Following a decade of radical advances in the areas of integrated photonics and computing architectures, we discuss the use of optics in the current computing landscape attempting to redefine and refine their role based on the progress in both research fields. We present the current set of critical challenges faced by the computing industry and provide a thorough review of photonic Network-on-Chip (pNoC) architectures and experimental demonstrations, concluding to the main obstacles that still impede the materialization of these concepts. We propose the employment of optics in chip-to-chip (C2C) computing architectures rather than on-chip layouts toward reaping their benefits while avoiding technology limitations on the way to manycore set-ups. We identify multisocket boards as the most prominent application area and present recent advances in optically enabled multisocket boards, revealing successful 40 Gb/s transceiver and routing capabilities via integrated photonics. These results indicate the potential to bring energy consumption down by more than 60% compared to current QuickPath Interconnect (QPI) protocol, while turning multisocket architectures into a single-hop low-latency setup for even more than four interconnected sockets, which form currently the electronic baseline. We go one step further and demonstrate how optically-enabled eight-socket boards can be combined via a 256 × 256 Hipoλaos Optical Packet Switch into a powerful 256-node disaggregated system with less than 335 ns latency, forming a highly promising solution for the latency-critical rack-scale memory disaggregation era. Finally, we discuss the perspective for disintegrated computing via optical technologies as a mean to increase the number of synergized high-performance cores overcoming die area constraints, introducing also the concept of cache disintegration via the use of future off-die ultrafast optical cache memory chiplets.

Analog fronthauling is currently promoted as a bandwidth and energy-efficient solution that can meet the requirements of the Fifth Generation (5G) vision for low latency, high data rates and energy efficiency. In this paper, we propose an analog optical fronthaul 5G architecture, fully aligned with the emerging Centralized-Radio Access Network (C-RAN) concept. The proposed architecture exploits the wavelength division multiplexing (WDM) technique and multicarrier intermediate-frequency-over-fiber (IFoF) signal generation per wavelength in order to satisfy the demanding needs of hotspot areas. Particularly, the fronthaul link employs photonic integrated circuit (PIC)-based WDM optical transmitters (Txs) at the baseband unit (BBU), while novel reconfigurable optical add-drop multiplexers (ROADMs) cascaded in an optical bus are used at the remote radio head (RRH) site, to facilitate reconfigurable wavelength switching functionalities up to 4 wavelengths. An aggregate capacity of 96 Gb/s has been reported by exploiting two WDM links carrying multi-IF band orthogonal frequency division multiplexing (OFDM) signals at a baud rate of 0.5 Gbd with sub-carrier (SC) modulation of 64-QAM. All signals exhibited error vector magnitude (EVM) values within the acceptable 3rd Generation Partnership Project (3GPP) limits of 8%. The longest reach to place the BBU away from the hotspot was also investigated, revealing acceptable EVM performance for fiber lengths up to 4.8 km.

In this article, we present for the first time to our knowledge, a DWDM 16×16 silicon photonic cyclic-frequency AWGR device. The AWGR features a channel spacing of 1.06 nm (189 GHz), a free spectral range of 17.80 nm (3.16 THz) and a 3-dB channel bandwidth of 0.65 nm (116 GHz). Its proper cyclic-frequency operation was experimentally verified for all 16×16 channels with channel peak insertion loss values in the range of 3.90 dB to 8.37 dB, yielding a channel loss non-uniformity of 4.47 dB. Its compact footprint of 0.27 mm × 0.71 mm and its low crosstalk value of 21.65 dB highlight its potential for employment in future AWGR-based interconnect schemes in O-band.

We propose a crosstalk (XT)-aware wavelength ( $\lambda $ )-detuning scheme to overcome the non-optimal in-band XT limiting all-to-all optical interconnects based on sub-optimal arrayed waveguide grating routers (AWGR). Experimental validation is demonstrated employing an $8 \times 8$ O-band cyclic Si-AWGR with 10 nm channel spacing and a 5.5 nm channel bandwidth, where the in-band XT limits the simultaneous use of the same $\lambda $ in only 2 out of the 8 possible AWGR I/O routing scenarios, yielding a 10 ⁻⁷ error-floor when >3 channels at the same $\lambda $ are launched. We present experimentally an $8 \times 25$ Gb/s $\lambda $ -switched interconnect exploiting slightly detuned $\lambda $ s around the central $\lambda $ for realizing all 8 possible I/O routing scenarios, enabling an error-free all-to-all interconnect even for sub-optimal AWGRs. Error-free simultaneous transmission via the AWGR of all 8 channels using $4\lambda \text{s}$ and a $\lambda $ utilization factor 2 is shown, with a max. power penalty of 4.64 dB at 10 ⁻⁹ error-rate.

We report an experimental characterization of the long-term stability of a local oscillator (LO) signal generation module for satellite communications based on the optical heterodyning of two InP/Si ₃ N ₄ hybrid integrated lasers. Specifically, we analyze simultaneously the wavelength drift of both lasers in the optical domain and the electrical drift of the generated radiofrequency (RF) beat between the lasers. We identify the main factors that contribute to the overall drift such as the electrical noise coming from the supply sources, the thermal noise coming from the temperature control unit, and the thermal crosstalk between lasers. This investigation compares the stability of an RF signal generated by two approaches. Using two integrated lasers and using a combination of one integrated laser with one external laser. This work also analyzes the difference when lasers are mounted over a standard breadboard with a thermally-controlled. The results show that the best approach is to use both integrated lasers on the same chip. We report a 25 GHz RF signal with long-term stability of 30 MHz measured over 10 hours under free-running operation mode. Moreover, we experimentally demonstrate that integrated lasers can be stabilized by using injection locking. We experimentally demonstrate the optical injection locking of two hybrid InP/Si3N4 lasers when the leader laser and the follower laser are both hybrid integrated lasers of same system on chip. When the laser is locked, we report a locking range of 1.86 GHz.