Enhanced Computational Sampling of Perylene and Perylothiophene Packing with Rigid-Body Models

Abstract

Molecular simulations have the potential to advance the understanding of how the structure of organic materials can be engineered through the choice of chemical components but are limited by computational costs. The computational costs can be significantly lowered through the use of modeling approximations that capture the relevant features of a system, while lowering algorithmic complexity or by decreasing the degrees of freedom that must be integrated. Such methods include coarse-graining techniques, approximating long-range electrostatics with short-range potentials, and the use of rigid bodies to replace flexible bonded constraints between atoms. To understand whether and to what degree these techniques can be leveraged to enhance the understanding of planar organic molecules, we investigate the morphologies predicted by molecular dynamic simulations using simplified molecular models of perylene and perylothiophene. Approximately, 10 000 wall-clock hours of graphics processing unit-accelerated simulations are performed using both rigid and flexible models to test their efficiency and predictive capability with the two chemistries. We characterize the 1191 resulting morphologies using simulated X-ray diffraction and cluster analysis to distinguish structural transitions, summarized by four phase diagrams. We find that the morphologies generated by the rigid model of perylene and perylothiophene match with those generated by the flexible model. We find that ordered, hexagonally packed columnar phases are thermodynamically favored over a wide range of densities and temperatures for both molecules, in qualitative agreement with experiments. Furthermore, we find the rigid model to be more computationally efficient for both molecules, providing more samples per second and shorter times to equilibrium. Owing to the structural accuracy and improved computational efficiency of modeling polyaromatic groups as rigid bodies, we recommend this modeling choice for enhancing the sampling in polyaromatic molecular simulations.

Computationally connecting organic photovoltaic performance to atomistic arrangements and bulk morphology

“…Subsequent relaxation of bonded constriants is required for each molecule to find a realistic atomistic morphology [69,71], unless an internal coordiante system is maintained to permit constrains such as the expected tor-sional rotation of a thiophene ring or a phenyl group [70,80] to be enforced during fine-graining. In some cases, it can be useful to treat each group as a rigid body, keeping constant the relative positions and orientations of atoms within the group according to a geometrically-relaxed template [81,82]. After the initial projection has been completed, there may be several unphysical molecular conformations -particularly for long, flexible polymer chains where backbones may wrap around and result in overlapping atoms.…”

Computational Study on Homogeneous Melting of Benzene Phase I

“…It was demonstrated that this force field reproduces the crystal structure of benzene phase I [35,40]. Other studies have also claimed that partial charges affect the orientation of benzene dimer [41] and the phase diagram for the other polycyclic aromatic hydrocarbons [42]. Intermolecular interactions are truncated at 1.20 nm.…”

Optimization and Validation of Efficient Models for Predicting Polythiophene Self-Assembly

“…The system volume is then reduced during another short simulation (1.8 ns and 1300 K) until the target density is reached. This process of "initializing", "mixing", and "shrinking" has been previously used to initialize independent snapshots at arbitrary densities [28,49,50,64]. Unless otherwise specified, every simulation presented herein is instantaneously quenched from high temperature to the target temperature for the duration of its NVT simulation.…”

“…This level of coarse-graining is convenient for modeling OPV materials, as the reduction of simulation elements from 25 atoms to 11 united-atom sites per monomer reduces computational cost [45][46][47], while simplifying back-mapping of atomic coordinates for charge transport calculations [35]. United-atom models have been used to successfully predict structures for a variety of systems including: polymers [48,49], proteins [46], and small molecules [50,51]. The base units of mass M = 32 amu, energy E = 0.32 kcal/mol, and length L = 3.905 Å, used to describe interactions within the simulation, are adapted from the OPLS-UA force-field [52].…”

“…Each simulation performed herein utilizes one of two simulation protocols to sample microstates at the target state-point. Protocol (1) ignores solvent evaporation: disordered initial configurations at the target density ρ and solvent quality ε s are instantaneously quenched from T = 1300 to the target temperature, after which equilibration progress is monitored [50]. Protocol (2) is an extremely simplified, qualitative model of solvent evaporation that helps to sample configurations at experimental densities (ρ = 1.11 g/cm 3 ): First, a system is equilibrated at ρ = 0.72 g/cm 3 and the target temperature and ε s using Protocol (1), followed by a linear compression to ρ = 1.11 g/cm 3 over 280 ns.…”