Powering AI at the edge: A robust, memristor-based binarized neural network with near-memory computing and miniaturized solar cell.

Memristor-based neural networks provide an exceptional energy-efficient platform for artificial intelligence (AI), presenting the possibility of self-powered operation when paired with energy harvesters. However, most memristor-based networks rely on analog in-memory computing, necessitating a stable and precise power supply, which is incompatible with the inherently unstable and unreliable energy harvesters. In this work, we fabricated a robust binarized neural network comprising 32,768 memristors, powered by a miniature wide-bandgap solar cell optimized for edge applications. Our circuit employs a resilient digital near-memory computing approach, featuring complementarily programmed memristors and logic-in-sense-amplifier. This design eliminates the need for compensation or calibration, operating effectively under diverse conditions. Under high illumination, the circuit achieves inference performance comparable to that of a lab bench power supply. In low illumination scenarios, it remains functional with slightly reduced accuracy, seamlessly transitioning to an approximate computing mode. Through image classification neural network simulations, we demonstrate that misclassified images under low illumination are primarily difficult-to-classify cases. Our approach lays the groundwork for self-powered AI and the creation of intelligent sensors for various applications in health, safety, and environment monitoring.

Investigation of antireflective and hydrophobic properties in polycrystalline GaN films with dual porosity produced by CVD.

We optimized the deposition conditions of polycrystalline nanoporousGaN coatings produced by Chemical Vapor Deposition on Si substrates, by exploring the effect produced by the Ga holder shape, the initial amount of Ga, the reaction deposition time and the metallic catalyst used. Such polycrystalline films probed to act as antireflective coatings by reducing the reflectance of Si substrates by 50% or more, and that of flat GaN samples by 40% in the UV and 83% in the visible, at the same time that they exhibit an almost constant reflectance from 400 to 800 nm, important to develop UV sensors with enhanced sensitivity. Furthermore, the polycrystalline nanoporous coatings we developed exhibit hydrophobic behaviour, with a static contact angle of 119°, and a contact angle hysteresis of 4.5°, which might contribute to enlarge the durability of such functional films, by the self cleaning effect induced.

Fully porous GaN p-n junction diodes fabricated by chemical vapor deposition.

Porous GaN based LEDs produced by corrosion etching techniques demonstrated enhanced light extraction efficiency in the past. However, these fabrication techniques require further postgrown processing steps, which increases the price of the final system. Also, the penetration depth of these etching techniques is limited, and affects not only the semiconductor but also the other elements constituting the LED when applied to the final device. In this paper, we present the fabrication of fully porous GaN p-n junctions directly during growth, using a sequential chemical vapor deposition (CVD) process to produce the different layers that form the p-n junction. We characterized their diode behavior from room temperature to 673 K and demonstrated their ability as current rectifiers, thus proving the potential of these fully porous p-n junctions for diode and LEDs applications. The electrical and luminescence characterization confirm that high electronic quality porous structures can be obtained by this method, and we believe this investigation can be extended to other III-N materials for the development of white light LEDs, or to reduce reflection losses and narrowing the output light cone for improved LED external quantum efficiencies.

Reduced workfunction intermetallic seed layers allow growth of porous n-GaN and low resistivity, ohmic electron transport.

Porous GaN crystals have been successfully grown and electrically contacted simultaneously on Pt- and Au-coated silicon substrates as porous crystals and as porous layers. By the direct reaction of metallic Ga and NH(3) gas through chemical vapor deposition, intermetallic metal-Ga alloys form at the GaN-metal interface, allowing vapor-solid-solid seeding and subsequent growth of porous GaN. Current-voltage and capacitance-voltage measurements confirm that the intermetallic seed layers prevent interface oxidation and give a high-quality reduced workfunction contact that allows exceptionally low contact resistivities. Additionally, the simultaneous formation of a lower workfunction intermetallic permits ohmic electron transport to n-type GaN grown using high workfunction metals that best catalyze the formation of porous GaN layers and may be employed to seed and ohmically contact a range of III-N compounds and alloys for broadband absorption and emission.

Chemical vapor deposition of porous GaN particles on silicon.

We present a technique for the direct deposition of nanoporous GaN particles on Si substrates without requiring any post-growth treatment. The internal morphology of the nanoporous GaN particles deposited on Si substrates by using a simple chemical vapor deposition approach was investigated, and straight nanopores with diameters ranging between 50 and 100 nm were observed. Cathodoluminescence characterization revealed a sharp and well-defined near band-edge emission at ∼365 nm. This approach simplifies other methods used for this purpose, such as etching and corrosion techniques that can damage the semiconductor structure and modify its properties.