1 |
gezelter |
3524 |
\documentclass[11pt]{article} |
2 |
|
|
\usepackage{amsmath} |
3 |
|
|
\usepackage{amssymb} |
4 |
|
|
\usepackage{setspace} |
5 |
|
|
\usepackage{endfloat} |
6 |
|
|
\usepackage{caption} |
7 |
|
|
%\usepackage{tabularx} |
8 |
|
|
\usepackage{graphicx} |
9 |
gezelter |
3609 |
\usepackage{multirow} |
10 |
gezelter |
3524 |
%\usepackage{booktabs} |
11 |
|
|
%\usepackage{bibentry} |
12 |
|
|
%\usepackage{mathrsfs} |
13 |
|
|
\usepackage[ref]{overcite} |
14 |
|
|
\pagestyle{plain} \pagenumbering{arabic} \oddsidemargin 0.0cm |
15 |
|
|
\evensidemargin 0.0cm \topmargin -21pt \headsep 10pt \textheight |
16 |
|
|
9.0in \textwidth 6.5in \brokenpenalty=10000 |
17 |
|
|
|
18 |
|
|
% double space list of tables and figures |
19 |
|
|
\AtBeginDelayedFloats{\renewcommand{\baselinestretch}{1.66}} |
20 |
|
|
\setlength{\abovecaptionskip}{20 pt} |
21 |
|
|
\setlength{\belowcaptionskip}{30 pt} |
22 |
|
|
|
23 |
|
|
\renewcommand\citemid{\ } % no comma in optional referenc note |
24 |
|
|
|
25 |
|
|
\begin{document} |
26 |
|
|
|
27 |
|
|
\title{A gentler approach to RNEMD: Non-isotropic Velocity Scaling for computing Thermal Conductivity and Shear Viscosity} |
28 |
|
|
|
29 |
|
|
\author{Shenyu Kuang and J. Daniel |
30 |
|
|
Gezelter\footnote{Corresponding author. \ Electronic mail: gezelter@nd.edu} \\ |
31 |
|
|
Department of Chemistry and Biochemistry,\\ |
32 |
|
|
University of Notre Dame\\ |
33 |
|
|
Notre Dame, Indiana 46556} |
34 |
|
|
|
35 |
|
|
\date{\today} |
36 |
|
|
|
37 |
|
|
\maketitle |
38 |
|
|
|
39 |
|
|
\begin{doublespace} |
40 |
|
|
|
41 |
|
|
\begin{abstract} |
42 |
gezelter |
3583 |
We present a new method for introducing stable non-equilibrium |
43 |
|
|
velocity and temperature distributions in molecular dynamics |
44 |
gezelter |
3609 |
simulations of heterogeneous systems. This method extends earlier |
45 |
|
|
Reverse Non-Equilibrium Molecular Dynamics (RNEMD) methods which use |
46 |
|
|
momentum exchange swapping moves that can create non-thermal |
47 |
|
|
velocity distributions and are difficult to use for interfacial |
48 |
|
|
calculations. By using non-isotropic velocity scaling (NIVS) on the |
49 |
|
|
molecules in specific regions of a system, it is possible to impose |
50 |
|
|
momentum or thermal flux between regions of a simulation and stable |
51 |
|
|
thermal and momentum gradients can then be established. The scaling |
52 |
|
|
method we have developed conserves the total linear momentum and |
53 |
|
|
total energy of the system. To test the methods, we have computed |
54 |
|
|
the thermal conductivity of model liquid and solid systems as well |
55 |
|
|
as the interfacial thermal conductivity of a metal-water interface. |
56 |
|
|
We find that the NIVS-RNEMD improves the problematic velocity |
57 |
|
|
distributions that develop in other RNEMD methods. |
58 |
gezelter |
3524 |
\end{abstract} |
59 |
|
|
|
60 |
|
|
\newpage |
61 |
|
|
|
62 |
|
|
%\narrowtext |
63 |
|
|
|
64 |
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% |
65 |
|
|
% BODY OF TEXT |
66 |
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% |
67 |
|
|
|
68 |
|
|
\section{Introduction} |
69 |
|
|
The original formulation of Reverse Non-equilibrium Molecular Dynamics |
70 |
|
|
(RNEMD) obtains transport coefficients (thermal conductivity and shear |
71 |
|
|
viscosity) in a fluid by imposing an artificial momentum flux between |
72 |
|
|
two thin parallel slabs of material that are spatially separated in |
73 |
skuang |
3534 |
the simulation cell.\cite{MullerPlathe:1997xw,ISI:000080382700030} The |
74 |
gezelter |
3583 |
artificial flux is typically created by periodically ``swapping'' |
75 |
|
|
either the entire momentum vector $\vec{p}$ or single components of |
76 |
|
|
this vector ($p_x$) between molecules in each of the two slabs. If |
77 |
|
|
the two slabs are separated along the $z$ coordinate, the imposed flux |
78 |
|
|
is either directional ($j_z(p_x)$) or isotropic ($J_z$), and the |
79 |
|
|
response of a simulated system to the imposed momentum flux will |
80 |
|
|
typically be a velocity or thermal gradient (Fig. \ref{thermalDemo}). |
81 |
|
|
The shear viscosity ($\eta$) and thermal conductivity ($\lambda$) are |
82 |
|
|
easily obtained by assuming linear response of the system, |
83 |
gezelter |
3524 |
\begin{eqnarray} |
84 |
skuang |
3532 |
j_z(p_x) & = & -\eta \frac{\partial v_x}{\partial z}\\ |
85 |
skuang |
3575 |
J_z & = & \lambda \frac{\partial T}{\partial z} |
86 |
gezelter |
3524 |
\end{eqnarray} |
87 |
gezelter |
3600 |
RNEMD has been widely used to provide computational estimates of |
88 |
|
|
thermal conductivities and shear viscosities in a wide range of |
89 |
|
|
materials, from liquid copper to both monatomic and molecular fluids |
90 |
|
|
(e.g. ionic |
91 |
|
|
liquids).\cite{Bedrov:2000-1,Bedrov:2000,Muller-Plathe:2002,ISI:000184808400018,ISI:000231042800044,Maginn:2007,Muller-Plathe:2008,ISI:000258460400020,ISI:000258840700015,ISI:000261835100054} |
92 |
gezelter |
3524 |
|
93 |
skuang |
3574 |
\begin{figure} |
94 |
|
|
\includegraphics[width=\linewidth]{thermalDemo} |
95 |
gezelter |
3583 |
\caption{RNEMD methods impose an unphysical transfer of momentum or |
96 |
|
|
kinetic energy between a ``hot'' slab and a ``cold'' slab in the |
97 |
|
|
simulation box. The molecular system responds to this imposed flux |
98 |
|
|
by generating a momentum or temperature gradient. The slope of the |
99 |
|
|
gradient can then be used to compute transport properties (e.g. |
100 |
|
|
shear viscosity and thermal conductivity).} |
101 |
skuang |
3574 |
\label{thermalDemo} |
102 |
|
|
\end{figure} |
103 |
|
|
|
104 |
skuang |
3591 |
RNEMD is preferable in many ways to the forward NEMD |
105 |
skuang |
3592 |
methods\cite{ISI:A1988Q205300014,hess:209,Vasquez:2004fk,backer:154503,ISI:000266247600008} |
106 |
|
|
because it imposes what is typically difficult to measure (a flux or |
107 |
gezelter |
3600 |
stress) and it is typically much easier to compute the response |
108 |
gezelter |
3609 |
(momentum gradients or strains). For similar reasons, RNEMD is also |
109 |
skuang |
3592 |
preferable to slowly-converging equilibrium methods for measuring |
110 |
|
|
thermal conductivity and shear viscosity (using Green-Kubo |
111 |
skuang |
3591 |
relations\cite{daivis:541,mondello:9327} or the Helfand moment |
112 |
|
|
approach of Viscardy {\it et |
113 |
skuang |
3527 |
al.}\cite{Viscardy:2007bh,Viscardy:2007lq}) because these rely on |
114 |
gezelter |
3524 |
computing difficult to measure quantities. |
115 |
|
|
|
116 |
|
|
Another attractive feature of RNEMD is that it conserves both total |
117 |
|
|
linear momentum and total energy during the swaps (as long as the two |
118 |
|
|
molecules have the same identity), so the swapped configurations are |
119 |
|
|
typically samples from the same manifold of states in the |
120 |
|
|
microcanonical ensemble. |
121 |
|
|
|
122 |
skuang |
3588 |
Recently, Tenney and Maginn\cite{Maginn:2010} have discovered |
123 |
skuang |
3565 |
some problems with the original RNEMD swap technique. Notably, large |
124 |
|
|
momentum fluxes (equivalent to frequent momentum swaps between the |
125 |
skuang |
3575 |
slabs) can result in ``notched'', ``peaked'' and generally non-thermal |
126 |
|
|
momentum distributions in the two slabs, as well as non-linear thermal |
127 |
|
|
and velocity distributions along the direction of the imposed flux |
128 |
|
|
($z$). Tenney and Maginn obtained reasonable limits on imposed flux |
129 |
|
|
and self-adjusting metrics for retaining the usability of the method. |
130 |
gezelter |
3524 |
|
131 |
|
|
In this paper, we develop and test a method for non-isotropic velocity |
132 |
gezelter |
3600 |
scaling (NIVS) which retains the desirable features of RNEMD |
133 |
gezelter |
3524 |
(conservation of linear momentum and total energy, compatibility with |
134 |
|
|
periodic boundary conditions) while establishing true thermal |
135 |
gezelter |
3600 |
distributions in each of the two slabs. In the next section, we |
136 |
gezelter |
3583 |
present the method for determining the scaling constraints. We then |
137 |
gezelter |
3600 |
test the method on both liquids and solids as well as a non-isotropic |
138 |
|
|
liquid-solid interface and show that it is capable of providing |
139 |
gezelter |
3524 |
reasonable estimates of the thermal conductivity and shear viscosity |
140 |
gezelter |
3600 |
in all of these cases. |
141 |
gezelter |
3524 |
|
142 |
|
|
\section{Methodology} |
143 |
gezelter |
3583 |
We retain the basic idea of M\"{u}ller-Plathe's RNEMD method; the |
144 |
|
|
periodic system is partitioned into a series of thin slabs along one |
145 |
gezelter |
3524 |
axis ($z$). One of the slabs at the end of the periodic box is |
146 |
|
|
designated the ``hot'' slab, while the slab in the center of the box |
147 |
|
|
is designated the ``cold'' slab. The artificial momentum flux will be |
148 |
|
|
established by transferring momentum from the cold slab and into the |
149 |
|
|
hot slab. |
150 |
|
|
|
151 |
|
|
Rather than using momentum swaps, we use a series of velocity scaling |
152 |
gezelter |
3583 |
moves. For molecules $\{i\}$ located within the cold slab, |
153 |
gezelter |
3524 |
\begin{equation} |
154 |
skuang |
3565 |
\vec{v}_i \leftarrow \left( \begin{array}{ccc} |
155 |
|
|
x & 0 & 0 \\ |
156 |
|
|
0 & y & 0 \\ |
157 |
|
|
0 & 0 & z \\ |
158 |
gezelter |
3524 |
\end{array} \right) \cdot \vec{v}_i |
159 |
|
|
\end{equation} |
160 |
gezelter |
3600 |
where ${x, y, z}$ are a set of 3 velocity-scaling variables for each |
161 |
|
|
of the three directions in the system. Likewise, the molecules |
162 |
|
|
$\{j\}$ located in the hot slab will see a concomitant scaling of |
163 |
|
|
velocities, |
164 |
gezelter |
3524 |
\begin{equation} |
165 |
skuang |
3565 |
\vec{v}_j \leftarrow \left( \begin{array}{ccc} |
166 |
|
|
x^\prime & 0 & 0 \\ |
167 |
|
|
0 & y^\prime & 0 \\ |
168 |
|
|
0 & 0 & z^\prime \\ |
169 |
gezelter |
3524 |
\end{array} \right) \cdot \vec{v}_j |
170 |
|
|
\end{equation} |
171 |
|
|
|
172 |
|
|
Conservation of linear momentum in each of the three directions |
173 |
gezelter |
3583 |
($\alpha = x,y,z$) ties the values of the hot and cold scaling |
174 |
gezelter |
3524 |
parameters together: |
175 |
|
|
\begin{equation} |
176 |
skuang |
3528 |
P_h^\alpha + P_c^\alpha = \alpha^\prime P_h^\alpha + \alpha P_c^\alpha |
177 |
gezelter |
3524 |
\end{equation} |
178 |
|
|
where |
179 |
skuang |
3565 |
\begin{eqnarray} |
180 |
skuang |
3528 |
P_c^\alpha & = & \sum_{i = 1}^{N_c} m_i \left[\vec{v}_i\right]_\alpha \\ |
181 |
skuang |
3565 |
P_h^\alpha & = & \sum_{j = 1}^{N_h} m_j \left[\vec{v}_j\right]_\alpha |
182 |
gezelter |
3524 |
\label{eq:momentumdef} |
183 |
skuang |
3565 |
\end{eqnarray} |
184 |
skuang |
3528 |
Therefore, for each of the three directions, the hot scaling |
185 |
|
|
parameters are a simple function of the cold scaling parameters and |
186 |
gezelter |
3524 |
the instantaneous linear momentum in each of the two slabs. |
187 |
|
|
\begin{equation} |
188 |
skuang |
3528 |
\alpha^\prime = 1 + (1 - \alpha) p_\alpha |
189 |
gezelter |
3524 |
\label{eq:hotcoldscaling} |
190 |
|
|
\end{equation} |
191 |
skuang |
3528 |
where |
192 |
|
|
\begin{equation} |
193 |
|
|
p_\alpha = \frac{P_c^\alpha}{P_h^\alpha} |
194 |
|
|
\end{equation} |
195 |
|
|
for convenience. |
196 |
gezelter |
3524 |
|
197 |
|
|
Conservation of total energy also places constraints on the scaling: |
198 |
|
|
\begin{equation} |
199 |
|
|
\sum_{\alpha = x,y,z} K_h^\alpha + K_c^\alpha = \sum_{\alpha = x,y,z} |
200 |
skuang |
3565 |
\left(\alpha^\prime\right)^2 K_h^\alpha + \alpha^2 K_c^\alpha |
201 |
gezelter |
3524 |
\end{equation} |
202 |
skuang |
3575 |
where the translational kinetic energies, $K_h^\alpha$ and |
203 |
|
|
$K_c^\alpha$, are computed for each of the three directions in a |
204 |
|
|
similar manner to the linear momenta (Eq. \ref{eq:momentumdef}). |
205 |
|
|
Substituting in the expressions for the hot scaling parameters |
206 |
|
|
($\alpha^\prime$) from Eq. (\ref{eq:hotcoldscaling}), we obtain the |
207 |
gezelter |
3583 |
{\it constraint ellipsoid}: |
208 |
gezelter |
3524 |
\begin{equation} |
209 |
gezelter |
3600 |
\sum_{\alpha = x,y,z} \left( a_\alpha \alpha^2 + b_\alpha \alpha + |
210 |
|
|
c_\alpha \right) = 0 |
211 |
gezelter |
3524 |
\label{eq:constraintEllipsoid} |
212 |
|
|
\end{equation} |
213 |
|
|
where the constants are obtained from the instantaneous values of the |
214 |
|
|
linear momenta and kinetic energies for the hot and cold slabs, |
215 |
skuang |
3565 |
\begin{eqnarray} |
216 |
skuang |
3528 |
a_\alpha & = &\left(K_c^\alpha + K_h^\alpha |
217 |
|
|
\left(p_\alpha\right)^2\right) \\ |
218 |
|
|
b_\alpha & = & -2 K_h^\alpha p_\alpha \left( 1 + p_\alpha\right) \\ |
219 |
skuang |
3565 |
c_\alpha & = & K_h^\alpha p_\alpha^2 + 2 K_h^\alpha p_\alpha - K_c^\alpha |
220 |
gezelter |
3524 |
\label{eq:constraintEllipsoidConsts} |
221 |
skuang |
3565 |
\end{eqnarray} |
222 |
gezelter |
3583 |
This ellipsoid (Eq. \ref{eq:constraintEllipsoid}) defines the set of |
223 |
gezelter |
3600 |
cold slab scaling parameters which, when applied, preserve the linear |
224 |
|
|
momentum of the system in all three directions as well as total |
225 |
|
|
kinetic energy. |
226 |
gezelter |
3524 |
|
227 |
gezelter |
3600 |
The goal of using these velocity scaling variables is to transfer |
228 |
gezelter |
3609 |
kinetic energy from the cold slab to the hot slab. If the hot and |
229 |
|
|
cold slabs are separated along the z-axis, the energy flux is given |
230 |
|
|
simply by the decrease in kinetic energy of the cold bin: |
231 |
gezelter |
3524 |
\begin{equation} |
232 |
skuang |
3534 |
(1-x^2) K_c^x + (1-y^2) K_c^y + (1-z^2) K_c^z = J_z\Delta t |
233 |
gezelter |
3524 |
\end{equation} |
234 |
|
|
The expression for the energy flux can be re-written as another |
235 |
|
|
ellipsoid centered on $(x,y,z) = 0$: |
236 |
|
|
\begin{equation} |
237 |
skuang |
3604 |
\sum_{\alpha = x,y,z} K_c^\alpha \alpha^2 = \sum_{\alpha = x,y,z} |
238 |
|
|
K_c^\alpha -J_z \Delta t |
239 |
gezelter |
3524 |
\label{eq:fluxEllipsoid} |
240 |
|
|
\end{equation} |
241 |
gezelter |
3583 |
The spatial extent of the {\it thermal flux ellipsoid} is governed |
242 |
gezelter |
3600 |
both by the target flux, $J_z$ as well as the instantaneous values of |
243 |
|
|
the kinetic energy components in the cold bin. |
244 |
gezelter |
3524 |
|
245 |
|
|
To satisfy an energetic flux as well as the conservation constraints, |
246 |
gezelter |
3600 |
we must determine the points ${x,y,z}$ that lie on both the constraint |
247 |
|
|
ellipsoid (Eq. \ref{eq:constraintEllipsoid}) and the flux ellipsoid |
248 |
|
|
(Eq. \ref{eq:fluxEllipsoid}), i.e. the intersection of the two |
249 |
|
|
ellipsoids in 3-dimensional space. |
250 |
gezelter |
3524 |
|
251 |
gezelter |
3569 |
\begin{figure} |
252 |
|
|
\includegraphics[width=\linewidth]{ellipsoids} |
253 |
gezelter |
3600 |
\caption{Velocity scaling coefficients which maintain both constant |
254 |
|
|
energy and constant linear momentum of the system lie on the surface |
255 |
|
|
of the {\it constraint ellipsoid} while points which generate the |
256 |
|
|
target momentum flux lie on the surface of the {\it flux ellipsoid}. |
257 |
|
|
The velocity distributions in the cold bin are scaled by only those |
258 |
|
|
points which lie on both ellipsoids.} |
259 |
gezelter |
3569 |
\label{ellipsoids} |
260 |
|
|
\end{figure} |
261 |
|
|
|
262 |
gezelter |
3600 |
Since ellipsoids can be expressed as polynomials up to second order in |
263 |
|
|
each of the three coordinates, finding the the intersection points of |
264 |
|
|
two ellipsoids is isomorphic to finding the roots a polynomial of |
265 |
|
|
degree 16. There are a number of polynomial root-finding methods in |
266 |
|
|
the literature, [CITATIONS NEEDED] but numerically finding the roots |
267 |
|
|
of high-degree polynomials is generally an ill-conditioned |
268 |
gezelter |
3609 |
problem.[CITATION NEEDED] One simplification is to maintain velocity |
269 |
|
|
scalings that are {\it as isotropic as possible}. To do this, we |
270 |
|
|
impose $x=y$, and to treat both the constraint and flux ellipsoids as |
271 |
|
|
2-dimensional ellipses. In reduced dimensionality, the |
272 |
gezelter |
3600 |
intersecting-ellipse problem reduces to finding the roots of |
273 |
gezelter |
3609 |
polynomials of degree 4. |
274 |
gezelter |
3600 |
|
275 |
|
|
Depending on the target flux and current velocity distributions, the |
276 |
|
|
ellipsoids can have between 0 and 4 intersection points. If there are |
277 |
|
|
no intersection points, it is not possible to satisfy the constraints |
278 |
|
|
while performing a non-equilibrium scaling move, and no change is made |
279 |
|
|
to the dynamics. |
280 |
|
|
|
281 |
|
|
With multiple intersection points, any of the scaling points will |
282 |
|
|
conserve the linear momentum and kinetic energy of the system and will |
283 |
|
|
generate the correct target flux. Although this method is inherently |
284 |
|
|
non-isotropic, the goal is still to maintain the system as close to an |
285 |
|
|
isotropic fluid as possible. With this in mind, we would like the |
286 |
|
|
kinetic energies in the three different directions could become as |
287 |
|
|
close as each other as possible after each scaling. Simultaneously, |
288 |
|
|
one would also like each scaling as gentle as possible, i.e. ${x,y,z |
289 |
|
|
\rightarrow 1}$, in order to avoid large perturbation to the system. |
290 |
gezelter |
3609 |
To do this, we pick the intersection point which maintains the three |
291 |
|
|
scaling variables ${x, y, z}$ as well as the ratio of kinetic energies |
292 |
gezelter |
3600 |
${K_c^z/K_c^x, K_c^z/K_c^y}$ as close as possible to 1. |
293 |
|
|
|
294 |
|
|
After the valid scaling parameters are arrived at by solving geometric |
295 |
|
|
intersection problems in $x, y, z$ space in order to obtain cold slab |
296 |
|
|
scaling parameters, Eq. (\ref{eq:hotcoldscaling}) is used to |
297 |
|
|
determine the conjugate hot slab scaling variables. |
298 |
|
|
|
299 |
|
|
\subsection{Introducing shear stress via velocity scaling} |
300 |
gezelter |
3609 |
It is also possible to use this method to magnify the random |
301 |
|
|
fluctuations of the average momentum in each of the bins to induce a |
302 |
|
|
momentum flux. Doing this repeatedly will create a shear stress on |
303 |
|
|
the system which will respond with an easily-measured strain. The |
304 |
|
|
momentum flux (say along the $x$-direction) may be defined as: |
305 |
gezelter |
3524 |
\begin{equation} |
306 |
skuang |
3565 |
(1-x) P_c^x = j_z(p_x)\Delta t |
307 |
skuang |
3531 |
\label{eq:fluxPlane} |
308 |
gezelter |
3524 |
\end{equation} |
309 |
gezelter |
3600 |
This {\it momentum flux plane} is perpendicular to the $x$-axis, with |
310 |
|
|
its position governed both by a target value, $j_z(p_x)$ as well as |
311 |
|
|
the instantaneous value of the momentum along the $x$-direction. |
312 |
gezelter |
3524 |
|
313 |
gezelter |
3583 |
In order to satisfy a momentum flux as well as the conservation |
314 |
|
|
constraints, we must determine the points ${x,y,z}$ which lie on both |
315 |
|
|
the constraint ellipsoid (Eq. \ref{eq:constraintEllipsoid}) and the |
316 |
|
|
flux plane (Eq. \ref{eq:fluxPlane}), i.e. the intersection of an |
317 |
gezelter |
3600 |
ellipsoid and a plane in 3-dimensional space. |
318 |
gezelter |
3524 |
|
319 |
gezelter |
3600 |
In the case of momentum flux transfer, we also impose another |
320 |
gezelter |
3609 |
constraint to set the kinetic energy transfer as zero. In other |
321 |
|
|
words, we apply Eq. \ref{eq:fluxEllipsoid} and let ${J_z = 0}$. With |
322 |
gezelter |
3600 |
one variable fixed by Eq. \ref{eq:fluxPlane}, one now have a similar |
323 |
|
|
set of quartic equations to the above kinetic energy transfer problem. |
324 |
gezelter |
3524 |
|
325 |
gezelter |
3600 |
\section{Computational Details} |
326 |
gezelter |
3583 |
|
327 |
gezelter |
3609 |
We have implemented this methodology in our molecular dynamics code, |
328 |
|
|
OpenMD,\cite{Meineke:2005gd,openmd} performing the NIVS scaling moves |
329 |
|
|
after each MD step. We have tested the method in a variety of |
330 |
|
|
different systems, including homogeneous fluids (Lennard-Jones and |
331 |
|
|
SPC/E water), crystalline solids ({\sc eam}~\cite{PhysRevB.33.7983} and |
332 |
|
|
quantum Sutton-Chen ({\sc q-sc})~\cite{PhysRevB.59.3527} |
333 |
|
|
models for Gold), and heterogeneous interfaces (QSC gold - SPC/E |
334 |
|
|
water). The last of these systems would have been difficult to study |
335 |
|
|
using previous RNEMD methods, but using velocity scaling moves, we can |
336 |
|
|
even obtain estimates of the interfacial thermal conductivities ($G$). |
337 |
gezelter |
3524 |
|
338 |
gezelter |
3609 |
\subsection{Simulation Cells} |
339 |
gezelter |
3524 |
|
340 |
gezelter |
3609 |
In each of the systems studied, the dynamics was carried out in a |
341 |
|
|
rectangular simulation cell using periodic boundary conditions in all |
342 |
|
|
three dimensions. The cells were longer along the $z$ axis and the |
343 |
|
|
space was divided into $N$ slabs along this axis (typically $N=20$). |
344 |
|
|
The top slab ($n=1$) was designated the ``cold'' slab, while the |
345 |
|
|
central slab ($n= N/2 + 1$) was designated the ``hot'' slab. In all |
346 |
|
|
cases, simulations were first thermalized in canonical ensemble (NVT) |
347 |
|
|
using a Nos\'{e}-Hoover thermostat.\cite{Hoover85} then equilibrated in |
348 |
gezelter |
3600 |
microcanonical ensemble (NVE) before introducing any non-equilibrium |
349 |
|
|
method. |
350 |
skuang |
3531 |
|
351 |
gezelter |
3609 |
\subsection{RNEMD with M\"{u}ller-Plathe swaps} |
352 |
skuang |
3531 |
|
353 |
gezelter |
3609 |
In order to compare our new methodology with the original |
354 |
|
|
M\"{u}ller-Plathe swapping algorithm,\cite{ISI:000080382700030} we |
355 |
|
|
first performed simulations using the original technique. |
356 |
skuang |
3531 |
|
357 |
gezelter |
3609 |
\subsection{RNEMD with NIVS scaling} |
358 |
|
|
|
359 |
|
|
For each simulation utilizing the swapping method, a corresponding |
360 |
|
|
NIVS-RNEMD simulation was carried out using a target momentum flux set |
361 |
|
|
to produce a the same momentum or energy flux exhibited in the |
362 |
|
|
swapping simulation. |
363 |
|
|
|
364 |
|
|
To test the temperature homogeneity (and to compute transport |
365 |
|
|
coefficients), directional momentum and temperature distributions were |
366 |
|
|
accumulated for molecules in each of the slabs. |
367 |
|
|
|
368 |
|
|
\subsection{Shear viscosities} |
369 |
|
|
|
370 |
|
|
The momentum flux was calculated using the total non-physical momentum |
371 |
|
|
transferred (${P_x}$) and the data collection time ($t$): |
372 |
skuang |
3534 |
\begin{equation} |
373 |
|
|
j_z(p_x) = \frac{P_x}{2 t L_x L_y} |
374 |
|
|
\end{equation} |
375 |
gezelter |
3609 |
where $L_x$ and $L_y$ denote the $x$ and $y$ lengths of the simulation |
376 |
|
|
box. The factor of two in the denominator is present because physical |
377 |
|
|
momentum transfer occurs in two directions due to our periodic |
378 |
|
|
boundary conditions. The velocity gradient ${\langle \partial v_x |
379 |
|
|
/\partial z \rangle}$ was obtained using linear regression of the |
380 |
|
|
velocity profiles in the bins. For Lennard-Jones simulations, shear |
381 |
|
|
viscosities are reporte in reduced units (${\eta^* = \eta \sigma^2 |
382 |
|
|
(\varepsilon m)^{-1/2}}$). |
383 |
skuang |
3532 |
|
384 |
gezelter |
3609 |
\subsection{Thermal Conductivities} |
385 |
skuang |
3534 |
|
386 |
gezelter |
3609 |
The energy flux was calculated similarly to the momentum flux, using |
387 |
|
|
the total non-physical energy transferred (${E_{total}}$) and the data |
388 |
|
|
collection time $t$: |
389 |
skuang |
3534 |
\begin{equation} |
390 |
|
|
J_z = \frac{E_{total}}{2 t L_x L_y} |
391 |
|
|
\end{equation} |
392 |
gezelter |
3609 |
The temperature gradient ${\langle\partial T/\partial z\rangle}$ was |
393 |
|
|
obtained by a linear regression of the temperature profile. For |
394 |
|
|
Lennard-Jones simulations, thermal conductivities are reported in |
395 |
|
|
reduced units (${\lambda^*=\lambda \sigma^2 m^{1/2} |
396 |
|
|
k_B^{-1}\varepsilon^{-1/2}}$). |
397 |
skuang |
3534 |
|
398 |
gezelter |
3609 |
\subsection{Interfacial Thermal Conductivities} |
399 |
skuang |
3563 |
|
400 |
gezelter |
3609 |
For materials with a relatively low interfacial conductance, and in |
401 |
|
|
cases where the flux between the materials is small, the bulk regions |
402 |
|
|
on either side of an interface rapidly come to a state in which the |
403 |
|
|
two phases have relatively homogeneous (but distinct) temperatures. |
404 |
|
|
In calculating the interfacial thermal conductivity $G$, this |
405 |
|
|
assumption was made, and the conductance can be approximated as: |
406 |
skuang |
3573 |
|
407 |
|
|
\begin{equation} |
408 |
|
|
G = \frac{E_{total}}{2 t L_x L_y \left( \langle T_{gold}\rangle - |
409 |
|
|
\langle T_{water}\rangle \right)} |
410 |
|
|
\label{interfaceCalc} |
411 |
|
|
\end{equation} |
412 |
gezelter |
3609 |
where ${E_{total}}$ is the imposed non-physical kinetic energy |
413 |
|
|
transfer and ${\langle T_{gold}\rangle}$ and ${\langle |
414 |
|
|
T_{water}\rangle}$ are the average observed temperature of gold and |
415 |
|
|
water phases respectively. |
416 |
skuang |
3573 |
|
417 |
gezelter |
3609 |
\section{Results} |
418 |
skuang |
3538 |
|
419 |
gezelter |
3609 |
\subsection{Lennard-Jones Fluid} |
420 |
|
|
2592 Lennard-Jones atoms were placed in an orthorhombic cell |
421 |
|
|
${10.06\sigma \times 10.06\sigma \times 30.18\sigma}$ on a side. The |
422 |
|
|
reduced density ${\rho^* = \rho\sigma^3}$ was thus 0.85, which enabled |
423 |
|
|
direct comparison between our results and previous methods. These |
424 |
|
|
simulations were carried out with a reduced timestep ${\tau^* = |
425 |
|
|
4.6\times10^{-4}}$. For the shear viscosity calculations, the mean |
426 |
|
|
temperature was ${T^* = k_B T/\varepsilon = 0.72}$. For thermal |
427 |
|
|
conductivity calculations, simulations were first run under reduced |
428 |
|
|
temperature ${\langle T^*\rangle = 0.72}$ in the microcanonical |
429 |
|
|
ensemble, but other temperatures ([XXX, YYY, and ZZZ]) were also |
430 |
|
|
sampled. The simulations included $10^5$ steps of equilibration |
431 |
|
|
without any momentum flux, $10^5$ steps of stablization with an |
432 |
|
|
imposed momentum transfer to create a gradient, and $10^6$ steps of |
433 |
|
|
data collection under RNEMD. |
434 |
|
|
|
435 |
|
|
Our thermal conductivity calculations show that the NIVS method agrees |
436 |
|
|
well with the swapping method. Four different swap intervals were |
437 |
|
|
tested (Table \ref{thermalLJRes}). With a fixed 10 fs [WHY NOT REDUCED |
438 |
|
|
UNITS???] scaling interval, the target exchange kinetic energy |
439 |
|
|
produced equivalent kinetic energy flux as in the swapping method. |
440 |
|
|
Similar thermal gradients were observed with similar thermal flux |
441 |
|
|
under the two different methods (Figure \ref{thermalGrad}). |
442 |
|
|
|
443 |
skuang |
3563 |
\begin{table*} |
444 |
gezelter |
3609 |
\begin{minipage}{\linewidth} |
445 |
|
|
\begin{center} |
446 |
skuang |
3538 |
|
447 |
gezelter |
3609 |
\caption{Thermal conductivity (in reduced units) of a |
448 |
|
|
Lennard-Jones fluid at ${\langle T^* \rangle = 0.72}$ and |
449 |
|
|
${\rho^* = 0.85}$ for the swapping and scaling methods at |
450 |
|
|
various kinetic energy exchange rates. Uncertainties are |
451 |
|
|
indicated in parentheses.} |
452 |
|
|
|
453 |
|
|
\begin{tabular}{|cc|cc|} |
454 |
|
|
\hline |
455 |
|
|
\multicolumn{2}{|c|}{Swapping RNEMD} & |
456 |
|
|
\multicolumn{2}{|c|}{NIVS-RNEMD} \\ |
457 |
|
|
\hline |
458 |
|
|
Swap Interval (fs) & $\lambda^*_{swap}$ & Equilvalent $J_z^*$ & $\lambda^*_{scale}$\\ |
459 |
|
|
\hline |
460 |
|
|
250 & 7.03(0.34) & 0.16 & 7.30(0.10)\\ |
461 |
|
|
500 & 7.03(0.14) & 0.09 & 6.95(0.09)\\ |
462 |
|
|
1000 & 6.91(0.42) & 0.047 & 7.19(0.07)\\ |
463 |
|
|
2000 & 7.52(0.15) & 0.024 & 7.19(0.28)\\ |
464 |
|
|
\hline |
465 |
|
|
\end{tabular} |
466 |
|
|
\label{thermalLJRes} |
467 |
|
|
\end{center} |
468 |
|
|
\end{minipage} |
469 |
skuang |
3563 |
\end{table*} |
470 |
|
|
|
471 |
|
|
\begin{figure} |
472 |
skuang |
3567 |
\includegraphics[width=\linewidth]{thermalGrad} |
473 |
gezelter |
3609 |
\caption{NIVS-RNEMD method creates similar temperature gradients |
474 |
|
|
compared with the swapping method under a variety of imposed kinetic |
475 |
|
|
energy flux values.} |
476 |
skuang |
3567 |
\label{thermalGrad} |
477 |
skuang |
3563 |
\end{figure} |
478 |
|
|
|
479 |
gezelter |
3609 |
During these simulations, velocities were recorded every 1000 steps |
480 |
|
|
and was used to produce distributions of both velocity and speed in |
481 |
|
|
each of the slabs. From these distributions, we observed that under |
482 |
|
|
relatively high non-physical kinetic energy flux, the spee of |
483 |
|
|
molecules in slabs where swapping occured could deviate from the |
484 |
|
|
Maxwell-Boltzmann distribution. This behavior was also noted by Tenney |
485 |
|
|
and Maginn.\cite{Maginn:2010} Figure \ref{thermalHist} shows how these |
486 |
|
|
distributions deviate from an ideal distribution. In the ``hot'' slab, |
487 |
|
|
the probability density is notched at low speeds and has a substantial |
488 |
|
|
shoulder at higher speeds relative to the ideal MB distribution. In |
489 |
|
|
the cold slab, the opposite notching and shouldering occurs. This |
490 |
|
|
phenomenon is more obvious at higher swapping rates. |
491 |
skuang |
3563 |
|
492 |
gezelter |
3609 |
In the velocity distributions, the ideal Gaussian peak is |
493 |
|
|
substantially flattened in the hot slab, and is overly sharp (with |
494 |
|
|
truncated wings) in the cold slab. This problem is rooted in the |
495 |
|
|
mechanism of the swapping method. Continually depleting low (high) |
496 |
|
|
speed particles in the high (low) temperature slab is not complemented |
497 |
|
|
by diffusions of low (high) speed particles from neighboring slabs, |
498 |
|
|
unless the swapping rate is sufficiently small. Simutaneously, surplus |
499 |
|
|
low speed particles in the low temperature slab do not have sufficient |
500 |
|
|
time to diffuse to neighboring slabs. Since the thermal exchange rate |
501 |
|
|
must reach a minimum level to produce an observable thermal gradient, |
502 |
|
|
the swapping-method RNEMD has a relatively narrow choice of exchange |
503 |
|
|
times that can be utilized. |
504 |
skuang |
3578 |
|
505 |
gezelter |
3609 |
For comparison, NIVS-RNEMD produces a speed distribution closer to the |
506 |
|
|
Maxwell-Boltzmann curve (Figure \ref{thermalHist}). The reason for |
507 |
|
|
this is simple; upon velocity scaling, a Gaussian distribution remains |
508 |
|
|
Gaussian. Although a single scaling move is non-isotropic in three |
509 |
|
|
dimensions, our criteria for choosing a set of scaling coefficients |
510 |
|
|
helps maintain the distributions as close to isotropic as possible. |
511 |
|
|
Consequently, NIVS-RNEMD is able to impose an unphysical thermal flux |
512 |
|
|
as the previous RNEMD methods but without large perturbations to the |
513 |
|
|
velocity distributions in the two slabs. |
514 |
|
|
|
515 |
skuang |
3568 |
\begin{figure} |
516 |
skuang |
3589 |
\includegraphics[width=\linewidth]{thermalHist} |
517 |
|
|
\caption{Speed distribution for thermal conductivity using a) |
518 |
|
|
``swapping'' and b) NIVS- RNEMD methods. Shown is from the |
519 |
|
|
simulations with an exchange or equilvalent exchange interval of 250 |
520 |
skuang |
3593 |
fs. In circled areas, distributions from ``swapping'' RNEMD |
521 |
|
|
simulation have deviation from ideal Maxwell-Boltzmann distribution |
522 |
|
|
(curves fit for each distribution).} |
523 |
skuang |
3589 |
\label{thermalHist} |
524 |
skuang |
3568 |
\end{figure} |
525 |
|
|
|
526 |
gezelter |
3609 |
\subsection{Bulk SPC/E water} |
527 |
|
|
|
528 |
|
|
We compared the thermal conductivity of SPC/E water using NIVS-RNEMD |
529 |
|
|
to the work of Bedrov {\it et al.}\cite{Bedrov:2000}, which employed |
530 |
|
|
the original swapping RNEMD method. Bedrov {\it et |
531 |
gezelter |
3594 |
al.}\cite{Bedrov:2000} argued that exchange of the molecule |
532 |
skuang |
3579 |
center-of-mass velocities instead of single atom velocities in a |
533 |
gezelter |
3609 |
molecule conserves the total kinetic energy and linear momentum. This |
534 |
|
|
principle is also adopted in our simulations. Scaling was applied to |
535 |
|
|
the center-of-mass velocities of the rigid bodies of SPC/E model water |
536 |
|
|
molecules. |
537 |
skuang |
3563 |
|
538 |
gezelter |
3609 |
To construct the simulations, a simulation box consisting of 1000 |
539 |
|
|
molecules were first equilibrated under ambient pressure and |
540 |
|
|
temperature conditions using the isobaric-isothermal (NPT) |
541 |
|
|
ensemble.\cite{melchionna93} A fixed volume was chosen to match the |
542 |
|
|
average volume observed in the NPT simulations, and this was followed |
543 |
|
|
by equilibration, first in the canonical (NVT) ensemble, followed by a |
544 |
|
|
[XXX ps] period under constant-NVE conditions without any momentum |
545 |
|
|
flux. [YYY ps] was allowed to stabilize the system with an imposed |
546 |
|
|
momentum transfer to create a gradient, and [ZZZ ps] was alotted for |
547 |
|
|
data collection under RNEMD. |
548 |
|
|
|
549 |
|
|
As shown in Figure \ref{spceGrad}, temperature gradients were |
550 |
|
|
established similar to the previous work. However, the average |
551 |
|
|
temperature of our system is 300K, while that in Bedrov {\it et al.} |
552 |
|
|
is 318K, which would be attributed for part of the difference between |
553 |
|
|
the final calculation results (Table \ref{spceThermal}). [WHY DIDN'T |
554 |
|
|
WE DO 318 K?] Both methods yield values in reasonable agreement with |
555 |
|
|
experiment [DONE AT WHAT TEMPERATURE?] |
556 |
|
|
|
557 |
skuang |
3570 |
\begin{figure} |
558 |
gezelter |
3609 |
\includegraphics[width=\linewidth]{spceGrad} |
559 |
|
|
\caption{Temperature gradients in SPC/E water thermal conductivity |
560 |
|
|
simulations.} |
561 |
|
|
\label{spceGrad} |
562 |
skuang |
3570 |
\end{figure} |
563 |
|
|
|
564 |
|
|
\begin{table*} |
565 |
gezelter |
3609 |
\begin{minipage}{\linewidth} |
566 |
|
|
\begin{center} |
567 |
|
|
|
568 |
|
|
\caption{Thermal conductivity of SPC/E water under various |
569 |
|
|
imposed thermal gradients. Uncertainties are indicated in |
570 |
|
|
parentheses.} |
571 |
|
|
|
572 |
|
|
\begin{tabular}{|c|ccc|} |
573 |
|
|
\hline |
574 |
|
|
$\langle dT/dz\rangle$(K/\AA) & \multicolumn{3}{|c|}{$\lambda |
575 |
|
|
(\mathrm{W m}^{-1} \mathrm{K}^{-1})$} \\ |
576 |
|
|
& This work (300K) & Previous simulations (318K)\cite{Bedrov:2000} & |
577 |
|
|
Experiment\cite{WagnerKruse}\\ |
578 |
|
|
\hline |
579 |
|
|
0.38 & 0.816(0.044) & & 0.64\\ |
580 |
|
|
0.81 & 0.770(0.008) & 0.784 & \\ |
581 |
|
|
1.54 & 0.813(0.007) & 0.730 & \\ |
582 |
|
|
\hline |
583 |
|
|
\end{tabular} |
584 |
|
|
\label{spceThermal} |
585 |
|
|
\end{center} |
586 |
|
|
\end{minipage} |
587 |
|
|
\end{table*} |
588 |
skuang |
3570 |
|
589 |
gezelter |
3609 |
\subsection{Crystalline Gold} |
590 |
skuang |
3570 |
|
591 |
gezelter |
3609 |
To see how the method performed in a solid, we calculated thermal |
592 |
|
|
conductivities using two atomistic models for gold. Several different |
593 |
|
|
potential models have been developed that reasonably describe |
594 |
|
|
interactions in transition metals. In particular, the Embedded Atom |
595 |
|
|
Model ({\sc eam})~\cite{PhysRevB.33.7983} and Sutton-Chen ({\sc |
596 |
|
|
sc})~\cite{Chen90} potential have been used to study a wide range of |
597 |
|
|
phenomena in both bulk materials and |
598 |
|
|
nanoparticles.\cite{ISI:000207079300006,Clancy:1992,Vardeman:2008fk,Vardeman-II:2001jn,ShibataT._ja026764r,Sankaranarayanan:2005lr,Chui:2003fk,Wang:2005qy,Medasani:2007uq} |
599 |
|
|
Both potentials are based on a model of a metal which treats the |
600 |
|
|
nuclei and core electrons as pseudo-atoms embedded in the electron |
601 |
|
|
density due to the valence electrons on all of the other atoms in the |
602 |
|
|
system. The {\sc sc} potential has a simple form that closely |
603 |
|
|
resembles the Lennard Jones potential, |
604 |
|
|
\begin{equation} |
605 |
|
|
\label{eq:SCP1} |
606 |
|
|
U_{tot}=\sum _{i}\left[ \frac{1}{2}\sum _{j\neq i}D_{ij}V^{pair}_{ij}(r_{ij})-c_{i}D_{ii}\sqrt{\rho_{i}}\right] , |
607 |
|
|
\end{equation} |
608 |
|
|
where $V^{pair}_{ij}$ and $\rho_{i}$ are given by |
609 |
|
|
\begin{equation} |
610 |
|
|
\label{eq:SCP2} |
611 |
|
|
V^{pair}_{ij}(r)=\left( \frac{\alpha_{ij}}{r_{ij}}\right)^{n_{ij}}, \rho_{i}=\sum_{j\neq i}\left( \frac{\alpha_{ij}}{r_{ij}}\right) ^{m_{ij}}. |
612 |
|
|
\end{equation} |
613 |
|
|
$V^{pair}_{ij}$ is a repulsive pairwise potential that accounts for |
614 |
|
|
interactions between the pseudoatom cores. The $\sqrt{\rho_i}$ term in |
615 |
|
|
Eq. (\ref{eq:SCP1}) is an attractive many-body potential that models |
616 |
|
|
the interactions between the valence electrons and the cores of the |
617 |
|
|
pseudo-atoms. $D_{ij}$, $D_{ii}$ set the appropriate overall energy |
618 |
|
|
scale, $c_i$ scales the attractive portion of the potential relative |
619 |
|
|
to the repulsive interaction and $\alpha_{ij}$ is a length parameter |
620 |
|
|
that assures a dimensionless form for $\rho$. These parameters are |
621 |
|
|
tuned to various experimental properties such as the density, cohesive |
622 |
|
|
energy, and elastic moduli for FCC transition metals. The quantum |
623 |
|
|
Sutton-Chen ({\sc q-sc}) formulation matches these properties while |
624 |
|
|
including zero-point quantum corrections for different transition |
625 |
|
|
metals.\cite{PhysRevB.59.3527} The {\sc eam} functional forms differ |
626 |
|
|
slightly from {\sc sc} but the overall method is very similar. |
627 |
skuang |
3570 |
|
628 |
gezelter |
3609 |
In this work, we have utilized both the {\sc eam} and the {\sc q-sc} |
629 |
|
|
potentials to test the behavior of scaling RNEMD. |
630 |
skuang |
3570 |
|
631 |
gezelter |
3609 |
A face-centered-cubic (FCC) lattice was prepared containing 2880 Au |
632 |
|
|
atoms. [LxMxN UNIT CELLS]. Simulations were run both with and |
633 |
|
|
without isobaric-isothermal (NPT)~\cite{melchionna93} |
634 |
|
|
pre-equilibration at a target pressure of 1 atm. When equilibrated |
635 |
|
|
under NPT conditions, our simulation box expanded by approximately 1\% |
636 |
gezelter |
3610 |
[IN VOLUME OR LINEAR DIMENSIONS ?]. Following adjustment of the box |
637 |
|
|
volume, equilibrations in both the canonical and microcanonical |
638 |
|
|
ensembles were carried out. With the simulation cell divided evenly |
639 |
|
|
into 10 slabs, different thermal gradients were established by |
640 |
|
|
applying a set of target thermal transfer fluxes. |
641 |
skuang |
3570 |
|
642 |
gezelter |
3609 |
The results for the thermal conductivity of gold are shown in Table |
643 |
|
|
\ref{AuThermal}. In these calculations, the end and middle slabs were |
644 |
gezelter |
3610 |
excluded in thermal gradient linear regession. {\sc eam} predicts |
645 |
gezelter |
3609 |
slightly larger thermal conductivities than {\sc q-sc}. However, both |
646 |
|
|
values are smaller than experimental value by a factor of more than |
647 |
|
|
200. This behavior has been observed previously by Richardson and |
648 |
|
|
Clancy, and has been attributed to the lack of electronic effects in |
649 |
|
|
these force fields.\cite{Clancy:1992} The non-equilibrium MD method |
650 |
gezelter |
3610 |
employed in their simulations produced comparable results to ours. It |
651 |
|
|
should be noted that the density of the metal being simulated also |
652 |
|
|
greatly affects the thermal conductivity. With an expanded lattice, |
653 |
|
|
lower thermal conductance is expected (and observed). We also observed |
654 |
|
|
a decrease in thermal conductance at higher temperatures, a trend that |
655 |
|
|
agrees with experimental measurements [PAGE |
656 |
gezelter |
3609 |
NUMBERS?].\cite{AshcroftMermin} |
657 |
skuang |
3570 |
|
658 |
gezelter |
3609 |
\begin{table*} |
659 |
|
|
\begin{minipage}{\linewidth} |
660 |
|
|
\begin{center} |
661 |
|
|
|
662 |
|
|
\caption{Calculated thermal conductivity of crystalline gold |
663 |
|
|
using two related force fields. Calculations were done at both |
664 |
|
|
experimental and equilibrated densities and at a range of |
665 |
|
|
temperatures and thermal flux rates. Uncertainties are |
666 |
|
|
indicated in parentheses. [CLANCY COMPARISON? SWAPPING |
667 |
|
|
COMPARISON?]} |
668 |
|
|
|
669 |
|
|
\begin{tabular}{|c|c|c|cc|} |
670 |
|
|
\hline |
671 |
|
|
Force Field & $\rho$ (g/cm$^3$) & ${\langle T\rangle}$ (K) & |
672 |
|
|
$\langle dT/dz\rangle$ (K/\AA) & $\lambda (\mathrm{W m}^{-1} \mathrm{K}^{-1})$\\ |
673 |
|
|
\hline |
674 |
|
|
\multirow{7}{*}{\sc q-sc} & \multirow{3}{*}{19.188} & \multirow{3}{*}{300} & 1.44 & 1.10(0.06)\\ |
675 |
|
|
& & & 2.86 & 1.08(0.05)\\ |
676 |
|
|
& & & 5.14 & 1.15(0.07)\\\cline{2-5} |
677 |
|
|
& \multirow{4}{*}{19.263} & \multirow{2}{*}{300} & 2.31 & 1.25(0.06)\\ |
678 |
|
|
& & & 3.02 & 1.26(0.05)\\\cline{3-5} |
679 |
|
|
& & \multirow{2}{*}{575} & 3.02 & 1.02(0.07)\\ |
680 |
|
|
& & & 4.84 & 0.92(0.05)\\ |
681 |
|
|
\hline |
682 |
|
|
\multirow{8}{*}{\sc eam} & \multirow{3}{*}{19.045} & \multirow{3}{*}{300} & 1.24 & 1.24(0.16)\\ |
683 |
|
|
& & & 2.06 & 1.37(0.04)\\ |
684 |
|
|
& & & 2.55 & 1.41(0.07)\\\cline{2-5} |
685 |
|
|
& \multirow{5}{*}{19.263} & \multirow{3}{*}{300} & 1.06 & 1.45(0.13)\\ |
686 |
|
|
& & & 2.04 & 1.41(0.07)\\ |
687 |
|
|
& & & 2.41 & 1.53(0.10)\\\cline{3-5} |
688 |
|
|
& & \multirow{2}{*}{575} & 2.82 & 1.08(0.03)\\ |
689 |
|
|
& & & 4.14 & 1.08(0.05)\\ |
690 |
|
|
\hline |
691 |
|
|
\end{tabular} |
692 |
|
|
\label{AuThermal} |
693 |
|
|
\end{center} |
694 |
|
|
\end{minipage} |
695 |
skuang |
3580 |
\end{table*} |
696 |
|
|
|
697 |
gezelter |
3609 |
\subsection{Thermal Conductance at the Au/H$_2$O interface} |
698 |
|
|
The most attractive aspect of the scaling approach for RNEMD is the |
699 |
|
|
ability to use the method in non-homogeneous systems, where molecules |
700 |
|
|
of different identities are segregated in different slabs. To test |
701 |
|
|
this application, we simulated a Gold (111) / water interface. To |
702 |
|
|
construct the interface, a box containing a lattice of 1188 Au atoms |
703 |
|
|
(with the 111 surface in the +z and -z directions) was allowed to |
704 |
|
|
relax under ambient temperature and pressure. A separate (but |
705 |
|
|
identically sized) box of SPC/E water was also equilibrated at ambient |
706 |
|
|
conditions. The two boxes were combined by removing all water |
707 |
|
|
molecules withing 3 \AA radius of any gold atom. The final |
708 |
|
|
configuration contained 1862 SPC/E water molecules. |
709 |
skuang |
3580 |
|
710 |
gezelter |
3609 |
After simulations of bulk water and crystal gold, a mixture system was |
711 |
|
|
constructed, consisting of 1188 Au atoms and 1862 H$_2$O |
712 |
|
|
molecules. Spohr potential was adopted in depicting the interaction |
713 |
|
|
between metal atom and water molecule.\cite{ISI:000167766600035} A |
714 |
|
|
similar protocol of equilibration was followed. Several thermal |
715 |
|
|
gradients was built under different target thermal flux. It was found |
716 |
|
|
out that compared to our previous simulation systems, the two phases |
717 |
|
|
could have large temperature difference even under a relatively low |
718 |
|
|
thermal flux. |
719 |
|
|
|
720 |
|
|
|
721 |
skuang |
3581 |
After simulations of homogeneous water and gold systems using |
722 |
|
|
NIVS-RNEMD method were proved valid, calculation of gold/water |
723 |
|
|
interfacial thermal conductivity was followed. It is found out that |
724 |
|
|
the low interfacial conductance is probably due to the hydrophobic |
725 |
skuang |
3595 |
surface in our system. Figure \ref{interface} (a) demonstrates mass |
726 |
skuang |
3581 |
density change along $z$-axis, which is perpendicular to the |
727 |
|
|
gold/water interface. It is observed that water density significantly |
728 |
|
|
decreases when approaching the surface. Under this low thermal |
729 |
|
|
conductance, both gold and water phase have sufficient time to |
730 |
|
|
eliminate temperature difference inside respectively (Figure |
731 |
skuang |
3595 |
\ref{interface} b). With indistinguishable temperature difference |
732 |
skuang |
3581 |
within respective phase, it is valid to assume that the temperature |
733 |
|
|
difference between gold and water on surface would be approximately |
734 |
|
|
the same as the difference between the gold and water phase. This |
735 |
|
|
assumption enables convenient calculation of $G$ using |
736 |
|
|
Eq. \ref{interfaceCalc} instead of measuring temperatures of thin |
737 |
|
|
layer of water and gold close enough to surface, which would have |
738 |
|
|
greater fluctuation and lower accuracy. Reported results (Table |
739 |
|
|
\ref{interfaceRes}) are of two orders of magnitude smaller than our |
740 |
|
|
calculations on homogeneous systems, and thus have larger relative |
741 |
|
|
errors than our calculation results on homogeneous systems. |
742 |
skuang |
3573 |
|
743 |
skuang |
3571 |
\begin{figure} |
744 |
skuang |
3595 |
\includegraphics[width=\linewidth]{interface} |
745 |
|
|
\caption{Simulation results for Gold/Water interfacial thermal |
746 |
|
|
conductivity: (a) Significant water density decrease is observed on |
747 |
skuang |
3597 |
crystalline gold surface, which indicates low surface contact and |
748 |
|
|
leads to low thermal conductance. (b) Temperature profiles for a |
749 |
|
|
series of simulations. Temperatures of different slabs in the same |
750 |
|
|
phase show no significant differences.} |
751 |
skuang |
3595 |
\label{interface} |
752 |
skuang |
3571 |
\end{figure} |
753 |
|
|
|
754 |
skuang |
3572 |
\begin{table*} |
755 |
|
|
\begin{minipage}{\linewidth} |
756 |
|
|
\begin{center} |
757 |
|
|
|
758 |
|
|
\caption{Calculation results for interfacial thermal conductivity |
759 |
|
|
at ${\langle T\rangle \sim}$ 300K at various thermal exchange |
760 |
|
|
rates. Errors of calculations in parentheses. } |
761 |
|
|
|
762 |
|
|
\begin{tabular}{cccc} |
763 |
|
|
\hline |
764 |
skuang |
3606 |
$J_z$ (MW/m$^2$) & $T_{gold}$ (K) & $T_{water}$ (K) & $G$ |
765 |
|
|
(MW/m$^2$/K)\\ |
766 |
skuang |
3572 |
\hline |
767 |
skuang |
3573 |
98.0 & 355.2 & 295.8 & 1.65(0.21) \\ |
768 |
|
|
78.8 & 343.8 & 298.0 & 1.72(0.32) \\ |
769 |
|
|
73.6 & 344.3 & 298.0 & 1.59(0.24) \\ |
770 |
|
|
49.2 & 330.1 & 300.4 & 1.65(0.35) \\ |
771 |
skuang |
3572 |
\hline |
772 |
|
|
\end{tabular} |
773 |
skuang |
3574 |
\label{interfaceRes} |
774 |
skuang |
3572 |
\end{center} |
775 |
|
|
\end{minipage} |
776 |
|
|
\end{table*} |
777 |
|
|
|
778 |
skuang |
3576 |
\subsection{Shear Viscosity} |
779 |
|
|
Our calculations (Table \ref{shearRate}) shows that scale RNEMD method |
780 |
|
|
produced comparable shear viscosity to swap RNEMD method. In Table |
781 |
|
|
\ref{shearRate}, the names of the calculated samples are devided into |
782 |
|
|
two parts. The first number refers to total slabs in one simulation |
783 |
|
|
box. The second number refers to the swapping interval in swap method, or |
784 |
|
|
in scale method the equilvalent swapping interval that the same |
785 |
|
|
momentum flux would theoretically result in swap method. All the scale |
786 |
|
|
method results were from simulations that had a scaling interval of 10 |
787 |
|
|
time steps. The average molecular momentum gradients of these samples |
788 |
skuang |
3590 |
are shown in Figure \ref{shear} (a) and (b). |
789 |
skuang |
3576 |
|
790 |
|
|
\begin{table*} |
791 |
|
|
\begin{minipage}{\linewidth} |
792 |
|
|
\begin{center} |
793 |
|
|
|
794 |
|
|
\caption{Calculation results for shear viscosity of Lennard-Jones |
795 |
|
|
fluid at ${T^* = 0.72}$ and ${\rho^* = 0.85}$, with swap and scale |
796 |
|
|
methods at various momentum exchange rates. Results in reduced |
797 |
|
|
unit. Errors of calculations in parentheses. } |
798 |
|
|
|
799 |
skuang |
3601 |
\begin{tabular}{ccccc} |
800 |
|
|
Swapping method & & & NIVS-RNEMD & \\ |
801 |
skuang |
3576 |
\hline |
802 |
skuang |
3601 |
Swap Interval (fs) & $\eta^*_{swap}$ & & Equilvalent $j_p^*(v_x)$ & |
803 |
skuang |
3598 |
$\eta^*_{scale}$\\ |
804 |
skuang |
3576 |
\hline |
805 |
skuang |
3601 |
500 & 3.64(0.05) & & 0.09 & 3.76(0.09)\\ |
806 |
|
|
1000 & 3.52(0.16) & & 0.046 & 3.66(0.06)\\ |
807 |
|
|
2000 & 3.72(0.05) & & 0.024 & 3.32(0.18)\\ |
808 |
|
|
2500 & 3.42(0.06) & & 0.019 & 3.43(0.08)\\ |
809 |
skuang |
3576 |
\hline |
810 |
|
|
\end{tabular} |
811 |
|
|
\label{shearRate} |
812 |
|
|
\end{center} |
813 |
|
|
\end{minipage} |
814 |
|
|
\end{table*} |
815 |
|
|
|
816 |
|
|
\begin{figure} |
817 |
skuang |
3590 |
\includegraphics[width=\linewidth]{shear} |
818 |
|
|
\caption{Average momentum gradients in shear viscosity simulations, |
819 |
|
|
using (a) ``swapping'' method and (b) NIVS-RNEMD method |
820 |
|
|
respectively. (c) Temperature difference among x and y, z dimensions |
821 |
|
|
observed when using NIVS-RNEMD with equivalent exchange interval of |
822 |
|
|
500 fs.} |
823 |
|
|
\label{shear} |
824 |
skuang |
3576 |
\end{figure} |
825 |
|
|
|
826 |
|
|
However, observations of temperatures along three dimensions show that |
827 |
|
|
inhomogeneity occurs in scaling RNEMD simulations, particularly in the |
828 |
skuang |
3590 |
two slabs which were scaled. Figure \ref{shear} (c) indicate that with |
829 |
skuang |
3576 |
relatively large imposed momentum flux, the temperature difference among $x$ |
830 |
|
|
and the other two dimensions was significant. This would result from the |
831 |
|
|
algorithm of scaling method. From Eq. \ref{eq:fluxPlane}, after |
832 |
|
|
momentum gradient is set up, $P_c^x$ would be roughly stable |
833 |
|
|
($<0$). Consequently, scaling factor $x$ would most probably larger |
834 |
|
|
than 1. Therefore, the kinetic energy in $x$-dimension $K_c^x$ would |
835 |
|
|
keep increase after most scaling steps. And if there is not enough time |
836 |
|
|
for the kinetic energy to exchange among different dimensions and |
837 |
|
|
different slabs, the system would finally build up temperature |
838 |
|
|
(kinetic energy) difference among the three dimensions. Also, between |
839 |
|
|
$y$ and $z$ dimensions in the scaled slabs, temperatures of $z$-axis |
840 |
|
|
are closer to neighbor slabs. This is due to momentum transfer along |
841 |
|
|
$z$ dimension between slabs. |
842 |
|
|
|
843 |
|
|
Although results between scaling and swapping methods are comparable, |
844 |
|
|
the inherent temperature inhomogeneity even in relatively low imposed |
845 |
|
|
exchange momentum flux simulations makes scaling RNEMD method less |
846 |
|
|
attractive than swapping RNEMD in shear viscosity calculation. |
847 |
|
|
|
848 |
skuang |
3574 |
\section{Conclusions} |
849 |
|
|
NIVS-RNEMD simulation method is developed and tested on various |
850 |
skuang |
3581 |
systems. Simulation results demonstrate its validity in thermal |
851 |
|
|
conductivity calculations, from Lennard-Jones fluid to multi-atom |
852 |
|
|
molecule like water and metal crystals. NIVS-RNEMD improves |
853 |
|
|
non-Boltzmann-Maxwell distributions, which exist in previous RNEMD |
854 |
|
|
methods. Furthermore, it develops a valid means for unphysical thermal |
855 |
|
|
transfer between different species of molecules, and thus extends its |
856 |
|
|
applicability to interfacial systems. Our calculation of gold/water |
857 |
|
|
interfacial thermal conductivity demonstrates this advantage over |
858 |
|
|
previous RNEMD methods. NIVS-RNEMD has also limited application on |
859 |
|
|
shear viscosity calculations, but could cause temperature difference |
860 |
|
|
among different dimensions under high momentum flux. Modification is |
861 |
|
|
necessary to extend the applicability of NIVS-RNEMD in shear viscosity |
862 |
|
|
calculations. |
863 |
skuang |
3572 |
|
864 |
gezelter |
3524 |
\section{Acknowledgments} |
865 |
|
|
Support for this project was provided by the National Science |
866 |
|
|
Foundation under grant CHE-0848243. Computational time was provided by |
867 |
|
|
the Center for Research Computing (CRC) at the University of Notre |
868 |
|
|
Dame. \newpage |
869 |
|
|
|
870 |
gezelter |
3583 |
\bibliographystyle{aip} |
871 |
gezelter |
3524 |
\bibliography{nivsRnemd} |
872 |
gezelter |
3583 |
|
873 |
gezelter |
3524 |
\end{doublespace} |
874 |
|
|
\end{document} |
875 |
|
|
|