Hybrid machine-learning-assisted stochastic nano-indentation behaviour of twisted bilayer graphene. (August 2022)

• Record Type: Journal Article
• Authors: Gupta, Kritesh Kumar; Roy, Lintu; Dey, Sudip
• Is Part Of: Journal of physics and chemistry of solids. Volume 167(2022)
• Publication Date: 2022-08
• Subjects: Twisted bilayer graphene -- Molecular dynamics -- Nano-indentation -- Uncertainty analysis -- PC-Kriging metamodel; Solids -- Periodicals; Solides -- Périodiques; Solids; Periodicals; 530.41
• DOI: 10.1016/j.jpcs.2022.110711
• Languages: English
• ISSNs: 0022-3697

Abstract

Abstract: We present herein a polynomial chaos-Kriging (PC-Kriging)-based molecular dynamics (MD) simulation framework of twisted bilayer graphene (tBLG) structures to investigate the influence of stochastic parametric variations on their nano-indentation behaviour. The relative rotation angle (RRA) and operating temperature were taken as input parameters, which were randomly distributed in the ranges 0–30° and 100–900 K, respectively. Considering Monte Carlo sampling (MCS), a series of MD simulations of nano-indentation was performed to obtain the critical indentation force (Fcr ) and critical indentation depth (δcr ) in each instance. The dataset generated by MCS-driven MD simulation was employed to train and validate the PC-Kriging-based metamodel. The generalization capability of the constructed model was ensured by implementing a leave-points-out (LpO) cross-validation scheme, and by minimizing prediction errors by adopting a sufficient size of samples for model training and validation. The constructed computationally efficient PC-Kriging-based metamodel has been used to perform data-driven uncertainty and probabilistic analysis. The hybrid machine-learning-based stochastic nano-indentation behaviour of tBLG structures has been described, considering practically relevant uncertain irregularities in RRA and temperature. The present analysis aims to capture the continuous parametric range of input parameters so as to allow detailed probabilistic investigation of the nano-indentation behaviour of tBLG nanostructures. Graphical abstract: Image 1 Highlights: Data-driven analysis is carried out to characterize the probabilistic nano-indentation behavior of twisted bilayer graphene. The successful integration of MD simulation and PC-Kriging metamodel is demonstrated. The inherent challenge of computational expense associated with MD simulation is mitigated. The influence of combined and individual stochastic parametric variations on the indentation responses is revealed.