Search

Biological neural networks effortlessly tackle complex computational problems and excel at predicting outcomes from noisy, incomplete data, a task that poses significant challenges to traditional processors. Artificial neural networks (ANNs), inspired by these biological counterparts, have emerged as powerful tools for deciphering intricate data patterns and making predictions. However, conventional ANNs can be viewed as "point estimates" that do not capture the uncertainty of prediction, which is an inherently probabilistic process. In contrast, treating an ANN as a probabilistic model derived via Bayesian inference poses significant challenges for conventional deterministic computing architectures. Here, we use chaotic light in combination with incoherent photonic data processing to enable high-speed probabilistic computation and uncertainty quantification. Since both the chaotic light source and the photonic crossbar support multiple independent computational wavelength channels, we sample from the output distributions in parallel at a sampling rate of 70.4 GS/s, limited only by the electronic interface. We exploit the photonic probabilistic architecture to simultaneously perform image classification and uncertainty prediction via a Bayesian neural network. Our prototype demonstrates the seamless cointegration of a physical entropy source and a computational architecture that enables ultrafast probabilistic computation by parallel sampling.

Plastic self-adaptation, nonlinear recurrent dynamics and multi-scale memory are desired features in hardware implementations of neural networks, because they enable them to learn, adapt and process information similarly to the way biological brains do. In this work, we experimentally demonstrate these properties occurring in arrays of photonic neurons. Importantly, this is realised autonomously in an emergent fashion, without the need for an external controller setting weights and without explicit feedback of a global reward signal. Using a hierarchy of such arrays coupled to a backpropagation-free training algorithm based on simple logistic regression, we are able to achieve a performance of 98.2% on the MNIST task, a popular benchmark task looking at classification of written digits. The plastic nodes consist of silicon photonics microring resonators covered by a patch of phase-change material that implements nonvolatile memory. The system is compact, robust, and straightforward to scale up through the use of multiple wavelengths. Moreover, it constitutes a unique platform to test and efficiently implement biologically plausible learning schemes at a high processing speed.

Integrated photonic platforms have proliferated in recent years, each demonstrating its own unique strengths and shortcomings. However, given the processing incompatibilities of different platforms, a formidable challenge in the field of integrated photonics still remains for combining the strength of different optical materials in one hybrid integrated platform. Silicon carbide is a material of great interest because of its high refractive index, strong second and third-order non-linearities and broad transparecy window in the visible and near infrared. However, integrating SiC has been difficult, and current approaches rely on transfer bonding techniques, that are time consuming, expensive and lacking precision in layer thickness. Here, we demonstrate high index Amorphous Silicon Carbide (a-SiC) films deposited at 150$^{\circ}$C and verify the high performance of the platform by fabricating standard photonic waveguides and ring resonators. The intrinsic quality factors of single-mode ring resonators were in the range of $Q_{int} = (4.7-5.7)\times10^5$ corresponding to optical losses between 0.78-1.06 dB/cm. We then demonstrate the potential of this platform for future heterogeneous integration with ultralow loss thin SiN and LiNbO$_3$ platforms.

Electronically reprogrammable photonic circuits based on phase-change chalcogenides present an avenue to resolve the von-Neumann bottleneck; however, implementation of such hybrid photonic-electronic processing has not achieved computational success. Here, we achieve this milestone by demonstrating an in-memory photonic-electronic dot-product engine, one that decouples electronic programming of phase-change materials (PCMs) and photonic computation. Specifically, we develop non-volatile electronically reprogrammable PCM memory cells with a record-high 4-bit weight encoding, the lowest energy consumption per unit modulation depth (1.7 nJ per dB) for Erase operation (crystallization), and a high switching contrast (158.5%) using non-resonant silicon-on-insulator waveguide microheater devices. This enables us to perform parallel multiplications for image processing with a superior contrast-to-noise ratio (greater than 87.36) that leads to an enhanced computing accuracy (standard deviation less than 0.007). An in-memory hybrid computing system is developed in hardware for convolutional processing for recognizing images from the MNIST database with inferencing accuracies of 86% and 87%.

Neural processing on devices and circuits is fast becoming a popular approach to emulate biological neural networks. Elaborate CMOS and memristive technologies have been employed to achieve this, including chalcogenide-based in-memory computing concepts. Here we show that nano-scaled films of chalcogenide semiconductors can serve as building-blocks for novel types of neural computations where their tunable electronic and optical properties are jointly exploited. We demonstrate that ultrathin photoactive cavities of Ge-doped Selenide can emulate the computationally powerful non-linear operations of three-factor neo-Hebbian plasticity and the shunting inhibition. We apply this property to solve a maze game through reinforcement learning, as well as a single-neuron solution to the XOR, which is a linearly inseparable problem with point-neurons. Our results point to a new breed of memristors with broad implications for neuromorphic computing.

Modern computation based on the von Neumann architecture is today a mature cutting-edge science. In the Von Neumann architecture, processing and memory units are implemented as separate blocks interchanging data intensively and continuously. This data transfer is responsible for a large part of the power consumption. The next generation computer technology is expected to solve problems at the exascale with 1018 calculations each second. Even though these future computers will be incredibly powerful, if they are based on von Neumann type architectures, they will consume between 20 and 30 megawatts of power and will not have intrinsic physically built-in capabilities to learn or deal with complex data as our brain does. These needs can be addressed by neuromorphic computing systems which are inspired by the biological concepts of the human brain. This new generation of computers has the potential to be used for the storage and processing of large amounts of digital information with much lower power consumption than conventional processors. Among their potential future applications, an important niche is moving the control from data centers to edge devices. The aim of this Roadmap is to present a snapshot of the present state of neuromorphic technology and provide an opinion on the challenges and opportunities that the future holds in the major areas of neuromorphic technology, namely materials, devices, neuromorphic circuits, neuromorphic algorithms, applications, and ethics. The Roadmap is a collection of perspectives where leading researchers in the neuromorphic community provide their own view about the current state and the future challenges. We hope that this Roadmap will be a useful resource to readers outside this field, for those who are just entering the field, and for those who are well established in the neuromorphic community. https://doi.org/10.1088/2634-4386/ac4a83

Over a century ago, Ivan P. Pavlov, in a classic experiment, demonstrated how dogs can learn to associate a ringing bell with food, thereby causing a ring to result in salivation. Today, it is rare to find the use of Pavlovian type associative learning for artificial intelligence (AI) applications even though other learning concepts, in particular backpropagation on artificial neural networks (ANNs) have flourished. However, training using the backpropagation method on 'conventional' ANNs, especially in the form of modern deep neural networks (DNNs), is computationally and energy intensive. Here we experimentally demonstrate a form of backpropagation-free learning using a single (or monadic) associative hardware element. We realize this on an integrated photonic platform using phase-change materials combined with on-chip cascaded directional couplers. We then develop a scaled-up circuit network using our monadic Pavlovian photonic hardware that delivers a distinct machine-learning framework based on single-element associations and, importantly, using backpropagation-free architectures to address general learning tasks. Our approach reduces the computational burden imposed by learning in conventional neural network approaches, thereby increasing speed, whilst also offering higher bandwidth inherent to our photonic implementation., Comment: 24 pages, 5 figures

Research in photonic computing has flourished due to the proliferation of optoelectronic components on photonic integration platforms. Photonic integrated circuits have enabled ultrafast artificial neural networks, providing a framework for a new class of information processing machines. Algorithms running on such hardware have the potential to address the growing demand for machine learning and artificial intelligence, in areas such as medical diagnosis, telecommunications, and high-performance and scientific computing. In parallel, the development of neuromorphic electronics has highlighted challenges in that domain, in particular, related to processor latency. Neuromorphic photonics offers sub-nanosecond latencies, providing a complementary opportunity to extend the domain of artificial intelligence. Here, we review recent advances in integrated photonic neuromorphic systems, discuss current and future challenges, and outline the advances in science and technology needed to meet those challenges., Comment: 21 pages, 6 figures

The use of metals of nanometer dimensions to enhance and manipulate light-matter interactions for a range of emerging plasmonics-enabled nanophotonic and optoelectronic applications is an interesting, yet not highly explored area of research outside of plasmonics1,2. Even more importantly, the concept of an active metal, i.e. a metal that can undergo an optical non-volatile transition has not been explored. Nanostructure-based applications would have unprecedented impact on both the existing and future of optics with the development of active and nonlinear optical tunabilities in single elemental metals3-5. Compared to alloys, pure metals have the material simplicity and uniformity; however single elemental metals have not been viewed as tunable optical materials, although they have been explored as viable electrically switchable materials. In this paper we demonstrate for the first time that antimony (Sb), a pure metal, is optically distinguishable between two programmable states as nanoscale thin films. We then show that these states are stable at room temperature, and the states correspond to the crystalline and amorphous phases of the metal. Crucially from an application standpoint, we demonstrate both its optoelectronic modulation capabilities as well as speed of switching using single sub-picosecond (ps) pulses. The simplicity of depositing a single metal portends its potential for use in applications ranging from high speed active metamaterials to photonic neuromorphic computing, and opens up the possibility for its use in any optoelectronic application where metallic conductors with an actively tunable state is important.

With the proliferation of ultra-high-speed mobile networks and internet-connected devices, along with the rise of artificial intelligence, the world is generating exponentially increasing amounts of data - data that needs to be processed in a fast, efficient and smart way. These developments are pushing the limits of existing computing paradigms, and highly parallelized, fast and scalable hardware concepts are becoming progressively more important. Here, we demonstrate a computational specific integrated photonic tensor core - the optical analog of an ASIC-capable of operating at Tera-Multiply-Accumulate per second (TMAC/s) speeds. The photonic core achieves parallelized photonic in-memory computing using phase-change memory arrays and photonic chip-based optical frequency combs (soliton microcombs). The computation is reduced to measuring the optical transmission of reconfigurable and non-resonant passive components and can operate at a bandwidth exceeding 14 GHz, limited only by the speed of the modulators and photodetectors. Given recent advances in hybrid integration of soliton microcombs at microwave line rates, ultra-low loss silicon nitride waveguides, and high speed on-chip detectors and modulators, our approach provides a path towards full CMOS wafer-scale integration of the photonic tensor core. While we focus on convolution processing, more generally our results indicate the major potential of integrated photonics for parallel, fast, and efficient computational hardware in demanding AI applications such as autonomous driving, live video processing, and next generation cloud computing services.

For many applications, a method for controlling the optical properties of a solid-state film over a broad wavelength range is highly desirable and could have significant commercial impact. One such application is smart glazing technology where it is necessary to harvest near-infrared solar radiation in the winter and reflect it in the summer--an impossibility for materials with fixed thermal and optical properties. Here, we experimentally demonstrate a smart window which uses a thin-film coating containing GeTe, a bi-stable, chalcogenide-based phase-change material which can modulate near-infrared absorption while maintaining neutral-colouration and constant transmission of light at visible wavelengths. We additionally demonstrate controlled down-conversion of absorbed near-infrared energy to far-infrared radiation which can be used to heat a building's interior and show that these thin-films also serve as low-emissivity coatings, reducing heat transfer between a building and its external environment throughout the year. Finally, we demonstrate fast, sub-millisecond switching using transparent electrical heaters integrated on glass substrates. These combined properties result in a smart window that is efficient, affordable, and aesthetically pleasing--three aspects which are crucial for successful adoption of green technology.

We present the first demonstration of an integrated photonic phase-change memory using GeSbTe-225 on silicon-on-insulator and demonstrate reliable multilevel operation with a single programming pulse. We also compare our results on silicon with previous demonstrations on silicon nitride. Crucially, achieving this on silicon enables tighter integration of traditional electronics with photonic memories in future, making phase-change photonic memory a viable and integrable technology.

Modern-day computers use electrical signaling for processing and storing data which is bandwidth limited and power-hungry. These limitations are bypassed in the field of communications, where optical signaling is the norm. To exploit optical signaling in computing, however, new on-chip devices that work seamlessly in both electrical and optical domains are needed. Phase change devices can in principle provide such functionality, but doing so in a single device has proved elusive due to conflicting requirements of size-limited electrical switching and diffraction-limited photonic devices. Here, we combine plasmonics, photonics and electronics to deliver a novel integrated phase-change memory and computing cell that can be electrically or optically switched between binary or multilevel states, and read-out in either mode, thus merging computing and communications technologies.

Collocated data processing and storage are the norm in biological systems. Indeed, the von Neumann computing architecture, that physically and temporally separates processing and memory, was born more of pragmatism based on available technology. As our ability to create better hardware improves, new computational paradigms are being explored. Integrated photonic circuits are regarded as an attractive solution for on-chip computing using only light, leveraging the increased speed and bandwidth potential of working in the optical domain, and importantly, removing the need for time and energy sapping electro-optical conversions. Here we show that we can combine the emerging area of integrated optics with collocated data storage and processing to enable all-photonic in-memory computations. By employing non-volatile photonic elements based on the phase-change material, Ge2Sb2Te5, we are able to achieve direct scalar multiplication on single devices. Featuring a novel single-shot Write/Erase and a drift-free process, such elements can multiply two scalar numbers by mapping their values to the energy of an input pulse and to the transmittance of the device, codified in the crystallographic state of the element. The output pulse, carrying the information of the light-matter interaction, is the result of the computation. Our all-optical approach is novel, easy to fabricate and operate, and sets the stage for development of entirely photonic computers.

The unique physical and optical properties of nanoparticles make their reliable and rapid assembly an important goal for nanotechnology. To this end, many competing driving forces have been targeted to control the self-assembly, though issues with reliability and efficiency indicate that interfaces on the nanoscale are not yet fully understood. In this letter we report for the first time that the flow of water media from the interface between chemically nanopatterned hydrophobic surfaces and adjacent hydrophilic regions (a Janus interface) can effectively be employed for rapid nanoparticle assembly with single nanoparticle resolution. We find that the dewetting of water molecules from this Janus interface drives the assembly of nanoparticles from electrostatically stabilised nanoparticle colloids onto the hydrophobic areas, expedited by removing the resistance to adhesion between wetted interfaces exerted by any hydration barrier, such that nanoparticle assembly is achieved in under a minute. We call this method of controlled nanoparticle assembly Janus Interface Nanoparticle Assembly - JINA. We then demonstrate control of this mechanism through pH switching and prove its effectiveness for fabrication by rapidly creating single-nanoparticle sites and single particle wide lines whilst simultaneously densely coating macro-scale features. Not only does this assembly method have far-reaching implications in nanomanufacturing, the unanticipated nature of this mechanism also challenges many commonly held assumptions about the interfacial interactions of nanoparticle self-assembly., Comment: 19 pages, 5 figures

Most mechanical resonators are treated as simple linear oscillators. Nonlinearity in the resonance behavior of nanoelectromechanical systems (NEMS) has only lately attracted significant interest. Most recently, cubic-order nonlinearity has been used to explain anomalies in the resonance frequency behaviors in the frequency domain. Particularly, such nonlinearities were explained using cubic nonlinearity in the restoring force (Duffing nonlinearity) or damping (van der Pol nonlinearity). Understanding the limits of linear resonant behavior is particularly important in NEMS, as they are frequently studied for their potential in ultrasensitive sensing and detection, applications that most commonly assume a linear behavior to transduce motion into a detected signal. In this paper, we report that even at low excitation, cubic nonlinearity is insufficient to explain nonlinearity in graphene NEMS. Rather, we observe that higher order, in particular, the fifth order effects need to be considered even for systems at room temperature with modest quality factors. These are particularly important results that could determine the limits of linear detection in such systems and quite possibly present unconventional avenues for ultrasensitive detection paradigms using nonlinear dynamics. Such intriguing possibilities, however, hinge crucially on a superior understanding and exploitation of these inherent nonlinearities as opposed to modeling them as approximated linear or cubic systems., Comment: 10 pages, 6 figures

Graphene is an ideal material for fabricating atomically thin nanometre spaced electrodes. Recently, carbon-based nanoelectrodes have been employed to create single-molecule transistors and phase change memory devices. In spite of the significant recent interest in their use in a range of nanoscale devices from phase change memories to molecular electronics, the operating and scaling limits of these electrodes are completely unknown. In this paper, we report on our observations of consistent voltage driven resistance switching in sub-5 nm graphene nanogaps. We find that we are able to reversibly cycle between a low and a high resistance state using feedback-controlled voltage ramps.We attribute this unexplained switching in the gap to the formation and breakdown of carbon filaments.By increasing the gap, we find that such intrinsic resistance switching of graphene nanogaps imposes a scaling limit of 10 nm (approx.) on the gap-size for devices with operating voltages of 1 to 2 volts.

Techniques to reliably pick and place single nanoparticles into functional assemblies are required to incorporate exotic nanoparticles into standard electronic circuits. In this paper we explore the use of electric fields to drive and direct the assembly process, which has the advantage of being able to control the nano-assembly process at the single nanoparticle level. To achieve this, we design an electrostatic gating system, thus enabling a voltage controllable nanoparticle picking technique. Using the nonlinear Poisson-Boltzmann equation, we can successfully characterise the parameters required for single-particle placement, the key being single particle selectivity, in effect designing a system that can achieve this controllably. We then present the optimum design parameters required for successful single nanoparticle placements at ambient temperatures, an important requirement for nanomanufacturing processes., Comment: 15 pages, 6 figures

We present the instrumentation and measurement scheme of a new Casimir force probe that bridges Casimir force measurements at microscale and macroscale. A metallized high Q silicon nitride nanomembrane resonator is employed as a sensitive force probe. The high tensile stress present in the nanomembrane not only enhances the quality factor but also maintains high flatness over large area serving as the bottom electrode in a sphere-plane configuration. A fiber interferometer is used to readout the oscillation of the nanomembrane and a phase-locked loop scheme is applied to track the change of the resonance frequency. Because of the high quality factor of the nanomembrane and the high stability of the setup, a frequency resolution down to $2\times10^{-9}$ and a corresponding force gradient resolution of 3 $\mu$N/m is achieved. Besides sensitive measurement of Casimir force, our measurement technique simultaneously offers Kelvin probe measurement capability that allows in situ imaging of the surface potentials.

We propose an all-photonic, non-volatile memory and processing element based on phase-change thin-films deposited onto nanophotonic waveguides. Using photonic microring resonators partially covered with Ge2Sb2Te5 (GST) multi-level memory operation in integrated photonic circuits can be achieved. GST provides a dramatic change in refractive index upon transition from the amorphous to crystalline state, which is exploited to reversibly control both the extinction ratio and resonance wavelength of the microcavity with an additional gating port in analogy to optical transistors. Our analysis shows excellent sensitivity to the degree of crystallization inside the GST, thus providing the basis for non-von Neuman neuromorphic computing.

We present Casimir force measurements in a sphere-plate configuration that consists of a high quality nanomembrane resonator and a millimeter sized gold coated sphere. The nanomembrane is fabricated from stoichiometric silicon nitride metallized with gold. A Kelvin probe method is used in situ to image the surface potentials to minimize the distance-dependent residual force. Resonance-enhanced frequency-domain measurements of the nanomembrane motion allow for very high resolution measurements of the Casimir force gradient (down to a force gradient sensitivity of 3 uN/m). Using this technique, the Casimir force in the range of 100 nm to 2 um is accurately measured. Experimental data thus obtained indicate that the device system in the measured range is best described with the Drude model.

We report the piezoresisitivity in magnetic thin films of FeGa and their use for fabricating self transducing microcantilevers. The actuation occurs as a consequence of both the ferromagnetic and magnetostrictive property of FeGa thin films, while the deflection readout is achieved by exploiting the piezoresisitivity of these films. This self-sensing, self-actuating micromechanical system involves a very simple bilayer structure, which eliminates the need for the more complex piezoelectric stack that is commonly used in active cantilevers. Thus, it potentially opens opportunities for remotely actuated, cantilever-based sensors., Comment: 10 Pages, 3 Figures