Structural plasticity for neuromorphic networks with electropolymerized dendritic PEDOT connections.

Neural networks are powerful tools for solving complex problems, but finding the right network topology for a given task remains an open question. Biology uses neurogenesis and structural plasticity to solve this problem. Advanced neural network algorithms are mostly relying on synaptic plasticity and learning. The main limitation in reconciling these two approaches is the lack of a viable hardware solution that could reproduce the bottom-up development of biological neural networks. Here, we show how the dendritic growth of PEDOT:PSS-based fibers through AC electropolymerization can implement structural plasticity during network development. We find that this strategy follows Hebbian principles and is able to define topologies that leverage better computing performances with sparse synaptic connectivity for solving non-trivial tasks. This approach is validated in software simulation, and offers up to 61% better network sparsity on classification and 50% in signal reconstruction tasks.

Electropolymerization processing of side-chain engineered EDOT for high performance microelectrode arrays.

Microelectrode Arrays (MEAs) are popular tools for in vitro extracellular recording. They are often optimized by surface engineering to improve affinity with neurons and guarantee higher recording quality and stability. Recently, PEDOT:PSS has been used to coat microelectrodes due to its good biocompatibility and low impedance, which enhances neural coupling. Herein, we investigate on electro-co-polymerization of EDOT with its triglymated derivative to control valence between monomer units and hydrophilic functions on a conducting polymer. Molecular packing, cation complexation, dopant stoichiometry are governed by the glycolation degree of the electro-active coating of the microelectrodes. Optimal monomer ratio allows fine-tuning the material hydrophilicity and biocompatibility without compromising the electrochemical impedance of microelectrodes nor their stability while interfaced with a neural cell culture. After incubation, sensing readout on the modified electrodes shows higher performances with respect to unmodified electropolymerized PEDOT, with higher signal-to-noise ratio (SNR) and higher spike counts on the same neural culture. Reported SNR values are superior to that of state-of-the-art PEDOT microelectrodes and close to that of state-of-the-art 3D microelectrodes, with a reduced fabrication complexity. Thanks to this versatile technique and its impact on the surface chemistry of the microelectrode, we show that electro-co-polymerization trades with many-compound properties to easily gather them into single macromolecular structures. Applied on sensor arrays, it holds great potential for the customization of neurosensors to adapt to environmental boundaries and to optimize extracted sensing features.

Precision of neuronal localization in 2D cell cultures by using high-performance electropolymerized microelectrode arrays correlated with optical imaging.

Recently, the development of electronic devices to extracellularly record the simultaneous electrical activities of numerous neurons has been blooming, opening new possibilities to interface and decode neuronal activity. In this work, we tested how the use of EDOT electropolymerization to tune post-fabrication materials could optimize the cell/electrode interface of such devices. Our results showed an improved signal-to-noise ratio, better biocompatibility, and a higher number of neurons detected in comparison with gold electrodes. Then, using such enhanced recordings with 2D neuronal cultures combined with fluorescent optical imaging, we checked the extent to which the positions of the recorded neurons could be estimated solely via their extracellular signatures. Our results showed that assuming neurons behave as monopoles, positions could be estimated with a precision of approximately tens of micrometers.

Voltage-dependent synaptic plasticity: Unsupervised probabilistic Hebbian plasticity rule based on neurons membrane potential.

This study proposes voltage-dependent-synaptic plasticity (VDSP), a novel brain-inspired unsupervised local learning rule for the online implementation of Hebb's plasticity mechanism on neuromorphic hardware. The proposed VDSP learning rule updates the synaptic conductance on the spike of the postsynaptic neuron only, which reduces by a factor of two the number of updates with respect to standard spike timing dependent plasticity (STDP). This update is dependent on the membrane potential of the presynaptic neuron, which is readily available as part of neuron implementation and hence does not require additional memory for storage. Moreover, the update is also regularized on synaptic weight and prevents explosion or vanishing of weights on repeated stimulation. Rigorous mathematical analysis is performed to draw an equivalence between VDSP and STDP. To validate the system-level performance of VDSP, we train a single-layer spiking neural network (SNN) for the recognition of handwritten digits. We report 85.01 ± 0.76% (Mean ± SD) accuracy for a network of 100 output neurons on the MNIST dataset. The performance improves when scaling the network size (89.93 ± 0.41% for 400 output neurons, 90.56 ± 0.27 for 500 neurons), which validates the applicability of the proposed learning rule for spatial pattern recognition tasks. Future work will consider more complicated tasks. Interestingly, the learning rule better adapts than STDP to the frequency of input signal and does not require hand-tuning of hyperparameters.

Theoretical modeling of dendrite growth from conductive wire electro-polymerization.

Electropolymerization is a bottom-up materials engineering process of micro/nano-scale that utilizes electrical signals to deposit conducting dendrites morphologies by a redox reaction in the liquid phase. It resembles synaptogenesis in the brain, in which the electrical stimulation in the brain causes the formation of synapses from the cellular neural composites. The strategy has been recently explored for neuromorphic engineering by establishing link between the electrical signals and the dendrites' shapes. Since the geometry of these structures determines their electrochemical properties, understanding the mechanisms that regulate polymer assembly under electrically programmed conditions is an important aspect. In this manuscript, we simulate this phenomenon using mesoscale simulations, taking into account the important features of spatial-temporal potential mapping based on the time-varying signal, the motion of charged particles in the liquid due to the electric field, and the attachment of particles on the electrode. The study helps in visualizing the motion of the charged particles in different electrical conditions, which is not possible to probe experimentally. Consistent with the experiments, the higher AC frequency of electrical activities favors linear wire-like growth, while lower frequency leads to more dense and fractal dendrites' growth, and voltage offset leads to asymmetrical growth. We find that dendrites' shape and growth process systematically depend on particle concentration and random scattering. We discover that the different dendrites' architectures are associated with different Laplace and diffusion fields, which govern the monomers' trajectory and subsequent dendrites' growth. Such unconventional engineering routes could have a variety of applications from neuromorphic engineering to bottom-up computing strategies.

Analog programing of conducting-polymer dendritic interconnections and control of their morphology.

Although materials and processes are different from biological cells', brain mimicries led to tremendous achievements in parallel information processing via neuromorphic engineering. Inexistent in electronics, we emulate dendritic morphogenesis by electropolymerization in water, aiming in operando material modification for hardware learning. Systematic study of applied voltage-pulse parameters details on tuning independently morphological aspects of micrometric dendrites': fractal number, branching degree, asymmetry, density or length. Growths time-lapse image processing shows spatial features to be dynamically dependent, and expand distinctively before and after conductive bridging with two electro-generated dendrites. Circuit-element analysis and impedance spectroscopy confirms their morphological control in temporal windows where growth kinetics is finely perturbed by the input frequency and duty cycle. By the emulation of one's most preponderant mechanisms for brain's long-term memory, its implementation in vicinity of sensing arrays, neural probes or biochips shall greatly optimize computational costs and recognition required to classify high-dimensional patterns from complex environments.

Dendritic Organic Electrochemical Transistors Grown by Electropolymerization for 3D Neuromorphic Engineering.

One of the major limitations of standard top-down technologies used in today's neuromorphic engineering is their inability to map the 3D nature of biological brains. Here, it is shown how bipolar electropolymerization can be used to engineer 3D networks of PEDOT:PSS dendritic fibers. By controlling the growth conditions of the electropolymerized material, it is investigated how dendritic fibers can reproduce structural plasticity by creating structures of controllable shape. Gradual topologies evolution is demonstrated in a multielectrode configuration. A detailed electrical characterization of the PEDOT:PSS dendrites is conducted through DC and impedance spectroscopy measurements and it is shown how organic electrochemical transistors (OECT) can be realized with these structures. These measurements reveal that quasi-static and transient response of OECTs can be adjusted by controlling dendrites' morphologies. The unique properties of organic dendrites are used to demonstrate short-term, long-term, and structural plasticity, which are essential features required for future neuromorphic hardware development.

Oxygen vacancy engineering of TaOx-based resistive memories by Zr doping for improved variability and synaptic behavior.

Resistive switching (RS) devices are promising forms of non-volatile memory. However, one of the biggest challenges for RS memory applications is the device-to-device (D2D) variability, which is related to the intrinsic stochastic formation and configuration of oxygen vacancy (VO) conductive filaments (CFs). In order to reduce the D2D variability, control over the formation and configuration of oxygen vacancies is paramount. In this study, we report on the Zr doping of TaOx-based RS devices prepared by pulsed-laser deposition as an efficient means of reducing the VOformation energy and increasing the confinement of CFs, thus reducing D2D variability. Our findings were supported by XPS, spectroscopic ellipsometry and electronic transport analysis. Zr-doped films showed increased VOconcentration and more localized VOs, due to the interaction with Zr. DC and pulse mode electrical characterization showed that the D2D variability was decreased by a factor of seven, the resistance window was doubled, and a more gradual and monotonic long-term potentiation/depression in pulse switching was achieved in forming-free Zr:TaOxdevices, thus displaying promising performance for artificial synapse applications.

Conductive filament evolution dynamics revealed by cryogenic (1.5 K) multilevel switching of CMOS-compatible Al2O3/TiO2 resistive memories.

Non-volatile resistive switching devices are considered as prime candidates for next-generation memory applications operating at room temperature and above, such as resistive random-access memories or brain-inspired in-memory computing. However, their operability in cryogenic conditions remains to be mastered to adopt these devices as building blocks enabling large-scale quantum technologies via quantum-classical electronics co-integration. This study demonstrates multilevel switching at 1.5 K of Al2O3/TiO2-x resistive memory devices fabricated with complementary metal-oxide-semiconducto-compatible processes and materials. The I-V characteristics exhibit a negative differential resistance (NDR) effect due to a Joule-heating-induced metal-insulator transition of the Ti4O7 conductive filament. Carrier transport analysis of all multilevel switching I-V curves show that while the insulating regime follows the space charge limited current (SCLC) model for all resistance states, the conduction in the metallic regime is dominated by SCLC and trap-assisted tunneling for low- and high-resistance states respectively. A non-monotonic conductance evolution is observed in the insulating regime, as opposed to the continuous and gradual conductance increase and decrease obtained in the metallic regime during the multilevel SET and RESET operations. Cryogenic transport analysis coupled to an analytical model accounting for the metal-insulator-transition-induced NDR effects and the resistance states of the device provide new insights on the conductive filament evolution dynamics and resistive switching mechanisms. Our findings suggest that the non-monotonic conductance evolution in the insulating regime is due to the combined effects of longitudinal and radial variations of the Ti4O7 conductive filament during the switching. This behavior results from the interplay between temperature- and field-dependent geometrical and physical characteristics of the filament.

Concentric-Electrode Organic Electrochemical Transistors: Case Study for Selective Hydrazine Sensing.

We report on hydrazine-sensing organic electrochemical transistors (OECTs) with a design consisting of concentric annular electrodes. The design engineering of these OECTs was motivated by the great potential of using OECT sensing arrays in fields such as bioelectronics. In this work, poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS)-based OECTs have been studied as aqueous sensors that are specifically sensitive to the lethal hydrazine molecule. These amperometric sensors have many relevant features for the development of hydrazine sensors, such as a sensitivity down to 10-5 M of hydrazine in water, an order of magnitude higher selectivity for hydrazine than for nine other water-soluble common analytes, the capability to entirely recover its base signal after water flushing, and a very low operation voltage. The specificity for hydrazine to be sensed by our OECTs is caused by its catalytic oxidation at the gate electrode, and enables an increase in the output current modulation of the devices. This has permitted the device-geometry study of the whole series of 80 micrometric OECT devices with sub-20-nm PEDOT:PSS layers, channel lengths down to 1 µm, and a specific device geometry of coplanar and concentric electrodes. The numerous geometries unravel new aspects of the OECT mechanisms governing the electrochemical sensing behaviours of the device-more particularly the effect of the contacts which are inherent at the micro-scale. By lowering the device cross-talk, micrometric gate-integrated radial OECTs shall contribute to the diminishing of the readout invasiveness and therefore further promote the development of OECT biosensors.

Interplay of multiple synaptic plasticity features in filamentary memristive devices for neuromorphic computing.

Bio-inspired computing represents today a major challenge at different levels ranging from material science for the design of innovative devices and circuits to computer science for the understanding of the key features required for processing of natural data. In this paper, we propose a detail analysis of resistive switching dynamics in electrochemical metallization cells for synaptic plasticity implementation. We show how filament stability associated to joule effect during switching can be used to emulate key synaptic features such as short term to long term plasticity transition and spike timing dependent plasticity. Furthermore, an interplay between these different synaptic features is demonstrated for object motion detection in a spike-based neuromorphic circuit. System level simulation presents robust learning and promising synaptic operation paving the way to complex bio-inspired computing systems composed of innovative memory devices.

Plasticity in memristive devices for spiking neural networks.

Memristive devices present a new device technology allowing for the realization of compact non-volatile memories. Some of them are already in the process of industrialization. Additionally, they exhibit complex multilevel and plastic behaviors, which make them good candidates for the implementation of artificial synapses in neuromorphic engineering. However, memristive effects rely on diverse physical mechanisms, and their plastic behaviors differ strongly from one technology to another. Here, we present measurements performed on different memristive devices and the opportunities that they provide. We show that they can be used to implement different learning rules whose properties emerge directly from device physics: real time or accelerated operation, deterministic or stochastic behavior, long term or short term plasticity. We then discuss how such devices might be integrated into a complete architecture. These results highlight that there is no unique way to exploit memristive devices in neuromorphic systems. Understanding and embracing device physics is the key for their optimal use.

Filamentary switching: synaptic plasticity through device volatility.

Replicating the computational functionalities and performances of the brain remains one of the biggest challenges for the future of information and communication technologies. Such an ambitious goal requires research efforts from the architecture level to the basic device level (i.e., investigating the opportunities offered by emerging nanotechnologies to build such systems). Nanodevices, or, more precisely, memory or memristive devices, have been proposed for the implementation of synaptic functions, offering the required features and integration in a single component. In this paper, we demonstrate that the basic physics involved in the filamentary switching of electrochemical metallization cells can reproduce important biological synaptic functions that are key mechanisms for information processing and storage. The transition from short- to long-term plasticity has been reported as a direct consequence of filament growth (i.e., increased conductance) in filamentary memory devices. In this paper, we show that a more complex filament shape, such as dendritic paths of variable density and width, can permit the short- and long-term processes to be controlled independently. Our solid-state device is strongly analogous to biological synapses, as indicated by the interpretation of the results from the framework of a phenomenological model developed for biological synapses. We describe a single memristive element containing a rich panel of features, which will be of benefit to future neuromorphic hardware systems.

Modeling and Experimental Demonstration of a Hopfield Network Analog-to-Digital Converter with Hybrid CMOS/Memristor Circuits.

The purpose of this work was to demonstrate the feasibility of building recurrent artificial neural networks with hybrid complementary metal oxide semiconductor (CMOS)/memristor circuits. To do so, we modeled a Hopfield network implementing an analog-to-digital converter (ADC) with up to 8 bits of precision. Major shortcomings affecting the ADC's precision, such as the non-ideal behavior of CMOS circuitry and the specific limitations of memristors, were investigated and an effective solution was proposed, capitalizing on the in-field programmability of memristors. The theoretical work was validated experimentally by demonstrating the successful operation of a 4-bit ADC circuit implemented with discrete Pt/TiO2- x /Pt memristors and CMOS integrated circuit components.

Pattern classification by memristive crossbar circuits using ex situ and in situ training.

Memristors are memory resistors that promise the efficient implementation of synaptic weights in artificial neural networks. Whereas demonstrations of the synaptic operation of memristors already exist, the implementation of even simple networks is more challenging and has yet to be reported. Here we demonstrate pattern classification using a single-layer perceptron network implemented with a memrisitive crossbar circuit and trained using the perceptron learning rule by ex situ and in situ methods. In the first case, synaptic weights, which are realized as conductances of titanium dioxide memristors, are calculated on a precursor software-based network and then imported sequentially into the crossbar circuit. In the second case, training is implemented in situ, so the weights are adjusted in parallel. Both methods work satisfactorily despite significant variations in the switching behaviour of the memristors. These results give hope for the anticipated efficient implementation of artificial neuromorphic networks and pave the way for dense, high-performance information processing systems.

Ionically-mediated electromechanical hysteresis in transition metal oxides.

Nanoscale electromechanical activity, remanent polarization states, and hysteresis loops in paraelectric TiO(2) and SrTiO(3) thin films are observed using scanning probe microscopy. The coupling between the ionic dynamics and incipient ferroelectricity in these materials is analyzed using extended Landau-Ginzburg-Devonshire (LGD) theory. The possible origins of electromechanical coupling including ionic dynamics, surface-charge induced electrostriction, and ionically induced ferroelectricity are identified. For the latter, the ionic contribution can change the sign of first order LGD expansion coefficient, rendering material effectively ferroelectric. The lifetime of these ionically induced ferroelectric states is then controlled by the transport time of the mobile ionic species and well above that of polarization switching. These studies provide possible explanation for ferroelectric-like behavior in centrosymmetric transition metal oxides.

High precision tuning of state for memristive devices by adaptable variation-tolerant algorithm.

Using memristive properties common for titanium dioxide thin film devices, we designed a simple write algorithm to tune device conductance at a specific bias point to 1% relative accuracy (which is roughly equivalent to seven-bit precision) within its dynamic range even in the presence of large variations in switching behavior. The high precision state is nonvolatile and the results are likely to be sustained for nanoscale memristive devices because of the inherent filamentary nature of the resistive switching. The proposed functionality of memristive devices is especially attractive for analog computing with low precision data. As one representative example we demonstrate hybrid circuitry consisting of an integrated circuit summing amplifier and two memristive devices to perform the analog multiply-and-add (dot-product) computation, which is a typical bottleneck operation in information processing.