A First-Principles Study On The Migration Of Eb On The Surface Of ZN Catalysts

Ziegler-Natta（ZN）catalysts occupy 90%of the polypropylene catalyst market due to their high activity and controllable stereo-selectivity.However,there are still many controversies about the catalytic mechanism,such as surface composition of the support,Ti active center structure and electron donor modulation mechanism.Even for ethyl benzoate（EB）,one of the simplest electron donor,there is no consensus in experimental investigations on how it migrates on the surface and whether it can directly participate in the active center through Ti-O bonding.On the other hand,most of the existing theoretical studies are based on small cluster models,which are limited in size and cannot answer the above questions.In this paper,we studied various migration ways of EB on the MgCl₂（110）surface by the first-principle density functional theory（DFT）and molecular dynamics（MD）simulations,combined with NEB transition state search and free energy calculations.The possibility of direct EB participation in the Ti active center was explored.The results show that 1)the activation energy of EB interlayer migration is much lower than that of intra-layer migration,indicating the necessity of using a periodic surface model;2)EB in the neighboring Ti active center can migrate to Ti and poison the active center by forming Ti-O bond;3)EB in the poisoned active center can migrate to Mg again to restore the activity,while it is also possible to form a new active center with direct participation of EB by inducing the migration of chlorine atom;4)MD simulations show that the fluctuation of Mg-O bonds is larger than that of Ti-O bonds.Due to the limitation of computational resources,the time and space scales that can be achieved by current DFT-MD simulations are far from what is needed for catalytic research,while MD simulations based on traditional force fields cannot describe bond formation and break.Thus,we also attempt to perform machine learning to establish the relationship between the structure and species information among atoms and the energies/forces calculated by DFT,resulting a deep learning force field.The force field has the advantage of combining the accuracy of DFT with the speed of conventional force fields.