Exercise 4: Determination of Dynamical Properties
This exercise is a collection of some exaples on how to determine and tune dynamical properties in CPMD.
Calculation of Vibrational Spectra
Geometry Optimization: 1-h2o-opt.inp
Vibrational Frequencies via Finite Differences: 2-h2o-vib.inp
Fake Gaussian output for visualization with Molden/Molekel: VIB1.log
Vibrational Frequencies via Perturbation Theory: 3-h2o-pert.inp
.. to be completed
The Dragging Effect
We now want to to have a look at the consequences of the 'dragging effect' of the fictitious dynamics during a CP-MD run. For this we look at 3 pre-calculated trajectories (1-2ps) of a single isolated water molecule:
a Born-Oppenheimer MD run with a time step of 10 a.u.
Since there is no 'dragging' in BO-MD, this is the reference run (3-h2o-bomd-run.inp).
a Car-Parrinello MD run with a time step of 4 a.u. and a fictitious mass of 400 a.u. (3-h2o-cp-400au-run.inp)
a Car-Parrinello MD run with a time step of 2 a.u. and a fictitious mass of 200 a.u. (3-h2o-cp-200au-run.inp)
You can look up the inputs and outputs of the simulation in the reference data section. If you look at the evolution of the various energies during the simulation (ENERGIES-bomd, ENERGIES-cp-200au, ENERGIES-cp-400au), you get the impression, that all trajectories seem to cover almost the same phase space, if the initial kinetic energy added to the system takes into account the extra amount needed for the fictitious dynamics of the electronic system (which has been determined empirically in this case).
During the simulation also an analysis of the current dipole moment was performed and recorded (DIPOLE-bomd, DIPOLE-cp-200au, DIPOLE-cp-400au). These file can be used to calculate the infrared spectral densities using the provided fourier transform program . This code also adds several (optional) corrections to the spectra. It is recommended to look at the data in the first and the fifth column. You can see, that there is a noticeable red shift for some peaks in the spectral densities. The shift itself depends on the fictitious mass and on the mode (which makes it a bit tricky to compensate for it).