SBIR Phase I: Developing Energy-Efficient 3D Memory Using Advanced Indium Gallium Zinc Oxide Transistors for Next-Generation AI Chips

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project lies in addressing the urgent need for energy-efficient hardware to support the rapid growth of artificial intelligence (AI). As AI applications expand to devices like smartphones, wearables, and autonomous systems, the energy inefficiency of current hardware limits deployment. This project introduces a new type of ultra-dense memory, called 3D Gain-Cell Random Access Memory (GCRAM), built using a novel vertical transistor structure. By enabling computing and memory functions in the same location, this technology reduces energy losses and supports real-time, low-power AI processing. The innovation is projected to significantly exceed the energy efficiency of existing processors. This advancement may enable on-device AI that will improve privacy, responsiveness, and sustainability by reducing reliance on cloud infrastructure. The proposed technology has a clear commercial path through licensing to chip designers and foundries. By year three of production, the company aims to reach AI markets in mobile devices and robotics. This Small Business Innovation Research (SBIR) Phase I project focuses on developing a new three-transistor (3T) memory cell architecture using vertical indium gallium zinc oxide (IGZO) transistors fabricated via atomic layer deposition. This Back End of Line (BEOL)-compatible process enables high-density monolithic 3D integration, addressing the memory bottleneck in edge AI chips. The project will demonstrate the feasibility of this novel structure through the fabrication of a stacked IGZO vertical transistor, benchmarking key performance metrics such as mobility, threshold voltage, and subthreshold slope. A device model using simulations will be developed and validated against experimental measurements. A machine learning framework will be implemented to predict device performance based on process and material parameters and optimize fabrication conditions. This physics-informed AI model will identify process-awareness failure modes and suggest optimization strategies, speeding up design iterations and lowering development costs. The proposed work lays the foundation for Phase II efforts in reliability modeling, multi-layer device scaling, and integration with commercial AI systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.