Scenario

A copolymer's properties hinge on comonomer ratio, and the composition space is large. Ranking compositions and predicting the resulting glass transition finds the window that meets the target without a long series of trial polymers.

Worked example. A packaging copolymer needs a Tg in a defined range for the right stiffness-and-toughness balance.

Agent workflow

1. Read the monomer options and the property target.
2. Copolymer composition ranking — rank comonomer ratios against the target.
3. Polymer Tg screen — predict Tg for the top compositions.
4. Short-list compositions that land in the property window.

Demo output

A ranked composition table with predicted Tg for each candidate and the ones inside the target window highlighted. The magic moment: a composition off the obvious ratio hits the Tg target with a better sustainability profile.

Deterministic vs LLM

• Deterministic — the composition ranking and Tg prediction.
• LLM — the composition-strategy narrative.

Every prediction is a validated calculation. The agent reads and routes; the engine decides.

Limits

Screening-level composition design, not a substitute for polymerization and measured Tg. It focuses the recipe; the lab confirms it.