Image shows a nanocomposite surface (from experiments) and spherical particle position as predicted by the Bayesian algorithm.
Animation shows the algorithm working with error reduction for each step.Fiber-reinforced composites are a mainstay in transport, aerospace, marine, and sporting goods. Nanofillers like graphene, carbon nanotubes and nanosilica improve impact resistance and toughness. To get the best out of them, they must be embedded uniformly within the matrix – but determining whether this is the case is challenging. Now researchers have developed an algorithm that can predict the location of nanosilica particles in a nanocomposite [Thiem et al., Composites Science and Technology 218 (2022) 109205, https://doi.org/10.1016/j.compscitech.2021.109205 ].
“We combined nanoscale experiments, simulations and data science algorithms to determine the relative position of nanoparticles to the nanocomposite surface, which is not possible from experiments,” explains Aniruddh Vashisth of the University of Washington, who led the study with colleagues from CCDC Army Research Laboratory in Maryland, Georgia Institute of Technology, Ansys Inc. in San Jose and Pennsylvania State University.
Quantifying the distribution of nanofillers in a composite matrix, which determines physical properties such as tensile modulus, fracture toughness, abrasion resistance and is key to understanding failure behavior, is difficult. Most methods, such as optical, electron and probe microscopy, spectroscopy, conductivity and fluorescence measurements, are qualitative rather than quantitative.
“This knowledge is essential to figure out how fracture propagates in a nanocomposite and how it is influenced by the nanofiller,” adds Vashisth.
Using atomic force microscopy (AFM) experimental data from nanosilica-polymer composites, the researchers generated virtual data using finite element simulations (FES) of surfaces with different features. The actual and virtual data is then fed into a Bayesian algorithm to generate the position of nanosilica particles relative to the nanocomposite surface. The algorithm consistently predicts the particle position to within 3 nm of the known actual position.
“This is the first time that both experimental AFM data and finite element data has been used with a data science framework to pin-point the position of nanoparticles on the surface of a composite,” says Vashisth.
The approach could prove a powerful tool for mapping particle distribution, enabling a better understanding of the effects of matrix chemistry variations, surface functionality, processing conditions and nanofiller particle properties on composites. For now, the algorithm only works for spherical fillers like nanosilica and carbon black but could be trained for 1D carbon nanotubes or nanofibers and 2D graphene or MXene particles, although this would require more experimental data and computing power.
“As more and more nanomaterials are adopted by the industry, it is necessary to understand the fundamental interactions of these nanoparticles with polymer matrices,” points out Vashisth. “This algorithm [can] determine what type of surface modification leads to a stiff or a soft hand-shake [binding] between the nanofiller and matrix, potentially helping us design better nanocomposites.”