Reproducing Results for Ultrafast Demagnetization (Evans 2015)

[marco menarini]

Hi, 
I am having trouble to reproduce the results from Dr. Evans paper Quantitative simulation of temperature-dependent magnetization dynamics and equilibrium properties of elemental ferromagnets (PHYSICAL REVIEW B 91, 144425 (2015)) on Vampire5.

It seems that during the thermalization process at room temperature (T=300 for 20ps), the material almost demagnetize if I choose material[1]:damping-constant=0.001 (the value given in the paper), while it stays to ~0.9 (the correct value in the paper) if a critical damping is chosen. Moreover, I cannot seem to reproduce the ultrafast demagnetization during the pulse lifetime (I choose the value given in the paper for gamma,Cl, and G, and a fluence of 20mJ/cm^3)

I tried the simulations using both Vampire5 and Vampire4 and the result is similar in both cases. I attached the input file and the material file (only difference between the two simulation is the damping constant), as well as the plot of the magnetization and the temperature profile obtained

Comparison: I considered the results obtained using Vampire5 and Vampire4. On the top it shows the magnetization dynamics during the thermalization period (T=300K), on the bottom the behavior after the pulse us shown. On the left are the results for damping=0.001 and on the right the dynamics for damping=1.0

Temperature dynamics 2 temperature model

Input file:

#------------------------------------------

# Sample vampire input file to perform

# benchmark calculation for v4.0

#

#------------------------------------------

#------------------------------------------

# Creation attributes:

#------------------------------------------

create:periodic-boundaries-x

create:periodic-boundaries-y

create:periodic-boundaries-z

create:crystal-structure=fcc

#------------------------------------------

# System Dimensions:

#------------------------------------------

#dimensions:unit-cell-size = 3.54 !A

dimensions:system-size-x=8 !nm

dimensions:system-size-y=8 !nm

dimensions:system-size-z=8 !nm

#------------------------------------------

# Material Files:

#------------------------------------------

material:file=Ni.mat

#------------------------------------------

# Simulation attributes:

#------------------------------------------

sim:equilibration-time-steps = 20000

sim:time-steps-increment = 10

sim:total-time-steps=40000

#sim:program=curie-temperature

sim:program=laser-pulse

#sim:integrator=monte-carlo

sim:integrator= llg-heun

# default is 20 mj/cm^2

sim:laser-pulse-power=20

sim:laser-pulse-time=100.0e-15

sim:two-temperature-electron-heat-capacity=800.0

sim:two-temperature-phonon-heat-capacity=4.0e6

sim:two-temperature-electron-phonon-coupling=12.0e17

sim:laser-pulse-time=60e-15

#------------------------------------------

# data output

#------------------------------------------

output:real-time

output:temperature

output:mean-magnetisation-length

output:output-rate = 10

output:mean-exchange-energy

config:atoms

config:atoms-output-rate=1

screen:real-time

screen:temperature

screen:mean-magnetisation-length

screen:output-rate = 1

screen:mean-exchange-energy

Material file:

#===================================================

# Sample vampire material file V5

#===================================================

#---------------------------------------------------

# Number of Materials

#---------------------------------------------------

material:num-materials=1

#---------------------------------------------------

# Material 1 Cobalt Generic

#---------------------------------------------------

material[1]:material-name=Ni

material[1]:damping-constant=1.00

material[1]:exchange-matrix[1]=2.757e-21

material[1]:atomic-spin-moment=0.606 !muB

material[1]:uniaxial-anisotropy-constant=5.47e-26

material[1]:material-element=Ag

material[1]:minimum-height=0.0

material[1]:maximum-height=1.0

material[1]:temperature-rescaling-curie-temperature=635

material[1]:temperature-rescaling-exponent=2.322

[Olle Heinonen]

Hi. Did you get this sorted out by any chance? I just pulled and built vampire (I have used it a lot before, but not with temperature pulses). I am using basically your input file but the output electron and phonon temperatures are just 300 no matter what I do with the laser power or pulse length. Any insight....?

Cheers,
Olle

[Ruth Li]

The output "mean-magnetisation-length" should be "magnetisation", and the magnetization from source are barely unit vector, we should modify codes by multiplying magnetization unit vector with their magnetisation-length

[AMRENDRA KUMAR]

Dear Vampire Users,

Does this issue is sorted ?
I was able to get the ultrafast demagnetization curve. But i could not match it exactly with the results given in the paper(PHYSICAL REVIEW B 91, 144425 (2015). I have attached input and material files. Please @R. F. L. Evans, have a look at my input and material file, and let me the reason why I am not able to get the desired result as given in your research paper.
Thanks and regards,
Amrendra Kumar

[gabo...@gmail.com]

In your material.mat, one line you have is:

material[1]:damping-constant=1

On page 5 in [1], there is:

λ = 0.001 is the phenomenological Gilbert damping parameter

That may correspond with:

material[1]:damping-constant=0.001

In addition, on page 6 in [1], there is:

The sLLG is solved numerically using the time dependent electron temperature rescaled using Eq. 9 with the Heun numerical scheme and a timestep of  Δt = 1×10−16 s.

To account for Δt , I believe you may need to add to your input file the line:

sim:time-step=1.0e-16

The paper [1] is from 2015.  It looks like there have been three bug fixes or changes to the temperature_pulse.cpp since then [2].   In April 2016, it looks like one of those changes was:

Bugfix: Corrected implementation of two temperature pulse calculations to include realistic pump power parameter (now specified in mJ/cm^2)

That's only from looking at temperature_pulse.cpp.  There could be changes to other areas of the source code.  Thus, you potentially could get outputted results that are not exactly the same as in [1] when using latest version of VAMPIRE.

There is also a possible issue is that I'm not seeing a mJ/cm^2 fluence given in [1] for trying to reproduce the results.

There is a FIG. 2 in the Phys. Rev. Lett. 76, 4250 (1996) article [3] that has 7 mJ/cm^2.  The [3] is cited within [1], however, results from VAMPIRE don't seem to match with FIG. 4 in [1] with 7 mJ/cm^2.  Using a fluence of 100 mJ/cm^2 seems give a result closer to matching with that in FIG. 4.  As using the attached input and material.mat files, it gave me the plot:

In FIG. 4, though, I'm guessing the paper authors chose to use the 20 ps equilibrated point as their zero reference point.  So, likely shifted the curve in the plot in time to make the 20 ps point be at 0 (i.e., used attached file MvsWithTimeShift with gnuplot for making the plot):

username@computername:~/Desktop/Ni$ ~/vampire-develop/vampire-serial
                                                _          
                                               (_)        
                    __   ____ _ _ __ ___  _ __  _ _ __ ___
                    \ \ / / _` | '_ ` _ \| '_ \| | '__/ _ \
                     \ V / (_| | | | | | | |_) | | | |  __/
                      \_/ \__,_|_| |_| |_| .__/|_|_|  \___|
                                         | |              
                                         |_|              

                      Version 7.0.0 Jun 10 2024 22:53:33

             Git commit: e38fe604be3d1f5abbbe34a5e2f501b9e7510a47

  Licensed under the GNU Public License(v2). See licence file for details.

  Developers:   Richard F L Evans, Sarah Jenkins, Andrea Meo,
                Daniel Meilak, Andrew Naden, Matthew Ellis,
                Oscar Arbelaez, Sam Morris, Rory Pond, Weijia Fan,
                Phanwadee Chureemart, Pawel Sobieszczyk, Joe Barker,
                Thomas Ostler, Andreas Biternas, Roy W Chantrell,
                Wu Hong-Ye, Razvan Ababei, Sam Westmoreland,
                Milton Persson
 
                Compiler Flags:

  Vampire includes a copy of the qhull library from C.B. Barber and The
  Geometry Center and may be obtained via http from www.qhull.org.

================================================================================
Tue Jun 18 06:08:28 2024
================================================================================
Initialising system variables
Creating system
Generating neighbour list..........done!
Copying system data to optimised data structures.
Using generic/normalised form of exchange interaction with 48 total interactions.
Number of atoms generated: 48668
Starting Simulation with Program Temperature-Pulse...
Simulation run time [s]: 1506.95
Simulation ended gracefully.
username@computername:~/Desktop/Ni$ gnuplot --persist MvsTime
username@computername:~/Desktop/Ni$ gnuplot --persist MvsWithTimeShift

[1] https://eprints.whiterose.ac.uk/86590/1/paper_39.pdf

[2] https://github.com/richard-evans/vampire/commits/main/src/program/temperature_pulse.cpp

[3] https://doi.org/10.1103/PhysRevLett.76.4250

Kind Regards,

Gavin

VAMPIRE user

On Wednesday, June 12, 2024 at 5:28:08 AM UTC-6 amrendra...@gmail.com wrote:

Dear Vampire Users,

Does this issue is sorted ?
I was able to get the ultrafast demagnetization curve. But i could not match it exactly with the results given in the paper(PHYSICAL REVIEW B 91, 144425 (2015). I have attached input and material files. Please @R. F. L. Evans, have a look at my input and material file, and let me the reason why I am not able to get the desired result as given in your research paper.
Thanks and regards,
Amrendra Kumar

 

On Friday 2 April 2021 at 02:13:29 UTC+5:30 menarin...@gmail.com wrote:

Hi, 
I am having trouble to reproduce the results from Dr. Evans paper Quantitative simulation of temperature-dependent magnetization dynamics and equilibrium properties of elemental ferromagnets (PHYSICAL REVIEW B 91, 144425 (2015)) on Vampire5.

It seems that during the thermalization process at room temperature (T=300 for 20ps), the material almost demagnetize if I choose material[1]:damping-constant=0.001 (the value given in the paper), while it stays to ~0.9 (the correct value in the paper) if a critical damping is chosen. Moreover, I cannot seem to reproduce the ultrafast demagnetization during the pulse lifetime (I choose the value given in the paper for gamma,Cl, and G, and a fluence of 20mJ/cm^3)

I tried the simulations using both Vampire5 and Vampire4 and the result is similar in both cases. I attached the input file and the material file (only difference between the two simulation is the damping constant), as well as the plot of the magnetization and the temperature profile obtained

Comparison: I considered the results obtained using Vampire5 and Vampire4. On the top it shows the magnetization dynamics during the thermalization period (T=300K), on the bottom the behavior after the pulse us shown. On the left are the results for damping=0.001 and on the right the dynamics for damping=1.0

Temperature dynamics 2 temperature model