Research Digest — 2026-11-09¶
Solid Electrolytes & Interfaces¶
1. Coupling electrolyte mechanical reinforcement and interfacial stress relaxation for near-ambient sulfide all-solid-state lithium metal batteries¶
Source: Nature Communications (s41467-026-79393-w) · 📅 2026-11-07 · ↗ Open paper
Demonstrates room-temperature all-solid-state lithium metal batteries running at stack pressures as low as 0.1-0.3 MPa through a combined electrolyte-current-collector design. A self-polymerizable thioctic acid-modified composite sulfide electrolyte suppresses internal lithium creep with high stress tolerance, while a porous carbon matrix current collector provides localized stress relaxation and hosts lithium deposition inside its voids, enabling stable long-term cycling without external pressure.
Relevance to DENG.Group
Directly relevant to our dendrite and interface modelling: Shoutong Jin's phase-field dendrite work can use the stress-partitioning picture (electrolyte reinforcement vs. collector stress relaxation) to set mechanically informed boundary conditions, and Yan Li's electrolyte degradation work gets a data point on how composite sulfide mechanical design controls contact loss. The low-stack-pressure operation criterion is a useful validation target for coupled electro-chemo-mechanical simulations.
2. Unraveling the Impact of Surface Lithiation on the Initial Solid Electrolyte Interphase Formation on Graphite via Machine Learning Molecular Dynamics¶
Source: ACS Nano (acsnano.6c13417) · 📅 2026-11-06 · ↗ Open paper
Develops a reactive machine learning potential to simulate electrolyte reduction at the graphite anode/electrolyte interface on nanosecond timescales. By systematically varying the local Li loading at edge sites (a proxy for electrode potential), they map SEI formation pathways: low Li loading leaves reduction incomplete and lets intermediates escape into the electrolyte, while high Li loading forms inorganic-rich interfacial clusters that passivate the surface and suppress further solvent decomposition.
Relevance to DENG.Group
This is the exact tool class Yanhao Deng is building — reactive ML potentials for electrochemical interfaces — and the Li-loading-as-potential protocol is a clean trick for our interface work: Umang Agarwal's electrolyte/electrode interface simulations could adopt it to mimic applied potential without explicit electron reservoirs, and the passivation-versus-escape competition is a testable mechanism for SEI growth models.
3. Quantifying the Dissolution of Organic and Inorganic Components in Solid Electrolyte Interphase for Sodium-Ion Batteries¶
Source: Angewandte Chemie (ange.5211841) · 📅 2026-11-05 · ↗ Open paper
Builds a component-resolved framework (ion chromatography, mass spectrometry, EQCM, XPS) to quantify dissolution and retention of individual organic and inorganic SEI species in sodium-ion batteries. Tetraglyme-based electrolytes cut total SEI dissolution from 75% to 53% and raise inorganic F-containing species retention from 18% to 30% via an anion-enriched Na+ solvation shell, yielding 91.2% capacity retention over 500 cycles in 1.25 Ah pouch cells.
Relevance to DENG.Group
Relevant to our Na-halide electrolyte interests (Yan Li's halide degradation and transport): Na SEI compounds are markedly more soluble than Li analogues, and these absolute dissolution numbers give the first quantitative dataset our thermodynamic stability models (solubility/Pourbaix-style analyses) could aim to reproduce for interphase failure prediction.
ML Interatomic Potentials¶
4. MACE-MAX: A fine-tuned machine learning interatomic potential for high-throughput screening and experimental validation of high-entropy M2AlC MAX phases¶
Source: Chemical Engineering Journal (cej.183974) · 📅 2026-11-01 · ↗ Open paper
Presents MACE-MAX, a machine learning interatomic potential fine-tuned from the universal MACE-MP foundation model on a compact, hierarchically structured dataset spanning unary, binary, and nine representative ternary M2AlC compositions. It enables high-throughput thermodynamic screening of high-entropy MAX phases across a nine-element design space (Ti, V, Cr, Zr, Nb, Mo, Hf, Ta, W), with experimental validation of the predictions.
Relevance to DENG.Group
The fine-tuning recipe mirrors Yanhao Deng's ML-potential workflow for battery materials: the hierarchical dataset construction (pure elements -> binaries -> selected ternaries) and the screening-to-experimental-validation loop are directly transferable to screening halide and oxide solid electrolyte compositions with fine-tuned foundation potentials.
Microstructure Evolution & Grain Boundaries¶
5. A general sharp-interface simulation framework for microstructure evolution¶
Source: arXiv:2611.04582 · 📅 2026-11-03 · ↗ Open paper
Introduces Gaussian-projected kinetics (GPK), a sharp-interface framework where each cell stores an integer grain/phase label and a temporary Gaussian projection supplies interface normal and curvature. Capillarity, chemistry, stored energy, transport and elasticity move boundaries in one topology-preserving update, handling grain growth, recrystallisation, multiphase evolution and elastic-coupled solidification in 2D and 3D — running 8.5-96x faster than phase-field at matched accuracy in their benchmarks.
Relevance to DENG.Group
A sharp-interface alternative to the phase-field methods Shoutong Jin uses for dendrite growth, with an order-of-magnitude speedup at matched accuracy; the label-based scheme also suits Cheng Peng's grain boundary migration problems in solid electrolytes. Worth benchmarking GPK against our phase-field dendrite and GB models — ensemble studies over many microstructures become feasible.
6. Surrogate Models for Stochastic Al-Cu Phase-Field Solidification¶
Source: arXiv:2611.04695 · 📅 2026-11-03 · ↗ Open paper
Trains a variance-preserving denoising diffusion bridge model to autoregressively predict phase and concentration fields in 2D Al-Cu phase-field solidification given the current state and temperature field. It beats deterministic U-Net and ConvNeXt baselines on long-horizon errors, retains fine dendritic structure where deterministic surrogates smooth it away, and runs ~73x faster than the phase-field solver.
Relevance to DENG.Group
Diffusion-bridge emulation of dendritic growth maps one-to-one onto Shoutong Jin's phase-field dendrite project: the stochastic ensemble capability is exactly what is needed for statistics of dendrite tip selection and morphology distributions, and the paper's failure analysis of deterministic surrogates is a practical guide for training such emulators on our own dendrite simulation data.
7. Insight into the two competing migration mechanisms of Σ5(120) grain boundary in face-centered cubic nickel¶
Source: Acta Materialia (actamat.2026.122921) · 📅 2026-11-01 · ↗ Open paper
Atomistic study of the Σ5(120) grain boundary in fcc nickel that resolves two competing migration mechanisms and identifies the conditions under which each dominates the boundary's motion.
Relevance to DENG.Group
Cheng Peng's grain boundary work in solid electrolytes faces the same question — which microscopic mechanism carries GB migration under given driving forces — and this paper's approach to separating competing mechanisms in molecular dynamics carries over directly to GB kinetics in ionic conductors, where coupling to ion transport adds a further selectivity criterion.
MOFs & Gas Kinetics¶
8. Molecular dynamics simulation of gas diffusion in metal-organic frameworks based on a classical force field and machine-learned interatomic potential¶
Source: Microporous and Mesoporous Materials (micromeso.114447) · 📅 2026-11-01 · ↗ Open paper
MD study of gas diffusion in metal-organic frameworks comparing diffusivities from a classical force field against a machine-learned interatomic potential, addressing when flexible-framework effects demand the more expensive ML description.
Relevance to DENG.Group
Jason Li's MOF adsorption and gas kinetics project lives on exactly this comparison: the paper's FF-versus-MLIP diffusivity benchmark informs whether classical force fields suffice for his gas transport simulations or whether framework flexibility needs a machine-learned potential.