During GaN vertical power device simulation, faces convergence problem at breakdown point. I used voltage, current boundary condition and curvetrace method. But can not get stable solution beyond 1000V. I also tuned the methods parameter for very low bias point increment like, method newton climit=1e-9 maxtrap=40 itlimit=40 dvmax=1e8 ir.tol=1e-40 cr.tol=1e-40 ix.tol=1e-40 px.tol=1e-30 pr.tol=1e-45 cx.tol=1e-30.
I also try high simflags for numerical precision like ~160bit, 256bit precision
During 256 bit precision get very long simulation time, although used all available system like "-P all"
You can try a "slow" transient solution trick with COMPLIANCE parameter, like:
solve vdrain=100e06 ramptime=100e6 tfinal=100e6 dt=100 compliance=5e-2 cname=drain previous
Dear Rashid Jamal
welcome!
The matter of instability stems from the operating point stabilization.
You have to bias the transistor with constant current source IDS which you can scan to get the I-VDS curve for specific gate voltage VGS . It is so that the breakdown curve may have an unlimited current growth at the breakdown point. So in fact you will have many operating points of you worked with constant VDS voltage source.
Best wishes
To Prof. Zekry and Arteev
Basically i simulate the GaN vertical trench MOSFET. I am looking for off-characteristics of devices. How much reverse voltage can sustained by device without breakdown(on to off current in nano-amp).
So, I fixed the gate voltage at minimum off voltage (~0 V) of device and then sweep the drain voltage. I face problems above >1000 V in reverse mode.
problems may be due to,
(a) due to low intrinsic concentration of GaN(~10^(-10))
(b) GaN high field mobility has negative differential resistance region.
My question is how to overcome above mentioned problems (a,b) in silvaco atlas/blaze ? Abdelhalim abdelnaby Zekry Dmitri Arteev
Dear Rashid Jamal
Can you show a measured curve for such device?
I know this problem very well and I want that you reed elaborately my first proposal.
One of the solutions is not to scan the device b=voltage but have to force current in the device and solve for the voltage.
If you can not then you have to ass a series resistance to your device such that you obtain the I-V characteristics of the two devices in series and then you can easily subtract the voltage drop on the resistance.
The resistance is here to limit the device current when it enters in the break down.
As for the negative differential resistance it prefer the use of the consatnt voltage source.
So, adding the resistance may stabilize the operation of the device.
If you show me the measured I-V curve I can precise my answer more.
Best wishes
Rashid Jamal , as I've said before, if you face a convergence problem in breakdown simulation, a transient solution with COMPLIANCE statement could do a trick (read the manual).
Also check if Selberherr impact ionization model is enabled.
(a) an 'ideal' GaN crystal has a very low intrinsic carrier concentration; generally it may cause convergence problem and extended precision should be used.
But real GaN is n-type of 1e16-1e17 cm-3 or compensated with deep acceptor levels/traps. So it is not quite accurate to model GaN as 'true' intrinsic material.
Moreover, high-purity intrinsic semiconductors (even wide bandgap) can effectively conduct an electric current due to carrier injection (see, for example, Current Injection in Solids by Lampert and Mark).
(b) Field-dependent mobility model for GaN could really cause convergence problems. Try to switch it off and see if the solution can be obtained.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts