While there are ample resources online that detail it, to cut the long story short, BOMD is molecular dynamics of particles, in which dynamics of electrons is treated at the quantum mechanics level of theory, while the ions in the system are treated using the classical (newtonian/lagrangian) formalism.

Initial details of this baseline explanation can be followed here:

http://en.wikipedia.org/wiki/Born%E2%80%93Oppenheimer_approximation

Basically, what people usually mean by BOMD is to run a classical molecular dynamics but the energy and the forces are computed at the quantum level using any electronic structure method (ab initio, DFT, etc.). To test such method, you can use for example the GNU part of Amber 14 (AmberTools) which contains in the sander program a "QM/MM" option (=if you do a QM/MM MD without an MM part, it's BOMD!) up to a dedicated SEBOMD (SemiEmpirical Born-Oppenheimer Molecular Dynamics) part.

www.ambermd.org

Gerald Monard Sir, Can the stability of a molecule be checked in either ADMP/BOMD methods implimented in gaussian? I have run a ADMP simulation of 1ps with optimized formaldehyde only to get the periodic dissociations and re-associations of structure. The molecule should have been stable. I have got similar results for benzene molecule also.

Stability of a molecule can be decided upon its optimization, followed by vibrational frequency analysis. So, in order to check the stability, you will have to perform vibrational analysis after optimization, and not molecular dynamics.

You should not see a simple molecule like benzene or formaldehyde to dissociate during BOMD. Check temperature, it should not increase too much during the MD. Check also the timestep that you use, it should be small (like 0.1-0.5fs), not more. I have no specific experience using the ADMP/BOMD implementation in Gaussian, but these simple tests should help you.

You can run BOMD also using Molpro (see molpro.net). There you'll find some simple examples.

As was already said here: The idea of BOMD is to use classical mechanics for the time evolution of coordinates and momenta of the nuclei, while the electronic problem is solved by means of quantum mechanics, within Born-Oppenheimer Approximation. 

Time step for numerical integration, also mentioned before by G. Monard, must be small enough to warrant energy conservation and large enough to allow you to get what are you looking for in your simulation.

Once the dynamics is classical, you'll also need to select initial conditions for your molecular system.

Dear Sk Jahiruddin,

I've tried too the ADMP / BOMD methods implemented in Gaussian09, and the molecule i tried to simulate dissociated unphysically during the ADMP run.

Anyway, during the BOMD run everything went as expected, and I was also able to recover a correct vibrational density of states (VDoS).

Also, fixing the time-step at an exact value can give issues.

So I think the BOMD routines can be safely used in production run, as long as one check the timestep and the temperature.

Francesco Muniz-Miranda,

Thank you very much for the answer. I have been experimenting with different molecules with both ADMP and BOMD since last few days and I have reached exactly same conclusion.

You could try the software CP2K. It's friendly to user and combine Gaussian and plane wave method. The website is www.cp2k.org

J. Y. Yang, Thank you. I'm also trying to learn CP2K.