Published in IET Circuits, Devices & Systems Received on 31st March 2010 Revised on 8th October 2010 doi: 10.1049/iet-cds.2010.0105 Special Issue on Optical Computing Circuits, Devices and Systems ISSN 1751-858X Optical logic elementary circuits A. Bogoni 1,2 L. Potı` 1 A.E. Willner 2 P. Ghelfi 1 C. Porzi 3 M. Scaffardi 1 G. Meloni 3 G. Berrettini 3 F. Fresi 3 E. Lazzeri 3 X. Wu 2 1 CNIT-Photonic Networks Laboratory, via Moruzzi 1, 56124, Pisa, Italy 2 University of Southern California, Optical Communications Laboratory, 3740 McClintock, 90089, Los Angeles, USA 3 Scuola Superiore Sant’Anna-CEIICP, via Moruzzi 1, 56124, Pisa, Italy E-mail: antonella.bogoni@cnit.it Abstract: Elementary blocks, performing logic operations, are the building elements for more complex subsystems implementing all-optical digital processing. They can potentially enable next generation optical networks and optical computing, overcoming the limitations of the electronics bandwidth, also guaranteeing scalability, transparency, easy reconfigurability and modularity. Finally, integrated technologies can reduce power consumption, footprint and cost. 1 Introduction The penetration of high-performance computing systems is increasing rapidly and multicore processors are distributed everywhere nowadays because of the computational demand coming from all kinds of application (internet gaming, high- definition television etc.). In order to fulfil the growing need, the easiest (and cheapest) approach is based on well- established technologies and has the objective of increasing the processing capabilities trying to overcome, step by step, the limitations imposed by electronic circuits. The market is ready for 100 parallel core processors where research has to deal with device speed, density, high-speed interconnections, power dissipation and power consumption. On the other side, one possible mid–long-term alternative solution can take benefit from photonic technologies and the fast evolution of optical digital processing. The main advantage of photonic processing with respect to electronic processing is the high speed [1, 2], combined with scalability, transparency, easy reconfigurability and modularity. Nevertheless, photonics is in its early stage and the realisation of complete all-optical computing system is still far. The main reason for that is the immature technology and the lack of key subsystems such as all-optical memories. The research in the last 15 years is focused on two lines: ‘subsystems’ [3–7] and new ‘technologies’ [8–10]. In the former most of the efforts have been made in order to exploit fibre or device non-linearities in order to demonstrate the feasibility for simple as well as more complex logical operation mainly considering operation speed but without taking into account power consumption, complexity and as a consequence, integrability. The latter, on the other side, is focused on new material where non-linear effects can take place with higher effectiveness in a smaller footprint and with low power consumption. In this case the compatibility among different materials and, at the end, with fibre communication is neither obvious nor guaranteed. Despite the present limitations, photonic digital processing shows promise in the applications where fast computation speed is required. One of these scenarios is represented by the all-optical short-range photonic interconnection networks. The improvement of the present high-performance computing systems is hampered by the bottleneck of the chip-to-chip and chip-to-memory communication. The limits are the high wiring density, the high power consumption and the limited throughput [11]. Photonic interconnection networks can overcome the limitations of the electronic interconnection networks by enabling high bit rate communications, data format transparency and electromagnetic field immunity. Moreover, they can reduce the wiring density and the power consumption. In such networks, photonic digital processing can be the most suitable paradigm for simple and ultra-fast control and switching operations, since it reduces the packet latency to the optical time-of-flight. In this paper, some of the fundamental subsystems required for photonic digital processing are reviewed and best implementations in terms of speed operation are reported. They exploit the non-linear optical media available in the last decade, even if still with technological issues. Finally, promising technologies for on-chip-based solutions able to guarantee low power consumption, low cost and high integration density are mentioned. The paper is organised as follows: in Section 2 photonic logic operation are considered and a demonstration at 160 Gb/s is reported using a periodically poled lithium niobate (PPLN) waveguide; in Section 3 we describe an optical comparator where logic operations are cascaded together with an optical serial to parallel converter; optical adder is demonstrated in Section 4 whereas analogue-to-digital (A/D) and digital-to- analogue (D/A) conversions are described in Section 5. Section 6 summarises examples of applications of photonic logic operations for high-capacity optical networks and Section 7 mentions the evolution of the available integrated technologies to guarantee low power consumption, low cost and high integration density solutions. 76 IET Circuits Devices Syst., 2011, Vol. 5, Iss. 2, pp. 76–83 & The Institution of Engineering and Technology 2011 doi: 10.1049/iet-cds.2010.0105 www.ietdl.org