Research Digest — 2026-08-31¶
ML Interatomic Potentials¶
1. Benchmarking of Fast and Interpretable UF Machine Learning Potentials¶
Source: arXiv (2608.27277) · 📅 2026-08-27 · ↗ Open paper
The Hennig group benchmarks the ultra-fast force field (UF3) potential, a linear-regression MLIP with cubic B-spline two- and three-body terms, against GAP, MTP, NNP, and qSNAP. UF3 reaches comparable accuracy and predicts melting points of simple metals including Li within ~6%, but fails for covalent/strongly angular systems (Si, Ge) due to its three-body truncation. A key selling point is interpretability: the spline formulation allows direct visualization of learned interactions to spot unphysical behavior that black-box potentials hide.
Relevance to DENG.Group
Yanhao Deng: a useful fast, interpretable baseline when choosing MLIP architectures for solid electrolyte/electrode workflows; the Li melting benchmark and failure-mode analysis (when 3-body truncation breaks down) directly inform potential selection and validation strategy for the group's MLIP projects.
AI-Driven Solid Electrolyte Discovery¶
2. A Hierarchical Synergistic Deep Learning Framework Integrating Composition, Structure, and Ionic Transport for Solid-State Electrolyte Discovery¶
Source: arXiv (2608.25592) · 📅 2026-08-26 · ↗ Open paper
A four-module deep-learning pipeline (graph-DCNN thermodynamic screening, multi-fidelity DenseGNN property evaluation, MatterSim transport pre-assessment, and DeePMD kinetic validation) screens 30.4 million ICSD/Alexandria-derived candidates for inorganic solid electrolytes. It yields 97 candidates with 0.1-59 mS/cm room-temperature conductivity, 94 of them halides, with 76 falling in experimentally validated high-conductivity structural regions. Analysis shows Li+ jump-network connectivity, not the sheer number of geometric Li sites, is the core determinant of conductivity, and suggests an intrinsic performance ceiling for oxide frameworks.
Relevance to DENG.Group
Group-wide: the halide-dominated hit list and structure-transport descriptors are directly relevant to Yan Li and Mengke Li's halide electrolyte transport/degradation studies, while the DeePMD-based kinetic validation mirrors Yanhao Deng's MLIP pipeline; the jump-network connectivity metric is a useful screening descriptor for Naibing Wu's composite electrolyte transport work.
Grain Boundaries & Interfaces¶
3. Grain Boundary Phase Transitions Enable Diffusionless Climb of Disconnections¶
Source: arXiv (2608.27314) · 📅 2026-08-27 · ↗ Open paper
Combining bicrystallography and MD of shear dislocation loops impinging on an Al symmetric tilt grain boundary, the authors show dislocation absorption launches a mobile extrinsic disconnection that climbs conservatively along the interface with no long-range point-defect transport. The climb is powered by localized GB phase transformations: cooperative atomic rearrangements cycle the boundary through successive metastable microstates. This establishes a direct coupling between lattice dislocations and GB phase evolution, distinct from classical vacancy-mediated climb.
Relevance to DENG.Group
Cheng Peng: GB microstate transitions and disconnection-GB coupling are core mechanics for understanding GB-mediated degradation and transport in solid electrolytes; the MD methodology for tracking GB phase evolution is directly transferable to GB simulations in the group.
4. Grain-Boundary Premelting in High-Entropy Transition Metal Carbides¶
Source: arXiv (2608.27273) · 📅 2026-08-27 · ↗ Open paper
MC/MD simulations with the universal MACE-OMAT-0 MLIP probe GB segregation and thermal disordering in high-entropy transition-metal carbides. Group-VI elements and Zr segregate strongly to boundaries (~45 at.%), and species-resolved Lindemann-index analysis shows Cr-rich boundaries premelt first (~1390 C) while grain interiors stay solid. The chemically random reference carbide resists boundary disordering ~60 C longer, highlighting how interfacial chemistry controls premelting onset and carbon-driven GB mobility.
Relevance to DENG.Group
Cheng Peng: a template for using universal MLIPs (MACE-OMAT-0) with MC sampling to study chemistry-dependent GB disordering — directly applicable to GB premelting/softening questions in solid electrolytes; also a methods reference for Yanhao Deng on universal-potential transferability at segregation profiles.
5. When chemical potential continuity fails: kinetic interface models for hydrogen isotope transport¶
Source: arXiv (2608.26231) · 📅 2026-08-26 · ↗ Open paper
The authors replace the standard local-thermodynamic-equilibrium (chemical potential continuity) boundary condition at material interfaces with reversible mass-action reaction channels, implemented in the festim finite-element code, with detailed balance tying rate constants to thermodynamic data. For hydrogen transport across a nickel/molten-salt interface, kinetic partitioning between molecular and fluoride carriers makes the apparent interfacial law drift between Sieverts and Henry regimes, and LTE underestimates steady permeation flux. Damköhler numbers and branching ratios delimit when each regime applies, and per-species LTE becomes ill-posed with multiple isotopes and carriers.
Relevance to DENG.Group
Umang Agarwal: a rigorous, transferable framework for interface boundary conditions when local equilibrium fails — directly applicable to Li transfer kinetics across heterogeneous electrolyte/electrode and electrolyte/interphase interfaces in the group's solid-state battery models; the Damköhler analysis offers a clean way to justify (or reject) equilibrium boundary conditions in continuum SSLB models.
Phase-Field & Microstructure Evolution¶
6. A novel Grand-Potential Phase-Field Lattice-Boltzmann model for multi-phase solidification with convection¶
Source: arXiv (2608.26908) · 📅 2026-08-27 · ↗ Open paper
A thermodynamically consistent coupling of the grand-potential phase-field method with Lattice-Boltzmann flow simulation enforces no-slip conditions at evolving solid-liquid interfaces in multi-phase, multi-component solidification. Demonstrations on dendritic and eutectic growth under natural convection show convection alters solute segregation, destabilizes growth fronts, and produces oscillatory growth modes. The unified formulation cleanly handles the interface-solute-flow triad that plagues ad hoc couplings.
Relevance to DENG.Group
Shoutong Jin: the grand-potential + LBM coupling is a methodological upgrade for his dendrite-growth phase-field work — the same machinery extends to electrodeposition where electrolyte convection and dendritic Li growth interact; also a reference implementation for thermodynamically consistent flow-coupled phase-field.
Battery Systems Modeling¶
7. Battery Recycling: Mechanistic Modelling of LiCoO2 Leaching with Coupled Diffusion-Reaction Kinetics and Film Passivation¶
Source: arXiv (2608.25492) · 📅 2026-08-26 · ↗ Open paper
A full mechanistic model of acidic-reductive LiCoO2 leaching tracks particle conversion, shrinking radius, acid/H2O2 consumption, and formation-dissolution of a Co3O4 passivation film, validated against literature Li/Co recovery data across multiple acid and reductant concentrations. Film passivation reproduces the reduced recovery observed without H2O2, and the complete equation/parameter set is published for reuse. The framework is designed to extend to other cathode chemistries and leaching systems.
Relevance to DENG.Group
Broader group interest: the coupled diffusion-reaction + passivation-film modeling mirrors the physics of interphase growth and ion-transport limitation the group studies for SEIs in solid-state batteries, and adds a battery-sustainability angle for grant framing; useful reference for Naibing Wu on mechanistic continuum model construction and validation.