Transiesta calsulation not converging due to 'charge'

Dear SIESTA users,

I am currently working on finding zero-bias transmission using TranSIESTA. My calculations are converged but the output is showing not converged. I guess it is due to ‘dq’ (charge not being converged). What can be the possible solution or source of error for charge not converging.
THe error part of the output is as follows:

SCF Convergence by DM+H criterion
max |DM_out - DM_in| : 0.0000506445
max |H_out - H_in| (eV) : 0.0000649631
SCF cycle converged after 616 iterations
SCF_NOT_CONV: SCF did not converge in maximum number of steps (required).
Geom step, scf iteration, dmax: 0 616 0.000005
Geom step, scf iteration, dq: 0 616 7.437410

Shubham Tygai
Ph.D. Scholar

When completing the SCF step, it misses 7.43 electrons in the device.

There could be many reasons. One, often encountered, reason could be that you don’t have enough screening layers towards your electrodes.

Thank you for the reply. As per your suggestion, I increased the electrode region. I am working on HfSSe and for the time being I just interested in a simple I-V i.e. my electrode and channel are the same (HfSSe throughout). With 45 (about 1825 in length*breadth) supercell as the electrode, I am still having the same issue. I think this much should be sufficient for electrodes to act as bulk. As you said there can be many reasons, what can be the other reasons?
Also, I am currently trying by defining a certain buffer region. But since I have HfSSe throughout, I am not sure if this would help.

Thank you

2 things.

  1. Please don’t do bias calculations on pristine systems. I don’t think it has any physical relevance.

  2. TranSiesta and many other NEGF schemes are based on metallic electrodes (not semi-conductors). To use semi-conducting electrodes requires much more delicate use.

The above are some of those many reasons :slight_smile:

Thank you for the information.

I wanted to study the change in current value with gas adsorption. Most of the papers report current with and without gas adsorption to make a comparison. That was what I was also trying to do. Is there any other way that you can suggest?

Also, even if using a semiconductor electrode, the change in charge should be zero. Like how can a system be short of electrons? Can we correct it or instruct siesta on how to deal with it?

While there are reports of doing bias (I-V) characteristics for bulk systems I don’t find them particularly valuable. The potential drop is highly fictious. I.e. one cannot experimentally create a system where you’ll have the potential drop over the pristine sample.

As for semi-conducting electrodes. The real problem for SC electrodes is 1) the Fermi level is not well-defined, and 2) the screening length for the SC material is excessively long meaning that ones device system should be several nm, close to um long.