Comparison of ReaxFF, DFTB, and DFT for phenolic pyrolysis. 1. Molecular dynamics simulations.

A systematic comparison of atomistic modeling methods including density functional theory (DFT), the self-consistent charge density-functional tight-binding (SCC-DFTB), and ReaxFF is presented for simulating the initial stages of phenolic polymer pyrolysis. A phenolic polymer system is simulated for several hundred picoseconds within a temperature range of 2500 to 3500 K. The time evolution of major pyrolysis products including small-molecule species and char is examined. Two temperature zones are observed which demark cross-linking versus fragmentation. The dominant chemical products for all methods are similar, but the yields for each product differ. At 3500 K, DFTB overestimates CO production (300-400%) and underestimates free H (~30%) and small C(m)H(n)O molecules (~70%) compared with DFT. At 3500 K, ReaxFF underestimates free H (~60%) and fused carbon rings (~70%) relative to DFT. Heterocyclic oxygen-containing five- and six-membered carbon rings are observed at 2500 K. Formation mechanisms for H2O, CO, and char are discussed. Additional calculations using a semiclassical method for incorporating quantum nuclear energies of molecules were also performed. These results suggest that chemical equilibrium can be affected by quantum nuclear effects at temperatures of 2500 K and below. Pyrolysis reaction mechanisms and energetics are examined in detail in a companion manuscript.

Comparison of ReaxFF, DFTB, and DFT for phenolic pyrolysis. 2. Elementary reaction paths.

Reaction paths for the loss of CO, H2, and H2O from atomistic models of phenolic resin are determined using the hybrid B3LYP approach. B3LYP energetics are confirmed using CCSD(T). The energetics along the B3LYP paths are also evaluated using the PW91 generalized gradient approximation (GGA), the more approximate self-consistent charge density functional tight binding (SCC-DFTB), and the reactive force field (ReaxFF). Compared with the CCSD(T)/cc-pVTZ level for bond and reaction energies and barrier heights, the B3LYP, PW91, DFTB(mio), DFTB(pbc), and ReaxFF have average absolute errors of 3.8, 5.1, 17.4, 13.2, and 19.6 kcal/mol, respectively. The PW91 is only slightly less accurate than the B3LYP approach, while the more approximate approaches yield somewhat larger errors. The SCC-DFTB paths are in better agreement with B3LYP than are those obtained with ReaxFF.

Polymer chain dynamics at interfaces: role of boundary conditions at solid interface.

Using classical molecular dynamics simulations, we study the dynamical properties of a single polymer chain dissolved in an explicit solvent and strongly adsorbed at solid-liquid interface. To circumvent a serious challenge posed by finite size effects due to long-range hydrodynamic effects, we developed a correction procedure that substantially limits the finite size effects. Concurrently, we provide an analysis of distinctly different size effects in the directions transverse and normal to the interface. We find that on analytically smooth interfaces, corresponding to the slip boundary condition, the motions of the polymer chain and the surrounding solvent are hydrodynamically coupled. This leads to the chain diffusion coefficient D scaling with the chain degree of polymerization N as D approximately N(-3/4), consistent with the Zimm dynamics for strongly adsorbed chains. Introduction of transverse forces at the interface results in loss of correlation between the motion of the polymer chain and the solvent. Consequently, D approximately N(-1), which is a characteristic of the Rouse dynamics.

Modeling diffusion of adsorbed polymer with explicit solvent.

Computer simulations of a polymer chain of length N strongly adsorbed at the solid-liquid interface in the presence of explicit solvent are used to delineate the factors affecting the N dependence of the polymer lateral diffusion coefficient, D(||). We find that surface roughness has a large influence, and D(||) scales as D(||) approximately N(-x), with x approximately 3/4 and x approximately 1 for ideal smooth and corrugated surfaces, respectively. The first result is consistent with the hydrodynamics of a "particle" of radius of gyration R(G) approximately N(nu) (nu=0.75) translating parallel to a planar interface, while the second implies that the friction of the adsorbed chains dominates. These results are discussed in the context of recent measurements.

Molecular-dynamics simulations of the transport properties of a single polymer chain in two dimensions.

Molecular-dynamics simulations are conducted to elucidate the critical factors affecting the transport properties of isolated polymer chains in strictly two dimensions. The relevance of surface inhomogeneity is critically examined. We unequivocally find that surface inhomogeneity is critical in obtaining transport behavior consistent with the recent measurements of surface diffusion for polymers adsorbed at the solid-liquid interface. For a systematic investigation of this point, heterogeneity was introduced by decorating the surface with impenetrable elements and we find that chain diffusivity crossed over from Rouse-type behavior to reptationlike with increasing surface coverage of obstacles. This transition in behavior occurred when the mean distance between obstacles is approximately equal to the end-to-end distance, Re, of the two-dimensional chain. Our results underscore the importance of surface disorder (not only literal obstacles but by reasonable extension also to other types of disorder) in determining the transport behavior of chains adsorbed to solids.