Multiple modern applications of electronics call for inexpensive chips that can perform complex operations on natural data with limited energy. A vision for accomplishing this is implementing hardware neural networks, which fuse computation and memory, with low cost organic electronics. A challenge, however, is the implementation of synapses (analog memories) composed of such materials. In this work, we introduce robust, fastly programmable, nonvolatile organic memristive nanodevices based on electrografted redox complexes that implement synapses thanks to a wide range of accessible intermediate conductivity states. We demonstrate experimentally an elementary neural network, capable of learning functions, which combines four pairs of organic memristors as synapses and conventional electronics as neurons. Our architecture is highly resilient to issues caused by imperfect devices. It tolerates inter-device variability and an adaptable learning rule offers immunity against asymmetries in device switching. Highly compliant with conventional fabrication processes, the system can be extended to larger computing systems capable of complex cognitive tasks, as demonstrated in complementary simulations.

The integration of memristive nanodevices within transistor-based electronic systems offers the potential for computing structures smaller, lower power and cheaper than traditional high-performance systems. Among emerging memristive technologies, a novel device based on organic materials distinguishes itself, in that it can feature several threshold voltages on the same die, and possesses unipolar behavior. In this work, we highlight that these two features can be beneficial for neural network-inspired learning systems. An on-chip supervised learning method for hybrid memristors / CMOS systems -an analogue synaptic array paired with a hybrid learning cell -is extended to the case of this novel organic memristor device. The organic device can be trained with only one pulse per row (two for the entire array) per presentation of input-as compared to four for a bipolar memristor array. The device also works universally-in both the synaptic grid as well as learning cellpaving the way to single die integration. The proposed scheme learns successfully, even while incorporating non-ideal circuit phenomena such as a wide range of parasitic wire resistances and associated sneak paths. These encouraging first results suggest that these multi-threshold, unipolar organic memristive devices are a useful species for inclusion in adaptive next generation electronic systems.

International audienceThe capabilities of memristors to serve as artificial synapses in neural network type of circuits have been recently recognized. These two-terminal analog memory devices offer valuable advantages in terms of circuit architectures. In particular, with their room temperature processes and large diversity coming from chemistry, organic memristors represent a chance to develop devices that can be densely integrated above-IC. In this article, we present a new class of organic resistive memory based on a robust electrografted redox thin film as active material integrated in a planar metal/organic/metal topology. The combination of a specific redox polymer and of the electro-grafting technique leading to fully covalent films makes such organic memristors particularly robust. The devices display high RMAX/RMIN ratio, long retention time and multi-level conductivity. The potential of these devices to store analog synaptic weights in neural network circuit strategies is shown by demonstrating their compatibility with the Spike Timing Dependence Plasticity (STDP) learning rule and by implementing the associative memory function

International audienceOrganic memristors are promising molecular electronic devices for neuro-inspired on-chip learning applications. In this paper, we present a numerically efficient compact model suitable for Fe(bpy)2+3 organic memristors operating according to an intramolecular charge transfer switching mechanism. This compact model, being physics-based and relying on electrical characterizations and parametric extractions performed on test structures, is especially efficient in pulsed mode and describes the conductance variations for both SET and RESET regimes. Using this model, a dynamic multiscale simulation approach has been set-up to extend the model from individual devices to larger model systems that learn progressively through time. To verify the soundness and highlight emergent properties of the organic memristors, instances of the compact model have been simulated within a simple neuromorphic design that co-integrates with CMOS neurons. In addition, a larger supervised learning system using the new compact model is demonstrated. These successful tests suggest our model might be of interest to neuromorphic designers