ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/group/trunk/mattDisertation/Introduction.tex
(Generate patch)

Comparing trunk/mattDisertation/Introduction.tex (file contents):
Revision 977 by mmeineke, Thu Jan 22 21:13:55 2004 UTC vs.
Revision 1001 by mmeineke, Sat Jan 31 22:10:21 2004 UTC

# Line 30 | Line 30 | which method best suits that objective.
30   thermodynamic properties of the system are being probed, then chose
31   which method best suits that objective.
32  
33 < \subsection{\label{introSec:statThermo}Statistical Thermodynamics}
33 > \subsection{\label{introSec:statThermo}Statistical Mechanics}
34  
35 < ergodic hypothesis
35 > The following section serves as a brief introduction to some of the
36 > Statistical Mechanics concepts present in this dissertation.  What
37 > follows is a brief derivation of Blotzman weighted statistics, and an
38 > explanation of how one can use the information to calculate an
39 > observable for a system.  This section then concludes with a brief
40 > discussion of the ergodic hypothesis and its relevance to this
41 > research.
42  
43 < enesemble averages
43 > \subsection{\label{introSec:boltzman}Boltzman weighted statistics}
44  
45 + Consider a system, $\gamma$, with some total energy,, $E_{\gamma}$.
46 + Let $\Omega(E_{gamma})$ represent the number of degenerate ways
47 + $\boldsymbol{\Gamma}$, the collection of positions and conjugate
48 + momenta of system $\gamma$, can be configured to give
49 + $E_{\gamma}$. Further, if $\gamma$ is in contact with a bath system
50 + where energy is exchanged between the two systems, $\Omega(E)$, where
51 + $E$ is the total energy of both systems, can be represented as
52 + \begin{equation}
53 + eq here
54 + \label{introEq:SM1}
55 + \end{equation}
56 + Or additively as
57 + \begin{equation}
58 + eq here
59 + \label{introEq:SM2}
60 + \end{equation}
61 +
62 + The solution to Eq.~\ref{introEq:SM2} maximizes the number of
63 + degenerative configurations in $E$. \cite{fix}
64 + This gives
65 + \begin{equation}
66 + eq here
67 + \label{introEq:SM3}
68 + \end{equation}
69 + Where $E_{\text{bath}}$ is $E-E_{\gamma}$, and
70 + $\frac{partialE_{\text{bath}}}{\partial E_{\gamma}}$ is
71 + $-1$. Eq.~\ref{introEq:SM3} becomes
72 + \begin{equation}
73 + eq here
74 + \label{introEq:SM4}
75 + \end{equation}
76 +
77 + At this point, one can draw a relationship between the maximization of
78 + degeneracy in Eq.~\ref{introEq:SM3} and the second law of
79 + thermodynamics.  Namely, that for a closed system, entropy wil
80 + increase for an irreversible process.\cite{fix} Here the
81 + process is the partitioning of energy among the two systems.  This
82 + allows the following definition of entropy:
83 + \begin{equation}
84 + eq here
85 + \label{introEq:SM5}
86 + \end{equation}
87 + Where $k_B$ is the Boltzman constant.  Having defined entropy, one can
88 + also define the temperature of the system using the relation
89 + \begin{equation}
90 + eq here
91 + \label{introEq:SM6}
92 + \end{equation}
93 + The temperature in the system $\gamma$ is then
94 + \begin{equation}
95 + eq here
96 + \label{introEq:SM7}
97 + \end{equation}
98 + Applying this to Eq.~\ref{introEq:SM4} gives the following
99 + \begin{equation}
100 + eq here
101 + \label{introEq:SM8}
102 + \end{equation}
103 + Showing that the partitioning of energy between the two systems is
104 + actually a process of thermal equilibration. \cite{fix}
105 +
106 + An application of these results is to formulate the form of an
107 + expectation value of an observable, $A$, in the canonical ensemble. In
108 + the canonical ensemble, the number of particles, $N$, the volume, $V$,
109 + and the temperature, $T$, are all held constant while the energy, $E$,
110 + is allowed to fluctuate. Returning to the previous example, the bath
111 + system is now an infinitly large thermal bath, whose exchange of
112 + energy with the system $\gamma$ holds teh temperature constant.  The
113 + partitioning of energy in the bath system is then related to the total
114 + energy of both systems and the fluctuations in $E_{\gamma}}$:
115 + \begin{equation}
116 + eq here
117 + \label{introEq:SM9}
118 + \end{equation}
119 + As for the expectation value, it can be defined
120 + \begin{equation}
121 + eq here
122 + \label{introEq:SM10}
123 + \end{eequation}
124 + Where $\int_{\boldsymbol{\Gamma}} d\Boldsymbol{\Gamma}$ denotes an
125 + integration over all accessable phase space, $P_{\gamma}$ is the
126 + probability of being in a given phase state and
127 + $A(\boldsymbol{\Gamma})$ is some observable that is a function of the
128 + phase state.
129 +
130 + Because entropy seeks to maximize the number of degenerate states at a
131 + given energy, the probability of being in a particular state in
132 + $\gamma$ will be directly proportional to the number of allowable
133 + states the coupled system is able to assume. Namely,
134 + \begin{equation}
135 + eq here
136 + \label{introEq:SM11}
137 + \end{equation}
138 + With $E_{\gamma} \lE$, $\ln \Omega$ can be expanded in a Taylor series:
139 + \begin{equation}
140 + eq here
141 + \label{introEq:SM12}
142 + \end{equation}
143 + Higher order terms are omitted as $E$ is an infinite thermal
144 + bath. Further, using Eq.~\ref{introEq:SM7}, Eq.~\ref{introEq:SM11} can
145 + be rewritten:
146 + \begin{equation}
147 + eq here
148 + \label{introEq:SM13}
149 + \end{equation}
150 + Where $\ln \Omega(E)$ has been factored out of the porpotionality as a
151 + constant.  Normalizing the probability ($\int_{\boldsymbol{\Gamma}}
152 + d\boldsymbol{\Gamma} P_{\gamma} =1$ gives
153 + \begin{equation}
154 + eq here
155 + \label{introEq:SM14}
156 + \end{equation}
157 + This result is the standard Boltzman statistical distribution.
158 + Applying it to Eq.~\ref{introEq:SM10} one can obtain the following relation for ensemble averages:
159 + \begin{equation}
160 + eq here
161 + \label{introEq:SM15}
162 + \end{equation}
163 +
164 + \subsection{\label{introSec:ergodic}The Ergodic Hypothesis}
165 +
166 + One last important consideration is that of ergodicity. Ergodicity is
167 + the assumption that given an infinite amount of time, a system will
168 + visit every available point in phase space.\cite{fix} For most
169 + systems, this is a valid assumption, except in cases where the system
170 + may be trapped in a local feature (\emph{i.~e.~glasses}). When valid,
171 + ergodicity allows the unification of a time averaged observation and
172 + an ensemble averged one. If an observation is averaged over a
173 + sufficiently long time, the system is assumed to visit all
174 + appropriately available points in phase space, giving a properly
175 + weighted statistical average. This allows the researcher freedom of
176 + choice when deciding how best to measure a given observable.  When an
177 + ensemble averaged approach seems most logical, the Monte Carlo
178 + techniques described in Sec.~\ref{introSec:MC} can be utilized.
179 + Conversely, if a problem lends itself to a time averaging approach,
180 + the Molecular Dynamics techniques in Sec.~\ref{introSec:MD} can be
181 + employed.
182 +
183   \subsection{\label{introSec:monteCarlo}Monte Carlo Simulations}
184  
185   The Monte Carlo method was developed by Metropolis and Ulam for their
# Line 243 | Line 387 | techniques, is normally decided by the observables in
387  
388   The choice of when to use molecular dynamics over Monte Carlo
389   techniques, is normally decided by the observables in which the
390 < researcher is interested.  If the observabvles depend on momenta in
390 > researcher is interested.  If the observables depend on momenta in
391   any fashion, then the only choice is molecular dynamics in some form.
392   However, when the observable is dependent only on the configuration,
393   then most of the time Monte Carlo techniques will be more efficent.
# Line 306 | Line 450 | a disproportionate number of the particles will feel t
450  
451   Another consideration one must resolve, is that in a given simulation
452   a disproportionate number of the particles will feel the effects of
453 < the surface. \cite{fix} For a cubic system of 1000 particles arranged
453 > the surface.\cite{fix} For a cubic system of 1000 particles arranged
454   in a $10x10x10$ cube, 488 particles will be exposed to the surface.
455   Unless one is simulating an isolated particle group in a vacuum, the
456   behavior of the system will be far from the desired bulk
457   charecteristics.  To offset this, simulations employ the use of
458 < periodic boundary images. \cite{fix}
458 > periodic boundary images.\cite{fix}
459  
460   The technique involves the use of an algorithm that replicates the
461   simulation box on an infinite lattice in cartesian space.  Any given
# Line 329 | Line 473 | predetermined distance, $fix$, are not included in the
473   cutoff radius be employed.  Using a cutoff radius improves the
474   efficiency of the force evaluation, as particles farther than a
475   predetermined distance, $fix$, are not included in the
476 < calculation. \cite{fix} In a simultation with periodic images, $fix$
476 > calculation.\cite{fix} In a simultation with periodic images, $fix$
477   has a maximum value of $fix$.  Fig.~\ref{fix} illustrates how using an
478   $fix$ larger than this value, or in the extreme limit of no $fix$ at
479   all, the corners of the simulation box are unequally weighted due to
480   the lack of particle images in the $x$, $y$, or $z$ directions past a
481   disance of $fix$.
338
339 With the use of an $fix$, however, comes a discontinuity in the potential energy curve (Fig.~\ref{fix}).
482  
483 + With the use of an $fix$, however, comes a discontinuity in the
484 + potential energy curve (Fig.~\ref{fix}). To fix this discontinuity,
485 + one calculates the potential energy at the $r_{\text{cut}}$, and add
486 + that value to the potential.  This causes the function to go smoothly
487 + to zero at the cutoff radius.  This ensures conservation of energy
488 + when integrating the Newtonian equations of motion.
489  
490 < \section{\label{introSec:chapterLayout}Chapter Layout}
491 <
492 < \subsection{\label{introSec:RSA}Random Sequential Adsorption}
490 > The second main simplification used in this research is the Verlet
491 > neighbor list. \cite{allen87:csl} In the Verlet method, one generates
492 > a list of all neighbor atoms, $j$, surrounding atom $i$ within some
493 > cutoff $r_{\text{list}}$, where $r_{\text{list}}>r_{\text{cut}}$.
494 > This list is created the first time forces are evaluated, then on
495 > subsequent force evaluations, pair calculations are only calculated
496 > from the neighbor lists.  The lists are updated if any given particle
497 > in the system moves farther than $r_{\text{list}}-r_{\text{cut}}$,
498 > giving rise to the possibility that a particle has left or joined a
499 > neighbor list.
500  
501 < \subsection{\label{introSec:OOPSE}The OOPSE Simulation Package}
501 > \subsection{\label{introSec:MDintegrate} Integration of the equations of motion}
502 >
503 > A starting point for the discussion of molecular dynamics integrators
504 > is the Verlet algorithm. \cite{Frenkel1996} It begins with a Taylor
505 > expansion of position in time:
506 > \begin{equation}
507 > eq here
508 > \label{introEq:verletForward}
509 > \end{equation}
510 > As well as,
511 > \begin{equation}
512 > eq here
513 > \label{introEq:verletBack}
514 > \end{equation}
515 > Adding together Eq.~\ref{introEq:verletForward} and
516 > Eq.~\ref{introEq:verletBack} results in,
517 > \begin{equation}
518 > eq here
519 > \label{introEq:verletSum}
520 > \end{equation}
521 > Or equivalently,
522 > \begin{equation}
523 > eq here
524 > \label{introEq:verletFinal}
525 > \end{equation}
526 > Which contains an error in the estimate of the new positions on the
527 > order of $\Delta t^4$.
528 >
529 > In practice, however, the simulations in this research were integrated
530 > with a velocity reformulation of teh Verlet method.\cite{allen87:csl}
531 > \begin{equation}
532 > eq here
533 > \label{introEq:MDvelVerletPos}
534 > \end{equation}
535 > \begin{equation}
536 > eq here
537 > \label{introEq:MDvelVerletVel}
538 > \end{equation}
539 > The original Verlet algorithm can be regained by substituting the
540 > velocity back into Eq.~\ref{introEq:MDvelVerletPos}.  The Verlet
541 > formulations are chosen in this research because the algorithms have
542 > very little long term drift in energy conservation.  Energy
543 > conservation in a molecular dynamics simulation is of extreme
544 > importance, as it is a measure of how closely one is following the
545 > ``true'' trajectory wtih the finite integration scheme.  An exact
546 > solution to the integration will conserve area in phase space, as well
547 > as be reversible in time, that is, the trajectory integrated forward
548 > or backwards will exactly match itself.  Having a finite algorithm
549 > that both conserves area in phase space and is time reversible,
550 > therefore increases, but does not guarantee the ``correctness'' or the
551 > integrated trajectory.
552  
553 < \subsection{\label{introSec:bilayers}A Mesoscale Model for Phospholipid Bilayers}
553 > It can be shown,\cite{Frenkel1996} that although the Verlet algorithm
554 > does not rigorously preserve the actual Hamiltonian, it does preserve
555 > a pseudo-Hamiltonian which shadows the real one in phase space.  This
556 > pseudo-Hamiltonian is proveably area-conserving as well as time
557 > reversible.  The fact that it shadows the true Hamiltonian in phase
558 > space is acceptable in actual simulations as one is interested in the
559 > ensemble average of the observable being measured.  From the ergodic
560 > hypothesis (Sec.~\ref{introSec:StatThermo}), it is known that the time
561 > average will match the ensemble average, therefore two similar
562 > trajectories in phase space should give matching statistical averages.
563 >
564 > \subsection{\label{introSec:MDfurther}Further Considerations}
565 > In the simulations presented in this research, a few additional
566 > parameters are needed to describe the motions.  The simulations
567 > involving water and phospholipids in Chapt.~\ref{chaptLipids} are
568 > required to integrate the equations of motions for dipoles on atoms.
569 > This involves an additional three parameters be specified for each
570 > dipole atom: $\phi$, $\theta$, and $\psi$.  These three angles are
571 > taken to be the Euler angles, where $\phi$ is a rotation about the
572 > $z$-axis, and $\theta$ is a rotation about the new $x$-axis, and
573 > $\psi$ is a final rotation about the new $z$-axis (see
574 > Fig.~\ref{introFig:euleerAngles}).  This sequence of rotations can be
575 > accumulated into a single $3 \times 3$ matrix $\mathbf{A}$
576 > defined as follows:
577 > \begin{equation}
578 > eq here
579 > \label{introEq:EulerRotMat}
580 > \end{equation}
581 >
582 > The equations of motion for Euler angles can be written down as
583 > \cite{allen87:csl}
584 > \begin{equation}
585 > eq here
586 > \label{introEq:MDeuleeerPsi}
587 > \end{equation}
588 > Where $\omega^s_i$ is the angular velocity in the lab space frame
589 > along cartesian coordinate $i$.  However, a difficulty arises when
590 > attempting to integrate Eq.~\ref{introEq:MDeulerPhi} and
591 > Eq.~\ref{introEq:MDeulerPsi}. The $\frac{1}{\sin \theta}$ present in
592 > both equations means there is a non-physical instability present when
593 > $\theta$ is 0 or $\pi$.
594 >
595 > To correct for this, the simulations integrate the rotation matrix,
596 > $\mathbf{A}$, directly, thus avoiding the instability.
597 > This method was proposed by Dullwebber
598 > \emph{et. al.}\cite{Dullwebber:1997}, and is presented in
599 > Sec.~\ref{introSec:MDsymplecticRot}.
600 >
601 > \subsubsection{\label{introSec:MDliouville}Liouville Propagator}
602 >
603 > Before discussing the integration of the rotation matrix, it is
604 > necessary to understand the construction of a ``good'' integration
605 > scheme.  It has been previously
606 > discussed(Sec.~\ref{introSec:MDintegrate}) how it is desirable for an
607 > integrator to be symplectic, or time reversible.  The following is an
608 > outline of the Trotter factorization of the Liouville Propagator as a
609 > scheme for generating symplectic integrators. \cite{Tuckerman:1992}
610 >
611 > For a system with $f$ degrees of freedom the Liouville operator can be
612 > defined as,
613 > \begin{equation}
614 > eq here
615 > \label{introEq:LiouvilleOperator}
616 > \end{equation}
617 > Here, $r_j$ and $p_j$ are the position and conjugate momenta of a
618 > degree of freedom, and $f_j$ is the force on that degree of freedom.
619 > $\Gamma$ is defined as the set of all positions nad conjugate momenta,
620 > $\{r_j,p_j\}$, and the propagator, $U(t)$, is defined
621 > \begin {equation}
622 > eq here
623 > \label{introEq:Lpropagator}
624 > \end{equation}
625 > This allows the specification of $\Gamma$ at any time $t$ as
626 > \begin{equation}
627 > eq here
628 > \label{introEq:Lp2}
629 > \end{equation}
630 > It is important to note, $U(t)$ is a unitary operator meaning
631 > \begin{equation}
632 > U(-t)=U^{-1}(t)
633 > \label{introEq:Lp3}
634 > \end{equation}
635 >
636 > Decomposing $L$ into two parts, $iL_1$ and $iL_2$, one can use the
637 > Trotter theorem to yield
638 > \begin{equation}
639 > eq here
640 > \label{introEq:Lp4}
641 > \end{equation}
642 > Where $\Delta t = \frac{t}{P}$.
643 > With this, a discrete time operator $G(\Delta t)$ can be defined:
644 > \begin{equation}
645 > eq here
646 > \label{introEq:Lp5}
647 > \end{equation}
648 > Because $U_1(t)$ and $U_2(t)$ are unitary, $G|\Delta t)$ is also
649 > unitary.  Meaning an integrator based on this factorization will be
650 > reversible in time.
651 >
652 > As an example, consider the following decomposition of $L$:
653 > \begin{equation}
654 > eq here
655 > \label{introEq:Lp6}
656 > \end{equation}
657 > Operating $G(\Delta t)$ on $\Gamma)0)$, and utilizing the operator property
658 > \begin{equation}
659 > eq here
660 > \label{introEq:Lp8}
661 > \end{equation}
662 > Where $c$ is independent of $q$.  One obtains the following:  
663 > \begin{equation}
664 > eq here
665 > \label{introEq:Lp8}
666 > \end{equation}
667 > Or written another way,
668 > \begin{equation}
669 > eq here
670 > \label{intorEq:Lp9}
671 > \end{equation}
672 > This is the velocity Verlet formulation presented in
673 > Sec.~\ref{introSec:MDintegrate}.  Because this integration scheme is
674 > comprised of unitary propagators, it is symplectic, and therefore area
675 > preserving in phase space.  From the preceeding fatorization, one can
676 > see that the integration of the equations of motion would follow:
677 > \begin{enumerate}
678 > \item calculate the velocities at the half step, $\frac{\Delta t}{2}$, from the forces calculated at the initial position.
679 >
680 > \item Use the half step velocities to move positions one whole step, $\Delta t$.
681 >
682 > \item Evaluate the forces at the new positions, $\mathbf{r}(\Delta t)$, and use the new forces to complete the velocity move.
683 >
684 > \item Repeat from step 1 with the new position, velocities, and forces assuming the roles of the initial values.
685 > \end{enumerate}
686 >
687 > \subsubsection{\label{introSec:MDsymplecticRot} Symplectic Propagation of the Rotation Matrix}
688 >
689 > Based on the factorization from the previous section,
690 > Dullweber\emph{et al.}\cite{Dullweber:1997}~ proposed a scheme for the
691 > symplectic propagation of the rotation matrix, $\mathbf{A}$, as an
692 > alternative method for the integration of orientational degrees of
693 > freedom. The method starts with a straightforward splitting of the
694 > Liouville operator:
695 > \begin{equation}
696 > eq here
697 > \label{introEq:SR1}
698 > \end{equation}
699 > Where $\boldsymbol{\tau}(\mathbf{A})$ are the tourques of the system
700 > due to the configuration, and $\boldsymbol{/pi}$ are the conjugate
701 > angular momenta of the system. The propagator, $G(\Delta t)$, becomes
702 > \begin{equation}
703 > eq here
704 > \label{introEq:SR2}
705 > \end{equation}
706 > Propagation fo the linear and angular momenta follows as in the Verlet
707 > scheme.  The propagation of positions also follows the verlet scheme
708 > with the addition of a further symplectic splitting of the rotation
709 > matrix propagation, $\mathcal{G}_{\text{rot}}(\Delta t)$.
710 > \begin{equation}
711 > eq here
712 > \label{introEq:SR3}
713 > \end{equation}
714 > Where $\mathcal{G}_j$ is a unitary rotation of $\mathbf{A}$ and
715 > $\boldsymbol{\pi}$ about each axis $j$.  As all propagations are now
716 > unitary and symplectic, the entire integration scheme is also
717 > symplectic and time reversible.
718 >
719 > \section{\label{introSec:layout}Dissertation Layout}
720 >
721 > This dissertation is divided as follows:Chapt.~\ref{chapt:RSA}
722 > presents the random sequential adsorption simulations of related
723 > pthalocyanines on a gold (111) surface. Chapt.~\ref{chapt:OOPSE}
724 > is about the writing of the molecular dynamics simulation package
725 > {\sc oopse}, Chapt.~\ref{chapt:lipid} regards the simulations of
726 > phospholipid bilayers using a mesoscale model, and lastly,
727 > Chapt.~\ref{chapt:conclusion} concludes this dissertation with a
728 > summary of all results. The chapters are arranged in chronological
729 > order, and reflect the progression of techniques I employed during my
730 > research.  
731 >
732 > The chapter concerning random sequential adsorption
733 > simulations is a study in applying the principles of theoretical
734 > research in order to obtain a simple model capaable of explaining the
735 > results.  My advisor, Dr. Gezelter, and I were approached by a
736 > colleague, Dr. Lieberman, about possible explanations for partial
737 > coverge of a gold surface by a particular compound of hers. We
738 > suggested it might be due to the statistical packing fraction of disks
739 > on a plane, and set about to simulate this system.  As the events in
740 > our model were not dynamic in nature, a Monte Carlo method was
741 > emplyed.  Here, if a molecule landed on the surface without
742 > overlapping another, then its landing was accepted.  However, if there
743 > was overlap, the landing we rejected and a new random landing location
744 > was chosen.  This defined our acceptance rules and allowed us to
745 > construct a Markov chain whose limiting distribution was the surface
746 > coverage in which we were interested.
747 >
748 > The following chapter, about the simulation package {\sc oopse},
749 > describes in detail the large body of scientific code that had to be
750 > written in order to study phospholipid bilayer.  Although there are
751 > pre-existing molecular dynamic simulation packages available, none
752 > were capable of implementing the models we were developing.{\sc oopse}
753 > is a unique package capable of not only integrating the equations of
754 > motion in cartesian space, but is also able to integrate the
755 > rotational motion of rigid bodies and dipoles.  Add to this the
756 > ability to perform calculations across parallel processors and a
757 > flexible script syntax for creating systems, and {\sc oopse} becomes a
758 > very powerful scientific instrument for the exploration of our model.
759 >
760 > Bringing us to Chapt.~\ref{chapt:lipid}. Using {\sc oopse}, I have been
761 > able to parametrize a mesoscale model for phospholipid simulations.
762 > This model retains information about solvent ordering about the
763 > bilayer, as well as information regarding the interaction of the
764 > phospholipid head groups' dipole with each other and the surrounding
765 > solvent.  These simulations give us insight into the dynamic events
766 > that lead to the formation of phospholipid bilayers, as well as
767 > provide the foundation for future exploration of bilayer phase
768 > behavior with this model.  
769 >
770 > Which leads into the last chapter, where I discuss future directions
771 > for both{\sc oopse} and this mesoscale model.  Additionally, I will
772 > give a summary of results for this dissertation.
773 >
774 >

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines