Research Digest — 2026-10-27¶
ML Potentials & Grain Boundaries¶
1. Machine-Learning-Potential-Driven Volcano Relationship in Grain Boundary Amorphicity and Ionic Transport of Antiperovskite Solid Electrolytes¶
Source: Advanced Materials (10.1002/adma.75492) · 📅 2026-10-23 · ↗ Open paper
An actively learned machine-learning interatomic potential bridges the quantum-accuracy/MD-scale gap for halide-tunable antiperovskite solid electrolytes. By varying the local Cl/Br ratio and quantifying interfacial amorphicity and ion flux, the authors identify a volcano relationship between structural disorder and ionic transport, with Cl-enriched grain boundaries optimizing conductivity.
Relevance to DENG.Group
Sits exactly at the intersection of Yanhao Deng's MLIP development and Cheng Peng's grain-boundary transport work. The active-learning workflow and the disorder-vs-transport volcano design rule are directly transferable to our GB and halide-electrolyte simulation projects.
2. Tracking the lithiation state of LixSi from machine-learned XPS binding energies¶
Source: npj Computational Materials (s41524-026-02302-8) · 📅 2026-10-17 · ↗ Open paper
The Ceriotti group combines machine-learned core-level binding-energy prediction with large-scale grand-canonical atomistic simulations to compute XPS spectra of disordered LixSi phases. The framework quantitatively tracks lithiation through the amorphous and crystal-amorphous transformations that are otherwise hard to characterize.
Relevance to DENG.Group
A template for connecting our atomistic simulations to spectroscopic observables: Yanhao Deng's MLIPs could supply the configurations, and the ML-XPS pipeline transfers directly to our SEI/interphase and degradation studies (Yan Li, Tzu-Chen Liu).
Defects & Ionic Conductivity¶
3. High Defect Formation Energies Limit Ionic Conductivity in Na3Ga2Li3F12 and Na3Al2Li3F12¶
Source: Chemistry of Materials (10.1021/acs.chemmater.6c01542) · 📅 2026-10-23 · ↗ Open paper
Combining structural characterization with first-principles calculations, this study shows fluoride garnets are highly ordered with low Li-ion migration barriers (0.16-0.29 eV) yet intrinsic conductivities of only ~2e-11 S/cm at 100 °C. The culprit is high lithium Frenkel/antisite defect formation energies, which starve the lattice of mobile carriers — defect chemistry, not the diffusion framework, limits transport.
Relevance to DENG.Group
A textbook defect-chemistry argument for why low migration barriers do not guarantee high conductivity — directly relevant to Mengke Li and Yan Li's halide defect/transport mechanisms and to Cheng Peng's point-defect work on grain boundaries. Good reference when arguing carrier concentration vs. barrier height in our own papers.
4. Vacancy-mediated ultrafast oxide-ion conduction in excess-oxygen Sillén oxybromides¶
Source: Nature Communications (s41467-026-77848-8) · 📅 2026-10-14 · ↗ Open paper
Bi1.9Te0.1LaO4.05Br achieves 0.01 S/cm oxide-ion conductivity at just 233 °C via excess-oxygen vacancy/interstitial chemistry, outperforming YSZ, LSGM, and leading proton conductors in the Norby gap, with negligible chemical expansion and high chemical stability.
Relevance to DENG.Group
Vacancy-mediated anion conduction in a layered oxyhalide framework connects to Mengke Li's halide ion-transport mechanisms and Tzu-Chen Liu's defect-mediated transport studies — a useful counterpoint on how layered halide-containing structures accommodate mobile anions.
Interfaces & Protective Coatings¶
5. Combining Simulations and Spectroscopy to Design Protective Coatings on Sulfide Solid Electrolytes¶
Source: arXiv:2610.32905 · 📅 2026-10-23 · ↗ Open paper
Sulfide solid electrolytes degrade electrochemically and on moisture exposure; molecular surface coatings are a scalable fix. This preprint pairs atomistic simulations with XPS to uncover the molecular-level factors governing coating-electrolyte surface interactions and to rationally design humidity-tolerant coatings.
Relevance to DENG.Group
Directly parallels Yan Li's halide degradation and transport work — the combined simulation/spectroscopy methodology and molecular-coating strategy transfer naturally to stabilizing halide electrolyte surfaces in our projects.
Thin-Film Growth Kinetics¶
6. Growth Kinetics of Gallium Selenide on Graphene: Bridging Atomistic and Mesoscale Processes¶
Source: ACS Applied Materials & Interfaces (10.1021/acsami.6c14237) · 📅 2026-10-22 · ↗ Open paper
DFT and kinetic Monte Carlo are combined to model early-stage 2D GaSe growth on graphene, mapping how domain morphology, growth rate, and nucleation density depend on Ga/Se atomic fluxes and substrate temperature. Ga pre-adsorption on the substrate is key to nucleation, while sufficient Se flux enables uniform 2D growth.
Relevance to DENG.Group
The same DFT-to-kMC multiscale pipeline Jingyuan Lin uses for thin-film growth modelling — their flux/temperature-dependent nucleation maps offer a concrete example of fitting kMC rates from first-principles energies that he can benchmark against.
MOFs: Adsorption, Diffusion & ML¶
7. CO2-Induced Shifts in the Confined Phase Transformation of CH4 in Metal-Organic Frameworks¶
Source: arXiv:2610.32918 · 📅 2026-10-23 · ↗ Open paper
Molecular simulations show that co-adsorbed CO2 shifts the sharp, pressure-triggered confined phase transformation of CH4 in MOFs and reshapes the associated metastable adsorption states. This matters for predicting gas storage and separation where mixtures, not pure gases, flow through the pores.
Relevance to DENG.Group
Mixture effects on confined phase transitions are core to Jason Li's MOF adsorption and gas kinetics project — a concrete example of how a secondary adsorbate tunes uptake behavior and hysteresis in nanoporous confinement.
8. AtomMOF: All-Atom Flow Matching for MOF-Adsorbate Structure Prediction¶
Source: arXiv:2602.07351 (v2) · 📅 2026-02-07 · ↗ Open paper
A generative flow-matching model predicts full all-atom MOF-adsorbate configurations without rigid-body assumptions or a known host structure, aimed at replacing expensive structural sampling in MOF screening for applications like direct air capture. The v2 revision was re-announced on this week's arXiv feed.
Relevance to DENG.Group
A potentially useful ML tool for Jason Li's MOF work: all-atom adsorbate-placement generation could seed our DFT/GCMC adsorption calculations much faster than current docking or manual structure-building approaches.