Wen Zhu and co-authors from Oxford, Exeter and Muenster provide a nice summary (in the MRS Bulletin) of the state-of-the art in terms of programming schemes for integrated phase-change photonic devices, as used for in-memory and neuromorphic computing. Specificaly we compare performances of PCM devices using optoelectronic programming schemes and show that energy consumption can be significantly reduced to 60 pJ using picosecond (ps) optical pulse programming and plasmonic nanogap devices with a programming speed approaching 1 GHz (remember the readout or inference speed of our devices is 10s of GHz, but the programming speed has been significantly slower than that to date). When integrated into processing systems and compared with digital electronic accelerators – application-specific integrated circuits (ASICs) and graphics processing units (GPUs) – these Fun-COMP photonic systems promise 1−3 orders higher compute density and energy efficiency, although much more work toward commercialization is still required. For more information see https://link.springer.com/article/10.1557/s43577-022-00358-7 for the full article
https://www.nature.com/articles/s41586-020-03070-1The Fun-COMP team have continued to improve the design and operation of their photonic tensor core (PTC) processor, that we reported in Nature last year (see https://www.nature.com/articles/s41586-020-03070-1). In work recently reported in the Nanophotonics journal, Fun-COMP researchers showcased an optimized layout of their photonic tensor core – one that is designed to perform real valued matrix vector multiplications and operates at telecommunication wavelengths. We deploy the well-studied phase-change material Ge2Sb2Te5 (GST) as an optical attenuator to perform single positive valued multiplications. In order to generalize the multiplication to arbitrary real factors, we develop a novel symmetric multiplication unit which directly includes a reference-computation branch. The variable GST attenuator enables a modulation depth of 5 dB over a wavelength range of 100 nm with a wavelength dependency below 0.8 dB. The passive photonic circuit itself ensures equal coupling to the main-computation and reference-computation branch over the complete wavelength range. For the first time, we integrate wavelength multiplexers (MUX) together with a photonic crossbar array on-chip, paving the way towards fully integrated systems. For more details see the paper at https://www.degruyter.com/document/doi/10.1515/nanoph-2021-0752/html
The ever-increasing demands for data processing and storage will require seamless monolithic co-integration of electronics and photonics. The extreme size disparity however between CMOS electronics and dielectric photonics inhibits the realization of efficient and compact electrically driven photonic switches, logic and routing elements. The Fun-COMP team have however recently reached an important milestone in harmonizing the two domains, by demonstrating an electrically reconfigurable, ultra-compact and nonvolatile memory that is optically accessible. The platform relies on localized heat, generated within a plasmonic structure; this uniquely allows for both optical and electrical readout signals to be interlocked with the material state of the PCM while still ensuring that the writing operation is electrically decoupled. Through miniaturization and effective thermal engineering, the Fun-COMP team also achieved unprecedented energy efficiency, opening up a path towards low-energy optoelectronic hardware for neuromorphic and in-memory computing. The work was recently reported in the journal Advanced Science, see https://onlinelibrary.wiley.com/doi/10.1002/advs.202200383
Fun-COMP researchers have recently showcased a new type of photonic processing engine that detects temporal correlations in data – with applications ranging from social media analysis, to financial forecasting, the detection of hacking threats and much more. The work has been published in the leading journal Science Advances – see https://www.science.org/doi/10.1126/sciadv.abn3243 for detailed information.
In January, Johannes and colleagues published an extremely interesting article in Nature, showcasing some of the results of the Fun-COMP research, and highlighting the efficiency and computational speed which such a class of devices can achieve. The article can be read at this address.
Additionally, the University of Exeter published an interview with Prof. Wright, which can be read here.
The current circumstances indeed took a toll on how research is carried out, not only on the Fun-COMP project collaborators, but on a global scale. Nevertheless, we’ve been able to continue our work, solving or working around the various issues the social distancing compels to tackle. The list of recent publications from the researchers collaborating to this project proofs the dedication and commitment to their work. We’d like to share their success, and congratulate with each one of them for their accomplishments.
A note of merit goes to Xuan and Johannes, which along with their research partners published extremely interesting results concerning the investigation of the potential and scalability of the Non-von Neumann optical unit cell.
Xuan’s work investigates the behavior and potential range of applications of the NvN unit cell on different photonic platforms: silicon, and silicon nitride. The results help to identify the issues and advantages deriving from the adoption of either of the concepts, which is of high importance for the foreseeable future of the research project.
Johannes’s work explores the scalability of the NvN unit cell as a memory platform towards realistic scenarios. In particular, the authors developed a test photonic device capable to store and retrieve in an all-optical fashion up to 512 bits, implementing 256 individually addressed unit cells.
Here follows a shortlist of all the publications from Fun-COMP research collaborators since January 2020.
Lugnan, Alessio, Andrew Katumba, Floris Laporte, Matthias Freiberger, Stijn Sackesyn, C. Ma, Emmanuel Gooskens, Joni Dambre, and Peter Bienstman. “Photonic neuromorphic information processing and reservoir computing.” APL Photonics 5, no. 2 (2020): 020901. DOI: 10.1063/1.5129762
Harkhoe, Krishan, Guy Verschaffelt, Andrew Katumba, Peter Bienstman, and Guy Van der Sande. “Demonstrating delay-based reservoir computing using a compact photonic integrated chip.” Optics Express 28, no. 3 (2020): 3086-3096. DOI: 10.1364/OE.382556
Faneca, Joaquin, Santiago G-C. Carrillo, Emanuele Gemo, Carlota Ruiz de Galarreta, Thalía Domínguez Bucio, Frederic Y. Gardes, Harish Bhaskaran, Wolfram HP Pernice, C. David Wright, and Anna Baldycheva. “Performance characteristics of phase-change integrated silicon nitride photonic devices in the O and C telecommunications bands.” Optical Materials Express 10, no. 8 (2020): 1778-1791. DOI: 10.1364/OME.10.001778
Gemo, Emanuele, Sameer V. Kesava, Carlota Ruíz de Galarreta, Liam Trimby, Santiago G-C. Carrillo, Moritz Riede, Anna Baldycheva, Arseny M. Alexeev, and C. David Wright “Simple technique for determining the refractive index of phase-change materials using near-infrared reflectometry” Optical Material Express, vol. 10, no. 3 (2020): 1675-1686. DOI: 10.1364/OME.395353
Li, Xuan, Nathan Youngblood, Zengguang Cheng, Santiago García-Cuevas Carrillo, Emanuele Gemo, Wolfram H.P. Penice, C. David Wright, and Harish Bhaskaran. “Experimental investigation of silicon and silicon nitride platforms for phase change photonic in-memory computing” Optica 7, no. 3 (2020): 218-225. DOI: 10.1364/OPTICA.379228
Feldmann, Johannes, Nathan Youngblood, Xuan Li, C. David Wright, Harish Bhaskaran, and Wolfram HP Pernice. “Integrated 256 cell photonic phase-change memory with 512-bit capacity.” IEEE Journal of Selected Topics in Quantum Electronics 26, no. 2 (2019): 1-7. DOI: 10.1109/JSTQE.2019.2956871