Powerful simulation of the Sandia supercomputer shows the melting of diamonds at high pressure

This multibillion-dollar simulation of shock wave propagation into an initially uncompressed diamond (blue) uses a high-precision SNAP model from Sandia National Laboratories to predict that the final state (orange) is formed by recrystallizing amorphous cracks (red) that turn blue-green in color. and yellow compressed material. Credit: Image with added colors, provided by Sandia National Laboratories

Hardware and software improvements reduce “uptime” from year to day.

The simulation model of the Sandia National Laboratories supercomputer is called SNAP which quickly predicts the behavior of billions of interacting atoms, recorded the melting of diamond under compression by extreme pressure and temperature. In several million atmospheres the rigid carbon lattice of the most solid known substance on Earth is shown in the SNAP (Spectral Neighbor Analysis Potential) simulation for cracks, melting into amorphous carbon and then recrystallized. The work could help to understand the internal structure of carbon-based exoplanets and have important implications for fusion, which uses capsules with polycrystalline diamond.

Develop new materials and implications for the giant planets

“We can now study the reaction of many materials under the same extreme pressure,” said Sandia scientist Aidan Thompson, who created SNAP. “Applications include questions of planetary science – for example, what shock stress would lead to the formation of our Moon. It also opens the door for the design and production of new materials in extreme conditions. ”

The influence of extreme pressures and temperatures on materials is also important for the development of interior models of giant planets. Powerful Ministry of Health facilities such as the Z Sandia Machine and Lawrence Livermore National Laboratory’s National Ignition Unit can recreate almost identical conditions of these worlds in terrestrial experiments that offer close-ups of radically compressed materials. But even these uniquely powerful machines cannot pinpoint the key microscopic mechanisms of change in these extreme conditions due to limitations in diagnostics at the atomic level.

“Only computer simulations can do that,” Thompson said.

Gordon Bell paper finalist on “a piece of compressed diamond the size of a micron”

A technical document describing the simulation, was selected as a finalist for the Gordon Bell Award, which is sponsored annually by the Computer Science Association. The specific diamond modeling, which took just one day on the Summit supercomputer (the fastest in the US) at the Oak Ridge National Laboratory, was led by Professor Ivan Oleynik of the University of South Florida. In addition to Sandia and USF, the joint team also includes software developers from the National Research and Computing Center of the Department of Energy and NVIDIA Corporation.

The team simulators relied on SNAP, one of the leading descriptions of interatomic interactions through machine learning, to model and solve a very important problem, Thompson said.

“We’ve created a giant simulation of a micron piece of compressed diamond,” Thompson said. “To do this, we track the movement of billions of atoms, repeatedly calculating the atomic forces over very, very small, periods of time.”

Machine learning combined with quantum mechanical calculations

SNAP used machine learning and other data science techniques to teach a surrogate model that accurately reproduced the correct atomic forces. They were calculated using high[{” attribute=””>accuracy quantum mechanical calculations, which are only possible for systems containing a few hundred atoms.  The surrogate model was then scaled up to predict forces and accelerations for systems containing billions of atoms.  All local atomic structures that emerged in the large-scale simulations were well-represented in the small-scale training data, a necessary condition for accuracy.

Another critical part of the final result was performance optimization of the software to run efficiently on GPU-based supercomputers like Summit, said Thompson. “Since 2018, just by improving the software, we have been able to make the SNAP code over 30 times faster, shortening the time for these kinds of simulations by 97%. At the same time, each generation of hardware is more powerful than the last. As a result, calculations that might have until recently taken an entire year can now be run in a day on Summit.”

Computational Speed SNAP Model

The graph demonstrates the dramatic improvement in computational speed achieved by Sandia National Laboratories’ SNAP model from 2018 to 2021. Credit: Sandia National Laboratories

Run time shortened by 97 percent

“Since supercomputer time is expensive and highly competitive,” said Thompson, “each shortening of SNAP’s run time saves money and increases the usefulness of the model.”

Sandia researchers Stan Moore and Mitchell Wood made important contributions to the SNAP model and the dramatic performance improvements.

The optimized software for running SNAP on supercomputers is available in the open source distribution of Sandia’s LAMMPS molecular dynamics code.  The Sandia FitSNAP software for building new SNAP models is also publicly available.

The first version of SNAP was created in 2012 with support from Sandia’s Laboratory Directed Research and Development program. Software improvement has been supported continuously since 2017 by the DOE Exascale Computing Project, a collaborative effort of the U.S. Department of Energy Office of Science and the National Nuclear Security Administration. Powerful simulation of the Sandia supercomputer shows the melting of diamonds at high pressure

Back to top button