yes, a convenient way is to do a scan over the sigma values (e.g. 0.05 - 0.5) by a set of static calculations, then you can compare the differences between the free energy and the total energy (entropy term TS), this difference should be small (e.g. 1meV/atom depending on the accuracy you are going for), typically a larger sigma value results in faster convergence, so usually a compromise between calculation time and accuracy has to be used

you can also have a look at: http://cms.mpi.univie.ac.at/vasp/vasp/Number_k_points_method_smearing.html

Oh, I see. Thank you for the reply. But what if you are studying a hypothetical system and you don't know yet what type (whether insulating/semiconducting/metallic) it will be? Or what if you have a slab with a layers of metal and insulator? (e.g. a metal on an insulator substrate)? Thanks again.

I see. Thank you very much! Your answers were big help!

A good starting point, if you are unsure what smearing to use for relaxations, is usually Gaussian smearing (ismear 0) with a small sigma value (0.05, 0.1), the sigma value should be tested and in general Gaussian smearing is bit more tricky than Methfessel-Paxton (MP) (see vasp guide, and if your system shows metallic behaviour you should use MP). You can then also use other smearings (MP, -5,...) to see if any significant  effects occur (structure, DOS,...)

For static calculations to obtain accurate total energies you should always use the tetrahedron method with Blöchl corrections (ismear -5) (as long as the number of k points you are using is sufficient for this method) for all materials. Of course, you have to start from the supposed ground state structure.

Ismear -5 can get the forces wrong in metals which is problematic in relaxations.  Instead of relaxtions you can also do a series of static calculations with variations in the structural parameters, of course, this can mean the need for a huge number of static calc to scan all possibilities with sufficient accuracy

I would like to ask, in the same regard, what should I do with SIGMA if I am doing microcanonical (NVE) molecular dynamics? An initial temperature fixes the energy, but the temperature reduces to a half usually (equipartition of energy), so the thumb rule SIGMA=kB*T of canonical ensemble simulations (NVT) does not work, since I don't know the temperature the system will converge to. So how to choose? Should I change SIGMA every time I change the temperature (energy) of the ions system, or the same SIGMA for all temperatures?

Thanks for the reply, Gregor Feldbauer. By testing, do mean that I vary the sigma value, get the resulting energy/forces, and choose the one which would give a small energy or force difference within the desired tolerance? Thanks.

yes, a convenient way is to do a scan over the sigma values (e.g. 0.05 - 0.5) by a set of static calculations, then you can compare the differences between the free energy and the total energy (entropy term TS), this difference should be small (e.g. 1meV/atom depending on the accuracy you are going for), typically a larger sigma value results in faster convergence, so usually a compromise between calculation time and accuracy has to be used

you can also have a look at: http://cms.mpi.univie.ac.at/vasp/vasp/Number_k_points_method_smearing.html

Thank you very much for your answers.

Electrical insulators as a new phase state of quantum matter with some bulk gap and some odd number of Dirac formations have recently been entangled with a two-dimensional model by us through an external magnetic field. The magnetic field used has its own impurities so much so that the eccentricity resulting from the turning point of magnetic critical insulation leads to aspects of compilation that have never been revealed by the “approximate approximation calculus” mostly used in a run under such circumstances.

Crystal and electronic structures have given us the topology that have made it likely for us to depend more on over-stuck service rather than on projections that go into manifolds and submanifolds of various dimensions. In additions, fulton heaters in this specific setting of complex systems where technique does not exist for efficient approximation, has led us to find even lower than submanifolds: something like up-submanifolds and down-submanifolds whereby the cloesed set of combining topological insulators itself has become the target in insulating the whole bulk without any regard for the gapless and/or edge services states. This means that the topological insulators that we are working on have been both gapless at surface and without such states so that they might be able to carry a pure spin current. However, this pure spin current is the point where we have made our own initiative provided through the closure of those areas where points cannot satisfy every path from one origin-set of the insulator to another point of it simulacrum. Therefore, we are actually depending on, so to speak, closure models for creating metainsulators, giving yet more models of simulations of magnetic insulators.

The surface of state of spin resolved ARPES works so that the focus of one direction is in Riemannian contrast to the zones that consequently appear throughout the folding holes as borderlines between submanifolds. For this, without any doubt, means that the dimensions of submanifolds where we are working on our different froms have been entangled with ergodicity as the totality of anisotropy that rises from insulation through coupling strength. The simplest form is, of course, happening within a two-dimensional partially anisotropic plane. In the most complex forms, they can even go into the quantum mechanic levels such as N-Hilbert dimension.

Roya Attarzadeh

Don Senior Prof. Reza Sanaye