Genetically optimized all-dielectric metasurfaces

Егоров, В. Н.; Eitan, Michal; Scheuer, Jacob

doi:10.1364/oe.25.002583

Cited by 59 publications

(50 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Evolutionary strategies represent classical optimization strategies to global optimization. One such algorithm is the genetic algorithm, which have been applied to many types of photonic design problems, including metasurface design [40]. Compared to our approach, genetic algorithms are not efficient and require many thousands of iterations to search for even a simple optimal device structure.…”

Section: Comparison With Evolution Strategies (Es)mentioning

confidence: 99%

Simulator-based training of generative neural networks for the inverse design of metasurfaces

2019

View full text Add to dashboard Cite

AbstractMetasurfaces are subwavelength-structured artificial media that can shape and localize electromagnetic waves in unique ways. The inverse design of these devices is a non-convex optimization problem in a high dimensional space, making global optimization a major challenge. We present a new type of population-based global optimization algorithm for metasurfaces that is enabled by the training of a generative neural network. The loss function used for backpropagation depends on the generated pattern layouts, their efficiencies, and efficiency gradients, which are calculated by the adjoint variables method using forward and adjoint electromagnetic simulations. We observe that the distribution of devices generated by the network continuously shifts towards high performance design space regions over the course of optimization. Upon training completion, the best generated devices have efficiencies comparable to or exceeding the best devices designed using standard topology optimization. Our proposed global optimization algorithm can generally apply to other gradient-based optimization problems in optics, mechanics, and electronics.

show abstract

Section: Comparison With Evolution Strategies (Es)mentioning

confidence: 99%

Simulator-based training of generative neural networks for the inverse design of metasurfaces

2019

View full text Add to dashboard Cite

show abstract

“…The technique has been used to design electromagnetic and 2 optical devices, and revolutionary silicon photonics components [18][19][20][21][22][23][24][25][26][27][28]. Recently, researchers have recognized the utility of inverse-design-based approaches for metasurface optics [29][30][31], among other methods [32]. Such work has led to designs for polychromatic and high-incidenceangle metasurface lenses and gratings [29,[33][34][35], and has facilitated the exploration of novel fabrication platforms, such as arrangements of Mie-scattering dielectric spheres [36].…”

Section: Introductionmentioning

confidence: 99%

Computational inverse design for cascaded systems of metasurface optics

Backer

2019

Opt. Express

View full text Add to dashboard Cite

Metasurfaces are an emerging technology that may supplant many of the conventional optics found in imaging devices, displays, and precision scientific instruments. Here, we develop a method for designing optical systems composed of multiple unique metasurfaces aligned in sequence. Our approach is based on computational inverse design, also known as the adjoint-gradient method. This technique enables thousands or millions of independent design variables to be optimized in parallel, with little or no intervention required by the user. To demonstrate the broad applicability of our method, we use it to design an achromatic doublet metasurface lens, a spectrally-multiplexed holographic element, and an ultra-compact optical neural network for classifying handwritten digits.

show abstract

“…This technique is similar to the Extended Local Periodicity (ELP) approach [18] proposed for reflectarray design. Specifically, we focus our attention to a beam deflector element which is a commonly used structure for benchmarking metasurfaces design strategies [17,19,20] as the transmission efficiencies can be compared across designs. The beam deflector element changes the direction of an normally incident plane wave by giving it a predesigned inclination with respect to the normal.…”

Section: Introductionmentioning

confidence: 99%

“…The beam deflector element changes the direction of an normally incident plane wave by giving it a predesigned inclination with respect to the normal. Although a simple element, the beam deflector, in addition to being an important element in itself, can be combined to produced metasurfaces with more advanced functionality like high numerical aperture and multi-wavelength focusing lenses and holograms [17,[20][21][22]. The beam deflector elements considered in this paper do not use the local phase approach [20] which is based on the unit-cell approximation but move beyond this by considering an extended unit-cell consisting of a larger number of nanoantenna.…”

Section: Introductionmentioning

confidence: 99%

“…In comparison to the beam deflector synthesis method reported by Byrnes et al [17] which uses local optimization methods, our method which relies on global optimization methods, as the Genetic Algorithm and Artificial Bee Colony, shows significant improvement in efficiency by not getting stuck at local optima. In comparison to [20] which has studied the application of Genetic Algorithm (GA) to the beam deflector design problem, we have explored unit-cells which do not restrict the design to a rectangular lattice and cylindrical elements. Additionally, our Artificial Bee Colony method is seen to outperform GA method in speed by cutting the convergence time by 35%.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Rapid Design of Wide-Area Heterogeneous Electromagnetic Metasurfaces Beyond the Unit-Cell Approximation

Donda¹,

Hegde²

2017

PIER M

View full text Add to dashboard Cite

Abstract-We propose a novel numerical approach for the optimal design of wide-area heterogeneous electromagnetic metasurfaces beyond the conventionally used unit-cell approximation. The proposed method exploits the combination of Rigorous Coupled Wave Analysis (RCWA) and global optimization techniques (two evolutionary algorithms namely the Genetic Algorithm (GA) and a modified form of the Artificial Bee Colony (ABC with memetic search phase method) are considered). As a specific example, we consider the design of beam deflectors using all-dielectric nanoantennae for operation in the visible wavelength region; beam deflectors can serve as building blocks for other more complicated devices like metalenses. Compared to previous reports using local optimization approaches, our approach improves device efficiency; transmission efficiency is especially improved for wide deflection angle beam deflectors. The ABC method with memetic search phase is also an improvement over the more commonly used GA as it reaches similar efficiency levels with a 35% reduction in computation time. The method described here is of interest for the rapid design of a wide variety of electromagnetic metasurfaces irrespective of their operational wavelength.

show abstract

Genetically optimized all-dielectric metasurfaces

Cited by 59 publications

References 30 publications

Simulator-based training of generative neural networks for the inverse design of metasurfaces

Simulator-based training of generative neural networks for the inverse design of metasurfaces

Computational inverse design for cascaded systems of metasurface optics

Rapid Design of Wide-Area Heterogeneous Electromagnetic Metasurfaces Beyond the Unit-Cell Approximation

Contact Info

Product

Resources

About