38 |
|
\begin{doublespace} |
39 |
|
|
40 |
|
\begin{abstract} |
41 |
< |
|
41 |
> |
A molecular simulation method is developed based on the previous RNEMD |
42 |
> |
algorithm. With scaling the velocities of molecules in specific |
43 |
> |
regions of a system, unphysical thermal transfer can be achieved and |
44 |
> |
thermal / momentum gradient can be established, as well as |
45 |
> |
conservation of linear momentum and translational kinetic |
46 |
> |
energy. Thermal conductivity calculations of Lennard-Jones fluid water |
47 |
> |
(SPC/E model) and crystal gold (QSC and EAM model) are performed and |
48 |
> |
demostrate the validity of the method. Furthermore, NIVS-RNEMD |
49 |
> |
improves the non-Maxwell-Boltzmann velocity distributions in previous |
50 |
> |
RNEMD methods, where unphysical momentum transfer occurs. NIVS-RNEMD |
51 |
> |
also extends its application to interfacial thermal conductivity |
52 |
> |
calculations, thanks to its novel means in kinetic energy transfer. |
53 |
|
\end{abstract} |
54 |
|
|
55 |
|
\newpage |
658 |
|
|
659 |
|
|
660 |
|
\subsection{Interfaciel Thermal Conductivity} |
661 |
< |
After valid simulations of homogeneous water and gold systems using |
662 |
< |
NIVS-RNEMD method, calculation of gold/water interfacial thermal |
663 |
< |
conductivity was followed. It is found out that the interfacial |
664 |
< |
conductance is low due to a hydrophobic surface in our system. Figure |
665 |
< |
\ref{interfaceDensity} demonstrates this observance. Consequently, our |
666 |
< |
approximation in $G$ calculation (Eq. \ref{interfaceCalc}) maintains |
667 |
< |
valid. Reported results (Table \ref{interfaceRes}) are of two orders of |
668 |
< |
magnitude smaller than our calculations on homogeneous systems, and |
669 |
< |
thus have larger relative errors than our calculation results on |
670 |
< |
homogeneous systems. |
661 |
> |
After simulations of homogeneous water and gold systems using |
662 |
> |
NIVS-RNEMD method were proved valid, calculation of gold/water |
663 |
> |
interfacial thermal conductivity was followed. It is found out that |
664 |
> |
the low interfacial conductance is probably due to the hydrophobic |
665 |
> |
surface in our system. Figure \ref{interfaceDensity} demonstrates mass |
666 |
> |
density change along $z$-axis, which is perpendicular to the |
667 |
> |
gold/water interface. It is observed that water density significantly |
668 |
> |
decreases when approaching the surface. Under this low thermal |
669 |
> |
conductance, both gold and water phase have sufficient time to |
670 |
> |
eliminate temperature difference inside respectively (Figure |
671 |
> |
\ref{interfaceGrad}). With indistinguishable temperature difference |
672 |
> |
within respective phase, it is valid to assume that the temperature |
673 |
> |
difference between gold and water on surface would be approximately |
674 |
> |
the same as the difference between the gold and water phase. This |
675 |
> |
assumption enables convenient calculation of $G$ using |
676 |
> |
Eq. \ref{interfaceCalc} instead of measuring temperatures of thin |
677 |
> |
layer of water and gold close enough to surface, which would have |
678 |
> |
greater fluctuation and lower accuracy. Reported results (Table |
679 |
> |
\ref{interfaceRes}) are of two orders of magnitude smaller than our |
680 |
> |
calculations on homogeneous systems, and thus have larger relative |
681 |
> |
errors than our calculation results on homogeneous systems. |
682 |
|
|
683 |
|
\begin{figure} |
684 |
|
\includegraphics[width=\linewidth]{interfaceDensity} |
685 |
|
\caption{Density profile for interfacial thermal conductivity |
686 |
< |
simulation box.} |
686 |
> |
simulation box. Significant water density decrease is observed on |
687 |
> |
gold surface.} |
688 |
|
\label{interfaceDensity} |
689 |
|
\end{figure} |
690 |
|
|
691 |
|
\begin{figure} |
692 |
|
\includegraphics[width=\linewidth]{interfaceGrad} |
693 |
|
\caption{Temperature profiles for interfacial thermal conductivity |
694 |
< |
simulation box.} |
694 |
> |
simulation box. Temperatures of different slabs in the same phase |
695 |
> |
show no significant difference.} |
696 |
|
\label{interfaceGrad} |
697 |
|
\end{figure} |
698 |
|
|
790 |
|
|
791 |
|
\section{Conclusions} |
792 |
|
NIVS-RNEMD simulation method is developed and tested on various |
793 |
< |
systems. Simulation results demonstrate its validity of thermal |
794 |
< |
conductivity calculations. NIVS-RNEMD improves non-Boltzmann-Maxwell |
795 |
< |
distributions existing in previous RNEMD methods, and extends its |
796 |
< |
applicability to interfacial systems. NIVS-RNEMD has also limited |
797 |
< |
application on shear viscosity calculations, but under high momentum |
798 |
< |
flux, it could cause temperature difference among different |
799 |
< |
dimensions. Modification is necessary to extend the applicability of |
800 |
< |
NIVS-RNEMD in shear viscosity calculations. |
793 |
> |
systems. Simulation results demonstrate its validity in thermal |
794 |
> |
conductivity calculations, from Lennard-Jones fluid to multi-atom |
795 |
> |
molecule like water and metal crystals. NIVS-RNEMD improves |
796 |
> |
non-Boltzmann-Maxwell distributions, which exist in previous RNEMD |
797 |
> |
methods. Furthermore, it develops a valid means for unphysical thermal |
798 |
> |
transfer between different species of molecules, and thus extends its |
799 |
> |
applicability to interfacial systems. Our calculation of gold/water |
800 |
> |
interfacial thermal conductivity demonstrates this advantage over |
801 |
> |
previous RNEMD methods. NIVS-RNEMD has also limited application on |
802 |
> |
shear viscosity calculations, but could cause temperature difference |
803 |
> |
among different dimensions under high momentum flux. Modification is |
804 |
> |
necessary to extend the applicability of NIVS-RNEMD in shear viscosity |
805 |
> |
calculations. |
806 |
|
|
807 |
|
\section{Acknowledgments} |
808 |
|
Support for this project was provided by the National Science |