Messages

91

General Questions and Answers / Re: crystal field splitting

« on: September 12, 2017, 09:40 »

Could you please state which exact figure are you referring to? The DOS analysis carried out in the paper can be done using the new ProjectedDensityOfStates, available in ATK 2017:

https://www.youtube.com/watch?v=6lw3nurLCcg

Regards,

Daniele.

92

General Questions and Answers / Re: device simulation error

« on: September 12, 2017, 09:36 »

Hi,

the log file does not provide sufficient information to understand what has caused the crash, but I suspect it is due to insufficient memory, because of incorrect load balancing. Since you have 96 contour points and only 70 processes, the first 26 processes are occupied twice as much as the others.

Could you try to re-run the simulation using:

equilibrium_method = GreensFunction(
    processes_per_contour_point=2,
    )

Regards,
Daniele.

93

General Questions and Answers / Re: Are these EDP curves correct?

« on: September 12, 2017, 09:12 »

Hi,

The potential looks flat-enough. However, I agree that the behavior of this device is quite weird, and I think it is mostly related to the fact that the device is very short, even including doping.
For bulk silicon, the depletion region at the doping that you have considered is around 20 nanometers. For a 2D material, screening will be less effective, and therefore the depletion region will be even larger.
If you don't want to increase further the size of the device, the only solution would be to keep increasing the doping, until the potential in the central region is fully screened. But since your device is very short, you might need to go to quite large values of doping.

Regards,
Daniele

94

General Questions and Answers / Re: Are these EDP curves correct?

« on: September 11, 2017, 09:34 »

Hi,

by doping, do you mean explicit doping by including dopant atoms in the structure? The HDP looks reasonable, and it shows that the charge redistribution is different between the two systems.
Also, notice that, apart from the oscillation due to the atomic cores, the potential should be flat close to C = 0.0 and C = 1.0.

Regards,
Daniele.

95

General Questions and Answers / Re: Prediction of a strucure

« on: September 11, 2017, 09:27 »

Hi,

this can be done by performing a geometry optimization of the structure, by allowing the unit cell to relax with a target pressure of 0 GPa.

Regards,
Daniele

96

General Questions and Answers / Re: Gilbert damping

« on: September 5, 2017, 15:45 »

Hello,

No, unfortunately it is not possible at the moment.

Regards,
Daniele.

97

General Questions and Answers / Re: Running batch jobs in command line

« on: August 22, 2017, 13:54 »

The script depends a lot on the environment where you want to run the calculation.

If you run on a cluster with e.g. PBS, you should be able to preview the submission script by clicking on the corresponding machine in the Machine Manager, then on Edit->Preview Script. It should be possible to use the same script also from the terminal.

If it is Linux on your local machine, use the following script "mpi_atkpython".

#!/bin/bash
export OMP_NUM_THREADS=1
export OMP_DYNAMIC=FALSE

MY_MPI=/home/stradi/QuantumWise/VNL-ATK-2017.0/libexec/mpiexec.hydra
MY_PYTHON=/home/stradi/QuantumWise/VNL-ATK-2017.0/bin/atkpython

N="-n $1"

$MY_MPI $N  $MY_PYTHON $2

The variables MY_MPI and MY_PYTHON should point at the MPI executable and at the atkpython executable, respectively.

To run the script "calculation.py" on 4 processors, type:

mpi_atkpython 4  calculation.py

Regards,
Daniele.

98

General Questions and Answers / Re: Transmission Eigenstates - linear combination of the Bloch states

« on: August 17, 2017, 13:49 »

Hi,

If I understood the question correctly, I do not think that what you propose is possible, because the transmission eigenchannels do not have the same symmetry as the Bloch states, see PRB 76 115117 (2007).

However, you can still calculate the transmission eigenstates for different energies and k-points, and then compare them directly.

Regards,
Daniele.

99

General Questions and Answers / Re: band structure under electric field with LCAO basis

« on: August 14, 2017, 16:14 »

Hello,

Then, 'solvent_dielectric_constant' is for material?

Yes, it is.

For example, if I have MoS2, then, I need to set  dielectric constant of MoS2 to 'solvent_dielectric_constant'?

You should set the value of the dielectric constant according to that of the solvent, not that of the material you want to simulate.
If the solvent_dielectric_constant = 1, it means that the region around the material you simulate is vacuum.

And if I use plane wave basis set instead of LCAO, I found that I cannot modify variable 'solvent_dielectric_constant'.

To enable the use of a solvent dielectric constant, you should use a Possion solver other than FFT and FFT2D, for example "Multi-Grid"

One more question is if I want to do device simulation to get IV in monolayer MoS2 MOSFET, how do I set the dielectric constant for monolayer MoS2?
Device characteristic should be affected by dielectric constant of channel material MoS2.

In a device calculation in ATK, the channel material is treated at the atomistic level, so it is not necessary to set any dielectric constant.

Regards,
Daniele.

100

General Questions and Answers / Re: Structural phase of MoS2

« on: August 14, 2017, 09:57 »

Just to clarify, if you are trying to create the 1T phase of MoS2, it is very easy to build it in VNL starting from the MoS2 1H phase contained in the structures database. You can import this structure from the database into the Builder, and then use the 'Coordinate Tools -> Translate' to shift one of the S atoms and create the 1T structure. For me, the following shift worked (in fractional coordinates):

A : -0.3333
B :   0.3333
C :   0.0000

Here is a tutorial on using the Builder together with the database:
http://docs.quantumwise.com/tutorials/builder_manual/builder_import/builder_import.html

Regards,
Daniele.

101

General Questions and Answers / Re: How to perform Bader charge analysis for metallic interface? Edit

« on: August 11, 2017, 15:16 »

I think you are mixing up things here. To calculate the CDD, you need:

1) the Al/Fe cell;
2) the Al fragment (using the same cell as 1);
3) the Fe fragment (using the same cell as 1).

Therefore, if you are able to calculate the Al/Fe system, you should also be able to calculate the fragments to extract the CDD. The required cutoff for 2 and 3 will be that used for 1. 

Daniele.

102

General Questions and Answers / Re: Bug of version 2017.0？

« on: August 11, 2017, 11:21 »

Hi,

the errors you experienced are due to the fact that the tutorial was strongly outdated (it was done using ATK 2015.1), and did not reflect anymore the current implementation of the methods based on the Boltzmann Transport Equation to calculate the phonon-limited mobility. More specifically, the most important improvement is that ATK now makes full use of sparse matrices for the evaluation and storage of the electron-phonon coupling elements.

We have updated the tutorial, which now focuses specifically on the use of ATK for the efficient calculation of the phonon-limited mobility:
http://docs.quantumwise.com/tutorials/mobility/mobility.html


As you will see, the section on the use of the electron-phonon coupling analyzer for the analysis of the electron-phonon couplings is gone. The reason is that the workflow for this analysis has slightly changed with respect to ATK 2015.1, and will be part of a separate tutorial, which has already been planned. If you still want to perform this analysis before the tutorial is online, I suggest you to read carefully the chapter "Usage Examples" in the section about the ElectronPhononCoupling class in the ATK manual:
http://docs.quantumwise.com/manuals/Types/ElectronPhononCoupling/ElectronPhononCoupling.html


Citing literally:
"to visually ananlyze the coupling matrix elements with the ElectronPhononCoupling analyzer, it is advantagerous to calculate the matrix elements for all (k,q)-points. This is achieved by specifying a large energy tolerance, ‘energy_tolerance=100*eV’, like in the script below. In that case it will also be advantagerous to use a dense matrix representation. This can be specified by setting ‘store_dense_coupling_matrices=True’."

Regards,
Daniele.

103

General Questions and Answers / Re: graphene phonon dispersion

« on: August 10, 2017, 08:42 »

Hi,

When I add graphene from database and I repeat A and B to make a large size, is it the same result like Brillouin Zone folding?

Yes, repeating the cell along A and B will result in a folding of the phonon bands in the Brillouin zone.

What I want to do is the phonon dispersion graphene on a substrate as function of different interface bonding.
If I make an interface between graphene and a substrate, I can't repeat its structure to obtain phonon dispersion.
Am I right?

For a repeated structure, you will still obtain a phonon dispersion, but the bands will be folded, hence the analysis might be complicated.
For a 1x1 structure, the analysis of the phonon band structure of graphene on a substrate is simpler, see Nano Lett. 10, 4335 (2010)
http://pubs.acs.org/doi/abs/10.1021/nl101657v

How can I have phonon dispersion without BZ folding with many atoms?

In a repeated structure, you will still get the correct band dispersion, but the phonon bands will be folded and difficult to analyze.
If you want to unfold the phonon bands of the supercell and map them into those of a 1x1 unit cell, unfortunately this is not possible at the moment in VNL. 

Regards,
Daniele.

104

General Questions and Answers / Re: Fermi level not midgap in semiconductor nanowire

« on: August 10, 2017, 08:11 »

Hi,

This oscillatory behavior means that the chemical potential is not yet converged with respect to the number of k-points. You may need many more k-points in the C direction, since the determination of the chemical potential becomes increasingly difficult as the band-gap increases. This is why you did not observe this behavior using GGA, which predicts a smaller band-gap.

I suggest you to check your results up using 401 k-points at least.

Regards,
Daniele

105

General Questions and Answers / Re: Restart a device calculation

« on: August 9, 2017, 10:41 »

Hi,

your script should work fine, but it seems that the file you nlread  does not have an updated calculator:

The argument configuration_1 is the same object as configuration_2, however, the object does not seem to have an updated calculator.

I assume you have read the tutorial:
http://docs.quantumwise.com/tutorials/restarting_calculations/restarting_calculations.html

Could you attach the hdf5 file you try to read in and the input of the calculation?

Regards,
Daniele.