MSGID: 1:317/3 649e5a8a   
   PID: hpt/lnx 1.9.0-cur 2019-01-08   
   TID: hpt/lnx 1.9.0-cur 2019-01-08   
    Transferring data with many colors of light simultaneously    
    The new photonic chip enables exponentially faster and more energy-   
   efficient artificial intelligence    
      
     Date:   
         June 29, 2023   
     Source:   
         Columbia University School of Engineering and Applied Science   
     Summary:   
         Scientists have developed a fast and extremely efficient method   
         for transferring huge amounts of data. The technique uses dozens   
         of frequencies of light to transfer several streams of information   
         over a fiber optic cable simultaneously.   
      
      
         Facebook Twitter Pinterest LinkedIN Email   
      
   ==========================================================================   
   FULL STORY   
   ==========================================================================   
   The data centers and high-performance computers that run artificial   
   intelligence programs, such as large language models, aren't limited by   
   the sheer computational power of their individual nodes. It's another   
   problem - - the amount of data they can transfer among the nodes --   
   that underlies the "bandwidth bottleneck" that currently limits the   
   performance and scaling of these systems.   
      
   The nodes in these systems can be separated by more than one   
   kilometer. Since metal wires dissipate electrical signals as heat when   
   transferring data at high speeds, these systems transfer data via   
   fiber-optic cables. Unfortunately, a lot of energy is wasted in the   
   process of converting electrical data into optical data (and back again)   
   as signals are sent from one node to another.   
      
   In a study published today in Nature Photonics, researchers at Columbia   
   Engineering demonstrate an energy-efficient method for transferring   
   larger quantities of data over the fiber-optic cables that connect the   
   nodes. This new technology improves on previous attempts to transmit   
   multiple signals simultaneously over the same fiber-optic cables. Instead   
   of using a different laser to generate each wavelength of light, the   
   new chips require only a single laser to generate hundreds of distinct   
   wavelengths of light that can simultaneously transfer independent streams   
   of data.   
      
   A simpler, more energy-efficient method for data transfer The   
   millimeter-scale system employs a technique called wavelength-division   
   multiplexing (WDM) and devices called Kerr frequency combs that take a   
   single color of light at the input and create many new colors of light   
   at the output.   
      
   The critical Kerr frequency combs developed by Michal Lipson, Higgins   
   Professor of Electrical Engineering and Professor of Applied Physics,   
   and Alexander Gaeta, David M. Rickey Professor of Applied Physics and   
   Materials Science and Professor of Electrical Engineering, allowed the   
   researchers to send clear signals through separate and precise wavelengths   
   of light, with space in between them.   
      
   "We recognized that these devices make ideal sources for optical   
   communications, where one can encode independent information channels on   
   each color of light and propagate them over a single optical fiber," says   
   senior author Keren Bergman, Charles Batchelor Professor of Electrical   
   Engineering at Columbia Engineering, where she also serves as the faculty   
   director of the Columbia Nano Initiative. This breakthrough could allow   
   systems to transfer exponentially more data without using proportionately   
   more energy.   
      
   The team miniaturized all of the optical components onto chips roughly   
   a few millimeters on each edge for generating light, encoded them with   
   electrical data, and then converted the optical data back into an   
   electrical signal at the target node. They devised a novel photonic   
   circuit architecture that allows each channel to be individually   
   encoded with data while having minimal interference with neighboring   
   channels. That means the signals sent in each color of light don't   
   become muddled and difficult for the receiver to interpret and convert   
   back into electronic data.   
      
   "In this way, our approach is much more compact and energy-efficient   
   than comparable approaches," says the study's lead author Anthony Rizzo,   
   who conducted this work while a PhD student in the Bergman lab and   
   is now a research scientist at the U.S. Air Force Research Laboratory   
   Information Directorate. "It is also cheaper and easier to scale since   
   the silicon nitride comb generation chips can be fabricated in standard   
   CMOS foundries used to fabricate microelectronics chips rather than   
   in expensive dedicated III- V foundries."  The compact nature of these   
   chips enables them to directly interface with computer electronics chips,   
   greatly reducing the total energy consumption since the electrical data   
   signals only have to propagate over millimeters of distance rather than   
   tens of centimeters.   
      
   Bergman noted, "What this work shows is a viable path towards both   
   dramatically reducing the system energy consumption while simultaneously   
   increasing the computing power by orders of magnitude, allowing artificial   
   intelligence applications to continue to grow at an exponential rate   
   with minimal environmental impact."  Exciting results pave the way to   
   real-world deployment In experiments, the researchers managed to transmit   
   16 gigabits per second per wavelength for 32 distinct wavelengths of   
   light for a total single-fiber bandwidth of 512 Gb/s with less than   
   one bit in error out of one trillion transmitted bits of data. These   
   are incredibly high levels of speed and efficiency. The silicon chip   
   transmitting the data measured just 4 mm x 1 mm, while the chip that   
   received the optical signal and converted it into an electrical signal   
   measured just 3 mm x 1 mm -- both smaller than a human fingernail.   
      
   "While we used 32 wavelength channels in the proof-of-principle   
   demonstration, our architecture can be scaled to accommodate over 100   
   channels, which is well within the reach of standard Kerr comb designs,"   
   Rizzo adds.   
      
   These chips can be fabricated using the same facilities used to make   
   the microelectronics chips found in a standard consumer laptop or   
   cellphone, providing a straightforward path to volume scaling and   
   real-world deployment.   
      
   The next step in this research is to integrate the photonics with   
   chip-scale driving and control electronics to further miniaturize   
   the system.   
      
       * RELATED_TOPICS   
             o Matter_&_Energy   
                   # Optics # Energy_Technology # Microarrays #   
                   Telecommunications # Physics # Energy_and_Resources #   
                   Engineering # Consumer_Electronics   
       * RELATED_TERMS   
             o Scientific_method o Earth_science o Radiocarbon_dating o   
             Concorde o Copper o Natural_gas o Methane o Combustion   
      
   ==========================================================================   
   Story Source: Materials provided by   
   Columbia_University_School_of_Engineering_and_Applied Science. Note:   
   Content may be edited for style and length.   
      
      
   ==========================================================================   
   Journal Reference:   
      1. Anthony Rizzo, Asher Novick, Vignesh Gopal, Bok Young Kim,   
      Xingchen Ji,   
         Stuart Daudlin, Yoshitomo Okawachi, Qixiang Cheng, Michal Lipson,   
         Alexander L. Gaeta, Keren Bergman. Massively scalable Kerr   
         comb-driven silicon photonic link. Nature Photonics, 2023; DOI:   
         10.1038/s41566-023- 01244-7   
   ==========================================================================   
      
   Link to news story:   
   https://www.sciencedaily.com/releases/2023/06/230629125713.htm   
      
   --- up 1 year, 17 weeks, 3 days, 10 hours, 50 minutes   
    * Origin: -=> Castle Rock BBS <=- Now Husky HPT Powered! (1:317/3)   
   SEEN-BY: 15/0 106/201 114/705 123/120 153/7715 218/700 226/30 227/114   
   SEEN-BY: 229/110 112 113 307 317 400 426 428 470 664 700 291/111 292/854   
   SEEN-BY: 298/25 305/3 317/3 320/219 396/45 5075/35   
   PATH: 317/3 229/426