QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: bubble on May 20, 2015, 14:01

Title: spin-dependent Seeback coefficients
Post by: bubble on May 20, 2015, 14:01
Hi all,
      If I perform a spin-polarized calculation for a magnetic device, can I get the spin-dependent Seeback coefficients? It seems that ATK can only give the total Seeback coefficients through the 'Thermoelectronic Coefficients' plugin in VNL.
Thank you in advance.
Title: Re: spin-dependent Seeback coefficients
Post by: Jess Wellendorff on May 20, 2015, 14:56
You are right, VNL currently calculates the Seebeck coefficient from the total electronic transmission spectrum (the sum of spin-up and -down spectra). So no, you are currently not able to get spin-resolved Seebeck coefficients.

- Jess
Title: Re: spin-dependent Seeback coefficients
Post by: Anders Blom on May 21, 2015, 19:51
... in the plugin.

But you can of course always evaluate it yourself from the spin-dependent transmission spectrum.
Title: Re: spin-dependent Seeback coefficients
Post by: bubble on May 24, 2015, 14:09
Quote from: Anders Blom on May 21, 2015, 19:51
... in the plugin.

But you can of course always evaluate it yourself from the spin-dependent transmission spectrum.

Thank you for your reply, Blom.
However, can you please say it more clearly? Where can i get and install the plugin. It seems that the plugin is not in the "Addon manager".
Thank you in advance.
Title: Re: spin-dependent Seeback coefficients
Post by: Anders Blom on May 24, 2015, 19:26
I assumed you were already familiar with our tutorial on computing the thermoelectric coefficients?
http://quantumwise.com/documents/tutorials/latest/Phonon/index.html/chap.thermoelectric.html
Title: Re: spin-dependent Seeback coefficients
Post by: bubble on May 25, 2015, 03:48
Quote from: Anders Blom on May 24, 2015, 19:26
I assumed you were already familiar with our tutorial on computing the thermoelectric coefficients?
http://quantumwise.com/documents/tutorials/latest/Phonon/index.html/chap.thermoelectric.html
Yes. However, what i want to know is the "spin-dependent" Seeback coefficients.
How can i seperate the spin-up and -down transmission spetra in one file in VNL? So we can calculate the spin-dependent coeffients seperately. are there some plugins?
Thank you!
Title: Re: spin-dependent Seeback coefficients
Post by: Jess Wellendorff on May 25, 2015, 12:25
The attached script shows how one should be able to calculate the Seebeck coefficient for each spin channel. You'll need an nc-file containing the electronic TransmissionSpectrum. The script has been only moderately tested, but I think it should work.
Title: Re: spin-dependent Seeback coefficients
Post by: bubble on May 25, 2015, 15:55
Quote from: Jess Wellendorff on May 25, 2015, 12:25
The attached script shows how one should be able to calculate the Seebeck coefficient for each spin channel. You'll need an nc-file containing the electronic TransmissionSpectrum. The script has been only moderately tested, but I think it should work.

Thank you for your help, Jess.
I will test it. If I have questions, I will contact you. ;)
Title: Re: spin-dependent Seeback coefficients
Post by: bubble on May 26, 2015, 01:54
Thank you for your help, Jess. :D
As an example, I performed a spin-polarized calculations on a carbon nanoribbons with LSDA (the obtained transsmission spetra are therefore symmetric about spin-up and -down components, and the spin seebeck coefficent (SSC) should be the same for both spins).
I tested the 'seebeck.py' file that you provide, and found two questions:
Q1. Indeed, the spin-up and spin-down SSC are the same. I also calculated the total seebeck coefficent through 'thermalelectric coefficent', however, I found that the three values, spin-up, spin-down and the total, are the same. i.e. S1=S2=S. I think S should be equal to S1+S2?
Q2. If I want to get the relationship about S1~T, S2~T (T is the absolute temperature), what shall I do? (I understand that the default temperature is the electron temperature of two electrodes, for example, both are 300 K. How to get S1 and S2 under other temperatures?)
Thank you in advance!
Title: Re: spin-dependent Seeback coefficients
Post by: Jess Wellendorff on May 26, 2015, 10:55
Q1) The Thermoelectric Coefficients plugin gives you the average Seebeck coefficient (averaged over both spin channels). You therefore have the relation S = (S1+S2)/2.
Q2) To get the temperature dependence, the VNL plugin offers you the option of setting the temperature (300 K is default). You can do the same with the script: 300 K is default, but you can set the temperature like this:
*******
from NL.CommonConcepts.PhysicalQuantity import Kelvin
T = 400*Kelvin
S = evaluateThermoelectricTransportCoefficients(transmission_spectrum, spin, temperature=T)
*******
Title: Re: spin-dependent Seeback coefficients
Post by: bubble on May 27, 2015, 03:46
Hi, Jess, thank you very much for your reply.
As a further estimation, I calculated a spin polarized model according to the tutorial (i.e. http://www.quantumwise.com/publications/tutorials/item/112-transmission-spectrum-of-a-spin-polarized-atomic-chain)
For the parallel case, the spin-polarized spetra were well re-produced.
However, when I caculated the spin-seeback coefficients and total seeback coefficients of the system at 300K, the results are:

*************
spin=1 : Seebeck coefficient=-1.530e-06 V/K
spin=2 : Seebeck coefficient=1.236e-03 V/K
*************
Peltier Coefficient   -0.0004587   V
Seebeck Coefficient   -1.529e-06   V/K
Thermal Conductance (electrons)   8.206e-10   W/K
Thermal Conductance (phonons)   -   W/K
Thermal Conductance (total)   -   W/K
ZT   -
*************
I have two questions:

Q1.  It seems that S2 is larger than S1 by about three orders of magnitude (absolute values). however, the total S is only closed to S1. According to your above reply, I think S=(S1+S2)/2=S2 rather than S1. Is the result right?

Q2. This question is another independent question.
       I further caculated the spin-polarized transsimisson spetra of the sytems at finite bias (Vb=0.2 V) . How to get the spin-polarized current?  Through the Tool-Custom Analyzer-IVcurve, it seems that only total current can be obtained, although the  transsimisson spetra are spin polarized .

**********The results from IV Curve Plugin*****************
+----------------------------------------------------------+
| Current Report                                           |
+----------------------------------------------------------+
| Applied bias        = -0.200 Volt                        |
| Bias window         = [0.000 Volt, 0.200 Volt]           |
| Bias window covered = 100.00 %                           |
| Calculated current  = -1.51330794906e-05 A               |
| Spin                = Sum                                |
+----------------------------------------------------------+

Data for interpolation
Voltage list (Volt)
[-0.2]
Current list (nano Ampere)
[-15133.07949056]
**********************************************

In addition, the nc files are large (20-30 MB), so I can not upload them. But the script is uploded. See next!
Thank you in advance!
Title: Re: spin-dependent Seeback coefficients
Post by: bubble on May 27, 2015, 12:52
Here is the script
Title: Re: spin-dependent Seeback coefficients
Post by: Jess Wellendorff on May 27, 2015, 15:55
Q1
OK, I agree that the script I made does not quite cut it.
So I had a closer look at the theory of calculating the Seebeck coefficient (http://arxiv.org/pdf/1107.5687.pdf, page 3, top part of left column), and realized I had made an error :(
This should be fixed in the attached script. My output from the carbon chain transmission spectrum is:
**********
spin=1 : Seebeck coefficient = -1.530e-06 V/K
spin=2 : Seebeck coefficient = +7.274e-10 V/K
      total Seebeck coefficient = -1.529e-06 V/K
**********
The sum of the two spin contributions ("total") now matches the VNL result from the Thermoelectric Coefficients plugin.

Q2
For the IV curve plotting, I recommend you use the "IV-Generator" and "IV-Plot..." plugins (available in the right-hand panel next to LabFloor) instead of the Custom Analyzers.
Recipe:

Best,
Jess
Title: Re: spin-dependent Seeback coefficients
Post by: wot19920302 on October 7, 2016, 05:30
dear Jess Wellendorff ：
        I inspect the script you offered above and can't understand this block: ???
   
Code
 t1 = transmission_spectrum.evaluate(spin=spin)
                 t1 = SplineInterpolation1D(E, t1)
                 t2 = transmission_spectrum.evaluate(spin=Spin.Up)+transmission_spectrum.evaluate(spin=Spin.Down)
                 t2 = SplineInterpolation1D(E, t2)

          It seems that you think spin-resolved seebeck coefficient should be calculated by 1/eT*(K1/K0), where K1 contains single spin transmission spectrum and K0 contains the sum of transmission of spin.up and spin.down. But in this paper:  Spin-dependent Seebeck effects in graphene-based molecular junctions (PHYSICAL REVIEW B 93, 195426 (2016)
),K1 and K0 seems only contains one type spin(up or down), do I misunderstand the script you offered?
                Best
Title: Re: spin-dependent Seeback coefficients
Post by: Jess Wellendorff on October 13, 2016, 11:47
This is an interesting question, and we are looking into it.
Title: Re: spin-dependent Seeback coefficients
Post by: Petr Khomyakov on November 22, 2016, 14:35
The spin-dependent Seebeck coefficient for a given spin channel is defined in terms of K0 and K1 calculated for this spin channel, which corresponds to either spin_up or spin_down transmission spectrum. This is now corrected in the 'seebeck_coefficient.py' script enclosed.

The total Seebeck coefficient is not defined correctly in the PRB 2016 paper cited in a previous post. It is neither (S_up + S_down) nor (S_up + S_down)/2 as defined in the PRB paper. The total thermopower is given as  (conductance_up S_up + conductance_down S_down)/(conductance_up + conductance_down), see two slides (in the enclosed zip file) with a formal derivation of this formula. These slides also contain a complete set of equations for calculating S, S_up, S_down, conductance_up, conductance_down as well as useful references used to derive these physical quantities.

Enclosed you can find an updated python script that calculates the spin-dependent and total Seebeck coefficient as well as the conductance for spin up and down channels. An important update is that the definition of the Seebeck coefficient sign has been corrected; it is now S = -K1/(K0*e*T), where the elemental charge e>0 (positive). This definition of the thermopower is now consistent with the definition given in a textbook on Solid State Physics by Ashcroft & Mermin, see Chapter 13.

To test the script, one can use an nc-file (Device_New_Configuration.nc in the zip-file) with the spin-dependent transmission spectrum calculated for a 1D carbon chain with parallel magnetization configuration of the electrodes. The description of how to run the script is given in the script header.

NOTE: we are working on updating the thermoelectrics plugin in the VNL to fix the sign problem in the definition of the Seebeck coefficient. 



 

 
Title: Re: spin-dependent Seeback coefficients
Post by: wot19920302 on November 23, 2016, 15:11
thanks for your work！ ;D
Title: Re: spin-dependent Seeback coefficients
Post by: Petr Khomyakov on November 24, 2016, 10:47
UPDATE: I have added a Device_New_Configuration.py to the zip-file in case someone wants to redo the calculation for the carbon chain, and then run the seebeck_coefficient.py with the corresponding Device_New_Configuration.nc file generated.  I note that the Device_New_Configuration.nc in the updated zip-file has been generated using the ATK2016.3 version.