| 1 |
\chapter{\label{chap:conclusion}CONCLUSION} |
| 2 |
|
| 3 |
The preceding chapters and included appendices discuss the primary |
| 4 |
aspects of the research I have performed and been involved with over |
| 5 |
the last several years. Rather than presenting the topics in a |
| 6 |
chronological fashion, they were arranged to form a series where the |
| 7 |
later topics apply and extend the findings of the former topics. This |
| 8 |
layout is more instructive and provides a more cohesive progression of |
| 9 |
research efforts. |
| 10 |
|
| 11 |
The first chapter laid out the foundation from which the research in |
| 12 |
the later chapters is built upon, primarily the technique of molecular |
| 13 |
dynamics. This chapter also introduces {\sc oopse}, the object |
| 14 |
oriented parallel simulation engine, the unified code-base developed in |
| 15 |
our lab for performing molecular simulations. Starting out as a |
| 16 |
collection of separate programs written by different group members, |
| 17 |
{\sc oopse} has developed into one of the few parallel molecular |
| 18 |
dynamics packages capable of accurately integrating rigid bodies, |
| 19 |
point multipoles, and metallic potentials.\cite{Meineke05} |
| 20 |
|
| 21 |
The second chapter discussed correction techniques for handling the |
| 22 |
long-ranged electrostatic interactions common in molecular |
| 23 |
simulations, in particular our shifted-force ({\sc sf}) modification |
| 24 |
of the damped shifted Coulombic summation method developed by Wolf |
| 25 |
{\it et al.}\cite{Wolf99} In the work outlined here, we showed {\sc |
| 26 |
sf} to be equivalent to the more prevalent Ewald summation in |
| 27 |
simulations of condensed phases, and since it is pairwise, it scales |
| 28 |
as $\mathcal{O}(N)$ and lacks periodicity artifacts introduced through |
| 29 |
heavy reliance on the reciprocal-space portion of the Ewald sum. We |
| 30 |
extended the electrostatic damping technique used with {\sc sf} beyond |
| 31 |
simple charge-charge interactions to include point-multipoles, and we |
| 32 |
also identified optimal damping parameter settings to ensure proper |
| 33 |
depiction of the dielectric behavior of molecular systems. Presenting |
| 34 |
this technique early enables us to apply it in the systems discussed |
| 35 |
in the later chapters and show how it can improve the quality of |
| 36 |
various molecular simulations. |
| 37 |
|
| 38 |
The third chapter focused on simple water models, specifically the |
| 39 |
single-point soft sticky dipole (SSD) model for water. We implemented |
| 40 |
this model and realized that we need to reparametrize it in order to |
| 41 |
use it in our simulations. This lead to the development of SSD/RF and |
| 42 |
SSD/E, new variants of the SSD model optimized for simulations with |
| 43 |
and without a reaction field correction. These new single-point models |
| 44 |
are more efficient than the common multi-point partial charge models |
| 45 |
and better capture the dynamic properties of water. We also showed |
| 46 |
that SSD/RF can be successfully used with damped {\sc sf} through our |
| 47 |
multipolar extension of the technique. For the sake of completeness, |
| 48 |
we also developed the two-point tetrahedrally restructured elongated |
| 49 |
dipole (TRED) water model, which is optimized for use with the damped |
| 50 |
{\sc sf} technique. Though there remain some algorithmic complexities |
| 51 |
that need to be addressed (logic for neglecting charge-quadrupole |
| 52 |
interactions between other TRED molecules) to use this model in |
| 53 |
general simulations, it is approximately twice as efficient as the |
| 54 |
commonly used three-point charge water models (i.e. TIP3P and |
| 55 |
SPC/E). This work succeeds in extending the limits of the |
| 56 |
computational efficiency of water models that can capture the |
| 57 |
thermodynamic and dynamic properties of liquid water. |
| 58 |
|
| 59 |
The final chapter deals with a unique polymorph of ice that we |
| 60 |
discovered while performing water simulations with the fast simple |
| 61 |
water models discussed in the previous chapter. This form of ice, |
| 62 |
which we called ``imaginary ice'' (Ice-$i$), has a low-density |
| 63 |
structure which is different from any known polymorph from either |
| 64 |
experiment or other simulations. The free energy analysis performed |
| 65 |
here shows that this structure is in fact the thermodynamically |
| 66 |
preferred form of ice for both the single-point and commonly used |
| 67 |
multi-point water models under the chosen simulation conditions. We |
| 68 |
then showed that inclusion of electrostatic corrections is necessary |
| 69 |
to obtain more realistic results; however, the free energies of the |
| 70 |
various polymorphs (both imaginary and real) in many of these models |
| 71 |
was shown to be so similar that choice of system properties, like the |
| 72 |
volume in $NVT$ simulations, will directly influence the expressed ice |
| 73 |
polymorph. This work shows that researchers ought to be wary of using |
| 74 |
these simplistic water models in the study of complex phase behavior |
| 75 |
where the choice of a water model that includes many-body effects, |
| 76 |
such as polarizability, might be more appropriate. |
| 77 |
|
| 78 |
The work presented in this dissertation includes advancements in |
| 79 |
simulation techniques, improved molecular models, and applications of |
| 80 |
both in simulations of novel molecular systems. In addition to |
| 81 |
answering interesting questions related to these topics, this work |
| 82 |
opens up new routes which other researchers can utilize to extend and |
| 83 |
improve their own work. Though specific in focus, through pathways |
| 84 |
such as these, this work can gain wider utility and expand our |
| 85 |
understanding of natural physical and chemical processes. |
