Learning Forum

[bb63w6kf]

Can I modify the parameters of the structure in parallel in the script? I now use the adjoint method to optimize a metalens with a 100*100 meta-atoms. Each time  updating the geometric size of the meta-atoms, the script update it one by one(based on two for loops), similar to serial, which is a waste of time. I would like to ask if there is any way to update the meta-atoms in parallel at the same time to reduce the time?

[Dimitris Polyzos]

Hello 

For start, the benefit of the adjoint method is that you can get the derivative of the FOM with respect to the parameters using only 2 simulations. For gradient based optimization you would typically need 2N+1 simulations with N being the number of parameters. At the top of my head, I don't think that parallel modification of a structure in a script can be done easily.  However, increasing the number of loops is something that you could try to introduce more meta-atom structures in the optimization process.  Also other optimization methods could be used such as PID and grating order optimization but at the moment it is not supported and will not be released in the immediate future.

Also, you could utilize structure groups, where you can simplify parameter sweeps, allowing you to loop through one of the  parameters of a meta-atom. Each time the set script command is called, the setup script runs, editing the array with the new lattice parameter. These values can be used as parameter sweep variables or as part of the figure of merit in optimizations. Also, you could cosnider using HPC resources to make the optimization process faster

I hope I helped

Regards