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.

In this paper, we study both theoretically and experimentally the sensitivity of bimodal interferometric sensors where interference occurs between two plasmonic modes with different properties propagating in the same physical waveguide. In contrast to the well-known Mach–Zehnder interferometric (MZI) sensor, we show for the first time that the sensitivity of the bimodal sensor is independent of the sensing area length. This is validated by applying the theory to an integrated plasmo-photonic bimodal sensor that comprises an aluminum (Al) plasmonic stripe waveguide co-integrated between two accessible SU-8 photonic waveguides. A series of such bimodal sensors utilizing plasmonic stripes of different lengths were numerically simulated, demonstrating bulk refractive index (RI) sensitivities around 5700 nm/RIU for all sensor variants, confirming the theoretical results. The theoretical and numerical results were also validated experimentally through chip-level RI sensing experiments on three fabricated SU-8/Al bimodal sensors with plasmonic sensing lengths of 50, 75, and 100 μm. The obtained experimental RI sensitivities were found to be very close and equal to 4464, 4386, and 4362 nm/RIU, respectively, confirming that the sensing length has no effect on the bimodal sensor sensitivity. The above outcome alleviates the design and optical loss constraints, paving the way for more compact and powerful sensors that can achieve high sensitivity values at ultra-short sensing lengths.

Plasmonic refractive index sensors are essential for detecting subtle variations in the ambient environment through surface plasmon interactions. Current efforts utilizing CMOS-compatible, plasmo-photonic Mach-Zehnder interferometers with active power balancing exhibit high sensitivities at the cost of fabrication and measurement complexity. Alternatively, passive bimodal plasmonic interferometers based on SU-8 waveguides offer cost-effectiveness and a smaller device footprint. However, their waveguides are exposed to ambient conditions and currently lack the necessary opto-mechanical isolation. In this work, we introduce the new SX AR LWL 2.0 polymer cladding material, which provides a high refractive index contrast to the SU-8 core, while ensuring compatibility with SU-8 waveguides. Our sensors consist of an aluminum plasmonic stripe and a bilayer SU-8 photonic waveguide core which is cladded with the SX AR LWL 2.0 polymer. They achieve a sensitivity of (6108 ± 13) nm/RIU (refractive index unit), where the error indicates the data fitting tolerance, and a limit of detection of 1.6 × 10⁻⁷ RIU, in aqueous solutions, surpassing both traditional and polymer-based bimodal plasmo-photonic sensors. This new cladding material paves the way for more efficient and integrated bimodal plasmonic refractive index sensing platforms for a wide range of applications.

In this work we experimentally demonstrate a Si ₃ N ₄ photonic integrated circuit which offers row decoding and RAM addressing functionalities. The passive integrated structure comprises a MRR-based wavelength filtering bank scheme in a 2 × 4 configuration, which reveals a suppression ratio in the range of 12–25 dB. The performance of the optical circuit has been evaluated in a system-level testbed, where successful addressing in one RAM row has been achieved. Error-free operation has been accomplished for all cases under study, with the whole row decoder system’s performance to offer a total power penalty of 2.5 dB.

This work presents a dual plasmo-photonic branch Mach-Zehnder interferometer refractive index sensor integrated on a Si₃N₄ platform, with both the sensor and reference arms hosting identical aluminum plasmonic metal stripes. Experimental evaluation of the configuration reveals a bulk sensitivity up to 8801 nm/RIU and an increased environmental noise resilience compared to state-of-the-art plasmo-photonic MZI sensors with the plasmonic section residing only on the sensor arm. The enhanced noise resilience along with the high sensitivity results in an experimentally obtained low detection limit value of 4.4 × 10⁻⁶ RIU, im-proved by an order of magnitude compared to the detection limit of 5 × 10⁻⁵ RIU that was measured in a reference MZI sensor with a plasmonic stripe only at its sensor arm.