Research Digest — 2026-11-11¶
ML Interatomic Potentials¶
1. Overcoming speed-accuracy trade-offs with medium-sized interatomic potentials¶
Source: arXiv:2611.11544 · 📅 2026-11-08 · ↗ Open paper
Shows that an intermediate-size machine-learned interatomic potential — an atomic cluster expansion that learns the radial components of each polynomial as independent linear transforms and embeds many densities with learned exponents — can approach the transferable accuracy of the largest models at a fraction of the cost. Demonstrated on He-bearing W-Ta-Hf-Cr-V-Ti alloys, with validation error close to leading potentials despite cheaper training and inference.
Relevance to DENG.Group
Directly informs Yanhao Deng's ML-potential project: the 'medium-sized' ACE design is a concrete recipe for fitting accurate electrolyte/electrode potentials without huge models. The cluster-expansion flavour also connects to Tzu-Chen Liu's cluster-expansion work on amorphous electrolytes.
2. Physics Informed Priors for Machine Learning Interatomic Potentials with Uncertainty Quantification¶
Source: arXiv:2611.09325 · 📅 2026-11-06 · ↗ Open paper
Introduces a model-misspecification-aware uncertainty scheme for MLIPs, building on POPS-Regression but letting the user enforce physical priors across the ensemble of models so uncertainty estimates stay physical. Demonstrated for phonon vibrations and elastic properties with both linear and message-passing neural network potentials.
Relevance to DENG.Group
Gives Yanhao Deng a practical way to attach trustworthy error bars to MLIP-predicted phonon and elastic properties of electrolytes — useful when validating our force fields against DFT or experiment before long MD runs.
3. Defects and Structural Evolution in CrSBr: Machine-Learning Potentials for Atomistic Simulations¶
Source: arXiv:2611.12267 · 📅 2026-11-09 · ↗ Open paper
Develops and benchmarks machine-learning interatomic potentials for the Cr-S-Br chemical space, reaching near-DFT accuracy for the layered van der Waals magnet CrSBr. The potentials enable large-scale simulations of defect formation, migration, and structural evolution at length- and timescales inaccessible to first-principles methods.
Relevance to DENG.Group
A clean template workflow — DFT data generation, potential fitting, defect benchmarking — for materials-specific MLIPs. Directly reusable for our defect and grain-boundary simulations (Cheng Peng's GB models, Yanhao Deng's potential development).
Ion Transport & Electrolyte Physics¶
4. A Rule That's Made to be Broken? Reframing the Arrhenius Law for Ionic Transport¶
Source: arXiv:2611.11302 · 📅 2026-11-08 · ↗ Open paper
Re-examines what 'activation energy' from Arrhenius fits of self-diffusion coefficients actually means for candidate solid electrolytes. By identifying individual diffusive steps in trajectories, the authors show how a physically significant activation energy can be extracted, and discuss when and why non-Arrhenius behaviour emerges.
Relevance to DENG.Group
Core methodology for Yan Li and Mengke Li's halide-ion transport analyses — it sharpens how we extract migration barriers from MD/kMC trajectories (Tzu-Chen Liu's kMC work) and how we should report conductivity activation energies vs temperature.
Interfaces, SEI & Halide Electrolytes¶
5. Machine learning for accelerated screening of sodium anode/chalcogenide electrolyte interfaces for all-solid-state sodium batteries¶
Source: Electrochimica Acta (10.1016/j.electacta.2026.149766) · 📅 2026-11-01 · ↗ Open paper
Builds a first-principles-guided machine-learning framework to screen Na anode / sulfide-selenide electrolyte interfaces, combining formation thermodynamics, electronic coupling, and Na-ion migration descriptors. Cheap compositional, structural, migration-network, and BVSE-like descriptors pre-rank Na, Na-alloy, and co-doped interfaces before relaxed DFT, dimer, and grand-potential calculations validate the top picks.
Relevance to DENG.Group
A multi-descriptor screening workflow directly analogous to what Umang Agarwal needs for electrolyte/electrode interface stability — the descriptor set (interface formation energy + electronic coupling + migration network) is a good starting template for our own interface-screening projects, including the Na analogues.
6. First-principles study of chemical species adsorption and solid electrolyte interphase formation on defective sodium metal anodes¶
Source: Applied Surface Science (10.1016/j.apsusc.2026.167352) · 📅 2026-11-01 · ↗ Open paper
DFT study of 22 electrolyte-derived species (solvents, salts, decomposition products, SEI components) adsorbing on defective Na(110) surfaces with adatom clusters and multi-atom vacancies. Undercoordinated defect sites strongly polarize and activate adsorbed species, fragmenting salt-anion intermediates into inorganic phases — explaining why the SEI grows inorganic-rich exactly at surface irregularities, while carbonate solvents remain largely intact. The authors derive electrolyte-design rules that favour compact inorganic nucleation at defects and suppress dendrite-promoting protrusions.
Relevance to DENG.Group
Site-specific interface chemistry that Umang Agarwal can mirror for electrode/electrolyte heterointerfaces: it shows how defect geometry alone dictates SEI composition — a mechanism worth testing in our interface models, and useful context for Yan Li's halide-degradation surfaces.
7. Sulfide-free halide interlayer for stable Li-In anodes and halide solid electrolytes in all-solid-state lithium batteries¶
Source: Journal of Energy Chemistry (10.1016/j.jechem.2026.07.052) · 📅 2026-11-01 · ↗ Open paper
Reports a sulfide-free halide interlayer, Li2.5Y0.5Hf0.5Cl6, that is both electrochemically stable against Li-In anodes and chemically compatible with halide electrolytes — avoiding the interfacial degradation that sulfide interlayers cause in halide stacks. Li-In symmetric cells reach a critical current density of 10 mA/cm2, and Li-In|LiTaInCl|NCM811 full cells retain 96.4% capacity over 300 cycles.
Relevance to DENG.Group
Direct input for Yan Li and Mengke Li's halide degradation and transport projects: the Hf-substitution strategy and its interfacial thermodynamics give concrete composition candidates for modelling reduction stability at halide/anode interfaces.
Thin-Film Growth & Pattern Formation¶
8. Multifractality in sputtered crystalline and amorphous transition metal nitride thin films¶
Source: arXiv:2611.10427 · 📅 2026-11-07 · ↗ Open paper
Presents evidence for multifractal growth in sputtered TiN and NbN thin films, with dendrite- and snowflake-like morphologies observed in crystalline films on both crystalline and amorphous substrates and in amorphous films on crystalline substrates. Two-dimensional multifractal detrended fluctuation analysis of SEM images shows fractal dimensions controlled by processing parameters and film-substrate interface conditions.
Relevance to DENG.Group
Relevant to Jingyuan Lin's thin-film growth kinetics: multifractal surface analysis is a quantitative handle for comparing growth-model morphologies (including kMC) against experiment. The dendritic pattern language also resonates with Shoutong Jin's phase-field dendrite work.