Electrically pumped all photonic crystal 2^nd order DFB lasers arrays emitting at 2.3 μm

The development of new high sensitivity techniques in absorption spectroscopy represents a high potential for industrial applications, environmental monitoring, medical diagnostics, agronomy, etc. Tunable diode laser absorption spectroscopy (TDLAS) is one of these high sensitivity techniques that are currently extensively studied. Its main advantages are the potential compacity and the robustness of laser diodes that lie at its heart. This application motivates a large background work on laser diodes’ development at the mid-IR wavelength. These developments include GaSb-based laser diodes that offer the best material properties to make high quality lasers in the 2-3 μm range. This range is of particular interest for air quality monitoring which requires multi-species detection,¹ such as CO₂ as greenhouse gas, CO due to incomplete combustion processes, CH₄ as greenhouse gas, marker of anthropic activities (livestock, biomass energy), nitrous oxides associated with asthma, lung cancers, and respiratory diseases due to combustion and industrial processes, SO₂ due to the combustion of sulfides.

The development of single frequency GaSb laser-diode sources for such applications has been tremendous in the last years. It included intra-cavity photonic crystal mirrors,² metallic Distributed Feedback (DFB) laser diodes with complex coupling (losses) strategies,³ lateral waveguide corrugations,^4,5 which have also been fabricated using low-cost nanoimprint lithography,⁶ or even buried grating structures.⁷ Very recently, DFB lasers using heterogeneously integrated GaSb type II quantum-wells (QWs) have also been demonstrated on silicon.⁸ However, such developments do not address yet the second strategic requirement for TDLAS applications which is the tunability spectral range of the laser diode. Indeed, whilst single absorption-line monitoring does not require a multi-nanometer spectral range of emission, analysis of complex gas mixtures will require the recording of absorption spectra that cover several absorption-lines to enable gas discrimination.

In this paper, we investigate GaSb, edge-emitting, single-mode laser-diode integrated arrays in order to control the emission over a large spectral range. To fulfill this goal, 2D Photonic Crystal (PhC) 2^nd order DFB laser diodes are particularly attractive. Indeed, the use of 2^nd order DFB gratings increases the yield of single-mode DFB lasers.⁹ Moreover, 2D DFB gratings offer enhanced modal selectivity.¹⁰ 2D Photonic Crystal (PhC) 2^nd order DFBs thus open the way to high fabrication yield of arrays of single-mode DFB laser diodes.

All PhC 2^nd order DFB lasers were designed following the double optimization scheme depicted in Refs. 11 and 12. In this approach, the lateral confinement of the laser mode is achieved using a defect waveguide. It is formed inside a triangular-lattice PhC made of air holes etched in a high-index vertical stack by omitting a given number of rows of holes (see Fig. 1). The properties of this waveguide are then tuned by deforming both the transverse lattice constant of the PhC and the width of the defect waveguide. We name ϵ the deformation of the transverse lattice, and α the waveguide deformation (see Fig. 1(a)). For practical purposes, we choose to use large W5 waveguides (5 missing rows of holes along the $ΓK$ direction), providing a good compromise between lateral confinement of the optical mode and large active area required for thermal management imposed by electrical pumping. The waveguide stack under investigation was composed of a 1.5-μm-thick top cladding layer (optical index n_sup = 3.38), a 0.8-μm-thick core layer (optical index n_c = 3.615), and a 0.6-μm-thick bottom cladding (optical index n_bottom = 3.17). These optical indexes correspond to AlGaAsSb with 65% and 90% Al-content for the top and bottom claddings. The core refractive index corresponds to the equivalent index of the confinement heterostructure of a standard 3-quantum-wells (QWs)2.3-μm-emitting GaSb laser diode. Based on previous studies,¹² we choose to work with a waveguide deformation of α=1.24 and PhC deformation ϵ=0.78. We use 3D-finite-difference time domain (FDTD) modelling to determine the evolution of the Q factors of the 2 DFB modes as a function of the holes’ depth across the vertical stack. For calculations, we consider a single period of the PhC lattice, with periodic boundary conditions along the waveguide direction and absorbing boundaries (Perfectly Matched Layers—PML) along the other edges of the simulation cell. Such simulations give access to the two eigen-modes of the infinitely long waveguide, as shown in Figure 1(a) and the associated Q-factors (see Fig. 1(b)). The two DFB eigen-modes, labelled n and a, exhibit, respectively, a node and an antinode at the center of the cell. For deeply etched PhC holes, such a deformation provides a high modal selectivity between the 2 modes.¹² As seen in Fig. 1(b), the a mode reaches Q factors higher than 10⁷ whilst the n mode’s Q factor stagnate at 10⁵. However such deep holes are difficult to fabricate technologically, and they require etching through the QWs, which degrades significantly the optical gain due to increased non-radiative carriers’ recombination rate. We have thus investigated the modal selectivity of this design for reduced hole depths with 3D-FDTD modelling, as presented in Fig. 1(c). As expected, reducing the depth lowers the Q factors of both modes.¹³ However, for holes etched down to just above the waveguide core, the a mode exhibits Q-factors as high as 4 × 10⁴, almost one order of magnitude larger than that of the n mode. We thus chose a design for the W5 waveguide with holes etched down to 300 nm above the core waveguide (solid red line in Figure 1(b)). Arrays of waveguides with periods ranging from 654 to 694 nm by 4-nm steps were fabricated. The PhC waveguide is surrounded by narrow isolation trenches lying 1 μm away from the last PhC holes to prevent current leakage from the active area through the contact layer.

The laser structure was grown by molecular beam epitaxy (MBE) in a system equipped with valved cracker cells for both Sb and As. The growth was carried out on an n-type GaSb substrate at a temperature of 480 °C. After the n-GaSb buffer layer, a 100-nm graded Al_xGa_1−xAs_ySb_1−y layer was grown to insure a linear transition from GaSb to the 1.2-μm-thick n-Al_0.9Ga_0.10As_0.07Sb_0.93 cladding layer. The active region is made of two 10-nm-thick Ga_0.66In_0.34As_0.06Sb_0.94 quantum wells (QWs) separated by 30-nm-thick Al_0.25Ga_0.75As_0.02Sb_0.98 barriers. The QWs are embedded between two 400-nm Al_0.25Ga_0.75As_0.02Sb_0.98 waveguide layers. The top of the structure consists of a 1.5 μm-thick p-Al_0.65Ga_0.35As_0.06Sb_0.94 cladding layer, a graded AlGaAsSb layer, and a 300-nm-thick p-GaSb contact layer.

After growth, a 200-nm thick SiO₂ hard mask was deposited by PECVD. E-beam lithography with a ZEP520 electron sensitive resist followed by CHF₃ dry etching was used to transfer the patterns in the hard mask. For AlGaAsSb deep etching, we used a dry etching cyclic process inspired from the silicon Bosch process, similar to the process depicted in Ref. 14. The etching parameters were optimized for the 65% aluminum composition used in the upper claddings of the laser diode heterostructure. As compared to the process described in Ref. 14, argon was added to the Cl₂/N₂ etch gas mix, with respective fluxes of 45 sccm Cl₂, 15 sccm N₂, and 5 sccm Ar. The plasma excitation power was decreased to 100 W, almost equal to the acceleration field power of 60 W, reaction conditions extremely close to that of Capacitive Coupled Plasma (CCP)-Reactive Ion Etching. The next two steps, O₂ plasma cleaning and N₂ plasma passivation, were kept as described in Ref. 14. The whole cycle was repeated 6 times, leading to 1.35-μm-deep holes. Taking into account the 300 nm thick contact layer, the holes’ bottom lies 450 nm above the confinement layer of the mode (dashed line in Fig. 1(b)).

These first process steps were followed by a standard laser diode process. A 200-nm-thick SiO₂ passivation layer was deposited by PECVD, followed by dry etching for the contact areas on top of the waveguide. The last steps were Ti/Au metal deposition for p contact, sample thinning, and Ti/Au back contact deposition. 1-mm-long laser bars were then cleaved. Fig. 1(c) shows a scanning electron microscope image of the facet of a fabricated component, showing that the holes’ bottom is 450 nm above the waveguide core. As can been seen in Figure 1(b), this depth (red dashed line in Figure 1(b)) is slightly lower than the targeted one, resulting in lower Q factors for both modes. However, for this depth, the spectral selectivity is high enough to ensure single-mode emission as shown in Sec. III. As shown in Figure 1(c), good sidewall smoothness was achieved, and some silica deposition occurred on sidewalls and at the bottom of each hole during the insulating layer deposition. Finally, the arrays incorporating 12 lasers (11 DFB lasers and 1 Fabry-Perot (FP) laser) separated laterally by 400 μm were reported on a printed circuit board (PCB) submount (Fig. 2(a)).

Fabricated samples were tested on a temperature-controlled test-bench under CW excitation, the edge emission being collected by a calibrated photodiode or sent to a spectrometer. Fig. 2(b) shows the light-current curves recorded for PhC periods between 670 and 690 nm and for a wide-area Fabry-Perot laser delimited only by the isolation lateral trenches. Each laser exhibits a clear threshold in the [140:190] mA range ([3.2:4.3] kA/cm² range). The differential efficiency at threshold is 0.011 W/A. The emission spectra were measured using a 26-cm Cornerstone monochromator from Newport, with a 300 lines/mm grating used at first order. The 40 μm in and out slits give to the monochromator a theoretical spectral resolution of 0.5 nm.

As seen in Fig. 2(c), PhC lasers exhibit narrowband emission spectra with a large suppression of the Amplified Spontaneous Emission (ASE) as compared to the FP laser. We attribute this strong suppression to the high selectivity of 2^nd order DFB PhC waveguides.¹⁰ The emission is single-mode and the best side-mode suppression ratio is estimated to be 30 dB.

These narrow emission lines can also explain some of the dents and other deviations from a straight line observed above threshold on the light-current curves (particularly visible for a = 682 nm and a = 678 nm). As the narrow emission gets spectrally shifted with increasing pumping current, it passes through the absorption lines of water vapour present in this spectral range.

In Fig. 3 we report the emission wavelength of the PhC lasers versus the PhC period (lattice constant a). As can be seen, two distinct linear branches are obtained above and below $a=682$ nm. For $a≤682$ nm, the emission wavelength scales linearly with the PhC period, with a $λ/a$ ratio sensibly equal to $λ/a=3.53$⁠, in good agreement with $λ/a=3.51$ calculated from the value of the effective index of the fundamental mode. Given the lattice constant difference of $Δa=4$ nm between adjacent lasers in the array, the emitted wavelengths form a coarse comb with a 15 nm spacing.

It can also bse seen that for $a≥682$ nm, the emission jumps to a second linear branch, with a lower effective index of 3.41. We attribute this second branch to the excitation of the first transverse high-order mode in the defect waveguide, which also exhibits a DFB behavior, but with a slightly lower effective index. As the PhC periods increase, the emission wavelength of the fundamental mode steps outside from the gain region whilst the first excited mode wavelength steps inside,giving rise to this mode jump. On such DFB lasers, emission wavelength control can be achieved with a submount temperature or injected current. The emission wavelength sensitivity to temperature was measured to be 0.21 nm/°C, whilst its tuning with current is 0.026 nm/mA (see Fig. 4). Such a tuning rate would require 550 mA-current modulation amplitude to cover the 15-nm spacing between adjacent PhC lasers. Thus, a re-design of the PhC array with smaller spacing is needed. A lower PhC lattice constant change between adjacent lasers of $Δa=1$ nm (instead of $Δa=4$ nm) would allow continuous spectrum coverage with a reasonable current modulation. Such a period pitch can be attained using modern nano-lithography tools.¹⁵ Moreover, the center lattice constant could also be adapted to fit the whole spectral range of our 11 lasers’ array to the gain range achievable with our QWs’ design.

To investigate the potential for multi-gas sensing of the lasers, we have used sample calibrated cells (from Wavelength References) filled with a controlled pressure of pure CH₄, CO, and HF. One laser among the 12 (11 DFB lasers and 1 FP) of the array was selected. We have worked with a laser having a grating period of a = 678 nm. Its wavelength was then current-tuned to cross different absorption lines at a fixed temperature of 15 °C, demonstrating the possibility of a multi-species detection. As can be observed in Figure 5, the measured lines (red curves) are well defined, with a very good spectral fit and a transmission amplitude approaching the theoretical value (gray curves) given by a simulation of the cells’ transmission with HITRAN-PC software.¹⁶ We explain the fluctuations between the experimental and theoretical curves by the misestimation of the cell windows’ transmission. These first results on one laser show that the spectral properties are adequate for TDLAS. However, for the moment we have not exploited all the potential of our approach. Indeed, using simultaneously all the sources of a re-designed array would allow a continuous coverage of a 40 nm-spectral range at $2.3μ$m, opening the way to the analysis of complex gas mixtures.

To conclude, we have demonstrated all-photonic-crystal, electrically pumped 2^ndorder DFB laser diodes emitting at 2.3 μm. These lasers cavities are based on the so-called double optimization approach using asymmetrically etched holes which provide a compromise between low-loss PhC designs and QWs gain preservation. Such a design is more readily fabricated than deeply etched holes and still provides modal selectivity allowing the demonstration of arrays of single mode DFB lasers. The fabricated lasers exhibit relatively high thresholds that could be lowered using optimized fabrication processes and slightly deeper holes. The emission is narrow enough to allow TDLAS in standard test cells with current tuning of the wavelength. Continuous spectral coverage of a 40 nm band using 10 lasers seems an achievable goal using laser arrays with PhC period variations of 1 nm from laser to laser.