TGMin-MLP: a new package for predicting the structures of the mono- and poly-nuclear clusters by the machine learning...

Nuclear power provides a sustainable energy with high efficiency and develops fast in the past decades. To process the nuclear wastes, the behaviors, structures, coordination number, and average radius of solvated actinide ions are required to be explored theoretically due to the high-level radioactivity. Herein, we developed the TGMin-MLP package and trained the machine learning potentials (MLP) for the mono-nuclear americium-water clusters. Then, the trained MLP was applied to searching the most stable structures of both mono-nuclear Am(H₂O)_n³⁺ (n = 1 ~ 20) and poly-nuclear Am_n(H₂O)₅₀³ⁿ⁺ (n = 1 ~ 6) actinide-water clusters. The results reveal the saturated coordination number of the Am_n³ⁿ⁺ core is 9, 16, 20, 23, 28 and 32, respectively. In addition, the average radius of solvated Am_n³ⁿ⁺ core is ranked as Am³⁺ (2.52 Å) < Am₂⁶⁺ (3.19 Å) < Am₃⁹⁺ (3.40 Å) < Am₄¹²⁺ (3.62 Å) < Am₅¹⁵⁺ (3.85 Å) < Am₆¹⁸⁺ (4.10 Å). This work trained the MLP and searched the most stable structures of clusters, providing new insights on the actinide solution chemistry, especially the structures, solvation number and average radius of solvated Am_n³ⁿ⁺ cores. Overall, the TGMin-MLP package and the findings on actinide-water clusters advance both computational actinide chemistry and practical nuclear waste treatment, opening new avenues for the high-throughput study of hydrated actinide clusters, polynuclear speciation, and solution-phase behavior of radioactive heavy elements.

We developed the TGMin-MLP package to search the most stable structures of clusters by using the well trained MLP, and investigate the coordination number and average radius of Am³⁺ ions both for the mono- and poly-nuclear americium-water clusters Am_n(H₂O)_m³ⁿ⁺. The results predict the possible structures, saturated solvation numbers and radius of Am_n(H₂O)_m³ⁿ⁺ clusters. The GM structures were also re-optimized at the DFT ZORA PBE0-D4/def2-TZVP (H, O)/SARC-ZORA-TZVP (Am) level of theory with dispersion, solvation and relativistic effects corrections by using the ORCA (v6.1.0) package.