Crystal structures, energies and search databases for "Bayesian Active Crystal Hopping: Machine-Learned Potentials and Surrogate-Guided Sampling for Crystal Structure Prediction of π-Extended Organic Molecules"

Primary data for three crystal-structure-prediction searches on tris(4-methoxyphenyl)amine-anthracene (TMPA-An, C42H36N2O4, Z = 4): the BACH active-learning search reported in the article (314 relaxed structures), an ablation of the same workflow with the Gaussian-process surrogate switched off (220 structures), and an evolutionary algorithm baseline run with USPEX (1009 local relaxations) under the same potential and relaxation protocol. Each search is deposited with its relaxed structures, its search database, its configuration and its LAMMPS input script, together with the calculations of the Cc/C2/c energy difference discussed in the article. The software that produced these data, CrYAL, is archived separately (DOI10.5281/zenodo.21896733). The experimental reference structure is available from the CCDC under deposition number 1493451. Acknowledgments Raul Flores acknowledges SECIHTI for the postdoctoral fellowship (CVU: 365229). The authors gratefully acknowledge the computing time granted by LANCAD and SECIHTI on the supercomputer Miztli at DGTIC UNAM.