[Pw_forum] question on constraint magnetic moment calculation

[Pang Rui]

Dear Gabriele Sclauzero:
I covered the output, so I gave a new one with the same problems.The input
file can be seen in the end. The only difference is the angle and
amplitude. The scf converged after 59 steps(In default conv_thr). The
output of the magnetic moment is as followed

==============================================================================
     atom number    1 relative position :    0.3750   0.0000   0.0000
     charge :    13.367288
     magnetization :          0.000122    0.000076    2.717489
     magnetization/charge:    0.000009    0.000006    0.203294
     polar coord.: r, theta, phi [deg] :     2.717489    0.003039  
31.837662
     constrained moment :     0.000000    0.000000    2.800000


==============================================================================


==============================================================================
     atom number    2 relative position :    0.6250   0.0000   0.0000
     charge :    13.367507
     magnetization :          0.000165    0.000174    2.717217
     magnetization/charge:    0.000012    0.000013    0.203270
     polar coord.: r, theta, phi [deg] :     2.717217    0.005053  
46.399234
     constrained moment :     0.000000    0.000000    2.800000


==============================================================================
But the constraint energy (Ryd) = 2.04918119.
The question is, the lambda has been quite large, scf problems will occur
if it is increased. But the constraint energy is still large. How can I get
both constraint energy and scf converged in a constraint magnetic moment
calculation?
Thanks for the reply. 
 
INPUTFILE:
 &control
    pseudo_dir = "~/pr/QE5/pseudo"
    outdir="./",
    calculation="scf",
 /
 &system
    ibrav=  0, nat=  2, ntyp= 2,
    ecutwfc = 280.0,
    occupations='smearing', 
    degauss=0.001,
    smearing='gauss' 
    starting_magnetization(1)=2.8
    starting_magnetization(2)=2.8
    angle1(1)=0.0
    angle1(2)=0.0
    angle2(1)=0.0
    angle2(2)=0.0
    constrained_magnetization="atomic"
    noncolin=.ture.
    lambda=150.0
    nosym=.true.
/
 &electrons
    mixing_beta = 0.1
    electron_maxstep=200
 /
ATOMIC_SPECIES
Fe1 56 Fe.pbe-sp-hgh.UPF
Fe2 56 Fe.pbe-sp-hgh.UPF
ATOMIC_POSITIONS angstrom
Fe1 3.0 0.0 0.0
Fe2 5.0 0.0 0.0
K_POINTS automatic
1 1 1 0 0 0
CELL_PARAMETERS angstrom
    8.0    0.0000000000000000    0.0000000000000000
    0.0 8.0    0.0000000000000000
    0.0 0.0 8.0


On Fri, 4 Jul 2014 13:36:09 +0200, Gabriele Sclauzero
<gabriele.sclauzero at mat.ethz.ch> wrote:
> Dear Pang Rui,
> 
>      It's difficult to judge just from the input. Has the scf converged?

> What are the final values of theta for the two atoms? Can you report the

> relevant part of the output?
> 
> GS
> 
> On 06/28/2014 11:40 AM, Pang Rui wrote:
>> Dear all
>> Could anyone gave me some suggestion to do the constraint magnetic
moment
>> calculation?
>> I followed the constraint magnetic moment calculation of the following
>> handson with PWSCF.
>> http://www.vasp.at/vasp-workshop/slides/handsonIV.pdf
>> VASP uses the same method of PWSCF on constraint MM. However, I found
the
>> constraint energy is difficult to go zero. In the handson, they use
>> lambda=50eV, got a constraint energy of 0.22591E-03eV. I used
>> lambda=150Ry,
>> but got a constraint energy=0.35Ry. Even PWSCF is different with VASP
on
>> some detail, I think it is not reasonable for two codes showing so
large
>> difference. So could anyone point the unreasonable part of my input
file?
>> Or give me some suggestion? This puzzled me for quite a long time.
>> The following is the input file.
>> &system
>>      ibrav=  0, nat=  2, ntyp= 2,
>>      ecutwfc = 80.0,
>>      occupations='smearing',
>>      degauss=0.001,
>>      smearing='gauss'
>>      starting_magnetization(1)=2.5
>>      starting_magnetization(2)=2.5
>>      angle1(1)=0.0
>>      angle1(2)=45.0
>>      angle2(1)=0.0
>>      angle2(2)=90.0
>>      constrained_magnetization="atomic direction"
>>      noncolin=.ture.
>>      lambda=150.0
>>      nosym=.true.
>> /
>>   &electrons
>>      mixing_beta = 0.1
>>      electron_maxstep=200
>>      startingwfc='file'
>>      startingpot='file'
>>   /
>> ATOMIC_SPECIES
>>   Fe1 56 Fe.pbe-spn-kjpaw_psl.0.2.1.UPF
>>   Fe2 56 Fe.pbe-spn-kjpaw_psl.0.2.1.UPF
>> ATOMIC_POSITIONS angstrom
>> Fe1 3.0 0.0 0.0
>> Fe2 5.0 0.0 0.0
>> K_POINTS automatic
>> 1 1 1 0 0 0
>> CELL_PARAMETERS angstrom
>>      8.0    0.0000000000000000    0.0000000000000000
>>      0.0 8.0    0.0000000000000000
>>      0.0 0.0 8.0
>>
>> PS1: I found the PWSCF used starting_magnetization as
>> magnetization/charge
>> for generating initial wave function,but used it as magnetization in
>> constraint MM. So I set 2.5 for them.
>> PS2: The startingwfc is of a smaller lambda calculation.

-- 
PostDoc
Department of Physics, South University of Science and Technology of China