Abstract

Different optimization algorithms have recently been utilized to design and improve the performance of many nanophotonic structures. We present the design of a compact photonic structure by an approach based on machine learning. Three-dimensional finite-difference time-domain method is integrated with a machine learning algorithm in order to design a photonic structure. In particular, a subwavelength focusing lens structure that operates at telecom wavelengths is designed to have desired beam properties such as subwavelength full-width at half-maximum value of 0.155 λ and suppressed side-lobe levels at focal point, where λ denotes the wavelength of incident light and equals to 1550 nm. The designed compact lens structure has the footprint of ${\text{2}}\times {\text{1}}\,{\mu} {\text{m}}^{\text{2}}$ with a slab thickness of 280 nm, which is the smallest photonic lens for subwavelength focusing of light to date comparing to its conventional ones. The focusing mechanism of designed lens structure is explained with the help of applying discrete Fourier transform to the two-dimensional dielectric distribution of the structure. It is also shown that, due to its strong light confinement property, the designed lens structure can be used as a waveguide-to-waveguide optical coupling device with a beamwidth compression ratio of 10:1 by integrating a nanowaveguide with the width of 200 nm to the output surface of lens structure. Normalized transmission efficiency of the optical coupling device is calculated as high as 0.62 at the wavelength of 1550 nm. The outcomes of the presented study show that machine learning can be beneficial for designing efficient compact photonic structures.

© 2018 IEEE

PDF Article

References

You do not have subscription access to this journal. Citation lists with outbound citation links are available to subscribers only. You may subscribe either as an OSA member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access OSA Member Subscription

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an OSA member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access OSA Member Subscription