• PDF / 18,734,273 Bytes
• 7 Pages / 585 x 783 pts Page_size
• 39 Downloads / 236 Views

DOWNLOAD

REPORT

---

Introduction Optical fibers have, for several decades, carried most of our long-distance communications. This is because optical signaling combines ultrahigh bandwidths, low losses, and high energy efficiencies, to deliver an overall performance (for signal transmission) far superior to that available in the electrical domain. Short-haul optical signaling, especially for rack-to-rack connections in data centers, is also now commonplace. The next obvious step is to move to interchip, and even intrachip, optical signaling. This could alleviate the bandwidth and power density limitations that currently act as bottlenecks to the performance of traditional semiconductor processing systems. This migration of optical signaling, from long-haul to on-chip applications, has benefited from rapid recent progress in silicon photonics that now offers a wide range of complementary metal oxide semiconductor (CMOS)-compatible integrated photonic components, including sources, detectors, modulators, switches, and couplers.1 Indeed, a single-chip processor that communicates directly using light has already been demonstrated.2 If interchip and intrachip signaling are likely to be lightbased, it is pertinent to ask whether we should aim to carry out processing tasks, such as memory, arithmetic, and logic, directly in the optical domain. Indeed, one might imagine the goal of realizing integrated photonic processors and memories

that supplement, or in some applications replace, electronic implementations. How realistic is such a goal? One possible approach, that we discuss here, is by combining chalcogenide phase-change materials (PCMs) with standard silicon photonic devices and circuits. By this method, it is possible to deliver photonic devices for nonvolatile binary and multilevel memory, for arithmetic and logic processing, and for neuronal and synaptic hardware mimics. Moreover, these devices can be readily combined into larger-scale systems to provide, for example, all-optical non-von Neumann arithmetic and brainlike (neuromorphic) computing.

Device and system concepts Binary and multilevel memory Probably the simplest integrated phase-change photonic device is that for the provision of memory. While previous promising approaches to the realization of integrated photonic memories have been reported,3–5 such devices were essentially inherently volatile. However, by combining chalcogenide PCMs with integrated photonics, we can achieve truly nonvolatile storage, with stored states persevering for years, if not decades.6 A schematic of the simplest integrated phase-change photonic memory device7,8 is given in Figure 1a–b. A small (micrometer- or submicrometer-size) cell of PCM, in this

C. David Wright, Department of Engineering, University of Exeter, UK; [email protected] Harish Bhaskaran, Department of Materials, University of Oxford, UK; [email protected] Wolfram H.P. Pernice, Department of Physics, University of Münster, Germany; [email protected] doi:10.1557/mrs.2019.203

• VOLUMECore • www.