Abstract
The advent of 5G communication and Internet of Things applications is driving the demand for faster and more efficient computers and communication devices. The stalling of Moore’s law due to CMOS scalability issues is also propelling research into novel device architectures and computing approaches. Semiconductor devices involved in computing and sensing technologies are commonly based on p-i-n junctions. p-i-n Tunnel Field Effect Transistors (TFETs) are one of the major candidates for sub-thermal operation of transistors, while p-i-n Avalanche Photodiodes (APDs) are widely deployed in applications like single photon detection, fiber optic communication and LIDAR systems due to their high gain-bandwidth product and low noise. An ideal III-V planar TFET operates on ballistic transport, but their performance is compromised by higher-order scattering processes. The desired transport mechanism in TFETs is band to band tunneling, with Auger generation (impact ionization) being undesirable. This research initially investigates the effects of the TFET non-idealities. The later chapters explore and establish the origins of observed low excess noise in p-i-n III-V digital alloy APDs. Design principles are then proposed based on these observations. In these APDs, impact ionization and unipolar transport are the key mechanisms for attaining gain, while tunneling promotes bipolarity and is detrimental to their performance. Since the materials and structures of the two devices are similar, it is possible to extend the simulation tools developed for TFETs to APDs.
To investigate the effect of TFET non-idealities, a chemistry-based analytical model is clearly needed that has a proper tunneling equation which accounts for multiple transverse modes, and includes accurate material chemistry, electrostatics and temperature effects. Presently, there is no such good model that captures all these effects. A quasi-analytical model for planar III-V TFETs is developed to study the effect of these higher-order processes, It is then demonstrated using the model that the observed discrepancy between experimental and theoretical TFET devices is due to the presence of non-ideal processes. Furthermore, the model shows that the minimum subthreshold swing in planar TFETs is limited by the Auger generation process.
Currently, the underlying physics of low noise III-V APDs are not well understood. A solid understanding of these physics will enable us to design better-performing APDs. Thus, the material bandstructures of the III-V digital alloys are studied to investigate the origin of the low excess noise, utilizing an Environment-Dependent Tight Binding Model coupled with a band unfolding technique. It is shown that a combination of ”minigaps”, increased effective mass, and large separation between light- hole and split-off bands lead to reduced excess noise. Thereafter, both quantum kinetic Non-Equilibrium Green’s Function and Boltzmann transport formalisms are used to show that these properties prevent hole ionization in many of these digital alloys. The resulting unipolarity creates low excess noise in the electron injected APDs. Based on these simulations, this thesis proposes some empirical design criteria for attaining low excess noise using digital alloys. Also, this work attributes the low noise in quaternary random alloys to large effective masses and separation between light-hole and split-off bands using bandstructure studies. Furthermore, it is demonstrated that strain generates opposite movement of bands in the alloy binary constituents which lead to the formation of the minigaps. It is then possible to control the minigap sizes by modulation of the strain, which can possibly lead to lower excess noise in APDs.
Optoelectronic devices, like APDs, are increasingly being incorporated into photonic integrated circuits. To accurately model the digital alloy APDs in these circuits, a physics-based multiscale compact model for these APDs is developed. The model provides a framework to study the digital alloys starting from material properties to device transport and eventually their performance in circuits. Lastly, in this research work, a simple one-dimensional NonEquilibrium Green’s Function model incorporating impact ionization is developed, forming a matrix-based theoretical quantum transport framework for studying APDs. This method automatically captures quantum effects like tunneling which are not accounted for by semiclassical tools and can incorporate the effects of complicated bandstructures that generate non-parabolic, energy and voltage-dependent effective mass tensors, as seen in digital alloy superlattices. This framework will allow the development of quantum transport based accurate APD models that are more advanced than state-of-the-art semi-classical approach based model.
