Machine-learning-based flow-synthesis of platinum-group-metal high-entropy alloy nanoparticles

Synthesis of size-controlled and compositionally homogeneous alloy nanoparticles (NPs) is challenging but important, because the particle size and elemental distribution significantly affect NP performance. This is particularly important for high-entropy alloy (HEA) NPs with highly complex compositions and structures. In this study, we established a method for synthesizing size-controlled and homogeneously alloyed RuRhPdIrPt HEA NPs by integrating flow synthesis with interpretable machine learning. The alloy homogeneity was evaluated using compositions derived from X-ray fluorescence and lattice constants from powder X-ray diffraction patterns as target variables, revealing the threshold of reaction temperature required for homogeneous alloying through decision tree classification. Partial least squares regression enabled particle-size predictions via the synthesis parameters (cross-validated R 2 value of 0.73) and the quantification of parameter contributions. This study translates empirical knowledge into interpretable mathematical models, thereby providing quantitative guidelines for the reproducible synthesis of complex HEA NPs.