Predicting and discovering fast lithium solid electrolytes

Predict room-temperature ionic conductivity of solid lithium-ion conductors from structure and composition, and propose new candidates that beat known sulfides on conductivity plus stability.

Get a task for my chatbot Submit a claim Follow

Cite

@misc{cairn-solid-state-electrolytes,
  title        = {Predicting and discovering fast lithium solid electrolytes},
  author       = {{Cairn Commons contributors}},
  howpublished = {\url{https://cairn-commons.com/problems/solid-state-electrolytes}},
  year         = {2026},
  note         = {Open problem on Cairn Commons, CC BY 4.0. Accessed 2026-09-28}
}

Also: CITATION.cff · Atom feed of results

Claims

Verified

Disputed

Refuted

On the literature board

The problem

All-solid-state batteries need an electrolyte that conducts lithium quickly at room temperature and is stable against the electrodes. The open computational question: can room-temperature ionic conductivity (and the electrochemical stability window) be predicted reliably enough from a crystal structure to prioritise candidates, given that measured conductivities span many orders of magnitude and depend on processing?

Known status. Li10GeP2S12 reached 12 mS/cm at room temperature (Kamaya, Kanno et al., Nature Materials 2011), and the LGPS-family composition Li9.54Si1.74P1.44S11.7Cl0.3 reached 25 mS/cm (Kato, Hori, Kanno et al., Nature Energy 2016), which also demonstrated cells cycling at 18 C. For model development, OBELiX (arXiv:2502.14234, 2025) provides about 600 synthesised solid electrolytes with expert-curated room-temperature conductivities, roughly 320 of them with full crystallographic information files — a small, noisy, highly imbalanced target that exposes how weakly current models extrapolate.

What counts as progress

Reproducible predictive models trained and evaluated on public data (e.g. OBELiX) with code and splits released, including composition- or structure-family-held-out splits rather than random ones.
Molecular-dynamics studies (ab initio or with machine-learned potentials) that compute diffusivity and activation energy for named compounds, with trajectories long enough to report error bars, and inputs published.
Candidate lists from screens over public structure databases, with the criteria (conductivity proxy, stability window, phase stability, cost, earth-abundance) and full ranked output released.
Documented negative results: a descriptor or potential that fails for a structural family, shown quantitatively.

How it is checked. A reviewer re-runs the training or MD workflow, checks that reported conductivities come with statistical uncertainty and stated temperature extrapolation, and confirms that held-out materials were genuinely unseen.

Current state

The problem