” Only computer system simulations can do that,” stated Thompson.
Gordon Bell paper finalist is about a micron-sized hunk of compressed diamond
A technical paper describing the simulation was picked as a finalist for the Gordon Bell reward, sponsored each year by the Association of Computing Machinery. The diamond-specific modeling, which took just a day on the Summit supercomputer (the fastest in the U.S.) at Oak Ridge National Laboratory, was led by Prof. Ivan Oleynik at the University of South Florida. In addition to Sandia and USF, the collective group likewise included software designers at the Department of Energys National Energy Research Scientific Computing Center and the NVIDIA Corporation.
The groups simulations depend on SNAP, among the leading machine-learning descriptions of interatomic interactions, to model and solve an extremely important issue, said Thompson.
” We developed gigantic simulations of a micron-sized hunk of compressed diamond,” said Thompson. “To do this, we track the motion of billions of atoms by repeatedly determining the atomic forces over many, exceptionally small, intervals of time.”
Machine-learning bridged with quantum mechanical computations
SNAP utilized machine-learning and other information science methods to train a surrogate model that consistently replicated the correct atomic forces. These were determined utilizing high-accuracy quantum mechanical computations, which are just possible for systems including a few hundred atoms. The surrogate model was then scaled as much as anticipate forces and accelerations for systems including billions of atoms. All regional atomic structures that emerged in the large-scale simulations were well-represented in the small training data, a necessary condition for precision.
Another crucial part of the final result was efficiency optimization of the software to run efficiently on GPU-based supercomputers like Summit, said Thompson. “Since 2018, simply by improving the software application, we have been able to make the SNAP code over 30 times quicker, shortening the time for these kinds of simulations by 97%.
The chart shows the dramatic enhancement in computational speed attained by Sandia National Laboratories SNAP design from 2018 to 2021. Credit: Sandia National Laboratories
Run time shortened by 97 percent
” Since supercomputer time is costly and extremely competitive,” stated Thompson, “each shortening of SNAPs run time saves money and increases the usefulness of the model.”
Sandia scientists Stan Moore and Mitchell Wood made essential contributions to the SNAP design and the dramatic performance enhancements.
The optimized software for running SNAP on supercomputers is offered outdoors source distribution of Sandias LAMMPS molecular characteristics code. The Sandia FitSNAP software application for developing new SNAP designs is also publicly available.
The very first variation of SNAP was created in 2012 with assistance from Sandias Laboratory Directed Research and Development program. Software enhancement has been supported continuously because 2017 by the DOE Exascale Computing Project, a collective effort of the U.S. Department of Energy Office of Science and the National Nuclear Security Administration.
A Sandia National Laboratories supercomputer simulation model called SNAP that quickly anticipates the habits of billions of interacting atoms has captured the melting of diamond when compressed by severe pressures and temperature levels. Effective DOE centers like Sandias Z machine and Lawrence Livermore National Laboratorys National Ignition Facility can recreate near-identical conditions of these worlds in earthly experiments that provide close-up assessments of significantly compressed materials. In addition to Sandia and USF, the collective team also included software developers at the Department of Energys National Energy Research Scientific Computing Center and the NVIDIA Corporation.
SNAP utilized other and machine-learning information science strategies to train a surrogate model that faithfully replicated the proper atomic forces. “Since 2018, simply by enhancing the software, we have actually been able to make the SNAP code over 30 times quicker, shortening the time for these kinds of simulations by 97%.
This multi-billion atom simulation of shockwave propagation into initially uncompressed diamond (blue) utilizes a high-accuracy SNAP model from Sandia National Laboratories to anticipate that the final state (orange) is formed by recrystallization of amorphous cracks (red) that take shape in the light blue, green and yellow compressed material. Credit: Image with colors added, courtesy Sandia National Laboratories
Hardware and software improvements reduce run time from year to a day.
A Sandia National Laboratories supercomputer simulation model called SNAP that rapidly predicts the habits of billions of communicating atoms has actually caught the melting of diamond when compressed by severe pressures and temperature levels. At several million environments, the stiff carbon lattice of the hardest known substance in the world is shown in SNAP (Spectral Neighbor Analysis Potential) simulations to break, melt into amorphous carbon and then recrystallize. The work might help understanding of the internal structure of carbon-based exoplanets and have important implications for nuclear fusion efforts that utilize pills made from polycrystalline diamond.
Designing unique products and implications for huge planets
” We can now study the action of numerous products under the exact same severe pressures,” said Sandia researcher Aidan Thompson, who stemmed SNAP. “Applications include planetary science concerns– for instance, what type of impact tension would have resulted in the development of our moon. It also opens the door to design and manufacture of unique products at extreme conditions.”
The impact of severe pressures and temperature levels on products likewise is very important for creating interior designs of giant planets. Powerful DOE facilities like Sandias Z maker and Lawrence Livermore National Laboratorys National Ignition Facility can recreate near-identical conditions of these worlds in earthly experiments that use close-up evaluations of significantly compressed materials. Even these distinctively powerful makers can not identify essential tiny mechanisms of change under these extreme conditions, due to limitations in diagnostics at the level of atoms.