How do you freeze the crystal?

I put ionic constraints on the atoms.

You have run only 1.4 ps, which is probably too small for the system to relax. especially when you have a lattice. Can you not run longer?

You need to run for a longer time period. Additionally, the increment in total energy is only 14 ev, which is not very significant.

I have a few questions to help clarify the issue:

- Is this a plot of the total energy, or the conserved quantity?

- How many atoms do you have? This will help us to determine the scale of the energy drift (some energy drift is inevitable)

- How did you determine your time-step?

- Have you already subtracted the equilibration time, or is this still equilibrating?

I would also try without fixing the atoms, in case the constraints are causing issues. Temperature is defined in terms of the distribution of kinetic energy amongst the degrees of freedom, so changing the degrees of freedom using constraints will change the dynamics.

All the best,

Phil Hasnip

(CASTEP Developer)

Thanks for your reply. @Philia James Hasnip

1. The plot is for the total Energy.

2. The unit cell has 1038 atoms including hydrogens.

3. I did not determine the time step so far. I used the same time step (1 fs) without any constraints. In that case, I did not see this kind of energy drift. I have got very nice NMR parameters (from CASTEP) for the complex, where I have used the X-ray diffraction structure and optimised the positions of all atoms. However, the anisotropy of 13C for the acetone molecules are quite different from the experiments. This is most likely due to dynamics of acetone molecules. Therefore, I thought of freezing the crystal accept for the acetone molecules.

4. I did not subtract the equilibrium time, the system is still equilibrating.

Update:

Thanks your earlier suggestions. I have continued the NVT MD simulations, from 1.3 ps to 5.4 ps now. The system seems be equilibrating now (see Fig. Constant of motion). However, the temperature is fluctuating between 210K and 390 K (See Fig).

1. I want to calculate NMR chemical shift for this system at 300K. Should I first run NVE MD for that?

2. Is the temperature fluctuation is normal in NVT simulation?

Yes, temperature is a statistical property and should fluctuate. For N atoms in your simulation, the size of the fluctuations, dT, should obey:

dT/T ~= 2/(3 sqrt(N))