AI Interatomic Potentials vs DFT: Crystal Relaxation Lab |…

What AI Interatomic Potentials Are

AI interatomic potentials are machine learning models designed to predict the interactions between atoms based on their positions. These models, such as M3GNet, CGCNN, and CHGNet, use graph neural networks to learn from large datasets of atomic configurations and forces.

These potentials can be used in molecular dynamics simulations, where they replace computationally expensive density functional theory (DFT) calculations with faster approximations, making the simulation of complex materials more feasible.

How DFT Works

Density Functional Theory (DFT) is a quantum mechanical model used to investigate the electronic structure of many-body systems, especially atoms, molecules, and the condensed phases. It relies on solving the Kohn-Sham equations, which approximate the exact ground-state wavefunction of electrons in a material.

The accuracy of DFT lies in its ability to balance computational efficiency with physical realism, but it can be computationally intensive for large systems or complex materials.

live demo · related simulation● LIVE

Crystal Relaxation and Its Importance

In crystal relaxation simulations, the goal is to find the lowest energy configuration of a crystal lattice. This process involves minimizing the potential energy by adjusting atomic positions until equilibrium is reached. Crystal relaxation is crucial for understanding material properties such as stability, phase transitions, and defect formation.

By comparing AI interatomic potentials with DFT results, researchers can assess the accuracy and reliability of these machine learning models in predicting atomic interactions.

Why Training Steps Matter

The training steps slider controls how many iterations the model undergoes to refine its predictions. More training steps generally lead to a more accurate force prediction, but also increase computational cost. The goal is to find an optimal balance between accuracy and efficiency.

By observing the convergence of atoms during relaxation, one can determine if the model has learned enough from the DFT data or if further training is necessary.

Frequently asked questions

What does the initial-disorder slider do?

The initial-disorder slider introduces perturbations to the crystal lattice, allowing you to observe how well the AI model can relax a non-equilibrium structure back to its ground state.

Why compare M3GNet/CGCNN/CHGNet with DFT?

Comparing these models with DFT helps validate their accuracy and reliability. DFT is considered the gold standard in computational materials science, so any model that can closely match its results is valuable for practical applications.

Can AI interatomic potentials replace DFT entirely?

While AI interatomic potentials offer significant speedup, they are not yet as accurate as DFT. They are particularly useful in large-scale simulations where the computational cost of DFT is prohibitive.

What are some real-world applications of crystal relaxation?

Crystal relaxation is used in developing new materials for electronics, understanding phase transitions in alloys, and predicting defect formation in semiconductors. It plays a crucial role in the design of advanced materials with tailored properties.

Try it live

Everything above runs in your browser — open AI Interatomic Potentials vs DFT: Crystal Relaxation Lab and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.

▶ Open AI Interatomic Potentials vs DFT: Crystal Relaxation Lab simulation