Research Digest — 2026-05-31¶
Defects & Interfaces / Dendrites¶
1. Electrochemical Corrosion Accompanies Dendrite Growth in Solid Electrolytes¶
Source: Nature (s41586-026-10279-z) · 📅 2026-03-25 · ↗ Open paper
Using operando birefringence microscopy, the authors directly measure stresses around growing Li dendrites in garnet LLZTO and show that dendrites can propagate at far lower stresses than the fracture stress of the solid electrolyte. As current densities and dendrite velocities increase, the measured stresses paradoxically decrease, revealing that dendrite propagation transitions from mechanically-driven fracture to an electrochemical corrosion mechanism at higher current densities. This two-stage process — mechanical cracking at low rates and electrochemical corrosion at high rates — fundamentally reframes the dendrite failure problem in solid-state batteries.
Relevance to DENG.Group
Directly relevant to Yanhao Deng's phase-field and computational work on dendrite growth. The discovery that dendrites propagate via electrochemical corrosion rather than purely mechanical fracture at practical current densities challenges existing phase-field models that focus only on mechanical driving forces. This finding should inform the group's future computational models of dendrite penetration, particularly the coupling between electrochemical and mechanical degradation modes.
2. Mechanically Driven Li Dendrite Penetration in Garnet Solid Electrolyte¶
Source: Nature (s41586-026-10415-9) · 📅 2026-04-22 · ↗ Open paper
Using cryogenic electron microscopy and micromechanical fracture modelling, the authors investigate both intergranular and transgranular fracture events driven by Li dendrites in LLZTO. No isolated Li nuclei were detected ahead of the dendrite tip by cryo-STEM, ruling out the electron-leakage nucleation hypothesis. Small crystal lattice rotations were observed only at the Li/LLZTO interface, indicating a nearly hydrostatic stress state within the dendrite interior. Based on the mechanically-driven mechanism, the authors propose a mechanics-informed strategy to redirect dendrite propagation through geometrically engineered voids in LLZTO.
Relevance to DENG.Group
Highly relevant to the group's dendrite and interface stability research. The cryo-EM evidence ruling out isolated Li nucleation ahead of the dendrite tip provides important validation for computational models of dendrite growth mechanisms. The proposed void-engineering strategy to redirect dendrites is a novel design concept that could be explored computationally using the group's phase-field and MLIP tools. The detailed fracture mode characterization (intergranular vs transgranular) also provides benchmarks for grain boundary modeling work.
Solid Electrolytes & Ion Transport¶
3. Superionic Composite Electrolytes with Continuously Perpendicular-Aligned Pathways for Pressure-Less All-Solid-State Lithium Batteries¶
Source: Nature Nanotechnology (s41565-025-02106-9) · 📅 2026-05-27 · ↗ Open paper
The authors engineer highly ionically conductive and flexible solid-state composite electrolytes by alternately stacking inorganic LixMyPS3 (M = Cd or Mn) nanosheets with lithium-containing polymer layers. The design creates continuously perpendicular-aligned superionic pathways that decouple ion conduction from mechanical flexibility, achieving high room-temperature conductivity without requiring external stack pressure. This approach eliminates the classic conductivity–flexibility trade-off in composite electrolytes by using superionic nanosheets as the primary ion conductor within a deformable polymer framework.
Relevance to DENG.Group
Relevant to the group's interest in solid electrolytes and composite designs. The perpendicular-aligned pathway architecture represents a new design principle for composite electrolytes that could potentially be adapted using halide SSE fillers instead of sulfide nanosheets. The pressure-less operation is especially notable for practical applications and could inform the group's thinking on electrode–electrolyte interface design. The concept of decoupling conductivity from mechanical properties through nanostructuring may inspire computational studies on ion transport in layered composite structures.
ML Interatomic Potentials & Workflows¶
4. Constructing Machine Learning Interatomic Potentials with Minimum Amount of Ab Initio Data¶
Source: npj Computational Materials (s41524-026-02023-y) · 📅 2026-03-17 · ↗ Open paper
Presents a uMLIP-based workflow (MACE-MD) that dramatically reduces the amount of new DFT data needed to construct reliable MLIPs for specific systems. By leveraging pretrained MACE foundation models and selectively adding only a small number of DFT calculations, the method achieves accurate MD simulations for solid-state electrolytes at a fraction of the conventional active-learning cost. Validated across three representative SSE systems, the pretrained-MACE simulations demonstrate performance comparable to fully trained potentials while requiring orders of magnitude less ab initio data.
Relevance to DENG.Group
Directly relevant to Yanhao Deng's MLIP development work. This paper provides a practical recipe for building system-specific MLIPs for halide and oxide electrolytes with minimal DFT cost by starting from pretrained MACE models. The group can immediately adopt this workflow to accelerate their MLIP development for new electrolyte chemistries. The demonstrated accuracy on SSE systems and the dramatic reduction in DFT data requirements could fundamentally change how the group approaches MLIP construction for battery materials.
5. Battery-Sim-Agent: Leveraging LLM-Agent for Inverse Battery Parameter Estimation¶
Source: arXiv:2605.29560 · 📅 2026-05-28 · ↗ Open paper
Introduces Battery-Sim-Agent, the first framework to deploy an LLM agent in a closed loop with a high-fidelity battery simulator (PyBaMM) for inverse parameter estimation. Instead of treating the simulator as a black-box optimizer, the agent interprets multi-modal feedback, forms physically-grounded hypotheses to explain discrepancies, and proposes structured parameter updates — mimicking a human scientist's workflow. On a systematic benchmark spanning diverse chemistries and conditions, the agent significantly outperforms Bayesian optimization baselines, and is further demonstrated on complex degradation fitting and real-world battery datasets.
Relevance to DENG.Group
Relevant to the group's computational workflow and battery modeling efforts. The LLM-agent approach to parameter estimation could be adapted by the group for automating the fitting of MLIPs, calibrating phase-field models, and extracting transport parameters from electrochemical data. The closed-loop reasoning paradigm — where the AI forms hypotheses about physical discrepancies — mirrors how the group currently debugs simulations manually. While focused on cell-level parameters rather than atomistic simulations, the framework's philosophy of physics-informed AI optimization is transferable to computational materials design.