Learning Ligand-Controlled Hot Electron Relaxation in Metal-Rich CdSe Quantum Dots

DATA AND REPRODUCIBILITY ARCHIVE Associated manuscript:"Learning Ligand-Controlled Hot Electron Relaxation in Metal-Rich CdSe Quantum Dots" This archive contains the principal structural data, computational inputs, molecular-dynamics trajectories, nonadiabatic molecular dynamics data, processed electronic data, and ALIGNN/FNA data used in the manuscript. The archive is intended to provide the data required to reproduce or verify the principal calculations and analyses while avoiding unnecessary large restart files and redundant intermediate outputs. ======================================================================DIRECTORY CONTENTS====================================================================== 01_OPTIMIZED_STRUCTURES/------------------------Contains the optimized structures used in this work. The directory includes the optimized Cd35Se20X30 quantum dots(X = F, Cl, Br, and I) used for the principal calculations. It also contains: Cl_configurational_check/ Contains the reference Cd35Se20Cl30 structure together with 10 additional configurationally distinct Cd35Se20Cl30 structures. generation_scripts/ Scripts used to generate symmetry-distinct bridging-Cl configurations, select structurally diverse candidates, and check atomic coordination. additional_configurations/ Final DFT-optimized structures for the 10 additional Cd35Se20Cl30 configurations. reference_structure/ Reference Cd35Se20Cl30 structure used in the main calculations. Larger_QD_Cd84Se56Cl56/ Contains five optimized Cd84Se56Cl56 structures used for the larger-quantum-dot structural check. 02_VASP_INPUTS/---------------Contains the principal VASP input files used for the electronic-structureand ab initio molecular dynamics calculations. The files document the main computational parameters used in thecalculations. Large restart and wavefunction files such as WAVECAR and CHGCAR arenot included. 03_AIMD_TRAJECTORIES/---------------------Contains the ab initio molecular dynamics (AIMD) trajectories used forthe electronic-structure, energy-gap, nonadiabatic-coupling, andsubsequent nonadiabatic molecular dynamics analyses. Trajectory data are provided for the F-, Cl-, Br-, and I-passivatedCd35Se20X30 systems. Additional extended trajectory data used for the Cl and Br samplinganalysis are also included where applicable. 04_NAMD/--------Contains the PYXAID/nonadiabatic molecular dynamics (NAMD) inputs andprocessed population data used to obtain the reported hot-electronrelaxation dynamics. The directory also contains the analysis scripts and numerical outputsused for: - population analysis- relaxation-lifetime fitting- fitting-window convergence analysis- alternative fitting-function comparison- block-bootstrap uncertainty analysis The NAMD calculations were performed using multiple initial conditionsand decoherence-induced surface hopping (DISH). Large Hamiltonian directories and individual stochastic trajectoryoutputs are not included because the processed population data requiredfor the reported analysis are supplied directly. 05_processed_electronic_data/-----------------------------Contains the processed electronic quantities used in the analysis. gap_NAC/ Contains the processed LUMO+1-LUMO energy-gap and nonadiabatic- coupling data obtained from the AIMD trajectories. Data from the primary trajectories and the extended Cl/Br trajectories are included where applicable. dephasing/ Contains the processed decoherence-rate dataset and scripts used to calculate the corresponding dephasing times. The combined decoherence-rate dataset contains the individual decoherence rates used in the analysis for all four halide systems. 06_ALIGNN_FNA/--------------Contains the system-specific data and analysis files used for theAtomistic Line Graph Neural Network (ALIGNN) and feature-nullificationanalysis (FNA). datasets/ Contains the labeled structure-property datasets for the F, Cl, Br, and I systems. Each file maps a structure identifier to its target nonadiabatic- coupling value. training/ Contains the system-specific ensemble-training submission scripts, the ensemble-RMSE analysis script, and final ensemble-performance results. test_predictions/ Contains the test-set predictions from the 20 independently trained ALIGNN models for each halide system. The files include the structure identifier, target value, and predicted value for the corresponding test set. FNA/ Contains the scripts used for atom-resolved feature-nullification analysis and the final atom-importance results. The ALIGNN architecture, ensemble workflow, and FNA methodology followthe public GitHub framework: kushalsamanta/cdse-qds-gnn-framework ======================================================================SOFTWARE====================================================================== Electronic-structure calculations and AIMD:VASP Nonadiabatic molecular dynamics:PYXAID with decoherence-induced surface hopping (DISH) Machine-learning analysis:ALIGNN Feature analysis:Feature-nullification analysis (FNA) ======================================================================DATA ORGANIZATION====================================================================== The archive follows the computational workflow: optimized structures | vVASP electronic-structure/AIMD calculations | venergy-gap and nonadiabatic-coupling analysis | vPYXAID nonadiabatic molecular dynamics | vpopulation and relaxation-lifetime analysis The ALIGNN/FNA directory contains the complementary machine-learninganalysis of the nonadiabatic-coupling data. ======================================================================NOTES====================================================================== 1. Only data and scripts needed to reproduce or verify the principal analyses reported in the manuscript are included. 2. Large temporary files, restart files, wavefunction files, redundant plotting files, and unnecessary intermediate outputs are intentionally omitted. 3. Processed numerical datasets are supplied where retaining complete raw program output would result in unnecessary duplication or very large file sizes. 4. Additional methodological details are provided in the manuscript and Supporting Information. 5. The public ALIGNN/FNA framework is available from the GitHub repository kushalsamanta/cdse-qds-gnn-framework.