Messages

31

General Questions and Answers / Re: Query regarding thermoelectric calculation

« on: August 4, 2020, 11:25 »

Hi,

1) In general you need to do a very careful optimization (relaxation) of the atomic structure when calculating phonons. In your case, the device seems to be a simple continuation of the electrodes, so it should be enough to do a fine relaxation of the electrode structure using OptimizeGeometry with a Force tolerance of 0.01 eV/Ang. At this point I would suggest to calculate the PhononBandstructure for the electrode to check that it looks reasonable. There shouldn't be phonon modes with negative energy.
From the relaxed electrode structure you make the device and calculate the DynamicalMatrix for the device.

You should only expect integer phonon transmissions for each individual transverse k-point. In your case the structure is periodic in the B-direction and if you calculate the phonon transmission for multiple k-points in the B-direction, you should not expect to get integer values. This is only true for a single k-point - or for a one-dimensional structure (where you only have the Gamma-point).

2) The thermoelectric properties are changing with doping concentration. So you should either plot the properties vs. doping (or vs. Fermi level) or pick a relevant one.

3) That depends on your doping level (position of Fermi level) - see answer to 2). The peaks closes to the valence/conduction band edges gives you the maximal ZT for p/n doped structures.

Regards,
Troels

32

General Questions and Answers / Re: Regarding phonon transmission calculation using quantumATK

« on: August 3, 2020, 13:41 »

Hi,
Sorry for the late reply, but this is vacation times...

I didn't see any obvious errors in your script. Also did you see any error messages - there are none in the Phonon_Trans.log file.

You can maybe try with this slightly modified script:

Code

path = '/home/utpal/Utpal/Jyoti/ATK/Thermoelectric_BeX2/BeX/device_pentag/DFT/Device_Pentag_DM.hdf5'
dynamical_matrix = nlread(path, object_id='dynamical_matrix')[0]
dynamical_matrix.update()

configuration = nlread(path, object_id='DeviceConfiguration_0')[0]

qpoint_grid = MonkhorstPackGrid()

phonon_transmission_spectrum = PhononTransmissionSpectrum(
    configuration=configuration,
    dynamical_matrix=dynamical_matrix,
    energies=numpy.linspace(0,0.5,101)*eV,
    qpoints=qpoint_grid,
    infinitesimal=1e-06*eV,
    self_energy_calculator=RecursionSelfEnergy(),
    )
nlsave('Phonon_Trans.hdf5', phonon_transmission_spectrum)
nlprint(phonon_transmission_spectrum)

33

General Questions and Answers / Re: mobility_error

« on: July 9, 2020, 14:14 »

Hi,

This looks strange and should not happen. How many processes did you run with?  I was unable to reproduce the error, with a slightly modified script. Perhaps try to run with fewer processes. Also, I noticed that your k-point sampling for the ElectronPhononCoupling is around the Gamma-point (ka_range=[-0.1, 0.1], kb_range=[-0.1, 0.1]), while I guess you would rather have it around the K-point, i.e.

Code

kpoints = RegularKpointGrid(
    ka_range=[0.23, 0.43],
    kb_range=[0.23, 0.43],
    kc_range=[0.0, 0.0],
    na=20,
    nb=20,
    nc=1,
    )

Regards,
Troels

34

General Questions and Answers / Re: Query Regarding thermoelectric coefficient calculation

« on: July 7, 2020, 12:19 »

Hi,

Currently the ZT is only shown in the table if both the phonon transmission and electron transmission spectra are available. However, the ZT which is shown in the plot is calculated from the usual formula ZT=G*S^2*T/k_tot,
where k_tot = k_el + k_ph. In the case where the phonon transmission spectrum is not calculated, the phonon thermal conductance, k_ph, is just set to zero.

Notice that often k_ph > k_el, which means that the ZT calculated without the phonon contribution is way too large. However, the values in the plot of ZT are fine, given that you are aware of the overestimation of ZT. In scientific papers, this quantity if sometimes denoted ZT_el, highlighting that it is calculated only from the electronic transmission. The "real" ZT is the ZT  = ZT_el * k_el / (k_el + k_ph).

Regards,
Troels

35

General Questions and Answers / Re: Regarding phonon transmission calculation using quantumATK

« on: July 7, 2020, 09:46 »

You can (and should) run the calculation of the DynamicalMatrix in parallel. This is the most expensive part of the calculation. When the DynamicalMatrix is calculated, you run a different script in serial to do the phonon transmission. The phonon transmission calculation can easily be done in serial for 144 atoms.

Regarding device optimization: If you simply repeat the unit cell and don't introduce defects, if should not be necessary to relax the device configuration.

Concerning the K-point sampling, this will depend on the system and in general convergence  with respect to number of K-points should be checked. But usually the default values in QuantumATK are good.

36

General Questions and Answers / Re: Semi-emprical methods

« on: July 7, 2020, 09:33 »

Hi,

If you are interested in Mobility calculations for the present system containing C, Al, S, and P atoms, I think you need to use an LCAOCalculator instead. Currently, we don't support HamiltonianDerivatices for Extended Huckel and we also miss the pair potentials needed to calculate Forces (for the DynamicalMatrix).

Regards,
Troels

37

General Questions and Answers / Re: print data of fatband

« on: July 3, 2020, 16:17 »

Hi,

Once you have a FatBandstructure object, you can get the projection weights as

Code

weights = fat_bandstructure.weights()

This will return a matrix of dimention (n_projections, n_kpoints, n_bands).

See also https://docs.quantumatk.com/manual/Types/FatBandstructure/FatBandstructure.html

Regards,
Troels

38

General Questions and Answers / Re: Regarding phonon transmission calculation using quantumATK

« on: July 3, 2020, 16:01 »

Hi,

This is probably caused by a bug that we have only recently fixed, which can cause the phonon transmission to hang, when running parallel calculations. Try to run the phonon transmission calculation in serial instead. This issue is solved in the upcoming 2020.09 version.

Regards,
Troels

39

General Questions and Answers / Re: Comparing DOS contribution from various orbitals

« on: June 25, 2020, 10:38 »

Hi,

The calculations can be run with something like

Code

ProjectOnShellsByElementAndSpin = ProjectionGenerator(atoms=ElementsProjection, l_quantum_numbers=True, spin=UpDownProjection)

projected_density_of_states = ProjectedDensityOfStates(
    configuration=bulk_configuration,
    kpoints=kpoint_grid,
    projections=ProjectOnShellsByElementAndSpin,
    energies=numpy.linspace(-10, 4, 1000)*eV,
    energy_zero_parameter=FermiLevel,
    bands_above_fermi_level=All,
    spectrum_method=TetrahedronMethod,
)

where I defined a new projection generator to project on elements, shells (s, p, d, etc.) and spin up/down.

Regarding the plotting, it will definitely be possible with the upcoming 2020.09 release. There you can rather easily customize plot to e.g. flip the sign of a given curve (in this case the spin-down component). For now, you might need to make the plots yourself by getting each individual pdos using p_xx = projected_density_of_states.evaluate(projection_index=xx), where xx is the index of the specific projection. And the plot the results using pylab or matplotlib.

40

General Questions and Answers / Re: Comparing DOS contribution from various orbitals

« on: June 23, 2020, 08:42 »

Hi,

You need to use the ProjectedDensityOfStates analysis object.
When you in the ProjectedDensityOfStates select projections to be  "Elements and Orbitals" you will get the individual contribution from each orbital. In the case of the p-shell you will get three curves (p_x, p_y, p_z).

Regards,
Troels

41

General Questions and Answers / Re: error in IdVg Calculation

« on: June 19, 2020, 11:35 »

Hi,
I think the problem is in your python script:

Code

# Add external potential
external_potential = AtomicCompensationCharge([
    ('doping_0', 0.0168772635404),
    ('doping_1', 1.66714432533e+98)
    ])

Clearly, something is wrong for the 'doping_1' value.

Regards,
Troels

42

General Questions and Answers / Re: Error in initial state energy durin electron_phonon_coupling

« on: June 15, 2020, 09:53 »

Hi,

The initial state energy range depends on your system and what you want to study. If your system is a metal, you are usually interested in an energy range around the Fermi level, so e.g. [-0.1, 0.1]*eV. If your system is a semi-conductor, you will usually want to study energies around the conduction band minimum (CBM) or the valence band maximum (VBM).

I suggest that you first do a Bandstructure calculation or a DensityOfStates calculation of your system and determine the relevant energies. Once you have determined the relevant energy range, you can use that in the ElectronPhononCoupling object. And then you will also need to increase the k-point sampling (kpoints_fractional) in ElectronPhononCoupling, such that you cover the energy range. For a semi-conductor, it can often be advantageous to use a RegularKpointGrid covering only the k-points around the CBM or VBM, if these are located around a single valley.

In your case, the CBM is located around the M point (0.5, 0.5, 0) and if you want to study the electron mobility I would then suggest a RegularKpointGrid of e.g.

Code

kpoints= RegularKpointGrid(
   ka_range=[0.45, 0.55],
   kb_range=[0.45, 0.55],
   kc_range=[0.0, 0.0],
   na=7,
   nb=7,
   nc=1,
)

and then set the initial_state_energy_range=[0.93,0.96]*eV.

The next you should determine is the qpoints. If you want to study both intra-valley and inter-valey scattering it might be best to use a MonkhorstPackGrid with a fine sampling, something like MonkhorstPackGrid(na=100, nb=100, nc=1). However, you can also focus on the intra-valley scattering taking place only around the M point. In this case I would suggest to start with

Code

qpoints= RegularKpointGrid(
   ka_range=[-0.1, 0.1],
   kb_range=[0.1, 0.1],
   kc_range=[0.0, 0.0],
   na=11,
   nb=11,
   nc=1,
)

You need to check your results for convergence wrt. both the k- and q-point ranges as well as the number of k-and q-points.
 
Regards,
Troels

43

General Questions and Answers / Re: Orbital contributions in bandstructure

« on: June 10, 2020, 13:44 »

Hi,

I think this is what you want:

Code

ProjectOnOrbitalsAndSpin = ProjectionGenerator(
    spin=UpDownProjection, l_quantum_numbers=True, m_quantum_numbers=True)

fat_bandstructure = FatBandstructure(
    configuration=bulk_configuration,
    route=['G', 'X', 'W', 'L', 'G', 'K', 'X', 'U', 'W', 'K', 'L'],
    points_per_segment=20,
    bands_above_fermi_level=All,
    projections=ProjectOnOrbitalsAndSpin
)

Regards,
Troels

44

General Questions and Answers / Re: Differene between zero bias conductance in bulk mode and device mode

« on: May 19, 2020, 09:04 »

Right. I'm not surprised that you get good agreement for pristine graphene. This is an 'easy' case for Huckel since there is only a single element. When you add dopant (nitrogen) atoms the interactions becomes more complicated and the Huckel parameters are probably not optimized for this situation. DFT is more transferable and probably gives a better description of the doped graphene.

Regards,
Troels

45

General Questions and Answers / Re: Differene between zero bias conductance in bulk mode and device mode

« on: May 18, 2020, 08:56 »

Hi,

There has unfortunately been a bug in the way QuantumATK calculated the transmission for a bulk configuration, which in some case could lead to wrong transmission values. This was solved in the 2019.12 version.
1) Concerning the comparison between DFT and extented Hückel, you cannot in general expect to get the same results. In the attached scripts, the device is calculated with Hückel, while the bulk configuration with DFT. Usually this will give different results, at least quantitatively different. There might be qualitative agreement. The use of force fields for the phonons is usually fine.
2) Yes, the electrode size can matter, also for a perfect device. You can use the "Electrode Validator" to the right of the LabFloor to validate if a certain bulk configuration is a valid electrode.

Regards,
Troels