Can we use Non equilibrium molecular dynamics to get dynamical matrix?

[wot19920302]

Dear QuantumWise staffs:
           I noticed that most of calculations for phonon transport in this link: http://quantumwise.com/documents/ATK_Publication_List.html were based on NEMD(Non equilibrium molecular dynamics) with Tersoff potential. So my questions are:
          1. Can we use NEMD with specific potentials in ATK?
          2. If question 1 is "no", what method we use to calculate dynamical matrix(NEGF+ATK_DFT or NEGF+ATK_classical)?
          Best wishes
         
         

[Julian Schneider]

1. Yes, you can use the Non-equilibrium MD with all potentials available in ATK-Classical. Have a look at www.docs.quantumwise.com/tutorials/interfacial_thermal_conductance.html and www.quantumwise.com/documents/manuals/latest/ReferenceManual/index.html/ref.nonequilibriummomentumexchange.html.

2. Alternatively, you can also use NEGF to calculate the phonon transmission and the thermal transport, using either ATK-Classical or ATK-DFT (if the system is sufficiently small).
This approach is described in www.docs.quantumwise.com/tutorials/thermoelectrics_cnt_isotope.html

[wot19920302]

    Thank you ,I will try .

[wot19920302]

1. Yes, you can use the Non-equilibrium MD with all potentials available in ATK-Classical. Have a look at www.docs.quantumwise.com/tutorials/interfacial_thermal_conductance.html and www.quantumwise.com/documents/manuals/latest/ReferenceManual/index.html/ref.nonequilibriummomentumexchange.html.

2. Alternatively, you can also use NEGF to calculate the phonon transmission and the thermal transport, using either ATK-Classical or ATK-DFT (if the system is sufficiently small).
This approach is described in www.docs.quantumwise.com/tutorials/thermoelectrics_cnt_isotope.html

Dear Julian Schneider:
           Thanks for your reply,after reading the toturials you provided, I have a few confusions:
          1) Does it mean that we can calculate the thermal conductivity  of a system by two methods: NEGF and NEMD?  The former requires the device configuration and the latter requires the bulk configuration. Whether NEGF or NEMD need Calculators, so there are several combinations to calculate  thermal conductivity: NEGF+ATK_DFT,  NEGF+ATK_classical,  NEMD+ATK_DFT,  NEMD+ATK_classical, is that correct?
          2) Can we use ATK_SE(Extended huckel or Slater-Koster)  for NEGF and NEMD to save time?
          3) What are advantages of NEGF/NEMD for conductivity calculation, in other word: when should we use NEGF and when should we use NEMD? In fact, I am going to calculate thermoelectric coefficeints like figure of merit(ZT), and I know I can realise this function using NEGF in this tutorials:  http://docs.quantumwise.com/tutorials/thermoelectrics_cnt_isotope.html,  can we use NEMD to realise this function?
           Best regards
           

[Julian Schneider]

1) Does it mean that we can calculate the thermal conductivity  of a system by two methods: NEGF and NEMD?  The former requires the device configuration and the latter requires the bulk configuration. Whether NEGF or NEMD need Calculators, so there are several combinations to calculate  thermal conductivity: NEGF+ATK_DFT,  NEGF+ATK_classical,  NEMD+ATK_DFT,  NEMD+ATK_classical, is that correct?

Yes, that is correct.

2) Can we use ATK_SE(Extended huckel or Slater-Koster)  for NEGF and NEMD to save time?

In principle, yes, that always depends on whether the specific SE-method has pair-potentials so that the forces can be calculated. However, even in that case, you should carefully evaluate if the SE-method is suitable for such calculations, e.g., by calculating the PhononBandstructure and compare to experimental values or reference calculations.

3) What are advantages of NEGF/NEMD for conductivity calculation, in other word: when should we use NEGF and when should we use NEMD?

NEGF considers ballistic phonon transport and does not account for anharmonic effects, such as phonon-phonon-scattering, which means that in particular for perfect crystals, the conductivity will probably be overestimated.
NEMD does not have this limitation, in principle all scattering mechanisms are accounted for. However, in practice one might have to extrapolate to large large systems, to avoid finite size when the systems size is smaller than the phonon mean-free-path. This limitation concerns primarily calculations of bulk conductivity, for interfacial conductance it seems to be less critical.

In fact, I am going to calculate thermoelectric coefficeints like figure of merit(ZT), and I know I can realise this function using NEGF in this tutorials:  http://docs.quantumwise.com/tutorials/thermoelectrics_cnt_isotope.html,  can we use NEMD to realise this function?

NEMD is based on classical mechanics, that means it doesn't consider electrons, which is why the electronic conductivity cannot be calculated by this approach.
The thermoelectric coefficients plugin can therefore only be used with the NEGF method.
Alternatively, one could calculate the thermal conductivity via NEMD and the electronic contributions via NEGF and then manually insert the results in the equation to get the figure of  merit.

[wot19920302]

Thank you, sir! An excellent answer for this！

[wot19920302]

NEGF considers ballistic phonon transport and does not account for anharmonic effects, such as phonon-phonon-scattering, which means that in particular for perfect crystals, the conductivity will probably be overestimated.

     
Dear Julian Schneider:
          Sorry to bother you again.
          1. From your reply on differences between NEGF and NEMD, I learn that NEGF is suitable for ballistic transport, this theory just considers phonon-boundary scattering and ignores anharmonic effects (phonon-phonon iteraction), am I correct?.So how about defect(such as vacancy,doping,grain) and interface?  Can we use NEGF to get thermal conductance for a defect or interface system? 
          2. By learning this tutorials  http://docs.quantumwise.com/tutorials/phonon_calcs.html, I think the calculation of phonon transport requires more accurate optimization than the calculation of electron transport. A way to check if the relaxed configuration is suitable for phonon calculation is to observe phononbands (negetive bands or not). Is that right?
           3. I know that the calculation for phonon transport with ATK_DFT is time-consuming, so I wanna useＡTK_Classical，can I import some specific potentials to ATK? If your anwser is "yes", where I can find these  potentials?
            Best regards

[Julian Schneider]

         
1. From your reply on differences between NEGF and NEMD, I learn that NEGF is suitable for ballistic transport, this theory just considers phonon-boundary scattering and ignores anharmonic effects (phonon-phonon iteraction), am I correct?.So how about defect(such as vacancy,doping,grain) and interface?  Can we use NEGF to get thermal conductance for a defect or interface system? 

Yes, NEGF accounts for phonon scattering at defects and interfaces.

          2. By learning this tutorials  http://docs.quantumwise.com/tutorials/phonon_calcs.html, I think the calculation of phonon transport requires more accurate optimization than the calculation of electron transport. A way to check if the relaxed configuration is suitable for phonon calculation is to observe phononbands (negetive bands or not). Is that right?

Generally, before you calculate the phonons you should make sure that your system is optimized, i.e. in a local minimum.
If you encounter "negative" (i.e. imaginary) frequencies in the phonon bandstructure, this can either mean that your system is not properly optimized, or that you are at a saddle point. In the latter case you can try and making some small displacements to the system to break the symmetry and then optimize again. If that still does not help, be aware that, when calculating the dynamical matrix, a super-cell is generated (look for repetitions keyword in the DynamicalMatrix object and in the log), which means that even if the unit cell itself is at a minimum, there might be cases where super-cell is at a saddle point, which might require to use a repetition of the original unit-cell which can accomodate this relaxation. However, those cases are the exception and for devices, this concerns only the lateral repetitions. 

3. I know that the calculation for phonon transport with ATK_DFT is time-consuming, so I wanna useＡTK_Classical，can I import some specific potentials to ATK? If your anwser is "yes", where I can find these  potentials?

All the pre-defined potentials, that match the elements in your system, can be selected in the NewCalculator-widget, if you select ATK-Classical. If you have a potential, which is not available in this list, it can often be implemented "by hand" if the underlying potential functions are available ATK-Classical (see http://docs.quantumwise.com/tutorials/combining_potentials.html and http://www.quantumwise.com/documents/manuals/latest/ReferenceManual/index.html/ref.tremolox.html#ref.nanolanguage.potentialclasses).

[wot19920302]

thank you so much for yor your kindness,sir!    Best wishes to you!