ML Interatomic Potentials With Electrostatic Embedding for Boundary Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning interatomic potential (MLIP) methods face scalability issues with large systems, fail to capture long-range effects, and struggle with accurate charge assignment and boundary condition simulations, leading to high computational costs and inaccuracies.
Innovation Solution
Implement a fixed shell simulation setup using machine learning interatomic potentials (MLIP) with a central high-fidelity region and optional shells to manage arbitrary geometric and chemical boundary conditions, incorporating electrostatic embedding and hybrid simulation methods to address long-range interactions and charge assignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ab-initio calculations (DFT) are used to obtain accurate estimations of interaction energies, then measurement precision is improved, but use of energy worsens due to high computational cost
Solution Approach 1:
The patent creates a machine learned interatomic potential (MLIP) model that copies the accuracy characteristics of expensive ab-initio DFT calculations while using computationally inexpensive machine learning algorithms. The MLIP is trained on DFT data to replicate its predictive accuracy for interaction energies, forces, and stresses, enabling high-precision simulations at a fraction of the computational cost.
Solution Approach 2:
The patent changes the fundamental parameter used for energy calculation from first-principles quantum mechanical solving to machine learning inference. By training neural networks or Gaussian process models on DFT training sets, the system replaces the computationally intensive DFT solver with a fast ML-based energy evaluation function that maintains similar accuracy for various material properties.
2Use of energy by moving object
If machine learned interatomic potentials (MLIPs) are used to reduce computational costs, then use of energy is improved, but measurement precision worsens due to scalability issues and failure to capture long-range effects
Solution Approach 1:
The patent introduces an intermediary electrostatic field model that mediates between the MLIP region and the MM region. This electrostatic embedding approach allows the MLIP to accurately capture long-range electrostatic interactions by incorporating charges from the MM region into the MLIP calculations, thereby maintaining measurement precision for charged systems while retaining MLIP computational efficiency.
Solution Approach 2:
The patent segments the simulation system into distinct MLIP and MM regions with clear interface definitions. This segmentation allows each region to be treated with appropriate methods: MLIP for high-accuracy atomistic simulations in the core region, and MM for efficient treatment of surrounding environments, achieving both computational efficiency and measurement precision through spatial decomposition.
3Measurement precision
If fixed shell simulation setup with electrostatic embedding is implemented, then measurement precision is improved for boundary conditions, but device complexity increases
Solution Approach 1:
The patent applies local quality by implementing different levels of detail in different spatial regions. The central MLIP region uses high-accuracy machine learning potentials for atomistic simulations, while the surrounding MM region uses simplified molecular mechanics force fields. This spatial variation in model quality optimizes computational resources and maintains measurement precision where it matters most at the MLIP-MM interface.
Solution Approach 2:
The patent creates a universal hybrid simulation framework that can handle multiple system types and boundary conditions through a single electrostatic embedding approach. The same MLIP-MM interface treatment works for various geometries, charged systems, and boundary conditions, reducing device complexity by providing a multi-functional solution rather than requiring separate methods for different scenarios.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Achieves high-fidelity atomistic simulations with reduced computational costs and accurate boundary conditions, enabling efficient simulation of complex systems by leveraging short-range topology-free MLIPs and hybrid models.
Implementation Method 1
assigning charges to a first subset of atoms in a molecular mechanics region via the MLIP to generate an electric field in a machine learning region to determine the physical state of interaction between the atoms
Data Source
AI summary
A machine learning interatomic potential (MLIP) method of determining a physical state of interaction between atoms from one or more physical properties of the atoms is disclosed. The method includes assigning a charge qi to a first subset of atoms via the MLIP dependent on a nonconstant field generated by a second subset of atoms having a charge qj to determine the physical state of interaction between the atoms. The method may include using the physical state of interaction between the atoms to control the chemical system.