Modulating Roughness and Ripple Dynamics in Hexagonal Boron Nitride via Defect Engineering

Two-dimensional materials like hexagonal boron nitride (hBN) display intrinsic roughness and thermal rippling that profoundly affect their mechanical, electronic, and thermal properties, yet the influence of atomic defects on these dynamics is poorly understood. In this work, we introduce a machine learning-based Atomic Cluster Expansion (ACE) potential that accurately models hBN and captures defect energies and dynamics -- a limitation of existing hBN potentials. Our simulations predict that the spontaneous formation of square-octagon (4|8) defects as triplets in an otherwise pristine sheet is more favorable than a previously predicted isolated $4|8$ defect that terminates a dislocation. Similar to graphene, we uncover a defect concentration-dependent crossover from spatially random rippling in pristine hBN to diminished fluctuations around a fixed corrugation at elevated defect densities with larger surface roughness. Unlike graphene, where divacancy defects most strongly modulate sheet rippling, hBN vacancies (N-monovacancies, B-monovacancies and divacancies) and N-interstitials exert negligible influence on corrugation and rippling; instead, the local stress induced by Stone-Wales (5|7) defects dominates this crossover, followed by square-octagon (4|8) and B-interstitial defects. The differences with graphene presumably arise from the heterogeneous composition of hBN and its polar-covalent bonding character, contrasting with graphene's homogeneous nature and purely covalent bonding. These results underscore material-specific sensitivity of rippling to defects and open avenues for targeted defect engineering to control hBN's flexibility in nanoelectronics, membranes, and beyond.