I'm a beginner when it comes to DFT and Quantum Espresso.

Consider I have a lattice which has associated magnetic properties. So for geometric optimization should I take into account the magnetic character while doing VC-relax in Quantum ESPRESSO. Or can I get reasonable good predictions for DOS and band-structure calculations if I do the VC-relax without taking the magnetic effects into consideration?

Apart for that if I am to do VC-relax with the magnetic effects taken into account (nspin = 2) what pseudo-potential should I use?

Thanks in Advance.


It is a bit difficult to answer this question, due to the information provided. If you are a beginner, using vc-relax will have its pros and cons.

With regards to using nspin=2:

Recall that when you use nspin=2, you are saying that you will define the initial magnetization for the involved species. So, just keep in mind that spin-polarized calculations will be more expensive than non-polarized...and if you perform fully-relativistic calculations then those will be more expensive than spin-polarized.

If magnetic properties are really important in your work and if there may be some non-collinear magnetism, spin-polarized would be a good start and then even compare with spin-orbit coupling as well. Be cautious about which atoms to consider for this.

With regards to vc-relax

Variable cell relaxation should be used with caution, and making sure that all parameters are properly optimized and after convergence tests have been performed. This depends on the symmetry of your system and how well you are treating the system to begin with.

With vc-relax, you control the degrees of freedom that your crystal will have. You can, for example, say that you only want the diagonal elements to change (a1, b2, c3) but preserve symmetry (cell-shape), or you can fix the volume and give the cell-shape freedom (if not careful, this will lead you AWAY from the global minimum!), you can also go "all out" by letting both the shape and volume change.

If careful, you can save as lot of time. If not careful, you'll find yourself spending more time trying to get it right and might end up "manually" optimizing a cell (analyze the system's energy as a function of Volume and then fit an equation of state).

Final remarks and things to think about as a beginner!

Is there experimental data available for the material you are modeling?

If there is available crystallographic data for that material, I would recommend keeping it simple at first. Start with the experimental structure, run a vc-relax that fixes symmetry but will vary volume. Do this without spin-polarization. If it is a bulk structure, adding a vdW correction won't do you any harm (e.g. Grimme-D3). Run this calculation and compare your computed unit cell with the experimental data. If all is well... you should be within 3%. If you think there may be some improvement, then include nspin=2... but I don't think it will be all that significant.

In my experience, for many materials optimizing with PBE-D3-SOC (SOC = Spin Orbit Coupling gives almost identical geometries to PBE-D3).

I would recommend trying to save time during optimization and save that time for the modeling of properties, with SP and SOC.


If the system is magnetic you need to use nspin=2. You can use any pseudopotential, unless you want to see SOC, in which case you need to use a fully relativistic pseudopotential.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.