Neural network force field computational algorithms for molecular dynamics computer simulations

Patent metadata

Field	Value
Publication number	US-11455439-B2
Application number	US-201816202890-A
Country	US
Kind code	B2
Filing date	Nov 28, 2018
Priority date	Nov 28, 2018
Publication date	Sep 27, 2022
Grant date	Sep 27, 2022

Abstract

A computational method for simulating the motion of elements within a multi-element system using a neural network force field (NNFF). The method includes receiving a combination of a number of rotationally-invariant features and a number of rotationally-covariant features of a local environment of the multi-element system; and predicting a force vector for each element within the multi-element system based on the combination of the number of rotationally-invariant features, the number of rotationally-covariant features, and the NNFF, to obtain a simulated motion of the elements within the multi-element system.

First claim

What is claimed is:

1. A computational method for simulating the motion of elements within a multi-element system using a neural network force field (NNFF), the method comprising:
receiving a combination of a number of rotationally-invariant features and a number of rotationally-covariant features of a local environment of the multi-element system;
predicting a force vector for each element within the multi-element system based on the combination of the number of rotationally-invariant features and the number of rotationally-covariant features, and the NNFF, to obtain a simulated motion of the elements within the multi-element system; and
training a first neural network (NN1) based on the number of rotationally-invariant features to predict a number of rotationally-invariant intermediate states.

2. The computational method of claim 1 , further comprising training a second neural network (NN2) based on the number of rotationally-invariant intermediate states and the number of rotationally-covariant features to predict a number of rotationally-covariant outputs.

3. The computational method of claim 2 , further comprising concentrating NN1 and NN2 into a third neural network (NN3) used to obtain the simulated motion of the elements within the multi-element system.

4. The computational method of claim 3 , wherein the multi-element system is a material system, the elements are atoms within the material system, and the local environment is an atomic local environment.

5. The computational method of claim 4 , wherein the number of rotationally-covariant features include a set of internal axes {{right arrow over (A ι )}}.

6. The computational method of claim 5 , wherein the number of rotationally-invariant intermediate states includes a scalar projection of an atomic force experienced by a center atom i ({right arrow over (F ι )}) to the set of internal axes {{right arrow over (A ι )}}.

7. The computational method of claim 6 , wherein the number of rotationally-covariant output includes a set of Cartesian components of the atomic force experiences by the center atom i ({right arrow over (F ι )}).

8. The computational method of claim 1 , wherein the number of rotationally-invariant features include one or more Behler-Parrinello type fingerprint feature vectors.

9. The computational method of claim 1 , wherein the multi-element system is a portion of: a fuel cell, a water desalination system, a catalysis system, a coating system, or a battery.

10. A non-transitory computer-readable medium tangibly embodying computer readable instructions for a software program, the software program being executable by a processor of a computing device to provide operations comprising:
receiving a combination of a number of rotationally-invariant features and a number of rotationally-covariant features of a local environment of a multi-element system having elements;
predicting a force vector for each element within the multi-element system based on the combination of the number of rotationally-invariant features and the number of rotationally-covariant features, and a neural network force field (NNFF), to obtain a simulated motion of the elements within the multi-element system; and
training a first neural network (NN1) based on the number of rotationally-invariant features to predict a number of rotationally-invariant intermediate states.

11. The non-transitory computer-readable medium of claim 10 , wherein the software program is executable by the processor of the computing device to provide the further operation of training a second neural network (NN2) based on the number of rotationally-invariant intermediate states and the number of rotationally-covariant features to predict a number of rotationally-covariant outputs.

12. The non-transitory computer-readable medium of claim 11 , wherein the software program is executable by the processor of the computing device to provide the further operation of concentrating NN1 and NN2 into a third neural network (NN3) used to obtain the simulated motion of the elements within the multi-element system.

13. The non-transitory computer-readable medium of claim 10 , wherein the number of rotationally-invariant features include one or more Behler-Parrinello type fingerprint feature vectors.

14. A computer system for simulating the motion of elements within a multi-element system using a neural network force field (NNFF) including a simulation computer having a processor for executing computer-readable instructions and a memory for maintaining the computer-executable instructions, the computer-executable instructions when executed by the processor perform the following functions:
receiving a combination of a number of rotationally-invariant features and a number of rotationally-covariant features of a local environment of the multi-element system having elements;
predicting a force vector for each element within the multi-element system based on the combination of the number of rotationally-invariant features and the number of rotationally-covariant features, and the NNFF, to obtain a simulated motion of the elements within the multi-element system; and
training a first neural network (NN1) based on the number of rotationally-invariant features to predict a number of rotationally-invariant intermediate states.

15. The computer system of claim 14 , wherein the computer-executable instructions when executed by the processor performs the further function of training a second neural network (NN2) based on the number of rotationally-invariant intermediate states and the number of rotationally-covariant features to predict a number of rotationally-covariant outputs.

16. The computer system of claim 15 , wherein the computer-executable instructions when executed by the processor performs the further function of concentrating NN1 and NN2 into a third neural network (NN3) used to obtain the simulated motion of the elements within the multi-element system.

17. The computer system of claim 16 , wherein the multi-element system is a material system, the elements are atoms within the material system, and the local environment is an atomic local environment.

18. The computational method of claim 17 , wherein the number of rotationally-covariant features include a set of internal axes {{right arrow over (A ι )}}.

19. The computational method of claim 18 , wherein the number of rotationally-invariant intermediate states includes a scalar projection of an atomic force experienced by a center atom i ({right arrow over (F ι )}) to the set of internal axes {{right arrow over (A ι )}}.

20. The computer system of claim 14 , wherein the number of rotationally-invariant features include one or more Behler-Parrinello type fingerprint feature vectors.

Assignees

• Bosch Gmbh Robert

Inventors

• Mailoa Jonathan
• Kornbluth Mordechai
• Samsonidze Georgy
• Kozinsky Boris
• Craig Nathan

Classifications

• G06N3/045 Primary

  Combinations of networks · CPC title

• G06N3/096

  Transfer learning · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/08

  Learning methods · CPC title