ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/group/trunk/ssdePaper/nptSSD.tex
Revision: 777
Committed: Fri Sep 19 19:29:24 2003 UTC (20 years, 9 months ago) by chrisfen
Content type: application/x-tex
File size: 48772 byte(s)
Log Message:
I must have changed something, but for the life of me I can't recall it...

File Contents

# User Rev Content
1 chrisfen 743 \documentclass[prb,aps,times,twocolumn,tabularx]{revtex4}
2     %\documentclass[prb,aps,times,tabularx,preprint]{revtex4}
3     \usepackage{amsmath}
4     \usepackage{graphicx}
5    
6     %\usepackage{endfloat}
7     %\usepackage{berkeley}
8     %\usepackage{epsf}
9     %\usepackage[ref]{overcite}
10     %\usepackage{setspace}
11     %\usepackage{tabularx}
12     %\usepackage{graphicx}
13     %\usepackage{curves}
14     %\usepackage{amsmath}
15     %\pagestyle{plain}
16     %\pagenumbering{arabic}
17     %\oddsidemargin 0.0cm \evensidemargin 0.0cm
18     %\topmargin -21pt \headsep 10pt
19     %\textheight 9.0in \textwidth 6.5in
20     %\brokenpenalty=10000
21     %\renewcommand{\baselinestretch}{1.2}
22     %\renewcommand\citemid{\ } % no comma in optional reference note
23    
24     \begin{document}
25    
26 chrisfen 759 \title{On the temperature dependent properties of the soft sticky dipole (SSD) and related single point water models}
27 chrisfen 743
28     \author{Christopher J. Fennell and J. Daniel Gezelter{\thefootnote}
29     \footnote[1]{Corresponding author. \ Electronic mail: gezelter@nd.edu}}
30    
31     \address{Department of Chemistry and Biochemistry\\ University of Notre Dame\\
32     Notre Dame, Indiana 46556}
33    
34     \date{\today}
35    
36     \begin{abstract}
37     NVE and NPT molecular dynamics simulations were performed in order to
38     investigate the density maximum and temperature dependent transport
39     for the SSD water model, both with and without the use of reaction
40     field. The constant pressure simulations of the melting of both $I_h$
41     and $I_c$ ice showed a density maximum near 260 K. In most cases, the
42     calculated densities were significantly lower than the densities
43     calculated in simulations of other water models. Analysis of particle
44     diffusion showed SSD to capture the transport properties of
45     experimental very well in both the normal and super-cooled liquid
46     regimes. In order to correct the density behavior, SSD was
47     reparameterized for use both with and without a long-range interaction
48     correction, SSD/RF and SSD/E respectively. In addition to correcting
49     the abnormally low densities, these new versions were show to maintain
50     or improve upon the transport and structural features of the original
51     water model.
52     \end{abstract}
53    
54     \maketitle
55    
56     %\narrowtext
57    
58    
59     %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
60     % BODY OF TEXT
61     %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
62    
63     \section{Introduction}
64    
65     One of the most important tasks in simulations of biochemical systems
66     is the proper depiction of water and water solvation. In fact, the
67     bulk of the calculations performed in solvated simulations are of
68     interactions with or between solvent molecules. Thus, the outcomes of
69     these types of simulations are highly dependent on the physical
70     properties of water, both as individual molecules and in
71     groups/bulk. Due to the fact that explicit solvent accounts for a
72     massive portion of the calculations, it necessary to simplify the
73     solvent to some extent in order to complete simulations in a
74     reasonable amount of time. In the case of simulating water in
75     bio-molecular studies, the balance between accurate properties and
76     computational efficiency is especially delicate, and it has resulted
77     in a variety of different water
78     models.\cite{Jorgensen83,Berendsen87,Jorgensen00} Many of these models
79     get specific properties correct or better than their predecessors, but
80     this is often at a cost of some other properties or of computer
81     time. As an example, compare TIP3P or TIP4P to TIP5P. TIP5P succeeds
82     in improving the structural and transport properties over its
83     predecessors, yet this comes at a greater than 50\% increase in
84     computational cost.\cite{Jorgensen01,Jorgensen00} One recently
85     developed model that succeeds in both retaining accuracy of system
86     properties and simplifying calculations to increase computational
87     efficiency is the Soft Sticky Dipole water model.\cite{Ichiye96}
88    
89     The Soft Sticky Dipole (SSD)\ water model was developed by Ichiye
90     \emph{et al.} as a modified form of the hard-sphere water model
91     proposed by Bratko, Blum, and Luzar.\cite{Bratko85,Bratko95} SSD
92     consists of a single point dipole with a Lennard-Jones core and a
93     sticky potential that directs the particles to assume the proper
94     hydrogen bond orientation in the first solvation shell. Thus, the
95     interaction between two SSD water molecules \emph{i} and \emph{j} is
96     given by the potential
97     \begin{equation}
98     u_{ij} = u_{ij}^{LJ} (r_{ij})\ + u_{ij}^{dp}
99     (\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j)\ +
100     u_{ij}^{sp}
101     (\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j),
102     \end{equation}
103     where the $\mathbf{r}_{ij}$ is the position vector between molecules
104     \emph{i} and \emph{j} with magnitude equal to the distance $r_ij$, and
105     $\boldsymbol{\Omega}_i$ and $\boldsymbol{\Omega}_j$ represent the
106     orientations of the respective molecules. The Lennard-Jones, dipole,
107     and sticky parts of the potential are giving by the following
108     equations,
109     \begin{equation}
110     u_{ij}^{LJ}(r_{ij}) = 4\epsilon \left[\left(\frac{\sigma}{r_{ij}}\right)^{12}-\left(\frac{\sigma}{r_{ij}}\right)^{6}\right],
111     \end{equation}
112     \begin{equation}
113     u_{ij}^{dp} = \frac{\boldsymbol{\mu}_i\cdot\boldsymbol{\mu}_j}{r_{ij}^3}-\frac{3(\boldsymbol{\mu}_i\cdot\mathbf{r}_{ij})(\boldsymbol{\mu}_j\cdot\mathbf{r}_{ij})}{r_{ij}^5}\ ,
114     \end{equation}
115     \begin{equation}
116     \begin{split}
117     u_{ij}^{sp}
118     (\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j)
119     &=
120     \frac{\nu_0}{2}[s(r_{ij})w(\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j)\\
121     & \quad \ +
122     s^\prime(r_{ij})w^\prime(\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j)]\ ,
123     \end{split}
124     \end{equation}
125     where $\boldsymbol{\mu}_i$ and $\boldsymbol{\mu}_j$ are the dipole
126     unit vectors of particles \emph{i} and \emph{j} with magnitude 2.35 D,
127     $\nu_0$ scales the strength of the overall sticky potential, $s$ and
128     $s^\prime$ are cubic switching functions. The $w$ and $w^\prime$
129     functions take the following forms,
130     \begin{equation}
131     w(\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j)=\sin\theta_{ij}\sin2\theta_{ij}\cos2\phi_{ij},
132     \end{equation}
133     \begin{equation}
134     w^\prime(\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j) = (\cos\theta_{ij}-0.6)^2(\cos\theta_{ij}+0.8)^2-w^0,
135     \end{equation}
136     where $w^0 = 0.07715$. The $w$ function is the tetrahedral attractive
137     term that promotes hydrogen bonding orientations within the first
138     solvation shell, and $w^\prime$ is a dipolar repulsion term that
139     repels unrealistic dipolar arrangements within the first solvation
140     shell. A more detailed description of the functional parts and
141     variables in this potential can be found in other
142     articles.\cite{Ichiye96,Ichiye99}
143    
144     Being that this is a one-site point dipole model, the actual force
145     calculations are simplified significantly. In the original Monte Carlo
146     simulations using this model, Ichiye \emph{et al.} reported a
147     calculation speed up of up to an order of magnitude over other
148     comparable models while maintaining the structural behavior of
149 chrisfen 777 water.\cite{Ichiye96} In the original molecular dynamics studies, it
150     was shown that SSD improves on the prediction of many of water's
151     dynamical properties over TIP3P and SPC/E.\cite{Ichiye99} This
152 chrisfen 743 attractive combination of speed and accurate depiction of solvent
153     properties makes SSD a model of interest for the simulation of large
154     scale biological systems, such as membrane phase behavior, a specific
155     interest within our group.
156    
157 chrisfen 757 One of the key limitations of this water model, however, is that it
158     has been parameterized for use with the Ewald Sum technique for the
159     handling of long-ranged interactions. When studying very large
160     systems, the Ewald summation and even particle-mesh Ewald become
161     computational burdens with their respective ideal $N^\frac{3}{2}$ and
162     $N\log N$ calculation scaling orders for $N$ particles.\cite{Darden99}
163 chrisfen 759 In applying this water model in these types of systems, it would be
164     useful to know its properties and behavior with the more
165     computationally efficient reaction field (RF) technique, and even with
166     a cutoff that lacks any form of long range correction. This study
167     addresses these issues by looking at the structural and transport
168     behavior of SSD over a variety of temperatures, with the purpose of
169     utilizing the RF correction technique. Towards the end, we suggest
170     alterations to the parameters that result in more water-like
171     behavior. It should be noted that in a recent publication, some the
172     original investigators of the SSD water model have put forth
173     adjustments to the original SSD water model to address abnormal
174     density behavior (also observed here), calling the corrected model
175     SSD1.\cite{Ichiye03} This study will consider this new model's
176     behavior as well, and hopefully improve upon its depiction of water
177     under conditions without the Ewald Sum.
178 chrisfen 757
179 chrisfen 743 \section{Methods}
180    
181     As stated previously, in this study the long-range dipole-dipole
182     interactions were accounted for using the reaction field method. The
183     magnitude of the reaction field acting on dipole \emph{i} is given by
184     \begin{equation}
185     \mathcal{E}_{i} = \frac{2(\varepsilon_{s} - 1)}{2\varepsilon_{s} + 1}
186     \frac{1}{r_{c}^{3}} \sum_{j\in{\mathcal{R}}} \boldsymbol{\mu}_{j} f(r_{ij})\ ,
187     \label{rfequation}
188     \end{equation}
189     where $\mathcal{R}$ is the cavity defined by the cutoff radius
190     ($r_{c}$), $\varepsilon_{s}$ is the dielectric constant imposed on the
191     system (80 in this case), $\boldsymbol{\mu}_{j}$ is the dipole moment
192     vector of particle \emph{j}, and $f(r_{ij})$ is a cubic switching
193     function.\cite{AllenTildesley} The reaction field contribution to the
194     total energy by particle \emph{i} is given by
195     $-\frac{1}{2}\boldsymbol{\mu}_{i}\cdot\mathcal{E}_{i}$ and the torque
196     on dipole \emph{i} by
197     $\boldsymbol{\mu}_{i}\times\mathcal{E}_{i}$.\cite{AllenTildesley} Use
198     of reaction field is known to alter the orientational dynamic
199     properties, such as the dielectric relaxation time, based on changes
200     in the length of the cutoff radius.\cite{Berendsen98} This variable
201     behavior makes reaction field a less attractive method than other
202     methods, like the Ewald summation; however, for the simulation of
203     large-scale system, the computational cost benefit of reaction field
204     is dramatic. To address some of the dynamical property alterations due
205     to the use of reaction field, simulations were also performed without
206     a surrounding dielectric and suggestions are proposed on how to make
207     SSD more compatible with a reaction field.
208 chrisfen 777
209 chrisfen 743 Simulations were performed in both the isobaric-isothermal and
210     microcanonical ensembles. The constant pressure simulations were
211     implemented using an integral thermostat and barostat as outlined by
212 chrisfen 777 Hoover.\cite{Hoover85,Hoover86} All particles were treated as
213     non-linear rigid bodies. Vibrational constraints are not necessary in
214     simulations of SSD, because there are no explicit hydrogen atoms, and
215     thus no molecular vibrational modes need to be considered.
216 chrisfen 743
217     Integration of the equations of motion was carried out using the
218     symplectic splitting method proposed by Dullweber \emph{et
219     al.}.\cite{Dullweber1997} The reason for this integrator selection
220     deals with poor energy conservation of rigid body systems using
221     quaternions. While quaternions work well for orientational motion in
222     alternate ensembles, the microcanonical ensemble has a constant energy
223 chrisfen 777 requirement that is quite sensitive to errors in the equations of
224     motion. The original implementation of this code utilized quaternions
225     for rotational motion propagation; however, a detailed investigation
226     showed that they resulted in a steady drift in the total energy,
227     something that has been observed by others.\cite{Laird97}
228 chrisfen 743
229     The key difference in the integration method proposed by Dullweber
230     \emph{et al.} is that the entire rotation matrix is propagated from
231     one time step to the next. In the past, this would not have been as
232     feasible a option, being that the rotation matrix for a single body is
233     nine elements long as opposed to 3 or 4 elements for Euler angles and
234     quaternions respectively. System memory has become much less of an
235     issue in recent times, and this has resulted in substantial benefits
236 chrisfen 759 in energy conservation. There is still the issue of 5 or 6 additional
237     elements for describing the orientation of each particle, which will
238     increase dump files substantially. Simply translating the rotation
239     matrix into its component Euler angles or quaternions for storage
240     purposes relieves this burden.
241 chrisfen 743
242     The symplectic splitting method allows for Verlet style integration of
243     both linear and angular motion of rigid bodies. In the integration
244     method, the orientational propagation involves a sequence of matrix
245     evaluations to update the rotation matrix.\cite{Dullweber1997} These
246     matrix rotations end up being more costly computationally than the
247 chrisfen 777 simpler arithmetic quaternion propagation. With the same time step, a
248     1000 SSD particle simulation shows an average 7\% increase in
249     computation time using the symplectic step method in place of
250     quaternions. This cost is more than justified when comparing the
251     energy conservation of the two methods as illustrated in figure
252     \ref{timestep}.
253 chrisfen 743
254     \begin{figure}
255     \includegraphics[width=61mm, angle=-90]{timeStep.epsi}
256     \caption{Energy conservation using quaternion based integration versus
257     the symplectic step method proposed by Dullweber \emph{et al.} with
258     increasing time step. For each time step, the dotted line is total
259     energy using the symplectic step integrator, and the solid line comes
260     from the quaternion integrator. The larger time step plots are shifted
261     up from the true energy baseline for clarity.}
262     \label{timestep}
263     \end{figure}
264    
265     In figure \ref{timestep}, the resulting energy drift at various time
266     steps for both the symplectic step and quaternion integration schemes
267     is compared. All of the 1000 SSD particle simulations started with the
268     same configuration, and the only difference was the method for
269     handling rotational motion. At time steps of 0.1 and 0.5 fs, both
270     methods for propagating particle rotation conserve energy fairly well,
271     with the quaternion method showing a slight energy drift over time in
272     the 0.5 fs time step simulation. At time steps of 1 and 2 fs, the
273     energy conservation benefits of the symplectic step method are clearly
274 chrisfen 759 demonstrated. Thus, while maintaining the same degree of energy
275     conservation, one can take considerably longer time steps, leading to
276     an overall reduction in computation time.
277 chrisfen 743
278     Energy drift in these SSD particle simulations was unnoticeable for
279     time steps up to three femtoseconds. A slight energy drift on the
280     order of 0.012 kcal/mol per nanosecond was observed at a time step of
281     four femtoseconds, and as expected, this drift increases dramatically
282     with increasing time step. To insure accuracy in the constant energy
283     simulations, time steps were set at 2 fs and kept at this value for
284     constant pressure simulations as well.
285    
286     Ice crystals in both the $I_h$ and $I_c$ lattices were generated as
287     starting points for all the simulations. The $I_h$ crystals were
288     formed by first arranging the center of masses of the SSD particles
289     into a ``hexagonal'' ice lattice of 1024 particles. Because of the
290     crystal structure of $I_h$ ice, the simulation box assumed a
291     rectangular shape with a edge length ratio of approximately
292     1.00$\times$1.06$\times$1.23. The particles were then allowed to
293     orient freely about fixed positions with angular momenta randomized at
294     400 K for varying times. The rotational temperature was then scaled
295     down in stages to slowly cool the crystals down to 25 K. The particles
296     were then allowed translate with fixed orientations at a constant
297     pressure of 1 atm for 50 ps at 25 K. Finally, all constraints were
298     removed and the ice crystals were allowed to equilibrate for 50 ps at
299     25 K and a constant pressure of 1 atm. This procedure resulted in
300     structurally stable $I_h$ ice crystals that obey the Bernal-Fowler
301     rules\cite{Bernal33,Rahman72}. This method was also utilized in the
302     making of diamond lattice $I_c$ ice crystals, with each cubic
303     simulation box consisting of either 512 or 1000 particles. Only
304     isotropic volume fluctuations were performed under constant pressure,
305     so the ratio of edge lengths remained constant throughout the
306     simulations.
307    
308     \section{Results and discussion}
309    
310     Melting studies were performed on the randomized ice crystals using
311     constant pressure and temperature dynamics. This involved an initial
312     randomization of velocities about the starting temperature of 25 K for
313     varying amounts of time. The systems were all equilibrated for 100 ps
314     prior to a 200 ps data collection run at each temperature setting,
315     ranging from 25 to 400 K, with a maximum degree increment of 25 K. For
316     regions of interest along this stepwise progression, the temperature
317     increment was decreased from 25 K to 10 and then 5 K. The above
318     equilibration and production times were sufficient in that the system
319     volume fluctuations dampened out in all but the very cold simulations
320 chrisfen 777 (below 225 K). In order to further improve statistics, an ensemble
321     average was accumulated from five separate simulation progressions,
322     each starting from a different ice crystal.
323 chrisfen 743
324     \subsection{Density Behavior}
325     In the initial average density versus temperature plot, the density
326     maximum clearly appears between 255 and 265 K. The calculated
327     densities within this range were nearly indistinguishable, as can be
328     seen in the zoom of this region of interest, shown in figure
329     \ref{dense1}. The greater certainty of the average value at 260 K makes
330     a good argument for the actual density maximum residing at this
331     midpoint value. Figure \ref{dense1} was constructed using ice $I_h$
332     crystals for the initial configuration; and though not pictured, the
333     simulations starting from ice $I_c$ crystal configurations showed
334     similar results, with a liquid-phase density maximum in this same
335     region (between 255 and 260 K). In addition, the $I_c$ crystals are
336     more fragile than the $I_h$ crystals, leading them to deform into a
337     dense glassy state at lower temperatures. This resulted in an overall
338     low temperature density maximum at 200 K, but they still retained a
339     common liquid state density maximum with the $I_h$ simulations.
340    
341     \begin{figure}
342     \includegraphics[width=65mm,angle=-90]{dense2.eps}
343     \caption{Density versus temperature for TIP4P\cite{Jorgensen98b},
344     TIP3P\cite{Jorgensen98b}, SPC/E\cite{Clancy94}, SSD without Reaction
345     Field, SSD, and Experiment\cite{CRC80}. }
346     \label{dense2}
347     \end{figure}
348    
349     The density maximum for SSD actually compares quite favorably to other
350     simple water models. Figure \ref{dense2} shows a plot of these
351     findings with the density progression of several other models and
352     experiment obtained from other
353     sources.\cite{Jorgensen98b,Clancy94,CRC80} Of the listed simple water
354     models, SSD has results closest to the experimentally observed water
355     density maximum. Of the listed water models, TIP4P has a density
356     maximum behavior most like that seen in SSD. Though not shown, it is
357     useful to note that TIP5P has a water density maximum nearly identical
358     to experiment.
359    
360     Possibly of more importance is the density scaling of SSD relative to
361     other common models at any given temperature (Fig. \ref{dense2}). Note
362     that the SSD model assumes a lower density than any of the other
363     listed models at the same pressure, behavior which is especially
364     apparent at temperatures greater than 300 K. Lower than expected
365     densities have been observed for other systems with the use of a
366     reaction field for long-range electrostatic interactions, so the most
367     likely reason for these significantly lower densities in these
368     simulations is the presence of the reaction field.\cite{Berendsen98}
369     In order to test the effect of the reaction field on the density of
370     the systems, the simulations were repeated for the temperature region
371     of interest without a reaction field present. The results of these
372     simulations are also displayed in figure \ref{dense2}. Without
373     reaction field, these densities increase considerably to more
374     experimentally reasonable values, especially around the freezing point
375     of liquid water. The shape of the curve is similar to the curve
376     produced from SSD simulations using reaction field, specifically the
377     rapidly decreasing densities at higher temperatures; however, a slight
378     shift in the density maximum location, down to 245 K, is
379     observed. This is probably a more accurate comparison to the other
380     listed water models in that no long range corrections were applied in
381     those simulations.\cite{Clancy94,Jorgensen98b}
382    
383     It has been observed that densities are dependent on the cutoff radius
384     used for a variety of water models in simulations both with and
385     without the use of reaction field.\cite{Berendsen98} In order to
386     address the possible affect of cutoff radius, simulations were
387     performed with a dipolar cutoff radius of 12.0 \AA\ to compliment the
388     previous SSD simulations, all performed with a cutoff of 9.0 \AA. All
389     the resulting densities overlapped within error and showed no
390     significant trend in lower or higher densities as a function of cutoff
391     radius, both for simulations with and without reaction field. These
392     results indicate that there is no major benefit in choosing a longer
393     cutoff radius in simulations using SSD. This is comforting in that the
394     use of a longer cutoff radius results in a near doubling of the time
395     required to compute a single trajectory.
396    
397     \subsection{Transport Behavior}
398     Of importance in these types of studies are the transport properties
399     of the particles and how they change when altering the environmental
400     conditions. In order to probe transport, constant energy simulations
401     were performed about the average density uncovered by the constant
402     pressure simulations. Simulations started with randomized velocities
403     and underwent 50 ps of temperature scaling and 50 ps of constant
404     energy equilibration before obtaining a 200 ps trajectory. Diffusion
405     constants were calculated via root-mean square deviation analysis. The
406     averaged results from 5 sets of these NVE simulations is displayed in
407     figure \ref{diffuse}, alongside experimental, SPC/E, and TIP5P
408     results.\cite{Gillen72,Mills73,Clancy94,Jorgensen01}
409    
410     \begin{figure}
411     \includegraphics[width=65mm, angle=-90]{betterDiffuse.epsi}
412     \caption{Average diffusion coefficient over increasing temperature for
413     SSD, SPC/E\cite{Clancy94}, TIP5P\cite{Jorgensen01}, and Experimental
414     data from Gillen \emph{et al.}\cite{Gillen72}, and from
415     Mills\cite{Mills73}.}
416     \label{diffuse}
417     \end{figure}
418    
419     The observed values for the diffusion constant point out one of the
420     strengths of the SSD model. Of the three experimental models shown,
421     the SSD model has the most accurate depiction of the diffusion trend
422     seen in experiment in both the supercooled and normal regimes. SPC/E
423     does a respectable job by getting similar values as SSD and experiment
424     around 290 K; however, it deviates at both higher and lower
425     temperatures, failing to predict the experimental trend. TIP5P and SSD
426     both start off low at the colder temperatures and tend to diffuse too
427     rapidly at the higher temperatures. This type of trend at the higher
428     temperatures is not surprising in that the densities of both TIP5P and
429     SSD are lower than experimental water at temperatures higher than room
430     temperature. When calculating the diffusion coefficients for SSD at
431     experimental densities, the resulting values fall more in line with
432     experiment at these temperatures, albeit not at standard
433     pressure. Results under these conditions can be found later in this
434     paper.
435    
436     \subsection{Structural Changes and Characterization}
437     By starting the simulations from the crystalline state, the melting
438     transition and the ice structure can be studied along with the liquid
439     phase behavior beyond the melting point. To locate the melting
440     transition, the constant pressure heat capacity (C$_\text{p}$) was
441     monitored in each of the simulations. In the melting simulations of
442     the 1024 particle ice $I_h$ simulations, a large spike in C$_\text{p}$
443     occurs at 245 K, indicating a first order phase transition for the
444     melting of these ice crystals. When the reaction field is turned off,
445     the melting transition occurs at 235 K. These melting transitions are
446     considerably lower than the experimental value, but this is not
447     surprising in that SSD is a simple rigid body model with a fixed
448     dipole.
449    
450     \begin{figure}
451     \includegraphics[width=85mm]{fullContours.eps}
452     \caption{Contour plots of 2D angular g($r$)'s for 512 SSD systems at
453     100 K (A \& B) and 300 K (C \& D). Contour colors are inverted for
454     clarity: dark areas signify peaks while light areas signify
455     depressions. White areas have g(\emph{r}) values below 0.5 and black
456     areas have values above 1.5.}
457     \label{contour}
458     \end{figure}
459    
460     Additional analyses for understanding the melting phase-transition
461     process were performed via two-dimensional structure and dipole angle
462     correlations. Expressions for the correlations are as follows:
463    
464     \begin{figure}
465     \includegraphics[width=45mm]{corrDiag.eps}
466     \caption{Two dimensional illustration of the angles involved in the
467     correlations observed in figure \ref{contour}.}
468     \label{corrAngle}
469     \end{figure}
470    
471     \begin{multline}
472     g_{\text{AB}}(r,\cos\theta) = \\
473     \frac{V}{N_\text{A}N_\text{B}}\langle\sum_{i\in\text{A}}\sum_{j\in\text{B}}\delta(\cos\theta-\cos\theta_{ij})\delta(r-\left|\mathbf{r}_{ij}\right|)\rangle\ ,
474     \end{multline}
475     \begin{multline}
476     g_{\text{AB}}(r,\cos\omega) = \\
477     \frac{V}{N_\text{A}N_\text{B}}\langle\sum_{i\in\text{A}}\sum_{j\in\text{B}}\delta(\cos\omega-\cos\omega_{ij})\delta(r-\left|\mathbf{r}_{ij}\right|)\rangle\ ,
478     \end{multline}
479     where $\theta$ and $\omega$ refer to the angles shown in the above
480     illustration. By binning over both distance and the cosine of the
481     desired angle between the two dipoles, the g(\emph{r}) can be
482     dissected to determine the common dipole arrangements that constitute
483     the peaks and troughs. Frames A and B of figure \ref{contour} show a
484     relatively crystalline state of an ice $I_c$ simulation. The first
485     peak of the g(\emph{r}) primarily consists of the preferred hydrogen
486     bonding arrangements as dictated by the tetrahedral sticky potential,
487     one peak for the donating and the other for the accepting hydrogen
488     bonds. Due to the high degree of crystallinity of the sample, the
489     second and third solvation shells show a repeated peak arrangement
490     which decays at distances around the fourth solvation shell, near the
491     imposed cutoff for the Lennard-Jones and dipole-dipole interactions.
492     In the higher temperature simulation shown in frames C and D, the
493     repeated peak features are significantly blurred. The first solvation
494     shell still shows the strong effect of the sticky-potential, although
495     it covers a larger area, extending to include a fraction of aligned
496     dipole peaks within the first solvation shell. The latter peaks lose
497     definition as thermal motion and the competing dipole force overcomes
498     the sticky potential's tight tetrahedral structuring of the fluid.
499    
500     This complex interplay between dipole and sticky interactions was
501     remarked upon as a possible reason for the split second peak in the
502     oxygen-oxygen g(\emph{r}).\cite{Ichiye96} At low temperatures, the
503     second solvation shell peak appears to have two distinct parts that
504     blend together to form one observable peak. At higher temperatures,
505     this split character alters to show the leading 4 \AA\ peak dominated
506     by equatorial anti-parallel dipole orientations, and there is tightly
507     bunched group of axially arranged dipoles that most likely consist of
508     the smaller fraction aligned dipole pairs. The trailing part of the
509     split peak at 5 \AA\ is dominated by aligned dipoles that range
510     primarily within the axial to the chief hydrogen bond arrangements
511     similar to those seen in the first solvation shell. This evidence
512     indicates that the dipole pair interaction begins to dominate outside
513     of the range of the dipolar repulsion term, with the primary
514     energetically favorable dipole arrangements populating the region
515     immediately outside of it's range (around 4 \AA), and arrangements
516     that seek to ideally satisfy both the sticky and dipole forces locate
517     themselves just beyond this region (around 5 \AA).
518    
519     From these findings, the split second peak is primarily the product of
520     the dipolar repulsion term of the sticky potential. In fact, the
521     leading of the two peaks can be pushed out and merged with the outer
522     split peak just by extending the switching function cutoff
523     ($s^\prime(r_{ij})$) from its normal 4.0 \AA\ to values of 4.5 or even
524     5 \AA. This type of correction is not recommended for improving the
525     liquid structure, because the second solvation shell will still be
526     shifted too far out. In addition, this would have an even more
527     detrimental effect on the system densities, leading to a liquid with a
528     more open structure and a density considerably lower than the normal
529     SSD behavior shown previously. A better correction would be to include
530     the quadrupole-quadrupole interactions for the water particles outside
531     of the first solvation shell, but this reduces the simplicity and
532     speed advantage of SSD, so it is not the most desirable path to take.
533    
534     \subsection{Adjusted Potentials: SSD/E and SSD/RF}
535     The propensity of SSD to adopt lower than expected densities under
536     varying conditions is troubling, especially at higher temperatures. In
537     order to correct this behavior, it's necessary to adjust the force
538     field parameters for the primary intermolecular interactions. In
539     undergoing a reparameterization, it is important not to focus on just
540     one property and neglect the other important properties. In this case,
541     it would be ideal to correct the densities while maintaining the
542     accurate transport properties.
543    
544     The possible parameters for tuning include the $\sigma$ and $\epsilon$
545     Lennard-Jones parameters, the dipole strength ($\mu$), and the sticky
546     attractive and dipole repulsive terms with their respective
547     cutoffs. To alter the attractive and repulsive terms of the sticky
548     potential independently, it is necessary to separate the terms as
549     follows:
550     \begin{equation}
551     \begin{split}
552     u_{ij}^{sp}
553     (\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j) &=
554     \frac{\nu_0}{2}[s(r_{ij})w(\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j)]\\
555     & \quad \ + \frac{\nu_0^\prime}{2}
556     [s^\prime(r_{ij})w^\prime(\mathbf{r}_{ij},\boldsymbol{\Omega}_i,\boldsymbol{\Omega}_j)],
557     \end{split}
558     \end{equation}
559    
560     where $\nu_0$ scales the strength of the tetrahedral attraction and
561     $\nu_0^\prime$ acts in an identical fashion on the dipole repulsion
562     term. For purposes of the reparameterization, the separation was
563     performed, but the final parameters were adjusted so that it is
564     unnecessary to separate the terms when implementing the adjusted water
565     potentials. The results of the reparameterizations are shown in table
566     \ref{params}. Note that both the tetrahedral attractive and dipolar
567     repulsive don't share the same lower cutoff ($r_l$) in the newly
568     parameterized potentials - soft sticky dipole enhanced (SSD/E) and
569     soft sticky dipole reaction field (SSD/RF).
570    
571     \begin{table}
572     \caption{Parameters for the original and adjusted models}
573     \begin{tabular}{ l c c c }
574     \hline \\[-3mm]
575     \ Parameters & \ \ \ SSD$^\dagger$\ \ \ \ & \ SSD/E\ \ & \ SSD/RF\ \ \\
576     \hline \\[-3mm]
577     \ \ \ $\sigma$ (\AA) & 3.051 & 3.035 & 3.019\\
578     \ \ \ $\epsilon$ (kcal/mol)\ \ & 0.152 & 0.152 & 0.152\\
579     \ \ \ $\mu$ (D) & 2.35 & 2.418 & 2.480\\
580     \ \ \ $\nu_0$ (kcal/mol)\ \ & 3.7284 & 3.90 & 3.90\\
581     \ \ \ $r_l$ (\AA) & 2.75 & 2.40 & 2.40\\
582     \ \ \ $r_u$ (\AA) & 3.35 & 3.80 & 3.80\\
583     \ \ \ $\nu_0^\prime$ (kcal/mol)\ \ & 3.7284 & 3.90 & 3.90\\
584     \ \ \ $r_l^\prime$ (\AA) & 2.75 & 2.75 & 2.75\\
585     \ \ \ $r_u^\prime$ (\AA) & 4.00 & 3.35 & 3.35\\
586     \\[-2mm]$^\dagger$ ref. \onlinecite{Ichiye96}
587     \end{tabular}
588     \label{params}
589     \end{table}
590    
591     \begin{figure}
592     \includegraphics[width=85mm]{gofrCompare.epsi}
593     \caption{Plots comparing experiment\cite{Head-Gordon00_1} with SSD/E
594     and SSD without reaction field (top), as well as SSD/RF and SSD with
595     reaction field turned on (bottom). The insets show the respective
596     first peaks in detail. Solid Line - experiment, dashed line - SSD/E
597     and SSD/RF, and dotted line - SSD (with and without reaction field).}
598     \label{grcompare}
599     \end{figure}
600    
601     \begin{figure}
602     \includegraphics[width=85mm]{dualsticky.ps}
603     \caption{Isosurfaces of the sticky potential for SSD (left) and SSD/E \&
604     SSD/RF (right). Light areas correspond to the tetrahedral attractive
605     part, and the darker areas correspond to the dipolar repulsive part.}
606     \label{isosurface}
607     \end{figure}
608    
609     In the paper detailing the development of SSD, Liu and Ichiye placed
610     particular emphasis on an accurate description of the first solvation
611     shell. This resulted in a somewhat tall and sharp first peak that
612     integrated to give similar coordination numbers to the experimental
613     data obtained by Soper and Phillips.\cite{Ichiye96,Soper86} New
614     experimental x-ray scattering data from the Head-Gordon lab indicates
615     a slightly lower and shifted first peak in the g$_\mathrm{OO}(r)$, so
616     adjustments to SSD were made while taking into consideration the new
617     experimental findings.\cite{Head-Gordon00_1} Figure \ref{grcompare}
618     shows the relocation of the first peak of the oxygen-oxygen
619     g(\emph{r}) by comparing the original SSD (with and without reaction
620     field), SSD-E, and SSD-RF to the new experimental results. Both the
621     modified water models have shorter peaks that are brought in more
622     closely to the experimental peak (as seen in the insets of figure
623     \ref{grcompare}). This structural alteration was accomplished by a
624     reduction in the Lennard-Jones $\sigma$ variable as well as adjustment
625     of the sticky potential strength and cutoffs. The cutoffs for the
626     tetrahedral attractive and dipolar repulsive terms were nearly swapped
627     with each other. Isosurfaces of the original and modified sticky
628     potentials are shown in figure \cite{isosurface}. In these
629     isosurfaces, it is easy to see how altering the cutoffs changes the
630     repulsive and attractive character of the particles. With a reduced
631     repulsive surface (the darker region), the particles can move closer
632     to one another, increasing the density for the overall system. This
633     change in interaction cutoff also results in a more gradual
634     orientational motion by allowing the particles to maintain preferred
635     dipolar arrangements before they begin to feel the pull of the
636     tetrahedral restructuring. Upon moving closer together, the dipolar
637     repulsion term becomes active and excludes the unphysical
638     arrangements. This compares with the original SSD's excluding dipolar
639     before the particles feel the pull of the ``hydrogen bonds''. Aside
640     from improving the shape of the first peak in the g(\emph{r}), this
641     improves the densities considerably by allowing the persistence of
642     full dipolar character below the previous 4.0 \AA\ cutoff.
643    
644     While adjusting the location and shape of the first peak of
645     g(\emph{r}) improves the densities to some degree, these changes alone
646     are insufficient to bring the system densities up to the values
647     observed experimentally. To finish bringing up the densities, the
648     dipole moments were increased in both the adjusted models. Being a
649     dipole based model, the structure and transport are very sensitive to
650     changes in the dipole moment. The original SSD simply used the dipole
651     moment calculated from the TIP3P water model, which at 2.35 D is
652     significantly greater than the experimental gas phase value of 1.84
653     D. The larger dipole moment is a more realistic value and improve the
654     dielectric properties of the fluid. Both theoretical and experimental
655     measurements indicate a liquid phase dipole moment ranging from 2.4 D
656     to values as high as 3.11 D, so there is quite a range available for
657     adjusting the dipole
658     moment.\cite{Sprik91,Kusalik02,Badyal00,Barriol64} The increasing of
659     the dipole moments to 2.418 and 2.48 D for SSD/E and SSD/RF
660     respectively is moderate in the range of the experimental values;
661     however, it leads to significant changes in the density and transport
662     of the water models.
663    
664     In order to demonstrate the benefits of this reparameterization, a
665     series of NPT and NVE simulations were performed to probe the density
666     and transport properties of the adapted models and compare the results
667     to the original SSD model. This comparison involved full NPT melting
668     sequences for both SSD/E and SSD/RF, as well as NVE transport
669     calculations at both self-consistent and experimental
670     densities. Again, the results come from five separate simulations of
671     1024 particle systems, and the melting sequences were started from
672     different ice $I_h$ crystals constructed as stated previously. Like
673     before, all of the NPT simulations were equilibrated for 100 ps before
674     a 200 ps data collection run, and they used the previous temperature's
675     final configuration as a starting point. All of the NVE simulations
676     had the same thermalization, equilibration, and data collection times
677     stated earlier in this paper.
678    
679     \begin{figure}
680     \includegraphics[width=85mm]{ssdecompare.epsi}
681     \caption{Comparison of densities calculated with SSD/E to SSD without a
682     reaction field, TIP4P\cite{Jorgensen98b}, TIP3P\cite{Jorgensen98b},
683     SPC/E\cite{Clancy94}, and Experiment\cite{CRC80}. The upper plot
684     includes error bars, and the calculated results from the other
685     references were removed for clarity.}
686     \label{ssdedense}
687     \end{figure}
688    
689     Figure \ref{ssdedense} shows the density profile for the SSD/E water
690     model in comparison to the original SSD without a reaction field,
691     experiment, and the other common water models considered
692     previously. The calculated densities have increased significantly over
693     the original SSD model and match the experimental value just below 298
694     K. At 298 K, the density of SSD/E is 0.995$\pm$0.001 g/cm$^3$, which
695     compares well with the experimental value of 0.997 g/cm$^3$ and is
696     considerably better than the SSD value of 0.967$\pm$0.003
697     g/cm$^3$. The increased dipole moment in SSD/E has helped to flatten
698     out the curve at higher temperatures, only the improvement is marginal
699     at best. This steep drop in densities is due to the dipolar rather
700     than charge based interactions which decay more rapidly at longer
701     distances.
702    
703     By monitoring C$\text{p}$ throughout these simulations, the melting
704     transition for SSD/E was observed at 230 K, about 5 degrees lower than
705     SSD. The resulting density maximum is located at 240 K, again about 5
706     degrees lower than the SSD value of 245 K. Though there is a decrease
707     in both of these values, the corrected densities near room temperature
708     justify the modifications taken.
709    
710     \begin{figure}
711     \includegraphics[width=85mm]{ssdrfcompare.epsi}
712     \caption{Comparison of densities calculated with SSD/RF to SSD with a
713     reaction field, TIP4P\cite{Jorgensen98b}, TIP3P\cite{Jorgensen98b},
714     SPC/E\cite{Clancy94}, and Experiment\cite{CRC80}. The upper plot
715     includes error bars, and the calculated results from the other
716     references were removed for clarity.}
717     \label{ssdrfdense}
718     \end{figure}
719    
720     Figure \ref{ssdrfdense} shows a density comparison between SSD/RF and
721     SSD with an active reaction field. Like in the simulations of SSD/E,
722     the densities show a dramatic increase over normal SSD. At 298 K,
723     SSD/RF has a density of 0.997$\pm$0.001 g/cm$^3$, right in line with
724     experiment and considerably better than the SSD value of
725     0.941$\pm$0.001 g/cm$^3$. The melting point is observed at 240 K,
726     which is 5 degrees lower than SSD with a reaction field, and the
727     density maximum at 255 K, again 5 degrees lower than SSD. The density
728     at higher temperature still drops off more rapidly than the charge
729     based models but is in better agreement than SSD/E.
730    
731     The reparameterization of the SSD water model, both for use with and
732     without an applied long-range correction, brought the densities up to
733     what is expected for simulating liquid water. In addition to improving
734     the densities, it is important that particle transport be maintained
735     or improved. Figure \ref{ssdediffuse} compares the temperature
736     dependence of the diffusion constant of SSD/E to SSD without an active
737     reaction field, both at the densities calculated at 1 atm and at the
738     experimentally calculated densities for super-cooled and liquid
739     water. In the upper plot, the diffusion constant for SSD/E is
740     consistently a little faster than experiment, while SSD starts off
741     slower than experiment and crosses to merge with SSD/E at high
742     temperatures. Both models follow the experimental trend well, but
743     diffuse too rapidly at higher temperatures. This abnormally fast
744     diffusion is caused by the decreased system density. Since the
745     densities of SSD/E don't deviate as much from experiment as those of
746     SSD, it follows the experimental trend more closely. This observation
747     is backed up by looking at the lower plot. The diffusion constants for
748     SSD/E track with the experimental values while SSD deviates on the low
749     side of the trend with increasing temperature. This is again a product
750     of SSD/E having densities closer to experiment, and not deviating to
751     lower densities with increasing temperature as rapidly.
752    
753     \begin{figure}
754     \includegraphics[width=85mm]{ssdediffuse.epsi}
755     \caption{Plots of the diffusion constants calculated from SSD/E and SSD,
756     both without a reaction field along with experimental results from
757     Gillen \emph{et al.}\cite{Gillen72} and Mills\cite{Mills73}. The
758     upper plot is at densities calculated from the NPT simulations at a
759     pressure of 1 atm, while the lower plot is at the experimentally
760     calculated densities.}
761     \label{ssdediffuse}
762     \end{figure}
763    
764     \begin{figure}
765     \includegraphics[width=85mm]{ssdrfdiffuse.epsi}
766     \caption{Plots of the diffusion constants calculated from SSD/RF and SSD,
767     both with an active reaction field along with experimental results
768     from Gillen \emph{et al.}\cite{Gillen72} and Mills\cite{Mills73}. The
769     upper plot is at densities calculated from the NPT simulations at a
770     pressure of 1 atm, while the lower plot is at the experimentally
771     calculated densities.}
772     \label{ssdrfdiffuse}
773     \end{figure}
774    
775     In figure \ref{ssdrfdiffuse}, the diffusion constants for SSD/RF are
776     compared with SSD with an active reaction field. In the upper plot,
777     SSD/RF tracks with the experimental results incredibly well, identical
778     within error throughout the temperature range and only showing a
779     slight increasing trend at higher temperatures. SSD also tracks
780     experiment well, only it tends to diffuse a little more slowly at low
781     temperatures and deviates to diffuse too rapidly at high
782     temperatures. As was stated in the SSD/E comparisons, this deviation
783     away from the ideal trend is due to a rapid decrease in density at
784     higher temperatures. SSD/RF doesn't suffer from this problem as much
785     as SSD, because the calculated densities are more true to
786     experiment. This is again emphasized in the lower plot, where SSD/RF
787     tracks the experimental diffusion exactly while SSD's diffusion
788     constants are slightly too low due to its need for a lower density at
789     the specified temperature.
790    
791     \subsection{Additional Observations}
792    
793     While performing the melting sequences of SSD/E, some interesting
794     observations were made. After melting at 230 K, two of the systems
795     underwent crystallization events near 245 K. As the heating process
796     continued, the two systems remained crystalline until finally melting
797     between 320 and 330 K. These simulations were excluded from the data
798     set shown in figure \ref{ssdedense} and replaced with two additional
799     melting sequences that did not undergo this anomalous phase
800     transition, while this crystallization event was investigated
801     separately.
802    
803     \begin{figure}
804     \includegraphics[width=85mm]{povIce.ps}
805     \caption{Crystal structure of an ice 0 lattice shown from the (001) face.}
806     \label{weirdice}
807     \end{figure}
808    
809     The final configurations of these two melting sequences shows an
810     expanded zeolite-like crystal structure that does not correspond to
811     any known form of ice. For convenience and to help distinguish it from
812     the experimentally observed forms of ice, this crystal structure will
813     henceforth be referred to as ice-zero (ice 0). The crystallinity was
814     extensive enough than a near ideal crystal structure could be
815     obtained. Figure \ref{weirdice} shows the repeating crystal structure
816     of a typical crystal at 5 K. The unit cell contains eight molecules,
817     and figure \ref{unitcell} shows a unit cell built from the water
818     particle center of masses that can be used to construct a repeating
819     lattice, similar to figure \ref{weirdice}. Each molecule is hydrogen
820     bonded to four other water molecules; however, the hydrogen bonds are
821     flexed rather than perfectly straight. This results in a skewed
822     tetrahedral geometry about the central molecule. Looking back at
823     figure \ref{isosurface}, it is easy to see how these flexed hydrogen
824     bonds are allowed in that the attractive regions are conical in shape,
825     with the greatest attraction in the central region. Though not ideal,
826     these flexed hydrogen bonds are favorable enough to stabilize an
827     entire crystal generated around them. In fact, the imperfect ice 0
828     crystals were so stable that they melted at greater than room
829     temperature.
830    
831     \begin{figure}
832     \includegraphics[width=65mm]{ice0cell.eps}
833     \caption{Simple unit cell for constructing ice 0. In this cell, $c$ is
834     equal to $0.4714\times a$, and a typical value for $a$ is 8.25 \AA.}
835     \label{unitcell}
836     \end{figure}
837    
838     The initial simulations indicated that ice 0 is the preferred ice
839     structure for at least SSD/E. To verify this, a comparison was made
840     between near ideal crystals of ice $I_h$, ice $I_c$, and ice 0 at
841     constant pressure with SSD/E, SSD/RF, and SSD. Near ideal versions of
842     the three types of crystals were cooled to ~1 K, and the potential
843     energies of each were compared using all three water models. With
844     every water model, ice 0 turned out to have the lowest potential
845     energy: 5\% lower than $I_h$ with SSD, 6.5\% lower with SSD/E, and
846     7.5\% lower with SSD/RF. In all three of these water models, ice $I_c$
847     was observed to be ~2\% less stable than ice $I_h$. In addition to
848     having the lowest potential energy, ice 0 was the most expanded of the
849     three ice crystals, ~5\% less dense than ice $I_h$ with all of the
850     water models. In all three water models, ice $I_c$ was observed to be
851     ~2\% more dense than ice $I_h$.
852    
853     In addition to the low temperature comparisons, melting sequences were
854     performed with ice 0 as the initial configuration using SSD/E, SSD/RF,
855     and SSD both with and without a reaction field. The melting
856     transitions for both SSD/E and SSD without a reaction field occurred
857     at temperature in excess of 375 K. SSD/RF and SSD with a reaction
858     field had more reasonable melting transitions, down near 325 K. These
859     melting point observations emphasize how preferred this crystal
860     structure is over the most common types of ice when using these single
861     point water models.
862    
863     Recognizing that the above tests show ice 0 to be both the most stable
864     and lowest density crystal structure for these single point water
865     models, it is interesting to speculate on the favorability of this
866     crystal structure with the different charge based models. As a quick
867     test, these 3 crystal types were converted from SSD type particles to
868     TIP3P waters and read into CHARMM.\cite{Karplus83} Identical energy
869     minimizations were performed on all of these crystals to compare the
870     system energies. Again, ice 0 was observed to have the lowest total
871     system energy. The total energy of ice 0 was ~2\% lower than ice
872     $I_h$, which was in turn ~3\% lower than ice $I_c$. From these initial
873     results, we would not be surprised if results from the other common
874     water models show ice 0 to be the lowest energy crystal structure. A
875     continuation on work studing ice 0 with multipoint water models will
876     be published in a coming article.
877    
878     \section{Conclusions}
879     The density maximum and temperature dependent transport for the SSD
880     water model, both with and without the use of reaction field, were
881     studied via a series of NPT and NVE simulations. The constant pressure
882     simulations of the melting of both $I_h$ and $I_c$ ice showed a
883     density maximum near 260 K. In most cases, the calculated densities
884     were significantly lower than the densities calculated in simulations
885     of other water models. Analysis of particle diffusion showed SSD to
886     capture the transport properties of experimental very well in both the
887     normal and super-cooled liquid regimes. In order to correct the
888     density behavior, SSD was reparameterized for use both with and
889     without a long-range interaction correction, SSD/RF and SSD/E
890     respectively. In addition to correcting the abnormally low densities,
891     these new versions were show to maintain or improve upon the transport
892     and structural features of the original water model, all while
893     maintaining the fast performance of the SSD water model. This work
894     shows these simple water models, and in particular SSD/E and SSD/RF,
895     to be excellent choices to represent explicit water in future
896     simulations of biochemical systems.
897    
898     \section{Acknowledgments}
899 chrisfen 777 Support for this project was provided by the National Science
900     Foundation under grant CHE-0134881. Computation time was provided by
901     the Notre Dame Bunch-of-Boxes (B.o.B) computer cluster under NSF grant
902     DMR 00 79647.
903 chrisfen 743
904     \bibliographystyle{jcp}
905    
906     \bibliography{nptSSD}
907    
908     %\pagebreak
909    
910     \end{document}