QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: yangzw1985 on February 18, 2009, 10:34
-
Hi, everyone!
I want to calculate the transimission spectrum for a li-zno-li two probe system. But all of the coefficients are zero when I set 200 energies and the points from -12 to 12 !! when I set the energies 5000 and points from -100-200, some results is not zero appear in the area 150-200 points, but it was very small.
The parameters for the method menu are as follows:szp,2x2x50 for brillouin zone,mesh cut off 100Ry,LDA.PZ, Tolerance 1e-004,100 points and 100Ry cut off for the two center integral. Is there anything wrong?
Can anyone give me some suggestions on " how should I set the parameters in order to get the right transimission spectrum?".
Thanks very much！
-
It appears your system has converged to zero charge, if the transmission is zero for all energies.
I can try to help your own, but I need a geometry or a hint of the geometry to try out a few things....
Is it a wire or ?
-
Do you by any chance have the log output for the selfconsistent part of the calculation?
-
Attached geometry structures have been optimized. Distance between the electrodes and zno nanowire was determined by calculating the total energy to min.
-
Hi,Nordland, Attached is the liznoliclusterwur72(1V) script and results. I think this may be useful ! Is there something inappropriate ?
Thank you!
-
I only have one computer and along queue of fun calculation, but it is grinding at the problem, so I dont have a final answer for you,
but I got an idea what went wrong for you, since the contour parameters was perhaps not conservative enough you might have lost all
the electrons of the system, and hence there is no transmission for any energy.
As soon as my computer has ended the calculation, and converged to something meaningful I will report it here :)
-
Hi,Nordland!I really need your help to check out this problem.I want to know what is the reason for the results? have you got enough time to solve this problem?
-
Hi,Nordland! If you have checked out the reasons for this problem,please tell me as soon as possible. I really need your help to solve the problem?
thanks!
-
Hey Yang.
I will try to take a look at it, as soon as my calculations on graphene ribbons ends ;)
-
Thank you!
-
Hi,Nordland!
Have you finished your calculation on graphene ribbons? I want to know when you will free to help to check out this problem. Because I have tried lots of methods to solve this problem, but the results are unreasonable. I really need to solve this problem as quickly as possible, I hope you can help me to change the situations .
Thank you very much!
I have been looking forward to hearing your good news!
-
Okay, I am trying to do a couple of runs right now.
You should not change the default values for the twocenter integral values. It only leads to bad behavior.
So those I have modifed back to default values.
You should also always try zero bias first, since it is more stable and faster to obtain a good idea about which parameters to use for this system.
I have also tried to double the electrode length, and made some minor modifications and tunning of the parameters.
I will post once the calculations are done.
-
The calculation is taking quite long!
I have attached the script here for you to try it - I have modified parameters into somewhat more reasonable values.
-
Hi,Nordland!
I have tried the calculation with different parameters, attached is the calculated results!
As we can see from the log out file, if the results are reasonable, the Fermi Energy = -0.39002 Ry for left and right electrodes. For the two probe model,the Fermi Energy = -0.34528 Ry .
The transmission coefficient is 0 from -12 to -2.23. This means the the transmission coefficient is 0 for the fermi level. How can I explain this?
And in the bias window [-1,1], is the transmission coefficient right?
what's more, I also can not distinguish where is the electrode and where is the nanowire.If it does not contain the electrode, why? I have read lots of papers, and the transmission spectrs shoud be divided in to three parts:left electrode, center scattering region, right electrode.
Thank you !
-
Trying to answer the questions as best as I can.
1) When you calculate the transmission spectrum, the transmission spectrum is shifted, such that the E=0 corresponds to the fermi level (-0.34528 Ry in your case)
2) In your case the transmission coefficient is roughly 0.05 - 0.10 around the fermi level, hence there is transport at the fermi level.
3) If the transmission does not look unphysical, so the transmission spectrum is properly correct. However it should be noted, that it is not exactly a bias window you are looking at when you are looking at the transmission spectrum. It is the transmission probability for a given energy. It is true that this can be related to NON-responsive current calculation.
4) I might be wrong, but I have never seen a transmission spectrum divided into spatial regions, as it is not part of the standard formulation of the problem.
Do you have reference for a paper where this division is done?
-
Hi,Nordland! In your opinion, the results is right, ok?
you can have a look at this papers: physical review B 71,165316 (2005) and nano letters 2008,8(11),pp 3662-3667.
Thank you !
-
I have looked through the papers, but as I can see, they do not divide the transmission spectrum into spatial regions. Is it a specific figure you are talking about?
-
In physical review B 71,165316 (2005), is figure 4, 6, 7. and in nano letters 2008,8(11),pp 3662-3667 is figure 3.
-
As we can see from my calculations, the transmission spectra is unsymmetrical. If I choose -6 to 6 as the X axis, in the bias window, there will be lots of values of zero. I don't know the reason, and I also can not decide the results is right or not!
In that two papers, the transmission specrta are basically symmetry when the x axis is zero.
I don't know the reason.
-
As we can see from my calculations, the transmission spectra is unsymmetrical.
In general all transmission spectra is unsymmetrical, it is only in very special cases that is symmetrical.
The only case I have ever seen, is in a nearest neighbour tightbinding approximation for carbon nanotubes, which simple overlap model. The transmission spectrum can only be expected to be symmetrical, if the band structure is symmetrical around the fermi level, and the system is homogenious and has no spatial variation in structure. Since your system has a spatial varation with ZnO in the middle,
it is the most natural thing, that the transmission spectrum is unsymmetrical.
Look at this simple system in the manual (link) (http://www.quantumwise.com/documents/manuals/ATK-2008.10/chap.litwoprobe.html#sect2.litwoprobe.spectrum).
You can look at transmission spectrum which is perfect and unsymmetrical.
If I choose -6 to 6 as the X axis, in the bias window, there will be lots of values of zero. I don't know the reason, and I also can not decide the results is right or not!
The transmission spectrum is not a bias window. The transmission spectrum is plot of the probability T of a electron with a certain energy E in the electrode being transmitted through the scattering region.
So if the transmission has a value 0.325 at the E=-3.0 eV, it means that a incident electron with an energy for -3.0 eV relative to the fermi level, have a probability of 0.325 of being transmitted through the scattering region.
If the value is zero in the transmission spectrum for a given energy (fx E= -7 eV) it means one of the two following things:
1) Either there is no electrons ( no allowed eigenstates ) in the electrode with this energy relative to the fermi level.
2) There is electrons in the electrode with this energy relative to fermi level, but the electrons at this energy has no chance for being transmitted through the device, therefore there is a transmission probability of zero.
Following up the note about the bias window - it is important to understand that the transmission spectrum is only related to the bias/current, that you can get the current from integrating the transmission spectrum from -V/2 to V/2 if you have bias of V taking into account the fermi distribution of the two electrodes
Therefore if you have a bias of 6 Volt, then you must integrate the transmission spectrum from -3 eV to 3 eV weight the fermi function. In order to be correct, the calculation must then be done at 6.0 V bias as well.
In your case the transmission spectrum could indicate that the the current will grow linearly with the bias until a certain thresshold, where it would enter a maximum current until the bias becomes very high.
Therefore
-
Having look through the paper, I might have found the cause of confusion.
In Physical Review B 71,165316, there is a figure 2 - this is NOT a transmission spectrum, but an IV curve.
Figure 3 is a transmission spectrum, just like the one you have calculated.
-
Thank you, Nordland! Thanks for your answers and suggestions!
Good job!
When you have finished your calculation on the transmission spectra, please post here to make a comparision!
Thanks again!
-
Hi, everyone! I want to know something about the applied voltage.
Now, I would like to calculate the current and the transmission spectra of the two probe system. I have calculated it with the applied voltages of 1,2,3,4,5,6V. Now I was wondering whether the increase of voltage is too large.
If I perform the calculated with the applied voltage of 0.5,1,1.5,2,2.5,3V. Would it be more reasonable? If not, please give me the details.
Thanks!
-
Well there are two side of it - one is the convergence, the other is the sampling of your IV curve.
If your system always converges, and you are not interested in the details of the IV curve, then go for 1,2,3,4,5,6.
However if you are interested in smaller features of the IV curve, or you having convergence issues, then I would go for 0.5, 1.0, 1.5....
-
Hi,Nordland! I heve try to perform the calculation, now there is a problem occured.
Under the applied voltage of 1V-6V,the calculation convergenced normally. But under the applied voltaged of 0V, the calculation have been spending a long time, and not finished. while the SC is not zero, and the calculation is still running.
In my opinion, under the applied voltage of 0V, it should be cost short time to finished the calculation. What is the reason for the long time?
-
Hmmm.... what is the converge charge of the other biases?
-
Hi,Nordland,attached is the the running log out window. The SC have reached 280, is it meanfulless to continute this calculation? Under the applied voltages, the calculation convergenced when the SC reached 60-80, with q=1000e-1200e.I don't know the reason.
-
Strange! I think you should kill it and try again with a slightly higher diagonal mixing parameter.
-
Hi,Nordland! I have performed the calculation with diagonal mixing parameter as 0.1, 0.2, 0.5, But it seems that the question is still! I will give it more time to continue the calculation, and if you have find out any other suggestions, please tell me as soon as possible!Thanks!
-
I am sure that the questions is still, because the SC for all of diagonal mixing parameter as 0.1, 0.2 0.5 have larger than that of 150, and the calculation is still running. I do not know the reason. Under the applied voltage of 0V, it should be easy to convergence. But the calculation is abnormal!
-
Agreed that something is fishy - it should be the most easy to converge it under 0 V bias. I have never seen this before, where a system could converge under finite bias, but not under zero bias.
-
But the question is just like that!!
I also can not find out the reason. All of my two probe model was made by the same methods.
Now, there is something strange occured.
The zno nanostructure with 72 and 144 atoms convergenced normally under applied voltages of 0 V. While the zno nanostructure with 48, 96 and 120 atoms can not convergence under 0V.
I am really confusing!!
what should I do to solve the problem?