Double-Tier ML Atom Simulation for Accuracy-Cost Balance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning interatomic potential (MLIP) methods face challenges such as scalability issues, inability to capture long-range effects, and inaccuracies in capturing discontinuities and physical charges, leading to high computational costs and inefficiencies in simulating complex chemical systems.
Innovation Solution
Implementing a double-tier machine learning approach with a central high-fidelity region and a surrounding low-fidelity region, using MLIPs to dynamically evolve atoms and interpolate between them, while incorporating electrostatic embedding and fixed shell simulations to manage boundary conditions, thereby reducing computational costs and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ab-initio calculations (DFT) are used to obtain accurate interaction energies, then measurement precision is improved, but computational cost increases significantly
Solution Approach 1:
The simulation system is divided into multiple regions with different fidelity levels: high-fidelity regions using ab-initio calculations for accurate interaction energies, and low-fidelity regions using classical force fields for computationally efficient simulations. This segmentation allows the system to maintain accuracy where needed while reducing overall computational cost.
Solution Approach 2:
Different quality levels of computational models are applied to different spatial regions of the simulation. High-fidelity ab-initio models are used locally in regions requiring accurate interaction energies, while lower-fidelity models are used in regions where approximate results are sufficient, optimizing the balance between accuracy and computational efficiency.
2Productivity
If machine learned interatomic potentials are used to reduce computational costs, then productivity is improved, but measurement precision deteriorates due to inability to capture long-range effects and discontinuities
Solution Approach 1:
Electrostatic embedding is introduced as an intermediary mechanism that allows the high-fidelity region to account for long-range electrostatic effects from the low-fidelity region. This mediator enables the machine learned potentials to capture long-range effects and discontinuities by incorporating electrostatic potentials from surrounding regions, thereby improving accuracy without sacrificing computational efficiency.
3Measurement precision
If a single high-fidelity model is used throughout the simulation, then measurement precision is improved, but device complexity and computational cost increase
Solution Approach 1:
The simulation system is divided into multiple regions with different fidelity levels: high-fidelity regions using ab-initio calculations for accurate interaction energies, and low-fidelity regions using classical force fields for computationally efficient simulations. This segmentation allows the system to maintain accuracy where needed while reducing overall computational cost.
Solution Approach 2:
Instead of applying high-fidelity calculations to the entire system, the method applies high-fidelity ab-initio calculations only to specific regions where accurate interaction energies are critical, while using lower-fidelity models in other regions. This partial application of high-fidelity modeling reduces computational complexity while maintaining necessary accuracy.
Data Source
Figure 1~3
Figure 4A~5
Figure 6A~7
AI summary
A machine learning simulation method of determining a physical state of interaction between atoms from one or more physical properties of the atoms is disclosed. The method including dynamically evolving a first subset of atoms via a first machine learning model within a central high-fidelity region based on the one or more physical properties of the atoms. The method further includes dynamically evolving a second subset of the atoms via a second machine learning model with a remaining low-fidelity region based on the one or more physical properties of the atoms. The method also includes dynamically evolving a third subset of atoms located between the central high-fidelity region and the remaining low-fidelity region based on an interpolation of the first and second machine learning models to determine the physical state between the atoms.