Neuromorphic electronics aims to create intelligent systems that mimic the nervous system's ability to process diverse signals. Emulating the brain's building blocks (synapses and neurons) allows bio-inspired processing and computation, with applications in edge computing, wearables, point-of-care diagnostics, bioelectronics, biorobotics, and environmental intelligence. Various materials have been used, including metal oxides, ferroelectrics, ferromagnetics, phase-change materials, and 2D materials. Organic materials are particularly interesting due to their biological resemblance. Neuromorphic electronics using organic mixed ionic-electronic conductors (OMIECs) can realistically emulate biological phenomena because they respond to biological information carriers (ions, neurotransmitters, neuromodulators). Their operation in wet environments enables neuromorphic biosensors and bioelectronics. Previous work has demonstrated artificial synapses and networks using OMIECs, and organic electrochemical transistors (OECTs). Recently, organic electrochemical artificial neurons (OANs) have emerged, with various circuit implementations. These OANs realistically mimic biological phenomena by responding to ions, electrolyte noise, and biological conditions. Different OAN designs have shown regular firing, firing frequency responsiveness to ion concentration, biorealistic firing properties, neuronal excitability, and biohybrid integration. Despite these advancements, rigorous investigation of circuit operation and quantitative models incorporating soft-matter parameters and biological wetware are lacking. This knowledge gap hinders the simulation of larger-scale circuits and the development of integrated organic neuromorphic electronics, biohybrids, and intelligent bioelectronics. This study addresses this gap by combining experimental, simulation, and analytical approaches to unravel OAN operation.

The existing literature extensively covers neuromorphic computing using various materials. Research on silicon-based neuromorphic devices, while mature, often lacks the biocompatibility and soft-matter integration capabilities desired for biohybrid systems. Inorganic memristors, based on materials like metal oxides, have demonstrated impressive performance in terms of switching speed and endurance. However, these materials typically require high voltages and are not readily compatible with biological systems. The use of organic materials, specifically OMIECs, offers a unique advantage. Their inherent biocompatibility and ability to operate in aqueous environments make them ideally suited for integration with biological systems. Several studies have demonstrated the use of OECTs as building blocks for artificial synapses and neural networks. These studies highlighted the potential of organic materials in neuromorphic computing but often lacked a comprehensive understanding of the underlying physical mechanisms and the ability to precisely control the device behavior. This paper aims to build upon this existing knowledge by developing a quantitative model and a deeper understanding of OAN operation.

This work uses a multi-pronged approach to investigate organic artificial neurons exhibiting S-shape negative differential resistance (S-NDR). The OAN consists of an organic electrochemical non-linear device (OEND) coupled to a biasing network (resistor R₁, capacitor C₁, DC voltage generator V_IN). The OEND uses two OECTs (T₁ and T₂, with PEDOT:PSS and p(g2T-TT) as channel materials, respectively). T₁ is a depletion-mode OECT, and T₂ is an accumulation-mode OECT. The OECT electrical characteristics are modeled, accounting for volumetric capacitance (C_v), ion-concentration-dependent threshold voltage (V_TH), energy disorder (γ), and channel-length modulation. This model was implemented in a circuit simulator to reproduce the measured electrical characteristics of the OEND. The OEND was characterized in both voltage and current modes, revealing a hysteretic characteristic in voltage mode and S-NDR in current mode. Nonlinear transient simulations of the full OAN were performed using the OECT model, accurately reproducing measured current and voltage output oscillations. Analytical expressions for V_ON and V_OFF (the voltages at which the OEND switches) were derived as functions of material, geometrical, and device parameters. These expressions allowed for the calculation of the amplitude of voltage oscillations (A_vspike = V_ON - V_OFF). Furthermore, an analytical expression for the spiking frequency (f_spike) was derived, considering the OEND characteristics, input voltage, and load resistor. Parametric analysis was performed using numerical simulations to evaluate the impact of threshold voltage (V_TH1, V_TH2), transconductance (g_m1, g_m2), volumetric capacitance (C_V1, C_V2), width (W₁, W₂), length (L₁, L₂), thickness (t₁, t₂), and resistance (R₁, R₂) on OAN performance. Excitability was investigated by injecting a sinusoidal excitation signal into the electrolyte medium of T₁, analyzing the resulting phase-locked bursting activity. Noise-induced activity was studied by applying white noise to the electrolyte, observing the transitions from tonic to bursting firing. Neuromorphic ion sensing was performed by varying ion concentrations (Na⁺, K⁺, Ca²⁺) in the electrolyte, measuring the resulting changes in spiking frequency. Finally, biointerfacing was demonstrated by integrating a cellular barrier between the channel and gate of T₁, analyzing the effect of barrier integrity on OAN spiking activity using electrochemical impedance spectroscopy (EIS).

The study provides a comprehensive understanding of OAN operation. The OEND's hysteretic behavior in voltage mode and S-NDR in current mode are crucial for spiking behavior. Analytical expressions for V_ON, V_OFF, and f_spike accurately predict experimental results and offer a tool for designing OANs. Parametric analysis revealed the significant influence of material and device parameters on spiking frequency (f_spike), voltage amplitude (A_vspike), current amplitude (A_tspike), power consumption (P_OAN), and energy per spike (E_spike). OAN excitability is tunable from microvolts to millivolts, and noise can induce transitions between tonic and bursting firing modes. The OAN exhibits neuromorphic ion sensing, with spiking frequency modulated by Na⁺, K⁺, and Ca²⁺ concentrations within the physiological range. The spiking frequency response shows high robustness to noise, contrasting with traditional OECT-based ion sensing. Ion-selective OANs were demonstrated using ionophore-based selective membranes, enabling ion-specific spiking activity. Biointerfacing with a cellular barrier showed that the barrier's integrity modulates OAN excitability and spiking, providing a method for monitoring barrier functionality. The impact of membrane resistance (R_M) and capacitance (C_M) on signal attenuation was also quantified.

This research significantly advances the field of neuromorphic bioelectronics by providing a detailed, quantitative understanding of OAN operation. The combination of experimental data, physics-based simulations, and analytical models allows for the rational design and optimization of OANs. The findings demonstrate the potential of OANs for creating biocompatible, high-performance neuromorphic devices. The ability to precisely control spiking frequency and excitability through material and device parameters opens possibilities for building complex neural networks. The demonstrated neuromorphic ion sensing capability has implications for developing next-generation biosensors for physiological monitoring. The biohybrid integration with cellular barriers further underscores the potential of OANs for studying cellular processes and creating novel bioelectronic interfaces. The robustness of the OAN to noise, contrasting with traditional OECT sensors, highlights a key advantage for real-world applications. This study provides a strong foundation for future research on OAN-based systems for applications in bioelectronics, neuromorphic computing, and biohybrid systems.

This study provides a comprehensive understanding of organic artificial neuron operation, bridging the gap between experimental observations and theoretical models. The developed analytical expressions and numerical framework enable rational design and optimization of OAN parameters for targeted functionalities. The work highlights the potential of OANs in neuromorphic ion sensing, biohybrid integration, and advancing next-generation neuromorphic bioelectronics. Future research should focus on reducing the energy per spike to approach biological levels and exploring more complex neural network architectures using OANs.

While this study provides a thorough investigation, limitations exist. The model simplifies some aspects of the OECT operation and the bio-interface. Further refinement of the model might enhance accuracy. The current work focuses on single neuron devices; scaling up to larger networks requires further investigation. The energy consumption per spike of the OAN is currently higher than biological neurons, a challenge to be addressed in future work. The specific cellular barrier used in this study may influence the results and different cell types or barrier architectures may exhibit diverse behaviors.