It depends on which type of MD simulation you propose to do.

For example, plain MD which implemented classical force field, you need to know: bonded parameters (bond, angle, dihedral) and nonbonded parameters ( charge, vdw)

MD with polarize force field or Car-Parrinello is more complex

Hi,

basically you need to know about the charges on the atoms. This is necessary because the MD uses "forces" on the atoms. This force is changes with time in the dynamical sense. Secondly, you need to know about the initial velocities of the atoms which can either be choosen as random or calculated from Boltzmann distribution at temperature T. So, you have the force and the initial velocity to run the dynamics by Verlet algorithm etc.

In some non-user-friendly softwares it is possible to define your own force field or it may be required some Lennard-Jones parameters for any atom. In these cases you should calculate Lennard-Jones parameters etc and implement them into the force field before running your MD code.

The answer of Berkay Sütay is highly missleading and doesn't really hit the point. As Quyen V. Vu pointed out, it depends on what you need. If you need real chemistry happening in only small systems, you need QM. If you are ok with predefined bonds and residues and want large systems, using classical forcefield is of choice.

What do you need to know about an atom:

In QM you need a Basisset and often a Pseudopotential only special for an element, not a specific atom, that's it. You just build your system out of "elements".

In FF, you usually already have most elements covered by a variety of atom types specifacally parameterized to represent elements in various chemical enviroments and compounds. These atomtypes usually only have mass and Lennard-Jones parameters. Also, between atomtypes, there are bond, angle and dihedral parameters defined. From those atom types, there are built residues, such as amino acids, nucleic acids or solvents. In such a residue, each atom is defined by e.g. name, atomtype, coordinates, bonds and the charge. These partial charges are the only thing usually required to define new, for new residues. Have a look at e.g Amber GAFF for instructions on how to build new compounds. Finally, you build a whole molecular simulation system, like a large Protein solvated in water from only such predefined residues without any need to ever touch any parameters.

If you tell us, what compounds and phenomena your simulations should represent, we can give you a better advice on the efford that may be required to describe your system.

Dear Norman Geist ,

I want to simulate NaCl crystal. For this I choose LAMMPS. For typical input file in lammps, I noticed that there I have to define atom style, lattice style, and so many things. From that, some questions arise in my mind that, "is there any prerequisite about the atoms to simulate".

I also notice that the force field(potential) vary from bond to bond. So is there any way to find out which potential file will suitable for different bond style?

Computational chemistry (either QM calculations or MD simulations) does not depend on a black box process. If we are talking about available force fields and user friendly softwares Dear Mr. Geist's answer is OK. But if you try to create your own force field or try to figure out what happens behind the black box computational process you need to know about the parameters that i have mentioned above.

For NaCl in MD, you should only need charges and LJ parameters. There are several parameter sets for those ions with sometimes refined properties on solubility and clustering. Away from that, to create a periodically continuing crystal along all or some periodic box vectors, these box vectors have to follow the respective crystal parameters, such as angles and side lengths. E.g. the box side length needs to be a multiple of the unit cell.

In QM as already pointed out, you just need to describe the elements themselves, so the wavefunction. Often only the valence electrons are used and a core potential mimics the atom nucleus and the remaining electrons. Additionally, the same care needs to be taken when setting up the periodic box, so it fits the crystal lattice in angles and side lengths.

To make it simple, you need two parameters.

1. Initial position of the atoms(from NaClcrystal structure): NaCl structure is well known and easy to build it lammps.

2. Inter-atomic force, Try to find it from here(https://www.ctcms.nist.gov/potentials/)