Over the past 20 years, a number of investigators have applied constant pressure molecular dynamics simulation to estimate the glass transition temperature Tg of polymers by determining the intersection of the characteristic low and high temperature regions of the resulting volume-temperature curves in a manner analogous to experimental dilatometry experiments. While simulations do generally show qualitatively the well-known experimental behavior, conclusions differ over the accuracy and general applicability of the method. In this context it can be remarked that the duration of a specific volume measurement in a simulation is many orders of magnitude shorter than in a typical experiment (e.g. nanoseconds vs seconds), and in consequence the simulations effectively determine the equivalent of a high frequency Tg, which theoretically should be expected to be at least a few tens of degrees higher than an experimental Tg made at a frequency of order 1Hz. Results of simulations however indicate apparent Tg values ranging from below to significantly above the experimental Tg.

A number of explanations can be advanced for this lack of general agreement, from use of different force fields of unknown accuracy, to differences in simulation methology and duration and the manner of analyzing results such as the precise method in which low and high temperature regions of the V-T curve are identified.

In recent studies we have accordingly employed a force field recognized for its ability to predict PVT behavior to conduct an in-depth investigation of one particular aspect of simulation methodology, namely the effective sample cooling rate, using atactic polystyrene oligomer systems for which there exists a significant amount of consistent experimental data reported by different authors. All simulations have commenced with thoroughly equilibrated melt systems and have performed cooling in a series of stages covering a temperature span of 300 degrees with a final temperature well below the experimental Tg. The resulting data, for simulation runs extending up to tens of nanoseconds have then been analyzed to estimate the apparent Tg as a function of cooling rate. Moreover since experimentally the molecular weight dependence of the Tg of polystyrene has been extensively measured, we have performed simulations for two molecular weights M ~ 940 and M ~ 8320, for which experimental Tg's differ by more than 70 degrees.

Since all simulations commence well above Tg, we have also examined the resulting curves for the occurrence of other reproducible transitions above the known Tg. Specifically this presentation will report on evidence of the so-called liquid-liquid transition Tll, experimental observation of which has been reported by a large number of workers.