UC Santa Barbara

Vertical-cavity surface-emitting lasers (VCSELs) are a special class of laser diode that use top-side and bottom-side parallel mirrors to emit a laser beam vertically from the top surface, as compared to conventional edge-emitting lasers that emit light from the sides of the devices. Compared to edge-emitting lasers, VCSELs have many unique attributes, such as a low threshold current that leads to low power consumption, high-speed direct modulation, superior beam quality, and they can be easily arranged into two-dimensional VCSEL arrays for scalable power. This leads to several exciting applications; for example, VCSELs are the key component in the Apple iPhone X that enables Face ID, which allows users to securely unlock their devices simply by looking at their phones. However, the VCSEL market is currently limited to red-emitting and infrared-emitting VCSELs that use GaAs-based and InP-based systems. If shorter wavelength emitting VCSELs could be created, it would open up a whole new world of untapped and exciting applications. For example, blue and green VCSELs could be paired with red VCSELs to create next-generation display and projector technology. The low power consumption and high beam quality of VCSEL-based displays would be particularly promising for virtual and augmented reality systems. Shorter wavelength VCSELs can be created using GaN-based materials, but these devices have been very challenging to create. The first GaN-based VCSEL was demonstrated in 2008, and only eight research groups have demonstrated these devices over the following decade. With the first report in 2012, the University of California, Santa Barbara (UCSB) has been the only group in the United States to create GaN-based VCSELs. Compared to the c-plane GaN VCSELs from all of the other groups, VCSELs from UCSB have used m-plane GaN, which uniquely provides 100% polarized emission for both individual VCSELs and VCSEL arrays. The main problem with GaN VCSELs from UCSB has been their inability to lase under continuous-wave (CW) operation. They could only lase under pulsed operation, which means that these VCSELs would fail if turned on for longer than a fraction of a millisecond. This has been a severe limitation that has prevented most practical applications of these devices. Therefore, the ultimate goal of the research described in this thesis has been to achieve CW lasing. This has been a tremendously difficult goal, and the initial VCSEL designs failed to lase, even under pulsed operation. Despite these initial discouraging results, this experiment was important because it led to an extensive failure analysis that revealed several key problems with the VCSEL design. Surface roughness prior to the DBR mirror deposition turned out to be a major problem that inhibited lasing due to scattering loss and reduced mirror reflectance. After performing experiments to improve the surface morphology, the surface roughness on the p-side was reduced by utilizing an indium flux during MBE tunnel junction regrowth, and the roughness on the n-side was reduced by removing an oxide residue that formed after the photoelectrochemical (PEC) undercut etch to remove the m-plane GaN growth substrate. Another significant problem was VCSEL yield in which only a small percentage of devices successfully transferred onto the flip-chip substrate, and most of those devices were cracked. A series of flip-chip bonding experiments were conducted to optimize the Au-Au thermocompression bond, but the yield only marginally improved at first. The flip-chip bond was also responsible for a severe thermal issue that prevented CW operation in previous devices. Based on thermal modeling in COMSOL, heat generated in the active region could not flow directly downward due to the thermally-insulating bottom DBR, so there was a bottleneck in heat transport through a relatively thin gold contact along the sidewall of the bottom DBR toward the flip-chip substrate. Focused ion beam (FIB) cross-sectioning revealed cracks in that thin metal contact, which were found to significantly impair the VCSEL thermal performance based on COMSOL simulations. Both the VCSEL yield and thermal performance were improved by implementing a new flip-chip bonding design. Instead of Au-Au thermocompression bonding, Au-In solid liquid interdiffusion (SLID) bonding was performed at a much lower temperature and pressure, which greatly improved the VCSEL yield. Furthermore, Au-In SLID bonding significantly improved the VCSEL thermal performance by incorporated a liquid phase during bonding so that the entire bottom DBR was embedded within metal. This led to the world’s first demonstration of CW operation for m-plane GaN VCSELs, and they were able to lase under CW operation for over 20 minutes of continuous testing.