Dark current calculation with Device


#1

I used Device to simulate the dark current of the vertical Ge-PD for 2V reverse bias voltage and for various length of the PD.

The simulated results are far from my measurement results (simulation results were much smaller than the measurement results for the length longer than 10um). I modified the bandgap parameter of the Ge and it improved the results. However, I still have noticeable differences for long length PDs.
I also try to use the interface material model, but I noticed that by adding the Si-Al interface and Si-Ge interface the dark current degrades while I expected the dark current increases.(the degradation of the dark current with Si-Ge interface is noticeable). Is there any specific setting for the interface models?
Further, when I add the Si-Al interface and consider the electrical surface recombination velocity for majority carriers also, the simulation stop due to the errors:
" Initialization failed to converge electrostatic potential update. Error: there was an unknown parallel error. the error code is 9002," what do these errors mean?

I have noticed that the simulated slope of the variation of the dark current for various PD’s area is smaller than the slope of the measurement results. I would appreciate if you suggest what parameter I should change to solve that issue. more precisely, I want that the dark current increases faster with increasing the PD’s area.

Thank you!


#2

Hi. I am guessing that you are using the example file from this example Vertical Photodetector? Please note that in this example silicon is used as a contact rather than a semiconducting layer. So when you add surface recombination, it basically implies an imperfect contact which will reduce the current. The recombination mechanisms contributing to dark current in this example file would be the bulk recombination (SRH, radiative, Auger) and any surface recombination between Ge and SiO2. The parameters for these models as well as the mobility, bandgap etc. for the semiconductor models need to be consistent with your deposited device. Also, the doping profile should be as accurate as possible compared to the actual device in order to ensure a good match. When you are trying to model the effect of length, how are you doing it? Are you doing a 3D simulation or are you simply changing the “norm length” in your 2D simulation?


#3

Hi Aalam,

Thanks for the explanations
I do the 3D simulation for each length and define the norm length based on the 3D simulation. For example, if I want to simulate the dark current for 20um length I draw the 20um length PD and also set the norm length to 20um

Thanks,
Monir


#4

Hi Monir,

Thanks for the reply. Just to be sure, you are still doing the 2D solver mode in DEVICE with different values for “norm length” right? Please note that if you are doing 2D simulations, then the length of the structure (geometry) in the third dimension does not really matter as long as you set the norm length to the right value. One thing to be careful about here is that even though we can justify using a 2D simulation in DEVICE with different norm length, the generation rate from FDTD should be calculated by a 3D simulation for the actual length of the PD (so you will have to run different simulations for different lengths). If you have done only one simulation in FDTD with a certain length (say 10 micron) and then are using that generation rate for all DEVICE simulation, then you will end up with much larger current at larger norm length values.


#5

Hi aalam,

I have two questions, and I would appreciate your help on that

  1. For the dark current simulation, we disable the generation rate file imported from FDTD. Since with enabling file, it will calculate the photocurrent. Therefore, I didn’t expect that the file affects the results. is there any extra information that we can import from FDTD?

  2. As you mentioned before about the example setup for the dark current simulation,
    "The recombination mechanisms contributing to dark current in this example file would be the bulk recombination (SRH, radiative, Auger) and any surface recombination between Ge and SiO2."
    Is there any simulation setup the I can add the Si-Al interface and Si-Ge interface without degradation effect I explained in my first message?

Thanks for your help in advance,


#6

You should not see any effect of the optical generation rate on the dark current since the import generation object will be disabled and will not have any effect on the simulation. Are you somehow seeing some effect of the generation object on dark current while it is disabled?

You can define a surface recombination velocity at semiconductor - metal interface as well. However, this is used to model the quality of the contact and a smaller value of surface recombination velocity will mean that the contact is a poor one and the number of minority carriers collected by the contact will be reduced. However please note that the surface recombination velocity at metal-semiconductor interface only gets applied to minority carriers and not to majority carriers.


#7

Thanks For the response,

I don’t see any generation effect when the generation object is disabled,
I ask this question which is related to your reply on August 8,

"If you have done only one simulation in FDTD with a certain length (say 10 microns) and then are using that generation rate for all DEVICE simulation, then you will end up with much larger current at larger norm length values."
I was wondering with disabling file why it is important to repeat the FDTD simulation for different lengths to simulate the dark current.

Is it correct to say that to simulate the dark current we cannot add the Si-Al interface and Si-Ge interface since this will reduce the conductivity of the Si-[contact] and Metal, and as a result, it reduces the measured dark current?
In general, we expect that this interface increases the dark current


#8

I mentioned that point as a caution simply to ensure that when you are doing a photocurrent simulation you will pay attention to the norm length value. Sorry I wasn’t very clear about it.


#9

Hi,

Just to confirm with you that, in 2D simulation, does the dark current scales (changes) with the norm length while the dark current density remains unchanged? I found the dark currents in my simulations are far larger than the measurement results in papers which I’m thinking this may be due to the long norm length I used in my 2D simulation. Is these right?

I appreciate your support.


#10

In a 2D simulation the reported current varies linearly with norm length. You will have to make sure that the norm length in your simulation is equal to the length of your fabricated device in order to compare the dark current between them.