The development of advanced brain-machine interfaces, prosthetics, and soft robotics necessitates seamless integration of artificial neuromorphic devices with biological systems. Current silicon-based neuromorphic implementations face significant challenges, including rigidity, poor biocompatibility, high energy consumption, and fundamentally different operating principles compared to biological systems. These differences hinder effective bio-integration. Organic semiconductors offer a promising alternative due to their biocompatibility, processability, flexibility, and ability to mimic biological signal modulation. While organic materials have shown success in emulating neuromorphic functions, creating and bio-integrating artificial neurons capable of spike-based information encoding, closely mirroring biological systems, remains a challenge. Existing organic field-effect transistor (OECT)-based artificial neurons often require high operating voltages, further limiting bio-integration. Organic electrochemical transistors (OECTs), modulated by gate-driven ionic doping/de-doping, offer a closer resemblance to biological ion-driven processes. Their low voltage operation (<1 V), high transconductance, and biocompatibility make them ideal candidates for creating printed, biocompatible artificial spiking neural circuits. This research focuses on addressing the need for bio-integrable artificial neurons by developing and characterizing OECNs and demonstrating their successful bio-integration.

The paper reviews existing literature on neuromorphic systems and their limitations. Silicon-based neuromorphic systems, while prevalent, suffer from biocompatibility issues and operational discrepancies from biological systems. The advantages of organic semiconductors in bioelectronics are highlighted, including their biocompatibility, processability, flexibility, and ability to support both ionic and electronic signal transport. The authors note previous work on organic synapses and neuromorphic devices but emphasize the limited success in fabricating and bio-integrating artificial neurons capable of spike-based information encoding. The superior properties of OECTs compared to OFETs in terms of biocompatibility, lower operating voltage, and closer resemblance to biological mechanisms are discussed, setting the stage for the proposed OECN design.

The study details the fabrication of all-printed complementary OECTs using screen printing and spray coating techniques on polyethylene terephthalate (PET) substrates. The OECTs utilize n-type and p-type transistors (P(BZT-T) and BBL, respectively) with lateral Ag/AgCl gates. The OECNs are constructed using these OECTs, along with a capacitor (Cmem) for current integration and an amplifier block. The operational mechanism of the OECNs is described, drawing parallels to biological neurons. The OECSs are fabricated using a similar printed electrode architecture, employing electropolymerization of a zwitterionic monomer (ETE-PC) to form the synaptic channel. Electrical characterization involves using a semiconductor parameter analyzer to measure OECT characteristics. SPICE modeling is used to simulate OECTs and OECNs. Bio-integration is demonstrated by interfacing the OECNs with Venus flytraps (Dionaea muscipula) using Ag/AgCl electrodes. The experimental setup for the Venus flytrap experiments is described in detail. The methodology also encompasses the characterization of short-term and long-term plasticity in the OECSs, including paired-pulse facilitation and long-term potentiation/depression.

The researchers successfully fabricated and characterized all-printed OECNs exhibiting concentration-dependent spiking, mirroring biological neurons. These OECNs respond to a wide range of input currents (0.1–10 μA), with frequency modulation exceeding 450%. Bio-integration with a Venus flytrap demonstrated successful induction of lobe closure via electrical stimulation from the OECN. The OECNs, operating at under 0.6 V, showed two distinct modes of operation based on ionic doping and electrochemical-induced conductivity modulation, facilitating short-term and long-term plasticity. Integration with OECSs enabled the demonstration of Hebbian learning, with long-term plasticity showing retention exceeding 1000 s. The spiking frequency of the OECNs is modulated by varying input current, membrane capacitance, and electrolyte concentration. A fully printed version of the neuron showed similar characteristics. The study demonstrates a simple neuro-synaptic system with Hebbian learning using a single synaptic transistor connected to the OECN, showcasing the potential for local learning capabilities.

The findings address the need for bio-integrable artificial neurons by presenting a fully printed, low-voltage, biocompatible OECN that closely mimics biological neuron behavior and enables effective bio-integration. The successful bio-integration with the Venus flytrap showcases the potential for interfacing artificial neural systems with biological organisms. The demonstration of Hebbian learning using the integrated OECN-OECS system highlights the potential for creating neuromorphic systems with local learning capabilities, reducing the reliance on complex external computing systems. The ability to modulate the spiking frequency through various parameters demonstrates the versatility and controllability of the OECNs, making them adaptable to various bio-integration scenarios. The low power consumption of the OECN makes it suitable for implantable applications. These results open new avenues for creating advanced bioelectronic systems that can interact with living organisms at a neural level. The advantages of using organic materials over silicon are repeatedly highlighted throughout the study.

This work presents the first successful demonstration of all-printed, bio-integrated OECNs with ion-modulated spiking and integrated OECSs capable of Hebbian learning. The soft, flexible, low-power, and biocompatible nature of these devices opens exciting possibilities for brain-machine interfaces, prosthetics, and soft robotics. The successful bio-integration with a Venus flytrap exemplifies the potential of these devices for interacting with diverse biological systems. Future research could explore the integration of more complex neural networks and the development of more sophisticated learning algorithms within this organic platform.

While the study demonstrates significant progress in bio-integrated neuromorphic devices, there are limitations. The long-term stability of the OECSs and the robustness of the bio-integration in vivo require further investigation. The current study focuses on a single neuron and synapse; scaling up to larger, more complex networks would require further development. The current implementation uses a simplified model of Hebbian learning, and more complex learning paradigms might require additional modifications to the system.