ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/group/stokes/stokes.tex
Revision: 3775
Committed: Thu Dec 8 01:34:21 2011 UTC (12 years, 9 months ago) by skuang
Content type: application/x-tex
File size: 43488 byte(s)
Log Message:
a rought draft of simulation details and a figure added.

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 REPLACE ABSTRACT HERE
47 With the Non-Isotropic Velocity Scaling (NIVS) approach to Reverse
48 Non-Equilibrium Molecular Dynamics (RNEMD) it is possible to impose
49 an unphysical thermal flux between different regions of
50 inhomogeneous systems such as solid / liquid interfaces. We have
51 applied NIVS to compute the interfacial thermal conductance at a
52 metal / organic solvent interface that has been chemically capped by
53 butanethiol molecules. Our calculations suggest that coupling
54 between the metal and liquid phases is enhanced by the capping
55 agents, leading to a greatly enhanced conductivity at the interface.
56 Specifically, the chemical bond between the metal and the capping
57 agent introduces a vibrational overlap that is not present without
58 the capping agent, and the overlap between the vibrational spectra
59 (metal to cap, cap to solvent) provides a mechanism for rapid
60 thermal transport across the interface. Our calculations also
61 suggest that this is a non-monotonic function of the fractional
62 coverage of the surface, as moderate coverages allow diffusive heat
63 transport of solvent molecules that have been in close contact with
64 the capping agent.
65
66 \end{abstract}
67
68 \newpage
69
70 %\narrowtext
71
72 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
73 % BODY OF TEXT
74 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
75
76 \section{Introduction}
77 [REFINE LATER, ADD MORE REF.S]
78 Imposed-flux methods in Molecular Dynamics (MD)
79 simulations\cite{MullerPlathe:1997xw} can establish steady state
80 systems with a set applied flux vs a corresponding gradient that can
81 be measured. These methods does not need many trajectories to provide
82 information of transport properties of a given system. Thus, they are
83 utilized in computing thermal and mechanical transfer of homogeneous
84 or bulk systems as well as heterogeneous systems such as liquid-solid
85 interfaces.\cite{kuang:AuThl}
86
87 The Reverse Non-Equilibrium MD (RNEMD) methods adopt constraints that
88 satisfy linear momentum and total energy conservation of a system when
89 imposing fluxes in a simulation. Thus they are compatible with various
90 ensembles, including the micro-canonical (NVE) ensemble, without the
91 need of an external thermostat. The original approaches by
92 M\"{u}ller-Plathe {\it et
93 al.}\cite{MullerPlathe:1997xw,ISI:000080382700030} utilize simple
94 momentum swapping for generating energy/momentum fluxes, which is also
95 compatible with particles of different identities. Although simple to
96 implement in a simulation, this approach can create nonthermal
97 velocity distributions, as discovered by Tenney and
98 Maginn\cite{Maginn:2010}. Furthermore, this approach to kinetic energy
99 transfer between particles of different identities is less efficient
100 when the mass difference between the particles becomes significant,
101 which also limits its application on heterogeneous interfacial
102 systems.
103
104 Recently, we developed a different approach, using Non-Isotropic
105 Velocity Scaling (NIVS) \cite{kuang:164101} algorithm to impose
106 fluxes. Compared to the momentum swapping move, it scales the velocity
107 vectors in two separate regions of a simulated system with respective
108 diagonal scaling matrices. These matrices are determined by solving a
109 set of equations including linear momentum and kinetic energy
110 conservation constraints and target flux satisfaction. This method is
111 able to effectively impose a wide range of kinetic energy fluxes
112 without obvious perturbation to the velocity distributions of the
113 simulated systems, regardless of the presence of heterogeneous
114 interfaces. We have successfully applied this approach in studying the
115 interfacial thermal conductance at metal-solvent
116 interfaces.\cite{kuang:AuThl}
117
118 However, the NIVS approach limits its application in imposing momentum
119 fluxes. Temperature anisotropy can happen under high momentum fluxes,
120 due to the nature of the algorithm. Thus, combining thermal and
121 momentum flux is also difficult to implement with this
122 approach. However, such combination may provide a means to simulate
123 thermal/momentum gradient coupled processes such as freeze
124 desalination. Therefore, developing novel approaches to extend the
125 application of imposed-flux method is desired.
126
127 In this paper, we improve the NIVS method and propose a novel approach
128 to impose fluxes. This approach separate the means of applying
129 momentum and thermal flux with operations in one time step and thus is
130 able to simutaneously impose thermal and momentum flux. Furthermore,
131 the approach retains desirable features of previous RNEMD approaches
132 and is simpler to implement compared to the NIVS method. In what
133 follows, we first present the method to implement the method in a
134 simulation. Then we compare the method on bulk fluids to previous
135 methods. Also, interfacial frictions are computed for a series of
136 interfaces.
137
138 \section{Methodology}
139 Similar to the NIVS methodology,\cite{kuang:164101} we consider a
140 periodic system divided into a series of slabs along a certain axis
141 (e.g. $z$). The unphysical thermal and/or momentum flux is designated
142 from the center slab to one of the end slabs, and thus the center slab
143 would have a lower temperature than the end slab (unless the thermal
144 flux is negative). Therefore, the center slab is denoted as ``$c$''
145 while the end slab as ``$h$''.
146
147 To impose these fluxes, we periodically apply separate operations to
148 velocities of particles {$i$} within the center slab and of particles
149 {$j$} within the end slab:
150 \begin{eqnarray}
151 \vec{v}_i & \leftarrow & c\cdot\left(\vec{v}_i - \langle\vec{v}_c
152 \rangle\right) + \left(\langle\vec{v}_c\rangle + \vec{a}_c\right) \\
153 \vec{v}_j & \leftarrow & h\cdot\left(\vec{v}_j - \langle\vec{v}_h
154 \rangle\right) + \left(\langle\vec{v}_h\rangle + \vec{a}_h\right)
155 \end{eqnarray}
156 where $\langle\vec{v}_c\rangle$ and $\langle\vec{v}_h\rangle$ denotes
157 the instantaneous bulk velocity of slabs $c$ and $h$ respectively
158 before an operation occurs. When a momentum flux $\vec{j}_z(\vec{p})$
159 presents, these bulk velocities would have a corresponding change
160 ($\vec{a}_c$ and $\vec{a}_h$ respectively) according to Newton's
161 second law:
162 \begin{eqnarray}
163 M_c \vec{a}_c & = & -\vec{j}_z(\vec{p}) \Delta t \\
164 M_h \vec{a}_h & = & \vec{j}_z(\vec{p}) \Delta t
165 \end{eqnarray}
166 where
167 \begin{eqnarray}
168 M_c & = & \sum_{i = 1}^{N_c} m_i \\
169 M_h & = & \sum_{j = 1}^{N_h} m_j
170 \end{eqnarray}
171 and $\Delta t$ is the interval between two operations.
172
173 The above operations conserve the linear momentum of a periodic
174 system. To satisfy total energy conservation as well as to impose a
175 thermal flux $J_z$, one would have
176 [SUPPORT INFO MIGHT BE NECESSARY TO PUT EXTRA MATH IN]
177 \begin{eqnarray}
178 K_c - J_z\Delta t & = & c^2 (K_c - \frac{1}{2}M_c \langle\vec{v}_c
179 \rangle^2) + \frac{1}{2}M_c (\langle \vec{v}_c \rangle + \vec{a}_c)^2 \\
180 K_h + J_z\Delta t & = & h^2 (K_h - \frac{1}{2}M_h \langle\vec{v}_h
181 \rangle^2) + \frac{1}{2}M_h (\langle \vec{v}_h \rangle + \vec{a}_h)^2
182 \end{eqnarray}
183 where $K_c$ and $K_h$ denotes translational kinetic energy of slabs
184 $c$ and $h$ respectively before an operation occurs. These
185 translational kinetic energy conservation equations are sufficient to
186 ensure total energy conservation, as the operations applied do not
187 change the potential energy of a system, given that the potential
188 energy does not depend on particle velocity.
189
190 The above sets of equations are sufficient to determine the velocity
191 scaling coefficients ($c$ and $h$) as well as $\vec{a}_c$ and
192 $\vec{a}_h$. Note that two roots of $c$ and $h$ exist
193 respectively. However, to avoid dramatic perturbations to a system,
194 the positive roots (which are closer to 1) are chosen. Figure
195 \ref{method} illustrates the implementation of this algorithm in an
196 individual step.
197
198 \begin{figure}
199 \includegraphics[width=\linewidth]{method}
200 \caption{Illustration of the implementation of the algorithm in a
201 single step. Starting from an ideal velocity distribution, the
202 transformation is used to apply both thermal and momentum flux from
203 the ``c'' slab to the ``h'' slab. As the figure shows, the thermal
204 distributions preserve after this operation.}
205 \label{method}
206 \end{figure}
207
208 By implementing these operations at a certain frequency, a steady
209 thermal and/or momentum flux can be applied and the corresponding
210 temperature and/or momentum gradients can be established.
211
212 This approach is more computationaly efficient compared to the
213 previous NIVS method, in that only quadratic equations are involved,
214 while the NIVS method needs to solve a quartic equations. Furthermore,
215 the method implements isotropic scaling of velocities in respective
216 slabs, unlike the NIVS, where an extra criteria function is necessary
217 to choose a set of coefficients that performs the most isotropic
218 scaling. More importantly, separating the momentum flux imposing from
219 velocity scaling avoids the underlying cause that NIVS produced
220 thermal anisotropy when applying a momentum flux.
221 %NEW METHOD DOESN'T CAUSE UNDESIRED CONCOMITENT MOMENTUM FLUX WHEN
222 %IMPOSING A THERMAL FLUX
223
224 The advantages of the approach over the original momentum swapping
225 approach lies in its nature to preserve a Gaussian
226 distribution. Because the momentum swapping tends to render a
227 nonthermal distribution, when the imposed flux is relatively large,
228 diffusion of the neighboring slabs could no longer remedy this effect,
229 and nonthermal distributions would be observed. Results in later
230 section will illustrate this effect.
231 %NEW METHOD (AND NIVS) HAVE LESS PERTURBATION THAN MOMENTUM SWAPPING
232
233 \section{Computational Details}
234 The algorithm has been implemented in our MD simulation code,
235 OpenMD\cite{Meineke:2005gd,openmd}. We compare the method with
236 previous RNEMD methods or equilibrium MD methods in homogeneous fluids
237 (Lennard-Jones and SPC/E water). And taking advantage of the method,
238 we simulate the interfacial friction of different heterogeneous
239 interfaces (gold-organic solvent and gold-SPC/E water and ice-liquid
240 water).
241
242 \subsection{Simulation Protocols}
243 The systems to be investigated are set up in a orthorhombic simulation
244 cell with periodic boundary conditions in all three dimensions. The
245 $z$ axis of these cells were longer and was set as the gradient axis
246 of temperature and/or momentum. Thus the cells were divided into $N$
247 slabs along this axis, with various $N$ depending on individual
248 system. The $x$ and $y$ axis were usually of the same length in
249 homogeneous systems or close to each other where interfaces
250 presents. In all cases, before introducing a nonequilibrium method to
251 establish steady thermal and/or momentum gradients for further
252 measurements and calculations, canonical ensemble with a Nos\'e-Hoover
253 thermostat\cite{hoover85} and microcanonical ensemble equilibrations
254 were used to prepare systems ready for data
255 collections. Isobaric-isothermal equilibrations are performed before
256 this for SPC/E water systems to reach normal pressure (1 bar), while
257 similar equilibrations are used for interfacial systems to relax the
258 surface tensions.
259
260 While homogeneous fluid systems can be set up with random
261 configurations, our interfacial systems needs extra steps to ensure
262 the interfaces be established properly for computations. The
263 preparation and equilibration of butanethiol covered gold (111)
264 surface and further solvation and equilibration process is described
265 as in reference \cite{kuang:AuThl}.
266
267 As for the ice/liquid water interfaces, the basal surface of ice
268 lattice was first constructed. Hirsch {\it et
269 al.}\cite{doi:10.1021/jp048434u} explored the energetics of ice
270 lattices with different proton orders. We refer to their results and
271 choose the configuration of the lowest energy after geometry
272 optimization as the unit cells of our ice lattices. Although
273 experimental solid/liquid coexistant temperature near normal pressure
274 is 273K, Bryk and Haymet's simulations of ice/liquid water interfaces
275 with different models suggest that for SPC/E, the most stable
276 interface is observed at 225$\pm$5K. Therefore, all our ice/liquid
277 water simulations were carried out under 225K. To have extra
278 protection of the ice lattice during initial equilibration (when the
279 randomly generated liquid phase configuration could release large
280 amount of energy in relaxation), a constraint method (REF?) was
281 adopted until the high energy configuration was relaxed.
282 [MAY ADD A FIGURE HERE FOR BASAL PLANE, MAY INCLUDE PRISM IF POSSIBLE]
283
284 \subsection{Force Field Parameters}
285 For comparison of our new method with previous work, we retain our
286 force field parameters consistent with the results we will compare
287 with. The Lennard-Jones fluid used here for argon , and reduced unit
288 results are reported for direct comparison purpose.
289
290 As for our water simulations, SPC/E model is used throughout this work
291 for consistency. Previous work for transport properties of SPC/E water
292 model is available\cite{Bedrov:2000,10.1063/1.3330544,Medina2011} so
293 that unnecessary repetition of previous methods can be avoided.
294
295 The Au-Au interaction parameters in all simulations are described by
296 the quantum Sutton-Chen (QSC) formulation.\cite{PhysRevB.59.3527} The
297 QSC potentials include zero-point quantum corrections and are
298 reparametrized for accurate surface energies compared to the
299 Sutton-Chen potentials.\cite{Chen90} For gold/water interfaces, the
300 Spohr potential was adopted\cite{ISI:000167766600035} to depict
301 Au-H$_2$O interactions.
302
303 The small organic molecules included in our simulations are the Au
304 surface capping agent butanethiol and liquid hexane and toluene. The
305 United-Atom
306 models\cite{TraPPE-UA.thiols,TraPPE-UA.alkanes,TraPPE-UA.alkylbenzenes}
307 for these components were used in this work for better computational
308 efficiency, while maintaining good accuracy. We refer readers to our
309 previous work\cite{kuang:AuThl} for further details of these models,
310 as well as the interactions between Au and the above organic molecule
311 components.
312
313 \subsection{Thermal conductivities}
314 When $\vec{j}_z(\vec{p})$ is set to zero and a target $J_z$ is set to
315 impose kinetic energy transfer, the method can be used for thermal
316 conductivity computations. Similar to previous RNEMD methods, we
317 assume linear response of the temperature gradient with respect to the
318 thermal flux in general case. And the thermal conductivity ($\lambda$)
319 can be obtained with the imposed kinetic energy flux and the measured
320 thermal gradient:
321 \begin{equation}
322 J_z = -\lambda \frac{\partial T}{\partial z}
323 \end{equation}
324 Like other imposed-flux methods, the energy flux was calculated using
325 the total non-physical energy transferred (${E_{total}}$) from slab
326 ``c'' to slab ``h'', which is recorded throughout a simulation, and
327 the time for data collection $t$:
328 \begin{equation}
329 J_z = \frac{E_{total}}{2 t L_x L_y}
330 \end{equation}
331 where $L_x$ and $L_y$ denotes the dimensions of the plane in a
332 simulation cell perpendicular to the thermal gradient, and a factor of
333 two in the denominator is present for the heat transport occurs in
334 both $+z$ and $-z$ directions. The temperature gradient
335 ${\langle\partial T/\partial z\rangle}$ can be obtained by a linear
336 regression of the temperature profile, which is recorded during a
337 simulation for each slab in a cell. For Lennard-Jones simulations,
338 thermal conductivities are reported in reduced units
339 (${\lambda^*=\lambda \sigma^2 m^{1/2} k_B^{-1}\varepsilon^{-1/2}}$).
340
341 \subsection{Shear viscosities}
342 Alternatively, the method can carry out shear viscosity calculations
343 by switching off $J_z$. One can specify the vector
344 $\vec{j}_z(\vec{p})$ by choosing the three components
345 respectively. For shear viscosity simulations, $j_z(p_z)$ is usually
346 set to zero. Although for isotropic systems, the direction of
347 $\vec{j}_z(\vec{p})$ on the $xy$ plane should not matter, the ability
348 of arbitarily specifying the vector direction in our method provides
349 convenience in anisotropic simulations.
350
351 Similar to thermal conductivity computations, linear response of the
352 momentum gradient with respect to the shear stress is assumed, and the
353 shear viscosity ($\eta$) can be obtained with the imposed momentum
354 flux (e.g. in $x$ direction) and the measured gradient:
355 \begin{equation}
356 j_z(p_x) = -\eta \frac{\partial v_x}{\partial z}
357 \end{equation}
358 where the flux is similarly defined:
359 \begin{equation}
360 j_z(p_x) = \frac{P_x}{2 t L_x L_y}
361 \end{equation}
362 with $P_x$ being the total non-physical momentum transferred within
363 the data collection time. Also, the velocity gradient
364 ${\langle\partial v_x/\partial z\rangle}$ can be obtained using linear
365 regression of the $x$ component of the mean velocity, $\langle
366 v_x\rangle$, in each of the bins. For Lennard-Jones simulations, shear
367 viscosities are reported in reduced units
368 (${\eta^*=\eta\sigma^2(\varepsilon m)^{-1/2}}$).
369
370 \subsection{Interfacial friction and Slip length}
371 While the shear stress results in a velocity gradient within bulk
372 fluid phase, its effect at a solid-liquid interface could vary due to
373 the interaction strength between the two phases. The interfacial
374 friction coefficient $\kappa$ is defined to relate the shear stress
375 (e.g. along $x$-axis) and the relative fluid velocity tangent to the
376 interface:
377 \begin{equation}
378 j_z(p_x)|_{interface} = \kappa\Delta v_x|_{interface}
379 \end{equation}
380 Under ``stick'' boundary condition, $\Delta v_x|_{interface}
381 \rightarrow 0$, which leads to $\kappa\rightarrow\infty$. However, for
382 ``slip'' boundary condition at the solid-liquid interface, $\kappa$
383 becomes finite. To characterize the interfacial boundary conditions,
384 slip length ($\delta$) is defined using $\kappa$ and the shear
385 viscocity of liquid phase ($\eta$):
386 \begin{equation}
387 \delta = \frac{\eta}{\kappa}
388 \end{equation}
389 so that $\delta\rightarrow 0$ in the ``no-slip'' boundary condition,
390 and depicts how ``slippery'' an interface is. Figure \ref{slipLength}
391 illustrates how this quantity is defined and computed for a
392 solid-liquid interface.
393
394 \begin{figure}
395 \includegraphics[width=\linewidth]{defDelta}
396 \caption{The slip length $\delta$ can be obtained from a velocity
397 profile of a solid-liquid interface. An example of Au/hexane
398 interfaces is shown.}
399 \label{slipLength}
400 \end{figure}
401
402 In our method, a shear stress can be applied similar to shear
403 viscosity computations by applying an unphysical momentum flux
404 (e.g. $j_z(p_x)$). A corresponding velocity profile can be obtained as
405 shown in Figure \ref{slipLength}, in which the velocity gradients
406 within liquid phase and velocity difference at the liquid-solid
407 interface can be measured respectively. Further calculations and
408 characterizations of the interface can be carried out using these
409 data.
410 [MENTION IN RESULTS THAT ETA OBTAINED HERE DOES NOT NECESSARILY EQUAL
411 TO BULK VALUES]
412
413 \section{Results}
414 [L-J COMPARED TO RNEMD NIVS; WATER COMPARED TO RNEMD NIVS AND EMD;
415 SLIP BOUNDARY VS STICK BOUNDARY; ICE-WATER INTERFACES]
416
417 There are many factors contributing to the measured interfacial
418 conductance; some of these factors are physically motivated
419 (e.g. coverage of the surface by the capping agent coverage and
420 solvent identity), while some are governed by parameters of the
421 methodology (e.g. applied flux and the formulas used to obtain the
422 conductance). In this section we discuss the major physical and
423 calculational effects on the computed conductivity.
424
425 \subsection{Effects due to capping agent coverage}
426
427 A series of different initial conditions with a range of surface
428 coverages was prepared and solvated with various with both of the
429 solvent molecules. These systems were then equilibrated and their
430 interfacial thermal conductivity was measured with the NIVS
431 algorithm. Figure \ref{coverage} demonstrates the trend of conductance
432 with respect to surface coverage.
433
434 \begin{figure}
435 \includegraphics[width=\linewidth]{coverage}
436 \caption{The interfacial thermal conductivity ($G$) has a
437 non-monotonic dependence on the degree of surface capping. This
438 data is for the Au(111) / butanethiol / solvent interface with
439 various UA force fields at $\langle T\rangle \sim $200K.}
440 \label{coverage}
441 \end{figure}
442
443 In partially covered surfaces, the derivative definition for
444 $G^\prime$ (Eq. \ref{derivativeG}) becomes difficult to apply, as the
445 location of maximum change of $\lambda$ becomes washed out. The
446 discrete definition (Eq. \ref{discreteG}) is easier to apply, as the
447 Gibbs dividing surface is still well-defined. Therefore, $G$ (not
448 $G^\prime$) was used in this section.
449
450 From Figure \ref{coverage}, one can see the significance of the
451 presence of capping agents. When even a small fraction of the Au(111)
452 surface sites are covered with butanethiols, the conductivity exhibits
453 an enhancement by at least a factor of 3. Capping agents are clearly
454 playing a major role in thermal transport at metal / organic solvent
455 surfaces.
456
457 We note a non-monotonic behavior in the interfacial conductance as a
458 function of surface coverage. The maximum conductance (largest $G$)
459 happens when the surfaces are about 75\% covered with butanethiol
460 caps. The reason for this behavior is not entirely clear. One
461 explanation is that incomplete butanethiol coverage allows small gaps
462 between butanethiols to form. These gaps can be filled by transient
463 solvent molecules. These solvent molecules couple very strongly with
464 the hot capping agent molecules near the surface, and can then carry
465 away (diffusively) the excess thermal energy from the surface.
466
467 There appears to be a competition between the conduction of the
468 thermal energy away from the surface by the capping agents (enhanced
469 by greater coverage) and the coupling of the capping agents with the
470 solvent (enhanced by interdigitation at lower coverages). This
471 competition would lead to the non-monotonic coverage behavior observed
472 here.
473
474 Results for rigid body toluene solvent, as well as the UA hexane, are
475 within the ranges expected from prior experimental
476 work.\cite{Wilson:2002uq,cahill:793,PhysRevB.80.195406} This suggests
477 that explicit hydrogen atoms might not be required for modeling
478 thermal transport in these systems. C-H vibrational modes do not see
479 significant excited state population at low temperatures, and are not
480 likely to carry lower frequency excitations from the solid layer into
481 the bulk liquid.
482
483 The toluene solvent does not exhibit the same behavior as hexane in
484 that $G$ remains at approximately the same magnitude when the capping
485 coverage increases from 25\% to 75\%. Toluene, as a rigid planar
486 molecule, cannot occupy the relatively small gaps between the capping
487 agents as easily as the chain-like {\it n}-hexane. The effect of
488 solvent coupling to the capping agent is therefore weaker in toluene
489 except at the very lowest coverage levels. This effect counters the
490 coverage-dependent conduction of heat away from the metal surface,
491 leading to a much flatter $G$ vs. coverage trend than is observed in
492 {\it n}-hexane.
493
494 \subsection{Effects due to Solvent \& Solvent Models}
495 In addition to UA solvent and capping agent models, AA models have
496 also been included in our simulations. In most of this work, the same
497 (UA or AA) model for solvent and capping agent was used, but it is
498 also possible to utilize different models for different components.
499 We have also included isotopic substitutions (Hydrogen to Deuterium)
500 to decrease the explicit vibrational overlap between solvent and
501 capping agent. Table \ref{modelTest} summarizes the results of these
502 studies.
503
504 \begin{table*}
505 \begin{minipage}{\linewidth}
506 \begin{center}
507
508 \caption{Computed interfacial thermal conductance ($G$ and
509 $G^\prime$) values for interfaces using various models for
510 solvent and capping agent (or without capping agent) at
511 $\langle T\rangle\sim$200K. Here ``D'' stands for deuterated
512 solvent or capping agent molecules. Error estimates are
513 indicated in parentheses.}
514
515 \begin{tabular}{llccc}
516 \hline\hline
517 Butanethiol model & Solvent & $G$ & $G^\prime$ \\
518 (or bare surface) & model & \multicolumn{2}{c}{(MW/m$^2$/K)} \\
519 \hline
520 UA & UA hexane & 131(9) & 87(10) \\
521 & UA hexane(D) & 153(5) & 136(13) \\
522 & AA hexane & 131(6) & 122(10) \\
523 & UA toluene & 187(16) & 151(11) \\
524 & AA toluene & 200(36) & 149(53) \\
525 \hline
526 AA & UA hexane & 116(9) & 129(8) \\
527 & AA hexane & 442(14) & 356(31) \\
528 & AA hexane(D) & 222(12) & 234(54) \\
529 & UA toluene & 125(25) & 97(60) \\
530 & AA toluene & 487(56) & 290(42) \\
531 \hline
532 AA(D) & UA hexane & 158(25) & 172(4) \\
533 & AA hexane & 243(29) & 191(11) \\
534 & AA toluene & 364(36) & 322(67) \\
535 \hline
536 bare & UA hexane & 46.5(3.2) & 49.4(4.5) \\
537 & UA hexane(D) & 43.9(4.6) & 43.0(2.0) \\
538 & AA hexane & 31.0(1.4) & 29.4(1.3) \\
539 & UA toluene & 70.1(1.3) & 65.8(0.5) \\
540 \hline\hline
541 \end{tabular}
542 \label{modelTest}
543 \end{center}
544 \end{minipage}
545 \end{table*}
546
547 To facilitate direct comparison between force fields, systems with the
548 same capping agent and solvent were prepared with the same length
549 scales for the simulation cells.
550
551 On bare metal / solvent surfaces, different force field models for
552 hexane yield similar results for both $G$ and $G^\prime$, and these
553 two definitions agree with each other very well. This is primarily an
554 indicator of weak interactions between the metal and the solvent.
555
556 For the fully-covered surfaces, the choice of force field for the
557 capping agent and solvent has a large impact on the calculated values
558 of $G$ and $G^\prime$. The AA thiol to AA solvent conductivities are
559 much larger than their UA to UA counterparts, and these values exceed
560 the experimental estimates by a large measure. The AA force field
561 allows significant energy to go into C-H (or C-D) stretching modes,
562 and since these modes are high frequency, this non-quantum behavior is
563 likely responsible for the overestimate of the conductivity. Compared
564 to the AA model, the UA model yields more reasonable conductivity
565 values with much higher computational efficiency.
566
567 \subsubsection{Are electronic excitations in the metal important?}
568 Because they lack electronic excitations, the QSC and related embedded
569 atom method (EAM) models for gold are known to predict unreasonably
570 low values for bulk conductivity
571 ($\lambda$).\cite{kuang:164101,ISI:000207079300006,Clancy:1992} If the
572 conductance between the phases ($G$) is governed primarily by phonon
573 excitation (and not electronic degrees of freedom), one would expect a
574 classical model to capture most of the interfacial thermal
575 conductance. Our results for $G$ and $G^\prime$ indicate that this is
576 indeed the case, and suggest that the modeling of interfacial thermal
577 transport depends primarily on the description of the interactions
578 between the various components at the interface. When the metal is
579 chemically capped, the primary barrier to thermal conductivity appears
580 to be the interface between the capping agent and the surrounding
581 solvent, so the excitations in the metal have little impact on the
582 value of $G$.
583
584 \subsection{Effects due to methodology and simulation parameters}
585
586 We have varied the parameters of the simulations in order to
587 investigate how these factors would affect the computation of $G$. Of
588 particular interest are: 1) the length scale for the applied thermal
589 gradient (modified by increasing the amount of solvent in the system),
590 2) the sign and magnitude of the applied thermal flux, 3) the average
591 temperature of the simulation (which alters the solvent density during
592 equilibration), and 4) the definition of the interfacial conductance
593 (Eqs. (\ref{discreteG}) or (\ref{derivativeG})) used in the
594 calculation.
595
596 Systems of different lengths were prepared by altering the number of
597 solvent molecules and extending the length of the box along the $z$
598 axis to accomodate the extra solvent. Equilibration at the same
599 temperature and pressure conditions led to nearly identical surface
600 areas ($L_x$ and $L_y$) available to the metal and capping agent,
601 while the extra solvent served mainly to lengthen the axis that was
602 used to apply the thermal flux. For a given value of the applied
603 flux, the different $z$ length scale has only a weak effect on the
604 computed conductivities.
605
606 \subsubsection{Effects of applied flux}
607 The NIVS algorithm allows changes in both the sign and magnitude of
608 the applied flux. It is possible to reverse the direction of heat
609 flow simply by changing the sign of the flux, and thermal gradients
610 which would be difficult to obtain experimentally ($5$ K/\AA) can be
611 easily simulated. However, the magnitude of the applied flux is not
612 arbitrary if one aims to obtain a stable and reliable thermal gradient.
613 A temperature gradient can be lost in the noise if $|J_z|$ is too
614 small, and excessive $|J_z|$ values can cause phase transitions if the
615 extremes of the simulation cell become widely separated in
616 temperature. Also, if $|J_z|$ is too large for the bulk conductivity
617 of the materials, the thermal gradient will never reach a stable
618 state.
619
620 Within a reasonable range of $J_z$ values, we were able to study how
621 $G$ changes as a function of this flux. In what follows, we use
622 positive $J_z$ values to denote the case where energy is being
623 transferred by the method from the metal phase and into the liquid.
624 The resulting gradient therefore has a higher temperature in the
625 liquid phase. Negative flux values reverse this transfer, and result
626 in higher temperature metal phases. The conductance measured under
627 different applied $J_z$ values is listed in Tables 2 and 3 in the
628 supporting information. These results do not indicate that $G$ depends
629 strongly on $J_z$ within this flux range. The linear response of flux
630 to thermal gradient simplifies our investigations in that we can rely
631 on $G$ measurement with only a small number $J_z$ values.
632
633 The sign of $J_z$ is a different matter, however, as this can alter
634 the temperature on the two sides of the interface. The average
635 temperature values reported are for the entire system, and not for the
636 liquid phase, so at a given $\langle T \rangle$, the system with
637 positive $J_z$ has a warmer liquid phase. This means that if the
638 liquid carries thermal energy via diffusive transport, {\it positive}
639 $J_z$ values will result in increased molecular motion on the liquid
640 side of the interface, and this will increase the measured
641 conductivity.
642
643 \subsubsection{Effects due to average temperature}
644
645 We also studied the effect of average system temperature on the
646 interfacial conductance. The simulations are first equilibrated in
647 the NPT ensemble at 1 atm. The TraPPE-UA model for hexane tends to
648 predict a lower boiling point (and liquid state density) than
649 experiments. This lower-density liquid phase leads to reduced contact
650 between the hexane and butanethiol, and this accounts for our
651 observation of lower conductance at higher temperatures. In raising
652 the average temperature from 200K to 250K, the density drop of
653 $\sim$20\% in the solvent phase leads to a $\sim$40\% drop in the
654 conductance.
655
656 Similar behavior is observed in the TraPPE-UA model for toluene,
657 although this model has better agreement with the experimental
658 densities of toluene. The expansion of the toluene liquid phase is
659 not as significant as that of the hexane (8.3\% over 100K), and this
660 limits the effect to $\sim$20\% drop in thermal conductivity.
661
662 Although we have not mapped out the behavior at a large number of
663 temperatures, is clear that there will be a strong temperature
664 dependence in the interfacial conductance when the physical properties
665 of one side of the interface (notably the density) change rapidly as a
666 function of temperature.
667
668 Besides the lower interfacial thermal conductance, surfaces at
669 relatively high temperatures are susceptible to reconstructions,
670 particularly when butanethiols fully cover the Au(111) surface. These
671 reconstructions include surface Au atoms which migrate outward to the
672 S atom layer, and butanethiol molecules which embed into the surface
673 Au layer. The driving force for this behavior is the strong Au-S
674 interactions which are modeled here with a deep Lennard-Jones
675 potential. This phenomenon agrees with reconstructions that have been
676 experimentally
677 observed.\cite{doi:10.1021/j100035a033,doi:10.1021/la026493y}. Vlugt
678 {\it et al.} kept their Au(111) slab rigid so that their simulations
679 could reach 300K without surface
680 reconstructions.\cite{vlugt:cpc2007154} Since surface reconstructions
681 blur the interface, the measurement of $G$ becomes more difficult to
682 conduct at higher temperatures. For this reason, most of our
683 measurements are undertaken at $\langle T\rangle\sim$200K where
684 reconstruction is minimized.
685
686 However, when the surface is not completely covered by butanethiols,
687 the simulated system appears to be more resistent to the
688 reconstruction. Our Au / butanethiol / toluene system had the Au(111)
689 surfaces 90\% covered by butanethiols, but did not see this above
690 phenomena even at $\langle T\rangle\sim$300K. That said, we did
691 observe butanethiols migrating to neighboring three-fold sites during
692 a simulation. Since the interface persisted in these simulations, we
693 were able to obtain $G$'s for these interfaces even at a relatively
694 high temperature without being affected by surface reconstructions.
695
696 \section{Discussion}
697 [COMBINE W. RESULTS]
698 The primary result of this work is that the capping agent acts as an
699 efficient thermal coupler between solid and solvent phases. One of
700 the ways the capping agent can carry out this role is to down-shift
701 between the phonon vibrations in the solid (which carry the heat from
702 the gold) and the molecular vibrations in the liquid (which carry some
703 of the heat in the solvent).
704
705 To investigate the mechanism of interfacial thermal conductance, the
706 vibrational power spectrum was computed. Power spectra were taken for
707 individual components in different simulations. To obtain these
708 spectra, simulations were run after equilibration in the
709 microcanonical (NVE) ensemble and without a thermal
710 gradient. Snapshots of configurations were collected at a frequency
711 that is higher than that of the fastest vibrations occurring in the
712 simulations. With these configurations, the velocity auto-correlation
713 functions can be computed:
714 \begin{equation}
715 C_A (t) = \langle\vec{v}_A (t)\cdot\vec{v}_A (0)\rangle
716 \label{vCorr}
717 \end{equation}
718 The power spectrum is constructed via a Fourier transform of the
719 symmetrized velocity autocorrelation function,
720 \begin{equation}
721 \hat{f}(\omega) =
722 \int_{-\infty}^{\infty} C_A (t) e^{-2\pi it\omega}\,dt
723 \label{fourier}
724 \end{equation}
725
726 \subsection{The role of specific vibrations}
727 The vibrational spectra for gold slabs in different environments are
728 shown as in Figure \ref{specAu}. Regardless of the presence of
729 solvent, the gold surfaces which are covered by butanethiol molecules
730 exhibit an additional peak observed at a frequency of
731 $\sim$165cm$^{-1}$. We attribute this peak to the S-Au bonding
732 vibration. This vibration enables efficient thermal coupling of the
733 surface Au layer to the capping agents. Therefore, in our simulations,
734 the Au / S interfaces do not appear to be the primary barrier to
735 thermal transport when compared with the butanethiol / solvent
736 interfaces. This supports the results of Luo {\it et
737 al.}\cite{Luo20101}, who reported $G$ for Au-SAM junctions roughly
738 twice as large as what we have computed for the thiol-liquid
739 interfaces.
740
741 \begin{figure}
742 \includegraphics[width=\linewidth]{vibration}
743 \caption{The vibrational power spectrum for thiol-capped gold has an
744 additional vibrational peak at $\sim $165cm$^{-1}$. Bare gold
745 surfaces (both with and without a solvent over-layer) are missing
746 this peak. A similar peak at $\sim $165cm$^{-1}$ also appears in
747 the vibrational power spectrum for the butanethiol capping agents.}
748 \label{specAu}
749 \end{figure}
750
751 Also in this figure, we show the vibrational power spectrum for the
752 bound butanethiol molecules, which also exhibits the same
753 $\sim$165cm$^{-1}$ peak.
754
755 \subsection{Overlap of power spectra}
756 A comparison of the results obtained from the two different organic
757 solvents can also provide useful information of the interfacial
758 thermal transport process. In particular, the vibrational overlap
759 between the butanethiol and the organic solvents suggests a highly
760 efficient thermal exchange between these components. Very high
761 thermal conductivity was observed when AA models were used and C-H
762 vibrations were treated classically. The presence of extra degrees of
763 freedom in the AA force field yields higher heat exchange rates
764 between the two phases and results in a much higher conductivity than
765 in the UA force field. The all-atom classical models include high
766 frequency modes which should be unpopulated at our relatively low
767 temperatures. This artifact is likely the cause of the high thermal
768 conductance in all-atom MD simulations.
769
770 The similarity in the vibrational modes available to solvent and
771 capping agent can be reduced by deuterating one of the two components
772 (Fig. \ref{aahxntln}). Once either the hexanes or the butanethiols
773 are deuterated, one can observe a significantly lower $G$ and
774 $G^\prime$ values (Table \ref{modelTest}).
775
776 \begin{figure}
777 \includegraphics[width=\linewidth]{aahxntln}
778 \caption{Spectra obtained for all-atom (AA) Au / butanethiol / solvent
779 systems. When butanethiol is deuterated (lower left), its
780 vibrational overlap with hexane decreases significantly. Since
781 aromatic molecules and the butanethiol are vibrationally dissimilar,
782 the change is not as dramatic when toluene is the solvent (right).}
783 \label{aahxntln}
784 \end{figure}
785
786 For the Au / butanethiol / toluene interfaces, having the AA
787 butanethiol deuterated did not yield a significant change in the
788 measured conductance. Compared to the C-H vibrational overlap between
789 hexane and butanethiol, both of which have alkyl chains, the overlap
790 between toluene and butanethiol is not as significant and thus does
791 not contribute as much to the heat exchange process.
792
793 Previous observations of Zhang {\it et al.}\cite{hase:2010} indicate
794 that the {\it intra}molecular heat transport due to alkylthiols is
795 highly efficient. Combining our observations with those of Zhang {\it
796 et al.}, it appears that butanethiol acts as a channel to expedite
797 heat flow from the gold surface and into the alkyl chain. The
798 vibrational coupling between the metal and the liquid phase can
799 therefore be enhanced with the presence of suitable capping agents.
800
801 Deuterated models in the UA force field did not decouple the thermal
802 transport as well as in the AA force field. The UA models, even
803 though they have eliminated the high frequency C-H vibrational
804 overlap, still have significant overlap in the lower-frequency
805 portions of the infrared spectra (Figure \ref{uahxnua}). Deuterating
806 the UA models did not decouple the low frequency region enough to
807 produce an observable difference for the results of $G$ (Table
808 \ref{modelTest}).
809
810 \begin{figure}
811 \includegraphics[width=\linewidth]{uahxnua}
812 \caption{Vibrational power spectra for UA models for the butanethiol
813 and hexane solvent (upper panel) show the high degree of overlap
814 between these two molecules, particularly at lower frequencies.
815 Deuterating a UA model for the solvent (lower panel) does not
816 decouple the two spectra to the same degree as in the AA force
817 field (see Fig \ref{aahxntln}).}
818 \label{uahxnua}
819 \end{figure}
820
821 \section{Conclusions}
822 The NIVS algorithm has been applied to simulations of
823 butanethiol-capped Au(111) surfaces in the presence of organic
824 solvents. This algorithm allows the application of unphysical thermal
825 flux to transfer heat between the metal and the liquid phase. With the
826 flux applied, we were able to measure the corresponding thermal
827 gradients and to obtain interfacial thermal conductivities. Under
828 steady states, 2-3 ns trajectory simulations are sufficient for
829 computation of this quantity.
830
831 Our simulations have seen significant conductance enhancement in the
832 presence of capping agent, compared with the bare gold / liquid
833 interfaces. The vibrational coupling between the metal and the liquid
834 phase is enhanced by a chemically-bonded capping agent. Furthermore,
835 the coverage percentage of the capping agent plays an important role
836 in the interfacial thermal transport process. Moderately low coverages
837 allow higher contact between capping agent and solvent, and thus could
838 further enhance the heat transfer process, giving a non-monotonic
839 behavior of conductance with increasing coverage.
840
841 Our results, particularly using the UA models, agree well with
842 available experimental data. The AA models tend to overestimate the
843 interfacial thermal conductance in that the classically treated C-H
844 vibrations become too easily populated. Compared to the AA models, the
845 UA models have higher computational efficiency with satisfactory
846 accuracy, and thus are preferable in modeling interfacial thermal
847 transport.
848
849 Of the two definitions for $G$, the discrete form
850 (Eq. \ref{discreteG}) was easier to use and gives out relatively
851 consistent results, while the derivative form (Eq. \ref{derivativeG})
852 is not as versatile. Although $G^\prime$ gives out comparable results
853 and follows similar trend with $G$ when measuring close to fully
854 covered or bare surfaces, the spatial resolution of $T$ profile
855 required for the use of a derivative form is limited by the number of
856 bins and the sampling required to obtain thermal gradient information.
857
858 Vlugt {\it et al.} have investigated the surface thiol structures for
859 nanocrystalline gold and pointed out that they differ from those of
860 the Au(111) surface.\cite{landman:1998,vlugt:cpc2007154} This
861 difference could also cause differences in the interfacial thermal
862 transport behavior. To investigate this problem, one would need an
863 effective method for applying thermal gradients in non-planar
864 (i.e. spherical) geometries.
865
866 \section{Acknowledgments}
867 Support for this project was provided by the National Science
868 Foundation under grant CHE-0848243. Computational time was provided by
869 the Center for Research Computing (CRC) at the University of Notre
870 Dame.
871
872 \newpage
873
874 \bibliography{stokes}
875
876 \end{doublespace}
877 \end{document}
878