OptoGPT: 10M+ Multilayer Thin Film Designs for Optical Inverse Modeling

A collection of between 10 million and 100 million records for multilayer thin film inverse design, created by Taigao Ma for the OptoGPT foundation model in 2024. It consists of multiple CSV partitions designed to train transformer-based models to predict optical properties and structural configurations. The data supports the research published in 'OptoGPT: A Foundation Model for Multilayer Thin Film Inverse Design'.

Use Cases

• Training transformer models for inverse design of multilayer thin films
• Predicting optical responses from structural sequences provided in the CSV files
• Benchmarking foundation models in the domain of materials science and nanophotonics

Strengths

• Scale of 10M to 100M records for specialized optical design
• Partitioned into multiple CSV files for compatibility with Dask and Polars
• Directly linked to a peer-reviewed foundation model for materials science

Limitations

• Column headers and units are not explicitly defined in the dataset metadata
• High domain specificity requires knowledge of photonics and thin-film physics

Provenance

Source

Opto-Electronic Advances (2024) publication by Taigao Ma et al.

Collection Method

Likely synthetic generation via optical simulations for thin-film structures

Freshness

Last updated June 2025; corresponds to 2024 research publication.