Research Digest — 2026-09-05¶
AI-Driven Electrolyte Discovery & Data Quality¶
1. Prototype-guided transfer of sparse literature knowledge for electrolyte additive discovery¶
Source: arXiv (2609.02209) · 📅 2026-09-02 · ↗ Open paper
Introduces ProtoMI, a literature-driven molecular intelligence framework that learns transferable structural priors from sparse, experimentally validated electrolyte additives and uses them to prioritize candidates across vast unlabeled chemical spaces. Demonstrated on boron-containing additives where performance arises from coupled interfacial reactions rather than single molecular properties.
Relevance to DENG.Group
Relevant to the group's growing AI-for-materials thread - the transfer-learning-from-sparse-data strategy complements simulation-driven screening and could seed hypotheses for Yan Li and Mengke Li's halide additive/degradation studies.
2. When Literature Data Mislead Artificial Intelligence in Materials Discovery¶
Source: arXiv (2609.01621) · 📅 2026-09-02 · ↗ Open paper
Uses solid electrolyte conductivity data as a case study to expose systematic pitfalls in literature-derived datasets: text-figure mismatches, ambiguous axis annotations, unit inconsistencies, and missing measurement context. Shows how these errors propagate into AI models and proposes safeguards for building reliable materials databases.
Relevance to DENG.Group
Essential methodological caution for the whole group, especially Yanhao Deng (ML potentials) and anyone training models on literature SE conductivity values - the same data-hygiene issues directly affect the reliability of the group's ML-screening pipelines.
Transport, Interfaces & Mechano-Electrochemical Coupling¶
3. Low-Temperature Transport in Li-Ion Battery EC/EMC/FEC Electrolytes: Molecular Dynamics and Machine-Learning Modeling¶
Source: arXiv (2609.02824) · 📅 2026-09-02 · ↗ Open paper
Combines classical MD and machine-learning modeling of 1 M LiPF6 EC/EMC electrolytes with 0-10 mol% FEC additive from 298 down to 233 K, quantifying Li+ self-diffusion and Nernst-Einstein vs Green-Kubo conductivity in the low-temperature regime where solvent composition and fluorinated additives couple strongly.
Relevance to DENG.Group
Useful methods reference for Naibing Wu's electrolyte transport simulations - the NE vs GK conductivity comparison and FEC-additive sensitivity analysis transfer directly to polymer and composite electrolyte studies, and offer Yanhao Deng a benchmark problem for ML-accelerated transport prediction.
4. Battery open-circuit voltage is not purely chemical¶
Source: arXiv (2609.02968) · 📅 2026-09-02 · ↗ Open paper
Shows via finite-element simulation and asymptotic analysis that OCV curves in ion-insertion batteries depend on mechanical state and microstructure: particle swelling and external loads create particle-particle contacts whose compressive stresses shift the inserted-ion chemical potential, with a nonlinear Hertzian-contact scaling correction.
Relevance to DENG.Group
Strong conceptual overlap with the group's mechano-electrochemical themes - relevant to Shoutong Jin's phase-field dendrite/pressure work and Umang Agarwal's interface mechanics, providing continuum-theory backing for how stack pressure feeds back into electrochemical potentials.
5. Grain boundary evolution in nanoparticles¶
Source: arXiv (2609.03307) · 📅 2026-09-03 · ↗ Open paper
Reviews and analyzes atomic-scale mechanisms of grain boundary evolution in nanoparticles, where free surfaces, rotational degrees of freedom, and shape relaxation alter GB migration and annihilation pathways relative to bulk. Motivated by GB engineering for functional properties including catalysis.
Relevance to DENG.Group
Relevant to Cheng Peng's grain boundary simulations in solid electrolytes - nanocrystalline SEs have high GB density and the finite-size/surface-coupled GB evolution mechanisms here extend bulk GB thermodynamics and kinetics intuition to the nanoscale regime.