Physics-Based Approach for a Neural Networks Enabled Design of All-Dielectric Metasurfaces
Condensed Matter - Mesoscale and Nanoscale Physics
Mesoscale and Nanoscale Physics (cond-mat.mes-hall)
FOS: Physical sciences
Physics - Applied Physics
Applied Physics (physics.app-ph)
02 engineering and technology
0210 nano-technology
DOI:
10.1021/acsphotonics.0c00663
Publication Date:
2020-07-24T16:48:11Z
AUTHORS (3)
ABSTRACT
Machine learning methods have found novel application areas in various disciplines as they offer low-computational cost solutions to complex problems. Recently, metasurface design has joined among these applications, and neural networks enabled significant improvements within a short period of time. However, there are still outstanding challenges that needs be overcome. Here, we propose data pre-processing approach based on the governing laws physical problem eliminate dimensional mismatch between high optical response low feature space metasurfaces. We train forward inverse models predict responses cylindrical meta-atoms retrieve their geometric parameters for desired response, respectively. Our provides accurate prediction capability even outside training spectral range. Finally, using our model, demonstrate focusing metalens proof-of-concept application, thus validating proposed approach. believe method will pave way towards practical learning-based solve more complicated photonic
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (37)
CITATIONS (66)
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....