A universal augmentation framework for long-range electrostatics in machine learning interatomic potentials 1. This novel study introduces the Latent Ewald Summation (LES) method, a universal framework to incorporate long-range electrostatics into machine learning interatomic potentials (MLIPs) without needing explicit charge labels. The LES method infers electrostatic interactions directly from energy and force data, significantly enhancing the accuracy of MLIPs for systems with significant electrostatics. 2. The LES framework is designed as a standalone library compatible with various short-range MLIPs, including MACE, NequIP, CACE, and CHGNet. It demonstrates remarkable improvements in predicting Born effective charges (BECs) and dipole moments, even when trained exclusively on energy and force data. This capability is crucial for accurately simulating systems with significant dielectric response, such as polar materials and charged molecules. 3. The study benchmarks LES-enhanced models on diverse systems, including bulk water, polar dipeptides, and gold dimer adsorption on defective substrates. Across all systems, LES not only reduces energy and force prediction errors but also accurately captures physical observables like BECs and adsorption energies. This highlights the robustness and versatility of the LES method. 4. The LES framework is scalable to large and chemically diverse datasets. The authors demonstrate this by training MACELES-OFF on the SPICE dataset, which includes organic molecules and biomolecules. MACELES-OFF outperforms its short-range counterpart in predicting bulk liquid properties and electrical response properties like IR spectra. 5. The LES method addresses a core limitation of current MLIPs by enabling efficient long-range electrostatics without additional training on electrical properties. This opens the door for developing universal MLIPs with full electrostatic physics, which can be applied across a wide range of chemical and biological systems. 💻Code: github.com/ChengUCB/les 📜Paper: arxiv.org/abs/2507.14302 #MachineLearning #Electrostatics #InteratomicPotentials #MaterialsScience #ComputationalChemistry #LongRangeInteractions

Guess what? By learning from energies and forces, machine learning interatomic potentials can now infer electrical responses like polarization and BECs! This means we can perform MLIP MD simulations under electric fields! arxiv.org/pdf/2504.05169

GPU-accelerated NumPy/SciPy library

1/9 🚨 New Paper Alert: Cross-Entropy Loss is NOT What You Need! 🚨 We introduce harmonic loss as alternative to the standard CE loss for training neural networks and LLMs! Harmonic loss achieves 🛠️significantly better interpretability, ⚡faster convergence, and ⏳less grokking!

Our recent work in @PNASNews Transferable #ML CVs streamline the process & reproduce trend in experimental #k_cat for several #enzyme-substrate pairs. Promising to speed up #DrugDiscovery @GroupParrinello @uraucci87 #TransferLearning #Catalysis #compchem pnas.org/doi/10.1073/pnas.24…

Collective variables without descriptors? In the latest preprint by Jintu Zhang, @LuigiBonati and @TrizioEnrico we used graph neural networks to design CVs for enhanced sampling directly as a function of atomic positions 🧵⤵️ arxiv.org/abs/2409.07339 #compchem #mlcolvar #plumed

Excited to finally announce the publication of MatterGen on Nature. MatterGen represents a new paradigm of materials design with generative AI. We are releasing the training and inference code of MatterGen under MIT license. Look forward to seeing how the community will use the tool and build on top of it.

Can universal MLIPs handle complex atomic environments? 🤔 Our study uncovers systematic PES softening in M3GNet, CHGNet, & MACE-MP-0, linking it to biased pre-training datasets. 🌐 A path forward: better PES sampling! Read here 👇 nature.com/articles/s41524-0…

Coshare Science Review highlight: Research progress on identification of the high Tc superconductivity and superconducting phase in La₃Ni₂O₇ under pressure cosharescience.com/articles/… DOI: doi.org/10.61109/cs.202412.1… #bilayernickelate #high_Tc_superconductivity

Learning atomic charges from energy and forces! Check the superior performance in various material systems with latent Ewald summation

In this ⁦@ACSEnergyLett⁩, we screened a large set of Halide Bilayer Separators for All-Solid-State Batteries with ⁦@ClementGroupSB⁩ . ⁩ We showed that some halides are stable with Li3PO4 extending their stabilities at low potentials. pubs.acs.org/doi/10.1021/acs…

Super excited to preprint our work on developing a Biomolecular Emulator (BioEmu): Scalable emulation of protein equilibrium ensembles with generative deep learning from @MSFTResearch AI for Science. #ML #AI #NeuralNetworks #Biology #AI4Science biorxiv.org/content/10.1101/…

👏 Celebrate excellence in materials research at the #F24MRS Graduate Student Awards Special Talk Sessions! 🌟 Join us to support award finalists as they showcase their groundbreaking research! 🧪🎓 🗓️ Dec 3, 12:15–3:15 PM ET 📍 Marriott, 1st Floor #MRSCommunity

Thrilled to preview NeuralPLexer3 Beta (NP3 Beta), the next generation of AI structure biology model our team at @iambic_ai has been building. 🧵 :

We’ve released an improved version of SynFlowNet! SynFlowNet can now handle large action spaces (up to 220k building blocks) and we experiment with training different GFlowNet backward policies. (1/3) arXiv: arxiv.org/abs/2405.01155 code: github.com/mirunacrt/synflow…

Introducing Meta’s Open Materials 2024 (OMat24) Dataset and Models! All under permissive open licenses for commercial and non-commercial use! Paper: arxiv.org/abs/2410.12771 Dataset: huggingface.co/datasets/fair… Models: huggingface.co/fairchem/OMAT… 🧵1/x

The BIDMaP Postdoc Fellows Program is now open for applications from recent PhDs interested in working at the interface of machine learning and natural sciences. Great opportunity to work on AI for Science! bidmap.berkeley.edu/fellowsh…

should have mentioned earlier since people had issues using CHGNet with numpy v2: v0.4.0 is out now and supports both numpy 1 and 2. the Trainer class also now comes with built-in wandb logging. just pass it an account and a project and you're ready to log. full release notes: github.com/CederGroupHub/chg… pip install chgnet==0.4.0

CP-VASP 2.0 is out! Make your constant-potential DFT calculations faster and more accurate! @ACSCOMP @ACSCATLDivision github.com/yuanyue-liu-group…

Meet GUD: our new Generative Unified Diffusion model inspired by the Renormalization Group. It generalizes standard diffusion by a choice of basis, a prior and separate noise schedules per dimension. Main author Mathis Gerdes and w/ Miranda Cheng. arxiv.org/abs/2410.02667