2020 Spring Meeting
Electronics, magnetics and photonicsY
Neuro-inspired information processing: from novel materials concepts for neuromorphic computing to local processing of biological signals
The widely anticipated end to Moore’s law and the growing demand for low-power computing systems capable of learning, pattern recognition and real-time analysis of large streams of unstructured data has spurred intense interest in devices with basic forms of neuroplasticity as building blocks for efficient neuromorphic computing systems.
The latest advancements of inorganic and organic neuromorphic devices will be broadly covered in this symposium. The symposium will offer an overview the desired properties of bio-inspired or neuromorphic devices and systems, including the merged processing and storage capabilities, adaptivity, delocalized or spatially correlated features, biocompatibility, generic classification and learning. Key showcases of novel neuromorphic devices and materials systems will be highlighted, that are oriented to a range of applications that span from traditional neuromorphic computing and efficient hardware-implemented neural networks to emulate biological neural network behavior and various concepts of neuromorphic sensing in bioelectronics.
The rapidly expanding field of adaptable biointerfacing through the merging of bioelectronics and neuromorphic sensing / actuation will also be covered in this symposium. The field of bioelectronics has made an enormous progress towards the development of concepts, materials and devices that are capable of bi-directional interaction with a biological environment by incorporating concepts such as drug delivery and electrical / chemical stimulation. Nevertheless, fully autonomous applications in the field of organic bioelectronics demand not only the acquisition of biological signals but also local data processing, storage and the extraction of specific features of merit. As such, materials, devices and architectures with bio-inspired features, can offer promising solutions for the manipulation and the processing of biological signals spanning from brain-computer-interfaces and robotics to bioinformatics and the definition of novel computational paradigms at the interface with biology.
This symposium aspires to bring together world-wide experts in the fields of neuro-inspired computing and bioelectronics in order to enhance transdisciplinary interactions and bridge the gaps between memristive devices and neuroscience. The envisioned forum purports the exploitation of the wide range of novel materials (e.g. diffusive memristors, novel 2D materials, organics, transition metals) and device properties towards novel applications in this increasingly promising field.
Hot topics to be covered by the symposium:
- Bio-inspired information processing
- Neuromorphic computing
- Inorganic and organic neuromorphic devices
- Novel device systems (multi-terminal, hybrid devices etc.)
- Memristive materials / devices at the interface with biology
- Neuromorphic sensing
- Neural interface devices
- Adaptable / trainable biointerfacing
- Systems neuroscience
|Start at||Subject View All||Num.|
|08:30||WELCOME (08:30-08:45): Paschalis Gkoupidenis, Yoeri van de Burgt, Jessamyn Fairfield, Duygu Kuzum|
Session 1 (08:45-10:00) - Spintronics for neuromorphic computing : Chair - Nawrocki
Authors : J. Grollier et al
Affiliations : Unité Mixte de Physique CNRS/Thales, Palaiseau, France
Resume : The purpose of neuromorphic computing is to take inspiration from the brain to build hardware neural networks that can learn to perform useful tasks with low energy consumption . In this talk, I will show that spintronic nano-oscillators based on magnetic tunnel junctions can act as artificial neurons . I will present our first results of pattern recognition with small networks of coupled oscillators . I will then show that these microwave nano-neurons open the path to wireless deep learning.  J. Grollier, D. Querlioz, et M. D. Stiles, « Spintronic Nanodevices for Bioinspired Computing », Proc. IEEE, vol. 104, no 10, p. 2024?2039, oct. 2016.  J. Torrejon et al., « Neuromorphic computing with nanoscale spintronic oscillators », Nature, vol. 547, no 7664, p. 428?431, juill. 2017.  M. Romera et al., « Vowel recognition with four coupled spin-torque nano-oscillators », Nature, vol. 563, no 7730, p. 230, nov. 2018.
Authors : Nathan Leroux, Danijela Markovic, Erwann Martin, Téodora Pétrisor, Alice Mizrahi, Julie Grollier
Affiliations : Unité Mixte de Physique CNRS/Thales, CNRS, University of Paris-Sud, Palaiseau, France;Thales Research and Technology, Palaiseau, France
Resume : Spin-Torque Nano-Oscillators (STNOs) can act both like microwave emitters (oscillators) and microwave rectifiers (diodes). We showed previously that STNOs acting as microwave emitters can emulate neurons and be used for Reservoir Computing. In this work, we show that STNOs acting as diodes can be used as trainable synaptic connections, and that we can assemble a Deep Neural Network (DNN) based on microwave detection and emission. In order to demonstrate the Multiply And Accumulate operation (MAC), we use numerical differential equations resolution to simulate a layer of oscillators sending their signals to an array of diodes. We prove that the voltage of each diode is proportional to the power emitted by one of the oscillator (Multiply), and that the total rectified voltage is the sum of the voltage of each diode (Accumulate). We show that each synaptic weight can be chosen by tuning the resonance frequency of each diode. Then, we use an analytical expression of the diode effect to simulate a multilayer network. We test the network under different conditions, and show that It can classify images of handwritten digits This novel concept of DNN relying on microwave signals allows us to route the information between neuron and synapses with a simplified spatial architecture, thus achieving a high density of connections. This work was supported by the European Research Council ERC under Grant bioSPINspired 682955, the French ANR project SPIN-IA (ANR-18-ASTR-0015) and the French Ministry of Defense (DGA).
Authors : Karin Everschor-Sitte
Affiliations : Institute of Physics, Johannes Gutenberg-University Mainz, Germany
Resume : Novel computational paradigms in combination with proper hardware solutions are required to overcome the limitations of our state-of-the-art computer technology, in particular regarding energy consumption. Due to the inherent complex and non-linear nature, spintronics offers the possibility towards energy efficient, non-volatile hardware solutions for various unconventional computing schemes. In this talk, I will address the potential of topologically stabilised magnetic whirls ? so called skyrmions ? for two unconventional computing schemes ? reservoir computing and stochastic computing. Reservoir computing is a computational scheme that allows to drastically simplify spatial-temporal recognition tasks. We have shown that random skyrmion fabrics provide a suitable physical implementation of the reservoir [2,3] and allow to classify patterns via their complex resistance responses either by tracing the signal over time or by a single spatially resolved measurement.  Stochastic computing is a computational paradigm which allows to speed up a calculation while trading for numerical precision. Information is encoded in terms of bit-streams as a probability. A key requirement and simultaneously a challenge is that the incoming bitstreams are uncorrelated. The Brownian motion of magnetic skyrmions allows to create a device that reshuffles the bit-streams. [5,6] References  G. Finocchio et al., arXiv:1907.04601  D. Prychynenko et al., Phys. Rev. Appl. 9, 014034 (2018)  G. Bourianoff et al., AIP Adv. 8, 055602 (2018)  D. Pinna et al., arXiv:1811.12623  D. Pinna et al., Phys. Rev. Appl. 9, 064018 (2018)  J. Zazvorka et al., Nat. Nanotech. 14, 658 (2019)
|10:00||COFFEE BREAK (10:00-10:30)|
Session 2 (10:30-12:30) - Neuromorphic computing: fundamentals and applications : Chair - Alibart
Authors : R. Tetzlaff (1), A. Ascoli (1) , I. Messaris (1), S. Kang (2), and L. O. Chua (3)
Affiliations : (1) Institute of Circuits and Systems, TU Dresden, Dresden, Germany (2) Jack Baskin School of Engineering, University of California Santa Cruz, Santa Cruz, CA 95064 USA (3) Department of Electrical Engineering and Computer Sciences, University of California Berkeley, Berkeley, CA 94720 USA
Resume : The importance of bio-inspired computing arrays for the development of programmable non-Von-Neumann computer architectures that avoid the separation of computing and memory structures has been demonstrated in recent investigations. Since the introduction by Chua and Yang in 1988, Cellular Neural Networks (CNN) have become a paradigm for complexity. These bio-inspired networks, which are characterized by local couplings of nonlinear dynamic systems of comparatively low complexity, are regarded as the basic structure in a CNN Universal Machine (CNN-UM) architecture, which is a complete dynamic array stored program computer, very often realized on a single chip with optical sensors in CMOS technology. CNN-UM high-speed computing systems are programmed by CNN templates as instructions e.g. in image processing applications or in multidimensional medical signal processing. Since CMOS implementations of programmable cellular sensor-processor array architectures cannot exploit the full potential of this paradigm and do not exceed a resolution of 176X144 cells, M-CNN structures have been proposed that can be implemented with an increased resolution as compared to the pure CMOS counterparts. M-CNN are proposed in this contribution as a bio-inspired paradigm for universal mem-computing in future sensor-processor systems. A detailed introduction to the theory of memristors and MCNN is given in the presentation. Especially, newly developed methods  for programming such structures are presented and discussed in detail.  R. Tetzlaff, A. Ascoli, I. Messaris, and L.O. Chua, ?Theoretical Foundations of Memristor Cellular Nonlinear Networks: Memcomputing with Bistable-like Memristors,? IEEE Trans. on Circuits and Systems?I: Regular Papers, 2019, 10.1109/TCSI.2019.2940909  A. Ascoli, R. Tetzlaff, I. Messaris, and L.O. Chua, ?Theoretical foundations of Memristor Cellular Nonlinear Networks: Stability Analysis with Dynamic Memristors,? IEEE Trans. Circuits and Systems?I: Regular Papers, 2019, in press  A. Ascoli, R. Tetzlaff, S.M ?Steve? Kang, and L.O. Chua, ?Theoretical foundations of Memristor Cellular Nonlinear Networks: A DRM2-based Method to Design Memcomputers with Dynamic Memristors, under review
Authors : Alejandro Fernández-Rodríguez1, Jordi Alcalà1, Jordi Suñe2, Anna Palau1 and Narcis Mestres1
Affiliations : 1. Institut de Ciència de Materials de Barcelona, ICMAB-CSIC, Campus UAB, 08193 Bellaterra, Barcelona, Spain; 2. Departament d?Enginyeria Electrònica, Universitat Autònoma de Barcelona, 08193 Bellaterra, Barcelona, Spain
Resume : Memristive devices are attracting a great deal of attention for memory, logic, neural networks, and sensing applications due to their simple structure, high density integration, low-power consumption, and fast operation. In particular, multi-terminal structures controlled by active gates, able to process and manipulate information in parallel, would certainly provide novel concepts for neuromorphic systems. Hence, transistor-based synaptic devices may be designed, where the synaptic weight in the postsynaptic membrane is encoded in a source-drain channel and modified by presynaptic terminals (gates). In this work, we show the potential of reversible field-induced metal-insulator transition (MIT) in strongly correlated cuprates for the design of robust and adjustable multi-terminal memristive transistor-like devices. We have studied different structures patterned on YBa2Cu3O7-? films, which are able to display gate modulated non-volatile volume MIT, driven by field-induced oxygen diffusion within the system. The key advantage of these materials is the possibility to homogeneously tune the oxygen diffusion not only in a confined filament or interface, as observed in widely explored binary and complex oxides, but also in the whole material volume. We show several device configurations in which the lateral conduction in a drain-source channel is effectively controlled by active gate-tunable volume resistance changes, thus emulating the synaptic weight. Large design flexibility can be obtained by changing the switching performance of different gates, thus offering the possibility to locally adjust the conductance response as required to implement neuromorphic functionalities.
Authors : Joksas, D., Buckwell, M., Ng, W.H., Kenyon, A.J., Mehonic, A.
Affiliations : University College London, United Kingdom
Resume : The ever-increasing power demands of machine learning algorithms, such as artificial neural networks (ANNs), make our conventional hardware systems seemingly inefficient. Memristors are a promising candidate for physical implementation of ANNs but various device- and system-level non-idealities usually prevent such networks from achieving high inference accuracy. We present an in-depth analysis of the most common non-idealities of memristor devices and their systems: faulty devices, programming non-linearities, random telegraph noise, device-to-device variability and line resistance. Our results show that device-to-device variability and line resistance have the most detrimental effect on accuracy. However, by employing the concept of a committee machine to make multiple physically implemented ANNs work together, we show significant increases in accuracy. Importantly, we demonstrate that we can improve the accuracy without increasing the number of devices—replacing a large ANN with several smaller ANNs results in higher accuracy.
Authors : Manan Suri
Affiliations : Indian Institute of Technology - Delhi
Resume : We live in an era which is more memory-centric than ever. Factors that contribute to the ever increasing importance of memory are ? (i) Saturation of Moore?s law, (ii) ease of generating enormous amounts of data and (iii) exciting new material properties. The nature of present day data intensive applications is such that, excellence in computational performance cannot be achieved alone on the basis of raw transistor scaling or linearly increasing the number of processing cores. A fundamental shift in the vastly successful Von Neumann computational paradigm is needed to overcome the bottlenecks associated with data-intensive real time applications. This is where next generation Non-Volatile Memory (NVM) begins to play a very significant role. Our research group at IIT-D, has been actively working on exploiting the characteristics of emerging NVM nanodevices and nanomaterials for a multitude of novel applications. We have considered emerging technologies such as OxRAM, CBRAM, PCM, RRAM, STT-MRAM etc. building an entire memory-centric application ecosystem comprising of various hybrid CMOS-NVM circuits. Applications realized include: Supervized and Unsupervised Learning (SNNs, CNNs, BNNs), AI Edge-Inference & Training, Sensing, Security, and in-Memory Computing.
Authors : Li, S.* (1,2), Liu, X. (3), Nandi, S.K. (1), Nath, S.K.(1), Grollier J. (2), Elliman, R.G. (1)
Affiliations : (1) Department of Electronic Material Engineering, Research School of Physics, The Australian National University, Canberra, ACT 2601, Australia (2) Unité Mixte de Physique, CNRS, Thales, Université Paris-Sud, Université Paris-Saclay, Palaiseau, France (3) Tianjin Key Laboratory of Low Dimensional Materials Physics and Preparation Technology, Faculty of Science, Tianjin University, Tianjin 300354, China * lead presenter
Resume : Locally active memristors exhibit reversible volatile resistive switching and display current-controlled negative differential resistance (NDR). The nonlinear current-voltage characteristics provide the fundamental basis of rich dynamical behaviors, such as all-or-nothing spiking, periodic and chaotic oscillations. The dynamic features of active memristors has enabled the emulation of neural functions and realization of relaxation oscillators which offers great potential to be used in building biologically-inspired neuromorphic hardware. Recently, several NDR behaviors, including the continuous S-type and discontinuous snap-back responses, have been reported. However, achieving the desired NDR properties for technological implementation is limited by the understanding on the switching mechanisms and design criteria. Here, we propose a material-independent model with experimental demonstration to explain and predict a broad spectrum of NDR characteristics. The model shows that the S-type and snap-back responses are resulted from a non-uniform current distribution in the oxide film and its impact on the effective circuit of the device driven by the same physical process. Furthermore, we show that the S-type and snap-back responses serves as the fundamental building blocks of realizing composite NDR behaviors with higher complexity and novel functionality. The results offer important insight and design flexibility of engineering active memristors for neuromorphic computing.
|12:30||LUNCH BREAK (12:30-14:00)|
Session 3 (14:00-15:30) - Resistive memories for neuromorphic computing : Chair - Suri
Authors : D. Ielmini, Z. Sun, G. Pedretti
Affiliations : Dipartimento di Elettronica, Informazione e Bioingegneria Politecnico di Milano, Milano (Italy)
Resume : Artificial intelligence (AI) is rising as a powerful enabling technology in many aspects of society such as industry, transport, business and health. To enable edge AI, new devices and architectures with high density and low power computation are needed. In this scenario, high density memories, such as resistive switching memory (RRAM) and phase change memory (PCM), have attracted interest based on their analogue storage, high density, and low voltage/current operation. The combination of RRAM/PCM with analogue in-memory computing architectures appears to considerably improve the speed and energy efficiency for neuromorphic computation . Here we present a novel architecture for one-shot learning and regression within a crosspoint memory array with feedback loop . The circuit can solve linear/logistic regression problems in one computational step, allowing fast learning by the computation of regression weights in multiple dimensions. The same approach can also be used for one-shot prediction and classification from the data. The impact of device non-idealities, such as non-linearity, variability and limited window are addressed. These results support in-memory computing as a strong candidate for edge AI.  Z. Sun, G. Pedretti, E. Ambrosi, A. Bricalli, W. Wang and D. Ielmini, ?Solving matrix equations in one step with crosspoint resistive arrays,? PNAS 116 (10) 4123-4128 (2019). doi/10.1073/pnas.1815682116  Z. Sun, G. Pedretti, A. Bricalli, D. Ielmini, ?One-step regression and classification with crosspoint resistive memory arrays,? Sci. Adv. 6:eaay2378 (2020).
Authors : Daniel J Mannion, Anthony J Kenyon
Affiliations : University College London
Resume : Memristors have been shown to produce chaotic oscillations and periodic spiking both of which can potentially be used to compute in the spike domain. [1,2,3] Combining this with their nanoscale dimensions and two terminal structure, they emerge as prime candidates to implement dense complex networks, such as in the dendritic connections occurring between synapses. However, from this complexity arises the question of how to design circuits with a specific computation in mind. In this work we address this question by constructing dendritic networks of passive two terminal components: resistors, capacitors and memristors, which perform predetermined operations. We generate these by applying the principles of genetic algorithms.  Genetic algorithms carry out iterative searches in a manner inspired by natural selection. We begin by designing a genetic code capable of defining the topology of the network and the components making it up. With this code, an initial population is generated and then characterised against our desired operation using a fitness function. The fitness function returns a score indicating how accurate the network is at replicating our target operation, essentially ranking members of the population. Genetic codes are then mixed and modified according to this ranking through two mechanisms, crossover and mutation. Crossover involves the reproduction between two members of the population. After each member’s fitness score is calculated, members reproduce with others in the population to produce offspring. Those with a higher fitness score have a higher chance of reproduction. The genetic codes of the two parents are exchanged in certain places to produce offspring with a modified genetic code. Importantly, genetic codes widely different from each other are not able to produce offspring due to too many mismatches in their genetic codes – this corresponds to widely different circuit designs which would be a challenge to mix. Once crossover has occurred, the second mechanism for change, mutation, takes over. Each gene has a low probability of being modified randomly whereby components and their values can be modified, or in rare cases, entirely new branches in the circuit can be introduced. The combination of crossover and mutations result in an offspring with a modified genetic code. This new genetic code represents a new circuit design which then enters the existing population and begins competing. Across generations the population converges and eventually replicates our target operation. In this talk we will present the genetic code used to define and generate dendrite-like inputs, as well as the performance of these circuits at replicating behaviours such as filtering and resonance.  A. Mehonic and A. J. Kenyon, “Emulating the Electrical Activity of the Neuron Using a Silicon Oxide RRAM Cell,” Front. Neurosci., vol. 10, Feb. 2016.  H. Wu, B. Bao, Z. Liu, Q. Xu, and P. Jiang, “Chaotic and periodic bursting phenomena in a memristive Wien-bridge oscillator,” Nonlinear Dyn., vol. 83, no. 1–2, pp. 893–903, Jan. 2016.  S. Li, X. Liu, S. K. Nandi, S. K. Nath, and R. G. Elliman, “Origin of Current‐Controlled Negative Differential Resistance Modes and the Emergence of Composite Characteristics with High Complexity,” Adv. Funct. Mater., vol. 29, no. 44, p. 1905060, Nov. 2019.  S. N. Sivanandam and S. N. Deepa, “Genetic Algorithms,” in Introduction to Genetic Algorithms, Berlin, Heidelberg: Springer Berlin Heidelberg, pp. 15–37.
Authors : Florian Maudet * (1), Veeresh Deshpande (1), Hanno Kröncke (1), Catherine Dubourdieu (1,2).
Affiliations : (1) Institute Functional Oxides for Energy-Efficient Information Technology (EM-IFOX), Helmholtz-Zentrum Berlin für Materialien und Energie, Hahn-Meitner-Platz 1, 14109 Berlin, Germany; (2) Freie Universitat Berlin, Physical Chemistry, Arnimallee 22, 14195 Berlin Germany
Resume : Resistance switching memories are promising candidates for low power consumption non-volatile memory applications owing to their dimensional scalability and fast switching speed. Furthermore, the possibility to obtain multiple stable states in such systems is of importance for applications in neuromorphic computing. Among the proposed systems, Cu/SiO2 based memory devices have demonstrated very promising performances and allow easy integration with CMOS technology. It is known that in such systems the different states originate from the formation of a Cu filament in the SiO2 solid electrolyte. Because of this, the devices are very sensitive to local inhomogeneities and defects in the SiO2 dielectric. As a result, the on/off resistance and the set/reset voltage can have significant device-to-device variations limiting their advantages for potential applications. Additionally, the endurance is also a limiting factor and needs to be improved. In order to quantify and understand the origin of the distribution of these characteristics we present a broad statistical analysis of Cu/SiO2/W memory behavior. The influence of different voltage sweep rates and device areas are investigated. This study allows identification of key parameters that can be optimized to reduce variance between samples. Finally, the influence of an oxide barrier deposited between W and SiO2 to improve the device characteristics, notably the endurance, is also investigated.
|15:15||COFFEE BREAK (15:15-16:00)|
Session 4 (16:00-17:30) - Novel devices for neuromorphics: photonic and electrochemical concepts : Chair - Georgiadou
Authors : Emil J.W. List-Kratochvil
Affiliations : Institut für Physik, Institut für Chemie & IRIS Adlershof, Humboldt-Universität zu Berlin, Brook-Taylor-Straße 6, 12489 Berlin, Germany Helmholtz-Zentrum für Materialien und Energie GmbH, Brook-Taylor-Straße 6, 12489 Berlin, Germany
Resume : Artificial neural networks take inspiration from the functionalities of the brain to solve complex tasks by learning. The hardware realization of such networks however still lags behind computer simulations, which improved significantly during the last decade. Here we demonstrate the capability to emulate synaptic functionalities by exploiting surface plasmon polaritons. Using light-induced reversible changes modulate the dielectric function of the photochromic molecules, which enables to optically remote control the surface plasmon polariton dispersion relation. This plasmonic device concept exhibits the fundamental functions of a synapse, such as potentiation, depression, and long-term plasticity. In order to integrate the concept into a photonic/plasmonic multi-bit optical storage device in a next step we are using this synaptic device in a polymeric waveguide network. This integration of such plasmonic devices in an optical artificial neural network paves the way towards neuro-inspired, photonic/plasmonic circuits for optical computing.
Authors : Spyros Stathopoulos, Ioulia Tzouvadaki, Themis Prodromakis
Affiliations : University of Southampton, Southampton, SO17 1BJ, UK
Resume : State-modulated resistive switching via a straightforward device programming is highly important for aptly exploiting memristors as analogue memories for brain-inspired and reconfigurable computing. We suggest a light-triggered memristor programming using organic semiconductor modified Metal-Insulator-Metal (MIM) memristors depicting efficient resistive switching using only UV-light illumination. The photosensitive MIM architecture consists of a < 50 nm thickness layer of TiO2 as solid electrolyte and Au top (10 nm) and bottom (20 nm) electrodes defined and realized through optical lithography, metal evaporation and liftoff processes. To introduce photosensitivity a thin layer of semiconducting polymer [poly(3,4ethylenedioxythiophene) poly (styrenesulfonate)] is deposited on the top electrode. UV light programming results in highly discernible resistive states (RS) without additional voltage stimulus. We study the effect of the light stimulus (wavelength of incident photons and light-pulse durations) on the resistive switching. The response of the system is monitored while under illumination from a Xe/QTH light source delivering monochromatic frequencies. The RS achieved were found in good agreement with their voltage-controlled counterparts. The suggested device stack and programming approach offer an alternative procedure to tune memristive devices and a valuable proof-of-concept towards novel optic and photo-energy memories and light-controlled neuromorphic applications.
Authors : Hea-Lim Park; Haeju Kim; Sungjin Park; Tae-Woo Lee
Affiliations : Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea; Department of Chemistry and Chemical Engineering, Inha University, Incheon, Republic of Korea; Department of Chemistry and Chemical Engineering, Inha University, Incheon, Republic of Korea; Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea
Resume : For emulating visual perception of retina, photonic synapses have attracted much attention due to high compactness merging two functions of sensing and synaptic elements in a single device [1,2]. With both functions of artificial synapses and photo-sensors, beyond a simple sensing function of sensors, the photonic synapses have abilities to process detected stimulation and track stimulation history including light intensity, exposure number, duration time, and frequency. Ultraviolet (UV) light from 10 nm to 400 nm wavelength is harmful to cause skin aging, skin cancer, macular degeneration, and cataract, but cannot be detected by eyes of humans. In this work, we demonstrated organic photonic synapses that selectively detect UV rays and process various optical stimuli. The photonic synapses use carbon nitride (C3N4) as a UV-responsive floating-gate layer in transistor geometry. C3N4 nanodots dominantly absorb UV light; this trait is the basis of UV selectivity in these photonic synapses. The presented devices consume only 18.06 fJ/synaptic event, which is comparable to the energy consumption of biological synapses. Furthermore, in-situ modulation of exposure to UV light is demonstrated by integrating the devices with UV transmittance modulators. These smart systems would be further developed to combine detection and dose-calculation to determine how and when to decrease UV transmittance for preventive health care.  H.-L. Park, Y. Lee, N. Kim, D.-G. Seo, G.-T. Go, and T.-W. Lee, Adv. Mater., 2020, 32, 1903558.  Y. Lee, T. W. Lee, Acc. Chem. Res., 2019, 52, 964.
Authors : Gunuk Wang
Affiliations : KU-KIST Graduate School of Converging Science and Technology, Korea University, 145 Anam-ro, Seongbuk-gu, Seoul, 02841, Republic of Korea
Resume : The complex neural circuits of the human brain comprising 10^11 neurons connected via 10^14 synapses efficiently process a vast amount of information, including cognitive functions and memory. Neurons generate electrical signals called spikes that trigger the release of neurotransmitters from synapses, which forms the basis of neural information processing in the brain. Especially, the human brain’s neural networks are sparsely connected via tunable and probabilistic synapses, which may be essential for performing energy-efficient cognitional and intellectual functions. Inspired by the efficient threshold-tunable and probabilistic rod-to-rod bipolar synapse in the visual system, we designed and fabricated a vertical form of a gate-tunable SiOx memristor synapse using a Si phase along with a graphene barristor, termed as a probabilistic synaptic barristor, then extended into 16 × 16 crossbar array for a neural network. Notably, the electrostatic gating from the barristor can modulate the Schottky barrier at the Si/graphene interface; thus, the switching-transition probability and the threshold for signal firing in the SiOx memristor could be actively regulated and its switching energy decreased. Furthermore, we constructed a drop-connect algorithm that can reflect the probabilistic connectivity in the neural network, and the shapes of several fashion items were successfully classified. The device can achieve ~94.4% recognition accuracy using only 15% switching-transition probability with ~2 pJ per single activity. Our probabilistic Si synaptic barristor can offer a distinctive and novel strategy for the energy-efficient neuromorphic computing application.
|Start at||Subject View All||Num.|
|08:45||PLENARY SESSION 1 - Prof. James Fraser Stoddart, Nobel Laureate in Chemistry (2016)|
|10:00||COFFEE BREAK (10:00-10:30)|
Session 5 (10:30-12:30) - Neuromorphic devices for processing, sensing and actuation (1) : Chair - van de Burgt
Authors : Alberto Salleo, Armantas Melianas, Tyler Quill, Hilbert van Loo, Scott Keene, Yaakov Tuchman, Alexander Giovannitti
Affiliations : Department of Materials Science and Engineering, Stanford University, Stanford CA 94305
Resume : The brain can perform massively parallel information processing while consuming only ~1- 100 fJ per synaptic event. In this talk I will describe a novel electrochemical neuromorphic device (ENODe) that, when properly scaled, switches in less than 20 ns consuming less than 100 fJ of energy per switching event and displays a large number of distinct, non-volatile conductance states within a ~1 V operating range. The tunable resistance behaves very linearly, allowing blind updates in a neural network when operated with the proper access device. ENODes also display outstanding endurance achieving over 109 switching events with very little degradation all the way to high temperature (up to 120°C). These properties are very promising in terms of the ability to integrate with Si electronics to demonstrate online learning and inference. ENODes are electrochemical devices where gated proton drift induces changes in the electronic states of a semiconductor channel. The peculiarities of the physics of these devices will be discussed along with their consequences on device design and performance.
Authors : Dae-Gyo Seo 1, Yeongjun Lee 1, 5, Hoichang Yang 2, Changduk Yang 3, Sang-Woo Kim 4, and Tae-Woo Lee 1
Affiliations : 1 Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea; 2 Department of Applied Organic Materials Engineering, Inha University, Incheon, Republic of Korea; 3 Department of Energy Engineering, Ulsan National Institute of Science and Technology (UNIST), Ulsan, Republic of Korea; 4 School of Advanced Materials Science & Engineering, Sungkyunkwan University (SKKU), Suwon, Gyeonggi-do, Republic of Korea; 5 Samsung Advanced Institute of Technology (SAIT), Suwon, South Korea
Resume : Unlike the inorganic neuromorphic electronics which can only implement the long-term plasticity, the organic neuromorphic electronics have paid attentions due to promising applications from next generation computing to neuroprosthetics by implementing not only the long-term potentiation (LTP) but also the short-term potentiation (STP). Recently, conjugated polymers such as poly(3-hexylthiophene-2,5-diyl) (P3HT) with electrolyte dielectrics has been studied to implement high-density computing and memory systems using less energy. Parallelly, diketopyrrolopyrroles (DPPs) based artificial synaptic devices have shown promising applications from neuroprosthetics to soft robotics. However, recently studied organic-based artificial synapses only focused on emulating single function of the biological counterparts; either memory or signal transitions. In this study, we introduce the engineering of synaptic properties by modulating morphological characteristics of organics semiconductors (OSCs) with fixed presynaptic spike forms in artificial synaptic devices without changing the OSCs. By achieving this, the relation between morphology of the film and the synaptic property have been revealed. Changing the morphological property of the OSCs, resulted in the clear transition from STP dominant synaptic property to LTP dominant synaptic properties. Furthermore, artificial neural networks (ANNs) simulation and artificial auditory sensory nervous system was successfully demonstrated. This achievement is imperative step toward development of versatile neuromorphic electronics and promising for various neuromorphic electronics application from neuromorphic computing to neural prosthetics, bio-interface device, and soft robotics which can immensely expand the field of neuromorphic electronics by providing understanding of the mechanisms of synaptic properties.  D.-G. Seo, et al., Nano Energy 2019, 65, 104035  Y. Kim, et al., Science. 2018, 360, 998  Y. Lee, et al., Sci. Adv. 2018, 4, eaat7387  W. Xu, et al., Sci. Adv. 2016, 2, e1501326
Authors : Soo Jin Kim, Jae Seung Jeong, Hyunsu Ju, Jung Ah Lim
Affiliations : Center for Opto-Electronic Materials and Devices Post-Silicon Semiconductor Institute, Korea Institute of Science and Technology, Seoul 136-791, Korea. Department of Materials Science and Engineering, Seoul National University, Seoul 08826, Korea.
Resume : Biocompatibility, mechanical flexibility and capability to control migration of ion species highlights ion-gel gated organic transistors as a promising artificial synaptic device. The presynaptic signal is applied on the gate connected with polymer channel through ion-gel. It triggers penetration of the ions into the polymer and changes the channel conductance via electrochemical doping process, which generates the postsynaptic responses. The ion-transport between the ion-gel and the polymeric semiconductor is related to short- (STP) and long-term plasticity (LTP) for emulating neurons and synapses. However, fundamentals on how the ions doped organic semiconductors (OSCs) and what the relation between doping condition of OSCs and synaptic plasticity properties of the device have not been fully understood. In this study, comprehensive investigations on the doping behaviors of typical polythiophenes (PT) and poythiophenes with functionalized side chain (func-PT) in ionic liquid were performed by FT-IR, Raman spectroscopy, X-ray scattering, AFM and cyclic voltammetry. It was found that the migration of anions or cations to the PT films at the certain redox condition caused the structural and morphological change and formation of polarons or bipolarons at the backbone of PTs. Additionally, the irreversible doping reaction of typical PT turns out not suitable for a reproducible artificial synapse, while the func-PT has reversible doping/dedoping reaction resulting in stable programing/erasing characteristics of neuromorphic transistor with synaptic plasticity. This studies on doping behavior of OSCs by ions helps to advanced neuromorphic devices.
Authors : S. Stathopoulos, I. Tzouvadaki, T. Abbey, A. Serb and T. Prodromakis
Affiliations : Centre for Electronics Frontiers, Zepler Institute for Photonics and Nanoelectronics, University of Southampton, Southampton SO17 1BJ, UK
Resume : Nanoscale resistive switching elements, also known as memristors, are nowadays regarded as a promising solution for establishing next-generation memory, due to their infinitesimal dimensions and their capacity to store multiple bits of information per element. While this emerging technology has provided the necessary disruption needed for enabling neuromorphic computing and reconfigurable systems, their underpinning dynamical behavior is well suited for supporting sensing applications ? across distinct sensing modalities. A plethora of materials and technologies has widely been investigated for their ability to demonstrate memristive behaviour and this work particularly focuses on metal-oxide devices that demonstrate analogue storage operation with supreme resolution. During this talk we shall present a detailed evaluation of the prototyped samples and how their performance can enable facilitate compressing sensing capabilities at the edge (device level). We will showcase a number of examples ranging from bio-signal processing as well as employing individual memristive devices as temperature, biochemical and light sensing devices. This showcases new application opportunities that go beyond conventional memory applications that are inspired by processes encountered in nature.
Authors : Vahl, A.* (1), Carstens, N. (1), Carstensen, J. (2), Kaps, S. (2), Strunskus, T.(1), Adelung, R. (2) & Faupel, F. (1).
Affiliations : (1) Kiel University, Institute for Materials Science, Chair for Multicomponent Materials, Kaiserstr. 2, 24143, Kiel, Germany; (2) Kiel University, Institute for Materials Science, Chair for Functional Nano Materials, Kaiserstr. 2, 24143 Kiel, Germany
Resume : Despite an unprecedented progress in neuromorphic engineering with memristive devices within the past decade, one particular aspect of complex neuronal systems has been mostly overlooked: The close connection between sensing (i.e. the detection of environmental signals) and data processing. In biological neuronal networks however, the incoming data are processed already at the location of data detection (e.g. by adaptation), which is an essential part of the high efficiency of neuronal networks. A recent approach to bridge this gap is the concept of memsensors, which combine memristive switching and sensing properties. In this work we report on a simple three-component memsensor model, which is capable of describing the inherited properties, pinched I-V hysteresis and stimulus dependent resistivity, as well as stimulus dependent hysteresis and a potential adaptation to an external stimulus. This adaptation shows striking similarities to adaptation in biological neuronal systems, making memsensors ideal candidates for applications in neuromorphic engineering. Furthermore, different switching characteristics of memristive devices (e.g. bipolar and diffusive switching) will be put into the context of the memsensor concept and potential applications will be discussed.
Authors : Dimitra G. Georgiadou, Thomas D. Anthopoulos, Themis Prodromakis
Affiliations : Centre for Electronics Frontiers, Zepler Institute for Photonics and Nanoelectronics, University of Southampton, SO17 1BJ Southampton, United Kingdom; Department of Physics & Centre for Plastic Electronics, Imperial College London, SW7 2AZ London, United Kingdom; Materials Science & Engineering, Division of Physical Sciences and Engineering, King Abdullah University of Science and Technology (KAUST), Kingdom of Saudi Arabia
Resume : Neuromorphic engineering is poised to revolutionise information technologies by developing electronic devices that emulate neuro-biological architectures at the hardware level. An example of neuromorphic device is the “artificial synapse”, which can be represented by non-volatile variable resistance memory elements or memristors. Another example is smart imaging, where light sensors are designed to mimic the spatio-temporal nature of human vision, not only by turning light into electrical signals but also by capturing and sending the useful-only information to the processing unit in an extremely efficient manner. The commonly employed device structure of optoelectronic and memristive devices is a vertically aligned configuration, where one or more layers of the active material(s) are “sandwiched” between the metal electrodes. To achieve fast response speed in light-sensing ability or resistive switching of the memristor, one has to perform extreme downscale of the active layer thickness. Alternatively, coplanar nanogap electrode architectures may be employed, which offer many advantages, such as lower power consumption, faster operating speed, greater sensitivity and higher-level of integration. Herein, we show high speed light-sensing and memristive devices based on coplanar nanogap (<15 nm) separated electrodes, fabricated with a high throughput, inexpensive nanopatterning technique, compatible with flexible substrates, named adhesion lithography. We demonstrate memristors based on metal oxide, organic and hybrid perovskite materials, discuss the mechanism dominating their operation and showcase their ability to be integrated in arrays in order to perform learning/training operations and other synaptic functionalities.
|12:30||LUNCH BREAK (12:30-14:00)|
Session 6 (14:00-15:30) - Neuromorphic devices for processing, sensing and actuation (2) : Chair - Nawrocki
Authors : Yeongjun Lee; Jin Young Oh; Yeongin Kim; Alex Chortos; Wentao Xu; Dae-Gyo Seo; Hea-Lim Park; Sungjin Park; Zhenan Bao; and Tae-Woo Lee*
Affiliations : Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea; Department of Chemical Engineering, Stanford University, Stanford, CA, USA; Department of Electrical Engineering, Stanford University, Stanford, CA, USA; Department of Materials Science and Engineering, Stanford University, Stanford, CA, USA; Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea; Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea; Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea; Department of Chemical Engineering, Department of Chemistry and Chemical Engineering, Inha University, Incheon, Republic of Korea; Stanford University, Stanford, CA, USA; ;Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea
Resume : Biological sensory nervous systems detect external stimuli and process the detected information. Thus, artificial sensory nervous systems emulating sensing and information processing functions of biological counterparts have abilities for processing detected stimulation and tracking stimulation history including various intensity, duration, and frequency of stimuli. These artificial sensory nerves can perform complicated functions like biological counterparts and can be applied for smart sensors of Internet of Things, biomedical electronics, soft robots, and neural prostheses [1,2]. Here, flexible and stretchable artificial sensory and sensorimotor nervous systems were developed by using organic electronics . For the artificial sensory nerve, pressure sensors (artificial mechanoreceptors), organic ring oscillators (artificial nerve fibers), and synaptic transistors were integrated to emulate the biological mechanoreceptor nerves. To verify the applicability for neural prostheses, a hybrid reflex arc composed of the artificial sensory nerve and biological motor nerves in a detached inset leg was demonstrated and the biological motor nerves were actuated depending on external pressure information. In addition, a stretchable artificial sensorimotor nervous system was developed by integrating a photodetector (an artificial light-sensory organ), a stretchable artificial synapse, and a polymer actuator (an artificial muscle) . The contraction of the artificial muscle was controlled by light stimulation, which means the contraction principle of the biological muscle by the neuromuscular junction was successfully emulated. Furthermore, for an artificial auditory nervous system, an organic artificial synapse was integrated with a triboelectric sensor . To emulate biological synapse facilitation recover time, we modulated morphology of organic semiconductors of the artificial synapses. Lastly, we developed photonic synapses that emulate functions of a retina by using ultraviolet (UV)-responsive 2-dimensional carbon nitride nanodot materials . By using the UV-responsive photonic synapses, a preventive health care system was demonstrated for in-situ modulation of exposure to UV rays depending on the degree of UV exposure and risk. The development of human-like robots, neural prostheses that replicate and expand the human sense, and preventive health care devices can benefit from our work.  H.-L. Park, Y. Lee, N. Kim, D.-G. Seo, G.-T. Go, and T.-W. Lee, Adv. Mater., 2020, 32, 1903558.  Y. Lee, T. W. Lee, Acc. Chem. Res., 2019, 52, 964.  Y. Kim, A. Chotors, W. Xu, Y. Liu, J. Y. Oh, D. Son, J. Kang, A. M. Foudeh, C. Zhu, Y. Lee, S. Niu, J. Liu, R. Pfattner, Z. Bao, T.-W. Lee, Science, 2018, 360, 998.  Y. Lee+, J. Y. Oh+, W. Xu, O. Kim, T. R. Kim, J. Kang, Y. Kim, D. Son, J. B.-H. Tok, M. J. Park, Z. Bao*, T.-W. Lee, Sci. Adv., 2018, 4, eaat7387  D.-G. Seo, Y. Lee, G.-T. Go, M. Pei, S. Jung, Y. H. Jeong, W. Lee, H.-L. Park, S.-W. Kim, H. Yang, C. Yang and T.-W. Lee, Nano Energy, 2019, 65, 104035.  H.-L. Park, H. Kim, D. Lim, H. Zhou, Y.-H. Kim, Y. Lee, S. Park, and T.-W. Lee, Adv. Mater., 2020, 32, 1906899.
Authors : Hongwei Tan1*, Quanzheng Tao2, Ishan Pande1, Sayani Majumdar1,3, Fu Liu4, Yifan Zhou1, Per O. Å. Persson2, Johanna Rosen2, Sebastiaan van Dijken1*
Affiliations : 1NanoSpin, Department of Applied Physics, Aalto University School of Science, P. O. Box 15100, FI-00076 Aalto, Finland 2Thin Film Physics, Department of Physics, Chemistry and Biology (IFM), Linköping University, SE-581 83 Linköping, Sweden 3VTT Technical Research Centre of Finland Ltd., P.O. Box 1000, FI-02044 VTT, Finland 4Department of Electronics and Nanoengineering, Aalto University, P.O. Box 15500, FI-00076 Aalto, Finland
Resume : The integration and cooperation of mechanoreceptors, neurons and synapses in somatosensory systems enable humans to efficiently sense and process tactile information. Inspired by biological somatosensory systems, we report an optoelectronic spiking afferent nerve with neural coding, perceptual learning and memorizing capabilities to mimic tactile sensing and processing. Our system senses pressure by MXene-based sensors, converts pressure information to light pulses by LED-coupled analog-to-digital circuits, then integrates light pulses using a synaptic photomemristor. With neural coding, our spiking nerve is capable of not only detecting simultaneous pressure inputs, but also recognizing Morse code, braille, and object movement. Furthermore, with dimensionality-reduced feature extraction and learning, our system can recognize and memorize handwritten alphabets and words, providing a promising approach towards e-skin, neurorobotics and human-machine interaction technologies.
Authors : Benjamin C.K. Tee
Affiliations : Materials Science and Engineering, National University of Singapore
Resume : Sensory inputs are critical for making intelligent decisions by humans. The five senses, including our sense our touch via skin, provide high-fidelity information about the state of our environment and convey them rapidly to the respective cortexes in the human brain via our nervous system. In artificial systems, such as robots or prosthetics, the ability to asynchronously process and transmit sensory information can help to provide similarly fast, real-time feedback in order to make control decisions. Moreover, asynchronously networked sensors can provide greater robustness to damage if each sensor node operates independently from one another. In my talk, I will discuss our recent work in developing neuromorphic electronic skins inspired by the somatosensory system(1). Such neuromorphic skins can enable greater scalability without sacrificing transmission speed. In addition, they can take advantage of machine learning algorithms to rapidly determine surface features for faster robotic controls. Recently demonstrated computer hardware such as neuromorphic chips can also be used to process the neuromorphic signals and lower the power requirements for processing large amounts of tactile information simultaneously. References: 1. Lee, W. W. et al. A neuro-inspired artificial peripheral nervous system for scalable electronic skins. Sci. Robot. 4, eaax2198 (2019).
|15:30||COFFEE BREAK (15:30-16:00)|
Authors : Kamal Asadi
Affiliations : Max Planck Institute for Polymer Research
Resume : Mimicking the ability of human brain to reconfigure and adopt the neuronal connections using an electronic device will be a major breakthrough, as it would enable realization of brain-like massively parallel and energy efficient computation. Electronic devices that show resistance switching are among candidates for realization of synapses. In this contribution we focus on resistance switching devices based on organic ferroelectrics. A brief overview of ferroelectric field-effect transistors, ferroelectric diodes and ferroelectric tunnel junctions will be given and the operation of the devices will be discussed. Plasticity in ferroelectric-based devices can be achieved only through partial polarization of the ferroelectric. We further discuss the partial polarizability and stability of the states.
Authors : Melkamu Belete *(1), Satender Kataria (1), Thorsten Wahlbrink (2), Olof Engström (2), Max C. Lemme (1,2)
Affiliations : (1)RWTH Aachen University, Germany; (2)AMO GmbH, Germany
Resume : Molybdenum disulfide (MoS2) is a layered two-dimensional (2D) transition metal dichalcogenide (TMD) material which is recently gaining considerable attention for exhibiting a resistive switching (RS) effect. However, the mechanism and origin for the RS behavior still remains unclear despite the claims and hand-waiving arguments made in research reports so far[1,2]. In this work, the presence and origin of a bipolar and forming-free RS of memristors based on polycrystalline MoS2 with vertically aligned layers is demonstrated using experimental evidences. Controlled switching tests under ambient and vacuum conditions indicate nonvolatile RS with stable endurance for at least 140 switching cycles and state-retention for at least 2500 s. From the results, it is inferred that bias-induced movements of mobile hydroxyl ions (OH-) along the vertically aligned MoS2 layers  give rise to the memristive behavior by tuning energy barriers at the silicon (Si)/MoS2 junction as was verified by analytical simulations. The OH- ions possibly originate from catalytic splitting of adsorbed water molecules in MoS2. The observed ion-based plasticity may also be exploited in other TMD materials. The MoS2 used in this work is grown directly on Si and the overall fabrication processes is also scalable and semiconductor manufacturing compatible. This favors integration of novel 2D materials-based memristors into existing Si-technologies to enable potential neuromorphic computing applications. (1) Sangwan et. al, Nat. Nanotechnol. 2015. (2) Kalita et. al, Sci. Rep. 2019. (3) Belete et. al, ACS Appl. Nano Mater. 2018. (4) Belete et. al, Adv. Electron. Mater. 2020. (5) Li et. al, ACS Catal. 2017.
Authors : Chun-Hsiu Chiang, Ling Lee, Tz-Yi Yang, Shin-Yi Tang, Ying-Chun Shen, Yu-Lun Chueh
Affiliations : Department of Material Science and Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan; Department of Material Science and Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan; Department of Material Science and Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan; Department of Material Science and Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan; Department of Material Science and Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan; Department of Material Science and Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan
Resume : Synapse-mimicking devices have attracted considerable attention recently. In a so-called analogue neurocomputer, each memristor can act as individual neurons and synapses, which not only reduce energy consumption but also speed up computation. Previous studies showed that transition metal dichalcogenides can be the host material for ionic modulation. As ions go into this kind of materials, intercalation happens, leading to structural and conductance change, namely hexagonal (2H) structure with semiconducting behavior and tetragonal (1T) structure with metallic behavior. However, most of the intercalations are carried out by immersing samples in organolithium reagent and IV measurement can only be conducted in glove box. In this study, manipulation and phase engineering in MoSe2 lamellar structure for synapse-mimicking devices was demonstrated where all the measurements can be done under atmospheric environment. MoSe2 patterns were synthesized by a specially designed selenization chamber assisted by plasma under low temperature to reduce the heat damage and can reach nowadays electronics requirements for low processing temperature. By using unsymmetrical electrodes design, electric field can induce active metal ions into MoSe2 lamellar structure and change local arrangement thus exhibits conductivity variation. With Raman spectroscopy analysis, MoSe2 pattern has a hybrid structure of 1T (289.4 cm-1) and 2H (165, 236.2 cm-1) phase. Using different bias conditions, ions will migrate and different local distribution of 1T and 2H. When applying DC bias, this work shows great endurance of more than 100 cycles, high stability for retention test (more than 10^4 s) and an on-off ratio about 900 %.
|Start at||Subject View All||Num.|
|08:45||PLENARY SESSION 2 - Prof. Andre Geim, Nobel Laureate in Physics (2010)|
|10:00||COFFEE BREAK (10:00-10:30)|
Authors : Gyeong-Tak Go a, Yeongjun Lee a, Dae-Gyo Seo a, Mingyuan Pei b, Wanhee Lee c, Hoichang Yang b, Tae Woo Lee* a
Affiliations : a Department of Material Science and Engineering, Seoul National University, Korea b Department of Chemical Engineering, Inha University, Korea c Department of Physics and Astronomy, Seoul National University Korea
Resume : ABSTRACT Organic neuromorphic electronics using intrinsic organic semiconductors have emerged to overcome the limitation of conventional electronics by demonstrating various synaptic functions. Although organic synaptic transistors have succeeded in mimic biological synapses, especially short-term plasticity, devices emulated only limited long-term plasticity with short retention behaviors. The long-term retention behavior is essential for non-volatile and long-term memory characteristic in neuromorphic computing. Correlating the synaptic responses with the microstructures of polymer semiconductor is a crucial step to achieve long-term retention. The fundamental study on synaptic responses related to microstructures of organic materials has not been discovered yet. Here, we show that the long-term retention in ion-gel gated organic synaptic transistors (IGOSTs) can be achieved by controlling the microstructure of organic semiconductors. We controlled the crystallinity of poly(3-hexylthiophene-2,5-diyl) (P3HT) in films spun-cast on bare and self-assembled monolayer, before and after thermal treatments. The long-term retention has been significantly prolonged in the IGOSTs, which tends to be elongated as the crystallinity of the semiconductor increased. The microstructure and crystallinity are determined by grazing-incidence X-ray diffraction, optical analyses, and atomic force microscopy. We also evaluate the synaptic current decay behaviors according to a de-doping mechanism of the polymer semiconductor over time. We simulated the recognition of handwritten digits of IGOSTs based on highly-crystalline P3HT, thus achieved higher classification accuracy (>92%) compared with IGOSTs based on low-crystalline P3HT. Our study provides fundamental information about the effects of microstructure on the synaptic responses and strategy to design the IGOSTs for neuromorphic electronics. REFERENCES  W. Xu, S.-Y. Min, H. Hwang, & T.‐W. Lee, Sci. Adv. 2, e1501326 (2016).  Y. Lee, J. Y. Oh, W. Xu, O. Kim, T. R. Kim, J. Kang, Y. Kim, D. Son, J. B.-H. Tok, M. J. Park, Z. Bao and T.-W. Lee, Sci. Adv., 4, eaat7387 (2018).  Y. Kim, A. Chortos, W. Xu, Y. Liu, J. Y. Oh, D. Son, J. Kang, A. M. Foudeh, C. Zhu, Y. Lee, S. Niu, J. Liu, R. Pfattner, Z. Bao & T.-W. Lee, Science, 360, 998–1003 (2018)  Y. Lee, T.-W. Lee, Acc. Chem. Res., 52, 964–974 (2019)
Authors : Mohammad Javad Mirshojaeian Hosseini (1), Elisa Donati (2), Giacomo Indiveri (2), Takao Someya (3), Robert A Nawrocki (1)
Affiliations : (1) School of Engineering Technology, Purdue University, 401 North Grant St, West Lafayette, IN 47907, USA (2) Institute of Neuroinformatics, University of Zurich and ETH Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland (3) Department of Electrical Engineering and Information Systems, School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan
Resume : Spiking neural networks (SNNs) are a type of artificial neural network that encodes information in a frequency domain, in the form of neural spike trains. These networks typically use Integrate-and-Fire (I&F) spiking neurons, hence they mimic biological neural networks more closely than the artificial neural networks based on neurons with graded response. Typically, hardware SNN implementations are based on CMOS circuits that are very fast and implemented on hard and rigid silicon substrates. In contrast, biological neural networks are slow and soft. This creates many levels of difficulty for implantable brain-machine interfaces. Organic electronics are new types of electronics based on organic and polymeric materials. Compared with silicon electronics, they are much slower, but softer and biocompatible. They can be used to implement neurons with compatible speeds and can be fabricated as large-scale, ultra-thin films, making for much better implementation of implantable brain-machine interfaces (BMI). We report the construction of Axon-Hillock (AH) I&F neuron using organic electronics. The circuit employs both n- and p-type organic transistors, along with resistors and capacitors. We demonstrate AH circuit producing series of spikes with frequency proportional to magnitude of input current, akin to the functioning of biological neurons. We also show a basic current summing capability of the neuron. The work proposed shows a path towards long-term implantable SNNs for BMI.
Authors : Jae Seung Jeong, Soo Jin Kim, Jung Ah Lim, Hyunsu Ju
Affiliations : Korea Institute of Science and Technology, University of Science and Technology of Korea ;Korea Institute of Science and Technology, Seoul National University; Korea Institute of Science and Technology; Korea Institute of Science and Technology
Resume : Inspired by neurotransmission of a biological nerve system, ion-gel gated organic transistors are suitable to mimic the neural signal processing activities in the spike neural network. Additionally, the ion-gel transistors are incorporated with ion trapping inside the organic transistor’s channel, in order to implement the synaptic property for the artificial neural network. This bio-inspired artificial neuromorphic device based on the ion-gel organic transistor is fabricated into a fabric form, which enables to literally imitate the neurons connected with each another by the synapses and to inherently simulate the convolutional signal propagation between the neural layers. In this study, the fibrous organic transistors with ion-gel gating are fabricated to form a unit neural connection. The ion-gel gate acts as presynaptic neuron and source-to-drain behaves as postsynaptic neuron, summing the weighted spikes from the presynaptic neurons. The fibrous neuromorphic devices are investigated to implement the neuron behavior, leaky-integrate-and-fire and to adjust the non-volatile synaptic weights to the desired values. Convoluting the presynaptic spike stimuli with the adjustable synaptic weight at the ion-gel gate, the channel conductance correspondingly reacts to represent the postsynaptic neuron’s membrane potential. With these desirable characteristics of the fibrous organic neuromorphic devices, voice recognition is performed to classify the spoken 10-digits.
|12:15||LUNCH BREAK (12:15-14:00)|
Session 9 (14:00-15:30) - Electrochemical concepts for neuromorphic devices : Chair - Vuillaume
Authors : Magnus Berggren, Simone Fabiano, Daniel Simon, Eleni Stavrinidou, Isak Engquist, and Igor Zozoulenko
Affiliations : Laboratory of Organic Electronics, ITN, Linköping University, SE-601 74 Norrköping, Sweden
Resume : Organic electronics can process electronic and ionic signals in a highly coupled manner. Based on this mode of operation feature, several different electrochemical devices can be generated that record and actuate neuronal signals. This opens up for several engineering pathways to mimic and also to connect to the different neuronal systems of biology. Here, the fundamental ion-electron coupling concepts of utilised materials along with the modes of operation of devices will be reported. Also, the development and evolution of circuits will be presented together with some specific applications, targeting novel neuronal interfaces and signal processing concepts.
Authors : Janzakova, K. (1), Kumar, A.* (1), Coffinier, Y. (1), Guérin, D. (1), Alibart, F.(1-2).& Pecqueur, S. (1)
Affiliations : (1) Institut d?Électronique, Microélectronique et Nanotechnologie (IEMN), CNRS, UMR 8520, F-59652 Villeneuve d?Ascq, France. (2) Laboratoire Nanotechnologies & Nanosystèmes (LN2), CNRS, Université de Sherbrooke, J1X0A5, Sherbrooke, Canada. * : corresponding author : email@example.com
Resume : Electropolymerization is an interesting bottom-up strategy to structure conducting materials at the micro/nano-scale in liquid phase that offers a wide morphological versatility. Since the geometry of these structures governs their electrochemical properties, it is fundamental to decipher the mechanisms that rule the polymer assembling upon the electrically-programmed growth to use this phenomenon as a neuro-inspired building block for unconventional information processing. Herein, we investigate various chemical and electrical parameters of electropolymerization affecting the conducting polymer network geometry. We find that various structures such as dendrites, trees, fractals as well as low-fractality cables can be obtained based on applied voltage amplitude, biasing symmetry, bias frequency, concentration of monomers and electrode configurations. We qualitatively and quantitatively study the relationship between the electrical and chemical parameters affecting geometrical parameters of the conducting polymer network as well as electropolymerization dynamics through video and image processing. We find that the different network architectures are associated with different Laplace and diffusion fields governing the monomers motion and in turn electropolymerized network geometry. Such unconventional engineering route could have a variety of applications from neuromorphic engineering to bottom-up computing strategies.
Authors : Alibart, F.* (1) (3)., Ghazal, M. (1), Janzakova, K. (1), Kumar, A., Susloparova, A. (1), Halliez, S. (2), Colin, M. (2), Buée, L. (2), Coffinier, Y (1), Dargent, T. (1), Guérin, D., Pecqueur, S. (1)
Affiliations : (1) Institut d?Électronique, Microélectronique et Nanotechnologie (IEMN), CNRS, UMR 8520, F-59652 Villeneuve d?Ascq, France. (2) Jean-Pierre Aubert Research Centre (JPARC, UMR - S 1172), Université de Lille, , Inserm, CHU-Lille, , 59045 Lille, France (3) Laboratoire Nanotechnologies & Nanosystèmes (LN2), CNRS, Université de Sherbrooke, J1X0A5, Sherbrooke, Canada.
Resume : Most of today?s strategies to interface biology with electronic hardware are based on layered architectures where the front-end of sensing is optimized separately from the back-end for processing/computing signals. Alternatively, biological systems are capitalizing on distributed architecture where both sensing and computing are mixed together and co-optimized. In this talk, we will present our strategy to implement bio-sensing of electroactive cells in a neuromorphic perspective. We will present how organic electrochemical transistors can be used to record electrical signals from neural cells. We will show various strategies capitalizing on the versatility of organic materials synthesis and organic device fabrication to tune and adapt the functionalities of such bio-sensors. We will then present how these strategies can be efficiently used to realize computing functions directly at the interface with biology. Notably, we will illustrate how a network of ionic sensors can implement the reservoir computing concept, a powerful neuromorphic computing approach of particular interest for dynamical signal processing.
|15:30||COFFEE BREAK (15:30-16:00)|
Session 10 (16:00-17:30) - Bioelectronics and adaptive biointerfacing : Chair - Fairfield
Authors : Stefano Buccelli, Yannick Bornat, Timothée Levi, Alberto Averna, David Guggenmos, Randolph J.Nudo, Michela Chiappalone
Affiliations : Stefano Buccelli, Rehab Technologies, Istituto Italiano di Tecnologia, Genova Italy; Yannick Bornat, Laboratoire de l'Intégration du Matériau au Système (IMS), University of Bordeaux, Bordeaux INP, Talence Cedex, France; Timothée Levi, Laboratoire de l'Intégration du Matériau au Système (IMS), University of Bordeaux, Bordeaux INP, Talence Cedex, France & LIMMS CNRS-IIS, The University of Tokyo, Tokyo, Japan; Alberto Averna, Rehab Technologies, Istituto Italiano di Tecnologia, Genova Italy; David Guggenmos, Department of Physical Medicine and Rehabilitation, University of Kansas Medical Center, Kansas City, KS, USA & Landon Center on Aging, University of Kansas Medical Center, Kansas City, KS, USA; Randolph J.Nudo, Department of Physical Medicine and Rehabilitation, University of Kansas Medical Center, Kansas City, KS, USA & Landon Center on Aging, University of Kansas Medical Center, Kansas City, KS, USA; Michela Chiappalone, Rehab Technologies, Istituto Italiano di Tecnologia, Genova Italy;
Resume : Neuroprostheses are neuroengineering devices that have an interface with the nervous system and supplement or substitute functionality in disabled people. In the collective imagination, neuroprotheses are primarily used to restore sensory (e.g. acoustic prostheses) or motor capabilities (e.g. artificial limbs), but in the recent years new devices to be applied directly at the brain level are taking place. To realize energy-efficient and real-time processing devices, closed-loop neuromorphic systems are envisaged as the core of next-generation neuroprosthetics for brain repair. In this talk, I will present the first exploitation of a real-time hardware neuromorphic prosthesis to restore bidirectional interactions between two neuronal populations. We developed an in vitro experimental model of neuronal sub-populations to mimic the mutual interaction between neuronal assemblies and performed a focal lesion to functionally disconnect them. Then, we employed our neuroprosthesis for two potential clinical applications: bidirectional bridging to artificially reconnect two disconnected populations and hybrid bidirectional bridging to replace the activity of one population with a real-time neuromorphic Spiking Neural Network. Further examples of closed-loop neuroprostheses for promoting neuroplasticity in vivo are introduced and discussed. Closed-loop neuroprosthetics based on neuromorphic computation will form the base of novel bioelectrical therapeutics for healthcare.
Authors : Francesca Santoro
Affiliations : Tissue Electronics, Istituto Italiano di Tecnologia, Naples, Italy
Resume : The interface between biological cells and non-biological materials has profound influences on cellular activities, chronic tissue responses, and ultimately the success of medical implants and bioelectronic devices. The optimal coupling between cells, i.e. neurons, and materials is mainly based on surface interaction, electrical communication and sensing. In the last years, many efforts have been devoted to the engineering of materials to recapitulate both the environment ( i.e. dimensionality, curvature, dinamicity) and the functionalities (i.e. long and short term plasticity) of the neuronal tissue to ensure a better integration of the bioelectronic platform and cells. On the one hand, here we explore how the transition from planar to pseudo-3D nanopatterned inorganic and organic materials have introduced a new strategy of integrating bioelectronic platforms with biological cells under static and dynamic conditions. Although a spontaneous penetration does not occur, adhesion processes are such that a very intimate contact can be achieved. On the other hand, we investigate how organic semiconductors can be exploited for recapitulating electrical neuronal functions such as long term and short term potentiation. In this way, both the topology and the material functionalities can be exploited for achieving in vitro biohybrid platforms for neuronal network interfacing.
Authors : Christos Markos , Abubakar I. Adamu , Rune W. Berg 
Affiliations :  DTU Fotonik, Technical University of Denmark, Kgs., DK, 2800, Lyngby, Denmark;  Department of Neuroscience, Faculty of Health and Medical Sciences, University of Copenhagen, Blegdamsvej 3, DK-2200 Copenhagen N, Denmark
Resume : Optical fibers constitute perhaps the most vital component in modern communication technologies. While their usage spans from basic telecommunication applications to high power lasers and medical devices, they have recently attracted significant attention as an enabling technology towards minimally invasive neural implants for in vivo stimulation and recording of biological signals. The development of optical fibers relies upon a thermal-drawing process in which the materials that constitute the macroscopic template of the final fiber - called preform- are molten together in a furnace and drawn down to the final fiber with diameter from millimeters to a few micrometers. The ability of combining novel optical materials with distinct functionalities in the preform stage allows for the development of novel multifunctional fiber-based neural tools for optical, chemical and electrical interrogation of biological signals. In this work, we will present our recent progress on the development of multi-material optical fibers towards novel neural implants. We will show how a co-extrusion and a rod-in-tube fabrication process can be used to combine high performance thermoplastics with low-loss optical polymers, thereby fabricating optical fibers suitable for neuro-stimulation using a 6-meter Draw Tower facility. We will also present recent results on developing multi-material fibers with integrated metal microwires that can act as electrodes for electrical signal recording. Preliminary electro-physiology experiments in mouse brain based on the reported fibers will be also presented.
Authors : Yohan Cho, Young-Geun Park, Jang-Ung Park
Affiliations : Yonsei University, Yonsei-IBS Institution
Resume : Electrophysiological signals of the body provide critical information to assess the functionalities of organs or to diagnose diseases. In order to detect minute biological signals effectively, three-dimensional (3D) electrodes are required to minimize the tissue-electrode impedance and to form conformal interfaces with tissues. However, conventional approaches to form 3D-structured electrodes are based on photolithographic patterning and etching, and these processes result in the low variety of 3D structures limited to certain crystallographic planes of materials. Also, the poor selectivity of materials and their high stiffness easily damage tissues and obstruct the effective signal recording. Herein, the 3D electrode arrays composed of tissue-compatible soft metals have been formed by high-resolution direct printing. Various form factors of 3D electrodes with different heights (20 to 100 μm), diameters (5 to 30 μm) and the tip morphologies (sharpness) are printed and studied in terms of the tissue-electrode interfaces. While conventional materials for 3d microelectrode arrays have compatibility problems such as inflammation or the detachment of tissues because of a modulus mismatch, our 3D electrodes based on soft metals have Young’s modulus at least five orders of magnitude lower than that of platinum and silicon which are conventional materials for bioelectrodes. When human dermal fibroblasts are directly seeded to test the viability of the soft 3D electrodes, over 80% of cells were viable after 7 days with adhesion to the surfaces of electrodes. The improvement of signal transfer, including the signal-to-noise ratio of biopotentials and the persistency of signals at the tissue-electrode interface, can be found by an in-vitro recording of neurons directly cultured on the 3D soft electrode arrays with a comparison to the platinum-based two-dimensional electrode arrays.
Poster session (16:30-18:30) : Chair(s) - Ascoli
Authors : Balzer, F.(1), Abdullaeva, O.S.(2), Maderitsch, A.(2), Schulz, M.(3), Lützen, A.(3), Schiek, M.(2)
Affiliations : (1) Mads Clausen Institute, University of Southern Denmark, Sønderborg, DK (2) Institute of Physics, University of Oldenburg, D (3) Kekulé Institute for Organic Chemistry and Biochemistry, University of Bonn, D
Resume : Interfacing optoelectronics with biology is of emergent interest to design medical tools for personalized healthcare. In particular, optical control of neuroelectronic signaling is an approach that eliminates the need for electrical wiring. Knowledge of the device properties on cellular dimensions are paramount for the design of tailored bionic neural interfaces. Here, local polarized surface photovoltage (SPV) and UV-vis spectroscopy are used to characterize a squaraine:fullerene (SQIB:PCBM) photovoltaic layer blend, which has shown potential to act as neurostimulating platform . The molecular model squaraine donor material SQIB is known to crystallize into two polymorphs upon thermal annealing with distinct polycrystalline thin film texture. For the orthorhombic polymorph, the anisotropic optical response is dominated by the Davydov-split J-type absorption into an upper (UDC) and lower (LDC) Davydov component within the deep red. Kelvin probe force microscopy (KPFM) maps the differential SPV of the active layer on the nanoscale without complications by interfaces, and is spatially correlated with the pleochroic optical response of the thin film. The SPV shows a wavelength-dependent, bichromatic change upon rotating the polarization axis of the illuminating light . With that, subtler nanoscaled optoelectronic sensing platforms become possible for future bioelectronic applications.  O.S. Abdullaeva, F. Balzer, M. Schulz, J. Parisi, A. Lützen, K. Dedek, M. Schiek, Adv. Funct. Mater. 29 (2019) 1805177  F. Balzer, O. S. Abdullaeva, A. Maderitsch, M. Schulz, A. Lützen, M. Schiek, Phys. Status Solidi B (2020), doi: 10.1002/pssb.201900570
Authors : Ying-Chun Shen, Yu-Lun Chueh
Affiliations : Department of Material Science and Engineering, National Tsing Hua University
Resume : As a promising candidate of the emerging devices, resistive switching (RS)-based memristors possess several advantages such as fast switching speed, low power consumption and shrinking size that serve highly potential applications in the next generation nonvolatile memory and neuromorphic computing system. However, it is difficult to realize the neuromorphic computing in filamentary resistive random access memory (RRAM). Previous researches show that disordered oxygen vacancy distribution and multiple-filaments type RRAM give rise to analog switching. In this study, we embedded the nano-pillar Al2O3 architecture in the HfO2-based RRAM to obtain the multiple-filaments. Nano-pillar Al2O3 architecture was fabricated by glancing angle e-beam evaporation to confine the oxygen vacancy regions. HfO2 layer was fabricated by atomic layer deposition, served as the main insulator layer. In filament type HfO2-based RRAM, the set voltage is around -1.5 V, unstable HRS about 10^4-10^6 Ω, and remain strong filament formation and rupture by the pulse measurement. With the nano-pillar Al2O3 architecture, the set voltage can be reduced to -0.5 V, more stable high resistance state (HRS) around 10^5-10^6 Ω, and multi-level resistance state by pulse measurement. The neuromorphic computing can be realized in this novel structure. Keywords: RRAM, neuromorphic computing, nano-pillar architecture, multiple filament
Authors : E.R.W. van Doremaele, Y. van de Burgt
Affiliations : Eindhoven University of Technology, The Netherlands
Resume : Traditional computing systems are unable to capture the capability of the brain in real world information processing as evidenced by the anticipated end to Moore’s law. Organic materials have recently emerged as building blocks of neural processing  and possess basic forms of neuroplasticity and can emulate brain-like functionality at the device level . The excellent characteristics of organic electronic materials, such as low energy operation and tunability, allows these materials to be used as a first step towards efficient neuromorphic computing systems . Nevertheless, fully autonomous bioelectronic applications demand not only the acquisition of biological signals, but also local data processing, storage and the extraction of specific features of merit. As a fist proof of concept, here we will show a simple biosensor based on a neuromorphic array which can be trained to classify a model disease (e.g. cystic fibrosis). We use ion selective sensors that are able to detect physiological levels of potassium and chloride and serve as the input to the hardware-implemented neural network. This example paves the way for more complex trainable biosensors and shows the potential of adaptable neuromorphic devices in a biological environment. 1. van de Burgt et al. Nature Materials, 2017 2. Gkoupidenis et al. Advanced Materials, 2015 3. van de Burgt et al. Nature Electronics, 2018
Authors : Suyoun Lee, Milim Lee, Seong Won Cho, Seon Jeong Kim, Joon Young Kwak, Hyunsoo Ju, Byung-ki Cheong
Affiliations : Korea Institute of Science and Technology, Korea University of Science and Technology
Resume : As an essential building block for developing a large-scale brain-inspired computing system, we present a highly scalable and energy-efficient neuron-mimicking device composed of an Ovonic Threshold Switch (OTS) and a few passive electrical components. It shows not only the basic integrate-and-fire (I&F) function and the rate coding ability, but also a few advanced functions such as the spike-frequency adaptation (SFA) property and the chaotic activity. The latter two, being the most common features found in the mammalian cortex, are particularly essential for the realization of the energy-efficient signal processing and adaptation to environments, but have been challenging goals to achieve up to now. Furthermore, with our OTS-based neuron device employing the reservoir computing technique, spoken-digit recognition task has been performed with a considerable degree of recognition accuracy (~94 %). These results demonstrate that our OTS-based artificial neuron device is promising for the application in the development of a large-scale brain-inspired computing system.
Authors : Filippo Melloni, Dr. Mauro Sassi, Dr. Luca Beverina, Dr. Mario Caironi
Affiliations : Politecnico di Milano; Istituto Italiano di Tecnologia; Università degli Studi di Milano Bicocca
Resume : Edible electronics is an emerging technology that exploits the use of bioinspired and natural materials to manufacture devices that can be safely ingested or disposed in the environment. By overcoming the safety constraints associated with standard ingestible electronics, this technology potentially enables a large scale of applications in medicine and food industry, including biosensing within the gastrointestinal tract, tracking and quality control of foodstuff. Within such highly potential framework, the possibility to process information in a Neuro-inspired way could be disruptive. The power of this alternative computing approach allows to disengage from the classic Von Neumann bus communication paradigm, enabling a highly power efficient electronics, which is a key element for a technology designed for distributed applications and compatible with safe operation in the gastrointestinal tract. Here we show an organic field-effect device gated by an edible and solid electrolyte, showing memristive properties and programmable in a Neuro-inspired way. We demonstrate the possibility to programme such device with a slow train of low-voltage pulses, with pulse amplitude under 500mV, width <10ms and firing frequency in the order of seconds (duty cycle ~1%). In addition, we introduce a proof-of-principle application for this kind of technology by realizing a simple and robust neuromorphic-logic circuit able to elaborate simultaneously two independent and asynchronous input signals.
Authors : Setareh Kazemzadeh3, Scott T. Keene1, Claudia Lubrano2, Armantas Melianas1, Yaakov Tuchman1, Giuseppina Polino2, Lucio Cinà4, Alberto Salleo1, Yoeri van de Burgt3, Francesca Santoro2
Affiliations : 1 Department of Materials Science and Engineering, Stanford University, USA 2 Center for Advanced Biomaterials for Healthcare, Istituto Italiano di Tecnologia, Italy 3 Institute for Complex Molecular Systems, Eindhoven University of Technology, 5612AJ Eindhoven, The Netherlands 4 Cicci Research, Italy
Resume : Recently developed low-power organic neuromorphic systems show promise for combining machine learning with bioelectronic devices and systems. These trainable devices have potential to perform learning and prediction locally using direct feedback from interfacing synapses. The first step towards biologically integrated neuromorphic systems is to develop a bio-feedback mechanism based on chemical as well as electrical activity. In this work, we bridge the gap between artificial and biological neurons by developing a neurotransmitter-sensitive organic neuromorphic device comprising microfabricated artificial synapses and a microfluidic platform. The presented neuromorphic device acts as the postsynaptic side of our hybrid synapse, where a PEDOT:PSS electrode converts chemical signals from biological cells into an long-term electrical response by selective dopamine (DA) oxidation reaction. We characterize the device performance in response to both electrical and chemical inputs to optimize the performance for cell monitoring and recording. Finally, we demonstrate the translation of chemical signals from a PC-12 cell line to resistive changes in the neuromorphic channel and monitor the dynamic response of the neuromorphic device when operated under microfluidic flow to mimic the exocytosis behavior of the cells. To the best of our knowledge, this is the first demonstration of a neuromorphic device with long-term potentiation/depression regulated by neurotransmitter detection. The device presented in this work could constitute a fundamental building block for artificial neural networks which directly adapt based on biological feedback.
Authors : Dilruba Hasina*(1, 2), Tapobrata Som(1, 2)
Affiliations : (1) SUNAG Laboratory, Institute of Physics, Sachivalaya Marg, Bhubaneswar 751 005, India, (2) Homi Bhabha National Institute, Training School Complex, Anushakti Nagar, Mumbai 400 085, India, *lead presenter
Resume : A nociceptor is a critical and a special receptor in the sensory neuron of a human body that responds to damaging stimuli from an external source (like electrical, temperature, pressure, mechanical, etc.) by sending possible threat signals to central nervous system. The key features and functions of a nociceptor are threshold, relaxation, no adaption, sensitization, etc. which are different from other common sensory receptors. Now-a-days the biomimetic behavior of a memristor as an artificial synapse and neuron, has inspired the advent of new information technology in neuromorphic computing. Recently, these artificial nerves are drawing attention for their use as an electronic receptor in the humanoid robots. However, the nociceptive behavior of a memristive device often suffers from the issue of stability which poses challenges for researchers to design neuromorphic devices. In this work, we report on highly stable nociceptive behavior in a Au-ion implanted well-characterized radio-frequency (RF) sputter-deposited TiO2/p -Si memristor. Conductive atomic force microscopy (cAFM)-based current-voltage characteristics of Pt/TiO2/p -Si devices show a prominent and highly stable loop opening with a high ON current/OFF current ratio of ~100 and good switching endurance (>100 cycles). The current conduction mechanism is attributed to the charge trapping/detrapping to/from the traps in TiO2 layers. Further, the electrical-stimuli-induced fundamental nanoscale nociceptive phenomenon such as a threshold, relaxation, allodynia, and hyperalgesia are found in the present devices under a self-biased condition which provides an interesting pathway towards neuromorphic computing which would be useful for artificial intelligence systems like humanoid robots.
Authors : Ghazal, M.* (1), Susloparova, A. (1), Halliez, S. (2), Colin, M. (2), Buée, L. (2), Coffinier, Y (1), Pecqueur S. (1), Dargent, T. (1)& Alibart, F. (1,3).
Affiliations : (1) Institut d’Électronique, Microélectronique et Nanotechnologie (IEMN), CNRS, UMR 8520, F-59652 Villeneuve d’Ascq, France. (2) Jean-Pierre Aubert Research Centre (JPARC, UMR - S 1172), Université de Lille, Inserm, CHU-Lille, 59045 Lille, France (3) Laboratoire Nanotechnologies & Nanosystèmes (LN2), CNRS, Université de Sherbrooke, J1X0A5, Sherbrooke, Canada.
Resume : Organic electrochemical transistors (OECTs) have been recently proposed to record extracellular potentials in electoactive cells culture both in-vitro and in-vivo. In addition to unique properties of interest for electrophysiology such as biocompatibility, transparency, flexibility and high transconductance, OECTs operating principle is based on the transduction of ionic currents in the biological medium into electronic currents in the organic semiconductor (PEDOT:PSS, for instance) via electrochemical coupling. Optimization and control of this iono-electronic coupling is crucial to better sense signals from neural cells and to explore bio-inspired signal transduction. Here, we study both top-down and bottom-up optimization of OECTs devices for in-vitro neural cells culture recording. Neural cells / OECTs coupling is optimized by tuning the transistor geometry for impedance matching. We report an analysis of impedance modification by transistor aspect ratio and effective area tuning in the context of bio-sensing. Bottom-up modification of the organic semiconductor by electropolymerization is a second route that we explore to adjust neural cells / OECTs coupling. In addition to bio-compatibility assessment between primary neural cells culture and various monomers, we show how OECTs impedance can be optimized by electropolymerization of thiophene-derivatives on standard PEDOT:PSS.
Authors : Boncheol Ku, Yu-Rim Jeon, Do Hee Lee, and Changhwan Choi
Affiliations : Division of Materials Science and Engineering, Hanyang University, Seoul 04763, Korea
Resume : Recent work in the field of neuromorphic devices has made great progress, but evaluation of various materials is still ongoing. Various Inorganic and organic materials are being studied for synaptic devices. Inorganic-based synaptic materials offer the advantages of low cost, high stability, and high electrical mobility, but they require high temperature processing, which limits the choice of materials and subsequent processing. Organic-based synaptic materials have an advantage in process cost due to easy solution process, but have difficulties in synthesis and device reliability. Therefore, the materials used to hybridize organic and inorganic materials are approaches to maximize their respective advantages. In this study, we synthesis CH3NH3PbI3 and evaluate Ag/MAPbI3/FTO device as for potential synapse device. We observe analogue characteristics at the low voltage operation, which are important to determine synaptic function, and β-AgI phase at the Ag/MAPbI3 interface is turned out to be important factor to modulate the resistance level (current level) gradually. By careful modulation of this barrier layer, we could see Spike Rate Dependent Plasticity (SRDP), Paired Pulse Facilitation (PPF), Post-Tetanic Potentiation (PTP), Short-Term Memory (STM) to Long-Term Memory (LTM) transition, and Spike Timing Dependent Plasticity (STDP). Our results suggest that hybrid perovskite material can be adopted for the potential synaptic device application. [Acknowledgements] This research was supported by the Nano Material Technology Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of science, ICT & Future Planning. (NRF-2016M3A7B4910426) as well as by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2019R1F1A1057243)
Authors : Chun Yan Gao1,2, Yerin Kim1, Yuri Kim1, Jeongui Lee1, Hyoung Jin Choi2*, and Hoichang Yang1*
Affiliations : 1Department of Chemical Engineering, Inha University, Incheon 22212, Republic of Korea, 2Department of Polymer Science and Engineering, Inha University, Incheon 22212, Republic of Korea
Resume : Ion-gel gated organic synaptic transistors (IGOSTs) provide effective ion-injection to modulate the bulk conductivity of organic semiconductor-based channel layers under a low power consumption, similar to with the biological synapse system, due to their high transconductance. Here, we use 2,3,5,6-tetrafluoro-7,7,8,8-tetracyanoquinodimethane (F4TCNQ, acceptor)-doped regioregular poly (3-hexylthiophene) (RR PHT) semiconducting layers as the channel materials for IGOSTs. Doping level (referred to as [F4TCNQ]/[hexyl thiophene]) is varied from 0 to 0.3 and optimized to control the ordered structures of RR P3HT chains in cast films. Via properly inducing the integer charge transfer (ICT) between RR PHT and F4TCNQ after simple solution blending, the ionized dopants can be allocated in both hexyl side-chain regions and π-conjugated backbones in RR PHT crystallites with an increase in MDR. Also, it is found that the dopants positioned to the intermolecular backbones of P3HT can form a co-crystallite, which has a shorter π-π overlapping structure, in comparison to the undoped RR PHT system. For conventional thin film transistors, a doped RR P3HT film with MDR = 0.03 showed the highest carrier mobility value of up to 0.20 ± 0.06 cm2V-1s-1, which is 7 times higher than 0.028 ± 0.004 cm2V-1s-1 of the undoped system optimized with an additional processing. Also, the optimized doped RR PHT film shows excellent paired-pulse facilitation and long-term plasticity based on spike-number dependent plasticity. The simple blending of the dopant and semiconducting polymer solutions provides the systematic guideline and feasibility of ordering structures and synaptic plasticity in organic artificial synapses application.
Authors : Rhim, S.-Y.*(1), Ligorio, G.(1), Hildebrandt, J.(2), Pätzel, M.(2), Hecht, S.(2, 3, 4) & List-Kratochvil, E.J.W.(1, 5)
Affiliations : (1) Institut für Physik, Institut für Chemie & IRIS Adlershof, Humboldt-Universität zu Berlin, Brook-Taylor-Straße 6, 12489 Berlin, Germany; (2) Institut für Chemie & IRIS Adlershof, Humboldt-Universität zu Berlin, Brook-Taylor-Straße 2, 12489 Berlin, Germany; (3) DWI – Leibniz Institut für interaktive Materialien e.V., RWTH Aachen University, Forckenbeckstraße 50, 52056 Aachen, Germany; (4) Institut für Technische und Makromolekulare Chemie, RWTH Aachen University, Worringerweg 2, 52074 Aachen, Germany; (5) Helmholtz-Zentrum für Materialien und Energie GmbH, Brook-Taylor-Straße 6, 12489 Berlin, Germany;
Resume : Artificial intelligence neural networks are already established in our everyday life and span from rail network planning to revealing chemical structures through data mining. The self-learning as the crucial property of an artificial neural network enables not only to recognize various patterns but also to create new patterns and solve new tasks with training sets but without prewritten codes. Since these digitally realized artificial networks are inspired from the biological nervous system, the programmed networks consist of linked device units called perceptrons similar to the nervous cells. While these simulations are suitable for error analysis and characterization of the network properties, the operation could be more efficient in energy consumption by using analogous realized networks. Therefore, we demonstrate a full optically driven perceptron by using polymeric waveguides. We exploit the total internal reflection effect to generate surface plasmon polaritons on the metal cladding of the waveguide. By depositing photo switchable molecules, we reversibly modulate the dielectric environment near the metal cladding surface and thus the output spectrum by illuminating the switchable molecules with different wavelengths. Using the example of this polymeric waveguide perceptron, we envision various constellations of the perceptron connections enable to learn simple logic operations like AND and OR.
Authors : *Wataru Namiki(1)(2), Takashi Tsuchiya(1), Makoto Takayanagi(1)(2), Tohru Higuchi(2), and Kazuya Terabe(1)
Affiliations : (1) International Center for Materials Nanoarchitectonics (WPI-MANA), National Institute for Materials Science (NIMS), Japan (2) Tokyo University of Science, Japan
Resume : There has been a number of studies to control magnetic anisotropy in magnetic materials toward application to magnetic random-access memory (MRAM) and spintronics-based neuromorphic computing.[1,2] Among various methods (e.g. carrier doping, mechanical stress introduction, and spin current injection),[3-5] electric field control is a promising approach due to low power consumption and relatively simple structure. However, serious problems are still remained. The manipulatable angle on electrostatic carrier doping using ferroelectric is particularly small (~10 degrees) because doped electronic carrier density is too low to tune magnetic properties in typical ferromagnetic materials, in which electronic carrier density is far above 10^20 cm^-3. On the other hand, carrier doping by using solid electrolyte can widely tune magnetic properties in ferromagnetic materials since electronic carrier density doped by electrochemical ion insertion/desertion (reduction and oxidation, redox) is extremely high (> 10^21 cm^-3). Here, we report the modulation of magnetic anisotropy at room temperature (RT), achieved with all-solid-state redox transistor using solid electrolyte. Our all-solid-state redox transistor consists of magnetite with spontaneous magnetization at RT, Li2O-ZrO2-SiO2 electrolyte (Li ion conducting oxide), and LiCoO2 gate electrode thin films. MgO (110) single crystal was used for epitaxial growth of magnetite thin film. Pt was deposited as current electrode with hall-bar shape and current collector. To evaluate magnetization direction variation, planar hall resistance was measured at various gate voltage (voltage applied between magnetite and LiCoO2). Planar hall resistance can be obtained by measuring transverse resistance for current direction under rotating magnetic field clockwise. Magnetic field and electrical current were set at 0.47 T and 7 micro A, respectively. Easy direction of magnetization and magnetic anisotropy energy were obtained by applying local minimum conditions for magnetic free energy and uniaxial magnetic anisotropy to the result. When gate voltage was changed from 0.0 to 2.0 V, electronic current in magnetite increased by about 26 % due to electron doping caused by charge compensation with inserted Li ion. Magnetization aligned to [-111] in the initial state of 0.0 V and switched to [-11-1] beyond [1-10] in the final state of 2.0 V. The corresponding modulation of easy direction of magnetization reached 55.5 degrees, which was never obtained so far only on the basis of carrier doping. Reduction of magnetic anisotropy energy (~38 %) was also observed. It is known that Li ion insertion contributes to not only carrier doping but also large displacement of atomic structure in magnetite. This giant modulation in the present study thus results from magnetic moment modulation through doped carrier and the atomic displacement involved with Li ion insertion. Reference 1. A. Mizrahi et al. Nat. Commun. 9, 1533 (2018). 2. J. Cai et al. Phys. Rev. Appl. 11, 034015 (2019). 3. D. Chiba et al., Nature 455, 25 (2008). 4. S. Zhang et al., Sci. Rep. 4, 3727 (2014). 5. T. Yang et al., Nat. Phys. 4, 851 (2008). 6. T. Tsuchiya et al., ACS Nano 10, 1655 (2016). 7. C. N. Lininger et al., Chem. Mater. 30, 7922 (2018).
Authors : Maxim Zhuk, Sergei Zarubin, Igor Karateev, Yury Matveyev, Evgeny Gornev, Gennady Krasnikov, Dmitrii Negrov and Andrei Zenkevich
Affiliations : Maxim Zhuk, Laboratory of Functional Materials and Devices for Nanoelectronics, Moscow Institute of Physics and Technology (National Research University), Dolgoprudny, Moscow region, Russia; Sergei Zarubin, Laboratory of Functional Materials and Devices for Nanoelectronics, Moscow Institute of Physics and Technology (National Research University), Dolgoprudny, Moscow region, Russia; Igor Karateev, National Research Center “Kurchatov Institute”, Moscow, Russia; Yury Matveyev, Deutsches Elektronen-Synchrotron, Hamburg, Germany; Evgeny Gornev, Molecular Electronics Research Institute (MERI), Moscow, Russia; Gennady Krasnikov, Molecular Electronics Research Institute (MERI), Moscow, Russia; Dmitrii Negrov, Laboratory of Neurocomputing Systems, Moscow Institute of Physics and Technology (National Research University), Dolgoprudny, Moscow region, Russia; Andrei Zenkevich, Laboratory of Functional Materials and Devices for Nanoelectronics, Moscow Institute of Physics and Technology (National Research University), Dolgoprudny, Moscow region, Russia;
Resume : The development of highly integrated electrophysiological devices working in direct contact with living neuron tissue opens new exciting prospects in fields of neurophysiology and medicine, but impose tight requirements on the power dissipated by electronics. On-chip pre-processing of neuronal signals can substantially decrease the power dissipated by external data interfaces and the addition of embedded non-volatile memory would significantly improve the performance of a co-processor in real-time processing of the incoming information stream from the neuron tissue. Here, we evaluate the parameters of TaOx based resistive switching (RS) memory devices produced by magnetron sputtering technique and integrated with the 180 nm CMOS field-effect transistors as possible candidate for on-chip memory in the hybrid neurointerface under development. The electrical parameters of the optimized 1T-1R devices, such as the switching voltage (~ ±1 V), the uniformity of Roff/Ron ratio (~10), read/write speed (<40 ns) and the number of the writing cycles (up to 1010), are satisfactory. The energy for writing and reading out a bit ~30 pJ and ~0.1 pJ, respectively, are also suitable for the desired in vitro neurointerfaces, but is still far too high once the prospective in vivo applications are considered. Challenges arising in the course of the prospective fabrication of the proposed TaOx based RS devices in the back-end-of-line process are identified.
Authors : Sujaya Kumar Vishwanath1, Benny Febriansyah2, Sien Ng1, Rohit Abraham John1, Metikoti Jagadeeswararo2, Naveen Tiwari1, Srilakshmi Subramanian Periyal1, Amoolya Nirmal1, Nripan Mathews1, 2*
Affiliations : 1School of Materials Science & Engineering, Nanyang Technological University, Singapore 639798. 2Energy Research Institute @ NTU (ERI@N), Nanyang Technological University, Singapore 637553.
Resume : Although there have been demonstrations of memristive behavior with organic–inorganic halide perovskite (HPs), no clear pathway for materials development and device design exists for their applications in neuromorphic computing. Present approaches are mostly limited to single element dot-array structures, fall behind in terms of switching reliability and scalability, and fail to map out the analog programmability of such devices. In this work, we systematically design and evaluate robust one-dimensional HP crossbar memristors as weighted synaptic connections for artificial neural networks (ANNs). Utilizing (propyl)pyridinium lead iodide (PrPyrPbI3) as the active switching matrix in a sandwich device configuration, we propose and demonstrate outstanding resistive switching behaviors in low-dimensional HPs in both dot point and crossbar configurations on large-scale flexible substrates. Harnessing the advantages of a forming-less compliance-free operation in the pulsed mode, our devices display analog switching transitions with a soft boundary, sufficing the hardware requirements of weighted synaptic connections in ANNs. Furthermore, capping with an atomic layer deposited (ALD) oxides shows reliable resistive switching devices with long-term stability and reliable, reproducible programmable memory characteristics.
Authors : Sujaya Kumar Vishwanath1, Sien Ng1, Nidhi Tiwari2, Anil Tanwat2, Rohit Abraham John1, Nripan Mathews1, 2*
Affiliations : 1School of Materials Science & Engineering, Nanyang Technological University, Singapore 639798. 2Energy Research Institute @ NTU (ERI@N), Nanyang Technological University, Singapore 637553.
Resume : Brain-inspired artificial neural networks (ANNs) that consists electronic synapses (memristors) are believed as a promising technology for energy efficient artificial intelligence application. Therefore, in recent years, many efforts have been showed to mimic brain-spiking activity in biological synapses by using resistive switching devices. In this work, flexible memristors, based on different functional amorphous oxides were used as synaptic cleft. Further, this devices are investigated as an artificial bio- synapse and essential synaptic functions such as long term potentiation (LTP), long term depression (LTD), paired-pulse facilitation (PPF), post-tetanic potential (PTP) and spike-timing dependent plasticity (STDP) characteristics are well emulated and analogues to the biological synapse behavior. The bending test to assess influence of stress/strain induction on the RS properties is also demonstrated. The modification of the number of states in conductance is realized by applying different pulse schemes are highly suitable for hardware-based neuromorphic applications.
Authors : Hajar Mousavi, Esma Ismailovia, Fabrice Wendling, Mariam Alharrach
Affiliations : Department of Bioelectronics, Ecole Nationale Supérieure des Mines de Saint Etienne, CMP-EMSE, MOC, 13541 Gardanne, France. firstname.lastname@example.org
Resume : In drug-resistant partial epilepsies, resective surgery is the treatment of choice to suppress seizures, provided that the epileptogenic zone (EZ) is clearly identified and that it can be safely removed. In this context, the capacity to rely on objective biomarkers of the EZ is fundamental to define the optimal surgical approach in the specific context of each patient. One of the most important biomarkers are electrophysiological ones which represent local field potential variations generated by a network of neurons. Such variations can be captured by small-scale electrodes implanted in the brain tissue. However, the smaller the electrodes are the higher their impedance will be. One of the strategies to overcome this problem, is to improve the impedance of the electrodes with conductive and biocompatible polymers such PEDOT:PSS. In this work, microwire electrodes are coated with thin layer of PEDOT through electro-polymerization process to achieve high resolution recordings of neural activities. This process enables a controlled tuning of the electrode surface impedance which is essential in the brain – electrode interfacing. Coated electrodes are characterized in term of in vitro impedance and their stability during time as well as the thickness and appearance of the coated layer. Finally, these electrodes are tested in vivo in order to record High Frequency Oscillations (HFOs), which are transient brief interictal signals with a frequency band of 200-600Hz, from the rat hippocampus. PEDOT:PSS coated electrodes are also compared with classical ones to demonstrate their higher performance in HFOs detection.
Authors : Yanxi Zhang, Eveline R.W. van Doremaele, Setareh Kazemzadeh, Yoeri van de Burgt
Affiliations : Microsystems, Institute for Complex Molecular Systems, Eindhoven University of Technology, The Netherlands
Resume : A new class of neuromorphic devices are based on easy-to-tune organic electronic materials. For these “ENODe’s” the conducting polymer poly(3,4-ethylene-dioxythiophene):polystyrene sulfonate (PEDOT:PSS) is often used as the active layer. The devices switch at sub-1V with low power consumption. Meanwhile, numerous individual, non-volatile conductance states can be achieved with relatively long retention time. Besides the active organic conducting layer, the gating electrolyte is essential for the ENODe performance, as well as for future applications. Moreover, all-solid-state devices are crucial for flexible electronics. In this work, we replaced the aqueous electrolyte solution with a solid, nontoxic, biodegradable poly(vinyl alcohol) (PVA) as the gating electrolyte, resulting in all-solid-state devices. We characterized the switching time as well as long-term stability of the devices showing their potential application in neuromorphic computing.
Authors : Janzakova, K.* (1), Ghazal, M. (1), Kumar, A. (1), Coffinier, Y. (1), Guérin, D. (1), Pecqueur, S. (1), Alibart, F. (1-2)
Affiliations : (1) Univ. Lille, CNRS, Centrale Lille, Yncréa ISEN, Univ. Polytechnique Hauts-de-France, UMR 8520 - IEMN, F-59000 Lille, France (2) Laboratoire Nanotechnologies & Nanosystèmes (LN2), CNRS, Université de Sherbrooke, J1X0A5, Sherbrooke, Canada. *lead presenter: email@example.com
Resume : Currently, there is an increasing interest in developing organic electrochemical transistors (OECTs), mostly due to their ability to transduce efficiently ionic signals into electronic ones, which makes them promising tools for applications in various fields such as neurosensing and neuromorphic computing. OECTs’ typical structure includes source, drain, gate electrodes, with conductive organic polymer thin-film acting as a channel. Mastering device geometry and material properties directly affect OECTs performances. Today, as in standard microelectronics, the most common way to fabricate OECTs is based on top-down methods such as lithography and/or printing technologies. In this work, we explore an innovative strategy with bottom-up assembled OECT where the channel is implemented with unconventional dendritic geometry similar to dendritic branching in neural networks. Conducting-polymer based dendritic structures with different morphology are synthesized in a two-electrode setup by pulsed voltage-driven electropolymerization derived from state-of-the-art bipolar AC-electrochemical synthetic methods. Post-fabrication tests with a third electrode show the operability of the polymer micro-objects as the active-element of p-type depletion-mode OECTs with encouraging stability in electrical properties and material integrity.
Authors : Filatov, D.O.(1), Novikov, A.S.(1), Tabakov, O.V.(1), Baranova, V.N.(1), Vrzheshch, D.V.(1), Antonov, D.A.(1), Kruglov, A.V.(1), Belov, A.I.(1), Antonov, I.N.(1), Zdoroveishchev, A.V.(1), Sharkov, V.V.(1), Koryazhkina, M.N.*(1, 2), Mikhaylov, A.N.(1), Gorshkov, O.N.(1), Dubkov, A.A.(1), Carollo, A.(1, 2), Spagnolo, B.(1, 2)
Affiliations : (1)National Research Lobachevsky State University of Nizhny Novgorod, Nizhny Novgorod, Russia (2)University of Palermo, Palermo, Italy
Resume : The development of a new type of computer non-volatile memory (Resistive Random-Access Memory) and its key elements ‒ memristors  or memristive devices ‒ is a topical R&D problem in modern information technologies. The operation of memristive structures is based on the phenomenon of resistive switching (RS) between at least two stable states: high-resistance state and low-resistance state. It should be emphasized that the RS is a stochastic process . In the present work, we have investigated experimentally the response of metal-oxide memristive devices based on ZrO2(Y)/Ta2O5 to a white Gaussian noise synthesized by using analog-to-digital converter. For comparison, we also have investigated the local RS of virtual memristors based on thin ZrO2(Y) film by using atomic force microscopy (AFM), where an AFM probe acts as one of the electrodes. The experimental results obtained in the present study show the applicability of the formalism used in statistical physics to describe the impact of noise on nonlinear multistable systems for understanding the stochastic resistive switching of memristors. The results of the present work indicate that the memristor has a complex multistable potential. The study was supported by the Government of the Russian Federation (Agreement No. 074-02-2018-330 (2)) and the Ministry of Education, University and Research of Italian Government. 1. Chua L O 1971 IEEE Trans. Circuit Theory 18 507 2. Hull R et al 2018 Appl.Phys.Rev. 5 011302
Authors : Gerasimova, S.A.(1), Koryazhkina, M.N.*(1, 2), Belov, A.I.(1), Korolev, D.S.(1), Guseinov, D.V.(1), Mikhaylov, A.N.(1), Kazantsev, V.B.(1), Spagnolo, B.(1, 2)
Affiliations : (1)National Research Lobachevsky State University of Nizhny Novgorod, Nizhny Novgorod, Russia (2)University of Palermo, Palermo, Italy
Resume : Memristor  or memristive device is a simple two-terminal device that can be realized as a capacitor-like thin film stack demonstrating the effect of resistive switching. On the other hand, memristor is a complex nonlinear system, whose dynamic response to external electrical action is well suited for creating the component base of artificial neuromorphic and neurohybrid systems . In the present work, we have investigated the dynamics of optimized memristive device based on ZrO2(Y)/Ta2O5  in response to neuron-like signals and coupling electronic neurons via memristive device both experimentally and theoretically. The simplest experimental system consists of electronic circuit based on the FitzHugh–Nagumo neuron model and memristive device. Real time dynamics of this system is based on stochastic plasticity of memristive device that allows presenting it as an active synapse. The physics-based dynamical model is developed to describe and predict the experimentally observed regularities of synchronization of neuron-like oscillators important for creating self-organized spiking neural networks. The study was supported by the Government of the Russian Federation (Agreement No. 074-02-2018-330 (2)), RFBR (grant No. 18-29-23001) and the Ministry of Education, University and Research of Italian Government. 1. Chua L O 1971 IEEE Trans. Circuit Theory 18 507 2. Mikhaylov A N et al 2018 IEEE TETCI 2 371 3. Mikhaylov A N et al 2019 Adv. Mat. Tech. 5 1900607
Authors : Chansoo Yoon*, Ji Hye Lee*, Sangik Lee*, Jihoon Jeon*, Jun Tae Jang**, Dae Hwan Kim**, Young Heon Kim***, Bae Ho Park*
Affiliations : * Division of Quantum Phases and Devices, Konkuk University, Seoul 05029, Korea ** School of Electrical Engineering, Kookmin University, Seoul 136-702, Korea ***Korea Research Institute of Standards and Science, Daejeon 305-304, Korea
Resume : Selectively activated inorganic synaptic devices, showing a high on/off ratio, ultrasmall dimensions, low power consumption, and short programming time, are required to emulate the functions of high-capacity and energy-efficient reconfigurable human neural systems combining information storage and processing. Here, we demonstrate that such a synaptic device is realized using a Ag/PbZr0.52Ti0.48O3 (PZT)/La0.8Sr0.2MnO3 (LSMO) ferroelectric tunnel junction (FTJ) with ultrathin PZT (thickness of ∼4 nm). Ag ion migration through the very thin FTJ enables a large on/off ratio (107) and low energy consumption (potentiation energy consumption = ∼22 aJ and depression energy consumption = ∼2.5 pJ). In addition, the simple alignment of the downward polarization in PZT selectively activates the synaptic plasticity of the FTJ and the transition from short-term plasticity to long-term potentiation.
|18:30||AWARD CEREMONY followed by SOCIAL EVENT|
|Start at||Subject View All||Num.|
|08:45||PLENARY SESSION 3 - Prof. Ulrike Diebold|
|10:00||COFFEE BREAK (10:00-10:30)|
Session 12 (10:30-12:30) - Devices and systems for neuromorphic computing and sensing : Chair - Fairfield
Authors : T. Venkatesan*^ Sreebrata Goswami* and Sreetosh Goswami*
Affiliations : * NUSNNI, National University of Singapore, Singapore 117411 ^ Azometrix, Washington DC 20008
Resume : Artificial intelligence (AI) has been heralded as the flagbearer of the fourth industrial revolution. But it comes with a cost and that is computing power. It is projected that by 2040, we will need more computing energy than the total energy we can produce now. So, we need devices that can offer higher computing/ storage density with low energy consumption like neuronal computation. We are addressing these challenges using a molecular-electronic route. Historically, organic electronic devices have stimulated scientific excitements in OLEDs but are yet to make any other significant technological impact. The reasons behind their limited success are their poor robustness, stability, endurance and most importantly, the lack of mechanistic understanding that restricts the emergence of approaches to solve these problems. We have overcome each of these difficulties in our memristors based on transition metal complexes of azo-aromatic ligands that exhibit high reproducibility (~350 devices), fast switching (?30 ns), excellent endurance (~10^12 cycles), stability (>10^6 s) and scalability (down to ~60nm^2)1, 2. Using in-situ Raman spectroscopy we are able to track the electronic changes in molecules in-operando at every point of our voltage sweep providing a clear picture of our molecular mechanism that enables us to do different molecular and device engineering to achieve targeted functionalities. Using devices of this genre we are addressing the existing computing challenges via three routes, (i) By designing devices with ultra-low power: We can design memristors with switching voltage as low as 70mV, with energy ~36aJ/ 60nm2. The current and voltage levels of these devices meet the requirements specified in ITRS road map. (ii) By designing memristors and memcapacitors with multiple discrete plateaus3: We have developed memristors with 3- 4 distinct conducting plateaus which also shows mem-capacitance. Their concomitant occurrence is enabled by symmetry breaking of our film-molecules driven by voltage, a new paradigm in condensed matter physics. (iii) Brain inspired computing: Using devices that exhibit concomitant memristive and memcapacitive functions we can simulate biological actions such as neuronal action potential and even cardiac myocyte pulsing. References 1. Goswami S, Matula AJ, Rath SP, Hedström S, Saha S, Annamalai M, et al. Robust resistive memory devices using solution-processable metal-coordinated azo aromatics. Nature materials 2017, 16(12): 1216. 2. Valov I, Kozicki M. Non-volatile memories: Organic memristors come of age. Nature materials 2017, 16(12): 1170. 3. Sreetosh Goswami, Santi P. Rath, Damien Thompson, Svante Hedström, Meenakshi Annamalai, Rajib Pramanick, B. Robert Ilic, Soumya Sarkar, Christian A. Nijhuis, Jens Martin, Sreebrata Goswami, and T. Venkatesan, A Ternary Resistive Memory Device Based on Charge Disproportionate Molecular Redox, Nature Nanotechnology, (2020)
Authors : Ioulia Tzouvadaki, Spyros Stathopoulos, Themis Prodromakis
Affiliations : University of Southampton, Southampton, SO17 1BJ, UK
Resume : Memristive effect may be highlighted as a novel and extremely valuable tool for the monitoring of biological processes opening new perspectives in the biosensing status quo. In this framework, arrays of metal-insulator-metal (MIM) memristive devices recently introduced as multibit memory elements depicting highly packed yet clearly discernible memory states are now challenged for their potential as innovative bio-sensing prototypes. The sensing mechanism is based on the monitoring of the resistive states (RS) over time and during the sensing procedure. For memristors, when the amplitude of input stimulation (e.g. input voltage) exceeds certain thresholds, the occurrence of the event is instantly recorded as a change in the RS. Following this paradigm, RS change occurs as a result of a chemical/biological input that, in this case, has the role of the excitation parameter. MIM devices are fabricated with platinum top and bottom electrodes (10 nm) and reactive magnetron sputtered TiO2-x (25 nm)/AlxOy (4 nm) active layer and are the converted to memristor-based sensing arrays through surface functionalization with receptor molecules i.e. anti-Prostate Specific Antigen (PSA) antibodies, and thereafter exposed to the target molecules (PSA). Fast, label-free sensing procedure is achieved thanks to the in-house memristor characterization platform ArC ONE. Overall, this scheme offers a highly dense sensing platform that paves the way for a robust sensing scheme for disease biomarkers.
Authors : Growney, E.A.*(1,2), Meyer, H. (3), Selby, A. (2), Fairfield, J. (1,2)
Affiliations : (1) CÙRAM, Centre for Research in Medical Devices, National University of Ireland, Galway (2) Department of Physics, National University of Ireland, Galway (3) School of Psychology, National University of Ireland, Galway
Resume : Neuromorphic technology has the potential to overcome size and sensitivity barriers of currently marketed neural electrodes. Titanium dioxide nanowires (NWs) are environmentally sensitive nanomaterials with inherent memristive properties as the current flow can be measurably recorded and deleted within the nanowires as an inorganic analogy to biological neural systems. Neuromorphic behaviour has been demonstrated in a variety of device geometries, but very little has been shown on biological integration with memristive randomly-aligned titanium dioxide NW networks (NWNs) toward electrical brain interfaces. This study examines the biological behaviour of multiple cell types when cultured in vitro on randomly-aligned NW networks. NWNs were physically and electrically characterized followed by addition of naturally-derived substrates known to enhance neural growth: Poly-L-Lysine, Poly-D-Lysine, Collagen I, Laminin-III, and Silk Fibroin. Methodologies for integrating NWNs into biological cultures were developed to preserve the NWN properties while enhancing neural growth, followed by further characterization of biological capabilities using rat embryonic dorsal root ganglion (DRG) explant cultures. Cytotoxicity and neurite outgrowth measurements on multiple neural cell types indicate a preference for NWs with a mainly cytotoxic effect on larger cell types such as fibroblasts. Finally, calcium signaling was performed to analyse neural activity of DRGs when cultured on NWNs. Overall, the results indicate a significant preference for NWNs over control groups regardless of substrate with a higher preference in Laminin-111/NW substrates for neurite outgrowth and reduction in astrocyte migration.
Authors : Fabio Biscarini1,2*, Martina Giordani3, Matteo Sensi2, Marcello Berto2, Michele Di Lauro1, Carlo Augusto Bortolotti2, Henrique L. Gomes4, 5, 2, Michele Zoli3, Francesco Zerbetto6, Luciano Fadiga1,7
Affiliations : 1. Center for Translational Neurophysiology - Istituto Italiano di Tecnologia, Via Fossato di Mortara 17-19, 44100 Ferrara (Italy) 2. Dipartimento di Scienze della Vita - Università di Modena e Reggio Emilia, Via Campi 103, 41125 Modena (Italy) 3. Dipartimento di Scienze Biomediche, Metaboliche e Neuroscienze, Centro di Neuroscienze e Neurotecnologie (CFNN) - Università di Modena e Reggio Emilia, Via Campi 287, 41125 Modena (Italy) 4. Instituto de Telecomunicações, Avenida Rovisco, Pais 1, 1049-001 Lisboa, Portugal 5.Universidade do Algarve, Faculdade de Ciências e Tecnologia, 8005-139 Faro, Portugal. 6. Dipartimento di Chimica “G. Ciamician”, Università di Bologna, Via F. Selmi 2, 40126 Bologna (Italy) 7. Dipartimento di Scienze Biomediche e Chirurgiche Specialistiche, Sezione di Fisiologia Umana, Università di Ferrara, Via Fossato di Mortara 17-19 44100 Ferrara (Italy)
Resume : Specific detection of dopamine (DA) is achieved with organic neuromorphic devices with no specific recognition function towards DA in an electrolyte solution. The response to voltage pulses consists of amplitude-depressed current spiking that mimics the short-term plasticity (STP) of neuronal synapses. A simple equivalent circuit hints that the STP timescale of the device arises from capacitance and resistance of the PEDOT:PSS, the latter in parallel to the electrolyte resistance. Both capacitance and resistance of PEDOT:PSS change with the composition of the buffer solution. Dose curves are constructed from the STP characteristic timescale for each DA metabolite across a range of molar concentrations from 1 pM to 1 mM. Remarkably, STP response of DA is distinctive with respect to the other metabolites even when the latter differ from DA only by one functional group. Both the STP timescale and the sensitivity to DA solutions are two-to-five times larger across the patho-physiological range (0.1-10 nM) with respect to those of the solutions with DA metabolites. Density Functional Theory calculations on clusters hint to a stronger hydrogen bond pattern of DA ammonium end group compared to that of the cationic metabolites. The exponential correlation between the experimental STP timescale and the binding energy of DA metabolites interacting with PEDOT:PSS indicates that the slower dynamics of ionic species in and out PEDOT:PSS upon voltage pulsing is the origin of the neuromorphic STP response of the device. Our sensing framework can discriminate differences of non-specific interactions with the active material as small as a few kcal/mol, else corresponding to one functional group in the molecular structure.
Authors : Giovanni Ligorio (1), Sebastian Nau (2), Johannes Kofler (2), Stefan Sax (2), Norbert Koch (1, 3) and Emil J.W. List-Kratochvil (1, 3)
Affiliations : (1) Institut für Physik, Institut für Chemie & IRIS Adlershof, Humboldt-Universität zu Berlin, Brook-Taylor-Str. 6, D-12489, Berlin, Germany (2) NanoTecCenter Weiz Forschungsgesellschaft mbH, Franz-Pichler-Straße 32, A-8160, Weiz, Austria (3) Helmholtz Zentrum Berlin für Materialien und Energie GmbH, Albert-Einstein-Str. 15, D-12489, Berlin, Germany
Resume : Amongst the different candidates competing to be the next new-generation memory technology, resistive switching non-volatile memories (R-NVMs) are increasingly appealing and are on the edge of commercialization. An R-NVM is a two-terminal device made by an interlayer material sandwiched between two metal electrodes. In order to successfully match low manufacturing costs with the need for high-density memory modules, the size-miniaturization of organic R-NVM must be further investigated and verified whether the resistance-switch mechanism occurs at the lateral nanometric scale. In this contribution, we report on the size miniaturization of organic R-NVMs on the nanometric scale. First, we employed glancing angle deposition to realize vertical nanocolumn-structures resembling nano-devices. The devices based on small molecule semiconductors were fabricated with a radius smaller than 50 nm. It is worth notice that the devices could not have been reversibly switched between both logic states because of technical limitations given by the measurement setup. Therefore, to overcome these technical limitations and to verify that reversible switching is possible at the nanometric scale, analogous nano-devices (40 nm x 40 nm) were fabricated via electron-beam lithography. Standard current vs.voltage electric characterization was performed confirming the ability to reversibly switch the device in both logic states. In conclusion, resistive switching at the nanometric scale of organic NVMs was investigated and it is confirmed that filament formation switching takes place also within nanometric devices. High-density organic NVM arrays were therefore fabricated achieving information density of more than 1 GB/cm2.
Authors : Mikko Nisula(1), Antti J. Karttunen(2), Christophe Detavernier(1)
Affiliations : (1)Department of Solid State Sciences, Ghent University, Belgium (2)Department of Chemistry and Materials Science, Aalto University, Finland
Resume : Recently, conductive polymers have been shown to be a highly promising material candidate for memristive synaptic devices. Metal-organic coordination polymers present an interesting extension for this material group. Similar to conductive polymers, their electronic conductivity can be modulated electrochemically by controlling the oxidation state of the material. The interplay of the metal and organic constituents enables the tuning of the properties towards low device conductance and low power consumption. Here we investigate Cu-dithiooxamide based thin films made with molecular layer deposition (MLD), a derivative of atomic layer deposition. We show that depending on the oxidation state of the organic ligand, the electric conductivity of the material can be tuned between 10-10 and 10-2 S/cm. Simple two-terminal devices relying on ambient humidity as the redox pair are manufactured. As the conductance switching is governed by the redox-reaction potential, the cycle-to-cycle variation is extremely low. The conductance can be tuned in continuous, linear fashion, and apparent interplay of ion and electron conduction allows for realization of simultaneous short- and long-term memory effects in a single device. The MLD-based approach results in high quality thin films with excellent thickness uniformity and low surface roughness over large area substrates. As a solvent-free, gas-phase processing technique it has potential in integration with existing CMOS processing methods.
|12:30||LUNCH BREAK (12:30-14:00)|
Session 13 (14:00-16:00) - Neuromorphic devices : Chair - Ligorio : Chair - Ligorio
Authors : Victor Erokhin
Affiliations : Italian National Council of Researches, Institute of Materials for Electronics and Magnetism, Parco Area delle Scienze 37/A, 43124, Parma Italy National Research Center ?Kurchatov Institute?, 123182, Moscow, Russia
Resume : Polyaniline-based memristive devices working principle is based on the significant ratio of the conductivity in reduced and oxidized states and they have some properties similar to those of synapses. In this presentation we will consider two areas of their applications. The first one is connected to the artificial neuronal networks. We will consider the realization of logic gates with memory, as well as single  and double  layer perceptrons. The second part will be dedicated to bio-mimicking applications. We will consider a reproduction of a part of the nervous system of pond snail. Then, a system, imitating Pavlov's dog STDP-like learning, using spiking signals , will be described. Finally, experiments on the synapse-like connections of two live nervous cells from rat cortex through organic memristive devices will be considered . Advantages and drawbacks of this system will be analysed in comparison with other memristive devices and systems. 1. V.A. Demin et al., Org. Electronics, vol. 25, pp. 16-20 (2015). 2. A.V. Emelyanov et al., AIP Adv., vol. 6, p. 111301 (2016). 3. A.A. Minnekhanov et al., Sci. Rep., vol. 9, p. 10800 (2019). 4. E. Juzekaeva et al.,Adv. Mater. Technol., vol. 4, p. 1800350 (2019).
Authors : Alon Ascoli1, Ronald Tetzlaff1, Ioannis Messaris1, Steve Kang2, and Leon Chua3
Affiliations : 1 Chair of Fundamentals of Electrical Engineering, Institute of Circuits and Systems, Faculty of Electrical and Computer Engineering, Technische Universität Dresden, Dresden, Germany 2 Jack Baskin School of Engineering, University of California Santa Cruz, Santa Cruz, CA 95064, USA 3 Department of Electrical Engineering and Computer Sciences, University of California Berkeley, Berkeley, CA 94720 USA
Resume : As CMOS scaling is approaching atomic boundaries, the end of Moore?s era  seems inevitable. It is for this reason that engineers worldwide are investigating new strategies  to keep the integrated circuit (IC) performance growth rate, predicted by Moore, in the years to come, without the need to shrink the transistor size any further. The search for novel devices, capable to combine within a single physical nanoscale volume multiple functionalities, is one of the approaches of greater interest. In this respect, the memory resistor, memristor for short , offers extraordinary opportunities for the electronics of the future. Given the inherently-nonlinear mechanisms at the origin of the dynamical behaviors of memristors -, the applicability of linear system-theoretic methodologies for the investigation and design of circuits based upon them is rather limited . In this presentation we show how the generalization of a powerful nonlinear system-theoretic technique, known as Dynamic Route Map (DRM)  and instrumental to analyze first-order mathematical descriptions, such as those governing the dynamics of memristors with scalar state, to ordinary differential equations with two degrees of freedom  allows to investigate  bio-inspired memcomputing memristor circuits. Furthermore, shaping the generalized DRM of a second-order memristive circuit, through the application of a comprehensive stability analysis to its mathematical description, allows to guide the circuit solutions from prescribed initial conditions toward desired attractors, allowing the implementation of a predefined memory or computing task . This study reveals the significant role, which nonlinear system theory is expected to play in the years to come, for the development of a systematic approach to memristor circuit design. Bibliography:  G.E. Moore, ?Cramming more components onto integrated circuits,? Electronics, vol. 38, no. 8, pp. 114-117, 1965  R.S. Williams, ?What's next? [The end of Moore's law],? IEEE Computing in Science & Engineering, vol. 19, no. 2, pp. 7-13, 2017, DOI: 10.1109/MCSE.2017.31  L.O. Chua, "Five Non-Volatile Memristor Enigmas Solved," Applied Physics A, vol. 124, no. 8, 563(43pp.), 2018  A. Ascoli, S. Slesazeck, H. Mähne, R. Tetzlaff, and T. Mikolajick, ?Nonlinear dynamics of a locally-active memristor,? IEEE Trans. Circuits and Systems--I (TCAS--I): Regular Papers, vol. 62, no. 4, pp. 1165-1174, 2015  A. Ascoli, R. Tetzlaff, L.O. Chua, J.P. Strachan, and R.S. Williams, "History Erase Effect in a Non-Volatile Memristor," IEEE Trans. on Circuits and Systems-I (TCAS-I): Regular Papers, vol. 63, no. 3, pp. 389-400, 2016  A. Ascoli, R. Tetzlaff, and M. Biey, "Memristor and Memristor Circuit Modelling based on Methods of Nonlinear System Theory," Springer Lecture Notes on Nonlinear Dynamics in Computational Neuroscience, F. Corinto, and A. Torcini eds., pp. 99-132, 2018, DOI: https://doi.org/10.1007/978-3-319-71048-8 7  R. Tetzlaff, A. Ascoli, I. Messaris, and L.O. Chua, "Theoretical Foundations of Memristor Cellular Nonlinear Networks: Memcomputing with Bistable-like Memristors," IEEE Trans. on Circuits and Systems-I: Regular Papers (TCAS-I), 2019, DOI: 10.1109/TCSI.2019.2940909  A. Ascoli, I.Messaris, R. Tetzlaff, and L.O. Chua, "Theoretical Foundations of Memristor Cellular Nonlinear Networks: Stability Analysis with Dynamic Memristors," IEEE Trans. on Circuits and Systems-I: Regular Papers (TCAS-I), 2019, DOI: 10.1109/TCSI.2019.2957813  A. Ascoli, R. Tetzlaff, S. Kang, and L.O. Chua, ?Theoretical Foundations of Memristor Cellular Nonlinear Networks: a DRM2-based method to design memcomputers with dynamic memristors,? IEEE Trans. on Circuits and Systems-I: Regular Papers (TCAS-I), 2020
Authors : Asal Kiazadeh*, Maria Pereira, Jonas Deuermeier, Joana Luis, João Paulo, Rodrigo Martins, Elvira Fortunato *e-mail: firstname.lastname@example.org
Affiliations : i3N/CENIMAT, Department of Materials Science, Faculty of Science and Technology, Universidade NOVA de Lisboa and CEMOP/UNINOVA, Campus de Caparica, 2829-516 Caparica, Portugal
Resume : Amorphous oxide semiconductors (AOS) are applied in industry for pixel driver thin-film transistors (TFT) of flat panel displays, hence they are ideal resistive switching materials for hardware artificial intelligence systems on panel. For the realization of high-density crossbar arrays, asymmetric memdiodes are appealing, because the rectification allows for a selector-less configuration. Here we present a noble-metal-free Schottky contact to indium-gallium-zinc oxide (IGZO). The base material for the Schottky bottom and the Ohmic top contact is identical and commonly used for source and drain contacts of AOS TFTs – molybdenum. The barrier formation is induced by the oxidation of the molybdenum surface during the sputter deposition of IGZO. The device shows well-controllable multi resistance states by keeping the rectification template with great endurance and reproducibility under multiple cycles of both DC sweeps (<500 DC sweeps) and pulses. Several synaptic functions are shown, such as synaptic potentiation and depression as response to programmed pulses, short to long term plasticity transition (STP to LTP) and “learning experience” properties. The Mo/IGZO/Mo memdiode shows potential application of an electronic synapse: an approach towards merging computing and memory application for an intelligent machine.
Authors : Morteza Hassanpour Amiri, Kamal Asadi (Correspond Author )
Affiliations : Molecular Electronic Department, Max Planck Institute for polymer research, Mainz, Germany
Resume : Brain's ability to reconfigure the neuronal connections and adapt as the result of experience is called neuroplasticity. Mimicking this capability is the milestone for realization of Brain-like massively parallel and energy efficient silicon intelligence. Hereby, we present a memory element made of graphene ferroelectric field effect transistor capable of modulate the conductance of the channel in some distinguishable states, demonstrating similar behavior of the brain. Using empirical description of ferroelectricity in P(VDF-TrFE), combined with physics of Graphene-FETs enabled us to derive an analytic model for the realized device.
Authors : Michal Harasimiuk, Noushin Rasti, Piers Barnes
Affiliations : Imperial College
Resume : We focus on the I-V characteristics of planar metal halide perovskite devices operating under a low voltage regime, and show that these devices exhibit strong mem-diode characteristics. We present a macroscopic model explaining the observed effects, using an ionically gated transistor interface circuit model. The model is used to analyse the behaviour of the device under a number of input voltage profiles, from simple pulsed and periodic inputs, to more complex ones, which demonstrate responses analogous to synaptic neuroplasticity. The model data are compared to experimental observations and the results of numerical drift diffusion simulations incorporating the presence of mobile ions. We also consider the reversibility and stability of devices when subjected to prolonged and repeated exposure to the input voltage profiles. We close our discussion with examples of how such devices can be used in constructing simple neuromorphic circuits.
No abstract for this day
Department of Electrical and Computer Engineering, San Diego, USA+1 858 534 2985
School of Physics, Galway, Ireland+353 91 492494
Department of Molecular Electronics, Mainz, Germany+49 6131379 605
Department of Mechanical Engineering, Eindhoven, The Netherlands+31402474419