1. Introduction

Mode-locked fiber lasers are valuable sources of high-power ultrashort light pulses, widely used in science and increasingly exploited in industrial applications. While semiconductor-based saturable absorbing mirrors [1] often provide excellent performance, in-fiber nonlinear modulation is attractive because of the femtosecond response time of silica and the high-damage threshold operation of fiber waveguides. Two mechanisms are generally exploited to provide for in-fiber passive modulation, leading to the selective transmission of high optical intensities in the laser. The first one involves the use of a nonlinear interferometer (e.g., Sagnac), where an intensity-dependent phase shift results in cutting of the low amplitude pulse wings. Soliton fiber lasers based on the nonlinear optical loop mirror [2] and the nonlinear amplifying loop mirror (NALM) [3] are based on this concept [4]. The other mechanism involves the use of a polarizer to cut the low-amplitude pulse wings [5], exploiting intensity-dependent birefringence that causes nonlinear polarization rotation [6]. Although passively mode-locked fiber lasers are simple, the pulse formation mechanism can be erratic – by slight changes in surrounding conditions the laser can go from not being self-starting to generating multiple pulses per roundtrip at random times [4,6]. These problems as well as a relatively narrow parameter range where stable mode-locking is achieved, limit the wider use of such lasers in commercial applications.

By introducing an active element in the cavity, mode-locking can be guaranteed. Laser systems employing a LiNbO₃ modulator in a fiber loop have been described [7,8], seeding the laser with a short duration pulse that evolves into a soliton. More recently, the replacement of the ferroelectric crystal in a laser by all-fiber acousto-optical [9,10] or electro-optic [11] modulators has been reported in the literature. These modulators work in the few MHz range in order to match the roundtrip of the laser and allow for the generation of relatively long pulses (tens to hundreds of picoseconds).

In this paper, the use is described of an electro-optic fiber in a Sagnac loop to actively mode-lock an all-fiber laser. The Sagnac produces a nanosecond open-time window creating a pulse inside a fiber laser ring cavity that is amplified and further compressed to the femtosecond range, exploiting nonlinear polarization rotation. Although the modulator operates in the MHz regime, the laser pulses produced are orders of magnitude shorter than the gating window, making it unnecessary to increase the modulator response to GHz for femtosecond pulse generation.

2. The active modulator

Active intracavity modulation is achieved here with an all-fiber 4-m long Sagnac interferometer containing a phase modulator.

2.1. The electro-optic fiber phase-modulator

This phase modulator consists of a 78-cm long piece of silica fiber that is made electro- optic by electrostatic charging, as described in [12]. The 125 µm fiber has a core diameter of 5.1 µm and a 1.38 µm wavelength cut-off. The two holes are 27 µm in diameter. The thermal poling [13] of the fiber [14–16] is accomplished by biasing the internal electrodes with + 5 kV for 195 minutes at 265 °C. After cooling to room temperature, this fiber exhibits a second-order nonlinear coefficient for the electro-optic effect χ⁽²⁾_eff ~0.26 pm/V at 1.5 µm. The electro-optic fiber is spliced to standard telecommunications fiber (SMF28), wound in a 4.5-cm diameter loop and packaged. It has a ~4.5 dB total insertion loss at 1.5 µm including splices. Typical splice loss between 2-hole fiber and SMF28 is 0.5 dB. The polarization dependent loss of the 78 cm 2-hole fiber with electrodes is ~1.7 dB. A schematic representation of the packaged phase modulator is displayed in Fig. 1 together with a scanning electron microscope (SEM) image of the fiber used and a picture of the packaged component.

 

Fig. 1 Sketch of the electro- optic phase modulator (center), together with a SEM image of the 2-hole fiber cross-section (left) and a photograph of the component with a ball pen as reference (right). This 2-hole fiber has 78 cm electrodes and is spliced with SMF28 fiber leads.

Download Full Size | PDF

The V_π of this device is ~140 V and has remained unchanged for over five years. The phase modulator acts as a capacitance with an RC time-constant of 10 ns when biased with a voltage step (which has a 10-90% risetime of 1.5 ns) through a 50 Ω coaxial cable. This implies that the phase modulator has a theoretical electrical 3 dB cutoff frequency ~16 MHz. Heat develops as the periodic voltage pulses travel along the finite-resistance internal fiber electrodes (~3 Ω/cm). Therefore, the removal of this heat limits the repetition rate of the phase modulator.

The component used in this work was originally manufactured for low repetition rate applications and mounted on a plastic substrate with low thermal conductivity. This sets a relatively low limit of operation to a few MHz with the application of V_π voltage steps.

2.2. The Sagnac as an amplitude modulator

Amplitude modulation is achieved by placing the phase modulator in a Sagnac loop and driving it with short voltage pulses generated by a CMOS circuit. Typical traces of the applied voltage and the resulting optical transmission of the Sagnac loop are displayed in Fig. 2 .

 

Fig. 2 Optical transmission of the Sagnac loop (blue) when 70 V voltage pulses (red) are applied to the phase modulator @ 1.88MHz.

Download Full Size | PDF

The driving pulses have tens of volts amplitude, 4 ns risetime and ~8 ns duration, and are triggered by a 5-V pulse generator with nominal <25 ps root-mean-square jitter. The Sagnac loop is all spliced, but not polarization maintaining (PM), and therefore it also incorporates a mechanical polarization controller. By placing the phase modulator slightly off-center in the Sagnac loop [17], a ~8 ns transmission window FWHM is triggered by the pulse generator.

The modulation depth achieved with the voltage pulses can be determined by examining the reflected light of the Sagnac loop.

3. Experimental setup

A schematic diagram of the actively mode-locked fiber ring laser is displayed in Fig. 3 . The gain medium consists of 22-m Erbium doped fiber (0.7 at%), which is core-pumped by a 100 mW laser diode emitting at 976 nm wavelength through a wavelength division multiplexer (WDM). The Er-doped fiber absorbs 88% of the pump power. Single polarization lasing and clockwise propagation are ensured by using a fiber polarizer together with a polarization controller (PC1) and an isolator. A fiber coupler provides 70% feedback to the cavity and removes 30% as laser output. A second polarization controller PC2 ensures that light with optimal polarization enters the Sagnac loop, which consists of a 3 dB fiber coupler, the phase modulator, and a third polarization controller. Here, the Sagnac loop acts as an amplitude modulator with an 8 ns open time-window, and not as a nonlinear loop mirror. Finally, a 38 m piece of SMF28 renders the cavity dispersion anomalous. The extra SMF28 also keeps the repetition rate relatively low (~2.6 MHz). The total fiber length of the laser is 80 m.

 

Fig. 3 Schematic of the mode-locked ring fiber laser with total length of 80 m. The phase modulator is placed off center in the Sagnac loop that is marked with the dotted line. The transmission window of the loop is displayed in the inset.

Download Full Size | PDF

The output is characterized with an optical spectrum analyzer (OSA), a commercial noncollinear intensity autocorrelator and a 1.2 GHz bandwidth photodiode connected to an electrical spectrum analyzer (ESA). An oscilloscope is also used for general observation of the pulse.

4. Results

Mode-locking is observed when the pump power exceeds 9 mW. Stable soliton pulses are generated when the PCs are properly set, the modulation depth is between 20 and 40% and the repetition rate is matched to the fundamental frequency by a few Hz (2.625348 MHz) at threshold. A relatively constant pulse-duration of 460 fs FWHM (assuming sech² pulse shape) is obtained for various operating conditions, e.g. different pump power or switching voltages. The soliton pulses are generated upon re-starting the phase modulator or the laser diode pump. No optical pulses are observed without driving the phase-modulator.

With improper PC settings, stable noise-like sub nanosecond pulses appear [18,19]. These pulses have a Gaussian shaped spectrum and a coherence spike corresponding to its spectral width.

Typical output traces for the correct settings for soliton generation are displayed in Fig. 4 . The measurements are made with 25 mW pump power and 62 V peak switching voltage, corresponding to 25% modulation depth or 60° phase shift. The pulse duration and spectral FWHM are 460 fs and 5.3 nm respectively. These values give a time-bandwidth product 0.30, close to the expected 0.315, for sech² pulses. The side lobes seen in the spectrum of the inset in Fig. 4a are characteristic of solitons that are periodically amplified [20].

 

Fig. 4 (a) Typical intensity autocorrelation trace displays a pulse with 460 fs deconvoluted FWHM. The inset shows the optical spectrum. (b) Normalized RF-spectrum of laser output (blue), at the fundamental frequency (2.6 MHz), and the electrical signal sent to the phase modulator (red). The inset shows the frequency comb of the optical signal.

Download Full Size | PDF

The RF-spectrum around the fundamental frequency (2.625 MHz) is displayed in Fig. 4b (blue), measured with the photodiode connected to the ESA with 100 Hz resolution bandwidth. It shows broad sidebands indicating some amplitude modulation. The RF-spectrum of the voltage pulses applied to the phase modulator is displayed as a reference (red). It is measured with 40 dB attenuation (electrical) and has a noise-floor 20 dB (electrical) lower than that of the photodiode. The inset in Fig. 4b displays the characteristic frequency comb of the mode-locked laser.

The average laser output power and peak power of the pulses are displayed as a function of pump power in Fig. 5a . A linear dependence is observed between the average output power and the pump power. As expected, the soliton peak power does not follow this linear dependence with the pump since it cannot grow indefinitely, shedding energy to the dispersive wave at higher pump levels.

 

Fig. 5 (a) Average output power (blue), and peak power of output pulses (green), as a function of pump power. The peak power is a relative measurement from the autocorrelator. (b) Consecutive autocorrelation measurement over 30 min.

Download Full Size | PDF

The laser is sensitive to temperature fluctuations since it relies on non-PM fibers. However placing most of the laser inside an enclosure removes the influence of airflows. With this simple measure it exhibits stable operation for at least 30 min of consecutive measurements, as shown in Fig. 5b, without further adjustments [21]. The peak power in this measurement has a relative standard deviation of 2%.

When the switching voltage is varied, the temperature of the phase modulator shifts accordingly. To compensate for this drift the frequency and PC settings must be adjusted. Stable mode-locking is easily achieved between 20 and 40% modulation depth. Outside of this region, stable mode-locking can still be achieved. However, the laser is then closer to a state of Q-switched mode-locking.

The allowed frequency detuning for the soliton generation is rather small due to the short open time window. When the frequency is detuned by a few tens of Hz, the laser jumps to a different mode of operation for which the PC settings are not optimized.

Since the laser is actively modulated it is straightforward to utilize harmonic mode-locking to increase the pulse repetition frequency. Here, the laser is tested at frequencies as high as 108 MHz. However, due to limited frequency response of the CMOS circuit the phase modulator is driven directly by a 5 V pulse generator. The laser still generates femtosecond pulses up to the 10^th harmonic (27 MHz). At even higher harmonics sub-picosecond pulses are only obtained in a Q-switched mode-locked regime. This is attributed to the limited pump power available and the small modulation depth (<1%) of the modulator driven with the 5-V pulses.

5. Discussions and conclusion

The phase modulator is mainly driven with voltages producing 20-40% amplitude modulation depth in the cavity. Larger modulation can easily be achieved by increasing the applied signal. However, this is both unnecessary for mode-locking and is accomplished at the expense of increased heating, which lowers the repetition rate limit. Better thermal handling than implemented here should bring the maximum repetition rate of the phase modulator close to the theoretical limit (16 MHz).

It should be possible to extend the results presented here to a mode-locked fiber laser based on PM fibers, since the presence of the electrodes gives the phase modulator a noticeable birefringence (∆n~10⁻⁵). The cavity configuration should then be re-designed to exploit for example self-phase modulation as pulse-shortening mechanism.

The use of an electro-optic fiber in the cavity allows for tuning the cavity length with a DC bias. Although this is not exploited here, such active adjustment could be used as an inexpensive stabilization-mechanism for the laser.

The present phase modulator operates at 1.5 µm wavelength which experiences a loss ~4 dB due to the extended mode-field outside the core interacting with the electrodes. A device operating at 1 µm wavelength however has lower loss due to better mode confinement. Additionally, to achieve the same relative phase difference at 1 µm, the electro-optic fiber length can be reduced by one-third. This should lower the loss even further as well as increase the cutoff frequency for the device. Work is ongoing to extend the results to the 1 µm wavelength range.