[Pw_forum] Technique for converging Ecut and K-points?

[Nicola Marzari]

On 28/02/2013 00:56, Ben Palmer wrote:
> May I ask how to converge forces with PWscf?  I've tried adding
> tprnfor=.TRUE., to my input file, but for scf, relax and vc-relax it's
> printing zero forces for all.

If you have a unit cell with 1 atom, forces will always be zero by symmetry.

Construct a simple cubic unit cell with 4 atoms, displace one by a
significant amount (15% of the al-al distance), and monitor force
as a function of smearing, for different k-points meshes. At high
smearing, all k-point meshes will give the same results (high smearing
make the integration easy, but you are calculated the "wrong" "hot"
properties). At small smearing, you need a lot of k-points to converge
(to the "right", "cold" result). At intermediate smearing, you might
hit the sweet spot - you have a smearing temperature that is not too
hot (so the infinite k-sampling result would be a good approx to the
zero temperature infinite k-smapling result), but hot enough that
you do not need infinite k-points to converge.

Try energies and forces with regular smearings (gaussian, fermi-dirac),
or with "cold" ones (methfessel-paxton, marzari-vanderbilt).

				nicola



----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL