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 $i$. |
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 |
|
|
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 |
66 |
< |
that if all forces are conservative, Energy $E = T + V$ is |
67 |
< |
conserved. All of these conserved quantities are important factors |
68 |
< |
to determine the quality of numerical integration scheme for rigid |
69 |
< |
body \cite{Dullweber1997}. |
66 |
> |
that if all forces are conservative, Energy |
67 |
> |
\begin{equation}E = T + V \label{introEquation:energyConservation} |
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}. |
72 |
|
|
73 |
|
\subsection{\label{introSection:lagrangian}Lagrangian Mechanics} |
74 |
|
|
93 |
|
The actual trajectory, along which a dynamical system may move from |
94 |
|
one point to another within a specified time, is derived by finding |
95 |
|
the path which minimizes the time integral of the difference between |
96 |
< |
the kinetic, $K$, and potential energies, $U$ \cite{tolman79}. |
96 |
> |
the kinetic, $K$, and potential energies, $U$ \cite{Tolman1979}. |
97 |
|
\begin{equation} |
98 |
|
\delta \int_{t_1 }^{t_2 } {(K - U)dt = 0} , |
99 |
|
\label{introEquation:halmitonianPrinciple1} |
117 |
|
\subsubsection{\label{introSection:equationOfMotionLagrangian}The |
118 |
|
Equations of Motion in Lagrangian Mechanics} |
119 |
|
|
120 |
< |
for a holonomic system of $f$ degrees of freedom, the equations of |
120 |
> |
For a holonomic system of $f$ degrees of freedom, the equations of |
121 |
|
motion in the Lagrangian form is |
122 |
|
\begin{equation} |
123 |
|
\frac{d}{{dt}}\frac{{\partial L}}{{\partial \dot q_i }} - |
189 |
|
Eq.~\ref{introEquation:motionHamiltonianCoordinate} and |
190 |
|
Eq.~\ref{introEquation:motionHamiltonianMomentum} are Hamilton's |
191 |
|
equation of motion. Due to their symmetrical formula, they are also |
192 |
< |
known as the canonical equations of motions \cite{Goldstein01}. |
192 |
> |
known as the canonical equations of motions \cite{Goldstein2001}. |
193 |
|
|
194 |
|
An important difference between Lagrangian approach and the |
195 |
|
Hamiltonian approach is that the Lagrangian is considered to be a |
200 |
|
appropriate for application to statistical mechanics and quantum |
201 |
|
mechanics, since it treats the coordinate and its time derivative as |
202 |
|
independent variables and it only works with 1st-order differential |
203 |
< |
equations\cite{Marion90}. |
203 |
> |
equations\cite{Marion1990}. |
204 |
|
|
205 |
< |
When studying Hamiltonian system, it is more convenient to use |
206 |
< |
notation |
205 |
> |
In Newtonian Mechanics, a system described by conservative forces |
206 |
> |
conserves the total energy \ref{introEquation:energyConservation}. |
207 |
> |
It follows that Hamilton's equations of motion conserve the total |
208 |
> |
Hamiltonian. |
209 |
|
\begin{equation} |
210 |
< |
r = r(q,p)^T |
210 |
> |
\frac{{dH}}{{dt}} = \sum\limits_i {\left( {\frac{{\partial |
211 |
> |
H}}{{\partial q_i }}\dot q_i + \frac{{\partial H}}{{\partial p_i |
212 |
> |
}}\dot p_i } \right)} = \sum\limits_i {\left( {\frac{{\partial |
213 |
> |
H}}{{\partial q_i }}\frac{{\partial H}}{{\partial p_i }} - |
214 |
> |
\frac{{\partial H}}{{\partial p_i }}\frac{{\partial H}}{{\partial |
215 |
> |
q_i }}} \right) = 0} \label{introEquation:conserveHalmitonian} |
216 |
|
\end{equation} |
208 |
– |
and to introduce a $2n \times 2n$ canonical structure matrix $J$, |
209 |
– |
\begin{equation} |
210 |
– |
J = \left( {\begin{array}{*{20}c} |
211 |
– |
0 & I \\ |
212 |
– |
{ - I} & 0 \\ |
213 |
– |
\end{array}} \right) |
214 |
– |
\label{introEquation:canonicalMatrix} |
215 |
– |
\end{equation} |
216 |
– |
Thus, Hamiltonian system can be rewritten as, |
217 |
– |
\begin{equation} |
218 |
– |
\frac{d}{{dt}}r = J\nabla _r H(r) |
219 |
– |
\label{introEquation:compactHamiltonian} |
220 |
– |
\end{equation} |
221 |
– |
where $I$ is an identity matrix and $J$ is a skew-symmetrix matrix |
222 |
– |
($ J^T = - J $). |
217 |
|
|
224 |
– |
%\subsection{\label{introSection:canonicalTransformation}Canonical |
225 |
– |
Transformation} |
226 |
– |
|
227 |
– |
\section{\label{introSection:geometricIntegratos}Geometric Integrators} |
228 |
– |
|
229 |
– |
\subsection{\label{introSection:symplecticMaps}Symplectic Maps and Methods} |
230 |
– |
|
231 |
– |
\subsection{\label{Construction of Symplectic Methods}} |
232 |
– |
|
218 |
|
\section{\label{introSection:statisticalMechanics}Statistical |
219 |
|
Mechanics} |
220 |
|
|
221 |
|
The thermodynamic behaviors and properties of Molecular Dynamics |
222 |
|
simulation are governed by the principle of Statistical Mechanics. |
223 |
|
The following section will give a brief introduction to some of the |
224 |
< |
Statistical Mechanics concepts presented in this dissertation. |
225 |
< |
|
241 |
< |
\subsection{\label{introSection::ensemble}Ensemble} |
242 |
< |
|
243 |
< |
\subsection{\label{introSection:ergodic}The Ergodic Hypothesis} |
224 |
> |
Statistical Mechanics concepts and theorem presented in this |
225 |
> |
dissertation. |
226 |
|
|
227 |
+ |
\subsection{\label{introSection:ensemble}Phase Space and Ensemble} |
228 |
|
|
229 |
< |
\section{\label{introSection:molecularDynamics}Molecular Dynamics} |
229 |
> |
Mathematically, phase space is the space which represents all |
230 |
> |
possible states. Each possible state of the system corresponds to |
231 |
> |
one unique point in the phase space. For mechanical systems, the |
232 |
> |
phase space usually consists of all possible values of position and |
233 |
> |
momentum variables. Consider a dynamic system in a cartesian space, |
234 |
> |
where each of the $6f$ coordinates and momenta is assigned to one of |
235 |
> |
$6f$ mutually orthogonal axes, the phase space of this system is a |
236 |
> |
$6f$ dimensional space. A point, $x = (q_1 , \ldots ,q_f ,p_1 , |
237 |
> |
\ldots ,p_f )$, with a unique set of values of $6f$ coordinates and |
238 |
> |
momenta is a phase space vector. |
239 |
|
|
240 |
< |
As a special discipline of molecular modeling, Molecular dynamics |
241 |
< |
has proven to be a powerful tool for studying the functions of |
242 |
< |
biological systems, providing structural, thermodynamic and |
243 |
< |
dynamical information. |
240 |
> |
A microscopic state or microstate of a classical system is |
241 |
> |
specification of the complete phase space vector of a system at any |
242 |
> |
instant in time. An ensemble is defined as a collection of systems |
243 |
> |
sharing one or more macroscopic characteristics but each being in a |
244 |
> |
unique microstate. The complete ensemble is specified by giving all |
245 |
> |
systems or microstates consistent with the common macroscopic |
246 |
> |
characteristics of the ensemble. Although the state of each |
247 |
> |
individual system in the ensemble could be precisely described at |
248 |
> |
any instance in time by a suitable phase space vector, when using |
249 |
> |
ensembles for statistical purposes, there is no need to maintain |
250 |
> |
distinctions between individual systems, since the numbers of |
251 |
> |
systems at any time in the different states which correspond to |
252 |
> |
different regions of the phase space are more interesting. Moreover, |
253 |
> |
in the point of view of statistical mechanics, one would prefer to |
254 |
> |
use ensembles containing a large enough population of separate |
255 |
> |
members so that the numbers of systems in such different states can |
256 |
> |
be regarded as changing continuously as we traverse different |
257 |
> |
regions of the phase space. The condition of an ensemble at any time |
258 |
> |
can be regarded as appropriately specified by the density $\rho$ |
259 |
> |
with which representative points are distributed over the phase |
260 |
> |
space. The density of distribution for an ensemble with $f$ degrees |
261 |
> |
of freedom is defined as, |
262 |
> |
\begin{equation} |
263 |
> |
\rho = \rho (q_1 , \ldots ,q_f ,p_1 , \ldots ,p_f ,t). |
264 |
> |
\label{introEquation:densityDistribution} |
265 |
> |
\end{equation} |
266 |
> |
Governed by the principles of mechanics, the phase points change |
267 |
> |
their value which would change the density at any time at phase |
268 |
> |
space. Hence, the density of distribution is also to be taken as a |
269 |
> |
function of the time. |
270 |
|
|
271 |
< |
\subsection{\label{introSec:mdInit}Initialization} |
272 |
< |
|
273 |
< |
\subsection{\label{introSection:mdIntegration} Integration of the Equations of Motion} |
274 |
< |
|
275 |
< |
\section{\label{introSection:rigidBody}Dynamics of Rigid Bodies} |
276 |
< |
|
277 |
< |
A rigid body is a body in which the distance between any two given |
278 |
< |
points of a rigid body remains constant regardless of external |
279 |
< |
forces exerted on it. A rigid body therefore conserves its shape |
280 |
< |
during its motion. |
271 |
> |
The number of systems $\delta N$ at time $t$ can be determined by, |
272 |
> |
\begin{equation} |
273 |
> |
\delta N = \rho (q,p,t)dq_1 \ldots dq_f dp_1 \ldots dp_f. |
274 |
> |
\label{introEquation:deltaN} |
275 |
> |
\end{equation} |
276 |
> |
Assuming a large enough population of systems are exploited, we can |
277 |
> |
sufficiently approximate $\delta N$ without introducing |
278 |
> |
discontinuity when we go from one region in the phase space to |
279 |
> |
another. By integrating over the whole phase space, |
280 |
> |
\begin{equation} |
281 |
> |
N = \int { \ldots \int {\rho (q,p,t)dq_1 } ...dq_f dp_1 } ...dp_f |
282 |
> |
\label{introEquation:totalNumberSystem} |
283 |
> |
\end{equation} |
284 |
> |
gives us an expression for the total number of the systems. Hence, |
285 |
> |
the probability per unit in the phase space can be obtained by, |
286 |
> |
\begin{equation} |
287 |
> |
\frac{{\rho (q,p,t)}}{N} = \frac{{\rho (q,p,t)}}{{\int { \ldots \int |
288 |
> |
{\rho (q,p,t)dq_1 } ...dq_f dp_1 } ...dp_f }}. |
289 |
> |
\label{introEquation:unitProbability} |
290 |
> |
\end{equation} |
291 |
> |
With the help of Equation(\ref{introEquation:unitProbability}) and |
292 |
> |
the knowledge of the system, it is possible to calculate the average |
293 |
> |
value of any desired quantity which depends on the coordinates and |
294 |
> |
momenta of the system. Even when the dynamics of the real system is |
295 |
> |
complex, or stochastic, or even discontinuous, the average |
296 |
> |
properties of the ensemble of possibilities as a whole may still |
297 |
> |
remain well defined. For a classical system in thermal equilibrium |
298 |
> |
with its environment, the ensemble average of a mechanical quantity, |
299 |
> |
$\langle A(q , p) \rangle_t$, takes the form of an integral over the |
300 |
> |
phase space of the system, |
301 |
> |
\begin{equation} |
302 |
> |
\langle A(q , p) \rangle_t = \frac{{\int { \ldots \int {A(q,p)\rho |
303 |
> |
(q,p,t)dq_1 } ...dq_f dp_1 } ...dp_f }}{{\int { \ldots \int {\rho |
304 |
> |
(q,p,t)dq_1 } ...dq_f dp_1 } ...dp_f }} |
305 |
> |
\label{introEquation:ensembelAverage} |
306 |
> |
\end{equation} |
307 |
|
|
308 |
< |
Applications of dynamics of rigid bodies. |
308 |
> |
There are several different types of ensembles with different |
309 |
> |
statistical characteristics. As a function of macroscopic |
310 |
> |
parameters, such as temperature \textit{etc}, partition function can |
311 |
> |
be used to describe the statistical properties of a system in |
312 |
> |
thermodynamic equilibrium. |
313 |
> |
|
314 |
> |
As an ensemble of systems, each of which is known to be thermally |
315 |
> |
isolated and conserve energy, Microcanonical ensemble(NVE) has a |
316 |
> |
partition function like, |
317 |
> |
\begin{equation} |
318 |
> |
\Omega (N,V,E) = e^{\beta TS} \label{introEquation:NVEPartition}. |
319 |
> |
\end{equation} |
320 |
> |
A canonical ensemble(NVT)is an ensemble of systems, each of which |
321 |
> |
can share its energy with a large heat reservoir. The distribution |
322 |
> |
of the total energy amongst the possible dynamical states is given |
323 |
> |
by the partition function, |
324 |
> |
\begin{equation} |
325 |
> |
\Omega (N,V,T) = e^{ - \beta A} |
326 |
> |
\label{introEquation:NVTPartition} |
327 |
> |
\end{equation} |
328 |
> |
Here, $A$ is the Helmholtz free energy which is defined as $ A = U - |
329 |
> |
TS$. Since most experiment are carried out under constant pressure |
330 |
> |
condition, isothermal-isobaric ensemble(NPT) play a very important |
331 |
> |
role in molecular simulation. The isothermal-isobaric ensemble allow |
332 |
> |
the system to exchange energy with a heat bath of temperature $T$ |
333 |
> |
and to change the volume as well. Its partition function is given as |
334 |
> |
\begin{equation} |
335 |
> |
\Delta (N,P,T) = - e^{\beta G}. |
336 |
> |
\label{introEquation:NPTPartition} |
337 |
> |
\end{equation} |
338 |
> |
Here, $G = U - TS + PV$, and $G$ is called Gibbs free energy. |
339 |
> |
|
340 |
> |
\subsection{\label{introSection:liouville}Liouville's theorem} |
341 |
> |
|
342 |
> |
The Liouville's theorem is the foundation on which statistical |
343 |
> |
mechanics rests. It describes the time evolution of phase space |
344 |
> |
distribution function. In order to calculate the rate of change of |
345 |
> |
$\rho$, we begin from Equation(\ref{introEquation:deltaN}). If we |
346 |
> |
consider the two faces perpendicular to the $q_1$ axis, which are |
347 |
> |
located at $q_1$ and $q_1 + \delta q_1$, the number of phase points |
348 |
> |
leaving the opposite face is given by the expression, |
349 |
> |
\begin{equation} |
350 |
> |
\left( {\rho + \frac{{\partial \rho }}{{\partial q_1 }}\delta q_1 } |
351 |
> |
\right)\left( {\dot q_1 + \frac{{\partial \dot q_1 }}{{\partial q_1 |
352 |
> |
}}\delta q_1 } \right)\delta q_2 \ldots \delta q_f \delta p_1 |
353 |
> |
\ldots \delta p_f . |
354 |
> |
\end{equation} |
355 |
> |
Summing all over the phase space, we obtain |
356 |
> |
\begin{equation} |
357 |
> |
\frac{{d(\delta N)}}{{dt}} = - \sum\limits_{i = 1}^f {\left[ {\rho |
358 |
> |
\left( {\frac{{\partial \dot q_i }}{{\partial q_i }} + |
359 |
> |
\frac{{\partial \dot p_i }}{{\partial p_i }}} \right) + \left( |
360 |
> |
{\frac{{\partial \rho }}{{\partial q_i }}\dot q_i + \frac{{\partial |
361 |
> |
\rho }}{{\partial p_i }}\dot p_i } \right)} \right]} \delta q_1 |
362 |
> |
\ldots \delta q_f \delta p_1 \ldots \delta p_f . |
363 |
> |
\end{equation} |
364 |
> |
Differentiating the equations of motion in Hamiltonian formalism |
365 |
> |
(\ref{introEquation:motionHamiltonianCoordinate}, |
366 |
> |
\ref{introEquation:motionHamiltonianMomentum}), we can show, |
367 |
> |
\begin{equation} |
368 |
> |
\sum\limits_i {\left( {\frac{{\partial \dot q_i }}{{\partial q_i }} |
369 |
> |
+ \frac{{\partial \dot p_i }}{{\partial p_i }}} \right)} = 0 , |
370 |
> |
\end{equation} |
371 |
> |
which cancels the first terms of the right hand side. Furthermore, |
372 |
> |
divining $ \delta q_1 \ldots \delta q_f \delta p_1 \ldots \delta |
373 |
> |
p_f $ in both sides, we can write out Liouville's theorem in a |
374 |
> |
simple form, |
375 |
> |
\begin{equation} |
376 |
> |
\frac{{\partial \rho }}{{\partial t}} + \sum\limits_{i = 1}^f |
377 |
> |
{\left( {\frac{{\partial \rho }}{{\partial q_i }}\dot q_i + |
378 |
> |
\frac{{\partial \rho }}{{\partial p_i }}\dot p_i } \right)} = 0 . |
379 |
> |
\label{introEquation:liouvilleTheorem} |
380 |
> |
\end{equation} |
381 |
> |
|
382 |
> |
Liouville's theorem states that the distribution function is |
383 |
> |
constant along any trajectory in phase space. In classical |
384 |
> |
statistical mechanics, since the number of particles in the system |
385 |
> |
is huge, we may be able to believe the system is stationary, |
386 |
> |
\begin{equation} |
387 |
> |
\frac{{\partial \rho }}{{\partial t}} = 0. |
388 |
> |
\label{introEquation:stationary} |
389 |
> |
\end{equation} |
390 |
> |
In such stationary system, the density of distribution $\rho$ can be |
391 |
> |
connected to the Hamiltonian $H$ through Maxwell-Boltzmann |
392 |
> |
distribution, |
393 |
> |
\begin{equation} |
394 |
> |
\rho \propto e^{ - \beta H} |
395 |
> |
\label{introEquation:densityAndHamiltonian} |
396 |
> |
\end{equation} |
397 |
> |
|
398 |
> |
\subsubsection{\label{introSection:phaseSpaceConservation}Conservation of Phase Space} |
399 |
> |
Lets consider a region in the phase space, |
400 |
> |
\begin{equation} |
401 |
> |
\delta v = \int { \ldots \int {dq_1 } ...dq_f dp_1 } ..dp_f . |
402 |
> |
\end{equation} |
403 |
> |
If this region is small enough, the density $\rho$ can be regarded |
404 |
> |
as uniform over the whole phase space. Thus, the number of phase |
405 |
> |
points inside this region is given by, |
406 |
> |
\begin{equation} |
407 |
> |
\delta N = \rho \delta v = \rho \int { \ldots \int {dq_1 } ...dq_f |
408 |
> |
dp_1 } ..dp_f. |
409 |
> |
\end{equation} |
410 |
|
|
411 |
+ |
\begin{equation} |
412 |
+ |
\frac{{d(\delta N)}}{{dt}} = \frac{{d\rho }}{{dt}}\delta v + \rho |
413 |
+ |
\frac{d}{{dt}}(\delta v) = 0. |
414 |
+ |
\end{equation} |
415 |
+ |
With the help of stationary assumption |
416 |
+ |
(\ref{introEquation:stationary}), we obtain the principle of the |
417 |
+ |
\emph{conservation of extension in phase space}, |
418 |
+ |
\begin{equation} |
419 |
+ |
\frac{d}{{dt}}(\delta v) = \frac{d}{{dt}}\int { \ldots \int {dq_1 } |
420 |
+ |
...dq_f dp_1 } ..dp_f = 0. |
421 |
+ |
\label{introEquation:volumePreserving} |
422 |
+ |
\end{equation} |
423 |
|
|
424 |
< |
%\subsection{\label{introSection:poissonBrackets}Poisson Brackets} |
424 |
> |
\subsubsection{\label{introSection:liouvilleInOtherForms}Liouville's Theorem in Other Forms} |
425 |
|
|
426 |
< |
\section{\label{introSection:correlationFunctions}Correlation Functions} |
426 |
> |
Liouville's theorem can be expresses in a variety of different forms |
427 |
> |
which are convenient within different contexts. For any two function |
428 |
> |
$F$ and $G$ of the coordinates and momenta of a system, the Poisson |
429 |
> |
bracket ${F, G}$ is defined as |
430 |
> |
\begin{equation} |
431 |
> |
\left\{ {F,G} \right\} = \sum\limits_i {\left( {\frac{{\partial |
432 |
> |
F}}{{\partial q_i }}\frac{{\partial G}}{{\partial p_i }} - |
433 |
> |
\frac{{\partial F}}{{\partial p_i }}\frac{{\partial G}}{{\partial |
434 |
> |
q_i }}} \right)}. |
435 |
> |
\label{introEquation:poissonBracket} |
436 |
> |
\end{equation} |
437 |
> |
Substituting equations of motion in Hamiltonian formalism( |
438 |
> |
\ref{introEquation:motionHamiltonianCoordinate} , |
439 |
> |
\ref{introEquation:motionHamiltonianMomentum} ) into |
440 |
> |
(\ref{introEquation:liouvilleTheorem}), we can rewrite Liouville's |
441 |
> |
theorem using Poisson bracket notion, |
442 |
> |
\begin{equation} |
443 |
> |
\left( {\frac{{\partial \rho }}{{\partial t}}} \right) = - \left\{ |
444 |
> |
{\rho ,H} \right\}. |
445 |
> |
\label{introEquation:liouvilleTheromInPoissin} |
446 |
> |
\end{equation} |
447 |
> |
Moreover, the Liouville operator is defined as |
448 |
> |
\begin{equation} |
449 |
> |
iL = \sum\limits_{i = 1}^f {\left( {\frac{{\partial H}}{{\partial |
450 |
> |
p_i }}\frac{\partial }{{\partial q_i }} - \frac{{\partial |
451 |
> |
H}}{{\partial q_i }}\frac{\partial }{{\partial p_i }}} \right)} |
452 |
> |
\label{introEquation:liouvilleOperator} |
453 |
> |
\end{equation} |
454 |
> |
In terms of Liouville operator, Liouville's equation can also be |
455 |
> |
expressed as |
456 |
> |
\begin{equation} |
457 |
> |
\left( {\frac{{\partial \rho }}{{\partial t}}} \right) = - iL\rho |
458 |
> |
\label{introEquation:liouvilleTheoremInOperator} |
459 |
> |
\end{equation} |
460 |
|
|
461 |
< |
\section{\label{introSection:langevinDynamics}Langevin Dynamics} |
461 |
> |
\subsection{\label{introSection:ergodic}The Ergodic Hypothesis} |
462 |
|
|
463 |
< |
\subsection{\label{introSection:generalizedLangevinDynamics}Generalized Langevin Dynamics} |
463 |
> |
Various thermodynamic properties can be calculated from Molecular |
464 |
> |
Dynamics simulation. By comparing experimental values with the |
465 |
> |
calculated properties, one can determine the accuracy of the |
466 |
> |
simulation and the quality of the underlying model. However, both of |
467 |
> |
experiment and computer simulation are usually performed during a |
468 |
> |
certain time interval and the measurements are averaged over a |
469 |
> |
period of them which is different from the average behavior of |
470 |
> |
many-body system in Statistical Mechanics. Fortunately, Ergodic |
471 |
> |
Hypothesis is proposed to make a connection between time average and |
472 |
> |
ensemble average. It states that time average and average over the |
473 |
> |
statistical ensemble are identical \cite{Frenkel1996, Leach2001}. |
474 |
> |
\begin{equation} |
475 |
> |
\langle A(q , p) \rangle_t = \mathop {\lim }\limits_{t \to \infty } |
476 |
> |
\frac{1}{t}\int\limits_0^t {A(q(t),p(t))dt = \int\limits_\Gamma |
477 |
> |
{A(q(t),p(t))} } \rho (q(t), p(t)) dqdp |
478 |
> |
\end{equation} |
479 |
> |
where $\langle A(q , p) \rangle_t$ is an equilibrium value of a |
480 |
> |
physical quantity and $\rho (p(t), q(t))$ is the equilibrium |
481 |
> |
distribution function. If an observation is averaged over a |
482 |
> |
sufficiently long time (longer than relaxation time), all accessible |
483 |
> |
microstates in phase space are assumed to be equally probed, giving |
484 |
> |
a properly weighted statistical average. This allows the researcher |
485 |
> |
freedom of choice when deciding how best to measure a given |
486 |
> |
observable. In case an ensemble averaged approach sounds most |
487 |
> |
reasonable, the Monte Carlo techniques\cite{Metropolis1949} can be |
488 |
> |
utilized. Or if the system lends itself to a time averaging |
489 |
> |
approach, the Molecular Dynamics techniques in |
490 |
> |
Sec.~\ref{introSection:molecularDynamics} will be the best |
491 |
> |
choice\cite{Frenkel1996}. |
492 |
|
|
493 |
< |
\subsection{\label{introSection:hydroynamics}Hydrodynamics} |
493 |
> |
\section{\label{introSection:geometricIntegratos}Geometric Integrators} |
494 |
> |
A variety of numerical integrators were proposed to simulate the |
495 |
> |
motions. They usually begin with an initial conditionals and move |
496 |
> |
the objects in the direction governed by the differential equations. |
497 |
> |
However, most of them ignore the hidden physical law contained |
498 |
> |
within the equations. Since 1990, geometric integrators, which |
499 |
> |
preserve various phase-flow invariants such as symplectic structure, |
500 |
> |
volume and time reversal symmetry, are developed to address this |
501 |
> |
issue\cite{Dullweber1997, McLachlan1998, Leimkuhler1999}. The |
502 |
> |
velocity verlet method, which happens to be a simple example of |
503 |
> |
symplectic integrator, continues to gain its popularity in molecular |
504 |
> |
dynamics community. This fact can be partly explained by its |
505 |
> |
geometric nature. |
506 |
> |
|
507 |
> |
\subsection{\label{introSection:symplecticManifold}Symplectic Manifold} |
508 |
> |
A \emph{manifold} is an abstract mathematical space. It locally |
509 |
> |
looks like Euclidean space, but when viewed globally, it may have |
510 |
> |
more complicate structure. A good example of manifold is the surface |
511 |
> |
of Earth. It seems to be flat locally, but it is round if viewed as |
512 |
> |
a whole. A \emph{differentiable manifold} (also known as |
513 |
> |
\emph{smooth manifold}) is a manifold with an open cover in which |
514 |
> |
the covering neighborhoods are all smoothly isomorphic to one |
515 |
> |
another. In other words,it is possible to apply calculus on |
516 |
> |
\emph{differentiable manifold}. A \emph{symplectic manifold} is |
517 |
> |
defined as a pair $(M, \omega)$ which consisting of a |
518 |
> |
\emph{differentiable manifold} $M$ and a close, non-degenerated, |
519 |
> |
bilinear symplectic form, $\omega$. A symplectic form on a vector |
520 |
> |
space $V$ is a function $\omega(x, y)$ which satisfies |
521 |
> |
$\omega(\lambda_1x_1+\lambda_2x_2, y) = \lambda_1\omega(x_1, y)+ |
522 |
> |
\lambda_2\omega(x_2, y)$, $\omega(x, y) = - \omega(y, x)$ and |
523 |
> |
$\omega(x, x) = 0$. Cross product operation in vector field is an |
524 |
> |
example of symplectic form. |
525 |
> |
|
526 |
> |
One of the motivations to study \emph{symplectic manifold} in |
527 |
> |
Hamiltonian Mechanics is that a symplectic manifold can represent |
528 |
> |
all possible configurations of the system and the phase space of the |
529 |
> |
system can be described by it's cotangent bundle. Every symplectic |
530 |
> |
manifold is even dimensional. For instance, in Hamilton equations, |
531 |
> |
coordinate and momentum always appear in pairs. |
532 |
> |
|
533 |
> |
Let $(M,\omega)$ and $(N, \eta)$ be symplectic manifolds. A map |
534 |
> |
\[ |
535 |
> |
f : M \rightarrow N |
536 |
> |
\] |
537 |
> |
is a \emph{symplectomorphism} if it is a \emph{diffeomorphims} and |
538 |
> |
the \emph{pullback} of $\eta$ under f is equal to $\omega$. |
539 |
> |
Canonical transformation is an example of symplectomorphism in |
540 |
> |
classical mechanics. |
541 |
> |
|
542 |
> |
\subsection{\label{introSection:ODE}Ordinary Differential Equations} |
543 |
> |
|
544 |
> |
For a ordinary differential system defined as |
545 |
> |
\begin{equation} |
546 |
> |
\dot x = f(x) |
547 |
> |
\end{equation} |
548 |
> |
where $x = x(q,p)^T$, this system is canonical Hamiltonian, if |
549 |
> |
\begin{equation} |
550 |
> |
f(r) = J\nabla _x H(r). |
551 |
> |
\end{equation} |
552 |
> |
$H = H (q, p)$ is Hamiltonian function and $J$ is the skew-symmetric |
553 |
> |
matrix |
554 |
> |
\begin{equation} |
555 |
> |
J = \left( {\begin{array}{*{20}c} |
556 |
> |
0 & I \\ |
557 |
> |
{ - I} & 0 \\ |
558 |
> |
\end{array}} \right) |
559 |
> |
\label{introEquation:canonicalMatrix} |
560 |
> |
\end{equation} |
561 |
> |
where $I$ is an identity matrix. Using this notation, Hamiltonian |
562 |
> |
system can be rewritten as, |
563 |
> |
\begin{equation} |
564 |
> |
\frac{d}{{dt}}x = J\nabla _x H(x) |
565 |
> |
\label{introEquation:compactHamiltonian} |
566 |
> |
\end{equation}In this case, $f$ is |
567 |
> |
called a \emph{Hamiltonian vector field}. |
568 |
> |
|
569 |
> |
Another generalization of Hamiltonian dynamics is Poisson |
570 |
> |
Dynamics\cite{Olver1986}, |
571 |
> |
\begin{equation} |
572 |
> |
\dot x = J(x)\nabla _x H \label{introEquation:poissonHamiltonian} |
573 |
> |
\end{equation} |
574 |
> |
The most obvious change being that matrix $J$ now depends on $x$. |
575 |
> |
|
576 |
> |
\subsection{\label{introSection:exactFlow}Exact Flow} |
577 |
> |
|
578 |
> |
Let $x(t)$ be the exact solution of the ODE system, |
579 |
> |
\begin{equation} |
580 |
> |
\frac{{dx}}{{dt}} = f(x) \label{introEquation:ODE} |
581 |
> |
\end{equation} |
582 |
> |
The exact flow(solution) $\varphi_\tau$ is defined by |
583 |
> |
\[ |
584 |
> |
x(t+\tau) =\varphi_\tau(x(t)) |
585 |
> |
\] |
586 |
> |
where $\tau$ is a fixed time step and $\varphi$ is a map from phase |
587 |
> |
space to itself. The flow has the continuous group property, |
588 |
> |
\begin{equation} |
589 |
> |
\varphi _{\tau _1 } \circ \varphi _{\tau _2 } = \varphi _{\tau _1 |
590 |
> |
+ \tau _2 } . |
591 |
> |
\end{equation} |
592 |
> |
In particular, |
593 |
> |
\begin{equation} |
594 |
> |
\varphi _\tau \circ \varphi _{ - \tau } = I |
595 |
> |
\end{equation} |
596 |
> |
Therefore, the exact flow is self-adjoint, |
597 |
> |
\begin{equation} |
598 |
> |
\varphi _\tau = \varphi _{ - \tau }^{ - 1}. |
599 |
> |
\end{equation} |
600 |
> |
The exact flow can also be written in terms of the of an operator, |
601 |
> |
\begin{equation} |
602 |
> |
\varphi _\tau (x) = e^{\tau \sum\limits_i {f_i (x)\frac{\partial |
603 |
> |
}{{\partial x_i }}} } (x) \equiv \exp (\tau f)(x). |
604 |
> |
\label{introEquation:exponentialOperator} |
605 |
> |
\end{equation} |
606 |
> |
|
607 |
> |
In most cases, it is not easy to find the exact flow $\varphi_\tau$. |
608 |
> |
Instead, we use a approximate map, $\psi_\tau$, which is usually |
609 |
> |
called integrator. The order of an integrator $\psi_\tau$ is $p$, if |
610 |
> |
the Taylor series of $\psi_\tau$ agree to order $p$, |
611 |
> |
\begin{equation} |
612 |
> |
\psi_tau(x) = x + \tau f(x) + O(\tau^{p+1}) |
613 |
> |
\end{equation} |
614 |
> |
|
615 |
> |
\subsection{\label{introSection:geometricProperties}Geometric Properties} |
616 |
> |
|
617 |
> |
The hidden geometric properties\cite{Budd1999, Marsden1998} of ODE |
618 |
> |
and its flow play important roles in numerical studies. Many of them |
619 |
> |
can be found in systems which occur naturally in applications. |
620 |
> |
|
621 |
> |
Let $\varphi$ be the flow of Hamiltonian vector field, $\varphi$ is |
622 |
> |
a \emph{symplectic} flow if it satisfies, |
623 |
> |
\begin{equation} |
624 |
> |
{\varphi '}^T J \varphi ' = J. |
625 |
> |
\end{equation} |
626 |
> |
According to Liouville's theorem, the symplectic volume is invariant |
627 |
> |
under a Hamiltonian flow, which is the basis for classical |
628 |
> |
statistical mechanics. Furthermore, the flow of a Hamiltonian vector |
629 |
> |
field on a symplectic manifold can be shown to be a |
630 |
> |
symplectomorphism. As to the Poisson system, |
631 |
> |
\begin{equation} |
632 |
> |
{\varphi '}^T J \varphi ' = J \circ \varphi |
633 |
> |
\end{equation} |
634 |
> |
is the property must be preserved by the integrator. |
635 |
> |
|
636 |
> |
It is possible to construct a \emph{volume-preserving} flow for a |
637 |
> |
source free($ \nabla \cdot f = 0 $) ODE, if the flow satisfies $ |
638 |
> |
\det d\varphi = 1$. One can show easily that a symplectic flow will |
639 |
> |
be volume-preserving. |
640 |
> |
|
641 |
> |
Changing the variables $y = h(x)$ in a ODE\ref{introEquation:ODE} |
642 |
> |
will result in a new system, |
643 |
> |
\[ |
644 |
> |
\dot y = \tilde f(y) = ((dh \cdot f)h^{ - 1} )(y). |
645 |
> |
\] |
646 |
> |
The vector filed $f$ has reversing symmetry $h$ if $f = - \tilde f$. |
647 |
> |
In other words, the flow of this vector field is reversible if and |
648 |
> |
only if $ h \circ \varphi ^{ - 1} = \varphi \circ h $. |
649 |
> |
|
650 |
> |
A \emph{first integral}, or conserved quantity of a general |
651 |
> |
differential function is a function $ G:R^{2d} \to R^d $ which is |
652 |
> |
constant for all solutions of the ODE $\frac{{dx}}{{dt}} = f(x)$ , |
653 |
> |
\[ |
654 |
> |
\frac{{dG(x(t))}}{{dt}} = 0. |
655 |
> |
\] |
656 |
> |
Using chain rule, one may obtain, |
657 |
> |
\[ |
658 |
> |
\sum\limits_i {\frac{{dG}}{{dx_i }}} f_i (x) = f \bullet \nabla G, |
659 |
> |
\] |
660 |
> |
which is the condition for conserving \emph{first integral}. For a |
661 |
> |
canonical Hamiltonian system, the time evolution of an arbitrary |
662 |
> |
smooth function $G$ is given by, |
663 |
> |
|
664 |
> |
\begin{eqnarray} |
665 |
> |
\frac{{dG(x(t))}}{{dt}} & = & [\nabla _x G(x(t))]^T \dot x(t) \\ |
666 |
> |
& = & [\nabla _x G(x(t))]^T J\nabla _x H(x(t)). \\ |
667 |
> |
\label{introEquation:firstIntegral1} |
668 |
> |
\end{eqnarray} |
669 |
> |
|
670 |
> |
|
671 |
> |
Using poisson bracket notion, Equation |
672 |
> |
\ref{introEquation:firstIntegral1} can be rewritten as |
673 |
> |
\[ |
674 |
> |
\frac{d}{{dt}}G(x(t)) = \left\{ {G,H} \right\}(x(t)). |
675 |
> |
\] |
676 |
> |
Therefore, the sufficient condition for $G$ to be the \emph{first |
677 |
> |
integral} of a Hamiltonian system is |
678 |
> |
\[ |
679 |
> |
\left\{ {G,H} \right\} = 0. |
680 |
> |
\] |
681 |
> |
As well known, the Hamiltonian (or energy) H of a Hamiltonian system |
682 |
> |
is a \emph{first integral}, which is due to the fact $\{ H,H\} = |
683 |
> |
0$. |
684 |
> |
|
685 |
> |
When designing any numerical methods, one should always try to |
686 |
> |
preserve the structural properties of the original ODE and its flow. |
687 |
> |
|
688 |
> |
\subsection{\label{introSection:constructionSymplectic}Construction of Symplectic Methods} |
689 |
> |
A lot of well established and very effective numerical methods have |
690 |
> |
been successful precisely because of their symplecticities even |
691 |
> |
though this fact was not recognized when they were first |
692 |
> |
constructed. The most famous example is leapfrog methods in |
693 |
> |
molecular dynamics. In general, symplectic integrators can be |
694 |
> |
constructed using one of four different methods. |
695 |
> |
\begin{enumerate} |
696 |
> |
\item Generating functions |
697 |
> |
\item Variational methods |
698 |
> |
\item Runge-Kutta methods |
699 |
> |
\item Splitting methods |
700 |
> |
\end{enumerate} |
701 |
> |
|
702 |
> |
Generating function\cite{Channell1990} tends to lead to methods |
703 |
> |
which are cumbersome and difficult to use. In dissipative systems, |
704 |
> |
variational methods can capture the decay of energy |
705 |
> |
accurately\cite{Kane2000}. Since their geometrically unstable nature |
706 |
> |
against non-Hamiltonian perturbations, ordinary implicit Runge-Kutta |
707 |
> |
methods are not suitable for Hamiltonian system. Recently, various |
708 |
> |
high-order explicit Runge-Kutta methods |
709 |
> |
\cite{Owren1992,Chen2003}have been developed to overcome this |
710 |
> |
instability. However, due to computational penalty involved in |
711 |
> |
implementing the Runge-Kutta methods, they do not attract too much |
712 |
> |
attention from Molecular Dynamics community. Instead, splitting have |
713 |
> |
been widely accepted since they exploit natural decompositions of |
714 |
> |
the system\cite{Tuckerman1992, McLachlan1998}. |
715 |
> |
|
716 |
> |
\subsubsection{\label{introSection:splittingMethod}Splitting Method} |
717 |
> |
|
718 |
> |
The main idea behind splitting methods is to decompose the discrete |
719 |
> |
$\varphi_h$ as a composition of simpler flows, |
720 |
> |
\begin{equation} |
721 |
> |
\varphi _h = \varphi _{h_1 } \circ \varphi _{h_2 } \ldots \circ |
722 |
> |
\varphi _{h_n } |
723 |
> |
\label{introEquation:FlowDecomposition} |
724 |
> |
\end{equation} |
725 |
> |
where each of the sub-flow is chosen such that each represent a |
726 |
> |
simpler integration of the system. |
727 |
> |
|
728 |
> |
Suppose that a Hamiltonian system takes the form, |
729 |
> |
\[ |
730 |
> |
H = H_1 + H_2. |
731 |
> |
\] |
732 |
> |
Here, $H_1$ and $H_2$ may represent different physical processes of |
733 |
> |
the system. For instance, they may relate to kinetic and potential |
734 |
> |
energy respectively, which is a natural decomposition of the |
735 |
> |
problem. If $H_1$ and $H_2$ can be integrated using exact flows |
736 |
> |
$\varphi_1(t)$ and $\varphi_2(t)$, respectively, a simple first |
737 |
> |
order is then given by the Lie-Trotter formula |
738 |
> |
\begin{equation} |
739 |
> |
\varphi _h = \varphi _{1,h} \circ \varphi _{2,h}, |
740 |
> |
\label{introEquation:firstOrderSplitting} |
741 |
> |
\end{equation} |
742 |
> |
where $\varphi _h$ is the result of applying the corresponding |
743 |
> |
continuous $\varphi _i$ over a time $h$. By definition, as |
744 |
> |
$\varphi_i(t)$ is the exact solution of a Hamiltonian system, it |
745 |
> |
must follow that each operator $\varphi_i(t)$ is a symplectic map. |
746 |
> |
It is easy to show that any composition of symplectic flows yields a |
747 |
> |
symplectic map, |
748 |
> |
\begin{equation} |
749 |
> |
(\varphi '\phi ')^T J\varphi '\phi ' = \phi '^T \varphi '^T J\varphi |
750 |
> |
'\phi ' = \phi '^T J\phi ' = J, |
751 |
> |
\label{introEquation:SymplecticFlowComposition} |
752 |
> |
\end{equation} |
753 |
> |
where $\phi$ and $\psi$ both are symplectic maps. Thus operator |
754 |
> |
splitting in this context automatically generates a symplectic map. |
755 |
> |
|
756 |
> |
The Lie-Trotter splitting(\ref{introEquation:firstOrderSplitting}) |
757 |
> |
introduces local errors proportional to $h^2$, while Strang |
758 |
> |
splitting gives a second-order decomposition, |
759 |
> |
\begin{equation} |
760 |
> |
\varphi _h = \varphi _{1,h/2} \circ \varphi _{2,h} \circ \varphi |
761 |
> |
_{1,h/2} , \label{introEquation:secondOrderSplitting} |
762 |
> |
\end{equation} |
763 |
> |
which has a local error proportional to $h^3$. Sprang splitting's |
764 |
> |
popularity in molecular simulation community attribute to its |
765 |
> |
symmetric property, |
766 |
> |
\begin{equation} |
767 |
> |
\varphi _h^{ - 1} = \varphi _{ - h}. |
768 |
> |
\label{introEquation:timeReversible} |
769 |
> |
\end{equation} |
770 |
> |
|
771 |
> |
\subsubsection{\label{introSection:exampleSplittingMethod}Example of Splitting Method} |
772 |
> |
The classical equation for a system consisting of interacting |
773 |
> |
particles can be written in Hamiltonian form, |
774 |
> |
\[ |
775 |
> |
H = T + V |
776 |
> |
\] |
777 |
> |
where $T$ is the kinetic energy and $V$ is the potential energy. |
778 |
> |
Setting $H_1 = T, H_2 = V$ and applying Strang splitting, one |
779 |
> |
obtains the following: |
780 |
> |
\begin{align} |
781 |
> |
q(\Delta t) &= q(0) + \dot{q}(0)\Delta t + |
782 |
> |
\frac{F[q(0)]}{m}\frac{\Delta t^2}{2}, % |
783 |
> |
\label{introEquation:Lp10a} \\% |
784 |
> |
% |
785 |
> |
\dot{q}(\Delta t) &= \dot{q}(0) + \frac{\Delta t}{2m} |
786 |
> |
\biggl [F[q(0)] + F[q(\Delta t)] \biggr]. % |
787 |
> |
\label{introEquation:Lp10b} |
788 |
> |
\end{align} |
789 |
> |
where $F(t)$ is the force at time $t$. This integration scheme is |
790 |
> |
known as \emph{velocity verlet} which is |
791 |
> |
symplectic(\ref{introEquation:SymplecticFlowComposition}), |
792 |
> |
time-reversible(\ref{introEquation:timeReversible}) and |
793 |
> |
volume-preserving (\ref{introEquation:volumePreserving}). These |
794 |
> |
geometric properties attribute to its long-time stability and its |
795 |
> |
popularity in the community. However, the most commonly used |
796 |
> |
velocity verlet integration scheme is written as below, |
797 |
> |
\begin{align} |
798 |
> |
\dot{q}\biggl (\frac{\Delta t}{2}\biggr ) &= |
799 |
> |
\dot{q}(0) + \frac{\Delta t}{2m}\, F[q(0)], \label{introEquation:Lp9a}\\% |
800 |
> |
% |
801 |
> |
q(\Delta t) &= q(0) + \Delta t\, \dot{q}\biggl (\frac{\Delta t}{2}\biggr ),% |
802 |
> |
\label{introEquation:Lp9b}\\% |
803 |
> |
% |
804 |
> |
\dot{q}(\Delta t) &= \dot{q}\biggl (\frac{\Delta t}{2}\biggr ) + |
805 |
> |
\frac{\Delta t}{2m}\, F[q(0)]. \label{introEquation:Lp9c} |
806 |
> |
\end{align} |
807 |
> |
From the preceding splitting, one can see that the integration of |
808 |
> |
the equations of motion would follow: |
809 |
> |
\begin{enumerate} |
810 |
> |
\item calculate the velocities at the half step, $\frac{\Delta t}{2}$, from the forces calculated at the initial position. |
811 |
> |
|
812 |
> |
\item Use the half step velocities to move positions one whole step, $\Delta t$. |
813 |
> |
|
814 |
> |
\item Evaluate the forces at the new positions, $\mathbf{r}(\Delta t)$, and use the new forces to complete the velocity move. |
815 |
> |
|
816 |
> |
\item Repeat from step 1 with the new position, velocities, and forces assuming the roles of the initial values. |
817 |
> |
\end{enumerate} |
818 |
> |
|
819 |
> |
Simply switching the order of splitting and composing, a new |
820 |
> |
integrator, the \emph{position verlet} integrator, can be generated, |
821 |
> |
\begin{align} |
822 |
> |
\dot q(\Delta t) &= \dot q(0) + \Delta tF(q(0))\left[ {q(0) + |
823 |
> |
\frac{{\Delta t}}{{2m}}\dot q(0)} \right], % |
824 |
> |
\label{introEquation:positionVerlet1} \\% |
825 |
> |
% |
826 |
> |
q(\Delta t) &= q(0) + \frac{{\Delta t}}{2}\left[ {\dot q(0) + \dot |
827 |
> |
q(\Delta t)} \right]. % |
828 |
> |
\label{introEquation:positionVerlet2} |
829 |
> |
\end{align} |
830 |
> |
|
831 |
> |
\subsubsection{\label{introSection:errorAnalysis}Error Analysis and Higher Order Methods} |
832 |
> |
|
833 |
> |
Baker-Campbell-Hausdorff formula can be used to determine the local |
834 |
> |
error of splitting method in terms of commutator of the |
835 |
> |
operators(\ref{introEquation:exponentialOperator}) associated with |
836 |
> |
the sub-flow. For operators $hX$ and $hY$ which are associate to |
837 |
> |
$\varphi_1(t)$ and $\varphi_2(t)$ respectively , we have |
838 |
> |
\begin{equation} |
839 |
> |
\exp (hX + hY) = \exp (hZ) |
840 |
> |
\end{equation} |
841 |
> |
where |
842 |
> |
\begin{equation} |
843 |
> |
hZ = hX + hY + \frac{{h^2 }}{2}[X,Y] + \frac{{h^3 }}{2}\left( |
844 |
> |
{[X,[X,Y]] + [Y,[Y,X]]} \right) + \ldots . |
845 |
> |
\end{equation} |
846 |
> |
Here, $[X,Y]$ is the commutators of operator $X$ and $Y$ given by |
847 |
> |
\[ |
848 |
> |
[X,Y] = XY - YX . |
849 |
> |
\] |
850 |
> |
Applying Baker-Campbell-Hausdorff formula\cite{Varadarajan1974} to |
851 |
> |
Sprang splitting, we can obtain |
852 |
> |
\begin{eqnarray*} |
853 |
> |
\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 \\ |
854 |
> |
& & \mbox{} + h^2 [X,X]/8 + h^2 [Y,Y]/8 \\ |
855 |
> |
& & \mbox{} + h^3 [Y,[Y,X]]/12 - h^3[X,[X,Y]]/24 + \ldots ) |
856 |
> |
\end{eqnarray*} |
857 |
> |
Since \[ [X,Y] + [Y,X] = 0\] and \[ [X,X] = 0\], the dominant local |
858 |
> |
error of Spring splitting is proportional to $h^3$. The same |
859 |
> |
procedure can be applied to general splitting, of the form |
860 |
> |
\begin{equation} |
861 |
> |
\varphi _{b_m h}^2 \circ \varphi _{a_m h}^1 \circ \varphi _{b_{m - |
862 |
> |
1} h}^2 \circ \ldots \circ \varphi _{a_1 h}^1 . |
863 |
> |
\end{equation} |
864 |
> |
Careful choice of coefficient $a_1 \ldots b_m$ will lead to higher |
865 |
> |
order method. Yoshida proposed an elegant way to compose higher |
866 |
> |
order methods based on symmetric splitting\cite{Yoshida1990}. Given |
867 |
> |
a symmetric second order base method $ \varphi _h^{(2)} $, a |
868 |
> |
fourth-order symmetric method can be constructed by composing, |
869 |
> |
\[ |
870 |
> |
\varphi _h^{(4)} = \varphi _{\alpha h}^{(2)} \circ \varphi _{\beta |
871 |
> |
h}^{(2)} \circ \varphi _{\alpha h}^{(2)} |
872 |
> |
\] |
873 |
> |
where $ \alpha = - \frac{{2^{1/3} }}{{2 - 2^{1/3} }}$ and $ \beta |
874 |
> |
= \frac{{2^{1/3} }}{{2 - 2^{1/3} }}$. Moreover, a symmetric |
875 |
> |
integrator $ \varphi _h^{(2n + 2)}$ can be composed by |
876 |
> |
\begin{equation} |
877 |
> |
\varphi _h^{(2n + 2)} = \varphi _{\alpha h}^{(2n)} \circ \varphi |
878 |
> |
_{\beta h}^{(2n)} \circ \varphi _{\alpha h}^{(2n)} |
879 |
> |
\end{equation} |
880 |
> |
, if the weights are chosen as |
881 |
> |
\[ |
882 |
> |
\alpha = - \frac{{2^{1/(2n + 1)} }}{{2 - 2^{1/(2n + 1)} }},\beta = |
883 |
> |
\frac{{2^{1/(2n + 1)} }}{{2 - 2^{1/(2n + 1)} }} . |
884 |
> |
\] |
885 |
> |
|
886 |
> |
\section{\label{introSection:molecularDynamics}Molecular Dynamics} |
887 |
> |
|
888 |
> |
As one of the principal tools of molecular modeling, Molecular |
889 |
> |
dynamics has proven to be a powerful tool for studying the functions |
890 |
> |
of biological systems, providing structural, thermodynamic and |
891 |
> |
dynamical information. The basic idea of molecular dynamics is that |
892 |
> |
macroscopic properties are related to microscopic behavior and |
893 |
> |
microscopic behavior can be calculated from the trajectories in |
894 |
> |
simulations. For instance, instantaneous temperature of an |
895 |
> |
Hamiltonian system of $N$ particle can be measured by |
896 |
> |
\[ |
897 |
> |
T = \sum\limits_{i = 1}^N {\frac{{m_i v_i^2 }}{{fk_B }}} |
898 |
> |
\] |
899 |
> |
where $m_i$ and $v_i$ are the mass and velocity of $i$th particle |
900 |
> |
respectively, $f$ is the number of degrees of freedom, and $k_B$ is |
901 |
> |
the boltzman constant. |
902 |
> |
|
903 |
> |
A typical molecular dynamics run consists of three essential steps: |
904 |
> |
\begin{enumerate} |
905 |
> |
\item Initialization |
906 |
> |
\begin{enumerate} |
907 |
> |
\item Preliminary preparation |
908 |
> |
\item Minimization |
909 |
> |
\item Heating |
910 |
> |
\item Equilibration |
911 |
> |
\end{enumerate} |
912 |
> |
\item Production |
913 |
> |
\item Analysis |
914 |
> |
\end{enumerate} |
915 |
> |
These three individual steps will be covered in the following |
916 |
> |
sections. Sec.~\ref{introSec:initialSystemSettings} deals with the |
917 |
> |
initialization of a simulation. Sec.~\ref{introSec:production} will |
918 |
> |
discusses issues in production run. Sec.~\ref{introSection:Analysis} |
919 |
> |
provides the theoretical tools for trajectory analysis. |
920 |
> |
|
921 |
> |
\subsection{\label{introSec:initialSystemSettings}Initialization} |
922 |
> |
|
923 |
> |
\subsubsection{Preliminary preparation} |
924 |
> |
|
925 |
> |
When selecting the starting structure of a molecule for molecular |
926 |
> |
simulation, one may retrieve its Cartesian coordinates from public |
927 |
> |
databases, such as RCSB Protein Data Bank \textit{etc}. Although |
928 |
> |
thousands of crystal structures of molecules are discovered every |
929 |
> |
year, many more remain unknown due to the difficulties of |
930 |
> |
purification and crystallization. Even for the molecule with known |
931 |
> |
structure, some important information is missing. For example, the |
932 |
> |
missing hydrogen atom which acts as donor in hydrogen bonding must |
933 |
> |
be added. Moreover, in order to include electrostatic interaction, |
934 |
> |
one may need to specify the partial charges for individual atoms. |
935 |
> |
Under some circumstances, we may even need to prepare the system in |
936 |
> |
a special setup. For instance, when studying transport phenomenon in |
937 |
> |
membrane system, we may prepare the lipids in bilayer structure |
938 |
> |
instead of placing lipids randomly in solvent, since we are not |
939 |
> |
interested in self-aggregation and it takes a long time to happen. |
940 |
> |
|
941 |
> |
\subsubsection{Minimization} |
942 |
> |
|
943 |
> |
It is quite possible that some of molecules in the system from |
944 |
> |
preliminary preparation may be overlapped with each other. This |
945 |
> |
close proximity leads to high potential energy which consequently |
946 |
> |
jeopardizes any molecular dynamics simulations. To remove these |
947 |
> |
steric overlaps, one typically performs energy minimization to find |
948 |
> |
a more reasonable conformation. Several energy minimization methods |
949 |
> |
have been developed to exploit the energy surface and to locate the |
950 |
> |
local minimum. While converging slowly near the minimum, steepest |
951 |
> |
descent method is extremely robust when systems are far from |
952 |
> |
harmonic. Thus, it is often used to refine structure from |
953 |
> |
crystallographic data. Relied on the gradient or hessian, advanced |
954 |
> |
methods like conjugate gradient and Newton-Raphson converge rapidly |
955 |
> |
to a local minimum, while become unstable if the energy surface is |
956 |
> |
far from quadratic. Another factor must be taken into account, when |
957 |
> |
choosing energy minimization method, is the size of the system. |
958 |
> |
Steepest descent and conjugate gradient can deal with models of any |
959 |
> |
size. Because of the limit of computation power to calculate hessian |
960 |
> |
matrix and insufficient storage capacity to store them, most |
961 |
> |
Newton-Raphson methods can not be used with very large models. |
962 |
> |
|
963 |
> |
\subsubsection{Heating} |
964 |
> |
|
965 |
> |
Typically, Heating is performed by assigning random velocities |
966 |
> |
according to a Gaussian distribution for a temperature. Beginning at |
967 |
> |
a lower temperature and gradually increasing the temperature by |
968 |
> |
assigning greater random velocities, we end up with setting the |
969 |
> |
temperature of the system to a final temperature at which the |
970 |
> |
simulation will be conducted. In heating phase, we should also keep |
971 |
> |
the system from drifting or rotating as a whole. Equivalently, the |
972 |
> |
net linear momentum and angular momentum of the system should be |
973 |
> |
shifted to zero. |
974 |
> |
|
975 |
> |
\subsubsection{Equilibration} |
976 |
> |
|
977 |
> |
The purpose of equilibration is to allow the system to evolve |
978 |
> |
spontaneously for a period of time and reach equilibrium. The |
979 |
> |
procedure is continued until various statistical properties, such as |
980 |
> |
temperature, pressure, energy, volume and other structural |
981 |
> |
properties \textit{etc}, become independent of time. Strictly |
982 |
> |
speaking, minimization and heating are not necessary, provided the |
983 |
> |
equilibration process is long enough. However, these steps can serve |
984 |
> |
as a means to arrive at an equilibrated structure in an effective |
985 |
> |
way. |
986 |
> |
|
987 |
> |
\subsection{\label{introSection:production}Production} |
988 |
> |
|
989 |
> |
Production run is the most important step of the simulation, in |
990 |
> |
which the equilibrated structure is used as a starting point and the |
991 |
> |
motions of the molecules are collected for later analysis. In order |
992 |
> |
to capture the macroscopic properties of the system, the molecular |
993 |
> |
dynamics simulation must be performed in correct and efficient way. |
994 |
> |
|
995 |
> |
The most expensive part of a molecular dynamics simulation is the |
996 |
> |
calculation of non-bonded forces, such as van der Waals force and |
997 |
> |
Coulombic forces \textit{etc}. For a system of $N$ particles, the |
998 |
> |
complexity of the algorithm for pair-wise interactions is $O(N^2 )$, |
999 |
> |
which making large simulations prohibitive in the absence of any |
1000 |
> |
computation saving techniques. |
1001 |
> |
|
1002 |
> |
A natural approach to avoid system size issue is to represent the |
1003 |
> |
bulk behavior by a finite number of the particles. However, this |
1004 |
> |
approach will suffer from the surface effect. To offset this, |
1005 |
> |
\textit{Periodic boundary condition} (see Fig.~\ref{introFig:pbc}) |
1006 |
> |
is developed to simulate bulk properties with a relatively small |
1007 |
> |
number of particles. In this method, the simulation box is |
1008 |
> |
replicated throughout space to form an infinite lattice. During the |
1009 |
> |
simulation, when a particle moves in the primary cell, its image in |
1010 |
> |
other cells move in exactly the same direction with exactly the same |
1011 |
> |
orientation. Thus, as a particle leaves the primary cell, one of its |
1012 |
> |
images will enter through the opposite face. |
1013 |
> |
\begin{figure} |
1014 |
> |
\centering |
1015 |
> |
\includegraphics[width=\linewidth]{pbc.eps} |
1016 |
> |
\caption[An illustration of periodic boundary conditions]{A 2-D |
1017 |
> |
illustration of periodic boundary conditions. As one particle leaves |
1018 |
> |
the left of the simulation box, an image of it enters the right.} |
1019 |
> |
\label{introFig:pbc} |
1020 |
> |
\end{figure} |
1021 |
> |
|
1022 |
> |
%cutoff and minimum image convention |
1023 |
> |
Another important technique to improve the efficiency of force |
1024 |
> |
evaluation is to apply cutoff where particles farther than a |
1025 |
> |
predetermined distance, are not included in the calculation |
1026 |
> |
\cite{Frenkel1996}. The use of a cutoff radius will cause a |
1027 |
> |
discontinuity in the potential energy curve. Fortunately, one can |
1028 |
> |
shift the potential to ensure the potential curve go smoothly to |
1029 |
> |
zero at the cutoff radius. Cutoff strategy works pretty well for |
1030 |
> |
Lennard-Jones interaction because of its short range nature. |
1031 |
> |
However, simply truncating the electrostatic interaction with the |
1032 |
> |
use of cutoff has been shown to lead to severe artifacts in |
1033 |
> |
simulations. Ewald summation, in which the slowly conditionally |
1034 |
> |
convergent Coulomb potential is transformed into direct and |
1035 |
> |
reciprocal sums with rapid and absolute convergence, has proved to |
1036 |
> |
minimize the periodicity artifacts in liquid simulations. Taking the |
1037 |
> |
advantages of the fast Fourier transform (FFT) for calculating |
1038 |
> |
discrete Fourier transforms, the particle mesh-based |
1039 |
> |
methods\cite{Hockney1981,Shimada1993, Luty1994} are accelerated from |
1040 |
> |
$O(N^{3/2})$ to $O(N logN)$. An alternative approach is \emph{fast |
1041 |
> |
multipole method}\cite{Greengard1987, Greengard1994}, which treats |
1042 |
> |
Coulombic interaction exactly at short range, and approximate the |
1043 |
> |
potential at long range through multipolar expansion. In spite of |
1044 |
> |
their wide acceptances at the molecular simulation community, these |
1045 |
> |
two methods are hard to be implemented correctly and efficiently. |
1046 |
> |
Instead, we use a damped and charge-neutralized Coulomb potential |
1047 |
> |
method developed by Wolf and his coworkers\cite{Wolf1999}. The |
1048 |
> |
shifted Coulomb potential for particle $i$ and particle $j$ at |
1049 |
> |
distance $r_{rj}$ is given by: |
1050 |
> |
\begin{equation} |
1051 |
> |
V(r_{ij})= \frac{q_i q_j \textrm{erfc}(\alpha |
1052 |
> |
r_{ij})}{r_{ij}}-\lim_{r_{ij}\rightarrow |
1053 |
> |
R_\textrm{c}}\left\{\frac{q_iq_j \textrm{erfc}(\alpha |
1054 |
> |
r_{ij})}{r_{ij}}\right\}. \label{introEquation:shiftedCoulomb} |
1055 |
> |
\end{equation} |
1056 |
> |
where $\alpha$ is the convergence parameter. Due to the lack of |
1057 |
> |
inherent periodicity and rapid convergence,this method is extremely |
1058 |
> |
efficient and easy to implement. |
1059 |
> |
\begin{figure} |
1060 |
> |
\centering |
1061 |
> |
\includegraphics[width=\linewidth]{shifted_coulomb.eps} |
1062 |
> |
\caption[An illustration of shifted Coulomb potential]{An |
1063 |
> |
illustration of shifted Coulomb potential.} |
1064 |
> |
\label{introFigure:shiftedCoulomb} |
1065 |
> |
\end{figure} |
1066 |
> |
|
1067 |
> |
%multiple time step |
1068 |
> |
|
1069 |
> |
\subsection{\label{introSection:Analysis} Analysis} |
1070 |
> |
|
1071 |
> |
Recently, advanced visualization technique are widely applied to |
1072 |
> |
monitor the motions of molecules. Although the dynamics of the |
1073 |
> |
system can be described qualitatively from animation, quantitative |
1074 |
> |
trajectory analysis are more appreciable. According to the |
1075 |
> |
principles of Statistical Mechanics, |
1076 |
> |
Sec.~\ref{introSection:statisticalMechanics}, one can compute |
1077 |
> |
thermodynamics properties, analyze fluctuations of structural |
1078 |
> |
parameters, and investigate time-dependent processes of the molecule |
1079 |
> |
from the trajectories. |
1080 |
> |
|
1081 |
> |
\subsubsection{\label{introSection:thermodynamicsProperties}Thermodynamics Properties} |
1082 |
> |
|
1083 |
> |
Thermodynamics properties, which can be expressed in terms of some |
1084 |
> |
function of the coordinates and momenta of all particles in the |
1085 |
> |
system, can be directly computed from molecular dynamics. The usual |
1086 |
> |
way to measure the pressure is based on virial theorem of Clausius |
1087 |
> |
which states that the virial is equal to $-3Nk_BT$. For a system |
1088 |
> |
with forces between particles, the total virial, $W$, contains the |
1089 |
> |
contribution from external pressure and interaction between the |
1090 |
> |
particles: |
1091 |
> |
\[ |
1092 |
> |
W = - 3PV + \left\langle {\sum\limits_{i < j} {r{}_{ij} \cdot |
1093 |
> |
f_{ij} } } \right\rangle |
1094 |
> |
\] |
1095 |
> |
where $f_{ij}$ is the force between particle $i$ and $j$ at a |
1096 |
> |
distance $r_{ij}$. Thus, the expression for the pressure is given |
1097 |
> |
by: |
1098 |
> |
\begin{equation} |
1099 |
> |
P = \frac{{Nk_B T}}{V} - \frac{1}{{3V}}\left\langle {\sum\limits_{i |
1100 |
> |
< j} {r{}_{ij} \cdot f_{ij} } } \right\rangle |
1101 |
> |
\end{equation} |
1102 |
> |
|
1103 |
> |
\subsubsection{\label{introSection:structuralProperties}Structural Properties} |
1104 |
> |
|
1105 |
> |
Structural Properties of a simple fluid can be described by a set of |
1106 |
> |
distribution functions. Among these functions,\emph{pair |
1107 |
> |
distribution function}, also known as \emph{radial distribution |
1108 |
> |
function}, is of most fundamental importance to liquid-state theory. |
1109 |
> |
Pair distribution function can be gathered by Fourier transforming |
1110 |
> |
raw data from a series of neutron diffraction experiments and |
1111 |
> |
integrating over the surface factor \cite{Powles1973}. The |
1112 |
> |
experiment result can serve as a criterion to justify the |
1113 |
> |
correctness of the theory. Moreover, various equilibrium |
1114 |
> |
thermodynamic and structural properties can also be expressed in |
1115 |
> |
terms of radial distribution function \cite{Allen1987}. |
1116 |
> |
|
1117 |
> |
A pair distribution functions $g(r)$ gives the probability that a |
1118 |
> |
particle $i$ will be located at a distance $r$ from a another |
1119 |
> |
particle $j$ in the system |
1120 |
> |
\[ |
1121 |
> |
g(r) = \frac{V}{{N^2 }}\left\langle {\sum\limits_i {\sum\limits_{j |
1122 |
> |
\ne i} {\delta (r - r_{ij} )} } } \right\rangle. |
1123 |
> |
\] |
1124 |
> |
Note that the delta function can be replaced by a histogram in |
1125 |
> |
computer simulation. Figure |
1126 |
> |
\ref{introFigure:pairDistributionFunction} shows a typical pair |
1127 |
> |
distribution function for the liquid argon system. The occurrence of |
1128 |
> |
several peaks in the plot of $g(r)$ suggests that it is more likely |
1129 |
> |
to find particles at certain radial values than at others. This is a |
1130 |
> |
result of the attractive interaction at such distances. Because of |
1131 |
> |
the strong repulsive forces at short distance, the probability of |
1132 |
> |
locating particles at distances less than about 2.5{\AA} from each |
1133 |
> |
other is essentially zero. |
1134 |
> |
|
1135 |
> |
%\begin{figure} |
1136 |
> |
%\centering |
1137 |
> |
%\includegraphics[width=\linewidth]{pdf.eps} |
1138 |
> |
%\caption[Pair distribution function for the liquid argon |
1139 |
> |
%]{Pair distribution function for the liquid argon} |
1140 |
> |
%\label{introFigure:pairDistributionFunction} |
1141 |
> |
%\end{figure} |
1142 |
> |
|
1143 |
> |
\subsubsection{\label{introSection:timeDependentProperties}Time-dependent |
1144 |
> |
Properties} |
1145 |
> |
|
1146 |
> |
Time-dependent properties are usually calculated using \emph{time |
1147 |
> |
correlation function}, which correlates random variables $A$ and $B$ |
1148 |
> |
at two different time |
1149 |
> |
\begin{equation} |
1150 |
> |
C_{AB} (t) = \left\langle {A(t)B(0)} \right\rangle. |
1151 |
> |
\label{introEquation:timeCorrelationFunction} |
1152 |
> |
\end{equation} |
1153 |
> |
If $A$ and $B$ refer to same variable, this kind of correlation |
1154 |
> |
function is called \emph{auto correlation function}. One example of |
1155 |
> |
auto correlation function is velocity auto-correlation function |
1156 |
> |
which is directly related to transport properties of molecular |
1157 |
> |
liquids: |
1158 |
> |
\[ |
1159 |
> |
D = \frac{1}{3}\int\limits_0^\infty {\left\langle {v(t) \cdot v(0)} |
1160 |
> |
\right\rangle } dt |
1161 |
> |
\] |
1162 |
> |
where $D$ is diffusion constant. Unlike velocity autocorrelation |
1163 |
> |
function which is averaging over time origins and over all the |
1164 |
> |
atoms, dipole autocorrelation are calculated for the entire system. |
1165 |
> |
The dipole autocorrelation function is given by: |
1166 |
> |
\[ |
1167 |
> |
c_{dipole} = \left\langle {u_{tot} (t) \cdot u_{tot} (t)} |
1168 |
> |
\right\rangle |
1169 |
> |
\] |
1170 |
> |
Here $u_{tot}$ is the net dipole of the entire system and is given |
1171 |
> |
by |
1172 |
> |
\[ |
1173 |
> |
u_{tot} (t) = \sum\limits_i {u_i (t)} |
1174 |
> |
\] |
1175 |
> |
In principle, many time correlation functions can be related with |
1176 |
> |
Fourier transforms of the infrared, Raman, and inelastic neutron |
1177 |
> |
scattering spectra of molecular liquids. In practice, one can |
1178 |
> |
extract the IR spectrum from the intensity of dipole fluctuation at |
1179 |
> |
each frequency using the following relationship: |
1180 |
> |
\[ |
1181 |
> |
\hat c_{dipole} (v) = \int_{ - \infty }^\infty {c_{dipole} (t)e^{ - |
1182 |
> |
i2\pi vt} dt} |
1183 |
> |
\] |
1184 |
> |
|
1185 |
> |
\section{\label{introSection:rigidBody}Dynamics of Rigid Bodies} |
1186 |
> |
|
1187 |
> |
Rigid bodies are frequently involved in the modeling of different |
1188 |
> |
areas, from engineering, physics, to chemistry. For example, |
1189 |
> |
missiles and vehicle are usually modeled by rigid bodies. The |
1190 |
> |
movement of the objects in 3D gaming engine or other physics |
1191 |
> |
simulator is governed by the rigid body dynamics. In molecular |
1192 |
> |
simulation, rigid body is used to simplify the model in |
1193 |
> |
protein-protein docking study\cite{Gray2003}. |
1194 |
> |
|
1195 |
> |
It is very important to develop stable and efficient methods to |
1196 |
> |
integrate the equations of motion of orientational degrees of |
1197 |
> |
freedom. Euler angles are the nature choice to describe the |
1198 |
> |
rotational degrees of freedom. However, due to its singularity, the |
1199 |
> |
numerical integration of corresponding equations of motion is very |
1200 |
> |
inefficient and inaccurate. Although an alternative integrator using |
1201 |
> |
different sets of Euler angles can overcome this |
1202 |
> |
difficulty\cite{Barojas1973}, the computational penalty and the lost |
1203 |
> |
of angular momentum conservation still remain. A singularity free |
1204 |
> |
representation utilizing quaternions was developed by Evans in |
1205 |
> |
1977\cite{Evans1977}. Unfortunately, this approach suffer from the |
1206 |
> |
nonseparable Hamiltonian resulted from quaternion representation, |
1207 |
> |
which prevents the symplectic algorithm to be utilized. Another |
1208 |
> |
different approach is to apply holonomic constraints to the atoms |
1209 |
> |
belonging to the rigid body. Each atom moves independently under the |
1210 |
> |
normal forces deriving from potential energy and constraint forces |
1211 |
> |
which are used to guarantee the rigidness. However, due to their |
1212 |
> |
iterative nature, SHAKE and Rattle algorithm converge very slowly |
1213 |
> |
when the number of constraint increases\cite{Ryckaert1977, |
1214 |
> |
Andersen1983}. |
1215 |
> |
|
1216 |
> |
The break through in geometric literature suggests that, in order to |
1217 |
> |
develop a long-term integration scheme, one should preserve the |
1218 |
> |
symplectic structure of the flow. Introducing conjugate momentum to |
1219 |
> |
rotation matrix $Q$ and re-formulating Hamiltonian's equation, a |
1220 |
> |
symplectic integrator, RSHAKE\cite{Kol1997}, was proposed to evolve |
1221 |
> |
the Hamiltonian system in a constraint manifold by iteratively |
1222 |
> |
satisfying the orthogonality constraint $Q_T Q = 1$. An alternative |
1223 |
> |
method using quaternion representation was developed by |
1224 |
> |
Omelyan\cite{Omelyan1998}. However, both of these methods are |
1225 |
> |
iterative and inefficient. In this section, we will present a |
1226 |
> |
symplectic Lie-Poisson integrator for rigid body developed by |
1227 |
> |
Dullweber and his coworkers\cite{Dullweber1997} in depth. |
1228 |
> |
|
1229 |
> |
\subsection{\label{introSection:constrainedHamiltonianRB}Constrained Hamiltonian for Rigid Body} |
1230 |
> |
The motion of the rigid body is Hamiltonian with the Hamiltonian |
1231 |
> |
function |
1232 |
> |
\begin{equation} |
1233 |
> |
H = \frac{1}{2}(p^T m^{ - 1} p) + \frac{1}{2}tr(PJ^{ - 1} P) + |
1234 |
> |
V(q,Q) + \frac{1}{2}tr[(QQ^T - 1)\Lambda ]. |
1235 |
> |
\label{introEquation:RBHamiltonian} |
1236 |
> |
\end{equation} |
1237 |
> |
Here, $q$ and $Q$ are the position and rotation matrix for the |
1238 |
> |
rigid-body, $p$ and $P$ are conjugate momenta to $q$ and $Q$ , and |
1239 |
> |
$J$, a diagonal matrix, is defined by |
1240 |
> |
\[ |
1241 |
> |
I_{ii}^{ - 1} = \frac{1}{2}\sum\limits_{i \ne j} {J_{jj}^{ - 1} } |
1242 |
> |
\] |
1243 |
> |
where $I_{ii}$ is the diagonal element of the inertia tensor. This |
1244 |
> |
constrained Hamiltonian equation subjects to a holonomic constraint, |
1245 |
> |
\begin{equation} |
1246 |
> |
Q^T Q = 1, \label{introEquation:orthogonalConstraint} |
1247 |
> |
\end{equation} |
1248 |
> |
which is used to ensure rotation matrix's orthogonality. |
1249 |
> |
Differentiating \ref{introEquation:orthogonalConstraint} and using |
1250 |
> |
Equation \ref{introEquation:RBMotionMomentum}, one may obtain, |
1251 |
> |
\begin{equation} |
1252 |
> |
Q^T PJ^{ - 1} + J^{ - 1} P^T Q = 0 . \\ |
1253 |
> |
\label{introEquation:RBFirstOrderConstraint} |
1254 |
> |
\end{equation} |
1255 |
> |
|
1256 |
> |
Using Equation (\ref{introEquation:motionHamiltonianCoordinate}, |
1257 |
> |
\ref{introEquation:motionHamiltonianMomentum}), one can write down |
1258 |
> |
the equations of motion, |
1259 |
> |
|
1260 |
> |
\begin{eqnarray} |
1261 |
> |
\frac{{dq}}{{dt}} & = & \frac{p}{m} \label{introEquation:RBMotionPosition}\\ |
1262 |
> |
\frac{{dp}}{{dt}} & = & - \nabla _q V(q,Q) \label{introEquation:RBMotionMomentum}\\ |
1263 |
> |
\frac{{dQ}}{{dt}} & = & PJ^{ - 1} \label{introEquation:RBMotionRotation}\\ |
1264 |
> |
\frac{{dP}}{{dt}} & = & - \nabla _Q V(q,Q) - 2Q\Lambda . \label{introEquation:RBMotionP} |
1265 |
> |
\end{eqnarray} |
1266 |
> |
|
1267 |
> |
In general, there are two ways to satisfy the holonomic constraints. |
1268 |
> |
We can use constraint force provided by lagrange multiplier on the |
1269 |
> |
normal manifold to keep the motion on constraint space. Or we can |
1270 |
> |
simply evolve the system in constraint manifold. These two methods |
1271 |
> |
are proved to be equivalent. The holonomic constraint and equations |
1272 |
> |
of motions define a constraint manifold for rigid body |
1273 |
> |
\[ |
1274 |
> |
M = \left\{ {(Q,P):Q^T Q = 1,Q^T PJ^{ - 1} + J^{ - 1} P^T Q = 0} |
1275 |
> |
\right\}. |
1276 |
> |
\] |
1277 |
> |
|
1278 |
> |
Unfortunately, this constraint manifold is not the cotangent bundle |
1279 |
> |
$T_{\star}SO(3)$. However, it turns out that under symplectic |
1280 |
> |
transformation, the cotangent space and the phase space are |
1281 |
> |
diffeomorphic. Introducing |
1282 |
> |
\[ |
1283 |
> |
\tilde Q = Q,\tilde P = \frac{1}{2}\left( {P - QP^T Q} \right), |
1284 |
> |
\] |
1285 |
> |
the mechanical system subject to a holonomic constraint manifold $M$ |
1286 |
> |
can be re-formulated as a Hamiltonian system on the cotangent space |
1287 |
> |
\[ |
1288 |
> |
T^* SO(3) = \left\{ {(\tilde Q,\tilde P):\tilde Q^T \tilde Q = |
1289 |
> |
1,\tilde Q^T \tilde PJ^{ - 1} + J^{ - 1} P^T \tilde Q = 0} \right\} |
1290 |
> |
\] |
1291 |
> |
|
1292 |
> |
For a body fixed vector $X_i$ with respect to the center of mass of |
1293 |
> |
the rigid body, its corresponding lab fixed vector $X_0^{lab}$ is |
1294 |
> |
given as |
1295 |
> |
\begin{equation} |
1296 |
> |
X_i^{lab} = Q X_i + q. |
1297 |
> |
\end{equation} |
1298 |
> |
Therefore, potential energy $V(q,Q)$ is defined by |
1299 |
> |
\[ |
1300 |
> |
V(q,Q) = V(Q X_0 + q). |
1301 |
> |
\] |
1302 |
> |
Hence, the force and torque are given by |
1303 |
> |
\[ |
1304 |
> |
\nabla _q V(q,Q) = F(q,Q) = \sum\limits_i {F_i (q,Q)}, |
1305 |
> |
\] |
1306 |
> |
and |
1307 |
> |
\[ |
1308 |
> |
\nabla _Q V(q,Q) = F(q,Q)X_i^t |
1309 |
> |
\] |
1310 |
> |
respectively. |
1311 |
> |
|
1312 |
> |
As a common choice to describe the rotation dynamics of the rigid |
1313 |
> |
body, angular momentum on body frame $\Pi = Q^t P$ is introduced to |
1314 |
> |
rewrite the equations of motion, |
1315 |
> |
\begin{equation} |
1316 |
> |
\begin{array}{l} |
1317 |
> |
\mathop \Pi \limits^ \bullet = J^{ - 1} \Pi ^T \Pi + Q^T \sum\limits_i {F_i (q,Q)X_i^T } - \Lambda \\ |
1318 |
> |
\mathop Q\limits^{{\rm{ }} \bullet } = Q\Pi {\rm{ }}J^{ - 1} \\ |
1319 |
> |
\end{array} |
1320 |
> |
\label{introEqaution:RBMotionPI} |
1321 |
> |
\end{equation} |
1322 |
> |
, as well as holonomic constraints, |
1323 |
> |
\[ |
1324 |
> |
\begin{array}{l} |
1325 |
> |
\Pi J^{ - 1} + J^{ - 1} \Pi ^t = 0 \\ |
1326 |
> |
Q^T Q = 1 \\ |
1327 |
> |
\end{array} |
1328 |
> |
\] |
1329 |
> |
|
1330 |
> |
For a vector $v(v_1 ,v_2 ,v_3 ) \in R^3$ and a matrix $\hat v \in |
1331 |
> |
so(3)^ \star$, the hat-map isomorphism, |
1332 |
> |
\begin{equation} |
1333 |
> |
v(v_1 ,v_2 ,v_3 ) \Leftrightarrow \hat v = \left( |
1334 |
> |
{\begin{array}{*{20}c} |
1335 |
> |
0 & { - v_3 } & {v_2 } \\ |
1336 |
> |
{v_3 } & 0 & { - v_1 } \\ |
1337 |
> |
{ - v_2 } & {v_1 } & 0 \\ |
1338 |
> |
\end{array}} \right), |
1339 |
> |
\label{introEquation:hatmapIsomorphism} |
1340 |
> |
\end{equation} |
1341 |
> |
will let us associate the matrix products with traditional vector |
1342 |
> |
operations |
1343 |
> |
\[ |
1344 |
> |
\hat vu = v \times u |
1345 |
> |
\] |
1346 |
> |
|
1347 |
> |
Using \ref{introEqaution:RBMotionPI}, one can construct a skew |
1348 |
> |
matrix, |
1349 |
> |
\begin{equation} |
1350 |
> |
(\mathop \Pi \limits^ \bullet - \mathop \Pi \limits^ \bullet ^T |
1351 |
> |
){\rm{ }} = {\rm{ }}(\Pi - \Pi ^T ){\rm{ }}(J^{ - 1} \Pi + \Pi J^{ |
1352 |
> |
- 1} ) + \sum\limits_i {[Q^T F_i (r,Q)X_i^T - X_i F_i (r,Q)^T Q]} - |
1353 |
> |
(\Lambda - \Lambda ^T ) . \label{introEquation:skewMatrixPI} |
1354 |
> |
\end{equation} |
1355 |
> |
Since $\Lambda$ is symmetric, the last term of Equation |
1356 |
> |
\ref{introEquation:skewMatrixPI} is zero, which implies the Lagrange |
1357 |
> |
multiplier $\Lambda$ is absent from the equations of motion. This |
1358 |
> |
unique property eliminate the requirement of iterations which can |
1359 |
> |
not be avoided in other methods\cite{Kol1997, Omelyan1998}. |
1360 |
> |
|
1361 |
> |
Applying hat-map isomorphism, we obtain the equation of motion for |
1362 |
> |
angular momentum on body frame |
1363 |
> |
\begin{equation} |
1364 |
> |
\dot \pi = \pi \times I^{ - 1} \pi + \sum\limits_i {\left( {Q^T |
1365 |
> |
F_i (r,Q)} \right) \times X_i }. |
1366 |
> |
\label{introEquation:bodyAngularMotion} |
1367 |
> |
\end{equation} |
1368 |
> |
In the same manner, the equation of motion for rotation matrix is |
1369 |
> |
given by |
1370 |
> |
\[ |
1371 |
> |
\dot Q = Qskew(I^{ - 1} \pi ) |
1372 |
> |
\] |
1373 |
> |
|
1374 |
> |
\subsection{\label{introSection:SymplecticFreeRB}Symplectic |
1375 |
> |
Lie-Poisson Integrator for Free Rigid Body} |
1376 |
> |
|
1377 |
> |
If there is not external forces exerted on the rigid body, the only |
1378 |
> |
contribution to the rotational is from the kinetic potential (the |
1379 |
> |
first term of \ref{ introEquation:bodyAngularMotion}). The free |
1380 |
> |
rigid body is an example of Lie-Poisson system with Hamiltonian |
1381 |
> |
function |
1382 |
> |
\begin{equation} |
1383 |
> |
T^r (\pi ) = T_1 ^r (\pi _1 ) + T_2^r (\pi _2 ) + T_3^r (\pi _3 ) |
1384 |
> |
\label{introEquation:rotationalKineticRB} |
1385 |
> |
\end{equation} |
1386 |
> |
where $T_i^r (\pi _i ) = \frac{{\pi _i ^2 }}{{2I_i }}$ and |
1387 |
> |
Lie-Poisson structure matrix, |
1388 |
> |
\begin{equation} |
1389 |
> |
J(\pi ) = \left( {\begin{array}{*{20}c} |
1390 |
> |
0 & {\pi _3 } & { - \pi _2 } \\ |
1391 |
> |
{ - \pi _3 } & 0 & {\pi _1 } \\ |
1392 |
> |
{\pi _2 } & { - \pi _1 } & 0 \\ |
1393 |
> |
\end{array}} \right) |
1394 |
> |
\end{equation} |
1395 |
> |
Thus, the dynamics of free rigid body is governed by |
1396 |
> |
\begin{equation} |
1397 |
> |
\frac{d}{{dt}}\pi = J(\pi )\nabla _\pi T^r (\pi ) |
1398 |
> |
\end{equation} |
1399 |
> |
|
1400 |
> |
One may notice that each $T_i^r$ in Equation |
1401 |
> |
\ref{introEquation:rotationalKineticRB} can be solved exactly. For |
1402 |
> |
instance, the equations of motion due to $T_1^r$ are given by |
1403 |
> |
\begin{equation} |
1404 |
> |
\frac{d}{{dt}}\pi = R_1 \pi ,\frac{d}{{dt}}Q = QR_1 |
1405 |
> |
\label{introEqaution:RBMotionSingleTerm} |
1406 |
> |
\end{equation} |
1407 |
> |
where |
1408 |
> |
\[ R_1 = \left( {\begin{array}{*{20}c} |
1409 |
> |
0 & 0 & 0 \\ |
1410 |
> |
0 & 0 & {\pi _1 } \\ |
1411 |
> |
0 & { - \pi _1 } & 0 \\ |
1412 |
> |
\end{array}} \right). |
1413 |
> |
\] |
1414 |
> |
The solutions of Equation \ref{introEqaution:RBMotionSingleTerm} is |
1415 |
> |
\[ |
1416 |
> |
\pi (\Delta t) = e^{\Delta tR_1 } \pi (0),Q(\Delta t) = |
1417 |
> |
Q(0)e^{\Delta tR_1 } |
1418 |
> |
\] |
1419 |
> |
with |
1420 |
> |
\[ |
1421 |
> |
e^{\Delta tR_1 } = \left( {\begin{array}{*{20}c} |
1422 |
> |
0 & 0 & 0 \\ |
1423 |
> |
0 & {\cos \theta _1 } & {\sin \theta _1 } \\ |
1424 |
> |
0 & { - \sin \theta _1 } & {\cos \theta _1 } \\ |
1425 |
> |
\end{array}} \right),\theta _1 = \frac{{\pi _1 }}{{I_1 }}\Delta t. |
1426 |
> |
\] |
1427 |
> |
To reduce the cost of computing expensive functions in $e^{\Delta |
1428 |
> |
tR_1 }$, we can use Cayley transformation, |
1429 |
> |
\[ |
1430 |
> |
e^{\Delta tR_1 } \approx (1 - \Delta tR_1 )^{ - 1} (1 + \Delta tR_1 |
1431 |
> |
) |
1432 |
> |
\] |
1433 |
> |
The flow maps for $T_2^r$ and $T_3^r$ can be found in the same |
1434 |
> |
manner. |
1435 |
> |
|
1436 |
> |
In order to construct a second-order symplectic method, we split the |
1437 |
> |
angular kinetic Hamiltonian function can into five terms |
1438 |
> |
\[ |
1439 |
> |
T^r (\pi ) = \frac{1}{2}T_1 ^r (\pi _1 ) + \frac{1}{2}T_2^r (\pi _2 |
1440 |
> |
) + T_3^r (\pi _3 ) + \frac{1}{2}T_2^r (\pi _2 ) + \frac{1}{2}T_1 ^r |
1441 |
> |
(\pi _1 ) |
1442 |
> |
\]. |
1443 |
> |
Concatenating flows corresponding to these five terms, we can obtain |
1444 |
> |
an symplectic integrator, |
1445 |
> |
\[ |
1446 |
> |
\varphi _{\Delta t,T^r } = \varphi _{\Delta t/2,\pi _1 } \circ |
1447 |
> |
\varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t,\pi _3 } |
1448 |
> |
\circ \varphi _{\Delta t/2,\pi _2 } \circ \varphi _{\Delta t/2,\pi |
1449 |
> |
_1 }. |
1450 |
> |
\] |
1451 |
> |
|
1452 |
> |
The non-canonical Lie-Poisson bracket ${F, G}$ of two function |
1453 |
> |
$F(\pi )$ and $G(\pi )$ is defined by |
1454 |
> |
\[ |
1455 |
> |
\{ F,G\} (\pi ) = [\nabla _\pi F(\pi )]^T J(\pi )\nabla _\pi G(\pi |
1456 |
> |
) |
1457 |
> |
\] |
1458 |
> |
If the Poisson bracket of a function $F$ with an arbitrary smooth |
1459 |
> |
function $G$ is zero, $F$ is a \emph{Casimir}, which is the |
1460 |
> |
conserved quantity in Poisson system. We can easily verify that the |
1461 |
> |
norm of the angular momentum, $\parallel \pi |
1462 |
> |
\parallel$, is a \emph{Casimir}. Let$ F(\pi ) = S(\frac{{\parallel |
1463 |
> |
\pi \parallel ^2 }}{2})$ for an arbitrary function $ S:R \to R$ , |
1464 |
> |
then by the chain rule |
1465 |
> |
\[ |
1466 |
> |
\nabla _\pi F(\pi ) = S'(\frac{{\parallel \pi \parallel ^2 |
1467 |
> |
}}{2})\pi |
1468 |
> |
\] |
1469 |
> |
Thus $ [\nabla _\pi F(\pi )]^T J(\pi ) = - S'(\frac{{\parallel \pi |
1470 |
> |
\parallel ^2 }}{2})\pi \times \pi = 0 $. This explicit |
1471 |
> |
Lie-Poisson integrator is found to be extremely efficient and stable |
1472 |
> |
which can be explained by the fact the small angle approximation is |
1473 |
> |
used and the norm of the angular momentum is conserved. |
1474 |
> |
|
1475 |
> |
\subsection{\label{introSection:RBHamiltonianSplitting} Hamiltonian |
1476 |
> |
Splitting for Rigid Body} |
1477 |
> |
|
1478 |
> |
The Hamiltonian of rigid body can be separated in terms of kinetic |
1479 |
> |
energy and potential energy, |
1480 |
> |
\[ |
1481 |
> |
H = T(p,\pi ) + V(q,Q) |
1482 |
> |
\] |
1483 |
> |
The equations of motion corresponding to potential energy and |
1484 |
> |
kinetic energy are listed in the below table, |
1485 |
> |
\begin{table} |
1486 |
> |
\caption{Equations of motion due to Potential and Kinetic Energies} |
1487 |
> |
\begin{center} |
1488 |
> |
\begin{tabular}{|l|l|} |
1489 |
> |
\hline |
1490 |
> |
% after \\: \hline or \cline{col1-col2} \cline{col3-col4} ... |
1491 |
> |
Potential & Kinetic \\ |
1492 |
> |
$\frac{{dq}}{{dt}} = \frac{p}{m}$ & $\frac{d}{{dt}}q = p$ \\ |
1493 |
> |
$\frac{d}{{dt}}p = - \frac{{\partial V}}{{\partial q}}$ & $ \frac{d}{{dt}}p = 0$ \\ |
1494 |
> |
$\frac{d}{{dt}}Q = 0$ & $ \frac{d}{{dt}}Q = Qskew(I^{ - 1} j)$ \\ |
1495 |
> |
$ \frac{d}{{dt}}\pi = \sum\limits_i {\left( {Q^T F_i (r,Q)} \right) \times X_i }$ & $\frac{d}{{dt}}\pi = \pi \times I^{ - 1} \pi$\\ |
1496 |
> |
\hline |
1497 |
> |
\end{tabular} |
1498 |
> |
\end{center} |
1499 |
> |
\end{table} |
1500 |
> |
A second-order symplectic method is now obtained by the |
1501 |
> |
composition of the flow maps, |
1502 |
> |
\[ |
1503 |
> |
\varphi _{\Delta t} = \varphi _{\Delta t/2,V} \circ \varphi |
1504 |
> |
_{\Delta t,T} \circ \varphi _{\Delta t/2,V}. |
1505 |
> |
\] |
1506 |
> |
Moreover, $\varphi _{\Delta t/2,V}$ can be divided into two |
1507 |
> |
sub-flows which corresponding to force and torque respectively, |
1508 |
> |
\[ |
1509 |
> |
\varphi _{\Delta t/2,V} = \varphi _{\Delta t/2,F} \circ \varphi |
1510 |
> |
_{\Delta t/2,\tau }. |
1511 |
> |
\] |
1512 |
> |
Since the associated operators of $\varphi _{\Delta t/2,F} $ and |
1513 |
> |
$\circ \varphi _{\Delta t/2,\tau }$ are commuted, the composition |
1514 |
> |
order inside $\varphi _{\Delta t/2,V}$ does not matter. |
1515 |
> |
|
1516 |
> |
Furthermore, kinetic potential can be separated to translational |
1517 |
> |
kinetic term, $T^t (p)$, and rotational kinetic term, $T^r (\pi )$, |
1518 |
> |
\begin{equation} |
1519 |
> |
T(p,\pi ) =T^t (p) + T^r (\pi ). |
1520 |
> |
\end{equation} |
1521 |
> |
where $ T^t (p) = \frac{1}{2}p^T m^{ - 1} p $ and $T^r (\pi )$ is |
1522 |
> |
defined by \ref{introEquation:rotationalKineticRB}. Therefore, the |
1523 |
> |
corresponding flow maps are given by |
1524 |
> |
\[ |
1525 |
> |
\varphi _{\Delta t,T} = \varphi _{\Delta t,T^t } \circ \varphi |
1526 |
> |
_{\Delta t,T^r }. |
1527 |
> |
\] |
1528 |
> |
Finally, we obtain the overall symplectic flow maps for free moving |
1529 |
> |
rigid body |
1530 |
> |
\begin{equation} |
1531 |
> |
\begin{array}{c} |
1532 |
> |
\varphi _{\Delta t} = \varphi _{\Delta t/2,F} \circ \varphi _{\Delta t/2,\tau } \\ |
1533 |
> |
\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 } \\ |
1534 |
> |
\circ \varphi _{\Delta t/2,\tau } \circ \varphi _{\Delta t/2,F} .\\ |
1535 |
> |
\end{array} |
1536 |
> |
\label{introEquation:overallRBFlowMaps} |
1537 |
> |
\end{equation} |
1538 |
> |
|
1539 |
> |
\section{\label{introSection:langevinDynamics}Langevin Dynamics} |
1540 |
> |
As an alternative to newtonian dynamics, Langevin dynamics, which |
1541 |
> |
mimics a simple heat bath with stochastic and dissipative forces, |
1542 |
> |
has been applied in a variety of studies. This section will review |
1543 |
> |
the theory of Langevin dynamics simulation. A brief derivation of |
1544 |
> |
generalized Langevin equation will be given first. Follow that, we |
1545 |
> |
will discuss the physical meaning of the terms appearing in the |
1546 |
> |
equation as well as the calculation of friction tensor from |
1547 |
> |
hydrodynamics theory. |
1548 |
> |
|
1549 |
> |
\subsection{\label{introSection:generalizedLangevinDynamics}Derivation of Generalized Langevin Equation} |
1550 |
> |
|
1551 |
> |
Harmonic bath model, in which an effective set of harmonic |
1552 |
> |
oscillators are used to mimic the effect of a linearly responding |
1553 |
> |
environment, has been widely used in quantum chemistry and |
1554 |
> |
statistical mechanics. One of the successful applications of |
1555 |
> |
Harmonic bath model is the derivation of Deriving Generalized |
1556 |
> |
Langevin Dynamics. Lets consider a system, in which the degree of |
1557 |
> |
freedom $x$ is assumed to couple to the bath linearly, giving a |
1558 |
> |
Hamiltonian of the form |
1559 |
> |
\begin{equation} |
1560 |
> |
H = \frac{{p^2 }}{{2m}} + U(x) + H_B + \Delta U(x,x_1 , \ldots x_N) |
1561 |
> |
\label{introEquation:bathGLE}. |
1562 |
> |
\end{equation} |
1563 |
> |
Here $p$ is a momentum conjugate to $q$, $m$ is the mass associated |
1564 |
> |
with this degree of freedom, $H_B$ is harmonic bath Hamiltonian, |
1565 |
> |
\[ |
1566 |
> |
H_B = \sum\limits_{\alpha = 1}^N {\left\{ {\frac{{p_\alpha ^2 |
1567 |
> |
}}{{2m_\alpha }} + \frac{1}{2}m_\alpha \omega _\alpha ^2 } |
1568 |
> |
\right\}} |
1569 |
> |
\] |
1570 |
> |
where the index $\alpha$ runs over all the bath degrees of freedom, |
1571 |
> |
$\omega _\alpha$ are the harmonic bath frequencies, $m_\alpha$ are |
1572 |
> |
the harmonic bath masses, and $\Delta U$ is bilinear system-bath |
1573 |
> |
coupling, |
1574 |
> |
\[ |
1575 |
> |
\Delta U = - \sum\limits_{\alpha = 1}^N {g_\alpha x_\alpha x} |
1576 |
> |
\] |
1577 |
> |
where $g_\alpha$ are the coupling constants between the bath and the |
1578 |
> |
coordinate $x$. Introducing |
1579 |
> |
\[ |
1580 |
> |
W(x) = U(x) - \sum\limits_{\alpha = 1}^N {\frac{{g_\alpha ^2 |
1581 |
> |
}}{{2m_\alpha w_\alpha ^2 }}} x^2 |
1582 |
> |
\] and combining the last two terms in Equation |
1583 |
> |
\ref{introEquation:bathGLE}, we may rewrite the Harmonic bath |
1584 |
> |
Hamiltonian as |
1585 |
> |
\[ |
1586 |
> |
H = \frac{{p^2 }}{{2m}} + W(x) + \sum\limits_{\alpha = 1}^N |
1587 |
> |
{\left\{ {\frac{{p_\alpha ^2 }}{{2m_\alpha }} + \frac{1}{2}m_\alpha |
1588 |
> |
w_\alpha ^2 \left( {x_\alpha - \frac{{g_\alpha }}{{m_\alpha |
1589 |
> |
w_\alpha ^2 }}x} \right)^2 } \right\}} |
1590 |
> |
\] |
1591 |
> |
Since the first two terms of the new Hamiltonian depend only on the |
1592 |
> |
system coordinates, we can get the equations of motion for |
1593 |
> |
Generalized Langevin Dynamics by Hamilton's equations |
1594 |
> |
\ref{introEquation:motionHamiltonianCoordinate, |
1595 |
> |
introEquation:motionHamiltonianMomentum}, |
1596 |
> |
\begin{equation} |
1597 |
> |
m\ddot x = - \frac{{\partial W(x)}}{{\partial x}} - |
1598 |
> |
\sum\limits_{\alpha = 1}^N {g_\alpha \left( {x_\alpha - |
1599 |
> |
\frac{{g_\alpha }}{{m_\alpha w_\alpha ^2 }}x} \right)}, |
1600 |
> |
\label{introEquation:coorMotionGLE} |
1601 |
> |
\end{equation} |
1602 |
> |
and |
1603 |
> |
\begin{equation} |
1604 |
> |
m\ddot x_\alpha = - m_\alpha w_\alpha ^2 \left( {x_\alpha - |
1605 |
> |
\frac{{g_\alpha }}{{m_\alpha w_\alpha ^2 }}x} \right). |
1606 |
> |
\label{introEquation:bathMotionGLE} |
1607 |
> |
\end{equation} |
1608 |
> |
|
1609 |
> |
In order to derive an equation for $x$, the dynamics of the bath |
1610 |
> |
variables $x_\alpha$ must be solved exactly first. As an integral |
1611 |
> |
transform which is particularly useful in solving linear ordinary |
1612 |
> |
differential equations, Laplace transform is the appropriate tool to |
1613 |
> |
solve this problem. The basic idea is to transform the difficult |
1614 |
> |
differential equations into simple algebra problems which can be |
1615 |
> |
solved easily. Then applying inverse Laplace transform, also known |
1616 |
> |
as the Bromwich integral, we can retrieve the solutions of the |
1617 |
> |
original problems. |
1618 |
> |
|
1619 |
> |
Let $f(t)$ be a function defined on $ [0,\infty ) $. The Laplace |
1620 |
> |
transform of f(t) is a new function defined as |
1621 |
> |
\[ |
1622 |
> |
L(f(t)) \equiv F(p) = \int_0^\infty {f(t)e^{ - pt} dt} |
1623 |
> |
\] |
1624 |
> |
where $p$ is real and $L$ is called the Laplace Transform |
1625 |
> |
Operator. Below are some important properties of Laplace transform |
1626 |
> |
|
1627 |
> |
\begin{eqnarray*} |
1628 |
> |
L(x + y) & = & L(x) + L(y) \\ |
1629 |
> |
L(ax) & = & aL(x) \\ |
1630 |
> |
L(\dot x) & = & pL(x) - px(0) \\ |
1631 |
> |
L(\ddot x)& = & p^2 L(x) - px(0) - \dot x(0) \\ |
1632 |
> |
L\left( {\int_0^t {g(t - \tau )h(\tau )d\tau } } \right)& = & G(p)H(p) \\ |
1633 |
> |
\end{eqnarray*} |
1634 |
> |
|
1635 |
> |
|
1636 |
> |
Applying Laplace transform to the bath coordinates, we obtain |
1637 |
> |
\begin{eqnarray*} |
1638 |
> |
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) \\ |
1639 |
> |
L(x_\alpha ) & = & \frac{{\frac{{g_\alpha }}{{\omega _\alpha }}L(x) + px_\alpha (0) + \dot x_\alpha (0)}}{{p^2 + \omega _\alpha ^2 }} \\ |
1640 |
> |
\end{eqnarray*} |
1641 |
> |
|
1642 |
> |
By the same way, the system coordinates become |
1643 |
> |
\begin{eqnarray*} |
1644 |
> |
mL(\ddot x) & = & - \frac{1}{p}\frac{{\partial W(x)}}{{\partial x}} \\ |
1645 |
> |
& & \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\}} \\ |
1646 |
> |
\end{eqnarray*} |
1647 |
> |
|
1648 |
> |
With the help of some relatively important inverse Laplace |
1649 |
> |
transformations: |
1650 |
> |
\[ |
1651 |
> |
\begin{array}{c} |
1652 |
> |
L(\cos at) = \frac{p}{{p^2 + a^2 }} \\ |
1653 |
> |
L(\sin at) = \frac{a}{{p^2 + a^2 }} \\ |
1654 |
> |
L(1) = \frac{1}{p} \\ |
1655 |
> |
\end{array} |
1656 |
> |
\] |
1657 |
> |
, we obtain |
1658 |
> |
\begin{eqnarray*} |
1659 |
> |
m\ddot x & = & - \frac{{\partial W(x)}}{{\partial x}} - |
1660 |
> |
\sum\limits_{\alpha = 1}^N {\left\{ {\left( { - \frac{{g_\alpha ^2 |
1661 |
> |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\int_0^t {\cos (\omega |
1662 |
> |
_\alpha t)\dot x(t - \tau )d\tau } } \right\}} \\ |
1663 |
> |
& & + \sum\limits_{\alpha = 1}^N {\left\{ {\left[ {g_\alpha |
1664 |
> |
x_\alpha (0) - \frac{{g_\alpha }}{{m_\alpha \omega _\alpha }}} |
1665 |
> |
\right]\cos (\omega _\alpha t) + \frac{{g_\alpha \dot x_\alpha |
1666 |
> |
(0)}}{{\omega _\alpha }}\sin (\omega _\alpha t)} \right\}} |
1667 |
> |
\end{eqnarray*} |
1668 |
> |
\begin{eqnarray*} |
1669 |
> |
m\ddot x & = & - \frac{{\partial W(x)}}{{\partial x}} - \int_0^t |
1670 |
> |
{\sum\limits_{\alpha = 1}^N {\left( { - \frac{{g_\alpha ^2 |
1671 |
> |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\cos (\omega _\alpha |
1672 |
> |
t)\dot x(t - \tau )d} \tau } \\ |
1673 |
> |
& & + \sum\limits_{\alpha = 1}^N {\left\{ {\left[ {g_\alpha |
1674 |
> |
x_\alpha (0) - \frac{{g_\alpha }}{{m_\alpha \omega _\alpha }}} |
1675 |
> |
\right]\cos (\omega _\alpha t) + \frac{{g_\alpha \dot x_\alpha |
1676 |
> |
(0)}}{{\omega _\alpha }}\sin (\omega _\alpha t)} \right\}} |
1677 |
> |
\end{eqnarray*} |
1678 |
> |
Introducing a \emph{dynamic friction kernel} |
1679 |
> |
\begin{equation} |
1680 |
> |
\xi (t) = \sum\limits_{\alpha = 1}^N {\left( { - \frac{{g_\alpha ^2 |
1681 |
> |
}}{{m_\alpha \omega _\alpha ^2 }}} \right)\cos (\omega _\alpha t)} |
1682 |
> |
\label{introEquation:dynamicFrictionKernelDefinition} |
1683 |
> |
\end{equation} |
1684 |
> |
and \emph{a random force} |
1685 |
> |
\begin{equation} |
1686 |
> |
R(t) = \sum\limits_{\alpha = 1}^N {\left( {g_\alpha x_\alpha (0) |
1687 |
> |
- \frac{{g_\alpha ^2 }}{{m_\alpha \omega _\alpha ^2 }}x(0)} |
1688 |
> |
\right)\cos (\omega _\alpha t)} + \frac{{\dot x_\alpha |
1689 |
> |
(0)}}{{\omega _\alpha }}\sin (\omega _\alpha t), |
1690 |
> |
\label{introEquation:randomForceDefinition} |
1691 |
> |
\end{equation} |
1692 |
> |
the equation of motion can be rewritten as |
1693 |
> |
\begin{equation} |
1694 |
> |
m\ddot x = - \frac{{\partial W}}{{\partial x}} - \int_0^t {\xi |
1695 |
> |
(t)\dot x(t - \tau )d\tau } + R(t) |
1696 |
> |
\label{introEuqation:GeneralizedLangevinDynamics} |
1697 |
> |
\end{equation} |
1698 |
> |
which is known as the \emph{generalized Langevin equation}. |
1699 |
> |
|
1700 |
> |
\subsubsection{\label{introSection:randomForceDynamicFrictionKernel}Random Force and Dynamic Friction Kernel} |
1701 |
> |
|
1702 |
> |
One may notice that $R(t)$ depends only on initial conditions, which |
1703 |
> |
implies it is completely deterministic within the context of a |
1704 |
> |
harmonic bath. However, it is easy to verify that $R(t)$ is totally |
1705 |
> |
uncorrelated to $x$ and $\dot x$, |
1706 |
> |
\[ |
1707 |
> |
\begin{array}{l} |
1708 |
> |
\left\langle {x(t)R(t)} \right\rangle = 0, \\ |
1709 |
> |
\left\langle {\dot x(t)R(t)} \right\rangle = 0. \\ |
1710 |
> |
\end{array} |
1711 |
> |
\] |
1712 |
> |
This property is what we expect from a truly random process. As long |
1713 |
> |
as the model, which is gaussian distribution in general, chosen for |
1714 |
> |
$R(t)$ is a truly random process, the stochastic nature of the GLE |
1715 |
> |
still remains. |
1716 |
> |
|
1717 |
> |
%dynamic friction kernel |
1718 |
> |
The convolution integral |
1719 |
> |
\[ |
1720 |
> |
\int_0^t {\xi (t)\dot x(t - \tau )d\tau } |
1721 |
> |
\] |
1722 |
> |
depends on the entire history of the evolution of $x$, which implies |
1723 |
> |
that the bath retains memory of previous motions. In other words, |
1724 |
> |
the bath requires a finite time to respond to change in the motion |
1725 |
> |
of the system. For a sluggish bath which responds slowly to changes |
1726 |
> |
in the system coordinate, we may regard $\xi(t)$ as a constant |
1727 |
> |
$\xi(t) = \Xi_0$. Hence, the convolution integral becomes |
1728 |
> |
\[ |
1729 |
> |
\int_0^t {\xi (t)\dot x(t - \tau )d\tau } = \xi _0 (x(t) - x(0)) |
1730 |
> |
\] |
1731 |
> |
and Equation \ref{introEuqation:GeneralizedLangevinDynamics} becomes |
1732 |
> |
\[ |
1733 |
> |
m\ddot x = - \frac{\partial }{{\partial x}}\left( {W(x) + |
1734 |
> |
\frac{1}{2}\xi _0 (x - x_0 )^2 } \right) + R(t), |
1735 |
> |
\] |
1736 |
> |
which can be used to describe dynamic caging effect. The other |
1737 |
> |
extreme is the bath that responds infinitely quickly to motions in |
1738 |
> |
the system. Thus, $\xi (t)$ can be taken as a $delta$ function in |
1739 |
> |
time: |
1740 |
> |
\[ |
1741 |
> |
\xi (t) = 2\xi _0 \delta (t) |
1742 |
> |
\] |
1743 |
> |
Hence, the convolution integral becomes |
1744 |
> |
\[ |
1745 |
> |
\int_0^t {\xi (t)\dot x(t - \tau )d\tau } = 2\xi _0 \int_0^t |
1746 |
> |
{\delta (t)\dot x(t - \tau )d\tau } = \xi _0 \dot x(t), |
1747 |
> |
\] |
1748 |
> |
and Equation \ref{introEuqation:GeneralizedLangevinDynamics} becomes |
1749 |
> |
\begin{equation} |
1750 |
> |
m\ddot x = - \frac{{\partial W(x)}}{{\partial x}} - \xi _0 \dot |
1751 |
> |
x(t) + R(t) \label{introEquation:LangevinEquation} |
1752 |
> |
\end{equation} |
1753 |
> |
which is known as the Langevin equation. The static friction |
1754 |
> |
coefficient $\xi _0$ can either be calculated from spectral density |
1755 |
> |
or be determined by Stokes' law for regular shaped particles.A |
1756 |
> |
briefly review on calculating friction tensor for arbitrary shaped |
1757 |
> |
particles is given in Sec.~\ref{introSection:frictionTensor}. |
1758 |
> |
|
1759 |
> |
\subsubsection{\label{introSection:secondFluctuationDissipation}The Second Fluctuation Dissipation Theorem} |
1760 |
> |
|
1761 |
> |
Defining a new set of coordinates, |
1762 |
> |
\[ |
1763 |
> |
q_\alpha (t) = x_\alpha (t) - \frac{1}{{m_\alpha \omega _\alpha |
1764 |
> |
^2 }}x(0) |
1765 |
> |
\], |
1766 |
> |
we can rewrite $R(T)$ as |
1767 |
> |
\[ |
1768 |
> |
R(t) = \sum\limits_{\alpha = 1}^N {g_\alpha q_\alpha (t)}. |
1769 |
> |
\] |
1770 |
> |
And since the $q$ coordinates are harmonic oscillators, |
1771 |
> |
|
1772 |
> |
\begin{eqnarray*} |
1773 |
> |
\left\langle {q_\alpha ^2 } \right\rangle & = & \frac{{kT}}{{m_\alpha \omega _\alpha ^2 }} \\ |
1774 |
> |
\left\langle {q_\alpha (t)q_\alpha (0)} \right\rangle & = & \left\langle {q_\alpha ^2 (0)} \right\rangle \cos (\omega _\alpha t) \\ |
1775 |
> |
\left\langle {q_\alpha (t)q_\beta (0)} \right\rangle & = &\delta _{\alpha \beta } \left\langle {q_\alpha (t)q_\alpha (0)} \right\rangle \\ |
1776 |
> |
\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 } } \\ |
1777 |
> |
& = &\sum\limits_\alpha {g_\alpha ^2 \left\langle {q_\alpha ^2 (0)} \right\rangle \cos (\omega _\alpha t)} \\ |
1778 |
> |
& = &kT\xi (t) \\ |
1779 |
> |
\end{eqnarray*} |
1780 |
> |
|
1781 |
> |
Thus, we recover the \emph{second fluctuation dissipation theorem} |
1782 |
> |
\begin{equation} |
1783 |
> |
\xi (t) = \left\langle {R(t)R(0)} \right\rangle |
1784 |
> |
\label{introEquation:secondFluctuationDissipation}. |
1785 |
> |
\end{equation} |
1786 |
> |
In effect, it acts as a constraint on the possible ways in which one |
1787 |
> |
can model the random force and friction kernel. |
1788 |
> |
|
1789 |
> |
\subsection{\label{introSection:frictionTensor} Friction Tensor} |
1790 |
> |
Theoretically, the friction kernel can be determined using velocity |
1791 |
> |
autocorrelation function. However, this approach become impractical |
1792 |
> |
when the system become more and more complicate. Instead, various |
1793 |
> |
approaches based on hydrodynamics have been developed to calculate |
1794 |
> |
the friction coefficients. The friction effect is isotropic in |
1795 |
> |
Equation, $\zeta$ can be taken as a scalar. In general, friction |
1796 |
> |
tensor $\Xi$ is a $6\times 6$ matrix given by |
1797 |
> |
\[ |
1798 |
> |
\Xi = \left( {\begin{array}{*{20}c} |
1799 |
> |
{\Xi _{}^{tt} } & {\Xi _{}^{rt} } \\ |
1800 |
> |
{\Xi _{}^{tr} } & {\Xi _{}^{rr} } \\ |
1801 |
> |
\end{array}} \right). |
1802 |
> |
\] |
1803 |
> |
Here, $ {\Xi^{tt} }$ and $ {\Xi^{rr} }$ are translational friction |
1804 |
> |
tensor and rotational resistance (friction) tensor respectively, |
1805 |
> |
while ${\Xi^{tr} }$ is translation-rotation coupling tensor and $ |
1806 |
> |
{\Xi^{rt} }$ is rotation-translation coupling tensor. When a |
1807 |
> |
particle moves in a fluid, it may experience friction force or |
1808 |
> |
torque along the opposite direction of the velocity or angular |
1809 |
> |
velocity, |
1810 |
> |
\[ |
1811 |
> |
\left( \begin{array}{l} |
1812 |
> |
F_R \\ |
1813 |
> |
\tau _R \\ |
1814 |
> |
\end{array} \right) = - \left( {\begin{array}{*{20}c} |
1815 |
> |
{\Xi ^{tt} } & {\Xi ^{rt} } \\ |
1816 |
> |
{\Xi ^{tr} } & {\Xi ^{rr} } \\ |
1817 |
> |
\end{array}} \right)\left( \begin{array}{l} |
1818 |
> |
v \\ |
1819 |
> |
w \\ |
1820 |
> |
\end{array} \right) |
1821 |
> |
\] |
1822 |
> |
where $F_r$ is the friction force and $\tau _R$ is the friction |
1823 |
> |
toque. |
1824 |
> |
|
1825 |
> |
\subsubsection{\label{introSection:resistanceTensorRegular}The Resistance Tensor for Regular Shape} |
1826 |
> |
|
1827 |
> |
For a spherical particle, the translational and rotational friction |
1828 |
> |
constant can be calculated from Stoke's law, |
1829 |
> |
\[ |
1830 |
> |
\Xi ^{tt} = \left( {\begin{array}{*{20}c} |
1831 |
> |
{6\pi \eta R} & 0 & 0 \\ |
1832 |
> |
0 & {6\pi \eta R} & 0 \\ |
1833 |
> |
0 & 0 & {6\pi \eta R} \\ |
1834 |
> |
\end{array}} \right) |
1835 |
> |
\] |
1836 |
> |
and |
1837 |
> |
\[ |
1838 |
> |
\Xi ^{rr} = \left( {\begin{array}{*{20}c} |
1839 |
> |
{8\pi \eta R^3 } & 0 & 0 \\ |
1840 |
> |
0 & {8\pi \eta R^3 } & 0 \\ |
1841 |
> |
0 & 0 & {8\pi \eta R^3 } \\ |
1842 |
> |
\end{array}} \right) |
1843 |
> |
\] |
1844 |
> |
where $\eta$ is the viscosity of the solvent and $R$ is the |
1845 |
> |
hydrodynamics radius. |
1846 |
> |
|
1847 |
> |
Other non-spherical shape, such as cylinder and ellipsoid |
1848 |
> |
\textit{etc}, are widely used as reference for developing new |
1849 |
> |
hydrodynamics theory, because their properties can be calculated |
1850 |
> |
exactly. In 1936, Perrin extended Stokes's law to general ellipsoid, |
1851 |
> |
also called a triaxial ellipsoid, which is given in Cartesian |
1852 |
> |
coordinates by\cite{Perrin1934, Perrin1936} |
1853 |
> |
\[ |
1854 |
> |
\frac{{x^2 }}{{a^2 }} + \frac{{y^2 }}{{b^2 }} + \frac{{z^2 }}{{c^2 |
1855 |
> |
}} = 1 |
1856 |
> |
\] |
1857 |
> |
where the semi-axes are of lengths $a$, $b$, and $c$. Unfortunately, |
1858 |
> |
due to the complexity of the elliptic integral, only the ellipsoid |
1859 |
> |
with the restriction of two axes having to be equal, \textit{i.e.} |
1860 |
> |
prolate($ a \ge b = c$) and oblate ($ a < b = c $), can be solved |
1861 |
> |
exactly. Introducing an elliptic integral parameter $S$ for prolate, |
1862 |
> |
\[ |
1863 |
> |
S = \frac{2}{{\sqrt {a^2 - b^2 } }}\ln \frac{{a + \sqrt {a^2 - b^2 |
1864 |
> |
} }}{b}, |
1865 |
> |
\] |
1866 |
> |
and oblate, |
1867 |
> |
\[ |
1868 |
> |
S = \frac{2}{{\sqrt {b^2 - a^2 } }}arctg\frac{{\sqrt {b^2 - a^2 } |
1869 |
> |
}}{a} |
1870 |
> |
\], |
1871 |
> |
one can write down the translational and rotational resistance |
1872 |
> |
tensors |
1873 |
> |
\[ |
1874 |
> |
\begin{array}{l} |
1875 |
> |
\Xi _a^{tt} = 16\pi \eta \frac{{a^2 - b^2 }}{{(2a^2 - b^2 )S - 2a}} \\ |
1876 |
> |
\Xi _b^{tt} = \Xi _c^{tt} = 32\pi \eta \frac{{a^2 - b^2 }}{{(2a^2 - 3b^2 )S + 2a}} \\ |
1877 |
> |
\end{array}, |
1878 |
> |
\] |
1879 |
> |
and |
1880 |
> |
\[ |
1881 |
> |
\begin{array}{l} |
1882 |
> |
\Xi _a^{rr} = \frac{{32\pi }}{3}\eta \frac{{(a^2 - b^2 )b^2 }}{{2a - b^2 S}} \\ |
1883 |
> |
\Xi _b^{rr} = \Xi _c^{rr} = \frac{{32\pi }}{3}\eta \frac{{(a^4 - b^4 )}}{{(2a^2 - b^2 )S - 2a}} \\ |
1884 |
> |
\end{array}. |
1885 |
> |
\] |
1886 |
> |
|
1887 |
> |
\subsubsection{\label{introSection:resistanceTensorRegularArbitrary}The Resistance Tensor for Arbitrary Shape} |
1888 |
> |
|
1889 |
> |
Unlike spherical and other regular shaped molecules, there is not |
1890 |
> |
analytical solution for friction tensor of any arbitrary shaped |
1891 |
> |
rigid molecules. The ellipsoid of revolution model and general |
1892 |
> |
triaxial ellipsoid model have been used to approximate the |
1893 |
> |
hydrodynamic properties of rigid bodies. However, since the mapping |
1894 |
> |
from all possible ellipsoidal space, $r$-space, to all possible |
1895 |
> |
combination of rotational diffusion coefficients, $D$-space is not |
1896 |
> |
unique\cite{Wegener1979} as well as the intrinsic coupling between |
1897 |
> |
translational and rotational motion of rigid body, general ellipsoid |
1898 |
> |
is not always suitable for modeling arbitrarily shaped rigid |
1899 |
> |
molecule. A number of studies have been devoted to determine the |
1900 |
> |
friction tensor for irregularly shaped rigid bodies using more |
1901 |
> |
advanced method where the molecule of interest was modeled by |
1902 |
> |
combinations of spheres(beads)\cite{Carrasco1999} and the |
1903 |
> |
hydrodynamics properties of the molecule can be calculated using the |
1904 |
> |
hydrodynamic interaction tensor. Let us consider a rigid assembly of |
1905 |
> |
$N$ beads immersed in a continuous medium. Due to hydrodynamics |
1906 |
> |
interaction, the ``net'' velocity of $i$th bead, $v'_i$ is different |
1907 |
> |
than its unperturbed velocity $v_i$, |
1908 |
> |
\[ |
1909 |
> |
v'_i = v_i - \sum\limits_{j \ne i} {T_{ij} F_j } |
1910 |
> |
\] |
1911 |
> |
where $F_i$ is the frictional force, and $T_{ij}$ is the |
1912 |
> |
hydrodynamic interaction tensor. The friction force of $i$th bead is |
1913 |
> |
proportional to its ``net'' velocity |
1914 |
> |
\begin{equation} |
1915 |
> |
F_i = \zeta _i v_i - \zeta _i \sum\limits_{j \ne i} {T_{ij} F_j }. |
1916 |
> |
\label{introEquation:tensorExpression} |
1917 |
> |
\end{equation} |
1918 |
> |
This equation is the basis for deriving the hydrodynamic tensor. In |
1919 |
> |
1930, Oseen and Burgers gave a simple solution to Equation |
1920 |
> |
\ref{introEquation:tensorExpression} |
1921 |
> |
\begin{equation} |
1922 |
> |
T_{ij} = \frac{1}{{8\pi \eta r_{ij} }}\left( {I + \frac{{R_{ij} |
1923 |
> |
R_{ij}^T }}{{R_{ij}^2 }}} \right). |
1924 |
> |
\label{introEquation:oseenTensor} |
1925 |
> |
\end{equation} |
1926 |
> |
Here $R_{ij}$ is the distance vector between bead $i$ and bead $j$. |
1927 |
> |
A second order expression for element of different size was |
1928 |
> |
introduced by Rotne and Prager\cite{Rotne1969} and improved by |
1929 |
> |
Garc\'{i}a de la Torre and Bloomfield\cite{Torre1977}, |
1930 |
> |
\begin{equation} |
1931 |
> |
T_{ij} = \frac{1}{{8\pi \eta R_{ij} }}\left[ {\left( {I + |
1932 |
> |
\frac{{R_{ij} R_{ij}^T }}{{R_{ij}^2 }}} \right) + R\frac{{\sigma |
1933 |
> |
_i^2 + \sigma _j^2 }}{{r_{ij}^2 }}\left( {\frac{I}{3} - |
1934 |
> |
\frac{{R_{ij} R_{ij}^T }}{{R_{ij}^2 }}} \right)} \right]. |
1935 |
> |
\label{introEquation:RPTensorNonOverlapped} |
1936 |
> |
\end{equation} |
1937 |
> |
Both of the Equation \ref{introEquation:oseenTensor} and Equation |
1938 |
> |
\ref{introEquation:RPTensorNonOverlapped} have an assumption $R_{ij} |
1939 |
> |
\ge \sigma _i + \sigma _j$. An alternative expression for |
1940 |
> |
overlapping beads with the same radius, $\sigma$, is given by |
1941 |
> |
\begin{equation} |
1942 |
> |
T_{ij} = \frac{1}{{6\pi \eta R_{ij} }}\left[ {\left( {1 - |
1943 |
> |
\frac{2}{{32}}\frac{{R_{ij} }}{\sigma }} \right)I + |
1944 |
> |
\frac{2}{{32}}\frac{{R_{ij} R_{ij}^T }}{{R_{ij} \sigma }}} \right] |
1945 |
> |
\label{introEquation:RPTensorOverlapped} |
1946 |
> |
\end{equation} |
1947 |
> |
|
1948 |
> |
To calculate the resistance tensor at an arbitrary origin $O$, we |
1949 |
> |
construct a $3N \times 3N$ matrix consisting of $N \times N$ |
1950 |
> |
$B_{ij}$ blocks |
1951 |
> |
\begin{equation} |
1952 |
> |
B = \left( {\begin{array}{*{20}c} |
1953 |
> |
{B_{11} } & \ldots & {B_{1N} } \\ |
1954 |
> |
\vdots & \ddots & \vdots \\ |
1955 |
> |
{B_{N1} } & \cdots & {B_{NN} } \\ |
1956 |
> |
\end{array}} \right), |
1957 |
> |
\end{equation} |
1958 |
> |
where $B_{ij}$ is given by |
1959 |
> |
\[ |
1960 |
> |
B_{ij} = \delta _{ij} \frac{I}{{6\pi \eta R}} + (1 - \delta _{ij} |
1961 |
> |
)T_{ij} |
1962 |
> |
\] |
1963 |
> |
where $\delta _{ij}$ is Kronecker delta function. Inverting matrix |
1964 |
> |
$B$, we obtain |
1965 |
> |
|
1966 |
> |
\[ |
1967 |
> |
C = B^{ - 1} = \left( {\begin{array}{*{20}c} |
1968 |
> |
{C_{11} } & \ldots & {C_{1N} } \\ |
1969 |
> |
\vdots & \ddots & \vdots \\ |
1970 |
> |
{C_{N1} } & \cdots & {C_{NN} } \\ |
1971 |
> |
\end{array}} \right) |
1972 |
> |
\] |
1973 |
> |
, which can be partitioned into $N \times N$ $3 \times 3$ block |
1974 |
> |
$C_{ij}$. With the help of $C_{ij}$ and skew matrix $U_i$ |
1975 |
> |
\[ |
1976 |
> |
U_i = \left( {\begin{array}{*{20}c} |
1977 |
> |
0 & { - z_i } & {y_i } \\ |
1978 |
> |
{z_i } & 0 & { - x_i } \\ |
1979 |
> |
{ - y_i } & {x_i } & 0 \\ |
1980 |
> |
\end{array}} \right) |
1981 |
> |
\] |
1982 |
> |
where $x_i$, $y_i$, $z_i$ are the components of the vector joining |
1983 |
> |
bead $i$ and origin $O$. Hence, the elements of resistance tensor at |
1984 |
> |
arbitrary origin $O$ can be written as |
1985 |
> |
\begin{equation} |
1986 |
> |
\begin{array}{l} |
1987 |
> |
\Xi _{}^{tt} = \sum\limits_i {\sum\limits_j {C_{ij} } } , \\ |
1988 |
> |
\Xi _{}^{tr} = \Xi _{}^{rt} = \sum\limits_i {\sum\limits_j {U_i C_{ij} } } , \\ |
1989 |
> |
\Xi _{}^{rr} = - \sum\limits_i {\sum\limits_j {U_i C_{ij} } } U_j \\ |
1990 |
> |
\end{array} |
1991 |
> |
\label{introEquation:ResistanceTensorArbitraryOrigin} |
1992 |
> |
\end{equation} |
1993 |
> |
|
1994 |
> |
The resistance tensor depends on the origin to which they refer. The |
1995 |
> |
proper location for applying friction force is the center of |
1996 |
> |
resistance (reaction), at which the trace of rotational resistance |
1997 |
> |
tensor, $ \Xi ^{rr}$ reaches minimum. Mathematically, the center of |
1998 |
> |
resistance is defined as an unique point of the rigid body at which |
1999 |
> |
the translation-rotation coupling tensor are symmetric, |
2000 |
> |
\begin{equation} |
2001 |
> |
\Xi^{tr} = \left( {\Xi^{tr} } \right)^T |
2002 |
> |
\label{introEquation:definitionCR} |
2003 |
> |
\end{equation} |
2004 |
> |
Form Equation \ref{introEquation:ResistanceTensorArbitraryOrigin}, |
2005 |
> |
we can easily find out that the translational resistance tensor is |
2006 |
> |
origin independent, while the rotational resistance tensor and |
2007 |
> |
translation-rotation coupling resistance tensor depend on the |
2008 |
> |
origin. Given resistance tensor at an arbitrary origin $O$, and a |
2009 |
> |
vector ,$r_{OP}(x_{OP}, y_{OP}, z_{OP})$, from $O$ to $P$, we can |
2010 |
> |
obtain the resistance tensor at $P$ by |
2011 |
> |
\begin{equation} |
2012 |
> |
\begin{array}{l} |
2013 |
> |
\Xi _P^{tt} = \Xi _O^{tt} \\ |
2014 |
> |
\Xi _P^{tr} = \Xi _P^{rt} = \Xi _O^{tr} - U_{OP} \Xi _O^{tt} \\ |
2015 |
> |
\Xi _P^{rr} = \Xi _O^{rr} - U_{OP} \Xi _O^{tt} U_{OP} + \Xi _O^{tr} U_{OP} - U_{OP} \Xi _O^{tr} ^{^T } \\ |
2016 |
> |
\end{array} |
2017 |
> |
\label{introEquation:resistanceTensorTransformation} |
2018 |
> |
\end{equation} |
2019 |
> |
where |
2020 |
> |
\[ |
2021 |
> |
U_{OP} = \left( {\begin{array}{*{20}c} |
2022 |
> |
0 & { - z_{OP} } & {y_{OP} } \\ |
2023 |
> |
{z_i } & 0 & { - x_{OP} } \\ |
2024 |
> |
{ - y_{OP} } & {x_{OP} } & 0 \\ |
2025 |
> |
\end{array}} \right) |
2026 |
> |
\] |
2027 |
> |
Using Equations \ref{introEquation:definitionCR} and |
2028 |
> |
\ref{introEquation:resistanceTensorTransformation}, one can locate |
2029 |
> |
the position of center of resistance, |
2030 |
> |
\begin{eqnarray*} |
2031 |
> |
\left( \begin{array}{l} |
2032 |
> |
x_{OR} \\ |
2033 |
> |
y_{OR} \\ |
2034 |
> |
z_{OR} \\ |
2035 |
> |
\end{array} \right) & = &\left( {\begin{array}{*{20}c} |
2036 |
> |
{(\Xi _O^{rr} )_{yy} + (\Xi _O^{rr} )_{zz} } & { - (\Xi _O^{rr} )_{xy} } & { - (\Xi _O^{rr} )_{xz} } \\ |
2037 |
> |
{ - (\Xi _O^{rr} )_{xy} } & {(\Xi _O^{rr} )_{zz} + (\Xi _O^{rr} )_{xx} } & { - (\Xi _O^{rr} )_{yz} } \\ |
2038 |
> |
{ - (\Xi _O^{rr} )_{xz} } & { - (\Xi _O^{rr} )_{yz} } & {(\Xi _O^{rr} )_{xx} + (\Xi _O^{rr} )_{yy} } \\ |
2039 |
> |
\end{array}} \right)^{ - 1} \\ |
2040 |
> |
& & \left( \begin{array}{l} |
2041 |
> |
(\Xi _O^{tr} )_{yz} - (\Xi _O^{tr} )_{zy} \\ |
2042 |
> |
(\Xi _O^{tr} )_{zx} - (\Xi _O^{tr} )_{xz} \\ |
2043 |
> |
(\Xi _O^{tr} )_{xy} - (\Xi _O^{tr} )_{yx} \\ |
2044 |
> |
\end{array} \right) \\ |
2045 |
> |
\end{eqnarray*} |
2046 |
> |
|
2047 |
> |
|
2048 |
> |
|
2049 |
> |
where $x_OR$, $y_OR$, $z_OR$ are the components of the vector |
2050 |
> |
joining center of resistance $R$ and origin $O$. |