Accelerated Science

Machine learning can help accelerate materials science across a variety of climate change applications by learning patterns in experimental or operational data in order to guide future experiments/operations. Examples include:


 * The design of solar fuels, i.e., synthetic fuels produced from sunlight or solar heat.
 * The design or characterization of battery conducting fluids.
 * The design of alternatives to cement.
 * The design of better CO2 sorbents, i.e., new chemical sorbents that either bind to atmospheric CO2 with either greater selectivity or have lower energy input requirements.

Background Readings

 * "Machine learning for molecular and materials science" (2018) : A summary of "recent progress in machine learning for the chemical sciences" and proposed future directions for machine learning's use in the "design, synthesis, characterization and application of molecules and materials."
 * "Materials discovery and design using machine learning" (2017) : A review of machine learning for materials science.

Data

 * The Materials Project: "[C]omputed information on known and predicted materials as well as powerful analysis tools to inspire and design novel materials." Available here.
 * Inorganic Crystal Structure Database: "[T]he world's largest database for completely identified inorganic crystal structures." Available here.
 * SciFinder: Chemical and materials science database (paid), available here.
 * "Concrete Compressive Strength": Dataset of concrete compressive strength available here via the UCI Machine Learning Repository.
 * Open Catalyst Project: Dataset of 1.2 million molecular relaxations with results from over 250 million DFT calculations, aimed towards the discovery of new catalysts for use in renewable energy storage. Available here.