Learning Forum

[pengzheng97]

Hi,

I'm using lumopt to inverse design a 1X3 WDM, Imitate the official example “CWDM_splitter_1310_4ch_2D_TE_topology”，the 1X4 WDM.

But I encountered some problems during the binarization stage， like the following two picture. Binarization is too much, resulting in great performance degradation. And the next optimization can't make it up， So that the result is poor performance.

Is it possible that my size is not enough？ it's 6umx4um

Is there any way to improve it？ thank you very much！！

[Greg Baethge]

Thanks for posting your question and apologies for the delayed response. Topology optimization is always a bit tricky. First, it depends a lot on the initial guess you used as a starting point. It can be useful to try multiple ones (for example, a matrix of 1s that corresponds to the whole area at the core index, a matrix of 0s for cladding, or in-between).
Also, did you use the minimum feature constraint in your optimization?

[pengzheng97]

Thank you so much!´╝ü I will try your suggestion and optimize again using other initial structures.
Another question, I did use the minimum feature constraint filter_R = 200e-9
min_feature_size = filter_R
Is it possible that the minimun feature constraint is the problem, because I encountered a sharp degradation of performance during the last binarization operation in another 1X2 WDM optimization similar to this example.

[Greg Baethge]

You're very welcome,. This is possible: if the constraint is too large, some of the details of the structure would have to be removed to fulfill the constraint. This can degrade the performance. You can try different values. Note min_feature_size and filter_R can have different values (see this page for more information about the parameters).