1. INTRODUCTION

The development of laser sources featuring ultrashort pulses and high power at a repetition rate above 1 GHz has stirred a lot of interest in the scientific community over the past decade, mostly because these types of sources can be used in a broad field of applications, such as multiphoton imaging [1], high-resolution time-domain terahertz spectroscopy with asynchronous optical sampling [2], self-referenced gigahertz frequency combs [3], or ultrafast communication systems [4] to cite a few. The passive mode locking of a vertical-external-cavity surface-emitting laser (VECSEL) with a semiconductor saturable absorber mirror (SESAM) is a common technique used to generate short pulses with repetition rates ranging from 100 MHz up to 100 GHz [5] in a relatively compact and simple laser system. The colliding pulse mode-locking technique was first demonstrated with a dye laser [6] and was later used with a semiconductor laser [7], but its potential had yet to be investigated with a VECSEL. The VECSEL technology is directly competing with diode-pumped solid-state lasers (DPSSLs), which have shown outstanding performance in term of pulse width or low-noise frequency combs [8], but they require more complicated cavity geometries with complex pumping schemes and suffer from $Q$-switching instabilities at high repetition rates [9]. On the other hand, the quantum-well (QW) gain medium of a VECSEL can be directly pumped with low-cost pump diodes, and its short carrier lifetime (few nanoseconds) largely suppresses spontaneous $Q$ switching, allowing stable mode locking at multigigahertz repetition rates. Furthermore, VECSEL chips and SESAMs can be fabricated at a low cost using wafer-scale mass production, making it a technology of choice for industrial applications. The laser pulse duration is often a limiting factor for applications, as it sets the resolution in time-domain terahertz spectroscopy [2], whereas nonlinear imaging techniques are mostly limited by the peak power and repetition rate. A higher peak power provides a higher signal/noise ratio and increases the penetration depth in tomography [10], and the fluorescence emission could be increased if the repetition rate is higher than the inverse of the fluorescence lifetime [11].

To date, the shortest pulse duration obtained from a VECSEL in a fundamental mode-locking regime is 107 fs, with an average power of 3 mW and a repetition rate of 5.1 GHz [12], giving a peak power just below 5 W. Higher peak powers have also been reported with longer pulses, such as a VECSEL with 400 fs pulses and a peak power of 4.35 kW at a repetition rate of 1.67 GHz [13]. However, combining a high power with sub-200-fs pulses has remained a challenge, and the current combinations of pulse durations and power reported with a VECSEL have been insufficient for the direct generation of a coherent supercontinuum in a photonic crystal fiber, requiring prior amplification and compression of the pulses. Recent studies suggested that to generate sub-200-fs pulses with a high average output power ($>500\mathrm{mW}$), it is crucial to have the gain saturation fluence as high as possible ($>200\mathrm{\mu J}/{\mathrm{cm}}^2$) while keeping the saturation fluence of the absorber very low ($<5\mathrm{\mu J}/{\mathrm{cm}}^2$) [14].

Here, we demonstrate a novel VECSEL cavity concept, where the gain chip and the SESAM are placed in a ring cavity to generate colliding pulses in the absorber, significantly reducing the effective saturation fluence of the absorber while providing a symmetric gain for the two counterpropagating pulses. This new geometry is shown to improve the mode-locking stability and to provide higher output power when compared to a more standard V-shaped cavity. We demonstrate fundamental mode locking with a pulse duration of 195 fs and an average power of 225 mW per output beam, giving a record peak power of 460 W at a repetition rate of 2.2 GHz. The simple combination of the two output beams would thus provide a total peak power of 920 W with a sub-200-fs pulse duration. These results set a new milestone in comparison to the prior state of the art of conventional VECSEL cavities and pave the way to further developments.

2. COLLIDING PULSE VECSEL DESIGN

A. Advantages of Colliding Pulses

In a ring laser cavity without direction-selective elements, two beams can counterpropagate and interact with each other at selected locations. If the absorption of the saturable absorber is strong enough, a pulsed-shaping mechanism occurs and provides a mode-locked state where two counterpropagating pulses automatically synchronize their flight in the cavity to cross in the SESAM and minimize the losses. Since the counterpropagating beams are spatially coherent and share the same polarization, the superposition of the two pulses produces interference that results in a transverse transient carrier grating in the SESAM. The transient field intensity on the SESAM, when the pulses are perfectly synchronized, has a maximum that is 4 times higher than the intensity of one beam alone. A higher field intensity leads to a higher absorption, and thus to a smaller saturation fluence. Moreover, the field interference pattern leads to a spatially modulated absorption, where roughly half of the beam area (interference peaks) is strongly absorbed and the other half (valleys) is very weakly absorbed, leading to a further decrease in the saturation fluence, since only half of the area has to be saturated. It has been shown that a colliding pulse scheme reduces the absorber saturation fluence by a factor of 3 in a longitudinal interaction [15], but this is expected to be further reduced with a transverse interaction in a thin medium like a SESAM, since the beam propagates perpendicular to the grating. As a result, the standard design restriction for the mode area ratio between the VECSEL gain and the SESAM, which typically needs to be between 10 and 30 to provide stable mode locking [16], is largely relaxed without sacrificing the modulation depth of the absorption. We can thus increase the mode size on the absorber to reduce the thermal impedance and increase the damage threshold.

Since the gain structure is the main contributor to the nonlinear group delay dispersion (GDD), the dispersion of the cavity is reduced, since the pulses hit the structure only once per round trip instead of twice when it is placed as a folding mirror in a V-shaped cavity. However, it also reduces the modal gain by half, requiring a smaller output coupler transmission.

B. VECSEL Setup

The VECSEL structure consists of two consecutive semiconductor Bragg mirrors containing, respectively, 23 and 13 pairs of $\mathrm{AlAs}/{\mathrm{Al}}_{0.12}{\mathrm{Ga}}_{0.88}\mathrm{As}$ to reflect the lasing and the pump wavelength, followed by the active region, an InGaP cap layer, and a single $\lambda /4$ layer of ${\mathrm{Si}}_3{\mathrm{N}}_4$ for dispersion management. The antiresonant active region contains eight InGaAs QWs, placed in pairs on four antinodes of the field, and is pumped in the GaAsP barriers with a 790 nm fiber-coupled pump diode. The structure is grown as a bottom emitter and is bonded to diamond following the procedure described in [17] for optimal thermal management. The SESAM consists of 24 pairs of AlAs/GaAs, followed by a single InGaAs QW placed in close proximity to the surface to provide a higher carrier recombination velocity via tunneling to surface states. The wavelength of the SESAM QW absorption peak was blueshifted by 10 nm from the lasing wavelength to provide minimal absorption at room temperature, facilitating alignment. The measured GDD of the gain structure and SESAM at room temperature and normal incidence is shown in Fig. 1(b).

 

Fig. 1. (a) Schematic layout of the VECSEL device. OC, output coupler; HR, highly reflective; ROC, radius of curvature; Cu, copper. (b) Measured group delay dispersion of the SESAM and gain structure. Vertical bars indicate the standard deviation over 20 measures.

Download Full Size | PDF

To provide a symmetrical amplification of the two pulses, the VECSEL and SESAM are placed in the cavity according to the geometry shown in Fig. 1(a). The gain medium is placed a quarter of the total cavity length ($L=136\mathrm{mm}$) from the saturable absorber, assuring an equal pumping duration and gain recovery for both pulses. This geometrical rule is not very stringent as long as the gain recovers to its full value between two consecutive pulses. It has been shown that after the gain depletion from a pulse interaction, the population inversion in a VECSEL can recover to more than 80% of its initial value in a time as short as 5 ps [18]. The cavity is completed with a highly reflective concave mirror with a radius of curvature of 75 mm and a flat output coupler with a reflectivity of 99.2%. The angle of incidence on the SESAM was 7°, giving an interference period of 4 μm. This geometry provides a mode waist of 152 μm on the gain and 85 μm on the SESAM. The mode area ratio of only 3.2 is a factor of 3 smaller than the lower limit usually adopted [16].

3. EXPERIMENTAL RESULTS

For all the characterizations, the temperature of the gain element was kept at 5°C, and the temperature of the SESAM was varied to adjust the amount of absorption. The QW absorption band edge was close to 980 nm at room temperature, below the lasing wavelength of 990 nm, providing a maximum reflectivity for the cavity alignment. Once the cavity was lasing, the SESAM temperature was adjusted to a higher temperature, shifting the absorption spectra at a rate of $+0.3\mathrm{nm}/\mathrm{K}$. The VECSEL switched from CW operation to a mode-locked state at a SESAM temperature of 30°C. Figure 2 shows the output power from one output beam versus the pump power and SESAM temperature, with the corresponding pulse durations.

 

Fig. 2. Output power from one output beam versus incident pump power at different SESAM temperatures. The corresponding pulse durations are also plotted (right axis). The empty, full, and half-full squares represent, respectively, an unstable, a stable single-pulse, and a stable with a secondary pulse mode-locking regime. The two regimes circled and labeled “1” and “2” are further investigated.

Download Full Size | PDF

As expected, the highest average power was obtained at a SESAM temperature of 35°C, where we obtained an output power of 300 mW with a pulse duration of 250 fs at a repetition rate of 2.2 GHz, giving a peak power of 480 W (regime 1). At a higher pump power, the gain was higher and recovered faster, enabling a stable but smaller secondary pulse, lowering the main pulse peak power. When the SESAM temperature was increased to 45°C, the maximum power in a single pulse regime dropped to 225 mW and the pulse duration got as short as 195 fs, giving a peak power of 460 W (regime 2). The power, the spectrum, and the duration of the two output beams have been measured simultaneously and are identical. Figure 2 also shows the remarkable robustness of the mode-locking regime, which can be obtained over almost the entire range of pumping power and for different SESAM absorption levels. When the same gain element was used as a folding mirror in a V-shaped cavity, with the same SESAM and a mode area ratio of 11, we obtained a similar pulse duration of 220 fs, but with only 50 mW of output power. Moreover, the mode-locking regime was observed only with a limited range of pump power, between 4.5 and 6 W and between 11 and 13 W, and only with a SESAM temperature restricted to the [45°C–55°C] range.

Figure 3 displays the autocorrelation trace of the regime 2, recorded with a second harmonic noncollinear Femtochrome autocorrelator. The cross correlation between the two outputs recorded with a long-range noncollinear home-built correlator is also plotted for comparison. The FWHMs of the autocorrelation and cross correlation are, respectively, 300 and 314 fs, with an almost perfect ${\mathrm{sech}}^2$ pulse shape, showing the synchronization of the counterpropagative pulses with negligible timing jitter. This means that the simple combination of the two linearly polarized outputs would provide 920 W of peak power with a sub-200-fs pulse duration. We should note, however, that a coherent combination the two beams would require an identical carrier–envelope offset frequency, which is not shown here. Since a stronger or early pulse would experience more losses than the counterpropagating pulse from the SESAM, the power balance between the two beams and the pulse synchronization occur naturally. This energy transfer and phase shift of the pulses is a mechanism showing similarities to the recently demonstrated locking of a dual-comb mode-locked laser when the spatial and temporal pulse overlap is sufficient [19].

 

Fig. 3. Measure of the autocorrelation and cross correlation of the two counterpropagative beams. In the inset is the cross correlation on a long scanning range, showing fundamental mode locking without any side pulses.

Download Full Size | PDF

Figure 4 shows the microwave spectra of the pulses in regime 2. The measurement with a resolution bandwidth (RBW) of 500 Hz shows a sharp peak more than 65 dB above the background, with a resolution-limited linewidth and without side peaks. When the laser develops a side pulse at higher pump powers, side peaks are clearly visible in the microwave spectrum. The second harmonic amplitude appears slightly lower because of the limited bandwidth of the photodetector used. The microwave spectrum is almost identical for the two regimes investigated here. The optical spectra, however, are slightly different for the two regimes (Fig. 5). The spectra are relatively smooth and do not show any CW spikes. The lower SESAM temperature and lower pump power of regime 1 give an output spectrum centered at 991 nm, whereas the spectrum of regime 2 is slightly redshifted to 992 nm, with respective bandwidths of 4.6 and 6.8 nm. These correspond to time–bandwidth products of 0.35 and 0.4, indicating nearly bandwidth-limited ${\mathrm{sech}}^2$ pulses (1.1 and 1.3 times the Fourier limit).

 

Fig. 4. Microwave spectrum of regime 2, centered at 2.202 GHz with a RBW of 500 Hz. The inset shows all available spectra, with the second harmonic at a RBW of 100 kHz.

Download Full Size | PDF

 

Fig. 5. Optical spectra of the two regimes investigated.

Download Full Size | PDF

4. CONCLUSION

We presented the first (to our knowledge) colliding pulse mode locking of a VECSEL and established a new state of the art in term of pulse duration and peak power combination. The significant reduction of the saturation fluence of the absorber and the increased robustness provided by this cavity concept were exploited to demonstrate a total output peak power of 920 W for a sub-200-fs pulse duration. This result should enable the direct generation of an octave-spanning coherent supercontinuum in a photonic crystal fiber from a VECSEL. To reach several kilowatts of peak power at a repetition rate of several gigahertz, a possible approach is to further increase the average power using a larger transverse mode while maintaining a similar pulse fluence, and to further reduce the pulse duration by optimizing the GDD and third-order dispersion with the use of a multilayer antireflective coating and numerically optimized structures. The cavity geometry could also be adapted to realize an optical gyroscope using the Sagnac effect, similar to a DPSSL [20].