Dear all,
In vasp, we can use either finite displacement method (IBRION 5/6) or DFTP (IBRION7/8) to perform vibrational spectroscopy.
I am aware that currently vdw correction cannot be applied with IBRION 7/8, but other than that, could you please leave your comments on how to choose between these two approaches for my system? Will the results of the two be system-dependent?
Thanks,
Azade
In general, both methods can give comparable accuracies but have their own advantages and disadvantages. In both methods, for calculating phonon modes, one needs to calculate the so-called Hessian of the potential energy or simply speaking the first-derivative of force on all atoms.
In finite-difference (FD) or frozen phonon method, this derivative is taken using the FD method. In other words, the program first calculates the force constants (F) for finite displacements of atoms (dx) and then approximates the dynamical matrix dF/dx as in any FD approach. This method's advantage is that one can implement it in any scheme, including DFT, GW, DMFT, hybrid-functionals, etc. The disadvantage is that it requires construction of multiple large supercells, and its accuracy is dependant on their sizes (the larger, the more accurate).
In DFPT, instead of directly calculating dF/dx, the program calculates the second derivative of energy with respect to such displacements. Within DFT, this can be treated using first-derivatives of charge density functionals and requires no supercell construction. Accordingly, the DFPT calculations are computationally much less demanding than the FD ones. As you may guess already, the disadvantage is it cannot work with any scheme other than DFT (as it is very difficult to treat any derivative of charge density non-locally).
Mohammad Saeed Bahramy Thanks a lot for your answer. Very clear and helpful.
Most of the calculations that can be performed by the VASP package can also be performed by other freely available packages such as Quantum Espresso (QE), abinit etc., However, the major difference comes in the computational time when large supercells involved. VASP usually does things at a faster rate than QE, abinit etc., and implementation is done at a faster rate in VASP than QE. However, much work has been done on QE in the last few years, such as adding GPU support and mixed OpenMP/MPI parallelism with good parallel scalability as described in (ref1) and (ref2).
It is not like VASP is easier to use than other packages. Interfaces are available even for QE as well such as PWgui, which can simplify the life of the user from preparing input files. Importantly, whatever the package it is, the user should be able to understand the corresponding source codes that are given in the package for DFT simulations. In DFT calculations, the important things are how efficiently a given package can yield reliable results in a short amount of time, and how efficiently the package can run the codes in parallel without affecting the nature of the results.
- Badges
- Science topic
- Similar topics
- Psychology