Noise-resilient and high-speed deep learning with coherent silicon photonics.

The explosive growth of deep learning applications has triggered a new era in computing hardware, targeting the efficient deployment of multiply-and-accumulate operations. In this realm, integrated photonics have come to the foreground as a promising energy efficient deep learning technology platform for enabling ultra-high compute rates. However, despite integrated photonic neural network layouts have already penetrated successfully the deep learning era, their compute rate and noise-related characteristics are still far beyond their promise for high-speed photonic engines. Herein, we demonstrate experimentally a noise-resilient deep learning coherent photonic neural network layout that operates at 10GMAC/sec/axon compute rates and follows a noise-resilient training model. The coherent photonic neural network has been fabricated as a silicon photonic chip and its MNIST classification performance was experimentally evaluated to support accuracy values of >99% and >98% at 5 and 10GMAC/sec/axon, respectively, offering 6× higher on-chip compute rates and >7% accuracy improvement over state-of-the-art coherent implementations.

Ultra-sensitive refractive index sensor using CMOS plasmonic transducers on silicon photonic interferometric platform.

Optical refractive-index sensors exploiting selective co-integration of plasmonics with silicon photonics has emerged as an attractive technology for biosensing applications that can unleash unprecedented performance breakthroughs that reaps the benefits of both technologies. However, towards this direction, a major challenge remains their integration using exclusively CMOS-compatible materials. In this context, herein, we demonstrate, for the first time to our knowledge, a CMOS-compatible plasmo-photonic Mach-Zehnder-interferometer (MZI) based on aluminum and Si3N4 waveguides, exhibiting record-high bulk sensitivity of 4764 nm/RIU with clear potential to scale up the bulk sensitivity values by properly engineering the design parameters of the MZI. The proposed sensor is composed of Si3N4 waveguides butt-coupled with an aluminum stripe in one branch to realize the sensing transducer. The reference arm is built by Si3N4 waveguides, incorporating a thermo-optic phase shifter followed by an MZI-based variable optical attenuation stage to maximize extinction ratio up to 38 dB, hence optimizing the overall sensing performance. The proposed sensor exhibits the highest bulk sensitivity among all plasmo-photonic counterparts, while complying with CMOS manufacturing standards, enabling volume manufacturing.

Plasmonics co-integrated with silicon nitride photonics for high-sensitivity interferometric biosensing.

We demonstrate a photonic integrated Mach-Zehnder interferometric sensor, utilizing a plasmonic stripe waveguide in the sensing branch and a photonic variable optical attenuator and a phase shifter in the reference arm to optimize the interferometer operation. The plasmonic sensor is used to detect changes in the refractive index of the surrounding medium exploiting the accumulated phase change of the propagating Surface-Plasmon-Polariton (SPP) mode that is fully exposed in an aqueous buffer solution. The variable optical attenuation stage is incorporated in the reference Si3N4 branch, as the means to counter-balance the optical losses introduced by the plasmonic branch and optimize interference at the sensor output. Bulk sensitivity values of 1930 nm/RIU were experimentally measured for a Mach Zehnder Interferometer (MZI) with a Free Spectral Range of 24.8 nm, along with extinction ratio of more than 35 dB, demonstrating the functional benefits of the co-integration of plasmonic and photonic waveguides.

CMOS plasmonics in WDM data transmission: 200 Gb/s (8 × 25Gb/s) transmission over aluminum plasmonic waveguides.

We demonstrate wavelength-division-multiplexed (WDM) 200 Gb/s (8 × 25 Gb/s) data transmission over 100 µm long aluminum (Al) surface-plasmon-polariton (SPP) waveguides on a Si3N4 waveguide platform at telecom wavelengths. The Al SPP waveguide was evaluated in terms of signal integrity by performing bit-error-rate (BER) measurements that revealed error-free operation for all eight 25 Gb/s non-return-to-zero (NRZ) modulated data channels with power penalties not exceeding 0.2 dB at 10-9. To the best of our knowledge, this is the first demonstration of WDM enabled data transmission over complementary-metal-oxide-semiconductor (CMOS) SPP waveguides fueling future development of CMOS compatible plasmo-photonic devices for on-chip optical interconnections.