Generative discovery, symmetry-aware membership screening, and device-level evaluation of six-member-ring MoS2 polytypes
- 1University of Science and Technology Beijing Tianjin College
Abstract
Six-membered-ring (SMR) inorganic materials were recently unified into a class, but the class-level perspective lacks two enabling tools: an operational test of “SMR-ness” and a path from newly generated crystals to device-level figures of merit. Here we couple conditional generative modelling (MatterGen; Mo–S, Ti–S, W–Se; 512 cells per system) to a novelty/stability funnel, a symmetry-aware membership checker that casts the SMR definition as three calibrated criteria (6/6 positives, 6/6 trap negatives), full DFT relaxation with composition-anchored hull analysis, and a contract-based device evaluation injecting literature-anchored OCV and diffusivity into an electrochemical cell model. The campaign maps the AA-stacked polytype manifold across five TMD chemistries, confirming 61 candidates (12 MoS2, 7 TiS2, 18 WSe2, 16 MoSe2, 8 NbS2; e.g. the MoS2 series P6̄m2, c = 7.1–40.8 Å, Ehull = 0.019–0.022 eV atom−1), 56 of them novel against every queried database. A dedicated spin cross-check (ferromagnetic, antiferromagnetic and unpolarised initialisations) collapses to zero moments in every tested channel, and a three-probe stability screen (phonons, annealing dynamics, hull distance) shows the layered frameworks robust while the AA registry is metastable toward layer sliding, sharpening the synthesis list into kinetically stabilisable targets with two registry-robust entries. Two methodological findings transfer to any SMR screening effort: minimum-ring statistics collapse on TMD coordination (four-rings swallow six-rings), and membership is adjudicable only after relaxation. At the device end, material classes are discriminated by 0.639 V and polytypes by computed Li barriers (a 1.6× 10 8 diffusivity ratio separates fast from frozen at 0.5C); every number is traceable to a versioned experiment record.
Related
- A Full-Spectrum Generative Lead Discovery (FSGLD) pipeline via DRUG-GAN: A multiscale method for drug-like/target-specific compound library generation
- From Generative Molecular Design to Agentic Artificial Intelligence in Drug Discovery
- Microscale high-throughput screening-assisted discovery of a photo/copper dual catalytic system for benzylic C(sp2)–C(sp3) Suzuki–Miyaura coupling: data mining and mechanistic insights
- A Pharmacophore-Based Virtual Screening Approach for the Discovery of Novel P2Y ₁₂ Receptor Ligands
- A New Perspective in Chemical Probe Discovery: Expanding Virtual Screening into Metal Complex Chemical Space
BibTeX
@misc{chin2026generative,
title = {{Generative discovery, symmetry-aware membership screening, and device-level evaluation of six-member-ring MoS2 polytypes}},
author = {Michael Chin},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15010769/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010769/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Generative discovery, symmetry-aware membership screening, and device-level evaluation of six-member-ring MoS2 polytypes
AU - Michael Chin
PY - 2026
DA - 2026/10/21
PB - ChemRxiv
DO - 10.26434/chemrxiv.15010769/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010769/v1
AB - Six-membered-ring (SMR) inorganic materials were recently unified into a class, but the class-level perspective lacks two enabling tools: an operational test of “SMR-ness” and a path from newly generated crystals to device-level figures of merit. Here we couple conditional generative modelling (MatterGen; Mo–S, Ti–S, W–Se; 512 cells per system) to a novelty/stability funnel, a symmetry-aware membership checker that casts the SMR definition as three calibrated criteria (6/6 positives, 6/6 trap negatives), full DFT relaxation with composition-anchored hull analysis, and a contract-based device evaluation injecting literature-anchored OCV and diffusivity into an electrochemical cell model. The campaign maps the AA-stacked polytype manifold across five TMD chemistries, confirming 61 candidates (12 MoS2, 7 TiS2, 18 WSe2, 16 MoSe2, 8 NbS2; e.g. the MoS2 series P6̄m2, c = 7.1–40.8 Å, Ehull = 0.019–0.022 eV atom−1), 56 of them novel against every queried database. A dedicated spin cross-check (ferromagnetic, antiferromagnetic and unpolarised initialisations) collapses to zero moments in every tested channel, and a three-probe stability screen (phonons, annealing dynamics, hull distance) shows the layered frameworks robust while the AA registry is metastable toward layer sliding, sharpening the synthesis list into kinetically stabilisable targets with two registry-robust entries. Two methodological findings transfer to any SMR screening effort: minimum-ring statistics collapse on TMD coordination (four-rings swallow six-rings), and membership is adjudicable only after relaxation. At the device end, material classes are discriminated by 0.639 V and polytypes by computed Li barriers (a 1.6× 10 8 diffusivity ratio separates fast from frozen at 0.5C); every number is traceable to a versioned experiment record.
ER -