The presence of a dataset that covers a parametric space of materials and process conditions in a process-consistent manner is essential for the realization of catalyst informatics. Here, an important piece of progress is demonstrated for the oxidative coupling of methane. A high-throughput screening instrument is developed for enabling an automatic performance evaluation of 20 catalysts in 216 reaction conditions. This affords an oxidative coupling of methane dataset comprised of 12 708 data points for 59 catalysts in three successive operations. Based on a variety of data visualization analysis, important insights into catalysis and catalyst design are successfully extracted. In particular, the simultaneous optimization of the catalyst and reactor design is found to be essential for improving the C2 yield. The consistent dataset allows the accurate prediction of the C2 yield with the aid of nonlinear supervised machine learning.

Catalysis research is on the verge of experiencing a paradigm shift regarding how catalysts are designed and characterized due to the rise of catalyst informatics. The details of catalyst informatics are reviewed where the following three key concepts are proposed: catalyst data, catalyst data to catalyst design via data science, and catalyst platform. Additionally, progress and opportunities within catalyst informatics are explored and introduced. If the field of catalyst informatics grows in the appropriate manner, the methods and approaches taken for catalyst design would be fundamentally altered, leading towards great advancement within catalysis research.

Combinatorial catalyst design is hardly generalizable, and the empirical aspect of the research has biased the literature data toward accidentally found combinations. Here, 300 quaternary solid catalysts are randomly sampled from a materials space consisting of 36,540 catalysts, and their performance in the oxidative coupling of methane is evaluated by a high-throughput screening instrument. The obtained bias-free data set is analyzed to withdraw catalyst design guidelines. Even with random sampling, 51 catalysts out of the 300 provide a C 2 yield sufficiently superior to the noncatalytic free radical process. Data analysis suggests the significance of choosing synergistic combinations, and such combinations could be generalized based on the group in the periodic table. Decision tree classification is successfully implemented to facilitate efficient sampling of quaternary catalysts toward a better C 2 yield.

Undiscovered perovskite materials for applications in capturing solar lights are explored through the implementation of data science. In particular, 15000 perovskite materials data is analyzed where visualization of the data reveals hidden trends and clustering of data. Random forest classification within machine learning is used in order to predict the band gap of perovskite materials where 18 physical descriptors are revealed to determine the band gap. With trained random forest, 9328 perovskite materials with potential for applications in solar cell materials are predicted. The selected Li and Na based perovskite materials within predicted 9328 perovskite materials are evaluated with first principle calculations where 11 undiscovered Li(Na) based perovskite materials fall into the ideal band gap and formation energy ranges for solar cell applications. Thus, the implementation of data science accelerates the discovery of hidden perovskite materials and the approach can be applied to the materials science in general for searching undiscovered materials.

Determining the manner in which crystal structures are formed is considered a great mystery within materials science. Potential solutions have the possibility to be uncovered by revealing hidden patterns within the material data. Data science is therefore implemented in order to link the material data to the crystal structure. In particular, unsupervised and supervised machine learning techniques are used where the Gaussian mixture model is employed in order to understand the data structure of the materials database while random forest classification is used to predict the crystal structure. As a result, the unsupervised and supervised machine learning techniques reveal descriptors for determining the crystal structure via the materials database. In addition, predicting atomic combinations from the crystal structure is also achieved using a trained machine where the first-principles calculations confirm the stability of predicted materials. Thus, one can consider that the estimation of the crystal structure can be achieved in principle via the combination of material data and machine learning, thereby leading to the advancement of crystal structure prediction.