Collaborative Research: Novel Interface Driven Designs for Advanced Resistive Switching Schemes

Nontechnical Description Advanced memory technologies are vital for modern computing and future digital systems such as the Internet of Things and neuromorphic computing. One promising technology is the memristor, a device based on switching electrical resistance in metal-oxide thin films. Memristors can be used for data storage, in-memory computing, and brain-inspired computing. These capabilities are important for machine learning (ML) and artificial intelligence (AI). Despite recent progress, several challenges limit the performance of memristors. These include low device reliability, large device-to-device variation, limited control of resistive states, and a narrow range of tuning. This project investigates another approach. Recent studies have shown that vertically aligned nanocomposites (VANs) can form conductive filaments in a controlled way. This discovery creates new opportunities to achieve reliable, multi-state, and reconfigurable switching behavior. Purdue University and the University of Texas at Arlington (UTA) will conduct a joint research program to understand the mechanisms that govern multi-state switching and to identify the factors that control resistive-switching behavior. The team will focus on how interfacial strain, defect type, and defect distribution affect the performance of VANs for use in memristors. The project will also provide broad educational and workforce benefits. These efforts include: 1) working with industry partners to increase the impact of the research; 2) training students and early-career researchers in experimental, computational, and machine-learning methods; 3) introducing new concepts in materials science and engineering into courses at Purdue and UTA. 4) Sharing research results and machine-learning tools with the wider scientific and engineering community. Technical Description Resistive random-access memory (RRAM) has shown strong promises in multistate data storage, in-memory-computing, and neuromorphic computing that are critical for ML and AI. Commonly used metal/metal oxide/metal structures for RRAM have conduction filaments that are nucleated randomly, suffering from limited device reliability, severe device-to-device variation, and restricted conductance modulation range. The goal of this project is to design materials with engineered vertical interfaces in either vertically aligned nanocomposite or fine-grained columnar structures to guide the conduction channels to overcome these limitations. We propose to 1) design and grow various VAN systems with tailored interfacial structures and oxygen vacancy densities and couple phase-field (PF) modeling with experiments to explore the fundamental switching mechanisms in VAN-based RS devices; 2) develop ML algorithms to expediate the VAN design and prediction; 3) conduct masked VAN growth and multi-step VAN growth for reconfigurable RS devices. The novelty of the proposed study resides in exceptionally large vertical interfacial strain stabilized and filament density tailored by pillar density tuning in VANs that cannot be easily achieved in typical lateral RS films. Scientific impacts include: 1) the ideas of interface enabled multi-state switching in VANs can be adopted by others in the field of RS devices. 2) The close-loop scheme of high-throughput PF calculations, ML prediction, and experimental validation established for VAN-memristor designs can be used by the community for other complex nanocomposite designs and predictions. 3) the dynamic reconfigurable switching materials via VAN interface designs can be integrated with existing CMOS structures due to the direct deposition compatibility. The Technological Impacts include 1) the proposed innovative VAN approaches could lead to the demonstration of high-performance reliable multi-state and reconfigurable RS devices for neuromorphic computing; 2) these pre-defined VAN interfaces enable the formation of precise and non-random vertical RS partial conducting paths with high densities and thus could lead to low power devices by eliminating high voltage forming process; 3) precise oxygen vacancy control at VAN interfaces allows the high and low resistance states to be tuned and thus achieve low device-to-device variation and improved materials reliability. The project addresses the NSF priority areas in AI, Advanced Manufacturing, and Semiconductors. This project is jointly funded by the Electronic and Materials program in the Directorate for Mathematical and Physical Sciences and the Electronic, Photonic, Magnetic, and Quantum Devices program in the Directorate for Engineering. 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.