Some Background
As discussed in the lectures Molecular Dynamics is a technique for allowing a system to evolve in time according to Newton's second law F=ma. The atoms simply follow the trajectories that they would in reality and we can then compute properties as time averages of their behaviour.
MD is implemented as follows,
First of all an initial configuration and initial velocities need to be assigned - here the initial configuration will be that of ideal MgO and the velocities will be random but scaled to produce roughly the target temperature. Then ...
- Compute the forces on the atoms (F).
- Compute the accelerations a=F/m
- Update the velocities: Vnew = Vold + a * dt
- Update the positions of the atoms: Rnew = Rold + Vnew * dt
- Repeat until average properties like E and T settle down
- Once settled measure some properties.
A crucial parameter here is the timestep dt - this needs to be long enough for the calculation to be efficient but short enough for all of the possible vibrations of the atoms to be sampled reasonably well - typically you want about 10 steps along any given vibration. For MgO a time step of 1 femtosecond (10-15 sec.) seems to be adequate.
The other key issue is the size of the system to be simulated. In the previous investigations you have seen that MgO can be described by a primitive cell containing a single MgO unit (the assymetric unit). However, running MD on this cell would be meaningless - every cell of the crystal would be moving perfectly in phase which is not physical. In fact, using a single unit cell is just the same as sampling the phonons at just the k=(0,0,0) point - this is the point at which all cells are in phase.
If we doubled the cell along all three cell vectors to obtain a 2x2x2 cell we would have a cell containing 8 MgO units within which all of the vibrations sampled on a 2x2x2 k-point grid can be represented. The choice of cell size for MD can be established empirically simply by running larger and larger cells but MD is much more expensive than using the quasi-harmonic approximation so that is not practical in this laboratory. As a compromise between accuracy and efficiency a cell containing 32 MgO units will be used - this has already been created and is stored in MgO_32.str.
Note: The 32 unit MgO cell describes exactly the same static system as the primitive cell - it simply allows more flexibility when the system is vibrating.
Running MD Simulations
Load the MgO_32.str structure - check that it is simply the same MgO crystal but represented in a larger cell (eg: check Mg-O bond distances by selecting neighbouring Mg and O ions and checking the report in the main window).
Open the Execute Gulp panel and select Molecular Dynamics.
Open the MD Opts panel and edit the MD options...
Set ensemble to NPT - in this simulation we fix the number of particles (N), the external pressure (P) and the temperature (T) while allowing the volume of the system (V) to vary.
Set the Temperature - initially use 300 K
Set the Time Step = 1.0 femtoseconds
Set the number of Equilibration steps to 500.
Production Steps = 500 ; this is the number of steps run after equilibration is complete.
Set Sampling steps and Trajectory write steps to 5. These are, the number of time steps over which averages are made and, the number of steps betweem each writing of the geometry to a trajectory file for animations, respectively.
Make sure that you have selected ionic.lib for the potential model, that the Include Shells option is NOT selected, and that the DOS calculations are turned off under general options. Click Run to start the calculation.
The results
You can monitor the progress of the job by selecting it in the job list and recovering files whenever it says that the output file has changed.
Take a look at the Logfile. You will see the start of the MD equilibration when properties begin to be reported at each sampled time step (every 5 in the current job).
At each timestep the instantaneous and Averaged values of a number of properties are reported. You will see that the cell structure and its volumefluctuate from timestep to timestep but as the system settles down (equilibrates) the averaged value becomes more stable. A good sign is that the temperature becomes reasonably stable at around 300 K - if it doesn't then the equations of motion are not being integrated accurately and a smaller timestep is needed.
This is a relatively small cell and so fluctuations are quite large however after the system has equilibrated you should be able to identify an average cell volume (plus or minus a bit) for the temperature.
You can animate the movements of the ions for the production steps.
Display -> Animate -> Sequence
Note: you can use the Display->structure to alter how the structure looks while it is being animated.
Note: The MD runs take a look time to complete. You can get together with other groups to run different temperatures - machines can be left running overnight (put a note on the keyboard to ensure that the machine isn't rebooted).
Questions
- Replot your data from the quasi-harmonic approximation calculations as, cell volume per formula unit vs temperature, and add some points from the MD runs at a few suitable temperatures.
- How does the thermal expansion predicted by MD compare to that predicted by the quasi-harmonic approximation ?
- Why do the two methods produce different answers ? - how does the difference depend on temperature ?
An opportunity to speculate
- From your experience of computing an accurate Free Energy how large a cell should be used to perform reliably MD for MgO ?
- What would happen to the cell volume at high temperature in the quasi-harmonic approximation and in the MD ?