Title: Machine Learning Takes Material Modeling to New Heights
Researchers from the Center for Advanced Systems Understanding (CASUS) at the Helmholtz-Zentrum Dresden-Rossendorf (HZDR) in Germany and Sandia National Laboratories in the United States have made a groundbreaking advancement in the field of material modeling. Their new machine learning-based simulation method, known as Materials Learning Algorithms (MALA), has the potential to revolutionize applied research by providing unprecedented scalability and accuracy in electronic structure prediction.
The arrangement of electrons in matter, referred to as electronic structure, holds immense importance in various scientific disciplines such as drug design and energy storage. With the aim of overcoming the limitations of existing simulation techniques, which struggle to offer both high fidelity and scalability across different time and length scales, the researchers embarked on developing MALA.
This innovative software stack integrates machine learning with physics algorithms to accurately predict electronic structures. By employing a hybrid approach, MALA combines deep learning – a powerful machine learning method – to predict local quantities, and physics algorithms to compute global quantities of interest.
A key feature of MALA is its ability to handle systems of all sizes. Unlike conventional methods that suffer from computational limitations as the system size increases, MALA’s machine learning model remains independent of system size, making it suitable for simulations involving a vast number of atoms. The researchers demonstrated MALA’s capabilities by achieving a remarkable speedup of over 1,000 times for smaller system sizes. Moreover, MALA successfully conducted electronic structure calculations on large-scale systems containing over 100,000 atoms, using minimal computational effort.
Attila Cangi, the Acting Department Head of Matter under Extreme Conditions at CASUS, believes that this breakthrough will reshape electronic structure calculations and open up new possibilities for researchers. With the ability to simulate significantly larger systems at an unprecedented speed, MALA has the potential to tackle a broad range of societal challenges. This includes developing new vaccines, exploring novel materials for energy storage, conducting large-scale simulations of semiconductor devices, studying material defects, and investigating chemical reactions for converting carbon dioxide into climate-friendly minerals.
Furthermore, MALA is highly compatible with high-performance computing (HPC) systems. As the system size increases, MALA can effectively leverage HPC resources, particularly graphical processing units, through independent processing on the computational grid it utilizes. This allows for unparalleled speed and efficiency in electronic structure calculations.
MALA is already making waves in various institutions and companies, such as the Georgia Institute of Technology, the North Carolina A&T State University, Sambanova Systems Inc., and Nvidia Corp. With its ability to predict electronic structures at any length scale, MALA is set to transform the field of material modeling and accelerate progress in scientific research.
This remarkable achievement serves as a testament to the power of machine learning in reimagining material modeling. By combining the strengths of both physics-based approaches and machine learning techniques, researchers have unlocked new frontiers in our understanding of electronic structures. The era of the Electronic Renaissance has begun, bringing us closer to unlocking the full potential of materials and driving innovation in a multitude of industries.
