Why computer simulations?

• Simulations are relatively simple, inexpensive, and everything can be measured in principle.
• Not to reproduce experimental results (exception: testing the accuracy of potentials)
• Help to understand experimental results or to propose new experiments
• Test of theoretical predictions or theories
• Investigating systems on a level of detail which is not possible in real experiments or analytical theories (local structure, mechanics .., etc.)
• Create new materials

Computer simulations often use an atomistic approach to find answers to interesting effect, technical problems, and theoretical predictions. At atomistic level simulation, particles interacting with each other is described by two methods:

ab initio calculation:

- positions of atoms -> electronic

- structure of the system -> potential

- computationally very expensive

Effective potential:

- Forms of potential is ad hoc.

- Parameters are obtained from fitting experimental data or restuls from ab initio calculations

- Cheap

The bottleneck for effective potentials (classical MD simulation) is about the time scale and accuracy of the potential, which have to be overcome so that the simulation can develop in their full capacity.

LAMMPS - A free open-source MD package

LAMMPS stands for "Large-scale Atomic/Molecular Massively Parallel Simulator", which is a molecular dynamics simulation package distributed by National Sandia Lab. LAMMPS incorporate MPI so it could run in parallel or on single processor.

The reason that I chose LAMMPS to simulate my ionomers for several reasons:

1. It could run coarse-grained/united-atom simulations - LAMMPS includes many widely used empirical potentials.
2. All codes are written in C++, so I am able to go into the code and do modification - the dihedral potential that I use for my ionomers are not included in the potentials LAMMPS provides. So I modified a similar potential in LAMMPS to describe the dihedral interactions.
3. Maintained and developed by experts - This is the most important. LAMMPS has a newer version almost every year. Many new useful functions/features are added and useful force fields are incorporated. It also has an very active mail-list.

To keep LAMMPS a simple and fast simulator, LAMMPS does not help you post-process data. This is very different from other MD package like Gromacs or NAMD, which performs analyses of your simulation. Because it doesn't have a GUI interface, so LAMMPS won't visualize your simulation either. There are a few tools that allow you to do some pre/post-process, but I found it is much easier to do analyses in my own C++ codes.

To know more specifics about LAMMPS, the official website has all you need to know, and is always up-to-date.

Molecular dynamics simulation

MD (molecular dynamics) simulation is a computational technique that describes the movement of atoms based on their interactions.The interactions among atoms can be described by empirical potentials (force field) or quantum chemical models, or a mix of the two. The most commonly seen MD in polymer or bio- physics uses force fields, and follows classical mechanics: solving Newton's equations of motion numerically to get the atom trajectory. By tracking where the atoms have been, structural and dynamical information can be extracted.

This type of classical MD has force field describing both intra- and inter-molecular interactions. Usually the force field refers to a set of empirical equations to describe the most essential part of the interaction. For example, the bonding between two atoms are often described in harmonic potential:

Bonding potential = K ( x - x₀)²

Depending on the complexity of the potentials, MD simulation could achieve different levels of realism. The force fields that use harmonic equation for bonding is capable of agreeing to some extent the structural and dynamical information, but incapable of describing any chemical reactions. Ab initio based MD simulation could provide very accurate electron distribution and bond breaking/forming, but the trade of is that the large amount of computational time required.

It is important to know what properties one's interested in finding out using MD simulation, and choose the force field accordingly. There is always a compromise between accuracy and time/length scales of MD simulation.

Read more about MD: MD simulation on wikipedia