I am trying to find dielectric constant using the mentioned method (10.48550/arXiv.2211.15571). They implement AIMD in VASP. Actually, I am familier with Quantum Espresso (not VASP) and using QE, I have plotted dielectric constant for 0k. But to include temperature effect I need to apply MD to get structural configuration. As I didn't work in MD, I found very hard to understand the following process from that paper.

"Ab initio molecular dynamics are performed on a supercell structure consisting of 192 atoms constructed by 4 × 4 × 4 CeO2 primitive cells. For this size of supercell, only Γ point is computed to accelerate the calculation. After reaching thermal equilibrium under NPT ensemble (zero external pressure) with Langevin thermostat, we use 1000 more steps to get averaged lattice structure at each temperature. An illustration of our temperature control in NPT ensemble is shown in Fig. S1. Then, on relaxed structure we perform NVT ensemble simulations and after reaching equilibrium we use 2000 more steps to construct effective force constants [5] at a time step of 2 fs. Harmonic force constants (HFCs)consider a cutoff radius of 6.31 ˚A, the third-order force constants have 6 ˚A and the fourth-order force constants have 4 ˚A as cutoff. Optical properties of 4 × 4 × 4 CeO2 supercell are computed in a DPFT manner. We use Γ-centered k-grid of 2 × 2 × 2 and 2560 KS orbitals. This large number of empty bands for transitions is tested for convergence."

Is it possible to perform it in using an MD package in QE (instead of AIMD or VASP)?

  • 1
    $\begingroup$ The process involves two main parts: calculating dielectric constants using phonons for a given supercell, and determining the correct supercell dimensions at finite temperatures using molecular dynamics (MD). The term AIMD refers to performing molecular dynamics simulations with density functional theory (DFT) codes. Quantum ESPRESSO (QE) is capable of conducting MD simulations. In this context, MD simulations are performed to determine the volume of the supercell at a given temperature using the NPT ensemble. $\endgroup$ Commented May 22 at 7:29
  • 1
    $\begingroup$ Once the volume is known, one can perform density functional perturbation theory (DFPT) or finite displacement methods to calculate the force constants. These force constants are then used to calculate phonons and dielectric constants $\endgroup$ Commented May 22 at 7:30
  • $\begingroup$ @Pranavkumar Are you referring CPMD of QE to perform MD simulation? So normally, to get the force constant in QE, We need to use pw.x and ph.x. To run pw.x, we use atomic positions of atoms in crystal. But, once the supercell volume is known at a given temperature from MD simulation, How do we prepare scf.in (pw.x)? As I still don't know the atomic positions which are require to prepare the scf.in. $\endgroup$ Commented May 22 at 14:16
  • 1
    $\begingroup$ Yes, CPMD. we can use same relative lattice position (crystal not angstrom) as were at 0K remapped into new supercell dimesnsion at finite T and perform atomic relaxation once. $\endgroup$ Commented May 22 at 15:10
  • $\begingroup$ @Pranavkumar Does that mean that I have to use ibrav 0 (always) and lattice vectors have to be provided in card CELL_PARAMETERS? Also, I keep same ATOMIC_POSITIONS (crystal) all of the runs? $\endgroup$ Commented May 22 at 15:23


You must log in to answer this question.

Browse other questions tagged .