Dear Researchers,
I getting a convergence problem while simulating InGaN/GaN multiple quantum well solar cell. How can I solve it?
Note: The problem occur when I add quantum well region, without quantum well region the code runs fine.
Also I have adjusted the mesh so there is no problem in messing I guess,
additionally I use GUMMEL and NEWTON numerical technique. Any suggestion how to remove the convergence issue?
UPDATE:
When I remover calculated valued of permittivity and affinity from the code there is no convergence.
But adding this value gives rise to convergence. why is that?
Thank you
Dear Dhanu,
It is clear that the inclusion of the quantum well caused the problem of convergence.
So, i would propose that you solve for the quantum well alone with simplified boundary conditions. Then you will get feeling how to adjust the simulation parameters of the well to get fast convergence.
It seems that the device simulator solves the quantum mechanical equations in the well region while solving the semicondcutor equations in the other pin device region. You can also, treat the well from the macroscopic point of view after undermining its macroscopic parameters.
You can also try to change the dimensions of the well , it may relax the problem and leads to convergence.
My opinion is to make physical changes to facilitate the mathematical problem!!!
Best wishes
Hi Dhanu,
What is the convergence problem like? Do you get this message?
Warning: Convergence problem. Taking smaller bias
step(s). Bias step reduced 1 times.
When you say you add the quantum well region, do you mean adding QWELL flag to that layer when it exists already? Or just adding the whole layer causes the issue?
If it is a carrier transport issue (maybe some default parameter not set correctly?) then why not try very low In content first to see if the convergence issue is still present?
Alex
Dear Alex,
Here is the problem,
Warning: Solution algorithm cannot reduce residual(s).
Warning: Convergence problem. Taking smaller bias
step(s). Bias step reduced 10 times.
The problem occurs when I add new region of quantum well and barrier. In my earlier InGaN/GaN PIN structure.
I am taking 0.02 In composition (low)
Also, when I add QWELL in the well region, the simulation takes 5 days to complete and at the end it still gives me the same convergence issue.
Regards,
Dhanu Chettri
Dear Dhanu,
It is clear that the inclusion of the quantum well caused the problem of convergence.
So, i would propose that you solve for the quantum well alone with simplified boundary conditions. Then you will get feeling how to adjust the simulation parameters of the well to get fast convergence.
It seems that the device simulator solves the quantum mechanical equations in the well region while solving the semicondcutor equations in the other pin device region. You can also, treat the well from the macroscopic point of view after undermining its macroscopic parameters.
You can also try to change the dimensions of the well , it may relax the problem and leads to convergence.
My opinion is to make physical changes to facilitate the mathematical problem!!!
Best wishes
Dear Mr. Dhanu Chettri,
As per my experience, convergence problem can come due to so many reasons, like
- Due to large number node in the full structure (more than 20,000 for Atlas)
- Improper selection of meshing in the well area
- Less or large number of mesh in "QTREGION" statement in the well area
- Selection of improper numerical technique
- Selection of less number of iteration in the "METHOD" statement
- Selection of large bias step in "SOLVE" statement.
I hope it will to help to remove the convergence problem in SILVACO TCAD
Dear Researchers,
Thank you for your valuable comments and suggestion.
When I remover calculated valued of permittivity and affinity from the code there is no convergence.
But adding this value gives rise to convergence.
Regards,
Dhanu Chettri
Affinity value of which material gives convergence problem?
I think .. by adding affinity value of contact material convergence problem will come.
hy I have a same problem and , I Will insert a graded layers by using silvaco , How can Insert this layers because when i use x=0,25 is not a graded layers please help me, and which models I use for nanoparticul
For linear composition or doping grading you can use the built in parameters COMPX.TOP and COMPX.BOTTOM (you can read more about how they work in the Atlas manual, section 2.6.4, page 53 in the the 2016 version)
You can then set average values of physical quantities in the normal way if that is appropriate.
To set physical quantities which depend on the composition in the graded layer you should the powerful C-Interpreter function.
Dear Dhanu,
Go for the most optimized and efficient meshing along with smooth biasing.
Best wishes
M. Imam
- Badges
- Science topic