length between the light and the SA very short. So
the impact of the nonlinear effect of the SA on the pulse shaping was not evident. Whereas, in combination with the
evanescent wave operation mechanism, Luo
et al. had experimentally demonstrated the generation of 2 GHz harmonic mode-locked fiber laser by a
microfiber-based
$\hbox{Bi}_{2}\hbox{Te}_{3}$ SA
[15]. This structure could effectively increase the interaction length between
the light and the TIs. In this case, the nonlinear effect of the TI-based SA in fiber laser could be enhanced, making
it very suitable for generating HML pulse. For the wide applications as astronomical frequency combs of high repetition
rate pulse fiber lasers
[16], such a result was of great value. However, the modulation depth of the
fabricated SA device in that report was very low (
$\sim$1.7%)
, and the authors did not measure laser signal-noise-ratio (SNR)
, which was an important characteristic of the mode-locked fiber laser.
In this contribution, we would like to fill this gap. We have demonstrated an Erbium-doped fiber (EDF) laser
passively mode-locked by microfiber-based
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ SA. Different HML states with a repetition rate tunable from 232 MHz to 390 MHz were
achieved. The spectra exhibited perfect Gaussian profile and Kelly sidebands without any CW component. Under different
input pump powers, the 3-dB bandwidth of the spectra was kept at 2.1 nm without change and the pulse duration fixed at
1.32 ps unvaried. The measured SNR was 60 dB. The obtained SNR in mode locking fiber (based on nonlinear polarization
rotation, SESAM film, or
$\hbox{MoS}_{2}$ film) was usually between 50 dBm to 60 dBm
[2],
[5],
[17]. The reported SNR obtained in microfiber-based mode-locked laser was 50
dBm
[18]. Therefore, the obtained SBR in our experiment was on the average level
in fiber lasers, but a high value in microfiber-based lasers. In addition, we had observed the mode-locking state for 4
hours, and there was no significant degradation of the laser spectra, pulse train on oscilloscope, as well as the
output power. Compared with the conventional SESAM absorbers or film-based absorbers of other material, in the
Evanescent-Light-based absorber, only part of the transmitting light power acted with the absorber material, and it was
uneasy to damage the absorber device. So the potential lifetime of such absorber device could be extended dramatically.
Our result proved that the TI base SA possessed a high performance as a nonlinear photonic device in fiber laser
system.
SECTION 2.
Fabrication and Characteristic Measurement of the Microfiber-Based
$\hbox{Bi}_{2}\hbox{Te}_{3}$ SA
The first step was to fabricate the
$\hbox{Bi}_{2}\hbox{Te}_{3}$-polyvinyl alcohol (PVA) solution. The
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ nano-platelets were synthesized by hydrothermal intercalation. The
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ nano-platelets were dispersed in deionized water. Sodium dodecyl sulfate power was put
into the
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ dispersion using as a surfactant. The dispersion was ultrasonically agitated for 6 hours.
Then some PVA power was dissolved in deionized water with ultrasonic agitation at 90 °C for 3 hours. The
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ dispersion and PVA solution uniformly mixed and were prepared for using. The second act
was the fabrication of tapered fiber similar to those in
[18]. A bare single-mode-fiber (SMF, Corning SMF-28) was heated by an alcohol
flame and stretched at the same time. And the waist diameter of the fiber could be tapered down to 23.8
$\mu\hbox{m}$ as shown in
Fig. 1(a). Then the microfiber was fixed on a U-shaped frame. After the
preparation of the microfiber, we injected
a CW laser (home-made EDF laser: operating wavelength of 1563 nm as shown in Inset-I of
Fig. 1(b); output power of 18 mW) into the microfiber as shown in
Fig. 1(b). We observed it with an infrared viewer, and at the same time, we
tested the output power out of the microfiber with a pump power continuously. However, no evanescent wave and output
power change had been observed. Then, we placed a glassplate under the microfiber; the distance between the microfiber
and the surface of the glassplate was about 2 mm. After that, the
$\hbox{Bi}_{2}\hbox{Te}_{3}
$-PVA solution was dripped onto the glassplate until it covered the microfiber. Just at
this time, we still couldn't observe evanescent wave and the output power had no distinct decreasing. However, just
after several seconds, very weak evanescent wave light from the microfiber could be seen through infrared viewer, and
it became brighter and brighter. Meanwhile, the monitored output power began to decrease. When the output power
decreased to 16 mw, the brightness of the evanescent light no longer changed. The Inset-II of
Fig. 1(b) showed the observed evanescent light-spot (the upper bright one
was the evanescent light from the microfiber and the lower dim one was the corresponding image formed by the optical
platform), this picture was taken through the infrared viewer. The process took only about one minute. Afterwards, we
separated the microfiber from the
$\hbox{Bi}_{2}\hbox{Te}_{3}
$-PVA solution and cut off the CW light source. Compared with other similar reports
[19]–
[21], the deposition time was much shorter. After that, we transferred it to a
microscope, and observed the situation of the
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ deposition with a magnification of 50-fold objective lens (combination with a 10-fold
eyepiece). As shown in
Fig. 1(c), the deposition was a little matte and tightly surrounded the
tapped fiber. The diameter of the deposition was measured to be
$\sim\! 86\ \mu\hbox{m}$, the length of that was
$\sim\!\! 162\ \mu\hbox{m}
$. Finally, the as-fabricated microfiber-based
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ SA was evaporated at room temperature.
In order to further investigate characteristics of the fabricated microfiber-based
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ SA, we measured its linear and nonlinear transmission curve as shown in
Fig. 2. First, we injected a broadband light source into the microfiber and
measured its transmittance. Since the other part of the SA device was SMF-28 fiber, it was single mode only around 1.55
$\mu\hbox{m}
$, so we gave the results of the wavelengths ranging from 1300 to 1700 nm here. As shown in
Fig. 2(a), the linear transmittance of the SA device was about 23.1% at the
wavelength of 1564 nm. Due to the instability of the light source, the measured curve was rough and performed strong
spikes at the longer wavelength. Secondly, we tested the nonlinear transmittance of the SA device with a home-made
dispersion management mode-locked fiber laser whose central wavelength was
$\sim$1578 nm. As shown in
Fig. 2(b), the saturable absorption data of the microfiber-based SA and the
corresponding fitting curve were given as a function of the injected laser power. As can be seen in
Fig. 2, the modulation depth was
$\sim$4.8% and the nonsaturable loss was
$\sim$73.4%. Correspondingly, the inserting loss of the microfiber
based SA was about 7.34 dB. Compared with the results in Ref.
[15], the inserting loss was a little high, but the modulation depth had been
improved distinctly.
SECTION 3.
Experimental Setup
After having prepared the microfiber-based
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ SA, we inserted the mode-locking device into the fiber laser cavity. The configuration of
proposed fiber laser system was schematically shown in
Fig. 3.
One 976-nm single-mode laser diodes (LD) with maximum power of 650 mW was used to provide pump through a 980/1053 nm
wavelength-division-multiplexer (WDM). A 10-m EDF (Nufern with core diameter of around 8
$\mu\hbox{m}$, cladding diameter of 125
$\mu\hbox{m}
$, Numerical Aperture of 0.11, and Core Absorption of about 4.5 dB/m at 976 nm)
acted as the gain media in the oscillator. The other fibers in the laser system together with the pigtail of the
passive components were Corning SMF-28 with dispersion parameter D of 17 ps/nm/km. The total length of the oscillator
was
$\sim$37 m, which corresponded to a fundamental repetition of
$\sim$5.5 MHz. One polarization insensitive isolators (ISO)
was used to ensure unidirectional operation in the oscillator. One set of polarization controller (PC)
was used in the oscillator for adjusting the linear cavity birefringence and selecting laser wavelength. A fused
optical coupler (OC)
with 10% output was placed after it as the output port. The fabricated microfiber device worked as the absorber. In
some reports, a spectral filer was incorporated in the oscillator to reach stable mode locking
[5]. What's more, the filter could be used to controlling the pulse duration,
spectral width and wavelength of the output
[22],
[23]. However, there were also some reports that had reaching stable mode
locking without a filter
[2],
[17]. On the other hand, in our experiment, the bandwidth of the WDM, ISO, and
output coupler were 40 nm, which could also act as a filter. Therefore, we did not embed a filter in our experiment.
The output prosperities of the laser oscillator and amplifier were monitored by a power meter, an optical spectrum
analyzer (OSA), an autocorrelator (AC), a radio-frequency analyzer (RFA), and a 6-GHz digital oscilloscope together
with a home-made 2.5-GHz photodiode detector.
SECTION 4.
Experimental Results and Analysis
Under different pump power, Q-switched mode locking, soliton rains, and bunched solitons could all be obtained in
this experiment. However, we only paid attention to the HML state in this report. With appropriate PC orientation,
stable self-started mode-locking could be achieved when the pump power reached 80 mW. The average output power was
$\sim$5.3 mW. As the blue line showed in
Fig. 4(a), the mode-locked spectrum had a 3-dB bandwidth of
$\sim$2.1 nm with the center wavelength of 1564 nm. The spectrum
displayed smooth Gaussian profile and symmetric Kelly sidebands, which were typical characteristics of conventional
solitons. Moreover, there was no CW component on the spectrum so that we could speculate that the noise in our
experiment was very low. The pulse train detected with an oscilloscope was shown in
Fig. 4(b), in which pulses had relatively uniform intensity and temporal
interval. The corresponding autocorrelation trace exhibited smooth profile as shown in
Fig. 4(c), the blue curve was the experimental result, and the red one was
the Sech2-fit result. Since it was conventional soliton, the pulse should be Sech2 shape in temporal domain. The full
width at half maximum (FWHM) of the autocorrelation trace was determined to be
$\sim$2.06 ps, as shown. If a Sech
$^{2}
$ temporal profile was assumed, the pulse width was estimated as
$\sim$1.34 ps, the corresponding time-bandwidth product was
$\sim$0.345, which was close to the transmission limit. The RF
spectra in
Fig. 4(d) showed that SNR was
$\sim$60 dB and the fundamental peak was located at
$\sim$232.14 MHz as determined by the pulse train. It corresponded
to 42
th HML state. Above all we could confirm that the fiber laser worked at harmonic mode locking state not at
Q-switched or bunched soliton.
When increasing the pump power to 120 mW, the laser oscillator maintained mode-locking.
Fig. 4(a) showed the output spectra at different pump power. It was obvious
that the spectral intensity broadened monotonously with the enhancement of pump power, while the 3-dB bandwidth kept
almost unchanged. The measured pulse duration was still
$\sim$1.34 ps without distinct change. The pulse intensity became
more uniform. The pulse interval narrowed and the pulse repetition rate, respectively, increased to
$\sim$314.47 MHz (
$\hbox{Pump}=100\ \hbox{mW}$) and
$\sim$390.02 MHz (
$\hbox{Pump}=120\ \hbox{mW}$), as shown in
Fig. 5. They corresponded to 57th and 71st HML states.
Further increasing or decreasing the pump power out of the stable mode-locking region, CW component would arise on
the spectra as shown in
Fig. 6, and the pulse train became unstable.
In summary, we had demonstrated a passively HML fiber laser by inserting a microfiber-based
$\hbox{Bi}_{2}\hbox{Te}_{3}
$ SA. Taking advantage of the high nonlinearity and SA effect introduced by the SA device,
HML operation could be easily initiated. With different input power, the repetition rates could vary from 232 MHz to
390 MHz, which respectively corresponded to the 42nd and 57th HML states. The SNR was measured as 60 dB, which proved
the high performance of the SA device in our fiber laser.
Contents 
