Fine-Tuned Machine-Learning Potential for Accurate Description of Mn$_x$O$_y$H$_z$ Clusters on Cobalt Surfaces

In this document, we present a fine-tuned version of the universal machine-learning po- tential (uMLP) CHGNet to accurately model MnxOyHz clusters on fcc-Co Surfaces. The pdf file explaining the procedure is divided in three sections: the first specifies the density functional theory (DFT) settings employed for the single-point (SP) calculations used for structures labeling, the second is related to the creation of the structure database and the third to the training procedure. The structural database file used for the fine-tuning procedure is provided as both an ase .db and .json file.