Scenario

An optoelectronic or photovoltaic application needs a material in a target bandgap window. Screening candidates computationally — grounded in their crystal structure — narrows a large set to the few worth growing.

Worked example. A photovoltaic absorber needs a bandgap near 1.3 eV from a family of candidate compounds.

Agent workflow

1. Read the candidate compounds.
2. Crystal structure analysis — parse structures and validate symmetry / phase.
3. Bandgap semiconductor screen — estimate bandgaps and rank against the target window.
4. Short-list candidates in range.

Demo output

A ranked table of candidates with estimated bandgaps against the target window and the governing structural phase. The magic moment: a candidate outside the usual chemistry lands squarely in the target window and enters the short-list.

Deterministic vs LLM

• Deterministic — the bandgap screening and crystal-structure analysis.
• LLM — the material-selection narrative.

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

Limits

Screening-level estimates, not a substitute for measured optoelectronic characterization. It focuses the growth campaign; the fab confirms it.