GaSb superluminescent diodes with broadband emission at 2.55 lm

We report the development of superluminescent diodes (SLDs) emitting mW-level output power in a broad spectrum centered at a wavelength of 2.55 μm. The emitting structure consists of two compressively strained GaInAsSb/GaSb-quantum wells placed within a lattice-matched AlGaAsSb waveguide. An average output power of more than 3 mW and a peak power of 38 mW are demonstrated at room temperature under pulsed operation. A cavity suppression element is used to prevent lasing at high current injection allowing emission in a broad spectrum with a full width at half maximum (FWHM) of 124 nm. The measured far-field of the SLD confirms a good beam quality at different currents. These devices open further development possibilities in the field of spectroscopy, enabling, for example, detection of complex molecules and mixtures of gases that manifest a complex absorption spectrum over a broad spectral range.

We report the development of superluminescent diodes (SLDs) emitting mW-level output power in a broad spectrum centered at a wavelength of 2.55 lm. The emitting structure consists of two compressively strained GaInAsSb/GaSb-quantum wells placed within a lattice-matched AlGaAsSb waveguide. An average output power of more than 3 mW and a peak power of 38 mW are demonstrated at room temperature under pulsed operation. A cavity suppression element is used to prevent lasing at high current injection allowing emission in a broad spectrum with a full width at half maximum (FWHM) of 124 nm. The measured far-field of the SLD confirms a good beam quality at different currents. These devices open further development possibilities in the field of spectroscopy, enabling, for example, detection of complex molecules and mixtures of gases that manifest a complex absorption spectrum over a broad spectral range. V C 2018 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/1.5015974 Short wavelength infrared light sources emitting in the 2-3 lm wavelength range are very important for trace gas sensing, molecular spectroscopy, and chemical process monitoring. 1,2 In particular, light sources emitting at 2550 nm enable the detection of gases, 3 such as nitrous oxide (N 2 O), 4 acetylene (C 2 H 2 ), 5,6 carbon monoxide (CO), 3 and carbon dioxide (CO 2 ), 3 and the in-situ measurement of humidity (H 2 O), 7 utilizing their spectroscopic fingerprint in this spectral range. Monitoring C 2 H 2 , CO 2 , and N 2 O is very important in the quest of reducing their detrimental impact on climate change and air pollution. Moreover, monitoring H 2 O has important applications in optimizing different industrial processes, for example, in diagnosing the efficiency of a combustion engine. However, currently available light sources lack a suitable combination of performance in terms of attainable output power, spectral coverage, compactness, power consumption, and price. GaInAsSb/GaSb-based type-I quantum well (QW) lasers have shown good performance between 2 lm and 3 lm wavelengths. [8][9][10] However, since the typical wavelength coverage and tuning of a conventional diode laser are limited to a few nanometers, 11 broadband spectroscopy would require multiple diode lasers with different emission wavelengths and/or complicated external cavity elements for tuning to detect several gases simultaneously. On the other hand, superluminescent diodes (SLDs) exhibit broadband emission from single transverse mode ridge waveguides (RWGs), which is easy to collimate, focus, and couple into an optical fiber or a silicon waveguide. Despite these attractive features, the development of SLDs in spectral regions beyond 2 lm has seen little progress. This is partly due to relatively low spread of GaSb optoelectronics technology. Owing to the increased demand for spectroscopy, several groups have recently reported developments concerning SLDs with emission wavelengths up to 2.4 lm. The leading results are concerned with continuous wave (CW) operation of SLDs at room temperature (RT) with single transverse mode output powers up to 60 mW at 1.90 lm, 12 40 mW at 2.05 lm, and 5 mW at 2.38 lm (Ref. 13) emission wavelengths. The severe power degradation at longer wavelengths is mainly due to the thermally activated Auger recombination process, 14 which increases with the wavelength, and is the dominant non-radiative recombination process at room temperature. 14,15 Moreover, the SLD carrier density in the active region increases monotonically with the current 16,17 rapidly increasing the Auger recombination rate as this is proportional to the third power of carrier density. 18 In this letter, we report the development of a SLD at 2.55 lm, delivering mW-level average output power. In order to reduce the average carrier density, thus minimizing the carrier heating effects and the non-radiative Auger recombination, we use a pulsing scheme that allows us to reach mW-level average power while maintaining a high quality broadband spectrum. Operation at a high peak injection level without lasing, and hence high average power, is rendered possible by employing a cavity suppression (CS) element reported recently for the 2 lm wavelength. 12 High gain operation is particularly important for SLDs, as their output power is exponentially proportional to the modal gain. 19 The combination of the CS element and an advanced drive strategy allowed us to demonstrate an average output power of more than 3 mW at RT and a spectral full width at half maximum (FWHM) of 124 nm.
The epitaxial structure was grown on an n-GaSb substrate by molecular beam epitaxy. The structure comprised two 2% compressively strained Ga 0.54 In 0.46 As 0.13 Sb 0.83 QWs placed between a lattice-matched Al 0.25 Ga 0.75 As 0.02 Sb 0.98 waveguide with a total thickness of 540 nm. The compressive strain ensures the band alignment for type-I QWs with a 200 meV valence band offset. 20 The photoluminescence spectra measured for a separate QW-sample peaked at 2.55 lm. The active region was embedded in 2 lm thick p-and n-Al 0.6 Ga 0.4 As 0.04 Sb 0.96 a) Author to whom correspondence should be addressed: nouman.zia@tut.fi.
cladding layers. The doping levels were linearly graded from 2 Â 10 17 cm À3 to 2 Â 10 18 cm À3 in 500 nm thick cladding layers to minimize the free carrier absorption near the active region. The epitaxial structure was then capped with a 200 nm highly p-doped GaSb layer. The choice of the vertical structure was selected to provide a single mode waveguide in the growth direction and to maximize the modal gain provided by the QWs. Lateral optical confinement was ensured by employing a ridge waveguide (RWG) design to guide only the fundamental transverse mode. The single transverse mode operation of the ridge waveguide was first optimized by maximizing the confinement of the fundamental mode under the ridge with respect to the higher order transverse modes. 21 Transverse optical field profiles were obtained from solving the scalar Helmholtz eigenvalue problem with homogeneous Dirichlet boundary conditions using a finite difference scheme. 22 The boundaries were set far away so that they did not interfere with the solution validity. The epitaxial structure was optimized to ensure the high vertical confinement for the fundamental transverse mode than the other modes. The transverse mode confinement was optimized experimentally by selecting a ridge width and an etching depth combination to achieve a stable lobe free far-field profile and a kink-free light-current (L-I) characteristic.
In superluminescent diodes, the light reflection from the waveguide facets must be suppressed for low spectral modulations and broadband emission. This constraint requires a specific longitudinal design of the SLD waveguide. In general, cavity feedback suppression in SLDs is achieved by reducing the reflectivity of both 19 or one facet. 23 The most common approaches to prevent lasing inside the cavity are using ultralow reflectivity antireflection (AR) coatings, 24 adding a highly active 25 or a passive absorber 26 section into the resonator, or tilting the waveguide with respect to the cavity facets. 19 An AR-coating can only decrease the reflectivity to 10 À4 at a single wavelength; a multilayer AR-coating can reduce reflectivity over a relatively broad wavelength range, but the fabrication of multilayer AR-coatings makes the process difficult. A passive absorber section is not capable of completely suppressing the lasing inside the cavity. An active absorber section is a technique that is more effective, but it causes processing and biasing issues. A tilted waveguide is a rather simple approach, which can provide an extremely low facet reflectivity of below 10 À5 . Recently, 12 we introduced a waveguide design in which the lasing was suppressed in 1.9 lm SLD by adding a CS element at one facet of a tilted RWG SLD. In the current report, we have tilted the RWG 8 with respect to the cavity facet, as the larger tilt angles suppress the lasing more effectively at longer wavelengths. 13 The length of the CS element is 90 lm, and the overall length of the device, including the RWG, is 2.5 mm. The choice of a long RWG has two main reasons: (1) it enables the high power operation as the SLD's single-pass gain depends exponentially on the length of the cavity and (2) it decreases the carrier density, which in turn lowers the Auger recombination. The schematic of our device is shown in Fig. 1, with the corresponding geometrical parameters in Table. I.
The preparation of the waveguides was done using UV-contact lithography to define the 5 lm wide ridge waveguides etched 1800 nm into the p-cladding using an inductively coupled reactive ion etch system with Cl 2 /N 2 -plasma chemistry. These waveguides were passivated with a 100 nm thick SiN-layer deposited using the plasma enhanced chemical vapour deposition (PECVD) technique. The SiN-layer was removed from the top of the waveguide in order to open a path to inject the current, and the p-side contact was deposited on the SiN and on the opened area. The p-side contact consisted of a Ti/Pt/Au-layer structure. To allow subsequent cleaving of high quality facets, the substrate was thinned down to 140 lm thickness and the n-side of the sample was metallized to produce the n-contact. For this step, we used an annealed Ni/Au/Ge/Au layer stack. Prior to testing, the chips were mounted on AlN-submounts with an epoxy-adhesive containing silver particles.
Output power characteristics were measured under pulsed injection with a constant pulse-width of 500 ns. We measured the peak and the average output powers by varying the current, duty cycle (DC), and heat sink temperature. Figure 2 shows the average and peak output powers versus the duty cycle for a set of different peak currents. At low currents (up to 400 mA) and up to 35% DC, the average power is nearly a quasi-linear function of DC without any clear sign of thermal degradation, as observed in Fig. 2(a). By increasing the current, device heating becomes significant and a thermal roll-over appears. For high currents, the thermal rollover becomes sharper and the maximum average power shifts towards low duty cycles. The effect of the device heating can also be seen from the behavior of the peak power, as shown in Fig. 2(b). The peak power is high for higher currents at low duty cycles, which is not the case at high duty cycles. After the point of thermal roll-over, the peak power for the highest current (1400 mA) starts lagging the peak power for the lower current (1200 mA). This continues until the peak power for the lowest (300 mA) current becomes greater than for the highest current (1400 mA). For higher currents, the device heating starts to be more important and operating the device at higher duty cycles becomes more difficult. Figure 3(a) shows the peak (solid line) and average powers (dashed-dotted line) of the SLD versus DC, at different temperatures, under a constant pulse injection of 1000 mA current. By decreasing the heat sink temperature from 30 C to 10 C, we have observed a shift in the DC, corresponding to the maximum average output power, from 14% to 26%. This shift in the duty cycle is due to a decrease in the non-radiative Auger effect at low temperature, which permits the operations of the device at high DCs, and high average power, without early thermal roll-over. The light-current (L-I) characteristic curves at different temperatures and at their corresponding maximum DCs are shown in Fig. 3(b).
The L-I curve shows that the SLD emits maximum average power around 6 mW at 10 C and a corresponding peak power of 22 mW. The emission spectra of the SLDs were measured using a monochromator with 0.4 nm resolution, under pulsed input current at 20 C temperature. The FWHM of the spectrum measured at 600 mA current is 107 nm, which indicates that the device operates in the superluminescent regime. An increase in the current from 600 mA to 1200 mA changes the spectral FWHM from 107 nm to 124 nm, as shown in Fig. 4(a). This increase in the FWHM is due to the fact that the material gain bandwidth is growing faster than the square root of the effective device gain when the current density is increased. 27 The effect of temperature on the emission spectra is also examined and shown in Fig. 4(b). We can notice that an increase in temperature causes a red shift of 1.3 nm/ C in the spectral peak. This shift in the wavelength is attributed to the decrease in the bandgap with temperature, which for bulk GaInAsSb was estimated to be 1.5 nm/ C. The spectral broadening with the temperature increase is attributed to the gain broadening, as the carriers start to fill high-energy states due to thermal excitation. We have also observed dips in the trailing edge of spectra, as seen in Fig. 4. These spectral dips are caused by the absorption characteristics of air, because spectral measurements were conducted in a non-vacuum environment.
In order to confirm the quality of an emitted beam from SLD, we measured the shape of the far-field distribution by using a beam profiling pyroelectric camera (Pyrocam III, Ophir). A collimated beam was incident on the pyroelectric camera, and the far-field was measured at different pulse injection currents. Beam profiles, shown in Fig. 5, for different levels of injection, reveal side-lobe-free intensity patterns across the current range indicating transverse single mode operation.
In conclusion, we report the development of a roomtemperature mW-level broadband SLD at 2.55 lm using type-I GaInAsSb/GaSb-based compressively strained quantum wells. We used pulsed driving to minimize the carrier heating and thus the non-radiative Auger recombination rate. In particular, injection with sub-ls pulses leads to high peak gain values. In addition, for a high current injection and hence high output power, we employed a recently proposed cavity suppression element. SLD chips delivered up to 3.2 mW average power at 20 C heat sink temperature, under 1 A peak current and 22% duty cycle (440 kHz repetition rate). A 10 K drop in the mount temperature increased the power up to about 6 mW. For a lower duty cycle of 5%, the SLDs delivered 38 mW of peak power. Devices exhibited wide emission spectra of 107 nm-124 nm, centered around 2526 nm-2540 nm depending on the operating temperature and current. The far-field of a device shows good beam quality at high currents without any side lobes. Such a level of output power, broad emission spectrum, and beam quality are instrumental for simpler implementation of spectroscopic systems in the field of trace gas sensing, molecular spectroscopy, and chemical process monitoring, where multiple emission lines over broad wavelength coverage need to be monitored simultaneously.
This work was carried out as a part of EU Horizon 2020 program MIREGAS (Grant Agreement No. 644192) and Academy of Finland Key project funding MIRELIGHT (Grant Agreement No. 644192). The authors would like to thank Mr. Riku Koskinen for his contribution in epitaxial growths and Ms. Maija Karjalainen for her contribution to device processing.