Method and device for simulating atomic dynamics

Patent metadata

Field	Value
Publication number	US-11386248-B2
Application number	US-201916695834-A
Country	US
Kind code	B2
Filing date	Nov 26, 2019
Priority date	Nov 29, 2018
Publication date	Jul 12, 2022
Grant date	Jul 12, 2022

Abstract

A method and a device for simulating atomic dynamics includes setting initial positions for multiple specific atoms in a specific scene; calculating, based on the initial positions, positions of the multiple specific atoms at each time in a first time series by utilizing a Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) configured with respect to the specific scene, as real positions; calculating, based on the initial positions, positions of the multiple specific atoms at the same time in the first time series by utilizing a generative adversarial network (GAN), as predicted positions; improving a configuration of the GAN based on the real positions and the predicted positions at a same time. Initial positions are settable for multiple atoms to be simulated in a scene; positions of the multiple atoms to be simulated are calculated at each time in a second time series in the scene by utilizing the improved GAN.

First claim

The invention claimed is:

1. A method for a computer to simulate atomic dynamics, comprising:
setting initial positions for a plurality of specific atoms in a specific scene;
calculating, based on the initial positions, positions of the plurality of specific atoms at each time in a first time series in the specific scene to represent real positions by utilizing a software tool known as LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) configured with respect to the specific scene;
calculating, based on the initial positions, positions of the plurality of specific atoms at the same time in the first time series in the specific scene to represent predicted positions by utilizing a generative adversarial network (GAN);
training a configuration of the GAN based on the real positions and the predicted positions at the same time;
setting initial positions for a plurality of atoms to be simulated in a scene;
calculating positions of the plurality of atoms simulated in the scene at each time in a second time series by utilizing the trained GAN;
determining movement information of the plurality of atoms simulated by analyzing the positions of the plurality of atoms simulated at each time in the second time series; and
determining properties of a substance made up of the plurality of atoms simulated based on the movement information.

2. The method according to claim 1 , further comprising:
calculating the predicted positions of the plurality of specific atoms at the same time in the first time series by utilizing a generative unit in the GAN;
determining a probability that a predicted position among the predicted positions is a real position among the real positions at the same time by utilizing a discriminative unit in the GAN; and
training the configuration of the generative unit and the discriminative unit based on the predicted position and the real position at the same time, until the probability determined by the discriminative unit is 50%.

3. The method according to claim 2 , further comprising:
training the discriminative unit in such a manner that the discriminative unit is capable of determining whether the predicted position is the real position at the same time.

4. The method according to claim 3 , further comprising:
training the generative unit in such a manner that the discriminative unit is not capable of determining whether the predicted position calculated by the generative unit is the real position at the same time.

5. The method according to claim 3 , further comprising:
training the generative unit in a manner of minimizing a mean square error between the real position and the predicted position at the same time.

6. The method according to claim 1 , wherein the generative unit and the discriminative unit are implemented with neural networks.

7. A device to simulate atomic dynamics, the device comprising:
a memory, and a processor configured to:
set initial positions for a plurality of specific atoms in a specific scene;
calculate, based on the initial positions, positions of the plurality of specific atoms at each time in a first time series in the specific scene to represent real positions by utilizing a software tool known as LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) configured with respect to the specific scene;
calculate, based on the initial positions, positions of the plurality of specific atoms at the same time in the first time series in the specific scene to represent predicted positions by utilizing a generative adversarial network (GAN);
train a configuration of the GAN based on the real positions and the predicted positions at the same time;
set initial positions for a plurality of atoms to be simulated in a scene;
calculate positions of the plurality of atoms simulated in the scene at each time in a second time series by utilizing the trained GAN;
determine movement information of the plurality of atoms simulated by analyzing the positions of the plurality of atoms to be simulated at each time in the second time series; and
determine properties of a substance made up of the plurality of atoms simulated based on the movement information.

8. The device according to claim 7 , wherein the processor is further configured to:
calculate the predicted positions of the plurality of specific atoms at the same time in the first time series by utilizing a generative unit in the GAN;
determine a probability that a predicted position among the predicted positions is a real position among the real positions at the same time by utilizing a discriminative unit in the GAN; and
train the configuration of the generative unit and the discriminative unit based on the predicted position and the real position at the same time, until the probability determined by the discriminative unit is 50%.

9. The device according to claim 8 , wherein the processor is further configured to:
train the discriminative unit in such a manner that the discriminative unit is capable of determining whether the predicted position is the real position at the same time; and
train the generative unit in such a manner that the discriminative unit is not capable of determining whether the predicted position calculated by the generative unit is the real position at the same time.

Assignees

• Fujitsu Ltd

Inventors

• Wang Liuan
• Sun Jun

Classifications

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0475

  Generative networks · CPC title

• G06N3/094

  Adversarial learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title