Fully Coupled Molecular Dynamics–Generalised Fluctuating Hydrodynamics Model Of Nanofluids

Nanofluids can meet the high-heat-load heat transfer and cooling requirements of heat exchange systems in various fields of heat-producing equipment and have a broad application prospect and potential economic value.Understanding the micro-scale and nano-scale flow characteristics of nanofluids is important for optimizing nanofluid preparation and revealing the mechanism of micro-nano flow regulation.However,the huge amount of solvent molecules of nanofluids,complex flow fields,the need for atomic-scale information,and some special nanofluid characteristic behaviors slower than thermal fluctuation make the existing single-scale methods unable to simultaneously satisfy the above characteristics.The aim of this dissertation is to propose a multi-scale model for nanofluids,which has the advantage of using a molecular dynamics model in the region where atomic-scale information needs to be concerned,and a larger scale numerical model in the region where atomic-scale information is not required,such as in solvents.This ensures the fidelity of the model on the one hand;on the other hand,it will greatly save computational power and indirectly reduce resource consumption and energy waste.In this study,the generalized Landau-Lifshitz hydrodynamics multiscale model is used as a starting point,and the model is gradually developed into a multiscale model of nanofluids,and a complete parametric test with preliminary application to copper-water nanofluids is carried out.The main work and conclusions are as follows:1.A multiscale coupled model with the consistent thermostat effect is developed,which is based on the generalized Landau-Lifshitz hydrodynamic equation.The thermostat dissipation term is incorporated into the particle equation to form a particle dynamics equation in the format of Langevin thermostat equation;then the new particle source term of the hydrodynamic part of the multiscale model is re-derived,which ensures the consistency of the thermostat effect in the multiscale model;and further,local thermostat equations sensitive to phase concentration are proposed.Based on two versions of the model,namely,the Langevin local thermostat and the Langevin constant temperature thermostat models,the SPC/E water system was tested under equilibrium isothermal conditions and compared with the original global thermostat model and single-scale all-atom molecular dynamics simulations at different computational domain sizes.In contrast to the global thermostat model,the Langevin local thermostat and Langevin constant temperature thermostat models are less sensitive to the computational domain size and can not only reproduce the correct velocity and density fluctuation but also capture the local temperature in the pure particle region.In addition,it is shown that the Langevin local thermostat can accurately capture the density in the pure particle region,and therefore the pressure and the first hydrated layer and the first valley of the radial distribution function have been well reproduced.Compared to single-scale all-atom molecular dynamics models,the coupled model is significantly less computationally expensive and,closely related on the size of the computational domain,the coupled model is 5 to 20 times faster.2.The multiscale model is essentially a coupling between two algorithms with vastly different degrees of freedom,and the intense oscillations of the particle domain have high requirements for model stability.In order to reduce the parameter sensitivity of the thermostat-consistent multiscale coupling method and make the model more stable,the approach of reducing the particle source term noise using spatial averaging and supplemented with artificial viscosity is proposed;and further,the approach of updating forcing term with the explicit-implicit combination is proposed.Based on the above alogithms,the stability of the multiscale model is tested for comparison.The results show that the above methods make the stability of the multiscale coupling model significantly improved,making it applicable to the case with larger coupling parameters,which lays a good foundation for the subsequent non-equilibrium simulation.More importantly,the current coupling method makes the particle equations applicable to higher-order long-range force integration methods,including the Reaction-field method and the PME method;this makes the fluctuation information,as well as the temperature and density in the pure particle region,present more accurate results in comparison with the original Cut-off method.Compared with the single-scale all-atom molecular dynamics model,the computational cost of the coupled model is significantly reduced,and the speed of the coupled model is increased by a factor of 4 to 18 depending upon the size of the computational domain.3.Nanofluids are rarely at rest,so the multiscale model should be compatible with nonequilibrium flows.Firstly,a linearization scheme is proposed for the nonequilibrium shear flow,and the scheme is optimized by using the periodic Poiseuille and Couette flows in the LJ argon fluid system as the reference case.The shear flow simulations in the SPC/E water system show that the linearized scheme results in good agreement between the flow profiles and the analytical solutions of the Poiseuille flow as well as the Couette flow;the fluctuation characteristics of the model under shear flow also maintain good agreement with those from the equilibrium multiscale model,and the presence of shear flow does not affect the radial distribution characteristics of particles in the pure MD particle domain.In addition,the current multiscale model has efficiency advantage of 1.5 to 17 times compared with the single-scale all-atomic nonequilibrium molecular dynamics model in the same computational domain.After extending the multiscale model implementation to shear flow,the current multiscale model is further made compatible with acoustic wave transfer by imposing an incident plane acoustic wave analysis solution at the entrance boundary of the continuous medium domain.The results show that the particle domain density and velocity signals obtained from the multiscale model are in general agreement with the trend of the acoustic wave analytical solution,fully indicating that the current multiscale model correctly transfers information between the discrete particle phase and the continuous medium phase.4.The multiscale model is extended to the nanofluidic system by implementing the dynamic migration of the pure particle domain and the whole particle domain in the model.Based on the multiscale model,the diffusion and migration properties of copper nanoparticles in equilibrium and in the Poiseuille background flow were simulated.The results show that the diffusion coefficients of copper nanoparticles are very close to the solutions of single-scale all-atom molecular dynamics with an error of about 2.27%.In addition,an analytical model for the diffusion of nanoparticles in background flow is derived,and the flow-diffusion properties of copper nanoparticles are analyzed based on the equivalent particle convective velocity and the dimensionless number characterizing the lateral migration;the results show that copper nanoparticles will show diffusion enhancement in the flow direction due to the main flow,and the diffusion in the lateral and homogeneous directions will appear to be enhanced synergistically with the flow.In addition,there exists a critical Reynolds number that allows copper nanoparticles to undergo significant lateral migration.The paper has 40 figures,26 tables and 187 references.