Will it be sufficient to calculate the properties of a material with the atoms in the unit cell only, like in the attached, or do I have to make a super-cell? what difference does it have?
That depends on what you are looking for and what the material in question is. Quantum Espresso uses PAW potentials which are periodic, so setting the unit cell size also has a big effect on the results, along with the energy cutoff. Look at what your material phase and then optimize the energy cutoff and slowly increase unit cell size until the results are self-consistent.
In DFT, the output observables, i.e. total energy and properties, are calculated for the input structure that is given.
That being said, the researcher is responsible for verifying that the input used corresponds to what you want to study.
When you use a unit cell, this implies that you are modeling a material that is an infinite repetition of this given structure.
In certain cases, like for ground state properties of bulk solids, this is exactly what you want; in other cases, if you want to study defects, impurities, surfaces, symmetry breaking, etc, it is not.
Therefore, the answer is given in two steps: first you define your goal, what property you will study, and then secondly you certify that both the structure and the technical parameters you use are consistent with this goal.
Thank you Dallon Penney Shibghatullah Muhammady , Gabriel Ravanhani Schleder , and Omar Al-Obaidi ,
Then I have more doubts: What about the case of layered 2D materials, like graphene/BN/graphene layer, (cif file attached), Will it be okay to use the unit cell only in calculating its properties?
Second, will QE see three layers only or an infinite number of layers in the c direction? The same case to the hBN (cif file attached); does QE calculate properties for 1 layer BN or an infinite number of layers?
I belive you are asking this question due to the constrains in your computer architechture such as number of nodes, GPU accelarated or not, available memory etc ...
However, there is no short cut for this and as others mentioned you have to test your system for k-points, lattice parameters, plane wave energy cut off etc with respect to the ground state energy and other properties that you are interested of. Hope this may help.
Sylvester Makumi
Elaborating on my previous answer, the input structure to be given have to consider your study goal. This precedes the technical aspects of the simulation.
However, when we consider DFT implementations based on plane-waves and periodic boundary conditions (PBC) on 3 dimensions such as Quantum Espresso (in contrast to atom-centered basis sets and no- or reduced- boundary conditions), this is relevant. In these cases, you always calculate infinite periodic repetition of your unit cell, so if that is not your goal, you have to circumvent this.
For a more in-depth discussion regarding DFT implementations of different basis-sets for wave function expansion, I suggest our recent review article:
Article From DFT to Machine Learning: recent approaches to Materials...
Therefore, in practice, when you have PBC in the direction where you want a surface, you must include a region of vacuum. In this way, you model material slabs separated virtually by this vacuum added. From the technical aspect, you have to converge how much vacuum you will need so that you don't have artificial interactions between periodic images. This will usually be between 10 to 25 Angstroms for most properties.
In your files, you will have infinite repetitions of hBN and graphene.
For a graphene/BN/graphene configuration, I have modified your file to a reasonable initial input structure. Note that this is a starting point, you will have to converge the DFT calculation parameters first (many tutorials are available for different software), and then obtain the ground state geometry for this system, which includes lattice parameters and strain on which sub-system, and inter-layer separation, for which you will have to account for van der Waals interactions (different methods are available in different software).
Thank you Gabriel Ravanhani Schleder , the information is exactly what I wanted to know.
Sylvester Makumi You are welcome, I hope my answers are useful and you keep advancing in this area.
If I use 'assume_isolated='2D', will it have the same effect as adding the vacuum or should I use both?
Hi Sylvester,
It actually depends upon the properties you want to see. For example, if you are only interested in band structures, the density of states, mechanical properties, optical properties, and so on, the unit cell is enough for these calculations. Some calculations, however, including thermal properties calculation, phonon analysis through Phonopy, etc need the supercell instead.
When I run vcrelax the vacuum reduce so much, to c = about 6Ang for single layer, should I fix the c lattice and ignore the calculated value?
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Atomic, Molecular and Optical Physics
- Atoms