2023 Gordon Bell Prize Announced: Frontier Supercomputer’s “Quantum Level Accuracy” Materials Simulation Wins

Editor | Zenan, Du Wei

The ACM Gordon Bell Prize was established in 1987 and awarded by the American Computer Society. It is known as the super prize. The “Nobel Prize” in computing. The award is given annually to recognize excellence in high-performance computing. The $10,000 prize, awarded to Gordon Bell, a pioneer in high-performance and parallel computing, was awarded to an eight-person international team of U.S. and Indian researchers at the recent global supercomputing conference SC23. Winner of the 2023 ACM Gordon Bell Prize for their successful implementation of large-scale quantum-precision materials simulations. The name of this project is "Quantum-accurate large-scale materials modeling: ab initio simulations of quasicrystals and interaction-propagated defects in metal alloys"

The team members are from the University of Michigan, Oak Ridge National Laboratory, and Indian Institute of Science (Bangalore).

Award-winning team member.

The 2021 Gordon Bell Award was previously awarded to a 14-member Chinese supercomputing application team, with members from Zhijiang Laboratory and the National Supercomputing Wuxi Center, Tsinghua University, and Shanghai Quantum Science Research Center, in recognition of the team Based on the application of my country's new generation Sunway supercomputer, "ultra-large-scale quantum random circuit real-time simulation". Going forward, China's supercomputing application team also won the Gordon Bell Award for two consecutive years in 2016 and 2017.

Research Overview

Rewritten content: We learned that molecular dynamics is a method of using computer simulations to better understand the processes by which atoms and molecules move within a system. A branch of molecular dynamics is "Ab initio", a technique that has proven to be very effective in solving important problems in physics and chemistry, such as better understanding of microscopic mechanisms, obtaining new insights in materials science, and Verification of experimental data, etc.

Paper address:

https://dl.acm.org/doi/pdf/10.1145/3581784.3627037

By The study, led by Vikram Gavini, professor of mechanical engineering and materials science and engineering at the University of Michigan, used first-principles simulations using the Schrödinger equation using Frontier (1.14 exaflop HPE Cray EX supercomputer) at the U.S. Department of Energy's Oak Ridge National Laboratory. , this equation describes microscopic systems, including their probabilistic properties. According to reports, the results can be used to help design candidate materials for new alloys and promote other computational design efforts such as drug discovery.
Gavini’s team used an integrated computing framework on the Frontier and Summit supercomputers to simulate dislocations, or defects, in a magnesium system composed of nearly 75,000 atoms. Magnesium alloys are promising candidates for lightweight alloys, but misaligned vacancies in the magnesium atomic structure can lead to brittleness and cracking. Understanding dislocations in magnesium alloys could provide industry with lighter, more flexible alloys

This article compares to previous work

The team is also The National Energy Research Scientific Computing Center's Perlmutter supercomputer was used to study the stability of quasicrystals (an ordered but non-periodic structure) in ytterbium-cadmium alloys.

These calculations rely on density functional theory, a quantum mechanical method for calculating the atomic and electronic structure of materials, and use machine learning to approach the high level of precision of quantum many-body calculations. They used Frontier’s 8,000 nodes, with a maximum computing power of 659.7 petaflops

“As we strive to achieve greater accuracy, the number of available computing systems has dropped dramatically, Gavini said. "We use the results of quantum many-body calculations on smaller systems and use machine learning to infer universal constitutive relations for electrons, which can be used in larger density functional theory calculations. Combining these methods, we are able to use tools like Frontier The advantage of such a large machine, while approaching quantum accuracy."

#This article aims to outline an approach to achieve this through large-scale material simulations with quantum precision

The Frontier team’s latest findings are a new milestone in their decade-long effort. Prior to this, a 2019 study used Summit to simulate more than 10,000 magnesium atoms, and was nominated for the Gordon Bell Award

The alloy production process involves the melting and mixing of metals. Defects may appear during the solidification process and have a positive or negative impact on the material's properties. The material's atomic structure plays a crucial role in the behavior of these line-like defects, often called dislocations. Malleable metals like aluminum benefit from atomic structure, allowing the metal to adapt to the dislocations and their movement. . Magnesium's atomic structure cannot easily accommodate dislocations, making it more brittle in nature.

Under the right circumstances, these flaws can create unprecedented features, Gavini said. "Why do these defects form? How can we exploit these defects to bring about desirable rather than undesirable properties? In previous research, we explored the energy of individual dislocations in bulk magnesium. In this study, we investigate The result is the most detailed image yet of this structure, with near-quantum precision. Gavini hopes to apply these methods to a wide range of studies.

"If we can perform these large-scale calculations with near-quantum precision, it means we can design better materials through computational design, explore compounds for drug discovery, and understand nanoparticles at a new level. and the details of the properties of the material system,” Gavini said. "Without exascale computing and Frontier, we would not be able to perform these types of calculations. Now that we know how to do it, we can broadly apply these methods to explore other problems."

According to the research team Introduction, this method can be widely used in many scientific fields and can solve some long-standing challenging problems in fields from aerospace to medicine.

Reference content:

https://awards.acm.org/bell

https://news.engin.umich.edu/2023/11/material-simulation-with-quantum-accuracy-wins-gordon-bell-prize/

https://www.hpcwire.com/off-the-wire/ornls-frontier-achieves-near-quantum-accuracy-in-alloy-simulation-contends-for-gordon-bell-prize/

The above is the detailed content of 2023 Gordon Bell Prize Announced: Frontier Supercomputer’s “Quantum Level Accuracy” Materials Simulation Wins. For more information, please follow other related articles on the PHP Chinese website!