Capturing the nuclear quantum effects in molecular dynamics for lattice thermal conductivity calculations: Using ice as example

Molecular dynamics (MD) is a powerful (and the most viable) tool to compute the thermal conductivities of solid disordered materials. However, conventional classical MD fails to describe the nuclear quantum effects (NQEs), so it may give inaccurate results for light materials at low temperatures. While the importance of NQE has been widely acknowledged, yet we do not have a fully reliable method to account for NQE in the MD thermal conductivity calculations. In this work, we will investigate and analyze the performances of a number of path-integral-based quantum MD methods, using ordered ice as a test case. To establish the validity of these methods, we will compare the MD results with the lattice dynamics results, in both classical and quantum limits. Through such a comparison, we will show that methods such as ring polymer MD stand as a good approach for a complex solid with short phonon lifetimes but could be problematic when describing long-living acoustic phonons. In addition, we will show that the rigid water model, which is the state-of-the-art model in the studies of ice/water systems, fails to capture most of the NQEs in ice thermal conductivity. Neglecting librational and translational NQEs leads to essential errors, which clearly demonstrates the importance of a true quantum simulation method that treats all modes at a consistent quantum level. © 2020 Author(s).