6 |
|
Closely related to Classical Mechanics, Molecular Dynamics |
7 |
|
simulations are carried out by integrating the equations of motion |
8 |
|
for a given system of particles. There are three fundamental ideas |
9 |
< |
behind classical mechanics. Firstly, One can determine the state of |
9 |
> |
behind classical mechanics. Firstly, one can determine the state of |
10 |
|
a mechanical system at any time of interest; Secondly, all the |
11 |
|
mechanical properties of the system at that time can be determined |
12 |
|
by combining the knowledge of the properties of the system with the |
17 |
|
\subsection{\label{introSection:newtonian}Newtonian Mechanics} |
18 |
|
The discovery of Newton's three laws of mechanics which govern the |
19 |
|
motion of particles is the foundation of the classical mechanics. |
20 |
< |
Newton¡¯s first law defines a class of inertial frames. Inertial |
20 |
> |
Newton's first law defines a class of inertial frames. Inertial |
21 |
|
frames are reference frames where a particle not interacting with |
22 |
|
other bodies will move with constant speed in the same direction. |
23 |
< |
With respect to inertial frames Newton¡¯s second law has the form |
23 |
> |
With respect to inertial frames, Newton's second law has the form |
24 |
|
\begin{equation} |
25 |
< |
F = \frac {dp}{dt} = \frac {mv}{dt} |
25 |
> |
F = \frac {dp}{dt} = \frac {mdv}{dt} |
26 |
|
\label{introEquation:newtonSecondLaw} |
27 |
|
\end{equation} |
28 |
|
A point mass interacting with other bodies moves with the |
29 |
|
acceleration along the direction of the force acting on it. Let |
30 |
< |
$F_ij$ be the force that particle $i$ exerts on particle $j$, and |
31 |
< |
$F_ji$ be the force that particle $j$ exerts on particle $i$. |
32 |
< |
Newton¡¯s third law states that |
30 |
> |
$F_{ij}$ be the force that particle $i$ exerts on particle $j$, and |
31 |
> |
$F_{ji}$ be the force that particle $j$ exerts on particle $i$. |
32 |
> |
Newton's third law states that |
33 |
|
\begin{equation} |
34 |
< |
F_ij = -F_ji |
34 |
> |
F_{ij} = -F_{ji} |
35 |
|
\label{introEquation:newtonThirdLaw} |
36 |
|
\end{equation} |
37 |
|
|
46 |
|
\end{equation} |
47 |
|
The torque $\tau$ with respect to the same origin is defined to be |
48 |
|
\begin{equation} |
49 |
< |
N \equiv r \times F \label{introEquation:torqueDefinition} |
49 |
> |
\tau \equiv r \times F \label{introEquation:torqueDefinition} |
50 |
|
\end{equation} |
51 |
|
Differentiating Eq.~\ref{introEquation:angularMomentumDefinition}, |
52 |
|
\[ |
59 |
|
\] |
60 |
|
thus, |
61 |
|
\begin{equation} |
62 |
< |
\dot L = r \times \dot p = N |
62 |
> |
\dot L = r \times \dot p = \tau |
63 |
|
\end{equation} |
64 |
|
If there are no external torques acting on a body, the angular |
65 |
|
momentum of it is conserved. The last conservation theorem state |
68 |
|
\end{equation} |
69 |
|
is conserved. All of these conserved quantities are |
70 |
|
important factors to determine the quality of numerical integration |
71 |
< |
scheme for rigid body \cite{Dullweber1997}. |
71 |
> |
schemes for rigid bodies \cite{Dullweber1997}. |
72 |
|
|
73 |
|
\subsection{\label{introSection:lagrangian}Lagrangian Mechanics} |
74 |
|
|
75 |
< |
Newtonian Mechanics suffers from two important limitations: it |
76 |
< |
describes their motion in special cartesian coordinate systems. |
77 |
< |
Another limitation of Newtonian mechanics becomes obvious when we |
78 |
< |
try to describe systems with large numbers of particles. It becomes |
79 |
< |
very difficult to predict the properties of the system by carrying |
80 |
< |
out calculations involving the each individual interaction between |
81 |
< |
all the particles, even if we know all of the details of the |
82 |
< |
interaction. In order to overcome some of the practical difficulties |
83 |
< |
which arise in attempts to apply Newton's equation to complex |
84 |
< |
system, alternative procedures may be developed. |
75 |
> |
Newtonian Mechanics suffers from two important limitations: motions |
76 |
> |
can only be described in cartesian coordinate systems. Moreover, It |
77 |
> |
become impossible to predict analytically the properties of the |
78 |
> |
system even if we know all of the details of the interaction. In |
79 |
> |
order to overcome some of the practical difficulties which arise in |
80 |
> |
attempts to apply Newton's equation to complex system, approximate |
81 |
> |
numerical procedures may be developed. |
82 |
|
|
83 |
< |
\subsubsection{\label{introSection:halmiltonPrinciple}Hamilton's |
84 |
< |
Principle} |
83 |
> |
\subsubsection{\label{introSection:halmiltonPrinciple}\textbf{Hamilton's |
84 |
> |
Principle}} |
85 |
|
|
86 |
|
Hamilton introduced the dynamical principle upon which it is |
87 |
< |
possible to base all of mechanics and, indeed, most of classical |
88 |
< |
physics. Hamilton's Principle may be stated as follow, |
87 |
> |
possible to base all of mechanics and most of classical physics. |
88 |
> |
Hamilton's Principle may be stated as follows, |
89 |
|
|
90 |
|
The actual trajectory, along which a dynamical system may move from |
91 |
|
one point to another within a specified time, is derived by finding |
92 |
|
the path which minimizes the time integral of the difference between |
93 |
< |
the kinetic, $K$, and potential energies, $U$ \cite{tolman79}. |
93 |
> |
the kinetic, $K$, and potential energies, $U$. |
94 |
|
\begin{equation} |
95 |
|
\delta \int_{t_1 }^{t_2 } {(K - U)dt = 0} , |
96 |
|
\label{introEquation:halmitonianPrinciple1} |
97 |
|
\end{equation} |
98 |
|
|
99 |
|
For simple mechanical systems, where the forces acting on the |
100 |
< |
different part are derivable from a potential and the velocities are |
101 |
< |
small compared with that of light, the Lagrangian function $L$ can |
102 |
< |
be define as the difference between the kinetic energy of the system |
106 |
< |
and its potential energy, |
100 |
> |
different parts are derivable from a potential, the Lagrangian |
101 |
> |
function $L$ can be defined as the difference between the kinetic |
102 |
> |
energy of the system and its potential energy, |
103 |
|
\begin{equation} |
104 |
|
L \equiv K - U = L(q_i ,\dot q_i ) , |
105 |
|
\label{introEquation:lagrangianDef} |
110 |
|
\label{introEquation:halmitonianPrinciple2} |
111 |
|
\end{equation} |
112 |
|
|
113 |
< |
\subsubsection{\label{introSection:equationOfMotionLagrangian}The |
114 |
< |
Equations of Motion in Lagrangian Mechanics} |
113 |
> |
\subsubsection{\label{introSection:equationOfMotionLagrangian}\textbf{The |
114 |
> |
Equations of Motion in Lagrangian Mechanics}} |
115 |
|
|
116 |
< |
for a holonomic system of $f$ degrees of freedom, the equations of |
117 |
< |
motion in the Lagrangian form is |
116 |
> |
For a system of $f$ degrees of freedom, the equations of motion in |
117 |
> |
the Lagrangian form is |
118 |
|
\begin{equation} |
119 |
|
\frac{d}{{dt}}\frac{{\partial L}}{{\partial \dot q_i }} - |
120 |
|
\frac{{\partial L}}{{\partial q_i }} = 0,{\rm{ }}i = 1, \ldots,f |
128 |
|
Arising from Lagrangian Mechanics, Hamiltonian Mechanics was |
129 |
|
introduced by William Rowan Hamilton in 1833 as a re-formulation of |
130 |
|
classical mechanics. If the potential energy of a system is |
131 |
< |
independent of generalized velocities, the generalized momenta can |
136 |
< |
be defined as |
131 |
> |
independent of velocities, the momenta can be defined as |
132 |
|
\begin{equation} |
133 |
|
p_i = \frac{\partial L}{\partial \dot q_i} |
134 |
|
\label{introEquation:generalizedMomenta} |
167 |
|
By identifying the coefficients of $dq_k$, $dp_k$ and dt, we can |
168 |
|
find |
169 |
|
\begin{equation} |
170 |
< |
\frac{{\partial H}}{{\partial p_k }} = q_k |
170 |
> |
\frac{{\partial H}}{{\partial p_k }} = \dot {q_k} |
171 |
|
\label{introEquation:motionHamiltonianCoordinate} |
172 |
|
\end{equation} |
173 |
|
\begin{equation} |
174 |
< |
\frac{{\partial H}}{{\partial q_k }} = - p_k |
174 |
> |
\frac{{\partial H}}{{\partial q_k }} = - \dot {p_k} |
175 |
|
\label{introEquation:motionHamiltonianMomentum} |
176 |
|
\end{equation} |
177 |
|
and |
184 |
|
Eq.~\ref{introEquation:motionHamiltonianCoordinate} and |
185 |
|
Eq.~\ref{introEquation:motionHamiltonianMomentum} are Hamilton's |
186 |
|
equation of motion. Due to their symmetrical formula, they are also |
187 |
< |
known as the canonical equations of motions \cite{Goldstein01}. |
187 |
> |
known as the canonical equations of motions \cite{Goldstein2001}. |
188 |
|
|
189 |
|
An important difference between Lagrangian approach and the |
190 |
|
Hamiltonian approach is that the Lagrangian is considered to be a |
191 |
< |
function of the generalized velocities $\dot q_i$ and the |
192 |
< |
generalized coordinates $q_i$, while the Hamiltonian is considered |
193 |
< |
to be a function of the generalized momenta $p_i$ and the conjugate |
194 |
< |
generalized coordinate $q_i$. Hamiltonian Mechanics is more |
195 |
< |
appropriate for application to statistical mechanics and quantum |
196 |
< |
mechanics, since it treats the coordinate and its time derivative as |
197 |
< |
independent variables and it only works with 1st-order differential |
203 |
< |
equations\cite{Marion90}. |
191 |
> |
function of the generalized velocities $\dot q_i$ and coordinates |
192 |
> |
$q_i$, while the Hamiltonian is considered to be a function of the |
193 |
> |
generalized momenta $p_i$ and the conjugate coordinates $q_i$. |
194 |
> |
Hamiltonian Mechanics is more appropriate for application to |
195 |
> |
statistical mechanics and quantum mechanics, since it treats the |
196 |
> |
coordinate and its time derivative as independent variables and it |
197 |
> |
only works with 1st-order differential equations\cite{Marion1990}. |
198 |
|
|
199 |
|
In Newtonian Mechanics, a system described by conservative forces |
200 |
|
conserves the total energy \ref{introEquation:energyConservation}. |
206 |
|
}}\dot p_i } \right)} = \sum\limits_i {\left( {\frac{{\partial |
207 |
|
H}}{{\partial q_i }}\frac{{\partial H}}{{\partial p_i }} - |
208 |
|
\frac{{\partial H}}{{\partial p_i }}\frac{{\partial H}}{{\partial |
209 |
< |
q_i }}} \right) = 0} |
216 |
< |
\label{introEquation:conserveHalmitonian} |
209 |
> |
q_i }}} \right) = 0} \label{introEquation:conserveHalmitonian} |
210 |
|
\end{equation} |
211 |
|
|
219 |
– |
When studying Hamiltonian system, it is more convenient to use |
220 |
– |
notation |
221 |
– |
\begin{equation} |
222 |
– |
r = r(q,p)^T |
223 |
– |
\end{equation} |
224 |
– |
and to introduce a $2n \times 2n$ canonical structure matrix $J$, |
225 |
– |
\begin{equation} |
226 |
– |
J = \left( {\begin{array}{*{20}c} |
227 |
– |
0 & I \\ |
228 |
– |
{ - I} & 0 \\ |
229 |
– |
\end{array}} \right) |
230 |
– |
\label{introEquation:canonicalMatrix} |
231 |
– |
\end{equation} |
232 |
– |
where $I$ is a $n \times n$ identity matrix and $J$ is a |
233 |
– |
skew-symmetric matrix ($ J^T = - J $). Thus, Hamiltonian system |
234 |
– |
can be rewritten as, |
235 |
– |
\begin{equation} |
236 |
– |
\frac{d}{{dt}}r = J\nabla _r H(r) |
237 |
– |
\label{introEquation:compactHamiltonian} |
238 |
– |
\end{equation} |
239 |
– |
|
212 |
|
\section{\label{introSection:statisticalMechanics}Statistical |
213 |
|
Mechanics} |
214 |
|
|
215 |
|
The thermodynamic behaviors and properties of Molecular Dynamics |
216 |
|
simulation are governed by the principle of Statistical Mechanics. |
217 |
|
The following section will give a brief introduction to some of the |
218 |
< |
Statistical Mechanics concepts presented in this dissertation. |
218 |
> |
Statistical Mechanics concepts and theorem presented in this |
219 |
> |
dissertation. |
220 |
|
|
221 |
< |
\subsection{\label{introSection:ensemble}Ensemble and Phase Space} |
221 |
> |
\subsection{\label{introSection:ensemble}Phase Space and Ensemble} |
222 |
> |
|
223 |
> |
Mathematically, phase space is the space which represents all |
224 |
> |
possible states. Each possible state of the system corresponds to |
225 |
> |
one unique point in the phase space. For mechanical systems, the |
226 |
> |
phase space usually consists of all possible values of position and |
227 |
> |
momentum variables. Consider a dynamic system of $f$ particles in a |
228 |
> |
cartesian space, where each of the $6f$ coordinates and momenta is |
229 |
> |
assigned to one of $6f$ mutually orthogonal axes, the phase space of |
230 |
> |
this system is a $6f$ dimensional space. A point, $x = (q_1 , \ldots |
231 |
> |
,q_f ,p_1 , \ldots ,p_f )$, with a unique set of values of $6f$ |
232 |
> |
coordinates and momenta is a phase space vector. |
233 |
> |
|
234 |
> |
%%%fix me |
235 |
> |
A microscopic state or microstate of a classical system is |
236 |
> |
specification of the complete phase space vector of a system at any |
237 |
> |
instant in time. An ensemble is defined as a collection of systems |
238 |
> |
sharing one or more macroscopic characteristics but each being in a |
239 |
> |
unique microstate. The complete ensemble is specified by giving all |
240 |
> |
systems or microstates consistent with the common macroscopic |
241 |
> |
characteristics of the ensemble. Although the state of each |
242 |
> |
individual system in the ensemble could be precisely described at |
243 |
> |
any instance in time by a suitable phase space vector, when using |
244 |
> |
ensembles for statistical purposes, there is no need to maintain |
245 |
> |
distinctions between individual systems, since the numbers of |
246 |
> |
systems at any time in the different states which correspond to |
247 |
> |
different regions of the phase space are more interesting. Moreover, |
248 |
> |
in the point of view of statistical mechanics, one would prefer to |
249 |
> |
use ensembles containing a large enough population of separate |
250 |
> |
members so that the numbers of systems in such different states can |
251 |
> |
be regarded as changing continuously as we traverse different |
252 |
> |
regions of the phase space. The condition of an ensemble at any time |
253 |
> |
can be regarded as appropriately specified by the density $\rho$ |
254 |
> |
with which representative points are distributed over the phase |
255 |
> |
space. The density distribution for an ensemble with $f$ degrees of |
256 |
> |
freedom is defined as, |
257 |
> |
\begin{equation} |
258 |
> |
\rho = \rho (q_1 , \ldots ,q_f ,p_1 , \ldots ,p_f ,t). |
259 |
> |
\label{introEquation:densityDistribution} |
260 |
> |
\end{equation} |
261 |
> |
Governed by the principles of mechanics, the phase points change |
262 |
> |
their locations which would change the density at any time at phase |
263 |
> |
space. Hence, the density distribution is also to be taken as a |
264 |
> |
function of the time. |
265 |
> |
|
266 |
> |
The number of systems $\delta N$ at time $t$ can be determined by, |
267 |
> |
\begin{equation} |
268 |
> |
\delta N = \rho (q,p,t)dq_1 \ldots dq_f dp_1 \ldots dp_f. |
269 |
> |
\label{introEquation:deltaN} |
270 |
> |
\end{equation} |
271 |
> |
Assuming a large enough population of systems, we can sufficiently |
272 |
> |
approximate $\delta N$ without introducing discontinuity when we go |
273 |
> |
from one region in the phase space to another. By integrating over |
274 |
> |
the whole phase space, |
275 |
> |
\begin{equation} |
276 |
> |
N = \int { \ldots \int {\rho (q,p,t)dq_1 } ...dq_f dp_1 } ...dp_f |
277 |
> |
\label{introEquation:totalNumberSystem} |
278 |
> |
\end{equation} |
279 |
> |
gives us an expression for the total number of the systems. Hence, |
280 |
> |
the probability per unit in the phase space can be obtained by, |
281 |
> |
\begin{equation} |
282 |
> |
\frac{{\rho (q,p,t)}}{N} = \frac{{\rho (q,p,t)}}{{\int { \ldots \int |
283 |
> |
{\rho (q,p,t)dq_1 } ...dq_f dp_1 } ...dp_f }}. |
284 |
> |
\label{introEquation:unitProbability} |
285 |
> |
\end{equation} |
286 |
> |
With the help of Eq.~\ref{introEquation:unitProbability} and the |
287 |
> |
knowledge of the system, it is possible to calculate the average |
288 |
> |
value of any desired quantity which depends on the coordinates and |
289 |
> |
momenta of the system. Even when the dynamics of the real system is |
290 |
> |
complex, or stochastic, or even discontinuous, the average |
291 |
> |
properties of the ensemble of possibilities as a whole remaining |
292 |
> |
well defined. For a classical system in thermal equilibrium with its |
293 |
> |
environment, the ensemble average of a mechanical quantity, $\langle |
294 |
> |
A(q , p) \rangle_t$, takes the form of an integral over the phase |
295 |
> |
space of the system, |
296 |
> |
\begin{equation} |
297 |
> |
\langle A(q , p) \rangle_t = \frac{{\int { \ldots \int {A(q,p)\rho |
298 |
> |
(q,p,t)dq_1 } ...dq_f dp_1 } ...dp_f }}{{\int { \ldots \int {\rho |
299 |
> |
(q,p,t)dq_1 } ...dq_f dp_1 } ...dp_f }} |
300 |
> |
\label{introEquation:ensembelAverage} |
301 |
> |
\end{equation} |
302 |
> |
|
303 |
> |
There are several different types of ensembles with different |
304 |
> |
statistical characteristics. As a function of macroscopic |
305 |
> |
parameters, such as temperature \textit{etc}, the partition function |
306 |
> |
can be used to describe the statistical properties of a system in |
307 |
> |
thermodynamic equilibrium. |
308 |
> |
|
309 |
> |
As an ensemble of systems, each of which is known to be thermally |
310 |
> |
isolated and conserve energy, the Microcanonical ensemble (NVE) has |
311 |
> |
a partition function like, |
312 |
> |
\begin{equation} |
313 |
> |
\Omega (N,V,E) = e^{\beta TS} \label{introEquation:NVEPartition}. |
314 |
> |
\end{equation} |
315 |
> |
A canonical ensemble (NVT)is an ensemble of systems, each of which |
316 |
> |
can share its energy with a large heat reservoir. The distribution |
317 |
> |
of the total energy amongst the possible dynamical states is given |
318 |
> |
by the partition function, |
319 |
> |
\begin{equation} |
320 |
> |
\Omega (N,V,T) = e^{ - \beta A} |
321 |
> |
\label{introEquation:NVTPartition} |
322 |
> |
\end{equation} |
323 |
> |
Here, $A$ is the Helmholtz free energy which is defined as $ A = U - |
324 |
> |
TS$. Since most experiments are carried out under constant pressure |
325 |
> |
condition, the isothermal-isobaric ensemble (NPT) plays a very |
326 |
> |
important role in molecular simulations. The isothermal-isobaric |
327 |
> |
ensemble allow the system to exchange energy with a heat bath of |
328 |
> |
temperature $T$ and to change the volume as well. Its partition |
329 |
> |
function is given as |
330 |
> |
\begin{equation} |
331 |
> |
\Delta (N,P,T) = - e^{\beta G}. |
332 |
> |
\label{introEquation:NPTPartition} |
333 |
> |
\end{equation} |
334 |
> |
Here, $G = U - TS + PV$, and $G$ is called Gibbs free energy. |
335 |
> |
|
336 |
> |
\subsection{\label{introSection:liouville}Liouville's theorem} |
337 |
> |
|
338 |
> |
Liouville's theorem is the foundation on which statistical mechanics |
339 |
> |
rests. It describes the time evolution of the phase space |
340 |
> |
distribution function. In order to calculate the rate of change of |
341 |
> |
$\rho$, we begin from Eq.~\ref{introEquation:deltaN}. If we consider |
342 |
> |
the two faces perpendicular to the $q_1$ axis, which are located at |
343 |
> |
$q_1$ and $q_1 + \delta q_1$, the number of phase points leaving the |
344 |
> |
opposite face is given by the expression, |
345 |
> |
\begin{equation} |
346 |
> |
\left( {\rho + \frac{{\partial \rho }}{{\partial q_1 }}\delta q_1 } |
347 |
> |
\right)\left( {\dot q_1 + \frac{{\partial \dot q_1 }}{{\partial q_1 |
348 |
> |
}}\delta q_1 } \right)\delta q_2 \ldots \delta q_f \delta p_1 |
349 |
> |
\ldots \delta p_f . |
350 |
> |
\end{equation} |
351 |
> |
Summing all over the phase space, we obtain |
352 |
> |
\begin{equation} |
353 |
> |
\frac{{d(\delta N)}}{{dt}} = - \sum\limits_{i = 1}^f {\left[ {\rho |
354 |
> |
\left( {\frac{{\partial \dot q_i }}{{\partial q_i }} + |
355 |
> |
\frac{{\partial \dot p_i }}{{\partial p_i }}} \right) + \left( |
356 |
> |
{\frac{{\partial \rho }}{{\partial q_i }}\dot q_i + \frac{{\partial |
357 |
> |
\rho }}{{\partial p_i }}\dot p_i } \right)} \right]} \delta q_1 |
358 |
> |
\ldots \delta q_f \delta p_1 \ldots \delta p_f . |
359 |
> |
\end{equation} |
360 |
> |
Differentiating the equations of motion in Hamiltonian formalism |
361 |
> |
(\ref{introEquation:motionHamiltonianCoordinate}, |
362 |
> |
\ref{introEquation:motionHamiltonianMomentum}), we can show, |
363 |
> |
\begin{equation} |
364 |
> |
\sum\limits_i {\left( {\frac{{\partial \dot q_i }}{{\partial q_i }} |
365 |
> |
+ \frac{{\partial \dot p_i }}{{\partial p_i }}} \right)} = 0 , |
366 |
> |
\end{equation} |
367 |
> |
which cancels the first terms of the right hand side. Furthermore, |
368 |
> |
dividing $ \delta q_1 \ldots \delta q_f \delta p_1 \ldots \delta |
369 |
> |
p_f $ in both sides, we can write out Liouville's theorem in a |
370 |
> |
simple form, |
371 |
> |
\begin{equation} |
372 |
> |
\frac{{\partial \rho }}{{\partial t}} + \sum\limits_{i = 1}^f |
373 |
> |
{\left( {\frac{{\partial \rho }}{{\partial q_i }}\dot q_i + |
374 |
> |
\frac{{\partial \rho }}{{\partial p_i }}\dot p_i } \right)} = 0 . |
375 |
> |
\label{introEquation:liouvilleTheorem} |
376 |
> |
\end{equation} |
377 |
|
|
378 |
+ |
Liouville's theorem states that the distribution function is |
379 |
+ |
constant along any trajectory in phase space. In classical |
380 |
+ |
statistical mechanics, since the number of members in an ensemble is |
381 |
+ |
huge and constant, we can assume the local density has no reason |
382 |
+ |
(other than classical mechanics) to change, |
383 |
+ |
\begin{equation} |
384 |
+ |
\frac{{\partial \rho }}{{\partial t}} = 0. |
385 |
+ |
\label{introEquation:stationary} |
386 |
+ |
\end{equation} |
387 |
+ |
In such stationary system, the density of distribution $\rho$ can be |
388 |
+ |
connected to the Hamiltonian $H$ through Maxwell-Boltzmann |
389 |
+ |
distribution, |
390 |
+ |
\begin{equation} |
391 |
+ |
\rho \propto e^{ - \beta H} |
392 |
+ |
\label{introEquation:densityAndHamiltonian} |
393 |
+ |
\end{equation} |
394 |
+ |
|
395 |
+ |
\subsubsection{\label{introSection:phaseSpaceConservation}\textbf{Conservation of Phase Space}} |
396 |
+ |
Lets consider a region in the phase space, |
397 |
+ |
\begin{equation} |
398 |
+ |
\delta v = \int { \ldots \int {dq_1 } ...dq_f dp_1 } ..dp_f . |
399 |
+ |
\end{equation} |
400 |
+ |
If this region is small enough, the density $\rho$ can be regarded |
401 |
+ |
as uniform over the whole integral. Thus, the number of phase points |
402 |
+ |
inside this region is given by, |
403 |
+ |
\begin{equation} |
404 |
+ |
\delta N = \rho \delta v = \rho \int { \ldots \int {dq_1 } ...dq_f |
405 |
+ |
dp_1 } ..dp_f. |
406 |
+ |
\end{equation} |
407 |
+ |
|
408 |
+ |
\begin{equation} |
409 |
+ |
\frac{{d(\delta N)}}{{dt}} = \frac{{d\rho }}{{dt}}\delta v + \rho |
410 |
+ |
\frac{d}{{dt}}(\delta v) = 0. |
411 |
+ |
\end{equation} |
412 |
+ |
With the help of stationary assumption |
413 |
+ |
(\ref{introEquation:stationary}), we obtain the principle of the |
414 |
+ |
\emph{conservation of volume in phase space}, |
415 |
+ |
\begin{equation} |
416 |
+ |
\frac{d}{{dt}}(\delta v) = \frac{d}{{dt}}\int { \ldots \int {dq_1 } |
417 |
+ |
...dq_f dp_1 } ..dp_f = 0. |
418 |
+ |
\label{introEquation:volumePreserving} |
419 |
+ |
\end{equation} |
420 |
+ |
|
421 |
+ |
\subsubsection{\label{introSection:liouvilleInOtherForms}\textbf{Liouville's Theorem in Other Forms}} |
422 |
+ |
|
423 |
+ |
Liouville's theorem can be expresses in a variety of different forms |
424 |
+ |
which are convenient within different contexts. For any two function |
425 |
+ |
$F$ and $G$ of the coordinates and momenta of a system, the Poisson |
426 |
+ |
bracket ${F, G}$ is defined as |
427 |
+ |
\begin{equation} |
428 |
+ |
\left\{ {F,G} \right\} = \sum\limits_i {\left( {\frac{{\partial |
429 |
+ |
F}}{{\partial q_i }}\frac{{\partial G}}{{\partial p_i }} - |
430 |
+ |
\frac{{\partial F}}{{\partial p_i }}\frac{{\partial G}}{{\partial |
431 |
+ |
q_i }}} \right)}. |
432 |
+ |
\label{introEquation:poissonBracket} |
433 |
+ |
\end{equation} |
434 |
+ |
Substituting equations of motion in Hamiltonian formalism( |
435 |
+ |
Eq.~\ref{introEquation:motionHamiltonianCoordinate} , |
436 |
+ |
Eq.~\ref{introEquation:motionHamiltonianMomentum} ) into |
437 |
+ |
(Eq.~\ref{introEquation:liouvilleTheorem}), we can rewrite |
438 |
+ |
Liouville's theorem using Poisson bracket notion, |
439 |
+ |
\begin{equation} |
440 |
+ |
\left( {\frac{{\partial \rho }}{{\partial t}}} \right) = - \left\{ |
441 |
+ |
{\rho ,H} \right\}. |
442 |
+ |
\label{introEquation:liouvilleTheromInPoissin} |
443 |
+ |
\end{equation} |
444 |
+ |
Moreover, the Liouville operator is defined as |
445 |
+ |
\begin{equation} |
446 |
+ |
iL = \sum\limits_{i = 1}^f {\left( {\frac{{\partial H}}{{\partial |
447 |
+ |
p_i }}\frac{\partial }{{\partial q_i }} - \frac{{\partial |
448 |
+ |
H}}{{\partial q_i }}\frac{\partial }{{\partial p_i }}} \right)} |
449 |
+ |
\label{introEquation:liouvilleOperator} |
450 |
+ |
\end{equation} |
451 |
+ |
In terms of Liouville operator, Liouville's equation can also be |
452 |
+ |
expressed as |
453 |
+ |
\begin{equation} |
454 |
+ |
\left( {\frac{{\partial \rho }}{{\partial t}}} \right) = - iL\rho |
455 |
+ |
\label{introEquation:liouvilleTheoremInOperator} |
456 |
+ |
\end{equation} |
457 |
+ |
|
458 |
|
\subsection{\label{introSection:ergodic}The Ergodic Hypothesis} |
459 |
|
|
460 |
|
Various thermodynamic properties can be calculated from Molecular |
461 |
|
Dynamics simulation. By comparing experimental values with the |
462 |
|
calculated properties, one can determine the accuracy of the |
463 |
< |
simulation and the quality of the underlying model. However, both of |
464 |
< |
experiment and computer simulation are usually performed during a |
463 |
> |
simulation and the quality of the underlying model. However, both |
464 |
> |
experiments and computer simulations are usually performed during a |
465 |
|
certain time interval and the measurements are averaged over a |
466 |
|
period of them which is different from the average behavior of |
467 |
< |
many-body system in Statistical Mechanics. Fortunately, Ergodic |
468 |
< |
Hypothesis is proposed to make a connection between time average and |
469 |
< |
ensemble average. It states that time average and average over the |
470 |
< |
statistical ensemble are identical \cite{Frenkel1996, leach01:mm}. |
467 |
> |
many-body system in Statistical Mechanics. Fortunately, the Ergodic |
468 |
> |
Hypothesis makes a connection between time average and the ensemble |
469 |
> |
average. It states that the time average and average over the |
470 |
> |
statistical ensemble are identical \cite{Frenkel1996, Leach2001}. |
471 |
|
\begin{equation} |
472 |
< |
\langle A \rangle_t = \mathop {\lim }\limits_{t \to \infty } |
473 |
< |
\frac{1}{t}\int\limits_0^t {A(p(t),q(t))dt = \int\limits_\Gamma |
474 |
< |
{A(p(t),q(t))} } \rho (p(t), q(t)) dpdq |
472 |
> |
\langle A(q , p) \rangle_t = \mathop {\lim }\limits_{t \to \infty } |
473 |
> |
\frac{1}{t}\int\limits_0^t {A(q(t),p(t))dt = \int\limits_\Gamma |
474 |
> |
{A(q(t),p(t))} } \rho (q(t), p(t)) dqdp |
475 |
|
\end{equation} |
476 |
< |
where $\langle A \rangle_t$ is an equilibrium value of a physical |
477 |
< |
quantity and $\rho (p(t), q(t))$ is the equilibrium distribution |
478 |
< |
function. If an observation is averaged over a sufficiently long |
479 |
< |
time (longer than relaxation time), all accessible microstates in |
480 |
< |
phase space are assumed to be equally probed, giving a properly |
481 |
< |
weighted statistical average. This allows the researcher freedom of |
482 |
< |
choice when deciding how best to measure a given observable. In case |
483 |
< |
an ensemble averaged approach sounds most reasonable, the Monte |
484 |
< |
Carlo techniques\cite{metropolis:1949} can be utilized. Or if the |
485 |
< |
system lends itself to a time averaging approach, the Molecular |
486 |
< |
Dynamics techniques in Sec.~\ref{introSection:molecularDynamics} |
487 |
< |
will be the best choice\cite{Frenkel1996}. |
476 |
> |
where $\langle A(q , p) \rangle_t$ is an equilibrium value of a |
477 |
> |
physical quantity and $\rho (p(t), q(t))$ is the equilibrium |
478 |
> |
distribution function. If an observation is averaged over a |
479 |
> |
sufficiently long time (longer than relaxation time), all accessible |
480 |
> |
microstates in phase space are assumed to be equally probed, giving |
481 |
> |
a properly weighted statistical average. This allows the researcher |
482 |
> |
freedom of choice when deciding how best to measure a given |
483 |
> |
observable. In case an ensemble averaged approach sounds most |
484 |
> |
reasonable, the Monte Carlo techniques\cite{Metropolis1949} can be |
485 |
> |
utilized. Or if the system lends itself to a time averaging |
486 |
> |
approach, the Molecular Dynamics techniques in |
487 |
> |
Sec.~\ref{introSection:molecularDynamics} will be the best |
488 |
> |
choice\cite{Frenkel1996}. |
489 |
|
|
490 |
|
\section{\label{introSection:geometricIntegratos}Geometric Integrators} |
491 |
< |
A variety of numerical integrators were proposed to simulate the |
492 |
< |
motions. They usually begin with an initial conditionals and move |
493 |
< |
the objects in the direction governed by the differential equations. |
494 |
< |
However, most of them ignore the hidden physical law contained |
495 |
< |
within the equations. Since 1990, geometric integrators, which |
496 |
< |
preserve various phase-flow invariants such as symplectic structure, |
497 |
< |
volume and time reversal symmetry, are developed to address this |
498 |
< |
issue. The velocity verlet method, which happens to be a simple |
499 |
< |
example of symplectic integrator, continues to gain its popularity |
500 |
< |
in molecular dynamics community. This fact can be partly explained |
501 |
< |
by its geometric nature. |
491 |
> |
A variety of numerical integrators have been proposed to simulate |
492 |
> |
the motions of atoms in MD simulation. They usually begin with |
493 |
> |
initial conditionals and move the objects in the direction governed |
494 |
> |
by the differential equations. However, most of them ignore the |
495 |
> |
hidden physical laws contained within the equations. Since 1990, |
496 |
> |
geometric integrators, which preserve various phase-flow invariants |
497 |
> |
such as symplectic structure, volume and time reversal symmetry, are |
498 |
> |
developed to address this issue\cite{Dullweber1997, McLachlan1998, |
499 |
> |
Leimkuhler1999}. The velocity Verlet method, which happens to be a |
500 |
> |
simple example of symplectic integrator, continues to gain |
501 |
> |
popularity in the molecular dynamics community. This fact can be |
502 |
> |
partly explained by its geometric nature. |
503 |
|
|
504 |
< |
\subsection{\label{introSection:symplecticManifold}Symplectic Manifold} |
505 |
< |
A \emph{manifold} is an abstract mathematical space. It locally |
506 |
< |
looks like Euclidean space, but when viewed globally, it may have |
507 |
< |
more complicate structure. A good example of manifold is the surface |
508 |
< |
of Earth. It seems to be flat locally, but it is round if viewed as |
509 |
< |
a whole. A \emph{differentiable manifold} (also known as |
510 |
< |
\emph{smooth manifold}) is a manifold with an open cover in which |
511 |
< |
the covering neighborhoods are all smoothly isomorphic to one |
512 |
< |
another. In other words,it is possible to apply calculus on |
303 |
< |
\emph{differentiable manifold}. A \emph{symplectic manifold} is |
304 |
< |
defined as a pair $(M, \omega)$ which consisting of a |
504 |
> |
\subsection{\label{introSection:symplecticManifold}Symplectic Manifolds} |
505 |
> |
A \emph{manifold} is an abstract mathematical space. It looks |
506 |
> |
locally like Euclidean space, but when viewed globally, it may have |
507 |
> |
more complicated structure. A good example of manifold is the |
508 |
> |
surface of Earth. It seems to be flat locally, but it is round if |
509 |
> |
viewed as a whole. A \emph{differentiable manifold} (also known as |
510 |
> |
\emph{smooth manifold}) is a manifold on which it is possible to |
511 |
> |
apply calculus on \emph{differentiable manifold}. A \emph{symplectic |
512 |
> |
manifold} is defined as a pair $(M, \omega)$ which consists of a |
513 |
|
\emph{differentiable manifold} $M$ and a close, non-degenerated, |
514 |
|
bilinear symplectic form, $\omega$. A symplectic form on a vector |
515 |
|
space $V$ is a function $\omega(x, y)$ which satisfies |
516 |
|
$\omega(\lambda_1x_1+\lambda_2x_2, y) = \lambda_1\omega(x_1, y)+ |
517 |
|
\lambda_2\omega(x_2, y)$, $\omega(x, y) = - \omega(y, x)$ and |
518 |
< |
$\omega(x, x) = 0$. Cross product operation in vector field is an |
519 |
< |
example of symplectic form. |
518 |
> |
$\omega(x, x) = 0$. The cross product operation in vector field is |
519 |
> |
an example of symplectic form. |
520 |
|
|
521 |
< |
One of the motivations to study \emph{symplectic manifold} in |
521 |
> |
One of the motivations to study \emph{symplectic manifolds} in |
522 |
|
Hamiltonian Mechanics is that a symplectic manifold can represent |
523 |
|
all possible configurations of the system and the phase space of the |
524 |
|
system can be described by it's cotangent bundle. Every symplectic |
525 |
|
manifold is even dimensional. For instance, in Hamilton equations, |
526 |
|
coordinate and momentum always appear in pairs. |
527 |
|
|
528 |
< |
Let $(M,\omega)$ and $(N, \eta)$ be symplectic manifolds. A map |
321 |
< |
\[ |
322 |
< |
f : M \rightarrow N |
323 |
< |
\] |
324 |
< |
is a \emph{symplectomorphism} if it is a \emph{diffeomorphims} and |
325 |
< |
the \emph{pullback} of $\eta$ under f is equal to $\omega$. |
326 |
< |
Canonical transformation is an example of symplectomorphism in |
327 |
< |
classical mechanics. According to Liouville's theorem, the |
328 |
< |
Hamiltonian \emph{flow} or \emph{symplectomorphism} generated by the |
329 |
< |
Hamiltonian vector filed preserves the volume form on the phase |
330 |
< |
space, which is the basis of classical statistical mechanics. |
528 |
> |
\subsection{\label{introSection:ODE}Ordinary Differential Equations} |
529 |
|
|
530 |
< |
\subsection{\label{introSection:exactFlow}The Exact Flow of ODE} |
530 |
> |
For an ordinary differential system defined as |
531 |
> |
\begin{equation} |
532 |
> |
\dot x = f(x) |
533 |
> |
\end{equation} |
534 |
> |
where $x = x(q,p)^T$, this system is a canonical Hamiltonian, if |
535 |
> |
\begin{equation} |
536 |
> |
f(r) = J\nabla _x H(r). |
537 |
> |
\end{equation} |
538 |
> |
$H = H (q, p)$ is Hamiltonian function and $J$ is the skew-symmetric |
539 |
> |
matrix |
540 |
> |
\begin{equation} |
541 |
> |
J = \left( {\begin{array}{*{20}c} |
542 |
> |
0 & I \\ |
543 |
> |
{ - I} & 0 \\ |
544 |
> |
\end{array}} \right) |
545 |
> |
\label{introEquation:canonicalMatrix} |
546 |
> |
\end{equation} |
547 |
> |
where $I$ is an identity matrix. Using this notation, Hamiltonian |
548 |
> |
system can be rewritten as, |
549 |
> |
\begin{equation} |
550 |
> |
\frac{d}{{dt}}x = J\nabla _x H(x) |
551 |
> |
\label{introEquation:compactHamiltonian} |
552 |
> |
\end{equation}In this case, $f$ is |
553 |
> |
called a \emph{Hamiltonian vector field}. |
554 |
|
|
555 |
< |
\subsection{\label{introSection:hamiltonianSplitting}Hamiltonian Splitting} |
555 |
> |
Another generalization of Hamiltonian dynamics is Poisson |
556 |
> |
Dynamics\cite{Olver1986}, |
557 |
> |
\begin{equation} |
558 |
> |
\dot x = J(x)\nabla _x H \label{introEquation:poissonHamiltonian} |
559 |
> |
\end{equation} |
560 |
> |
The most obvious change being that matrix $J$ now depends on $x$. |
561 |
|
|
562 |
< |
\section{\label{introSection:molecularDynamics}Molecular Dynamics} |
562 |
> |
\subsection{\label{introSection:exactFlow}Exact Flow} |
563 |
|
|
564 |
< |
As a special discipline of molecular modeling, Molecular dynamics |
565 |
< |
has proven to be a powerful tool for studying the functions of |
566 |
< |
biological systems, providing structural, thermodynamic and |
567 |
< |
dynamical information. |
564 |
> |
Let $x(t)$ be the exact solution of the ODE system, |
565 |
> |
\begin{equation} |
566 |
> |
\frac{{dx}}{{dt}} = f(x) \label{introEquation:ODE} |
567 |
> |
\end{equation} |
568 |
> |
The exact flow(solution) $\varphi_\tau$ is defined by |
569 |
> |
\[ |
570 |
> |
x(t+\tau) =\varphi_\tau(x(t)) |
571 |
> |
\] |
572 |
> |
where $\tau$ is a fixed time step and $\varphi$ is a map from phase |
573 |
> |
space to itself. The flow has the continuous group property, |
574 |
> |
\begin{equation} |
575 |
> |
\varphi _{\tau _1 } \circ \varphi _{\tau _2 } = \varphi _{\tau _1 |
576 |
> |
+ \tau _2 } . |
577 |
> |
\end{equation} |
578 |
> |
In particular, |
579 |
> |
\begin{equation} |
580 |
> |
\varphi _\tau \circ \varphi _{ - \tau } = I |
581 |
> |
\end{equation} |
582 |
> |
Therefore, the exact flow is self-adjoint, |
583 |
> |
\begin{equation} |
584 |
> |
\varphi _\tau = \varphi _{ - \tau }^{ - 1}. |
585 |
> |
\end{equation} |
586 |
> |
The exact flow can also be written in terms of the of an operator, |
587 |
> |
\begin{equation} |
588 |
> |
\varphi _\tau (x) = e^{\tau \sum\limits_i {f_i (x)\frac{\partial |
589 |
> |
}{{\partial x_i }}} } (x) \equiv \exp (\tau f)(x). |
590 |
> |
\label{introEquation:exponentialOperator} |
591 |
> |
\end{equation} |
592 |
|
|
593 |
< |
\subsection{\label{introSec:mdInit}Initialization} |
593 |
> |
In most cases, it is not easy to find the exact flow $\varphi_\tau$. |
594 |
> |
Instead, we use an approximate map, $\psi_\tau$, which is usually |
595 |
> |
called integrator. The order of an integrator $\psi_\tau$ is $p$, if |
596 |
> |
the Taylor series of $\psi_\tau$ agree to order $p$, |
597 |
> |
\begin{equation} |
598 |
> |
\psi_\tau(x) = x + \tau f(x) + O(\tau^{p+1}) |
599 |
> |
\end{equation} |
600 |
|
|
601 |
< |
\subsection{\label{introSection:mdIntegration} Integration of the Equations of Motion} |
601 |
> |
\subsection{\label{introSection:geometricProperties}Geometric Properties} |
602 |
|
|
603 |
< |
\section{\label{introSection:rigidBody}Dynamics of Rigid Bodies} |
603 |
> |
The hidden geometric properties\cite{Budd1999, Marsden1998} of an |
604 |
> |
ODE and its flow play important roles in numerical studies. Many of |
605 |
> |
them can be found in systems which occur naturally in applications. |
606 |
|
|
607 |
< |
A rigid body is a body in which the distance between any two given |
608 |
< |
points of a rigid body remains constant regardless of external |
609 |
< |
forces exerted on it. A rigid body therefore conserves its shape |
610 |
< |
during its motion. |
607 |
> |
Let $\varphi$ be the flow of Hamiltonian vector field, $\varphi$ is |
608 |
> |
a \emph{symplectic} flow if it satisfies, |
609 |
> |
\begin{equation} |
610 |
> |
{\varphi '}^T J \varphi ' = J. |
611 |
> |
\end{equation} |
612 |
> |
According to Liouville's theorem, the symplectic volume is invariant |
613 |
> |
under a Hamiltonian flow, which is the basis for classical |
614 |
> |
statistical mechanics. Furthermore, the flow of a Hamiltonian vector |
615 |
> |
field on a symplectic manifold can be shown to be a |
616 |
> |
symplectomorphism. As to the Poisson system, |
617 |
> |
\begin{equation} |
618 |
> |
{\varphi '}^T J \varphi ' = J \circ \varphi |
619 |
> |
\end{equation} |
620 |
> |
is the property that must be preserved by the integrator. |
621 |
|
|
622 |
< |
Applications of dynamics of rigid bodies. |
622 |
> |
It is possible to construct a \emph{volume-preserving} flow for a |
623 |
> |
source free ODE ($ \nabla \cdot f = 0 $), if the flow satisfies $ |
624 |
> |
\det d\varphi = 1$. One can show easily that a symplectic flow will |
625 |
> |
be volume-preserving. |
626 |
|
|
627 |
< |
\subsection{\label{introSection:lieAlgebra}Lie Algebra} |
627 |
> |
Changing the variables $y = h(x)$ in an ODE |
628 |
> |
(Eq.~\ref{introEquation:ODE}) will result in a new system, |
629 |
> |
\[ |
630 |
> |
\dot y = \tilde f(y) = ((dh \cdot f)h^{ - 1} )(y). |
631 |
> |
\] |
632 |
> |
The vector filed $f$ has reversing symmetry $h$ if $f = - \tilde f$. |
633 |
> |
In other words, the flow of this vector field is reversible if and |
634 |
> |
only if $ h \circ \varphi ^{ - 1} = \varphi \circ h $. |
635 |
|
|
636 |
< |
\subsection{\label{introSection:DLMMotionEquation}The Euler Equations of Rigid Body Motion} |
636 |
> |
A \emph{first integral}, or conserved quantity of a general |
637 |
> |
differential function is a function $ G:R^{2d} \to R^d $ which is |
638 |
> |
constant for all solutions of the ODE $\frac{{dx}}{{dt}} = f(x)$ , |
639 |
> |
\[ |
640 |
> |
\frac{{dG(x(t))}}{{dt}} = 0. |
641 |
> |
\] |
642 |
> |
Using chain rule, one may obtain, |
643 |
> |
\[ |
644 |
> |
\sum\limits_i {\frac{{dG}}{{dx_i }}} f_i (x) = f \bullet \nabla G, |
645 |
> |
\] |
646 |
> |
which is the condition for conserving \emph{first integral}. For a |
647 |
> |
canonical Hamiltonian system, the time evolution of an arbitrary |
648 |
> |
smooth function $G$ is given by, |
649 |
|
|
650 |
< |
\subsection{\label{introSection:otherRBMotionEquation}Other Formulations for Rigid Body Motion} |
650 |
> |
\begin{eqnarray} |
651 |
> |
\frac{{dG(x(t))}}{{dt}} & = & [\nabla _x G(x(t))]^T \dot x(t) \\ |
652 |
> |
& = & [\nabla _x G(x(t))]^T J\nabla _x H(x(t)). \\ |
653 |
> |
\label{introEquation:firstIntegral1} |
654 |
> |
\end{eqnarray} |
655 |
|
|
362 |
– |
%\subsection{\label{introSection:poissonBrackets}Poisson Brackets} |
656 |
|
|
657 |
< |
\section{\label{introSection:correlationFunctions}Correlation Functions} |
657 |
> |
Using poisson bracket notion, Equation |
658 |
> |
\ref{introEquation:firstIntegral1} can be rewritten as |
659 |
> |
\[ |
660 |
> |
\frac{d}{{dt}}G(x(t)) = \left\{ {G,H} \right\}(x(t)). |
661 |
> |
\] |
662 |
> |
Therefore, the sufficient condition for $G$ to be the \emph{first |
663 |
> |
integral} of a Hamiltonian system is |
664 |
> |
\[ |
665 |
> |
\left\{ {G,H} \right\} = 0. |
666 |
> |
\] |
667 |
> |
As well known, the Hamiltonian (or energy) H of a Hamiltonian system |
668 |
> |
is a \emph{first integral}, which is due to the fact $\{ H,H\} = |
669 |
> |
0$. |
670 |
|
|
671 |
< |
\section{\label{introSection:langevinDynamics}Langevin Dynamics} |
671 |
> |
When designing any numerical methods, one should always try to |
672 |
> |
preserve the structural properties of the original ODE and its flow. |
673 |
|
|
674 |
< |
\subsection{\label{introSection:LDIntroduction}Introduction and application of Langevin Dynamics} |
674 |
> |
\subsection{\label{introSection:constructionSymplectic}Construction of Symplectic Methods} |
675 |
> |
A lot of well established and very effective numerical methods have |
676 |
> |
been successful precisely because of their symplecticities even |
677 |
> |
though this fact was not recognized when they were first |
678 |
> |
constructed. The most famous example is the Verlet-leapfrog method |
679 |
> |
in molecular dynamics. In general, symplectic integrators can be |
680 |
> |
constructed using one of four different methods. |
681 |
> |
\begin{enumerate} |
682 |
> |
\item Generating functions |
683 |
> |
\item Variational methods |
684 |
> |
\item Runge-Kutta methods |
685 |
> |
\item Splitting methods |
686 |
> |
\end{enumerate} |
687 |
|
|
688 |
< |
\subsection{\label{introSection:generalizedLangevinDynamics}Generalized Langevin Dynamics} |
688 |
> |
Generating function\cite{Channell1990} tends to lead to methods |
689 |
> |
which are cumbersome and difficult to use. In dissipative systems, |
690 |
> |
variational methods can capture the decay of energy |
691 |
> |
accurately\cite{Kane2000}. Since their geometrically unstable nature |
692 |
> |
against non-Hamiltonian perturbations, ordinary implicit Runge-Kutta |
693 |
> |
methods are not suitable for Hamiltonian system. Recently, various |
694 |
> |
high-order explicit Runge-Kutta methods |
695 |
> |
\cite{Owren1992,Chen2003}have been developed to overcome this |
696 |
> |
instability. However, due to computational penalty involved in |
697 |
> |
implementing the Runge-Kutta methods, they have not attracted much |
698 |
> |
attention from the Molecular Dynamics community. Instead, splitting |
699 |
> |
methods have been widely accepted since they exploit natural |
700 |
> |
decompositions of the system\cite{Tuckerman1992, McLachlan1998}. |
701 |
|
|
702 |
+ |
\subsubsection{\label{introSection:splittingMethod}\textbf{Splitting Methods}} |
703 |
+ |
|
704 |
+ |
The main idea behind splitting methods is to decompose the discrete |
705 |
+ |
$\varphi_h$ as a composition of simpler flows, |
706 |
|
\begin{equation} |
707 |
< |
H = \frac{{p^2 }}{{2m}} + U(x) + H_B + \Delta U(x,x_1 , \ldots x_N) |
708 |
< |
\label{introEquation:bathGLE} |
707 |
> |
\varphi _h = \varphi _{h_1 } \circ \varphi _{h_2 } \ldots \circ |
708 |
> |
\varphi _{h_n } |
709 |
> |
\label{introEquation:FlowDecomposition} |
710 |
|
\end{equation} |
711 |
< |
where $H_B$ is harmonic bath Hamiltonian, |
711 |
> |
where each of the sub-flow is chosen such that each represent a |
712 |
> |
simpler integration of the system. |
713 |
> |
|
714 |
> |
Suppose that a Hamiltonian system takes the form, |
715 |
|
\[ |
716 |
< |
H_B =\sum\limits_{\alpha = 1}^N {\left\{ {\frac{{p_\alpha ^2 |
379 |
< |
}}{{2m_\alpha }} + \frac{1}{2}m_\alpha w_\alpha ^2 } \right\}} |
716 |
> |
H = H_1 + H_2. |
717 |
|
\] |
718 |
< |
and $\Delta U$ is bilinear system-bath coupling, |
718 |
> |
Here, $H_1$ and $H_2$ may represent different physical processes of |
719 |
> |
the system. For instance, they may relate to kinetic and potential |
720 |
> |
energy respectively, which is a natural decomposition of the |
721 |
> |
problem. If $H_1$ and $H_2$ can be integrated using exact flows |
722 |
> |
$\varphi_1(t)$ and $\varphi_2(t)$, respectively, a simple first |
723 |
> |
order expression is then given by the Lie-Trotter formula |
724 |
> |
\begin{equation} |
725 |
> |
\varphi _h = \varphi _{1,h} \circ \varphi _{2,h}, |
726 |
> |
\label{introEquation:firstOrderSplitting} |
727 |
> |
\end{equation} |
728 |
> |
where $\varphi _h$ is the result of applying the corresponding |
729 |
> |
continuous $\varphi _i$ over a time $h$. By definition, as |
730 |
> |
$\varphi_i(t)$ is the exact solution of a Hamiltonian system, it |
731 |
> |
must follow that each operator $\varphi_i(t)$ is a symplectic map. |
732 |
> |
It is easy to show that any composition of symplectic flows yields a |
733 |
> |
symplectic map, |
734 |
> |
\begin{equation} |
735 |
> |
(\varphi '\phi ')^T J\varphi '\phi ' = \phi '^T \varphi '^T J\varphi |
736 |
> |
'\phi ' = \phi '^T J\phi ' = J, |
737 |
> |
\label{introEquation:SymplecticFlowComposition} |
738 |
> |
\end{equation} |
739 |
> |
where $\phi$ and $\psi$ both are symplectic maps. Thus operator |
740 |
> |
splitting in this context automatically generates a symplectic map. |
741 |
> |
|
742 |
> |
The Lie-Trotter splitting(\ref{introEquation:firstOrderSplitting}) |
743 |
> |
introduces local errors proportional to $h^2$, while Strang |
744 |
> |
splitting gives a second-order decomposition, |
745 |
> |
\begin{equation} |
746 |
> |
\varphi _h = \varphi _{1,h/2} \circ \varphi _{2,h} \circ \varphi |
747 |
> |
_{1,h/2} , \label{introEquation:secondOrderSplitting} |
748 |
> |
\end{equation} |
749 |
> |
which has a local error proportional to $h^3$. The Sprang |
750 |
> |
splitting's popularity in molecular simulation community attribute |
751 |
> |
to its symmetric property, |
752 |
> |
\begin{equation} |
753 |
> |
\varphi _h^{ - 1} = \varphi _{ - h}. |
754 |
> |
\label{introEquation:timeReversible} |
755 |
> |
\end{equation},appendixFig:architecture |
756 |
> |
|
757 |
> |
\subsubsection{\label{introSection:exampleSplittingMethod}\textbf{Examples of the Splitting Method}} |
758 |
> |
The classical equation for a system consisting of interacting |
759 |
> |
particles can be written in Hamiltonian form, |
760 |
|
\[ |
761 |
< |
\Delta U = - \sum\limits_{\alpha = 1}^N {g_\alpha x_\alpha x} |
761 |
> |
H = T + V |
762 |
|
\] |
763 |
< |
Completing the square, |
764 |
< |
\[ |
765 |
< |
H_B + \Delta U = \sum\limits_{\alpha = 1}^N {\left\{ |
388 |
< |
{\frac{{p_\alpha ^2 }}{{2m_\alpha }} + \frac{1}{2}m_\alpha |
389 |
< |
w_\alpha ^2 \left( {x_\alpha - \frac{{g_\alpha }}{{m_\alpha |
390 |
< |
w_\alpha ^2 }}x} \right)^2 } \right\}} - \sum\limits_{\alpha = |
391 |
< |
1}^N {\frac{{g_\alpha ^2 }}{{2m_\alpha w_\alpha ^2 }}} x^2 |
392 |
< |
\] |
393 |
< |
and putting it back into Eq.~\ref{introEquation:bathGLE}, |
394 |
< |
\[ |
395 |
< |
H = \frac{{p^2 }}{{2m}} + W(x) + \sum\limits_{\alpha = 1}^N |
396 |
< |
{\left\{ {\frac{{p_\alpha ^2 }}{{2m_\alpha }} + \frac{1}{2}m_\alpha |
397 |
< |
w_\alpha ^2 \left( {x_\alpha - \frac{{g_\alpha }}{{m_\alpha |
398 |
< |
w_\alpha ^2 }}x} \right)^2 } \right\}} |
399 |
< |
\] |
400 |
< |
where |
401 |
< |
\[ |
402 |
< |
W(x) = U(x) - \sum\limits_{\alpha = 1}^N {\frac{{g_\alpha ^2 |
403 |
< |
}}{{2m_\alpha w_\alpha ^2 }}} x^2 |
404 |
< |
\] |
405 |
< |
Since the first two terms of the new Hamiltonian depend only on the |
406 |
< |
system coordinates, we can get the equations of motion for |
407 |
< |
Generalized Langevin Dynamics by Hamilton's equations |
408 |
< |
\ref{introEquation:motionHamiltonianCoordinate, |
409 |
< |
introEquation:motionHamiltonianMomentum}, |
763 |
> |
where $T$ is the kinetic energy and $V$ is the potential energy. |
764 |
> |
Setting $H_1 = T, H_2 = V$ and applying the Strang splitting, one |
765 |
> |
obtains the following: |
766 |
|
\begin{align} |
767 |
< |
\dot p &= - \frac{{\partial H}}{{\partial x}} |
768 |
< |
&= m\ddot x |
769 |
< |
&= - \frac{{\partial W(x)}}{{\partial x}} - \sum\limits_{\alpha = 1}^N {g_\alpha \left( {x_\alpha - \frac{{g_\alpha }}{{m_\alpha w_\alpha ^2 }}x} \right)} |
770 |
< |
\label{introEq:Lp5} |
767 |
> |
q(\Delta t) &= q(0) + \dot{q}(0)\Delta t + |
768 |
> |
\frac{F[q(0)]}{m}\frac{\Delta t^2}{2}, % |
769 |
> |
\label{introEquation:Lp10a} \\% |
770 |
> |
% |
771 |
> |
\dot{q}(\Delta t) &= \dot{q}(0) + \frac{\Delta t}{2m} |
772 |
> |
\biggl [F[q(0)] + F[q(\Delta t)] \biggr]. % |
773 |
> |
\label{introEquation:Lp10b} |
774 |
|
\end{align} |
775 |
< |
, and |
775 |
> |
where $F(t)$ is the force at time $t$. This integration scheme is |
776 |
> |
known as \emph{velocity verlet} which is |
777 |
> |
symplectic(\ref{introEquation:SymplecticFlowComposition}), |
778 |
> |
time-reversible(\ref{introEquation:timeReversible}) and |
779 |
> |
volume-preserving (\ref{introEquation:volumePreserving}). These |
780 |
> |
geometric properties attribute to its long-time stability and its |
781 |
> |
popularity in the community. However, the most commonly used |
782 |
> |
velocity verlet integration scheme is written as below, |
783 |
|
\begin{align} |
784 |
< |
\dot p_\alpha &= - \frac{{\partial H}}{{\partial x_\alpha }} |
785 |
< |
&= m\ddot x_\alpha |
786 |
< |
&= \- m_\alpha w_\alpha ^2 \left( {x_\alpha - \frac{{g_\alpha}}{{m_\alpha w_\alpha ^2 }}x} \right) |
784 |
> |
\dot{q}\biggl (\frac{\Delta t}{2}\biggr ) &= |
785 |
> |
\dot{q}(0) + \frac{\Delta t}{2m}\, F[q(0)], \label{introEquation:Lp9a}\\% |
786 |
> |
% |
787 |
> |
q(\Delta t) &= q(0) + \Delta t\, \dot{q}\biggl (\frac{\Delta t}{2}\biggr ),% |
788 |
> |
\label{introEquation:Lp9b}\\% |
789 |
> |
% |
790 |
> |
\dot{q}(\Delta t) &= \dot{q}\biggl (\frac{\Delta t}{2}\biggr ) + |
791 |
> |
\frac{\Delta t}{2m}\, F[q(t)]. \label{introEquation:Lp9c} |
792 |
|
\end{align} |
793 |
+ |
From the preceding splitting, one can see that the integration of |
794 |
+ |
the equations of motion would follow: |
795 |
+ |
\begin{enumerate} |
796 |
+ |
\item calculate the velocities at the half step, $\frac{\Delta t}{2}$, from the forces calculated at the initial position. |
797 |
|
|
798 |
< |
\subsection{\label{introSection:laplaceTransform}The Laplace Transform} |
798 |
> |
\item Use the half step velocities to move positions one whole step, $\Delta t$. |
799 |
|
|
800 |
< |
\[ |
426 |
< |
L(x) = \int_0^\infty {x(t)e^{ - pt} dt} |
427 |
< |
\] |
800 |
> |
\item Evaluate the forces at the new positions, $\mathbf{q}(\Delta t)$, and use the new forces to complete the velocity move. |
801 |
|
|
802 |
< |
\[ |
803 |
< |
L(x + y) = L(x) + L(y) |
431 |
< |
\] |
802 |
> |
\item Repeat from step 1 with the new position, velocities, and forces assuming the roles of the initial values. |
803 |
> |
\end{enumerate} |
804 |
|
|
805 |
< |
\[ |
806 |
< |
L(ax) = aL(x) |
807 |
< |
\] |
805 |
> |
By simply switching the order of the propagators in the splitting |
806 |
> |
and composing a new integrator, the \emph{position verlet} |
807 |
> |
integrator, can be generated, |
808 |
> |
\begin{align} |
809 |
> |
\dot q(\Delta t) &= \dot q(0) + \Delta tF(q(0))\left[ {q(0) + |
810 |
> |
\frac{{\Delta t}}{{2m}}\dot q(0)} \right], % |
811 |
> |
\label{introEquation:positionVerlet1} \\% |
812 |
> |
% |
813 |
> |
q(\Delta t) &= q(0) + \frac{{\Delta t}}{2}\left[ {\dot q(0) + \dot |
814 |
> |
q(\Delta t)} \right]. % |
815 |
> |
\label{introEquation:positionVerlet2} |
816 |
> |
\end{align} |
817 |
|
|
818 |
< |
\[ |
438 |
< |
L(\dot x) = pL(x) - px(0) |
439 |
< |
\] |
818 |
> |
\subsubsection{\label{introSection:errorAnalysis}\textbf{Error Analysis and Higher Order Methods}} |
819 |
|
|
820 |
+ |
The Baker-Campbell-Hausdorff formula can be used to determine the |
821 |
+ |
local error of splitting method in terms of the commutator of the |
822 |
+ |
operators(\ref{introEquation:exponentialOperator}) associated with |
823 |
+ |
the sub-flow. For operators $hX$ and $hY$ which are associated with |
824 |
+ |
$\varphi_1(t)$ and $\varphi_2(t)$ respectively , we have |
825 |
+ |
\begin{equation} |
826 |
+ |
\exp (hX + hY) = \exp (hZ) |
827 |
+ |
\end{equation} |
828 |
+ |
where |
829 |
+ |
\begin{equation} |
830 |
+ |
hZ = hX + hY + \frac{{h^2 }}{2}[X,Y] + \frac{{h^3 }}{2}\left( |
831 |
+ |
{[X,[X,Y]] + [Y,[Y,X]]} \right) + \ldots . |
832 |
+ |
\end{equation} |
833 |
+ |
Here, $[X,Y]$ is the commutators of operator $X$ and $Y$ given by |
834 |
|
\[ |
835 |
< |
L(\ddot x) = p^2 L(x) - px(0) - \dot x(0) |
835 |
> |
[X,Y] = XY - YX . |
836 |
|
\] |
837 |
< |
|
837 |
> |
Applying the Baker-Campbell-Hausdorff formula\cite{Varadarajan1974} |
838 |
> |
to the Sprang splitting, we can obtain |
839 |
> |
\begin{eqnarray*} |
840 |
> |
\exp (h X/2)\exp (h Y)\exp (h X/2) & = & \exp (h X + h Y + h^2 [X,Y]/4 + h^2 [Y,X]/4 \\ |
841 |
> |
& & \mbox{} + h^2 [X,X]/8 + h^2 [Y,Y]/8 \\ |
842 |
> |
& & \mbox{} + h^3 [Y,[Y,X]]/12 - h^3[X,[X,Y]]/24 + \ldots ) |
843 |
> |
\end{eqnarray*} |
844 |
> |
Since \[ [X,Y] + [Y,X] = 0\] and \[ [X,X] = 0,\] the dominant local |
845 |
> |
error of Spring splitting is proportional to $h^3$. The same |
846 |
> |
procedure can be applied to a general splitting, of the form |
847 |
> |
\begin{equation} |
848 |
> |
\varphi _{b_m h}^2 \circ \varphi _{a_m h}^1 \circ \varphi _{b_{m - |
849 |
> |
1} h}^2 \circ \ldots \circ \varphi _{a_1 h}^1 . |
850 |
> |
\end{equation} |
851 |
> |
A careful choice of coefficient $a_1 \ldots b_m$ will lead to higher |
852 |
> |
order methods. Yoshida proposed an elegant way to compose higher |
853 |
> |
order methods based on symmetric splitting\cite{Yoshida1990}. Given |
854 |
> |
a symmetric second order base method $ \varphi _h^{(2)} $, a |
855 |
> |
fourth-order symmetric method can be constructed by composing, |
856 |
|
\[ |
857 |
< |
L\left( {\int_0^t {g(t - \tau )h(\tau )d\tau } } \right) = G(p)H(p) |
857 |
> |
\varphi _h^{(4)} = \varphi _{\alpha h}^{(2)} \circ \varphi _{\beta |
858 |
> |
h}^{(2)} \circ \varphi _{\alpha h}^{(2)} |
859 |
|
\] |
860 |
< |
|
861 |
< |
Some relatively important transformation, |
860 |
> |
where $ \alpha = - \frac{{2^{1/3} }}{{2 - 2^{1/3} }}$ and $ \beta |
861 |
> |
= \frac{{2^{1/3} }}{{2 - 2^{1/3} }}$. Moreover, a symmetric |
862 |
> |
integrator $ \varphi _h^{(2n + 2)}$ can be composed by |
863 |
> |
\begin{equation} |
864 |
> |
\varphi _h^{(2n + 2)} = \varphi _{\alpha h}^{(2n)} \circ \varphi |
865 |
> |
_{\beta h}^{(2n)} \circ \varphi _{\alpha h}^{(2n)}, |
866 |
> |
\end{equation} |
867 |
> |
if the weights are chosen as |
868 |
|
\[ |
869 |
< |
L(\cos at) = \frac{p}{{p^2 + a^2 }} |
869 |
> |
\alpha = - \frac{{2^{1/(2n + 1)} }}{{2 - 2^{1/(2n + 1)} }},\beta = |
870 |
> |
\frac{{2^{1/(2n + 1)} }}{{2 - 2^{1/(2n + 1)} }} . |
871 |
|
\] |
872 |
|
|
873 |
< |
\[ |
455 |
< |
L(\sin at) = \frac{a}{{p^2 + a^2 }} |
456 |
< |
\] |
873 |
> |
\section{\label{introSection:molecularDynamics}Molecular Dynamics} |
874 |
|
|
875 |
+ |
As one of the principal tools of molecular modeling, Molecular |
876 |
+ |
dynamics has proven to be a powerful tool for studying the functions |
877 |
+ |
of biological systems, providing structural, thermodynamic and |
878 |
+ |
dynamical information. The basic idea of molecular dynamics is that |
879 |
+ |
macroscopic properties are related to microscopic behavior and |
880 |
+ |
microscopic behavior can be calculated from the trajectories in |
881 |
+ |
simulations. For instance, instantaneous temperature of an |
882 |
+ |
Hamiltonian system of $N$ particle can be measured by |
883 |
|
\[ |
884 |
< |
L(1) = \frac{1}{p} |
884 |
> |
T = \sum\limits_{i = 1}^N {\frac{{m_i v_i^2 }}{{fk_B }}} |
885 |
|
\] |
886 |
+ |
where $m_i$ and $v_i$ are the mass and velocity of $i$th particle |
887 |
+ |
respectively, $f$ is the number of degrees of freedom, and $k_B$ is |
888 |
+ |
the boltzman constant. |
889 |
|
|
890 |
< |
First, the bath coordinates, |
891 |
< |
\[ |
892 |
< |
p^2 L(x_\alpha ) - px_\alpha (0) - \dot x_\alpha (0) = - \omega |
893 |
< |
_\alpha ^2 L(x_\alpha ) + \frac{{g_\alpha }}{{\omega _\alpha |
894 |
< |
}}L(x) |
890 |
> |
A typical molecular dynamics run consists of three essential steps: |
891 |
> |
\begin{enumerate} |
892 |
> |
\item Initialization |
893 |
> |
\begin{enumerate} |
894 |
> |
\item Preliminary preparation |
895 |
> |
\item Minimization |
896 |
> |
\item Heating |
897 |
> |
\item Equilibration |
898 |
> |
\end{enumerate} |
899 |
> |
\item Production |
900 |
> |
\item Analysis |
901 |
> |
\end{enumerate} |
902 |
> |
These three individual steps will be covered in the following |
903 |
> |
sections. Sec.~\ref{introSec:initialSystemSettings} deals with the |
904 |
> |
initialization of a simulation. Sec.~\ref{introSection:production} |
905 |
> |
will discusse issues in production run. |
906 |
> |
Sec.~\ref{introSection:Analysis} provides the theoretical tools for |
907 |
> |
trajectory analysis. |
908 |
> |
|
909 |
> |
\subsection{\label{introSec:initialSystemSettings}Initialization} |
910 |
> |
|
911 |
> |
\subsubsection{\textbf{Preliminary preparation}} |
912 |
> |
|
913 |
> |
When selecting the starting structure of a molecule for molecular |
914 |
> |
simulation, one may retrieve its Cartesian coordinates from public |
915 |
> |
databases, such as RCSB Protein Data Bank \textit{etc}. Although |
916 |
> |
thousands of crystal structures of molecules are discovered every |
917 |
> |
year, many more remain unknown due to the difficulties of |
918 |
> |
purification and crystallization. Even for molecules with known |
919 |
> |
structure, some important information is missing. For example, a |
920 |
> |
missing hydrogen atom which acts as donor in hydrogen bonding must |
921 |
> |
be added. Moreover, in order to include electrostatic interaction, |
922 |
> |
one may need to specify the partial charges for individual atoms. |
923 |
> |
Under some circumstances, we may even need to prepare the system in |
924 |
> |
a special configuration. For instance, when studying transport |
925 |
> |
phenomenon in membrane systems, we may prepare the lipids in a |
926 |
> |
bilayer structure instead of placing lipids randomly in solvent, |
927 |
> |
since we are not interested in the slow self-aggregation process. |
928 |
> |
|
929 |
> |
\subsubsection{\textbf{Minimization}} |
930 |
> |
|
931 |
> |
It is quite possible that some of molecules in the system from |
932 |
> |
preliminary preparation may be overlapping with each other. This |
933 |
> |
close proximity leads to high initial potential energy which |
934 |
> |
consequently jeopardizes any molecular dynamics simulations. To |
935 |
> |
remove these steric overlaps, one typically performs energy |
936 |
> |
minimization to find a more reasonable conformation. Several energy |
937 |
> |
minimization methods have been developed to exploit the energy |
938 |
> |
surface and to locate the local minimum. While converging slowly |
939 |
> |
near the minimum, steepest descent method is extremely robust when |
940 |
> |
systems are strongly anharmonic. Thus, it is often used to refine |
941 |
> |
structure from crystallographic data. Relied on the gradient or |
942 |
> |
hessian, advanced methods like Newton-Raphson converge rapidly to a |
943 |
> |
local minimum, but become unstable if the energy surface is far from |
944 |
> |
quadratic. Another factor that must be taken into account, when |
945 |
> |
choosing energy minimization method, is the size of the system. |
946 |
> |
Steepest descent and conjugate gradient can deal with models of any |
947 |
> |
size. Because of the limits on computer memory to store the hessian |
948 |
> |
matrix and the computing power needed to diagonalized these |
949 |
> |
matrices, most Newton-Raphson methods can not be used with very |
950 |
> |
large systems. |
951 |
> |
|
952 |
> |
\subsubsection{\textbf{Heating}} |
953 |
> |
|
954 |
> |
Typically, Heating is performed by assigning random velocities |
955 |
> |
according to a Maxwell-Boltzman distribution for a desired |
956 |
> |
temperature. Beginning at a lower temperature and gradually |
957 |
> |
increasing the temperature by assigning larger random velocities, we |
958 |
> |
end up with setting the temperature of the system to a final |
959 |
> |
temperature at which the simulation will be conducted. In heating |
960 |
> |
phase, we should also keep the system from drifting or rotating as a |
961 |
> |
whole. To do this, the net linear momentum and angular momentum of |
962 |
> |
the system is shifted to zero after each resampling from the Maxwell |
963 |
> |
-Boltzman distribution. |
964 |
> |
|
965 |
> |
\subsubsection{\textbf{Equilibration}} |
966 |
> |
|
967 |
> |
The purpose of equilibration is to allow the system to evolve |
968 |
> |
spontaneously for a period of time and reach equilibrium. The |
969 |
> |
procedure is continued until various statistical properties, such as |
970 |
> |
temperature, pressure, energy, volume and other structural |
971 |
> |
properties \textit{etc}, become independent of time. Strictly |
972 |
> |
speaking, minimization and heating are not necessary, provided the |
973 |
> |
equilibration process is long enough. However, these steps can serve |
974 |
> |
as a means to arrive at an equilibrated structure in an effective |
975 |
> |
way. |
976 |
> |
|
977 |
> |
\subsection{\label{introSection:production}Production} |
978 |
> |
|
979 |
> |
The production run is the most important step of the simulation, in |
980 |
> |
which the equilibrated structure is used as a starting point and the |
981 |
> |
motions of the molecules are collected for later analysis. In order |
982 |
> |
to capture the macroscopic properties of the system, the molecular |
983 |
> |
dynamics simulation must be performed by sampling correctly and |
984 |
> |
efficiently from the relevant thermodynamic ensemble. |
985 |
> |
|
986 |
> |
The most expensive part of a molecular dynamics simulation is the |
987 |
> |
calculation of non-bonded forces, such as van der Waals force and |
988 |
> |
Coulombic forces \textit{etc}. For a system of $N$ particles, the |
989 |
> |
complexity of the algorithm for pair-wise interactions is $O(N^2 )$, |
990 |
> |
which making large simulations prohibitive in the absence of any |
991 |
> |
algorithmic tricks. |
992 |
> |
|
993 |
> |
A natural approach to avoid system size issues is to represent the |
994 |
> |
bulk behavior by a finite number of the particles. However, this |
995 |
> |
approach will suffer from the surface effect at the edges of the |
996 |
> |
simulation. To offset this, \textit{Periodic boundary conditions} |
997 |
> |
(see Fig.~\ref{introFig:pbc}) is developed to simulate bulk |
998 |
> |
properties with a relatively small number of particles. In this |
999 |
> |
method, the simulation box is replicated throughout space to form an |
1000 |
> |
infinite lattice. During the simulation, when a particle moves in |
1001 |
> |
the primary cell, its image in other cells move in exactly the same |
1002 |
> |
direction with exactly the same orientation. Thus, as a particle |
1003 |
> |
leaves the primary cell, one of its images will enter through the |
1004 |
> |
opposite face. |
1005 |
> |
\begin{figure} |
1006 |
> |
\centering |
1007 |
> |
\includegraphics[width=\linewidth]{pbc.eps} |
1008 |
> |
\caption[An illustration of periodic boundary conditions]{A 2-D |
1009 |
> |
illustration of periodic boundary conditions. As one particle leaves |
1010 |
> |
the left of the simulation box, an image of it enters the right.} |
1011 |
> |
\label{introFig:pbc} |
1012 |
> |
\end{figure} |
1013 |
> |
|
1014 |
> |
%cutoff and minimum image convention |
1015 |
> |
Another important technique to improve the efficiency of force |
1016 |
> |
evaluation is to apply spherical cutoff where particles farther than |
1017 |
> |
a predetermined distance are not included in the calculation |
1018 |
> |
\cite{Frenkel1996}. The use of a cutoff radius will cause a |
1019 |
> |
discontinuity in the potential energy curve. Fortunately, one can |
1020 |
> |
shift simple radial potential to ensure the potential curve go |
1021 |
> |
smoothly to zero at the cutoff radius. The cutoff strategy works |
1022 |
> |
well for Lennard-Jones interaction because of its short range |
1023 |
> |
nature. However, simply truncating the electrostatic interaction |
1024 |
> |
with the use of cutoffs has been shown to lead to severe artifacts |
1025 |
> |
in simulations. The Ewald summation, in which the slowly decaying |
1026 |
> |
Coulomb potential is transformed into direct and reciprocal sums |
1027 |
> |
with rapid and absolute convergence, has proved to minimize the |
1028 |
> |
periodicity artifacts in liquid simulations. Taking the advantages |
1029 |
> |
of the fast Fourier transform (FFT) for calculating discrete Fourier |
1030 |
> |
transforms, the particle mesh-based |
1031 |
> |
methods\cite{Hockney1981,Shimada1993, Luty1994} are accelerated from |
1032 |
> |
$O(N^{3/2})$ to $O(N logN)$. An alternative approach is the |
1033 |
> |
\emph{fast multipole method}\cite{Greengard1987, Greengard1994}, |
1034 |
> |
which treats Coulombic interactions exactly at short range, and |
1035 |
> |
approximate the potential at long range through multipolar |
1036 |
> |
expansion. In spite of their wide acceptance at the molecular |
1037 |
> |
simulation community, these two methods are difficult to implement |
1038 |
> |
correctly and efficiently. Instead, we use a damped and |
1039 |
> |
charge-neutralized Coulomb potential method developed by Wolf and |
1040 |
> |
his coworkers\cite{Wolf1999}. The shifted Coulomb potential for |
1041 |
> |
particle $i$ and particle $j$ at distance $r_{rj}$ is given by: |
1042 |
> |
\begin{equation} |
1043 |
> |
V(r_{ij})= \frac{q_i q_j \textrm{erfc}(\alpha |
1044 |
> |
r_{ij})}{r_{ij}}-\lim_{r_{ij}\rightarrow |
1045 |
> |
R_\textrm{c}}\left\{\frac{q_iq_j \textrm{erfc}(\alpha |
1046 |
> |
r_{ij})}{r_{ij}}\right\}. \label{introEquation:shiftedCoulomb} |
1047 |
> |
\end{equation} |
1048 |
> |
where $\alpha$ is the convergence parameter. Due to the lack of |
1049 |
> |
inherent periodicity and rapid convergence,this method is extremely |
1050 |
> |
efficient and easy to implement. |
1051 |
> |
\begin{figure} |
1052 |
> |
\centering |
1053 |
> |
\includegraphics[width=\linewidth]{shifted_coulomb.eps} |
1054 |
> |
\caption[An illustration of shifted Coulomb potential]{An |
1055 |
> |
illustration of shifted Coulomb potential.} |
1056 |
> |
\label{introFigure:shiftedCoulomb} |
1057 |
> |
\end{figure} |
1058 |
> |
|
1059 |
> |
%multiple time step |
1060 |
> |
|
1061 |
> |
\subsection{\label{introSection:Analysis} Analysis} |
1062 |
> |
|
1063 |
> |
Recently, advanced visualization technique have become applied to |
1064 |
> |
monitor the motions of molecules. Although the dynamics of the |
1065 |
> |
system can be described qualitatively from animation, quantitative |
1066 |
> |
trajectory analysis are more useful. According to the principles of |
1067 |
> |
Statistical Mechanics, Sec.~\ref{introSection:statisticalMechanics}, |
1068 |
> |
one can compute thermodynamic properties, analyze fluctuations of |
1069 |
> |
structural parameters, and investigate time-dependent processes of |
1070 |
> |
the molecule from the trajectories. |
1071 |
> |
|
1072 |
> |
\subsubsection{\label{introSection:thermodynamicsProperties}\textbf{Thermodynamic Properties}} |
1073 |
> |
|
1074 |
> |
Thermodynamic properties, which can be expressed in terms of some |
1075 |
> |
function of the coordinates and momenta of all particles in the |
1076 |
> |
system, can be directly computed from molecular dynamics. The usual |
1077 |
> |
way to measure the pressure is based on virial theorem of Clausius |
1078 |
> |
which states that the virial is equal to $-3Nk_BT$. For a system |
1079 |
> |
with forces between particles, the total virial, $W$, contains the |
1080 |
> |
contribution from external pressure and interaction between the |
1081 |
> |
particles: |
1082 |
> |
\[ |
1083 |
> |
W = - 3PV + \left\langle {\sum\limits_{i < j} {r{}_{ij} \cdot |
1084 |
> |
f_{ij} } } \right\rangle |
1085 |
|
\] |
1086 |
+ |
where $f_{ij}$ is the force between particle $i$ and $j$ at a |
1087 |
+ |
distance $r_{ij}$. Thus, the expression for the pressure is given |
1088 |
+ |
by: |
1089 |
+ |
\begin{equation} |
1090 |
+ |
P = \frac{{Nk_B T}}{V} - \frac{1}{{3V}}\left\langle {\sum\limits_{i |
1091 |
+ |
< j} {r{}_{ij} \cdot f_{ij} } } \right\rangle |
1092 |
+ |
\end{equation} |
1093 |
+ |
|
1094 |
+ |
\subsubsection{\label{introSection:structuralProperties}\textbf{Structural Properties}} |
1095 |
+ |
|
1096 |
+ |
Structural Properties of a simple fluid can be described by a set of |
1097 |
+ |
distribution functions. Among these functions,the \emph{pair |
1098 |
+ |
distribution function}, also known as \emph{radial distribution |
1099 |
+ |
function}, is of most fundamental importance to liquid theory. |
1100 |
+ |
Experimentally, pair distribution function can be gathered by |
1101 |
+ |
Fourier transforming raw data from a series of neutron diffraction |
1102 |
+ |
experiments and integrating over the surface factor |
1103 |
+ |
\cite{Powles1973}. The experimental results can serve as a criterion |
1104 |
+ |
to justify the correctness of a liquid model. Moreover, various |
1105 |
+ |
equilibrium thermodynamic and structural properties can also be |
1106 |
+ |
expressed in terms of radial distribution function \cite{Allen1987}. |
1107 |
+ |
|
1108 |
+ |
The pair distribution functions $g(r)$ gives the probability that a |
1109 |
+ |
particle $i$ will be located at a distance $r$ from a another |
1110 |
+ |
particle $j$ in the system |
1111 |
|
\[ |
1112 |
< |
L(x_\alpha ) = \frac{{\frac{{g_\alpha }}{{\omega _\alpha }}L(x) + |
1113 |
< |
px_\alpha (0) + \dot x_\alpha (0)}}{{p^2 + \omega _\alpha ^2 }} |
1112 |
> |
g(r) = \frac{V}{{N^2 }}\left\langle {\sum\limits_i {\sum\limits_{j |
1113 |
> |
\ne i} {\delta (r - r_{ij} )} } } \right\rangle = \frac{\rho |
1114 |
> |
(r)}{\rho}. |
1115 |
|
\] |
1116 |
< |
Then, the system coordinates, |
1117 |
< |
\begin{align} |
1118 |
< |
mL(\ddot x) &= - \frac{1}{p}\frac{{\partial W(x)}}{{\partial x}} - |
1119 |
< |
\sum\limits_{\alpha = 1}^N {\left\{ {\frac{{\frac{{g_\alpha |
1120 |
< |
}}{{\omega _\alpha }}L(x) + px_\alpha (0) + \dot x_\alpha |
1121 |
< |
(0)}}{{p^2 + \omega _\alpha ^2 }} - \frac{{g_\alpha ^2 }}{{m_\alpha |
1122 |
< |
}}\omega _\alpha ^2 L(x)} \right\}} |
1123 |
< |
% |
1124 |
< |
&= - \frac{1}{p}\frac{{\partial W(x)}}{{\partial x}} - |
1125 |
< |
\sum\limits_{\alpha = 1}^N {\left\{ { - \frac{{g_\alpha ^2 }}{{m_\alpha \omega _\alpha ^2 }}\frac{p}{{p^2 + \omega _\alpha ^2 }}pL(x) |
482 |
< |
- \frac{p}{{p^2 + \omega _\alpha ^2 }}g_\alpha x_\alpha (0) |
483 |
< |
- \frac{1}{{p^2 + \omega _\alpha ^2 }}g_\alpha \dot x_\alpha (0)} \right\}} |
484 |
< |
\end{align} |
485 |
< |
Then, the inverse transform, |
1116 |
> |
Note that the delta function can be replaced by a histogram in |
1117 |
> |
computer simulation. Figure |
1118 |
> |
\ref{introFigure:pairDistributionFunction} shows a typical pair |
1119 |
> |
distribution function for the liquid argon system. The occurrence of |
1120 |
> |
several peaks in the plot of $g(r)$ suggests that it is more likely |
1121 |
> |
to find particles at certain radial values than at others. This is a |
1122 |
> |
result of the attractive interaction at such distances. Because of |
1123 |
> |
the strong repulsive forces at short distance, the probability of |
1124 |
> |
locating particles at distances less than about 3.7{\AA} from each |
1125 |
> |
other is essentially zero. |
1126 |
|
|
1127 |
< |
\begin{align} |
1128 |
< |
m\ddot x &= - \frac{{\partial W(x)}}{{\partial x}} - |
1129 |
< |
\sum\limits_{\alpha = 1}^N {\left\{ {\left( { - \frac{{g_\alpha ^2 |
1130 |
< |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\int_0^t {\cos (\omega |
1131 |
< |
_\alpha t)\dot x(t - \tau )d\tau - \left[ {g_\alpha x_\alpha (0) |
1132 |
< |
- \frac{{g_\alpha }}{{m_\alpha \omega _\alpha }}} \right]\cos |
1133 |
< |
(\omega _\alpha t) - \frac{{g_\alpha \dot x_\alpha (0)}}{{\omega |
494 |
< |
_\alpha }}\sin (\omega _\alpha t)} } \right\}} |
495 |
< |
% |
496 |
< |
&= - \frac{{\partial W(x)}}{{\partial x}} - \int_0^t |
497 |
< |
{\sum\limits_{\alpha = 1}^N {\left( { - \frac{{g_\alpha ^2 |
498 |
< |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\cos (\omega _\alpha |
499 |
< |
t)\dot x(t - \tau )d} \tau } + \sum\limits_{\alpha = 1}^N {\left\{ |
500 |
< |
{\left[ {g_\alpha x_\alpha (0) - \frac{{g_\alpha }}{{m_\alpha |
501 |
< |
\omega _\alpha }}} \right]\cos (\omega _\alpha t) + |
502 |
< |
\frac{{g_\alpha \dot x_\alpha (0)}}{{\omega _\alpha }}\sin |
503 |
< |
(\omega _\alpha t)} \right\}} |
504 |
< |
\end{align} |
1127 |
> |
%\begin{figure} |
1128 |
> |
%\centering |
1129 |
> |
%\includegraphics[width=\linewidth]{pdf.eps} |
1130 |
> |
%\caption[Pair distribution function for the liquid argon |
1131 |
> |
%]{Pair distribution function for the liquid argon} |
1132 |
> |
%\label{introFigure:pairDistributionFunction} |
1133 |
> |
%\end{figure} |
1134 |
|
|
1135 |
+ |
\subsubsection{\label{introSection:timeDependentProperties}\textbf{Time-dependent |
1136 |
+ |
Properties}} |
1137 |
+ |
|
1138 |
+ |
Time-dependent properties are usually calculated using \emph{time |
1139 |
+ |
correlation functions}, which correlate random variables $A$ and $B$ |
1140 |
+ |
at two different times, |
1141 |
|
\begin{equation} |
1142 |
< |
m\ddot x = - \frac{{\partial W}}{{\partial x}} - \int_0^t {\xi |
1143 |
< |
(t)\dot x(t - \tau )d\tau } + R(t) |
509 |
< |
\label{introEuqation:GeneralizedLangevinDynamics} |
1142 |
> |
C_{AB} (t) = \left\langle {A(t)B(0)} \right\rangle. |
1143 |
> |
\label{introEquation:timeCorrelationFunction} |
1144 |
|
\end{equation} |
1145 |
< |
%where $ {\xi (t)}$ is friction kernel, $R(t)$ is random force and |
1146 |
< |
%$W$ is the potential of mean force. $W(x) = - kT\ln p(x)$ |
1145 |
> |
If $A$ and $B$ refer to same variable, this kind of correlation |
1146 |
> |
function is called an \emph{autocorrelation function}. One example |
1147 |
> |
of an auto correlation function is the velocity auto-correlation |
1148 |
> |
function which is directly related to transport properties of |
1149 |
> |
molecular liquids: |
1150 |
|
\[ |
1151 |
< |
\xi (t) = \sum\limits_{\alpha = 1}^N {\left( { - \frac{{g_\alpha ^2 |
1152 |
< |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\cos (\omega _\alpha t)} |
1151 |
> |
D = \frac{1}{3}\int\limits_0^\infty {\left\langle {v(t) \cdot v(0)} |
1152 |
> |
\right\rangle } dt |
1153 |
|
\] |
1154 |
< |
For an infinite harmonic bath, we can use the spectral density and |
1155 |
< |
an integral over frequencies. |
1154 |
> |
where $D$ is diffusion constant. Unlike the velocity autocorrelation |
1155 |
> |
function, which is averaging over time origins and over all the |
1156 |
> |
atoms, the dipole autocorrelation functions are calculated for the |
1157 |
> |
entire system. The dipole autocorrelation function is given by: |
1158 |
> |
\[ |
1159 |
> |
c_{dipole} = \left\langle {u_{tot} (t) \cdot u_{tot} (t)} |
1160 |
> |
\right\rangle |
1161 |
> |
\] |
1162 |
> |
Here $u_{tot}$ is the net dipole of the entire system and is given |
1163 |
> |
by |
1164 |
> |
\[ |
1165 |
> |
u_{tot} (t) = \sum\limits_i {u_i (t)} |
1166 |
> |
\] |
1167 |
> |
In principle, many time correlation functions can be related with |
1168 |
> |
Fourier transforms of the infrared, Raman, and inelastic neutron |
1169 |
> |
scattering spectra of molecular liquids. In practice, one can |
1170 |
> |
extract the IR spectrum from the intensity of dipole fluctuation at |
1171 |
> |
each frequency using the following relationship: |
1172 |
> |
\[ |
1173 |
> |
\hat c_{dipole} (v) = \int_{ - \infty }^\infty {c_{dipole} (t)e^{ - |
1174 |
> |
i2\pi vt} dt} |
1175 |
> |
\] |
1176 |
|
|
1177 |
+ |
\section{\label{introSection:rigidBody}Dynamics of Rigid Bodies} |
1178 |
+ |
|
1179 |
+ |
Rigid bodies are frequently involved in the modeling of different |
1180 |
+ |
areas, from engineering, physics, to chemistry. For example, |
1181 |
+ |
missiles and vehicle are usually modeled by rigid bodies. The |
1182 |
+ |
movement of the objects in 3D gaming engine or other physics |
1183 |
+ |
simulator is governed by rigid body dynamics. In molecular |
1184 |
+ |
simulations, rigid bodies are used to simplify protein-protein |
1185 |
+ |
docking studies\cite{Gray2003}. |
1186 |
+ |
|
1187 |
+ |
It is very important to develop stable and efficient methods to |
1188 |
+ |
integrate the equations of motion for orientational degrees of |
1189 |
+ |
freedom. Euler angles are the natural choice to describe the |
1190 |
+ |
rotational degrees of freedom. However, due to $\frac {1}{sin |
1191 |
+ |
\theta}$ singularities, the numerical integration of corresponding |
1192 |
+ |
equations of motion is very inefficient and inaccurate. Although an |
1193 |
+ |
alternative integrator using multiple sets of Euler angles can |
1194 |
+ |
overcome this difficulty\cite{Barojas1973}, the computational |
1195 |
+ |
penalty and the loss of angular momentum conservation still remain. |
1196 |
+ |
A singularity-free representation utilizing quaternions was |
1197 |
+ |
developed by Evans in 1977\cite{Evans1977}. Unfortunately, this |
1198 |
+ |
approach uses a nonseparable Hamiltonian resulting from the |
1199 |
+ |
quaternion representation, which prevents the symplectic algorithm |
1200 |
+ |
to be utilized. Another different approach is to apply holonomic |
1201 |
+ |
constraints to the atoms belonging to the rigid body. Each atom |
1202 |
+ |
moves independently under the normal forces deriving from potential |
1203 |
+ |
energy and constraint forces which are used to guarantee the |
1204 |
+ |
rigidness. However, due to their iterative nature, the SHAKE and |
1205 |
+ |
Rattle algorithms also converge very slowly when the number of |
1206 |
+ |
constraints increases\cite{Ryckaert1977, Andersen1983}. |
1207 |
+ |
|
1208 |
+ |
A break-through in geometric literature suggests that, in order to |
1209 |
+ |
develop a long-term integration scheme, one should preserve the |
1210 |
+ |
symplectic structure of the flow. By introducing a conjugate |
1211 |
+ |
momentum to the rotation matrix $Q$ and re-formulating Hamiltonian's |
1212 |
+ |
equation, a symplectic integrator, RSHAKE\cite{Kol1997}, was |
1213 |
+ |
proposed to evolve the Hamiltonian system in a constraint manifold |
1214 |
+ |
by iteratively satisfying the orthogonality constraint $Q^T Q = 1$. |
1215 |
+ |
An alternative method using the quaternion representation was |
1216 |
+ |
developed by Omelyan\cite{Omelyan1998}. However, both of these |
1217 |
+ |
methods are iterative and inefficient. In this section, we descibe a |
1218 |
+ |
symplectic Lie-Poisson integrator for rigid body developed by |
1219 |
+ |
Dullweber and his coworkers\cite{Dullweber1997} in depth. |
1220 |
+ |
|
1221 |
+ |
\subsection{\label{introSection:constrainedHamiltonianRB}Constrained Hamiltonian for Rigid Bodies} |
1222 |
+ |
The motion of a rigid body is Hamiltonian with the Hamiltonian |
1223 |
+ |
function |
1224 |
+ |
\begin{equation} |
1225 |
+ |
H = \frac{1}{2}(p^T m^{ - 1} p) + \frac{1}{2}tr(PJ^{ - 1} P) + |
1226 |
+ |
V(q,Q) + \frac{1}{2}tr[(QQ^T - 1)\Lambda ]. |
1227 |
+ |
\label{introEquation:RBHamiltonian} |
1228 |
+ |
\end{equation} |
1229 |
+ |
Here, $q$ and $Q$ are the position and rotation matrix for the |
1230 |
+ |
rigid-body, $p$ and $P$ are conjugate momenta to $q$ and $Q$ , and |
1231 |
+ |
$J$, a diagonal matrix, is defined by |
1232 |
|
\[ |
1233 |
< |
R(t) = \sum\limits_{\alpha = 1}^N {\left( {g_\alpha x_\alpha (0) |
522 |
< |
- \frac{{g_\alpha ^2 }}{{m_\alpha \omega _\alpha ^2 }}x(0)} |
523 |
< |
\right)\cos (\omega _\alpha t)} + \frac{{\dot x_\alpha |
524 |
< |
(0)}}{{\omega _\alpha }}\sin (\omega _\alpha t) |
1233 |
> |
I_{ii}^{ - 1} = \frac{1}{2}\sum\limits_{i \ne j} {J_{jj}^{ - 1} } |
1234 |
|
\] |
1235 |
< |
The random forces depend only on initial conditions. |
1235 |
> |
where $I_{ii}$ is the diagonal element of the inertia tensor. This |
1236 |
> |
constrained Hamiltonian equation is subjected to a holonomic |
1237 |
> |
constraint, |
1238 |
> |
\begin{equation} |
1239 |
> |
Q^T Q = 1, \label{introEquation:orthogonalConstraint} |
1240 |
> |
\end{equation} |
1241 |
> |
which is used to ensure rotation matrix's unitarity. Differentiating |
1242 |
> |
\ref{introEquation:orthogonalConstraint} and using Equation |
1243 |
> |
\ref{introEquation:RBMotionMomentum}, one may obtain, |
1244 |
> |
\begin{equation} |
1245 |
> |
Q^T PJ^{ - 1} + J^{ - 1} P^T Q = 0 . \\ |
1246 |
> |
\label{introEquation:RBFirstOrderConstraint} |
1247 |
> |
\end{equation} |
1248 |
|
|
1249 |
< |
\subsubsection{\label{introSection:secondFluctuationDissipation}The Second Fluctuation Dissipation Theorem} |
1250 |
< |
So we can define a new set of coordinates, |
1249 |
> |
Using Equation (\ref{introEquation:motionHamiltonianCoordinate}, |
1250 |
> |
\ref{introEquation:motionHamiltonianMomentum}), one can write down |
1251 |
> |
the equations of motion, |
1252 |
> |
|
1253 |
> |
\begin{eqnarray} |
1254 |
> |
\frac{{dq}}{{dt}} & = & \frac{p}{m} \label{introEquation:RBMotionPosition}\\ |
1255 |
> |
\frac{{dp}}{{dt}} & = & - \nabla _q V(q,Q) \label{introEquation:RBMotionMomentum}\\ |
1256 |
> |
\frac{{dQ}}{{dt}} & = & PJ^{ - 1} \label{introEquation:RBMotionRotation}\\ |
1257 |
> |
\frac{{dP}}{{dt}} & = & - \nabla _Q V(q,Q) - 2Q\Lambda . \label{introEquation:RBMotionP} |
1258 |
> |
\end{eqnarray} |
1259 |
> |
|
1260 |
> |
In general, there are two ways to satisfy the holonomic constraints. |
1261 |
> |
We can use a constraint force provided by a Lagrange multiplier on |
1262 |
> |
the normal manifold to keep the motion on constraint space. Or we |
1263 |
> |
can simply evolve the system on the constraint manifold. These two |
1264 |
> |
methods have been proved to be equivalent. The holonomic constraint |
1265 |
> |
and equations of motions define a constraint manifold for rigid |
1266 |
> |
bodies |
1267 |
|
\[ |
1268 |
< |
q_\alpha (t) = x_\alpha (t) - \frac{1}{{m_\alpha \omega _\alpha |
1269 |
< |
^2 }}x(0) |
1268 |
> |
M = \left\{ {(Q,P):Q^T Q = 1,Q^T PJ^{ - 1} + J^{ - 1} P^T Q = 0} |
1269 |
> |
\right\}. |
1270 |
|
\] |
1271 |
< |
This makes |
1271 |
> |
|
1272 |
> |
Unfortunately, this constraint manifold is not the cotangent bundle |
1273 |
> |
$T_{\star}SO(3)$. However, it turns out that under symplectic |
1274 |
> |
transformation, the cotangent space and the phase space are |
1275 |
> |
diffeomorphic. By introducing |
1276 |
|
\[ |
1277 |
< |
R(t) = \sum\limits_{\alpha = 1}^N {g_\alpha q_\alpha (t)} |
1277 |
> |
\tilde Q = Q,\tilde P = \frac{1}{2}\left( {P - QP^T Q} \right), |
1278 |
|
\] |
1279 |
< |
And since the $q$ coordinates are harmonic oscillators, |
1279 |
> |
the mechanical system subject to a holonomic constraint manifold $M$ |
1280 |
> |
can be re-formulated as a Hamiltonian system on the cotangent space |
1281 |
|
\[ |
1282 |
< |
\begin{array}{l} |
1283 |
< |
\left\langle {q_\alpha (t)q_\alpha (0)} \right\rangle = \left\langle {q_\alpha ^2 (0)} \right\rangle \cos (\omega _\alpha t) \\ |
542 |
< |
\left\langle {q_\alpha (t)q_\beta (0)} \right\rangle = \delta _{\alpha \beta } \left\langle {q_\alpha (t)q_\alpha (0)} \right\rangle \\ |
543 |
< |
\end{array} |
1282 |
> |
T^* SO(3) = \left\{ {(\tilde Q,\tilde P):\tilde Q^T \tilde Q = |
1283 |
> |
1,\tilde Q^T \tilde PJ^{ - 1} + J^{ - 1} P^T \tilde Q = 0} \right\} |
1284 |
|
\] |
1285 |
|
|
1286 |
< |
\begin{align} |
1287 |
< |
\left\langle {R(t)R(0)} \right\rangle &= \sum\limits_\alpha |
1288 |
< |
{\sum\limits_\beta {g_\alpha g_\beta \left\langle {q_\alpha |
1289 |
< |
(t)q_\beta (0)} \right\rangle } } |
1290 |
< |
% |
1291 |
< |
&= \sum\limits_\alpha {g_\alpha ^2 \left\langle {q_\alpha ^2 (0)} |
1292 |
< |
\right\rangle \cos (\omega _\alpha t)} |
1293 |
< |
% |
1294 |
< |
&= kT\xi (t) |
1295 |
< |
\end{align} |
1286 |
> |
For a body fixed vector $X_i$ with respect to the center of mass of |
1287 |
> |
the rigid body, its corresponding lab fixed vector $X_0^{lab}$ is |
1288 |
> |
given as |
1289 |
> |
\begin{equation} |
1290 |
> |
X_i^{lab} = Q X_i + q. |
1291 |
> |
\end{equation} |
1292 |
> |
Therefore, potential energy $V(q,Q)$ is defined by |
1293 |
> |
\[ |
1294 |
> |
V(q,Q) = V(Q X_0 + q). |
1295 |
> |
\] |
1296 |
> |
Hence, the force and torque are given by |
1297 |
> |
\[ |
1298 |
> |
\nabla _q V(q,Q) = F(q,Q) = \sum\limits_i {F_i (q,Q)}, |
1299 |
> |
\] |
1300 |
> |
and |
1301 |
> |
\[ |
1302 |
> |
\nabla _Q V(q,Q) = F(q,Q)X_i^t |
1303 |
> |
\] |
1304 |
> |
respectively. |
1305 |
|
|
1306 |
+ |
As a common choice to describe the rotation dynamics of the rigid |
1307 |
+ |
body, the angular momentum on the body fixed frame $\Pi = Q^t P$ is |
1308 |
+ |
introduced to rewrite the equations of motion, |
1309 |
|
\begin{equation} |
1310 |
< |
\xi (t) = \left\langle {R(t)R(0)} \right\rangle |
1311 |
< |
\label{introEquation:secondFluctuationDissipation} |
1310 |
> |
\begin{array}{l} |
1311 |
> |
\mathop \Pi \limits^ \bullet = J^{ - 1} \Pi ^T \Pi + Q^T \sum\limits_i {F_i (q,Q)X_i^T } - \Lambda \\ |
1312 |
> |
\mathop Q\limits^{{\rm{ }} \bullet } = Q\Pi {\rm{ }}J^{ - 1} \\ |
1313 |
> |
\end{array} |
1314 |
> |
\label{introEqaution:RBMotionPI} |
1315 |
|
\end{equation} |
1316 |
+ |
, as well as holonomic constraints, |
1317 |
+ |
\[ |
1318 |
+ |
\begin{array}{l} |
1319 |
+ |
\Pi J^{ - 1} + J^{ - 1} \Pi ^t = 0 \\ |
1320 |
+ |
Q^T Q = 1 \\ |
1321 |
+ |
\end{array} |
1322 |
+ |
\] |
1323 |
|
|
1324 |
< |
\section{\label{introSection:hydroynamics}Hydrodynamics} |
1325 |
< |
|
1326 |
< |
\subsection{\label{introSection:frictionTensor} Friction Tensor} |
1327 |
< |
\subsection{\label{introSection:analyticalApproach}Analytical |
1328 |
< |
Approach} |
1329 |
< |
|
1330 |
< |
\subsection{\label{introSection:approximationApproach}Approximation |
1331 |
< |
Approach} |
1324 |
> |
For a vector $v(v_1 ,v_2 ,v_3 ) \in R^3$ and a matrix $\hat v \in |
1325 |
> |
so(3)^ \star$, the hat-map isomorphism, |
1326 |
> |
\begin{equation} |
1327 |
> |
v(v_1 ,v_2 ,v_3 ) \Leftrightarrow \hat v = \left( |
1328 |
> |
{\begin{array}{*{20}c} |
1329 |
> |
0 & { - v_3 } & {v_2 } \\ |
1330 |
> |
{v_3 } & 0 & { - v_1 } \\ |
1331 |
> |
{ - v_2 } & {v_1 } & 0 \\ |
1332 |
> |
\end{array}} \right), |
1333 |
> |
\label{introEquation:hatmapIsomorphism} |
1334 |
> |
\end{equation} |
1335 |
> |
will let us associate the matrix products with traditional vector |
1336 |
> |
operations |
1337 |
> |
\[ |
1338 |
> |
\hat vu = v \times u |
1339 |
> |
\] |
1340 |
> |
Using \ref{introEqaution:RBMotionPI}, one can construct a skew |
1341 |
> |
matrix, |
1342 |
> |
\begin{equation} |
1343 |
> |
(\mathop \Pi \limits^ \bullet - \mathop \Pi \limits^ {\bullet ^T} |
1344 |
> |
){\rm{ }} = {\rm{ }}(\Pi - \Pi ^T ){\rm{ }}(J^{ - 1} \Pi + \Pi J^{ |
1345 |
> |
- 1} ) + \sum\limits_i {[Q^T F_i (r,Q)X_i^T - X_i F_i (r,Q)^T Q]} - |
1346 |
> |
(\Lambda - \Lambda ^T ) . \label{introEquation:skewMatrixPI} |
1347 |
> |
\end{equation} |
1348 |
> |
Since $\Lambda$ is symmetric, the last term of Equation |
1349 |
> |
\ref{introEquation:skewMatrixPI} is zero, which implies the Lagrange |
1350 |
> |
multiplier $\Lambda$ is absent from the equations of motion. This |
1351 |
> |
unique property eliminates the requirement of iterations which can |
1352 |
> |
not be avoided in other methods\cite{Kol1997, Omelyan1998}. |
1353 |
|
|
1354 |
< |
\subsection{\label{introSection:centersRigidBody}Centers of Rigid |
1355 |
< |
Body} |
1354 |
> |
Applying the hat-map isomorphism, we obtain the equation of motion |
1355 |
> |
for angular momentum on body frame |
1356 |
> |
\begin{equation} |
1357 |
> |
\dot \pi = \pi \times I^{ - 1} \pi + \sum\limits_i {\left( {Q^T |
1358 |
> |
F_i (r,Q)} \right) \times X_i }. |
1359 |
> |
\label{introEquation:bodyAngularMotion} |
1360 |
> |
\end{equation} |
1361 |
> |
In the same manner, the equation of motion for rotation matrix is |
1362 |
> |
given by |
1363 |
> |
\[ |
1364 |
> |
\dot Q = Qskew(I^{ - 1} \pi ) |
1365 |
> |
\] |
1366 |
> |
|
1367 |
> |
\subsection{\label{introSection:SymplecticFreeRB}Symplectic |
1368 |
> |
Lie-Poisson Integrator for Free Rigid Body} |
1369 |
> |
|
1370 |
> |
If there are no external forces exerted on the rigid body, the only |
1371 |
> |
contribution to the rotational motion is from the kinetic energy |
1372 |
> |
(the first term of \ref{introEquation:bodyAngularMotion}). The free |
1373 |
> |
rigid body is an example of a Lie-Poisson system with Hamiltonian |
1374 |
> |
function |
1375 |
> |
\begin{equation} |
1376 |
> |
T^r (\pi ) = T_1 ^r (\pi _1 ) + T_2^r (\pi _2 ) + T_3^r (\pi _3 ) |
1377 |
> |
\label{introEquation:rotationalKineticRB} |
1378 |
> |
\end{equation} |
1379 |
> |
where $T_i^r (\pi _i ) = \frac{{\pi _i ^2 }}{{2I_i }}$ and |
1380 |
> |
Lie-Poisson structure matrix, |
1381 |
> |
\begin{equation} |
1382 |
> |
J(\pi ) = \left( {\begin{array}{*{20}c} |
1383 |
> |
0 & {\pi _3 } & { - \pi _2 } \\ |
1384 |
> |
{ - \pi _3 } & 0 & {\pi _1 } \\ |
1385 |
> |
{\pi _2 } & { - \pi _1 } & 0 \\ |
1386 |
> |
\end{array}} \right) |
1387 |
> |
\end{equation} |
1388 |
> |
Thus, the dynamics of free rigid body is governed by |
1389 |
> |
\begin{equation} |
1390 |
> |
\frac{d}{{dt}}\pi = J(\pi )\nabla _\pi T^r (\pi ) |
1391 |
> |
\end{equation} |
1392 |
> |
|
1393 |
> |
One may notice that each $T_i^r$ in Equation |
1394 |
> |
\ref{introEquation:rotationalKineticRB} can be solved exactly. For |
1395 |
> |
instance, the equations of motion due to $T_1^r$ are given by |
1396 |
> |
\begin{equation} |
1397 |
> |
\frac{d}{{dt}}\pi = R_1 \pi ,\frac{d}{{dt}}Q = QR_1 |
1398 |
> |
\label{introEqaution:RBMotionSingleTerm} |
1399 |
> |
\end{equation} |
1400 |
> |
where |
1401 |
> |
\[ R_1 = \left( {\begin{array}{*{20}c} |
1402 |
> |
0 & 0 & 0 \\ |
1403 |
> |
0 & 0 & {\pi _1 } \\ |
1404 |
> |
0 & { - \pi _1 } & 0 \\ |
1405 |
> |
\end{array}} \right). |
1406 |
> |
\] |
1407 |
> |
The solutions of Equation \ref{introEqaution:RBMotionSingleTerm} is |
1408 |
> |
\[ |
1409 |
> |
\pi (\Delta t) = e^{\Delta tR_1 } \pi (0),Q(\Delta t) = |
1410 |
> |
Q(0)e^{\Delta tR_1 } |
1411 |
> |
\] |
1412 |
> |
with |
1413 |
> |
\[ |
1414 |
> |
e^{\Delta tR_1 } = \left( {\begin{array}{*{20}c} |
1415 |
> |
0 & 0 & 0 \\ |
1416 |
> |
0 & {\cos \theta _1 } & {\sin \theta _1 } \\ |
1417 |
> |
0 & { - \sin \theta _1 } & {\cos \theta _1 } \\ |
1418 |
> |
\end{array}} \right),\theta _1 = \frac{{\pi _1 }}{{I_1 }}\Delta t. |
1419 |
> |
\] |
1420 |
> |
To reduce the cost of computing expensive functions in $e^{\Delta |
1421 |
> |
tR_1 }$, we can use Cayley transformation to obtain a single-aixs |
1422 |
> |
propagator, |
1423 |
> |
\[ |
1424 |
> |
e^{\Delta tR_1 } \approx (1 - \Delta tR_1 )^{ - 1} (1 + \Delta tR_1 |
1425 |
> |
) |
1426 |
> |
\] |
1427 |
> |
The flow maps for $T_2^r$ and $T_3^r$ can be found in the same |
1428 |
> |
manner. In order to construct a second-order symplectic method, we |
1429 |
> |
split the angular kinetic Hamiltonian function can into five terms |
1430 |
> |
\[ |
1431 |
> |
T^r (\pi ) = \frac{1}{2}T_1 ^r (\pi _1 ) + \frac{1}{2}T_2^r (\pi _2 |
1432 |
> |
) + T_3^r (\pi _3 ) + \frac{1}{2}T_2^r (\pi _2 ) + \frac{1}{2}T_1 ^r |
1433 |
> |
(\pi _1 ). |
1434 |
> |
\] |
1435 |
> |
By concatenating the propagators corresponding to these five terms, |
1436 |
> |
we can obtain an symplectic integrator, |
1437 |
> |
\[ |
1438 |
> |
\varphi _{\Delta t,T^r } = \varphi _{\Delta t/2,\pi _1 } \circ |
1439 |
> |
\varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t,\pi _3 } |
1440 |
> |
\circ \varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t/2,\pi |
1441 |
> |
_1 }. |
1442 |
> |
\] |
1443 |
> |
|
1444 |
> |
The non-canonical Lie-Poisson bracket ${F, G}$ of two function |
1445 |
> |
$F(\pi )$ and $G(\pi )$ is defined by |
1446 |
> |
\[ |
1447 |
> |
\{ F,G\} (\pi ) = [\nabla _\pi F(\pi )]^T J(\pi )\nabla _\pi G(\pi |
1448 |
> |
) |
1449 |
> |
\] |
1450 |
> |
If the Poisson bracket of a function $F$ with an arbitrary smooth |
1451 |
> |
function $G$ is zero, $F$ is a \emph{Casimir}, which is the |
1452 |
> |
conserved quantity in Poisson system. We can easily verify that the |
1453 |
> |
norm of the angular momentum, $\parallel \pi |
1454 |
> |
\parallel$, is a \emph{Casimir}. Let$ F(\pi ) = S(\frac{{\parallel |
1455 |
> |
\pi \parallel ^2 }}{2})$ for an arbitrary function $ S:R \to R$ , |
1456 |
> |
then by the chain rule |
1457 |
> |
\[ |
1458 |
> |
\nabla _\pi F(\pi ) = S'(\frac{{\parallel \pi \parallel ^2 |
1459 |
> |
}}{2})\pi |
1460 |
> |
\] |
1461 |
> |
Thus $ [\nabla _\pi F(\pi )]^T J(\pi ) = - S'(\frac{{\parallel \pi |
1462 |
> |
\parallel ^2 }}{2})\pi \times \pi = 0 $. This explicit |
1463 |
> |
Lie-Poisson integrator is found to be both extremely efficient and |
1464 |
> |
stable. These properties can be explained by the fact the small |
1465 |
> |
angle approximation is used and the norm of the angular momentum is |
1466 |
> |
conserved. |
1467 |
> |
|
1468 |
> |
\subsection{\label{introSection:RBHamiltonianSplitting} Hamiltonian |
1469 |
> |
Splitting for Rigid Body} |
1470 |
> |
|
1471 |
> |
The Hamiltonian of rigid body can be separated in terms of kinetic |
1472 |
> |
energy and potential energy, |
1473 |
> |
\[ |
1474 |
> |
H = T(p,\pi ) + V(q,Q) |
1475 |
> |
\] |
1476 |
> |
The equations of motion corresponding to potential energy and |
1477 |
> |
kinetic energy are listed in the below table, |
1478 |
> |
\begin{table} |
1479 |
> |
\caption{Equations of motion due to Potential and Kinetic Energies} |
1480 |
> |
\begin{center} |
1481 |
> |
\begin{tabular}{|l|l|} |
1482 |
> |
\hline |
1483 |
> |
% after \\: \hline or \cline{col1-col2} \cline{col3-col4} ... |
1484 |
> |
Potential & Kinetic \\ |
1485 |
> |
$\frac{{dq}}{{dt}} = \frac{p}{m}$ & $\frac{d}{{dt}}q = p$ \\ |
1486 |
> |
$\frac{d}{{dt}}p = - \frac{{\partial V}}{{\partial q}}$ & $ \frac{d}{{dt}}p = 0$ \\ |
1487 |
> |
$\frac{d}{{dt}}Q = 0$ & $ \frac{d}{{dt}}Q = Qskew(I^{ - 1} j)$ \\ |
1488 |
> |
$ \frac{d}{{dt}}\pi = \sum\limits_i {\left( {Q^T F_i (r,Q)} \right) \times X_i }$ & $\frac{d}{{dt}}\pi = \pi \times I^{ - 1} \pi$\\ |
1489 |
> |
\hline |
1490 |
> |
\end{tabular} |
1491 |
> |
\end{center} |
1492 |
> |
\end{table} |
1493 |
> |
A second-order symplectic method is now obtained by the composition |
1494 |
> |
of the position and velocity propagators, |
1495 |
> |
\[ |
1496 |
> |
\varphi _{\Delta t} = \varphi _{\Delta t/2,V} \circ \varphi |
1497 |
> |
_{\Delta t,T} \circ \varphi _{\Delta t/2,V}. |
1498 |
> |
\] |
1499 |
> |
Moreover, $\varphi _{\Delta t/2,V}$ can be divided into two |
1500 |
> |
sub-propagators which corresponding to force and torque |
1501 |
> |
respectively, |
1502 |
> |
\[ |
1503 |
> |
\varphi _{\Delta t/2,V} = \varphi _{\Delta t/2,F} \circ \varphi |
1504 |
> |
_{\Delta t/2,\tau }. |
1505 |
> |
\] |
1506 |
> |
Since the associated operators of $\varphi _{\Delta t/2,F} $ and |
1507 |
> |
$\circ \varphi _{\Delta t/2,\tau }$ commute, the composition order |
1508 |
> |
inside $\varphi _{\Delta t/2,V}$ does not matter. Furthermore, the |
1509 |
> |
kinetic energy can be separated to translational kinetic term, $T^t |
1510 |
> |
(p)$, and rotational kinetic term, $T^r (\pi )$, |
1511 |
> |
\begin{equation} |
1512 |
> |
T(p,\pi ) =T^t (p) + T^r (\pi ). |
1513 |
> |
\end{equation} |
1514 |
> |
where $ T^t (p) = \frac{1}{2}p^T m^{ - 1} p $ and $T^r (\pi )$ is |
1515 |
> |
defined by \ref{introEquation:rotationalKineticRB}. Therefore, the |
1516 |
> |
corresponding propagators are given by |
1517 |
> |
\[ |
1518 |
> |
\varphi _{\Delta t,T} = \varphi _{\Delta t,T^t } \circ \varphi |
1519 |
> |
_{\Delta t,T^r }. |
1520 |
> |
\] |
1521 |
> |
Finally, we obtain the overall symplectic propagators for freely |
1522 |
> |
moving rigid bodies |
1523 |
> |
\begin{equation} |
1524 |
> |
\begin{array}{c} |
1525 |
> |
\varphi _{\Delta t} = \varphi _{\Delta t/2,F} \circ \varphi _{\Delta t/2,\tau } \\ |
1526 |
> |
\circ \varphi _{\Delta t,T^t } \circ \varphi _{\Delta t/2,\pi _1 } \circ \varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t,\pi _3 } \circ \varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t/2,\pi _1 } \\ |
1527 |
> |
\circ \varphi _{\Delta t/2,\tau } \circ \varphi _{\Delta t/2,F} .\\ |
1528 |
> |
\end{array} |
1529 |
> |
\label{introEquation:overallRBFlowMaps} |
1530 |
> |
\end{equation} |
1531 |
> |
|
1532 |
> |
\section{\label{introSection:langevinDynamics}Langevin Dynamics} |
1533 |
> |
As an alternative to newtonian dynamics, Langevin dynamics, which |
1534 |
> |
mimics a simple heat bath with stochastic and dissipative forces, |
1535 |
> |
has been applied in a variety of studies. This section will review |
1536 |
> |
the theory of Langevin dynamics. A brief derivation of generalized |
1537 |
> |
Langevin equation will be given first. Following that, we will |
1538 |
> |
discuss the physical meaning of the terms appearing in the equation |
1539 |
> |
as well as the calculation of friction tensor from hydrodynamics |
1540 |
> |
theory. |
1541 |
> |
|
1542 |
> |
\subsection{\label{introSection:generalizedLangevinDynamics}Derivation of Generalized Langevin Equation} |
1543 |
> |
|
1544 |
> |
A harmonic bath model, in which an effective set of harmonic |
1545 |
> |
oscillators are used to mimic the effect of a linearly responding |
1546 |
> |
environment, has been widely used in quantum chemistry and |
1547 |
> |
statistical mechanics. One of the successful applications of |
1548 |
> |
Harmonic bath model is the derivation of the Generalized Langevin |
1549 |
> |
Dynamics (GLE). Lets consider a system, in which the degree of |
1550 |
> |
freedom $x$ is assumed to couple to the bath linearly, giving a |
1551 |
> |
Hamiltonian of the form |
1552 |
> |
\begin{equation} |
1553 |
> |
H = \frac{{p^2 }}{{2m}} + U(x) + H_B + \Delta U(x,x_1 , \ldots x_N) |
1554 |
> |
\label{introEquation:bathGLE}. |
1555 |
> |
\end{equation} |
1556 |
> |
Here $p$ is a momentum conjugate to $x$, $m$ is the mass associated |
1557 |
> |
with this degree of freedom, $H_B$ is a harmonic bath Hamiltonian, |
1558 |
> |
\[ |
1559 |
> |
H_B = \sum\limits_{\alpha = 1}^N {\left\{ {\frac{{p_\alpha ^2 |
1560 |
> |
}}{{2m_\alpha }} + \frac{1}{2}m_\alpha \omega _\alpha ^2 } |
1561 |
> |
\right\}} |
1562 |
> |
\] |
1563 |
> |
where the index $\alpha$ runs over all the bath degrees of freedom, |
1564 |
> |
$\omega _\alpha$ are the harmonic bath frequencies, $m_\alpha$ are |
1565 |
> |
the harmonic bath masses, and $\Delta U$ is a bilinear system-bath |
1566 |
> |
coupling, |
1567 |
> |
\[ |
1568 |
> |
\Delta U = - \sum\limits_{\alpha = 1}^N {g_\alpha x_\alpha x} |
1569 |
> |
\] |
1570 |
> |
where $g_\alpha$ are the coupling constants between the bath |
1571 |
> |
coordinates ($x_ \alpha$) and the system coordinate ($x$). |
1572 |
> |
Introducing |
1573 |
> |
\[ |
1574 |
> |
W(x) = U(x) - \sum\limits_{\alpha = 1}^N {\frac{{g_\alpha ^2 |
1575 |
> |
}}{{2m_\alpha w_\alpha ^2 }}} x^2 |
1576 |
> |
\] and combining the last two terms in Equation |
1577 |
> |
\ref{introEquation:bathGLE}, we may rewrite the Harmonic bath |
1578 |
> |
Hamiltonian as |
1579 |
> |
\[ |
1580 |
> |
H = \frac{{p^2 }}{{2m}} + W(x) + \sum\limits_{\alpha = 1}^N |
1581 |
> |
{\left\{ {\frac{{p_\alpha ^2 }}{{2m_\alpha }} + \frac{1}{2}m_\alpha |
1582 |
> |
w_\alpha ^2 \left( {x_\alpha - \frac{{g_\alpha }}{{m_\alpha |
1583 |
> |
w_\alpha ^2 }}x} \right)^2 } \right\}} |
1584 |
> |
\] |
1585 |
> |
Since the first two terms of the new Hamiltonian depend only on the |
1586 |
> |
system coordinates, we can get the equations of motion for |
1587 |
> |
Generalized Langevin Dynamics by Hamilton's equations, |
1588 |
> |
\begin{equation} |
1589 |
> |
m\ddot x = - \frac{{\partial W(x)}}{{\partial x}} - |
1590 |
> |
\sum\limits_{\alpha = 1}^N {g_\alpha \left( {x_\alpha - |
1591 |
> |
\frac{{g_\alpha }}{{m_\alpha w_\alpha ^2 }}x} \right)}, |
1592 |
> |
\label{introEquation:coorMotionGLE} |
1593 |
> |
\end{equation} |
1594 |
> |
and |
1595 |
> |
\begin{equation} |
1596 |
> |
m\ddot x_\alpha = - m_\alpha w_\alpha ^2 \left( {x_\alpha - |
1597 |
> |
\frac{{g_\alpha }}{{m_\alpha w_\alpha ^2 }}x} \right). |
1598 |
> |
\label{introEquation:bathMotionGLE} |
1599 |
> |
\end{equation} |
1600 |
> |
|
1601 |
> |
In order to derive an equation for $x$, the dynamics of the bath |
1602 |
> |
variables $x_\alpha$ must be solved exactly first. As an integral |
1603 |
> |
transform which is particularly useful in solving linear ordinary |
1604 |
> |
differential equations,the Laplace transform is the appropriate tool |
1605 |
> |
to solve this problem. The basic idea is to transform the difficult |
1606 |
> |
differential equations into simple algebra problems which can be |
1607 |
> |
solved easily. Then, by applying the inverse Laplace transform, also |
1608 |
> |
known as the Bromwich integral, we can retrieve the solutions of the |
1609 |
> |
original problems. |
1610 |
> |
|
1611 |
> |
Let $f(t)$ be a function defined on $ [0,\infty ) $. The Laplace |
1612 |
> |
transform of f(t) is a new function defined as |
1613 |
> |
\[ |
1614 |
> |
L(f(t)) \equiv F(p) = \int_0^\infty {f(t)e^{ - pt} dt} |
1615 |
> |
\] |
1616 |
> |
where $p$ is real and $L$ is called the Laplace Transform |
1617 |
> |
Operator. Below are some important properties of Laplace transform |
1618 |
> |
|
1619 |
> |
\begin{eqnarray*} |
1620 |
> |
L(x + y) & = & L(x) + L(y) \\ |
1621 |
> |
L(ax) & = & aL(x) \\ |
1622 |
> |
L(\dot x) & = & pL(x) - px(0) \\ |
1623 |
> |
L(\ddot x)& = & p^2 L(x) - px(0) - \dot x(0) \\ |
1624 |
> |
L\left( {\int_0^t {g(t - \tau )h(\tau )d\tau } } \right)& = & G(p)H(p) \\ |
1625 |
> |
\end{eqnarray*} |
1626 |
> |
|
1627 |
> |
|
1628 |
> |
Applying the Laplace transform to the bath coordinates, we obtain |
1629 |
> |
\begin{eqnarray*} |
1630 |
> |
p^2 L(x_\alpha ) - px_\alpha (0) - \dot x_\alpha (0) & = & - \omega _\alpha ^2 L(x_\alpha ) + \frac{{g_\alpha }}{{\omega _\alpha }}L(x) \\ |
1631 |
> |
L(x_\alpha ) & = & \frac{{\frac{{g_\alpha }}{{\omega _\alpha }}L(x) + px_\alpha (0) + \dot x_\alpha (0)}}{{p^2 + \omega _\alpha ^2 }} \\ |
1632 |
> |
\end{eqnarray*} |
1633 |
> |
|
1634 |
> |
By the same way, the system coordinates become |
1635 |
> |
\begin{eqnarray*} |
1636 |
> |
mL(\ddot x) & = & - \frac{1}{p}\frac{{\partial W(x)}}{{\partial x}} \\ |
1637 |
> |
& & \mbox{} - \sum\limits_{\alpha = 1}^N {\left\{ { - \frac{{g_\alpha ^2 }}{{m_\alpha \omega _\alpha ^2 }}\frac{p}{{p^2 + \omega _\alpha ^2 }}pL(x) - \frac{p}{{p^2 + \omega _\alpha ^2 }}g_\alpha x_\alpha (0) - \frac{1}{{p^2 + \omega _\alpha ^2 }}g_\alpha \dot x_\alpha (0)} \right\}} \\ |
1638 |
> |
\end{eqnarray*} |
1639 |
> |
|
1640 |
> |
With the help of some relatively important inverse Laplace |
1641 |
> |
transformations: |
1642 |
> |
\[ |
1643 |
> |
\begin{array}{c} |
1644 |
> |
L(\cos at) = \frac{p}{{p^2 + a^2 }} \\ |
1645 |
> |
L(\sin at) = \frac{a}{{p^2 + a^2 }} \\ |
1646 |
> |
L(1) = \frac{1}{p} \\ |
1647 |
> |
\end{array} |
1648 |
> |
\] |
1649 |
> |
, we obtain |
1650 |
> |
\begin{eqnarray*} |
1651 |
> |
m\ddot x & = & - \frac{{\partial W(x)}}{{\partial x}} - |
1652 |
> |
\sum\limits_{\alpha = 1}^N {\left\{ {\left( { - \frac{{g_\alpha ^2 |
1653 |
> |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\int_0^t {\cos (\omega |
1654 |
> |
_\alpha t)\dot x(t - \tau )d\tau } } \right\}} \\ |
1655 |
> |
& & + \sum\limits_{\alpha = 1}^N {\left\{ {\left[ {g_\alpha |
1656 |
> |
x_\alpha (0) - \frac{{g_\alpha }}{{m_\alpha \omega _\alpha }}} |
1657 |
> |
\right]\cos (\omega _\alpha t) + \frac{{g_\alpha \dot x_\alpha |
1658 |
> |
(0)}}{{\omega _\alpha }}\sin (\omega _\alpha t)} \right\}} |
1659 |
> |
\end{eqnarray*} |
1660 |
> |
\begin{eqnarray*} |
1661 |
> |
m\ddot x & = & - \frac{{\partial W(x)}}{{\partial x}} - \int_0^t |
1662 |
> |
{\sum\limits_{\alpha = 1}^N {\left( { - \frac{{g_\alpha ^2 |
1663 |
> |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\cos (\omega _\alpha |
1664 |
> |
t)\dot x(t - \tau )d} \tau } \\ |
1665 |
> |
& & + \sum\limits_{\alpha = 1}^N {\left\{ {\left[ {g_\alpha |
1666 |
> |
x_\alpha (0) - \frac{{g_\alpha }}{{m_\alpha \omega _\alpha }}} |
1667 |
> |
\right]\cos (\omega _\alpha t) + \frac{{g_\alpha \dot x_\alpha |
1668 |
> |
(0)}}{{\omega _\alpha }}\sin (\omega _\alpha t)} \right\}} |
1669 |
> |
\end{eqnarray*} |
1670 |
> |
Introducing a \emph{dynamic friction kernel} |
1671 |
> |
\begin{equation} |
1672 |
> |
\xi (t) = \sum\limits_{\alpha = 1}^N {\left( { - \frac{{g_\alpha ^2 |
1673 |
> |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\cos (\omega _\alpha t)} |
1674 |
> |
\label{introEquation:dynamicFrictionKernelDefinition} |
1675 |
> |
\end{equation} |
1676 |
> |
and \emph{a random force} |
1677 |
> |
\begin{equation} |
1678 |
> |
R(t) = \sum\limits_{\alpha = 1}^N {\left( {g_\alpha x_\alpha (0) |
1679 |
> |
- \frac{{g_\alpha ^2 }}{{m_\alpha \omega _\alpha ^2 }}x(0)} |
1680 |
> |
\right)\cos (\omega _\alpha t)} + \frac{{\dot x_\alpha |
1681 |
> |
(0)}}{{\omega _\alpha }}\sin (\omega _\alpha t), |
1682 |
> |
\label{introEquation:randomForceDefinition} |
1683 |
> |
\end{equation} |
1684 |
> |
the equation of motion can be rewritten as |
1685 |
> |
\begin{equation} |
1686 |
> |
m\ddot x = - \frac{{\partial W}}{{\partial x}} - \int_0^t {\xi |
1687 |
> |
(t)\dot x(t - \tau )d\tau } + R(t) |
1688 |
> |
\label{introEuqation:GeneralizedLangevinDynamics} |
1689 |
> |
\end{equation} |
1690 |
> |
which is known as the \emph{generalized Langevin equation}. |
1691 |
> |
|
1692 |
> |
\subsubsection{\label{introSection:randomForceDynamicFrictionKernel}\textbf{Random Force and Dynamic Friction Kernel}} |
1693 |
> |
|
1694 |
> |
One may notice that $R(t)$ depends only on initial conditions, which |
1695 |
> |
implies it is completely deterministic within the context of a |
1696 |
> |
harmonic bath. However, it is easy to verify that $R(t)$ is totally |
1697 |
> |
uncorrelated to $x$ and $\dot x$, |
1698 |
> |
\[ |
1699 |
> |
\begin{array}{l} |
1700 |
> |
\left\langle {x(t)R(t)} \right\rangle = 0, \\ |
1701 |
> |
\left\langle {\dot x(t)R(t)} \right\rangle = 0. \\ |
1702 |
> |
\end{array} |
1703 |
> |
\] |
1704 |
> |
This property is what we expect from a truly random process. As long |
1705 |
> |
as the model chosen for $R(t)$ was a gaussian distribution in |
1706 |
> |
general, the stochastic nature of the GLE still remains. |
1707 |
> |
|
1708 |
> |
%dynamic friction kernel |
1709 |
> |
The convolution integral |
1710 |
> |
\[ |
1711 |
> |
\int_0^t {\xi (t)\dot x(t - \tau )d\tau } |
1712 |
> |
\] |
1713 |
> |
depends on the entire history of the evolution of $x$, which implies |
1714 |
> |
that the bath retains memory of previous motions. In other words, |
1715 |
> |
the bath requires a finite time to respond to change in the motion |
1716 |
> |
of the system. For a sluggish bath which responds slowly to changes |
1717 |
> |
in the system coordinate, we may regard $\xi(t)$ as a constant |
1718 |
> |
$\xi(t) = \Xi_0$. Hence, the convolution integral becomes |
1719 |
> |
\[ |
1720 |
> |
\int_0^t {\xi (t)\dot x(t - \tau )d\tau } = \xi _0 (x(t) - x(0)) |
1721 |
> |
\] |
1722 |
> |
and Equation \ref{introEuqation:GeneralizedLangevinDynamics} becomes |
1723 |
> |
\[ |
1724 |
> |
m\ddot x = - \frac{\partial }{{\partial x}}\left( {W(x) + |
1725 |
> |
\frac{1}{2}\xi _0 (x - x_0 )^2 } \right) + R(t), |
1726 |
> |
\] |
1727 |
> |
which can be used to describe the effect of dynamic caging in |
1728 |
> |
viscous solvents. The other extreme is the bath that responds |
1729 |
> |
infinitely quickly to motions in the system. Thus, $\xi (t)$ can be |
1730 |
> |
taken as a $delta$ function in time: |
1731 |
> |
\[ |
1732 |
> |
\xi (t) = 2\xi _0 \delta (t) |
1733 |
> |
\] |
1734 |
> |
Hence, the convolution integral becomes |
1735 |
> |
\[ |
1736 |
> |
\int_0^t {\xi (t)\dot x(t - \tau )d\tau } = 2\xi _0 \int_0^t |
1737 |
> |
{\delta (t)\dot x(t - \tau )d\tau } = \xi _0 \dot x(t), |
1738 |
> |
\] |
1739 |
> |
and Equation \ref{introEuqation:GeneralizedLangevinDynamics} becomes |
1740 |
> |
\begin{equation} |
1741 |
> |
m\ddot x = - \frac{{\partial W(x)}}{{\partial x}} - \xi _0 \dot |
1742 |
> |
x(t) + R(t) \label{introEquation:LangevinEquation} |
1743 |
> |
\end{equation} |
1744 |
> |
which is known as the Langevin equation. The static friction |
1745 |
> |
coefficient $\xi _0$ can either be calculated from spectral density |
1746 |
> |
or be determined by Stokes' law for regular shaped particles. A |
1747 |
> |
briefly review on calculating friction tensor for arbitrary shaped |
1748 |
> |
particles is given in Sec.~\ref{introSection:frictionTensor}. |
1749 |
> |
|
1750 |
> |
\subsubsection{\label{introSection:secondFluctuationDissipation}\textbf{The Second Fluctuation Dissipation Theorem}} |
1751 |
> |
|
1752 |
> |
Defining a new set of coordinates, |
1753 |
> |
\[ |
1754 |
> |
q_\alpha (t) = x_\alpha (t) - \frac{1}{{m_\alpha \omega _\alpha |
1755 |
> |
^2 }}x(0) |
1756 |
> |
\], |
1757 |
> |
we can rewrite $R(T)$ as |
1758 |
> |
\[ |
1759 |
> |
R(t) = \sum\limits_{\alpha = 1}^N {g_\alpha q_\alpha (t)}. |
1760 |
> |
\] |
1761 |
> |
And since the $q$ coordinates are harmonic oscillators, |
1762 |
> |
|
1763 |
> |
\begin{eqnarray*} |
1764 |
> |
\left\langle {q_\alpha ^2 } \right\rangle & = & \frac{{kT}}{{m_\alpha \omega _\alpha ^2 }} \\ |
1765 |
> |
\left\langle {q_\alpha (t)q_\alpha (0)} \right\rangle & = & \left\langle {q_\alpha ^2 (0)} \right\rangle \cos (\omega _\alpha t) \\ |
1766 |
> |
\left\langle {q_\alpha (t)q_\beta (0)} \right\rangle & = &\delta _{\alpha \beta } \left\langle {q_\alpha (t)q_\alpha (0)} \right\rangle \\ |
1767 |
> |
\left\langle {R(t)R(0)} \right\rangle & = & \sum\limits_\alpha {\sum\limits_\beta {g_\alpha g_\beta \left\langle {q_\alpha (t)q_\beta (0)} \right\rangle } } \\ |
1768 |
> |
& = &\sum\limits_\alpha {g_\alpha ^2 \left\langle {q_\alpha ^2 (0)} \right\rangle \cos (\omega _\alpha t)} \\ |
1769 |
> |
& = &kT\xi (t) \\ |
1770 |
> |
\end{eqnarray*} |
1771 |
> |
|
1772 |
> |
Thus, we recover the \emph{second fluctuation dissipation theorem} |
1773 |
> |
\begin{equation} |
1774 |
> |
\xi (t) = \left\langle {R(t)R(0)} \right\rangle |
1775 |
> |
\label{introEquation:secondFluctuationDissipation}. |
1776 |
> |
\end{equation} |
1777 |
> |
In effect, it acts as a constraint on the possible ways in which one |
1778 |
> |
can model the random force and friction kernel. |