Combination of Linear Discriminant Analysis and All-Atom Molecular Dynamics Simulations for Probing the Electrokinetics in Polyelectrolyte-Brush-Grafted Nanochannels

Here we employ Linear Discriminant Analysis (LDA) based Machine Learning (ML) and all-atom Molecular Dynamics (MD) simulations to probe the non-linearly large electroosmotic (EOS) transport in nanochannels grafted with cationic [poly(2-(methacryloyloxy)ethyl) trimethylammonium chloride] (PMETAC) polyelectrolyte (PE) brushes. We first identify some basic features, namely different n_xys, or the number of atoms of type x around the atom of type y inside different bins (dividing the simulation domain into different bins). The resulting data points are subjected to LDA producing, corresponding to different values of the electric fields, data clusters in each bin and importance score for each n_xy in each bin and for each electric field. For the present case, the importance score for n_mon-mon (mon: monomer) and n_Cl-Ow (Cl: Chloride ion; O_w: water oxygen atom) in bins at the brush-bulk interface undergo maximum change with the electric field. On studying these quantities, we identify that a significant localization of the monomers, counterions, and water molecules at the brush-bulk interface at larger electric fields cause the non-linear increase in the EOS flow strength.