[Pw_forum] Constrained magnetization calculations on doped cell will not converge

[Jeffrey Mullen]

Greetings all,

I am attempting to run a supercell of MoS2 with a single Co atom where I
constrain the magnetization of the dopant atom in the cell. After following
the advice in several messages in the forum, I think I have the correct
configuration of input variables for what I want to do. I want to confine
the direction to be normal to the plane of the MoS2 layer, which is the
"angle1" variable. I have the following input in the system to accomplish
this

&SYSTEM
    lspinorb=.true.,
    noncolin = .true.,
    angle1(3)=0,
    constrained_magnetization='atomic direction',
    ibrav = 4,
    celldm(1) = 24.12,
    celldm(3) = 10,
    nat = 48,
    ntyp = 3,
 /

I want the net magnetic moment to be aligned along +z for the dopant atom
(ntype = 3) and, in a separate calculation, to be aligned along -z. I have
attempted calculations with and without starting_magnetization because I am
unsure if starting_magnetization, with constrained_magnetization='atomic
direction', is a value which the code attempts to constrain. I do not
believe so, but I cannot get the system to converge to view any results.
Among the configurations I have tried, all from previous emails on the
topic in the forum, are

1) denser k grid (than required for relaxation convergence) for the
constrained calculation

2) setting the starting_magnetization=1, but maintaining the
constrained_magnetization='atomic direction'

3) fixing constrained_magnetization='atomic direction' for all of the other
ntypes in order to limit the variability of the other atoms in the run (I
am sure this is terrible physically, but I am testing the scf convergence
trends)

4) setting LAMBDA to a wide array of values (1, 5, 10, 50)

The convergence threshold on my calculations is relaxed significantly in
order to decrease my computation time while testing.

In summary, the configurations never converge and a typical "estimated scf
accuracy" check from iteration to iteration looks something like
     estimated scf accuracy    <    9195.18118944 Ry
     estimated scf accuracy    <    9595.61384890 Ry
     estimated scf accuracy    <   10668.19123448 Ry
     estimated scf accuracy    <    8674.33184988 Ry
     estimated scf accuracy    <   10284.22871137 Ry
     estimated scf accuracy    <    7712.40566666 Ry
     estimated scf accuracy    <    8482.35738594 Ry
     estimated scf accuracy    <    9385.66444081 Ry
     estimated scf accuracy    <    8720.98552610 Ry
     estimated scf accuracy    <   10700.22801358 Ry
     estimated scf accuracy    <   11622.15976529 Ry
     estimated scf accuracy    <   10474.22752741 Ry
     estimated scf accuracy    <   12212.06974477 Ry
     estimated scf accuracy    <   10317.40668337 Ry
     estimated scf accuracy    <   11908.95640393 Ry
     estimated scf accuracy    <   11244.08847431 Ry
     estimated scf accuracy    <   10030.78189283 Ry
     estimated scf accuracy    <   11404.75835493 Ry
     estimated scf accuracy    <   10166.54181955 Ry
     estimated scf accuracy    <   11266.67685047 Ry
     estimated scf accuracy    <    9588.30548883 Ry
     estimated scf accuracy    <   11971.16137165 Ry
     estimated scf accuracy    <   10109.23777213 Ry
     estimated scf accuracy    <   10332.43063917 Ry
     estimated scf accuracy    <   10149.87556671 Ry
     estimated scf accuracy    <    9348.01440051 Ry
     estimated scf accuracy    <   13629.68937599 Ry
     estimated scf accuracy    <    9074.81346315 Ry
     estimated scf accuracy    <    9378.08640942 Ry
     estimated scf accuracy    <   10010.57416910 Ry
     estimated scf accuracy    <    8566.82065492 Ry
     estimated scf accuracy    <   12137.47227188 Ry

Can anyone correct my understanding, or point me towards a path for
convergence ?

Cheers
Jeff Mullen, PhD
North Carolina State University
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20171108/fe8cf43f/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mos2.in
Type: application/octet-stream
Size: 3714 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20171108/fe8cf43f/attachment.obj>