Problem about the simulation diverged in high electric field



dear support,
when I simulation about the transient, there is some problems that the simulation did diverged.

.To solve this problem , I changed the solve mode from transient to steady, and change the bc mode too(from 0- -1V at anode). . This work well in steady simulation. Then I changed the solve mode from steady to transient, and click the enable continuation. . At last, I ran the simulation, the result did diverged as always. I wonder to know how to get the correct result.vpd_electrical (6).ldev (6.9 MB)



Have you tried the solutions discussed in the following topics?


I have tried these solutions as follows, but the result did diverge as always. Is there any other ways to solve this problem?


The reason your simulation is diverging is that you are biasing the device in forward bias (which results in large currents) where as since this is a photodetector, it should be biased in reverse. So if you switch the polarity of the applied bias, the simulation should converge.


Okay, that is my fault. Thanks for your advise.



签名由 网易邮箱大师 定制


This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.



dear mmahpeykar
The result of -1V bias voltage is OK, but when I change the bias from -1V to -2V, and -3V, the results went to diverge.vpd_electrical (6).ldev (6.9 MB)


As you increase the voltage, due to the stronger electric field, the charge variation between each iteration becomes larger and this can cause the solver to diverge. To remedy this, you can reduce the min time step in transient settings of the CHARGE solver to reduce the charge variation between each transient simulation iteration.


It did converge when I reduce the min time. But for 4V bias, no matter how to reduce the min time, the result will diverge. I am very confused about that, if there is any ways to make the result right?4V bias.ldev (6.9 MB)

@2V bias, when I open the high field model of Si, the result will deverge too no matter how to reduce the min time.

-2V_with Si high field model.ldev (6.9 MB)


Are there any ways to solve that? Any suggestion would be greatly appreciated!


Having the high field model enabled for Si is not necessary in this simulation because the junction is located in Germanium not Si so the field in Si is not that high. Thus enabling the high field model for Si only makes the simulation numerically more complicated. I’d only have the high field model enabled for Ge.
Regarding the 4V bias, you might wanna try to apply the voltage in steps. This means you apply the voltage up to 2V for example and wait until the simulation reaches steadystate, then apply 4V and wait for steadystate again and finally apply the illumination. This needs careful timing of the bias and illumination application.


I think the high field is in Ge not in Si, the simulation result is as fllows.

Another thing is that it’s hard for simulation @4V bias to converge.


This confirms what I mentioned in my previous post so you don’t need to have high field model enabled for Si. Regarding the 4V case, the most reliable solution is what I suggested in my previous post.


But there the central region is Si, the electric field is higher than other region. Please confirmed that structure again.



签名由 网易邮箱大师 定制


Sorry I didn’t realize that Si is part of the detector. In that case try disabling the high field for Ge and have it enabled only for Si. Also, switch the driving field from “E dot J” to “grad phi” which is an easier scheme to converge.