How do I simulate loss due to sidewall roughness?



I have a structure that works well given the perfectly etched sidewalls. But I’m suspecting that it is prone to loss due to sidewall roughness, since the mode doesn’t always stay within the core in this particular design. So I’m looking for a way to introduce sidewall roughness to see how much of a loss it results in. I could switch to FDTD as well, if there’s a way to do this there.

The solution I can think of is to specify another material within a few nm of the waveguide sidewalls. This material would also have a non-zero imaginary part of its refractive index. But I would be almost getting back what I put in, since I’d be specifying the loss in that material. I’m really looking for a more elegant solution. Any ideas?

Thank you.



The following publication may be of interest to you since it compares simulated sidewall roughness scattering loss (using FDTD Solutions) to an analytical calculation:
Jaberansary, et al., “Scattering loss estimation using 2D Fourier analysis and modelling of sidewall roughness on optical waveguides.” IEEE Photonics Journal, 5, (3), (2013)

Depending on the size of the roughness features compared to the size of the structure, it may not always be practical to directly simulate the roughness since smaller feature sizes require a finer mesh which increases the memory requirements. However, even if it’s not feasible to simulate the roughness directly, you could look into the analytical calculation of scattering loss from the publication and see if that would apply for your structure.

Surface roughness and scattering loss calculation for plasmonic waveguide
Mode loss and waveguide loss

It would be fantastic if MODE had a special material “sidewall roughness” that could be drawn in the geometry. Then eigenmode would calculate the field profile (ignoring this material), find the field intensity at the sidewall, and use some analytic expressions to determine the additional loss from this extra roughness material. Perhaps this could just be implemented in a script, but it would be nice if the GUI included this already.

There are a few other papers on this topic. In particular, we’ve replicated the work in F. Grillot’s paper “Propagation Loss in Single-Mode Ultrasmall Square Silicon-on-Insulator Optical Waveguides” to predict the loss in a waveguide for a given set of parameters (correlation length, standard deviation of the roughness). The results look reasonable, but we haven’t compared them to our experimental results; this would be certainly very useful.

This paper only presents the model for the TE polarization. It would be nice to have an extension for the TM polarization as well.




Great suggestion. I’ve added this as a feature request.