Simple descriptor derived from symbolic regression accelerating the discovery of new perovskite catalysts

Abstract

Symbolic regression (SR) is an approach of interpretable machine learning for building mathematical formulas that best fit certain datasets. In this work, SR is used to guide the design of new oxide perovskite catalysts with improved oxygen evolution reaction (OER) activities. A simple descriptor, μ/t, where μ and t are the octahedral and tolerance factors, respectively, is identified, which accelerates the discovery of a series of new oxide perovskite catalysts with improved OER activity. We successfully synthesise five new oxide perovskites and characterise their OER activities. Remarkably, four of them, Cs 0.4 La 0.6 Mn 0.25 Co 0.75 O 3 , Cs 0.3 La 0.7 NiO 3 , SrNi 0.75 Co 0.25 O 3 , and Sr 0.25 Ba 0.75 NiO 3 , are among the oxide perovskite catalysts with the highest intrinsic activities. Our results demonstrate the potential of SR for accelerating the data-driven design and discovery of new materials with improved properties.

High‐Performance Perovskite Composite Electrocatalysts Enabled by Controllable Interface Engineering

“…LaSr2.7 gives a η 10 as low as 271 mV, about 90 and 130 mV smaller than the respective value for RP LaSr3 and SP LaSr2. Notably, this exceptional activity places LaSr2.7 among the best-in-class perovskite-based OER electrocatalysts reported to date, including both single-phase [5,8,10,[16][17][18][19][20][21][22] and dual-phase [23][24][25] materials (Figure S7 and Table S6, Supporting Information). In addition to catalytic efficiency, we show that the best-performing sample LaSr2.7 could operate stably at 10 mA − cm geo 2 for 200 h when supported on a carbon paper substrate (Figure S8, Supporting Information), implying its potential for practical use.…”

Artificial Intelligence to Power the Future of Materials Science and Engineering

“…Different chromosomes pass through mutation and heredity, and gradually iterate until the best form of function and parameter set for a given problem is found . GPSR is suitable for the field of material research with little prior knowledge and unclear relationship between related variables, such as the magic angle in graphene, the viscosity of normal hydrogen, and the search for descriptors of perovskite stability …”

Analogical discovery of disordered perovskite oxides by crystal structure information hidden in unsupervised material fingerprints

“…Nevertheless, adopting high-throughput ab initio methods 17 to model disordered solid solutions with mixed siteoccupations is limited due to extreme computational power demanded by large supercells and, several random structure generations and optimizations. Alternatively, machine learning (ML) approaches underpinned by experimental data have been applied to rapidly predict complex functional materials, eventually synthesized to exhibit excellent properties [18][19][20][21][22][23] . Meanwhile, the complexities brought by compositional disorder such as random atom occupancy of the disordered site, non-stoichiometry and the presence of fractional occupations in the average unit cell have hindered the exploration of the complete composition space for target-driven perovskites discovery.…”

“…Meanwhile, the complexities brought by compositional disorder such as random atom occupancy of the disordered site, non-stoichiometry and the presence of fractional occupations in the average unit cell have hindered the exploration of the complete composition space for target-driven perovskites discovery. Current supervised ML studies are limited to a minute fraction of disordered perovskites 19,20,24 , calling for a general strategy to inversely search the undiscovered territory. Fortunately, unsupervised learning techniques can extract hidden knowledge from raw input features and embed those information in a compressed numerical latent vector 25 .…”