comment on minimum lengthscale in dispersion problem

[stevengj]

Would be good to comment on the fact that, even though we imposed no minimum lengthscale in the dispersion-engineering problem, the final design has a minimum lengthscale of about 100nm(?). This means that it will remain a local optimum even if you impose any minimum lengthscale ≤ 100nm, since the constraints will be inactive.

[oskooi]

In particular, it was argued that dispersion-engineered designs such as slow-light waveguides do not generally include small features and thus below a certain imposed minimum feature size (i.e., 100 nm), the final designs would not vary significantly. It would be useful to investigate this claim further preferably using at least two different optimization methods/codes.

[stevengj]

I don't know if that's true "generally", but it seems to be true of the local optima found here.