威尼斯金沙欢乐娱人城

Skip to main content
Cornell University
We gratefully acknowledge support from
the Simons Foundation and member institutions.
arXiv.org > physics > arXiv:2005.00223

Help | Advanced Search

arXiv
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Physics > Computational Physics

arXiv:2005.00223 (physics)
[Submitted on 1 May 2020 (v1), last revised 14 Sep 2020 (this version, v3)]

Title:Pushing the limit of molecular dynamics with ab initio accuracy to 100 million atoms with machine learning

Authors:Weile Jia, Han Wang, Mohan Chen, Denghui Lu, Lin Lin, Roberto Car, Weinan E, Linfeng Zhang
Download PDF
Abstract: For 35 years, {\it ab initio} molecular dynamics (AIMD) has been the method of choice for modeling complex atomistic phenomena from first principles. However, most AIMD applications are limited by computational cost to systems with thousands of atoms at most. We report that a machine learning-based simulation protocol (Deep Potential Molecular Dynamics), while retaining {\it ab initio} accuracy, can simulate more than 1 nanosecond-long trajectory of over 100 million atoms per day, using a highly optimized code (GPU DeePMD-kit) on the Summit supercomputer. Our code can efficiently scale up to the entire Summit supercomputer, attaining $91$ PFLOPS in double precision ($45.5\%$ of the peak) and {$162$/$275$ PFLOPS in mixed-single/half precision}. The great accomplishment of this work is that it opens the door to simulating unprecedented size and time scales with {\it ab initio} accuracy. It also poses new challenges to the next-generation supercomputer for a better integration of machine learning and physical modeling.
Subjects: Computational Physics (physics.comp-ph)
Cite as: arXiv:2005.00223 [physics.comp-ph]
  (or arXiv:2005.00223v3 [physics.comp-ph] for this version)

Submission history

From: Weile Jia [view email]
[v1] Fri, 1 May 2020 04:46:52 UTC (8,946 KB)
[v2] Tue, 4 Aug 2020 21:18:32 UTC (24,070 KB)
[v3] Mon, 14 Sep 2020 05:00:56 UTC (25,843 KB)
Full-text links:

Download:

  • PDF
  • Other formats
(license)
Current browse context:
physics.comp-ph
< prev   |   next >
new | recent | 2005
Change to browse by:
physics

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar
a export bibtex citation Loading...

Bibtex formatted citation

×
Data provided by:

Bookmark

BibSonomy logo Mendeley logo Reddit logo ScienceWISE logo

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)

Code and Data Associated with this Article

arXiv Links to Code & Data (What is Links to Code & Data?)

Recommenders and Search Tools

Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs and how to get involved.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)
  • About
  • Help
  • 中国·金沙(威尼斯)欢乐娱人城 - Royal VIP ClubClick here to contact arXiv Contact
  • 中国·金沙(威尼斯)欢乐娱人城 - Royal VIP ClubClick here to subscribe Subscribe
  • Copyright
  • Privacy Policy
  • Web Accessibility Assistance
  • arXiv Operational Status
    Get status notifications via email or slack