Model dimensional scale effects on molecular dynamics simulations of bituminous materials

Accurate asphalt molecular dynamics (MD) simulations are essential for predicting its thermal, mechanical, and rheological properties. However, commonly used small-scale models (40 Å cubic) usually lead to significant fluctuations in simulation results. The fluctuations diminish with increasing model size, indicating the scale-dependent convergence of simulation results, a phenomenon that hasn't been systematically quantified in previous asphalt MD simulation studies. This study aims to investigate the convergence behavior and determine appropriate model sizes for key simulation parameters. Different sizes of cubic asphalt models based on AAA-1 and AAM-1 asphalt samples were constructed, and the simulation results were validated against experimental data. Results showed that as the model side length increased from 40 Å to 80 Å, the average density, solubility parameter, and shear viscosity gradually stabilized, approaching convergence beyond 60 Å. Structural stability also improved markedly, as evidenced by the radial distribution function. Diffusion coefficients increased by 42.8 % (AAA-1) and 11.8 % (AAM-1) as the model size grew to 60 Å, whereas the shear and Young's modulus in small-scale models showed strong anisotropy and only converged at 80 Å. Therefore, the model side lengths of 80 Å are required for mechanical simulations, while 60 Å models are sufficient for simulations of the other five parameters above, balancing computational cost with predictive reliability. Compared to previous studies, this work fills a critical gap in understanding the effects of model dimensional scale on asphalt molecular simulations and provides a practical reference for selecting model sizes, enabling reliable and efficient asphalt MD simulations.