Fig. 13: The predicted first-order frequencies comparisons between all-atom and CG results.
From: Coarse-grained machine learning potential for mesoscale multilayered graphene
a The predicted first-order frequencies of rectangular and b circular graphene membranes with different number of layers. The black square points, red circular points, blue upright triangular points and purple inverted triangular points joined by solid lines denote results obtained via all-atom, CGNEP (Zang = 4 Å), NEP-CG and CGNEP (Zang = 10 Å) calculations, respectively. In CGNEP (Zang = 10 Å), both angular and radial descriptor components consider adjacent CG graphene layers. In CGNEP(Zang = 4 Å), radial descriptor components consider adjacent CG graphene layers, and angular descriptor components only consider the single CG graphene layer and neglect the rest CG graphene layers.
