I-V curves for 2D materials

[BarneyCavs]

Hi,

I am currently doing some benchmarks for the adsorption behavior of 2D materials. According to the experiment and previous theoretical calculations, normally there is a increased resistance (or reduced current values) after the adsorption of gas molecules on the surface of 2D materials.

But for my results, there is barely a difference before and after the adsorption. Could anyone please tell what is possibly going wrong?

Thanks a lot.

[Petr Khomyakov]

There are a couple of issues with your calculations.

1. Two of the three electron transport calculations are done for perfect, infinite system. For that, please have a look at my replies to this forum discussion, http://quantumwise.com/forum/index.php?topic=4970.msg21537#msg21537.

2. One of your calculations assumes a scatter (molecule) on top of the 2D material in the central region. This molecule seems to be quite decoupled from the substrate, so it is not surprising that you see no difference in electron transmission, compared to perfect system calculations. I guess in your case the current essentially goes through the substrate. Since the current is carried by charge carriers near the Fermi level, the electronic structure of your material at and near the Fermi energy should be altered by the molecule adsorbed on the surface in order to affect the electron transmission through the central region. Apparently, it is not the case for the system as you have designed it. 

[BarneyCavs]

Thanks a lot, Petr.

1. I checked your previous post on the perfect, infinite system, I guess there is no problem if we calculate the transmission and I-V curve under zero bias rather than a finite bias. In my case, do I need to change the BP electrodes to metallic electrodes?

2. The adsorption configuration is the most stable configuration determined by DFT calculations, and indeed, there is no obvious modification to the DOS near the Fermi level. In the previous theoretical calculation about phosphorene, https://arxiv.org/ftp/arxiv/papers/1406/1406.2670.pdf, the NH3 molecule does not affect the substrate (only some minor modification on energy around -2 eV) but still a large reduction in the current values. In their simulation, the scattering region is even smaller and the density of molecules is higher than my simulation, and this cannot be achieved in VNL-ATK, because the length of electrode region cannot surpass that of scattering region. They used TranSiesta package instead.

I want to compare the resistance or conductance of the system before and after the adsorption (under zero bias), to illustrate the sensitivity of the substrate. So, if there any methods that I can do to increase the discrepancies for I-V curve before and after adsorption, such as change to a metallic electrodes, increase the density of molecules, or I just can't?

Thanks for your help.

[Petr Khomyakov]

- Adding metallic electrodes will make your calculation more realistic, but computationally more expensive. On the other hand, if you are just interested in seeing a difference between the electron transmission with and without a scatter (molecule), your setup might still be good enough for that purpose.

- VNL-ATK and TranSiesta are based on the same methodology. Using short electrode and electrode extension regions might give rise to convergence problems as well as to incorrect electron transmission. In the paper, they have one molecule in-between two electrodes, so the transmission should converge to a certain value upon increasing the size of the left and right electrode extensions.  If they used too short electrode extensions, the result could be wrong.

Did you mean the density of molecules defined with respect to lateral direction (perpendicular to the transport direction)? Otherwise, the density of molecules does not depend on the size of the central region.

In the paper, I do not see any strong effect of the molecule adsorption on the current, at least if I am looking at Figure 4.

I guess the idea is not to induce the discrepancy, but to figure out what happens in reality, meaning that you should build a reliable (but still simple enough)  model that reflects experimental conditions.   

 

[Anders Blom]

... In their simulation, the scattering region is even smaller and the density of molecules is higher than my simulation, and this cannot be achieved in VNL-ATK, because the length of electrode region cannot surpass that of scattering region ...

I think here you are mixing up the actual electrodes (repeated, semi-infinitely to either side) and the "electrode extensions", i.e. the part of the central region that looks the same as the electrodes. The confusion might come from the way TranSIESTA handles the definition of the electrodes compared to ATK, although I'm not sure since I'm not very familiar with that code.

Nevertheless, in either code, the electrodes should be large enough to capture all interactions, but making them longer than that has no effect, due to the finite range of the basis sets. The length of the electrode extension, on the other hand, is a critical parameter to converge, since this is where the effect of the "scattering region" (the part of the central region that does not look geometrically the same as the electrodes) is screened. ATK certainly allows you to make this as long as you need - we routinely run simulations with 400 Å central regions.

[BarneyCavs]

Thanks a lot, Petr and Anders.

Yes, the central region includes the left and right electrode extensions, and also another concept is the screening length.

I got to know the VNL-ATK uses the ballistic approximation to do the transport calculation, so the length along the transport direction does not affect so much to the computational workload. I simply want to compare the I-V curves, specifically the different responses for current before and after the adsorption. As you can seen in Figure 4 the paper above, there is a discrepancy (although it might not be large) for the current values after the adsorption of NH3, but barely no difference observed in my simulation results.

Based on your answer, my setup for the simulation seems okay, and in this way, I wonder if I could enlarge the differences, i.e., increase the molecule density in the transport direction (Z) or periodic direction (B) (suppose A is the vacuum)? Because in the experiment, there is a large reduction in the resistance, for this particular molecule is more than 50% reduction.

[Anders Blom]

Sure, try that Just add more molecules, but in case you do so and at the same time make the cell larger in the periodic direction, then nothing changes.

[BarneyCavs]

Sure, try that Just add more molecules, but in case you do so and at the same time make the cell larger in the periodic direction, then nothing changes.

Sorry, I don't quite understand. Do you mean nothing will change if we increase the density, here the density I mean, is to increase the number of molecules on the surface of same area rather than repeat the supercell along one direction.

Thanks.

[Anders Blom]

If you make the system 2x larger in the periodic direction and add one more molecule, that's the same density. So I'm just saying you need to add more molecules in the same system, without making it larger, if you want a bigger effect.

[BarneyCavs]

Thanks. That is what exactly I mean.

[BarneyCavs]

If you make the system 2x larger in the periodic direction and add one more molecule, that's the same density. So I'm just saying you need to add more molecules in the same system, without making it larger, if you want a bigger effect.

And one more question, back to the answer of Petr, if the molecule does not affect the DOS near the Fermi level, we do not expect a large change in the I-V curve, right? But for the example I mentioned above, NH3 only has minor modification around -2 eV and still the calculation results demonstrate a obvious change in the I-V curve before and after the adsorption.

Thank you.

[Anders Blom]

The DOS is just part of the story, of course. With the molecule you increase scattering, and that is not captured directly in the DOS. But the difference is small, which I guess is a reflection of the fact as you say that there is no strong DOS for the molecule close to the Fermi level.

[BarneyCavs]

Yes, from the DOS, I think my calculation result is right, because I get the similar DOS as in the reference, a minor modification around -2 eV. And for the transport calculation, my simulation is also along the armchair direction, there is barely difference for the I-V curve (as attached), while a obvious effect for adsorption in the reference.

So, I am really confused. Is it because we are studying a totally different molecule or something is go wrong with our simulation?

[Anders Blom]

Of course a different molecule can have a different effect (or just the position of the molecule on the surface), but besides that one needs to analyze the problem more carefully to dig deeper into the results.

[BarneyCavs]

Thanks a lot, Anders. I will check whether the density has a effect on the I-V characteristic.