ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/group/trunk/nivsRnemd/nivsRnemd.tex
Revision: 3577
Committed: Thu Mar 18 00:53:20 2010 UTC (14 years, 3 months ago) by skuang
Content type: application/x-tex
File size: 31023 byte(s)
Log Message:
modify results of LJ thermal.

File Contents

# Content
1 \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 %\usepackage{booktabs}
10 %\usepackage{bibentry}
11 %\usepackage{mathrsfs}
12 \usepackage[ref]{overcite}
13 \pagestyle{plain} \pagenumbering{arabic} \oddsidemargin 0.0cm
14 \evensidemargin 0.0cm \topmargin -21pt \headsep 10pt \textheight
15 9.0in \textwidth 6.5in \brokenpenalty=10000
16
17 % double space list of tables and figures
18 \AtBeginDelayedFloats{\renewcommand{\baselinestretch}{1.66}}
19 \setlength{\abovecaptionskip}{20 pt}
20 \setlength{\belowcaptionskip}{30 pt}
21
22 \renewcommand\citemid{\ } % no comma in optional referenc note
23
24 \begin{document}
25
26 \title{A gentler approach to RNEMD: Non-isotropic Velocity Scaling for computing Thermal Conductivity and Shear Viscosity}
27
28 \author{Shenyu Kuang and J. Daniel
29 Gezelter\footnote{Corresponding author. \ Electronic mail: gezelter@nd.edu} \\
30 Department of Chemistry and Biochemistry,\\
31 University of Notre Dame\\
32 Notre Dame, Indiana 46556}
33
34 \date{\today}
35
36 \maketitle
37
38 \begin{doublespace}
39
40 \begin{abstract}
41
42 \end{abstract}
43
44 \newpage
45
46 %\narrowtext
47
48 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
49 % BODY OF TEXT
50 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
51
52
53
54 \section{Introduction}
55 The original formulation of Reverse Non-equilibrium Molecular Dynamics
56 (RNEMD) obtains transport coefficients (thermal conductivity and shear
57 viscosity) in a fluid by imposing an artificial momentum flux between
58 two thin parallel slabs of material that are spatially separated in
59 the simulation cell.\cite{MullerPlathe:1997xw,ISI:000080382700030} The
60 artificial flux is typically created by periodically ``swapping'' either
61 the entire momentum vector $\vec{p}$ or single components of this
62 vector ($p_x$) between molecules in each of the two slabs. If the two
63 slabs are separated along the $z$ coordinate, the imposed flux is either
64 directional ($j_z(p_x)$) or isotropic ($J_z$), and the response of a
65 simulated system to the imposed momentum flux will typically be a
66 velocity or thermal gradient (Fig. \ref{thermalDemo}). The transport
67 coefficients (shear viscosity and thermal conductivity) are easily
68 obtained by assuming linear response of the system,
69 \begin{eqnarray}
70 j_z(p_x) & = & -\eta \frac{\partial v_x}{\partial z}\\
71 J_z & = & \lambda \frac{\partial T}{\partial z}
72 \end{eqnarray}
73 RNEMD has been widely used to provide computational estimates of thermal
74 conductivities and shear viscosities in a wide range of materials,
75 from liquid copper to monatomic liquids to molecular fluids
76 (e.g. ionic liquids).\cite{ISI:000246190100032}
77
78 \begin{figure}
79 \includegraphics[width=\linewidth]{thermalDemo}
80 \caption{Demostration of thermal gradient estalished by RNEMD
81 method. Physical thermal flow directs from high temperature region
82 to low temperature region. Unphysical thermal transfer counteracts
83 it and maintains a steady thermal gradient.}
84 \label{thermalDemo}
85 \end{figure}
86
87 RNEMD is preferable in many ways to the forward NEMD methods because
88 it imposes what is typically difficult to measure (a flux or stress)
89 and it is typically much easier to compute momentum gradients or
90 strains (the response). For similar reasons, RNEMD is also preferable
91 to slowly-converging equilibrium methods for measuring thermal
92 conductivity and shear viscosity (using Green-Kubo relations or the
93 Helfand moment approach of Viscardy {\it et
94 al.}\cite{Viscardy:2007bh,Viscardy:2007lq}) because these rely on
95 computing difficult to measure quantities.
96
97 Another attractive feature of RNEMD is that it conserves both total
98 linear momentum and total energy during the swaps (as long as the two
99 molecules have the same identity), so the swapped configurations are
100 typically samples from the same manifold of states in the
101 microcanonical ensemble.
102
103 Recently, Tenney and Maginn\cite{ISI:000273472300004} have discovered
104 some problems with the original RNEMD swap technique. Notably, large
105 momentum fluxes (equivalent to frequent momentum swaps between the
106 slabs) can result in ``notched'', ``peaked'' and generally non-thermal
107 momentum distributions in the two slabs, as well as non-linear thermal
108 and velocity distributions along the direction of the imposed flux
109 ($z$). Tenney and Maginn obtained reasonable limits on imposed flux
110 and self-adjusting metrics for retaining the usability of the method.
111
112 In this paper, we develop and test a method for non-isotropic velocity
113 scaling (NIVS-RNEMD) which retains the desirable features of RNEMD
114 (conservation of linear momentum and total energy, compatibility with
115 periodic boundary conditions) while establishing true thermal
116 distributions in each of the two slabs. In the next section, we
117 develop the method for determining the scaling constraints. We then
118 test the method on both single component, multi-component, and
119 non-isotropic mixtures and show that it is capable of providing
120 reasonable estimates of the thermal conductivity and shear viscosity
121 in these cases.
122
123 \section{Methodology}
124 We retain the basic idea of Muller-Plathe's RNEMD method; the periodic
125 system is partitioned into a series of thin slabs along a particular
126 axis ($z$). One of the slabs at the end of the periodic box is
127 designated the ``hot'' slab, while the slab in the center of the box
128 is designated the ``cold'' slab. The artificial momentum flux will be
129 established by transferring momentum from the cold slab and into the
130 hot slab.
131
132 Rather than using momentum swaps, we use a series of velocity scaling
133 moves. For molecules $\{i\}$ located within the cold slab,
134 \begin{equation}
135 \vec{v}_i \leftarrow \left( \begin{array}{ccc}
136 x & 0 & 0 \\
137 0 & y & 0 \\
138 0 & 0 & z \\
139 \end{array} \right) \cdot \vec{v}_i
140 \end{equation}
141 where ${x, y, z}$ are a set of 3 scaling variables for each of the
142 three directions in the system. Likewise, the molecules $\{j\}$
143 located in the hot slab will see a concomitant scaling of velocities,
144 \begin{equation}
145 \vec{v}_j \leftarrow \left( \begin{array}{ccc}
146 x^\prime & 0 & 0 \\
147 0 & y^\prime & 0 \\
148 0 & 0 & z^\prime \\
149 \end{array} \right) \cdot \vec{v}_j
150 \end{equation}
151
152 Conservation of linear momentum in each of the three directions
153 ($\alpha = x,y,z$) ties the values of the hot and cold bin scaling
154 parameters together:
155 \begin{equation}
156 P_h^\alpha + P_c^\alpha = \alpha^\prime P_h^\alpha + \alpha P_c^\alpha
157 \end{equation}
158 where
159 \begin{eqnarray}
160 P_c^\alpha & = & \sum_{i = 1}^{N_c} m_i \left[\vec{v}_i\right]_\alpha \\
161 P_h^\alpha & = & \sum_{j = 1}^{N_h} m_j \left[\vec{v}_j\right]_\alpha
162 \label{eq:momentumdef}
163 \end{eqnarray}
164 Therefore, for each of the three directions, the hot scaling
165 parameters are a simple function of the cold scaling parameters and
166 the instantaneous linear momentum in each of the two slabs.
167 \begin{equation}
168 \alpha^\prime = 1 + (1 - \alpha) p_\alpha
169 \label{eq:hotcoldscaling}
170 \end{equation}
171 where
172 \begin{equation}
173 p_\alpha = \frac{P_c^\alpha}{P_h^\alpha}
174 \end{equation}
175 for convenience.
176
177 Conservation of total energy also places constraints on the scaling:
178 \begin{equation}
179 \sum_{\alpha = x,y,z} K_h^\alpha + K_c^\alpha = \sum_{\alpha = x,y,z}
180 \left(\alpha^\prime\right)^2 K_h^\alpha + \alpha^2 K_c^\alpha
181 \end{equation}
182 where the translational kinetic energies, $K_h^\alpha$ and
183 $K_c^\alpha$, are computed for each of the three directions in a
184 similar manner to the linear momenta (Eq. \ref{eq:momentumdef}).
185 Substituting in the expressions for the hot scaling parameters
186 ($\alpha^\prime$) from Eq. (\ref{eq:hotcoldscaling}), we obtain the
187 {\it constraint ellipsoid equation}:
188 \begin{equation}
189 \sum_{\alpha = x,y,z} a_\alpha \alpha^2 + b_\alpha \alpha + c_\alpha = 0
190 \label{eq:constraintEllipsoid}
191 \end{equation}
192 where the constants are obtained from the instantaneous values of the
193 linear momenta and kinetic energies for the hot and cold slabs,
194 \begin{eqnarray}
195 a_\alpha & = &\left(K_c^\alpha + K_h^\alpha
196 \left(p_\alpha\right)^2\right) \\
197 b_\alpha & = & -2 K_h^\alpha p_\alpha \left( 1 + p_\alpha\right) \\
198 c_\alpha & = & K_h^\alpha p_\alpha^2 + 2 K_h^\alpha p_\alpha - K_c^\alpha
199 \label{eq:constraintEllipsoidConsts}
200 \end{eqnarray}
201 This ellipsoid equation defines the set of cold slab scaling
202 parameters which can be applied while preserving both linear momentum
203 in all three directions as well as kinetic energy.
204
205 The goal of using velocity scaling variables is to transfer linear
206 momentum or kinetic energy from the cold slab to the hot slab. If the
207 hot and cold slabs are separated along the z-axis, the energy flux is
208 given simply by the decrease in kinetic energy of the cold bin:
209 \begin{equation}
210 (1-x^2) K_c^x + (1-y^2) K_c^y + (1-z^2) K_c^z = J_z\Delta t
211 \end{equation}
212 The expression for the energy flux can be re-written as another
213 ellipsoid centered on $(x,y,z) = 0$:
214 \begin{equation}
215 x^2 K_c^x + y^2 K_c^y + z^2 K_c^z = K_c^x + K_c^y + K_c^z - J_z\Delta t
216 \label{eq:fluxEllipsoid}
217 \end{equation}
218 The spatial extent of the {\it thermal flux ellipsoid equation} is
219 governed both by a targetted value, $J_z$ as well as the instantaneous
220 values of the kinetic energy components in the cold bin.
221
222 To satisfy an energetic flux as well as the conservation constraints,
223 it is sufficient to determine the points ${x,y,z}$ which lie on both
224 the constraint ellipsoid (Eq. \ref{eq:constraintEllipsoid}) and the
225 flux ellipsoid (Eq. \ref{eq:fluxEllipsoid}), i.e. the intersection of
226 the two ellipsoids in 3-dimensional space.
227
228 \begin{figure}
229 \includegraphics[width=\linewidth]{ellipsoids}
230 \caption{Scaling points which maintain both constant energy and
231 constant linear momentum of the system lie on the surface of the
232 {\it constraint ellipsoid} while points which generate the target
233 momentum flux lie on the surface of the {\it flux ellipsoid}. The
234 velocity distributions in the cold bin are scaled by only those
235 points which lie on both ellipsoids.}
236 \label{ellipsoids}
237 \end{figure}
238
239 One may also define momentum flux (say along the $x$-direction) as:
240 \begin{equation}
241 (1-x) P_c^x = j_z(p_x)\Delta t
242 \label{eq:fluxPlane}
243 \end{equation}
244 The above {\it momentum flux equation} is essentially a plane which is
245 perpendicular to the $x$-axis, with its position governed both by a
246 target value, $j_z(p_x)$ as well as the instantaneous value of the
247 momentum along the $x$-direction.
248
249 Similarly, to satisfy a momentum flux as well as the conservation
250 constraints, it is sufficient to determine the points ${x,y,z}$ which
251 lie on both the constraint ellipsoid (Eq. \ref{eq:constraintEllipsoid})
252 and the flux plane (Eq. \ref{eq:fluxPlane}), i.e. the intersection of
253 an ellipsoid and a plane in 3-dimensional space.
254
255 To summarize, by solving respective equation sets, one can determine
256 possible sets of scaling variables for cold slab. And corresponding
257 sets of scaling variables for hot slab can be determine as well.
258
259 The following problem will be choosing an optimal set of scaling
260 variables among the possible sets. Although this method is inherently
261 non-isotropic, the goal is still to maintain the system as isotropic
262 as possible. Under this consideration, one would like the kinetic
263 energies in different directions could become as close as each other
264 after each scaling. Simultaneously, one would also like each scaling
265 as gentle as possible, i.e. ${x,y,z \rightarrow 1}$, in order to avoid
266 large perturbation to the system. Therefore, one approach to obtain the
267 scaling variables would be constructing an criteria function, with
268 constraints as above equation sets, and solving the function's minimum
269 by method like Lagrange multipliers.
270
271 In order to save computation time, we have a different approach to a
272 relatively good set of scaling variables with much less calculation
273 than above. Here is the detail of our simplification of the problem.
274
275 In the case of kinetic energy transfer, we impose another constraint
276 ${x = y}$, into the equation sets. Consequently, there are two
277 variables left. And now one only needs to solve a set of two {\it
278 ellipses equations}. This problem would be transformed into solving
279 one quartic equation for one of the two variables. There are known
280 generic methods that solve real roots of quartic equations. Then one
281 can determine the other variable and obtain sets of scaling
282 variables. Among these sets, one can apply the above criteria to
283 choose the best set, while much faster with only a few sets to choose.
284
285 In the case of momentum flux transfer, we impose another constraint to
286 set the kinetic energy transfer as zero. In another word, we apply
287 Eq. \ref{eq:fluxEllipsoid} and let ${J_z = 0}$. After that, with one
288 variable fixed by Eq. \ref{eq:fluxPlane}, one now have a similar set
289 of equations on the above kinetic energy transfer problem. Therefore,
290 an approach similar to the above would be sufficient for this as well.
291
292 \section{Computational Details}
293 \subsection{Lennard-Jones Fluid}
294 Our simulation consists of a series of systems. All of these
295 simulations were run with the OpenMD simulation software
296 package\cite{Meineke:2005gd} integrated with RNEMD codes.
297
298 A Lennard-Jones fluid system was built and tested first. In order to
299 compare our method with swapping RNEMD, a series of simulations were
300 performed to calculate the shear viscosity and thermal conductivity of
301 argon. 2592 atoms were in a orthorhombic cell, which was ${10.06\sigma
302 \times 10.06\sigma \times 30.18\sigma}$ by size. The reduced density
303 ${\rho^* = \rho\sigma^3}$ was thus 0.85, which enabled direct
304 comparison between our results and others. These simulations used
305 velocity Verlet algorithm with reduced timestep ${\tau^* =
306 4.6\times10^{-4}}$.
307
308 For shear viscosity calculation, the reduced temperature was ${T^* =
309 k_B T/\varepsilon = 0.72}$. Simulations were first equilibrated in canonical
310 ensemble (NVT), then equilibrated in microcanonical ensemble
311 (NVE). Establishing and stablizing momentum gradient were followed
312 also in NVE ensemble. For the swapping part, Muller-Plathe's algorithm was
313 adopted.\cite{ISI:000080382700030} The simulation box was under
314 periodic boundary condition, and devided into ${N = 20}$ slabs. In each swap,
315 the top slab ${(n = 1)}$ exchange the most negative $x$ momentum with the
316 most positive $x$ momentum in the center slab ${(n = N/2 + 1)}$. Referred
317 to Tenney {\it et al.}\cite{ISI:000273472300004}, a series of swapping
318 frequency were chosen. According to each result from swapping
319 RNEMD, scaling RNEMD simulations were run with the target momentum
320 flux set to produce a similar momentum flux, and consequently shear
321 rate. Furthermore, various scaling frequencies can be tested for one
322 single swapping rate. To test the temperature homogeneity in our
323 system of swapping and scaling methods, temperatures of different
324 dimensions in all the slabs were observed. Most of the simulations
325 include $10^5$ steps of equilibration without imposing momentum flux,
326 $10^5$ steps of stablization with imposing unphysical momentum
327 transfer, and $10^6$ steps of data collection under RNEMD. For
328 relatively high momentum flux simulations, ${5\times10^5}$ step data
329 collection is sufficient. For some low momentum flux simulations,
330 ${2\times10^6}$ steps were necessary.
331
332 After each simulation, the shear viscosity was calculated in reduced
333 unit. The momentum flux was calculated with total unphysical
334 transferred momentum ${P_x}$ and data collection time $t$:
335 \begin{equation}
336 j_z(p_x) = \frac{P_x}{2 t L_x L_y}
337 \end{equation}
338 where $L_x$ and $L_y$ denotes $x$ and $y$ lengths of the simulation
339 box, and physical momentum transfer occurs in two ways due to our
340 periodic boundary condition settings. And the velocity gradient
341 ${\langle \partial v_x /\partial z \rangle}$ can be obtained by a
342 linear regression of the velocity profile. From the shear viscosity
343 $\eta$ calculated with the above parameters, one can further convert
344 it into reduced unit ${\eta^* = \eta \sigma^2 (\varepsilon m)^{-1/2}}$.
345
346 For thermal conductivity calculations, simulations were first run under
347 reduced temperature ${\langle T^*\rangle = 0.72}$ in NVE
348 ensemble. Muller-Plathe's algorithm was adopted in the swapping
349 method. Under identical simulation box parameters with our shear
350 viscosity calculations, in each swap, the top slab exchanges all three
351 translational momentum components of the molecule with least kinetic
352 energy with the same components of the molecule in the center slab
353 with most kinetic energy, unless this ``coldest'' molecule in the
354 ``hot'' slab is still ``hotter'' than the ``hottest'' molecule in the
355 ``cold'' slab. According to swapping RNEMD results, target energy flux
356 for scaling RNEMD simulations can be set. Also, various scaling
357 frequencies can be tested for one target energy flux. To compare the
358 performance between swapping and scaling method, distributions of
359 velocity and speed in different slabs were observed.
360
361 For each swapping rate, thermal conductivity was calculated in reduced
362 unit. The energy flux was calculated similarly to the momentum flux,
363 with total unphysical transferred energy ${E_{total}}$ and data collection
364 time $t$:
365 \begin{equation}
366 J_z = \frac{E_{total}}{2 t L_x L_y}
367 \end{equation}
368 And the temperature gradient ${\langle\partial T/\partial z\rangle}$
369 can be obtained by a linear regression of the temperature
370 profile. From the thermal conductivity $\lambda$ calculated, one can
371 further convert it into reduced unit ${\lambda^*=\lambda \sigma^2
372 m^{1/2} k_B^{-1}\varepsilon^{-1/2}}$.
373
374 \subsection{ Water / Metal Thermal Conductivity}
375 Another series of our simulation is the calculation of interfacial
376 thermal conductivity of a Au/H$_2$O system. Respective calculations of
377 liquid water (SPC/E) and crystal gold (QSC) thermal conductivity were
378 performed and compared with current results to ensure the validity of
379 NIVS-RNEMD. After that, a mixture system was simulated.
380
381 For thermal conductivity calculation of bulk water, a simulation box
382 consisting of 1000 molecules were first equilibrated under ambient
383 pressure and temperature conditions using NPT ensemble, followed by
384 equilibration in fixed volume (NVT). The system was then equilibrated in
385 microcanonical ensemble (NVE). Also in NVE ensemble, establishing a
386 stable thermal gradient was followed. The simulation box was under
387 periodic boundary condition and devided into 10 slabs. Data collection
388 process was similar to Lennard-Jones fluid system.
389
390 Thermal conductivity calculation of bulk crystal gold used a similar
391 protocol. The face centered cubic crystal simulation box consists of
392 2880 Au atoms. The lattice was first allowed volume change to relax
393 under ambient temperature and pressure. Equilibrations in canonical and
394 microcanonical ensemble were followed in order. With the simulation
395 lattice devided evenly into 10 slabs, different thermal gradients were
396 established by applying a set of target thermal transfer flux. Data of
397 the series of thermal gradients was collected for calculation.
398
399 After simulations of bulk water and crystal gold, a mixture system was
400 constructed, consisting of 1188 Au atoms and 1862 H$_2$O
401 molecules. Spohr potential was adopted in depicting the interaction
402 between metal atom and water molecule.\cite{ISI:000167766600035} A
403 similar protocol of equilibration was followed. Several thermal
404 gradients was built under different target thermal flux. It was found
405 out that compared to our previous simulation systems, the two phases
406 could have large temperature difference even under a relatively low
407 thermal flux. Therefore, under our low flux conditions, it is assumed
408 that the metal and water phases have respectively homogeneous
409 temperature, excluding the surface regions. In calculating the
410 interfacial thermal conductivity $G$, this assumptioin was applied and
411 thus our formula becomes:
412
413 \begin{equation}
414 G = \frac{E_{total}}{2 t L_x L_y \left( \langle T_{gold}\rangle -
415 \langle T_{water}\rangle \right)}
416 \label{interfaceCalc}
417 \end{equation}
418 where ${E_{total}}$ is the imposed unphysical kinetic energy transfer
419 and ${\langle T_{gold}\rangle}$ and ${\langle T_{water}\rangle}$ are the
420 average observed temperature of gold and water phases respectively.
421
422 \section{Results And Discussions}
423 \subsection{Thermal Conductivity}
424 \subsubsection{Lennard-Jones Fluid}
425 Our thermal conductivity calculations show that scaling method results
426 agree with swapping method. Four different exchange intervals were
427 tested (Table \ref{thermalLJRes}) using swapping method. With a fixed
428 10fs exchange interval, target exchange kinetic energy was set to
429 produce equivalent kinetic energy flux as in swapping method. And
430 similar thermal gradients were observed with similar thermal flux in
431 two simulation methods (Figure \ref{thermalGrad}).
432
433 \begin{table*}
434 \begin{minipage}{\linewidth}
435 \begin{center}
436
437 \caption{Calculation results for thermal conductivity of Lennard-Jones
438 fluid at ${\langle T^* \rangle = 0.72}$ and ${\rho^* = 0.85}$, with
439 swap and scale methods at various kinetic energy exchange rates. Results
440 in reduced unit. Errors of calculations in parentheses.}
441
442 \begin{tabular}{ccc}
443 \hline
444 (Equilvalent) Exchange Interval (fs) & $\lambda^*_{swap}$ &
445 $\lambda^*_{scale}$\\
446 \hline
447 250 & 7.03(0.34) & 7.30(0.10)\\
448 500 & 7.03(0.14) & 6.95(0.09)\\
449 1000 & 6.91(0.42) & 7.19(0.07)\\
450 2000 & 7.52(0.15) & 7.19(0.28)\\
451 \hline
452 \end{tabular}
453 \label{thermalLJRes}
454 \end{center}
455 \end{minipage}
456 \end{table*}
457
458 \begin{figure}
459 \includegraphics[width=\linewidth]{thermalGrad}
460 \caption{Temperature gradients under various kinetic energy flux of
461 thermal conductivity simulations}
462 \label{thermalGrad}
463 \end{figure}
464
465 During these simulations, molecule velocities were recorded in 1000 of
466 all the snapshots. These velocity data were used to produce histograms
467 of velocity and speed distribution in different slabs. From these
468 histograms, it is observed that with increasing unphysical kinetic
469 energy flux, speed and velocity distribution of molecules in slabs
470 where swapping occured could deviate from Maxwell-Boltzmann
471 distribution. Figure \ref{histSwap} indicates how these distributions
472 deviate from ideal condition. In high temperature slabs, probability
473 density in low speed is confidently smaller than ideal distribution;
474 in low temperature slabs, probability density in high speed is smaller
475 than ideal. This phenomenon is observable even in our relatively low
476 swapping rate simulations. And this deviation could also leads to
477 deviation of distribution of velocity in various dimensions. One
478 feature of these deviated distribution is that in high temperature
479 slab, the ideal Gaussian peak was changed into a relatively flat
480 plateau; while in low temperature slab, that peak appears sharper.
481
482 \begin{figure}
483 \includegraphics[width=\linewidth]{histSwap}
484 \caption{Speed distribution for thermal conductivity using swapping RNEMD.}
485 \label{histSwap}
486 \end{figure}
487
488 \begin{figure}
489 \includegraphics[width=\linewidth]{histScale}
490 \caption{Speed distribution for thermal conductivity using scaling RNEMD.}
491 \label{histScale}
492 \end{figure}
493
494 \subsubsection{SPC/E Water}
495 Our results of SPC/E water thermal conductivity are comparable to
496 Bedrov {\it et al.}\cite{ISI:000090151400044}, which employed the
497 previous swapping RNEMD method for their calculation. Our simulations
498 were able to produce a similar temperature gradient to their
499 system. However, the average temperature of our system is 300K, while
500 theirs is 318K, which would be attributed for part of the difference
501 between the two series of results. Both methods yields values in
502 agreement with experiment. And this shows the applicability of our
503 method to multi-atom molecular system.
504
505 \begin{figure}
506 \includegraphics[width=\linewidth]{spceGrad}
507 \caption{Temperature gradients for SPC/E water thermal conductivity.}
508 \label{spceGrad}
509 \end{figure}
510
511 \begin{table*}
512 \begin{minipage}{\linewidth}
513 \begin{center}
514
515 \caption{Calculation results for thermal conductivity of SPC/E water
516 at ${\langle T\rangle}$ = 300K at various thermal exchange rates. Errors of
517 calculations in parentheses. }
518
519 \begin{tabular}{cccc}
520 \hline
521 $\langle dT/dz\rangle$(K/\AA) & & $\lambda$(W/m/K) & \\
522 & This work & Previous simulations\cite{ISI:000090151400044} &
523 Experiment$^a$\\
524 \hline
525 0.38 & 0.816(0.044) & & 0.64\\
526 0.81 & 0.770(0.008) & 0.784\\
527 1.54 & 0.813(0.007) & 0.730\\
528 \hline
529 \end{tabular}
530 \label{spceThermal}
531 \end{center}
532 \end{minipage}
533 \end{table*}
534
535 \subsubsection{Crystal Gold}
536 Our results of gold thermal conductivity used QSC force field are
537 shown in Table \ref{AuThermal}. Although our calculation is smaller
538 than experimental value by an order of more than 100, this difference
539 is mainly attributed to the lack of electron interaction
540 representation in our force field parameters. Richardson {\it et
541 al.}\cite{ISI:A1992HX37800010} used similar force field parameters
542 in their metal thermal conductivity calculations. The EMD method they
543 employed in their simulations produced comparable results to
544 ours. Therefore, it is confident to conclude that NIVS-RNEMD is
545 applicable to metal force field system.
546
547 \begin{figure}
548 \includegraphics[width=\linewidth]{AuGrad}
549 \caption{Temperature gradients for crystal gold thermal conductivity.}
550 \label{AuGrad}
551 \end{figure}
552
553 \begin{table*}
554 \begin{minipage}{\linewidth}
555 \begin{center}
556
557 \caption{Calculation results for thermal conductivity of crystal gold
558 at ${\langle T\rangle}$ = 300K at various thermal exchange rates. Errors of
559 calculations in parentheses. }
560
561 \begin{tabular}{ccc}
562 \hline
563 $\langle dT/dz\rangle$(K/\AA) & $\lambda$(W/m/K)\\
564 & This work & Previous simulations\cite{ISI:A1992HX37800010} \\
565 \hline
566 1.44 & 1.10(0.01) & \\
567 2.86 & 1.08(0.02) & \\
568 5.14 & 1.15(0.01) & \\
569 \hline
570 \end{tabular}
571 \label{AuThermal}
572 \end{center}
573 \end{minipage}
574 \end{table*}
575
576 \subsection{Interfaciel Thermal Conductivity}
577 After valid simulations of homogeneous water and gold systems using
578 NIVS-RNEMD method, calculation of gold/water interfacial thermal
579 conductivity was followed. It is found out that the interfacial
580 conductance is low due to a hydrophobic surface in our system. Figure
581 \ref{interfaceDensity} demonstrates this observance. Consequently, our
582 reported results (Table \ref{interfaceRes}) are of two orders of
583 magnitude smaller than our calculations on homogeneous systems.
584
585 \begin{figure}
586 \includegraphics[width=\linewidth]{interfaceDensity}
587 \caption{Density profile for interfacial thermal conductivity
588 simulation box.}
589 \label{interfaceDensity}
590 \end{figure}
591
592 \begin{figure}
593 \includegraphics[width=\linewidth]{interfaceGrad}
594 \caption{Temperature profiles for interfacial thermal conductivity
595 simulation box.}
596 \label{interfaceGrad}
597 \end{figure}
598
599 \begin{table*}
600 \begin{minipage}{\linewidth}
601 \begin{center}
602
603 \caption{Calculation results for interfacial thermal conductivity
604 at ${\langle T\rangle \sim}$ 300K at various thermal exchange
605 rates. Errors of calculations in parentheses. }
606
607 \begin{tabular}{cccc}
608 \hline
609 $J_z$(MW/m$^2$) & $T_{gold}$ & $T_{water}$ & $G$(MW/m$^2$/K)\\
610 \hline
611 98.0 & 355.2 & 295.8 & 1.65(0.21) \\
612 78.8 & 343.8 & 298.0 & 1.72(0.32) \\
613 73.6 & 344.3 & 298.0 & 1.59(0.24) \\
614 49.2 & 330.1 & 300.4 & 1.65(0.35) \\
615 \hline
616 \end{tabular}
617 \label{interfaceRes}
618 \end{center}
619 \end{minipage}
620 \end{table*}
621
622 \subsection{Shear Viscosity}
623 Our calculations (Table \ref{shearRate}) shows that scale RNEMD method
624 produced comparable shear viscosity to swap RNEMD method. In Table
625 \ref{shearRate}, the names of the calculated samples are devided into
626 two parts. The first number refers to total slabs in one simulation
627 box. The second number refers to the swapping interval in swap method, or
628 in scale method the equilvalent swapping interval that the same
629 momentum flux would theoretically result in swap method. All the scale
630 method results were from simulations that had a scaling interval of 10
631 time steps. The average molecular momentum gradients of these samples
632 are shown in Figure \ref{shearGrad}.
633
634 \begin{table*}
635 \begin{minipage}{\linewidth}
636 \begin{center}
637
638 \caption{Calculation results for shear viscosity of Lennard-Jones
639 fluid at ${T^* = 0.72}$ and ${\rho^* = 0.85}$, with swap and scale
640 methods at various momentum exchange rates. Results in reduced
641 unit. Errors of calculations in parentheses. }
642
643 \begin{tabular}{ccc}
644 \hline
645 Series & $\eta^*_{swap}$ & $\eta^*_{scale}$\\
646 \hline
647 20-500 & 3.64(0.05) & 3.76(0.09)\\
648 20-1000 & 3.52(0.16) & 3.66(0.06)\\
649 20-2000 & 3.72(0.05) & 3.32(0.18)\\
650 20-2500 & 3.42(0.06) & 3.43(0.08)\\
651 \hline
652 \end{tabular}
653 \label{shearRate}
654 \end{center}
655 \end{minipage}
656 \end{table*}
657
658 \begin{figure}
659 \includegraphics[width=\linewidth]{shearGrad}
660 \caption{Average momentum gradients of shear viscosity simulations}
661 \label{shearGrad}
662 \end{figure}
663
664 \begin{figure}
665 \includegraphics[width=\linewidth]{shearTempScale}
666 \caption{Temperature profile for scaling RNEMD simulation.}
667 \label{shearTempScale}
668 \end{figure}
669 However, observations of temperatures along three dimensions show that
670 inhomogeneity occurs in scaling RNEMD simulations, particularly in the
671 two slabs which were scaled. Figure \ref{shearTempScale} indicate that with
672 relatively large imposed momentum flux, the temperature difference among $x$
673 and the other two dimensions was significant. This would result from the
674 algorithm of scaling method. From Eq. \ref{eq:fluxPlane}, after
675 momentum gradient is set up, $P_c^x$ would be roughly stable
676 ($<0$). Consequently, scaling factor $x$ would most probably larger
677 than 1. Therefore, the kinetic energy in $x$-dimension $K_c^x$ would
678 keep increase after most scaling steps. And if there is not enough time
679 for the kinetic energy to exchange among different dimensions and
680 different slabs, the system would finally build up temperature
681 (kinetic energy) difference among the three dimensions. Also, between
682 $y$ and $z$ dimensions in the scaled slabs, temperatures of $z$-axis
683 are closer to neighbor slabs. This is due to momentum transfer along
684 $z$ dimension between slabs.
685
686 Although results between scaling and swapping methods are comparable,
687 the inherent temperature inhomogeneity even in relatively low imposed
688 exchange momentum flux simulations makes scaling RNEMD method less
689 attractive than swapping RNEMD in shear viscosity calculation.
690
691 \section{Conclusions}
692 NIVS-RNEMD simulation method is developed and tested on various
693 systems. Simulation results demonstrate its validity of thermal
694 conductivity calculations. NIVS-RNEMD improves non-Boltzmann-Maxwell
695 distributions existing in previous RNEMD methods, and extends its
696 applicability to interfacial systems. NIVS-RNEMD has also limited
697 application on shear viscosity calculations, but under high momentum
698 flux, it could cause temperature difference among different
699 dimensions. Modification is necessary to extend the applicability of
700 NIVS-RNEMD in shear viscosity calculations.
701
702 \section{Acknowledgments}
703 Support for this project was provided by the National Science
704 Foundation under grant CHE-0848243. Computational time was provided by
705 the Center for Research Computing (CRC) at the University of Notre
706 Dame. \newpage
707
708 \bibliographystyle{jcp2}
709 \bibliography{nivsRnemd}
710 \end{doublespace}
711 \end{document}
712