Dear siesta users
I am using transiesta code to get the I-V characteristics of Armchair MoS2 nanoribbon as scattering (with 8 unit cells) and electrode material. All the calculations (electrode, transiesta and tbtrans) run perfectly for the voltage range of 0-2V with good SCF convergence. However, in all voltages I get nearly zero transmission in the bias window and consequently very low current (about 10^-16 A).
Since my armchair MoS2 NR is a semiconductor with the bandgap of about 0.5 eV I expected that current should start to increase exponentially from 0.5 V bias and reach the order of micro Ampere. I have attached .fdf files of my scattering and electrodes.
I would appreciate if someone could help me to find the problem.
Best regards,
Samaneh Soleimani-Amiri
You can't reliably trust NEGF results for semi-conductors.
The reason is quite simple.
The NEGF method relies on "bulk-like" electrodes. And for semi-conductors the relaxation length is much longer than any typical scattering/device region. I.e. to fully relax the semi-conductor electrostatic potential one would need more than 20 Ang of electrode (and perhaps more!) material to properly screen off the device.
Secondly, if this is a pristine system, i.e. electrode and device being the same material, then I don't think applying a bias makes physical sense.
How would you experimentally apply a bias between two semi-infinite materials in some very small region of exactly the same material? So please don't do bias calculations on pristine systems. :)
Nick Rübner Papior
Dear Nick,
I did the same calculations in siesta for pristine AGNR with Eg=0.2 (with the same material contact) and got the expected answer. I mean the current was negligible before 0.2 and increased dramatically beyond 0.2 V. I thought that when we define a contact in siesta it extend it as a long contant to avoid screening. Now the question is: How can I analyze the transport in pristine armchair MoS2 nanoribbon as a semiconductor in siesta? I mean how can I define the suitable contact for armchair MoS2 nanoribbon to avoid screeining effect and get the expected results as I got in AGNR?
I look forward to hearing from you.
Samaneh Soleimani-Amiri
First; it is still non-physical to create a pristine system with an applied bias. TranSiesta has no way of screening off the device potential drop in the electrodes and thus it becomes non-physical, even for metals this is non-physical. Sure, you get something that is easily explained in terms of band-gaps and how the electronic structure looks. But generally the method is based on fully bulk electrodes. This is in contrast to pristine systems with applied bias. Again, you can't screen off the potential drop.
Secondly, you can still perfectly do EGF (V=0 NEGF) for pristine systems. However, transport properties of pristine systems are quite "boring" since you simply count the bands crossing at each energy. Effectively you'll get a step function for the transmission curves and then perform the k-averaging. To do this in tbtrans I would simply do an electrode calculation, then use https://github.com/zerothi/sisl to tile the structure 3 times and then do transport properties, this will work for metals and semiconductors since the Fermi-level is well-defined. However, it is not guaranteed that this will work if you use semi-conducting electrodes and a transiesta calculation. Since the Fermi-level may be arbitrarily determined in the band-gap you may end up with band-edge mis-alignments which are unphysical.
But if you use sisl, then this won't be a problem since everything comes from the same electronic structure calculation.
@ Nick Rübner Papior
Thank you. Do you use https://github.com/zerothi/sisl only for V=0 or you use it for the voltage range?(for example 0-1 V)?
- Badges
- Science topic
- Similar topics
- Condensed Matter Physics
- Zero