1. Introduction

Ultrashort optical pulses are widely used in various fields of scientific and engineering applications [1, 2]. These applications are mostly based on the mode-locked solid state or fiber lasers. These lasers use ion-doped solid state or glass materials as gain media, and hence they require optical pumping and their oscillation wavelengths are restricted to the near-infrared range. Ultrashort-pulse semiconductor laser diodes (LDs) are under development for several decades in anticipation of realizing compact ultrashort pulse sources using electrical pumping and a broad gain bandwidth. In particular, GaN-based LDs are advantageous for the direct generation of blue-violet ultrashort optical pulses since conventional mode-locked lasers need harmonic generation to generate ultrashort optical pulses in this wavelength range. GaN-based ultrashort-pulse lasers can reduce the size and cost of light sources for applications, like high-density volumetric optical recording [3] and next-generation refractive surgeries [4].

To realize ultrashort-pulse semiconductor LDs for practical use, several technical issues are to be solved. One such issue is reducing the pulse duration in semiconductor-based ultrashort-pulse lasers to address that the pulse energy available from the semiconductor gain medium is limited by the finite carrier lifetime. However, reducing the pulse duration is a major challenge in the development of ultrashort-pulse semiconductor LDs because the carrier-induced changes in the refractive index make the pulse formation mechanism complicated. This problem is known to be relaxed by using quantum dot active layers because the gain spectrum is broadened due to the dot size distribution. Rafailov et al. reported the generation of 390 fs duration optical pulses by a mode-locked LD (MLLD) with quantum dot active layers [5]. Dispersion compensation is another conventional method to reduce the pulse duration in mode-locked solid-state lasers [6]. The situation is, however, different for MLLDs. Intracavity dispersion compensation enables of broadening the spectral bandwidth, while extra-cavity dispersion compensation can reduce the pulse duration to the femtosecond range. Delfyett et al. reported the generation of 200-fs duration and 160-W peak power optical pulses using intra- and extra-cavity dispersion compensation optics and a semiconductor optical amplifier [7]. Balzer et al. reported the shortest pulse duration of 158 fs using a dispersion-compensated MLLD with an external pulse compressor [8] and the same group recently reported a MLLD with a self-optimization system of the intracavity dispersion using a spatial light modulator (SLM) [9].

We have previously reported the generation of blue-violet optical pulses of 200-fs duration by a dispersion-compensated GaInN MLLD with subsequent spectral filtering [10]. Our results were different from those for MLLDs so far reported since our femtosecond optical pulses were obtained using only spectral filtering, which means that some spectral components were almost free from chirp. However, due to the spectral filtering, the femtosecond optical pulses were only less than 10% of the total output from the GaInN MLLD. On the other hand, directly after the GaInN MLLD, the optical pulses showed good passive mode-locking operation characteristics since the intensity autocorrelation traces of the entire spectrum and spectral components at different wavelengths exhibited no coherence spike [11, 12]. These pulse characteristics suggest that the optical output from the GaInN MLLD was sufficiently coherent that the entire pulse could be expected to be compressed with an appropriate dispersion compensation.

In this study, we applied a nonlinear pulse compression technique to optical pulses from a dispersion-compensated GaInN MLLD in order to obtain high peak-power optical pulses. Using a nonlinear pulse compressor consisting of a holographic grating and an SLM [13], blue-violet optical pulses of 140-fs duration and 60-W peak power were obtained. The throughput of the nonlinear pulse compressor was about 50%. This is more than five times improvement in power transmission compared with femtosecond pulse generation by spectral filtering. To our knowledge, the obtained pulses had the shortest pulse duration ever obtained for MLLDs using edge-emitting type devices. The results show that nonlinear pulse compression is a promising technique for manipulating blue-violet optical pulses from a GaInN MLLD. This technique will assist in introducing chirped pulse amplification schemes into master oscillator power amplifier systems using high-power GaInN semiconductor optical amplifiers in order to generate kilowatt-peak-power and femtosecond-duration optical pulses [14].

2. Experimental setup

The experimental setup is shown in Fig. 1. Blue-violet optical pulses were generated by a GaInN bisectional laser diode (BS-LD) passively mode-locked in a dispersion-compensated external cavity. The intracavity group velocity dispersion (GVD) was set to a negative value. The optical configuration of the GaInN MLLD has been explained in detail in our previous report [10]. The bias current to the gain section was 130 mA and the reverse bias voltage to the saturable absorber (SA) section was 6.6 V. The average output power was 14.4 mW and the repetition rate was 917 MHz. The optical pulses from the GaInN MLLD were put into a pulse compressor consisting of a holographic grating (4320 grooves/mm), a Fourier lens (f = 150 mm) and an SLM (Hamamatsu, LCOS-SLM X10468-05). The SLM was a liquid crystal on silicon (LCOS) type modulator with a high-reflection dielectric coating at around 410 nm. The optical pulses after the pulse compressor were investigated by a laboratory-made intensity autocorrelation setup using second harmonic generation at the surface of a β-BaB₂O₄ crystal [15].

 

Fig. 1 Experimental setup of a dispersion-compensated GaInN MLLD and a nonlinear pulse compressor.

Download Full Size | PDF

In order to estimate phase dispersion due to the nonlinear pulse compressor, we calibrated the pixels on the LCOS with the wavelength spatially dispersed by the holographic grating. We operated the GaInN BS-LD under continuous oscillation with a reverse bias voltage of 0 V. The oscillation wavelength was tuned using a narrow bandpass filter (BPF) inserted in the external cavity and we confirmed the correspondence between the LCOS pixels and the wavelength. The distance between the grating, the Fourier lens and the LCOS were determined according to the focal length of the Fourier lens in order to reproduce the intensity autocorrelation trace of the optical pulses without the nonlinear pulse compresssor. The phase dispersion was calculated using the phases at each pixel as a function of the wavelength converted from the pixel position.

3. Results and discussions

The entire optical pulse from the GaInN MLLD was down-chirped and the chirp-characteristics were wavelength dependent according to our previous study on the optical pulses using a cross correlation technique [12]. These nonlinear chirp characteristics were experimentally confirmed by giving linear chirp compensation by creating a second-order phase dispersion in the nonlinear pulse compressor.

The autocorrelation traces of the optical pulses after the pulse compressor giving second-order phase dispersions are shown in Fig. 2(a). The inset shows the optical spectrum of the GaInN MLLD output. The full width at half maximum (FWHM) of the autocorrelation trace decreased with increasing GVD applied by the pulse compressor. The FWHM was reduced to 300 fs at a positive GVD of 0.03 ps², as shown in Fig. 2(b). The shapes of the autocorrelation traces in Fig. 2(a) did not change uniformly with the change in GVD. Even in the autocorrelation trace with the smallest FWHM, pedestals still remain. These autocorrelation traces show that the optical pulses were not efficiently compressed by the second-order phase dispersion, and thus higher-order phase dispersion was needed to be compensated for.

 

Fig. 2 (a) Intensity autocorrelation traces of the optical pulses after the nonlinear pulse compressor giving second-order phase dispersion. Inset: Corresponding optical spectrum. (b) Full width at half maximum of autocorrelation traces of the optical pulses after linear compression as a function of GVD.

Download Full Size | PDF

The nonlinear chirp characteristics of the optical pulses were investigated by spectro-temporal analysis using a BPF as previously reported [12]. We performed spectrally-resolved intensity autocorrelation measurement using a BPF and the time-bandwidth product for each spectral component is plotted in Fig. 3(a) as a function of wavelength. We performed these measurements with a wavelength step smaller than that in the previous report [12].

 

Fig. 3 (a) Time-bandwidth product for the optical pulses extracted by a bandpass filter at different transmission wavelengths (blue filled circles). (b) GVD needed to obtain Fourier transform-limited pulses of the corresponding spectral components shown in (a) (red filled circles), GVD due to the nonlinear pulse compressor (dashed-dotted trace), and phase dispersion corresponding to the GVD given by the nonlinear pulse compressor (dashed trace). The gray traces are a reproduction of the optical spectrum of the GaInN MLLD output.

Download Full Size | PDF

These time-bandwidth products are converted into the GVDs needed to obtain Fourier transform limited pulses of the each spectral component, as shown in Fig. 3(b). The GVD is small and constant at short wavelengths where femtosecond optical pulses were obtained by spectral filtering. From about 401 nm to about 402.5 nm, the GVD increases almost linearly with wavelength. For wavelengths longer than 402.5 nm, the slope becomes steeper because the spectral bandwidth is determined by the short wavelength edge of the BPF transmission and the long wavelength edge of the optical spectrum. The analysis shows that the phase dispersion needed for pulse compression can be mainly represented by a GVD that is linearly dependent on the wavelength and approximately equivalent to a third-order phase dispersion. The transmission bandwidth of the BPF was not sufficiently narrow, so the estimated GVD could be underestimated in realizing an optimized nonlinear pulse compression. The spectral width after the BPF was typically about 0.54 nm. The wavelengths where the slope of the GVD changes are important for optimizing the nonlinear phase dispersion for pulse compression.

Based on the estimated GVD shown in Fig. 3(b), we optimized the nonlinear phase dispersion. The autocorrelation trace of the optical pulses after the nonlinear pulse compressor with an optimized nonlinear phase dispersion is shown in Fig. 4. The autocorrelation trace of the optical pulses compressed by a second-order phase dispersion is also reproduced as the gray trace for comparison. Slow pulse components remaining in the linearly-compressed autocorrelation trace are well suppressed after the nonlinear pulse compression. The inset shows a reproduction of the intensity autocorrelation trace expanded in time scale. The pulse duration was estimated to be 140 fs assuming a sech² pulse shape shown by the dotted trace. The average power after the nonlinear pulse compressor was 7.5 mW, yielding a pulse energy of 8.2 pJ with a repetition rate of 917 MHz. The pulse peak power attained was about 60 W. The peak power was improved by about fifteen times compared with that in our previous report [10].

 

Fig. 4 Intensity autocorrelation trace of the optical pulses after the nonlinear pulse compressor with an optimized phase dispersion (red trace), and reproduction of the intensity autocorrelation trace of the optical pulses after the linear pulse compression (0.03ps²) shown in Fig. 2(a) (gray trace). Inset: Expansion of the intensity autocorrelation trace in the main figure (red trace), and example trace of sech² function (dotted trace).

Download Full Size | PDF

The wavelength dependent GVD used for the nonlinear pulse compression is shown by the dashed-dotted trace in Fig. 3(b). The nonlinear phase dispersion qualitatively follows the experimentally estimated GVD represented by the red circles. The discrepancy between the estimated and optimized GVDs increases as the GVD slope increases because the spectral bandwidth of the BPF was not sufficiently narrow compared with the increase in GVD.

4. Conclusion

In conclusion, we generated 140-fs duration and 60-W peak power optical pulses by a dispersion-compensated GaInN MLLD using a nonlinear pulse compressor. This is the shortest pulse duration ever obtained from an MLLD using an edge-emitting type device. The peak power was improved by about fifteen times compared with that in the previous reports. The obtained results show that the nonlinear pulse compression is a promising technique for manipulating temporal shapes of the optical pulses from a GaInN MLLD in order to generate high-peak-power femtosecond optical pulses. This technique is important for a chirped pulse amplification technique with a high power GaInN master oscillator power amplifier system.