physicsresearch news

Deep learning pushes the limits of optical information storage

By 24th January 2019 June 11th, 2019 No Comments

A team of researchers from CEMES and LAAS-CNRS, both in Toulouse has put forward a new ultrahigh-density optical storage technology that can store up to nine bits of information per diffraction-limited area. The technology works thanks to a deep learning approach and overcomes the physical constraints of compact disks (CDs), DVDs and Blu-ray disks, which can only store one bit per diffraction-limited area. It is based on silicon and could thus be produced on a large scale using mass-market-ready complementary metal-oxide-semiconductor (CMOS) technology.

The storage capacity of magnetic hard disk drives is reaching its limits. Optical data storage is a promising alternative because it can store much higher volumes of information (over 10 terabytes) and for longer periods (hundreds of years compared to just five to ten years for hard disk drives).

The storage capacity of current optical data storage technologies such as the CD, DVD and Blu-ray disk is, however, restricted by the diffraction limit of the laser used to write to the disk. This means that they can store just a single bit per diffraction-limited area (“0” or “1”) to reach 200 gigabytes per disk at the most.

The new technique, developed by a  team led by Guilhem Larrieu of LAAS-CNRS encodes multiple bits of digital information in subwavelength silicon nanostructures. Thanks to the high refractive index of silicon, these nanostructures scatter light with little loss. The researchers then retrieve the information stored in these particles using a machine-learning-based approach in which an artificial neuronal network (ANN) analyses their scattering spectra.

Read the research paper: Pushing the limits of optical information storage using deep learning, Peter R. Wiecha et al. Nature Nanotechnology