We
provide a fundamental understanding of the interaction between cis-1,4-polybtadiene (PB) and the alumina (001) surface
via a computational multi-stage methodology involving density functional
theory (DFT) calculations and atomistic simulations. The latter were
performed via newly derived, computationally efficient, pairwise-additive,
parametric atomistic models using machine learning (ML) algorithms.
First, DFT calculations that include van der Waals (vdW) interactions
were performed to investigate the adsorption of cis-2-butene (i.e., the repeat unit of cis-1,4-PB)
on the alumina surface. To obtain an accurate sampling of the complex cis-2-butene/alumina configurational space at the DFT level,
we studied several systems in which alumina is modeled as (a) a small
cluster, to quickly find the stable configurations of the system,
and be used as a guide in (b) realistic periodic surface models. We
found similar optimal (stable) structures in both models. We also
performed a series of constrained DFT simulations to enhance the configurational
space sampling between cis-2-butene and alumina for
the periodic alumina (001). The atomistic model for the cis-2-butene/alumina interaction was developed using ML algorithms that
find, using a specific functional representation, the optimum classical
model parameters, which reproduce at the atomistic level the DFT-calculated
adsorption energy. To examine the whole procedure, we developed atomistic
models, using Lennard–Jones or Morse potential functions, in
both all-atom and united-atom representations. Moreover, we provide
a detailed statistical estimation of parametric uncertainty of the
derived atomistic models by applying the Bootstrap method. We were
able to fit the data with a mean absolute error below 1 kcal/mol,
for all the models considered. As a last part of the proposed methodology,
we performed molecular dynamics simulations, using the newly developed
models, of a representative system of 30mer cis-1,4-PB
chains confined between an infinite (periodic) alumina slab of thickness
≈1.114 nm at a temperature of 413 K and a pressure of 1 atm.
We studied the molecular packing in the vicinity of the interface.
The results demonstrate that the new atomistic models can accurately
reproduce the structure of PB chains in the vicinity of the alumina
substrate. In addition, we compared our results with atomistic simulations
describing the PB/alumina interaction via standard mixing (Lorentz–Berthelot
combination) rules. We demonstrate that the latter are inadequate
in describing the complex interfacial polymer/solid interaction.