It is a very good question - I appreciate. Please try to understand the meaning of the word ab-initio. I think that will help.

Hi Shilendra,

Sometimes, since in MD simulations you are interested in statistics, the level of theory can be lower than in very accurate optimizations and you can use Molecular Mechanics. This is very common in biological systems, for example.

If you carry out "Ab Initio Molecular Dynamics" or AIMD (which is also "molecular dynamics") you calculate the potential energy surface at a quantum level in each step.The computational effort is huge (although the initial guess for the electronic density of each step can be taken from the previous one) and that is why AIMD is used only in small systems and time scales (until few ps)

So, the results are not the same if you are using different levels of theory, of course, but if your force field describe correctly your system, the results should be similar to the AIMD ones.

And your last question: it depends on what are you interested in, the size of the system and the time scale. For example, to study the reaction of H2+O2, I would use AIMD, whereas to explore the conformational space of a protein or see the evolution until the ns scale, MD with a force field is probably the best option.

Hope it helps,

Fernando

Dear Shilendra,

The difference between ab initio and standard molecular dynamics (MD) lies in the way the interatomic forces are calculated. In both methods, however, the motions of the atoms are computed by applying Newton's second law to the atomic coordinates, i.e., by treating them classically. I guess this is where the "MD" part of the name comes from.

In the standard MD approach, the instantaneous force on each atom is calculated as a gradient of a prescribed interatomic potential function (or, in other words, force field), which is generally a function of the atomic coordinates that you can equally well regard as the coordinates of their nuclei. In principle, the potential could involve many-particle terms, but many typical potential models involve terms only up to three-particle ones.

In the ab initio MD, instead of using a prescribed potential, one solves the interatomic forces at a given time instant as follows. From a quantum-mechanical perspective, the system at a fixed time can be parametrized in terms of the coordinates of the nuclei and the relevant electrons. By invoking the Born-Oppenheimer approximation, one can regard the nuclei fixed at the instantaneous positions of the atoms. This way, one can write a time-independent Schrödinger equation for the many-body wave function of the electrons. This Schrödinger equation is then solved using (time-independent) DFT to obtain the energy. The energy is then considered to be a function of the nuclear coordinates that were fixed earlier, and it can thus act as the interatomic potential that is needed to compute the forces in Newton's equation of motion for the nuclei. So, by computing the gradients of the DFT energy at this fixed point with respect to the nuclear coordinates, the forces are obtained and the nuclei are moved accordingly to get to the next time step. The DFT process is then repeated with these new nuclear coordinates.

Quite obviously, these two approaches differ significantly in terms of the computational costs, and thus the ab initio MD can only treat small system sizes. On the other hand, it enables calculating dynamics for systems for which the forces have not been parametrized in terms of force fields. The ab initio MD can also treat effects such as bonding, charge transfer, polarization, and many-body effects intrinsically, while in standard MD they usually must be artificially imposed. The choice of the method for a specific case should be made based on these factors.

The two methods are not likely to give identical results, since already standard MD results depend on the interatomic potentials chosen. Of course, one would hope that despite these differences, there would at least be satisfactory qualitative agreement so that meaningful conclusions can be drawn regardless of what specific model is used.

I hope my answer helped at least a little!

Regards,

Pekko

In classical MD force between the atoms / particles are calculated from empirical potential like LJ, Morse etc. then by numerical integration position and velocity of the particles can be calculated.

whereas in abinitio MD force is calculated from energy of the electrons(density ) by DFT technique and assuming nucleus is fixed. the by integration one can find the position and velocity of the particles can be calculated.

Dear fellow researchers,

I have been reading the answers above regarding the difference between abinitio and classical MD. Anyone has experience in calculating the defect formation energies using abinitio MD implemented in any DFT code, such as VASP? I understand that we can calculate the ground state defect formation energy using DFT codes. But anyone attempted it using abinitio MD ??

thanks in advance,

Pekko Kuopanportti gave a good answer to the first part of your question. Coming to the second part, "When should do ab-initio MD and when classical MD", It depends on the accuracy you want and the kind of problem you want to solve. I would like to make the following points:

1) The length-scale and time-scale of interest is one main point to consider. For example a collision cascade by a 10 keV primary knock-on atom in a metal would be impossible to simulate using ab-initio MD with the currently available super-computers, but can be simulated on a laptop using classical MD. Therefore, any dynamics problem involving say more than 10000 atoms and having time-scales of more than a pico-second would be difficult to simulate using ab-initio MD. Using classical MD several 100 million atom scan be simulated for several nano-seconds with the best available supercomputers. On a laptop one can simulate around a million atoms for several 100 pico-seconds in a day.

2) If you do not have a valid interatomic-potential for the atomic species you are interested in simulating, you cannot carry out classical MD simulations. Several researchers would then use several thousands of ab-initio simulations and construct a inter-atomic potential for classical MD simulations using tools like potfit (https://www.potfit.net/wiki/doku.php).

I hope this helps ...

Molecular Dynamics deals more with classical mechanics whiles Abi initio deals uses quantum mechanics.