Research Digest — 2026-09-21¶
Machine Learning Interatomic Potentials¶
1. Benchmarking Machine-Learning Interatomic Potentials for Dynamical Stability in Inorganic Semiconductor Nanocrystals: A CdSe Case Study¶
Source: arXiv · 📅 2026-09-14 · ↗ Open paper
Systematic benchmark of universal and fine-tuned MLIPs for capturing dynamical stability of inorganic semiconductor nanocrystals, using CdSe as the test case. Assesses how faithfully different potentials reproduce phonon spectra, thermal evolution, and the onset of structural instability relative to DFT references.
Relevance to DENG.Group
The dynamical-stability benchmarking protocol is a cheap pre-screen Yanhao Deng can apply before production MD with his MLIPs for solid electrolytes - unstable potentials quietly corrupt transport statistics, and this paper offers a practical validation checklist transferable to halide and polymer electrolyte force fields.
2. Orbital-Free DFT-Assisted Machine-Learned Molecular Dynamics for Electric-Field-Driven Ionic Transport¶
Source: arXiv · 📅 2026-09-03 · ↗ Open paper
Couples orbital-free DFT with machine-learned molecular dynamics to simulate ionic transport under explicitly applied electric fields, reaching length- and timescales beyond conventional ab initio MD while retaining an electronic-structure description of field-matter coupling.
Relevance to DENG.Group
Field-driven transport is the atomistic layer beneath Shoutong Jin's phase-field dendrite growth models (boundary conditions and flux definitions) and Umang Agarwal's electrolyte/electrode interface simulations - the orbital-free DFT + MLMD scheme suggests how to impose potential gradients in group simulations without full DFT cost.
Tools, Data & Computational Methods¶
3. The Roadmap of Inorganic Computational Materials Databases: Capabilities, Credibility, Coverage, and the Open Frontier¶
Source: arXiv · 📅 2026-09-17 · ↗ Open paper
A community-authored roadmap assessing inorganic computational materials databases on capabilities, credibility (provenance, verification, uncertainty quantification), coverage, and open frontiers - including their role as MLIP training reservoirs and the challenges of interoperability and long-term curation.
Relevance to DENG.Group
Group-wide reference: data credibility and provenance directly affect the quality of Yanhao Deng's MLIP training sets and the group's screening pipelines, and the curation/infrastructure discussion is useful background for Tim Pook in managing the cluster and the group's growing simulation data holdings.
4. Cracking the case: fluctuations enhance ductility in refractory alloys¶
Source: arXiv · 📅 2026-09-11 · ↗ Open paper
Combining large-scale machine-learning-potential simulations with micromechanical experiments shows that thermal fluctuations enhance ductility in refractory alloys: atomic shuffling around crack tips at finite temperature promotes crack blunting and suppresses brittle fracture that zero-temperature models predict.
Relevance to DENG.Group
The fluctuation-assisted fracture mechanism speaks directly to Cheng Peng's grain boundary work in solid electrolytes and Umang Agarwal's heterogeneous interface studies - ceramic ion conductors fail brittlely at GBs and interfaces, and temperature-dependent crack-tip mechanics is central to understanding dendrite-induced short circuits.