Re-examining the procedure for simulating polymer T g using molecular dynamics.

In this work, the poly(ethylene oxide) bulk as one example has been iteratively heated and cooled back using MD simulations to examine the effects of thermal history on the resulting Tg . It is demonstrated that, after the system is equilibrated once at the high temperatures, the simulated Tg does not exhibit a systematical shift with the thermal history, and the averaged Tg compares well with that for the single procedure, that is, adequately equilibrating at the highest temperature and cooling with the same rate to the lowest temperature. Additionally, the continuous and stepwise processes lead to almost identical Tg , density and volumetric expansive coefficients at both the glassy and rubbery states at 300 K and 1 atm. However, these results would somewhat vary with what (volume or density) are used and how to yield them. Furthermore, the stepwise processes allow one to obtain the time-dependent dynamical Tg values from the reorientation functions of the monomer vectors, which suggest greater differences within longer observation time. This work rationalizes the "golden standard" procedure to simulate polymer Tg using the MD method, and provides some key clues to obtain the reliable results (specially for comparisons). Graphical abstract The extensive molecular dynamics simulations show that the glass transition temperature (Tg ) values obtained from volumetric (vol.) or density (den.) data do not exhibit a systematic shift with the thermal history (Proc.) whereas the Tg values obtained from dynamical (dyn.) data decrease and exhibit greater difference with increasing the observation time (t*).