ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/group/stokes/stokes.tex
Revision: 3780
Committed: Sun Dec 11 03:22:47 2011 UTC (12 years, 9 months ago) by skuang
Content type: application/x-tex
File size: 36188 byte(s)
Log Message:
edit a figure, and a few more changes.

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{multirow}
10 %\usepackage{booktabs}
11 %\usepackage{bibentry}
12 %\usepackage{mathrsfs}
13 %\usepackage[ref]{overcite}
14 \usepackage[square, comma, sort&compress]{natbib}
15 \usepackage{url}
16 \pagestyle{plain} \pagenumbering{arabic} \oddsidemargin 0.0cm
17 \evensidemargin 0.0cm \topmargin -21pt \headsep 10pt \textheight
18 9.0in \textwidth 6.5in \brokenpenalty=10000
19
20 % double space list of tables and figures
21 \AtBeginDelayedFloats{\renewcommand{\baselinestretch}{1.66}}
22 \setlength{\abovecaptionskip}{20 pt}
23 \setlength{\belowcaptionskip}{30 pt}
24
25 %\renewcommand\citemid{\ } % no comma in optional reference note
26 \bibpunct{[}{]}{,}{n}{}{;}
27 \bibliographystyle{achemso}
28
29 \begin{document}
30
31 \title{ENTER TITLE HERE}
32
33 \author{Shenyu Kuang and J. Daniel
34 Gezelter\footnote{Corresponding author. \ Electronic mail: gezelter@nd.edu} \\
35 Department of Chemistry and Biochemistry,\\
36 University of Notre Dame\\
37 Notre Dame, Indiana 46556}
38
39 \date{\today}
40
41 \maketitle
42
43 \begin{doublespace}
44
45 \begin{abstract}
46 We present a new method for introducing stable nonequilibrium
47 velocity and temperature gradients in molecular dynamics simulations
48 of heterogeneous systems. This method conserves the linear momentum
49 and total energy of the system and improves previous Reverse
50 Non-Equilibrium Molecular Dynamics (RNEMD) methods and maintains
51 thermal velocity distributions. It also avoid thermal anisotropy
52 occured in NIVS simulations by using isotropic velocity scaling on
53 the molecules in specific regions of a system. To test the method,
54 we have computed the thermal conductivity and shear viscosity of
55 model liquid systems as well as the interfacial frictions of a
56 series of metal/liquid interfaces.
57
58 \end{abstract}
59
60 \newpage
61
62 %\narrowtext
63
64 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
65 % BODY OF TEXT
66 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
67
68 \section{Introduction}
69 [REFINE LATER, ADD MORE REF.S]
70 Imposed-flux methods in Molecular Dynamics (MD)
71 simulations\cite{MullerPlathe:1997xw} can establish steady state
72 systems with a set applied flux vs a corresponding gradient that can
73 be measured. These methods does not need many trajectories to provide
74 information of transport properties of a given system. Thus, they are
75 utilized in computing thermal and mechanical transfer of homogeneous
76 or bulk systems as well as heterogeneous systems such as liquid-solid
77 interfaces.\cite{kuang:AuThl}
78
79 The Reverse Non-Equilibrium MD (RNEMD) methods adopt constraints that
80 satisfy linear momentum and total energy conservation of a system when
81 imposing fluxes in a simulation. Thus they are compatible with various
82 ensembles, including the micro-canonical (NVE) ensemble, without the
83 need of an external thermostat. The original approaches by
84 M\"{u}ller-Plathe {\it et
85 al.}\cite{MullerPlathe:1997xw,ISI:000080382700030} utilize simple
86 momentum swapping for generating energy/momentum fluxes, which is also
87 compatible with particles of different identities. Although simple to
88 implement in a simulation, this approach can create nonthermal
89 velocity distributions, as discovered by Tenney and
90 Maginn\cite{Maginn:2010}. Furthermore, this approach to kinetic energy
91 transfer between particles of different identities is less efficient
92 when the mass difference between the particles becomes significant,
93 which also limits its application on heterogeneous interfacial
94 systems.
95
96 Recently, we developed a different approach, using Non-Isotropic
97 Velocity Scaling (NIVS) \cite{kuang:164101} algorithm to impose
98 fluxes. Compared to the momentum swapping move, it scales the velocity
99 vectors in two separate regions of a simulated system with respective
100 diagonal scaling matrices. These matrices are determined by solving a
101 set of equations including linear momentum and kinetic energy
102 conservation constraints and target flux satisfaction. This method is
103 able to effectively impose a wide range of kinetic energy fluxes
104 without obvious perturbation to the velocity distributions of the
105 simulated systems, regardless of the presence of heterogeneous
106 interfaces. We have successfully applied this approach in studying the
107 interfacial thermal conductance at metal-solvent
108 interfaces.\cite{kuang:AuThl}
109
110 However, the NIVS approach limits its application in imposing momentum
111 fluxes. Temperature anisotropy can happen under high momentum fluxes,
112 due to the nature of the algorithm. Thus, combining thermal and
113 momentum flux is also difficult to implement with this
114 approach. However, such combination may provide a means to simulate
115 thermal/momentum gradient coupled processes such as freeze
116 desalination. Therefore, developing novel approaches to extend the
117 application of imposed-flux method is desired.
118
119 In this paper, we improve the NIVS method and propose a novel approach
120 to impose fluxes. This approach separate the means of applying
121 momentum and thermal flux with operations in one time step and thus is
122 able to simutaneously impose thermal and momentum flux. Furthermore,
123 the approach retains desirable features of previous RNEMD approaches
124 and is simpler to implement compared to the NIVS method. In what
125 follows, we first present the method to implement the method in a
126 simulation. Then we compare the method on bulk fluids to previous
127 methods. Also, interfacial frictions are computed for a series of
128 interfaces.
129
130 \section{Methodology}
131 Similar to the NIVS methodology,\cite{kuang:164101} we consider a
132 periodic system divided into a series of slabs along a certain axis
133 (e.g. $z$). The unphysical thermal and/or momentum flux is designated
134 from the center slab to one of the end slabs, and thus the center slab
135 would have a lower temperature than the end slab (unless the thermal
136 flux is negative). Therefore, the center slab is denoted as ``$c$''
137 while the end slab as ``$h$''.
138
139 To impose these fluxes, we periodically apply separate operations to
140 velocities of particles {$i$} within the center slab and of particles
141 {$j$} within the end slab:
142 \begin{eqnarray}
143 \vec{v}_i & \leftarrow & c\cdot\left(\vec{v}_i - \langle\vec{v}_c
144 \rangle\right) + \left(\langle\vec{v}_c\rangle + \vec{a}_c\right) \\
145 \vec{v}_j & \leftarrow & h\cdot\left(\vec{v}_j - \langle\vec{v}_h
146 \rangle\right) + \left(\langle\vec{v}_h\rangle + \vec{a}_h\right)
147 \end{eqnarray}
148 where $\langle\vec{v}_c\rangle$ and $\langle\vec{v}_h\rangle$ denotes
149 the instantaneous bulk velocity of slabs $c$ and $h$ respectively
150 before an operation occurs. When a momentum flux $\vec{j}_z(\vec{p})$
151 presents, these bulk velocities would have a corresponding change
152 ($\vec{a}_c$ and $\vec{a}_h$ respectively) according to Newton's
153 second law:
154 \begin{eqnarray}
155 M_c \vec{a}_c & = & -\vec{j}_z(\vec{p}) \Delta t \\
156 M_h \vec{a}_h & = & \vec{j}_z(\vec{p}) \Delta t
157 \end{eqnarray}
158 where
159 \begin{eqnarray}
160 M_c & = & \sum_{i = 1}^{N_c} m_i \\
161 M_h & = & \sum_{j = 1}^{N_h} m_j
162 \end{eqnarray}
163 and $\Delta t$ is the interval between two operations.
164
165 The above operations conserve the linear momentum of a periodic
166 system. To satisfy total energy conservation as well as to impose a
167 thermal flux $J_z$, one would have
168 [SUPPORT INFO MIGHT BE NECESSARY TO PUT EXTRA MATH IN]
169 \begin{eqnarray}
170 K_c - J_z\Delta t & = & c^2 (K_c - \frac{1}{2}M_c \langle\vec{v}_c
171 \rangle^2) + \frac{1}{2}M_c (\langle \vec{v}_c \rangle + \vec{a}_c)^2 \\
172 K_h + J_z\Delta t & = & h^2 (K_h - \frac{1}{2}M_h \langle\vec{v}_h
173 \rangle^2) + \frac{1}{2}M_h (\langle \vec{v}_h \rangle + \vec{a}_h)^2
174 \end{eqnarray}
175 where $K_c$ and $K_h$ denotes translational kinetic energy of slabs
176 $c$ and $h$ respectively before an operation occurs. These
177 translational kinetic energy conservation equations are sufficient to
178 ensure total energy conservation, as the operations applied do not
179 change the potential energy of a system, given that the potential
180 energy does not depend on particle velocity.
181
182 The above sets of equations are sufficient to determine the velocity
183 scaling coefficients ($c$ and $h$) as well as $\vec{a}_c$ and
184 $\vec{a}_h$. Note that two roots of $c$ and $h$ exist
185 respectively. However, to avoid dramatic perturbations to a system,
186 the positive roots (which are closer to 1) are chosen. Figure
187 \ref{method} illustrates the implementation of this algorithm in an
188 individual step.
189
190 \begin{figure}
191 \includegraphics[width=\linewidth]{method}
192 \caption{Illustration of the implementation of the algorithm in a
193 single step. Starting from an ideal velocity distribution, the
194 transformation is used to apply both thermal and momentum flux from
195 the ``c'' slab to the ``h'' slab. As the figure shows, the thermal
196 distributions preserve after this operation.}
197 \label{method}
198 \end{figure}
199
200 By implementing these operations at a certain frequency, a steady
201 thermal and/or momentum flux can be applied and the corresponding
202 temperature and/or momentum gradients can be established.
203
204 This approach is more computationaly efficient compared to the
205 previous NIVS method, in that only quadratic equations are involved,
206 while the NIVS method needs to solve a quartic equations. Furthermore,
207 the method implements isotropic scaling of velocities in respective
208 slabs, unlike the NIVS, where an extra criteria function is necessary
209 to choose a set of coefficients that performs the most isotropic
210 scaling. More importantly, separating the momentum flux imposing from
211 velocity scaling avoids the underlying cause that NIVS produced
212 thermal anisotropy when applying a momentum flux.
213
214 The advantages of the approach over the original momentum swapping
215 approach lies in its nature to preserve a Gaussian
216 distribution. Because the momentum swapping tends to render a
217 nonthermal distribution, when the imposed flux is relatively large,
218 diffusion of the neighboring slabs could no longer remedy this effect,
219 and nonthermal distributions would be observed. Results in later
220 section will illustrate this effect.
221
222 \section{Computational Details}
223 The algorithm has been implemented in our MD simulation code,
224 OpenMD\cite{Meineke:2005gd,openmd}. We compare the method with
225 previous RNEMD methods or equilibrium MD methods in homogeneous fluids
226 (Lennard-Jones and SPC/E water). And taking advantage of the method,
227 we simulate the interfacial friction of different heterogeneous
228 interfaces (gold-organic solvent and gold-SPC/E water and ice-liquid
229 water).
230
231 \subsection{Simulation Protocols}
232 The systems to be investigated are set up in a orthorhombic simulation
233 cell with periodic boundary conditions in all three dimensions. The
234 $z$ axis of these cells were longer and was set as the gradient axis
235 of temperature and/or momentum. Thus the cells were divided into $N$
236 slabs along this axis, with various $N$ depending on individual
237 system. The $x$ and $y$ axis were usually of the same length in
238 homogeneous systems or close to each other where interfaces
239 presents. In all cases, before introducing a nonequilibrium method to
240 establish steady thermal and/or momentum gradients for further
241 measurements and calculations, canonical ensemble with a Nos\'e-Hoover
242 thermostat\cite{hoover85} and microcanonical ensemble equilibrations
243 were used to prepare systems ready for data
244 collections. Isobaric-isothermal equilibrations are performed before
245 this for SPC/E water systems to reach normal pressure (1 bar), while
246 similar equilibrations are used for interfacial systems to relax the
247 surface tensions.
248
249 While homogeneous fluid systems can be set up with random
250 configurations, our interfacial systems needs extra steps to ensure
251 the interfaces be established properly for computations. The
252 preparation and equilibration of butanethiol covered gold (111)
253 surface and further solvation and equilibration process is described
254 as in reference \cite{kuang:AuThl}.
255
256 As for the ice/liquid water interfaces, the basal surface of ice
257 lattice was first constructed. Hirsch {\it et
258 al.}\cite{doi:10.1021/jp048434u} explored the energetics of ice
259 lattices with different proton orders. We refer to their results and
260 choose the configuration of the lowest energy after geometry
261 optimization as the unit cells of our ice lattices. Although
262 experimental solid/liquid coexistant temperature near normal pressure
263 is 273K, Bryk and Haymet's simulations of ice/liquid water interfaces
264 with different models suggest that for SPC/E, the most stable
265 interface is observed at 225$\pm$5K. Therefore, all our ice/liquid
266 water simulations were carried out under 225K. To have extra
267 protection of the ice lattice during initial equilibration (when the
268 randomly generated liquid phase configuration could release large
269 amount of energy in relaxation), a constraint method (REF?) was
270 adopted until the high energy configuration was relaxed.
271 [MAY ADD A FIGURE HERE FOR BASAL PLANE, MAY INCLUDE PRISM IF POSSIBLE]
272
273 \subsection{Force Field Parameters}
274 For comparison of our new method with previous work, we retain our
275 force field parameters consistent with the results we will compare
276 with. The Lennard-Jones fluid used here for argon , and reduced unit
277 results are reported for direct comparison purpose.
278
279 As for our water simulations, SPC/E model is used throughout this work
280 for consistency. Previous work for transport properties of SPC/E water
281 model is available\cite{Bedrov:2000,10.1063/1.3330544,Medina2011} so
282 that unnecessary repetition of previous methods can be avoided.
283
284 The Au-Au interaction parameters in all simulations are described by
285 the quantum Sutton-Chen (QSC) formulation.\cite{PhysRevB.59.3527} The
286 QSC potentials include zero-point quantum corrections and are
287 reparametrized for accurate surface energies compared to the
288 Sutton-Chen potentials.\cite{Chen90} For gold/water interfaces, the
289 Spohr potential was adopted\cite{ISI:000167766600035} to depict
290 Au-H$_2$O interactions.
291
292 The small organic molecules included in our simulations are the Au
293 surface capping agent butanethiol and liquid hexane and toluene. The
294 United-Atom
295 models\cite{TraPPE-UA.thiols,TraPPE-UA.alkanes,TraPPE-UA.alkylbenzenes}
296 for these components were used in this work for better computational
297 efficiency, while maintaining good accuracy. We refer readers to our
298 previous work\cite{kuang:AuThl} for further details of these models,
299 as well as the interactions between Au and the above organic molecule
300 components.
301
302 \subsection{Thermal conductivities}
303 When $\vec{j}_z(\vec{p})$ is set to zero and a target $J_z$ is set to
304 impose kinetic energy transfer, the method can be used for thermal
305 conductivity computations. Similar to previous RNEMD methods, we
306 assume linear response of the temperature gradient with respect to the
307 thermal flux in general case. And the thermal conductivity ($\lambda$)
308 can be obtained with the imposed kinetic energy flux and the measured
309 thermal gradient:
310 \begin{equation}
311 J_z = -\lambda \frac{\partial T}{\partial z}
312 \end{equation}
313 Like other imposed-flux methods, the energy flux was calculated using
314 the total non-physical energy transferred (${E_{total}}$) from slab
315 ``c'' to slab ``h'', which is recorded throughout a simulation, and
316 the time for data collection $t$:
317 \begin{equation}
318 J_z = \frac{E_{total}}{2 t L_x L_y}
319 \end{equation}
320 where $L_x$ and $L_y$ denotes the dimensions of the plane in a
321 simulation cell perpendicular to the thermal gradient, and a factor of
322 two in the denominator is present for the heat transport occurs in
323 both $+z$ and $-z$ directions. The temperature gradient
324 ${\langle\partial T/\partial z\rangle}$ can be obtained by a linear
325 regression of the temperature profile, which is recorded during a
326 simulation for each slab in a cell. For Lennard-Jones simulations,
327 thermal conductivities are reported in reduced units
328 (${\lambda^*=\lambda \sigma^2 m^{1/2} k_B^{-1}\varepsilon^{-1/2}}$).
329
330 \subsection{Shear viscosities}
331 Alternatively, the method can carry out shear viscosity calculations
332 by switching off $J_z$. One can specify the vector
333 $\vec{j}_z(\vec{p})$ by choosing the three components
334 respectively. For shear viscosity simulations, $j_z(p_z)$ is usually
335 set to zero. Although for isotropic systems, the direction of
336 $\vec{j}_z(\vec{p})$ on the $xy$ plane should not matter, the ability
337 of arbitarily specifying the vector direction in our method provides
338 convenience in anisotropic simulations.
339
340 Similar to thermal conductivity computations, linear response of the
341 momentum gradient with respect to the shear stress is assumed, and the
342 shear viscosity ($\eta$) can be obtained with the imposed momentum
343 flux (e.g. in $x$ direction) and the measured gradient:
344 \begin{equation}
345 j_z(p_x) = -\eta \frac{\partial v_x}{\partial z}
346 \end{equation}
347 where the flux is similarly defined:
348 \begin{equation}
349 j_z(p_x) = \frac{P_x}{2 t L_x L_y}
350 \end{equation}
351 with $P_x$ being the total non-physical momentum transferred within
352 the data collection time. Also, the velocity gradient
353 ${\langle\partial v_x/\partial z\rangle}$ can be obtained using linear
354 regression of the $x$ component of the mean velocity, $\langle
355 v_x\rangle$, in each of the bins. For Lennard-Jones simulations, shear
356 viscosities are reported in reduced units
357 (${\eta^*=\eta\sigma^2(\varepsilon m)^{-1/2}}$).
358
359 \subsection{Interfacial friction and Slip length}
360 While the shear stress results in a velocity gradient within bulk
361 fluid phase, its effect at a solid-liquid interface could vary due to
362 the interaction strength between the two phases. The interfacial
363 friction coefficient $\kappa$ is defined to relate the shear stress
364 (e.g. along $x$-axis) and the relative fluid velocity tangent to the
365 interface:
366 \begin{equation}
367 j_z(p_x)|_{interface} = \kappa\Delta v_x|_{interface}
368 \end{equation}
369 Under ``stick'' boundary condition, $\Delta v_x|_{interface}
370 \rightarrow 0$, which leads to $\kappa\rightarrow\infty$. However, for
371 ``slip'' boundary condition at the solid-liquid interface, $\kappa$
372 becomes finite. To characterize the interfacial boundary conditions,
373 slip length ($\delta$) is defined using $\kappa$ and the shear
374 viscocity of liquid phase ($\eta$):
375 \begin{equation}
376 \delta = \frac{\eta}{\kappa}
377 \end{equation}
378 so that $\delta\rightarrow 0$ in the ``no-slip'' boundary condition,
379 and depicts how ``slippery'' an interface is. Figure \ref{slipLength}
380 illustrates how this quantity is defined and computed for a
381 solid-liquid interface. [MAY INCLUDE A SNAPSHOT IN FIGURE]
382
383 \begin{figure}
384 \includegraphics[width=\linewidth]{defDelta}
385 \caption{The slip length $\delta$ can be obtained from a velocity
386 profile of a solid-liquid interface simulation. An example of
387 Au/hexane interfaces is shown. Calculation for the left side is
388 illustrated. The right side is similar to the left side.}
389 \label{slipLength}
390 \end{figure}
391
392 In our method, a shear stress can be applied similar to shear
393 viscosity computations by applying an unphysical momentum flux
394 (e.g. $j_z(p_x)$). A corresponding velocity profile can be obtained as
395 shown in Figure \ref{slipLength}, in which the velocity gradients
396 within liquid phase and velocity difference at the liquid-solid
397 interface can be measured respectively. Further calculations and
398 characterizations of the interface can be carried out using these
399 data.
400
401 \section{Results and Discussions}
402 \subsection{Lennard-Jones fluid}
403 Our orthorhombic simulation cell of Lennard-Jones fluid has identical
404 parameters to our previous work\cite{kuang:164101} to facilitate
405 comparison. Thermal conductivitis and shear viscosities were computed
406 with the algorithm applied to the simulations. The results of thermal
407 conductivity are compared with our previous NIVS algorithm. However,
408 since the NIVS algorithm could produce temperature anisotropy for
409 shear viscocity computations, these results are instead compared to
410 the momentum swapping approaches. Table \ref{LJ} lists these
411 calculations with various fluxes in reduced units.
412
413 \begin{table*}
414 \begin{minipage}{\linewidth}
415 \begin{center}
416
417 \caption{Thermal conductivity ($\lambda^*$) and shear viscosity
418 ($\eta^*$) (in reduced units) of a Lennard-Jones fluid at
419 ${\langle T^* \rangle = 0.72}$ and ${\rho^* = 0.85}$ computed
420 at various momentum fluxes. The new method yields similar
421 results to previous RNEMD methods. All results are reported in
422 reduced unit. Uncertainties are indicated in parentheses.}
423
424 \begin{tabular}{cccccc}
425 \hline\hline
426 \multicolumn{2}{c}{Momentum Exchange} &
427 \multicolumn{2}{c}{$\lambda^*$} &
428 \multicolumn{2}{c}{$\eta^*$} \\
429 \hline
430 Swap Interval $t^*$ & Equivalent $J_z^*$ or $j_z^*(p_x)$ &
431 NIVS & This work & Swapping & This work \\
432 \hline
433 0.116 & 0.16 & 7.30(0.10) & 6.25(0.23) & 3.57(0.06) & 3.53(0.16)\\
434 0.232 & 0.09 & 6.95(0.09) & 8.02(0.56) & 3.64(0.05) & 3.43(0.17)\\
435 0.463 & 0.047 & 7.19(0.07) & 6.48(0.15) & 3.52(0.16) & 3.51(0.08)\\
436 0.926 & 0.024 & 7.19(0.28) & 7.02(0.13) & 3.72(0.05) & 3.79(0.11)\\
437 1.158 & 0.019 & 7.98(0.33) & 7.37(0.23) & 3.42(0.06) & 3.53(0.06)\\
438 \hline\hline
439 \end{tabular}
440 \label{LJ}
441 \end{center}
442 \end{minipage}
443 \end{table*}
444
445 \subsubsection{Thermal conductivity}
446 Our thermal conductivity calculations with this method yields
447 comparable results to the previous NIVS algorithm. This indicates that
448 the thermal gradients rendered using this method are also close to
449 previous RNEMD methods. Simulations with moderately higher thermal
450 fluxes tend to yield more reliable thermal gradients and thus avoid
451 large errors, while overly high thermal fluxes could introduce side
452 effects such as non-linear temperature gradient response or
453 inadvertent phase transitions.
454
455 Since the scaling operation is isotropic in this method, one does not
456 need extra care to ensure temperature isotropy between the $x$, $y$
457 and $z$ axes, while thermal anisotropy might happen if the criteria
458 function for choosing scaling coefficients does not perform as
459 expected. Furthermore, this method avoids inadvertent concomitant
460 momentum flux when only thermal flux is imposed, which could not be
461 achieved with swapping or NIVS approaches. The thermal energy exchange
462 in swapping ($\vec{p}_i$ in slab ``c'' with $\vec{p}_j$ in slab ``h'')
463 or NIVS (total slab momentum conponemts $P^\alpha$ scaled to $\alpha
464 P^\alpha$) would not obtain this result unless thermal flux vanishes
465 (i.e. $\vec{p}_i=\vec{p}_j$ or $\alpha=1$ which are trivial to apply a
466 thermal flux). In this sense, this method contributes to having
467 minimal perturbation to a simulation while imposing thermal flux.
468
469 \subsubsection{Shear viscosity}
470 Table \ref{LJ} also compares our shear viscosity results with momentum
471 swapping approach. Our calculations show that our method predicted
472 similar values for shear viscosities to the momentum swapping
473 approach, as well as the velocity gradient profiles. Moderately larger
474 momentum fluxes are helpful to reduce the errors of measured velocity
475 gradients and thus the final result. However, it is pointed out that
476 the momentum swapping approach tends to produce nonthermal velocity
477 distributions.\cite{Maginn:2010}
478
479 To examine that temperature isotropy holds in simulations using our
480 method, we measured the three one-dimensional temperatures in each of
481 the slabs (Figure \ref{tempXyz}). Note that the $x$-dimensional
482 temperatures were calculated after subtracting the effects from bulk
483 velocities of the slabs. The one-dimensional temperature profiles
484 showed no observable difference between the three dimensions. This
485 ensures that isotropic scaling automatically preserves temperature
486 isotropy and that our method is useful in shear viscosity
487 computations.
488
489 \begin{figure}
490 \includegraphics[width=\linewidth]{tempXyz}
491 \caption{Unlike the previous NIVS algorithm, the new method does not
492 produce a thermal anisotropy. No temperature difference between
493 different dimensions were observed beyond the magnitude of the error
494 bars. Note that the two ``hotter'' regions are caused by the shear
495 stress, as reported by Tenney and Maginn\cite{Maginn:2010}, but not
496 an effect that only observed in our methods.}
497 \label{tempXyz}
498 \end{figure}
499
500 Furthermore, the velocity distribution profiles are tested by imposing
501 a large shear stress into the simulations. Figure \ref{vDist}
502 demonstrates how our method is able to maintain thermal velocity
503 distributions against the momentum swapping approach even under large
504 imposed fluxes. Previous swapping methods tend to deplete particles of
505 positive velocities in the negative velocity slab (``c'') and vice
506 versa in slab ``h'', where the distributions leave a notch. This
507 problematic profiles become significant when the imposed-flux becomes
508 larger and diffusions from neighboring slabs could not offset the
509 depletion. Simutaneously, abnormal peaks appear corresponding to
510 excessive velocity swapped from the other slab. This nonthermal
511 distributions limit applications of the swapping approach in shear
512 stress simulations. Our method avoids the above problematic
513 distributions by altering the means of applying momentum
514 fluxes. Comparatively, velocity distributions recorded from
515 simulations with our method is so close to the ideal thermal
516 prediction that no observable difference is shown in Figure
517 \ref{vDist}. Conclusively, our method avoids problems happened in
518 previous RNEMD methods and provides a useful means for shear viscosity
519 computations.
520
521 \begin{figure}
522 \includegraphics[width=\linewidth]{velDist}
523 \caption{Velocity distributions that develop under the swapping and
524 our methods at high flux. These distributions were obtained from
525 Lennard-Jones simulations with $j_z^*(p_x)\sim 0.4$ (equivalent to a
526 swapping interval of 20 time steps). This is a relatively large flux
527 to demonstrate the nonthermal distributions that develop under the
528 swapping method. Distributions produced by our method are very close
529 to the ideal thermal situations.}
530 \label{vDist}
531 \end{figure}
532
533 \subsection{Bulk SPC/E water}
534 Since our method was in good performance of thermal conductivity and
535 shear viscosity computations for simple Lennard-Jones fluid, we extend
536 our applications of these simulations to complex fluid like SPC/E
537 water model. A simulation cell with 1000 molecules was set up in the
538 same manner as in \cite{kuang:164101}. For thermal conductivity
539 simulations, measurements were taken to compare with previous RNEMD
540 methods; for shear viscosity computations, simulations were run under
541 a series of temperatures (with corresponding pressure relaxation using
542 the isobaric-isothermal ensemble[CITE NIVS REF 32]), and results were
543 compared to available data from Equilibrium MD methods[CITATIONS].
544
545 \subsubsection{Thermal conductivity}
546 Table \ref{spceThermal} summarizes our thermal conductivity
547 computations under different temperatures and thermal gradients, in
548 comparison to the previous NIVS results\cite{kuang:164101} and
549 experimental measurements\cite{WagnerKruse}. Note that no appreciable
550 drift of total system energy or temperature was observed when our
551 method is applied, which indicates that our algorithm conserves total
552 energy even for systems involving electrostatic interactions.
553
554 Measurements using our method established similar temperature
555 gradients to the previous NIVS method. Our simulation results are in
556 good agreement with those from previous simulations. And both methods
557 yield values in reasonable agreement with experimental
558 values. Simulations using moderately higher thermal gradient or those
559 with longer gradient axis ($z$) for measurement seem to have better
560 accuracy, from our results.
561
562 \begin{table*}
563 \begin{minipage}{\linewidth}
564 \begin{center}
565
566 \caption{Thermal conductivity of SPC/E water under various
567 imposed thermal gradients. Uncertainties are indicated in
568 parentheses.}
569
570 \begin{tabular}{ccccc}
571 \hline\hline
572 $\langle T\rangle$ & $\langle dT/dz\rangle$ & \multicolumn{3}{c}
573 {$\lambda (\mathrm{W m}^{-1} \mathrm{K}^{-1})$} \\
574 (K) & (K/\AA) & This work & Previous NIVS\cite{kuang:164101} &
575 Experiment\cite{WagnerKruse} \\
576 \hline
577 300 & 0.8 & 0.815(0.027) & 0.770(0.008) & 0.61 \\
578 318 & 0.8 & 0.801(0.024) & 0.750(0.032) & 0.64 \\
579 & 1.6 & 0.766(0.007) & 0.778(0.019) & \\
580 & 0.8 & 0.786(0.009)\footnote{Simulation with $L_z$
581 twice as long.} & & \\
582 \hline\hline
583 \end{tabular}
584 \label{spceThermal}
585 \end{center}
586 \end{minipage}
587 \end{table*}
588
589 \subsubsection{Shear viscosity}
590 The improvement our method achieves for shear viscosity computations
591 enables us to apply it on SPC/E water models. The series of
592 temperatures under which our shear viscosity calculations were carried
593 out covers the liquid range under normal pressure. Our simulations
594 predict a similar trend of $\eta$ vs. $T$ to EMD results we refer to
595 (Table \ref{spceShear}). Considering subtlties such as temperature or
596 pressure/density errors in these two series of measurements, our
597 results show no significant difference from those with EMD
598 methods. Since each value reported using our method takes only one
599 single trajectory of simulation, instead of average from many
600 trajectories when using EMD, our method provides an effective means
601 for shear viscosity computations.
602
603 \begin{table*}
604 \begin{minipage}{\linewidth}
605 \begin{center}
606
607 \caption{Computed shear viscosity of SPC/E water under different
608 temperatures. Results are compared to those obtained with EMD
609 method[CITATION]. Uncertainties are indicated in parentheses.}
610
611 \begin{tabular}{cccc}
612 \hline\hline
613 $T$ & $\partial v_x/\partial z$ & \multicolumn{2}{c}
614 {$\eta (\mathrm{mPa}\cdot\mathrm{s})$} \\
615 (K) & (10$^{10}$s$^{-1}$) & This work & Previous simulations[CITATION]\\
616 \hline
617 273 & & 1.218(0.004) & \\
618 & & 1.140(0.012) & \\
619 303 & & 0.646(0.008) & \\
620 318 & & 0.536(0.007) & \\
621 & & 0.510(0.007) & \\
622 & & & \\
623 333 & & 0.428(0.002) & \\
624 363 & & 0.279(0.014) & \\
625 & & 0.306(0.001) & \\
626 \hline\hline
627 \end{tabular}
628 \label{spceShear}
629 \end{center}
630 \end{minipage}
631 \end{table*}
632
633 [MAY COMBINE JZPX AND JZKE TO RUN ONE JOB BUT GET ETA=F(T)]
634 [PUT RESULTS AND FIGURE HERE IF IT WORKS]
635 \subsection{Interfacial frictions and slip lengths}
636 An attractive aspect of our method is the ability to apply momentum
637 and/or thermal flux in nonhomogeneous systems, where molecules of
638 different identities (or phases) are segregated in different
639 regions. We have previously studied the interfacial thermal transport
640 of a series of metal gold-liquid
641 surfaces\cite{kuang:164101,kuang:AuThl}, and attemptions have been
642 made to investigate the relationship between this phenomenon and the
643 interfacial frictions.
644
645 Table \ref{etaKappaDelta} includes these computations and previous
646 calculations of corresponding interfacial thermal conductance. For
647 bare Au(111) surfaces, slip boundary conditions were observed for both
648 organic and aqueous liquid phases, corresponding to previously
649 computed low interfacial thermal conductance. Instead, the butanethiol
650 covered Au(111) surface appeared to be sticky to the organic liquid
651 molecules in our simulations. We have reported conductance enhancement
652 effect for this surface capping agent,\cite{kuang:AuThl} and these
653 observations have a qualitative agreement with the thermal conductance
654 results. This agreement also supports discussions on the relationship
655 between surface wetting and slip effect and thermal conductance of the
656 interface.[CITE BARRAT, GARDE]
657
658 \begin{table*}
659 \begin{minipage}{\linewidth}
660 \begin{center}
661
662 \caption{Computed interfacial friction coefficient values for
663 interfaces with various components for liquid and solid
664 phase. Error estimates are indicated in parentheses.}
665
666 \begin{tabular}{llcccccc}
667 \hline\hline
668 Solid & Liquid & $T$ & $j_z(p_x)$ & $\eta_{liquid}$ & $\kappa$
669 & $\delta$ & $G$\footnote{References \cite{kuang:AuThl} and
670 \cite{kuang:164101}.} \\
671 surface & molecules & K & MPa & mPa$\cdot$s & Pa$\cdot$s/m & nm
672 & MW/m$^2$/K \\
673 \hline
674 Au(111) & hexane & 200 & 1.08 & 0.20() & 5.3$\times$10$^4$() &
675 3.7 & 46.5 \\
676 & & & 2.15 & 0.14() & 5.3$\times$10$^4$() &
677 2.7 & \\
678 Au-SC$_4$H$_9$ & hexane & 200 & 2.16 & 0.29() & $\infty$ & 0 &
679 131 \\
680 & & & 5.39 & 0.32() & $\infty$ & 0 &
681 \\
682 \hline
683 Au(111) & toluene & 200 & 1.08 & 0.72() & 1.?$\times$10$^5$() &
684 4.6 & 70.1 \\
685 & & & 2.16 & 0.54() & 1.?$\times$10$^5$() &
686 4.9 & \\
687 Au-SC$_4$H$_9$ & toluene & 200 & 5.39 & 0.98() & $\infty$ & 0
688 & 187 \\
689 & & & 10.8 & 0.99() & $\infty$ & 0
690 & \\
691 \hline
692 Au(111) & water & 300 & 1.08 & 0.40() & 1.9$\times$10$^4$() &
693 20.7 & 1.65 \\
694 & & & 2.16 & 0.79() & 1.9$\times$10$^4$() &
695 41.9 & \\
696 \hline
697 ice(basal) & water & 225 & 19.4 & 15.8() & $\infty$ & 0 & \\
698 \hline\hline
699 \end{tabular}
700 \label{etaKappaDelta}
701 \end{center}
702 \end{minipage}
703 \end{table*}
704
705 An interesting effect alongside the surface friction change is
706 observed on the shear viscosity of liquids in the regions close to the
707 solid surface. Note that $\eta$ measured near a ``slip'' surface tends
708 to be smaller than that near a ``stick'' surface. This suggests that
709 an interface could affect the dynamic properties on its neighbor
710 regions. It is known that diffusions of solid particles in liquid
711 phase is affected by their surface conditions (stick or slip
712 boundary).[CITE SCHMIDT AND SKINNER] Our observations could provide
713 support to this phenomenon.
714
715 In addition to these previously studied interfaces, we attempt to
716 construct ice-water interfaces and the basal plane of ice lattice was
717 first studied. In contrast to the Au(111)/water interface, where the
718 friction coefficient is relatively small and large slip effect
719 presents, the ice/liquid water interface demonstrates strong
720 interactions and appears to be sticky. The supercooled liquid phase is
721 an order of magnitude viscous than measurements in previous
722 section. It would be of interst to investigate the effect of different
723 ice lattice planes (such as prism surface) on interfacial friction and
724 corresponding liquid viscosity.
725
726 \section{Conclusions}
727 Our simulations demonstrate the validity of our method in RNEMD
728 computations of thermal conductivity and shear viscosity in atomic and
729 molecular liquids. Our method maintains thermal velocity distributions
730 and avoids thermal anisotropy in previous NIVS shear stress
731 simulations, as well as retains attractive features of previous RNEMD
732 methods. There is no {\it a priori} restrictions to the method to be
733 applied in various ensembles, so prospective applications to
734 extended-system methods are possible.
735
736 Furthermore, using this method, investigations can be carried out to
737 characterize interfacial interactions. Our method is capable of
738 effectively imposing both thermal and momentum flux accross an
739 interface and thus facilitates studies that relates dynamic property
740 measurements to the chemical details of an interface.
741
742 Another attractive feature of our method is the ability of
743 simultaneously imposing thermal and momentum flux in a
744 system. potential researches that might be benefit include complex
745 systems that involve thermal and momentum gradients. For example, the
746 Soret effects under a velocity gradient would be of interest to
747 purification and separation researches.
748
749 \section{Acknowledgments}
750 Support for this project was provided by the National Science
751 Foundation under grant CHE-0848243. Computational time was provided by
752 the Center for Research Computing (CRC) at the University of Notre
753 Dame.
754
755 \newpage
756
757 \bibliography{stokes}
758
759 \end{doublespace}
760 \end{document}