Agriculture

DeepMind simulates matter on a nanoscale using artificial intelligence

  • In an article published in Science, DeepMind demonstrates how neural networks can improve density functional approximation (a method used to describe the interaction of electrons in chemical systems).
  • This illustrates the promise of deep learning in accurate modeling of matter at the quantum-mechanical level.
  • Along with the document, DeepMind will open source code to provide a research foundation on which others can build.

In an article published in a scientific journal science, DeepMind demonstrates how neural networks can be used to describe the interaction of electrons in chemical systems more accurately than existing methods.

Functional density theory, created in the 1960s, describes the mapping between electron density and interaction energy. For more than 50 years, the exact nature of the reflection between electron density and the interaction energy – the so-called density functional – has remained unknown. Significant progress in this area DeepMind has shown that neural networks can be used to create a more accurate map of density and interaction between electrons than was previously possible.

Expressing functionality as a neural network and incorporating accurate properties into training data, DeepMind was able to teach the model to assimilate functionalities without two important systematic errors – provocation errors and spin symmetry errors – resulting in a better description of a wide class of chemical reactions.

In the short term, this will give researchers the opportunity to improve the approximation of the exact density functionality for immediate use through the availability of our code. In the long run, this is another step that shows the prospect of deep learning in accurate modeling of matter at the quantum-mechanical level, which can allow materials to be developed in computers, allowing researchers to explore materials, drugs and catalysts at the nanoscale level.

“Understanding nanoscale technology is becoming increasingly important to help us solve some of the major challenges of the 21st century, from clean electricity to plastic pollution,” says James Kirkpatyk, a researcher at DeepMind. “This study is a step in the right direction so we can better understand the interactions between electrons, the glue that holds molecules together.”

In order to accelerate progress in this area DeepMind has released paper and open source code for free access.

Reference: “Breaking the Density of Density Electron Functionals by Solving Fractional Electron Problems” by James Kirkpatrick, Brendan McMorrow, David H. P. Turban, Alexander L. Gaunt, James S. Spencer, Alexander GDG Matthews, Annette Obika, Louis Fort, Thierry, May Pfau, Lara Roman Castellanas, Stig Petersen, Alexander UR Nelson, Puschmit Cole, Paula Mori-Sanchez, Demis Khasabis and Aaron J. Cohen, December 9, 2021, Science.
DOI: 10.1126 / science.abj6511

About DeepMind

DeepMind is a scientific discovery company that seeks to “solve intelligence to advance science and humanity.” Intelligence solutions require a diverse and interdisciplinary team working closely together – from scientists and designers, to engineers and ethicists – to begin developing advanced artificial intelligence.

The company’s breakthroughs include AlphaGo, AlphaFold, more than a thousand published scientific papers (including more than a dozen in Nature or Science), partnerships with research organizations and hundreds of contributions to Google products (everything from Android battery performance to text-to-speech).



https://scitechdaily.com/deepmind-simulates-matter-on-the-nanoscale-with-artificial-intelligence/ DeepMind simulates matter on a nanoscale using artificial intelligence

Back to top button