High-precision computer simulations of bond-fluctuation (BF) model (volume fraction φ=0.5) are performed in the transition region between the non-entanglement and the entanglement regime. Defects of the model called X-traps are newly found which result in serious errors in the long-time behavior of the system. For samples free from X-traps, diffusion coefficient D of the center of mass, relaxation times τα for Rouse coordinate Rα (α=1,2) and τL for the end-to-end vector L are determined for chains of length N=16–180 within few percent of statistical errors. Their N-dependence changes around N=100 at which entanglement coupling is supposed to begin. By comparing D obtained by the simulations with experimental data of polystyrene melts and solutions, the average chain length per entanglement Ne was estimated to be 89, which is much larger than 30–42 reported by Paul et al. (J Phys II 1991;1:37). To see the origin of the discrepancy, statistical errors and system size effects are studied in detail and it was found that, to determine D within few percent of error, the number of independent chain samples Nsample should be larger than 10 000 and the size of simulation cells ℓcell should be larger than 4Rg; these conditions are not satisfied in the previous simulations. Critical chain length Ncη for entanglement of BF model is estimated to be Ncη=170 using the empirical relationship Ncη/Ne=1.92 for polystyrene melts. It is argued that Ncη=170 is a universal parameter of entanglement but Ne=89 is a specific value for polystyrenes and it may change with the materials with which comparisons are made.
Read full abstract