Computers can make photons interact more strongly: enhancing optical frequency conversion via inverse design


Generally, photonics refers to the study of the properties of light, or rather the study of radiant energy. This science has found applications across all aspects of life. One vital and prominent application involves the conversion of light from one frequency (say infrared) to another (say visible), e.g. visible laser sources. Frequency conversion has been studied thoroughly, specifically in bulky optical systems. Nonetheless, little has been done within the scope of micro- and nano-scale structures, where the possibilities to confine (trap) light to length scales of the order smaller than its wavelength exists, allowing nonlinear effects to arise at small powers or with large efficiency. Existing studies of frequency conversion are based on the premise of observing enhanced nonlinear effects in structures that have the capability of supporting multiple resonances at far-away frequencies. Unfortunately, such conventional designs still fall short of simultaneously meeting the numerous design challenges that are associated with resonant frequency conversion, principally the need to support multiple nodes with highly concentrated fields, strong mode overlaps and exactly matched resonant frequencies.

Recently, Princeton University researchers: Chawin Sitawarin, Weiliang Jin, Zin Lin, and led by Prof. Alejandro Rodriguez from the department of Electrical Engineering showed that recently developed inverse-design (machine learning and related optimization) techniques could be successfully applied to automatically discover novel and unusual types of micro-structured fibers and meta-surfaces designed to achieve large nonlinear frequency-conversion efficiencies. In addition, they improved our knowledge on the potential of photonic optimization in nonlinear optics. Their work is currently published in the research journal, Photonics Research.

In brief, the research method employed commenced with a thorough overview and development of the novel inverse optimization approach, with reference to the typical optimization problem seeking to minimize or maximize an objective function, subject to certain constrains over a set of free variables. Next, the researchers applied the optimization technique to propose a more efficiency third-harmonic generation process at any desired wavelength. Lastly, using the technique mentioned earlier, they estimated the power requirements associated with the fiber designs they were working on.

The authors observed that the optimized structures demonstrated very high leaky-mode lifetimes at both fundamental and harmonic modes and orders-of-magnitude larger nonlinearities than observed in traditional intuition-based designs. Moreover, they noted that the inverse designs not only subdued the efficiency limitations of traditional index fibers and photonic crystal meta-surfaces, but also greatly reduced the challenges and drawbacks inherent to the design process.

In a nutshell, Alejandro Rodriguez and his research team demonstrated new optimization approach for the design of nonlinear photonic fibers and meta-surfaces. In general, they observed that by using their approach, the proposed all dielectric meta-surfaces exhibited negligible material losses and hence larger lifetimes, while ensuring larger optical nonlinearities (interactions among photonic modes at different frequencies). Altogether, their approach can be extended to consider terahertz frequency generation and other frequency conversion processes.

Our work is among the first to consider application of large-scale optimization methods (and related computer-driven design methodologies) to the problem of enhancing interactions between light waves of different wavelengths. This has implications to the design of large-scale photonic networks, quantum information processing, and generation of light sources at scarce wavelengths, such as Terahertz frequencies. Ultimately, I believe that many of these engineering problems will be solved using machine learning and computer algorithms that not only achieve optimal functionalities but also automate the entire process”, Said Alejandro Rodriguez, senior author of the research paper.


Chawin Sitawarin, Weiliang Jin, Zin Lin, Alejandro W. Rodriguez. Inverse-designed photonic fibers and meta-surfaces for nonlinear frequency conversion [Invited]. Volume 6, Number 5 / 2018 / Photonics Research.

Go To Photonics Research

Check Also

Distributed rotating consensus of second-order multi-agent systems with nonuniform delays