seeking steady state excess carrier concentration; no junction; no bias

[patrick.goley]

Hi,

I have an integrated silicon rib waveguide that has been excited with
visible wavelength light from the surface. I obtained the generation rate from FDTD. I then imported this generation rate in to DEVICE where I want to obtain the steady state excess free electron and hole concentrations assuming CW visible light excitation. I’ll note that the doping in the waveguide is low (uniform NA = 1e15 /cm^3) and the incident light intensity is relatively high, therefore I am expecting high injection conditions, i.e. that dn = dp >> 1e15/cm^3.

I am receiving the following error:
SIGETERMSRV(process 0): The program terminated due to an error: Initialization failed to converge electrostatic potential update.
SIGETERMSRV(process 0): Error: there was an unknown parallel error. The error code is 9002, the process number is 0

I have attached my project file. Any assistance would be greatly appreciated.
wg_absorption_DEVICE.ldev (6.2 MB)

Best,
Patrick

[patrick.goley]

I updated the doping object dimensions to extend beyond the edges of the silicon, thinking that might be the issue (new project file attached) - that didn’t work. I am also attaching the generation rate file from FDTD.

wg_absorption_DEVICE.ldev (6.2 MB)
CW_generation.mat (174.3 KB)

I see that the simulation runs fine if I disable the generation object.
Again, any help debugging this issue would be greatly appreciated.

[aalam]

Hi @patrick.goley, my apologies for the late reply. I have looked at your file and there are a few things that need attention.

1. All CHARGE simulation requires application of bias voltage. However the structures that get the bias applied (i.e. the contacts) are not included in the simulation and they are used to simply provide a boundary condition. In your simulation you therefore need to figure out where you want to apply the bias (for example where are your anode and cathode located).

2. I would recommend to not include the substrate in the simulation. It will not affect your result in the SOI but if you include it in the simulation then you will need to have additional bias applied to the substrate which will complicate the simulation unnecessarily.

3. The two electrical contacts should ideally have different names.

I have added two metal contacts in your simulation and have set them to zero volts. I have also changed the simulation region so that the substrate in not included. If you look at this attached file you will see that the simulation now runs without any problem.

wg_absorption_DEVICE_mod.ldev (6.2 MB)

I understand that this may not be the structure you have in mind but hopefully this will give you an idea as to what is needed.

One other thing to note here is that your excess carrier concentration is not necessarily >> 1e15. This is because your electron/hole lifetime is in the range of microsecond or even less. Even if you have a photogeneration rate of 1e21 / cm^3, the excess carriers recombine withing microseconds and as a result your steady state excess carrier concentration will be much lower. This can be seen when you run the file above. You will see that you can only detect the excess carrier for the minority carrier (n). For the majority carrier the excess carrier concentration is still much less than the doping value.

[patrick.goley]

Hi @aalam, thanks very much for your thorough reply. It all makes sense. You are absolutely right about the excess carrier concentration being so low. I was expecting a lifetime on the order of hundreds of microseconds. I see now that it was set to a few microseconds, which helps explain the result.