Search

Coarse graining (CG) enables the investigation of molecular properties for larger systems and at longer timescales than the ones attainable at the atomistic resolution. Machine learning techniques have been recently proposed to learn CG particle interactions, i.e. develop CG force fields. Graph representations of molecules and supervised training of a graph convolutional neural network architecture are used to learn the potential of mean force through a force matching scheme. In this work, the force acting on each CG particle is correlated to a learned representation of its local environment that goes under the name of SchNet, constructed via continuous filter convolutions. We explore the application of SchNet models to obtain a CG potential for liquid benzene, investigating the effect of model architecture and hyperparameters on the thermodynamic, dynamical, and structural properties of the simulated CG systems, reporting and discussing challenges encountered and future directions envisioned., Comment: Proceedings of 12th Conference on Artificial Intelligence (SETN), 2022

Due to the wide range of timescales that are present in macromolecular systems, hierarchical multiscale strategies are necessary for their computational study. Coarse-graining (CG) allows to establish a link between different system resolutions and provides the backbone for the development of robust multiscale simulations and analyses. The CG mapping process is typically system- and application-specific, and it relies on chemical intuition. In this work, we explored the application of a Machine Learning strategy, based on Variational Autoencoders, for the development of suitable mapping schemes from the atomistic to the coarse-grained space of molecules with increasing chemical complexity. An extensive evaluation of the effect of the model hyperparameters on the training process and on the final output was performed, and an existing method was extended with the definition of different loss functions and the implementation of a selection criterion that ensures physical consistency of the output. The relationship between the input feature choice and the reconstruction accuracy was analyzed, supporting the need to introduce rotational invariance into the system. Strengths and limitations of the approach, both in the mapping and in the backmapping steps, are highlighted and critically discussed., Comment: Proceedings of 12th Conference on Artificial Intelligence (SETN), 2022

A mesoscopic, mixed particle- and field-based Brownian dynamics methodology for the simulation of entangled polymer melts has been developed. Polymeric beads consist of several Kuhn segments, and their motion is dictated by the Helmholtz energy of the sample, which is a sum of the entropic elasticity of chain strands between beads, slip springs, and nonbonded interactions. The entanglement effect is introduced by the slip springs, which are springs connecting either nonsuccessive beads on the same chain or beads on different polymer chains. The terminal positions of slip springs are altered during the simulation through a kinetic Monte Carlo hopping scheme, with rate-controlled creation/destruction processes for the slip springs at chain ends. The rate constants are consistent with the free energy function employed and satisfy microscopic reversibility at equilibrium. The free energy of nonbonded interactions is derived from an appropriate equation of state, and it is computed as a functional of the local density by passing an orthogonal grid through the simulation box; accounting for it is necessary for reproducing the correct compressibility of the polymeric material. Parameters invoked by the mesoscopic model are derived from experimental volumetric and viscosity data or from atomistic molecular dynamics simulations, establishing a "bottom-up" predictive framework for conducting slip spring simulations of polymeric systems of specific chemistry. The mesoscopic simulation methodology is implemented for the case of cis-1,4-polyisoprene, whose structure, dynamics, thermodynamics, and linear rheology in the melt state are quantitatively predicted and validated without a posteriori fitting the results to experimental measurements., Comment: 80 pages, 17 figures

A method is formulated, based on combining self-consistent field theory with dynamically corrected transition state theory, for estimating the rates of adsorption and desorption of end-constrained chains (e.g. by crosslinks or entanglements) from a polymer melt onto a solid substrate. This approach is tested on a polyethylene/graphite system, where the whole methodology is parametrized by atomistically detailed molecular simulations. For short-chain melts, which can still be addressed by molecular dynamics simulations with reasonable computational resources, the self-consistent field approach gives predictions of the adsorption and desorption rate constants which are gratifyingly close to molecular dynamics estimates., Comment: 18 pages, 10 figures

The polymer dynamics of homogeneous C$_{60}$-polystyrene mixtures in the molten state are studied via molecular simulations using two interconnected levels of representation for polystyrene nanocomposites: (a) A coarse-grained representation, in which each polystyrene repeat unit is mapped into a single "superatom" and each fullerene is viewed as a spherical shell. Equilibration of coarse-grained polymer-nanoparticle systems at all length scales is achieved via connectivity-altering Monte Carlo simulations. (b) An atomistic representation, where both nanoparticles and polymer chains are represented in terms of united-atom forcefields. Initial configurations for atomistic Molecular Dynamics (MD) simulations are obtained by reverse mapping well-equilibrated coarse-grained configurations. By analyzing MD trajectories under constant energy, the segmental dynamics of polystyrene (for neat and filled systems) is characterized in terms of bond orientation time autocorrelation functions. Nanocomposite systems are found to exhibit slightly slower segmental dynamics than the unfilled ones, in good agreement with available experimental data. Moreover, by employing Voronoi tessellation of the simulation box, the mean-squared displacement of backbone carbon atoms is quantified in the vicinity of each fullerene molecule. Fullerenes are found to suppress the average motion of polymeric chains, in agreement with neutron scattering data, while slightly increasing the dynamic and stress heterogeneity of the melt. Atomic-level and local (per Voronoi cell) stress distributions are reported for the pure and the filled systems., Comment: 15 pages, 16 figures

The structural features of polystyrene brushes grafted on spherical silica nanoparticles immersed in polystyrene are investigated by means of a Monte Carlo methodology based on polymer mean field theory. The nanoparticle radii (either 8 nm or 13 nm) are held constant, while the grafting density and the lengths of grafted and matrix chains are varied systematically in a series of simulations. The primary objective of this work is to simulate realistic nanocomposite systems of specific chemistry at experimentally accessible length scales and study the structure and scaling of the grafted brush. The profiles of polymer density around the particles are examined; based on them, the brush thickness of grafted chains is estimated and its scaling behavior is compared against theoretical models and experimental findings. Then, neutron scattering spectra are predicted both from single grafted chains and from the entire grafted corona. It is found that increasing both the grafting density and the grafted chain molar mass drastically alters the brush dimensions, affecting the wetting behavior of the polymeric brush. On the contrary, especially for particles dispersed in high molecular weight matrix, variation of the matrix chain length causes an almost imperceptible change of the density around the particle surface., Comment: 12 pages, 11 figures

The structure of a polystyrene matrix filled with tightly cross-linked polystyrene nanoparticles, forming an athermal nanocomposite system, is investigated by means of a Monte Carlo sampling formalism. The polymer chains are represented as random walks and the system is described through a coarse grained Hamiltonian. This approach is related to self-consistent-field theory but does not invoke a saddle point approximation and is suitable for treating large three-dimensional systems. The local structure of the polymer matrix in the vicinity of the nanoparticles is found to be different in many ways from that of the corresponding bulk, both at the segment and the chain level. The local polymer density profile near to the particle displays a maximum and the bonds develop considerable orientation parallel to the nanoparticle surface. The depletion layer thickness is also analyzed. The chains orient with their longest dimension parallel to the surface of the particles. Their intrinsic shape, as characterized by spans and principal moments of inertia, is found to be a strong function of position relative to the interface. The dispersion of many nanoparticles in the polymeric matrix leads to extension of the chains when their size is similar to the radius of the dispersed particles., Comment: 9 pages, 13 figures

We present atomistic molecular dynamics simulations of two Polyethylene systems where all entanglements are trapped: a perfect network, and a melt with grafted chain ends. We examine microscopically at what level topological constraints can be considered as a collective entanglement effect, as in tube model theories, or as certain pairwise uncrossability interactions, as in slip-link models. A pairwise parameter, which varies between these limiting cases, shows that, for the systems studied, the character of the entanglement environment is more pairwise than collective. We employ a novel methodology, which analyzes entanglement constraints into a complete set of pairwise interactions, similar to slip links. Entanglement confinement is assembled by a plethora of links, with a spectrum of confinement strengths, from strong to weak. The strength of interactions is quantified through a link `persistence', which is the fraction of time for which the links are active. By weighting links according to their strength, we show that confinement is imposed mainly by the strong ones, and that the weak, trapped, uncrossability interactions cannot contribute to the low frequency modulus of an elastomer, or the plateau modulus of a melt. A self-consistent scheme for mapping topological constraints to specific, strong binary links, according to a given entanglement density, is proposed and validated. Our results demonstrate that slip links can be viewed as the strongest pairwise interactions of a collective entanglement environment. The methodology developed provides a basis for bridging the gap between atomistic simulations and mesoscopic slip link models., Comment: published in Macromolecules, 17 pages, 19 figures, 1 table, 123 references, 7 appendices, Supporting Information, 15 videos (two references added)

We introduce an algorithm for the reduction of computer generated atomistic polymer samples to networks of primitive paths. By examining network ensembles of Polyethylene and cis-1,4 Polybutadiene melts, we quantify the underlying topologies through the radial distribution function of entanglements and the distribution of the number of monomers between entanglements. A suitable scaling of acquired data leads to a unifying microscopic topological description of both melts., Comment: 4 pages 3 figures. Work presented at '5th International Discussion Meeting on Relaxations in Complex Systems (5IDMRCS)', Lille, France, July 7-13, 2005, and at 'Modeling and simulation of entangled polymeric liquids', CECAM, Lyon, France, July 18-21, 2005