1. Introduction

GaN-based wide-bandgap optoelectronic devices have undergone significant advancements with the advent of commercial light emitting diodes (LEDs) and edge-emitting lasers diodes (LDs) in the violet-blue spectral region. GaN-based light emitters are now ubiquitous in several lighting, communication, data storage, display, and sensing applications [1]. Among these, VCSELs have attracted significant attention in recent years due to their inherent advantages over edge-emitting lasers. The short cavity length typically enables single-longitudinal-mode operation. Furthermore, small aperture diameters may also enable single-transverse-mode operation in some cases. The surface-normal geometry and small device size are also conducive to the cost-effective formation of high-density 2D arrays. VCSELs also have a circular beam profile with low divergence for efficient fiber coupling and wafer-normal surface emission, simplify on-chip testing. Their relatively small cavity volumes results in low threshold current and high modulation bandwidth at low bias currents [2]. Typically, III-nitride VCSELs are fabricated on the polar c-plane orientation [3–6]. Polar c-plane devices suffer from polarization-related electric fields in the active region, lowering the per-pass gain and increasing the threshold conditions. The nonpolar m-plane orientation, on the other hand, eliminates internal electric fields, resulting in a near perfect overlap of the electron and hole wave functions [7–9]. The nonpolar orientation also exhibits in-plane gain anisotropy, enabling polarization-pinned emission along the a-direction, commonly sought for atomic clock applications [10–12].

Several research groups have demonstrated optically and electrically pumped GaN VCSELs [4–6,13–17]. One of the crucial challenges plaguing GaN-based VCSELs is the lack of high quality epitaxial DBR with high reflectance. Since a DBR consists of layers of alternating material, it is necessary that the two materials are lattice-matched to prevent the formation of dislocations. Al_0.82In_0.18N is the only suitable III-nitride alloy that has a lattice constant equal to the lattice constant of GaN [6,13]. Unfortunately, the refractive index difference between GaN and Al_0.82In_0.18N layers is only ~0.2, requiring more than 40-pairs to obtain a peak reflectance >99.9%. This results in long growth times in metal organic chemical vapor deposition (MOCVD) for the Al_0.82In_0.18N/GaN DBRs (~8 hours for 40-pairs) and maintaining the correct composition of ternary alloys in the group-III elements is very difficult. Dielectric DBRs are often used in III-nitride VCSELs as a substitute for epitaxial DBRs due to the wide variety of materials available with high index contrast. Hence, dielectric layers can be easily deposited to form high reflectance mirrors with few pairs [17]. However, the use of dielectric DBRs involves added fabrication complexity to gain access to the backside of the cavity. Also, the non-conductive nature of dielectric materials results in devices with poor thermal and electrical conduction characteristics [18].

Another method for obtaining lattice-matched epitaxial DBRs with high refractive index contrast is to introduce subwavelength air-voids or nanopores in the alternating layers of the same material. Nanoporous layers can be achieved by the controlled anodic electrochemical (EC) etching of highly doped n-type GaN in acids [19]. The selective formation of the nanopores in the doped layers lowers the effective refractive index compared to the adjacent undoped GaN layers. We have previously demonstrated nanoporous DBRs on free-standing nonpolar m-plane GaN substrates obtaining peak mirror reflectance greater than 98% at 450 nm and investigated the effect of the doping concentration and the EC bias on the nanopore formation while also verifying the polarization sustainability of the DBRs [20]. Although, optically pumped VCSELs with top and bottom nanoporous DBRs have been demonstrated previously on c-plane GaN [15,16], in this work, we demonstrate the first hybrid-cavity VCSEL with a bottom nanoporous DBR in the nonpolar orientation, leveraging the anisotropic gain properties observed in m-plane GaN. The optical characteristics of nonpolar GaN have been thoroughly investigated in [7–11]. The presence of unbalanced biaxial strain in the nonpolar QWs causes the two topmost valence subbands to split unevenly. The upper and the lower valence subbands exhibit the highest transition strength for an electric field polarized parallel to the a-direction and c-direction, respectively. Since the two subbands are separated in energy, carriers from the upper subband will recombine before the lower subband, resulting in polarized emission along the a-direction. The VCSEL structure presented here is characterized using CW optical pumping at RT, and single-mode stimulated emission is observed in several locations, all of which are polarization-pinned in the a-direction.

2. Experimental details

2.1. Growth

The free-standing m-plane GaN substrates were purchased from Mitsubishi Chemical Corporation (MCC) with doping concentration of ~10¹⁷ cm^-3 and defect density ~10⁵ cm⁻². Using MOCVD, a 3-µm-thick n-GaN template was first grown with a Si concentration of ~7 × 10¹⁷ cm⁻³ to ensure smooth surface morphology with an RMS surface roughness ~0.2 nm. Next, 16-pairs of alternating undoped and doped GaN was grown to form the bottom DBR structure with Si concentration ~1 × 10¹⁹ cm⁻³ confirmed using secondary ion mass spectroscopy (SIMS). The n-side undoped GaN cladding layer was grown next followed by the active region which consisted of 5 pairs of QW/barriers (InGaN/GaN). Finally, the active region was capped with the thin p-side undoped cladding layer. Figure 1(a) shows the cross-sectional schematic of the VCSEL structure and Fig. 1(b) shows the corresponding refractive index profile and the 1D transmission matrix method (TMM) normalized mode intensity plot of the cavity. The thicknesses of the active region along with the thickness of the cladding layers were set such that the losses are minimized while maximizing the gain enhancement factor (Γ_enh ≈1.8) by aligning the peak of the standing wave profile with the active region. This produced a minimum threshold modal gain of ~213 cm⁻¹ and an effective cavity length of 2.1-λ (L_eff × n_eff ≈ 529.5 nm × 1.84). Table 1 list the layers in the cavity, along with measured layer thicknesses, refractive indices, and absorption coefficients used in plotting Fig. 1(b).

 

Fig. 1 (a) Cross-sectional schematic of the hybrid VCSEL structure. (b) 1D TMM simulation showing the corresponding refractive index profile and the normalized mode intensity.

Download Full Size | PDF

Table 1. List of all the materials used in the VCSEL cavity for the TMM simulation. The refractive indices are given for a wavelength of 462 nm and the absorption loss values are rough estimates used to determine threshold modal gain [21,22].

View Table

2.2. Fabrication

The samples were first patterned using standard contact lithography and etched using ICP till the n-side cladding layer to form the device mesa. Then, a ~100 nm layer of SiO₂ was deposited using CVD to passivate the active region sidewalls and to prevent surface roughening from subsequent processing steps. Next, mesa stripes of width 130 µm with 10 µm separation were patterned along the c-direction and etched using a combination of both RIE and ICP to form the EC etch trenches. This exposes the sidewalls of the n++ GaN layers for EC etching. Next, an indium contact was soldered at the corner of the sample and the sample (anode) was then submerged in a solution of 0.3 M oxalic acid (stirred at 100 rpm), ensuring that the contact does not touch the solution. A platinum wire mesh was placed in the solution as a cathode, and the two electrodes were connected to a 5.5 V DC source (shown in Fig. 2(a)) and etched for ~8 hours. This etch bias was used to attain a certain pore size from a given pore density determined by the layer doping concentration [15,16,20]. After the EC etch, the SiO₂ layer was stripped using buffer HF and rinsed in deionized water. This was followed by the blanket CVD deposition of the top dielectric DBR which consisted of 20 pairs of alternating SiN_x and SiO₂ at 100°C. SiN_x and SiO₂ were readily available in our deposition system, but we note that other DBR materials (e.g., Ta₂O₅, HfO₂) would offer reduced absorption loss.

 

Fig. 2 (a) Schematic of the EC etch setup. (b) µ-PL setup where the samples are pumped by a 405 nm laser diode from the top and the emissions are collected from the bottom.

Download Full Size | PDF

2.3. Characterization

The epilayer film thicknesses while growth was monitored in situ using the reactor pyrometry. The fabrication etch depths were measured using profilometry and scanning electron microscope (SEM) imaging. The dielectric film thickness, refractive index, and absorption losses were measured using ellipsometry. The reflectance spectra of the dielectric DBR were characterized using a UV-VIS spectrophotometer which yielded a peak reflectance of 99.9% at 462 nm. The nanoporous DBR reflectance was measured using a combination of previous µ-reflectance data and volume average theory (VAT) estimation [20]. In the VAT, SEM images of the cross section of the porous layers are digitized to calculate the porosity and estimate the layer refractive index based on the index of bulk GaN at the desired wavelength. Then using µ-reflectance, the bottom nanoporous GaN DBR was measured to have a peak reflectance of ~98% at 462 nm. The µ-reflectance setup was then converted to transmission mode µ-photoluminescence (µ-PL) setup to allow optical characterization from the backside while viewing and pumping the devices from the top. Figure 2(b) shows a schematic of the setup which includes a 405 nm 400 mW laser diode with a spot size at the surface of ~8 µm in diameter, thermally conductive SiC sample mount to mitigate self-heating, 0.1 OD filter, 405 nm high pass filter, polarizer, and a spectrometer. All components are mounted on xyz-micro manipulator stages for optimized signal collection.

3. Results

The peak reflectance of the top dielectric DBR was greater than the reflectance of the bottom nanoporous DBR, hence stimulated emission was observed from the bottom of the wafer. The PL spectrum of the emission at various pump power densities is shown in Fig. 3(a) where a clear single-mode stimulated emission peak at 462 nm was observed. Translating the spectrum to an integrated PL plot (Fig. 3(b)), a clear non-linear lasing L-L plot characteristic was observed with a threshold pump density of ~5 kW/cm² with no apparent thermal roll-over. The emission linewidth is also plotted in Fig. 3(b). The gradual drop in the FWHM is the result of double peaking during spontaneous emission at 458 nm and 462 nm. Once stimulated emission commences, the mode at 462 nm subdues the 458 nm peak and the FWHM drops to ~0.12 nm. Note that the maximum output power was limited by the µ-PL setup, where the pump laser had already reached at its maximum output power (6.5 kW/cm²). Additionally, saturation of the spectrometer caused a slight increase in linewidth beyond threshold. Figure 3(c) shows the polar diagram of the integrated PL at three different lasing spots on the wafer. The spectrum was recorded after the polarizer was rotated in 10-degree increments and integrated over the entire spectrum (spontaneous and stimulated emission). The polarization ratios for spot 1, spot 2, and spot 3 were 82%, 72%, and 84%, respectively. The relatively smaller polarization ratios compared to a previously published work on m-plane VCSELs are due to the inclusion of spontaneous emission in the integration of the PL [23]. We note that removing the spontaneous emission and integrating the spectrum across only the stimulated emission region would results in a polarization ratio of ~99%. Polarization-pinned emission in the a-direction was also confirmed showing that the scattering effects from the nanoporous layers was negligible. This is further evidenced in our previous study of the nanoporous DBRs where we concluded that the nanoporous DBRs maintain the polarization properties of an incident polarized source regardless of the pore alignment in the a- or c-directions [20].

 

Fig. 3 (a) µ-PL emission spectra of a VCSEL at various pump power densities. (b) Corresponding integrated PL and FHWM plot. (c) Polar plot showing polarization pinning at 3 different lasing spots.

Download Full Size | PDF

Several other lasing peaks were observed at various regions across the sample. Figure 4(a) shows the PL spectrum of four such spots. It was also observed that the lower index porous layers are smaller in thickness compared to the non-porous GaN layers, shown in the SEM image in Fig. 4(b). To compensate this error in growth, an etch bias of 5.5 V was used to intentionally increase the refractive index difference between the layers by causing the pores to agglomerate to form larger pores. As a result, localized regions of nonuniform effective cavity lengths were formed [20], producing different lasing peaks throughout the sample. The non-uniform reflectance caused by the changing cavity lengths is evident by the colorful streaks from the microscope image in Fig. 4(c). If the layer thickness were correct, EC etch could be performed at lower bias voltages to achieve much more uniform porosification.

 

Fig. 4 (a) Emission spectra of several VCSEL devices showing lasing emission at different wavelengths. (b) Cross-sectional SEM image of the nanoporous DBR, where higher index contrast was required to counter the incorrect epilayer thicknesses. (c) Microscope image of a device mesa with non-uniform in spectra due to large pores sizes.

Download Full Size | PDF

4. Conclusion

To summarize, we have extended our previous work on nonpolar nanoporous DBRs by implementing an active region and forming a hybrid cavity using a bottom nanoporous DBR and a top dielectric DBR. We demonstrated an optically pumped VCSEL with single-mode lasing emission at 462 nm with a linewidth of ~0.12 nm. The stimulated emission threshold was ~5 kW/cm². Several lasing spots were examined for polarization pinning, and all spots were found to be locked in the a-direction, characteristic of m-plane nonpolar GaN devices. The uniformity of the lasing peaks can be improved greatly by growing the DBR stacks in their corresponding λ/4 thicknesses and forming high density small pores in the doped layers. Overall, these results indicate that the nanoporous approach is a strong candidate for solving some of the issues affecting III-nitride VCSELs and achieving high performance devices. Nanoporous DRBs also maintain the advantageous polarization-pinning enabled by the nonpolar orientation.