Pushing the limits of optical information storage using deep learning

Abstract

Diffraction drastically limits the bit density in optical data storage. To increase the storage density, alternative strategies involving supplementary recording dimensions and robust read-out schemes must be explored. Here, we propose to encode multiple bits of information in the geometry of subwavelength dielectric nanostructures. A crucial problem in high-density information storage concepts is the robustness of the information readout with respect to fabrication errors and experimental noise. Using a machine-learning based approach in which the scattering spectra are analyzed by an artificial neural network, we achieve quasi error free read-out of sequences of up to 9 bit, encoded in top-down fabricated silicon nanostructures. We demonstrate that probing few wavelengths instead of the entire spectrum is sufficient for robust information retrieval and that the readout can be further simplified, exploiting the RGB values from microscopy images. Our work paves the way towards high-density optical information storage using planar silicon nanostructures, compatible with mass-production ready CMOS technology.Optical information storage promises perennial longevity, high information densities and low energy consumption compared to magnetic storage media. 1,2The compact disc (CD) and its successors, the DVD and the blue-ray disc, broadly established optical storage in our society. 3,4 Those media are based on storing a single binary digit per diffraction limited area ("zero" or "one"). Several concepts have been proposed to increase the information density in optical storage. Examples are schemes exploiting polarization-sensitive digits, 5 near-field optical recording, 6 the use of fluorescent dyes 7 or three-dimensional approaches like two-photon point-excitation 8 . Yet, all these alternatives suffer from major drawbacks. Either they are hardly superior to commercial planar solutions (polarization-sensitive patterns) or they require very complex storage media (fluorescence) or sophisticated read-out schemes (nearfield recording, two-photon point-excitation). The most promising alternative seemed to be holographic memory, which was proposed in the early 1960ies and makes use of the volume of the storage medium. To date, however, there is still no commercial product available, despite several announcements in the past 20 years. 9,10 In the last decades, photonic nanostructures emerged as powerful instruments to control light at the nanometer scale. 11,12 Localized surface plasmons (LSP) in metal nanoparticles 13 or Mie-type resonances in high-index dielectric structures 14 cover the entire visible spectrum and can be tuned by designing appropriate geometric features. 15,16 Furthermore, the high scattering efficiencies of photonic nanostructures render single-particle spectroscopy relatively easy. In consequence, the idea has been raised to encode information in the rich scattering spectra of plasmonic nanostructures, denser than a single data bit. 17-21 The information density might be even further increased by addre...

Enhanced light–matter interactions in dielectric nanostructures via machine-learning approach

“…40 Recently, deep learning approaches, based on the artificial neural networks (ANNs), have emerged as a revolutionary and robust methodology in nanophotonics. [41][42][43][44][45][46][47][48][49][50][51][52][53][54][55] Indeed, applying the deep learning algorithms to the nanophotonic inverse design can introduce remarkable design flexibility that can go far beyond that of the conventional methods. The inverse design approach works based one the training process, that enables fast prediction of complex optical properties of nanostructures with intricate architectures.…”

Accelerating optics design optimizations with deep learning

“…2 The current resurgence of neural computing in the form of deep learning (DL) 3 has raised the intriguing possibility of artificial intelligencebased methods that could potentially overcome the "curse of dimensionality." The application of DL has shown early promise in the design of optical thin-films, 4 nanostructures, [5][6][7][8] metasurfaces, [9][10][11] and integrated photonics. 12,13 Optics design differs from pattern recognition problems (a space where DL has achieved remarkable success) in many ways: (1) performance is often very sensitive to variations in design parameters; (2) large datasets are difficult to generate although labeling of data is automatic; (3) performance requirements are often stringent and, hence, uncertainties in the model are not acceptable; and (4) a given response can be realized through multiple designs, while a single design has a unique response (nonuniqueness).…”

A Bidirectional Deep Neural Network for Accurate Silicon Color Design

“…The application of ANNs has emerged recently showing power new capabilities in various scenario including computer vision, speech and face recognition, language and image processing, etc. A deep learning approach can analyze the scattering spectra of silicon nanostructure within the diffraction‐limited area and output digital information, which breaks the limits of optical information storage and already beats the Blu‐ray Disk technology . In the field of nanophotonics, computational inverse design can reshape the landscape and techniques available to complex and emerging applications .…”

“…A deep learning approach can analyze the scattering spectra of silicon nanostructure within the diffraction-limited area and output digital information, which breaks the limits of optical information storage and already beats the Blu-ray Disk technology. [8] In the field of nanophotonics, computational inverse design can reshape the landscape and techniques available to complex and emerging applications. [9] Recent advancements in deep neural networks (DNNs) have demonstrated efficient forward-modeling that can predict resonance spectrum accurately, and perform the inverse design of photonic device structures.…”