I am simulating a PIN device in Sentaurus Workbench (TCAD), and my simulation results show a huge electric field peak. (I apply a negative DC voltage of 1300V to see if the PIN rectifier can withstand it).
I have been trying to model the real device better (try to get a reduced electric field distribution in my simulation) by modifying the resistance of certain regions (ion implanted regions) but my simulation has problems converging whenever I change any parameters (the resistance of regions or if I add interface charge to it). I can add and modify the slightest thing and then it will not converge.
Would you have any suggestions for me as to how to make my solutions converge? I am using Quasistationary Newton convergence algorithm. Using simple Poisson coupling. As of now, I am playing around with the maximum step size and initial step size in the SDEVICE code, but that does not seem to work.
Thanks a lot!
Isha,
What is the equation you solve for the reverse biased PIN diode? Do you solve the Poison equation, or do you solve the poison and continuity equations simultaneously? What do yo mean by changing the resistance of certain regions?Do you mean you change the doping of the end regions of diode or you change the doping in certain region in the i region.
The first step in investigating the divergence problem is to make an approximate analytical solution to see the critical parameters which can lead to the divergence.
If you work the device near the breakdown and you make full solution including the current voltage relation ship, then physically at breakdown the current can increase indefinitely to very large values. If you solve for the I-V in the breakdown region you have to push a current and calculate the voltage. In this way you control the unlimited growth of the current at the breakdown.
These some are guide lines that show how divergence can occur.
From the point of view of device engineer, the best electric field distribution is achieved when the middle region is intrinsic. So, as you reduce the i-layer doping your filed in the i region will be more homogeneous.
The most important thing then is to not to bias the diode by constant voltage but a constant current.
Best wishes
This would help you a lot if you provide what you did in Sentaurus device TCAD.
And at what bias point did you get failed?
Please let me know if you are struggling and could not find any methods to solve that.
Best regards,
Jun-Sik Yoon
Hi Isha,
I think convergence problem is mostly due to meshing refinement and the sensitive areas present in the device. Use CNormPrint command to know the areas where your device encounters problem. Hope it will help!
I faced something similar with PiN device with breakdown of around 770V, in the forward bias though with simple Poisson coupling. I also played with max step size, iteration etcetera, the convergence was not achieved. In my case, the trick to solve the problem was to use transient instead of quasistationary in both forward and reserve bias (both voltage and current ramping).
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Electrodes