Learning Forum

[mhu]

Hello,
I am designing a photodetector. First, I did a 3D FDTD simulation and generated optical generation rate to put in 2D charge simulation (followed vertical photodetector example). Then I performed 3D charge simulation but the photocurrent is very poor now. I want the same photocurrent in 3D charge simulation that I got from 2D charge simulation. Moreover, the photodiode is covered with silicon-dioxide (Box) layer.
What can I do to get the same photocurrent?

[Guilin Sun]

The solver internally calculates the current density J. So the current I is post processed from the simulated result (integral over the line or surface). In 2D simulation the norm length (by default it is 1cm) is used to get the proper current. In 3D simulation, I believe the 3rd dimension in um is very small. This is why the current value is so small. You can rescale the current if you wish.

[mhu]

Thanks for your answer.
Actually in 2D simulation I already fixed my norm length and kept the same length in 3rd dimension in 3D simulation. Still my photocurrent is low.
In 2D simulation I just have to give the norm length and no background material. But in 3D simulation without background material the simulation do not runs and shows job error. To fix this error, I need to add background material, and I added SiO2 as a background material in 3D simulation. What is the background material actually? Is it affecting photocurrent?
And I am using the same script used in vertical photo-detector to get the photocurrent and responsivity.

[Guilin Sun]

In both 2D and 3D simulations, the background material must be assigned since in side the simulation region a proper material must be assigned for any geometry, including the background. Please check if you set the same material in both cases.
Please note that CHARGE only calculates the charge/current density etc inside the semiconductor materials. SiO2 is an insulator no charge can penetrate and move inside it.
I would suggest that you use only ONE file for 2D and 3D simulation to make sure they use the same materials, except that you need to modify the simulation region and dimension.
When you say that "Actually in 2D simulation I already fixed my norm length and kept the same length in 3rd dimension in 3D simulation" I assume that the 3rd dimension uses the norm length in 2D/
Once more thing: if you used 2D data by summing up the 3rd dimension (optical generation rate), the two results should be very similar, except the recombination effect. However, if you only used one slice of the 3rd dimension for 2D simulation, it is reasonable to have different results, since the current density is usually not uniform in 3rd dimension. Please check.

[mhu]

"When you say that "Actually in 2D simulation I already fixed my norm length and kept the same length in 3rd dimension in 3D simulation" I assume that the 3rd dimension uses the norm length in 2D/" - yes! I meant that.
"Once more thing: if you used 2D data by summing up the 3rd dimension (optical generation rate), the two results should be very similar, except the recombination effect. However, if you only used one slice of the 3rd dimension for 2D simulation, it is reasonable to have different results, since the current density is usually not uniform in 3rd dimension. Please check."
- Here I am confused. In optical generation rate analysis section, when I imported matlab file for 2D charge simulate, I kept period to '1' and average to 'x'. But for 3D charge simulation I kept period to '1' and average to 'none' in optical generation analysis. Is this the point where I am wrong?
I checked in all the photodetector examples given by lumerical that for 2D simulation there is no background material selected. So, I did not added any background material in 2D simulation but for 3D simulation I must have to add this otherwise the simulation donÔÇÖt runs. My photodetector geometry is fixed with some fixed materials (silicon, germanium, silicon nitride and aluminium for contacts) . I am confused that what type of material should I use as background material?

[Guilin Sun]

From your description, it is correct way to get the generation rate in 2D and 3D simulation.
I guess it is from the background material: in 2D, it cuts through the device so it does not have the background material; in 3D, in addition to the waveguides, there is space that needs to fill a material. Taking this online example as an illustration: https://support.lumerical.com/hc/en-us/articles/360042332014-InP-InGaAs-uni-traveling-carrier-photodetector
It has air and superstrate. For 2D simulation, it cuts through the section where the superstrate fills the simulation region:



which is not the physical geometry:

compare the two, they are different.
in 3D simulation, you are plotting the physical device. Therefore you need to specify the geometry with a material, eg, the air part. So regarding to the background material, I think it is clear now.

As for the different result, If the device is the same except the substrate/air part, the result should be similar. So please use ONE single file to set the simulation so 2D and 3D will use the same device configuration as suggested before to avoid any possible geometry/material discrepancy. You can modify the simulation region and CHARGE region for 2D and 3D simulations. The recombination loss should be small, or better to turn them off.

[mhu]

Thanks for your quick reply. Let me try and I will update results.

[mhu]

Hello The photocurrent I got in 2D simulation was in 0.001 range but in 3D simulation it is near 0.0001. But it is much better than previous. As you told in 3D space the current is not uniform in all the device area so, it might be the reason.
Now I increased the doping concentration and started 3D simulation again. After some moment it stopped and showed errors. please see the attached error.
then I checked the online reference as given in the error notification. It took me to the Charge setting option. Below I have attached my own charge setting for the photodiode simulation.
can you please tell me why the simulation is not running for higher doping concentration?

[mhu]

some information I must add. I run my simulation from 1 to -2 bias voltage. The simulation stopped after a specific voltage and it can show result till that voltage. You can see from the picture:
and the error details in the previous comment I got from this job manager.

[Guilin Sun]

Please write a new post for the new issues.
If the difference is 10 times, you may need to check the norm length and the 3D 3rd dimension span, as the recombination loss should not be that large. Have you turned off all the recombination mechanism? to isolate the cause of the discrepancy, I strongly suggest to disable unrelated settings. Have unified the simulation file?

[mhu]

3D 3rd dimension span and norm length of optical generation rate is the same.
And I did not added recombination mechanism in the simulation.

"Have unified the simulation file?"
-yes in the same file I ran 2D & 3D simulation.

[Guilin Sun]

Please calculate the total number of charges from the generation by spatial integration for the 2D and 3D and check the result. If they are very close, then we can try to find other causes.

[mhu]

"Please calculate the total number of charges from the generation by spatial integration for the 2D and 3D and check the result."
can I calculate total number of charge from this CHARGE setting?

[Guilin Sun]

You should calculate it from optical generation, which is the raw data for CHARGE. Those charge quantities are simulated in CHARGE which you have already seen the difference, right?

[mhu]

"Please calculate the total number of charges from the generation by spatial integration for the 2D and 3D and check the result."
From 3D FDTD simulation generation rate analysis I can get G (electrical generation rate as a function of x,y,z (units: charge pairs/m^3/s). are you talking about this?

[Guilin Sun]

Yes, integrate over space and check the 2D and 3D results.