Thermostats and thermostat strategies for molecular dynamics simulations of nanofluidics

The thermostats in molecular dynamics (MD) simulations of highly confined channel flow may have significant influences on the fidelity of transport phenomena. In this study, we exploit non-equilibrium MD simulations to generate Couette flows with different combinations of thermostat algorithms and strategies. We provide a comprehensive analysis on the effectiveness of three thermostat algorithms Nosé-Hoover chain (NHC), Langevin (LGV) and dissipative particle dynamics (DPD) when applied in three thermostat strategies, thermostating either walls (TW) or fluid (TF), and thermostating both the wall and fluid (TWTF). Our results of thermal and mechanical properties show that the TW strategy more closely resembles experimental conditions. The TF and TWTF systems also produce considerably similar behaviors in weakly sheared systems, but deviate the dynamics in strongly sheared systems due to the isothermal condition. The LGV and DPD thermostats used in the TF and TWTF systems provide vital ways to yield correct dynamics in coarse-grained systems by tuning the fluid transport coefficients. Using conventional NHC thermostat to thermostat fluid only produces correct thermal behaviors in weakly sheared systems, and breaks down due to significant thermal inhomogeneity in strongly sheared systems.