17 |
|
\textwidth 6.5in |
18 |
|
\brokenpenalty=10000 |
19 |
|
\renewcommand{\baselinestretch}{1.2} |
20 |
+ |
\usepackage[square, comma, sort&compress]{natbib} |
21 |
+ |
\bibpunct{[}{]}{,}{n}{}{;} |
22 |
|
|
23 |
+ |
|
24 |
|
%\renewcommand\citemid{\ } % no comma in optional reference note |
25 |
|
\lstset{language=C,frame=TB,basicstyle=\tiny,basicstyle=\ttfamily, % |
26 |
|
xleftmargin=0.25in, xrightmargin=0.25in,captionpos=b, % |
47 |
|
\newcolumntype{M}{p{1.55in}} |
48 |
|
|
49 |
|
|
50 |
< |
\title{{\sc OpenMD}: Molecular Dynamics in the Open} |
50 |
> |
\title{{\sc OpenMD-2}: Molecular Dynamics in the Open} |
51 |
|
|
52 |
< |
\author{Shenyu Kuang, Charles F. Vardeman II, \\ |
53 |
< |
Teng Lin, Christopher J. Fennell, Xiuquan Sun, \\ |
52 |
> |
\author{Shenyu Kuang, Joseph Michalka, Kelsey Stocker, James Marr, \\ |
53 |
> |
Teng Lin, Charles F. Vardeman II, Christopher J. Fennell, Xiuquan Sun, \\ |
54 |
|
Chunlei Li, Kyle Daily, Yang Zheng, Matthew A. Meineke, and \\ |
55 |
|
J. Daniel Gezelter \\ |
56 |
|
Department of Chemistry and Biochemistry\\ |
499 |
|
\endhead |
500 |
|
\hline |
501 |
|
\endfoot |
502 |
< |
{\tt forceField} & string & Sets the force field. & Possible force |
503 |
< |
fields are DUFF, WATER, LJ, EAM, SC, and CLAY. \\ |
502 |
> |
{\tt forceField} & string & Sets the base name for the force field file & |
503 |
> |
OpenMD appends a {\tt .frc} to the end of this to look for a force |
504 |
> |
field file.\\ |
505 |
|
{\tt component} & & Defines the molecular components of the system & |
506 |
|
Every {\tt $<$MetaData$>$} block must have a component statement. \\ |
507 |
|
{\tt minimizer} & string & Chooses a minimizer & Possible minimizers |
2943 |
|
\end{longtable} |
2944 |
|
|
2945 |
|
% \chapter{\label{section:restraints}Restraints} |
2946 |
< |
% Restraints are external potentials that are added to a system to keep |
2947 |
< |
% particular molecules or collections of particles close to some |
2946 |
> |
% Restraints are external potentials that are added to a system to |
2947 |
> |
% keep particular molecules or collections of particles close to some |
2948 |
|
% reference structure. A restraint can be a collective |
2949 |
|
|
2950 |
|
\chapter{\label{section:thermInt}Thermodynamic Integration} |
3088 |
|
\chapter{\label{section:rnemd}RNEMD} |
3089 |
|
|
3090 |
|
There are many ways to compute transport properties from molecular |
3091 |
< |
dynamic simulations. Equilibrium Molecular Dynamics (EMD) simulations |
3092 |
< |
can be used by computing relevant time correlation functions and |
3093 |
< |
assuming linear response theory holds. These approaches are generally |
3094 |
< |
subject to noise and poor convergence of the relevant correlation |
3095 |
< |
functions. Traditional Non-equilibrium Molecular Dynamics (NEMD) |
3096 |
< |
methods impose a gradient (e.g. thermal or momentum) on a simulation. |
3097 |
< |
However, the resulting flux is often difficult to |
3091 |
> |
dynamics simulations. Equilibrium Molecular Dynamics (EMD) |
3092 |
> |
simulations can be used by computing relevant time correlation |
3093 |
> |
functions and assuming linear response theory holds. These approaches |
3094 |
> |
are generally subject to noise and poor convergence of the relevant |
3095 |
> |
correlation functions. Traditional Non-equilibrium Molecular Dynamics |
3096 |
> |
(NEMD) methods impose a gradient (e.g. thermal or momentum) on a |
3097 |
> |
simulation. However, the resulting flux is often difficult to |
3098 |
|
measure. Furthermore, problems arise for NEMD simulations of |
3099 |
|
heterogeneous systems, such as phase-phase boundaries or interfaces, |
3100 |
|
where the type of gradient to enforce at the boundary between |
3101 |
|
materials is unclear. |
3102 |
|
|
3103 |
< |
{\it Reverse} Non-Equilibrium Molecular Dynamics (RNEMD) methods adopt a |
3104 |
< |
different approach in that an unphysical {\it flux} is imposed between |
3105 |
< |
different regions or ``slabs'' of the simulation box. The response of |
3106 |
< |
the system is to develop a temperature or momentum {\it gradient} |
3107 |
< |
between the two regions. Since the amount of the applied flux is known |
3108 |
< |
exactly, and the measurement of gradient is generally less |
3109 |
< |
complicated, imposed-flux methods typically take shorter simulation |
3110 |
< |
times to obtain converged results for transport properties. |
3103 |
> |
{\it Reverse} Non-Equilibrium Molecular Dynamics (RNEMD) methods adopt |
3104 |
> |
a different approach in that an unphysical {\it flux} is imposed |
3105 |
> |
between different regions or ``slabs'' of the simulation box. The |
3106 |
> |
response of the system is to develop a temperature or momentum {\it |
3107 |
> |
gradient} between the two regions. Since the amount of the applied |
3108 |
> |
flux is known exactly, and the measurement of gradient is generally |
3109 |
> |
less complicated, imposed-flux methods typically take shorter |
3110 |
> |
simulation times to obtain converged results for transport properties. |
3111 |
|
|
3112 |
< |
%RNEMD figure |
3112 |
> |
\begin{figure} |
3113 |
> |
\includegraphics[width=\linewidth]{rnemdDemo} |
3114 |
> |
\caption{The (VSS) RNEMD approach imposes unphysical transfer of both |
3115 |
> |
linear momentum and kinetic energy between a ``hot'' slab and a |
3116 |
> |
``cold'' slab in the simulation box. The system responds to this |
3117 |
> |
imposed flux by generating both momentum and temperature gradients. |
3118 |
> |
The slope of the gradients can then be used to compute transport |
3119 |
> |
properties (e.g. shear viscosity and thermal conductivity).} |
3120 |
> |
\label{rnemdDemo} |
3121 |
> |
\end{figure} |
3122 |
> |
|
3123 |
> |
The original ``swapping'' approaches by M\"{u}ller-Plathe {\it et |
3124 |
> |
al.}\cite{ISI:000080382700030,MullerPlathe:1997xw} can be understood |
3125 |
> |
as a sequence of imaginary elastic collisions between particles in |
3126 |
> |
opposite slabs. In each collision, the entire momentum vectors of |
3127 |
> |
both particles may be exchanged to generate a thermal |
3128 |
> |
flux. Alternatively, a single component of the momentum vectors may be |
3129 |
> |
exchanged to generate a shear flux. This algorithm turns out to be |
3130 |
> |
quite useful in many simulations. However, the M\"{u}ller-Plathe |
3131 |
> |
swapping approach perturbs the system away from ideal |
3132 |
> |
Maxwell-Boltzmann distributions, and this may leads to undesirable |
3133 |
> |
side-effects when the applied flux becomes large.\cite{Maginn:2010} |
3134 |
> |
This limits the application of the swapping algorithm, so in OpenMD, |
3135 |
> |
we implement two additional algorithms for RNEMD in addition to the |
3136 |
> |
original swapping approach. |
3137 |
|
|
3138 |
+ |
{\bf Non-Isotropic Velocity Scaling (NIVS):}\cite{kuang:164101} |
3139 |
+ |
Instead of having momentum exchange between {\it individual particles} |
3140 |
+ |
in each slab, the NIVS algorithm applies velocity scaling to all of |
3141 |
+ |
the selected particles in both slabs. A combination of linear |
3142 |
+ |
momentum, kinetic energy, and flux constraint equations governs the |
3143 |
+ |
amount of velocity scaling performed at each step. Interested readers |
3144 |
+ |
should consult ref. \citealp{kuang:164101} for further details on the |
3145 |
+ |
methodology. |
3146 |
|
|
3147 |
< |
RNEMD methods further its advantages by utilizing momentum- and |
3148 |
< |
energy-conserving approaches to apply fluxes. The original |
3149 |
< |
``swapping'' approach by Muller-Plathe {\it et al.} %CITATIONS |
3150 |
< |
can be seen as an imaginary elastic collision between selected |
3151 |
< |
particles for each momentum exchange. This simple to implement |
3152 |
< |
algorithm turned out to be quite useful in many simulations. However, |
3117 |
< |
the approach inherently perturbs the ideal Maxwell-Boltzmann |
3118 |
< |
distributions, which leads to undesirable side-effects when the |
3119 |
< |
applied exchanged flux becomes quite large. %CITATION |
3120 |
< |
This limits the range of flux available to the method, and also its |
3121 |
< |
applications. |
3147 |
> |
NIVS has been shown to be very effective at producing thermal |
3148 |
> |
gradients and for computing thermal conductivities, particularly for |
3149 |
> |
heterogeneous interfaces. Although the NIVS algorithm can also be |
3150 |
> |
applied to impose a directional momentum flux, thermal anisotropy was |
3151 |
> |
observed in relatively high flux simulations, and the method is not |
3152 |
> |
suitable for imposing a shear flux. |
3153 |
|
|
3154 |
< |
In OpenMD, we improve the above method by introducing two alternative |
3155 |
< |
approaches: |
3154 |
> |
{\bf Velocity Shearing and Scaling (VSS)}:\cite{2012MolPh.110..691K} |
3155 |
> |
The third RNEMD algorithm implemented in OpenMD utilizes a series of |
3156 |
> |
simultaneous velocity shearing and scaling exchanges between the two |
3157 |
> |
slabs. This method results in a set of simpler equations to satisfy |
3158 |
> |
the conservation constraints while creating a desired flux between the |
3159 |
> |
two slabs. |
3160 |
|
|
3161 |
< |
Non-Isotropic Velocity Scaling (NIVS): %CITATION |
3162 |
< |
Instead of have two individual particles involved in momentum |
3163 |
< |
exchange, this algorithm applies scaling to all the particles in |
3164 |
< |
particular regions: |
3161 |
> |
The VSS approach is versatile in that it may be used to implement both |
3162 |
> |
thermal and shear transport either separately or simultaneously. |
3163 |
> |
Perturbations of velocities away from the ideal Maxwell-Boltzmann |
3164 |
> |
distributions are minimal, and thermal anisotropy is kept to a |
3165 |
> |
minimum. This ability to generate simultaneous thermal and shear |
3166 |
> |
fluxes has been utilized to map out the shear viscosity of SPC/E water |
3167 |
> |
in a wide range of temperature (90~K) just with a single simulation. |
3168 |
> |
VSS-RNEMD also allows the directional momentum flux to have quite |
3169 |
> |
arbitary directions, which could aid in the study of anisotropic solid |
3170 |
> |
surfaces in contact with liquid environments. |
3171 |
|
|
3172 |
< |
%NIVS equations |
3173 |
< |
|
3174 |
< |
Although the above matrices can be diagonal as shown, these |
3175 |
< |
coefficients cannot be always the same, in order to satisfy the linear |
3176 |
< |
momentum and kinetic energy conservation constraints: |
3172 |
> |
{\bf What the user needs to specify:} To carry out a RNEMD simulation, |
3173 |
> |
a user must specify a number of parameters in the MetaData (.md) file. |
3174 |
> |
Because the RNEMD methods have a large number of parameters, these |
3175 |
> |
must be enclosed in a {\tt RNEMD\{...\}} block. The most important |
3176 |
> |
parameters to specify are the {\tt useRNEMD}, {\tt fluxType} and flux |
3177 |
> |
parameters. Most other parameters (summarized in table |
3178 |
> |
\ref{table:rnemd}) have reasonable default values. {\tt fluxType} |
3179 |
> |
sets up the kind of exchange that will be carried out between the two |
3180 |
> |
slabs (either Kinetic Energy ({\tt KE}) or momentum ({\tt Px, Py, Pz, |
3181 |
> |
Pvector}), or some combination of these). The flux is specified |
3182 |
> |
with the use of three possible parameters: {\tt kineticFlux} for |
3183 |
> |
kinetic energy exchange, as well as {\tt momentumFlux} or {\tt |
3184 |
> |
momentumFluxVector} for simulations with directional exchange. |
3185 |
|
|
3186 |
< |
%Conservation equations |
3186 |
> |
{\bf How to process the results:} OpenMD will generate a {\tt .rnemd} |
3187 |
> |
file with the same prefix as the original {\tt .md} file. This file |
3188 |
> |
contains a running average of properties of interest computed within a |
3189 |
> |
set of bins that divide the simulation cell along the $z$-axis. The |
3190 |
> |
first column of the {\tt .rnemd} file is the $z$ coordinate of the |
3191 |
> |
center of each bin, while following columns may contain the average |
3192 |
> |
temperature, velocity, or density within each bin. The output format |
3193 |
> |
in the {\tt .rnemd} file can be altered with the {\tt outputFields}, |
3194 |
> |
{\tt outputBins}, and {\tt outputFileName} parameters. A report at |
3195 |
> |
the top of the {\tt .rnemd} file contains the current exchange totals |
3196 |
> |
as well as the average flux applied during the simulation. Using the |
3197 |
> |
slope of the temperature or velocity gradient obtaine from the {\tt |
3198 |
> |
.rnemd} file along with the applied flux, the user can very simply |
3199 |
> |
arrive at estimates of thermal conductivities ($\lambda$), |
3200 |
> |
\begin{equation} |
3201 |
> |
J_z = -\lambda \frac{\partial T}{\partial z}, |
3202 |
> |
\end{equation} |
3203 |
> |
and shear viscosities ($\eta$), |
3204 |
> |
\begin{equation} |
3205 |
> |
j_z(p_x) = -\eta \frac{\partial \langle v_x \rangle}{\partial z}. |
3206 |
> |
\end{equation} |
3207 |
> |
Here, the quantities on the left hand side are the actual flux values |
3208 |
> |
(in the header of the {\tt .rnemd} file), while the slopes are |
3209 |
> |
obtained from linear fits to the gradients observed in the {\tt |
3210 |
> |
.rnemd} file. |
3211 |
|
|
3212 |
< |
And to apply a kinetic energy exchange between the two regions, the |
3213 |
< |
following should be satisfied as well: |
3214 |
< |
|
3215 |
< |
%Flux equations |
3216 |
< |
|
3217 |
< |
Mathematically, any points in the 3-dimensional space of the solution |
3218 |
< |
set would satisfy the equations. However, to avoid solving an |
3219 |
< |
ill-conditioned high-order polynomial in actual practice, another |
3220 |
< |
constraint, ${x_c=y_c}$, is applied, taking into consideration of its |
3221 |
< |
physical relevance. Therefore, a quartic equation is solved in actual |
3222 |
< |
practice to determine the sets of possible coefficients. To determine |
3223 |
< |
which set is actually used to perform the scaling, two criteria are |
3224 |
< |
mainly considered: 1. ${x,y,z\rightarrow 1}$ so that the perturbation |
3225 |
< |
could be as gentle as possible. 2. ${K^x, K^y, K^z}$ have minimal |
3226 |
< |
difference among each other, so that the anisotropy introduced by the |
3227 |
< |
algorithm can be offset to some extend. One set of scaling |
3228 |
< |
coefficients is chosen against these criteria, and the best one is |
3229 |
< |
used to perform the scaling for that particular step. However, if no |
3157 |
< |
solution found, the NIVS move is not performed in that step. |
3158 |
< |
|
3159 |
< |
Although the NIVS algorithm can also be applied to impose a |
3160 |
< |
directional momentum flux, thermal anisotropy was observed in |
3161 |
< |
relatively high flux simulations. %This is because... |
3162 |
< |
However, the gentleness and ability to apply a wide range of kinetic |
3163 |
< |
energy flux makes the method useful in thermal transport simulations, |
3164 |
< |
particularly for complex and heterogeneous systems including |
3165 |
< |
interfaces. %CITATION |
3166 |
< |
|
3167 |
< |
Velocity Shearing and Scaling (VSS): %CITATION |
3168 |
< |
Learning from NIVS that imposing directional momentum flux by velocity |
3169 |
< |
scaling could cause problem, we shift the approach to combine the move |
3170 |
< |
of velocity shearing and scaling: |
3171 |
< |
|
3172 |
< |
%VSS equations |
3173 |
< |
|
3174 |
< |
It turned out that this approach results in a set of simpler-to-solve |
3175 |
< |
equations for conservation and to satisfy momentum exchange: |
3176 |
< |
|
3177 |
< |
%conservation equations |
3178 |
< |
|
3179 |
< |
Furthermore, isotropic scaling is now possible, with the presence of |
3180 |
< |
velocity shearing quantities. Only a set of simple quadratic equations |
3181 |
< |
need to be solved, and the positive set of coefficients are chosen, in |
3182 |
< |
order to reach minimal perturbations. Similar to the NIVS method, no |
3183 |
< |
VSS is performed in a step given that no solution can be found. |
3184 |
< |
|
3185 |
< |
The VSS approach turned out to be versatile in both thermal and |
3186 |
< |
directional momentum transport simulations. It is found that the |
3187 |
< |
perturbation is minimal and undesired side-effects like thermal |
3188 |
< |
anisotropy can be avoided. Another nice feature of VSS is its ability |
3189 |
< |
to combine a thermal and a directional momentum flux. This feature has |
3190 |
< |
been utilized to map out the shear viscosity of SPC/E water in a wide |
3191 |
< |
range of temperature (90~K) just with one single simulation. Possible |
3192 |
< |
applications may also include the studies of thermal-momentum coupled |
3193 |
< |
transport phenomena. VSS also allows the directional momentum flux to |
3194 |
< |
have quite arbitary directions, which could benefit researches of |
3195 |
< |
anisotropic systems. |
3212 |
> |
More complicated simulations (including interfaces) require a bit more |
3213 |
> |
care. Here the second derivative may be required to compute the |
3214 |
> |
interfacial thermal conductance, |
3215 |
> |
\begin{align} |
3216 |
> |
G^\prime &= \left(\nabla\lambda \cdot \mathbf{\hat{n}}\right)_{z_0} \\ |
3217 |
> |
&= \frac{\partial}{\partial z}\left(-\frac{J_z}{ |
3218 |
> |
\left(\frac{\partial T}{\partial z}\right)}\right)_{z_0} \\ |
3219 |
> |
&= J_z\left(\frac{\partial^2 T}{\partial z^2}\right)_{z_0} \Big/ |
3220 |
> |
\left(\frac{\partial T}{\partial z}\right)_{z_0}^2. |
3221 |
> |
\label{derivativeG} |
3222 |
> |
\end{align} |
3223 |
> |
where $z_0$ is the location of the interface between two materials and |
3224 |
> |
$\mathbf{\hat{n}}$ is a unit vector normal to the interface. We |
3225 |
> |
suggest that users interested in interfacial conductance consult |
3226 |
> |
reference \citealp{kuang:AuThl} for other approaches to computing $G$. |
3227 |
> |
Users interested in {\it friction coefficients} at heterogeneous |
3228 |
> |
interfaces may also find reference \citealp{2012MolPh.110..691K} |
3229 |
> |
useful. |
3230 |
|
|
3231 |
< |
Table \ref{table:rnemd} summarizes the parameters used in RNEMD |
3198 |
< |
simulations. |
3231 |
> |
\newpage |
3232 |
|
|
3233 |
|
\begin{longtable}[c]{JKLM} |
3234 |
|
\caption{The following keywords must be enclosed inside a {\tt RNEMD\{\}} block} |
3258 |
|
{\tt outputBins} & int & number of $z$-bins in the output histogram & |
3259 |
|
default is 20 \\ |
3260 |
|
{\tt outputFields} & string & columns to print in the {\tt .rnemd} |
3261 |
< |
file where each column is separated by a pipe ($\mid$) symbol. & Allowed column names are: {\sc z, temperature, velocity, density}} \\ |
3261 |
> |
file where each column is separated by a pipe ($\mid$) symbol. & Allowed column names are: {\sc z, temperature, velocity, density} \\ |
3262 |
|
\label{table:rnemd} |
3263 |
|
\end{longtable} |
3264 |
|
|
3936 |
|
DMR-0079647. |
3937 |
|
|
3938 |
|
|
3939 |
< |
\bibliographystyle{jcc} |
3939 |
> |
\bibliographystyle{aip} |
3940 |
|
\bibliography{openmdDoc} |
3941 |
|
|
3942 |
|
\end{document} |