Design And Process For Three Dimensional Heterogeneous Integration

Since the invention of the integrated circuit (IC) in the late 1950s, the semiconductor industry has experienced dramatic growth driven by both technology and manufacturing improvements. Over the past 40 years, the industry's growth trend has been predicted by Moore's law, and driven by the constant electrical field scaling design methodology. While the intrinsic performance of each device improves over generations, the corresponding interconnects do not. To alleviate this interconnect issue, a three-dimensional (3D) integration concept of transforming longer side to side interconnects into shorter vertical vias by using multiple active layers has attracted much attention. The focus of this thesis is on providing the foundation for 3D heterogeneous integration by investigating methods of growing single crystal materials on the silicon platform and the subsequent low-temperature process flow, through experimental demonstration, theoretical modeling and device structure simplification. First, thin film single crystal GaAs and GaSb were grown on dielectric layers on bulk silicon substrates by the rapid melt growth (RMG) method, using both rapid thermal annealing (RTA) and laser annealing. The relationship between stoichiometry and the crystal structure is discussed according to the theoretical phase diagram and the experimental results. A modified RMG structure is also proposed and demonstrated to solve the potential issue involved in integrating the RMG method into a three-dimensional integrated circuits (3D-IC) process with thick isolation layers. In order to estimate the outcome of the crystallization and to provide further understanding of the physics behind this RMG process, compact models are derived based on classical crystallization theory. Mathematical models including the geometry, the thermal environment and the outcome of the crystallization are built. The initial cooling rate is identified as the key factor for the RMG process. With the ability of integrating multiple materials on silicon substrates, the subsequent process flows using low-temperature-fabrication or simplified device structures are proposed and evaluated to achieve high density 3D integration. A "bonding substrate/monolithic contact" approach is proposed to relieve the thermal constraint from getting the starting single crystal layer without sacrificing the interconnect performance. A low-temperature process using germanium as the channel material is also discussed. Finally, gated thin film resistor structures are designed and compared to the conventional MOSFET structure with a focus on their relative performance and process complexity trade-off for future 3D-IC implementation.