A simplest saturable absorber, in the form of an unpumped section, is introduced into a Fabry-Perot semiconductor laser with a strongly asymmetric broadened waveguide structure incorporating a relatively thick (80 nm) active layer. This allows for suppression of trailing oscillations and a decrease in the optical pulse width compared to the uniformly biased structure. Single optical pulses of ~80 ps full width at half maximum (FWHM) and ~35 W peak power (~3 nJ pulse energy, Eopt), practically without trailing edge oscillations, were experimentally achieved under room temperature conditions by absorber-assisted gain-switching, using pumping current pulses of ~1.3 ns FWHM and ~17 A amplitude. The laser emission has a narrow (13 degrees FWHM in the transverse direction) far field.
© 2013 Optical Society of America
There has been a considerable interest in picosecond-range (~100 ps) high energy (significantly above 1 nJ) optical pulse generation with semiconductor lasers, since they are compact optical sources with many applications including high-precision laser radar , three-dimensional (3-D) time imaging, spectroscopy and lifetime studies.
A well-established technique which allows the generation of short optical pulses is gain-switching (see for example a recent review in , and references therein , and references therein). However, the optical energy achievable with commercial structures operated in the gain-switched mode is modest, with a noticeable part of the energy located in the trailing pulse train. To overcome the power limitation and to suppress trailing oscillations, a number of structure modifications have been suggested in literature. In the context of injection-pumped semiconductor lasers, an increase in peak-power was achieved by several methods including forming a saturable absorber inside a single-heterostructure (SH) laser cavity by deep implantation of heavy ions , spectral filtering of the emission from a SH laser , and a laser structure with field assisted gain control by introducing an extra potential barrier between the junction and the active layer . Unfortunately, SH gain-switched lasers have required high-density current pulses to pump, have shown difficulties in reproducibility and, in case of spectral filtering, are laborious to adjust.
One promising structure of a gain-switched laser diode utilizes an asymmetric waveguide, with the refractive index step between the optical confinement layer (OCL) and the n- cladding much smaller than the one at the OCL- p-cladding interface. Such a structure can combine a relatively large active layer width (da) with a small optical confinement factor (Γa) thus ensuring a very large equivalent spot size (da/Γa) .
The simplest qualitative explanation of the main advantage of a structure with a large da/Γa for gain-switching is that in such a structure, the start of the optical pulse is delayed compared to a more traditional construction, because of the slower growth of the laser emission from the spontaneous seed. This allows for a high excess carrier density above threshold to be accumulated, which is then emitted as a short, high-energy optical pulse. More detailed, quantitative theoretical analysis can be found in our earlier papers [7–9].
While the asymmetric waveguide of the type used here is not the only construction to allow a large da/Γa ratio, it has a number of advantages over other edge-emitting structures using the same fundamental principle. Most importantly, our asymmetric waveguide supports only a single transverse mode regardless of the laser stripe width; it also enables low leakage of electrons from the waveguide layer into the p-cladding (due to the high barrier seen by the electrons) and low optical losses (due to the small penetration of the mode intensity into the lossy p-cladding).
An asymmetric waveguide structure was designed to operate in the gain-switching mode and was subsequently tested [10,11]. Laser pulses with an energy of more than 3 nJ in the single transverse mode, corresponding to a peak power level of approximately 30 W and a pulse length of about 100 ps, but with some trailing edge oscillations were obtained at room temperature with pumping current pulses having a duration of ~1.5 ns and an amplitude up to ~17 A. The driver circuit of ~1cm2 in size utilizes a commercial silicon avalanche transistor as electrical switch which fulfills the criteria of compactness and ease-of-use.
Here, we describe experiments that make use of a similar semiconductor laser structure operated in the saturable absorber assisted gain-switching mode.
Previously, several groups (see e.g [12,13].) demonstrated trailing pulse free pulse generation by gain-switching of a laser diode with a saturable absorber (SA). Here we show that the implementation of a saturable absorber section of an experimentally optimized length into the cavity of a laser with very large value da/Γa efficiently attenuates the low energy wings of the trailing edge of the pulse. A clean single optical pulse of ~80 ps / ~35 W (about 3 nJ) at current pulse parameters of 1.3 ns duration and ~17 A amplitude is achieved, which is beneficial, for example, for use with a single-photon avalanche diode (SPAD) detector, allowing for better timing accuracy.
2. Laser diode structure/measurement setup
Molecular beam epitaxy (MBE) was used to grow the semiconductor laser diode structure. The structure (similar to that shown in Fig. 1 and used earlier in ) has a broadened strongly asymmetric waveguide incorporating an 80 nm thick bulk n-GaAs active layer surrounded by n-AlGaAs optical confinement layers (OCL) of 20 nm thickness on the p-side and 1.8 µm thickness on the n-side. The blue line in Fig. 1 corresponds to the modal intensity distribution of the fundamental (single) transverse mode, while the black line represents the refractive index variation across the layers (the profile of the structure).
In subsequent processing, an oxide stripe 128 µm wide was created [Fig. 2]. The stripe was formed by depositing an insulating silicon dioxide layer on the semiconductor surface and by patterning the layer using UV lithography and wet etching prior to deposition of metal contacts. The laser chips were shortened to a length of 1.4 mm. A low reflection (LR) coating with a reflectance RLR ≈0.06 was applied to the front facet and a high reflection (HR) coating with RHR ≈0.94, to the rear facet of the laser cavity.
The intracavity saturable absorber was implemented by using focused ion beam (FIB) technique to remove a part of the p-type electrode in close proximity to the laser diode front facet [Fig. 2], thus forming a current non-injection region which becomes the saturable absorber. The amount of saturable absorption introduced can be adjusted by the length of this electrically isolated region (lSA).
Experimental results described below are based on test samples some of which had a saturable absorber of 20 µm length and the others, of 30 µm length [Fig. 2]. Prior to that, we tested several SA lengths and found that 20 µm and 30 µm were the best for achieving trailing-pulse-free gain switched pulses of a sufficiently high energy.
Temporal and spectral characteristics of the transient lasing behavior were investigated at room temperature before and after saturable absorber implementation. The lasers were pumped with current pulses of ~(1.3-1.5) ns duration and ~(9-17) A amplitude generated by an optimized silicon avalanche transistor based driver circuit  at a pulse repetition rate of 1 kHz. All experiments were performed with the same driver circuit without changing the operating capacitor to keep the pump pulse shapes as constant as possible. Current pulses were deduced from the voltage drop measured across the load resistor with a 30 GHz oscilloscope (LeCroy WM830Zi-A).
For a set of driving current pulse amplitudes, optical pulses were generated and their time-resolved spectra were measured with a spectrograph equipped streak camera (Hamamatsu C5680). Images were recorded in the analog integrated mode (number of exposures: 50) to reduce noise for high precision measurements, and the background was subtracted.
The spectrometer, assembled in front of the streak camera, utilizes a blaze grating with a groove density of 150 grooves per mm, which gives a nominal reciprocal linear dispersion (resolution) of 20 nm per mm. The spectral resolution can be improved by narrowing the spectrometer entrance slit width; in our experiments, the slit width was set to 150 µm, to approximately resemble the laser diode oxide stripe width, resulting in a spectral resolution of 3 nm. To ensure the correct position of the spectral emission distribution, the measurement was repeated with the entrance slit width set to 20 µm, which corresponds to 0.4 nm resolution. The acquired image itself can resolve ~0.08 nm per pixel.
The input slit width of the streak camera sweep unit, in its turn, affects the time resolution, with the precise value depending on sweep time range. The slit width was adjusted to 30 µm, which lies in the manufacturer-specified range for high time resolution measurements. The sweep time range was set to ~2 ns allowing an acquired image time resolution of ~3.5 ps/pixel.
In another setup, to determine the time resolved transient lasing behavior we utilized a 25-GHz broadband InGaAs PIN-type photodetector with a 21 µm detector diameter. The photodetector was connected via an 18 GHz cable to the 30 GHz oscilloscope.
A silicon photodiode (PIN-10DF, UDT Sensors, Inc.) was used to determine the average optical power. Therefore the detector was placed in close proximity to the laser diode front facet. The peak power was extracted from the lasing pulse intensity-time profile and the average optical power taking into account the 1 kHz pulse repetition rate.
3. Improvement in gain switching pulse shape by saturable absorber implementation
The structure has a rather high steady state threshold current, Ith~3.45 A, which is mainly due to the following reasons: i) an intentionally small confinement factor Γa (as in CW high power lasers, see e.g .) and ii) a moderate injection efficiency ηi~0.5.
Lasers with and without saturable absorbers, but otherwise identical, were analyzed and compared.
The pump current pulse amplitude in steps of 1 A from 10 A (1.5 ns FWHM) to 17 A (1.3 ns FWHM) is shown in Fig. 3(a). For such short current pulses, the effective lasing threshold, defined here as the current pulse amplitude at which lasing starts, was found to be around Ipulse.th_(no SA) ~9.5 A in the case of no SA. After SA implementation, the effective laser threshold increased by a factor of 1.2-1.4, depending on the fraction of the p-electrode which got electrically isolated. Structures with a current non-injection region of 20 µm or 30 µm (1.4% or 2.1% of the cavity length) had effective threshold currents of Ipulse.th_20 ~11.5 A or Ipulse.th_30 ~13.5 A, respectively.
Time-resolved optical output power profiles shown in Fig. 3 are extracted from streak camera measurements with the total optical pulse energy Eopt measured with a silicon photodiode with a large aperture, positioned near the laser facet. Figure 3(b) and 3(c) show optical responses of the laser diode to current pulses presented in Fig. 3(a). The output power in watts is drawn on a nanosecond timescale. Typical laser pulse profiles before (solid lines) and after implementing an intra-cavity absorber region of 20 µm [Fig. 3(b)] or 30 µm [Fig. 3(c)] length (filled curves) are presented. Optical pulse characteristics such as pulse shape, pulse width and peak power are compared. Besides the total pulse energy, Eopt, we also quote later in the text the useful energy, Euseful which is defined as the energy contained within the first pulse, up to the first minimum in the optical output. The “useless” energy after the first minimum of the optical pulse is called here trailing edge energy, Etrail. The total pulse energy is simply described by .
Close to the effective lasing threshold Ipulse.th_(no SA) clean and rather symmetric optical pulse shapes are achieved from lasers without the SA. However, higher injected pumping current causes additional optical oscillations seen as trailing edge shoulder in the temporal pulse profile. As described in our earlier work [7–9], this can be seen as a stage of the transition from proper gain-switching (emission of a single optical pulse) to a relaxation oscillation train that characterizes the turn-on of quasi-CW operation. Put simply, a trailing optical pulse emerges when the carrier density in the active layer, depleted by the initial optical pulse, is able to recover back to the threshold value before the end of the pump pulse. The carrier density recovery speeds up as the pump amplitude is increased, while the time available for the recovery becomes longer, as the initial optical pulse moves from the trailing edge of the pump pulse towards the leading edge. Therefore, at high pump pulse amplitudes, and with the pulse durations typically available with Si electronics (> 1 ns), the eventual emergence of trailing pulses is, in principle, unavoidable. However, optimizing the laser construction may push their onset to pump amplitudes beyond the practically available range, and/or drastically decrease their amplitude. This is what happens in a laser with an intracavity SA.
In lasers with a SA, as stated above, the effective threshold current increases with the SA length lSA. Close to the effective threshold current (I pulse.th_30, for lSA = 30 µm) a clean lasing pulse is generated, with a higher peak power than that obtained from lasers without the SA close to their threshold. The passage of light through an intra-cavity saturable absorber leads to slight pulse shaping, as expected.
Of particular interest was the laser performance when pumped with a high current pulse (1.3 ns duration, 17 A amplitude, the highest available from our electrical source). We found that as the absorber length was increased from zero, the pulse width, at the current pulse amplitude of 17 A, initially decreased due to the shaping action of the absorber nonlinearity suppressing the low-energy parts of the output. The same effect led to the trailing edges gradually weakening, with Etrail decreasing. With lSA = 30 µm, the trailing edges disappeared completely, and the pulse was at its shortest. Further increase in the absorber length produced pulses that were still trailing edge free, but had a lower energy (which can be attributed to the increase in the effective threshold) and longer duration (which can be explained by the lower energy leading to less efficient saturation of the absorber). The absorber length of 30 µm was thus judged to be the optimal for the laser diode structure and the current pumping conditions used, since it produced a trailing edge free pulse (Eopt ≈Euseful) with a relatively high energy (Eopt = 3.9 nJ) and the lowest duration obtained (FWHM = 81 ps). A structure with 20 µm long SA showed still slight “afterpulsing”, where Eopt = 4.8 nJ (FWHM = 90 ps) of which the trailing edge oscillation energy, Etrail = 0.6 nJ. The pulse energy of each pulse at these and other pumping conditions are visualized later on in detail in Fig. 7. As can be expected, the pulse energy decreased after absorber implementation, which may be partly due to linear effects (an increased effective threshold) and partly due to the nonlinear pulse shaping (the energy within the low shoulder of the trailing edge initially contributed to the total energy emission but is now suppressed by the absorber as it recovers from being saturated with the main pulse). The peak power calculated from the measurements for a structure with SA is almost unchanged however, as the reduced area under the curve compensates for the decreased optical pulse energy.
Next, we analyzed the spectral properties of the pulses shown in Fig. 3(b) and 3(c). Time resolved spectra were measured at room temperature with a spectrometer equipped streak camera. The optical responses of the structures with (a) lSA = 0, (b) lSA = 20 µm, and (c) lSA = 30 µm, to a 17 A, 1.3 ns pumping current pulse are represented in Fig. 4(a)-4(c) by contour plots with 100 intensity levels. The longer the unpumped section of the laser diode, the more the emission shifts towards shorter wavelength (higher energy), from initially ~852 nm (a) to ~846 nm (b) and 843 nm (c), respectively and pulse width narrowing can be observed. Figure 4(d) shows spectral lasing profiles for the three types of structures (a) lSA = 0, (b) lSA = 20 µm, and (c) lSA = 30 µm as response to pumping current pulse amplitudes of (10 – 17) A, (12 – 17) A, and (14 – 17) A respectively, which can be viewed from Fig. 3(a).
The short-wavelength shift can be attributed mainly to the higher dynamic threshold leading to a higher threshold carrier density. Since the gain peak is known to shift towards shorter wavelengths with carrier density, this translates into shorter operating wavelength.
We turn next to the spatial properties of the pulse generated.
The far field (the relative intensity versus transverse angle of the laser diode) was measured with a silicon photodiode of 1 cm2 active area at a distance of 50 cm from the laser diode facet. Figure 5 shows the measurement results for a laser with 30 µm long SA in the plane normal to the junction (fast axis), at the pumping pulse amplitude of 17 A. The broad transverse modal distribution shown in Fig. 1 ensures excellent transverse far field properties, with the FWHM of the fast axis radiation distribution of just 13 degrees and the full transverse aperture angle containing 95% of the laser output of about 28 degrees.
Some information on the time-resolved in-plane near field properties of the laser diode can be deduced from measurements of the time resolved transient lasing behavior [Fig. 6] with a 25-GHz broadband InGaAs PIN-type photodetector with a 21 µm opening diameter, positioned at twice the focal distance (2f) of a collimating lens and 4f from the laser facet. The detector opening was about six times smaller than the laser stripe width. This means that only a small part of the emission which is located close to the lateral (and vertical) intensity peak can enter the PIN detector. The limited detector diameter acts as kind of spatial filter which directly influences the observed width of the optical pulse. To measure the pulse energy, the broadband photodetector was replaced by a large area (1 cm2) silicon photodiode positioned near the laser facet as in streak camera measurements.
Figure 6 shows the optical response of a laser diode of reproducible technology to a ~17 A pumping current pulse before (solid line) and after implementation of a 20 µm (dashed line), 40 µm (filled curve) long saturable absorber. The delay in emission is induced by the SA. Here, lasing from the structure with no SA shows significant relaxation oscillations which reduce drastically after absorber region implementation. Depending on the extent of those oscillations observed in the emission of an initial laser diode structure, the SA will cause a more or less pronounced improvement on the lasing pulse shape due to the increase in dynamic threshold current and the shortening action of the SA nonlinearity. This measurement, which allows only a part of the broad stripe emission to enter, shows a much narrower optical pulse width of ~52 ps FWHM [Fig. 6] compared to streak camera measurements ~81 ps, [Fig. 3(b) and 3(c)] where the whole emission is coupled into the system. The inhomogeneity in emission along the wide stripe appears to broaden the optical pulse, presumably because pulses emitted by the different spatial areas of a broad laser stripe are not synchronized with each other. This measurement suggests that using a laser diode with a narrower stripe, or with some control of lateral modal properties to achieve more homogeneous near-field profile of the laser emission will provide a shorter pulse. This will be particularly advantageous for use with standard single-photon avalanche diode (SPAD) detectors, whose resolution is about 50 ps.
The graphs displayed in Figs. 7 to 9 are based on values extracted from Fig. 3 to point out the effects of the absorber region with regards to the optical pulse energy, pulse shape, and pulse width which in turn affects on the optical peak power.
Figure 7 shows the dependence of the optical pulse energy on current pulse amplitude for the absorber length lSA = 0, 20 µm, and 30 µm. The top family of curves show the total pulse energy Eopt, and the bottom family of curves, the energy in the trailing part of the pulse Etrail. The figures clearly show that the absorber implementation can largely (for lSA = 20 µm) or virtually completely (for lSA = 30 µm) eliminate the trailing part of the pulse output, which is important since the trailing pulses can have a detrimental effect on measurement applications. Their suppression is achieved at the cost of only a small penalty (no more than 5-15%) in the useful energy of the pulse Eopt – Etrail caused by the absorber-induced increase in the effective threshold current.
The dependence of the optical pulse width for the same structures on the pumping current pulse amplitude ~(13 – 17) A (i.e. those pulse amplitudes where the absorber use is beneficial) can be seen in Fig. 8. Compared to the structures with no SA, pulse width narrowing can be observed at a current pulse amplitude >13 A for the 20 µm shorter p-type electrode and at >14 A for the 30 µm long SA. At a current pulse amplitude of 17 A the pulse width decreases from ~100 ps (lSA = 0), to ~90 ps (lSA = 20 µm), to ~81 ps (lSA = 30 µm).
Figure 9 shows the dependence of optical peak power on the current pulse amplitude. Whilst at relatively lower pumping current amplitude, the structures with the SA show weaker pulses due to the increased effective threshold, high enough above threshold, the absorber saturation ensures that the peak powers in the structures with the absorber virtually catch up with those from SA-free (lSA = 0) structures. At 17 A current pulse amplitude, a very similar peak power of ~35 W is achieved in the lasers with the SA compared to the absorber-free ones, but in a narrower, single pulse, as shown in Fig. 8.
4. Summary and conclusions
In summary, we have implemented a saturable absorber, in the simplest form of an unpumped section of a laser cavity next to the front facet, in broad asymmetric waveguide gain-switched lasers pumped by a current pulse source based on standard avalanche pulsing techniques using commercially available components.
The optimal length of the SA, which nearly completely suppresses the trailing part of the optical output while generating an intense, short single optical pulse, is experimentally determined to be 30 µm for a 1.4 mm long laser cavity and the current pulse of 17 A in amplitude and 1.3 ns in duration. This yields a single optical pulse with an energy of about 4 nJ, only about 5-15% less than the useful part of the energy obtained from the structure without the SA, while the pulse duration was about 80 ps (compared to 100 ps in the absorber-free structure).
Measurements with a PIN detector (21 µm in diameter), filtering out the output from the center of the laser stripe, showed shorter pulses compared to streak camera measurements integrating the output from the entire stripe, which suggests that specialized structures with an improved uniformity of the near field profile may yield a further narrowing of the optical pulses while preserving the high energy, which can be beneficial in applications involving SPAD detectors.
The authors would like to thank the ORC at Tampere University of Technology for the laser diode growth; the CMNT institute at the University of Oulu for FIB processing, and Matti Polojärvi and Lauri Hallman for technological support. This work has been financially supported by Infotech Oulu GS, Tauno Tönning Foundation, Emil Aaltonen Foundation, and Ulla Tuominen Foundation. Financial support from the Academy of Finland is also gratefully acknowledged (contract nos. 255359, 251751 and 263705).
References and links
1. A. Kilpelä, R. Pennala, and J. Kostamovaara, “Precise pulsed time-of-flight laser range finder for industrial distance measurements,” Rev. Sci. Instrum. 72(4), 2197–2202 (2001). [CrossRef]
2. E. U. Rafailov and E. A. Avrutin, “Ultrafast pulse generation by semiconductor lasers,” in Semiconductor lasers, A. Baranov and E. Tournie, (Woodhead, Oxford-Cambridge-Philadelphia-New Delhi, 2013), pp. 149–160.
3. K. Y. Lau, “Gain-switching in semiconductor injection lasers,” Appl. Phys. Lett. 52(4), 257–259 (1988). [CrossRef]
4. E. L. Portnoi, G. B. Venus, A. A. Khazan, I. M. Gadjiev, A. Yu. Shmarcev, J. Frahm, and D. Kuhl, “Superhigh-power picosecond optical pulses from q-switched diode laser,” IEEE J. Sel. Top. Quantum Electron. 3(2), 256–260 (1997). [CrossRef]
5. S. N. Vainshtein and J. T. Kostamovaara, “Spectral filtering for time isolation of intensive picosecond optical pulses from a Q-switched laser diode,” J. Appl. Phys. 84(4), 1843–1847 (1998). [CrossRef]
6. S. Vainshtein, J. Kostamovaara, L. Shestak, M. Sverdlov, and V. Tretyakov, “Laser diode structure for the generation of high-power picosecond optical pulses,” Appl. Phys. Lett. 80(24), 4483–4485 (2002). [CrossRef]
7. B. Ryvkin, E. Avrutin, and J. Kostamovaara, “Asymmetric-waveguide laser diode for high-power optical pulse generation by gain switching,” IEEE J. Lightwave Technol. 27(12), 2125–2131 (2009). [CrossRef]
8. B. Ryvkin, E. Avrutin, and J. Kostamovaara, “Quantum well laser with an extremely large active layer width to optical confinement factor ratio for high-energy single picosecond pulse generation by gain switching,” Semicond. Sci. Technol. 26(4), 045010 (2011). [CrossRef]
9. B. Ryvkin, E. Avrutin, and J. Kostamovaara, “Vertical cavity surface emitting lasers with the active layer position detuned from standing wave antinode for picosecond pulse generation by gain switching,” J. Appl. Phys. 110(12), 123101 (2011). [CrossRef]
10. L. W. Hallman, B. Ryvkin, K. Haring, S. Ranta, T. Leinonen, and J. Kostamovaara, “Asymmetric waveguide laser diode operated in gain switching mode with high-power optical pulse generation,” Electron. Lett. 46(1), 65–66 (2010). [CrossRef]
11. L. W. Hallman, K. Haring, L. Toikkanen, T. Leinonen, B. S. Ryvkin, and J. T. Kostamovaara, “3 nJ, 100 ps laser pulses generated with an asymmetric waveguide laser diode for a single-photon avalanche diode time-of-flight (SPAD TOF) rangefinder application,” Meas. Sci. Technol. 23(2), 025202 (2012). [CrossRef]
12. J. Ohya, G. Tohmon, K. Yamamoto, T. Taniuchi, and M. Kume, “Generation of picosecond blue light pulse by frequency doubling of a gain-switched GaAlAs laser diode having saturable absorber,” IEEE J. Quantum Electron. 27(8), 2050–2059 (1991). [CrossRef]
13. G. B. Venus, A. Gubenko, V. Dashevskii, E. L. Portnoi, E. A. Avrutin, J. Frahm, J. Kubler, S. Schelhase, and A. Paraskevopoulos, “Multi-section ion-implantation-induced saturable absorbers for high-power 1.5-µm picosecond laser diodes,” in Proceedings of 17th IEEE International Semiconductor Laser Conference, (Monterey, CA, 2000), pp. 145–146.
14. P. Crump, G. Erbert, H. Wenzel, C. Frevert, C. M. Schultz, K.-H. Hasler, R. Staske, B. Sumpf, A. Maaßdorf, F. Bugge, S. Knigge, and G. Tränkle, “Efficient high power laser diodes,” IEEE J. Sel. Top. Quantum Electron. 19(4), 1501211 (2013). [CrossRef]