ADVANCED MATERIALS & PROCESSES | SEPTEMBER 2024 5 MACHINE LEARNING | AI NEW STANDARD COVERS AI FOR MATERIALS A prerequisite for artificial intelligence (AI) in materials research is largescale use and exchange of data on materials, which is facilitated by a broad international standard. A major international collaboration has developed an extended version of the OPTIMADE standard. Many demanding simulations are now performed on supercomputers that describe how electrons move in materials, giving rise to different properties. These advanced calculations yield large amounts of data that can be used to train machine learning models. These AI models can then predict responses to new calculations that have not yet been made, and by extension predict the properties of new materials. However, huge amounts of data are required to train the models. “We’re moving into an era where we want to train models on all data that exist,” says researcher Rickard Armiento of Linköping University, Sweden. Data from large-scale simulations, as well as general data about materials, are collected in large data- bases. Many such databases have emerged over time from various research groups—all working differently and using properties defined in numerous ways. The OPTIMADE (open data- bases integration for materials design) standard has been developed over the past eight years. Behind the standard is a large international network with over 30 institutions worldwide and large materials databases in Europe and the U.S. The goal is to give users easier access to both leading and lesser-known materials data- bases. A new version of the standard (v1.2) is now being released and is described in an article published in Digital Discovery. One of the biggest changes in the new version is a greatly enhanced ability to accurately describe different materials properties and other data using common definitions. The collaboration spans the EU, U.K., U.S., Mexico, Japan, and China. www.liu.se. ALGORITHM DESIGNS OPTICAL FILMS OptoGPT, developed by engineers at the University of Michigan, employs the computer architecture underpinning ChatGPT to work backward from desired optical properties to the material structure that can provide them. The new algorithm designs optical multilayer film structures that can serve a variety of purposes. Well-designed multi- layer structures can maximize light absorption in a solar cell, optimize reflection in a telescope, and improve semiconductor manufacturing with extreme UV light. From left, Oskar Andersson and Rickard Armiento work on supercomputers at Linköping University to simulate how atoms in different materials behave. OptoGPT produces designs for multilayer film structures within 0.1 seconds. In addition, the algorithm’s designs contain six fewer layers on average compared to previous models, making its results easier to manufacture. To automate the design process for optical structures, the research team tailored a transformer architecture— the framework used in large language models like OpenAI’s ChatGPT—for their own purposes. The model treats materials at a certain thickness as words, also encoding their associated optical properties as inputs. Seeking out correlations between these “words,” the model predicts the next word to create a “phrase”—in this case a design for an optical multilayer film structure—that achieves the desired property such as high reflection. Researchers tested the new model’s performance using a validation dataset containing 1000 known design structures including their material composition, thickness, and optical properties. When comparing OptoGPT’s designs to the validation set, the difference between the two was only 2.58%, lower than the closest optical properties in the training dataset at 2.96%. umich.edu. OptoGPT’s process combines an undetermined layer’s possible materials and thicknesses into a format that can be run through the program to choose the best possible combination. Courtesy of L. Jay Guo Laboratory.
RkJQdWJsaXNoZXIy MjMzNTA5MA==