Mechanochemical disassembly pathways of self-assembled PdnL2n supramolecular architectures

This Zenodo record contains data associated with the paper “Mechanochemical disassembly pathways of self-assembled PdnL2n supramolecular architectures.” It includes the fine-tuned Universal Model for Atoms S (UMA-S-PdN) provided as a PyTorch checkpoint (.pt), as well as training/validation/test datasets in ASE LMDB format (.aselmdb). Example trajectories for the two systems studied in this work are provided in XYZ format (.xyz) and include every 100th frame from the original trajectories. Additionally, the zip archive xyz_files.zip contains the first and last frame of every MD trajectory. Minimal Python code is included to load the .aselmdb files. Two Jupyter notebooks reproduce the analyses presented in the paper: 01_kinetic_fitting.ipynb performs kinetic fitting, and 02_fragment_analysis.ipynb plots the time-resolved fragment decomposition of Pd12L24 during disassembly. The corresponding post-processed data is provided as CSV files containing rupture times for Pd2L4 and Pd12L24 cages across multiple pulling geometries and force ramps, as well as the time-resolved fragment composition during Pd12L24 disassembly. The source code of the MD simulation engine used to run the pulling simulations is included as minirauche.zip.