Friction characteristics of doublewall carbon nanotube-based nanobearings: a comparative molecular dynamics study using deep learning and reactive force fields

A convergence in scientific advancements associated with molecular biology and nanofabrication technologies now offers the potential of engineering functional hybrid organic/inorganic devices on a nanometer scale. As such, the creation of nanomechanical systems powered by biological motors is within reach. Doublewall carbon nanotubes (DWCNT) offer great potential as nanobearings due intrinsically atomically smooth surfaces enabling easy inter-shell sliding. Although simulation of friction in DWCNT bearing systems has been a popular topic, quantitative estimates of friction reported in literature encompass a wide range of magnitudes. Moreover, the computational cost of accurate quantum-mechanical calculations proves prohibitive for larger simulations. In recent years, increased efforts to overcome such hurdle using Deep learning Neural Networks Potential (DNNP) models, where only a reduced set of reference calculations is required to accurately predict force fields, has been undertaken. Molecular Dynamics (MD) simulations could greatly benefit from such artificial intelligence method to achieve more accurate predictions of properties in relation to empirically parameterized force fields, namely in the atomistic simulation of tribological properties.