USW
Materials ScienceCrystal DiscoveryHardSeed

Discover a stable ternary oxide for solid-state batteries

Propose ternary oxide compositions, relax structures, and rank by energy above hull.

Dr. Wei Zhang
Materials Discovery Lab
Scaling deep learning for materials discovery
Nature · 2023
~8 h
registered 2025-06-30
materialsstabilityscreening

End goal

Nominate ≥5 candidates with energy above hull < 30 meV/atom.

Overview

The agent enumerates candidate compositions, predicts and relaxes structures with a universal interatomic potential, and nominates thermodynamically stable, conductive candidates.

Tools allowed

1
MatGL / M3GNet·Terminal

Workflow

1-step protocol

Each step is verified against the scientist's targets. Open any simulation to test it live.

  1. 1

    Screen & relax

    Step 1 / 1

    Predict and relax candidate structures.

    Protocol

    1. aEnumerate compositions.
    2. bRelax with the ML potential.
    3. cRank by e-above-hull.

    Targets

    Stable candidates≥5count
    Hull accuracy≤25meV/atom MAE
    Expected output

    A ranked list of stable candidates.

    Simulations · click to test