Research Digest — 2026-05-17¶
ML Interatomic Potentials & Active Learning¶
1. End-to-End Language-to-Machine Learning Interatomic Potential Development with Autonomous Agentic Workflows¶
Source: arXiv:2605.14527 · 📅 2026-05-14 · ↗ Open paper
Proposes Lang2MLIP, a multi-agent framework that takes natural-language input and formulates end-to-end MLIP development as a sequential decision-making problem solved by large language models. At each step, a decision-making agent observes the current dataset, model, evaluation results, and execution log, and automatically selects an appropriate action to improve the model — removing the need for a predefined pipeline and enabling self-correction. The framework is evaluated on a solid electrolyte interphase (SEI) system with multiple components and interfaces.
Relevance to DENG.Group
Highly relevant to Yanhao Deng's MLIP development work. The multi-agent autonomous workflow directly addresses the bottleneck of MLIP training for complex battery interfaces. The SEI application is particularly pertinent to the group's solid-state battery modeling. This could significantly reduce the expertise barrier for developing MLIPs for the group's halide and oxide electrolyte systems.
2. Force-Aware Neural Tangent Kernels for Scalable and Robust Active Learning of Machine-Learning Interatomic Potentials¶
Source: arXiv:2605.13788 · 📅 2026-05-14 · ↗ Open paper
Introduces a force-aware Neural Tangent Kernel (NTK) framework for active learning of MLIPs that scales linearly with candidate pool size. The method uses chunked feature-space posterior-variance shortlisting to screen ~200k structures within hours, and extends NTKs to handle energy-force supervision via mixed parameter-coordinate derivatives. On the OC20 dataset, force-aware acquisition achieves the lowest energy and force MAE across all metrics and distribution splits, while remaining robust to candidate-pool distribution shifts where committee-based methods show higher variance.
Relevance to DENG.Group
Directly relevant to Yanhao Deng's MLIP development. Active learning is critical for efficiently training accurate potentials with minimal DFT calculations. The force-aware NTK approach could replace committee-based uncertainty estimation in the group's existing workflows, offering better scalability and robustness. The linear scaling to large candidate pools is essential for the high-throughput screening of battery materials.
Defects & Grain Boundaries¶
3. Grain Boundary-Driven Lattice Dynamics in a Solid-State Li-Ion Conductor¶
Source: Advanced Science (10.1002/advs.75295) · 📅 2026-04-17 · ↗ Open paper
Performs first-principles phonon calculations on grain boundaries in anti-perovskite Li3OCl to explore their influence on lattice dynamics and Li-ion transport. The study finds Li-ion vibrational hardening within grain boundaries — indicating increased transport barriers — while the anion sublattice undergoes vibrational softening, linked to increased motion and potential for detrimental side reactions leading to structural degradation. The alignment between Li-ion migration pathways and vibrational eigenvectors is poorer in grain boundaries compared to bulk, providing a phonon-level explanation for grain boundary resistance.
Relevance to DENG.Group
Directly relevant to Cheng Peng's grain boundary research. This is the first study linking bulk phonon properties to grain boundary resistance in solid electrolytes. The finding that anion softening at grain boundaries correlates with degradation risk provides a new computational screening criterion. The anti-perovskite Li3OCl system complements the group's halide electrolyte work and the methodology could be applied to Li3YCl6 and other halide systems.
Polymer Electrolytes¶
4. Chemically Driven Conformational Rearrangement of PVDF-Based Polymer Electrolyte to Improve Ionic Conductivity for Long-Cycling Lithium Metal Batteries¶
Source: Energy & Environmental Science (10.1039/D6EE00044D) · 📅 2026-05-11 · ↗ Open paper
Proposes a chemically driven approach to induce molecular chain conformational rearrangement of PVDF, generating more polar β-phase to promote Li salt dissociation. An interchain Li+ transport pathway with low energy barrier is designed through synergy of strong coordinated Li+···-SO3⁻ (in Nafion) and weak dipole -C-F···Li+ interaction, achieving ionic conductivity of 1.81 mS cm⁻¹. 6Li NMR tracing shows the proportion of Li+ transport through the designed low-energy-barrier pathway increases from 44% to 79%. Li||Li symmetric cells cycle for 7800 h at 0.1 mA cm⁻², and a 1.3 Ah pouch cell passes nail penetration testing.
Relevance to DENG.Group
Relevant to the group's interest in polymer and composite electrolytes. The conformational rearrangement strategy and interchain pathway design via synergistic coordination provide molecular-level design principles that could be explored computationally. The Nafion-PVDF combination and the β-phase promotion concept could inform the group's work on polymer-ceramic composite electrolytes for solid-state batteries.
Solid Electrolyte Reviews & Fundamentals¶
5. Modern Solid Electrolytes for All-Solid-State Batteries: Materials Chemistry, Structure, and Transport¶
Source: arXiv:2604.17380 · 📅 2026-04-24 · ↗ Open paper
A comprehensive 46-page review examining inorganic solid-state electrolytes through the lens of structure-property relationships, covering oxides, sulfides, and halides as three major framework chemistries. The review argues that long-range ion transport is increasingly understood not as motion along a single idealized pathway, but as the macroscopic outcome of statistically connected low-barrier local migration events distributed across the structure. It covers mixed anion halides and anti-perovskite-related materials, and discusses both experimental and computational approaches for establishing multiscale structure-property relationships.
Relevance to DENG.Group
A valuable reference for the entire Deng group. The multiscale view of ion transport as statistically connected local events aligns with the group's computational approach. The systematic comparison of oxide, sulfide, and halide transport mechanisms provides a framework for computational screening. Particularly useful for new group members as a comprehensive introduction to solid electrolyte design principles.