9 OREGON STATE ENGINEERING WINTER 2025 various nonlinear functions, materials with wider-range memory effect, and strategies for integration with existing silicon technology,” Chae said. BREAKTHROUGH IN PHOTONICS Another significant research initiative addresses energy usage in data centers and supercomputers through advancements in photonic chips. This collaboration with Baylor University researchers focuses on a new method that compensates for temperature variations that can degrade the performance of photonic chips, which utilize light particles instead of electrons for data transmission. John Conley, a professor of electrical and computer engineering whose research specializes in electronic materials and devices, worked with colleagues to develop prototypes that control gate voltage, potentially reducing the energy required for temperature control by a factor of one million. This breakthrough addresses a primary shortcoming of photonic chips, which previously required substantial energy to maintain optimal temperatures for performance. “Our method is much more acceptable for the planet. It will perform AI tasks more energy efficiently,” Chae said. This research, funded by the National Science Foundation, used memristors based on a new material system known as entropy-stabilized oxides, which allow for finely tuned memory capabilities by incorporating multiple elements. This optimization not only improves energy efficiency but also enables artificial neural networks to process time-dependent information, such as audio and video data. Chae’s findings were recently published in Nature Electronics, highlighting the significant potential for these advanced chips to reshape the landscape of AI technology. Chae also published a “Perspective” article touting the memristors’ potential for efficient data processing in machine learning tasks in the Aug. 12 issue of Applied Physics Letters. Chae’s research on memristors began while she was working toward her doctorate in materials science and engineering at the University of Michigan, which she completed in 2022. She joined the faculty at Oregon State in December 2023, where she will continue to create solutions for AI and other intensive computing applications. “For the future, I plan to develop new materials that can mimic one day allow data centers to keep getting faster and more powerful while using less energy, enabling applications driven by machine learning — such as ChatGPT — without feeling guilty,” Conley said. The implications of these projects extend far beyond the lab. Creating more efficient computer chips and improving thermal management not only lowers operational costs but also decreases the environmental footprint of these technologies. The success of these initiatives relies heavily on collaboration between academia, industry, and government. Oregon State University is actively seeking partnerships with tech companies and governmental agencies to translate research findings into practical applications. “Working together with industry leaders, we can accelerate the adoption of energy-efficient technologies and make a real impact on global energy consumption,” Conley said. OPPOSITE Sieun Chae (left) and graduate student Dipannita Ghosh. “Our method is much more acceptable for the planet. It will one day allow data centers to keep getting faster and more powerful while using less energy, enabling applications driven by machine learning — such as ChatGPT — without feeling guilty.” John Conley, professor of electrical and computer engineering
RkJQdWJsaXNoZXIy MTcxMjMwNg==