Research Digest — 2026-05-05¶
Halide Solid Electrolytes¶
1. Mechanistic Study of Mixed Lithium Halides Solid State Electrolytes¶
Source: arXiv:2511.15402 · 📅 2025-11 · ↗ Open paper
Used the PET-MAD universal MLIP to study the Li₃YCl₆₋ₓBr₆₍₁₋ₓ₎ halide system, finding that Cl/Br distribution is weakly correlated and that alloying primarily modulates lattice parameters. Constant-volume vs constant-pressure simulations disentangle chemical composition from volumetric effects on conductivity, showing compensation between the two. Y→In substitution increases conductivity at 25% In content for the C2/m phase.
Relevance to DENG.Group
Directly relevant to Mengke Li and Yan Li's work on halide electrolytes. The PET-MAD approach to screening halide compositions via MLIPs is a methodology the group could adopt for exploring new halide chemistries.
ML Interatomic Potentials¶
2. An AI-Ready Fine-Tuning Framework for Accurate Machine-Learning Interatomic Potentials in Solid-Solid Battery Interfaces¶
Source: arXiv:2601.17847 · 📅 2026-01 · ↗ Open paper
Proposes FIRE (Fine-tuning with Integrated Replay and Efficiency), a framework for adapting universal MLIPs to solid-solid battery interfaces. Achieves <1 meV/atom energy RMSE and ~20 meV/Å force RMSE across six interface systems, an order-of-magnitude improvement over existing models while using only 10% of original datasets.
Relevance to DENG.Group
Highly relevant to Yanhao Deng's ML potential work and Umang Agarwal's interface studies. The FIRE framework could be directly applied to fine-tune universal models for the group's electrolyte/electrode interface systems.
3. Importance of Electronic Entropy for Machine Learning Interatomic Potentials¶
Source: arXiv:2603.26471 · 📅 2026-03 · ↗ Open paper
Shows that conventional MLIPs fail for mixed-valence materials (NaFePO₄ cathode) because structural optimization leads to incorrect Fe²⁺/Fe³⁺ charge assignments. Introduces charge-state-aware MLIP representations that distinguish oxidation states during training, fixing convex hull predictions for CHGNet, cPaiNN, and MACE.
Relevance to DENG.Group
Important methodological insight for Yanhao Deng's ML potential development. If the group works on any mixed-valence systems (e.g., transition-metal-containing halides), charge-state-aware representations will be essential.
4. Design Space of Self-Consistent Electrostatic Machine Learning Interatomic Potentials¶
Source: arXiv:2603.14700 · 📅 2026-03 · ↗ Open paper
Presents a unified framework for understanding electrostatics in MLIPs by viewing existing models as coarse-grained approximations to DFT. Implements variants within the MACE architecture and evaluates on metal-water interfaces and charged vacancies in SiO₂, demonstrating that more expressive self-consistent models are needed for systems with charge transfer and polarization.
Relevance to DENG.Group
Relevant to the group's ML potential work, especially for modeling charged interfaces and solid electrolytes where long-range electrostatics matter. The framework clarifies design choices for next-generation MLIPs.
Grain Boundaries & Defects¶
5. Grain Boundaries in Ceramic Solid-State Lithium Metal Batteries: A Review¶
Source: Ind. Eng. Chem. Res. 65, 1-26 (2026), arXiv:2508.06866 · 📅 2025-08 · ↗ Open paper
Comprehensive 67-page review covering grain boundary influence on ionic/electronic transport, dendrite and void formation in ceramic solid electrolytes. Discusses space charge layer formation, defect chemistry, and conditions where grain boundaries serve as fast-ion pathways or failure initiation sites across different electrolyte classes.
Relevance to DENG.Group
Essential reading for Cheng Peng, who works on grain boundaries in solid electrolytes. Provides a thorough overview of modeling approaches, experimental characterization, and processing techniques relevant to his research.
Dendrite Growth & Phase Field¶
6. Phase Field Simulation of Dendrite Growth in Solid-State Lithium Batteries Based on Mechanical-Thermo-Electrochemical Coupling¶
Source: arXiv:2509.02013 · 📅 2025-09 · ↗ Open paper
Develops a multi-physics phase-field model coupling mechanical stress, thermal, and electrochemical fields to simulate Li dendrite growth in solid-state batteries. Finds that higher temperature and external pressure suppress dendrite growth, with combined effects promoting flatter and denser Li deposits. Stress concentrates at dendrite roots, promoting lateral growth.
Relevance to DENG.Group
Directly relevant to Shoutong Jin's phase-field dendrite simulation work. The mechanical-thermo-electrochemical coupling approach and findings on temperature/pressure effects provide useful methodological and physical insights.
7. An Analytical Model of Critical and Subcritical Alkali Metal Dendrite Growth in Ceramic Solid Electrolytes¶
Source: arXiv:2603.20113 · 📅 2026-03 · ↗ Open paper
Derives an analytical model for dendrite penetration in ceramic solid electrolytes based on minimal power dissipation, predicting J_crit ∝ c_max^(3/2) where c_max is the longest pre-existing interfacial defect. Extends to subcritical growth via stress-corrosion-cracking from residual electron conduction, predicting Weibull-distributed dendrite growth across samples.
Relevance to DENG.Group
Provides theoretical foundation for Shoutong Jin's phase-field simulations. The analytical J_crit scaling with defect size and Weibull statistics prediction could validate or benchmark simulation results.