In recent years, studies of vertical-external-cavity-surface-emitting lasers (VECSELs) have shown increasing promise for this class of semiconductor laser, combining advantages such as power scalability [1], flexible emission wavelength [2–4], and transform-limited beam quality. The maturity of these lasers was recently highlighted by the demonstration of more than 100 W average output power in continuous wave (CW) operation [1], peak powers of kWs in mode-locked operation [5,6], and the generation of pulses in the 100 fs regime with appreciable power levels [7].

Despite successful ultrafast results so far, however, the large gain bandwidth of semiconductor heterostructures should, in principle, allow for generating even shorter pulses with durations of a few tens of femtoseconds. The shortest pulse duration achieved with VECSELs in a fundamental mode-locked operation currently is 107 fs with very low output power [8], as well as 147 fs with up to 100 mW [7]. These durations are within the known timescale for intraband carrier scattering in these media [9,10]. This suggests that the dynamically changing, nonequilibrium distributions will not relax to the quasi-equilibrium Fermi–Dirac distributions during the pulse itself. Therefore, nonequilibrium effects such as kinetic-hole burning and filling are expected to play a major role in pulse formation dynamics, as computational results predict [9]. A better understanding of these effects may open a pathway to advanced models of the ultrafast gain dynamics needed to design structures capable of producing pulses which make full use of the available gain-bandwidth of the semiconductor gain medium.

Sieber et al. [11] performed detailed studies of the gain saturation and gain dispersion, and the impact on pulse shaping mechanisms by numerical calculation. In their study, they showed that the total cavity dispersion significantly affects the pulse duration and is one of the challenges faced when engineering shorter pulses. However, the gain model used is based only on a long-time gain recovery with an exponential time constant. While this approach is sufficient to optimize VECSEL structures for the generation of pulses on the longer edge of the ultrafast regime where kinetic hole filling becomes negligible, faster relaxation mechanics such as carrier heating, scattering from higher excited states, etc. (as have been studied carefully for edge emitting lasers and amplifiers [12,13]) complicate the practical use of this model for pulses approaching 100 fs in duration. Due to a partial recovery of the gain on a timescale much faster than the repetition rate of the laser cavity, multi-pulse events or pulse elongation may occur as the saturation fluence increases with pulse duration [14]. Including these characteristics of VECSEL lasers in computational models will help to more rigorously predict pulse stability and duration.

In this Letter, we demonstrate in situ measurements of VECSEL gain dynamics while a device is in stable mode-locked operation. To probe the semiconductor structure, we utilize a femtosecond fiber laser with an emission wavelength centered at 1040 nm. To overlap the probe wavelength with the gain structures under observation and to shorten the duration of our probe pulse, we spectrally broaden the probe laser pulses in a single-mode fiber to a bandwidth exceeding 100 nm. We employ an asynchronous optical sampling technique using mismatched repetition rates to passively scan the time domain of the interaction through the increasing shot-to-shot delay between probe and in situ pulses. An adjustable offset frequency between the lasers allows for tuning of the measurement resolution to a level suitable for the round-trip time of the VECSEL cavity. An introduction to this technique can be found in [15]. As no mechanical delay line is required to cover the nanosecond-scale temporal window, the measurement precision can be improved from standard pump/probe techniques, with sub-fs resolution possible given proper stability of the time jitter between the lasers.

One of the primary advantages of this technique compared to traditional pump/probe techniques is that no assumptions are necessary to connect the measurement result with real cavity parameters. The gain dynamics probed must be the same dynamics experienced by the lasing pulse. This includes the effect of the rapid gain depletion when the intracavity pulse interacts with the quantum wells, in addition to fast and long-term gain recovery resulting from, e.g., intraband scattering and the absorption of pump photons, respectively. As the repetition rate of VECSELs lies typically in the multi-GHz range, the carrier density will not reach a steady state situation as it would in the case of a typical extra-cavity pump and probe experiment. To demonstrate the efficacy of this technique, we measure the time-resolved reflectivity change of two VECSELs in stable mode-locked operation with sub-1 ps pulses. Taking these gain dynamics into account should improve future modeling of pulse shaping mechanisms significantly.

The VECSEL gain chips measured in the experiments were grown by metal-organic vapor phase epitaxy and consist of a periodic gain structure with multiple InGaAs quantum wells (QWs) between GaAsP barriers, an InGaP window layer, and an AlGaAs-AlAs Bragg mirror, and are similar to the design in [1]. In this Letter, we show results from two different structures: (1) a VECSEL chip with ten QWs in a resonant configuration which is optimized for an emission wavelength of 1030 nm; (2) a chip with eight QWs in an antiresonant configuration and an emission wavelength near 980 nm. An anti-reflective coating is applied to both structures to reduce the gain-narrowing effect of the micro-cavity and to flatten the dispersion of the chip. For mode-locking, we employ semiconductor saturable absorber mirrors (SESAMs) with a single QW placed near the semiconductor-air interface to minimize the recovery time as discussed in [16]. Traditional pump/probe measurements of the differential reflectivity of one SESAM using a 100 fs probe laser centered at 1030 nm is shown in Fig. 1. As can be seen in the figure, the reflectivity recovery is dominated by a fast component in the 500 fs range, and the reflectivity recovers fully within a few ps.

 

Fig. 1. (a) Time resolved differential reflectivity of a SESAM used in the experiments. (b) Resulting pulse duration of the mode-locked VECSELs for the case of the two different gain chips.

Download Full Size | PDF

Both VECSELs were placed at the fold mirror in respective V-cavities, with 1% curved output couplers and SESAMs composed of the end mirrors. The cavity lengths were adjusted such that the repetition rates of the VECSELs were close to 1680 MHz, which corresponds to the 21st harmonic of the probe laser. The autocorrelation traces of the VECSEL outputs are shown in Fig. 1(b). In this configuration, the VECSEL sample at 1030 nm emits pulses with an FWHM in the range of 800 fs, while the structure at 980 nm emits pulses with about 300 fs FWHM. The average output power in both cases is about 200 mW.

Our probe laser is an amplified Yb-doped fiber oscillator with an emission wavelength centered at 1040 nm which emits 12 nJ pulses with 120 fs duration. To probe both VECSEL structures, which emit at different wavelengths, we broaden the spectrum of the fiber laser in a single-mode fiber to an FWHM of about 100 nm as shown in Fig. 2. We then temporally compress the output with a pulse shaper down to sub-20 fs pulses, which is close to the transform limit. The cavity length of the probe laser can be adjusted with a linear stage and a fast piezo to control the repetition rate in a window of about 100 kHz.

 

Fig. 2. Spectrum of the probe laser after spectrally broadening and the spectra of the two investigated VECSEL structures.

Download Full Size | PDF

To provide both an adjustable frequency offset and help stabilize the timing jitter between the VECSELs and probe, we electronically mix the repetition rate of the 1.6 GHz VECSEL (measured with a fast photodiode) with a frequency synthesizer near 720 MHz. The resulting signal near 960 MHz is then mixed with the 12th harmonic of the probe laser to obtain an error signal for control of the relative timing between the pulse trains. By adjusting the synthesizer, a detuning of the pulse repetition rates between the VECSEL and the 21st harmonic of the probe laser can be set between 0 Hz and $\sim 100\mathrm{kHz}$. While the short probe pulse would, in principle, allow for a 20 fs time resolution, the main limitation in our scheme is the timing jitter between the two pulse trains and the electronics involved in the data acquisition. We estimate the actual time resolution of the setup by performing cross-correlation measurements and find it is on the order of 100 fs.

An illustration of this scheme is shown in Fig. 3(a). For a calibrated time axis, we furthermore generate a scan signal by mixing the 21st harmonic of the probe laser’s repetition rate with the VECSEL’s fundamental rate. After a 1 MHz low pass filter, this signal will correspond to the relative phase of the two lasers and can be used to either calibrate the time axis or as a trigger to average the data.

 

Fig. 3. Illustration of the (a) frequency control scheme and (b) in situ pump and probe setup. The probe laser is locked to a multiple of the repetition rate of the VECSEL cavity and focused on the VECSEL chip. A differential detector is used to enhance the signal to noise ratio.

Download Full Size | PDF

We focus the probe laser on the VECSEL sample with a beam diameter of about 150 μm, which is smaller than the diameter of the pumped region (400 μm). This allows us to probe the central area of the resonator mode. For enhanced sensitivity, we split the probe beam to create a reference and use a differential photodiode to sample the probe signal. To avoid a perturbation of the mode-locked VECSEL operation, we attenuate the probe laser incident onto the VECSEL structure to 10 mW, which corresponds to fluences less than $1\mathrm{\mu J}/{\mathrm{cm}}^2$. A scheme of the setup is shown in Fig. 3(b).

Figure 4 shows the scan signal which we use to calibrate the time axis. As this signal represents the relative phase between VECSEL laser and probe, each oscillation corresponds to a cavity roundtrip of the VECSEL laser (600 ps). The measured differential reflection seen by the probe laser is shown in Fig. 4(b) with a calibrated time axis. As the VECSEL serves as the folding mirror in the V-cavity, the intracavity pulse interacts with the gain chip twice during a single round trip. Due to the different round-trip times of the two cavity arms (200 and 400 ps), the carrier distribution and, thus, the gain seen by the probe, at each event differs.

 

Fig. 4. (a) Scan signal showing the phase between VECSEL and probe laser. (b) Scan of the reflectivity change of the gain medium as a function of the time. In each round trip, the intracavity pulse hits the gain medium in two different conditions due to the different length of the two cavity arms. Thus, during a round trip, two different carrier distributions are present.

Download Full Size | PDF

Figure 5 shows a shorter time window of the scan. When the intracavity pulse interacts with the gain medium, a rapid depletion of the gain seen by the probe laser pulse occurs on the timescale of the intracavity pulse. Afterward, a rapid recovery with time constant in the few ps range takes place. This can be attributed to multiple effects: intraband scattering into the kinetically burned hole in the carrier distribution, carrier heating changing the Fermi distribution, and refilling from higher excited states as well as from the barrier. These effects, observed in semiconductor edge emitters, have recovery time constants in the range from several femtoseconds to a few picoseconds [3,4] which agree with our observations. While the gain depletion is significantly faster in the case of the 300 fs VECSEL pulse compared to the 800 fs one, as expected, the measurements show that the recovery also occurs on faster timescales. This might be explained by the fact that the longer intracavity pulse burns only a shallow hole into the carrier distribution. The hole is already being filled substantially during the pulse duration, while the shorter pulse—which burns a deeper hole—leaves a larger number of unoccupied states into which carriers can scatter. Numerical simulations of the gain dynamics also indicate that a faster gain recovery can take place in the case of a shorter pump pulse, but further work needs to be done to provide a clear understanding of these observations.

 

Fig. 5. (a) Scan of the reflectivity change of the gain medium at 1040 nm for a 1.5 ns time window. (b) Comparison of the normalized signals obtained for the 980 and 1040 nm VECSEL chips. Finer time resolution provides a more visible display of faster carrier restoration timescales.

Download Full Size | PDF

After several picoseconds, gain recovery becomes dominated by a much slower mechanism, most likely by the continuous supply of carriers by the CW pumping, lasting until the completion of the cavity round trip. Of interest is the magnitude ratio of these two recovery components. The reflectivity drop seen by the probe laser recovers by more than 80% in a time window of 5 ps and, thus, the net gain recovery for the lasing pulse can be expected to match this behavior. While these measurements cannot distinguish between changes in reflectivity due to saturation of gain or light induced absorption (LIA), the latter will occur for the intracavity pulse as well, so the depiction of the gain dynamics felt by the intracavity pulse is still accurate. Moreover, the time duration of the intracavity pulse is matched closely by the time duration of the depletion edge of the reflectivity spikes, suggesting that while coherent effects may be present during that time window of the measurement, the recovery edge is purely carrier based.

As the fast gain recovery component taking place on a timescale of a few ps—much faster than the repetition rate of the laser cavity—resembles a significant fraction of the overall total gain recovery per roundtrip, it will affect the mode-locking results noticeably. Simulations performed using a round-trip model as presented in [9], which considers these ultrafast carrier dynamics as observed in our measurements, show that the inclusion of these fast gain characteristics strongly affects the stability of the mode-locked regime. Only for a delicate parameter space can stable mode-locked results with short pulse durations be obtained without breakup into multi-pulses [9]—an agreement with experimental observation. Studying the effects of the ultrafast gain dynamics on the pulse forming process numerically is part of an ongoing investigation.

The data collection for the full scans in Figs. 4 and 5 was taken in fractions of a second, contrasting sharply with traditional pump/probe measurements where similar scan windows at fs-scale resolution would take minutes to hours. In addition, the absence of a mechanical delay line removes measurement artifacts due to optical misalignment over long scan windows, a significant challenge in traditional pump/probe techniques. The fact that multiple cavity scans can be taken in fractions of a second allows for the averaging of multiple round trips when scans are properly calibrated by the VECSEL/probe phase signal. Our in situ scans also do not display the long-tail features of carrier saturation which occur on several ns-durations due to the carrier lifetime. Thus, the gain does not fully recover to its steady state between the interactions with the intracavity pulses.

It should be noted that the broad spectrum of the probe laser only partially overlaps with the emission spectrum of the VECSELs as shown in Fig. 2. Therefore, the measured response is not completely identical to the response for a bandwidth of the lasing pulse alone. That said, our use of a pulse-shaping device allowed for bandwidth filtering to investigate whether spectral effects were present. Using 10 nm bands with adjustable center wavelengths to test varying amounts of spectral overlap between probe and VECSEL, we determined that spectral overlap is necessary for reflectivity dynamics to be measured, but ultimately results in a scaling of amplitude alone. The timescales of the measured reflectivity changes are not obscured by spectral effects.

In conclusion, we propose a novel method of studying mode-locked VECSEL gain dynamics utilizing an in situ probing technique. We employ asynchronous optical sampling by phase locking the repetition rate of a probe laser to that of the VECSEL cavity with an adjustable offset. This allows for sensitive, high-resolution measurements of gain depletion and recovery during a full round trip of the VECSEL. An additional major advantage of this technique is the necessarily real depiction of mode-locked cavity carrier dynamics. We show that, in addition to the long recovery due to the constant pump fluence, a significant part of the gain recovery takes place within a few picoseconds. Therefore, pulse breakup or other pulse elongation effects can occur if the balance of the total net gain within the cavity is not carefully optimized [9]. Further studies will focus on increasing the time resolution by probing VECSELs producing shorter pulses, as well as VECSEL cavities with different repetition rates.