Researchers from Lawrence Berkeley National Laboratory, including Phillip S. Thomas, Minh Nguyen, Dimitri Bazile, Tucker Allen, Barry Y. Li, Wenfei Li, Daniel Neuhauser, Mauro Del Ben, and Jack Deslippe, have developed a new computational tool that could significantly impact the energy industry’s approach to material science and energy storage.
The team has introduced an updated version of their code, StochasticGW-GPU, designed to calculate accurate Quasi-Particle (QP) energies for molecules and material systems using the GW approximation. The GW approximation is a computational quantum mechanics method used to predict electronic properties of materials. The new implementation leverages the power of Graphics Processing Units (GPUs) to substantially improve performance over previous versions.
StochasticGW-GPU utilizes a technique called stochastic Resolution of the Identity (sROI) to enable massively parallel processing. This approach reduces the computational costs, allowing the method to handle systems with tens of thousands of electrons. The researchers demonstrated the capabilities of their new code by computing the band gaps of hydrogenated silicon clusters containing up to 10,001 atoms and 35,144 electrons. They achieved a statistical precision of better than ±0.03 eV with computation times on the order of minutes.
The practical applications for the energy sector are significant. Accurate computation of electronic properties can aid in the design and development of new materials for energy storage, such as batteries and supercapacitors. It can also enhance the understanding of semiconductor materials used in solar cells and other renewable energy technologies. The ability to rapidly and accurately compute QP energies for large molecular systems can accelerate the discovery and optimization of materials critical to advancing energy technologies.
This research was published in the journal npj Computational Materials.
This article is based on research available at arXiv.