Optimization of metasurfaces under geometrical uncertainty using statistical learning

Area of Science:

• Nanophotonics and Metasurface Technology
• Computational Modeling and Simulation
• Materials Science and Engineering

Background:

• Experimental metasurface performance frequently deviates from numerical predictions.
• These discrepancies stem from fabrication uncertainties and nanoscale imperfections affecting metasurface building blocks.
• Understanding efficiency loss due to geometric variations is vital for advancing metasurface applications.

Purpose of the Study:

• To develop a novel optimization methodology that incorporates manufacturing errors into metasurface designs.
• To enhance the robustness and reliability of metasurfaces in practical applications.
• To reduce the computational cost associated with optimizing metasurface designs.

Main Methods:

• Utilizing probabilistic surrogate models for accurate predictions, minimizing the need for extensive numerical simulations.
• Implementing a new optimization approach to systematically account for fabrication uncertainties.
• Applying the methodology to optimize a standard beam-steering metasurface composed of cylindrical nanopillars.

Main Results:

• The proposed optimization methodology successfully generated a metasurface design.
• The optimized design demonstrated double the robustness compared to conventionally designed metasurfaces.
• The use of probabilistic surrogate models significantly reduced the number of required numerical simulations.

Conclusions:

• The developed methodology effectively addresses fabrication uncertainties in metasurface design.
• This approach leads to significantly more robust metasurface designs, enhancing their practical viability.
• The findings pave the way for more reliable and widely applicable metasurface technologies.