1. Introduction

The generation of ultrafast high intensity optical pulses by means of superradiance (SR) has been investigated both theoretically and experimentally since the concept was first proposed by Dicke in 1954 [1, 2]. Superradiance is different from other methods of short pulse generation by laser diode devices, such as Q-switching and mode-locking, because of its ability to generate femtosecond pulses on demand with peak optical powers on occasion exceeding 200 W [2]. It has been previously established [3] that for the generation of ultrashort superradiant pulses two conditions must simultaneously be fulfilled. The first condition, in practice a most challenging one, is the creation of a very large e–h density, much larger than typical densities required for lasing. The second one is the presence of a resonant electromagnetic field which can assist the pairing of electrons and holes located at the bottom of the conduction band and the top of the valence band. These requirements can be met in multisection diode laser devices where a built-in saturable absorber frustrates lasing and provides the required profile of gain and absorption in the cavity [3].

Until recently, superradiance has only been reported experimentally in near infra-red semiconductor GaAs/AlGaAs/InP heterostructures at room temperature at wavelengths between 880 nm and 1580 nm [2, 4, 5]. There is however much interest in generating short optical pulses from InGaN devices in the 400-450 nm wavelength range for high density storage, sensing and medical applications [6–8]. To date Kono et al. [9] have reported 10 ps, 12 W pulse generation at a repetition rate of 100 kHz from a gain-switched GaInN laser with an emission wavelength of 405 nm. Using post-amplification, Koda et al. [10] generated a 3 ps, 100 W pulse train at repetition rate of 1 GHz from a GaInN passively mode-locked master oscillator power amplifier system within an external cavity at an emission wavelength of 404 nm.

In this work, we have for the first time experimentally demonstrated superradiant emission from an InGaN semiconductor laser diode device operated at room temperature. The generated optical pulses have powers in excess of 2.8 W and durations as short as 1.4 ps at repetition rates up to 10 MHz at a wavelength of 408 nm

This paper describes the characteristic features of superradiant pulse generation in semiconductor lasers, which are a much higher peak intensity, a spectral red shift upon entering the superradiance, anomalously large timing jitter [2] and greatly reduced time bandwidth product, significantly lower than that found in the Q-switched regime [11].

2. Experimental setup description

The laser diode used in this experiment is a commercially available single transverse mode device (SHARP GH04125A2A) mounted on a 5.6 mm TO can. Focused ion beam (FIB) etching is used to modify the p-contact metallisation of the uncapped laser to produce a two-section device. The etching forms a 10 µm wide gap between a 100 µm long absorber section located at the rear of the device and a 285 µm long gain section at the front facet. This creates a structure with the absorber representing 25% of the total cavity length. The measured intercontact resistance between the gain and absorber section is 50 kΩ. In operation, the absorber section is biased with a maximum voltage of −8.0 V. The gain section is driven with electrical pulses of up to 500 mA peak, which are several nanoseconds long, from an HP 214B pulse generator at repetition rates between 1 and 10 MHz, the highest value limiting the SR repetition rate achievable. The output power is either coupled using a lensed fibre to a New Focus 1004 40 GHz visible photodiode or collimated using free space optics into a single-shot streak camera with a temporal resolution of approximately 900 fs. In addition, the optical spectrum of the laser is recorded with a spectrometer with a resolution of 1 nm. After etching, the threshold current is 29 mA and a typical slope efficiency of 1.7 W/A is measured for absorber and gain section electrically connected together.

3. Experimental results

The 40 GHz bandwidth visible wavelength photodiode and a 40GHz bandwidth digital sampling oscilloscope are used to record the time domain waveforms. Initially, an otherwise identical single contact laser diode is studied in the gain-switched regime and is found generate pulses with a minimum duration of 22 ps at a repetition rate of 600 MHz with typical time bandwidth products greater than 20 [12]. Subsequently the two-section structure created using FIB etching is studied. The gain section is biased with an electrical pulse of 190 mA amplitude and 9 ns duration (at full width half maximum). The device is operated at constant temperature of 20 °C. The temporal waveform for 9 ns long electrical pulse applied to the gain section is provided in Fig. 1a . A comparison of optical waveforms in the superradiance regime for reverse bias levels of −1.9 V, −3.4 V and −3.7 V to the absorber section is shown in Fig. 1b, 1c and 1d. At low reverse bias of −1.9 V, 8 superradiant pulses are generated (Fig. 1b) and with increasing reverse bias the number of pulses is reduced from 3 (Fig. 1c) to a single superradiant pulse (Fig. 1d). This operation is similar to that observed from near infra-red devices operating in the superradiance regime [2]. Reverse bias applied to the absorber section introduces a band-edge shift due to Quantum Confinement Stark Effect (QCSE). Therefore the increased reverse bias increases the initial waveguide losses and hence the onset of optical pulse(s) is delayed. Furthermore, with increased reverse bias the times required for full gain recovery and the emission of further optical pulses increase (Fig. 1).

 

Fig. 1 (a) Electrical pulse waveform with amplitude of 190 mA, pulsewidth equal to 9 ns applied to the gain section and output waveforms for gain section biased with 190 mA, 9 ns long el. pulse and absorber biased with (b) −1.9 V, (c) −3.4 V and (d) −3.7 V.

Download Full Size | PDF

When the reverse bias is increased up to −3.7 V (Fig. 1d), superradiance is achieved with the first optical pulse being delayed by more than 2 ns. As the carrier concentration builds to high values with the delay of optical emission and the strong current drive, the optical pulse, when it is generated, has a large optical peak power in excess of 2.5 W. The large energy of the optical pulse however causes very strong depletion of the optical gain, and as a result the generation of additional pulses does not take place even if the electrical drive pulse is a as long as 9 ns. Several GaN/InGaN devices have been tested for fixed amplitude of gain current. In all instances values of initial pulse delay and reverse bias were found to be of the same order of magnitude. Such a delay enables an estimated increase in e-h pair density of approximately 8 × 10¹⁹ cm⁻³ assuming 100% efficiency. In AlGaAs/GaAs heterostructures an increase in e-h pair density over 4 × 10¹⁸ cm⁻³ has been reported in 880 nm superradiant lasers [2]. In Nitride heterostures such as GaN/InGaN, order of magnitude higher densities are expected owing to higher effective masses of both electron and hole. Peak powers are observed to increase from 70 mW in the gain-switched regime to more than 2.5 W in the superradiance regime.

More accurate measurements of the pulsewidth have been carried out using a single-shot streak camera. Depending on the nanosecond current pulse amplitude and the reverse bias, output superradiant pulse trains consist of bursts of up to 10 individual optical pulses. By increasing the reverse bias up to −8.0 V, single superradiant pulses are generated for each electrical pulse. The estimated energy of each burst of superradiant pulses is typically around 40 pJ. Due to quantum-mechanical fluctuations of the build-up of the superradiant pulses [2], individual realizations of each superradiant pulse are intrinsically different. Figure 2 presents two superradiant pulses detected from the violet GaN/InGaN device under test at higher reverse biases.

 

Fig. 2 Streak camera photos and superradiant pulse traces from a laser diode biased at 250 mA current and –7.0 V reverse bias (a) and from a laser diode biased at 270 mA and –5.4 V (b).

Download Full Size | PDF

The typical pulsewidth of the individual pulses in the bursts is less than 5 ps. Upon deconvolving the 900 fs response time of the streak camera, the shortest deconvolved pulsewidth is less than 1.4 ps (being much less than the round trip time of the device which is 8.6 ps), the peak power is over 2.8 W and the time bandwidth product is 3.7. The peak power is calculated after subtracting the energy in the background of the pulses. The calculated time bandwidth product from the two-section laser diode is larger than typical time bandwidth products for three-section bulk 880 nm lasers of between 1 and 3 [2], but significantly less than both that for gain switched operation reported earlier in this paper and that reported by Kono et al. [9].

Enhanced timing jitter is another characteristic feature of superradiant emission [2]. We compare here the jitter observed in gain-switched and superradiance regimes measured from the same devices at different driving conditions. Figure 3 illustrates the gain-switched (left) and superradiant (right) regimes detected using the photodiode and the sampling oscilloscope. The gain-switched pulse is clearly seen on the screen, whereas superradiant pulses produce a blurred picture.

 

Fig. 3 Pulse train for (a) single contact device and pulse current equal to 140 mA and (b) a two-section device at biased with pulsed current equal to 190 mA and reverse biased absorber at −3.5 V.

Download Full Size | PDF

More accurate measurements of the jitter are presented in Fig. 4 . Figure 4a shows the jitter of gain-switched pulses versus the driving current amplitude. The jitter drops from around 25 ps down to below 10 ps as the current increases. These values include the intrinsic jitter of the pulse driver and oscilloscope. By contrast, the measured jitter increases from 25 ps up to 200 ps as the superradiant emission develops with increasing reverse bias on the absorber section (Fig. 4b). The physical reason of enhanced timing jitter of superradiant pulses has been explained elsewhere [2] and originates from fundamental low-frequency fluctuations of the radiative decay of the coherent macroscopic cooperative e-h state.

 

Fig. 4 Jitter measured for (a) single contact laser diode as a function of increased pulsed current amplitude (gain-switching) for (b) two-section laser diode with fixed pulsed current amplitude and increased reverse voltage (superradiant regime).

Download Full Size | PDF

A comparison of the emission optical spectra in the gain-switched and superradiance regimes is provided in Fig. 5 . This shows that with increased reverse bias, the peak optical wavelength is red-shifted by 2.7 nm (4.86 THz). This is similar to the red shifts reported in other material systems. For instance, Xia et al. [5] reported a red-shift in InP quantum well laser diode from 1550 nm to 1590 nm (4.87 THz) and Vasil’ev [2] measured red-shift in 880 nm bulk lasers from 874 nm to 887 nm (5.03 THz).

 

Fig. 5 Optical spectra acquired for gain-switched (black) and superradiance regime (red)

Download Full Size | PDF

4. Discussion and conclusion

In this paper we have demonstrated for the first time the generation of ultrafast pulses by superradiance in a violet GaN/InGaN laser. This two-section device, operating at 408 nm, produces short optical pulses with peak powers in excess of 2.8 W with durations of 1.4 ps on-demand, up to repetition rates up to 10 MHz, limited by the pulse driver. This flexible method of pulse production is suited to high density storage and biological imaging applications, provided that the large timing jitter is strongly decreased by external means.