UC Santa Barbara

In the field of semiconductor devices, the III-nitride material system, which is mainly made up of Indium Nitride (InN), Gallium Nitride (GaN) and Aluminum Nitride (AlN), has seen a great deal of attention over the past decade. Despite the maturity of this field of research, the growth mechanics and physics that govern the behavior of these devices is still poorly understood.

For all the devices mentioned, there exist regions called heterojunctions, which can be defined as the interface between two materials of different band gaps. In the case of LEDs, these heterojunctions are typically at the interface of alloy regions, which are the Quantum Well (QWs) layer and the Electron-Blocking Layer (EBL). These regions have compositional fluctuations due to the random distribution of the atomic constituent in the alloy. This phenomenon, known as alloy disorder, has largely been insignificant in the studies of other III-V semiconductors. However, due to the higher effective mass and larger band gaps of the nitrides, disorder plays a significant role in understanding the carrier transport behavior within nitride devices.

My research involves examining each of these layers and studying the hole transport behavior within these two types of heterostructures to better understand their electrical behavior. However, typical studies use LEDs as test structures, which are bipolar devices and are subject to recombination mechanisms. By using unipolar heterostructures, we can focus solely on the carrier transport within these structures without recombination complicating the analysis of the system, making them ideal test vehicles for theoretical models. My study involves simulating a three-dimensional unipolar p-type heterostructure that incorporates the fluctuations of the alloy composition within the alloy region. This would normally require solving for the wavefunctions of the system via Schrödinger’s equation. However, solving this equation in 3D is a computationally expensive task and could take months to obtain results. By using a mathematical theory called the Localization Landscape theory, we can simplify Schrödinger’s equation and converge to a solution three orders of magnitude faster than current simulation techniques. This allows us to viably give LEDs the full 3-dimensional treatment and obtain band structure information as well as current-voltage characteristics. These simulations were then compared to experimental realizations of these structures, which were grown by ammonia-assisted molecular beam epitaxy (NH3-MBE) and fabricated into devices for electrical measurements. The simulation results are verified by experiments using unipolar vertical hole transport structures enabled by n-to-p tunnel junctions (TJs) grown by ammonia molecular-beam epitaxy (NH3-MBE). The experimental results show that even a thin UID AlxGa1-xN (x = 14%, 13 nm) introduces an asymmetric barrier to the hole transport; A nearly 100% increase in drop voltages induced by a thin UID AlGaN at 50 A/cm2 in reverse direction is compared to only 25% corresponding increase in the forward direction. Furthermore, p-type doping of the AlGaN layer results in a drastic drop in the potential barrier to hole transport in both directions. Following that, a similar study was conducted for InGaN double heteostructures and quantum wells. The results indicate that increasing the UID In¬0.1Ga0.9N layer thickness from 15 nm to 30 nm increases the forward bias voltage drop (~2 V at 500 A/cm2) more than the reverse bias voltage drop (~ 0.2 V at 500A/cm2). For the QW structures, increasing the number of QWs from 1 to 3 increases the voltage penalty similarly in forward and reverse directions (~ 0.25 V per QW at 500 A/cm2).

Since the demonstration of III-nitride based transistors and diodes, their progress has also been limited by different challenges, one of which includes the presence of extended defects such as high densities of threading dislocations in the material grown on lattice-mismatched foreign substrates, which results in the degradation of device performance. This is demonstrated for GaN p-n diodes, in which threading dislocations behave as leakage pathways under both forward and reverse biases. In addition to this, high-voltage power switches require thick drift regions (on the order of 10 μm) with low background doping levels (on the order of 1015 – 1016 cm-3) to realize high blocking voltages. Hence, there is also a need for growth methods and optimized conditions to enhance the growth rate, maintaining the low background doping.

We conducted a study to investigate the effects of threading dislocation density on the transport properties of vertical GaN p-n junctions. Vertical GaN p-n diodes grown by NH3-MBE have been shown to be severely affected by threading dislocation density in terms of leakage currents. To study the effect of threading dislocations, finite element simulations were conducted to compare GaN p-n diodes with and without a dislocation. At zero bias, the depletion width and the maximum electric field were significantly reduced near the dislocation line. The reduction in the diffusion barrier for electrons and holes was asymmetric due to the asymmetric nature of the dislocation induced band bending related to the doping and trap parameters. This reduction in diffusion barrier facilitated the diffusion of electrons and holes in forward bias. Finally, this diffusion barrier reduction carrier resulted in an additional leakage mechanism via Shockley-Read-Hall non-radiative recombination mediated by a high np-product and trap state density near the intersection of the dislocation with the junction. In the reverse bias case, it was found that the defects coalesced by the dislocation strain field will mediate electron-hole pair generation by a trap-assisted tunneling mechanism occurring at a peak electric field in the junction near the dislocation. These electron-hole pairs are then swept away from the junction by the strong, reverse bias electric field thereby resulting in a reverse bias leakage current mediated by the dislocation trap states.

To improve the surface morphology in epitaxial growth, surfactants are commonly employed. For the case of group III nitrides, indium has been shown to be a highly effective surfactant. Typically, surfactants alter the surface morphology by modifying the surface energy and/or the adatom mobility. I investigated the effects of indium as a surfactant and other growth conditions on the surface morphology during NH3-MBE growth of unintentionally-doped (UID) GaN under fast growth rates (1 µm/hour). The surface morphology was characterized using atomic probe tomography and the impurity concentration within the UID GaN was obtained using secondary ion mass spectroscopy (SIMS). It was found that, through a series of optimizations, indium was able to improve the surface morphology during high flux growth. Moreover, it was also observed that indium suppresses the background Si impurity concentration in the film. The improvements in the surface morphology while maintaining low background impurity levels for fast growth rates would provide a path toward high-quality thick drift regions growths with smooth morphologies for regrowth-free high-voltage vertical devices for power switching applications.