Skip to content

Research Digest — 2026-11-05

Halide Solid Electrolytes & Interfaces

1. Turning Insulators into Accelerators: Deciphering the Interfacial Ionic Conductivity Boost in Composite Solid-State Electrolytes through Machine Learning Molecular Dynamics Simulations

Source: ACS Nano (10.1021/acsnano.6c16367) · 📅 2026-11-02 · ↗ Open paper

Machine-learned force field simulations resolve why adding ionically insulating ZrO2 nanoparticles boosts the conductivity of Li2ZrCl6 halide electrolytes. The atomistic origin of the interfacial conductivity enhancement in halide/ZrO2 heterostructures is mapped across spatiotemporal scales inaccessible to plain DFT-MD.

Relevance to DENG.Group

Directly at the intersection of our halide electrolyte work and Yanhao Deng's ML potential development: the same neuroevolution-trained force-field route could be applied to our Li/Na halide interface models, and it gives Yan Li's halide degradation project a concrete atomistic picture of how insulating nanoinclusions modify interfacial transport. Also relevant to Naibing Wu's composite electrolyte simulations.

ML Interatomic Potentials

2. Fine-Tuning Universal Machine Learning Potentials Enables Efficient Computation of Li Electrode Potentials in Battery Electrolytes

Source: Journal of the American Chemical Society (10.1021/jacs.6c12519) · 📅 2026-10-31 · ↗ Open paper

Jun Cheng's group combines a universal machine learning potential with data-efficient fine-tuning (~200 extra DFT calculations per electrolyte) to compute Li metal electrode potentials at ab initio accuracy, avoiding the cost of full AIMD thermodynamic sampling. They also connect the computed potentials to local Li+ solvation structure.

Relevance to DENG.Group

A practical workflow template for Yanhao Deng's ML potential project: fine-tuned universal MLIPs could replace AIMD in our electrolyte stability and interface screening, and the electrode-potential formalism is directly usable when we compare candidate halide or polymer electrolytes against Li and Na metal anodes.

3. When do machine-learning potentials need a configuration-dependent kernel? Learning long-range interactions with Fourier neural operators

Source: arXiv (2611.03678) · 📅 2026-10-21 · ↗ Open paper

The authors attach a Fourier neural operator to a MACE backbone to learn how latent charge-like source fields couple across the simulation cell, testing whether long-range interactions need a configuration-dependent kernel. Under inhomogeneous screening only the nonlinear FNO captures the physics, cutting force errors by 27-41% versus configuration-independent models, with benchmarks on liquid water, LLZO, and Au2/MgO(001).

Relevance to DENG.Group

Methodological guidance for Yanhao Deng's ML potential work on ion-conducting materials: LLZO is one of the benchmarks, and the message that heterogeneous screening (exactly our electrolyte/electrode interfaces) demands learned, configuration-dependent long-range coupling should inform how we add electrostatics to our solid-electrolyte potentials.

4. Agentic, Autonomous Design Evolution of Machine-Learned Interatomic Potentials

Source: Precision Chemistry (10.1021/prechem.6c00099) · 📅 2026-10-29 · ↗ Open paper

AutoResearch-MLIP lets an LLM agent autonomously propose, implement, and evaluate MLIP architectures over 40 generations, selecting parent models by balancing exploration and exploitation. Evolved designs span message-passing, radial representations, and body-order terms, showing agents can run the MLIP design loop end-to-end.

Relevance to DENG.Group

For Yanhao Deng's potential-development workflow and the group's broader interest in agentic tooling: this is a concrete demonstration that our own MLIP architecture searches and benchmarking loops could be partially delegated to an agent, freeing student time for physics analysis rather than hyperparameter babysitting.

Composite & Polymer Electrolytes

5. Super-Toughness Self-Healing MOF-Based Composite Solid Electrolyte with Reaction-Transport Coupling Regulates Transportation of Ions and Polysulfides in High-Performance Solid-State Li-S Batteries

Source: Nano Letters (10.1021/acs.nanolett.6c04006) · 📅 2026-10-29 · ↗ Open paper

A self-healing composite solid electrolyte of PVDF-HFP, graphene oxide, UiO-66, and boronic-ester cross-links reaches 1.7e-4 S/cm ionic conductivity, 335% elongation at break, and 9.7 N puncture resistance, while nanoconfinement suppresses the polysulfide shuttle in Li-S cells. The dynamic cross-linking network couples mechanical recovery with ion and polysulfide transport control.

Relevance to DENG.Group

Directly relevant to Naibing Wu's solid polymer electrolyte and composite simulation work: the UiO-66/polymer interface is exactly the kind of filler-polymer boundary our MD models probe, and the reported conductivity-mechanics trade-off gives quantitative targets for our composite electrolyte design hypotheses.

Sulfide & Superionic Electrolytes

6. Strain-Glass Dynamics in the Superionic Conductor Na\(_3\)PS\(_4\)

Source: arXiv (2611.04547) · 📅 2026-11-03 · ↗ Open paper

Bragg X-ray photon correlation spectroscopy on a single ball-milled Na3PS4 particle shows the cubic superionic phase behaves as a strain glass, with stress-driven domain-wall motion following scale-free avalanche statistics. Host-lattice dynamics occur on timescales comparable to ionic diffusion, meaning the static-lattice assumption breaks down exactly where transport matters; the tetragonal polymorph relaxes 50x more slowly.

Relevance to DENG.Group

Speaks to a core assumption in our sulfide electrolyte modeling (Yanhao Deng's potentials, Tzu-Chen Liu's amorphous-electrolyte and kMC work): if the host framework is not static on transport timescales, our potential fits and hopping barriers should explicitly include lattice dynamics, and the strain-glass picture offers a framework for how processing-induced disorder enhances conductivity.

Thin-Film Growth & Interfaces

7. Capturing the Hidden Short-Range Order in Amorphous Oxide Interfaces That Directs Quasi-Epitaxial Thin-Film Growth

Source: ACS Nano (10.1021/acsnano.6c13264) · 📅 2026-10-30 · ↗ Open paper

Combining 4D-STEM, fluctuation electron microscopy, and multislice ptychography reveals that nominally amorphous Ta2O5 interlayers carry hidden short-range order that templates quasi-epitaxial Ru thin-film growth. The buried structural motifs direct film texture and defect formation despite the absence of a crystalline substrate surface.

Relevance to DENG.Group

For Jingyuan Lin's thin-film growth kinetics modelling: this provides experimental evidence that growth templates need not be crystalline, suggesting kMC growth models should encode the amorphous template's short-range order rather than a flat periodic substrate; the amorphous-SRO characterization also resonates with Tzu-Chen Liu's amorphous electrolyte structure work.

Metal-Organic Frameworks

Source: Journal of the American Chemical Society (10.1021/jacs.6c16378) · 📅 2026-11-02 · ↗ Open paper

ZJNU-600, a Hopf-link MOF with two mechanically interlocked sql nets and thiocyanate hinge linkers, replaces thermodynamic selectivity with diffusion gating: C2H2 and CO2 adsorb strongly but diffuse slowly, enabling one-step polymer-grade C2H4 purification plus inverse C2H2/CO2 separation. The adaptive framework's gating kinetics, not binding energies, set the separation outcome.

Relevance to DENG.Group

A design principle for Jason Li's MOF adsorption and gas-kinetics projects: kinetically gated separations mean transport barriers and framework flexibility, not just adsorption enthalpies, must be in our GCMC/kMC workflow - a concrete case where computing diffusion coefficients through a flexible pore is the decisive quantity.