Skip to content
Apex
All recipes
Materials

Screen a million candidate compositions for a target property

DFT + ML-potential active-learning screen over 10⁶ candidate compositions for high ionic conductivity at room temperature. Top-100 candidates flagged for synthesis.

Artefact returned
Ranked composition list · uncertainty · synthesis routes
Typical runtime
~10 h wall · 2,400 CU
Cost
≈ $7.7k per screen + Pro seat
github.com/apex-dev/cookbookAPI surface
01 · The recipe

The expanded snippet.

The version that lands in production — with the annotations a colleague would want when picking up the work. Copy-paste, swap in your inputs, run.

from apex import Aether

simos = Aether(api_key=os.environ["APEX_API_KEY"])

# DFT + ML-potential active-learning screen over a composition space. Returns
# a ranked list of top candidates with predicted property, uncertainty, and a
# proposed synthesis route — ready for wet-lab confirmation.
screen = simos.materials_screen(
    target_property="ionic_conductivity",
    target_value={"min_S_per_cm": 1e-2, "temp_K": 300},
    composition_space="Li-La-Zr-O + dopants(Mg, Al, Ga, In)",
    method="dft+ml_potential+active_learning",
    constraints={
        "max_formation_energy_eV": 0.05,
        "stable_phase": True,
        "band_gap_min_eV": 4.0,
    },
    n_candidates=1_000_000,
    n_to_surface=100,
    propose_synthesis=True,
    confidence_floor=0.85,
)

for c in screen.top():
    print(
        f"{c.composition:24s}  "
        f"sigma={c.predicted_conductivity:.2e} S/cm  "
        f"ci=[{c.ci_low:.2e}, {c.ci_high:.2e}]  "
        f"route={c.synthesis_route_short}"
    )
02 · Reproduction notes

The numbers in this recipe are real.

What we ran. What we used for inputs. Where the reproduction kit lives, and what reproduces within published bounds.

  • 01Composition space: any chemical formula with dopant specification. The cookbook repo ships a small open garnet example.
  • 02Method: dft alone, ml-potential alone, or the active-learning combination. The combination is the default.
  • 03Reproduces published screening methodology against the literature panel.
  • 04Wall time scales with composition space × method; estimate uses 10⁶ candidates with the active-learning method.
03 · How to adapt this

Make it your workload.

The recipe is a starting point. The four variations below are the ones we get asked about most. Anything else, write to docs@apexworldlabs.com.

  • Different target: any property the runtime knows about. Custom targets via register_property().
  • Tighter constraints: tighten the band gap or formation energy and the screen returns fewer, higher-confidence candidates.
  • Closed loop: pair this screen with the wet-lab cookbook (autonomous-lab-protocol). The screen surfaces; the lab confirms.
More recipes

Other workloads worth looking at.

Adapt this to your stack.

Swap in your geometry / target / repo and run it. If your workload is novel, send it and we will co-author the recipe with you.

Get an API keyAll recipes