ab initio– tag –
-
DeepH-pack Unites Ab Initio Calculations and Deep Learning to Accelerate AI-Driven Electronic Structure Modeling
Facebook (reposting about DeepH-pack) USA Overview DeepH-pack has been introduced as a general-purpose neural network package that combines ab initio calculations with deep learning to accelerate electronic structure modeling. This tool ... -
Graph Neural Network Accelerates Novel Material Discovery with Reduced Computational Cost for Catalysts and Batteries
PRX Intelligence USA Overview A new Graph Neural Network (GNN) model published in PRX Intelligence significantly accelerates novel material prediction and discovery while drastically lowering computational costs. The model leverages proj... -
MACE and SevenNet Data Efficiency Evaluated for Material-Specific MLIP Construction: Achieving Ab Initio Accuracy with 2,000 AIMD Configurations
arXiv International Overview The amount of ab initio molecular dynamics (AIMD) data required to fine-tune universal machine-learned interatomic potentials (MLIPs) for material-specific applications has been quantified. Research indicates... -
LAMMPS Accelerates MLIP Integration: Revolutionizing MD Simulations with DeePMD-kit, MACE, CHGNet Foundation Models
LAMMPS Molecular Dynamics Simulator USA Overview The LAMMPS molecular dynamics simulator is enhancing its integration with several machine-learned interatomic potential (MLIP) frameworks, including DeePMD-kit, NequIP, Allegro, MACE, Seve... -
UniFFBench Reveals Critical Role of System-Specific Fine-Tuning for Universal ML Force Fields via Rigorous Experimental Benchmarking of 6 EGraFF Algorithms
ResearchGate International Overview The UniFFBench study rigorously evaluates universal machine learning interatomic potentials (uMLIPs) from six prominent EGraFF algorithms—NequIP, Allegro, BOTNet, MACE, Equiformer, and TorchMDNet—again... -
GPUMD 4.0 Achieves High-Performance Versatile Materials Simulations with Integrated Machine-Learned Potentials
Materials Genome Engineering Advances Global Overview The high-performance molecular dynamics package GPUMD 4.0 has been released, integrating advanced machine-learning potentials (MLPs) based on the NeuroEvolution Potential (NEP) framew... -
arXiv: New ML Potential ‘TWIN’ Achieves Ab Initio Accuracy and Transferability in Biomolecular Simulations
arXiv Global Overview A new arXiv preprint introduces TWIN (Transferable Water Implicit Network), a highly transferable implicit solvent machine-learning potential (MLP) for biomolecular systems. TWIN, trained solely on ab initio and exp... -
MLIP Enhanced Sampling Simulations Uncover Dynamic Conformations and Catalytic Implications of Au9 Nanocluster Confined in UiO-66-NH2 MOF
ChemRxiv USA Overview This research meticulously investigated the structural and dynamic behavior of an Au9 nanocluster confined within a UiO-66-NH2 MOF using machine learning interatomic potential (MLIP)-driven enhanced sampling simulat... -
Hugging Face Papers Unveil Electronic Density Generative Framework Combining 3D Convolutional Autoencoders and Latent Diffusion Models
Hugging Face International Overview A generative framework for learning electronic density's latent space dynamics has been introduced in Hugging Face's paper collection. This framework combines 3D convolutional autoencoders with latent ... -
arXiv Paper Presents ‘AutoPot’: Automated, Massively Parallel Workflow for Constructing Machine-Learning Potentials
arXiv USA Overview A new preprint on arXiv introduces 'AutoPot,' an automated and massively parallelized workflow for constructing Machine Learning Interatomic Potentials (MLIPs). MLIPs bring quantum accuracy to atomic modeling, enabling...
1