Multi-Instruments Autonomous Discovery of Phase Change Memory Materials for Nonvolatile Memory Devices

Autonomous experimentation enabled by artificial intelligence (AI) facilitates acceleration of materials discovery and design. Investigating materials of complex compositional and structural landscapes requires resource-intensive high-throughput experimentations. Conventionally, we study materials by analyzing data collected from multiple characterization and measurements at different times. However, analyzing these data streams separately cannot exploit shared trends to boost analysis, prediction, and decision making. Herein, a novel approach is proposed to explore structural and functional properties synchronously. The autonomous robot is able to explore phase maps and exploit property optimization simultaneously in the high dimensional space. Multi-instruments run in parallel. The central AI exploits shared trends from the diverse measurement data sources and then employs a multi-objective active learning process to make decisions on the next iteration. The goal is accelerate discovery of optimal materials by learning and utilizing knowledge of composition-structure-property relationships. With this approach, we successfully discover novel materials in the titanium-selenium-tellurium system (TST), which is considered a promising phase change memory (PCM) candidate used in random-access memory. Combining x-ray diffraction and four point probe measurements, structural and functional property maps are established in terms of combinatorial wafer composed of hundreds of various compositions. The algorithm combines active phase mapping of a live x-ray diffraction data stream and multi-objective optimization of functional properties from additional live data streams. We identify a set of materials with identical structural signatures as well as two important features required for nonvolatile memory—large contrast of resistance between amorphous and crystalline states and large resistance in crystalline states. This novel approach again confirms that AI accelerates materials discovery and design.