Caroline Bishop
November 19, 2024 00:32
NVIDIA launched ALCHEMI to revolutionize AI-based materials discovery with the goal of accelerating R&D through machine learning interatomic potential and high-throughput simulations.
For a breakthrough in materials science, NVIDIA has unveiled its AI Lab for Chemistry and Materials Innovation, known as ALCHEMI, to accelerate the discovery of new materials through artificial intelligence. According to NVIDIA, this initiative will transform the traditional materials discovery process, which often takes decades, into a streamlined task that can be accomplished in just months.
AI-accelerated workflow
The AI-based workflow for materials discovery consists of four main steps: hypothesis generation, solution space definition, property prediction, and experimental validation. Each step is designed to maximize the efficiency and accuracy of new material discovery using AI.
During hypothesis generation, large language models (LLMs) trained on the chemical literature help scientists synthesize insights and formulate hypotheses. The solution space definition phase uses generative AI to explore new chemical structures, while property prediction uses machine learning interatomic potential (MLIP) and density functional theory (DFT) simulations to verify properties. Finally, the experimental validation phase leverages AI to recommend candidates for laboratory testing and optimize the balance between known chemistry and unexplored potential.
Innovative tools and technologies
NVIDIA’s ALCHEMI provides APIs and microservices that enable developers to deploy generative AI models and AI surrogate models. These tools are critical to efficiently map material properties and perform simulations, which are essential for high-throughput screening and innovation.
ALCHEMI introduces Machine Learning Interatomic Potentials (MLIP), which provides a cost-effective and accurate method for predicting material properties. This technology has a variety of applications across chemistry, materials science, and biology, enabling large-scale simulations previously impossible due to high computational costs.
Impact on Research and Development
NVIDIA Batch Geometry Relaxation The NVIDIA Inference Microservice (NIM) significantly accelerates the geometry relaxation process, showing speedups of 800x in some scenarios. This advancement increases the throughput of materials discovery by allowing numerous simulations to be processed simultaneously.
SES AI, a leader in lithium metal battery technology, is exploring the use of NVIDIA’s ALCHEMI NIM microservice to accelerate the identification of new electrolyte materials. By mapping 100,000 molecules in just half a day, SES AI demonstrates the transformative potential of AI-accelerated materials discovery.
future prospects
Going forward, NVIDIA aims to further enhance ALCHEMI’s capabilities to enable mapping of up to 10 billion molecules in the next few years. This ambitious goal highlights the potential for AI to make breakthroughs in materials science, fostering a more sustainable and innovative future.
For more information about NVIDIA’s ALCHEMI, visit the official NVIDIA blog.
Image source: Shutterstock