Dear all,
I did some test about the magnetization of fcc bulk Pd and got very strange results.
Details refer to another post in VASP forum.
http://cms.mpi.univie.ac.at/vasp-forum/viewtopic.php?f=4&t=17553
Any comments are highly appreciated.
Best,
Binglun
Hi Binglun,
This forum is a place for Materials Project questions regarding our data or methodology. Please keep questions regarding VASP usage in your own research to the VASP forum.
Dear Shyamd,
Thanks for your reply.
It is a kind of problem of methodology, since the result of Pd published on the website is affected by your VASP settings.
According to your current result, Pd is non-magnetic. But if you increase the density of k-mesh, you will get a magnetic result.
Does this mean that your result of Pd is not converged?
Best,
Binglun
Hi Binglun,
Fundamentally, no property is guaranteed to converge monotonically with K-points when using DFT. The magnetic description in DFT is also not the best for magnetic properties, e.g. its a scalar description and not a vector description.
One can ALWAYS do a higher order calculation and claim it is more accurate. Our calculation methodology is heavily benchmarked to experiment where available. These are not a substitute for careful and detailed calculations if that is your goal, but a series of computed properties that do as well as experiment if not a bit better in general.
As for Pd itself, it’s entirely possible that there is a magnetic ground state and it may be a simple ordering or it may be more complex. We use a set of parameters that are benchmarked for reaction energies, and while we don’t capture every little nuance we do a good job across the board in finding appropriate ground states. We’re currently in the process of developing a magnetic ordering workflow that is better at this.
To add to @shyamd’s comments, the magnetic moment you’re reporting for Pd is very small compared to other magnetic materials. If it’s close to the Stoner transition, it’s not only going to be more difficult to converge, but it’ll also be very sensitive to on-site Coulomb interactions. Therefore, the fact you get a different result from PBEsol for example is not at all surprising.
As a general tip, make sure you’re setting your initial magnetic moments (‘MAGMOM’ tag) to sensible values and that you’re reading your total cell magnetization rather than atom-projected magnetization too, since these can sometimes vary.
Does this mean that your result of Pd is not converged?
The question is if it’s sufficiently converged for a particular property or application. It’s possible it’s not sufficiently converged to distinguish between a small-but-non-zero magnetic moment and a zero magnetic moment, but it should be sufficiently converged for total energy and lattice constants for example.