Learning Forum

[mmarch]

Hi, I am having some convergence problems when simulating the structure described in this paper: https://aip.scitation.org/doi/10.1063/1.350913

I have created the exact same structure in Lumerical CHARGE and I want to obtain the I-V curve to compare it with their results. Here is my structure

[Khashayar Ghaffari]

Hi Thank you for reaching out. As you might know convergence issues in CHARGE can occur due to a variety of reasons. Here I have the following suggestions:
If you have a large voltage step in your boundaries (e.g. 1 V) try reducing this to a smaller value like 0.25V or 0.5V. This can help the convergence.
In case you are able to get partial results for some voltages in forward bias (or even at 0V), inspect the available current results. It is possible that the convergence is due to high values of current; this can occur in simulation of forward bias pn junctions. If this is the case try increasing the iteration limit in the advanced settings. This might help the simulation converge for larger values of current.
If the issue persists inspect the mesh. Is the mesh refined enough to resolve the smaller features in your structure? For example the thin layer visible in the middle of your device. You can mesh the structure and visualize grid to inspect the generated mesh. Using a local mesh constraint can help refining the mesh in these areas while keeping the overall mesh coarse and the simulation time reasonable.
Best regards

[mmarch]

Hi,

Thanks for your reply. I would like to sweep the voltage from 0 - 1.4V and I am currently doing this in 50 steps to ensure that it goes smoothly. However, I have noted that whether I sweep from 0-1.4V, 0-0.5V or 0-0.25V, my simulation will always run until 13% completion before getting stuck. So independently of my voltage rang, at 13% progress it will fail to converge which seems very strange to me. Do you have an explanation for this?
I have had a look at the mesh and it seems adequately defined. Here is a photo:

[mmarch]

Hi
Do you have any advice?

[Khashayar Ghaffari]

Hi Sorry for the long delay. You are correct the mesh doesn't seem to be too coarse. I have two suggestions:
Advanced tab settings: Revisit the advanced settings of your CHARGE solver. For example: enable Fermi statistics if you are using high doping values, increase the global iteration limit to see its effect on convergence, try changing the gradient mixing selection if high field mobility is used, etc. This article might be helpful to you.
Available results: Try to look into the available results, specifically current and electric field. These will be especially important at the last successfully run bias before the simulation failed to converge. For example is the current rapidly increasing? In that case increasing the iteration limit could be helpful. Or is the electric field extremely high at places? If so, does this the match the expected operation for the device? If this occurs in between very thin layers or at corners, further increasing the mesh refinement can be helpful.
Best regards