Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - Julian Schneider

Pages: 1 [2] 3 4 ... 10
16
Future Releases / Re: Combining potentials of classical
« on: July 24, 2017, 16:34 »
You cannot select particle types when you set up a ReaxFF potential (see http://docs.quantumwise.com/manuals/Types/ReaxFFPotential/ReaxFFPotential.html#reaxffpotential-c). It automatically acts on all particles in the system, meaning that the reax parameter file must contain parameters for at least all elements in the configuration.
If you wanted the potential not to act on some of the elements, you would have to add these as dummy elements with zero interaction to the reax parameter file.

Once you have set up a ReaxFF potential for your entire system, you can add (almost) any other potential on top of the ReaxFF potential to extend the potential if necessary.

17
Future Releases / Re: Combining potentials of classical
« on: July 22, 2017, 12:09 »
In principle you can combine all potentials, not only Lennard Jones and Tersoff.
There are some limitations for EAM and ReaxFF, which means that a ReaxFF or EAM potential cannot be used for only a part of the configuration, but they must be defined for all atoms in the system (for EAM one can make a workaround, if necessary).
Still, other potentials, such as Lennard Jones, can be added to a ReaxFF or EAM potential, as well.

18
With these settings you should in principle be able to import your trajectory in the labfloor in VNL.
VNL looks for the log file that has been generated by lammps,  make sure that this file ends with ".lammps", otherwise VNL will not detect it as lammps file. Rename the file if it has a different extension.
From this log file, VNL will read the name of the lammps trajectory file and import the trajectory.
The trajectory should show up under the name "xxx" if "xxx.lammps" is the name of the lammps log file.
You can treat it as a normal VNL MD-Trajectory. i.e. open it in the Viewer / MovieTool / MD-Analyzer, etc.
 If you want to send a snapshot of the trajectory to the Script Generator, open the lammps trajectory with the MovieTool, play the movie until you have reached the desired snapshot, and then use the blue "Send To" arrow in the lower right-hand corner to send the configuration to the ScriptGenerator.
If you haven't done so already, you should read the tutorial http://docs.quantumwise.com/tutorials/lammps_trajectories/lammps_trajectories.html, everything should be explained there.

19
Questions and Answers / Re: Stress Error
« on: July 12, 2017, 13:58 »
If you use the ScriptGenerator:
You will only do a stress optimization if you uncheck the "Constrain Lattice Vectors" box, no matter what stress error tolerance value you give.

If you set up the calculation in a script:
Whether you do a stress optimization or not, depends on if you set the parameter
Code: [Select]
disable_stress
to True (no stress optimization) or False (so a stress optimization).
If you don't set this keyword, the default is False, that means by default you do a stress optimization (see http://docs.quantumwise.com/manuals/Types/OptimizeGeometry/OptimizeGeometry.html#optimizegeometry-f).
That also means you will perform a stress optimization, even if you don't explicitly give the stress error tolerance (i.e. via the
Code: [Select]
max_stress
keyword). In that case the default stress tolerance of 0.1*GPa will be used.
Same with the
Code: [Select]
target_stress
, here the default is 0.0*GPa, unless you specify something else.

20
Questions and Answers / Re: Stress Error
« on: July 10, 2017, 12:49 »
The stress error tells you by how much the current stress deviates from the given target stress. The stress error must be below the given stress tolerance to consider the stress optimization converged.

21
Questions and Answers / Re: Phonon transmission sprectrum
« on: July 2, 2017, 14:12 »
I would guess the problem comes from the fact that the Tersoff-potential only describes the local bonded interactions within the graphene-layers, but does not account for the longer-ranged dispersion interactions between adjacent layers. This is  a known short-coming of Tersoff- and similar potentials. If there is no interactions between adjacent layers, then phonons cannot be transmitted across different layers.
This is typically corrected by adding a Lennard-Jones potential between Carbon-atoms, that accounts for the dispersion interactions, as e.g. in http://dx.doi.org/10.1103/PhysRevB.88.045444 .

22
Questions and Answers / Re: About trajectory files
« on: May 17, 2017, 09:29 »
Hi,

This is a known shortcoming for nc trajectories with a small interval.
Most of these issues will be solved in the upcoming ATK-2017 version, where we use the HDF5 format to save MD trajectories. There will also be an option to disable storing velocities, forces, and stress to make the file size smaller.
For ATK-2016 the workaround to make it at least somewhat faster for such cases is that you switch off  "Save trajectory" (or set
Code: [Select]
trajectory_filename=None
in the script in the MolecularDynamics function), and instead make sure that you check "Save" in the Molecular Dynamics widget (or add a line 
Code: [Select]
nlsave('trajectory.nc', md_trajectory)
after the MolecularDynamics block in the script).
Then the trajectory is not saved on-the-fly during the simulation, but as a whole after the simulation has completed, which should be faster.

23
If you select only M, then the voidsize calculation is performed only on the M atoms in the system, i.e. all other atoms are ignored.

24
The voidsizes are a measure for how evenly the atoms in a material are distributed. This is primarily of interest for amorphous materials, as the atoms are arranged in an irregular manner.
What you measure is essentially the volume of the void space between the atoms.
This volume is calculated by first creating a Delauney traingulation (https://en.wikipedia.org/wiki/Delaunay_triangulation) of the atoms in the system, and then calculating the tetrahedral volume for each simplex. The size of a void is then calculated as the diameter (in Angstrom) of a sphere with equivalent volume.
The Voidsize Distribution analysis gives the distribution of these sizes in the system.
The Element selection lets you select only a subset of elements for whcih the voidsaize should be calculated, but the most common case would be that would calculate the voidsize for all elements in the system.
The Resolution selection lets you choose the bin size that is used when calculating the histogram of the void sizes.

25
Anharmonic effects are not included in phonon transmission calculated via non-equilibrium Green's functions. Everything is purely harmonic. Therefore NEGF is primarily suited for interfacial thermal conductance (where scattering at the interface is the main contribution to thermal resistance) and not so much for bulk conductivity (where inelastic phonon-phonon-scattering is the main contribution to resistance).

Non-equilibrium MD, on the other hand fully includes anharmonic effects and can therefore be used to calculate both interface conductance and bulk conductivity.
So what you observe makes perfect sense.

Have a look at our recent webinar on thermal transport simulations http://docs.quantumwise.com/webinars/webinars.html

26
VNL calculates the Young's modulus as Y_i = 1/S_ii  (i=x,y,z) where S is the inverse of the elastic constants matrix.
We do not calculate the bulk modulus and the shear modulus directly, but from the elastic constants, as well. Since there are several conventions how to do this exactly (Hill, Voigt, Reuss) which all use slightly different formulas, it is more straightforward to calculate the Young's modulus from the elastic compliance matrix, since that does not rely on any of these conventions.

27
Sorry for the late answer, I took me some to look into your system.

Quote
1. What do you think the suitable momentum exchange interval? I used 150 intervals for the simulation. When we used 150 intervals, the temperature was increased.
150 should be fine in general, although it depends a bit on your system. Generally, the exchange interval allows you to modify the magnitude of the thermal current. For a system with a small cross section, the same thermal current can lead to a larger thermal flux and thus a larger temperature gradient. The temperature gradient should ideally not be too large, on the other hand large enough so that it can be measured with sufficient accuracy. 

Quote
2. How can we decide the convergence or steady state of the temperature profile? Are there any criteria for that? For my case, the temperature gradient is changed in a short time, and it is hard to figure out the convergence of the temperature profile.

You should plot your temperature profile at different time intervals during the simulation. If you don't see a difference between the temperature profiles of subsequent time intervals then you can assume that your temperature profile has reached a steady-state.

Quote
3. In the tutorial of the calculating interfacial thermal conductance using molecular dynamics, you compared the two temperature profiles. Are there a proper time range for the integration of the temperature profile? The temperature gradient is changed when we modify the time range.

At first, I would recommend you to use a smaller log interval than 10 000, e.g. something like 500 or 1000. This will make the temperature profile a lot smoother, as you will have more snapshots to average over. Remember, for your system you have to calculate the profile along the B-direction, not the default C-direction.
Yes, the temperature profile will depend on the chosen time interval. On the one hand, as for a small time interval your average might not be very good and the profile can become fuzzy. On the other hand, as long as you haven't reached a steady state, changing the length of the time interval will also include snapshots from different phases of the simulation, which will result in a different profile, that's just the nature of the non-equilibrium simulation. So, as a rule of thumb, you could use intervals of 50-100 ps to check for convergence, and once you know that you have reached a steady-sate, you should average the profile over several 100 ps, to get a smooth profile.
If your profile is too fuzzy you can also try and increase the bin width a bit.

Quote
4. Can you tell me the reason why the increase of the system temperature during the simulation of the non-equilibrium momentum exchange? In the case, the system is isolated, and no energy can be provided.

I can reproduce your problem, and it actually seems that the total energy increases during time. Although I'm not 100% sure, I would guess a major part of the problem is that the time step is too large and the conservation of total energy is not given any more. In this case a smaller time step, maybe 0.5 fs, might at least fix the problem to some extent. Furhtermore, the system is very soft and buckles quite a lot during the simulation which might make it additionally difficult.

Quote
5. What is the meaning of "truncated content"? I saw the warning in my log file of the simulation.

That is just a message that the output line in the log file is longer than the width of the terminal. You can ignore that.

28
The kinetic energy should not increase significantly  during the simulation. By increasing the time step, do you mean the total number of MD steps or the actual time step  (e.g. 1 fs)?
The latter should not be increased. Increasing it to e.g. 5 fs can in fact lead to energy conservation being violated.
If you just increased the number if MD steps and the energy increases, then the system may not be sufficiently equilibrated before the NEMD simulation. Or your temperature gradient has become so large that part of the system experiences such a high temperature that it runs out of equilibrium.

29
The NVTNoseHooveChain in ATK-2015 is essentially the same as the new NVTNoseHoover thermostat in ATK-2016.
The old NVTNoseHoover in ATK-2015 thermostat does not exists any more in ATK-2016, as it was using an older algorithm which is not state-of-the-art.
In ATK-2015 you should use the NVTNoseHooverChain thermostat.

30
Quote
1)    I know that the nanosheet is 2D and periodic in the y and z directions. Is it required to set the periodic boundary condition for the top and bottom side additionally? If the boundary condition is periodic, there are no scattering effects related on the phonon or electrons because the top and bottom side are not terminated anymore, Right? I expect for the case; there are no dependencies on the width of the structure for the thermal conductivity. I want to know about this.
With ATK-Classical you cannot switch off the periodic boundary conditions, so you should make sure that you have a sufficiently large vacuum buffer on top of th graphene layer if you want to simulate an isolated layer.

Quote
2)    For the non-equilibrium momentum exchange, should we set the fixed boundary at the left and right side of the structure besides the heat sink and source?
You have two options how to set up the boundaries in transport direction:
a) Add some vacuum buffer in transport direction and fix the outermost terminating atoms. This way you'd only have periodicity in the in-plane direction perpendicular to the transport direction.
b) You treat the system periodic in the transport direction, as well. Then the thermal current runs in both directions. Here you in principle don't need to fix anything.

Quote
3)     For the Nose-Hoover thermostat, the temperature gradient can be generated using reservoir temperature. Is there are no problem to obtain the thermal conductivity of the graphene layer although some synchronization issues are between the heat source and sink.
In principle you could just set two thermostats on heat sink and source to obtain the gradient. However, in this case you'd have no straightforward way to measure the magnitude of the thermal flux. So, I'd recommend using the NonEquilibriumMomentumExchange method.

Quote
4)    I read a several papers about NEMD for the calculation of the thermal conductivity by LAMMPS. In that case, the method is similar methods kinds of setting the different reservoir temperature on both sides for the Nose-Hoover thermostat.


According to recent papers, the alternative way to using NonEquilibriumMomentumExchange  not really setting reservoir temperatures, but to set achieve a specified thermal by heating and cooling the heat and sink with a constant heating power, which at steady state has the same magnitude as the thermal curent in the system. This method will be available in the upcoming ATK-2017 release.

Quote
5)    Please comment about the procedure of simulation, some notes and missing point and so on.


To obtain reliable bulk conductivity values, you may have to run a series of simulations at increasing system sizes and then extrapolate to infinite lengths, as described e.g. in http://aip.scitation.org/doi/abs/10.1063/1.4767516.
As, Petr said, for more explanations and details, you may want to look at our tutorials and the webinar.

Pages: 1 [2] 3 4 ... 10