how to optimize time steps and equilibration steps

[sira...@gmail.com]

i ma trying to simulate the coercivity of 2D material. The problem comes with how to optimize the equilibration and time steps for simulation because both these terms are very sensitive to the coercivity.

secondly if I don't have any experimental data for comparison then how i would be say that these equilibration and time steps are showing my result accurate. 

And how to set these steps for temperature dependent coercivity

i need some suggestions what the better way to resolve this timsteps issue for accurate results  

below is my input and mat file

create:crystal-structure=hcp

create:periodic-boundaries-x
create:periodic-boundaries-y

#------------------------------------------
# System Dimensions:
#------------------------------------------
dimensions:unit-cell-size=3.29 !A
dimensions:unit-cell-size=3.29 !A
dimensions:unit-cell-size=13.19 !A

dimensions:system-size-x= 32.9 !A
dimensions:system-size-y= 32.9  !A
dimensions:system-size-z=8.75 !A
#------------------------------------------
# Material Files:
#------------------------------------------
material:file=VSe2.mat

#------------------------------------------
# Simulation attributes:
#------------------------------------------

#------------------------------------------
# Program and integrator details
#----------------------------------------
sim:time-step=1E-15
sim:maximum-applied-field-strength= 0.2 !T
sim:minimum-applied-field-strength=-0.2 !T
sim:applied-field-strength-increment= 0.01 !T
sim:applied-field-angle-phi = 0.1    # (degrees from z)
sim:temperature=0
sim:equilibration-time-steps=100000
sim:loop-time-steps         =100000
sim:program=hysteresis-loop
sim:integrator=llg-heun

#------------------------------------------
# data output
#------------------------------------------
output:real-time
output:temperature
output:applied-field-strength
output:mean-magnetisation-length
output:magnetisation

screen:real-time
screen:temperature
screen:applied-field-strength
screen:mean-magnetisation-length
screen:magnetisation

and material file are as follow

# Material 1
#---------------------------------------------------
material[1]:material-name=V1
material[1]:damping-constant=1.0
material[1]:atomic-spin-moment=1.27 !muB
material[1]:uniaxial-anisotropy-constant= -1.6e-22
material[1]:material-element=V1
material[1]:minimum-height=0.0
material[1]:maximum-height=0.5
material[1]:initial-spin-direction=1,0,0
#---------------------------------------------------
# Material 2
#---------------------------------------------------
material[2]:material-name=V2
material[2]:damping-constant=1.0
material[2]:atomic-spin-moment=1.27 !muB
material[2]:uniaxial-anisotropy-constant= -1.6e-22
material[2]:material-element=V2
material[2]:minimum-height=0.5
material[2]:maximum-height=1.0
material[2]:initial-spin-direction=1,0,0

[sira...@gmail.com]

[sira...@gmail.com]

[gabo...@gmail.com]

To get accurate results from numerical calculations, convergence tests may be needed.

As an example, if you are familiar with Density Functional Theory (DFT) numerical calculations, a parameter of interest is the energy (ENE) and convergence test on the energy value is done by varying the number of k-points.  For example, see Figure 1 at [1].

In your case, it looks like Coercivity (Hc) value could be taken as the parameter of interest.

In the attached "input" file, everything was kept the same except for the line:

sim:loop-time-steps         =x

That line I changed and I ran multiple calculation varying the loop-time-steps with x = 10000, 50000, 100000, 150000, 200000, 250000, 300000, 350000, and 400000.

The results from that are shown in the attached "loop-time-step convergence.pdf".

For "loop-time-steps = 10000", you can see the hysteresis loop is inaccurate with Hc of 3.995 T.  However, it looks like the convergence and accuracy gets better as the loop-time-step is increased, where loops with 100000 and 400000 loop-time-steps start to look about the same having an Hc of about 1 T.

That behavior from increasing the loop-time-steps seems to follow the statement given at [2], which is:

Increasing the number of averaging steps (set by [sim:loop-time-steps]) will generally lead to smoother data.

Similarly, for equilibration-time-steps, you could fix the loop-time-steps value and make a plot of equilibration-time-steps versus Hc.

[1] https://www.beilstein-journals.org/bjnano/content/supplementary/2190-4286-2-45-S1.pdf

[2] https://vampire.york.ac.uk/tutorials/simulation/curie-temperature/

Kind Regards,

Gavin

VAMPIRE user