Phase Transition in Materials

[Juan]

Dear all,

I'm studying the BaTiO3 material. Right now, I'm trying to replicate the phase transition from tetragonal (room temperature) to cubic (aprox. 393 K). Using MD, I'm using the ATK-Classical with the parameter set Pedone_2006Fe3 (it also provides the F2). The additional info says that the Brenner and EMT calculators don't have parameters for Barium (same additional info for all the parameters set). I'm using the NVT Berendsen type of MD with the recommended parameters in the tutorial. My questions are:

-If the parameter set I'm using does not have information on the parameters of one of the atoms in my structure, will this affect the results on the calculations? (Lets say Total Energy)

-Is it adequate to use the ATK-DFT calculator using molecular dynamics? I tried using it (just putting the calculator and leaving it with default parameters) but the program was still running after five days! I know some calculations are long but I am not sure if this combination can have something to do with this prolonged time(I'm using a fast machine for this type of calculations). I ran the calculation with the ATK-Classical and it not took so much time.

-I know if there is a change in total energy, there may be a phase change. I would like to know how can I make a graph of TotalEnergy(Temperature). I looked for this in the forum but the only things I found were the ones you provide in the MD tutorial (V(t) and P(t) for the NPT Melchionna) and for the Langevin type, you use the following script:

# Get the kinetic energies
kinetic_energies = md_trajectory.kineticEnergies()
# The number of constrained atoms
number_of_constraints = 10
# Calculate the degrees of freedom in the system
degrees_of_freedom = bulk_configuration.numberOfAtoms() - number_of_constraints
# Calculate the temperatures from the kinetic energies
temperatures = [ke/(1.5*degrees_of_freedom)/boltzmann_constant for ke in kinetic_energies]

I don't know if this can be used for my purpose.

I eagerly wait for your response.

Regards
Juan

[kstokbro]

For the Classical potential you need to check if it was developed for your system,  else you will need to use DFT. You can use ATK-Classical to find an appropriate time-step and check other parameters of the MD simulation, I recommend the tutorial,
http://www.quantumwise.com/publications/tutorials/item/732-basic-molecular-dynamics-tutorial

With the DFT you should save the trajectory and then you can follow the simulation by inspecting the trajectory.
To see a 1. order phase transition, there must be a discontinuity in the potential energy as a function of temperature.  Thus,  you can forinstance slowly heat up the system, and if there is no phase transition both the temperature and potential energy will increase slowly, upon the phase transition temperature will be constant and the potential energy will change suddently.

[Julian Schneider]

Some additional remarks:

If you want to see the phase transition in an MD simulation you need to use a constant pressure simulation, in this case NPT Melchionna. Using NVT you will never see the phase transformation because the cell size won't change. Currently, NPT does not support increasing the temperature during the simulation, so you have to run a new simulation a each temperature point.
A good general example how phase transformations can be simulated using MD can be found in this paper:
http://dx.doi.org/10.1063/1.2038747.

Before you run the simulation you should check if your calculator/parameter set can reasonably well reproduce all phases that you expect to see, i.e. that they are reasonably stable under a geometry and cell optimization. If not, you cannot expect to encounter this structure in a dynamical simulation.

To check the total energy, the simplest way is to view your MDTrajectory in the Movie Tool, where all energy contributions, as well as the temperature are automatically plotted.