New AI model promises rapid discovery of advanced photovoltaic materials
By Mrigakshi Dixit,
5 hours ago
Artificial intelligence could make discovering new materials for advanced technologies like solar cells and quantum computers a simple process in the coming years.
Researchers from Tohoku University and the Massachusetts Institute of Technology (MIT) have developed a new AI model (named GNNOpt) for understanding materials and their properties.
Interestingly, it can predict the optical properties of materials with the same accuracy as quantum simulations, but at a speed that is a “million times faster.”
This is a significant development that could accelerate the development of new photovoltaic and quantum materials.
Predicting optical properties
The semiconductor industry’s recent growth is fueled by the development of optoelectronic devices, which require an extensive understanding of materials ‘ optical properties. These devices include LEDs, solar cells, photodetectors, and photonic integrated circuits.
Standard methods of calculating the optical properties of materials are time-consuming and computationally intensive. This new AI model, however, can predict these properties with incredible accuracy in a fraction of the time.
Notably, the AI model uses a material’s crystal structure as input and can predict its optical properties across a wide range of light frequencies.
“Optics is a fascinating aspect of condensed matter physics, governed by the causal relationship known as the Kramers-Krönig (KK) relation,” said Nguyen Tuan Hung, an assistant professor at the Frontier Institute for Interdisciplinary Science (FRIS), Tohoku University.
Nguyen added: “Once one optical property is known, all other optical properties can be derived using the KK relation. It is intriguing to observe how AI models can grasp physics concepts through this relation.”
Mode’s technique
The other typical methods of obtaining optical spectra — such as experiments using lasers — are limited by the available wavelengths.
Moreover, simulations are computationally expensive and require strict convergence criteria. This spurred a long-standing search for alternative and fast methods to predict the optical spectra of diverse materials.
“Machine-learning models utilized for optical prediction are called graph neural networks (GNNs). GNNs provide a natural representation of molecules and materials by representing atoms as graph nodes and interatomic bonds as graph edges,” said Ryotaro Okabe, a chemistry graduate student at MIT.
But there is a drawback: GNNs struggle to capture the complex structural nuances of crystals, limiting their universality in predicting material properties.
The key to this new model’s success is a technique called “ensemble embedding.” This method allows the AI to learn from multiple data representations, making it more accurate and versatile.
“This ensemble embedding goes beyond human intuition but is broadly applicable to improve prediction accuracy without affecting neural network structures,” said Abhijatmedhi Chotrattanapituk, an electrical engineering and computer science graduate student at MIT.
This method accurately predicts optical properties from crystal structures, opening doors to a wide range of applications, especially materials screening for advanced solar cells and quantum materials .
As per the press release, the researchers further plan to create comprehensive databases containing diverse material characteristics, such as mechanical and magnetic properties. This will help expand the AI model’s capabilities.
Get updates delivered to you daily. Free and customizable.
It’s essential to note our commitment to transparency:
Our Terms of Use acknowledge that our services may not always be error-free, and our Community Standards emphasize our discretion in enforcing policies. As a platform hosting over 100,000 pieces of content published daily, we cannot pre-vet content, but we strive to foster a dynamic environment for free expression and robust discourse through safety guardrails of human and AI moderation.
Comments / 0