Simulating Metal-Semiconductor Junctions in SILVACO TCAD

Discussion in 'Physics' started by RicUser, Dec 24, 2014.

  1. RicUser

    Thread Starter New Member

    Dec 24, 2014
    2
    1
    Hi,

    Sorry if it is too specific. I could not find a more specific group in the site. Would really appreciate if someone who has come across this already can find me a clue.

    I stumbled upon a strange result while playing with SILVACO TCAD, to simulate a metal semiconductor junction.

    I made a Metal-GaN structure (1D, nearly, along x,) [GaN thickness along x was 6.4 micron (just to keep consistency with an experimental result) and the electrode was 0.4 micron on either side of the sample. The dimension along Y was 100 microns.]

    GaN was doped to a level of 1E14

    There was no work function specified at electrode: manual calls this an ohmic contact
    Expected consequences:
    i) No charge transfer: The semiconductor and the metal electrode remain charge-neutral: no band bending.
    ii) Voltage should be zero in both of them


    During simulation I had to use "solve init" command, which forced the electrodes on either side to assume a voltage of zero volts. That should not change the situation, since the voltage should have been already zero in absence of any charge transfer.

    After simulation of the zero volt situation I found that there was a constant potential inside GaN (about +1.428 V) with respect to the metal electrode.

    The doubt is: If there are no net charges in semiconductor & electrode, why should there be any potential in either of them?

    Manual [version release date: Feb 2012] says this voltage is the actual electrostatic potential as in Poisson’s equation [‘psi ’ in page 102, referring to eqn. 3.1 in Page 96: div (epsilon*grad (psi)) =-row]. I was wondering where is this potential coming from?

    The magnitude of the potential happens to be the same as the Fermi energy shift [divided by the electronic charge] due to doping in GaN. Is it possible that the doping-induced change in Fermi level is interpreted as due to an increase in voltage? In that case, when we require the true electrostatic potential, should we subtract the doping induced Fermi level shift from the simulated value of the potential?

    Please let me know if you have come across this. I am attaching the image of the str file and the potential cutline [horizontally through the middle of the sample, 50 micron deep, from the top].

    If you know of an existing thread on this, please do let me know.

    Thanks in advance
     
    Last edited: Dec 24, 2014
  2. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,692
    2,756
    Wow. And this is why I am an applications engineer...

    We have some brilliant people here. Maybe someone knows the answer.
     
  3. Monty_786

    New Member

    Feb 17, 2015
    2
    0
    Hi Ric, I was actually looking for the same problem on the internet. But no luck.
    Can someone help me identify the error. Ric, please reply if you've got your answer. My whole project is stuck on this problem.
     
  4. RicUser

    Thread Starter New Member

    Dec 24, 2014
    2
    1
    Hi, Monty,

    Sorry for the delay. I am also sorry that I am still looking for the answer.

    As I wrote in my previous post, the potential difference between the metal and the doped semiconductor is actually equal to the Fermi level difference between the un-doped semiconductor and the doped semiconductor (divided by the electronic charge, of course). Looks like, whenever we do not specify any work-function in the metal, the software automatically assigns an work-function to the metal which is equal to that of the semiconductor, but in un-doped state. The doping induced shift in the Fermi level (and in turn the work-function) of the semiconductor is not considered, and according to my understanding, it appears as a voltage drop across the interface.

    Until I am more certain about the software, I am going to live with this, and subtract the Fermi-level induced voltage difference from the actual one if at all it is needed. Also, I would, in most cases specify some value for the work-function at all contacts. If Ohmic contact is desired, I will either use a small value of the work-function, or use a local high doping in the semiconductor around the Ohmic electrode/semiconductor interface.

    I do not know if it helped. Please keep me posted if you come across some more clues. Thanks and once again apologies for the delay in replying.
     
  5. Monty_786

    New Member

    Feb 17, 2015
    2
    0
    Thanx for replying on the thread and for your help. Silvaco does it on purpose to ensure the bandgap narrowing model.
     
  6. zeeshan002

    New Member

    Apr 18, 2015
    1
    0
    Until I am more certain about the software, I am going to live with this, and subtract the Fermi-level induced voltage difference from the actual one if at all it is needed. Also, I would, in most cases specify some value for the work-function at all contacts. If Ohmic contact is desired, I will either use a small value of the work-function, or use a local high doping in the semiconductor around the Ohmic electrode/semiconductor interface.

    _____________________
    aliiii
     
Loading...