An efficient algorithm to accelerate the discovery of complex material formulations.

Area of Science:

• Materials Science
• Computational Materials Discovery
• Chemical Engineering

Background:

• Identifying optimal multicomponent material formulations with desired properties is a significant challenge in materials discovery.
• High-dimensional composition spaces require efficient sampling and property measurement to identify effective formulations.
• Constraints such as mass fraction must be considered during the materials discovery process.

Purpose of the Study:

• To introduce random sampling-high dimensional model representation (RS-HDMR) as an algorithmic tool for nonlinear multivariate problems in materials discovery.
• To demonstrate the efficiency and accuracy of RS-HDMR in interpolating properties over sampled materials.
• To evaluate the applicability of RS-HDMR to existing materials databases.

Main Methods:

• The study introduces and simulates the application of random sampling-high dimensional model representation (RS-HDMR).
• RS-HDMR is utilized for interpolating material properties within high-dimensional composition spaces.
• The technique was tested on existing databases for a seven-component phosphor and a four-component deNO(x) catalyst.

Main Results:

• Simulations show that RS-HDMR can efficiently and accurately interpolate material properties.
• The number of sampled materials required for a given prediction accuracy is largely independent of the number of components.
• RS-HDMR was successfully applied to real-world datasets, demonstrating its practical utility.

Conclusions:

• RS-HDMR is a powerful algorithmic tool for accelerating the discovery of complex multicomponent materials.
• The method offers efficient property prediction and is scalable with respect to the number of components.
• RS-HDMR facilitates guided, iterative materials design through random sampling and property observation.