Machine learning has taken materials modeling to a new era with a deep learning approach that enables accurate electronic structure calculations at large scales. This breakthrough has been made possible by the Materials Learning Algorithms (MALA) software stack developed by researchers from the Center for Advanced Systems Understanding (CASUS) at the Helmholtz-Zentrum Dresden-Rossendorf (HZDR) in Germany and Sandia National Laboratories in the US.
The arrangement of electrons in matter, known as the electronic structure, plays a crucial role in fundamental and applied research, including drug design and energy storage. However, the lack of a simulation technique that combines high fidelity and scalability across different time and length scales has hindered progress in these areas.
Scientific challenges are increasingly being addressed through computational modeling and simulation, leveraging the power of high-performance computing. However, achieving realistic simulations with quantum precision has been a major obstacle due to the absence of a predictive modeling technique that offers both high accuracy and scalability. Classical atomistic simulation methods can handle large and complex systems but lack the inclusion of quantum electronic structure, limiting their applicability. On the other hand, first principles methods, which do provide high fidelity, are computationally demanding.
To overcome these limitations, the team of researchers introduced MALA, a novel simulation method that integrates machine learning with physics-based approaches. MALA uses an established machine learning method called deep learning to accurately predict the electronic structure of materials. It combines physics algorithms with machine learning models to compute both local and global quantities of interest.
MALA takes the arrangement of atoms as input and generates fingerprints known as bispectrum components, which encode the spatial arrangement of atoms around a Cartesian grid point. The machine learning model in MALA is trained to predict the electronic structure based on this atomic neighborhood. One of the key advantages of MALA is its ability to be independent of the system size, allowing it to be trained on small systems and deployed at any scale.
The researchers demonstrated the effectiveness of MALA by achieving a significant speedup of over 1,000 times for smaller system sizes compared to traditional algorithms. They also showed that MALA can accurately perform electronic structure calculations at a large scale, involving over 100,000 atoms, with modest computational effort. This outperforms conventional density functional theory (DFT) codes, which become impractical as the system size increases.
With the introduction of MALA, electronic structure calculations are expected to undergo a transformation, enabling researchers to simulate significantly larger systems at unprecedented speeds. This breakthrough opens up computational possibilities for addressing a wide range of societal challenges, such as developing vaccines, creating novel materials for energy storage, simulating semiconductor devices, studying material defects, and exploring chemical reactions for climate-friendly solutions.
MALA is particularly well-suited for high-performance computing (HPC) as it allows for independent processing on the computational grid it utilizes, effectively leveraging HPC resources, especially graphical processing units. This enables unparalleled speed and efficiency in electronic structure calculations.
Institutions and companies such as the Georgia Institute of Technology, the North Carolina A&T State University, Sambanova Systems Inc., and Nvidia Corp. are already using MALA to enhance their research and development efforts.
The introduction of MALA marks a significant milestone in the field of materials modeling and opens up new possibilities for advancements in various scientific disciplines. With its ability to combine high fidelity and scalability, MALA is set to revolutionize electronic structure calculations and accelerate progress in scientific research and technological innovations.
